Adherence to various dietary quality indices of centenarian offspring in the New England Centenarian Study

Abstract

Background

Many centenarian offspring (CO) show survival and health advantages compared to population controls, yet little is known about their dietary patterns. The New England Centenarian Study (NECS) offers an opportunity to characterize diet quality in this unique longevity-enriched population.

Objective

To characterize overall and component-level diet quality among CO in the New England Centenarian Study (NECS) using four established indices and to contextualize these patterns relative to published benchmarks from large U.S. cohorts of older adults.

Design

We analyzed data from 457 NECS participants who completed a 131-item food frequency questionnaire in 2005. We computed dietary scores using the Alternative Healthy Eating Index (AHEI), Healthy Eating Index (HEI), Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diet, and Planetary Health Diet Index (PHDI). We performed linear regression to examine whether these scores differ by sex, age, education, and marital status.

Results

Participants’ mean (SD) age was 73.6 (9.2) years; 55.1% were women. Overall mean (SD) index scores were: AHEI 51.9 (11.0), HEI 70.1 (9.2), MIND 8.6 (1.9), and PHDI 87.1 (12.0), indicating moderate overall diet quality. NECS participants generally met or approached targets for intakes of fruits, greens/beans, and protein-food quality (including seafood), as well as for moderation components such as sodium, added sugar, and refined grains. However, they fell short on intakes of legumes/soy/nuts and whole grains. Compared with nationally representative studies, NECS participants had modestly higher overall dietary scores across the four indices (P < 0.001). NECS participants had higher component scores for intakes of fruits, vegetables, and omega-3 s, but lower scores for whole grains, legumes, and soy. Higher education was consistently associated with healthier diet scores, while younger age and female sex corresponded to more favorable component patterns.

Conclusion

Centenarian offspring exhibit moderately higher diet quality than average U.S. older adults, with clear strengths and persistent gaps. These findings provide the first reference profile of dietary patterns in a longevity-enriched population and establish a foundation for future longitudinal research on the interaction of nutrition with inherited resilience to support healthy aging.

1. Introduction

Information about the dietary habits of centenarians—the longest-lived individuals in the population—is limited [1,2]. Although diet is a well-established modifiable determinant of healthy aging and longevity [[3], [4], [5], [6], [7], [8], [9], [10], [11], [12]], empirical evidence on centenarians’ dietary habits is remarkably sparse [1,2,[13], [14], [15], [16], [17]]. The few existing studies characterizing centenarians' diets include a small number of participants, inquire about a limited number of foods, and are often retrospective and/or rely on self-reports collected late in life when recall bias and survival-related changes in eating behavior may distort lifetime patterns [[13], [14], [15], [16], [17]], making it difficult to identify dietary determinants for their exceptional survival. This lack of dietary information represents a major gap in longevity research. While the genetic component of exceptional longevity has been increasingly well characterized, far less is known about the lifestyle factors, including diet, that may contribute to or interact with these biological advantages [[18], [19], [20], [21], [22]].

Centenarian offspring (CO) thus provide a critical opportunity to address this gap. Many CO display the same advantages observed in centenarians: delayed morbidity and disability, and a higher likelihood of achieving extreme old age themselves because they share approximately half of their parents’ longevity-associated variants and often early-life environments [[23], [24], [25], [26], [27], [28], [29], [30]]. Importantly, unlike centenarians, CO are studied decades before the onset of survival-related dietary changes, potentially offering an accurate view of habitual nutritional behaviors in families enriched for biological resilience. Understanding these patterns is foundational for advancing research on nutritional determinants of healthy aging and clarifying how diet contributes to exceptional longevity. Despite their scientific importance, virtually no empirical data exist on the dietary patterns of centenarian offspring. Only one cohort study examined offspring of the oldest-old Jewish adults aged ≥95 years and reported no difference in dietary patterns with peers [31]. To date, no study has evaluated dietary quality or adherence to established nutritional patterns in a large, well-characterized CO cohort.

To address this gap, this study characterized the diets of CO in the New England Centenarian Study (NECS), one of the world’s largest and longest-running family-based studies of exceptional longevity, using four complementary dietary indices: the Healthy Eating Index (HEI), Alternative Healthy Eating Index (AHEI), Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diet, and Planetary Health Diet Index (PHDI), capturing guideline adherence, chronic disease prevention, cognitive health, and environmental sustainability. Together, these indices provide a multidimensional profile of diet quality in a cohort genetically enriched for longevity, establishing foundational evidence for future work linking nutrition, biological aging, and healthy longevity.

2. Methods

2.1. NECS study design

NECS has enrolled centenarians and their offspring since 1995 and is the largest study of centenarian families worldwide [32]. Age was validated by linking birth certificates or archival documents dated close to the individual’s birth, ensuring consistent identification across major life events (e.g., marriage, children’s births). Additional records and family reconstitution were required for more extreme age claims (e.g., 110+ years) [33,34]. Demographic, health, and family history data, as well as physical and cognitive function data, were collected at least once for the majority of study participants. The Harvard Food Frequency Questionnaire (FFQ)—a widely used semi-quantitative dietary assessment tool designed to estimate usual food and nutrient intake—was completed by participants in 2005, and represents the only dietary assessment conducted within the NECS cohort.

2.2. Sample selection

The dataset included 465 NECS participants who completed the FFQ in 2005. After excluding seven study participants with implausible caloric intakes (caloric intake < 600 or > 4200 kcal) and one with missing demographic data, 457 participants remained in the analysis. This study was reviewed and approved by the Institutional Review Boards of Boston University and Tufts University. All participants provided informed consent at the time of data collection. The study was conducted in accordance with the ethical guidelines of human subjects’ research.

2.3. Dietary intake

The Harvard FFQ captures portion sizes and consumption frequency of 131 food and beverage items in 9 categories. Participants can select never or <1/month, 1–3 per month, 1 per week, 2–4 per week, 5–6 per week, 1 per day, 2–3 per day, 4–5 per day, or 6+ per day, for the number of servings of each food. These responses were transformed into servings per day. Nutrient data (e.g., sodium, polyunsaturated fatty acids [PUFAs]) for the FFQ were extracted from the Harvard University Food Composition Database, which multiplies consumption frequency by each item’s nutrient content. The validity and reliability of the FFQ for measuring intakes of foods, nutrients, and dietary patterns have been reported previously [35]. The FFQ has been extensively validated in adult and older adult populations using repeated dietary records, 24-h recalls, and objective biomarkers, demonstrating moderate to strong validity for ranking individuals and assessing habitual dietary patterns [36].

2.4. Healthy Eating Index (HEI)

Our primary outcome was HEI, which reflects adherence to the Dietary Guidelines for Americans (DGA). Our study used the most updated scoring system (2020) to reflect updates in dietary guidance [37]. HEI-2020 has 13 components categorized into adequacy and moderation (See Supplemental Table 1). Each component was scored proportionally based on intake levels, with higher scores indicating greater adherence to dietary recommendations – i.e., greater intake of adequacy components or lower intake of moderation components. The final HEI score was derived by summing the individual component scores, with a total possible score of 0–100.

2.5. Alternative Healthy Eating Index (AHEI)

AHEI was developed as an alternative to HEI with an emphasis on chronic disease risk prediction [38]. We created a modified 10-component AHEI score (see Supplemental Table 2), excluding intakes of trans fats due to data unavailability [39,40]. Each component was scored on a 0–10 scale, with higher scores indicating healthier consumption.

2.6. Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diet

The MIND diet score was developed in three stages: 1) determination of dietary components of the Mediterranean and DASH diets including the foods and nutrients shown to be important to incident dementia and cognitive decline through literature reviews, 2) selection of FFQ items that were relevant to each MIND diet component, and 3) determination of daily servings to be assigned to component scores guided by published studies on diet and dementia [21]. Olive oil consumption was scored 1 if identified as the primary oil used at home. For all other diet score components, the frequency of consumption of each food item was summed, and a concordance score of 0, 0.5, or 1 was assigned (see Supplemental Table 3). The total MIND diet score was computed by summing over all 15 of the component scores (range: 0–15).

2.7. The Planetary Health Diet Index (PHDI)

The PHDI was developed to operationalize the EAT-Lancet Commission’s reference diet, which aims to keep global dietary patterns within three key environmental boundaries—greenhouse gas emissions, cropland use, and freshwater use—while promoting human health [41]. This approach has also been applied in large cohorts (e.g., the Nurses’ Health Study and Health Professionals Follow-up Study) using a similar FFQ [42]. The PHDI categorizes components into four groups: adequacy, optimum, ratio, and moderation. Each PHDI component is scored proportionally (see Supplemental Table 4). The PHDI score was obtained by summing the component scores, with a total possible score of 140. We used vegetable fat intake as a proxy for added unsaturated fat and dairy fat as a proxy for added saturated fat [[43], [44], [45], [46], [47]].

2.8. Nutrient data processing

Some foods and nutrients were not directly provided in the FFQ output (e.g., whole grains, added sugars, dairy fat). We therefore manually derived the missing values by matching each FFQ line item to a comparable entry in USDA FoodData Central, assigning a gram weight per reported serving and multiplying by each participant’s reported frequency to obtain grams/day. Where reliable nutrient information was not available (e.g., incomplete nutrient panels for some branded foods in 2005 or subsequently discontinued items), we applied transparent rules: for whole grain, because the FFQ recorded brand names and whether a cereal was marketed as wholegrain, we adopted the American Association of Cereal Chemists International convention of 8 g whole grains per 30 g of labeled whole-grain cereal as reported previously [[48], [49], [50]]. For each reported serving of wholegrain cereal, we assumed a 30 g portion and assigned 8 g of whole grain, multiplying by the reported frequency to yield grams/day of whole grain from cereals. Whole grains from other foods were taken from FoodData Central entries. We expressed total whole-grain intake in grams/day and, when scoring indices required servings/ounce-equivalents, converted grams using the published index documentation [[48], [49], [50]]. For added sugars, we omitted the component of added sugars derived from breakfast cereals from the calculation, as no robust estimate could be applied across brands, resulting in a small underestimation of total added-sugar intakes.

2.9. Covariates

We used demographic data, including age, sex, education, and marital status, to adjust for potential confounders in the analysis of dietary patterns.

Age: Age in 2005 was calculated by subtracting each participant’s year of birth from 2005.

Education: Participants were categorized into three education levels based on total years of schooling: ≤12 years (high school or less), 13–16 years (some college or bachelor's degree), and >16 years (graduate education).

Marital Status: Participants self-reported their marital status, which was recoded into four categories: Married, Divorced, Single, and Widowed.

Participant classification Type: Participants were classified as CO or Non-CO.

2.10. Statistical analysis

First, we summarized the four dietary index scores, stratified by CO participant status. We compared the dietary index scores derived for the NECS sample to those of age-matched national cohorts. To contextualize dietary patterns observed in NECS, we benchmarked NECS scores against previously published values from large age-matched U.S. cohorts, including the Health and Retirement Study 2011, Health Care and Nutrition Study 2013 (HRS/HCNS; n ≈ 6000; age ≥ 60 y; 57 % women) [51,52], National Health and Nutrition Examination Survey 2015–2020 cycle (NHANES; n ≈ 5300; age ≥ 60 y; 54 % women) [53], Nurses’ Health Study 2006 (NHS; n ≈ 58,700; age ≥ 60 y; 100 % women), and Health Professional Follow-up Study 2006 (HPFS; n ≈ 36,600; age ≥ 60 y; 100 % men) [42]. To facilitate comparison of the AHEI sodium component score, which was scored based on the sodium intake distribution of HPFS and NHS participants [38], we applied the same sodium intake cut-points to derive intake deciles in the present study, consistent with studies in the HRS [52]. We used the Wald test to examine whether the dietary indices or component scores differences were statistically significant between NECS and the benchmarked cohort(s). In a sensitivity analysis to evaluate the extent to which sociodemographic factors—particularly educational attainment—could account for observed differences in diet quality, we conducted additional benchmarking using the Health and Retirement Study (HRS), a nationally representative cohort with comparable FFQ-based dietary assessment. The HRS sample was restricted to participants aged 60 years and older to closely match the age and sex distribution of NECS. HEI-2020 scores were generated using identical methods and examined overall and stratified by educational attainment categories aligned with those used in NECS (High school graduates, Bachelor’s degrees, and Graduate’s degrees).

We then performed linear regression analyses to investigate the determinants of dietary quality indices. The independent variables were sex (female and male), age groups (quartiles), education level (high school graduates, bachelor’s degree, graduate’s degree), marital status (married vs non-married), and CO status (CO vs non-CO). We modeled these using ordinary least squares. We also examined potential effect modification to assess whether associations between participant classification status and dietary quality differed by key demographic characteristics.

We used the unbiased forward stepwise procedure to determine which interaction terms to retain in the final model. We evaluated all possible two-way interaction terms between age, sex, education, marital status, and CO status. For model selection, we used SLE = 0.05 (Significance Level for Entry) and SLS = 0.05 (Significance Level of Stay). SLE specifies the P-value threshold a variable must meet to be added to the model; SLS specifies the P-value threshold a variable must meet to remain in the model. In the stepwise process, if any effect in the current model did not meet the SLS threshold, the least significant effect was removed before adding new effects. Only after all necessary deletions were made could another effect be considered. At each step, the effect that produced the most significant F statistic was added, and the process was repeated. The algorithm stopped when no effects outside the model meet SLE and all effects in the model meet SLS. The five main effects were forced to remain in the model at all times, ensuring that model selection only tested the added explanatory value of each interaction term.

All analyses were conducted using SAS 9.4 (Cary, NC), and significance was determined at P < 0.05.

3. Results

3.1. Sample characteristics

The demographic profiles of the NECS cohort are shown in Table 1, stratified by participant classification type. The cohort included 44.9% men and 55.1% women. The mean age at FFQ collection was 73.6 (range: 42–92). Nearly half (45.3%) of the study sample held a Bachelor’s degree, and about one-third (33.6%) had a graduate degree. Most participants were currently married (71.3%), with 28.7% classified as non-married. The CO group was 64.3% male, and the spousal group was 37.3% male. CO participants were more educated than their non-CO, and more than one-third of CO participants held graduate degrees.

Table 1.

Descriptive Statistics of the included participants from the New England Centenarian Study.

Subject Type	Total sample (N = 457)	Centenarian offspring (N = 331)	Non-CO (N = 126)
	Mean (SD)/N (%)	Mean (SD)/N (%)	Mean (SD)/N (%)
Dietary quality
HEI (100 total)	70.1 (9.1)	70.1 (8.9)	70.2 (9.6)
AHEI (100 total)	51.9 (11.1)	51.7 (11.2)	52.3 (10.9)
MIND (15 total)	8.6 (1.9)	8.5 (2.0)	8.6 (1.8)
PHDI (140 total)	87.1 (11.9)	86.8 (11.9)	87.9 (12.0)
Gender
Men	206 (45.1)	124 (37.5)	82 (65.1)
Women	251 (54.9)	207 (62.5)	44 (34.9)
Age, mean (SD), years
1st quartile	64.1 (4.5)	64.0 (4.6)	64.3 (4.2)
2nd quartile	72.2 (1.4)	72.3 (1.4)	72.1(1.6)
3rd quartile	76.4 (1.2)	76.5 (1.2)	76.3 (1.1)
4th quartile	81.8 (2.6)	81.8 (2.7)	81.9 (2.5)
Educational attainment
High school graduate	95 (20.8)	61 (18.4)	34 (27.0)
Bachelor’s degree	208 (45.5)	147 (44.4)	61 (48.4)
Graduate’s degree	154 (33.7)	123 (37.2)	31 (24.6)
Marital Status
Married	327 (71.6)	224 (67.7)	103 (81.8)
Non-married	130 (28.4)	107 (32.3)	23 (18.2)

3.2. HEI scores

HEI total scores averaged 70.1 (SD = 9.2) (Table 2) for the entire sample out of 100 possible points. Component level data (Fig. 1) showed NECS participants had high adherence to HEI’s recommendation for greens and beans (4.4 out of 5.0), total fruits (4.2 out of 5.0), whole fruits (4.4 out of 5.0), total protein foods (4.9 out of 5.0), seafood and plant proteins (4.8 out of 5.0), refined grains (9.0 out of 10.0), and sodium (9.4 out of 10.0). More than half of the possible points were awarded for total vegetables, dairy, and saturated fats. The mean scores for whole grains and fatty acids were comparatively lower—the average whole grain score was only 2.0 out of 10, and the fatty acids score was 4.3 out of 10.

Table 2.

Linear regression of centenarian-offspring status and sociodemographic factors in relation to four dietary quality indices.

Dietary index	Alternative healthy eating index		Healthy eating index		MIND		Planetary health diet index
	Coefficient (SE)	P-value	Coefficient (SE)	P-value	Coefficient (SE)	P-value	Coefficient (SE)	P-value
Intercept	60.969 (5.700)	<0.0001	82.653 (7.287)	<0.0001	10.060 (1.005)	<0.0001	91.518 (6.382)	<0.0001
Subject Type (ref = Non-CO)
Centenarian Offspring	−0.779 (1.190)	0.507	−4.058 (1.932)	0.035	−0.131 (0.266)	0.622	−5.663 (2.532)	0.025
Sex (ref = men)
Women	1.678 (1.087)	0.119	−17.364 (8.865)	0.051	0.299 (0.190)	0.116	1.286 (1.169)	0.271
Age (continuous), year	−0.207 (0.073)	0.004	−0.186 (0.093)	0.048	−0.031 (0.013)	0.014	−0.073 (−0.078)	0.352
Educational attainment (ref = high school graduate)
Bachelor’s degree	4.170 (1.360)	0.002	−0.681 (1.922)	0.721	0.787 (0.235)	0.001	−0.175 (2.522)	0.944
Graduate’s degree	4.520 (1.448)	0.002	4.050 (2.218)	0.065	0.946 (0.254)	0.001	4.883 (2.913)	0.042
Marital Status (ref = married)
Non-married	0.443 (1.167)	0.702	−0.458 (0.957)	0.631	0.084 (0.200)	0.675	−0.393 (1.256)	0.755
CO * Education (ref = CO*high school graduate)
CO*Bachelor's			5.678 (2.340)	0.015			7.093 (3.070)	0.021
CO*Graduate's			0.709 (2.607)	0.783			2.176 (3.422)	0.521
Age*Sex (ref = Men)
Age * Women			0.269 (0.120)	0.025

Fig. 1.

Distribution of individual components of four dietary indices.

Regression analyses showed that two interaction coefficients were significant: Sex × Age (β = +0.27 points/year for women vs men, P = 0.025) and CO status × Education (CO × Bachelor’s β = +5.68, P = 0.015; CO × Graduate β = +0.71, P = 0.783). Among men, HEI declined with age (slope = −0.19 points/year, P = 0.048). Among women, the age slope was attenuated but increasing overall (−0.19 + 0.27 ≈ +0.08 points/year). CO participants have significantly lower dietary scores than non-CO participants only among those with high school degrees, but not among those with higher education levels. Marital status did not significantly influence HEI (P > 0.05). While both interactions (Sex × Age and CO × Education) reached nominal significance, they did not remain significant after multiple-COmparison correction. Thus, we interpret these findings as exploratory.

3.3. AHEI scores

The mean AHEI score (Table 1) for the overall cohort was 51.9 out of 100.0 possible points. Regarding AHEI components (Fig. 1), there was notable adherence to the index’s recommendation for long-chain fatty acid intake (specifically eicosapentaenoic acid and docosahexaenoic acid). The average score for this component was 8.2 out of 10.0 possible points. Approximately half of the possible points were awarded for most components, including vegetables, fruits, nuts, red and processed meats, polyunsaturated fat, sodium, and alcohol. The mean scores for whole grains, sugar-sweetened beverages, and fruit juice were comparatively lower (∼2.3 and 3.5, respectively).

AHEI scores did not vary between CO and non-CO, but regression analyses showed that older age was associated with significantly lower scores (β = –0.21 per year, P = 0.004), while higher educational attainment was associated with significantly higher AHEI: a bachelor’s degree (+4.17 points, P = 0.002) and a graduate degree (+4.52 points, P = 0.002) were each associated with higher scores compared to a high school education (Table 2). Neither sex, marital status, nor the included interaction terms showed significant associations with AHEI.

3.4. PHDI scores

The average PHDI score was 87.1 out of a total possible 140.0 points (Table 2). PHDI showed high scores for starchy vegetables, poultry, and eggs with a mean score of 8.3, 8.6, and 9.2 out of 10.0, respectively (Fig. 1). The mean scores for vegetables, fruits, dairy foods, fish, added unsaturated fat, and added sugars ranged from 6.9 to 7.6, with the scores at the 75th percentile being 10.0 or nearly 10.0 points. However, the NECS cohort scored low on whole grains, legumes, soy foods, nuts, and added saturated fats. CO scored 86.8 (SD = 11.9), while non-CO scored 87.9 (SD = 12.0), a nonsignificant difference (P = 0.61). Regression revealed no clear effect of age, sex, or marital status for PHDI (P = 0.12). The regression suggested a statistically significant interaction effect between CO status and education level (7.09, P = 0.02). For those who only had a high school education, CO participants had significantly lower PHDI scores than non-CO (−5.66, P = 0.03); however, this difference between CO and non-CO was not observed among those with a college or graduate level education. These results are again exploratory as they did not remain significant after correction for multiple testing.

3.5. MIND scores

The MIND average diet score was 8.6 (SD = 1.9) out of a possible 15 points. Regarding the specific components (Fig. 1), NECS participants generally scored high on other vegetables (0.8 out of 1.0), berries (0.8 out of 1.0), butter/margarine (0.9 out of 1.0), fish (0.8 out of 1.0), and fast fried food (0.8 out of 1.0). They received about half of the points for green leafy vegetables, nuts, cheese, poultry, and red/processed meat products. They scored relatively low on olive oil, whole grains, beans, pastries and sweets, and wine. Women and men had similar scores (P = 0.116). Age showed a small but significant negative association, with each additional year linked to a 0.03-point decrease (P = 0.0143). Individuals with a bachelor’s degree scored +0.79 points higher (P = 0.001), and those with a graduate degree scored +0.95 points higher (P = 0.001), relative to high school graduates. Neither marital status nor any examined interaction terms were associated with MIND scores.

To contextualize dietary quality observed in the NECS cohort, we benchmarked dietary index scores against published reference values from large, age-matched U.S. cohorts using peer-reviewed sources. Using Wald tests, NECS participants exhibited higher overall diet quality across multiple indices compared with previously published values for older adults. Specifically, when benchmarked against adults aged 60 years and older in NHANES 2015–2020, the NECS cohort achieved a substantially higher mean HEI-2020 score (70.1 vs. 62.9; P < 0.001). Mean AHEI and MIND scores were modestly higher in NECS compared with age-matched participants in the HRS/HCNS (AHEI: 51.9 vs. 48.6; P < 0.001; MIND: 8.6 vs. 7.1; P < 0.001), and previously reported mean PHDI scores in the Nurses’ Health Study and Health Professionals Follow-up Study were 83.8 and 79.5, whereas the mixed-sex NECS cohort averaged 87.1 (P < 0.001).

In our sensitivity analysis comparing the analytic sample restricted to matched educational categories in the HRS, NECS participants exhibited modestly higher HEI-2020 scores within each stratum. Specifically, mean HEI scores were 63.9 in HRS versus 67.3 in NECS among high school graduates (P = 0.002), 68.1 versus 70.6 among individuals with bachelor’s degrees (P = 0.001), and 69.2 versus 71.2 among those with graduate degrees (P = 0.023) (Supplemental Table 13). While statistically detectable, these differences were attenuated relative to unstratified cohort comparisons. Similar patterns were observed when stratifying by sex, with NECS participants exhibiting higher HEI-2020 scores among both women and men compared with HRS counterparts (Supplemental Table 13).

4. Discussion

This study provides the first comprehensive characterization of dietary quality in centenarian offspring (CO), a population genetically enriched for exceptional longevity. We described overall diet quality using four established indices (AHEI, HEI, MIND, PHDI) and mapped component-level patterns relevant to healthy aging. On average, participants exhibited moderate overall diet quality, consistently meeting or approaching targets for fruits, greens/beans, protein-food quality (including seafood), and several moderation items (e.g., sodium and refined grains), while showing consistent shortfalls in whole grains, plant-protein (legumes/soy/nuts), and, in some indices, fat quality (HEI fatty-acid ratio; PHDI added saturated fat). When compared to previously published age-matched cohorts of U.S. older adults, we observed that NECS participants had slightly higher total index scores. Notably, NECS participants exhibited higher intake scores of fruits, vegetables, long-chain omega-3 s, and scored higher on several items of moderation, such as added sugar, sodium, and butter components. However, CO families appeared to score lower on intakes of whole grains, legumes, soy, and nuts than their national peers.

When benchmarked against adults 60 years and older in NHANES 2015–2020, the NECS cohort achieved a markedly higher total HEI score (70.1 vs 62.9, P < 0.001), reflecting generally stronger alignment with the Dietary Guidelines [53]. Component-level contrasts show that NECS participants outperformed national peers on total fruit (4.2 vs 3.6) and greens & beans (4.4 vs 3.1). They also accumulated more points on moderation metrics, including grains, sodium, added sugar, and saturated fat, indicating lower intakes of these items. Conversely, NECS participants’ diets lagged on total vegetables (3.4 vs 3.8), whole fruits (4.4 vs 4.8), and whole grains (2.0 vs 3.6).

AHEI totals of our NECS sample were slightly higher than HRS/HCNS (NECS 51.9 vs 48.6, P < 0.001) [51]. NECS participants scored higher for intakes of fruits (4.4 vs 3.8) and sodium (7.5 vs 5.5). Conversely, they scored lower on whole grains (2.3 vs 3.5), nuts (5.1 vs 6.1), and scored lower for the polyunsaturated-to-saturated fat ratio (5.6 vs 6.1). In contrast, the NECS group excelled in long-chain n-3 fats (8.2 vs 4.2), which are known to positively contribute to cardiovascular health, lower inflammation, and neuroprotection.

PHDI is a relatively new metric with few published reference values in US populations. Using the same Harvard FFQ, the NHS and HPFS reported mean (SD) PHDI scores of 83.8 (13.1) and 79.5 (20.6) [42], respectively, while our mixed-sex NECS cohort averaged 87.1 (12.0). Component-level patterns were broadly comparable across cohorts, with participants in all three studies meeting or approaching the 10-point targets for fruit, vegetables, poultry, eggs, and fish, but lagging on legumes and soy (≤2 pts of 10 in every sample). For whole grains, NECS scored 2.3 pts versus 4.3 (NHS) and 3.8 (HPFS), well below the ≥75/90 g EAT Lancet benchmark. For added saturated fat & sugars, NECS scored higher for saturated fat (4.9 vs 3.0 NHS; 2.1 HPFS) and for added sugars/juice (7.6 vs 6.0 NHS; 4.5 HPFS). And for Red/processed meat, HPFS men incurred a larger penalty (4.5 pts) than NECS (5.8 pts) or NHS (5.7 pts), consistent with sex-specific dietary habits. These findings suggest that NECS participants had dietary patterns that offer not only higher nutritional values but also planetary sustainability when compared to high-literacy cohorts; still, opportunities remain to increase whole-grain, soy, and legume consumption. The low legume scores across all three well-characterized cohorts underscored a broader dietary gap even among populations with relatively good diet quality.

NECS participants also surpassed HRS on the MIND pattern scores (8.6 vs 7.1, P < 0.001). CO families scored higher on leafy greens (0.6 vs 0.4), other vegetables (0.8 vs 0.6), berries (0.8 vs 0.3), nuts (0.6 vs 0.4), olive oil use (0.4 vs 0.3), fish (0.8 vs 0.6), and poultry (0.6 vs 0.2) [52]. These differences, many linked to Mediterranean-style eating, were consistent with the cohort’s relatively high educational attainment, which often predicts diet quality. Areas for refinement mirror the HEI findings: whole grains (0.2 vs 0.3) and legumes (0.2 vs 0.4) remain low. NECS participants also scored worse on pastries/sweets. These comparisons indicate that centenarian families exhibit a modestly higher overall diet that provides neuroprotection than age-matched U.S. peers. Taken together, these cross-cohort patterns indicate that CO diets were modestly more favorable than those of age-matched U.S. peers, especially for items known to support metabolic, cardiovascular, and cognitive domains of healthy aging (e.g., higher intake of fish, fruits/vegetables; lower intake of sodium, added sugar, etc.).

Beyond these patterns, the sociodemographic differentials in the diet scores may inform future public health policy. Education emerged as the most consistent determinant of dietary quality at both the composite level (Table 2; Supplemental Tables 5–8) and the component level (Supplemental Tables 9–12). Participants with higher education scored higher on all four indices and also scored more favorably on produce (total vegetables, fruits), high-fiber foods (whole grains, legumes), and lean or plant-based proteins (fish, soy, poultry, eggs) (Supplemental Tables 9–12). Age, sex, and marital status also influenced dietary profiles. Both the AHEI and MIND indices declined with age, likely reflecting poorer adherence to dietary recommendations in older cohorts. Specifically, older participants consumed more red and processed meats and greater amounts of added sugars, butter, oils, and sodium. Although sex was not a significant predictor of overall index scores, women consistently reported higher consumption of vegetables and dairy, and lower intake of added sugars (Supplemental Tables 9–12). These results highlight the need to look beyond total index scores to detect important sex-specific dietary behaviors. Participants who were not married tend to score higher on red meat and sodium consumption, and score lower on vegetables than those who were married. It is important to note that we did not observe differences in dietary scores between CO and non-CO participants, most of the latter are spouses who are likely to share meals and food environments and to resemble one another through assortative mating (selection on education, culture, and health behaviors), which are all factors that may attenuate between-group contrasts in diet quality [54]. Therefore, these null differences should not be taken as evidence that familial longevity is unrelated to diet per se, and a stronger test will require non-spouse controls—peers without parental longevity (i.e., parents with average life expectancy)—matched on age, sex, and socioeconomic factors. Our results were consistent with broader evidence that sociodemographic variables often outweigh familial predispositions in eating habits [55].

Given the strong and well-documented association between educational attainment and dietary quality, and the relatively high educational attainment of the NECS cohort, it is reasonable to expect that differences in diet quality between NECS participants and population-based cohorts could be partially attributable to socioeconomic factors. To assess this possibility, we conducted sensitivity analyses using nationally representative HRS data stratified by educational attainment.

Under these stratified comparisons, differences in HEI scores between NECS and HRS were attenuated, indicating that education explains a large proportion of the observed cohort differences in diet quality (Supplemental Table 13). However, modest differences persisted within comparable education categories, suggesting that socioeconomic factors alone do not fully account for the dietary patterns observed among centenarian offspring. These residual differences may reflect shared familial, cultural, or behavioral characteristics within longevity-enriched families, though such interpretations should be made cautiously and require further examination in future research.

5. Strengths and limitations

NECS is a well-phenotyped sample of older adults who are offspring of centenarians, allowing us to examine diet in families enriched for exceptional-longevity genes. The use of validated Harvard 131-item FFQ provides a robust assessment of diet quality that is comparable to other large cohorts. The inclusion of four validated dietary indices also provides a comprehensive view of this cohort’s dietary profile. Several limitations warrant consideration. First, because dietary intake was assessed at a single time point, findings should be interpreted as descriptive of dietary patterns and the relative ranking of diet quality within the sample at the time of assessment. While the Harvard FFQ has been extensively validated in older adult populations, it was not specifically validated within the NECS cohort. Nevertheless, FFQs are well suited for characterizing habitual dietary patterns, which was the primary objective of this study. Second, the NECS cohort is a predominantly White, highly educated cohort; our findings may not extend to more diverse populations. Third, whole grain intake was estimated, and added sugar from cereal was omitted due to unavailable information on cereal brands. Thus, some component-level scores may not accurately reflect real intake in the NECS sample. For instance, the consistently lower score of whole grain consumption in comparison to the national averages may be caused by the conservative estimation we used. Fourth, as PHDI is a relatively novel index, the only references that we could find were the NHS and HPFS, which were not representative samples and may lack comparability to the NECS cohort. Finally, cross-study comparisons should be interpreted cautiously. Our NECS diet was assessed by a Harvard FFQ in 2005, whereas NHANES estimates commonly cited in the literature were based on 24-h recalls (2015–2020). Instrument differences (FFQ vs 24-h recalls) and also changing trends in the consumption of various food items can produce systematic differences in component scores independent of true intake. We present the diet scores from other cohorts only as contextual benchmarks and not formal statistical comparisons. Though interactions in our models reached nominal significance, none remained significant after false-discovery-rate adjustment; we therefore treat them as exploratory.

6. Conclusion

By examining dietary indices among NECS participants, we found that their overall dietary quality was moderate but surpassed that of the national averages. However, their diet is not uniformly optimal and instead reflects a profile of selective strengths and persistent gaps. The dietary components in which CO families excel, fruits, vegetables, plant proteins, long-chain omega-3 fats, and lower added sodium and sugars, correspond with patterns linked to lower chronic inflammation, improved metabolic homeostasis, and preserved functional capacity in aging. These elements may represent behavioral pathways that complement or amplify inherited biological resilience in this cohort. Conversely, the persistent shortfalls in whole grains, legumes, nuts, and soy may represent missed opportunities to optimize muscle, cardiometabolic, and cognitive health. Even in longevity-enriched families, dietary gaps mirror national trends, reinforcing that nutrition is a critical yet underutilized lever for promoting healthy longevity.

Education emerged as the strongest and most consistent determinant of diet quality, surpassing familial longevity status. Age and sex also shaped component-level patterns: older participants consumed more red/processed meat, added sugars, and sodium, whereas women reported higher intake of vegetables and dairy and lower intake of added sugars. These results reinforce that sociodemographic and environmental factors often outweigh familial predispositions in shaping nutritional behavior. From a public-health perspective, these findings highlight several priorities: (1) nutrition interventions for older adults should emphasize accessible education—such as label literacy and cooking skills—and develop gender-specific targeted strategies; and (2) even among longevity-enriched families with high health literacy, whole grain and legume intake remains consistently low, underscoring the need for policies that improve affordability and availability of these foods.

As the current CO cohort continues to age, with some of them eventually becoming centenarians themselves, these current baseline data become solid infrastructure for future longitudinal analysis, establishing associations between dietary indices and various domains of health among individuals predisposed to exceptional longevity. Studying long-lived individuals and how dietary quality and its interaction with other social and biological dimensions impact longevity remains an exciting area for exploration.

Funding sources

This study was supported by the USDA Agricultural Research Service under Cooperative Agreement No. 58-8050-3-003 and the National Institutes of Health (NIH), NIA cooperative agreementsU19-AG023122 [Longevity Consortium]. Andres Ardisson Korat was supported by NIH training grant 1K12TR004384.

Declaration of Generative AI and AI-assisted technologies in the writing process

We have nothing to disclose regarding generative AI use. All scientific content, data analyses, interpretations, and conclusions were developed, reviewed, and verified by the authors. The authors take full responsibility for the accuracy and integrity of the final manuscript.

Data sharing

Data described in the manuscript, code book, and analytic code will be made available upon request pending approval.

Declaration of interest statement

The authors have no conflicts of interest to declare.

Appendix A. Supplementary data

The following is Supplementary data to this article: