**4. Discussion**

In the present analysis of French community-dwelling older adults enrolled in the 3C-Bordeaux study, the frequency of DP consumption was not significantly associated with frailty, assessed using proxies of the frailty phenotype, in either cross-sectional or prospective analyses. In particular, total DP, milk, fresh DP and cheese were not associated with frailty prevalence at baseline. Similarly, these food groups were not associated with frailty risk at 10 years. Similar results were observed when frailty was assessed using the FRAIL scale, strengthening our conclusions.

Several studies have evaluated the association between DP and age-related chronic diseases and mortality [22,23,44]. Nevertheless, to our knowledge, very few studies have evaluated the relationship between DP and frailty and their results were mixed. A crosssectional study evaluated the association between dairy intake and physical function among1456 older women aged 70 to 85 years [45]. The authors observed that compared to those in the lowest tertile of dairy consumption, those in the highest tertile of consumption had significantly higher handgrip strength and lower odds for a poor Timed Up and Go while no differences were observed in the prevalence of falls. In a sample of 1871 Spanish older adults enrolled in the Seniors-ENRICA cohort [29], greater consumption of low-fat dairy products and low-fat milk in particular was associated with lower frailty risk over 3.5 years of follow-up, while no significant results were observed for whole milk, yogurt, cheese and low-fat yogurt. Contrary to the presented studies, our findings did not show any association with frailty prevalence or risk at the 10-year follow-up. Interestingly, our

results are similar to a recent analysis of the InCHIANTI study where the main objective was to evaluate the associations between adherence to a Mediterranean-type diet (MeDi) and frailty index at baseline and at the 10-year follow-up [46]. In a sub-analysis, the authors investigated the effect of individual components of the MeDi and frailty, and they observed that DP intake was not significantly associated with frailty in both the cross-sectional and prospective analyses.

Several possible explanations could justify the absence of associations between DP and frailty in our sample. First, the FFQ used to collect dietary data assessed the frequency of consumption only, while information about the quantities, which could have been interesting, were only assessed by a single 24-h dietary recall (which questions its relevance). Therefore, despite the higher consumption frequency, this intake might have been below what is recommended and therefore affected our results. In fact, in a recent analysis of the 3C-study participants describing their DP intake at baseline, it was observed that participants with the highest frequency of total DP per day consumed lower than the recommended intakes [31]. These results were in line with a previous national report where it was observed that 64% of participants aged 55 to 79 years old reported consumption below recommendations [47]. Second, in the 3C-study, total DP consumption and its sub-types have been previously shown to be associated with different eating patterns [31]. Although we have adjusted for diet quality in our analyses, we cannot exclude the possibility of some residual confounding that has led to an absence of significance. This is noteworthy as it was observed that higher total DP consumption was associated with a higher consumption of biscuits, sweets and cooked vegetables, and higher frequency of milk consumption was significantly associated with higher intakes of biscuit and sweets, a dietary pattern described as "biscuits and snacking" in the 3C-study. Moreover, it was observed that the highest frequency consumers of fresh DP had a low total energy intake. Finally, the highest frequency of cheese consumption was associated with a high consumption of cereals and grains, sweets, charcuterie, meat, poultry, and alcohol [31]. These results showed that the higher consumption of total DP or sub-types was associated with less-than-optimal diets, rich in sugar and saturated fatty acids, part of a western-type diet [48] and potential risk factors for frailty [14,49,50]. For instance, in a cross-sectional NHANES study including 4062 participants ≥50 y of age, a higher percentage of saturated fatty acid intake was associated with higher frailtyprevalence [50]. Therefore, we speculated that the null association between highest DP consumption and frailty observed here might be the result of possible positive effects of some favorable nutrients on frailty (i.e., higher protein and energy intake), but attenuated by the possible negative effects of saturated fatty acid intake and the overall diet quality, although we have controlled for components of the diet in the analyses. Third, no information was available about the quality of DP, whether they were natural or sweetened, or fermented or not. In fact, flavored milk, whole yogur<sup>t</sup> and fermented milk, dairy desserts and sweetened cheeses are all sources of added sugars, which were shown to be associated with an increased risk of frailty in an analysis of the Seniors ENRICA cohort [49]. The highest tertile of added sugars consumption was associated with a higher frailty risk (i.e., multiplied by 2.3) compared to those in the lowest tertile. Finally, unlike the analyses from the Seniors-ENRICA cohort study [29], we were not able to differentiate between types of DP consumed based on their fat content. Nevertheless, the 24-h dietary recall administered at baseline of the 3C-Bordeaux study showed that only 7% of the participants had whole-fat milk and among those, only 10% reported regular consumption of whole-fat milk while up to 25% of the sample consumed whole fresh DP and 19% consumed flavored fresh DP or yogur<sup>t</sup> with fruits (unpublished data). This might imply that factors other than the fat content of DP might play a role in the association between DP and frailty. Altogether, we speculated that the observed null association between highest DP consumption (whatever the subtypes) and frailty might be the results of interactions between the different concentrations of beneficial and harmful ingredients leading to an unbalanced quality of DP and of related dietary patterns.

The present study has some methodological limitations. First, as previously stated, we had no detailed information about the portion sizes and this would have affected our results as national recommendations emphasize the quantity consumed rather than frequency. Moreover, a high frequency consumption does not necessarily mean reaching the recommended levels, as older adults might have frequent but smaller intakes. Therefore, the inability to evaluate the portion sizes might have hindered any potential association of DP with frailty status. We also did not assess DP intake from mixed dishes, and this might have led to underestimation of the DP consumption frequency. This is an important issue to consider in future studies as milk is a recurrent constituent of several French recipes. Another limitation is that we did not adjust for important micronutrients related to DP and associated with frailty, namely, vitamin D [51,52]. Furthermore, recall bias cannot be excluded as it can lead to under or overestimation of DP intakes despite meticulous data collection. Regarding the assessment of frailty, we complemented slowness and handgrip strength with the Rosow–Breslau scale and the chair stand test, respectively, to minimize the loss of participants due to missing data. Indeed, the Rosow–Breslau scale has been shown to be strongly associated with walking [38] and the chair stand test was shown to be a good proxy for handgrip strength [39,53]. Furthermore, we were not able to check frailty incidence over 10-years at different waves of follow-up because the frailty phenotype could not be calculated at each time interval. Nevertheless, this limit was toned down when using the FRAIL scale, which identified a lower number of frail participants, but provided similar results on the DP-frailty associations in both the cross-sectional and prospective analyses. Despite these similarities, we speculate that the imbalance between frail and non-frail groups might have led to underpowered comparisons, hindering the observation of real differences if any. In addition, a selection bias cannot be dismissed, since not included participants (cross-sectional sample) were older, had lower educational levels and cognitive performance, had more frequent depressive symptoms, multi-morbidities, polypharmacy, and worse diet scores than included participants (data not shown). Finally, although we adjusted for several major covariates, some residual confounding factors cannot be dismissed. In fact, we acknowledge that the collected dietary data dates back to 2000, which is old, and this might affect the relevance of our results. Nevertheless, the French RDA applied to the year of data collection (2000–2001) is still applied till now and it has been previously reported that intakes of major food groups appeared to be relatively stable during follow-up in 3C Bordeaux [54].

Despite these limitations, the current study has several strengths. First, we focused our analyses on a large sample of French elderly consumers, known to exhibit distinctive DP consumption, notably cheese [31], within a population-based setting while adjusting for major confounders (note, less than 0.1% of 3C-Bordeaux participants were consumers of food supplements at baseline, which precluded using this data as a confounder). Second, survival analyses were performed to check if there is any competitive risk with death (data not shown). We observed that DP exposures were not significantly associated with mortality, eliminating the selection bias leading to survival effect often faced in prospective studies involving older adults. Moreover, we confirmed our main results using a different scale to assess frailty and when keeping participants who exhibited dependency in both cross-sectional and prospective studied samples, which allowed us to further decrease the selection bias (i.e., frailty being considered as a pre-dependency stage and risk factor for disability [35,55]).

In conclusion, we did not observe any association between DP consumption, whatever the sub-types, and frailty prevalence or incidence among this sample of French older adults. Studies on this topic are scarce and future studies are still needed while taking into consideration the identified limitations, such as the potential benefits/risks ratio of DP nutrient contents. In the meantime, and beyond frailty, older adults are encouraged to follow French nutritional recommendations for DP consumption (2 to 3 times/d) as their benefits on the general well-being of older adults to prevent osteoporosis and malnutrition are well established, and recent large-scale settings have also suggested their protective ability to prevent chronic diseases and mortality.

**Author Contributions:** Conceptualization, C.F.; Formal analysis, H.P.; Methodology, B.R., H.P. and C.F.; Project administration, C.H. and C.F.; Resources, C.S. and C.F.; Supervision, C.F.; Validation, C.F.; Writing—original draft, B.R. and H.P.; Writing—review & editing, V.C., C.H., C.S. and C.F. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** The 3C study was approved by the Consultative Committee for the Protection of Persons participating in Biomedical Research at Kremlin-Bicêtre.

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

**Data Availability Statement:** Data described in the manuscript, code book and analytic code will be made available upon request: http://www.three-city-study.com/ancillary-studies.php (accessed on 5 April 2021).

**Acknowledgments:** The Three-City Study is conducted under a partnership agreemen<sup>t</sup> between the Institut National de la Santé et de la Recherche Médicale (INSERM), Victor Segalen—Bordeaux2 University and the Sanofi-Synthélabo company. The Fondation pour la Recherche Médicale funded the preparation and beginning of the study. The 3C-Study is also sponsored by the Caisse Nationale Maladie des Travailleurs Salariés, Direction Générale de la Santé, Conseils <sup>R</sup>égionaux of Aquitaine and Bourgogne, Fondation de France, Ministry of Research-INSERM Program Cohortes et collections de données biologiques, the Fondation Plan Alzheimer (FCS 2009–2012), the Caisse Nationale pour la Solidarité et l'Autonomie (CNSA) and the "Programme Longévité et vieillissement", COGICARE 07-LVIE 003 01, the FRAILOMIC Initiative (FP7-HEALTH-2012-Proposal No. 305483-2).

**Conflicts of Interest:** The authors declare no conflict of interest regarding the current work.
