Changes in Diet Quality and Mortality
By Kevin C Maki, PhD
Sotos-Prieto and colleagues from the Departments of Nutrition and Epidemiology at Harvard University published an analysis of the relationships between changes in diet quality scores and mortality [all-cause, cardiovascular (CV), and cancer] in a recent issue of the New England Journal of Medicine.1 These relationships were examined in two large cohorts of health professionals, the US Nurses’ Health Study (women) and the US Health Professionals Follow-up Study (men).
Previous studies have shown that higher diet quality scores are associated with lower mortality. Of course, the potential problem with observational evidence is that it is difficult to know whether it is higher diet quality per se that is responsible for the relationship to lower mortality, or other differences between those with lower and higher diet quality scores. People who consume diets that conform to recommendations from health authorities often have other characteristics that might contribute to better outcomes. They tend to be more health conscious, have higher educational attainment, exercise more regularly, have lower body mass index, and be less likely to smoke, use illicit substances and consume alcohol to excess, just to name a few of the differences.
Epidemiologists attempt to adjust for these differences using statistical techniques to reduce the potential for bias and confounding by differences other than the exposure of interest; in this case diet quality score. It is often impossible to identify and adjust for all of the variables that may be relevant. The ultimate goal is to identify exposures that are causally related to disease status and modifiable, which can serve as the basis for public health actions, including recommendations for the characteristics of a healthy diet.
While far from perfect, it is useful to investigate the relationships between changes in exposure variables over time and disease risk. For example, it bolstered the case for a causal relationship between cigarette smoking and risks for lung cancer and heart disease when it was shown that risks declined over time in those who quit smoking, but did not in those who continued to smoke. Similarly, the Harvard group sought to assess whether changes in diet quality scores over time predicted risks for all-cause and cause-specific mortality.
In order to address this question, the researchers calculated changes over time in three scores: the Alternative Healthy Eating Index (aHEI), the Alternative Mediterranean Diet (aMED) score, and the Dietary Approaches to Stop Hypertension (DASH) score. Scores were calculated based on responses to the Willett Food Frequency Questionnaire that was administered every 4 years over long periods in both cohorts. For the main analysis, changes in diet quality were calculated from 1986 to 1998 and follow-up was through 2010 (12 years). The analysis included data from 73,739 participants (65% women).
The aHEI used scores from 0-10 for 11 food components selected on the basis of their relationships with chronic diseases. Thus, the score could be from 0 to 110. The aMED score included 9 components and were scored as 0 or 1 according to whether intake was above or below the cohort-specific median, allowing a score from 0 to 9. The DASH score included 8 components, each scored 1-5, thus producing scores ranging from 8 to 40. For all, a higher score indicated higher diet quality.
Participants who increased their diet quality scores reported increased intakes of whole grains, vegetables and omega-3 fatty acids, as well as reduced intakes of sodium. Compared to participants with relatively stable scores (middle quintile for change), those in the top 20% for increases (roughly 15-16 points for aHEI, 2-3 points for aMED, and 5-6 points for DASH), had 9-14% lower mortality risk, all p < 0.05 for the fifth vs. the third quintile in multivariable-adjusted models. Similar results were obtained when a 20-percentile increase in score was modeled (8-17% lower risk for mortality).
Results were somewhat less consistent for deaths from CV causes and cancer. A 20-percentile increase in aHEI was associated with 15% lower CV mortality (p < 0.05), while the same increase in aMED was associated with a 7% lower CV mortality (p < 0.05), and a 20-percentile increase in DASH score was associated with a non-significant 4% reduction in mortality. For cancer mortality, a 20-percentile increase was associated with 6-9% lower mortality, which was only statistically significant (p < 0.05) for the DASH score (9%).
Having constant high diet quality score was associated with reductions of 9-14% in all-cause mortality over 12 years compared to those with constant low scores. The investigators also looked at 8-year and 16-year changes. In general, the effect became more pronounced with longer periods. This was particularly evident for the aHEI, although that may be due to the greater range of values possible, which increases the variation in the populations studied.
This study adds support for the healthy eating patterns recommended in the Dietary Guidelines for Americans (2015-2020).2 Although confounding by unmeasured, or crudely measured, factors cannot be ruled out, the available data are consistent with a causal association between a healthy diet pattern and reduced risks for all-cause and CV mortality. The evidence for a reduction in cancer mortality with a healthy dietary pattern is less convincing. These results particularly support the recommendations for increased consumption of whole grains, fruits, vegetables, and fish/omega-3 fatty acids, compared with the average American diet, since these were the foods that were primarily responsible for changes in diet quality scores over time.
As I have stated repeatedly, public policy recommendations regarding diet often have to be based on evidence from observational studies assessing disease risk and intervention studies of biomarkers for disease risk, because few randomized, controlled dietary intervention studies have been completed to assess effects on disease incidence. The strongest recommendations should be reserved for those areas where we have alignment between results from all three types of studies.
The results from the Prevención con Dieta Mediterránea (PREDIMED) study support health benefits, including reduced incidence of CV events (particularly stroke) and diabetes associated with advice to consume a Mediterranean diet pattern supplemented with nuts or olive oil, compared to low-fat diet advice.3 PREDIMED was not a perfect trial. More randomized, controlled dietary intervention trials with outcomes of disease incidence are badly needed to answer questions about risks and benefits of various types of dietary advice. PREDIMED demonstrates the feasibility of completing such studies.
For now, despite a number of caveats and uncertainties, the best available evidence suggests that the dietary patterns recommended in the Dietary Guidelines for Americans (healthy US diet, Mediterranean diet, DASH diet) are associated with a variety of favorable outcomes, including reduced total and CV mortality.
- Sotos-Prieto M, Bhupathiraju SN, Mattei J, et al. Association of changes in diet quality with total and cause-specific mortality. N Engl J Med. 2017;377:143-153.
- US Department of Health and Human Services and US Department of Agriculture. Dietary Guidelines for Americans 2015-2020. 8th December 2016. Available at https://health.gov/dietaryguidelines/2015/resources/2015-2020_Dietary_Guidelines.pdf.
- Estruch R, Martínez-González MA, Corella D, et al.; PREDIMED Study Investigators. Effects of a Mediterranean-style diet on cardiovascular risk factors: a randomized trial. Ann Intern Med. 2006;145:1-11.