In our long-term assessment, in an elderly population at high CVD risk, we found that an increments in the consumption of some ultra-processed foods such as snacks, fast-foods, pre-prepared dishes, processed meat, and sweets was associated with changes in WC; whereas increased consumptions of high glycemic index (GI) foods (including white bread, refined grains, potatoes, and sweets) and red and processed meats were associated solely with weight gain. Alcohol intake was associated with higher gains in both weight and WC. In turn, increased intake of some low-fat dairy products was associated with less weight, and intake of vegetables and nuts with less WC gain. In case of the associations between some food items (i.e. potatoes, red and processed meats) and weight change, these associations were attenuated after mutual adjustment for each other or after adjustment for energy; in contrast, most associations with WC changes remained significant after these adjustments. Importantly, these findings were obtained after using repeated yearly measurements of concurrent changes in foods and anthropometry, both evaluated along the same timespan. This is a novelty because previous studies using repeated measurements of diet and weight to assess concurrent changes considered a wider timespan (4 years instead of 1 year) and they relied only in self-reported measures of weight, but not in objectively measured weight and waist [3, 4]. In the PREDIMED trial anthropometric variables were repeatedly collected in duplicate (the average of 2 measurements was the analyzed value) every year by dietitians, specifically trained to follow the measurement protocol of the trial.
This work emphasizes previous findings on importance of carbohydrate quality for obesity prevention [23, 24]. Foods that are low in fiber but high in refined carbohydrates or starches, with a high GI, i.e. white bread, refined grains, potatoes or sweets, were associated with higher weight gains. These results are consistent with previous longitudinal studies [3, 25, 26] and short-term trials . Thus, it is likely that the observed positive link between refined carbohydrates or starches and weight gain is mediated through nutritional factors, such as fiber content, GI and added simple sugars. High-GI foods are less satiating, inducing hunger and overconsumption, and enhance lipogenesis, as compared to low-GI and high-fiber foods [27, 28]. Nevertheless, mutual adjustment for other food groups attenuated the link between refined grains and potatoes with weight gain, suggesting that this association may be explained by combining effect of other unfavorable foods usually consumed with refined grains and potatoes i.e. red and processed meat. Similarly, after adjusting for changes in energy, these associations were attenuated, which may indicate that a plausible mechanism of action is related to alterations of the energy balance.
Low-fat dairy products have been included in American Dietary Guidelines for disease prevention and overall health. In line with these recommendations we found an inverse association between the intake of both low-fat milk and low-fat yogurt and changes in weight; while no association was found for whole-fat milk and yogurt. The plausible mechanism might be related to calcium, casein or biopeptides, whether it could be related to fat content deserves future investigations. For comparison, prior longitudinal studies evaluating dairy-obesity relationships showed protective roles for high-fat dairy products against weight gain, and reported null association for low-fat dairy in younger and healthier American cohorts [29, 30]. The discrepancy between the findings across longitudinal studies might be attributed to differences in exposure variable categorization and measurement (baseline intake or changes over time); or to the fact that whole-fat dairy is not consumed much in this cohort, potentially limiting the ability to detect possible associations.
In concordance with a majority of prior prospective reports [3, 4, 31], this study supports the hypothesis that red and processed meat predict long-term weight gains. This positive association might be attributed to high-energy density, high content of cholesterol, saturated fatty acid (SFA), sodium and other additives as nitrates (processed meat), as well as synergistic effect with detrimental dietary or lifestyle patterns associated with meat intake [26, 32]. It has been also postulated that intake of dietary protein may help to maintain lean mass in older adults , thus the use of body weight as outcome rather than adiposity might be misleading. In our previous analysis conducted in other cohort of older subjects with overweight/obesity and metabolic syndrome, we have postulated that alternative measures of WC might serve as plausible options to assess changes in adiposity, in case in which more precise methods cannot be applied. In this regard, our investigation found positive associations between processed meat and red meat (the latter statistically non-significant), with WC changes. This is in line with the results from a recent systematic review and meta-analysis of observational studies .
Regarding alcohol use, prior studies reported a positive association with weight gain among heavy drinkers and spirits drinkers, whereas light-to-moderate alcohol intake, especially wine, was found protective . In our investigation, overall alcohol intake was consistently and positively associated with weight and WC gain, as well as in two other large longitudinal studies among middle-aged Europeans and Americans [3, 5]. Our secondary analysis showed that the different alcohol subtypes showed similar associations with weight and waist. Despite high energy content, the plausible mechanisms may include appetite enhancement  and lipid oxidation reduction , but also consumption of alcohol might mirror unhealthy lifestyles that lead to overall and abdominal obesity. However, futures studies with comprehensive set of confounders and modifiers (frequency, subtypes, alcohol history and obesity tendency) are warranted.
Recently, an increased attention has been given to the detrimental association between SSB consumption and obesity or weight gain . In turn, in our analysis we could not support those findings. However, the point estimates and most of the confidence interval showed results that were compatible with increased weight and WC in association with SSB consumption. The wide confidence intervals could be attributed to the fact that previous studies included participants of younger age groups [3, 5], in which the consumption of SSB is usually higher than that observed in the PREDIMED cohort.
Convenient and palatable ultra-processed foods contribute to obesity epidemic, supplying excessive amount of SFA, trans fat, added sugars, and sodium [38, 39]. In this regard, in our study WC gain was associated with higher consumption of sweets, snacks, fast-foods and pre-prepared dishes. Similarly, prior large longitudinal studies of younger US and European citizens found that fast foods, French fries, potato chips, sweets were associated with higher WC [5, 40]. Thus, abdominal obesity prevention should put particular attention on these foods, which due to its hyper-palatability and conveniency tend to be consumed in excess replacing healthier and more satiating options.
Dietary fiber has been associated with less visceral fat , and the possible mechanisms include satiety effect released by increasing food volume and lowering glycemic and insulinemic response to a meal , as well as improvement of gut microbiome. An inverse association of vegetables intake with WC changes in our study is in accordance with previous longitudinal studies using baseline dietary information [5, 40]. However, unlike in those studies, we did not reveal an inverse association with fruits; this might be attributed to the seasonal fluctuations in fruits intake and warrants future studies. Furthermore, nuts are a special case of fiber-rich food that is also rich in fats, albeit mostly unsaturated. It is likely that despite fiber and bioactive compounds, vegetable protein and fatty acids found in nuts exert anti-obesity effect by increasing thermogenesis, resting energy expenditure and oxidation [43, 44]. Previous analysis based on the present cohort showed that an increase in vegetable fat intake from natural sources as nuts and EVOO, implemented in the setting of MedDiet, has significant effect on weight reduction and lesser age-related increases in WC . In this longitudinal analysis we found nuts negatively associated with changes in WC, but not weight. Future investigations should establish whether the association on abdominal obesity is due to specific nutrient properties or displacement of unfavorable foods.
We acknowledge several study limitations, as these analyses are exploratory within the PREDIMED trial, and findings are limited to white, elderly Spanish subjects at high CVD risk. Whilst changes in adiposity were objectively measured at each time point (WC and body weight), anthropometry is less accurate than direct methods (i.e. imaging techniques) to assess changes in body composition in elderly. Although food groups intake was assessed using validated FFQ, the self-reporting may always be biased. However, data on dietary intake were collected yearly to capture changes over follow-up, and a dietitian checked the FFQ with the participant to ensure that no missing data exist. Moreover, the use of FFQ does not capture sufficient details on how the food is prepared and consumed. Despite using portion sizes, residual and unmeasured variation in portion sizes might influence the associations. Due to the high number of food groups studied chance finding cannot be excluded, and as in any observational study the causality cannot be inferred.
Besides the long-term prospective design, direct measurements of anthropometry, and assessment of concurrent changes in diet with anthropometry evaluated along the same timespan, strengths of this study also include large sample size, wide spectrum of foods comprised only healthy or unhealthy items, use of standardized protocols and validated tools for anthropometry and dietary measurements, assessment of habitual dietary pattern (without energy-restrictions, supplementation or physical activity program), control for covariables, and inclusion of LOCF method. In contrary to prior observational studies [3,4,5], study sample included elderly participants with a great prevalence of obesity and health risk, who comprise increasing component in our societies.
In conclusion, this prospective study on concurrent changes in diet and anthropometry performed among elderly subjects at high CVD risk, revealed that rather than focusing on total calories amount, modifications in the consumption of specific food groups have a potential to prevent overall and abdominal obesity. Future studies with more precise imaging techniques are warranted to confirm findings on WC. Finally, the effect of dietary patterns rather than individual food groups, in combination with other lifestyle determinants should be studied, in order to better extrapolate these findings into public health recommendations.