Comparing Indices of Diet Quality With Chronic Disease Mortality Risk in Postmenopausal Women in the Women's Health Initiative Observational Study: Evidence to Inform National Dietary Guidance

Abstract

Cardiovascular disease (CVD) and cancer are the leading causes of death among women in the United States (1), and mortality rates from these diseases are increasing (2). Poor diet quality is thought to be a leading risk factor for years of life lost and years lived with chronic disability (2). To date, most dietary guidance to promote health is based largely on data about single foods and nutrients, but foods are not consumed in isolation. There has been growing interest in studying overall diet quality, an approach that takes into account the complexity of the diet and the potentially synergistic or antagonistic effects of all individual dietary components (3). Diet quality indices are increasingly being used in epidemiologic studies (3). However, the methods for measuring diet quality and analyzing its relationship with the risk of death vary across studies, and this has hampered the formulation of public health recommendations with regard to dietary patterns. The 2010 Dietary Guidelines for Americans dedicated an entire chapter to discussing the health benefits of eating a high-quality diet, but indicated that there was not yet enough evidence to provide a definitive recommendation related to diet quality and health outcomes (4).

The assessment of associations between a priori diet quality indices and health outcomes may inform policy with respect to dietary guidelines, because these measures provide a common metric for comparing findings across populations. In response to this critical need, the National Cancer Institute (Rockville, Maryland) initiated the Dietary Patterns Methods Project (DPMP) and formed a collaboration among investigators from 3 US cohorts—the Women's Health Initiative Observational Study (WHI OS), the National Institutes of Health–AARP Diet and Health Study, and the Multiethnic Cohort Study. The goal of the DPMP is to apply consistent, standardized methodologies and to systematically examine the associations of common diet quality indices and mortality risk. The purpose of this report is to present findings from the WHI OS.

The WHI OS presents a unique opportunity to study these relationships among postmenopausal women and inform the 2015 Dietary Guidelines for Americans for this population. Among 63,805 postmenopausal women, we examined how scores on 4 key commonly used a priori diet quality indices—the Healthy Eating Index 2010 (HEI) (5–7), the Alternative Healthy Eating Index 2010 (AHEI) (8, 9), the Alternate Mediterranean Diet (aMED) (10), and the Dietary Approaches to Stop Hypertension (DASH) (11)—are related to risks of death from any cause, from CVD, and from cancer. In all of these indices, higher scores correspond with a better-quality diet.

METHODS

The WHI has been previously described in depth (12–14). Briefly, between 1993 and 1998, through 40 clinical centers throughout the United States, postmenopausal women who were 50–79 years of age at study entry were recruited into either a clinical trials component (n = 68,132) or the WHI OS (n = 93,676 women). The clinical trials component and observational study conducted closeout activities in 2004–2005, and the participants were invited to continue being followed in the 2005–2010 WHI Extension Study through September 30, 2010. Written informed consent was obtained from all study participants. Procedures and protocols were approved by institutional review boards at all participating institutions. A standardized written protocol, centralized training of staff, and quality assurance visits by the clinical coordinating center were used to ensure uniformity of data collection.

The present sample was drawn from the 93,676 women participating in the WHI OS. Of these, we excluded those with incomplete diet data (n = 96), implausible energy intakes of less than 600 kcals/day or more than 5,000 kcals/day (n = 3,570), prior diagnosis of CVD or cancer (25,794), or missing information on diabetes status (n = 411), because the absence of diabetes information was highly associated with mortality risk in the WHI. Our sample for analysis was 63,805 women.

At enrollment, participants reported demographic characteristics, health behaviors, and medical histories using self-administered standardized questionnaires. We categorized risk factors as follows: race/ethnicity (white, black, Hispanic, other, or missing); educational level (high school or below, some college, college, postgraduate, or missing); marital status (married or living as married, single/widowed/divorced, or missing), smoking status (never, past, or current); diabetes (no or yes); and hypertension (no, yes, or missing) or high cholesterol that required pills (no, yes, or missing). In WHI's Measurement Precision Study, questionnaire items on demographic characteristics and medical conditions were shown to be reliable (weighted κ > 0.8) (13). Self-reported physical activity was measured using the WHI brief physical activity inventory, which has been shown to be reliable (weighted κ ranging from 0.67 to 0.71) (13) and valid when compared with accelerometer data (r = 0.73) (13). For each participant, we calculated metabolic equivalent (MET)-hours per week of recreational physical activity and categorized physical activity level (0, 0.1–3, 3.1–8.9, or ≥9 MET-hours/week, or missing), as described in detail previously in the WHI (15). The use of postmenopausal hormone therapies (unopposed estrogen and/or estrogen plus progesterone) via pills or patches was self-reported, and we classified women as never, past, or current users.

At the clinic visit, trained staff measured each participant's weight and height using a standardized protocol. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared, and we categorized BMI as less than 18.5, 18.5–24.9, 25.0–29.9, 30.0–34.9, or 35.0 or greater. Also at the clinic visit, trained staff measured each participant's waist circumference during expiration at the narrowest section of the torso (14).

RESULTS

Across diet quality indices, compared with women with poor-quality diets (quintile 1), women with better-quality diets (quintile 5) were older, had lower BMI values, engaged in more physical activity, and were more likely to be college educated, non-Hispanic white, and current users of postmenopausal hormone replacement therapy; they were also less likely to have diabetes (Tables 2 and ​and3).3). Women who were lost to follow-up during the study (n = 655) had slightly lower diet quality scores than women with complete follow-up (data not shown). Univariate correlations between the 4 diet quality indices were fairly strong, ranging from 0.55 to 0.70 (all P < 0.0001), with the strongest correlation between HEI and DASH (data not shown).

Over a median 12.9 years of follow-up, 5,692 deaths occurred, including 1,483 from CVD and 2,384 from cancer. As shown in Table 4, in multivariate models, across indices, having a better-quality diet was associated with lower risks of death from all causes and CVD (all P ≤ 0.017). A lower risk of cancer death was observed for women with higher HEI, aMED, and DASH scores (all P ≤ 0.001), but not for those with higher AHEI scores (P = 0.275). Additional adjustment for hypertension and high cholesterol did not result in changes to the hazard ratios (data not shown). Models without BMI as a covariate yielded similar results (data not shown). As shown in Table 5, among women with BMI values of less than 25.0 or between 25.0 and 29.9, higher scores on all indices were associated with a lower risk of all-cause mortality (all P ≤ 0.022), but among those with BMI values of 30 or more, only HEI was associated with all-cause mortality risk (P = 0.023).

Among women with BMI values less than 25.0, higher HEI and DASH scores were associated with a lower risk of CVD death (all P ≤ 0.038). Among women with BMI values between 25.0 and 29.9, higher HEI, AHEI, and aMED scores were associated with a lower risk of CVD death (all P ≤ 0.047). No scores were associated with CVD mortality risk among women with BMI values of 30.0 or higher.

Among women with BMI values less than 25.0 or between 25.0 and 29.9, higher HEI, aMED, and DASH scores, but not AHEI scores, were associated with a lower risk of cancer death (all P ≤ 0.032) (Table 5). No scores were associated with cancer mortality risk among women with BMI values of 30 or higher.

As shown in Table 6, among women with a waist circumference of 88 cm or less, higher HEI, aMED, and DASH scores were associated with lower risks of all-cause, CVD-specific, and cancer-specific mortality; higher AHEI score was associated with a lower risk of all-cause mortality but not CVD or cancer death. Among women with a waist circumference of more than 88 cm, all scores were associated with a lower risk of all-cause mortality, but no scores were associated with CVD death, and only HEI score was associated with a lower risk of cancer death (P = 0.011). A total of 64% of women with a waist circumference greater than 88 cm had BMI values of 30.0 or more, and 88% of women with BMI values of 30.0 or more had a waist circumference greater than 88 cm (data not shown). Statistical interactions by BMI were observed for AHEI and DASH for all-cause and cancer death (all P ≤ 0.04) and by waist circumference for aMED and all-cause mortality risk (P = 0.009).

DISCUSSION

This prospective study of approximately 64,000 postmenopausal women provided evidence that high scores on key diet quality indices were associated with an 18%–26% lower risk of death overall and from 2 of the most common diet-related chronic diseases—CVD and cancer. Our study is consistent with past research that has documented inverse associations between single a priori diet quality indices and all-cause mortality risk and CVD death (Mediterranean diet (29–31); HEI (32); and AHEI (33, 34)). In our study, AHEI score was not associated with cancer death, which is consistent with the results of 1 past study (33) but contrasts with those of another (34). The AHEI was designed to measure adherence to a high-quality diet predictive of major chronic disease risk (9); some of its components may be somewhat specific to CVD risk, which may explain its lack of robustness for predicting cancer mortality risk in our study. Our results are also consistent with all-cause mortality findings among populations with CVD (AHEI among 2,258 myocardial infarction patients (35); DASH among 3,215 postmenopausal heart failure patients (36)) and populations with cancer (HEI among 670 breast cancer survivors (22) and 2,317 breast cancer survivors (37); AHEI and DASH among 4,103 breast cancer survivors (38)).

This past literature is difficult to summarize because the indices and analytical methods varied from study to study. Most importantly, our findings build on recent results among older US adults reported by our DPMP collaborators (39); and, in tandem, the results from the DPMP show the robustness of these diet quality measures in predicting mortality risk in different populations. Because of its size, comprehensiveness, and standardized methodology, our study can inform those developing dietary guidelines for women of postmenopausal age of the potential survival benefit associated with consuming a diet in line with the indices explored in this paper.

To our knowledge, our study is the first to address the potential heterogeneity of the diet quality–mortality risk relationship by obesity status. We found that diet quality seems to be a weaker and less significant predictor of death at high levels of BMI. There are several potential explanations for this finding. First, it may be due to exposure misclassification and differential measurement error; having a high BMI is associated with underreporting of food intakes, and the foods most underreported are those that dietary guidelines advise populations to consume in moderation (e.g., high-fat, high-sugar, and high-calorie/low–nutrient-dense foods) (40). Alternatively, it is possible that once a person has a high BMI value, many aspects of beneficial diets that are captured in these indices are less likely to be inversely associated with mortality risk unless individuals also lose weight, because of the physiological impact of obesity. However, even if diet quality does not have a direct survival benefit for obese women, consuming a high-quality diet has been shown to be important for weight loss and maintenance in overweight and obese adults (41), and lower BMI values at midlife have been associated with a lower risk of death (42).

Among women with a waist circumference greater than 88 cm, associations between indices and CVD and cancer mortality risk were null, with the exception of HEI being inversely associated with cancer death. This finding seems to mirror the null cause-specific findings observed in this study among women with BMI values of 30.0 or more. However, for all-cause mortality, our conclusions about diet quality's predictive value across obesity levels differed on the basis of whether we defined obesity using BMI (overall obesity) or waist circumference (abdominal obesity). Although the directions of association were the same, the magnitude of association was stronger and more likely to be statistically significant among those with a high waist circumference than among those with a high BMI value. One possible reason is that waist circumference may be a more direct measure of body composition because a higher waist circumference indicates a greater extent of abdominal obesity, which is more tightly linked to metabolic dysfunction than BMI alone. Abdominal obesity is known to be associated with abnormal production of inflammatory and metabolic cytokines that are responsive to dietary manipulation (43). However, considering that only some of the interaction terms for BMI and waist circumference were significant, and that interpretation of the interaction terms is complex, replication of these findings in future research is needed.

Overall, the diet quality scores that we examined in this paper showed similar relationships with mortality risk, and they share some commonalities. All scores emphasize intakes of fruit, vegetables, whole grains, and plants or plant-based proteins. All scores except DASH emphasize consumption of polyunsaturated and monounsaturated fats over saturated fats. All except aMED include a sodium component because of its importance in lowering blood pressure and preventing coronary heart disease and stroke (44). All indices except HEI stress nut consumption. aMED and AHEI emphasize moderate consumption of alcohol because of its association with lower CVD risk (45), as well as low intake of red and processed meat, which is a risk factor for certain cancers (46–48). HEI and aMED include a seafood component. The comprehensive results presented here show that, although these indices were developed for slightly different purposes, include some different components, and vary in their definitions of optimal diet quality and scoring, they are also all capturing the essential elements of a healthy diet. This was evidenced by the fairly strong correlations between scores and the strikingly consistent magnitudes of hazard ratios when comparing scores in the highest and lowest quintiles across the indices, except for AHEI and cancer death.

Advantages of this study include the use of the multidimensional diet quality indices that capture the potentially synergistic nature of multiple important dietary components (49) and permit comparisons among study populations. Further strengths include objectively measured height, weight, and waist circumference, large sample size, the prospective nature of evaluations, long-term mortality follow-up, and central adjudication of deaths. Additionally, the WHI OS had high-quality data on covariates, allowing us to make simultaneous adjustments for factors known to alter mortality risk and to conduct a thorough analysis.

Study limitations include measurement error inherent to the FFQ (50) and to other self-reported measures of health behaviors, like physical activity (51). Measurement error is most keenly recognized for energy (52), but less is known about the extent and severity of measurement error for many of the dietary measures presented in this report. Because there are no quantitative/recovery biomarkers for these dietary components, it is not possible to infer that the same type of measurement error exists for these components as it does for energy, and future research is needed in this area. Additionally, although we had detailed data allowing us to carefully control for the major confounders and to show that associations were unlikely to be artifacts of reverse causation, given the observational nature of this study, it remains possible that those who chose a better-quality diet lived longer for reasons that we did not examine.

Overall, within this cohort of postmenopausal women, associations of diet quality and mortality outcomes were consistent, of similar magnitudes, and highly significant, regardless of the index of choice, with the exception of AHEI and death from cancer. Most importantly, this large study of nearly 64,000 women and 6,000 deaths provides needed data on diet quality and health—an important public health topic currently being considered in examining the scientific basis for the 2015 Dietary Guidelines for Americans. Our study suggests that postmenopausal women who consume a diet consistent with a priori–specified diet quality indices such as HEI, AHEI, aMED, and DASH may have a lower risk of death from chronic diseases.

ACKNOWLEDGMENTS

REFERENCES