Mid‐life adherence to the Dietary Approaches to Stop Hypertension (DASH) diet and late‐life subjective cognitive complaints in women

Abstract

INTRODUCTION

Evidence is limited on the role of mid‐life Dietary Approaches to Stop Hypertension (DASH) diet in late‐life subjective cognitive complaints (SCCs).

METHODS

We included 5116 women (mean age in 1985–1991: 46 years) from the New York University Women's Health Study. SCCs were assessed from 2018 to 2020 (mean age: 79 years) by a 6‐item questionnaire.

RESULTS

Compared to women in the bottom quartile of the DASH scores, the odds ratio (OR) for having two or more SCCs was 0.83 (95% confidence interval: 0.70–0.99) for women in the top quartile of DASH scores at baseline (P for trend = 0.019). The association was similar with multiple imputation and inverse probability weighting to account for potential selection bias. The inverse association was stronger in women without a history of cancer (P for interaction = 0.003).

DISCUSSION

Greater adherence to the DASH diet in mid‐life was associated with lower prevalence of late‐life SCCs in women.

1. INTRODUCTION

Dementia is characterized by decline in memory and other cognitive domains that interferes with daily activities. Alzheimer's disease (AD) is the most prevalent form of dementia. ¹ , ² Approximately 6.5 million Americans aged > 65 were diagnosed with AD in 2022, and that number is expected to more than double by 2060. ³ Almost two thirds of AD cases are women. ⁴ Subjective cognitive complaints (SCCs) are self‐reported impairments in daily cognitive performance characterized by a greater frequency of memory loss without objective signs of cognitive impairment from neuropsychological test performance. ⁵ , ⁶ Though SCCs have been associated with some non‐neurodegenerative factors (i.e., depression, anxiety), ⁷ , ⁸ , ⁹ it is increasingly considered indicative of mild cognitive impairment (MCI) and as a plausible predictor for incident neurocognitive disorders, including AD. ⁶ , ¹⁰ Research into risk factors for SCCs may offer insight into the pathology of AD/dementia and potentially improve risk assessment and early prevention.

Recent data suggest that health conditions, including obesity, hypertension, and diabetes, during mid‐life (40 to 60 years of age), rather than later in life, contribute to risk of late‐life cognitive impairment. ¹¹ , ¹² This is consistent with biological processes underlying neurodegeneration occuring over decades, ¹¹ and suggests that epidemiologic studies should assess risk factors in mid‐life. Diet is a modifiable risk factor that may help protect brain health. ¹³ Recent studies have focused on the effect of intake of specific food groups and nutrients on SCCs. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ These studies have shown the potential benefits of carotenoids, ¹⁵ flavonoids, ¹⁴ protein intake, ¹⁶ overall intake of energy and fat, ¹⁷ as well as the consumption of vegetables, fruits, and orange juice, ¹⁸ in maintaining cognitive function. Regarding dietary patterns, adherence to a Mediterranean diet at mid‐life was associated with better late‐life subjective cognitive function in men in a prospective study. ¹⁹ However, the role of other types of diet is not as well established. Prior studies have indicated that the Dietary Approaches to Stop Hypertension (DASH) diet—which includes a high consumption of plant‐based foods that are rich in potassium, calcium, and magnesium, and limits the intake of saturated fat, cholesterol, sodium, and sugar—prevents high blood pressure, an important risk factor for cognitive decline and AD. ²⁰ , ²¹ , ²² Only one prospective study, the Nurses’ Health Study (NHS), has assessed the effects of the DASH diet at mid‐life on late‐life SCCs. The results of this study support a beneficial role of long‐term adherence to DASH dietary pattern in subjective cognitive function among women. ²³ Studies are needed to confirm these findings and explore potential effect modifiers, such as demographic and lifestyle factors and comorbidities, such as cancer and depression, given their associations with cognitive function, AD, or dementia and/or dietary patterns. ⁷ , ⁸ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹

RESEARCH IN CONTEXT

1. Systematic review: We reviewed the literature on dietary patterns as a potential modifiable risk factor for subjective cognitive complaints (SCCs), a validated precursor of cognitive dysfunction. We noticed a gap and lack of evidence in the relationship between mid‐life Dietary Approaches to Stop Hypertension (DASH) diet and late‐life SCCs.

2. Interpretation: Greater adherence to the DASH diet in mid‐life was associated with lower late‐life SCCs among women. Our findings suggest that improvements in diet quality during mid‐life, especially the diet related to hypertension and cardiovascular profile, may have a role in maintaining an optimal subjective cognitive function among women.

3. Future directions: These findings highlight the important role of diet in cognitive function. Future studies are needed to assess: (1) the impact of DASH diet on objective measures of cognitive function; (2) the generalizability of the results across a variety of racial/ethnic groups.

The objective of the present study was to assess the association between the DASH score measured in mid‐life, and SCCs reported later in life in the New York University Women's Health Study (NYUWHS), a prospective cohort study with > 30 years of follow‐up.

2. METHOD

2.1. Study design

The NYUWHS recruited 14,274 women aged 35 to 65 at a mammography screening center in New York City between 1985 and 1991. ³⁰ Women were ineligible for enrollment if they had used hormonal medications or had been pregnant in the previous 6 months. At enrollment, each woman was asked to complete a questionnaire about demographics, physical activity, reproductive history, cancer history, and medication use in the previous 4 weeks. Participants also completed a food frequency questionnaire (FFQ), described below. Participants were followed up using questionnaires mailed every 3 to 5 years; participants who did not return questionnaires were contacted by phone. ³⁰ The follow‐up questionnaires asked about health events and updated information on lifestyle. As the cohort grew older, we included in our two most recent questionnaires (2018–present) questions on SCCs, adapted from other cohort questionnaires. ¹⁴ , ¹⁵ , ¹⁹ , ²³ The most recent follow‐up of the NYUWHS began in October 2020. As of October 2022, 7795 participants (54.6%) were still alive, of whom 1260 (16.2%) had previously withdrawn from active follow‐up and were therefore not contacted. Of the 6535 remaining women, 5116 (78.3%) women have completed at least one of the SCC surveys, while 3690 (56.5%) completed both SCC surveys. All NYUWHS study procedures were approved by the institutional review board of NYU School of Medicine.

2.2. Assessment of dietary intake

The dietary questionnaire completed by participants at enrollment was a semi‐quantitative validated FFQ ³¹ adapted from the Block FFQ, which was shown to have adequate temporal reproducibility in the NYUWHS. ³¹ , ³² Daily intake of foods and nutrients was calculated on the basis of food lists and food composition tables elaborated at the National Cancer Institute, with minor modifications. ³¹ We constructed the DASH score using FFQ data according to published guidelines with eight food components, ³³ defined as low intakes of red meat, sodium, sweets, and high intakes of fruits (including fruit juices), vegetables (excluding potatoes), legumes and nuts, low‐fat dairy, and grains (Table S1 in supporting information). ³⁴ Each food component was classified into quintiles with higher quintiles representing greater consumption. Red meat, sodium, and sweets are considered indicative of low adherence to the DASH diet. Therefore, the highest quintile was assigned the lowest score; for example, the top quintile was assigned 1 point while the bottom quintile was assigned 5 points. In contrast, high intakes of fruits, vegetables, legumes and nuts, low‐fat dairy, and grains are indicative of greater adherence to the DASH diet and the highest quintiles were thus given the highest score ( = 5). The total DASH score was determined by adding the scores of the eight DASH components, resulting in a potential range of 8 to 40, with higher DASH scores reflecting greater adherence to DASH diet. ³³ , ³⁵ , ³⁶ , ³⁷ The DASH score has been validated ³⁸ , ³⁹ , ⁴⁰ and considered in many studies of cancer and cardiovascular disease. ²⁹ , ⁴¹ We then used the DASH score distribution to classify women into quartiles, resulting in the following groupings: Q1 (10–20), Q2 (21–24), Q3 (25–27), Q4 (28–37). This quartile classification was determined so that we have sufficient power (0.80; estimated using Quanto) ⁴² to estimate approximately a minimal odds ratio (OR) of 0.80 in a comparison between any two quartiles.

2.3. Assessment of subjective cognitive decline

Six self‐completed yes/no questions were used to assess SCCs. These questions were similar to those used by other cohorts, ¹⁴ , ¹⁵ , ¹⁸ , ¹⁹ , ²³ , ⁴³ including the NHS ¹⁴ , ¹⁵ , ²³ and the Health Professionals Follow‐Up Study (HPFS). ¹⁴ , ¹⁸ , ¹⁹ The questions have been validated against objective features of dementia, clinically established cognitive testing questionnaires for memory loss, and the apolipoprotein E ɛ4 genotype. ⁴⁴ , ⁴⁵ Participants were asked whether they experienced the following memory complaints: (1) recent change in the ability to remember things; (2) difficulty remembering recent events; (3) difficulty remembering a short list of items, such as a shopping list; (4) difficulty understanding or following spoken instructions; (5) difficulty following a group conversation or a plot in a TV program; (6) difficulty navigating familiar streets. Every complaint was worth 1 point and then summed for a total score, ranging from 0 to 6. The frequencies of endorsement for each of the six complaints on the two SCC assessments are shown in Table S2 in supporting information.

2.4. Covariates

The following variables, collected at enrollment, were examined as potential confounders and/or effect modifiers: menopausal status (premenopausal or postmenopausal), race/ethnicity (White or other), educational attainment (< college, college graduate, or graduate school), smoking status (never vs. ever), total calories intake (continuous, kcal) and body mass index (BMI; < 23, 23–24.9, 25–29.9, or ≥30 kg/m²), which was computed using self‐reported height and weight. We also considered in our analysis several health conditions reported at the time of SCC collection, such as age (continuous, years) and history of depression (having ever taken depression medication, yes/no). ⁴³ History of cancer was determined based on self‐report, with confirmation by review of medical records, and linkages to tumor registries of New York, New Jersey, and Florida.

2.5. Statistical analysis

As we did in a previous analysis, ⁴³ we classified women with ≥ 2 SCCs as having some indication of cognitive loss, while women with < 2 SCCs were classified as having no sign of cognitive loss. This cut‐off point was chosen because ≈ 50% of participants reported having at least one complaint. For women who completed the two follow‐up questionnaires including questions on SCCs, reporting ≥ 2 SCCs on either one of these two questionnaires led to being classified as having some indication of cognitive loss. For descriptive analyses, we examined covariate distributions in the two SCC categories (< 2, ≥2) and by quartiles of DASH scores. We used the t test for continuous covariates, and Pearson chi‐square for categorical covariates.

To evaluate the relationship between DASH diet and SCCs, we conducted unconditional logistic regression and estimated ORs for having ≥ 2 SCCs, comparing the higher three quartiles of the DASH scores with the bottom quartile. We also examined linear trends across quartiles. We conducted similar analyses for each component of DASH. We first adjusted for age at the time of SCCs and other baseline potential confounders, including, menopausal status, race, education attainment, smoking status, BMI, and total caloric intake (Model 1). Additional adjustment for history of cancer and use of depression medication was conducted (Model 2), as these comorbidities may influence cognition and be associated with diet. ⁷ , ²⁴ , ²⁵ , ²⁷ , ²⁹ We hypothesized that the comorbidity conditions are potential confounders and/or effect modifiers. Data on physical activity at baseline were missing for a large number of participants in the main cohort (n = 4677) and in this subset (n = 800); therefore, physical activity was not included in the adjustments. Given that cardiometabolic diseases, including hypertension, diabetes, heart disease, and stroke may be in the pathway of DASH diet and SCCs, we did not control for these variables in the main model. For the eight DASH components, we conducted an additional model that adjusts for the variables in model 2 and all other DASH components simultaneously (model 3) to test the independent association between each component and SCCs.

Stratified analyses were conducted to assess potential effect modification by other factors. Age at the time of SCCs (≤78 vs. > 78 years) and total calories (≤1430.21 vs. > 1430.21 kcal) were dichotomized based on the median values among women included in this analysis, while BMI (≤25 vs. > 25 kg/m²) was dichotomized as overweight or not, based on the BMI classification criteria. ⁴⁶ Categories of educational attainment were combined (pre‐college vs. college or beyond), while categories of race/ethnicity included White, Black, Latina, and other (including Asians, Native Americans, and individuals with more than one race) to assess differences in the associations by these groups. Multiplicative interaction was assessed using the cross‐product between per DASH score and each dichotomized potential effect modifier.

It has been indicated that there were strong, linear trends of increasingly worse scores on objective cognitive tests with increasing numbers of SCCs. ⁴⁷ In exploratory analyses, we assess the relationship between per quartile difference in the DASH score and number of SCCs as a continuous outcome in linear regression. This analysis was limited to women who participated in both surveys (n = 3690), and the sum of SCC score from the two questionnaires, ranging from 0 to 12, was used as the dependent variable in a linear regression model. Previous studies from the HPFS ¹⁴ , ¹⁸ , ¹⁹ and NHS ¹⁴ , ¹⁵ , ²³ also computed the continuous SCC score as an average of the two assessments (2008 and 2012 for the HPFS, 2012 and 2014 for the NHS), which aligns consistently with our study approach.

The time difference between the two assessments was 2 to 3 years, not long enough to have substantial longitudinal changes in cognition. Therefore, we considered defining the endpoint using two assessments for women who had two assessments (72% of the study population) to increase the reproducibility of the endpoint. We also conducted sensitivity analyses using the first SCC assessment only (n = 4433) for logistic regression with dichotomized SCCs and for linear regression with continuous SCCs. We compared the distributions of demographic and health‐related characteristics in the study population by the availability of the two assessments, that is, when we considered the first assessment only (n = 4433), and when we considered women with at least one assessment (n = 5116), and two assessments for women who had two assessments (n = 3690).

Covariate missing data were handled in two ways. For the main analyses above, we created a “missing” category for each variable with missing values. We also conducted multiple imputation for missing values using Monte Carlo methods ⁴⁸ , ⁴⁹ based on baseline demographic variables including age, race, education level, BMI, and smoking status. As the present study could only include a subset of the NYUWHS participants, that is, women who completed at least one survey including questions on SCCs, we also conducted analyses using the inverse probability weighting (IPW) method to assess the impact of potential selection bias. ⁵⁰ , ⁵¹ , ⁵² In these analyses, we modeled the probability of inclusion in the present study by fitting a logistic regression model with demographic and lifestyle variables, and comorbidities, at baseline and/or follow‐up, as predictors.

All analyses were conducted using SAS version 9.4 and R Studio version 4.2.

3. RESULTS

For the 5116 participants included in this study, the mean age at baseline was 46.3 years, the mean follow‐up time was 32.6 years (range 27.1–37.0), and the mean age at completion of the SCC questionnaire was 78.9 years (range 62.7–100.6; Table 1). Overall, 32.5% of the study participants reported no SCC, 17.7% reported one, 32.9% reported two, and 16.9% reported three or more in at least one survey. According to Table S2, frequencies of endorsement for each of the six SCCs were similar between the 2018 and 2020 surveys. Of the six SCC complaints, women were more likely to report complaints related to memory, such as recent change in the ability to remember things (40%), difficulty remembering a short list of items (19%), and difficulty remembering recent events (17%). A small portion of women reported complaints related to other cognitive domains (SCCs 4–6; 3%–6%). Compared to participants with < 2 SCCs, those who reported ≥2 SCCs had lower educational attainment (P = 0.002), were more likely to have ever been smokers (P < 0.001), and had a higher BMI (P = 0.02) at enrollment. At the time of SCC assessment, they were more likely to be older (P < 0.001) and have a history of diabetes (P = 0.001), cardiovascular disease (P < 0.001), hypertension (P = 0.005), and/or depression (P < 0.001). The mean DASH score was 24 points, ranging from 10 to 37. Women with higher DASH scores were older at enrollment (P < 0.001), had a higher educational attainment (P < 0.001), had a lower BMI at baseline (P < 0.001), and were less likely to have a history of hypertension (P = 0.004, Table S3 in supporting information).

TABLE 1.

Distribution of demographic and health‐related characteristics by SCCs (≥2 versus < 2).

Characteristic mean (SD) or N (%)	Overall (n = 5116)	SCCs < 2 (n = 3486)	SCCs ≥ 2 (n = 1630)	P‐value g
Baseline demographic and lifestyle variables
Age at enrollment, mean (std), years	46.31 (7.16)	45.74 (6.90)	47.52 (7.57)	<0.001
Menopausal status at enrollment, n (%)				<0.001
Premenopausal	3802 (74.3%)	2680 (76.9%)	1122 (68.8%)
Postmenopausal	1314 (25.7%)	806 (23.1%)	508 (31.2%)
Race/ethnicity, n (%) a				0.473
White	3919 (81.5%)	2673 (82.0%)	1246 (80.4%)
Black	503 (10.5%)	336 (10.3%)	167 (10.8%)
Latina	306 (6.4%)	196 (6.0%)	110 (7.1%)
Asian/Native American/more than one race	82 (1.7%)	55 (1.7%)	27 (1.7%)
Education, n (%) b				0.002
< college	2038 (43.6%)	1328 (41.9%)	710 (47.2%)
College	1208 (25.8%)	835 (26.3%)	373 (24.8%)
Graduate school	1432 (30.6%)	1010 (31.8%)	422 (28.0%)
Smoking at enrollment, n (%) c				<0.001
Never	2370 (47.0%)	1667 (48.6%)	703 (43.5%)
Ever	2677 (53.0%)	1764 (51.4%)	913 (56.5%)
BMI at enrollment, n (%) d				0.022
Underweight, < 18.5	108 (2.1%)	85 (2.5%)	23 (1.4%)
Normal weight, 18.5 ≤ BMI < 25	3386 (66.6%)	2326 (67.1%)	1060 (65.4%)
Overweight, 25 ≤ BMI < 30	1134 (22.3%)	759 (21.9%)	375 (23.1%)
Obese, ≥30	459 (9.0%)	296 (8.5%)	163 (10.1%)
Variables from follow‐up
Age at completion of SCCs, mean (std), years	78.89 (7.60)	78.26 (7.30)	80.25 (8.06)	<0.001
History of diabetes, n (%)	798 (15.6%)	505 (14.5%)	293 (18.0%)	0.001
History of heart disease, n (%) e , h	389 (7.6%)	234 (6.7%)	155 (9.6%)	<0.001
History of stroke, n (%) f	174 (3.4%)	91 (2.6%)	83 (5.1%)	<0.001
History of cancer, n (%)	1610 (31.5%)	1074 (30.8%)	536 (32.9%)	0.137
History of hypertension, n (%) l	3059 (60.0%)	2040 (58.6%)	1019 (62.8%)	0.005
History of depression medications, n (%) m	963 (18.9%)	504 (14.5%)	459 (28.3%)	<0.001
History of cholesterol lowering medications, n (%) n	2734 (53.6%)	1820 (52.4%)	914 (56.3%)	0.008
DASH score i
Mean (SD)	24.00 (4.56)	24.08 (4.55)	23.84 (4.57)	0.088
Q1 (10–20), n (%)	1179 (23.1%)	775 (22.2%)	404 (24.8%)
Q2 (21–24), n (%)	1597 (31.2%)	1085 (31.1%)	512 (31.4%)
Q3 (25–27), n (%)	1155 (22.5%)	809 (23.2%)	346 (21.2%)
Q4 (28–37), n (%)	1185 (23.2%)	817 (23.5%)	368 (22.6%)
DASH components, g/day, mean (std)
Inverse Item or unhealthy j
Red and processed meat	33.64 (31.94)	33.54 (32.93)	33.86 (29.73)	0.724
Sodium	2.69 (1.39)	2.69 (1.39)	2.69(1.40)	0.918
Sweets	41.72 (56.51)	41.24 (59.44)	42.74 (49.68)	0.347
Direct Items or healthy k
Fruits (including fruit juices)	327.81 (264.69)	332.44 (274.58)	317.89 (241.99)	0.055
Vegetables excluding potatoes	192.06 (168.05)	194.32 (178.00)	187.22 (144.40)	0.129
Legumes and Nuts	25.11 (32.05)	25.35 (33.84)	24.58 (27.84)	0.392
Low fat dairy	205.88 (252.19)	206.95 (257.05)	203.61 (241.54)	0.653
Whole grains	29.95 (45.51)	30.18 (44.84)	29.46 (46.91)	0.604
Food groups/nutrients
Total calories (kcal)	1530.75 (692.52)	1530.24 (694.06)	1531.85 (689.41)	0.939

Abbreviations: BMI, body mass index; DASH, Dietary Approaches to Stop Hypertension; SCCs, subjective cognitive complaints; SD, standard deviation.

^a Missing value, n = 306.

^b Missing value, n = 438.

^c Missing value, n = 69.

^d Missing value, n = 29.

^e Missing value, n = 14.

^f Missing value, n = 11.

^g t test for continuous covariates, Pearson chi‐square for categorical covariates.

^h Heart disease includes heart attack, myocardial infarction, and congestive heart failure.

ⁱ DASH score ranged from 10 to 37, with a higher score indicating a healthier diet.

^j Indicative of low adherence to the DASH diet.

^k Indicative of high adherence to the DASH diet.

^l Missing value, n = 13.

^m Missing value, n = 17.

ⁿ Missing value, n = 17.

There was an inverse association between DASH adherence and SCCs. Women in the highest quartile of DASH intake at baseline had a 17% reduction in the odds of having ≥2 SCCs, compared to those in the bottom quartile (OR = 0.83; 95% confidence interval [CI]: 0.70–0.99; P‐trend = 0.016; model 1; Table 2). The association remained after adjustment for additional covariates (OR = 0.83; 95% CI: 0.70–0.99; P‐trend = 0.019; model 2). When we considered the DASH score a continuous variable, every quartile increase in DASH was associated with an OR of 0.93 (95% CI: 0.87–0.99) for reporting ≥2 SCCs in model 2. When we separately assessed the association between each of the eight DASH components and SCCs, only sweets intake (servings/day) had an apparent positive association with odds of having ≥2 SCCs (P‐trend for model 2 = 0.026). However, after controlling for all other components, the association between sweets alone and SCCs was attenuated (P‐value = 0.133), indicating that the association between sweets and SCCs was not independent of other DASH components.

TABLE 2.

ORs (95% CIs) for SCCs (< 2 vs. ≥ 2) associated with adherence to DASH and its components, from unconditional logistic regression (n = 5116).

	Quartile of intake OR (95% CI)					OR (95% CI) per quartile a
variables	Q1	Q2	Q3	Q4	P‐trend
DASH
Cases / N total	404 / 1179	512 / 1597	346 / 1155	368 / 1185
Model 1 b	1	0.91 (0.77–1.06)	0.79 (0.67–0.95)	0.83 (0.70–0.99)	0.016	0.93 (0.88–0.99)
Model 2 c	1	0.91 (0.77–1.07)	0.79 (0.66–0.95)	0.83 (0.70–0.99)	0.019	0.93 (0.87–0.99)
Red and processed meat d
Cases / N total	388 / 1279	412 / 1279	416 / 1279	414 / 1279
Model 1	1	1.06 (0.89–1.25)	1.07 (0.91–1.27)	1.06 (0.89–1.27)	0.473	1.02 (0.97–1.08)
Model 2	1	1.09 (0.91–1.29)	1.10 (0.93–1.31)	1.11 (0.93–1.33)	0.249	1.03 (0.98–1.09)
Model 3	1	1.06 (0.89–1.27)	1.07 (0.90–1.28)	1.05 (0.87–1.27)	0.618	1.02 (0.96–1.08)
Sodium
Cases / N total	424 / 1279	378 / 1279	423 / 1279	405 / 1279
Model 1	1	0.88 (0.73–1.05)	1.04 (0.85–1.27)	1.01 (0.78–1.31)	0.682	1.02 (0.94–1.11)
Model 2	1	0.86 (0.72–1.03)	1.05 (0.86–1.29)	1.03 (0.78–1.34)	0.484	1.03 (0.95–1.12)
Model 3	1	0.85 (0.71–1.02)	1.03 (0.84–1.26)	0.96 (0.73–1.28)	0.792	1.01 (0.93–1.11)
Sweets
Cases / N total	395 / 1279	396 / 1279	407 / 1279	432 / 1279
Model 1	1	1.05 (0.89–1.25)	1.10 (0.92–1.30)	1.22 (1.02–1.46)	0.032	1.07 (1.01–1.13)
Model 2	1	1.07 (0.90–1.27)	1.11 (0.94–1.33)	1.23 (1.02–1.48)	0.026	1.07 (1.01–1.13)
Model 3	1	1.05 (0.88–1.25)	1.08 (0.90–1.29)	1.16 (0.96–1.41)	0.133	1.05 (0.99–1.11)
Fruits, including Fruit Juices
Cases/ N total	411 / 1279	419 / 1279	402 / 1279	398 / 1279
Model 1	1	1.00 (0.84–1.18)	0.92 (0.77–1.09)	0.88 (0.73–1.05)	0.125	0.96 (0.90–1.01)
Model 2	1	1.00 (0.84–1.18)	0.94 (0.79–1.12)	0.89 (0.74–1.07)	0.180	0.96 (0.91–1.02)
Model 3	1	1.01 (0.85–1.20)	0.96 (0.80–1.15)	0.94 (0.77–1.15)	0.479	0.98 (0.92–1.04)
Vegetables without potatoes
Cases / N total	424 / 1279	402 / 1279	403 / 1279	401 / 1279
Model 1	1	0.90 (0.76–1.07)	0.90 (0.75–1.06)	0.89 (0.74–1.06)	0.197	0.96 (0.91–1.02)
Model 2	1	0.89 (0.75–1.05)	0.91 (0.77–1.09)	0.89 (0.74–1.07)	0.271	0.97 (0.91–1.03)
Model 3	1	0.90 (0.76–1.07)	0.94 (0.79–1.12)	0.95 (0.78–1.14)	0.669	0.99 (0.93–1.05)
Legumes and nuts
Cases/ N total	388 / 1279	418 / 1278	437 / 1280	387 / 1279
Model 1	1	1.14 (0.96–1.35)	1.19 (1.00–1.41)	1.00 (0.83–1.20)	0.871	1.01 (0.95–1.06)
Model 2	1	1.18 (0.99–1.40)	1.23 (1.03–1.46)	1.01 (0.84–1.22)	0.763	1.01 (0.95–1.07)
Model 3	1	1.19 (1.00–1.41)	1.25 (1.05–1.49)	1.05 (0.87–1.27)	0.454	1.02 (0.97–1.08)
Low‐fat dairy
Cases / N total	419 / 1278	382 / 1280	420 / 1280	409 / 1278
Model 1	1	0.88 (0.74–1.04)	0.96 (0.81–1.14)	0.92 (0.76–1.10)	0.566	0.98 (0.93–1.04)
Model 2	1	0.87 (0.73–1.03)	0.97 (0.82–1.16)	0.88 (0.73–1.06)	0.366	0.97 (0.92–1.03)
Model 3	1	0.88 (0.74–1.05)	0.99 (0.83–1.18)	0.89 (0.74–1.08)	0.482	0.98 (0.92–1.04)
Whole grains
Cases / N total	422 / 1279	418 / 1279	395 / 1279	395 / 1279
Model 1	1	0.99 (0.84–1.18)	0.92 (0.77–1.09)	0.90 (0.75–1.07)	0.153	0.96 (0.91–1.02)
Model 2	1	1.00 (0.84–1.19)	0.93 (0.78–1.10)	0.91 (0.76–1.09)	0.203	0.96 (0.91–1.02)
Model 3	1	1.01 (0.85–1.20)	0.94 (0.79–1.11)	0.93 (0.77–1.12)	0.317	0.97 (0.92–1.03)

Abbreviations: BMI, body mass index; CI, confidence interval; DASH, Dietary Approaches to Stop Hypertension; OR, odds ratio; SCCs, subjective cognitive complaints.

^a OR associated with per quartile difference in the scores of DASH and its components.

^b Model 1 adjusted for age (continuous; at the time of SCCs), menopausal status (premenopausal or postmenopausal; baseline), race (White or other; baseline), education (< college, college, or graduate school; baseline), smoking status (never or ever; baseline), BMI (< 23, 23–24.9, 25–29.9, or ≥30 kg/m2; baseline), and total caloric intake (continuous; baseline).

^c Model 2 had the same adjustments as model 1 plus history of cancer (yes or no; follow‐up) and depression medication (yes or no; follow‐up).

^d Model 3 had the same adjustments as model 2 plus all other 7 DASH components.

Results of stratified analyses, with the DASH score examined as a continuous variable, are presented in Figure 1. Upon examining the associations among different racial/ethnic groups, the association among Black women was stronger than that that in White women, though the difference was marginally significant (P‐value for interaction = 0.06). The inverse association was also stronger in women without a history of cancer, compared to those with a history of cancer (P‐value for interaction = 0.005). The association did not differ by other potential effect modifiers.

FIGURE 1.

Stratified analyses of DASH adherence and SCCs (< 2 vs. ≥ 2; n = 5116).* OR associated with per quartile difference in the DASH score. Model adjusted for all the variables in the figure except for the effect modifier. BMI, body mass index; CI, confidence interval; DASH, Dietary Approaches to Stop Hypertension; OR, odds ratio; SCCs, subjective cognitive complaints

In linear regression considering the sum of two SCC assessments as a continuous dependent variable (mean = 1.67, standard deviation [SD] = 2.19), we also observed a consistent inverse association between DASH diet and number of SCCs (Table 3). One SD difference in the DASH score was associated with a 0.083 lower number of SCCs (P‐value = 0.020). In the evaluation of each DASH component separately, higher consumption of fruits including fruit juices (P‐value = 0.049), higher consumption of vegetables excluding potatoes (P‐value = 0.013), and higher consumption of legumes and nuts (P‐value = 0.037) was associated with lower number of SCCs. Specifically, one SD difference in total fruit consumption, vegetables except potatoes, and legumes and nuts was associated with 0.076, 0.089, and 0.076 lower number of SCCs, respectively. However, after adjusting for all other components of DASH score simultaneously, the association only remained for total fruit intake (P‐value = 0.049). In sensitivity analyses considering data on the first SCC assessment only, the association was similar, from both the logistic regression with dichotomized SCCs and the linear regression with continuous SCCs (Table S4 in supporting information). For instance, the highest quartile of DASH score at baseline was associated with a 17% reduction in the odds of having ≥2 SCCs, compared to the bottom quartile (OR = 0.83; 95% CI: 0.68–1.00; P‐trend = 0.016; model 2). One SD difference in the DASH score was associated with a 0.045 lower number of SCCs (P‐value = 0.020). There were no systematic differences in the distribution of demographic and health‐related characteristics in the study population when we considered the first assessment, the first and/or second, or both assessments (Table S5 in supporting information).

TABLE 3.

Association of DASH and its components (quartiles) with sum of two SCCs scores, from linear regression (n = 3690).

Linear regression a	Linear regression for each component only			Linear regression adjusted for all other components simultaneously
Variables	β	Standard error	P‐value	β	Standard error	P‐value
DASH	–0.083	0.036	0.020	—	—	—
Red and processed meat	0.049	0.037	0.182	0.008	0.040	0.840
Sodium	–0.075	0.090	0.409	–0.136	0.120	0.257
Sweets	0.041	0.038	0.283	–0.019	0.044	0.657
Fruits total	–0.076	0.039	0.049	–0.095	0.048	0.049
Vegetables without potatoes	–0.089	0.036	0.013	–0.050	0.040	0.207
Legumes and nuts	–0.076	0.036	0.037	–0.052	0.040	0.187
Low‐fat dairy	–0.067	0.039	0.082	–0.078	0.041	0.060
Whole grains	–0.047	0.037	0.209	–0.028	0.038	0.464

Abbreviations: BMI, body mass index; DASH, Dietary Approaches to Stop Hypertension; SCCs, subjective cognitive complaints.

^a Linear regression of DASH and its components in association with SCCs among women who completed both SCCs assessments in 2018 and 2020 (n = 3690). Dependent variable is sum of the SCCs score from the two SCCs surveys; independent variable is per standard deviation difference in the DASH score, adjusting for age (continuous; at the time of SCCs), menopausal status (premenopausal or postmenopausal; baseline), race (White or other; baseline), smoking status (never or ever; baseline), total caloric intake (continuous; baseline), education (< college, college, or graduate school; baseline), BMI (< 23, 23–24.9, 25–29.9, or ≥ 30 kg/m2; baseline), history of cancer (yes or no; follow‐up), and depression medication (yes or no; follow‐up).

We observed a similar pattern of inverse association between DASH diet and SCCs with multiple imputations and IPW (Table 4). In analyses imputing missing values, while OR estimates for quartiles differed somewhat from estimates obtained in analyses in which a category was created for missing values, the OR corresponding to a one quartile increase was the same in both analyses (OR = 0.93, 95% CI: 0.87–0.99). In analyses using IPW to account for potential selection bias, the association became more statistically significant (P‐trend < 0.001). The OR corresponding to one SD increase was 0.91 (95% CI: 0.88–0.95).

TABLE 4.

Association between DASH adherence and SCCs (< 2 vs. ≥ 2) by multiple imputation and inverse probability weighting.

	Quantile of DASH scores					OR (95% CI) per quartile
Variables	Q1	Q2	Q3	Q4	P‐trend
Model a
Conventional logistic regression b	1	0.91 (0.77–1.07)	0.79 (0.66–0.95)	0.83 (0.70–0.99)	0.019	0.93 (0.87–0.99)
Multiple imputation model c	1	1.03 (0.94–1.14)	0.90 (0.81–1.01)	0.94 (0.84–1.05)	0.013	0.93 (0.87–0.99)
IPW model d	1	0.95 (0.86–1.06)	0.85 (0.76–0.95)	0.77 (0.69–0.86)	<0.001	0.91 (0.88–0.95)
MI + IPW model	1	1.08 (1.01–1.14)	0.94 (0.88–1.02)	0.87 (0.81–0.92)	<0.001	0.91 (0.87–0.94)

Abbreviations: BMI, body mass index; CI, confidence interval; DASH, Dietary Approaches to Stop Hypertension; IPW, inverse probability weighting; MI, multiple imputation; OR, odds ratio; SCCs, subjective cognitive complaints.

^a Model adjusted for age (continuous; at the time of SCCs), menopausal status (premenopausal or postmenopausal; baseline), race (White or other; baseline), smoking status (never or ever; baseline), total caloric intake (continuous; baseline), education (< college, college, or graduate school; baseline), BMI (< 23, 23–24.9, 25–29.9, or ≥30 kg/m2; baseline), history of cancer (yes or no; follow‐up), and depression medication (yes or no; follow‐up).

^b Missing values in conventional logistic regression model were incorporated as dummy variables.

^c Multiple imputation model predicted missing values using Monte Carlo multiple imputation method.

^d Inverse‐probability weighting model calculated weights using baseline variables, including age, menopausal status, race, smoking status, education, and BMI.

4. DISCUSSION

In this study of women from the NYUWHS, greater adherence to the DASH diet in mid‐life was associated with ≈ 20% lower odds of having ≥2 SCCs in later life, after adjusting for potential confounders. The inverse association between DASH scores and having ≥2 SCCs was stronger among Black women and among those without a history of cancer. The association remained similar with different methods used to handle missing data and after adjusting for potential selection bias.

Previous longitudinal studies generated inconsistent findings on the association between adherence to the DASH diet and the risk of objective cognitive decline or AD. ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ For instance, in a US study of 6425 postmenopausal women aged 65 to 79 years, the highest quartile of DASH diet adherence was associated with a 28% lower risk of MCI after 9.11 years of follow‐up. ⁵⁸ In a US study of 3580 men and women ≥ 65 years of age, the highest quartile of DASH diet scores was associated with better average cognition, but not with cognitive decline after 10.6 years of follow‐up. ⁵⁵ However, in three other studies from the United States (n = 4169, mean age 60.4 years), ⁵⁹ Sweden (n = 2223, aged ≥60), ⁶⁰ and Australia (n = 1037, aged 70–90 years), ⁶¹ no association was found after 9.5, 6, and 6 years of follow‐up, respectively. Of note, participants in all the aforementioned studies on the DASH diet were in old age at baseline.

As AD has a long preclinical stage (up to decades), the extent to which factors assessed in late life, and/or shortly before the onset of clinical symptoms, could be causally related to AD or a result of pathological changes, is a subject of intense debate. ⁶² There has been only one other study that explored the association between mid‐life adherence to the DASH diet and SCCs in later life. ²³ The NHS, including 49,493 women with a mean age of 48 years at baseline, used repeated measures of diet collected between 1984 and 2006 and the mean total SCC score assessed in 2012 and 2014. Compared to the NYUWHS, the NHS included an additional complaint (difficulty remembering things from 1 second to the next), a complaint associated with normal aging that was not related to any objectively assessed cognitive impairment. ⁴⁷ Classifying the SCC score (ranging from 0 to 7) into none (0 point), moderate (0.5–2.5 points), and severe (3–7 points), the NHS reported that the odds of moderate and severe subjective cognitive decline were 24% and 39% lower for participants in the top compared to the bottom quintile of DASH score, respectively. Our results are consistent with those of the NHS. In addition, the NYUWHS population is more diverse than the NHS, as 18.6% of the NYUWHS participants are non‐White, while 97.9% of the NHS are White. We observed that the association between DASH scores and SCCs was stronger among Black women. Studies that conducted stratified analyses by race were limited. One study reported a stronger protective effect of the Mediterranean diet on cognitive outcomes in Black, compared to White, study participants. ⁶³ , ⁶⁴ Future studies are thus warranted to investigate susceptibility related to cultural and social differences, including dietary patterns, among different ethnic groups. ⁶³ Also, participants in the NYUWHS were younger at diet assessment in our study (mean age 46 years; 95.9% < 60 years) with a long follow‐up (> 30 years), compared to the NHS. Our data thus suggest that mid‐life may be a window of opportunity for lifestyle modifications to improve later life cognitive function. We also observed a stronger inverse association of DASH scores with the odds of having ≥ 2 SCCs among those without a history of cancer. Cancer itself or cancer treatments may have masked the inverse association between DASH diet and SCCs among women with a history of cancer. Future studies with detailed treatment and post‐cancer dietary information are needed to compare the relationship between diet and cognitive function in cancer survivors and cancer‐free individuals. ⁶⁵

There are several putative mechanisms by which the DASH diet could impact cognitive function. Many essential nutrients and bioactive substances that are abundant in vegetables, fruits, legumes, and nuts have anti‐inflammatory and antioxidant properties and are hypothesized to reduce brain oxidative stress, promote neurogenesis, and improve neuronal connectivity. ⁶⁶ , ⁶⁷ , ⁶⁸ , ⁶⁹ The association may also be explained by the fact that such a diet reduces hypertension, a risk factor of cognitive decline, ⁷⁰ a healthy eating regimen may have an impact on this risk factor. Additionally, by keeping the consumption of red or processed meat and sweets low in the DASH diet, the deleterious effect of high fat/sugar on brain inflammation and the production of amyloid beta protein may be minimized. ⁷¹ Finally, perturbation or improvement of the gut microbiome by diet may also be a pathway, ⁴³ , ⁷² , ⁷³ with support from a prior NYUWHS study that found enrichment of pro‐inflammatory gut microbes in women reporting more SCCs. ⁴³ Our findings regarding individual dietary components in the DASH diet and SCC risk support previous research observations, albeit limited, that documented associations of higher intake of fruits and fruit juices, vegetables, and nuts and legumes with better subjective cognitive function. ¹⁸ , ¹⁹ , ²³ We also found a 23% higher odds of having ≥2 SCCs among women in the top quartile of sweets intake versus the bottom quartile, consistent with the NHS, which reported lower consumption of sweets is associated with a better verbal memory in later life. ⁷⁴

Strengths of this study include its large sample size; diverse study population; the prospective study design; the inclusion of middle‐aged women; the long‐term follow‐up with dietary exposure measured > 30 years prior to SCC assessment, during mid‐life; and the availability of detailed longitudinal data on lifestyle and comorbidities. However, several limitations should also be noted. First, dietary intake was self‐reported, and we only assessed dietary intake at baseline in the NYUWHS. However, the questionnaire has been shown to have adequate temporal reproducibility. ³¹ , ⁷⁵ , ⁷⁶ The time difference between the two SCC assessments was relatively short, which limits our ability to measure reliably longitudinal change in SCCs, which would be more indicative of an underlying dementing process. Second, we cannot exclude some potential selection bias due to death, non‐response, and loss to follow up. Because individuals who did not complete the follow‐up questionnaires that included questions on SCCs might be more likely to have died or have had health conditions, including dementia or cognitive difficulties, which are related to poor diet, the resulting differential bias would be expected to be toward the null. The observation of a stronger inverse association between DASH score and SCCs in an analysis adjusting for possible selection bias is consistent with such a toward‐the‐null bias. Third, there may be certain unmeasured confounders, such as other lifestyle factors and behaviors (e.g., social/cognitive engagement, sleep, alcohol/substance use, and physical activity), that were not accounted for in the study. These unmeasured factors could account in part for the observed associations. In addition, there might be potentially different drivers for subjective reporting of cognitive issues compared to drivers of the underlying cognitive impairment. For instance, personality traits, particularly higher neuroticism and lower conscientiousness, have been associated with SCCs ⁷⁷ , ⁷⁸ and unhealthy diet habits such as low consumption of fruit and vegetables, and the increased consumption of sugar and saturated fats. ⁷⁹ , ⁸⁰ We adjusted for history of depression in multivariable models to partially account for the potential influence of personality traits on self‐report of SCCs, given the accumulating evidence associating personality with depression. ⁸¹ , ⁸² Growing evidence also indicates an association between personality traits and AD/dementia ⁸³ , ⁸⁴ , ⁸⁵ as well as amyloid and tau neuropathology. ⁸⁶ Future studies are needed to elucidate the interplay among personality traits, diet, and risk of cognitive impairment. We also do not have information on family history of dementia or other neurodegenerative diseases, and therefore we were not able to assess if the associations differ by family history. Finally, while our focus was on subjective cognitive function using validated instruments, which is predictive of objective measures of memory loss, ⁴⁴ we did not evaluate if participants had MCI or dementia at the time of SCC assessment to exclude women with these conditions. However, almost all participants included in the present study filled out the questionnaires on their own, and the completeness of the questionnaires was checked. Family members or close relatives of the participants who were unable to fill out questionnaires due to cognitive issues or other conditions were asked to fill out an abbreviated questionnaire that did not include questions on SCCs.

5. CONCLUSION

We found that a higher level of adherence to the DASH diet in mid‐life was associated with lower SCCs later in life among women. These findings suggest that improvements in diet quality during mid‐life, especially the diet related to hypertension and cardiovascular profile, may have a role in maintaining an optimal subjective cognitive function among women. Future studies with objective measures of cognitive function and multiple racial/ethnic groups are needed to confirm the associations and evaluate the generalizability of the findings.

CONFLICT OF INTEREST STATEMENT

The authors disclose no conflicts. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All human subjects provided informed consent.

Supporting information

Supporting Information

Supporting Information

Song Y, Wu F, Sharma S, et al. Mid‐life adherence to the Dietary Approaches to Stop Hypertension (DASH) diet and late‐life subjective cognitive complaints in women. Alzheimer's Dement. 2024;20:1076–1088. 10.1002/alz.13468