*3.1. Participants*

Out of a total of 2389 individuals (1540 men at baseline and 849 women who participated in the 15-year assessment), 1071 (694 men and 377 women) were ≥60 years. Of these, 262 (163 men and 99 women) were excluded from this analysis because they met one or more of the exclusion criteria: 9 weighed ≥120 kg, 98 were affected by lower limb

prostheses, plates or screws, 2 were unilaterally affected by a lower limb amputation, 98 did not provide a full body scan, 112 did not perform a TUG test, 66 did not complete an FFQ, and 20 reported excessively high or low daily FFQ-derived energy intakes. Thus, analyses included data from 809 individuals (531 men and 278 women).

#### *3.2. Characteristics of Participants in the Study Sample*

Key characteristics are described pooled and by sex in Table 1. Participants' ages ranged from 60 to 95 years, with 34% identified as female. The DII scores for the sample ranged from −2.7 to 2.5. Median DII scores for women were 0.8 (interquartile range: −0.2 to 1.5) and 0.4 (interquartile range: −0.4 to 1.2) for men. Compared with men, women had higher TUG and DII scores (*p* = 0.02 and *p* = 0.003, respectively), lower ALM/height<sup>2</sup> and reported lower levels of mobility (*p* < 0.001 and *p* = 0.001, respectively). No differences were detected in proportions of men and women with low muscle mass and low muscle function combined (8.6% vs. 10.9%, *p* = 0.31).

**Table 1.** Key Characteristics of the Participants; Data are Shown for All, and According to Sex.


DII, dietary inflammatory index; ALM/h2, appendicular lean mass/height2; TUG, Timed-Up-and-Go. Data are presented as mean ± standard deviation, median (interquartile range) or *n* (%). Comparison of characteristics between male and female participants was performed using independent sample *t* test with parametric continuous variables, Mann–Whitney U test with non-parametric continuous variables, and Chi-square test with categorical variables. \* Missing values: 4 for mobility level. † ALM/height<sup>2</sup> cutpoints: <7.87 kg/m<sup>2</sup> for men, <6.07 kg/m<sup>2</sup> for women.

> Table 2 shows the total daily energy intake and nutrient intake of participants which was used to calculate DII scores. Calcium intake was similar between men and women; however, men had higher intakes of energy, protein, carbohydrate, fats and alcohol (*p* < 0.001 for all).

#### *3.3. Dietary Inflammatory Index and Muscle Mass and Muscle Function*

Table 3 shows the results of linear and logistic regression modelling for the association between DII and low muscle mass and low muscle function, alone and combined. A negative association was observed between DII and ALM/height<sup>2</sup> in the unadjusted model (β = −0.13, SE = 0.04 for model 1). This association persisted after adjustment for age, sex and body fat percentage (β = −0.05, SE = 0.02 for model 2) and for the interaction of age and sex (β = −0.05, SE = 0.02 for model 3). Repeating the statistical analysis with model 2 but including mobility as a covariate did not change the association (β = −0.05, SE = 0.02).


**Table 2.** Total Daily Energy and Nutrient Intake of Participants; Data is Shown for All, and According to Sex.

IQR, interquartile range. Comparison of daily dietary intakes between females and males was performed using Mann–Whitney U test.

> A positive association between DII score and ln (TUG) was observed in the unadjusted model (β = 0.03, SE = 0.01 for model 1). This association remained significant after adjustment for age, sex and body fat percentage (β = 0.02, SE = 0.01 for model 2) and for the interaction of age and sex (β = 0.02, SE = 0.01 for model 3). Repeating the statistical analysis with model 2 but including height as a covariate did not change the association (β = 0.01, SE = 0.01).

> Each one-unit increase in DII was positively associated with a 33% increase in combined low ALM/h<sup>2</sup> plus TUG > 10 s in the unadjusted and adjusted logistic model (OR 1.34, 95% CI 1.08 to 1.67 for model 1; OR 1.33, 95% CI 1.05 to 1.69 for model 2). There were no significant interactions found between covariates.


**Table 3.** Linear and Logistic Regression Results for the Association between DII Score and Low Muscle Mass and Low Muscle Function, Alone and Combined, for All Participants, Geelong Osteoporosis Study (GOS), 2001 to 2014.

β, standardised beta coefficient; SE, standard error; ALM/h2, appendicular lean mass/height2; ln(TUG), natural log-transformed Timed-Up-and-Go; OR, odds ratio; CI, confidence interval. Standardised beta coefficients and standard errors and odds ratios and confidence intervals are for DII scores. \* Model 1: unadjusted. † Model 2: adjusted for age, sex and body fat percentage. ‡ Model 3: adjusted for co-variables in model 2 as well as sex\*age interaction term.

## **4. Discussion**

In this cross-sectional study, higher DII score, indicating a more pro-inflammatory diet, was associated with lower muscle mass, poorer muscle function and higher likelihood for the combination of low muscle mass and low muscle function. The sex\*age interaction term identified that the relationship between DII and ALM/height<sup>2</sup> and ln (TUG) was different between men and women and that the size of this difference increased with increasing age.

In this study, higher DII score (indicating a more pro-inflammatory diet) was associated with lower ALM/height2, indicating lower muscle mass. Other studies examining the relationship between DII and muscle mass have reported similar results. In a prospective longitudinal study of 1098 individuals aged 50 to 79 years from the Tasmanian Older Adult Cohort Study (TASOAC), inverse associations were shown between energy-adjusted DII scores and appendicular lean mass in men but not in women after controlling for age and percent body fat (semi-adjusted model) [24]. Findings from a study of 466 Chinese boys and girls aged 6 to 9 years reported that DII score was inversely associated with relative appendicular skeletal muscle mass (ASM/height2) [40]. In a longitudinal study with 494 female participants aged 21 to 89 years from the GOS, while the DII was not predictive of skeletal muscle index (ALM/height2) significance increased with adjustment; thus, suggesting a higher DII score was associated with increases in skeletal muscle index [23]. Together, these findings highlight the potential role for overall diet quality based on the inflammatory potential of diet in the maintenance of skeletal muscle mass across the life course.

Other studies that have looked at anti-inflammatory dietary patterns like the Mediterranean diet and muscle mass have produced differing results [41,42]. In a cross-sectional study of women aged 18 to 79 years from the Twins UK study, higher adherence to a Mediterranean diet was associated with higher FFM% (fat-free mass/weight × 100) after adjustment for age, physical activity, smoking, energy and protein intake and misreporting; specifically, FFM% was 1.0% higher in the highest quartile (Q4) compared to the lowest quartile (Q1) [41]. In contrast, in a study conducted in Iran among community-dwelling men and women with an average age of 66 years, no differences in mean muscle mass were detected in the higher tertiles of a Mediterranean dietary pattern compared with the lower tertiles; although the direction of the association was as expected (i.e., lower adherence to a Mediterranean dietary pattern was associated with lower muscle mass) [42]. These inconsistencies may be due to a range of factors including insufficient sample size, the use of samples with different age ranges (e.g., some including both pre and postmenopausal women), different ranges of the DII scores and the different settings.

Another finding of our study was that higher DII score is associated with higher ln (TUG). Handgrip strength, a clinical marker of poor mobility, and gait speed can also be used to assess low muscle function for the diagnosis of sarcopenia [28]. Several studies have explored these measures, but results have been inconsistent. In a cohort study of 1948 individuals aged 60 years or older from the Seniors-ENRICA study, higher DII score was associated with slow gait speed, as a low score in the Short Physical Performance Battery (SPPB) test [20], which is somewhat comparable to our study findings. In a study of 321 individuals aged 70 to 85 years, low gait speed and low grip strength were positively associated with higher DII scores [43]. Furthermore, in a cross-sectional study of 78 frail individuals aged 65 years or older from South Korea, a higher SPPB score was associated with lower levels of TNFα, suggesting that improving muscle function may lower levels of inflammation [44]. Conversely, no significant associations have been observed between DII and gait speed or handgrip strength in other studies [22,24,40]. The inconsistency of results could be due to different methods used to assess muscle function, age-group differences and limited DII score ranges. More research is therefore required to determine the effects of dietary inflammation on muscle function in older adults.

The final component considered in this study was a combination of low muscle mass and low muscle function as a representation of sarcopenia. We found that higher DII score was associated with a higher likelihood for these components combined. Our findings are in agreemen<sup>t</sup> with a cross-sectional study of 300 individuals aged 55 years or older from Iran by Bagheri et al. [22], who found that those in the top tertile of DII had higher odds of sarcopenia than those in the bottom tertile. In a study of 1344 postmenopausal Korean women aged 50 years or older, a pro-inflammatory diet, as determined by DII score over the median, was associated with increased odds for sarcopenic obesity. However, this result was attenuated and did not reach statistical significance after adjustment for age, family income, regular exercise, education status, smoking and female hormone supplements [25]. Interestingly, a pro-inflammatory diet was associated with increased odds for osteosarcopenic obesity in the adjusted model [25]. However, a direct comparison between these results and ours is made difficult by several factors; sarcopenic obesity is a distinct condition [31], two different criteria were used to define sarcopenia (low muscle mass and function vs. low muscle mass alone) and muscle mass was adjusted for body size in different ways (ALM/height<sup>2</sup> vs. ASM/weight %). Cut-off values also differ because of ethnicity, body size, lifestyles and culture between European and Asian populations [45], and there is no consensus about which method is best for adjusting for body size [31].

To date, evidence that a pro-inflammatory diet is associated with sarcopenia has been limited. Previous studies have mainly focused on the association of "healthy eating", high fruit and vegetable intake, and Mediterranean anti-inflammatory dietary patterns with sarcopenia [42,46–49]. Our findings support those observed by Hashemi et al. [42] who found that a Mediterranean dietary pattern was associated with lower odds for EWGSOPdefined sarcopenia among community-dwelling men and women with an average age of 66 years. Given that the inflammatory potential of the Mediterranean diet is comparable to a DII score of −3.96, indicating a strong anti-inflammatory potential, in a similar way, these results are consistent with our study findings [50]. In contrast, Chan, Leung and Woo [47] found no association between Mediterranean Diet Score (MDS) and the Asian Working Group for Sarcopenia (AWGS)-defined sarcopenia in a prospective cohort study of community-dwelling Chinese men and women aged 65 years and older. The absence of associations may be due to the differences in the Chinese diet compared to the traditional Mediterranean diet. Additionally, cut points for muscle mass were lower (<7.0 kg/m<sup>2</sup> for men and <5.4 kg/m<sup>2</sup> for women) than those used in this study, which may have affected the case ascertainment of sarcopenia.

Consistent with the findings of this study, other studies have suggested that a proinflammatory diet, as measured by the DII, is associated with increased hip fracture risk and frailty, which are associated with loss of muscle mass and/or function [18,19,51,52]. Research indicates that chronic low-grade inflammation plays a role in the development of sarcopenia, and that diet plays a role in the regulation of chronic inflammation, supporting the findings of this study that the inflammatory potential of the diet may be a modifiable risk factor for sarcopenia [9,13,53].

There were several strengths to this study. The secondary analysis of existing data from the GOS allowed for access to a large data set. Not only was this efficient but the random sampling method used in the GOS strengthened the external validity of this study by achieving a sample that was representative of the underlying population [26]. Objective measures were used to assess muscle mass and muscle function. Furthermore, a systematic approach was adopted for addressing confounding and effect modification with adjustment for a number of variables. The validity of the Cancer Council Victoria FFQ has been assessed against weighed food records in Australian men and women ranging from 31 to 75 years [33] and in young to middle-aged women [34] with good agreement; thus, confirming that the FFQ used was a valid tool in the assessment of dietary intake in our study sample of Australian men and women.

Despite its strengths, our study had several limitations. The primary limitation of cross-sectional studies is the inability to account for temporality, and as a result, causality cannot be established. Reliance on long-term memory for some self-reported data may have affected the accuracy of dietary and lifestyle self-reported data, resulting in recall bias and increased random measurements error [54,55]. Despite using objective measures to confirm some self-reported data, biases may still exist. As well, the presence of selection bias due to non-response and attrition rates cannot be excluded. Additionally, the fact that data were pooled from different study periods for men and women may have introduced bias. Data also may have been affected by the exclusion criteria; as a consequence, the study findings may not be applicable to individuals who weigh ≥120 kg or who are affected by lower limb prostheses, plates or screws. The original definition of sarcopenia by EWGSOP focussed on the detection of low muscle mass. More recent definitions have turned attention to low muscle strength as the primary diagnostic criterion of sarcopenia [31,56]. In the absence of muscle strength measures in this data set, we have not adopted the latest version of the definition. Furthermore, the absence of data on 23 parameters may have limited the range of DII scores, which appear to be somewhat narrower than other studies [57]. This may have contributed to the narrow effective range of the DII score, which is about half of that normally observed in other studies that typically range from about −5 to +5 [57]. Increasing the effective range of the independent variable often increases magnitude of the observed effect [58]. Therefore, our results actually may underestimate the relationship between DII score and the combined low muscle mass and low muscle function components.
