**4. Discussion**

To the best of our knowledge, this is the first study that compares dietary protein intake patterns and the percentage of IPI per day and per meal among countries. While there are studies where dietary protein intake patterns were reported, direct evidence about their differences was lacking. This study demonstrates that some dietary protein intake patterns in older adults significantly differed among countries, but they shared traits in IPI. For instance, three countries showed a similar percentage of IPID-1.0 (>60%) and IPID-1.2 (>75%) (Figure 1). These high percentages of IPI per day might represent a higher risk of lower physical functioning [9,10,33] and developing frailty [34,35].

Furthermore, the main finding of this study was that participants from all the analyzed samples reported a high percentage of IPIM-30 at breakfast (>90%) and lunch (>60%) (Figure 2a). This is relevant due to IPI at breakfast might also be attributable to breakfast skipping [36]. Although breakfast skipping is less studied in older adults, evidence in their younger counterparts offers insight about what to expect. For example, breakfast skipping was associated with lower skeletal muscle mass in young adults [37], which seems feasible in older adults because meal skipping might lead to malnutrition [38] and skeletal muscle loss [39]. Additionally, breakfast skipping is associated with a higher risk of developing type 2 diabetes in adults of different ages (18–83 year) [40]. Moreover, breakfast has been previously reported as a low-energy containing meal [36]. For instance, Huseinovic et al. [28] reported

that breakfast composed the lowest contribution to the daily energy intake in di fferent European countries, which we speculate might include IPIM-30. On the other hand, insu fficient protein intake at lunch is also associated with lower functionality and muscle mass decline [18,19]. Both eating occasions might represent an opportunity for nutritional interventions in older adults in many countries; however, breakfast appears to be a priority due to its larger percentage of inadequate protein intake.

We observed that most di fferences among countries could be attributable to women because the pattern was very similar to that observed in the whole sample. However, while it appears to be a sex-derived pattern in some variables (e.g., protein at breakfast and lunch), it is important to highlight that there were more women than men in this analysis. Therefore, men could be underrepresented and di fferent patterns attributable to di fferent sample sizes. Further research among countries with larger men samples is recommended.

As previously mentioned, the evidence comparing eating patterns among countries is scarce, and the available studies are focused on eating occasions, meal frequency, and meal energy content, rather than protein intake per meal [28,36,41]. Therefore, it is di fficult to draw comparisons with the previous evidence. However, we can see that our results are supported by drawing inferences from other studies; i.e., IPI per day may di ffer among countries [42,43], but IPI at breakfast appears to be high [43,44]. Although an "all countries" comparison might not be reliable, we believe it would be of interest to compare countries sharing a geographical area (e.g., North America, South America, and Western Europe) and among states/provinces within countries. It would lead to data about the magnitude of the problem in specific regions to design concrete strategies to overcome this problem.

Another important finding is that about 90%, 70%, and 74% of the participants did not eat enough protein at breakfast, lunch, and dinner, respectively (Table 3). Therefore, there is interest in strategies to increase total dietary protein intake [4,5,45]. We [20] and others [15,21,46] have previously reported that reaching an adequate protein content per meal (i.e., ≥30 g) is significantly associated with higher daily protein intake and lower proportion of IPI per day (<1.2 g/kg/d). However, this practice is not usual. We reported here that only about 15% of participants ate two or three meals with adequate protein content (Table 3). This low proportion of compliance might be attributable to the lack of appetite and some chewing di fficulties commonly observed in older adults [5,6,47], thus making it challenging to reach this protein amount with simple foods [47]. Some strategies, like mincing meat [48] and food enrichment with protein [46], may help overcome the di fficulties mentioned earlier. On the other hand, about 45% of the participants did not eat enough protein at any meal (Table 3), which may be a concern because their muscle protein synthesis might not be stimulated to its maximum level at any meal [11]. This low protein intake might lead to a negative protein balance, muscle loss, and possibly accelerating the functional decline in the elderly [49,50].

In this study, the combined data suggested that about 40%, 60%, and 80% of the participants showed IPI per day for the <0.8, <1.0, and <1.2 g/kg criteria (Table 3). These results are similar to those recently reported by Hengeveld et al. [51]. In their meta-analysis, they found that when protein intake was divided by the actual body weight (g/kg of actual body weight/day), the IPI was 29.1, 54.3, and 75.7%, respectively. Nevertheless, when they used the adjusted body weight (g/kg adjusted body weight/day), the percentages of IPI were lower, 21.5%, 46.7%, and 70.8%, respectively. It is important to note that Hengeveld et al. analyzed data from adults aged ≥55 year, belonging to large cohorts (*n* = 410 to 2660) from di fferent countries (U.S.A., the Netherlands, U.K., Canada, Finland, and Italy) and that they performed complex statistics to came up with the previously mentioned percentages. However, despite these di fferences in comparison to our study, they showed similar results when the protein intake was reported with the actual body weight for the 1.0 and 1.2 g/kg/d criteria. Nonetheless, the authors sugges<sup>t</sup> using the adjusted body weight to decrease the chances of over- or underestimation [25,51].

It is important to consider this study's limitations. First, meal identification varied among studies (participant-identified [18,29]; time of the day [24]; time of the day and meal composition [14]); therefore, it impedes us to analyze the specific hour when occurred. Thus, meals could be close to one another in the time of the day but were categorized di fferently due to cultural traits [36]. Hence, it is di fficult to fully conclude di fferences among meals, at least at their respective timing (i.e., breakfast, lunch, and dinner) [36]. This is one of the most important limitations to consider when translating results regarding the timing of IPI and functionality in older adults [17,18]. Consequently, the number of meals per day that reach adequate protein content might o ffer a better approach than meal timing labelling. This number has shown significant associations with higher muscle mass, strength, and lower risk of physical disability, even in samples that showed di fferent protein eating patterns (e.g., eating more protein at dinner [15] vs. lunch [20]). Thus, it is less likely to be a ffected by di fferences in meal labelling [36].

Second, we did not analyze snacks because its categorization is more complex than with the main meals [36]. Nevertheless, further studies should include them because some protein-rich foods (e.g., dairy) might be commonly eaten at snacks [22,43]. In addition, snacks might represent as much energy as a main meal when merged [28], and possibly, the same is true with protein intake.

Third, we had no data about the foods consumed (e.g., meat, dairy, and grains). Protein sources are important because animal protein has a higher anabolic e ffect than vegetable proteins; this e ffect is mainly attributable to higher amino acid availability in animal-based foods than those coming from plant-based foods [52]. Therefore, the inclusion of data of ingested foods and food intake patterns along the determination of IPI per day and per meal might help to understand their combined role on health-related variables and which foods are associated with a lower probability of inadequacy [19,53–55].

Fourth, there were di fferent dietary assessment methods for all studies (Table 1), and each one addresses the information from a di fferent approach [56,57]. Moreover, the results should be interpreted cautiously because recall methods are prone to underreporting when compared to doubly labelled water, and two studies relied on one- and 2-days dietary recalls [20,29], which may poorly represent the actual nutritional intake [58,59]. Additionally, the di fferences in the nutritional analysis due to di fferent food databases must be considered. It is important to highlight that data reported in this study might di ffer from that in the original articles due to di fferences in how it was analyzed. For example, Gingrich et al. [14] calculated the PDCV for each day of the 7-day diary and then averaged them. Instead, we first averaged the 7-day protein intake, and then we calculated the PDCV. As mentioned in the methods section, all databases followed the same procedure to average protein intake and made comparisons possible.

Fifth, these results came from very localized samples and must not be deemed nationally representative (Table 1). Instead, these data should be considered a glance into how the dietary protein intake patterns are in these countries. In the case of NHANES, it is a nationally representative study; however, we analyzed a sample of the available cases (200 out of 1039), and data were not weighted according to the complex sampling design; therefore, it cannot be considered as nationally representative. Comparisons of data coming from nationally representative studies deserve further research [51]. Similarly, the combined data reported here should be regarded with caution for several reasons: (1) it came from very di fferent samples (Tables 1 and 2); (2) there are some studies reporting protein intake per meal that were not included here, and the lack of these data surely a ffect the reported percentages; and (3) the percentages were not weighted. A meta-analysis of the available evidence with weighted pooled prevalence would help to overcome this limitation. We hope this work stimulates the curiosity to carry out systematic reviews with meta-analysis in this area.

The recruitment period may be another limitation. Although the recruitment years were close to one another (about 3 years), this period may yield differences in dietary patterns within countries. Similarly, the recruitment method differed among studies (Table 1). While all participants were community-dwelling, there may be differences in sociodemographic (e.g., education and income) and health-related variables (e.g., functionality and diagnosed diseases) (not included here) that may explain the differences in the outcome variables instead of the geographical localization. The addition and comparison of these variables for further studies are recommended. Additionally, mean ages ranged from 71 to 79 year; this eight-year gap might have influenced the differences observed among countries as older adults tend to eat less protein [23,25]. Finally, the differences in sample sizes in some comparisons were considerable (38 vs. 200, min vs. max), affecting the statistical tests.
