Cancer and CVD are the two leading causes of death worldwide(1). Diet and lifestyle are associated with these and other diseases(Reference Behrens, Fischer and Kohler2–Reference Petersen, Johnsen and Olsen5). Increased attention is being placed on dietary patterns rather than individual food components in epidemiological investigations of associations of diet with mortality(Reference McCullough6). Various diet patterns, which contain multiple constituents acting and interacting along the same and different pathways, have been inversely associated with chronic disease risk and mortality(Reference Jankovic, Geelen and Streppel7–Reference Zazpe, Sanchez-Tainta and Toledo17), supporting further research into dietary patterns in chronic disease prevention(Reference Hu18). Physical activity, excess adiposity, smoking and other modifiable lifestyle factors have been associated with higher mortality(Reference Ding, Rogers and van der Ploeg3, Reference Gellert, Schottker and Brenner19–Reference Nocon, Hiemann and Muller-Riemenschneider25). Lifestyle factors coexist and may interact, and underlying causes of death may be multi-causal(Reference Petersen, Johnsen and Olsen5), suggesting that investigating associations of various lifestyle factors, in combination, with mortality may help with developing public health recommendations(Reference Ding, Rogers and van der Ploeg3, Reference Loef and Walach4, Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26–Reference Yun, Won and Kimm29). Recently, we developed a ‘Paleolithic diet’ pattern score and a modified Mediterranean diet pattern score and found both diet pattern scores to be strongly inversely associated with biomarkers of oxidative stress and inflammation(Reference Whalen, McCullough and Flanders30), colorectal adenoma(Reference Whalen, McCullough and Flanders31) and all-cause, all-CVD and all-cancer mortality(Reference Whalen, Judd and McCullough32).
The Paleolithic diet pattern was developed to address the increase in ‘diseases of civilisation’ (cancer, CVD, etc.) as being the possible consequence of evolutionary discordance between the general diets and lifestyles of Homo sapiens living in the range of environments of evolutionary adaptedness before the agricultural revolution and those during the modern era(Reference Konner and Eaton33). The Paleolithic diet pattern, which was estimated from anthropological studies of fossils and extant hunter–gatherer societies, is characterised as rich in a marked diversity of fruits and vegetables, lean meats, eggs and nuts; excluding grains, dairy products and refined fats and sugar; and very low in salt(Reference Eaton and Konner34). Furthermore, lifestyle patterns that are more ‘Paleolithic-like’ include high levels of physical activity, energy balances resulting in lean body masses, and no tobacco use(Reference Eaton and Konner34, Reference Eaton, Konner and Shostak35). Given the constraints in investigating Paleolithic diet and lifestyle patterns in the modern context (e.g. limited wild food intakes, food preparation methods, types of physical activity), herein we use the alternative terms evolutionary-concordance diet and lifestyle patterns rather than ‘Paleolithic’ (our re-termed evolutionary-concordance diet pattern score is identical to our previously reported Paleolithic diet score).
The Mediterranean diet has been consistently reported to be beneficial for health and longevity(Reference Trichopoulou36, Reference Garcia-Fernandez, Rico-Cabanas and Rosgaard37), and is characterised by high intakes of fruits, cereals, nuts, vegetables, legumes and olive oil; moderate intakes of fish and poultry; low intakes of eggs, dairy products, red and processed meats and sweets; and moderate consumption of alcohol with meals(Reference Willett, Sacks and Trichopoulou38, Reference Trichopoulou, Costacou and Bamia39). Prospective cohort studies and recent meta-analyses found adherence to the Mediterranean diet pattern to be inversely associated with all-CVD and all-cause mortality(Reference Knoops, de Groot and Kromhout8, Reference McNaughton, Bates and Mishra9, Reference Mitrou, Kipnis and Thiebaut10, Reference Reedy, Krebs-Smith and Miller11, Reference Tognon, Lissner and Saebye13, Reference Tong, Wareham and Khaw15, Reference Zazpe, Sanchez-Tainta and Toledo17, Reference Dinu, Pagliai and Casini40) and inconsistently inversely associated with all-cancer mortality(Reference Knoops, de Groot and Kromhout8, Reference Mitrou, Kipnis and Thiebaut10, Reference Reedy, Krebs-Smith and Miller11, Reference Tognon, Nilsson and Lissner14, Reference Vormund, Braun and Rohrmann16, Reference Dinu, Pagliai and Casini40). Thus, the Mediterranean diet pattern can serve as a reference for comparisons of associations of a more evolutionary-concordant diet with mortality.
We previously reported similar inverse associations of evolutionary-concordance and Mediterranean diet pattern scores with all-cause, all-CVD and all-cancer mortality in a prospective cohort of white and black adults(Reference Whalen, Judd and McCullough32). However, there are no reported studies of an association of an evolutionary-concordance lifestyle pattern score, alone or combined with an evolutionary-concordance or a Mediterranean diet pattern score, with mortality. Therefore, we addressed this in the prospective Iowa Women’s Health Study (IWHS).
Methods
Study population and data collection
As described previously(Reference Folsom, Kaye and Prineas41), the IWHS, established in 1986, is a prospective cohort study of 41 836 55- to 69-year-old Iowa women. In addition to the original survey, follow-up surveys were mailed in 1987, 1989, 1992, 1997 and 2004. The University of Minnesota Institutional Review Board (IRB) approved the study, and the Emory University IRB also approved the present analysis.
At baseline, detailed information on demographics, self-measured anthropometrics, lifestyle, medical and family history, diet and other factors was collected. Dietary intakes and vitamin and mineral supplement use were collected at baseline using a 127-item semi-quantitative Willett FFQ, for which the validity and reliability in the study population were previously reported(Reference Munger, Folsom and Kushi42). Total energy and nutrient intakes were calculated by adding energy and nutrients from all food sources using the dietary database developed by Willett et al. (Reference Willett, Sampson and Browne43). Physical activity was assessed via two questions regarding the frequency of partaking in moderate and vigorous physical activities(Reference Kushi, Fee and Folsom44) and categorised as low, medium and high (see Table 2 footnote §). Diet and physical activities were only comprehensively reassessed in 2004 when only 68·3 % of the participants remained alive (therefore, only baseline exposure information was used in the present analyses).
Deaths were identified through the State Health Registry of Iowa and the National Death Index for those who did not respond to the follow-up questionnaires or had emigrated from Iowa. Underlying cause of death was assigned and coded by state vital registries according to the International Classification of Diseases (ICD). CVD mortality was defined using ICD-9 codes 390–459 and ICD-10 codes I00–I99, and cancer mortality was defined using ICD-9 codes 140–239 and ICD-10 codes C00-D48. Follow-up time was calculated as the time from the date of completing the baseline questionnaire to the date of death, the last follow-up contact, or the end of follow-up (31 December 2012), whichever was first(Reference Klein and Moeschberger45).
Scores
The evolutionary-concordance and Mediterranean diet pattern scores were constructed in a similar manner as described previously(Reference Whalen, McCullough and Flanders30–Reference Whalen, Judd and McCullough32). Based on the distribution of all study participants’ baseline intakes, each participant was assigned a quintile rank (a corresponding score from 1 to 5) of intake for each food category. Higher scores were given for higher intakes of foods considered characteristic of the diet pattern, and lower scores were given for lower-to-no consumption of foods considered not characteristic of the diet pattern (Table 1). For the evolutionary-concordance diet score, two unique variables were created. The first, a fruit and vegetable diversity score, was created by summing the total number of different fruits and vegetables that participants reported consuming >1–3 servings/month. Second, since the Paleolithic diet had little dairy food but high amounts of Ca (from wild plant foods)(Reference Eaton and Konner34), to consider dietary Ca separately from dairy products, we used the residuals of a linear regression of total Ca intake on total dairy products consumption. A modified Mediterranean diet score, the alternative Mediterranean diet score, originally developed as an adaptation to FFQ in the USA(Reference Fung, McCullough and Newby46), was calculated according to previous literature(Reference Fung, McCullough and Newby46, Reference Reedy, Mitrou and Krebs-Smith47). However, instead of basing it on dichotomising the component dietary intake categories (high v. low, based on median intake) as is most common, we used quintiles of intakes to facilitate a more direct comparison of the two diet scores. The components of the dietary scores were not weighted because (1) in our previously reported studies of dietary scores, weighting made no difference(Reference Dash, Bostick and Goodman48–Reference Cheng, Um and Prizment50) and (2) the components of Mediterranean diet scores traditionally are not weighted. The points for each of the food groups comprising diet scores were summed. Therefore, the final, possible score ranges were 14–70 for the fourteen-component evolutionary-concordance diet score and 9–45 for the nine-component Mediterranean diet score, with higher scores indicating higher concordance with a dietary pattern.
Table 1 Constituents and construction of evolutionary-concordance and Mediterranean diet pattern scores in the prospective Iowa Women’s Health Study (n 35 221), 1986–2012*
* All constituents measured in servings/week unless otherwise indicated. Highest intake ‘best’: number of points assigned to each quintile = quintile rank (e.g. highest and lowest quintiles scored 5 and 1 points, respectively); lowest intake ‘best’: number of points assigned to each quintile = reverse quintile rank (e.g. highest and lowest quintiles scored 1 and 5 points, respectively). Other: alcohol intake 5–15 g/d scored 5 points, outside of the range scored 1 point.
† Consumption of nitrate-processed meats and non-lean red meat combined.
‡ The evolutionary-concordance diet pattern score had fourteen components; range of possible scores, 14–70.
§ Lean meats included skinless chicken or turkey and lean beef.
|| Baked goods included items such as cake, pie and other pastry-type foods.
¶ Fruit and vegetable diversity calculated by summing the total number of different fruits and vegetables items in the FFQ the participants indicated that they ate more than 1–3 times per month.
** Intake of Ca independent of non-Ca components of dairy foods; calculated as residuals from the linear regression of total Ca intake (mg/d) on dairy-food intake.
†† The Mediterranean diet pattern score had nine components; range of possible scores, 9–45.
For our novel, weighted, evolutionary-concordance lifestyle score, we combined physical activity, BMI (weight (kg)/height (m)2) and smoking status (Table 2). First, because there were only three categories for each lifestyle variable (rather than five categories as for the dietary variables), to put the lifestyle variables on the same initial scale as the dietary variables, each component was assigned a preliminary score of 1, 3 or 5, for, respectively, low/medium/high physical activity, BMI ≥30/25–<30/<25 kg/m2 and current/former/never smoking. Then, because the individual lifestyle factors are more strongly associated with all-cause, CVD and cancer mortality than are the individual dietary factors, the preliminary scores were weighted by dividing the two ‘most exposed’ category scores by summary relative risks from meta-analyses of associations of physical activity(Reference Li, Gu and Jing21, Reference Lollgen, Bockenhoff and Knapp22, Reference Nocon, Hiemann and Muller-Riemenschneider25), BMI(Reference McGee23) and smoking status(Reference Gellert, Schottker and Brenner19, Reference Jones, Tellez-Plaza and Navas-Acien20, Reference Mons, Muezzinler and Gellert24) with all-cause, CVD and cancer mortality. The summary relative risk values from these meta-analyses that were used for weighting the lifestyle score components are shown in Table 2. The final three-component lifestyle scores for all-cause, all-CVD and all-cancer mortality could range from 2·26 to 15·81, 2·10 to 17·69 and 2·34 to 17·25, respectively, with higher scores indicating a more evolutionary-concordant lifestyle.
Table 2 Constituents and construction of the evolutionary-concordance lifestyle pattern score in the prospective Iowa Women’s Health Study (n 35 221), 1986–2012
* Weights based on summary relative risks from reported meta-analyses of observational epidemiological studies of associations of physical activity(Reference Li, Gu and Jing21, Reference Lollgen, Bockenhoff and Knapp22, Reference Nocon, Hiemann and Muller-Riemenschneider25), smoking status(Reference Gellert, Schottker and Brenner19, Reference Jones, Tellez-Plaza and Navas-Acien20, Reference Mons, Muezzinler and Gellert24) and BMI(Reference McGee23) with all-cause, cardiovascular and cancer mortality; the initial points in the two most evolutionary concordant categories of physical activity (hypothesised to be inversely associated with risk) and the two least evolutionary concordant categories of smoking and BMI (hypothesised to be directly associated with risk) were divided by these values to yield the weighted values to be summed for the lifestyle score (e.g. for high physical activity and all-cause mortality: 5/0·69 = 7·25; for current smoker: 1/1·80 = 0·56).
† Least evolutionary concordant category.
‡ Most evolutionary concordant category.
§ Physical activity level derived from two questions regarding the frequency of moderate and vigorous physical activity(Reference Trichopoulou, Costacou and Bamia39), and categorised as high (vigorous activity twice a week or moderate activity >4 times/week), medium (vigorous activity once a week plus moderate activity once a week or moderate activity 2–4 times/week) and low.
|| Categories are for obese, overweight and normal/underweight, respectively, according to the WHO guidelines.
Statistical analysis
For our analyses, we excluded participants with a history of cancer (other than non-melanoma skin cancer) at baseline (n 3830) and those who left >10 % of the FFQ questions blank (n 2499) or had implausible total energy intakes (<2510 or >20920 kJ/d; n 286), leaving an analytic cohort of 35 221.
Participants’ characteristics, by score quintiles, were summarised and compared using the χ 2 test for categorical variables and ANOVA for continuous variables. Cox proportional hazards regression was used to calculate hazards ratios (HR) and 95 % CI to estimate the associations of the various scores with cause-specific and all-cause mortality. The scores were analysed as continuous and categorical variables (quintiles) based on the distributions of all participants’ scores at baseline. The median value of each diet score quintile was used for conducting all trend tests. Correlations between scores were assessed using Pearson correlation coefficients.
Based on previous literature and biological plausibility, the following variables were considered as potential confounders for the diet score models: age (years; continuous), smoking status (current, past, never smoker), education (<high school, high school, >high school), BMI (continuous), physical activity (low, medium, high), total energy intake (kJ/d; continuous), hormone replacement therapy (HRT) use (current, past, never), marital status (married, never married, widowed, divorced/separated) and chronic disease (yes/no). Participants with chronic disease were defined as those who had a self-reported history of diabetes, heart disease or cirrhosis. For the lifestyle score model, the evolutionary-concordance diet score was also considered as a covariate. Criteria for inclusion in the final models were biological plausibility and/or whether inclusion/exclusion of the variable from the model changed the adjusted HR for the primary exposure variable by ≥10 %. The covariates for the final adjusted models are noted in the footnotes of Tables 4 and 5.
To assess potential interaction of the two diet scores with the lifestyle score, we conducted joint/combined (cross-classification) analyses. In these analyses, the reference group was participants who were in the first quintile of both the lifestyle score and the diet score of interest.
To assess whether associations differed by categories of other risk factors not included in the scores, we conducted separate analyses within each category of age (≤/>median age of 61 years), education (≤high school/>high school), chronic disease (yes/no), total energy intake (≤/>median of 7186 kJ/d) and HRT use (current or past/never).
To assess the sensitivity of the associations with various considerations, we repeated the analyses with the following variations: (1) excluded participants who died within 1 or 2 years of follow-up, (2) used a lifestyle score composed of unweighted components and (3) used a Mediterranean diet score based on dichotomised components. Finally, we investigated whether removing and replacing each component of each score one at a time materially affected the observed associations.
All analyses were conducted using SAS statistical software, version 9.4 (SAS Institute Inc.). All P values were two-sided. A P value ≤0·05 or a 95 % CI that excluded 1·0 was considered statistically significant.
Results
Of the 18 687 participants who died during 310 762 person-years of follow-up over 26 years (interquartile range: 11·6–22·6 years), 7064 died of CVD and 4665 of cancer. The baseline characteristics of the study participants according to quintiles of the evolutionary-concordance and Mediterranean diet scores and the evolutionary-concordance lifestyle score are presented in Table 3. Participants in the highest relative to the lowest quintile of each diet score were more likely to be educated, use HRT, be physically active, and have higher mean total Ca and dietary fibre intakes. Participants in the highest quintile of the evolutionary-concordance diet score had lower mean total energy, alcohol, protein and carbohydrate intakes and were more likely to have a chronic disease, while those in the highest quintile of the Mediterranean diet score had higher mean total energy, alcohol, protein, carbohydrate and total fat intakes. Exclusive of variables included in the evolutionary-concordance lifestyle score, participants in the highest quintiles of the score were more likely to be educated, less likely to have a chronic disease, and had higher mean total energy, total Ca, total fat, dietary fibre, protein and carbohydrate intake and lower alcohol intake.
Table 3 Selected characteristics of participants according to quintiles of the evolutionary-concordance and Mediterranean diet and evolutionary-concordance lifestyle pattern scores at baseline in the Iowa Women’s Health Study (n 35 221), 1986–2012 (Mean values and standard deviations, or percentages)
HRT, hormone replacement therapy.
* Continuous variables presented as means and standard deviations, and categorical variables as percentage.
† P values calculated using the χ 2 test for categorical variables and one-way ANOVA for continuous variables.
‡ Physical activity level derived from two questions regarding the frequency of moderate and vigorous physical activity(Reference Kushi, Fee and Folsom44), and categorised as high (vigorous activity twice a week or moderate activity >4 times/week), medium (vigorous activity once a week plus moderate activity once a week or moderate activity 2–4 times/week) and low.
§ History of diabetes, heart disease or cirrhosis.
|| Total = diet+supplements.
The evolutionary-concordance and Mediterranean diet scores ranged from 19 to 68 and 9 to 45, respectively. The correlation between the two diet scores was r 0·54 (P <0·01). The lifestyle score for all-cause mortality ranged from 2·34 to 17·25. The correlations between the lifestyle score and the evolutionary-concordance and Mediterranean diet scores were r 0·17 and 0·19, respectively (both P <0·01).
Multivariable-adjusted associations of the scores with all-cause and cause-specific mortality are presented in Table 4. For each score, the findings for all-CVD and all-cancer mortality were similar to those for all-cause mortality, and the lifestyle score was more strongly inversely associated with mortality than was either diet pattern score. When the lifestyle score was treated as a continuous variable, each point increase was associated with statistically significant 7, 6 and 8 % lower risk of all-cause, all-CVD and all-cancer mortality, respectively; when the score was categorised as quintiles, the corresponding estimates for those in the highest relative to the lowest quintiles were for 48, 47 and 49 % lower risk, respectively (all point estimates and tests for trend were statistically significant). The evolutionary-concordance diet score, whether treated as a continuous or categorical variable, was minimally inversely, but not statistically significantly, associated with all-cancer and all-cause mortality. For those in the upper relative to the lowest quintile of the Mediterranean diet score, risk for all-cause, all-CVD and all-cancer mortality was estimated to be 15, 17 and 7 % lower, respectively (point estimates for all-cause and CVD mortality were statistically significant).
Table 4 Multivariable-adjusted associations of evolutionary-concordance and Mediterranean diet and evolutionary-concordance lifestyle pattern scores with total and cause-specific mortality in the Iowa Women’s Health Study (n 35 221), 1986–2012 (Numbers, hazard ratios (HR) and 95 % confidence intervals)
HR, hazards ratio.
* For score construction, see text and Table 1. HR from Cox proportional hazards models; covariates included age (years; continuous), smoking status (current, past, never smoker), education (<high school, high school, >high school), BMI (weight (kg)/height (m)2; continuous), physical activity (low, medium, high), total energy intake (kJ/d; continuous), hormone replacement therapy use (current, past, never), marital status (married, never married, widowed, divorced/separated) and chronic disease (yes/no).
† Includes smoking, physical activity and BMI; for score construction, see text. HR from Cox proportional hazards models; covariates included age (years; continuous), education (<high school, high school, >high school), total energy intake (kJ/d; continuous), hormone replacement therapy use (current, past, never), marital status (married, never married, widowed, divorced/separated), chronic disease (yes/no) and evolutionary-concordant diet score (quintiles).
‡ For all-cancer mortality, family history of cancer in a first-degree relative (yes/no) was added to the Cox proportional hazards models.
As shown in Table 5, being in the highest quintile of the lifestyle score jointly with being in either the highest quintile of the evolutionary-concordance diet score or the highest quintile of the Mediterranean diet score was associated with modestly lower risk for all-cause mortality (HR 0·46 (95 % CI 0·42, 0·50) and HR 0·45 (95 % CI 0·41, 0·49), respectively) than was being in the highest quintile of only one of the scores (both P interaction <0·01).
Table 5 Multivariable-adjusted joint/combined associations* of the evolutionary-concordance lifestyle score, and evolutionary-concordance and Mediterranean diet pattern scores with all-cause mortality in the Iowa Women’s Health Study (n 35 221), 1986–2012 (Hazard ratios (HR) and 95 % confidence intervals)
Ref., reference.
* HR from Cox proportional hazards models; covariates included age (years; continuous), education (<high school, high school, >high school), total energy intake (kJ/d; continuous), hormone replacement therapy use (current, past, never), marital status (married, never married, widowed, divorced/separated) and chronic disease (yes/no).
† P interaction <0·01; from lifestyle score × diet score interaction term in the Cox proportional hazards model.
There were no consistent, clear patterns of differences in associations of any of the scores with all-cause mortality according to age, education, total energy intake or HRT use (online Supplementary Table S1). However, whereas the association of the diet scores with all-cause mortality were null among those with a history of a chronic disease at baseline, there was a statistically significant trend for decreasing risk with an increasing score among those with no such history and for those in the upper relative to the lowest quintile of the evolutionary-concordance and the Mediterranean diet scores, the HR were 0·92 (95 % CI 0·87, 0·97) and 0·80 (95 % CI 0·76, 0·84) (P interaction <0·01 and P interaction = 0·05, respectively) (online Supplementary Table S1).
Removal of physical activity and BMI from the lifestyle score did not substantially change the association of the score with all-cause mortality (online Supplementary Table S2). However, removal of smoking status from the score attenuated the association: for those in the upper relative to the lowest quintile, the HR changed from 0·52 (95 % CI 0·50, 0·55) to 0·76 (95 % CI 0·72, 0·80). In addition, (1) exclusion of participants who died within 1 or 2 years of follow-up, (2) using an unweighted lifestyle score and (3) using a Mediterranean diet score based on dichotomised components did not materially alter the results (online Supplementary Tables S3–S5). In the sensitivity analyses, removal/replacement of each individual dietary score component one at a time did not materially change the results.
Discussion
Our findings suggest that (1) a more Mediterranean-like diet pattern and (2) a more evolutionary-concordant lifestyle pattern, alone and in interaction with a more evolutionary-concordant or Mediterranean-like diet pattern, may be inversely associated with all-cause, all-CVD and all-cancer mortality. Although our findings for a more evolutionary-concordant diet were close to the null, its association with all-cause mortality was more inverse and statistically significant among those without a history of a chronic disease at baseline, suggesting that the diet pattern may possibly more strongly influence preventing than ameliorating chronic diseases that lead to premature mortality.
More evolutionary-concordant and Mediterranean-like diet patterns and more evolutionary-concordant lifestyle patterns may reduce the risk of chronic diseases that lead to premature mortality by several plausible mechanisms. Both diet patterns include high intakes of fruits, vegetables and nuts, which are rich sources of a variety of nutrients that may reduce risk via modulating detoxification enzymes; stimulating the immune system; reducing platelet aggregation; modulating cholesterol synthesis and hormone metabolism; and reducing blood pressure and antioxidant, antibacterial and antiviral effects(Reference Lampe51, Reference Bao, Han and Hu52). Both diet patterns also include low intakes of red, processed and fatty meats, which are high in high-energy total and saturated fats, which contribute to a positive energy balance, oxidative stress, inflammation and mutagenic/mitogenic bile acids in the gut(Reference Santarelli, Pierre and Corpet53–Reference McAfee, McSorley and Cuskelly55) and have been linked to hypercholesterolaemia, endothelial dysfunction, atherosclerosis, insulin resistance, hypertension, CVD, type 2 diabetes and colorectal cancer(Reference McAfee, McSorley and Cuskelly55–Reference Cross, Leitzmann and Gail58). Physical inactivity, obesity and smoking individually have been consistently reported in epidemiological studies and systematic reviews with meta-analyses to be associated with higher risk of all-cause and cause-specific mortality(Reference Gellert, Schottker and Brenner19, Reference Jones, Tellez-Plaza and Navas-Acien20, Reference Li, Gu and Jing21–Reference Nocon, Hiemann and Muller-Riemenschneider25). Being more physically active was found to reduce obesity and risk of falling and associated injuries and to improve glucose metabolism, bone health, independent living and physical well-being, all of which may help reduce mortality risk(Reference Warburton, Glendhill and Quinney59). Obesity increases insulin resistance, inflammation and oxidative stress and contributes to the risk of death from some cancers, CVD and all causes combined(Reference Barness, Opitz and Gilbert-Barness60-Reference Solomon and Manson62). Smoking delivers known toxicants and carcinogens, which cause DNA damage leading to mutations and thus to multiple types of cancer; cell damage – especially in small airways – leading to chronic airway diseases; and endothelial dysfunction, dyslipidaemia and platelet activation – leading to vascular occlusion, thereby contributing to premature death(Reference Doll63, Reference Tsiara, Elisaf and Mikhailidis64).
Several reported studies investigated the possible health benefits of following a more evolutionary-concordant diet pattern. In four small, uncontrolled trials, four short-term randomised trials and one 2-year randomised trial, participants on an evolutionary-concordance diet pattern intervention, either short-term or long-term, had reductions in weight and waist circumference(Reference Genoni, Lyons-Wall and Lo65–Reference Otten, Stomby and Waling69); improved glucose control, lipid profiles and insulin sensitivity(Reference Lindeberg, Jonsson and Granfeldt66, Reference Mellberg, Sandberg and Ryberg67, Reference Otten, Stomby and Waling69–Reference Masharani, Sherchan and Schloetter74); and decreases in blood pressure(Reference Osterdahl, Kocturk and Koochek68, Reference Boers, Muskiet and Berkelaar71–Reference Jonsson, Granfeldt and Ahren73). An evolutionary-concordance diet pattern was inversely associated with inflammation and oxidative stress biomarkers in a cross-sectional study (n 646)(Reference Whalen, McCullough and Flanders30); with incident, sporadic colorectal adenoma in a case–control study (n 2301)(Reference Whalen, McCullough and Flanders31); and with all-cause, all-CVD and all-cancer mortality in the REasons for Geographic and Racial Differences in Stroke (REGARDS) prospective cohort study (n 21 423; 2513 deaths during follow-up)(Reference Whalen, Judd and McCullough32). In REGARDS, for participants in the highest relative to the lowest evolutionary-concordance (‘Paleolithic’) diet pattern score quintile, the HR for all-cause, all-CVD and all-cancer mortality were 0·77 (95 % CI 0·67, 0·89), 0·78 (95 % CI 0·61, 1·00) and 0·72 (95 % CI 0·55, 0·95), respectively; the associations did not substantially differ by sex. Our minimally inverse, non-statistically significant results may be explained by the relative homogeneity of the dietary exposures in the IWHS population. Examples of first and fifth quintile median intakes in the IWHS v. the REGARDS cohorts are total vegetables, 11·5 and 43 v. 6·5 and 52·6 servings/week; sugar-sweetened beverages, 0 and 4 v. 0 and 11·8 servings/week and alcohol, 0 and 12·1 v. 0 and 23·1 g/d (online Supplementary Table S6).
Our findings of inverse associations of a Mediterranean diet pattern with all-cause, all-CVD and all-cancer mortality are consistent with those reported in previous prospective studies. Of ten reported prospective cohort studies(Reference Knoops, de Groot and Kromhout8–Reference Tognon, Nilsson and Lissner14, Reference Vormund, Braun and Rohrmann16, Reference Zazpe, Sanchez-Tainta and Toledo17, Reference Prinelli, Yannakoulia and Anastasiou28), all found inverse associations with all-cause mortality, of which nine were statistically significant(Reference Knoops, de Groot and Kromhout8–Reference Tognon, Nilsson and Lissner14, Reference Zazpe, Sanchez-Tainta and Toledo17, Reference Prinelli, Yannakoulia and Anastasiou28); the estimated associations were generally stronger among men(Reference Mitrou, Kipnis and Thiebaut10, Reference Reedy, Krebs-Smith and Miller11, Reference Sotos-Prieto, Bhupathiraju and Mattei12, Reference Tognon, Nilsson and Lissner14, Reference Vormund, Braun and Rohrmann16). Of the eight reported prospective cohort studies(Reference Knoops, de Groot and Kromhout8, Reference Mitrou, Kipnis and Thiebaut10–Reference Vormund, Braun and Rohrmann16), all reported inverse associations with all-CVD mortality, of which seven were statistically significant(Reference Knoops, de Groot and Kromhout8, Reference Mitrou, Kipnis and Thiebaut10–Reference Tong, Wareham and Khaw15); five reported associations stratified by sex, of which two reported stronger, statistically significant, associations among women(Reference Reedy, Krebs-Smith and Miller11, Reference Tognon, Nilsson and Lissner14). Of the six reported prospective cohort studies of associations of Mediterranean diet pattern scores with all-cancer mortality(Reference Knoops, de Groot and Kromhout8, Reference Mitrou, Kipnis and Thiebaut10, Reference Reedy, Krebs-Smith and Miller11, Reference Sotos-Prieto, Bhupathiraju and Mattei12, Reference Tognon, Nilsson and Lissner14, Reference Vormund, Braun and Rohrmann16), all reported inverse associations, of which three were statistically significant(Reference Knoops, de Groot and Kromhout8, Reference Mitrou, Kipnis and Thiebaut10, Reference Reedy, Krebs-Smith and Miller11); five reported associations stratified by sex, of which one reported a stronger, statistically significant, association among women(Reference Reedy, Krebs-Smith and Miller11).
Eight prospective cohort studies reported associations of lifestyle scores (some studies considered diet as part of lifestyle) with all-cause and cause-specific mortality. Common score components included alcohol, dietary behaviour (primarily framed as adherence to existing dietary recommendations; for example, Mediterranean diet, Danish Dietary Recommendations, Dietary Guidelines to Australians, etc.), physical activity, sedentary behaviour, smoking, BMI and waist circumference(Reference Ding, Rogers and van der Ploeg3–Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26–Reference Yun, Won and Kimm29). Although these studies created different lifestyle scores, their results are consistent with ours in relation to all-cause(Reference Ding, Rogers and van der Ploeg3–Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26–Reference Yun, Won and Kimm29), all-CVD(Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26) and all-cancer(Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26, Reference Yun, Won and Kimm29) mortality. Of eight studies that investigated associations of lifestyle scores with all-cause mortality, all reported statistically significant, inverse associations(Reference Ding, Rogers and van der Ploeg3–Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26–Reference Yun, Won and Kimm29). Of the three studies of all-CVD mortality, all reported inverse, statistically significant associations(Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26). Of the four studies of all-cancer mortality, all reported statistically significant inverse associations(Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26, Reference Yun, Won and Kimm29). Of the three studies that reported results by sex, all found stronger inverse associations with all-cause and all-CVD mortality but not with all-cancer mortality, among women(Reference Ding, Rogers and van der Ploeg3, Reference Petersen, Johnsen and Olsen5, Reference Yun, Won and Kimm29).
In line with previous studies(Reference Ding, Rogers and van der Ploeg3–Reference Petersen, Johnsen and Olsen5, Reference Knoops, de Groot and Kromhout8, Reference Ford, Bergmann and Boeing26–Reference Yun, Won and Kimm29), our findings of statistically significant, albeit modest, estimated interactions between a more evolutionary-concordant or Mediterranean-like diet and the lifestyle score suggest that multiple lifestyle factors and diet may interact to influence mortality. Since the effects of adverse factors may accumulate throughout life, it is particularly essential for elderly people to adhere to healthy diets and lifestyles that minimise their risk of death(75).
Our study strengths include that this is the first study of associations of a weighted evolutionary-concordance lifestyle score with all-cause and cause-specific mortality, the long-term prospective study design and the large sample size and number of deaths. However, the study has certain limitations. The FFQ has known limitations (e.g. recall error, limited number of food items, complex task of estimating and calculating intake frequencies, etc.). There was also limited reassessment of diet and other key exposures during follow-up. Although we adjusted for many potential confounders, residual and unmeasured confounding cannot be ruled out. Our study population was limited to older, white Iowa women, limiting the generalisability of our findings. Furthermore, as noted above, there was more homogeneity of diet and other exposures in our study population than found in other cohorts. However, our findings are generally consistent with those of other investigators.
In conclusion, our findings, taken together with previous literature, suggest that (1) a more Mediterranean-like diet pattern and (2) a more evolutionary-concordant lifestyle pattern, alone and in interaction with a more evolutionary- or Mediterranean-concordant diet pattern, may be inversely associated with all-cause, all-CVD and all-cancer mortality. Also, our findings for a more evolutionary-concordant diet alone, considering (a) its statistically significant, albeit modest, estimated interaction with the lifestyle score, (b) that it was inversely associated with mortality among those without a history of a chronic disease at baseline (which may suggest possible stronger influence on preventing than on ameliorating chronic diseases that lead to premature mortality) and (c) previous findings that it was inversely associated with mortality in a population with more heterogeneous diets, suggest that a more evolutionary-concordant diet pattern may contribute to lower mortality risk. Given that this is the first study of an association of an evolutionary-concordance lifestyle score – alone or jointly with evolutionary concordance and Mediterranean diet pattern scores – with mortality, further study in other populations is indicated.
Acknowledgements
The study was supported by National Cancer Institute of the National Institutes of Health (grant R01 CA039742). The National Cancer Institute had no influence on the analysis and interpretation of the data, the decision to submit the manuscript for publication, or the writing of the manuscript.
The author’s responsibilities are as follows: E. C. and R. M. B. designed the research; A. P. and D. L. coordinated data collection and provided the dataset; E. C. and C. Y. U. analysed data; E. C. wrote the manuscript; C. Y. U., A. P. and D. L. provided comments on the manuscript; R. M. B. revised the manuscript, supervised the research and had primary responsibility for final content. All authors read and approved the final manuscript.
None of the authors has any conflicts of interest to declare.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114518003483
