Intake of dietary carotenoids and flavonoids at midlife and risk of physical frailty at late life: the Singapore Chinese Health Study Authors: Liyan Huang, Kevin Yiqiang Chua, Changzheng Yuan, Woon-Puay Koh Published: 2025 ABSTRACT AbstractBackgroundAlthough Western studies have suggested that dietary carotenoids and flavonoids may reduce the risk of ageing-related physical frailty, evidence from prospective Asian studies is limited. We examined the association between midlife intake of these phytochemicals and risk of physical frailty among Chinese adults in Singapore.MethodsThis study included 10,738 participants from the population-based prospective cohort of the Singapore Chinese Health Study. Dietary information was collected using a validated 165-item semi-quantitative food frequency questionnaire at recruitment from 1993 to 1998 (mean age 51 years, range 45–60 years). During follow-up 3 interviews from 2014 to 2017, physical frailty was assessed using a modified version of the Cardiovascular Health Study phenotype (mean age 72 years, range 63–84 years). Multivariable logistic regression models were used to calculate odds ratios (OR) and 95% confidence intervals (CI) for the associations between nutrient intake and physical frailty.ResultsAmong the 10,738 participants in the current study, 58.3% were female. Over a mean (SD) follow-up period of 20 (1.9) years, 1,220 participants (11.4%) were identified as being physically frail. Higher baseline consumption of carotenoids and flavonoids were associated with lower odds of physical frailty at follow-up 3. Compared to those in the lowest quintiles, the ORs (95% CIs) for the association with physical frailty for those in the highest quintiles were 0.69 (0.55–0.85) for total carotenoids (p-trend = 0.001) and 0.80 (0.64–0.99) for total flavonoids (p-trend = 0.019). Specifically, total carotenoid intake showed significant inverse association with weakness defined by weak handgrip; the OR (95% CI) comparing extreme quintiles was 0.60 (0.49–0.72) (p-trend = 0.001). Conversely, flavonoid intake was inversely associated with slow timed-up-and-go test, with an OR (95% CI) of 0.83 (0.67–1.03) for the extreme quintiles (p-trend = 0.021). For individual nutrients, significant inverse associations between consumption and odds of physical frailty were found for alpha-carotene, beta-carotene and lutein among carotenoids, and for flavan-3-ols, flavonese and flavonols among flavonoids.ConclusionHigher dietary intake of carotenoids and flavonoids during midlife may be beneficial in reducing physical frailty in later life.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12937-025-01259-6.Keywords: Diet, Carotenoid, Flavonoid, Physical frailty, Chinese, Ageing Introduction Physical frailty is a geriatric syndrome characterized by diminished strength, endurance, and decreased physiological function [1], and is a pivotal contributor to functional decline and mortality among older individuals [2]. A recent meta-analysis that included 240 studies reported a global prevalence of approximately 12% for physical frailty among older adults [3], while another meta-analysis that included 46 studies reported an incidence rate of 43 frailty cases per 1,000 person-years [4]. The prevalence of frailty increases with age, affecting 3.2%, 9.5%, and 25.7% of individuals aged 65–70, 75–79, and 85–89 years, respectively [5]. Since effective treatment remains lacking [6], understanding potentially modifiable factors that could either increase or reduce the risk of frailty in the general population is crucial in preventing, delaying or even in reversing the development of physical frailty in ageing. Several studies have shown that dietary patterns rich in plant-based foods could reduce the risk of ageing-related physical frailty by 50–70%, and this has been attributed to plant-derived antioxidant phytochemicals, such as carotenoids and flavonoids [7–9]. Notably, two recent prospective studies conducted in the United States have reported a protective effect in the consumption of carotenoids [6] and flavonoids [10] in preventing physical frailty. However, although studies on carotenoids have generally shown a beneficial effect, many of these studies are limited by cross-sectional designs, short follow-up periods, small sample sizes, and a focus primarily on Western populations [11]. Moreover, the benefits of dietary flavonoids are understudied and the results are conflicting [10, 12]. Hence, whether dietary carotenoids and flavonoids could affect the risk of physical frailty remains unclear, and further evidence is needed to fill the gap, especially among understudied Asian populations. Therefore, we aimed to determine the associations between dietary intake of carotenoids (alpha-carotene, beta-carotene, beta-cryptoxanthin, lycopene, and lutein) and flavonoids (anthocyanins, flavan-3-ols, flavanones, flavones, and flavonols) at midlife with the likelihood of physical frailty in later life within a population-based cohort of Chinese adults living in Singapore over a follow-up period of 20 years. Methods Study population We utilized data from the Singapore Chinese Health Study (SCHS) [13], which enrolled 63,257 participants aged 45–74 years at baseline from April 1993 to December 1998. These participants were all from two major Chinese dialect groups, the Hokkien and Cantonese, who originated from the Fujian and Guangdong provinces in southern China, respectively. Participants were followed up every 5 to 6 years over the phone or in person to update information on selected lifestyle factors and disease status. Follow-up 1 occurred from 1999 to 2004, followed by Follow-up 2 from 2006 to 2010. The Follow-up 3 interviews, conducted in person by trained interviewers, took place between July 2014 and December 2017. A total of 12,583 participants from Follow-up 3, who had complete data on all four criteria defining physical frailty, were initially included in the analysis. We excluded 30 participants with incomplete dietary information, 138 participants with implausible daily energy intake (< 700 or > 3700 kcal for men, < 600 or > 3000 kcal for women), and a further 1677 participants who were already aged > 60 years old at recruitment, and thus were left with 10,738 participants in the final analysis (Supplementary Fig. 1). The study was approved by the Institutional Review Board at the National University of Singapore, and written informed consent was obtained from all participants. Assessment of intake of carotenoids and flavonoids A 165-item semiquantitative food frequency questionnaire (FFQ) specifically designed for this cohort was used during baseline interviews to assess participants’ usual diet over the prior year. The FFQ was subsequently validated with 858 cohort subjects using two 24-hour dietary recalls about two months apart, and with another 164 subjects via a single 24-hour recall. Mean daily intakes of energy and selected nutrients from the 24-hour recall interviews and the structured FFQ were mostly within 10% of each other. Correlations for intake of energy and selected nutrients between these two methods ranged from 0.24 to 0.79, which were similar to results from other population calibration studies [13]. Specifically, of relevance to this study, the energy-adjusted correlation coefficients for beta-carotene ranged from 0.37 to 0.61 among the four groups of men and women in the two different dialect groups. For each food item, participants selected from eight intake frequency categories (ranging from “never or hardly ever” to “two or more times a day”) and three portion sizes (small, medium, or large), aided by pictures and actual plates, bowls and cups. The intake levels of nutrients and energy were computed by linking dietary intakes to the Singapore Food Composition Database, which was specifically developed for this cohort [13]. Additionally, the intake of all nutrients was adjusted for total daily energy consumption using the residual method [14]. In the current study, the nutrients of interest included carotenoids (alpha-carotene, beta-carotene, beta-cryptoxanthin, lycopene, and lutein) and flavonoids (anthocyanins, flavan-3-ols, flavanones, flavones, and flavonols). Assessment of physical frailty Physical frailty was assessed during the Follow-up 3 interviews between 2014 and 2017 using a modified version of the Cardiovascular Health Study (CHS) frailty phenotype [9]. Among the five criteria for physical frailty, we were only able to include weight loss, exhaustion, slowness, and weakness, but were not able to include low activity because relevant data on activity level was not collected during the third follow-up. Participants provided self-reported body weight at both the second (2006–2010) and third (2014–2017) follow-up interviews. Those who lost 10% or more of their body weight between these two time points, with an average interval of 7.3 (1.0) years between the interviews, were considered to have met the weight loss criterion. Participants were classified as experiencing exhaustion if they responded “No” to the question, “Do you feel full of energy?”. Slowness was determined based on participants’ performance on the Timed Up-and-Go (TUG) test, and those falling in the slowest sex-specified quintile were classified as being slow. Weakness was defined by handgrip strength, with participants in the weakest sex-specific quintile classified as being weak. Those meeting at least two of these criteria were defined as being physically frail. Assessment of other covariates At baseline, trained interviews used a structured questionnaire to collect information on participants’ sociodemographic characteristics, physician-diagnosed comorbidities (including hypertension, angina or heart attack, stroke, diabetes, and cancer), lifestyle factors (such as alcohol consumption, smoking history and average hours of sleep per day), as well as data on height, weight, and physical activity (including weekly hours of strenuous sports, vigorous work, and moderate activity) during midlife. Participants self-reported their height and weight, and body mass index (BMI) was calculated by dividing weight by the square of height. Statistical analysis Baseline characteristics of participants in each quintile of total carotenoid and total flavonoid intake were presented as means (SDs) for continuous variables and as frequencies (percentages) for categorical variables. We used one-way ANOVA to compare continuous variables, and the Chi-square test to compare categorical variables among different quintiles of nutrient intake. We used multivariable logistic regression models to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for the associations of carotenoids and flavonoids with physical frailty. To test for a linear trend across quintiles of intake, the median intake levels for consecutive quintiles were assessed as a continuous variable. In Model 1, we adjusted for basic sociodemographic characteristics [age at physical tests (years); sex; dialect group (Hokkien, Cantonese); level of education (no formal education, primary, secondary, diploma/university)] and total daily energy intake (kcal). In Model 2, we additionally adjusted for the following covariates assessed at the baseline interviews: history of physician-diagnosed comorbidities (hypertension, angina or heart attack, stroke, diabetes, cancer), history of smoking (never smoker, former smoker, current smoker), frequency of alcohol consumption (never, monthly, weekly, daily), body mass index group (< 18.5, 18.5–22.9, 23.0–27.4, ≥ 27.5 kg/m2), hours of sleep per day (≤ 5, 6–8, ≥ 9 h), hours of strenuous sports per week (0, 0.5–1.5, 2–3, ≥ 4 h), hours of vigorous work per week (0, 0.5–3.5, 4–6, ≥ 7 h), hours of moderate activity per week (0, 0.5–1.5, 2–3, 4–6, ≥ 7 h). In Model 3, we further adjusted for total daily protein intake (g), total daily caffeine intake (mg) [15], use of any vitamin or mineral supplements at least once a week (yes, no), and mutually adjusted for total carotenoid and total flavonoid intake (quintiles as categorical variables). Stratified analyses were conducted to explore the associations across subgroups defined by age (≥ 50, < 50 at baseline), sex, history of smoking (never, ever smoker), and BMI (< 23, ≥ 23 kg/m2). Interactions between carotenoid or flavonoid intake and these aforementioned factors were tested by including a product of the nutrient and factor in the multivariable-adjusted models. In secondary analyses, we examined the individual associations between total carotenoid intake and total flavonoid intake with specific physical frailty criteria, including slowness and weakness. These criteria reflect muscle strength and physical performance, and are also used as diagnostic indicators for sarcopenia [16]. Additionally, we explored the associations between physical frailty and specific carotenoids (alpha-carotene, beta-carotene, beta-cryptoxanthin, lycopene, and lutein), as well as specific flavonoids (anthocyanins, flavan-3-ols, flavanones, flavones, and flavonols). To assess the robustness of our associations, we conducted sensitivity analyses by (1) repeating the analysis after excluding participants who were likely to have had physical frailty at baseline due to a history of hypertension, cardiovascular disease, diabetes, or cancer, and (2) by applying sex-specific quintiles for carotenoid and flavonoid intake to account for potential sex-related differences in dietary habits. Statistical analyses were performed using R (version 4.2.3; The R Foundation for Statistical Computing), and 2-sided p-values below 0.05 were considered statistically significant. Results A total of 10,738 participants with a mean age of 51 years at baseline (range: 45–60 years) were included in the current study. The median baseline intake was 5.6 mg/d (interquartile range: 3.9–7.9 mg/d) for total carotenoids and 94.7 mg/d (interquartile range: 39.2–297 mg/d) for total flavonoids. The dietary intakes of total carotenoids and total flavonoids were weakly correlated (Spearman correlation coefficient = 0.17). Among the 5 carotenoids, pairwise correlation ranged from 0.18 (beta-cryptoxanthin and lutein) to 0.81 (beta-carotene and lutein) (Supplementary Table 1). Among the 5 flavonoids, pairwise correlation ranged from 0.09 (flavan-3-ols and anthocyanins) to 0.88 (flavan-3-ols and flavonols) (Supplementary Table 2). The characteristics of participants according extreme quintiles of carotenoid and flavonoid intake are shown in Table 1. Compared to those in the lowest quintiles, participants in the highest quintiles of carotenoid or flavonoid intake were more likely to be female, more highly educated, and to take supplements, but less likely to drink or smoke. Over a mean (SD) follow-up period of 20 (1.9) years, 1,220 participants (11.4%) were identified as being physically frail. Among these frail participants, the majority had weakness (69.2%), followed by slowness (60.5%), exhaustion (55.8%), and weight loss (39.4%). Table 1Characteristics of the study participants by quintiles of dietary carotenoid and flavonoid intakeOverallIntake of carotenoidIntake of flavonoidQ1Q5Q1Q5N10,7382148214821482148Mean age at physical test (SD) [years]72.1 (4.6)72.2 (4.6)71.5 (4.5)71.9 (4.6)72.1 (4.6)Female (%)6262 (58.3)777 (36.2)1390 (64.7)1016 (47.3)1066 (49.6)Level of education (%) No formal education1413 (13.2)274 (12.8)236 (11.0)278 (12.9)177 (8.2) Primary school4652 (43.3)1004 (46.7)866 (40.3)987 (45.9)889 (41.4) Secondary school3689 (34.4)684 (31.8)818 (38.1)721 (33.6)811 (37.8) Diploma/University984 (9.2)186 (8.7)228 (10.6)162 (7.5)271 (12.6) Hypertension (%)1867 (17.4)350 (16.3)401 (18.7)326 (15.2)449 (20.9) Heart attack (%)175 (1.6)33 (1.5)39 (1.8)32 (1.5)46 (2.1) Stroke (%)37 (0.3)7 (0.3)8 (0.4)9 (0.4)10 (0.5) Diabetes (%)449 (4.2)93 (4.3)88 (4.1)71 (3.3)105 (4.9) Cancer (%)164 (1.5)17 (0.8)42 (2.0)20 (0.9)33 (1.5)Frequency of alcohol consumption (%) Never/Monthly9476 (88.2)1776 (82.7)1926 (89.7)1828 (85.1)1858 (86.5) Weekly979 (9.1)249 (11.6)195 (9.1)241 (11.2)227 (10.6) Daily283 (2.6)123 (5.7)27 (1.3)79 (3.7)63 (2.9)History of smoking (%) Never smoker8560 (79.7)1401 (65.2)1808 (84.2)1539 (71.6)1643 (76.5) Ever smoker2178 (20.3)747 (34.8)340 (15.8)609 (28.4)505 (23.5)Body mass index [kg/m2] Underweight622 (5.8)151 (7.0)107 (5.0)149 (6.9)101 (4.7) Normal weight4767 (44.4)976 (45.4)948 (44.1)1041 (48.5)831 (38.7) Over weight4410 (41.1)851 (39.6)879 (40.9)799 (37.2)987 (45.9) Obesity939 (8.7)170 (7.9)214 (10.0)159 (7.4)229 (10.7)Hours of sleep per day (%) ≤ 5 h760 (7.1)159 (7.4)163 (7.6)143 (6.7)152 (7.1) 6–8 h9434 (87.9)1870 (87.1)1864 (86.8)1873 (87.2)1902 (88.5) ≥ 9 h544 (5.1)119 (5.5)121 (5.6)132 (6.1)94 (4.4)Hours of strenuous sports per week (%) 0 h9519 (88.6)1931 (89.9)1838 (85.6)1911 (89.0)1861 (86.6) 0.5–1 h572 (5.3)93 (4.3)139 (6.5)108 (5.0)128 (6.0) 2–3 h400 (3.7)78 (3.6)103 (4.8)75 (3.5)101 (4.7) ≥ 4 h247 (2.3)46 (2.1)68 (3.2)54 (2.5)58 (2.7)Hours of vigorous work per week (%) 0 h9708 (90.4)1804 (84.0)1999 (93.1)1862 (86.7)1924 (89.6) 0.5–3 h570 (5.3)162 (7.5)86 (4.0)143 (6.7)126 (5.9) 4–6 h169 (1.6)68 (3.2)19 (0.9)49 (2.3)33 (1.5) ≥ 7 h291 (2.7)114 (5.3)44 (2.0)94 (4.4)65 (3.0)Horus of moderate activity per week (%) 0 h8306 (77.4)1725 (80.3)1554 (72.3)1738 (80.9)1577 (73.4) 0.5–1 h761 (7.1)126 (5.9)180 (8.4)129 (6.0)182 (8.5) 2–3 h909 (8.5)158 (7.4)225 (10.5)156 (7.3)217 (10.1) 4–6 h472 (4.4)79 (3.7)103 (4.8)78 (3.6)114 (5.3) ≥ 7 h290 (2.7)60 (2.8)86 (4.0)47 (2.2)58 (2.7)Median total energy intake (IQR) [kcal/day]1612.5 (516.4)1842.2 (541.8)1686.6 (523.5)2006.1 (465.3)1607.7 (505.7)Median total protein intake (IQR) [g/day]59.7 (10.0)56.1 (10.7)61.3 (11.0)59.4 (12.3)59.8 (9.9)Median caffeine intake (IQR) [mg/day]144.2 (106.8)171.4 (118.6)135.1 (105.4)152.3 (111.9)194.8 (114.6)Weekly use of vitamin/mineral supplements (%)893 (8.3)142 (6.6)250 (11.6)139 (6.5)211 (9.8) Characteristics of the study participants by quintiles of dietary carotenoid and flavonoid intake After adjusting for demographic covariates, lifestyle factors, medical history, other dietary factors, and supplement use, as well as mutually adjusting for total carotenoid and total flavonoid intake (Model 3), higher intake of total carotenoids and total flavonoids was associated with lower odds of physical frailty in a stepwise manner (Table 2). Compared to participants in the lowest quintile of carotenoid intake at midlife [median (interquartile range): 3.24 (2.35–4.33) mg/day], those in the highest quintile of carotenoid intake [median (interquartile range): 10.73 (8.96–13.37) mg/day] had lower odds of frailty in later life [OR (95% CI): 0.69 (0.55, 0.85); p-trend = 0.001]. A similar association was observed for flavonoid intake; the OR (95% CI) for the highest [median (interquartile range): 552.1 (380.3–876.0) mg/day] versus lowest quartile [median (interquartile range): 34.6 (17.4–56.4) mg/day] of flavonoid intake was 0.80 (0.64, 0.99) (p-trend = 0.019) (Table 2). Furthermore, the associations between carotenoid or flavonoid intake and physical frailty did not significantly differ across subgroups defined by key demographic and lifestyle factors such as age, sex, smoking status, and BMI (Supplementary Fig. 2). Table 2Associations between total dietary carotenoids and flavonoids at midlife and physical frailty in later lifeQuintile 1Quintile 2Quintile 3Quintile 4Quintile 5p-trendTotal carotenoid Median intake (IQR) [mg/day] a3.24 [2.35, 4.33]4.01 [3.16, 5.09]5.16 [4.32, 6.26]6.85 [5.86, 7.99]10.73 [8.96, 13.37] Case/total number282/2148261/2147243/2148250/2147184/2148 Model 1Ref (1.0)0.87 (0.72, 1.05)0.81 (0.67, 0.98)0.89 (0.73, 1.08)0.68 (0.55, 0.83)0.001 Model 2Ref (1.0)0.89 (0.74, 1.08)0.83 (0.68, 1.01)0.90 (0.74, 1.10)0.68 (0.55, 0.84)0.001 Model 3Ref (1.0)0.89 (0.73, 1.08)0.82 (0.67, 1.00)0.90 (0.74, 1.10)0.69 (0.55, 0.85)0.001Total flavonoid Median intake (IQR) (mg/day]a34.6 [17.4, 56.4]39.1 [21.3, 65.6]78.1 [50.4, 112.7]222.6 [157.7, 299.6]552.1 [380.3, 876.0] Case/total number262/2148276/2147251/2148219/2147212/2148 Model 1Ref (1.0)0.96 (0.79, 1.17)0.87 (0.70, 1.07)0.83 (0.68, 1.01)0.76 (0.62, 0.93)0.004 Model 2Ref (1.0)0.98 (0.80, 1.20)0.87 (0.71, 1.08)0.82 (0.67, 1.00)0.73 (0.59, 0.90)0.001 Model 3Ref (1.0)1.00 (0.81, 1.22)0.90 (0.73, 1.12)0.85 (0.69, 1.05)0.80 (0.64, 0.99)0.019Model 1: adjusted for age at physical tests (years), gender, dialect group (Hokkien, Cantonese), level of education (none, primary, secondary, diploma/university), daily energy intake (kcal)Model 2: adjusted for Model 1 and the following covariates assessed at the baseline interviews: hypertension, angina or heart attack, stroke, diabetes, cancer, alcohol consumption (none, monthly, weekly, daily), smoking history (never, former, current), body mass index (< 18.5, 18.5–22.9, 23.0–27.4, ≥ 27.5 kg/m2), hours of sleep per day (≤ 5, 6–8, ≥ 9 h), hours of strenuous sports per week (0, 0.5–1.5, 2–3, ≥ 4 h), hours of vigorous work per week (0, 0.5–3.5, 4–6, ≥ 7 h), hours of moderate activity per week (0, 0.5–1.5, 2–3, 4–6, ≥ 7 h)Model 3: adjusted for Model 2 and the following covariates assessed at the baseline interviews: daily protein intake (g), daily caffeine intake (mg), weekly use of vitamin/mineral supplements, total carotenoid or total flavonoid intake (quintiles) [where applicable]aThe median intake value is based on absolute intake Associations between total dietary carotenoids and flavonoids at midlife and physical frailty in later life Model 1: adjusted for age at physical tests (years), gender, dialect group (Hokkien, Cantonese), level of education (none, primary, secondary, diploma/university), daily energy intake (kcal) Model 2: adjusted for Model 1 and the following covariates assessed at the baseline interviews: hypertension, angina or heart attack, stroke, diabetes, cancer, alcohol consumption (none, monthly, weekly, daily), smoking history (never, former, current), body mass index (< 18.5, 18.5–22.9, 23.0–27.4, ≥ 27.5 kg/m2), hours of sleep per day (≤ 5, 6–8, ≥ 9 h), hours of strenuous sports per week (0, 0.5–1.5, 2–3, ≥ 4 h), hours of vigorous work per week (0, 0.5–3.5, 4–6, ≥ 7 h), hours of moderate activity per week (0, 0.5–1.5, 2–3, 4–6, ≥ 7 h) Model 3: adjusted for Model 2 and the following covariates assessed at the baseline interviews: daily protein intake (g), daily caffeine intake (mg), weekly use of vitamin/mineral supplements, total carotenoid or total flavonoid intake (quintiles) [where applicable] aThe median intake value is based on absolute intake We then examined the associations of carotenoid and flavonoid intake at midlife with weakness and slowness within the physical frailty phenotype in later life. Total carotenoid intake was significantly inversely associated with the odds of weakness, as defined by handgrip strength being in the lowest sex-specific quintile. The OR (95% CI) comparing the highest with the lowest quintile of carotenoid intake was 0.60 (0.49, 0.72) (p-trend = 0.001). In contrast, higher intake of flavonoids was associated with a lower likelihood of slowness, as defined by TUG times being in the lowest sex-specific quintile, with the OR (95% CI) comparing the extreme quintiles of flavonoid intake being 0.83 (0.67, 1.03) (p-trend = 0.021) (Table 3). Table 3Associations between total dietary carotenoids and flavonoids at midlife with weakness and slowness in later lifeQuintile 1Quintile 2Quintile 3Quintile 4Quintile 5p-trendWeakness (handgrip strength in the weakest sex-specific quintile) Total carotenoidRef (1.0)0.86 (0.72, 1.01)0.75 (0.63, 0.89)0.80 (0.68, 0.96)0.60 (0.49, 0.72)0.001 Total flavonoidRef (1.0)0.93 (0.78, 1.11)0.94 (0.79, 1.14)0.79 (0.66, 0.95)0.90 (0.75, 1.09)0.32Slowness (TUG test time in the slowest sex-specific quintile) Total carotenoidRef (1.0)0.97 (0.79, 1.18)0.93 (0.75, 1.14)1.00 (0.81, 1.23)0.84 (0.68, 1.05)0.16 Total flavonoidRef (1.0)1.05 (0.85, 1.29)1.00 (0.80, 1.24)0.84 (0.68, 1.04)0.83 (0.67, 1.03)0.021Models were adjusted for age at physical tests (years), gender, dialect group (Hokkien, Cantonese), level of education (none, primary, secondary, diploma/university), daily energy intake (kcal), hypertension, angina or heart attack, stroke, diabetes, cancer, alcohol consumption (none, monthly, weekly, daily), smoking history (never, former, current), body mass index (< 18.5, 18.5–22.9, 23.0–27.4, ≥ 27.5 kg/m2), hours of sleep per day (≤ 5, 6–8, ≥ 9 h), hours of strenuous sports per week (0, 0.5–1.5, 2–3, ≥ 4 h), hours of vigorous work per week (0, 0.5–3.5, 4–6, ≥ 7 h), hours of moderate activity per week (0, 0.5–1.5, 2–3, 4–6, ≥ 7 h), daily protein intake (g), daily caffeine intake (mg), weekly use of vitamin/mineral supplements, total carotenoid or total flavonoid intake (quintiles) [where applicable] Associations between total dietary carotenoids and flavonoids at midlife with weakness and slowness in later life Models were adjusted for age at physical tests (years), gender, dialect group (Hokkien, Cantonese), level of education (none, primary, secondary, diploma/university), daily energy intake (kcal), hypertension, angina or heart attack, stroke, diabetes, cancer, alcohol consumption (none, monthly, weekly, daily), smoking history (never, former, current), body mass index (< 18.5, 18.5–22.9, 23.0–27.4, ≥ 27.5 kg/m2), hours of sleep per day (≤ 5, 6–8, ≥ 9 h), hours of strenuous sports per week (0, 0.5–1.5, 2–3, ≥ 4 h), hours of vigorous work per week (0, 0.5–3.5, 4–6, ≥ 7 h), hours of moderate activity per week (0, 0.5–1.5, 2–3, 4–6, ≥ 7 h), daily protein intake (g), daily caffeine intake (mg), weekly use of vitamin/mineral supplements, total carotenoid or total flavonoid intake (quintiles) [where applicable] The associations between physical frailty and specific subgroups of carotenoids and flavonoids were further evaluated (Fig. 1). Among the different subgroup of carotenoids, significant inverse associations with the likelihood of physical frailty were observed for intake of alpha-carotene (OR Q5 vs. Q1 = 0.78, 95% CI = 0.63, 0.95; p-trend = 0.037), beta-carotene (OR Q5 vs. Q1 = 0.65, 95% CI = 0.53, 0.81; p-trend < 0.001), and lutein (OR Q5 vs. Q1 = 0.73, 95% CI = 0.59, 0.90; p-trend = 0.003) (Supplementary Table 3). In terms of flavonoids, we found that the intake of flavan-3-ols (OR Q5 vs. Q1 = 0.73, 95% CI: 0.59–0.91, p-trend = 0.014), flavones (OR Q5 vs. Q1 = 0.80, 95% CI: 0.65–0.99, p-trend = 0.13), and flavonols (OR Q5 vs. Q1 = 0.75, 95% CI: 0.60–0.93, p-trend = 0.012) were inversely associated with odds of physical frailty. Additionally, even though the risk estimates and p-trend for lycopene, beta-cryptoxanthin, anthocyanins and flavanones did not reach statistical significance, their associations with physical frailty still demonstrated an overall inverse trend (Supplementary Table 4). Associations between subgroups of dietary carotenoids and flavonoids at midlife and physical frailty in later life. For each nutrient, the number in parenthesis in the first column represents median intake (mg/day) in that quintile. Models were adjusted for age at physical tests (years), gender, dialect group (Hokkien, Cantonese), level of education (none, primary, secondary, diploma/university), hypertension, angina or heart attack, stroke, diabetes, cancer, alcohol consumption (none, monthly, weekly, daily), daily energy intake (kcal), smoking history (never, former, current), body mass index (<18.5, 18.5–22.9, 23.0–27.4, ≥27.5 kg/m2), hours of sleep per day (≤5, 6-8, ≥9 hours), hours of strenuous sports per week (0, 0.5-1, 2-3, ≥4 hours), hours of vigorous work per week (0, 0.5-3, 4-6, ≥7 hours), hours of moderate activity per week (0, 0.5-1, 2-3, 4-6, ≥7 hours), daily protein intake (g), daily caffeine intake (mg), weekly use of vitamin/mineral supplements, total carotenoid or total flavonoid intake (quintiles) [where applicable] In sensitivity analyses, the associations between higher carotenoid and flavonoid intake and lower odds of physical frailty remained robust after excluding participants with a baseline history of hypertension, cardiovascular disease, diabetes, or cancer. In addition, the results remained essentially unchanged when sex-specific quintiles for nutrient intake were applied (Supplementary Table 5). Discussion In this population-based cohort of Chinese adults living in Singapore, higher dietary intake of carotenoids and flavonoids at midlife was associated with lower odds of physical frailty in subsequent years. Specifically, higher carotenoid intake was associated with a lower likelihood of weakness defined by handgrip strength, while greater flavonoid intake was associated with a lower likelihood of slowness during the TUG test. Among individual nutrients, alpha-carotene, beta-carotene, and lutein (from the carotenoid group), as well as flavan-3-ols, flavones, and flavonols (from the flavonoid group), appeared to be major contributors to these associations. Although the exact mechanisms remain complex and not fully understood, the primary protective effects of carotenoids and flavonoids are likely attributable to their potent antioxidant and anti-inflammatory properties, since oxidative stress and inflammation are central to age-related loss of muscle and physical function [17–19]. By efficiently scavenging reactive oxygen species (ROS), these compounds may help mitigate or prevent mitochondrial and cellular damage caused by oxidative stress [20], as well as maintain protein homeostasis [21] and skeletal muscle function [22], thereby protecting against age-related physical decline. Our finding that carotenoids reduce the risk of physical frailty in late life concurred with previous prospective studies that had examined the relationship between dietary or serum carotenoids and frailty outcomes. For instance, a cohort from the Framingham Offspring Study, with a mean baseline age of 60 years and a mean follow-up of 11 years, found that every 10-mg increase in total daily carotenoid intake was associated with a 16% lower odds of frailty [6]. Additionally, a previous cohort study found that individuals in the lowest quartile of serum carotenoid levels had 1.39 times (over 2–5 years of follow-up) higher odds of physical frailty [23]. Furthermore, our results suggest that the beneficial effect could extend beyond a single carotenoid, and this is consistent with a recent systematic review of 8 cross-sectional and 3 cohort studies, which demonstrated that higher serum levels of five major dietary carotenoids, including alpha-carotene, beta-carotene, lutein, lycopene, and beta-cryptoxanthin, were associated with better physical function-related outcomes [11]. Unlike evidence for dietary carotenoids, the evidence for the benefit of dietary flavonoids is relatively limited and inconsistent. Specifically, a study using data from the National Health and Nutrition Examination Surveys reported a significant U-shaped relationship between total flavonoid intake and frailty, with the prevalence of frailty decreasing with increased flavonoid intake until the lowest association with frailty was observed at an intake of 220 mg/d [12]. In contrast, in a prospective study based on the Framingham Heart Study cohort, no association was observed between total flavonoid intake and frailty onset; however, the authors of this study did report that higher intake of flavonols was linearly associated with lower odds of frailty onset, which concurred with the findings in our study [10]. Flavonoids are widely distributed across various fruits, vegetables, plants, and herbs, and include anthocyanidins (such as cyanidin, delphinidin, and malvidin, which are typically found in colored berries and red wine), flavan-3-ols (including catechin, epicatechin, and theaflavin, which are present in cocoa, apples, and grapes), flavanones (like eriodictyol, hesperidin, and naringenin, which are mainly found in citrus fruits), flavones (such as apigenin and luteolin, which are commonly found in celery, parsley, and chamomile tea), and flavonols (including kaempferol, quercetin, and myricetin, which are present in tea, broccoli, and various fruits) [24]. Each of these subclasses of flavonoids possess a distinct chemical structure, which can impact their antioxidant activity, bioavailability, metabolism, and the specific oxidative environments in which they function [25]. Furthermore, the presence of other compounds and nutrients in food sources may also influence their overall antioxidant potential [26]. Therefore, future studies should focus on evaluating the specific subclasses of flavonoids and their associations with the development of physical frailty. Interestingly, a recent meta-analysis of 20 randomized controlled trials (RCTs) found that flavonoid supplementation did not significantly improve handgrip strength, but did enhance performance in the 6-minute walk test and TUG test [27]. These findings, and ours, suggest that flavonoids may not improve handgrip strength but could play a beneficial role in maintaining the muscle function required for mobility. To the best of our knowledge, this is the first study that evaluated the association of dietary carotenoids and flavonoids with physical frailty in a Chinese population, and refined the previous results based on Western population. The strengths of the current study include its long-term follow-up within a population-based cohort, a relatively large sample size, validated dietary intake assessments, control for confounding factors and mutual adjustment for carotenoids and flavonoids. However, we acknowledge several limitations in our study. First, we did not assess physical frailty at baseline and could therefore have included frail subjects at recruitment. However, given that physical frailty is a geriatric syndrome associated with advancing age, we only included participants 60 years and younger, in whom the prevalence of frailty should still be very low [5]. Furthermore, we excluded individuals with comorbidities, as advanced age and the presence of chronic diseases are well-established risk factors for the onset and progression of physical frailty. Importantly, our results remained robust in sensitivity analyses. Second, due to the lack of data on participants’ physical activity level at follow-up 3, we used a modified definition of frailty that included only four available criteria. As a result, we could have inadvertently included pre-frail individuals with only two criteria among those considered to be frail, and thus underestimated the true nutrient-frailty association. Nevertheless, the prevalence of physical frailty among women aged 70–79 years in our cohort (14.4%) was comparable to that reported in three other studies using the original CHS phenotype [28]. Additionally, our method is supported by two prior studies that also omitted the physical activity component and classified participants as ‘frail’ or ‘non-frail’ using only the four remaining criteria [29, 30]. Third, our study could be affected by selection bias since it only included a subset of those initially recruited at baseline, and we could not include participants due to death or inability to participate in Follow-up 3 due to disability. In our previous analyses, participants who did not participate in the third follow-up interview were generally older, reported more comorbidities and had less healthy diet and lifestyles at recruitment than those who participated [9]. Hence, we could have excluded participants who had lower nutrient intake at midlife, but who were also more likely to have developed physical frailty later, and thus underestimated the nutrient-frailty association. Fourth, validation results were not specifically published for the intake of all carotenoids except beta-carotene, nor for flavonoids. Nonetheless, our previous publications have reported associations between the intake of various carotenoids and the risk of other diseases in this cohort [31, 32]. A dietary oxidative index, which included flavonoids, was also calculated from the FFQ, and our subsequent research had reported associations between this dietary antioxidative index and reduced risks of mortality [33] and late-life cognitive impairment [34] in this cohort. These findings provide indirect validation for the calculation of flavonoid intake in this group. In conclusion, our study suggests that higher dietary intake of carotenoids and flavonoids during midlife may help reduce the risk of developing physical frailty in later life. An implication of our study is that promoting a diet rich in carotenoids and flavonoids during midlife could be a strategic public health approach to mitigate the risk of physical frailty as people age. Future work should focus on identifying specific subclasses of carotenoids or flavonoids which may be beneficial for maintaining muscle strength and enhancing physical performance in older adults in interventional studies. Supplementary Information Supplementary Material 1. Acknowledgements We thank Siew-Hong Low of the National University of Singapore for supervising the fieldwork in the Singapore Chinese Health Study. Authors’ contributions WK designed research; LH performed the statistical analysis, and interpreted the data with WK; KC provided statistical support; LH developed the draft of the manuscript. WK, KC, and CY critically revised the manuscript for important intellectual content. WK had primary responsibility for the final content. All authors read and approved the final manuscript. Funding This work was supported by the Singapore National Medical Research Council [NMRC/CSA/0055/2013]; the United States National Cancer Institute, National Institutes of Health [UM1 CA182876, R01 CA144034]; and the Saw Swee Hock School of Public Health, National University of Singapore. W-P Koh is supported by the National Medical Research Council, Singapore [CSA-SI (MOH-000434)]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data availability Data for this study cannot be made available, due to legal or ethical confidentiality requirements. Declarations Ethics approval and consent to participate The study was approved by the institutional review board at the National University of Singapore (NUS-IRB), reference code LH-17-058. Written informed consent was obtained from all participants. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement Data for this study cannot be made available, due to legal or ethical confidentiality requirements.