**A High Protein Diet Is More Effective in Improving Insulin Resistance and Glycemic Variability Compared to a Mediterranean Diet—A Cross-Over Controlled Inpatient Dietary Study**

**Francesca Tettamanzi 1,†, Vincenzo Bagnardi 2,†, Panayiotis Louca 3, Ana Nogal 3, Gianna Serafina Monti 4, Sara P. Mambrini 5,6, Elisa Lucchetti 7, Sabrina Maestrini 7, Silvia Mazza 5, Ana Rodriguez-Mateos 8, Massimo Scacchi 5,9, Ana M. Valdes 10,‡, Cecilia Invitti 11,‡ and Cristina Menni 3,\*,‡**


**Abstract:** The optimal dietary pattern to improve metabolic function remains elusive. In a 21-day randomized controlled inpatient crossover feeding trial of 20 insulin-resistant obese women, we assessed the extent to which two isocaloric dietary interventions—Mediterranean (M) and high protein (HP)—improved metabolic parameters. Obese women were assigned to one of the following dietary sequences: M–HP or HP–M. Cardiometabolic parameters, body weight, glucose monitoring and gut microbiome composition were assessed. Sixteen women completed the study. Compared to the M diet, the HP diet was more effective in (i) reducing insulin resistance (insulin: Beta (95% CI) = −6.98 (−12.30, −1.65) μIU/mL, *p* = 0.01; HOMA-IR: −1.78 (95% CI: −3.03, −0.52), *<sup>p</sup>* = 9 <sup>×</sup> <sup>10</sup><sup>−</sup>3); and (ii) improving glycemic variability (−3.13 (−4.60, <sup>−</sup>1.67) mg/dL, *<sup>p</sup>* = 4 <sup>×</sup> <sup>10</sup><sup>−</sup>4), a risk factor for T2D development. We then identified a panel of 10 microbial genera predictive of the difference in glycemic variability between the two diets. These include the genera *Coprococcus* and *Lachnoclostridium*, previously associated with glucose homeostasis and insulin resistance. Our results suggest that morbidly obese women with insulin resistance can achieve better control of insulin resistance and glycemic variability on a high HP diet compared to an M diet.

**Citation:** Tettamanzi, F.; Bagnardi, V.; Louca, P.; Nogal, A.; Monti, G.S.; Mambrini, S.P.; Lucchetti, E.; Maestrini, S.; Mazza, S.; Rodriguez-Mateos, A.; et al. A High Protein Diet Is More Effective in Improving Insulin Resistance and Glycemic Variability Compared to a Mediterranean Diet—A Cross-Over Controlled Inpatient Dietary Study. *Nutrients* **2021**, *13*, 4380. https:// doi.org/10.3390/nu13124380

Academic Editors: Silvia V. Conde and Fatima O. Martins

Received: 15 November 2021 Accepted: 2 December 2021 Published: 7 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Keywords:** high protein diet; Mediterranean diet; insulin resistance; glycemic variability; obesity; gut microbiome; dietary intervention

#### **1. Introduction**

An obesity pandemic is gripping the globe, with higher demand and availability for energy-dense foods, accompanied by increasingly sedentary lifestyles [1–3]. This is a major public health concern, as obesity often confers an increased risk of developing a wide range of complex and life-changing diseases, including cardiovascular and cerebrovascular disease, type II diabetes and cancers [4–7]. Therefore, the development and implementation of effective and affordable measures to combat obesity is of utmost importance. As well as encouraging increased physical activity, many efforts to reduce obesity and its associated disorders have focused on the impact of diet and nutrition [8]. In particular, the Mediterranean (M) diet, a diet characterized by high levels of polyphenols, monoand polyunsaturated fatty acids (MUFAs and PUFAs), antioxidants, and fiber, as well as low levels of salt, sugar and saturated fatty acids [9], has been associated with improved health outcomes [9]. Greater adherence to the M diet has been associated with reduced risk of cardiovascular disease [9,10], which also supports weight loss [11]. A high-protein (HP) diet, comprising low carbohydrate, high fat and high protein intake, has also been suggested as a potential dietary intervention for obesity prevention [12] with HP diets corresponding to greater weight loss compared to similar isocaloric diets with standard protein content [13]. A HP diet has also been shown to lead to a greater weight loss compared to a high-carbohydrate diet, along with an improvement in insulin parameters, highlighting its power to lower the risk of type 2 diabetes and cardiovascular diseases [14,15]. Over the short term, a HP diet has been suggested to more effectively aid weight loss in contrast to a low-fat diet, and has been shown to change body composition in overweight or obese men [14]. The mechanisms supporting the HP diets' effects on weight loss efficacy is theorized to be related to increased satiety [16] and it has been suggested that this enhances an individual's metabolic rate [17]. Recent evidence suggests that the benefits of any dietary intervention are intrinsically linked to an individual's metabolic profile [18]. The exact role of the gut microbiome in nutrient metabolism is still unclear, but various studies have linked microbial diversity and specific bacteria to a propensity for obesity, as well as to the metabolism of dietary compounds found in the M diet, including omega-3 fatty acids and polyphenols [19,20]. Here, we aimed to explore the differential effects on metabolic parameters elicited by the M and HP diets. As these effects are reported to be exacerbated in obese individuals with impaired metabolic response, we conducted a 21-day randomized crossover controlled dietary trial in 20 insulin-resistant women with obesity.

#### **2. Materials and Methods**

A flowchart of the study design is presented in Figure 1.

**Figure 1.** Study flowchart.

#### *2.1. Study Design and Participants*

This is an open-label, single-center randomized crossover controlled dietary trial in an inpatient setting. Participants were assigned to receive, in a 1:1 ratio, one of the two following dietary sequences: hypocaloric M diet followed by hypocaloric HP diet (sequence M–HP) or vice versa (sequence HP–M). Each period of intervention lasted 10 days with no washout before the switch from the first to the second diet. Participants were at San Giuseppe Hospital, Piancavallo of the IRCCS Istituto Auxologico Italiano, where they were hospitalized throughout the duration of the trial. Patients eligible for the study were women aged 20 to 57 years with BMI 35–64 kg/m2, insulin resistant (HOMA-IR ≥ 3) and able to perform physical activity. Exclusion criteria included individuals suffering from type 2 diabetes mellitus (defined by the presence of occasional plasma glucose value of ≥200 mg/dL or a fasting plasma glucose of ≥126 mg/dl or an HbA1c ≥ 6.5% (≥48 mmol/mol Hb)), binge eating disorder, taking proton pump inhibitors, antibiotics, metformin or probiotics. Moreover, women included in the study were not following any specific dietary patterns in the 6 months preceding study enrolment, and were characterized by prandial hyperphagia, excessive carbohydrates, lipid and sodium consumption, poor fiber intake and insufficient hydration.

The study protocol was approved by the Institutional Ethical Committee (2018\_01\_30\_02) and all participants provided written informed consent before the trial.

#### *2.2. Study Procedures*

Eligible patients were randomly allocated into two groups on day 1: sequence M–HP or sequence HP–M. The M diet was composed of approximately 55% carbohydrates (whole wheat), 25% fat (PUFA from olive oil, almonds and pistachios) and 20% protein (fish, goat cheese and legumes). The HP diet was composed of approximately 40% carbohydrate, 30% fat and 30% protein. Both diets had the same caloric intake, which was 500 Kcal less than the individual daily caloric requirement, and a similar equally moderate glycemic load ranging between 11 and 19. Moreover, for both the M and HP diets, the energy derived from the consumption of simple carbohydrates (represented mainly by fruits and dairy products) was lower than 15% of the total energy. Animal and vegetable proteins were provided in both diets. In the M diet, protein consumption was in line with the relevant Food Guide Pyramid. Second courses included mainly white meat, bluefish, goat cheese and legumes. In the HP diet, proteins sources were mainly white meat, fish and eggs (Table S1).

On day 1, baseline measurements of clinical variables were obtained for each participant, including height, weight, waist and hip circumference, blood pressure, heart rate and body composition as measured by phase-sensitive, single-frequency bioimpedance analyzer (BIA 101, Akern, Pisa, Italy). Resting energy expenditure (REE) was assessed with indirect computerized calorimetry (Vmax 29, Sensor Medics, Yorba Linda, CA, USA), and the total energy expenditure (TEE) was estimated by multiplying the REE by Physical Activity Level (PAL), which was 1.2 for all, i.e., *TEE* = *REE* × 1.2. 500 kcal were subtracted from the individual TEE to determine the diet hypocaloric target. Additionally, fasting blood samples for insulin and lipids (total, LDL and HDL cholesterol, triglycerides) measurement and stool samples were collected. Fecal samples were immediately frozen at −20 ◦C. For the gut microbiota analysis, samples were stored at −80 ◦C directly until processing following 3–5 h refrigeration.

Measurements and sample collections were repeated during clinical visits on days 6, 11, 16, and 21. Adherence to the diet was closely monitored by the nurses. Throughout the study, glucose levels were monitored by flash continuous glucose monitoring (FSL-FGM; Free-Style Libre™; Abbott, Witney, Oxfordshire, UK).

#### *2.3. 16S rRNA Gut Microbiome*

Microbial 16S rRNA gene was extracted from fecal samples and sequenced using the Illumina MiSeq platform at the Genetic Laboratory, Erasmus Medical Centre in Rotterdam, the Netherlands. The Microbiota pipeline 25 was used to filter and cluster reads into Operational Taxonomic Units (OTU) based on 97% similarity against the SILVA database v132 [20,21]. Microbial diversity indices were calculated using the platform QIIME 2 (v2018.11) as the average value after rarefying the OTU table to 13678 reads. Shannon and Simpson indices were calculated to describe the alpha diversity (i.e., microbial diversity within individual samples) [22,23]:

*Shannon Index* <sup>=</sup> <sup>−</sup> <sup>∑</sup>*<sup>s</sup> <sup>i</sup>*=1(*pilog*<sup>2</sup> *pi*), where *s* is the number of OTUs and *pi* the proportion of the community represented by OTU *i*.

*Simpson Index* = <sup>1</sup> − <sup>∑</sup> *<sup>p</sup>*<sup>2</sup> *<sup>i</sup>* , where *pi* is the proportion of the community represented by OTU *i*.

#### *2.4. Study Outcomes*

The primary study outcomes were insulin and HOMA-IR measured as the change from baseline concentration during each diet (i.e., the difference between insulin, HOMA-IR from day 11 to day 1 for the first diet in the sequence, and between day 21 and day 11 for the second diet in the sequence) and glycemic variability. Individual HOMA-IR was computed as *HOMA-IR* = (*fasting insulin* × *fasting glucose*)/405 with glucose measured in mg/dl and insulin in μU/L.

Individual glycemic variability was measured for each diet as daily mean standard deviation (SD) of glucose concentration during continuous monitoring, that is, for each

individual the mean standard deviation was calculated as *SD* = 1/*dx* ∑ *SDd*, where *SDd* is standard deviation of each day's glucose measurements in HP or M diet. For overall evaluation of CGM data, we also computed (i) the difference between diets in mean blood glucose concentration and (ii) the percentage of time of sensor glucose concentration within target range (within 70–180 mg/dL), in hypoglycemic (below 70 mg/dL) and hyperglycemic (above 180 mg/dl) conditions. All CGM metrics were calculated using the R package iglu [24]. Secondary study outcomes included change from baseline in weight waist to hip ratio, fat to lean mass ratio, lipids, blood pressure, heart rate and microbial diversity metrics.

#### *2.5. Statistical Analysis*

The analysis dataset included all participants who completed the dietary sequence and had measurement of the main study outcomes at least at the beginning and at the end of each intervention period (i.e., day 1, 11 and 21) (Figure 1). According to the intentionto-treat principle, patients were analyzed in the dietary sequence assigned. Baseline characteristics of the study population were described as mean values along with their standard deviations (SD). To adjust for treatment period and sequence, a linear mixed effect regression model was fitted, which included as fixed predictors treatment type (HP over M), treatment period (P1 over 2) and sequence (HP–M over M–HP). The effect of the type of diet on glycemic variability was evaluated by calculating the difference in SD of glucose concentration between HP and M diets, and applying linear mixed effect regression model as described above. Additional study outcomes, including clinical, microbial and other glucose-related variables, were similarly analyzed. Estimates of unadjusted and adjusted mean values along with their 95% Confidence Intervals (CI) were calculated for each outcome. Mean differences and CI were standardized to obtain comparable effect sizes for considered variables, and were represented in forest plots.

Exploratory sub-analyses were performed to evaluate the association of baseline microbial taxa with the difference between HP and M diet in individual glucose variability (*SDHP* <sup>−</sup> *SDM*, with SD as defined above), by using a Lasso regression model with zero sum constraint to account for the compositional nature of microbial data [25]. Relative abundances (RA) of OTU agglomerated to genus level were calculated, filtered if sparse in less than 80% of the samples, and log transformed before the analysis. For variable selection, 5-fold cross-validation was applied to tune the regularization term lambda. Associations were expressed as beta regression coefficients.

Statistical analyses were performed using the statistical software SAS 9.4 (SAS Institute, Cary, NC, USA) and R version 3.6.2. A *p*-value less than 0.05 was considered statistically significant.

#### **3. Results**

Between April and December 2018, 20 patients were enrolled in the study at the Piancavallo Hospital, Italy. Of them, three patients decided not to take part in the study before randomization, and one patient assigned to sequence M–HP discontinued the study after the first diet; 16 participants completed the dietary sequence assigned and represented the analysis set (Figure 1). As depicted in Figure 1, participants were enrolled in the study and randomly allocated on day 1 to one of two dietary sequences: the HP–M indicates high protein diet followed by Mediterranean diet, and the M–HP indicates the Mediterranean diet followed by high protein diet. Crossover (C) to the second diet occurred on day 12.

The characteristics of the participants at study entry are presented in Table 1. On average, women were slightly younger, had a lower BMI, fasting glucose, insulin and HOMA-IR in the HP–M group compared to the M–HP group. However, differences were not statistically significant.


**Table 1.** Clinical and biochemical characteristics of obese women at baseline.

Mean (SD). HP: high protein diet; M: Mediterranean diet; BMI: body mass index; SBP: systolic blood pressure, DBP: diastolic blood pressure, LDL: low-density lipoprotein; HDL: high-density lipoprotein; HOMA-IR: homeostasis model assessment of insulin resistance; HbA1c: hemoglobin A1c; \* Statistically significant difference between two sequence groups according to *t*-test (fat to lean mass ratio: *p* = 0.05; HDL cholesterol: *p* = 0.01).

The HP and M diets led to a similar loss of body weight, with a mean change from baseline of −2.71 (95% CI: −3.59, −1.82) kg and −2.09 (95% CI: −2.71, −1.46) kg, respectively (Figure S1). Moreover, reduction in body weight was greater during the first period and diminished in the second part of the study, regardless of the dietary sequence (Figure S1). Changes in other biometric measures such as BMI, waist to hip ratio and fat to lean mass ratio lipids, blood pressure and in gut microbiome composition (Shannon and Simpson indexes) were also similar after the two diets (Figure S1).

#### *3.1. Improvement in Insulin Resistance and HOMA-IR*

In order to investigate whether an improvement in insulin resistance could be achieved after the two diets, we compared elicited effects on insulin and HOMA-IR variation using linear mixed effect regression models, with adjustment for treatment period and intervention sequence.

As shown in Figure 2A, the HP diet was more effective in reducing insulin levels, leading to a mean change from baseline of −3.50 (95% CI: −8.22, 1.21) μIU/mL, while higher levels were registered after the M diet with a value of 1.55 (95% CI: −1.08, 4.18) μIU/mL. Similarly, the HP diet led to a greater reduction in HOMA-IR with respect to the M diet with mean change from baseline of −0.996 (95% CI: −2.11, 0.12) and 0.32 (95% CI: −0.32, 0.96). Differences in the two outcomes between diets were statistically significant (*<sup>p</sup>* = 0.01, *<sup>p</sup>* = 9 × <sup>10</sup>−3). Reduction in glucose concentration was slightly greater in HP diet (−2.44 (95% CI: −6.02, 1.14) mg/dL) with respect to M diet (−1.88b (95% CI: −491, 1.16) mg/dL), however the difference between the two interventions was not statistically significant (*p* = 0.55).

#### *3.2. Effect of HP and M Diets on Glycemic Variability*

To further investigate the possible differential effect of the two dietary regimens on glucose variability, continuous monitoring data on glucose concentration were analyzed. Figure 2B shows 24 h sensor glucose profiles for each diet (individual patients' profiles are reported in Figure S2). Mean differences between interventions in SD of glucose concentration and other glucose summary outcomes, after adjustment for dietary sequence

and treatment period, are presented in Figure 2C along with related standardized effect size and *p*-values (means of glucose outcomes in the two groups are reported in Table S2).

Patients while on HP diet improved glycemic variability, showing a significant reduction in SD of glucose concentration (Figure 2C), with a mean of 14.79 (95% CI: 12.83, 16.75) mg/dL compared to 17.92 (95% CI: 15.96, 19.89) mg/dL observed during M diet (*<sup>p</sup>* = 4 × <sup>10</sup>−<sup>4</sup> for the difference) as reported in Table S2. Consistent results were also observed for both mean amplitude of glycemic excursions (MAGE) and the mean of daily differences (MODD) [26], Figure 2C. CGM data supported the previous indication that glucose levels were not affected by the type of diet, as the mean daily concentration of blood glucose was comparable in the two groups (Figure 2C).

Patients spent similar sensor time at glucose levels below 70 mg/dL during HP diet and M diet. Time spent in hyperglycemic conditions at glucose levels above 180 mg/dL was limited and comparable for the two diets. No differences were detected between diets in indices of low and high blood glucose risks (data not shown).

#### *3.3. Association of Baseline Gut Microbial Composition at Genus Level with Glucose Variability*

We further investigated whether the patients' gut microbiome composition could be related to the difference in glycemic variability observed in the two diets. After aggregating OTUs into 148 genera, with the use of a zero sum constraint regression model, we identified a panel of 10 microbial genera (Figure 3).

**Figure 3.** Microbial genera associated with the difference in glucose standard deviation between diets. Association between baseline microbial composition and mean difference of glucose index between HP and M diets was evaluated using zero sum constraint regression model. Results are reported in terms of beta regression coefficients, where the increase in the relative abundance of the selected genus at baseline is associated with an increased difference (red bars) or a decreased difference (blue bars) in SD of glucose concentration between HP and M diets. OTUs at baseline were first agglomerated to genus level. Relative abundances were filtered for sparsity and log transformed before the analysis. SD: standard deviation; (*N* = 16).

Of the 10 identified microbial taxa, 4 genera were annotated to the family of *Lachnospiraceae* (with opposite directions), one genus to *Ruminococcaceae* (with negative direction), *Peptostreptococcaceae* (with positive direction), *Acidaminococcaceae* (with negative direction), *Clostridiaceae* (with positive direction), *Coriobacteriaceae* (with negative direction) and *Desulfovibrionaceae* (with positive direction) families.

#### **4. Discussion**

In this 21-day randomized crossover controlled inpatient feeding trial, we found the HP diet to be more effective in reducing insulin resistance and in improving glycemic variability, compared to the M diet in 16 morbidly obese women with pre-diabetes. Moreover, we identified a panel of 10 microbial genera underlying the difference in glycemic variability between the two diets. These include microbes previously associated with the regulation of glucose homeostasis and insulin resistance [25,27].

We have also reported that both diets are equally effective in reducing weight (Figure S1) with participants consistently losing more weight during the first half of the study, compared to the second half. The lack of difference observed in weight, waist to hip ratio and BMI between the M and HP diets suggests that the weight loss observed may be primarily due to the isocaloric nature of the two diets, rather than specific dietary components. Moreover, participants may have benefitted overall from improved nutrition (increased fiber, PUFA, etc.) compared to their previous dietary habits; however, this information was unavailable for study.

Beneficial effects on health outcomes and metabolic functions have been reported in individuals adhering to both HP and M diets [28,29]. A HP diet has been linked with greater improvements in metabolic health and insulin sensitivity in individuals, mainly obese, overweight or insulin-resistant, when compared to alternate diets if weight loss is achieved [30,31]. Indeed, a greater decrease in HOMA-IR/insulin was reported in (i) obese women on an 8-week HP compared to those on a low protein diet [32]; (ii) in obese women on a 9-month isocaloric HP-low carbohydrate diet compared to those adhering to a standard isocaloric diet [33]; (iii) in overweight and obese women with the highest protein uptake in a 6-month calorie reduced diet with increasing protein content (20%, 27% or 35%) [33,34], among others. Our results are consistent and support, in this study group, a greater improvement in insulin resistance and related parameters after following an HP diet.

In our study, HP and M diets did not elicit an effect on the mean glucose level, both during clinical visits and when using 24 h glucose monitoring data. This observation is in line with several studies that reported no effect on the mean glucose levels after high protein intake in T2D patients [35]. However, we found a lower glycemic variability in morbidly obese women following the HP diet. Glucose variability is a risk factor for T2D development and complications [36], and increases in variability may be considered an additional parameter in the assessment of glucose homeostasis at the early stages of glucose dysregulation [37,38]. Reducing glucose variability by diet in non-diabetic patients may be clinically relevant because at early stages of dysglycemia; there is a decline of the cardiac autonomic function that is related to glucose variability and HOMA-IR [39]. In addition, glucose variability is associated with in-hospital complications and longer hospitalization following surgery [40–42] and with mortality in critically ill subjects [43].

Several studies have linked glycemic control to gut microbiome composition [36,44,45]. When we investigated the role of gut microbiome composition in our study, we found that *Eubacterium xylanophilum*, *Desulfovibrio*, *Terrisporobacter*, *Clostridium sensu stricto* and *Coprococcus* presented a positive effect on glycemic variability following the HP diet. This suggests that these genera might play an important role in improving host glucose homeostasis. For instance, *Coprococcus* might exert such a positive effect by its production of short-chain fatty acids (SCFA) [46]. Several studies have reported the benefits of SCFA in glucose homeostasis by regulating the blood glucose levels and glucose uptake [46]. Likewise, *Coprococcus* spp. might be able to metabolize the vitamins folate and biotin, which have been associated with lower plasma glucose levels [27]. On the other hand, a negative association was found with *Ruminococcus*, *Eggerthelia*, *Eubacterium hallii*, *Lachnoclostridium* and *Phascolarctobacterium*, suggesting that they might negatively impact glucose metabolism, and thus, type-2 diabetes. We have previously investigated the functional capabilities of *Lachnoclostridium spp*. and reported that *Lachnoclostridium* might metabolize both choline and phosphatidylethanolamine [27], which are precursors of trimethylamine

(TMA) and TMA-N-oxide (TMAO), thereby negatively regulating glucose metabolism and insulin sensitivity [37].

Our study benefits from a highly controlled nature; we can be confident in participants' adherence to the diet and its effect on the observed changes. However, we cannot infer any information about the long-term effects of either diet, as this study was only 21 days long, with participants spending 10 days on each diet. A study over a longer time course may reveal differences between the M and HP diets.

We also note some limitations. Our study has a small sample size with only 20 participants and a drop-out rate of 20%. Second, there was no wash-out period between the two diets. Third, study participants did not do an oral glucose tolerance test. Although the OGTT provides useful information about glucose tolerance, it does not infer on insulin sensitivity/resistance per se [47]. Moreover, under fasting conditions, basal insulin secretion determines a relatively constant level of insulinemia that will be lower or higher in accordance with insulin sensitivity such that hepatic glucose production matches whole body glucose disposal under fasting conditions. Thus, surrogate indexes based on fasting glucose and insulin concentrations, such as HOMA-IR, provide a greater reflection of primarily hepatic insulin sensitivity/resistance. Finally, only 11 participants provided stool samples for the entire duration of the study. Further work to investigate the role of gut microbiome composition and diversity in individual response to dietary intervention would be of considerable interest. Moreover, the study design lacked a wash-out period between dietary crossovers, as dietary effects may be brought about over a longer duration, some crossover effects of the previous diets may have been observed.

#### **5. Conclusions**

In conclusion, we find that the HP diet is more effective in reducing insulin resistance and in improving glycemic variability in morbidly obese women with pre-diabetes and have identified a panel of 10 microbes underlying the difference in glycemic variability between the two diets. Further investigation is required to elucidate the links between dietary interventions, the microbiome and clinical outcomes, as well as to identify measures that are predictive of individual response to intervention. Continued investigation of these interactions will contribute to the development of stratified intervention and prevention strategies for obesity and its associated health problems.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/nu13124380/s1, Table S1. Details on food provided with the two diets. Table S2. Difference between the two diets in glycemic exposure, control and variability. Figure S1. Effect of diet on the mean change from baseline of anthropometric measures, blood pressure, heart rate and lipids. Figure S2. Individual 24 h median sensor glucose profiles according to the type of diet.

**Author Contributions:** Conceptualization: A.M.V., C.I. and C.M. Project administration/recruitment: S.P.M., S.M. (Sabrina Maestrini), S.M. (Silvia Mazza), E.L. and M.S. Formal analysis: F.T. and V.B. Study coordination/supervision: M.S., C.I. and C.M. Resources: G.S.M. and A.R.-M. Writing original draft preparation: F.T., V.B., P.L., A.N., A.M.V. and C.M.; writing—review and editing: all. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is funded by the NIHR Nottingham BRC. C.M., A.N. and P.L. are funded by the Chronic Disease Research Foundation. AMV is supported by the National Institute for Health Research Nottingham Biomedical Research Centre.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of Instituto Auxologico Italiano.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Data for this study is deposited on Mendeley (Mendeley Data, V1, doi:10.17632/nsnm9tjrnt.1).

**Acknowledgments:** We thank all the participants for contributing and supporting our research.

**Conflicts of Interest:** AMV is a consultant for Zoe Global Ltd. (London, UK). All other authors declare no competing financial interests.

#### **References**


## *Article* **The Relationship between Macronutrient Distribution and Type 2 Diabetes in Asian Indians**

**Amisha Pandya \*, Mira Mehta and Kavitha Sankavaram**

Department of Nutrition and Food Science, University of Maryland, College Park, MD 20742, USA; mmehta@umd.edu (M.M.); kavitha@umd.edu (K.S.)

**\*** Correspondence: apdiabetesstudy@verizon.net; Tel.: +1-240-676-6594

**Abstract:** Asian Indians (AIs) are at increased risk for type 2 diabetes mellitus than other ethnic groups. AIs also have lower body mass index (BMI) values than other populations, so can benefit from strategies other than weight reduction. Macronutrient distributions are associated with improved glycemic control; however, no specific distribution is generally recommended. This study looks at whether a macronutrient distribution of 50:30:20 (percent of total calories from carbohydrates, fats, and protein) is related to diabetes status in AIs. Diet and Hemoglobin A1c (HbA1c) were assessed from convenience sample of AI adults in Maryland. A ratio of actual to needed calories using the 50:30:20 macronutrient distribution was then tested against diabetes status to identify associations. All groups except non-diabetic females, were in negative energy balance. The non-diabetic group consumed larger actual to needed ratios of protein than pre-diabetics and diabetics. However, all groups consumed protein at the lower end of the Acceptable Macronutrient Distribution Range (AMDR), and the quality of all macronutrients consumed was low. Therefore, weight loss may not be the recommendation for diabetes management for AIs. Increasing protein and insoluble fiber consumption, could play a critical role.

**Keywords:** macronutrient distribution; type 2 diabetes mellitus management; type 2 diabetes in Asian Indians immigrants to the US

#### **1. Introduction**

Asian Indians (AIs) in India as well as those who have emigrated to the United States and other western nations are seeing an increase in incidence and prevalence of type 2 diabetes mellitus (T2DM) [1]. Prevalence of T2DM among AIs is estimated at 9.3% across India [2], and 18% in the US [3]. In one Indian State, Gujarat, estimates are 7–14% [4]. Rates of undiagnosed T2DM in AIs in India as well as in other countries is estimated to be approximately 50% [5].

There are complex and poorly understood reasons for the increasing incidence and prevalence of T2DM in AIs, including pathophysiological and sociocultural characteristics specific to this population. AIs have unique physical attributes and cultural attitudes that increase their risk for T2DM compared to other ethnic groups [6]. For example, two commonly cited factors for insulin resistance (IR), a precursor to the development of T2DM, are obesity, as measured by body mass index (BMI), and adverse fat distribution, neither of which seem to be associated with high basal insulin levels in this population [7]. AIs have younger onset for T2DM, and lower BMI values as compared with other populations [1,8–11]. The age of onset for T2DM in AIs is estimated to occur 10 years earlier than in Europeans, and AIs require lower BMI cut-offs for effective identification of T2DM risk [5]. Additionally, AIs may be predisposed to IR and T2DM because AI children are born smaller, have more fat, and less lean muscle [6]. Reduced lean muscle mass at birth is significant because lean muscle mass contains more mitochondria than fat tissue and so is more metabolically efficient. However, because the secretion of insulin is triggered by adipose tissue upon consumption of food, individuals with higher adipose to lean muscle

**Citation:** Pandya, A.; Mehta, M.; Sankavaram, K. The Relationship between Macronutrient Distribution and Type 2 Diabetes in Asian Indians. *Nutrients* **2021**, *13*, 4406. https:// doi.org/10.3390/nu13124406

Academic Editors: Silvia V. Conde and Fatima O. Martins

Received: 12 November 2021 Accepted: 7 December 2021 Published: 9 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

mass ratios may have increased blood insulin levels, which can trigger a negative feedback response resulting in insulin receptor dysfunction leading to IR and T2DM.

The Asian Indian diet is also high in carbohydrates, and with urbanization and migration, there has been a growing tendency towards processed, refined and higher fat convenience foods, coupled with decreases in physical activity seen both in AIs living in India and abroad [12,13].

The purpose of this study is two-fold: (1) to describe the dietary intake of Asian Indian adults with and without T2DM, and (2) to determine whether there is an association between diabetes status and diet indicative of T2DM. Dietary intake was examined as both dietary quantity as well as dietary quality. Dietary quantity was measured by macronutrients (carbohydrates, fat, and protein) as a proportion of total kilocalories. Dietary quality was indicated by consumption of soluble and insoluble fiber, cholesterol, saturated fat, trans fat, and sugar; both were assessed relative to diabetes status.

#### **2. Materials and Methods**

#### *2.1. Study Population*

A convenience sample of 59 AI adults from Mangal Mandir, a Hindu temple in the Baltimore/Washington Metropolitan Area, was taken over a period of three months. Subjects included AI adults ≥ 18 years of age, literate in English, and residents of the US for >5 years. Participants' demographic information such as age, gender, education, income and number of years in US were collected.

#### *2.2. Data Collection*

Diabetes status was assessed by hemoglobin A1c readings obtained via physician ordered lab results during a health fair, run by community physicians, and held at the Mangal Mandir temple; an event used to initiate the study. For study participants with no physician's diagnosis of diabetes or with pre-diabetes, physician ordered labs for fasting blood glucose and hemoglobin A1c were not warranted and therefore were not available. Glucose and HbA1c point of care (POC) monitors were used for those that did not have fasting blood glucose and HbA1c readings from their physicians to confirm diabetes status. Cut-offs used for no-diabetes, pre-diabetes and diabetes are those widely accepted by the American Diabetes Association (ADA), the World Health Organization (WHO), and the International Diabetes Federation (IDF); >5.7% = no diabetes, ≥5.7% but <6.5% = pre-diabetes, and ≥6.5% = diabetes. Hemoglobin A1c level was chosen as the exclusive diagnostic variable because it was the most reliable measure of diabetes status. The collection of fasting blood glucose after 8 h of fasting was found not to be feasible and could not always be obtained from participant physician ordered lab results (about 13% of participants were missing this data point). Therefore, fasting blood glucose was dropped as a diagnostic variable. The accuracy of Bayer's A1cNow+ monitor was confirmed for non-clinical diagnoses of diabetes with up to 98% agreement, or non-significant difference, between the device and laboratory results [14–18]. Participants also provided self-report of their diabetes status which was compared to their HbA1c to determine the rate of undiagnosed diabetes in this population.

All participants completed a 163-item food frequency questionnaire (FFQ) validated with AIs that gave a one-year retrospective to their dietary intake [19,20]. Diet quantity was measured by the proportion of total kilocalories consumed daily in the form of the three energy producing macronutrients, carbohydrates, fats, and proteins contributing to metabolic efficiency and examined by diabetes status. Diet quality was assessed by the consumption of harmful fats such as saturated and trans-fat, dietary cholesterol, and fiber content of carbohydrates as seen in consumption of insoluble and soluble fiber. Basal metabolic rate (BMR) calculations were performed using five different methods, all varied slightly but they each generally followed the same trend by diabetes status.

The initial selection of BMR equations was based on methods used in various studies. There was consideration given to equations applied to Asian Indians in India, however, their use over other methods was not validated in the literature [21–23]. There are four equations commonly cited in the literature. They are the Owen, Mifflin-St. Jeor, Harris Benedict, and the WHO/FAO/UNU; there is varying agreement as to which method is the most reliable [24–27]. The BMR provided by the Tanita Scale (BC-558 Ironman Segmental Body Composition Monitor) used to capture bioelectrical impedance measures was added to previously mentioned four methods. Each method has its predictive variability, and none can be validated as the most accurate predictor in Asian Indian populations. Thus, all were considered in the analysis and correlational analysis confirmed that all five methods were significantly correlated with one another (*p* < 0.0001). For simplicity, the Owen method was selected for the remainder of the analyses. BMR establishes the minimum caloric intake needed to meet energy requirement assuming no physical activity. Energy requirements for each participant were calculated (BMR + kilocalories burned through physical activity).

#### *2.3. Macronutrient Distribution*

Although the United States Institute of Medicine's (IOM's) Acceptable Macronutrient Distribution Ranges (AMDR) offers recommended ranges of macronutrient intake, no specific recommendations exist to achieve optimal metabolic efficiency, however, there is some evidence that suggests that a macronutrient distribution of 50:30:20 (percent of total calories from carbohydrates, fats, and protein) can be metabolically favorable [28–31]. To establish a reference value for macronutrient distribution, total caloric energy requirement as determined from BMR + kilocalories burned through physical activity, was multiplied by a standard recommended proportion of total kilocalories for each macronutrient (50 percent of total kilocalories from carbohydrates, 30 percent of total kilocalories from fat, and 20 percent of total kilocalories from protein) to obtain the daily required grams of each macronutrient. A ratio of actual to needed kilocalories from each macronutrient was then calculated by dividing daily intake of macronutrients (carbohydrate, fat, and protein) by the daily required grams of each macronutrient. These ratios of actual to needed carbohydrates, fats and proteins were then tested against diabetes status to identify associations.

#### *2.4. Statistical Analysis*

Descriptive statistics such as mean, standard deviations and correlational analyses were used to describe the data and establish covariance for any variables. Univariate analyses of variance, including *t*-tests, ANOVAs, and linear and logistic regression models were used to determine the association of diet across diabetes status groups. Multiple linear logistic regression was used to determine the relationship between diet and diabetes status. Those participating as married couples, warranted the data being examined to determine any effect this may have had on variable confounding. A correlational analysis was performed between males and females of participant couples.

#### **3. Results**

#### *3.1. Participant Demographics*

Fifty-nine individuals initially expressed interest in participating in the study, and 39 participants completed the study (power = 0.76). Study participants were about equally divided by gender (49% male and 51% female) with an average age of 65.2 years (67.4 years for males and 63.0 years for females). About 72% of participants were married. Participants were predominantly Gujarati immigrants (95%) who had lived an average of 37 years in the United States (39 years for males and 35 years for females). Ninety-seven percent of all participants were born in India or Africa, and 3% were born in the United States. Almost two thirds (61.5%) of participants had earned a bachelor's degree or less and 38.5% had earned post graduate or professional degrees. Participants were almost equally split in terms of household income, with 56.4% earning less than or equal to USD 100,000 annually and 43.6% earning more than USD 100,000 annually. Almost three quarters of participants were vegetarian (58% of males and 90% of females, *p* < 0.0310). About 90% of participants, both male and female, consumed alcohol occasionally (1–2 times/month) or never, and most participants did not have a history of smoking (92% overall, 100% for females, and 84% for males). About 53% of males and 70% of females reported primarily being responsible for grocery shopping in their households, whereas most females were predominantly responsible for cooking (0% of males, and 90% of females, *p* < 0.0001).

#### *3.2. Diabetes Status*

Participant self-report was compared to HbA1c groupings for both diabetes and prediabetes to determine the rate of undiagnosed diabetes in this population. Although, 71% of participants correctly self-reported having diabetes and 88% self-reported having prediabetes, there were 12% undiagnosed for diabetes and 39% undiagnosed for pre-diabetes. Overall, there were 39% undiagnosed for diabetes and prediabetes, by gender that broke down to be 26% of males and 50% of females.

Participants were 18% non-diabetic (10% male and 8% female), 49% were pre-diabetic (13% male and 36% females), and 33% were diabetic (26% male and 8% female); the difference by gender across these three diabetes status groups is significant (F = 0.0017, *p* = 0.0197). Figure 1 shows the breakout of diabetes status group by gender.

**Figure 1.** Participants by diabetes status and gender (T2DM = Type 2 Diabetes Mellitus).

#### *3.3. Participant Diets*

Participants consumed an average of 1463.7 kilocalories per day (1600.9 for males and 1333.3 for females, *p* = 0.01). Figure 2 shows the macronutrient intake per day (in grams) for all participants as well as for males and females separately. Daily consumption of protein averaged 51.4 g across all participants (46.8 g for females and 56.4 g for males), carbohydrates averaged 201.3 g across all participants (181.3 g for females and 222.4 g for males), and fats averaged 49.5 g across all participants (46.4 g for females and 52.9 g for males).

Among married participants, couples consumed similar diets, however, participants were not significantly correlated on diabetes status. Therefore, comparisons by diabetes status or gender would still reveal real differences between Asian Indian male and female non-diabetics, pre-diabetics and diabetics.

Although males and females consumed significantly different amounts of total kilocalories per day, they did not differ significantly when each macronutrient was examined as the percent of total kilocalories, or when macronutrient components were examined as proportion of their respective macronutrient. Protein, as a percent of total kilocalories, was about 14% for both males and females, carbohydrates, as a percent of total kilocalories, was 56% for males and 55% for females, and fat, as a percent of total kilocalories, was 29% for males and 31% for females.

Although the percentages of total kilocalories from protein, carbohydrates, and fats for both males and females fell within acceptable macronutrient distribution ranges for daily consumption by adults in the US as recommended by the Food and Nutrition Board, Institute of Medicine, National Academies [32], protein as percent of total kilocalories was at the lower recommended ranges for daily consumption. Table 1 provides a comparison of the IOM's AMDR with participant daily intakes. Participants had lower than recommended daily ranges for total fiber consumption (16.10 g for females and 19.00 g for males, IOM recommendations ranged 21–25% for females and 30–38% for males) and dramatically less than the maximum allowances for sugar (<1% for both males and females; IOM maximum allowance is 25%). No recommendations were given by the IOM for monounsaturated fatty acids and the nutrient data for this study did not break out polyunsaturated fatty acids into n-6 (linoleic acid) and n-3 (α-linolenic acid), so those comparisons were not possible. Saturated fat and trans fatty acids were consumed in small quantities by participants (<1% and <<1% respectively), and cholesterol was consumed in relatively small quantities as well (5.35% of total kilocalories for females, and 6.63% for males).

Looking at these dietary components by diabetes status reveals that there was a significant difference between the proportion of trans fat consumed of total fat across groups. Non-diabetics consumed the largest proportion (1%), pre-diabetics consumed a smaller proportion (0.2%), and diabetics consumed the smallest proportion of all (0.08%), *p* = 0.036\*. There was a difference in the consumption of cholesterol as a proportion of total fat, but it was not significant (*p* = 0.2225). Non-diabetics again consumed the largest proportions (2.67) followed by diabetic (1.85), and then pre-diabetic (1.41), *p* = 0.0636. These differences were not seen when looking at males alone, but were when looking at females alone, albeit not significantly. Non-diabetic females consumed 1% of total fat from trans-fat, whereas pre-diabetic consumed 0.1%, and diabetic consumed 0.03%, *p* = 0.0522. For proportion of cholesterol, non-diabetic females consumed the most (3.03), followed by pre-diabetic females (1.40), and then diabetic females (0.74), *p* = 0.0867.


**Table 1.** Dietary Components as Compared to IOM's Acceptable Macronutrient Distribution Ranges (AMDR).

IOM: Institute of Medicine.

#### *3.4. BMR by Diabetes Status*

Table 2 provides the mean BMR by gender and diabetes status. BMR establishes the minimum caloric intake needed to meet energy requirement assuming no physical activity. Energy requirements for each participant were calculated (BMR + kilocalories burned through physical activity).

**Table 2.** Basal Metabolic Rate (BMR) Calculated Using Five Different Methods by Gender and Diabetes Status.


Figures 3–5 show the average BMR, average caloric intake needed to meet energy requirements and average caloric intake by diabetes status and gender. Correlational analysis shows that BMR is significantly correlated with total caloric intake for all participants (0.4251, *p* = 0.0070), but not for males only (0.0334, *p* = 0.8920), or females only (0.1088, *p* = 0.6478). However, BMR was significantly correlated with total caloric intake needed to meet energy requirements for all participants, males only and females only (0.7799, *p* < 0.0001; 0.6738, *p* = 0.0016; 0.6865, *p* = 0.0008, respectively). Total caloric intake to meet energy needs significantly exceeds total kilocalories consumed for all participants as well as for males only and females only (0.3015, *p* = 0.0625; 0.0444, *p* = 0.8569; 0.2550, *p* = 0.2779, respectively), except for female non-diabetics, whose caloric intake exceeds total kilocalories needed for energy needs.

**Figure 3.** BMR, energy requirement, and caloric intake for all participants by diabetes status.

**Figure 4.** BMR, energy requirement, and caloric intake for female participants by diabetes status.

#### *3.5. Diet Quantity—Macronutrient Distributions—Actual to Needed Calorie Ratios*

Kilocalories required to meet energy needs exceeded caloric intake for all groups, except non-diabetic females; caloric intakes for diabetic females were slightly higher than what is needed to meet energy needs.

The association between macronutrient distribution and diabetes status was examined by looking at the ratio of actual to needed total kilocalories, protein, carbohydrates and fats based on energy needs by gender and diabetes status (Figures 6–9).

There were no significant differences noted by diabetes status for total actual to needed kilocalories (Figure 6). However, non-diabetic females did exceed their total needed caloric intake (105%), whereas pre-diabetic and diabetic females consume 81% and 80%, respectively, of their needed kilocalories. Actual to needed ratios of total kilocalories were lower for non-diabetic men (78%) than pre-diabetic (81%) and diabetic men (80%).

Figure 7 shows non-diabetics consuming higher ratios of actual to needed protein than pre-diabetics and diabetics (64%, 59% and 60%, respectively); this difference was not statistically significant (*p* = 0.8450). Non-diabetic females consumed higher ratios of actual to needed protein than pre-diabetic and diabetic females (83%, 61%, and 57%, respectively); this difference was not statistically significant (*p* = 0.0699). Male diabetics consumed the highest ratios of actual to needed protein (61%), followed by pre-diabetics (55%) and then non-diabetics (50%); this difference was not statistically significant (*p* = 0.6786).

**Figure 7.** Actual to needed kilocalories from protein by diabetes status and gender.

**Figure 8.** Actual to needed kilocalories from carbohydrates by diabetes status and gender.

**Figure 9.** Actual to needed kilocalories from fat by diabetes status and gender.

Correlational analysis also showed a significant relationship between diabetes status and actual to needed kilocalories from protein for females (−0.45, *p* = 0.0458). This

relationship was not seen in males; however, correlational analysis did show a significant relationship for males only between HbA1c level and actual to needed kilocalories from protein (0.51, *p* = 0.0243); this relationship is in the reverse direction than seen in females.

Figure 8 shows a similar pattern for actual to needed kilocalories from carbohydrates as shown for actual to needed consumption of protein. Non-diabetic females were consuming higher ratios of carbohydrates than pre-diabetics (111% and 96%, respectively); diabetic females consumed the lowest ratios of carbohydrates (90%). The pattern for men is slightly different. Pre-diabetic males consumed the highest ratio of actual to needed carbohydrates (94%), followed by non-diabetic (88%) and finally diabetic males (87%).

Figure 9 shows actual to needed consumption of fat was highest for non-diabetic females, exceeding the daily caloric needed for fat (111%). Pre-diabetics consumed the second highest percent (91%), followed by diabetics (86%). For males, however, diabetics consumed the highest percent of actual to needed fat (81%), followed by non-diabetics (80%) and then pre-diabetics (77%).

#### *3.6. Diet Quality*

Additional findings from the correlational analysis showed a significant relationship between diabetes status and proportion of trans fats (−0.34034, *p* = 0.034) and a weak correlation between diabetes status and proportion of insoluble fiber (0.30208, *p* = 0.0616) for all participants, and a significant correlation between diabetes status and proportion of insoluble fiber (0.45653, *p* = 0.043) for females only. However, when correlations between HbA1c, macronutrients and other dietary components were examined, percent protein (0.33328, *p* = 0.0381), proportion of soluble fiber (0.31638, 0.0497), and total kilocalories consumed were significantly correlated with HbA1c levels for all participants. Males did not show any specific significant correlations between diabetes status and any macronutrient, however, when looking at the correlations between HbA1c, macronutrients and other dietary components, percent protein (0.50846, *p* = 0.0262), and proportion of insoluble fiber (0.50401, *p* = 0.0278), were significantly correlated, while proportion of soluble fiber was weakly correlated with diabetes status for males (0.45357, *p* = 0.0511).

#### *3.7. Predicting Diabetes Status*

A multiple linear regression model was performed to test how well macronutrient independent variables, that were correlated with diabetes status, could predict diabetes status. With diabetes status as the dependent variable the initial model included actual to needed kilocalories from protein, actual to needed kilocalories from carbohydrates, actual to needed kilocalories from fat, proportion of trans fat, proportion of cholesterol, proportion of soluble and insoluble fiber, percent protein as independent variables. The resulting model was significant (F = 2.58, *p* = 0.0282). All independent variables were significant predictors of diabetes status, except proportion of trans fat and proportion of soluble fiber. However, when these variables were removed from the model, the model was no longer significant, so these variables were retained in the overall model. Table 3 provides the test statics for this multiple regression model. Table 4 shows the results when the same independent variables were tested to predict HbA1c level, the overall model was also significant (F = 4.39, *p* = 0.0013). The model's performance was improved by removing proportion of cholesterol from the model. The result was a significant regression for the remaining variables (F = 2.66, *p* = 0.0012), with an R2 of 0.4165.


**Table 3.** Parameter Estimates for Predicting Diabetes Status.

\* Indicates a significant *p*-value (*p* < 0.05).



\* Indicates a significant *p*-value (*p* < 0.05); < indicate *p* < 0.005; << indicates *p* < 0.0005.

#### **4. Discussion**

This study established an association between diet as measured by actual to needed macronutrients and diabetes status. Pre-diabetics and diabetics consumed lower ratios of actual to needed protein, carbohydrates and fats relative their energy requirements than non-diabetics. This is consistent with recommendations normally given to patients with IR, pre-diabetes or diabetes; eat less to lose weight [33]. In addition, participants with pre-diabetes and diabetes ate smaller proportions of insoluble and higher proportions of soluble fiber. Consistent with typical AI diets, refined carbohydrates are preferred over whole grains, however, non-diabetics consumption of higher amounts of insoluble fiber may suggest that carbohydrate quality may be beneficial in this population.

This study also shows that Asian Indian pre-diabetics and diabetics recognize the benefit of physical activity and caloric restriction, as they relate to their diabetes status and as recommended for diabetes management by health care providers; and establishes a pattern of decreased overall caloric intake, and decreased intake for each macronutrient (protein, carbohydrate and fat), below required levels based on energy requirements for prediabetics and diabetics in the United States. Additionally, it is not known if the combination

of increased physical activity and decreased caloric intake discovered in this population can be viewed as contributing to health or negatively impacting diabetes status.

Understanding associations between diabetes status and dietary intake related to diabetes may provide evidenced based strategies to reduce risk of T2DM in this population. As would be the norm in traditional Indian families, females in this population were responsible for cooking and had control over dietary consumption for themselves as well as their families. Additionally, as 72% of participants were married couples, it was noted that of those participants, 70% of the females did not have diabetes or had pre-diabetes while their spouses had diabetes. Along with females being responsible for the household cooking, this suggests a level of control over dietary options that may have an impact on their spouses' diabetes status.

Contrary to the normal pattern of low physical activity among AIs, levels of physical activity by diabetes status in this study population suggest participants may be recognizing the need for physical activity as a strategy to manage diabetes.

These findings suggest that although AIs may be implementing strategies typically recommended to patients with IR, pre-diabetes and diabetes, those strategies may not be improving disease status in this population. Perhaps a focus on macronutrient balance with sufficient caloric intake to meet energy needs, with increased protein intake, and reduced intake of highly refined carbohydrates in this population would be more effective. A recent comparison of various diets, such as the Mediterranean, Dietary Approaches to Stop Hypertension (DASH), plant-based, low and very low carbohydrate, low-fat, and high protein, aimed at improving metabolic syndrome, T2DM, Cardiovascular Disease (CVD), and hypertension, illustrates that a balance of both macronutrient proportions and quality are needed. In particular, it can be inferred that the macronutrient proportions studied here (50:30:20) provide a starting point from which minor adjustments can be made for individuals' management of T2DM. In addition, the quality of those macronutrients play an important role (e.g., whole grains as compared to processed carbohydrates, fresh fruits and vegetables, plant-based proteins, and unsaturated fats) [31,34]. Additionally, because carbohydrates and fats are primarily implicated in metabolic disorders, studies vary fat and carbohydrate intake (high fat: low carbohydrate vs. low fat: high carbohydrate) and hold protein intake constant. These studies generally show little difference between these two intake conditions [35]. As suggested in this study, in terms of macronutrient proportions, the key may be in moderating carbohydrates and fats, while adjusting protein intake, as well as ensuring quality of macronutrients consumed. Similarly, one additional study looking at regional differences in dietary patterns in Asian Indians in India, suggests that this strategy may be the most beneficial in the management of T2DM [36].

Notable limitations of this study include that generalization of findings to all AIs, or Gujaratis, was not possible because this study used a convenience sample, and most participants were from a specific cohort of AIs; first generation older Gujarati adults with high levels of education, income and acculturation. Additionally, the cross-sectional design of the current study only allowed for examination of associations between diet and diabetes status, and do not imply causality.

Notable strengths of this study include that this is the first study to examine and identify a relationship between diabetes status and macronutrient composition; suggesting a greater adherence to an optimal macronutrient composition (50:30:20; for percent of total kilocalories from carbohydrates, fat and protein) in non-diabetics than in pre-diabetics or diabetics.

The prevalence of type 2 diabetes and its associated co-morbidities within AI populations, especially Gujaratis, in the US and elsewhere is at extremely high levels and is continuing to rise. As such, further research in this population is needed to more thoroughly study the relationship between diet quality and diabetes status to develop evidence-based strategies helpful in the prevention and management of diabetes in this high-risk population. Specifically, further examination is needed on whether the 50:30:20 macronutrient distribution with caloric intakes consistent with energy needs and improved

macronutrient quality has an impact on diabetes status and whether adjustments to diet in these ways could impact diabetes related outcomes.

#### **5. Conclusions**

AIs are at high risk for insulin resistance (IR) leading to impaired glucose tolerance (IGT) and T2DM and their sequelae. Future research is needed to establish the association between diet quality and diabetes status; however, the findings of this study suggest that those at risk for T2DM may benefit from adhering to a macronutrient distribution of approximately 50:30:20 percent of total kilocalories from carbohydrates, fat and protein, or from a higher intake of dietary protein, while ensuring the quality of those macronutrients. Finally, the implications of near universal female responsibility for cooking, in households where a spouse or other family members are diagnosed with either pre-diabetes or diabetes, are intriguing. Targeted education to AI females about preparing meals that adhere to optimal dietary choices, may have a potential benefit to the entire family including those with diabetes and prediabetes.

The current study is the first to investigate a relationship between the consumption of macronutrients in general and in specific relative proportion (50:30:20) and diabetes status in an AI population. These relationships can be further explored to develop a recommended diet for AIs that is quantitatively and qualitatively supportive metabolically for those with IR, pre-diabetes or diabetes.

#### **6. Study Limitations**

This study lacked sufficient funding to enable laboratory testing to establish hemoglobin A1c levels for group assignment. Therefore, the study leveraged an annual health fair event at the data collection site which provided for this laboratory testing at no cost to the participants. In cases where study participants did not participate in the health fair or their diabetes status did not indicate laboratory testing by a physician, Hemoglobin A1c levels were supplemented by use of the Bayer A1cNow+ POC monitor for group assignment by diabetes status.

**Author Contributions:** This research was conducted by A.P. as part of a graduate thesis project. M.M. advised and assisted in all aspects of the project. K.S. provided support on various aspects of the project as well. Conceptualization, A.P., M.M., and K.S.; methodology, A.P and M.M.; software, A.P.; validation, A.P.; formal analysis, A.P.; investigation, A.P.; resources, A.P.; data curation, A.P.; writing—original draft preparation, A.P.; writing—review and editing, M.M., A.P., K.S.; visualization, A.P.; supervision, M.M.; project administration, A.P.; funding acquisition, M.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding. The University of Maryland provided internal funds for publication.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of The University of Maryland (771334-3, 3 June 2016).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available upon request from the corresponding author.

**Acknowledgments:** This research project would not have been possible without the interest, support and participation of the leaders and community of the Mangal Mandir in Spencerville, Maryland, the consultation, support, and nutrient analysis services provided by the Population Health Research Institute (PHRI) located in Canada, the support and understanding of my loving family, and support from my advisor and committee. I wish to thank the religious leadership of the temple as well as all of the volunteer physicians providing free health screens and laboratory analysis for members of the AI community for all their time, encouragement, support and facilitation in recruiting participants for my study as well as providing a venue to administer my study. I wish to thank the Population Health Research Institute (PHRI) for providing both permission for me to use the Food Frequency Questionnaire (FFQ) for South Asians (SA), which PHRI adapted from the Study of Health Assessment and Risk in Ethnic groups (SHARE) FFQ instrument, and provision of the nutrient analysis for all participants who completed the FFQ-SA in my study. I wish to thank Dipika Desai for her consult and guidance through the process of instituting the student research data use agreement, which provided me the use of the instrument and the nutrient analysis at no charge, and Karleen Schulze, who provide both consult and guidance in the delivery and interpretation of the nutrient data files as well as consult on use of the FFQ to calculate Healthy Eating Index scores. I wish to thank Roche Diagnostics for donating glucometers and test strips for my study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Effects of Different Types of Carbohydrates on Arterial Stiffness: A Comparison of Isomaltulose and Sucrose**

**Ryota Kobayashi 1,\*, Miki Sakazaki 2, Yukie Nagai 2, Kenji Asaki 3, Takeo Hashiguchi <sup>4</sup> and Hideyuki Negoro 5,6**


**Abstract:** Increased arterial stiffness during acute hyperglycemia is a risk factor for cardiovascular disease, but the type of carbohydrate that inhibits it is unknown. The purpose of this study was to determine the efficacy of low-glycemic-index isomaltulose on arterial stiffness during hyperglycemia in middle-aged and older adults. Ten healthy middle-aged and older adult subjects orally ingested a solution containing 25 g of isomaltulose (ISI trial) and sucrose (SSI trial) in a crossover study. In the SSI trial, the brachial–ankle (ba) pulse wave velocity (PWV) increased 30, 60, and 90 min after ingestion compared with that before ingestion (*p* < 0.01); however, in the ISI trial, the baPWV did not change after ingestion compared with that before ingestion. Blood glucose levels 30 min after intake were lower in the ISI trial than in the SSI trial (*p* < 0.01). The baPWV and systolic blood pressure were positively correlated 90 min after isomaltulose and sucrose ingestion (*r* = 0.640, *p* < 0.05). These results indicate that isomaltulose intake inhibits an acute increase in arterial stiffness. The results of the present study may have significant clinical implications on the implementation of dietary programs for middle-aged and elderly patients.

**Keywords:** arterial stiffness; glucose ingestion; middle-aged and older patients; isomaltulose; sucrose

#### **1. Introduction**

Previous studies have reported acute hyperglycemia as an independent risk factor for cardiovascular disease [1]. An increased postprandial blood glucose level is also a risk factor for cardiovascular disease and exerts a greater effect than the fasting blood glucose level [2]. Moreover, increased arterial stiffness owing to impaired vascular endothelial function underlies the increased risk of cardiovascular disease in acute hyperglycemia [3]. Gordin et al. [4] suggested that arterial stiffness increased with increasing postprandial blood glucose levels in healthy middle-aged and older individuals. Furthermore, we previously demonstrated that systemic arterial stiffness increased in middle-aged and older people after the ingestion of a 25 g glucose solution [5]. Since postprandial blood glucose increases with age [6] and arterial stiffness progresses, there is a significant relationship between arterial stiffness and postprandial blood glucose levels [7]. Japan has a superaging society [8]; it is important to control the progression of arterial stiffness during acute hyperglycemia in older Japanese individuals.

The increase in arterial stiffness occurs immediately after food intake [7] and may be influenced by the glycemic index (GI) value of the cardiovascular disease indices [9]. In fact, vascular endothelial function, the underlying mechanism of arterial stiffness, varies with GI [10]. A previous study reported that arterial stiffness increased in middle-aged and older people consuming a glucose solution, but the changes differed between different

**Citation:** Kobayashi, R.; Sakazaki, M.; Nagai, Y.; Asaki, K.; Hashiguchi, T.; Negoro, H. Effects of Different Types of Carbohydrates on Arterial Stiffness: A Comparison of Isomaltulose and Sucrose. *Nutrients* **2021**, *13*, 4493. https://doi.org/ 10.3390/nu13124493

Academic Editors: Silvia V. Conde and Fatima O. Martins

Received: 10 November 2021 Accepted: 12 December 2021 Published: 15 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

carbohydrate intakes [5]. A high-GI diet has been shown to increase arterial stiffness compared with a low-GI diet. In other words, a low-GI diet may reduce the acute adverse effects on arterial stiffness [11]. For example, isomaltulose is a natural carbohydrate found in honey that has a low GI and is certified as a novel food (European Food Safety Authority) because of its nutritional quality [12]. Isomaltulose has similar amounts of sweetness and energy to sucrose; however, the rate at which it is broken down in the small intestine is slower than that of sucrose, which moderates the rise in blood glucose levels after ingestion [13]. In a previous study, a comparison of acute changes in blood glucose levels after the ingestion of isomaltulose or sucrose in 10 healthy subjects showed that the highest blood glucose levels were lower in patients who ingested isomaltulose than in those who ingested sucrose [14]. In addition, when 10 patients with type 2 diabetes were asked to consume either isomaltulose or sucrose and the changes in blood glucose levels after ingestion were examined, it was found that the blood glucose levels rose rapidly after sucrose ingestion and increased gradually after isomaltulose ingestion, with the peak values being lower for isomaltulose than for sucrose [13]. In other words, isomaltulose, which has a lower GI than sucrose, is expected to reduce the increase in arterial stiffness. However, the changes in arterial stiffness after ingestion of isomaltulose compared with that after the ingestion of sucrose are not sufficiently clear. Therefore, it is necessary to investigate whether arterial stiffness is altered after isomaltulose intake in healthy middle-aged and older people.

In this study, we hypothesized that sucrose intake will increase arterial stiffness with increasing blood glucose levels, but isomaltulose intake will not influence arterial stiffness. To test this hypothesis, we investigated the acute effects of isomaltulose and sucrose intake on arterial stiffness.

#### **2. Materials and Methods**

#### *2.1. Participants*

The participants were 10 healthy middle-aged and older adults (five men and five women). Participants were recruited by distributing flyers for research cooperation with residents of the Teikyo University of Science. Finally, we received 20 applications, from which we selected 10 participants who met the following conditions. All participants were normotensive (Japanese standard: <140/90 mmHg), non-smokers, no obvious disease on electrocardiogram or other diagnostic tests, and no exercise habit before the study according to the physical activity questionnaire. Patients with abnormalities in blood tests, urine tests, chest radiographs, or electrocardiograms in the year prior to the study; with diabetes mellitus (American Diabetes Association/ European Association for the Study of Diabetes diagnostic criteria); and who had problems with exercise (e.g., those with musculoskeletal injuries) were excluded from the study. This study was conducted in compliance with the Declaration of Helsinki in terms of ethics, human rights, and protection of participants' personal information. Ethical approval for this study was obtained from the Ethics Committee of Teikyo University of Science (approval no. 20A013). In addition, this study was registered with the University Hospital Medical Information Network Center (UMIN Center; Study No. UMIN000041622). All hardcopy (paper) study data were stored in a locked filing cabinet, and electronic data were stored on a secured network drive, accessible only to those working in the laboratory. The study was conducted in accordance with the guidelines for human experimentation published by the Institutional Review Board.

#### *2.2. Study Design*

The participants were 10 healthy middle-aged and older adults. They were instructed to maintain a normal diet and activities of daily living for the duration of the study. Intense exercise (training and activities of daily living), caffeine, and alcohol consumption were prohibited for 24 h prior to the experiment. Fasting (10–12 h) was started at 9:00 p.m. the day before the start of the experiment. Arterial stiffness, blood pressure (BP) at the

level of the brachial artery and at the ankle, heart rate (HR), and blood glucose (BG) levels were measured before (baseline) and 30, 60, and 90 min after 25-g isomaltulose or sucrose loading. Before each measurement, the subjects were asked to rest in a supine position (Figure 1).

**Figure 1.** Study design. Arterial stiffness, BP, HR, and BG levels were measured at baseline and at 30, 60, and 90 min after isomaltulose or sucrose ingestion. The participants rested in a supine position for 10 min before the test. BP, blood pressure; HR, heart rate; BG, blood glucose.

#### *2.3. Body Composition*

Height was measured using a height meter in increments of 0.1 cm. Body weight, body fat percentage, and body mass index (BMI) were measured in 0.1 kg increments using a precision instrument body-composition analyzer (WB-150 PMA, Tanita, Tokyo, Japan).

#### *2.4. Arterial Stiffness*

Pulse wave velocity (PWV) at the brachial and ankle (ba) and at the brachial and heart (hb) of all participants was measured using an automated oscillometric device (PWV/Ankle Brachial Index (ABI), Colin Medical Technology, Komaki, Japan) as previously described [15]. All measurements were performed in a supine position in a quiet room at baseline and 30, 60, and 90 min after isomaltulose solution and sucrose solution ingestion. The daily coefficients of variation in our laboratory were 3 ± 1% and 3 ± 2% for baPWV and hbPWV, respectively.

#### *2.5. Upper Arm and Ankle Blood Pressure*

Systolic blood pressure (SBP), mean blood pressure (MBP), diastolic blood pressure (DBP), and pulse pressure (PP) of the upper arm and ankle were measured in the supine position using an automated oscillometric PWV/ABI device (Omron Colin, Tokyo, Japan) over the brachial and posterior tibial arteries [15]. All measurements were performed in the supine position in a quiet room at baseline and 30, 60, and 90 min after isomaltulose solution and sucrose solution ingestion. The coefficients of variation per day in our laboratory were 2 ± 1% and 2 ± 2% for brachial blood pressure and ankle blood pressure, respectively.

#### *2.6. Heart Rate*

HR was measured in the supine position using an automated oscillometric PWV/ABI device (Omron Colin, Tokyo, Japan) [15]. All measurements were performed in the supine position in a quiet room at baseline and 30, 60, and 90 min after the ingestion of isomaltulose and sucrose solutions. The coefficient of variation per day in our laboratory was 2 ± 1%.

#### *2.7. Blood Glucose*

Venous blood was collected from the participants' left fingertips. Blood glucose levels were measured by the flavin-adenine dinucleotide glucose dehydrogenase method using a Glutest Neo Alpha glucometer (Sanwa Kagaku Kenkyusho, Tokyo, Japan) [16]. Measurements were taken before and 30, 60, and 90 min after ingestion of isomaltulose and sucrose solutions. The interday coefficient of variation of blood glucose levels was 3 ± 1%.

#### *2.8. Isomaltulose Solution and Sucrose Solution Ingestion*

Each participant orally ingested 25 g of isomaltulose (ISI trial) or 25 g of sucrose (SSI trial) in 200 mL of water within 5 min, since the new World Health Organization guidelines recommend that adults consume less than 25 g of free sugars per day [17]. Each subject waited approximately 3 days after the completion of one test before taking the next test.

#### *2.9. Statistical Analysis*

Data are presented as means ± standard deviation. Normality of the data and homogeneity of variance were examined using the Shapiro–Wilk and Levene tests, respectively. Changes in each measurement before and after the intervention are presented as mean values and 95% confidence intervals for each group. Parametric analysis was performed using two-way analysis of variance with repeated measures (time\*group) for the measurements taken. When the assumption of sphericity was violated (Mauchly's test), the analysis was adjusted using the Greenhouse–Geisser correction. The Bonferroni method was used with post hoc tests for changes in each intervention. The total area under the curve at 90 min (AUC) was calculated using the trapezoidal formula and analyzed using the corresponding *t*-test. The correlation between baPWV and brachial SBP levels 90 min after consumption was examined using the Pearson product-moment correlation coefficient. SPSS (version 25, IBM Corp., Armonk, NY, USA) was used for the statistical analysis. Statistical significance was set at α = 0.05, and all α values were two-sided. To examine the magnitude of the differences, the effect size was calculated based on Cohen's d.

#### **3. Results**

#### *3.1. Physical Characteristics*

The mean age of the participants was 62.8 ± 4.4 years; the mean height was 162.5 ± 2.9 cm; the mean weight was 60.9 ± 3.0 kg; the mean BMI was 23.1 ± 1.1 kg/m2; and the mean body fat percentage was 27.7 ± 2.7% (Table 1).


**Table 1.** Baseline characteristics of the participants.

Values are mean ± SD. BMI, body mass index; SBP, systolic blood pressure; SD, standard deviation.

#### *3.2. Arterial Stiffness*

In the SSI trial, the baPWV increased 30, 60, and 90 min after ingestion compared with that before ingestion (*p* < 0.01); however, in the ISI trial, the baPWV did not change after ingestion compared with that before ingestion. The baPWV was not significantly different between the trials before ingestion (Figure 2A). The baPWV AUC was lower (*p* < 0.01) in the ISI trial than in the SSI trial (Figure 2B).

The hbPWV did not change after ingestion compared with that before ingestion in both the SSI and ISI trials, and the hbPWV was not different between the two trials (Figure 2C). The hbPWV AUC did not differ between the trials (Figure 2D).

#### *3.3. Heart Rate*

The HR did not change after sucrose ingestion compared with that before sucrose ingestion. Moreover, the HR did not change after isomaltulose ingestion compared with that before isomaltulose ingestion. Furthermore, there was no difference between the trials (Table 2).

#### *3.4. Brachial Blood Pressure*

The SBP and PP of the upper arm in the SSI trial increased 90 min after ingestion compared with those before ingestion (*p* < 0.05), whereas the SBP and PP of the upper arm in the ISI trial did not change after ingestion compared with those before ingestion. There was no difference between the trials. The MBP and DBP of the upper arm in the SSI trial did not change after ingestion compared with those before ingestion. The MBP and DBP of the upper arm in the ISI trial did not change after ingestion compared with those before ingestion. There was no difference between trials (Table 2).


**Table 2.** Changes in brachial SBP, MBP, DBP, and HR before and after the ingestion of isomaltulose and sucrose.

Values are mean ± SD. \* *p* < 0.05, vs. baseline. SSI, sucrose solution intake; ISI, isomaltulose solution intake; SBP, systolic blood pressure; MBP, mean blood pressure; DBP, diastolic blood pressure; PP, pulse pressure; HR, heart rate; SD, standard deviation.

#### *3.5. Ankle Blood Pressure*

The SBP, MBP, and PP of the ankle in the SSI trial increased 90 min after ingestion compared with those before ingestion (*p* < 0.05), and the SBP, MBP, and PP of the ankle in the ISI trial did not change after ingestion compared with those before ingestion. There were no differences between the trials.

The DBP of the ankle in the SSI trial did not change after ingestion compared with that before ingestion, and the DBP and HR of the ankle in the ISI trial did not change after ingestion compared with those before ingestion. There were no differences between the trials (Table 3).


**Table 3.** Changes in ankle SBP, MBP, and DBP before and after the ingestion of isomaltulose and sucrose.

Values are mean ± SD. \* *p* < 0.05, vs. baseline. SSI, sucrose solution intake; ISI, isomaltulose solution intake; SBP, systolic blood pressure; MBP, mean blood pressure; DBP, diastolic blood pressure; PP, pulse pressure; SD, standard deviation.

#### *3.6. Blood Glucose*

The blood glucose levels in the SSI trial increased 30 and 60 min after ingestion compared with those before ingestion (*p* < 0.01). The blood glucose levels in the ISI trial increased 30 min after ingestion compared with those before ingestion (*p* < 0.05). The blood glucose levels 30 min after intake were lower in the ISI trial than in the SSI trial (*p* < 0.01, (Figure 3A). The AUC of blood glucose level was lower in the ISI trial than in the SSI trial (Figure 3B).

**Figure 3.** Changes in blood glucose at baseline and post-ingestion in both trials. Values are mean ± SD. \*\* *p* < 0.01 and \* *p* < 0.05, vs. baseline. † *p* < 0.05, vs. ISI. BG, blood glucose; SSI, sucrose solution intake; ISI, isomaltulose solution intake; SD, standard deviation; Figure 3A, blood glucose; Figure 3B, blood glucose AUC.

#### *3.7. Arterial Stiffness and Brachial SBP at 90 Min after Sucrose and Isomaltulose Solution Ingestion*

The baPWV and brachial SBP were positively correlated 90 min after isomaltulose and sucrose ingestion (*r* = 0.640, *p* < 0.05) (Figure 4).

**Figure 4.** Correlation between arterial stiffness and brachial SBP at 90 min after sucrose and isomaltulose intake. Values are mean ± SD. *r* = 0.640 and *p* = 0.046. SBP, systolic blood pressure; SSI, sucrose solution intake; ISI, isomaltulose solution intake; SD, standard deviation.

#### **4. Discussion**

The main finding of this study was that the baPWV and SBP did not change after isomaltulose intake compared to before. This confirms our hypothesis. These results suggest that isomaltulose could be used as an alternative to sucrose, given the neutral effect on the PWV and SBP.

A rapid increase in blood glucose levels after a meal is an independent risk factor for cardiovascular disease and a greater risk factor than fasting glucose [18]. Therefore, it is necessary to control the rapid increase in blood glucose levels after meals to prevent cardiovascular diseases. There is a consensus that eating high-GI foods results in rapid carbohydrate absorption, whereas low-GI foods result in milder carbohydrate absorption

and consequently milder insulin secretion [19]. For example, a previous study of 10 healthy individuals showed a slower increase in blood glucose and insulin levels after consuming isomaltulose compared with that after sucrose consumption [14]. Our results are in agreement with these findings. Blood glucose levels after ingestion were lower in the ISI trial than in the SSI trial. Therefore, isomaltulose may slow down the rise in blood glucose levels after a meal compared with sucrose, the main component of sugar.

A number of studies have shown that arterial stiffness increases during hyperglycemia [4,19,20]. Moreover, previous studies have shown that the baPWV increases during acute hyperglycemia [5]. Our previous study also found an increase in the baPWV after glucose ingestion. The present results are in agreement with these findings, in which the baPWV increased 30, 60, and 90 min after sucrose ingestion compared with before sucrose ingestion, but no increase was observed in the ISI trial. In addition, the AUC of the baPWV was lower in the ISI trial than in the SSI trial. Therefore, isomaltulose can be expected to inhibit the increase in arterial stiffness compared with the consumption of other carbohydrates, such as sucrose and glucose, making it possible to create food products that are both tasty and healthy.

Diabetes mellitus induces peripheral arterial disease in the limbs [21] and it has been found that peripheral arterial stiffness increases after the 75-g glucose-tolerance test compared with before the test [22]. Previous studies have reported that the peripheral arterial PWV, especially in the lower limb arteries, increases during acute hyperglycemia [23]. In the current SSI study, the baPWV increased after ingestion compared to before sucrose ingestion, while the hbPWV did not change. Previous studies have reported that the baPWV reflects arterial stiffness in the distal (mainly abdominal) aorta and lower limbs [24], while the hbPWV reflects arterial stiffness in the proximal aorta and upper limbs [25]. In addition, the MBP and PP, which reflect aortic and peripheral arterial stiffness, are elevated after sucrose ingestion. Therefore, in middle-aged and older people, an increase in arterial stiffness during acute hyperglycemia is likely to affect the abdominal aorta and lower-limb arteries. However, in the current study, we were unable to measure arterial stiffness in detail by site. In future studies, we plan to further investigate the increase in arterial stiffness during acute hyperglycemia by site.

This study did not examine the mechanism by which arterial stiffness was not altered after isomaltulose ingestion, but there are several possible explanations. In the present study, in the SSI test, the SBP increased 90 min after compared to before sucrose intake. In previous studies, the SBP and baPWV were found to be correlated [26]. In the present study, there was a correlation between the SBP and baPWV in the upper arm after 90 min of ingestion (*r* = 0.640, *p* < 0.05). This suggests that increased systemic arterial stiffness may be responsible for the increase in the SBP. In this study, there was no correlation between the baPWV and the blood glucose level at 90 min, when the increase in the baPWV was highest. In other words, the blood glucose level may not be directly involved in the increase in the baPWV during acute hyperglycemia. Furthermore, sympathetic hyperactivity, increased oxidative stress, and decreased vascular endothelial function associated with increased blood glucose levels may be related in parallel. Increased sympathetic nerve activity has been implicated in the increase in the baPWV [27]. Sympathetic nerve activity has been found to increase after eating [28]. The sympathetic ratio after a meal shows a sustained elevation lasting at least one hour, which has been suggested to be primarily due to a decrease in vagal activity [28]. Thus, the increase in the baPWV after sucrose consumption in the present study may be due to increased sympathetic nerve activity. However, since sympathetic nerve activity was not measured in this study, it should be assessed in the future. Decreased vascular function (PWV and FMD) after a meal has been proven to be dependent on oxidative stress [3]. For example, 2-thiobarbituric-acidreactive substances (TBARS), an indicator of oxidative stress, have been shown to increase after acute hyperglycemia [29]. Oxidative stress is thought to reduce vascular function by increasing asymmetric dimethylarginine (ADMA) [30]. Hyperglycemia-induced vascular dysfunction after an oral glucose challenge was found to be associated with increased

plasma ADMA/Arg [30]. Therefore, it is likely that acute hyperglycemia increased TBARS and induced vascular endothelial dysfunction via increased ADMA, which caused the increase in the PWV. However, oxidative stress was not measured in this study and should be measured in future studies.

Regarding the application of the study results, the use of isomaltulose as a sweetener in everyday cooking may reduce arterial stiffness and blood pressure increases during acute hyperglycemia compared with the use of other carbohydrates. We believe that the need for isomaltulose to prevent arteriosclerosis and elevated blood pressure will increase, especially as people are increasingly cooking for themselves as a way of preventing new coronavirus infections and as they become more conscious of nutritional balance.

Nevertheless, this study has certain limitations. One limitation was the relatively small number of participants. However, the sample size was statistically significant. Furthermore, we believe that the findings are not generalizable to different populations (e.g., young people and people with diabetes) because the study was conducted in older people, and we will therefore examine different groups of people in the future. In addition, although insulin and endothelial dysfunction may alter the PWV, insulin levels and endothelial function biomarkers were not measured in the present study.

#### **5. Conclusions**

The main finding of this study was that the baPWV and SBP did not change after isomaltulose intake compared to before. This confirms our hypothesis. These results suggest that isomaltulose could be used as an alternative to sucrose, given the neutral effect on the PWV and SBP.

**Author Contributions:** R.K. designed the study; R.K. and K.A. collected the data and conducted the study; R.K. performed the statistical analysis; R.K. drafted the manuscript; M.S., Y.N., K.A., T.H., and H.N. provided a critical review of the manuscript; R.K. had primary responsibility for the final content. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by a joint research grant from Mitsui Sugar Co., Ltd.

**Institutional Review Board Statement:** This study was conducted in compliance with the Declaration of Helsinki on the basis of ethics, human rights, and the protection of the personal information of participants. Ethical approval for this study was obtained from the Ethics Committee of Teikyo University of Science (approval number 20A024, September 30, 2020). This study was also registered at the University Hospital Medical Information Network Center (UMIN Center; Study No. UMIN000041622).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available from the corresponding author upon request.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Macronutrient Intake and Insulin Resistance in 5665 Randomly Selected, Non-Diabetic U.S. Adults**

**Larry A. Tucker**

College of Life Sciences, Brigham Young University, Provo, UT 84602, USA; tucker@byu.edu; Tel.: +1-801-422-4927

**Abstract:** The main goal of this investigation was to evaluate the relationships between several macronutrients and insulin resistance in 5665 non-diabetic U.S. adults. A secondary objective was to determine the extent to which the associations were influenced by multiple potential confounding variables. A cross-sectional design and 8 years of data from the 2011–2018 National Health and Nutrition Examination Survey (NHANES) were used to answer the research questions. Ten macronutrients were evaluated: total carbohydrate, starch, simple carbohydrate, dietary fiber, total protein, total fat, saturated, polyunsaturated, monounsaturated, and total unsaturated fat. The homeostatic model assessment (HOMA), based on fasting glucose and fasting insulin levels, was used to index insulin resistance. Age, sex, race, year of assessment, physical activity, cigarette smoking, alcohol use, and waist circumference were used as covariates. The relationships between total carbohydrate intake (F = 6.7, *p* = 0.0121), simple carbohydrate (F = 4.7, *p* = 0.0344) and HOMA-IR were linear and direct. The associations between fiber intake (F = 9.1, *p* = 0.0037), total protein (F = 4.4, *p* = 0.0393), total fat (F = 5.5, *p* = 0.0225), monounsaturated fat (F = 5.5, *p* = 0.0224), and total unsaturated fat (F = 6.5, *p* = 0.0132) were linear and inversely related to HOMA-IR, with 62 degrees of freedom. Starch, polyunsaturated fat, and saturated fat intakes were not related to HOMA-IR. In conclusion, in this nationally representative sample, several macronutrients were significant predictors of insulin resistance in U.S. adults.

**Keywords:** carbohydrate; protein; fat; unsaturated fat; saturated fat; sugar; starch; fiber; diabetes

#### **1. Introduction**

Insulin resistance is a pathological condition in which body cells manifest reduced sensitivity to insulin. Insulin promotes the distribution of glucose across muscle and fat tissues and decreases the liver's release of glucose by decreasing the breakdown of glycogen, and gluconeogenesis. Additionally, insulin represses the release of non-esterified fatty acids from fat tissue by decreasing lipolysis [1]. Because of insulin resistance, glucose is not transported into cells at the correct rate, resulting in elevated blood glucose levels. As blood glucose levels increase, the body responds by increasing circulating insulin levels. Consequently, individuals who are insulin resistant often have elevated insulin levels.

Some of the most common chronic diseases in developed societies are linked to insulin resistance. For example, the relationship between insulin resistance and the risk of developing cardiovascular disease is strong [2–4]. Endothelial dysfunction and the development of atherosclerosis is also closely tied to insulin resistance [5,6], along with stroke [7], hypertension [8], dyslipidemia [9], neurodegenerative diseases [10,11], metabolic syndrome [12], and type 2 diabetes [13], to name a few.

Although there are many factors that contribute to insulin resistance, obesity is one of the strongest driving factors [14], even in children [15]. Abdominal obesity seems to pose a greater risk of metabolic disease than an elevated body mass index [16]. Although challenging, weight loss increases insulin sensitivity and reduces the risk of developing insulin resistance and diabetes.

**Citation:** Tucker, L.A. Macronutrient Intake and Insulin Resistance in 5665 Randomly Selected, Non-Diabetic U.S. Adults. *Nutrients* **2022**, *14*, 918. https://doi.org/10.3390/ nu14050918

Academic Editors: Silvia V. Conde and Fatima O. Martins

Received: 14 January 2022 Accepted: 19 February 2022 Published: 22 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Regular physical activity also plays a significant role in reducing the likelihood of insulin resistance [17], even in those with abdominal obesity [18]. Furthermore, exercise and physical activity also reduce the risk of developing type 2 diabetes [19].

Although obesity and physical inactivity often lead to metabolic impairment, several other lifestyle factors also contribute to metabolic health. For example, research indicates that diet is a key modifiable variable that can be targeted to counter the rising rates of insulin resistance and diabetes. Research indicates that diet composition can be manipulated to improve insulin sensitivity and reduce the risk of diabetes.

To date, dozens of investigations have been conducted to evaluate the relationship between diet composition and metabolic disease, particularly insulin resistance. Study designs and methods have varied substantially, and findings have been mixed. Due to the inconsistent findings in the literature, additional investigations are needed to assess the relationship between diet composition and insulin resistance. Moreover, most investigations in this area have been conducted using small samples and have resulted in few generalizable findings. Hence, the chief goal of the present study was to determine the extent to which differences in diet composition, particularly macronutrient intake, account for differences in insulin resistance in a large sample of adults representing the U.S. population. Another objective was to ascertain the role of several potential confounding factors, including age, sex, race, year of assessment, smoking, alcohol, physical activity, and waist circumference, on the relationship between macronutrient intake and insulin resistance. Effect modification was also evaluated across tertiles of the primary macronutrients and levels of physical activity.

#### **2. Materials and Methods**

#### *2.1. Study Design and Sample*

The U.S. National Health and Nutrition Examination Survey (NHANES) database was used to answer the research questions. NHANES is an ongoing government-run survey, administered by the National Center for Health Statistics (NCHS) and the Centers for Disease Control and Prevention (CDC). NHANES data were gathered through the use of extensive interviews, questionnaires, blood samples, and physical examinations performed by trained professionals on individuals selected randomly from the U.S. population.

Before data were collected, each individual in the sample provided written informed consent. The Ethics Review Board for the NCHS approved the data collection protocol and the data files, containing no confidential information, that were published on the NHANES website for public use [20].

This investigation used NHANES data gathered during an 8-year period, from 2011– 2018. Data from 2019–2020 were not available due to the COVID pandemic limiting the NHANES data collection process. The codes signifying ethical approval for NHANES data collected from 2011–2018 are: Protocol #2011–17 and Protocol #2018-01.

A total of 5665 adults were included in this study, ages 18–75 years. A 4-stage strategy was employed to randomly select non-institutionalized, civilian U.S. adults. Census information was utilized so that counties, then blocks, then dwelling units, and finally individuals were randomly selected, so that the final data are nationally representative [21].

Individuals who reported that they were diabetic, or took oral medication or insulin to control their blood sugar levels, or were found to have a fasting blood glucose concentration of 126 mg/dL or higher, were excluded from the sample. Those with hypoglycemia (fasting glucose < 70 mg/dL) were also excluded. Additionally, subjects who did not consume any kilocalories (kcal) during either of the two 24-h dietary recall assessments (i.e., they fasted) were excluded. Participants with extreme HOMA-IR levels (≥4 standard deviations above the mean) were excluded, and individuals who were underweight (BMI < 18.5) were excluded because of the likelihood of an eating disorder, severe frailty, or a serious disease.

#### *2.2. Instrumentation and Measurement Methods*

Macronutrient intake was the exposure variable for this study. A total of 10 macronutrients were studied. The outcome variable was insulin resistance, indexed using the homeostatic model assessment for insulin resistance (HOMA-IR). Age, sex, race, year of assessment, physical activity, cigarette smoking, alcohol use, and waist circumference were included as covariates, so their influence on the association between the exposure and outcome variables could be minimized.

#### 2.2.1. Insulin Resistance

HOMA-IR is the most frequently used method of assessing insulin resistance. Over 18,600 published journal articles available in PUBMED include the term "HOMA" or "HOMA-IR." The HOMA-IR calculation was based on fasting plasma glucose and fasting insulin levels., specifically: fasting insulin (μU/mL) × fasting glucose (mg/dL) ÷ 405. Subjects who were randomly assigned to attend a morning data collection session were asked to fast for 9 h for the fasting blood draw. Comprehensive information is provided by NHANES on their website about the glucose and insulin measurement protocols [22,23].

#### 2.2.2. Macronutrient Intake

Two 24-h dietary recall assessments were used to gather the macronutrient data. The first diet interview occurred in-person. Data for the second recall interview was collected by telephone 3 to 10 days later. The average of the two assessments was used. Both dietary assessments gathered detailed information about all foods and beverages eaten during the 24-h period prior to the interview (midnight to midnight). Individuals reporting that they did not eat during one or both of the 24-h dietary assessment periods were not included in the analyses.

According to Willett, one 24-h dietary recall may be adequate if sample sizes are sufficiently large. He also states that, to estimate within-person variability, "it is statistically most efficient to increase the number of individuals in the sample, rather than to increase the number of days beyond 2 days per individual" (page 55) [24]. Given that the present study included over 5000 randomly selected adults, with each completing two 24-h dietary recalls 3–10 days apart, the assessment methods employed were more than satisfactory to secure quality estimates of dietary intake.

The diet recall interviewers were thoroughly trained in preparation for administering the diet assessments. They each had at least 10 college credits in nutrition courses, and each was a graduate in Food and Nutrition or Home Economics. Each of the individuals administering the diet recalls was bilingual and the dietary data collection occurred in a private setting in the NHANES mobile examination center (MEC). Interviewers were guided by scripts, and the computer-based program afforded a standard interview protocol. The diet assessments followed a multi-pass format called the Automated Multiple Pass Method (AMPM), available online [25]. The diet recall included food probes that have been used in previous United States Department of Agriculture (USDA) and NHANES surveys.

To help the participant, the in-person interviews included a number of real-life examples, such as different sized glasses, bowls, mugs, bottles, spoons, cups, plates, etc. After completing the first (in-person) diet interview, subjects were given sample cups, spoons, etc. and a food model booklet to take home to assist them during the telephone diet interview.

The present investigation included 10 macronutrients in order to study their relationship with insulin resistance: total carbohydrate, starch, simple carbohydrate, and fiber; total protein; total fat, polyunsaturated fat, monounsaturated fat, saturated fat, and total unsaturated fat. Except for dietary fiber intake, each of the macronutrients was reported as a percentage of the total energy consumed by the participant. Hence, total energy intake was taken into account for each macronutrient and for each participant. For fiber intake, the value was expressed as grams per 1000 kcal.

#### *2.3. Covariates*

NHANES classifies race into six categories: Mexican American, Non-Hispanic Black, Non-Hispanic White, Non-Hispanic Asian, Other Hispanic, or Other Race/Multi-racial. The NHANES racial categories were used as a covariate in the present study.

Waist circumference was utilized to index abdominal or central adiposity. Abdominal adiposity tends to be a better predictor of insulin resistance and diabetes than general obesity [16]. The measurement was taken by trained specialists. The procedures used by those performing the assessment were evaluated regularly to ensure high-quality performance. Measurement of the waist was taken in a customized room in the mobile examination center (MEC). Those taking the measurements were assisted by a trained recorder. The person taking the measurement and the assistant worked together to position, assess, and record the values precisely. The measuring tape was put around the body in a horizontal plane immediately above the top border of the ilium. To safeguard horizontal alignment of the tape, a wall mirror was utilized. The tape was placed snugly around the person, but the skin was not to be compressed. The measurement was finalized at the conclusion of a normal expiration [26].

Differences in physical activity levels were also used as a covariate. Physical activity was measured via interview. Participants were queried about the amount of time they spent in moderate and vigorous activities. Moderate activity was described as physical activity that results in small increases in breathing or heart rate, such as walking, carrying light loads or casual bike riding. Vigorous physical activity was explained as activity resulting in large increases in breathing or heart rate, such as jogging or running, walking up a moderate or steep incline, or lifting heavy loads.

Physical activity (PA) was assessed using specific questions asked by the NHANES interviewer: "In a typical week, on how many days do you do moderate-intensity sports, fitness, or recreational activities?" Also, "How much time do you spend doing moderateintensity sports, fitness, or recreational activities on a typical day?" These questions, with slight alterations, were also used to assess vigorous PA. For the two intensities, days and minutes were multiplied together to produce the total minutes of moderate and total minutes of vigorous physical activity. These minutes were summed together to give total time (minutes) spent doing moderate and vigorous physical activity (MVPA).

The relationship between each of the primary macronutrients and insulin resistance was evaluated within three different sex-specific categories of physical activity. Participants who reported 0–30 min of activity per week were placed into the Low physical activity category. This category comprised approximately 45% of the sample. The remaining participants were divided equally between the Moderate and High physical activity categories. Specifically, females who reported 40 or more minutes and less than 180 min of activity per week were placed in the Moderate physical activity category. Males who reported 40 or more minutes and less than 240 min of activity per week were also placed in the Moderate physical activity category. Women reporting 180 min or more and men reporting 240 min or more of activity were placed in the High physical activity category.

Statistical adjustments were also made for differences in alcohol use in the current study. As part of the two 24-h dietary recall interviews, participants were asked to report their alcoholic beverage consumption. Those who reported that they did not drink any beverages containing alcohol were given a zero and alcohol consumers were assigned values based on the percentage of their total energy intake derived from alcohol.

Differences in cigarette smoking were also controlled statistically in this study. Smoking was measured by assessing the typical number of cigarettes smoked per day during the past month. An NHANES interviewer specifically asked, "During the past 30 days, on the days that you smoked, about how many cigarettes did you smoke?" Non-smokers were given the value of 0, whereas smokers had values up to 95 [27].

#### *2.4. Data Analysis*

A multi-level sampling technique was employed in this study so that the findings can be generalized to the U.S. adult population. To accomplish this, strata, clusters and individual sample weights were included in each statistical model.

With a sample size of 5665 individuals, a high level of statistical power would be expected in each statistical model. However, given the sampling strategy employed, degrees of freedom (df) were based on the number of clusters (121) minus the number of strata (58), resulting in 62 df, instead of 5665 df in the denominator.

There was one outcome variable (HOMA-IR) and 10 exposure variables (macronutrient intake) evaluated separately using multiple regression. A number of covariates were controlled statistically to reduce their influence on the relationship between the macronutrients and insulin resistance, specifically age, sex, race, year of assessment, physical activity, smoking, alcohol use, and waist circumference. Macronutrients could not be employed as covariates to determine their influence on the key relationships because of multicollinearity, tested by using the SAS variance inflation factor option (VIF). The VIF was 3.8 for total carbohydrate intake and 3.5 for total fat intake when in the same model. A VIF ≥ 2.5 is considered problematic [28].

Two statistical analysis strategies were employed to measure the associations between the macronutrients and HOMA-IR. First, linear relationships were tested by treating both the exposure and outcome measures as continuous variables and using multiple regression and the SAS SurveyReg procedure. Second, one-way analysis of variance (ANOVA) was employed using the SAS SurveyReg procedure to measure mean differences in HOMA-IR across each macronutrient divided into tertiles. Tertile cut-points were: Total Carbohydrate: <44.10, 44.11–51.85, >51.85; Simple carbohydrate: <16.52, 16.53–23.59, >23.59%; Starch: <24.23, 24.24–29.51, >29.51; Fiber: <6.43, 6.44–9.31, >9.31; Total protein: <13.94, 13.95– 17.24, >17.24; Total fat: 31.47, 31.48–37.44, >37.44; Saturated fat: <9.70, 9.71–12.36, >12.36; Unsaturated fat: <20.93, 20.94–25.20, >25.20; Monounsaturated fat: <10.60, 10.61–13.07, >13.07; Polyunsaturated fat: <6.70, 6.71–8.82, >8.82. The macronutrient categories were based on tertiles calculated using percentage of total kilocalories, so energy intake was taken into account for each macronutrient and each participant. Tertiles of total fiber consumption were categorized based on grams consumed per 1000 kilocalories (kcal).

Covariates were controlled using multiple regression and partial correlation, and the Least Squares Means (LSMeans) procedure was used to produce adjusted means. Effect modification was tested by dividing the macronutrients into tertiles and then analyzing the relationships between each primary macronutrient and HOMA-IR within the individual tertiles. Effect modification was also evaluated with physical activity divided into three categories and then analyzing the associations between each primary macronutrient and HOMA-IR within the three categories separately. Physical activity could not be divided into tertiles because over 40% of the subjects reported participating in no regular physical activity each week.

SAS version 9.4 (SAS Institute, Inc., Cary, NC, USA) was used to organized and examine the data. All the statistical tests were two-sided, and alpha was fixed at <0.05 to define statistical significance.

#### **3. Results**

The sampling strategy used by NHANES included 59 strata and 121 clusters selected randomly. There were 5665 subjects in the sample spread evenly from 2011–2018. Mean age (± SE) was 43.7 ± 0.4 years. Mean (± SE) energy intake for the two independent 24-h recall assessments was 2130 ± 14.0 kilocalories (kcals) per 24 h. Average (± SE) HOMA-IR was 2.76 ± 0.05. Moreover, mean (± SE) intake of the primary macronutrients, expressed as a percentage of total energy intake, was: total carbohydrate (48.0 ± 0.2), total protein (16.1 ± 0.1), and total fat (34.4 ± 0.1). Fiber intake averaged (± SE) 8.5 ± 0.1 g per 1000 kcals. Mean (± SE) waist circumference was 98.4 (±0.4) cm. Table 1 shows the distribution of values across percentiles for each of the key continuous variables.


**Table 1.** Percentile values of the key variables for 5665 adults representing the U.S. population.

HOMA-IR is the homeostatic model of assessment. % of kcal is the percentage of total kilocalories derived from the listed macronutrient. MVPA is moderate to vigorous physical activity.

Race and sex were treated as categorical variables in this investigation. Results indicated that 66.2% were Non-Hispanic White, 11.0% were Non-Hispanic Black, 8.4% were Mexican American, 6.2% were Other Hispanic, 4.7% were Non-Hispanic Asian, and 3.4% were Other-race or Multi-racial. The sample was comprised of 52.4% women and 47.6% men.

#### *3.1. Dietary Carbohydrate and Insulin Resistance*

The linear relationship between carbohydrate intake and HOMA-IR was studied with carbohydrate intake and insulin resistance both treated as continuous variables. After adjusting for the covariates (i.e., age, sex, race, year of assessment, smoking, alcohol, physical activity, and waist circumference) total carbohydrate intake was a significant predictor of HOMA-IR with 62 degrees of freedom. Specifically, as percent of kilocalories (kcal) from total carbohydrate increased, HOMA-IR increased linearly (F = 6.7, *p* = 0.0121), as shown in Table 2. Level of simple carbohydrate consumption was also positively and linearly related to HOMA-IR (F = 4.7, *p* = 0.0344). However, percentage of kcals from starch was not predictive of insulin resistance (F = 1.8, *p* = 0.1886). Grams of total fiber intake per 1000 kcals were inversely associated with HOMA-IR (F = 9.1, *p* = 0.0037). As fiber consumption increased, insulin resistance tended to decrease in a linear fashion (Table 2), after adjusting for the covariates.

The association between total carbohydrate consumption and HOMA-IR was also evaluated by comparing mean differences in HOMA-IR across tertile levels of carbohydrate intake (Table 3). As displayed in Table 3, results showed that adults in the lowest tertile of total carbohydrate intake had significantly less insulin resistance compared to the other two tertiles (F = 10.4, *p* = 0.0001). Similarly, those in the lowest tertile of simple carbohydrate consumption had lower HOMA-IR levels than the other two tertiles (F = 4.8, *p* = 0.0117). However, mean HOMA-IR levels did not differ across tertiles of starch intake. Finally, those in the highest one-third of fiber intake had significantly less insulin resistance than the other two tertiles (F = 3.3, *p* = 0.0436), as displayed in Table 3.


**Table 2.** Linear relationship between macronutrient consumption and HOMA-IR in a randomly selected sample of 5665 adults representing the U.S. population.

Note: Each of the dietary intake variables was reported as the percentage of total kilocalories consumed by the participant. For example, if an individual consumed an average of 2000 kilocalories per day for the two 24-h dietary recall assessments, and the person averaged 250 g of carbohydrate intake per day, then the individual's total carbohydrate intake would be 50% of total kcals (250 g × 4 kcal = 1000 kcal; 1000/2000 = 0.50). The covariates for each model were age, sex, race, year of assessment, smoking, alcohol, physical activity, and waist circumference. There were 62 degrees of freedom (df) in the denominator for each model.


**Table 3.** Differences in mean HOMA-IR levels across tertiles of macronutrient intake in 5665 U.S. adults, after adjusting for covariates.

a,b Means on the same row with the same superscript letter do not differ significantly. The difference in HOMA-IR between the moderate and high fiber groups was borderline significant (*p* = 0.0915). All means were adjusted for differences in the covariates: age, sex, race, year of assessment, physical activity, smoking, alcohol, and waist circumference. Tertile cut-points were: Total Carbohydrate: <44.10%, 44.10–51.85%, >51.85; Sugar: <16.52%, 16.53–23.59%, >23.59%; Starch: <24.23, 24.24–29.51, >29.51; Fiber: <6.43, 6.44–9.31, >9.31; Total protein: <13.94, 13.95–17.24, >17.24; Total fat: 31.47, 31.48–37.44, >37.44; Saturated fat: <9.70, 9.71–12.36, >12.36; Unsaturated fat: <20.93, 20.94–25.20, >25.20; Monounsaturated fat: <10.60, 10.61–13.07, >13.07; Polyunsaturated fat: <6.70, 6.71–8.82, >8.82. The macronutrient categories were based on tertiles calculated using percentage of total kilocalories. Tertiles of fiber intake were categorized based on grams consumed per 1000 kilocalories.

#### *3.2. Dietary Protein and Insulin Resistance*

Percent of kcals derived from dietary protein was also a significant predictor of insulin resistance (F = 4.4, *p* = 0.0393). The association was linear and inverse. As total protein consumption increased, insulin resistance tended to decrease (Table 2). However, when mean HOMA-IR levels were compared across tertiles of total protein intake, mean differences were not significant (Table 3).

#### *3.3. Dietary Fat and Insulin Resistance*

As displayed in Table 2, total dietary fat intake, recorded as a percentage of total kcals, was a significant predictor of HOMA-IR (F = 5.5, *p* = 0.0225), with 62 df and both measures treated as continuous variables. The relationship was linear and inverse. Specifically, as total fat intake increased, insulin resistance decreased, after controlling for the potential confounding variables. Conversely, the association between polyunsaturated fat consumption and HOMA-IR was not statistically significant (F = 2.6, *p* = 0.1102). However, the relationship between monounsaturated fat and insulin resistance was significant and inverse (F = 5.5, *p* = 0.0224). Additionally, with polyunsaturated and monounsaturated fat intake combined, the association between unsaturated fat intake and HOMA-IR was significant and inverse (F = 6.5, *p* = 0.0132). Level of saturated fat consumption was not predictive of HOMA-IR (F = 2.1, *p* = 0.1513), with the covariates controlled.

With total fat consumption divided into equal categories, mean HOMA-IR levels differed significantly across the tertiles (F = 6.4, *p* = 0.0030), as shown in Table 3. Specifically, subjects in the highest tertile of total fat intake had significantly less insulin resistance compared to those in the middle or lowest tertile of fat consumption. However, mean levels of insulin resistance did not differ significantly across tertiles of saturated, polyunsaturated, or monounsaturated fat intakes (Table 3). However, with poly- and monounsaturated fat intakes combined, mean HOMA-IR levels differed significantly across tertiles of unsaturated fat (F = 4.1, *p* = 0.0218). Specifically, adults in the highest tertile of unsaturated fat intake had significantly lower levels of HOMA-IR compared to the other two tertiles.

#### *3.4. Effect Modification*

Isolating the association between macronutrient consumption and insulin resistance is challenging. Consuming more of some foods typically results in consuming less of others. Macronutrients are intercorrelated. Using multiple regression and partial correlation to control statistically for differences in a macronutrient, such as dietary fat intake, when studying the relationship between carbohydrate consumption and insulin resistance, typically results in problems associated with multicollinearity, making the findings invalid or uncertain. Although different from partial correlation, testing for effect modification can help clarify the interaction between each macronutrient and insulin resistance.

As displayed in Table 4, when delimited to individual tertiles of total dietary fat intake, carbohydrate consumption was not related to HOMA-IR. Similarly, total protein intake and dietary fat consumption were not associated with insulin resistance within any of the fat intake tertiles (Table 4).

**Table 4.** Linear relationships between the primary macronutrients and HOMA-IR within tertiles of dietary fat intake, after controlling for the covariates.


SE = standard error of the regression coefficient. Total dietary fat intake was divided into tertiles. The relationships between insulin resistance (HOMA-IR) and consumption of total carbohydrate, total protein, and total fat were evaluated separately within each of the three dietary fat tertiles. Age, sex, race, year of assessment, physical activity, smoking, alcohol, and waist circumference were the covariates. Each of the analyses had 62 degrees of freedom (df) in the denominator.

According to Table 5, none of the primary macronutrients were significantly related to insulin resistance within the lowest tertile of protein intake. Similarly, none were associated with insulin resistance when the sample was delimited to the middle tertile of protein consumption. On the other hand, total carbohydrate, protein, and fat intakes were each related to HOMA-IR in adults reporting a high protein diet. The carbohydrate and HOMA-IR association was positive (direct), and the protein and fat intake correlations were negative (inverse).


**Table 5.** Linear relationships between the primary macronutrients and HOMA-IR within tertiles of dietary protein intake, after controlling for the covariates.

SE = standard error of the regression coefficient. Total dietary protein intake was divided into tertiles. The relationships between insulin resistance (HOMA-IR) and consumption of total carbohydrate, total protein, and total fat were evaluated separately within each of the three dietary protein tertiles. Age, sex, race, year of assessment, physical activity, smoking, alcohol, and waist circumference were the covariates. Each of the analyses had 62 degrees of freedom (df) in the denominator.

As revealed in Table 6, among adults reporting a high carbohydrate diet (highest tertile), none of the primary macronutrients were related to HOMA-IR. With the sample delimited to adults in the middle tertile of carbohydrate intake, fat intake was inversely related to HOMA-IR. Among those eating a low carbohydrate diet, protein intake was inversely related and carbohydrate intake was directly associated with insulin resistance. The carbohydrate correlation was borderline significant.

**Table 6.** Linear relationships between the primary macronutrients and HOMA-IR within tertiles of dietary carbohydrate intake, after controlling for the covariates.


SE = standard error of the regression coefficient. Total carbohydrate intake was divided into tertiles. The relationships between insulin resistance (HOMA-IR) and consumption of total carbohydrate, total protein, and total fat were evaluated separately within each of the three dietary carbohydrate tertiles. Age, sex, race, year of assessment, physical activity, smoking, alcohol, and waist circumference were the covariates. Each of the analyses had 62 degrees of freedom (df) in the denominator.

As shown in Table 7, there were no significant relationships between macronutrient consumption and insulin resistance in the Moderate or High physical activity categories. However, among participants who reported Low physical activity, dietary fat intake was inversely related to HOMA-IR and total carbohydrate consumption was associated with HOMA-IR in a direct and significant relationship. Protein intake was not related to insulin resistance within any of the physical activity categories, Low, Moderate, or High.


**Table 7.** Linear relationships between the primary macronutrients and HOMA-IR within 3 categories of physical activity, after controlling for the covariates.

SE = standard error of the regression coefficient. Total physical activity was divided into 3 categories. Adults with Low weekly physical activity (30 min or less per week) comprised 45% of the sample. The remaining 55% was divided into sex-specific equal categories with approximately 27.5% of the sample in each. For females, the cut-point dividing between Moderate and High physical activity was 180 min per week. For males, the cut-point was 240 min per week. The relationships between the primary macronutrients and insulin resistance (HOMA-IR) were evaluated separately within each of the three physical activity categories. Age, sex, race, year of assessment, smoking, alcohol, and waist circumference were the covariates. Each of the analyses had 62 degrees of freedom (df) in the denominator.

#### **4. Discussion**

The chief objective of the present study was to evaluate the relationships between multiple dietary macronutrient intakes and insulin resistance, measured using HOMA-IR, in 5665 randomly selected U.S. adults. Several potential confounding variables were controlled statistically to help isolate the association between the macronutrient intakes and HOMA-IR. Effect modification across tertiles of the primary macronutrients and physical activity were also evaluated.

There were seven major outcomes in this study: (1) Macronutrient intake was predictive of insulin resistance measured by HOMA-IR in U.S. adults. (2) Both total carbohydrate and simple carbohydrate intakes were positively and linearly related to HOMA-IR, but starch consumption was not associated with HOMA-IR. (3) Fiber intake was inversely related to HOMA-IR. (4) Protein intake was inversely associated with HOMA-IR when both variables were treated as continuous, but protein consumption was not associated with HOMA-IR when protein intake was divided into tertiles. (5) Total fat consumption was linearly and inversely related to HOMA-IR, along with monounsaturated fat intake, when both were treated as continuous measures. However, polyunsaturated and saturated fats were not related to HOMA-IR. Unsaturated fat intake (poly- and monounsaturated fats combined) was linearly and inversely related to HOMA-IR, whether or not the association was analyzed with both variables treated as continuous or with unsaturated fat intake divided into tertiles. (6) With dietary fat intake divided into tertiles, none of the primary macronutrients were predictive of HOMA-IR when confined within the tertiles. (7) None of the primary macronutrients were predictive of HOMA-IR within the Moderate or High physical activity categories, but only within the Low physical activity group.

#### *4.1. Dietary Carbohydrate*

In the present investigation, as carbohydrate consumption increased, insulin resistance increased in a linear fashion. Simple carbohydrate intake followed a similar pattern, but starch intake was not related to insulin resistance. Numerous scientists have studied the relationship between carbohydrate intake and insulin resistance. However, findings focusing on this relationship vary widely [29].

In a cross-sectional analysis of the Framingham Offspring Study by McKeown et al. with 2834 participants, total carbohydrate intake was not associated with HOMA-IR [30], conflicting with the present findings. On the other hand, whole grain intake and fiber consumption were significantly related to lower levels of insulin resistance. When high glycemic index foods were consumed in larger amounts (i.e., typically indicating more simple carbohydrate intake), HOMA-IR was also significantly higher. On the other hand, in a cross-sectional study of 173 south Asian and European men, elevated insulin levels were directly related to carbohydrate intake (*p* = 0.001) [31], similar to this investigation.

Findings from the Inter99 study revealed that grams of sucrose, glucose, and fructose were each inversely related to HOMA-IR, whereas the association was positively associated with daily lactose intake [32]. Each of these relationships was significant or borderline significant. Note that in the Inter99 investigation, simple sugar intakes were expressed as grams per day, not as a percent of total energy intake.

Randomized controlled trials afford a different perspective. For example, Borkman et al. compared the effect of a high carbohydrate diet or a high fat diet (mostly saturated fat) on insulin sensitivity over three weeks [33]. The diets were administered in random order. Whole body glucose uptake using euglycemic glucose clamps revealed no change or difference in the effects of the diets on insulin sensitivity. Similarly, Garg et al. studied eight men with mild diabetes using a randomized cross-over investigation [34]. Diets were isocaloric and either high carbohydrate (60% of kcal) or low carbohydrate (35% of kcal). The low carbohydrate diet was high in monounsaturated fat. Plasma glucose and insulin responses were equal and insulin sensitivity via clamp at the end of each period were not different.

There is evidence that postprandial glycemic and insulinemic responses to foods differ based on the amount and characteristics of the carbohydrate ingested [35–37]. As shown in the McKeown investigation above, the glycemic index (GI) can be used to study the degree to which glucose levels are affected by carbohydrate type [36]

According to a randomized cross-over study in diabetic men by Rizkalla et al., fourweeks on a low glycemic diet produced lower postprandial glucose and insulin levels and areas under the curve than 4-weeks on a high GI diet [38]. Overall, whole-body glucose use, assessed using the euglycemic clamp, favored the low glycemic diet [38]. Similarly, Juanola-Falgarona et al. directed a six-month study of 122 overweight or obese adults to evaluate the effect of two moderate-carbohydrate diets or a low-fat diet, each with energy restriction, and different glycemic index scores, on a variety of cardio-metabolic outcomes, including HOMA-IR [39]. Results showed that insulin resistance was decreased more in the low-GI group than the low-fat group.

Several prospective investigations have studied the relationship between the dietary glycemic index and the development of type II diabetes, a common consequence of insulin resistance [40]. Specifically, Villegas et al. studied a cohort of over 64,000 Chinese women with no history of diabetes or other chronic disease for almost five years [41]. Divided into quintiles, the highest GI group had 21% greater risk of developing diabetes than the lowest quintile, and 34% greater risk based on glycemic load. When the focus was simply on carbohydrate intake rather than glycemic index, the highest quintile had 28% higher risk of developing diabetes than the lowest quintile [41]. On the other hand, in a smaller cohort study of Japanese men, Sakurai showed that men in the upper quintile of the glycemic index had 80% higher risk of developing diabetes compared to the lowest quintile, although the fourth and fifth quintiles based on glycemic load did not differ from the lowest quintile [42].

In a five-year prospective study of older Australians by Barclay et al., consumption of total carbohydrate, sugar, starch, and fiber, evaluated separately, were not predictive of the development of type 2 diabetes. However, in adults younger than 70 years old, higher intake levels on the glycemic index were predictive of higher incidence of diabetes over time [43].

Three large prospective cohort investigations, the Nurses' Health Study [44], the Health Professionals Follow-up Study [45], and the Iowa Women's Health Study [46] investigated the extent to which total carbohydrate intake influences risk of developing diabetes over time. None of these investigations found an association between total carbohydrate consumption and diabetes incidence. On the other hand, Swishburn et al. designed a five-year prospective cohort investigation which showed that a low fat, moderately high

carbohydrate diet was correlated with reduced insulin resistance and decreased risk of developing diabetes in adults with diminished glucose tolerance [47].

#### *4.2. Dietary Fiber*

Fiber intake and insulin resistance were strongly and inversely related. Similar to the current study, many investigations indicate that diets with high fiber content have beneficial effects on insulin sensitivity. Specifically, Fukagawa et al. studied 12 healthy individuals before and after 3–4 weeks of a high carbohydrate and high fiber diet [48]. Glucose disposal using the euglycemic clamp was measured. The high carbohydrate and high fiber diet reduced fasting glucose and insulin levels substantially and glucose disposal rates were also improved significantly. Similarly, Pereira et al. looked at the effect of a whole grain compared to a refined grain diet on insulin sensitivity using a randomized crossover design in 11 overweight, hyper-insulinemic adults [49]. Energy needs were balanced to prevent weight gain. Insulin sensitivity was significantly better when on the whole grain compared to the refined grain diet.

Using a cross-sectional design and baseline values from the Inter99 study, Lau et al. employed Danish adults to study the association between fiber consumption and HOMA-IR [32]. A food frequency questionnaire (FFQ) was utilized to assess fiber intake. Fiber was reported in total grams, not grams per 1000 kcal. Even after controlling for a variety of covariates, the relationship between fiber intake per day and HOMA-IR remained significant and inverse.

Lutsey et al. studied the association between whole grain intake and HOMA-IR in the MESA (Multi-Ethnic Study of Atherosclerosis) investigation [50]. Findings indicated that as whole grain intake increased, HOMA-IR decreased, even after adjusting for potential mediating factors.

Cross-sectional outcomes were also the focus of the Insulin Resistance Atherosclerosis Study by Liese et al. [51]. Fiber intake was measured by a food frequency questionnaire. Fiber consumption was not reported as grams per 1000 kcal. Findings revealed that fiber intake was associated directly with insulin sensitivity.

In general, it appears that the link between fiber intake and insulin resistance is strong and consistent. Current dietary recommendations in the United States encourage significant amounts of whole-grains and high-fiber foods, consistent with the literature [52].

#### *4.3. Dietary Protein*

Fewer studies have investigated the relationship between protein intake compared to carbohydrate consumption and insulin resistance. Overall, findings have been mixed. In a large, cross-sectional study that included 5675 non-diabetic subjects, a food frequency questionnaire (FFQ) was employed to measure total protein intake [32]. Across quartiles of insulin resistance, mean protein intake levels, reported as a percentage of total energy, increased as HOMA-IR increased (*p* = 0.001), after adjusting for potential confounders.

Protein consumption has an insulinotropic effect. In short, protein intake results in insulin release, which leads to increased glucose clearance from the blood. Despite these predictable outcomes, insulin resistance findings of short-term intervention trials vary, depending on the characteristics of the sample and the source of the protein. For example, according to Kahleova et al., in a 16-week randomized controlled trial, plant-based protein intake was predictive of reduced insulin resistance, whereas animal protein had the opposite effects [53]. However, most of the protein and insulin resistance associations of Kahleova's study were no longer significant after controlling for changes in BMI and energy intake, suggesting that nearly all the relationships were driven by weight loss. On the other hand, according to Adevia-Andany et al., the association between animal protein intake and insulin resistance is independent of body mass index [54].

Several other short-term trials show that protein intake has no effect on insulin resistance when the sample is healthy [55,56]. However, not all studies agree. Some indicate that protein consumption reduces plasma insulin levels [57].

In obese subjects, the influence of protein on insulin resistance is unpredictable. As shown in a review by Rietman et al., some investigations indicate that protein intake improves insulin resistance, whereas other studies indicate there is no effect [55]. Still others suggest that the outcome is dependent on whether or not weight is lost [55].

Prospective cohort investigations have also produced varied results. In a cohort study with 1205 subjects, Asghari et al. determined that intake of several individual branchchain amino acids was related to increased risk of developing insulin resistance over 2.3 years [58]. However, total branch-chain amino acid consumption was not related to incident insulin resistance.

In another prospective cohort investigation, Chen et al. followed 6822 participants for over 20 years [59]. Baseline protein intake, particularly animal protein consumption, was positively related to HOMA-IR, increased risk of pre-diabetes, and diabetes. Total plant protein was not associated with any of the metabolic problems, including insulin resistance or diabetes [59].

According to Rietman et al., increasing amino acid levels in the blood by consuming protein can lead to insulin resistance by preventing muscle glucose transport and phosphorylation of glucose with ensuing decreased glycogen synthesis [55]. Hence, protein intake can contribute to alterations in insulin sensitivity and promote insulin resistance [60]. Long-term investigations are likely to capture this process better than studies of short duration or cross-sectional studies.

#### *4.4. Dietary Fat*

Research findings also vary concerning the relationship between dietary fat intake and insulin resistance. In a high-quality intervention by Samaha et al., after six months on a high fat diet, insulin sensitivity improved among obese subjects [61]. In another investigation, Bisschop et al. designed a cross-over study using six healthy men [62]. Subjects ate each of three isocaloric diets for 11 days. The diets were low-fat with high carbohydrate, intermediate fat and intermediate carbohydrate, and high fat with low carbohydrate. Insulin sensitivity was measured using the clamp method. The ratio of fat to carbohydrate in the diets had no effect on glucose uptake. However, glucose disposal tended to increase as the fat to carbohydrate ratios in the diets increased.

Bradley et al. used a RCT to test the effect of a low-fat versus a low-carbohydrate weight reduction diet (−500 kcals) on insulin resistance in 24 overweight/obese adults over an 8-week period [63]. Insulin action was measured using the clamp method. Significant weight loss occurred in both groups but there was no difference between the two groups in insulin resistance at the conclusion of the study.

Using a cross-sectional design and 7-day weighed food records, no relationship was found between saturated fat intake and serum insulin concentrations in nearly 200 middleaged men [31]. However, in a sample of 389 older men, ages 70–89, intake of polyunsaturated fats was inversely associated with insulin levels and saturated fat intake was directly related to insulin concentrations [64]. Additionally, using cross-sectional data from the Inter99 study, total fat intake, expressed as a percentage of total kcal, was not related to quartiles of HOMA-IR [32].

According to Rivellese et al., animal studies indicate that insulin action is not only affected by the amount of fat consumed, but the type of fat also has a significant influence [65]. Specifically, saturated fatty acids tend to increase insulin resistance, whereas long- and short-chain omega-3 fatty acids tend to enhance insulin sensitivity.

As part of the multicenter KANWU study, 162 healthy subjects were fed an isoenergetic diet for 3-months with either a high level of saturated or monounsaturated fatty acids [66]. Insulin sensitivity was 12.5% lower on the saturated fat diet and 8.8% higher when the focus was on monounsaturated fat intake. The difference in insulin sensitivity was not present when total fat intake was high.

In a 16-week RCT by Kahleova and Fleeman et al., overweight subjects were randomized to follow a randomized low-fat vegan (*n* = 38) or control diet (*n* = 37) [67]. HOMA-IR

was used to index insulin resistance. The authors concluded that even after adjusting for differences in energy intake and body mass, decreased intakes of saturated and trans fats and increased consumption of polyunsaturated fats decreased fat mass and insulin resistance.

High quality clinical studies designed to assess the effect of dietary fat on insulin resistance typically use isocaloric substitution methods to prevent changes in body weight. This safeguard is crucial for accurately determining the influence of dietary fat on insulin sensitivity. However, one of the most important contributors to insulin resistance is weight gain and obesity. Hence, inhibiting weigh gain by using an isoenergetic diet may result in false judgments about the effect of dietary fat on insulin resistance because weight gain is prevented. In the general public, ad libitum, high fat diets often lead to weight gain [68–70]. In short, it is likely that many dietary fat and insulin resistance intervention studies have questionable external validity because of the use of this common substitution strategy.

#### *4.5. Effect Modification*

Rarely have researchers studied the relationship between diet composition and insulin resistance across tertiles of carbohydrate, protein, and fat intake considered separately. Nor have researchers evaluated the association between diet composition and insulin resistance across categories of physical activity. In the present study, there were several unique effect modification findings. For example, none of the primary macronutrients were related to insulin resistance when evaluated within fat intake tertiles. Specifically, total carbohydrate consumption was not related to insulin resistance in adults when the focus was only on those reporting a low-fat, a moderate-fat, or a high-fat diet. Yet, total carbohydrate consumption was directly associated with insulin resistance when participants were not restricted to fat intake tertiles. These results suggest that the relationship between carbohydrate intake and insulin resistance is partly due to the wide-ranging variation in fat intake across the U.S. diet. In short, when delimited to individual tertiles of fat intake, the shared variation necessary for the carbohydrate and insulin resistance association to manifest itself is not sufficient, but when differences in fat intake are not restricted, carbohydrate intake is a significant predictor of insulin resistance.

The same pattern appears to be true for both protein and fat intakes. Specifically, there was no relationship between either of these primary macronutrients and insulin resistance when the associations were confined to fat intake tertiles. However, like carbohydrate consumption, protein and fat intakes were each significant predictors of insulin resistance when the wide-spread variation in fat intake was not restricted to tertiles. Apparently, differences in total fat intake play an important role in the relationships between macronutrient consumption and insulin resistance in U.S. adults.

Effect modification across tertiles of carbohydrate and protein also seems to affect the macronutrient and insulin resistance associations. This is logical because of the intercorrelations among the macronutrients. However, confining subjects to tertiles of carbohydrate or tertiles of protein appears to have less influence on the macronutrient and HOMA-IR relationships than restricting subjects to tertiles of dietary fat.

Also of interest, each of the three primary macronutrients were related to HOMA-IR, but only within the highest protein intake tertile. Macronutrient consumption was not predictive of insulin resistance in adults reporting low or moderate protein intake.

Effect modification was also tested within three categories of physical activity. Only two of the nine associations were significant and both relationships, carbohydrate and fat, were within the low physical activity category. These findings suggest that macronutrient intake may not play as significant a role in insulin resistance when physical activity levels are moderate or high. It could be that if physical activity levels are sufficient, macronutrient consumption may be less important to the development of insulin resistance.

#### *4.6. Application of the Results*

As shown in Tables 2 and 3, total carbohydrate and simple carbohydrate consumption were related directly to insulin resistance in U.S. adults. However, fiber, protein, and dietary fat consumption, particularly unsaturated fat, were inversely related to HOMA-IR. Consequently, the temptation would be to conclude that a diet with less carbohydrate, particularly fewer simple carbohydrates, and with more fiber, protein, and unsaturated fat, should be recommended. Of course, such advice would be overreaching the cross-sectional design of this study. Instead, the present results should be considered as additional evidence supporting other investigations that have found similar dietary patterns associated with insulin resistance.

#### *4.7. Intricacies of the Diet and Insulin Resistance Relationship*

Dietary relationships are complex. An inherent issue associated with the study of diet composition is that higher consumption of one macronutrient usually translates into lower intake of another. For example, when dietary fat is decreased in the diet, carbohydrate intake is usually increased. Foods are not eaten in isolation. The obvious question is whether the outcome is caused by the increased intake of carbohydrates or the decreased consumption of dietary fat, or some other combination? Randomized controlled trials (RCT) are especially vulnerable to this issue because diets are assigned and manipulated. Compliance is also a concern and can have a significant influence on findings.

Insulin resistance is not a simple phenotype. It appears that different tissues have different levels of sensitivity to insulin. Without question, the complexity surrounding the biology of insulin action has led to multiple interpretations of the relationships between diet composition and insulin resistance [1].

There are a number of investigations in the literature about diet composition and insulin resistance. Comparing these investigations is challenging for a variety of reasons. For example, was the study design cross-sectional, prospective, or based on an intervention? Was it short-term or long-term? Was there a control group? What was the composition of the control diet? Did participants gain weight or lose weight? What was the protein source, animal or plant? What was the composition of the dietary fat, saturated or unsaturated? What was the composition of the carbohydrate, simple or complex? How much fiber was in the diet? Was the fiber soluble or insoluble? Were participants younger or older, normal weight, overweight, or obese? Were subjects diabetic or non-diabetic? Clearly, studies focusing on the relationship between diet composition and insulin resistance have produced mixed results partly because investigations that look similar on the surface often have important differences in their samples, designs, and measurement methods.

#### *4.8. Weaknesses and Strengths of the Study*

The present investigation had multiple weaknesses. Because the study was based on a cross-sectional design, cause-and-effect conclusions cannot be applied. Additionally, there was only a single measure of protein consumption, i.e., total protein intake. It was not divided by NHANES into animal-derived or plant-based protein. Similarly, fiber intake was not categorized by NHANES as soluble or insoluble, but only as total grams of fiber. Moreover, although two 24-h dietary recalls were obtained from each subject, and the sample was very large (>5000 individuals), in general, the more dietary assessments, the more representative the dietary data will tend to be.

This investigation also had several strengths. First, participants were randomly selected using a four-stage sampling model, making the results generalizable to the U.S. civilian, non-institutionalized adult population. Second, a large (*n* = 5665), multi-racial sample was utilized, including subjects who were Mexican American, Non-Hispanic Black, Non-Hispanic White, Non-Hispanic Asian, Other Hispanic, or Other Race/Multi-racial. Third, carbohydrate intake was divided into total carbohydrate, starch, simple carbohydrate, and fiber. Fat was broken into total fat, polyunsaturated, monounsaturated, unsaturated, and saturated fat. Fourth, a number of potential confounding factors, demographic and lifestyle, were controlled statistically to minimize their influence on the results. Fifth, although the cross-sectional design employed in this study has weaknesses, in correlational research, dietary intake can be studied using an ad libitum perspective because usual macronutrient

intake is the focus. The present investigation took advantage of this strength. Participants were required to report what they had eaten during the previous 24 h, on two different days. As a result, degree of insulin resistance was evaluated based on usual macronutrient consumption, without concern for artificial manipulation of intake, or lack of compliance among participants.

#### **5. Conclusions**

In conclusion, evidence from the present investigation, conducted using participants representing the U.S. adult population, clearly shows that diet composition accounts for differences in insulin resistance. Higher intakes of carbohydrate predicted higher levels of insulin resistance. However, a closer look indicated that starch consumption was not related to insulin resistance, and fiber intake fits hand-in-hand with increased insulin sensitivity. Furthermore, elevated consumption of simple carbohydrates was predictive of higher levels of insulin resistance. Additionally, higher intakes of protein predicted lower levels of insulin resistance, but not with protein intake divided into tertiles. Also, higher amounts of fat consumption, particularly unsaturated fat, predicted lower levels of insulin resistance. Adjusting for differences in many demographic and lifestyle variables seemed to have little influence on the relationships. However, testing for effect modification indicated that the wide-range of dietary fat intake in the U.S. may play an important role in the macronutrient and insulin resistance associations. Moreover, evaluation of effect modification also showed that moderate to high levels of physical activity may reduce the role of macronutrient intake on insulin resistance. Overall, because numerous studies focusing on diet composition and insulin resistance have been conducted to date with broadly differing results, more investigations are needed to untangle the complex associations between diet composition and insulin resistance.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of the National Center for Health Statistics, now referred to as the Ethics Review Board (ERB). The ethical approval codes for NHANES data collection for 2011–2018 are: Protocols #2011-17, #2018-01.

**Informed Consent Statement:** Written informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** All data supporting reported results can be found online as part of the National Health and Nutrition Examination Survey (NHANES). The data are free and can be found at the following website: https://wwwn.cdc.gov/nchs/nhanes/Default.aspx. (accessed on 21 February 2022).

**Acknowledgments:** Much appreciation is extended to the NHANES technicians who performed the measurements and gathered the data. Also, to those who participated as subjects in the survey.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


## *Review* **Taurine Supplementation as a Neuroprotective Strategy upon Brain Dysfunction in Metabolic Syndrome and Diabetes**

**Zeinab Rafiee 1,2, Alba M. García-Serrano 1,2 and João M. N. Duarte 1,2,\***


**\*** Correspondence: joao.duarte@med.lu.se

**Abstract:** Obesity, type 2 diabetes, and their associated comorbidities impact brain metabolism and function and constitute risk factors for cognitive impairment. Alterations to taurine homeostasis can impact a number of biological processes, such as osmolarity control, calcium homeostasis, and inhibitory neurotransmission, and have been reported in both metabolic and neurodegenerative disorders. Models of neurodegenerative disorders show reduced brain taurine concentrations. On the other hand, models of insulin-dependent diabetes, insulin resistance, and diet-induced obesity display taurine accumulation in the hippocampus. Given the possible cytoprotective actions of taurine, such cerebral accumulation of taurine might constitute a compensatory mechanism that attempts to prevent neurodegeneration. The present article provides an overview of brain taurine homeostasis and reviews the mechanisms by which taurine can afford neuroprotection in individuals with obesity and diabetes. We conclude that further research is needed for understanding taurine homeostasis in metabolic disorders with an impact on brain function.

**Keywords:** 2-aminoethanesulfonic acid; neurodegeneration; brain metabolism; diabetes; obesity

#### **1. Introduction**

Taurine, or 2-aminoethanesulfonic acid, was first isolated from ox bile in 1827, by Friedrich Tiedemann and Leopold Gmelin. Taurine is obtained from the diet or results from de novo synthesis through catabolism of the amino acid cysteine (Figure 1). Together with glycine, taurine is well known for bile acid amidation, producing bile salts for excretion. Taurine supplementation has been suggested to have beneficial effects on a number of disorders, for example, hypertension [1,2], congestive heart failure [3], ischemia– reperfusion myocardial injury [4], intracerebral hemorrhage [5], pulmonary fibrosis [6], obesity-induced low-grade inflammation [7]. The neuroprotective effects of taurine have received considerable attention, and there is a plethora of publications showing the ability of exogenously added taurine to prevent toxicity in neurons or astrocytes in vitro, as well as in animal models of neurological disorders (reviewed by Jakaria et al. [8]). Namely, taurine treatments have been shown to protect tissues and cells against oxidative stress (e.g., [9]), mitochondrial stress (e.g., [10]), or inflammation (e.g., [11]). In addition, brain taurine is known as an osmoregulator and neuromodulator [12,13] and is involved in numerous processes, such as the modulation of neuronal excitability, the cerebral control of the cardiorespiratory system, appetite regulation, resistance to hypoxia, osmoregulation, and anti-oxidation [14]. Enzymes that synthetize taurine show low activity in cats, dogs, and foxes, which develop pathologies when fed a taurine-deficient diet, namely, cardiomyopathy and myocardial dysfunction, retinal degeneration, neurological abnormalities, weakened immune response, pregnancy and fetal development complications, as well as gastrointestinal problems (see [15] and references therein). This is clear evidence advocating for the importance of taurine.

**Citation:** Rafiee, Z.; García-Serrano, A.M.; Duarte, J.M.N. Taurine Supplementation as a Neuroprotective Strategy upon Brain Dysfunction in Metabolic Syndrome and Diabetes. *Nutrients* **2022**, *14*, 1292. https:// doi.org/10.3390/nu14061292

Academic Editors: Silvia V. Conde and Fatima O. Martins

Received: 12 February 2022 Accepted: 15 March 2022 Published: 18 March 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Figure 1.** Synthesis of taurine in mammals from the sulfur amino acid cysteine.

Taurine is one of the most abundant metabolites in the central nervous system (CNS), whose levels show substantial variations across species, brain areas, and developmental stages (Figure 2). The particularly high concentration of taurine in the developing brain further suggests its developmental importance. Indeed, a relation between plasma taurine and neurodevelopment has been proposed [16]. This role of taurine in CNS development was made clear by experiments on cats fed a taurine-deficient diet [17]. More recent research proposes that taurine has neurotrophic effects, playing an important role in neurite outgrowth, synaptogenesis, and synaptic transmission during the early stages of brain development [18,19].

**Figure 2.** Concentrations of taurine in the plasma (**A**) and cerebral cortex of various species (**B**), in different areas of the mouse brain (**C**), and in the mouse cortex during development (**D**). Plasma taurine levels are indicated as mean and range for humans [2,20–27], guinea pigs [28,29], rat [30–36], and mice [37–41]. The plotted brain taurine concentration ranges are based on the concentrations reported in 1H MRS studies for humans [42–47], tree shrews [48], guinea pigs [49], Sprague–Dawley rats [50–55], and C57BL/6J mice [38,56–60].

#### **2. Taurine Homeostasis**

Dietary taurine is absorbed by the gut, released into the blood stream, and excreted by the kidney through urine and by the liver via conjugation to bile acids [14,61]. Submillimolar concentrations of taurine are observed in the plasma (Figure 2A), while much larger concentrations occur in organs with high energy metabolism rates, such as the heart [62–65].

#### *2.1. Brain Taurine Transport*

Taurine in the brain results from its transport from the periphery (believed to be the main source) and local de novo synthesis. In most mammals, taurine is mainly synthetized in the liver and then actively transported through the blood-brain barrier into the brain parenchyma.

Taurine, as well as hypotaurine, β-alanine, and other β-amino acids, are taken up through the blood–brain barrier into the brain by a high-affinity, low-capacity Na+- and Cl<sup>−</sup> dependent transport system [66,67]. The passive diffusion of taurine across the blood-brain barrier is negligible [14]. Taurine uptake or efflux at both luminal and albumen membranes has been proposed to be mediated by SLC6A6 transporter, also called TauT [68]. The blood-brain barrier also expresses the GABA transporter SLC6A13, known as GAT-2, which is capable of carrying taurine across membranes [69,70]. Both TauT and GAT-2 are also able to efficiently carry hypotaurine [71]. Genetic deletion of the taurine transporter (TauT) in mice reduces taurine concentrations in plasma and tissues, including the brain [37]. In contrast, genetic deletion of GAT-2 in mice increases brain taurine levels, suggesting that GAT-2 is mainly functioning as a brain-to-blood efflux system for taurine [69].

TauT is expressed in astrocytes and to a lesser extent in neurons [72,73]. GAT-2 expression appears restricted to leptomeninges and blood vessels [74]. Taurine is also transported by ubiquitously expressed volume-sensitive organic osmolyte–anion channels, commonly called volume-regulated anion channels (VRACs), that are activated by cell swelling (see [75] and references therein). Within the brain parenchyma, it has been proposed that taurine uptake is mediated by TauT, while taurine release is mostly mediated by VRACs. Furukawa et al., have shown that that taurine uptake is blocked by a TauT inhibitor and taurine release is blocked by a VRAC blocker in the developing mouse neocortex [76].

#### *2.2. Taurine Metabolism*

The synthesis of taurine occurs from the catabolism of cysteine in both neurons and astrocytes (Figure 1) and is limited by the oxidation of hypotaurine [77,78]. Cysteine dioxygenase and cysteine sulfinate decarboxylase are concerted to produce hypotaurine from cysteine. Genetic deletion of cysteine dioxygenase in mice depletes hypotaurine and taurine, while causing the accumulation of cysteine and cysteine-containing metabolites such as glutathione [48]. Genetic deletion of cysteine sulfinate decarboxylase also reduces taurine levels in the brain (four-fold less than in controls), as well as in the plasma and other tissues [79]. Either of these mouse models shows impaired development, including reduced brain volume. Cysteamine can also be converted to hypotaurine via cysteamine dioxygenase. The identity of the enzyme that catalyzes the biosynthesis of taurine from hypotaurine, which is denominated hypotaurine dehydrogenase, has remained elusive. Recently, Veeravalli et al., proposed that the oxygenation of hypotaurine to taurine is mainly catalyzed by flavin-containing monooxygenase 1 [80]. Accordingly, the developmental expression of this enzyme in the mouse brain [81] accompanies the developmental decay of brain taurine levels (Figure 2).

Neurons and astrocytes express taurine transporters (e.g., [82]) and release hypotaurine and/or taurine originating from cysteine oxidation [77,78]. However, it remains to be experimentally determined whether taurine metabolism is interdependently regulated by neurons and astrocytes, as proposed elsewhere (see discussion by Banerjee et al. [83]).

#### *2.3. Sulphur-Containing Amino Acids*

Taurine is not used for protein synthesis. In contrast, the sulphur-containing amino acids methionine and cysteine are protein components and play important roles in maintaining protein structure. While methionine is a very hydrophobic amino acid that contributes to interactions such as those between proteins and lipid bilayers, cysteine mainly participates in protein folding by the formation of disulfide bonds with other cysteine residues [84]. Methionine can be metabolized to the cofactor S-adenosylmethionine that participates in a number of metabolic pathways by acting as a methyl donor, including epigenetic regulation [85] and catecholamine metabolism (epinephrine synthesis) [86]. Such transmethylation reactions can be funneled to produce homocysteine that generates cysteine through transsulfuration [85,87]. Notably, both methionine and cysteine produced from protein degradation can generate taurine as an end-product [88].

#### **3. Taurine in Cellular Physiology**

#### *3.1. Osmoregulation by Taurine*

Cells swell and shrink when challenged with osmotic changes. The regulation of cell volume in response to extracellular or intracellular stimuli or osmotic changes is critical for cellular homeostasis. Neuronal activity is associated with changes in cell membrane polarization as a result of active ion fluxes and involves cell volume regulation (e.g., [89]). Pathological edema resulting from cellular swelling occurs in hypo-osmotic conditions or in the presence of cytotoxic ion imbalance. While water is taken up via aquaporin4 mainly expressed in astrocytes, it has been reported that both neurons and astrocytes swell during acute hypo-osmotic stress (e.g., [90]). As a reaction to cell swelling, several low-molecular-weight organic compounds will influence intracellular osmolarity.

Taurine occurs in its zwitterionic form over the physiological pH range, turning into an excellent metabolite for osmolarity regulation [14,91]. Indeed, neurons and astrocytes exposed to exogenous taurine up to 10 mmol/L are able to take up extracellular taurine without changes in cell volume [92]. Consistent with a tight regulation of taurine concentration for its action as an organic osmolyte, exposure of brain cells to cysteine or cysteamine results in elevated hypotaurine, but not taurine, levels [78]. Superfused acute mouse cerebral cortical slices regulate taurine release upon osmotic challenges [93]. Brain taurine levels decline over 2 weeks of hyponatremia in rats in vivo [94], while increasing during hypernatremia [95]. Accordingly, taurine synthesis is stimulated under hypertonic conditions in cultured neurons [78]. Astrocytes in a hyperosmotic medium accumulate taurine [96,97]. This is likely due to the increased expression of TauT for taurine uptake rather than to the stimulation of taurine synthesis [98]. In contrast, astrocytes cultured in a hypo-osmotic medium release taurine [99], a process likely mediated by VRAC [100]. While osmotic pressure is regulated by taurine, there are other effects of this compound on the balance of K+ and Ca2+, which might have implications for neurotransmission [92].

#### *3.2. Taurine as a Neurotransmitter*

Early work reported taurine uptake into synaptosomes and its release upon electrical stimulation [101,102], as well as taurine binding to synaptosomal membranes [103,104]. Such observations suggested a role of taurine as a neurotransmitter in the central nervous system (CNS); in fact, taurine turned out to be a modulator of inhibitory neurotransmission.

γ-Aminobutyric acid (GABA) and glycine are amino acids that mediate inhibitory transmission at chemical synapses. GABAergic synapses employ three types of postsynaptic receptors: the ionotropic GABAA and GABAC that are permeable to Cl− and the metabotropic GABAB. Glycine receptors are also permeable to Cl− upon ligand binding. Taurine is known to interact with GABAA, GABAB, and glycine receptors (Figure 3; [12,105]). While taurine binding to GABAA and GABAB is weaker than to GABA, taurine is a rather potent ligand of the glycine receptor [105].

Intracellular taurine concentration is estimated to be 400-fold higher than the concentration in the extracellular space [30]. Taurine concentration in the brain measured extracellularly using microdialysis is generally below 10 μmol/L, and increases by at least one order of magnitude upon depolarization [106–108]. After release, taurine acts on GABA and glycine receptors and is cleared through sodium-dependent transport (see above). Taurine release does not take place exclusively at synapses but can be of glial origin [109–112] and mediate astrocyte-to-neuron communication [110,113].

Concentrations of taurine below 1 mmol/L are rather selective for glycine receptors, as observed in neurons in the basolateral amygdala [114], supraoptic nucleus [115], hippocampus [116], nucleus accumbens [117], and inferior colliculus [118]. Above 1 mmol/L, taurine also activates GABA receptors. However, taurine was shown to act as an endogenous ligand for extra-synaptic GABAA receptors at concentrations ranging from 10 to 100 μmol/L [119].

While not modulating glutamatergic neurotransmission, taurine regulates cytoplasmic and intra-mitochondrial Ca2+ homeostasis. Therefore, taurine is able to dampen glutamateinduced Ca2+ transients in neurons, and thus intracellular Ca2+-dependent signaling mediators, and even prevent glutamate excitotoxicity [120–122]. Therefore, inhibitory actions of taurine on neuronal excitability might be attributed to a direct enhancement of GABAergic and glycinergic neurotransmission, as well as to the dampening glutamatergic neurotransmission via intracellular effects (discussed by El Idrissi and Trenkner [123]).

**Figure 3.** Schematic representation of activity-dependent taurine release modulation from neurons or astrocytes by glutamate and purines and action of taurine on inhibitory receptors. Taurine release is mainly mediated by volume-regulated anion channels (VRAC) that are activated by hypo-osmotic conditions and electrical activity and can be stimulated via glutamate metabotropic (mGluR) and ionotropic receptors (mainly NMDA and AMPA), adenosine A1 receptors (A1R), and metabotropic ATP receptors (P2Y). Taurine mediates its neuromodulatory effects by binding to GABAA, GABAB, and glycine receptors. Reuptake of taurine occurs vis the taurine transporter TauT.

#### *3.3. Modulation of Taurine Release in the CNS*

In the central nervous system, basal taurine release is largely independent of Ca2+, and a Ca2+-dependent component can be stimulated by glutamate and K+ [124–126]. The facilitation of glutamate-induced taurine release is slow and prolonged, varies across the life span, and is mediated by NMDA and AMPA receptors, as well as by kainate receptors in the developing brain [125]. Metabotropic glutamate receptors have also been proposed to modulate taurine release from acute hippocampal slices [127]. Adenosine has been proposed to modulate both basal and K+-stimulated taurine release from mouse hippocampal slices via A1 receptors [126]. While the activation of adenosine A1 receptors enhanced the basal taurine release and stimulated it in hippocampal slices from the developing mouse, it inhibited the basal but not the stimulated release in adults. Purinergic activation by ATP was also proposed to stimulate taurine efflux in cultured rat hippocampal neurons [128]. ATP caused a dose-dependent loss of taurine mediated by P2Y rather than P2X receptors, which could be blocked by a VRAC inhibitor. In sum, taurine release appears to be physiologically regulated by glutamatergic activity and their modulators (Figure 3), namely, purines.

#### *3.4. Taurine in Mitochondria*

Taurine concentrations in the brain mitochondria are in the same order of magnitude than those found in other subcellular compartments, such as synaptosomes [129]. Recently, in cultured HeLa cells, taurine concentrations in the mitochondrial matrix were also determined to be similar to those in the whole cell [130]. The authors further found that blocking the complex I with piericidin reduced taurine levels by 40%, but no substantial effects on taurine concentrations in the matrix were found when inhibiting complex II or ATP synthase [130].

Taurine amino group with a pKa of 8.6 at 37 ◦C is suitable for acting as a mitochondrial matrix pH buffer [131]. The regulation of mitochondrial pH is important for brain function, since mitochondrial metabolism in both neurons and astrocytes responds to brain activity (see [132] and references therein). The proton gradient and mitochondrial membrane potential are the drivers of the proton-motive force that produces ATP. Like other cells, neurons and astrocytes in culture show a mitochondrial matrix pH of 7.5–8 [133–135]. For example, the uptake of glutamate by astrocytes after synaptic release triggers intracellular acidification that spreads over the mitochondrial matrix [134]. The authors further showed that glutamate-induced mitochondrial matrix acidification exceeded cytosolic acidification and dissipated the cytosol-to-mitochondrial matrix pH gradient, which resulted in the modulation of metabolism and oxygen consumption [131,134,136]. On the other hand, the pH in the mitochondrial matrix of neurons increased upon exposure to excitotoxic levels of glutamate [133]. Taurine might counteract extreme mitochondrial pH fluctuations and help preserve mitochondrial physiology. Mohammadi et al., exposed mitochondria isolated from the mouse liver to a wide range of exogenous taurine concentrations and found that taurine participates in regulating mitochondrial potential, Ca2+-induced mitochondrial swelling, the activity of mitochondrial dehydrogenases, and ATP concentration [137]. Mitochondria isolated from the mouse brain or liver show inhibited mitochondrial dehydrogenases activity, collapse of mitochondrial membrane potential, induced mitochondrial swelling, and increased levels of reactive oxygen species upon exposure to ammonia, which are all mitigated by taurine [138].

Taurine is not able to act as a radical scavenger [139]. However, beneficial antioxidant effects of taurine in cells have mostly been linked to improved mitochondrial action and reduced generation of mitochondrial superoxide. Taurine administration to isolated mitochondria from liver or brain was shown to mitigate ammonia-induced mitochondrial dysfunction, including preventing or ameliorating the ammonia-induced collapse of mitochondrial membrane potential, mitochondrial swelling, ATP depletion, and increased reactive oxygen species and oxidative stress [138]. Taurine also decreased the activity of glutathione peroxidase and manganese-superoxide dismutase upon tamoxifen toxicity, which contributed to decreasing mitochondrial oxidative stress, measured through lipid peroxidation, protein carbonyl content, and superoxide radical generation [140].

Taurine is a component of mitochondrial tRNAs in taurine-containing modified uridines that are indispensable for protein translation [141,142]. This taurine modification is catalyzed by the enzyme mitochondrial optimization-1, whose deficiency impairs mitochondrial protein translation and ultimately the efficiency of respiration [143]. Several diseases have been directly associated with the lack taurine modification of mitochondrial tRNA [144,145].

In sum, taurine supplementation is proposed to improve the function of the mitochondria, contributing to the preservation of mitochondrial membrane potential, proton gradient, and matrix pH that are critical for energy metabolism and efficient oxidative phosphorylation, as well as intracellular calcium homeostasis.

#### *3.5. Taurine as an Inhibitor of Apoptosis*

Taurine was found to prevent apoptosis upon many noxious challenges (e.g., [146–148]). The most striking neuroprotective effects of taurine were observed on the reduction of apoptotic rates and the improvement of neurological outcomes upon brain ischemia. The

suggested mechanisms include the prevention of mitochondrial and endoplasmic reticulum (ER) stress. Taurine was found to attenuate mitochondria-dependent cell death in the ischemic core and penumbra of stroke models by stimulating the antioxidant machinery, preventing energy charge dampening, inhibiting the reduction of anti-apoptotic Bcl-xL and the increase of the pro-apoptotic Bax, preventing cytochrome C release from the mitochondria, and inhibiting the activation of calpain and caspase-3 [149–151]. Taurine was also found to prevent ischemia/hypoxia-induced endoplasmic reticulum (ER) stress by inhibiting the unfolded protein response via transcription factor 6 (ATF6), protein kinase R-like ER kinase (PERK), and inositol-requiring enzyme 1 (IRE1) pathways [152,153].

#### **4. Brain Taurine in Diabetes**

Diabetes and many factors of the metabolic syndrome impact the brain, leading to metabolic alterations, synaptic dysfunction, gliosis, and memory impairment [154,155]. MRS studies on rats rendered diabetic by streptozotocin administration showed increased taurine concentrations in the hippocampus (+23%) [156] and cortex (+8%) [157], which is consistent with increased brain taurine uptake in this model [31]. Non-obese, insulin resistant Goto-Kakizaki rats also display increased taurine concentration in the hippocampus (+22%), a brain area involved in learning and memory, relative to Wistar control rats [158]. Brain taurine alterations have also been reported in diet-induced obesity models. Namely, mice fed a lard-based 60%-fat-rich diet for 6 months showed increased taurine in the cortex (+7%), hypothalamus (+9%), and, most prominently, hippocampus (+12%), when compared to low-fat-fed mice [159]. Recently, we further demonstrated that a high-fat and high-sugar diet led to increased hippocampal levels of taurine after 4 weeks, which persisted for several months (ranging from +8% to +14% relative to low-fat-diet-fed controls), which were reversed by diet normalization [38]. Such increase in brain taurine levels in mice with diabetes might have resulted from a compensatory mechanism for cellular protection against metabolic syndrome.

While increased hippocampal taurine concentrations have been reported in the brain of diabetes models, that remains to be demonstrated in individuals with diabetes (reviewed and discussed in [160]). The lack of evidence on alterations of brain taurine levels in diabetes patients is inherent to the relatively low levels of taurine in the human brain (see Figure 2), and to the difficulty in distinguishing taurine peaks at the weak magnetic fields used in clinical MRS studies (discussed in [161]). However, MRS at higher magnetic fields, namely, at 7 T and above, improves the ability to examine taurine in the living human brain. While not many MRS studies on diabetes individuals are available, other neurodegenerative disorders have been more studied, including Alzheimer's disease (AD).

#### *4.1. Brain Taurine Levels in Subjects with Alzheimer's Disease*

There is a growing body of epidemiological evidence suggesting that obesity and insulin resistance increases the risk of developing age-related cognitive decline, mild cognitive impairment, vascular dementia, and AD, and molecular and metabolic mechanisms linking T2D and AD have been proposed [154,162,163]. While there are limited studies measuring brain taurine in patients with diabetes, research from the AD field might provide additional clues on taurine alterations upon neurodegeneration.

Little attention has been given to taurine concentrations measured by MRS in the brain of AD patients relative to those in healthy individuals ([164,165] and references therein). That is because most MRS studies were conducted at low magnetic fields. In a recent MRS study conducted at 7.0 T, Marja ´nska et al., found similar concentrations of taurine in AD individuals and age- and gender-matched cognitively healthy controls in the posterior cingulate cortex, a region known to be impacted by AD, and the occipital cortex [166]. Early studies on AD patients also found no substantial changes in cerebrospinal fluid (CSF) taurine levels [167,168] or post-mortem brain taurine levels [169,170]. These studies, however, might be biased by confounding effects from previous medications. Indeed, taurine levels were found reduced (up to −36%) in the CSF of individuals diagnosed with

dementia and probable AD who had never been treated with antidepressant or neuroleptic medications [171] and in individuals with advanced symptoms of AD [172]. In another study, CSF taurine levels in AD patients correlated significantly with cognitive scores [168]. Altogether, one might speculate that taurine loss in patients with AD is linked to worsened cognitive deterioration.

#### *4.2. Plasma Taurine Levels in Individuals with Dementia and Alzheimer's Disease*

Reduced levels of blood taurine (−23% to −40%) have been observed in subjects with Alzheimer's disease relative to subjects without neurodegenerative symptoms [173]. In another study, low taurine levels were associated with dementia risk but not with AD risk [174]. Therefore, the authors postulated that a low concentration of taurine might be linked to vascular dysfunction (possibly, vascular dementia) rather than to neurodegeneration. Accordingly, low levels of dietary taurine have been linked to hypertension [175], taurine supplementation in a mouse study was implicated in blood flow regulation [176], and a chronic taurine supplementation showed antihypertensive effects in a clinical trial [2]. However, not all studies associate low taurine levels to AD, and higher taurine levels in the plasma have actually been found in patients with mild cognitive impairment (+43%) and Alzheimer's disease (AD) (+49%) compared to control subjects [177].

#### *4.3. Brain Taurine Levels in AD Models*

The transgenic rat model of AD TgF344-AD rat has been reported to develop agedependent MRS alterations in brain metabolites, including increased taurine levels in the cortex (+35%) at 18 months of age, but not earlier [178]. Age-dependent increased taurine levels were also observed in the hippocampus (+16% to +21%) and cortex (+25%) of McGill-R-Thy1-APP rats, relative to controls [179]. One study on aged transgenic mice carrying the human Swedish APP mutant Tg2576 showed elevated taurine levels in the cortex (+21%) [180]. However, taurine levels were found unaltered during aging in the brain in many other studies on transgenic mouse models of AD (Refs. [181–183] and references therein). Altogether, we conclude that the current evidence points towards contrasting findings on brain taurine levels in AD patients and animal models of the disease.

#### **5. Neuroprotection by Taurine**

Neuroprotection by taurine has been reported for many models of brain injury and neurodegeneration. In animal models, taurine treatments have been reported to significantly improve functional recovery after traumatic brain injury [184,185] or ischemic stroke [149,176,186]. Not only taurine has beneficial effects against neurodegeneration, but also it can modulate inflammatory processes. Namely, it has been established that taurine dampens neuroinflammation in animal models of ischemic stroke and traumatic brain injury that develop severe gliosis (e.g., [11,184,186]).

Given its role as an inhibitory transmitter, taurine was shown to reduce seizures in a mouse model of kainite-induced epilepsy and prevent cell death in the hippocampus, as well as microgliosis and astrogliosis [187]. Furthermore, taurine was suggested to protect dopaminergic neurons in a mouse and rat models of Parkinson's disease, namely, by inhibiting neuroinflammation and microgliosis [188,189]. Taurine was found to ameliorate cellular and neurochemical alterations in the hippocampus of rodents exposed to chronic stress induced by repeated immobilization or noise exposure, with substantial improvements on memory performance [190,191]. Taurine supplementation was also suggested to afford neuroprotection and anti-apoptotic activity, as well as to reduce microglia activation, in a rat model of chronic inflammation induced by the repeated administration of lipopolysaccharide that mimics a bacterial infection [192].

In aging mice, taurine administration was reported to stimulate hippocampal neurogenesis by increasing the rate of progenitor cell formation and to induce a shift in microglia from activated to resting states [193].

Taurine has been shown to protect neurons against excitotoxicity induced by amyloidβ or glutamate in vitro [121,194]. Moreover, taurine supplementation was reported to recover spatial memory in the APP/PS1 mouse model [195] and to improve glutamatergic activity in the brain of the 5xFAD mouse model [196]. While in both models taurine failed to reduce the rate of amyloid-β deposition, taurine was reported to have the ability to decrease amyloid-β aggregation, while favoring the formation for tau protein fibrils [197].

#### *5.1. Taurine Affords Neuroprotection in Diabetes Models*

In streptozotocin-induced diabetic rats (insulin-deficient diabetes), treatment with taurine at a dose of 100 mg/kg i.p. during a month reduced oxidative stress, DNA damage, and inflammatory cytokine levels in the frontal cortex and hippocampus, contributing to improving memory performance [198,199]. A study by Agca et al. [200] demonstrated that a 2% (*w*/*v*) taurine supplementation in drinking water for 8 weeks administered to streptozotocin-treated rats ameliorated the diabetes-induced increase of the transcription factor NF-κβ, involved in inflammatory processes, and the diabetes-induced reduction of Nrf2 and glucose transporters Glut1 and Glut3 in the brain. Rahmeier et al. [201] further showed anti-apoptotic effects of taurine administration (100 mg/kg daily i.p.) in the brain of streptozotocin-treated rats. Li et al. [202] described taurine as a protector against myelin damage of the sciatic nerve in streptozotocin-treated rats through the inhibition of apoptosis of Schwann cells. In mice fed a fat-rich diet, which develop metabolic syndrome, we recently demonstrated that 3% (*w*/*v*) taurine supplemented in the drinking water for 2 months prevented memory impairment [203]. Furthermore, magnetic resonance spectroscopy (MRS) for metabolic profiling in vivo showed that taurine treatment prevented the obesityinduced reduction of the neuronal marker *N*-acetylaspartate in the hippocampus [203]. Energy metabolism impairments were also observed in the hippocampus of high-fat-dietfed mice in this study but could not be prevented by taurine. However, treatment with *N*-acetylcysteine, which acts as a cysteine donor for the synthesis of taurine as well as glutathione, fully prevented obesity-induced metabolic alterations in the hippocampus. Interestingly, it has also been proposed that taurine treatment increases brain insulin receptor density, in particular in the hippocampus [204], which could improve brain insulin sensitivity and thus have beneficial effects to counteract cognitive impairment [154,162,163]. Altogether, the available literature supports taurine administration as a way of preventing neuronal dysfunction in patients with obesity and diabetes.

#### *5.2. Taurine Effectiveness in Diabetes Management*

Taurine supplementation has shown beneficial effects on metabolic syndrome factors in both preclinical and clinical studies. We recently reported a taurine-induced improvement of glucose tolerance in female mice fed a high-fat diet during 2 months, compared to non-taurine-supplemented obese mice [202]. Similar results were described by Ribeiro et al. [205], who used 5% (*w*/*v*) taurine in drinking water for 6 months.

The plasma levels of taurine were found to be slightly lower in individuals with T2D than in healthy subjects [20,21]. Interestingly, plasma taurine was found to inversely correlate with fasting glycemia but not with glycated hemoglobin HbA1c levels [206] and to be independent of obesity or body mass index [20,22]. This suggests that taurine is involved in acute metabolic regulation and glucose homeostasis, but not in the etiology of diabetes. Indeed, plasma taurine is reduced during an euglycemic hyperinsulinemic clamp in healthy individuals [23] or during the metabolic response to exercise [207]. According to the roles of taurine in metabolic regulation, we previously observed that taurine concentration in the hippocampus of streptozotocin-treated diabetic rats could be reduced by acute glycemic normalization by means of insulin administration [156].

Given the lower levels of circulating taurine in subjects with diabetes, it has been speculated that dietary taurine supplementation might contribute to diabetes management. Accordingly, several studies on animal models of diabetes have indicated that taurine supplementation lowers glycaemia and improves insulin secretion and sensitivity

(e.g., [205,208–212]). Interestingly, it has been proposed that such effects could also be associated with taurine conjugation to bile acids, such as the formation of tauro–ursodeoxycholic acid [213].

Evidence from studies in humans remains controversial, and taurine supplementation has little or no effect on improving metabolic syndrome or T2D and its complications (reviewed in [214]). The source of controversy regarding taurine effects on diabetes might be the poor study design and the low number of subjects tested. For example, a sufficiently powered, double-blinded, randomized, crossover study, based on the administration of a daily taurine supplementation for 8 weeks found no effect on insulin secretion and action and on plasma lipid levels in overweight men with a positive history of T2D [215]. Nevertheless, the beneficial effects of taurine might contribute to protect the various bodily systems from diabetes complications.

#### **6. Conclusions**

Overfeeding and sedentary lifestyles drive the development of a systemic metabolic imbalance and the emergence of obesity and prediabetes that are strongly associated with all-cause dementia, Alzheimer's disease (AD), and vascular dementia (e.g., [216]). Obesity is associated with comorbidities such as hypertension, cardiovascular disease, metabolic syndrome, and insulin resistance or type 2 diabetes [216,217], which might modulate the genetic susceptibility to neurodegenerative disorders [218] and thus constitute a risk factor for cognitive decline [219,220]. The reported cytoprotective actions of taurine contribute to brain health improvements in subjects with obesity and diabetes through various mechanisms that improve neuronal function, such as the modulation of inhibitory neurotransmission and, therefore, the promotion of an excitatory–inhibitory balance, the stimulation of antioxidant systems, and the stabilization of mitochondria and thus of energy production and Ca2+ homeostasis. Taurine supplementation in experimental models of obesity and diabetes provides evidence for its effects in the prevention of metabolic syndrome-associated memory dysfunction, but the exact mechanisms of taurine action remain to be ascertained; this should be addressed in future studies. Based on this literature survey, we conclude that further research is indeed necessary for a clear understanding of taurine homeostasis in metabolic disorders with an impact on brain function.

In addition to taurine, the amino acids methionine and cysteine from which taurine can be produced (see Section 2.3) have been associated with obesity and metabolic syndrome [207,221,222], and the modulation of the bioavailability of sulphur-containing amino acids might provide further benefits, e.g., by stimulating the synthesis of the antioxidant glutathione (discussed in [203]).

**Funding:** The authors' research is supported by the Swedish foundation for International Cooperation in Research and Higher education (BR2019-8508), the Swedish Research council (2019-01130), the Diabetesfonden (Dia2019-440), the Direktör Albert Påhlssons Foundation, the Crafoord Foundation, the Tage Blücher Foundation, the Dementiafonden, and the Royal Physiographic Society of Lund. J.M.N.D. acknowledges generous financial support from The Knut and Alice Wallenberg foundation, the Faculty of Medicine at Lund University and Region Skåne. The authors acknowledge support from the Lund University Diabetes Centre, which is funded by the Swedish Research Council (Strategic Research Area EXODIAB, grant 2009-1039) and the Swedish Foundation for Strategic Research (grant IRC15-0067).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors have no relationships or activities that might constitute potential conflicts of interest with respect to the research, authorship, and publication of this article.

#### **Abbreviations**


#### **References**

