*3.2. Effects of Web-Based Weight Loss Programs on HEI-NVS*

The total HEI-NVS score increased significantly within the intervention group from baseline (t0) to t2 and t3, but not to t1 (Figure 2 and Table 2. Descriptively, the intervention group improved over the course of the study, particularly in the vegetables, fish and meat component (Table 3). The increase of the total score corresponded to a small effect from baseline to every measurement timepoint (Table 4). Compared with the control group, which deteriorated slightly from a descriptive point of view (Table 4), the statistical analysis showed a significantly different course from baseline to t2 and non-significant to t1 and t3 (Table 2). An analysis on the effects of the web-based programs on weight loss can be found elsewhere [56].

**Figure 2.** Mean and 95% confidence interval of the HEI-NVS for intervention (*n* = 78) and control (*n* = 75) (ITT analysis).

**Table 2.** Results of the robust linear mixed model of the HEI-NVS (ITT analysis) 1.


<sup>1</sup> Unstandardized regression coefficients with standard errors in parentheses.





<sup>1</sup> Intervention (*n* = 78) and control (*n* = 75) over four measurement time points (t0: 0 months, t1: 3 months, t2: 6 months after t1, t3: 12 months after t1). Data are presented as mean values (SD).

**Table 4.** Effect sizes (Cohen's d) with 95% confidence interval of the HEI-NVS (ITT analysis) 1.


<sup>1</sup> Interpretation: |d| = 0.2: small effect, |d| = 0.5: medium effect, |d| = 0.8: large effect.

#### *3.3. Associations between HEI-NVS and Dietary, Anthropometric and Cardiometabolic Variables*

The relationship between changes in HEI-NVS and other variables over the study period is presented in Table 5. Descriptive statistics of all variables used in the analysis can be found in Table S2. The changes in total HEI-NVS score correlated inversely with changes in energy density independent of group. Moreover, a weak positive correlation was observed between the change in HEI-NVS and the change in fasting blood glucose as well as a weak negative correlation with fat-free mass from t0 to t3. Apart from these findings, the analysis showed no further correlations between changes in HEI-NVS and other cardiovascular risk profile variables.

**Table 5.** Association between changes in HEI-NVS and changes in dietary, anthropometric, and cardiometabolic variables (ITT analysis) 1.



**Table 5.** *Cont.*

<sup>1</sup> Bootstrapped Pearson correlation with biased corrected confidence interval. \* Statistically significant correlation.

#### **4. Discussion**

The main finding of the present study was that a fully automated and interactive web-based health program focusing on the dietary energy density improved the total HEI-NVS score, thereby shifting participants' diets toward the DGE dietary recommendations. Descriptively, these improvements were primarily due to improvement in the vegetables, fish, and meat components and resulted in a small effect in HEI-NVS from baseline to all three measurement timepoints. Compared to the non-interactive web-based weight loss program, however, there was only a significant advantage after 6-month follow-up (t0 to t2) and a non-significant difference after the 12-week intervention (t0 to t1) and after 12-month follow-up (t0 to t3). According to the meta-analysis published by Beleigoli and colleagues [45], none of the investigated web-based weight loss intervention demonstrated a significant advantage in diet quality over the control group. It should be noted that the included studies used very different instruments to measure dietary quality [45], which makes comparability difficult. Another meta-analysis on individuals with noncommunicable diseases [44] showed benefits of eHealth interventions on healthy eating behaviors. The definition of healthy eating behaviors used in the studies included in this analysis had little overlap with diet quality. Thus, the outcomes used for inclusion were energy intake, macronutrient composition, and core food groups such as fruits or vegetables. While core food groups are often part of diet quality indices, diet quality is otherwise distinct from diet quantity and can only be represented to a limited extent, if at all, by macronutrient composition.

A recent systematic review demonstrated that weight loss interventions can improve diet quality as measured by a HEI [29]. Included studies covered in-person and mobile health interventions, which mostly resulted in an improvement between 4 to 7 points. In our study, mean improvements in the intervention group from baseline ranged from about 2.5 to 3.8 points, depending on the time of measurement. Thus, the improvement in this study tended to be lower than in the analysis by Cheng and colleagues [29]. However, it should be noted that the values cannot be directly compared because the review includes only studies using the U.S. versions HEI-2005, HEI-2010, and HEI-2015. These differ from each other and also from the HEI-NVS partly in components and evaluation system.

Besides the effect on the HEI-NVS of the web-based weight loss program focusing on reducing energy density, it was another important finding of the study that the change in HEI-NVS showed a weak to moderate inverse correlation with the change in energy density. An inverse relationship between energy density and diet quality has already been demonstrated in cross-sectional studies in various countries, e.g., Spain, Iran or Brazil [9–11], but to the best of our knowledge not yet in an intervention study. In contrast, improvements in HEI-NVS were not correlated with improvements in cardiovascular risk profile variables or changes in energy intake [57]. In contrast, the German National Nutrition Survey II, a representative cross-sectional study, found a positive association between HEI-NVS and BMI in women and in the 5th quintile in men [57]. In this longitudinal weight loss study, however, this was not confirmed with regard to body weight. Based on the absolute reference values of the HEI-NVS, it seems plausible that a higher score is associated with a higher energy intake and therefore a higher body weight or BMI. Also, with regard to the correlation between change in energy intake and HEI-NVS, this was not confirmed in this analysis.

Interestingly, this analysis revealed a weak positive correlation of change in HEI-NVS with change in fasting blood glucose as well as a weak negative correlation with change in fat-free mass. These results are surprising because it is assumed that improvement in a diet quality index is associated with better outcomes in cardiometabolic variables. There are several possible explanations for the lack of association between HEI-NVS and improvements in cardiometabolic variables. As previously reported, the observed effects on cardiometabolic variables by the two interventions were small. The effects on cardiometabolic variables have already been studied in the context of weight loss in people with overweight or obesity [58] and are consistent with the effects found in this study. Weight loss may already explain these effects, and the influence of diet quality may be insignificant and minor in the context of a weight loss intervention.

Furthermore, the construction of a diet quality index influences whether it correlates with diet quantity and thus with body weight and possibly other anthropometric variables. A negative association was shown between the change in HEI-NVS and the change in energy density, but not the change in energy intake. It is plausible that energy intake/dietary quantity and dietary quality may overlap if a diet quality index is not constructed appropriately. It is in fact possible that the HEI-NVS does not reflect diet quality independent of quantity. Due to the lack of reference to dietary quantity, as for example in the Healthy Eating Index-2015, a complete delineation to dietary quantity is compromised, as shown by the positive association between BMI and total score in women and partly in men [57]. Thus, the HEI-NVS measures compliance with the absolute amounts recommended in the German dietary guidelines rather than independent diet quality based on component reference values, which may tend to result in higher total scores if energy intake is high [57].

The following limitations must be considered when interpreting the results. First, the reference values of the HEI-NVS and the DGE, respectively, are based on systematic literature research and thus represent an important aspect for the preventive and therapeutic effects of nutrition in addition to the dietary quantity. However, considering other dietary indices and current findings in nutritional science, it is clear that important components of dietary quality such as sodium intake or carbohydrate quality such as intake of whole grains, refined grains, or added sugars are missing. The components and construction of the HEI-NVS may ultimately be responsible for the fact that improvements in HEI-NVS were not associated with improvements in cardiometabolic variables or, on the contrary, are partly even associated with negative effects on fasting blood glucose and fat-free mass. Thus, a differently constructed diet quality index might yield a different result.

Second, the reliability of self-reported data such as dietary records is limited. Recent data with reference data based on the doubly labeled water method suggests frequent underreporting, especially among people with overweight and obesity [59,60].

Third, compared to the real-world setting, both study groups may have been additionally motivated due to the activity tracker received as incentives as well as the free medical examinations provided by the study. In addition, complete blinding was not possible as subjects were likely to recognize their assigned program based on study information received in advance.

Fourth, the COVID-19 pandemic may also have influenced both groups in their dietary behavior. Because the COVID-19 pandemic and its limitations took a seasonal course, changes in dietary behavior are difficult to differentiate from seasonal changes and influence of the COVID-19 pandemic. The consumption of numerous food groups such as fruits, vegetables, or cereals, as well as energy intake, follows a seasonal pattern [61]. Simultaneously, an influence on dietary behavior could also be observed due to COVID-19 restrictions [62].

The elaborate implementation of seven-day dietary protocols is a strength of the present study. In addition, numerous anthropometric and cardiometabolic variables could be collected in a standardized manner in the clinical study. Therefore, this randomized controlled clinical trial provides more detailed insights into diet quality during web-based weight loss interventions and the association with dietary, anthropometric, and cardiometabolic variables.
