|Home | About | Journals | Submit | Contact Us | Français|
Previous studies have shown distinct associations between specific dietary fat and cardiovascular disease. However, evidence on specific dietary fats and mortality remains limited and inconsistent.
To examine the associations of specific dietary fats with total and cause-specific mortality in two large ongoing prospective cohort studies.
We investigated 83,349 women from the Nurses’ Health Study (1980-2012) and 42,884 men from the Health Professionals Follow-up Study (1986-2012) who were free from cardiovascular disease, cancer and diabetes at baseline. Dietary fat intake was assessed at baseline and updated every 2 to 4 years.
We performed systematic searches of the vital records of states and of the National Death Index, supplemented by reports from family members or postal authorities.
We documented 33,304 deaths during 3,439,954 person-years of follow-up. After adjustment for known and suspected risk factors, dietary total fat, compared to total carbohydrate, was inversely associated with total mortality (P for trend <0.001). The hazard ratios (HRs) of total mortality comparing extreme quintiles of specific dietary fats was 1.08, (95% confidence interval (CI), 1.03-1.14) for saturated fat, 0.81 (95% CI, 0.78-0.84) for polyunsaturated fat, 0.89 (95% CI, 0.84-0.94) for monounsaturated fat and 1.13 (95% CI, 1.07-1.18) for trans fat (P for trend <0.001 for all). Replacing 5% of energy from saturated fats with equivalent energy from polyunsaturated fats and monounsaturated fats was associated with 27% (HR =0.73, 95% CI, 0.70-0.77) and 13% (HR =0.87, 95% CI, 0.82-0.93) estimated reductions in total mortality, respectively. HR of total mortality comparing extreme quintiles of n-6 polyunsaturated fat intake was 0.85 (95% CI, 0.81-0.89). Intake of n-6 polyunsaturated fat, especially linoleic acid, was inversely associated with mortality due to most major causes, while marine n-3 polyunsaturated fat intake was associated with a modestly lower total mortality (HR comparing extreme quintiles =0.96, 95% CI, 0.93-1.00).
Different types of dietary fats have divergent associations with total and cause-specific mortality. These findings support current dietary recommendations to replace saturated and trans fat with unsaturated fats.
Health effects of different types of dietary fat have been a long-standing research interest for decades. 1 Vast literature clearly indicates that specific types of dietary fat have distinct effects on cardiovascular disease (CVD) risk and replacing saturated fats with unsaturated fats and avoidance of trans fat is widely recommended. 1-4 However, a recent meta-analysis concluded that dietary polyunsaturated fatty acids (PUFAs) or saturated fatty acids (SFAs) had no significant relationships with risk of coronary heart disease (CHD), but this meta-analysis failed to specify the macronutrient to which saturated fat was compared; 5 by default this is largely refined starch and sugar in Western diets. 6,7 In addition, existing evidence, especially those from relatively small clinical trials, regarding the effects of n-3 PUFAs on total and cause-specific mortality remains inconsistent. 8,9 Health benefits of n-6 PUFA intake are still contentious; concern has been raised over its hypothesized pro-inflammatory and pro-thrombotic effects. 10 The evidence for industrially produced trans fatty acids (TFAs) and risk of CHD is well established, 11 although little data are available on TFA intake and mortality.
Numerous controlled metabolic trials have documented the effects of dietary fatty acids on blood lipids. 3,12 However, these fatty acids also influence other mechanistic pathways such as insulin resistance, endothelial function, electrophysiology, carcinogenesis, and systemic inflammation. 9,13,14 In addition to CVD, various dietary fatty acids have been associated with incidence of other major chronic diseases including type 2 diabetes, cancer, multiple sclerosis, and respiratory diseases in prospective cohort studies. 15-18 To address the role of dietary fats in overall health, it would be informative to analyze total and cause-specific mortality as outcomes. Therefore, we prospectively examined the associations of specific dietary fats with total and cause-specific mortality in two large ongoing prospective cohort studies, the Nurses’ Health Study (NHS) and Health Professionals Follow-up Study (HPFS). The many repeated dietary assessments in the cohorts provided a unique assessment of diet over multiple decades. Also, this allowed us to estimate the consequences of isocalorically replacing one type of fat with another, thus describing the choices we make on a daily basis and providing a basis for dietary guidelines.
The NHS is a prospective cohort study of 121,700 registered female nurses aged 30 to 55 years in 1976; 92,468 participants responded to the semi-quantitative food frequency questionnaire (SFFQ) in 1980. The HPFS is a prospective cohort of 51,529 male health professionals aged 40 to 75 years in 1986. The baseline of this analysis was defined as 1980 for the NHS and 1986 for the HPFS. The two cohorts have been followed via biennial mailed questionnaires that inquire about lifestyle risk factors and other exposures of interest, as well as newly diagnosed diseases. We also collected information on race, marital status and family history of major chronic diseases. The cumulative follow-up of the two cohorts exceeds 90% of potential person-time. The study was approved by the human research committees at the Harvard T.H. Chan School of Public Health and the Brigham and Women’s Hospital.
We excluded participants who had a history of diabetes, CVD or cancer, did not provide information on dietary fat intake, or reported implausible SFFQ data (total energy intake <800 or >4200 kcal/d for men and <600 or >3500 kcal/d for women), at baseline (Table S1). After exclusions, the analytical population consisted of 83,349 women and 42,884 men.
Dietary information was collected with SFFQs. 19 In each SFFQ, we asked how often, on average, the participant had consumed a specified portion size of each food over the preceding year. The number of listed foods was 61 in 1980 and was expanded to 116-150 in 1984 and thereafter; additional frequently used foods were reported in an open-ended section. We also collected detailed information on type of fat or oil used in food preparation and brand/type of margarines by the SFFQ. Fatty acid and other nutrient values were calculated based on the Harvard University Food Composition Database, 20 which is updated regularly using external publications and direct analysis of fatty acids in commonly used margarines and processed foods to take into account changes in manufacturing. We calculated average daily nutrient and total energy intakes by multiplying the frequency of consumption of each item by its nutrient content and summing across all foods, taking into account specific brand and type of margarines and types of fat used in food preparation. The assessment of specific types of fat has been validated by comparison with multiple weighed 1-week dietary records and with fatty acid measurements in adipose tissue and plasma. 19,21-23 The correlations between energy-adjusted intakes assessed by the 1986 questionnaire and the mean of diet records collected in 1980 and 1986, corrected for variation in the records, were 0.67 for total fat, 0.70 for SFAs, 0.69 for monounsaturated fatty acids (MUFAs), and 0.64 for PUFAs. 19 Correlations increased when the mean of three SFFQs (1980, 1984, and 1986) was used; for example, for SFAs the correlation was 0.95. 19 The correlation between dietary fatty acid intake assessed by the SFFQ and the composition of fatty acids in adipose tissue were 0.51 for TFAs, 0.35 for linoleic acid, and 0.48 for marine n-3 PUFAs in women, 21 and 0.29 for TFAs, 0.48 for linoleic acid, and 0.47 for eicosapentaenoic acid in men. Alcohol intake was also estimated from the SFFQs.
In the NHS, dietary questionnaires used in this analysis were completed in 1980, 1984, 1986, and then every four years for a total of nine. In the HPFS, dietary questionnaires were completed in 1986 and then every 4 years for a total of seven.
We performed systematic searches of the vital records of states and of the National Death Index, supplemented by reports from family members or postal authorities. More than 98% of the deaths in each cohort were identified. 24 A physician reviewed death certificates and medical records to classify the cause of death according to the eighth and ninth revisions of the International Classification of Diseases (Table S2).
The percentage energy intakes from total fat and specific fats were calculated as cumulative average up to the start of each two or four-year follow-up interval to best represent long-term dietary intake and dampen within-person variation. We categorized participants into quintiles of intake levels. Person-years of follow-up were calculated from baseline to the earliest of time of death, loss to follow-up or the end of follow-up. The end of follow-up was defined as June 2012 for the NHS and December 2012 for the HPFS. Proportional hazard Cox models were applied to estimate hazard ratios (HRs) and their 95% confidence intervals (CIs) of mortality comparing participants in each quintile to the lowest quintile. To quantify a linear trend, we assigned the median within each quintile and modeled this variable continuously; Wald test was used for statistical significance. In addition to including percentages of energy from total and specific fat as quintiles, we also included them as continuous terms in the multivariable models. For multivariable analyses, we built isocaloric substitution models which simultaneously included energy intake, the percentages of energy derived from protein and specific types of fat and other potentially confounding variables (see the footnote of Table 2). The coefficients from these models can be interpreted as the estimated effect of substituting a certain percentage of energy from fat for equivalent energy from carbohydrates. For repeatedly measured covariates, we included their updated values as time-varying variables in the model. To minimize missing covariates, we replaced missing data with the last valid values. The covariates had very few missing values after the replacement. In addition, we included missing indicators for the remaining missing covariates in the model. To evaluate the effect of substituting specific types of fat for saturated fat, we treated intakes as continuous variables and calculated the difference in coefficients. As we hypothesized that the effects of total and specific n-3 PUFAs might be more acute, we conducted an additional analysis using the most recent data on n-3 PUFAs at the beginning of each biennial follow-up. We conducted sensitivity analyses to test the robustness of our findings. To address concern that chronic disease occurrence in the years that precede diagnosis may influence dietary behavior, we conducted lagged analyses by excluding the first 4 years follow-up data and adding a 4-year lag period between dietary fat intake assessment and each follow-up period. To address the possibility that our findings may be explained by underlying overall dietary pattern, we further adjusted for overall dietary pattern score (the Alternate Healthy Eating Index-2010 score 25 minus component scores for fatty acids). To minimize the influence of hypertension and hypercholesterolemia on our results, we performed additional analysis by excluding participants who reported hypertension and hypercholesterolemia at baseline. An inverse-variance-weighted, fixed-effect meta-analysis was used to combine the results across cohorts. All analyses were performed using SAS software, version 9.2 (SAS Institute, North Carolina), at a two-tailed P value of 0.05.
During 32 years of follow-up in the NHS (2,464,852 person-years), we documented 20,314 deaths; during 26 years of follow-up in the HPFS (975,102 person-years), we documented 12,990 deaths. At baseline, participants with higher SFA and MUFA intakes had higher levels of BMI, total energy intake, and dietary cholesterol intake, but were less likely to be physically active, use multivitamin and vitamin E supplementations, and report histories of hypercholesterolemia and hypertension (Table 1). The prevalence of current smoking was higher among men with higher SFA and MUFA intakes. Women with higher intakes of SFA, PUFA and MUFA were generally older.
Although total fat intake was positively associated with total mortality in age-adjusted models, an inverse association became apparent after adjusting for other potential confounding variables (P for trend <0.001, Table 2). When substituted for total carbohydrate, a higher intake of SFA was associated with a slightly higher total mortality (HR comparing extreme quintiles =1.08, 95% CI, 1.03-1.14, P for trend <0.001). For TFA intake, a significant positive association with total mortality was observed (HR comparing extreme quintiles =1.13, 95% CI, 1.07-1.18, P for trend <0.001). Intakes of PUFA and MUFA were inversely associated with total mortality (Figure 1); HRs comparing extreme quintiles were 0.81 (95% CI, 0.78-0.84) for PUFA and 0.89 (95% CI, 0.84-0.94) for MUFA (P for trend <0.001 for both). The inverse association between total PUFA and mortality was mainly driven by linoleic acid as illustrated by their high correlation (correlation coefficient =0.99, Table S3). The corresponding HR for linoleic acid intake was 0.82 (95% CI, 0.79-0.86, P for trend <0.001, Table 3). Intake of total n-3 PUFA was associated with modestly lower total mortality, which was mainly driven by the inverse association of marine n-3 PUFAs (docosahexaenoic acid and eicosapentaenoic acid) with total mortality. HRs comparing extreme quintiles were 0.95 (95% CI, 0.91-0.99, P for trend =0.03) for total n-3 PUFAs and 0.96 (95% CI, 0.93-1.00, P for trend =0.002) for marine n-3 PUFAs. The inverse associations became stronger in continuous analyses and when recent intakes of total and marine n-3 PUFAs were used (Table S14); HRs comparing extreme quintiles were 0.91 (95% CI, 0.87-0.95) and 0.90 (95% CI, 0.87-0.93, P for trend <0.001 for both) for recent intakes of total and marine n-3 PUFAs, respectively. The n-6/n-3 ratio was not significantly associated with total mortality (P for trend =0.29). The associations for total and specific types of fat remained largely unchanged in sensitivity analyses (Tables S15-17).
SFA intake, when substituted for total carbohydrates, was not significantly associated with CVD mortality (P for trend across quintiles =0.17, Table S5), while TFA intake was associated a 20% higher CVD mortality across quintiles (HR =1.20, 95% CI, 1.08-1.33, P for trend <0.001). PUFA intake was inversely associated with CVD mortality (P for trend <0.001). We observed an inverse association, primarily in women, between MUFA intake and CVD mortality (P for trend =0.01). Among specific PUFAs, linoleic acid intake was most strongly related to lower risk of CVD mortality (P for trend <0.001, Table S6).
Dietary intake of SFA, when substituted for total carbohydrate, was associated with slightly higher cancer mortality (HR comparing extreme quintiles =1.07, 95% CI, 0.98-1.17, P for trend =0.02), whereas PUFA intake, especially linoleic acid intake, was associated with modestly lower cancer mortality (HR comparing extreme quintiles of PUFA intake =0.93, 95% CI, 0.87-0.99, P for trend =0.02, Tables S7 and S8). Other major subclasses of dietary fat generally were not associated with cancer mortality except that α-linolenic acid (ALA) intake was associated with a slightly elevated risk of cancer mortality (Tables S7 and S8). However, this association was not significant when recent ALA intake was analyzed (Table S14). We observed inverse associations of PUFA and MUFA intakes and strong positive associations of TFA intake with both neurodegenerative (Table S9) and respiratory disease mortality (Table S11). Higher SFA intake was associated with a substantial increase of mortality due to respiratory disease (HR comparing extreme quintiles =1.56, 95% CI, 1.30-1.87, P for trend <0.001). Among major PUFAs, n-3 PUFA intake, primarily α-linolenic acid, was inversely associated with neurodegenerative disease mortality (Table S10). Marine n-3 PUFA intake was inversely associated with respiratory disease mortality (Table S12). Sensitivity analyses minimally changed these results for cause-specific mortality (Tables S15-17).
Figure 2 shows that replacing 5% of energy from SFAs with the same energy from PUFAs and MUFAs was associated with 27% (HR =0.73, 95% CI, 0.70-0.77, Table S13) and 13% (HR =0.87, 95% CI, 0.82-0.93, Table S13) estimated reduction in total mortality, respectively. Replacing SFAs with the same energy from PUFAs was associated with lower risk of mortality due to CVD, cancer and neurodegenerative disease. Replacement of 5% of energy from SFAs with 5% of energy from MUFAs was associated with 29% estimated reduction in neurodegenerative disease mortality (HR =0.71, 95% CI, 0.57-0.88, Table S13).
In two large cohorts with many repeated measures of diet and long duration of follow-up, we found that higher intakes of PUFA and MUFA were associated with lower mortality, whereas higher intakes of SFA and TFA were associated with increased mortality. Although the modest positive association between SFA intake and mortality suggests small health benefits of replacing SFAs by total carbohydrate, replacing SFAs with MUFAs and/or PUFAs was associated with a significantly lower risk of total mortality and cause-specific mortality due to several major chronic diseases. Dietary intake of total fat, compared to total carbohydrate, was inversely associated with total mortality. However, the association between total fat intake and mortality largely depends on specific types of fat. Intake of linoleic acid, the most abundant n-6 PUFA, showed strong inverse associations with total and most cause-specific mortality, without any evidence of detrimental effects. Interestingly, a higher ratio of n-6 to n-3 PUFAs was not associated with increased mortality, but with a slightly lower total, CVD, and cancer mortality.
Previous data on the associations between different types of n-6 PUFA intake and total mortality have been limited. Concordant with our findings, two prospective cohort studies using circulating biomarkers 26,27 found a lower total mortality associated with higher concentrations of n-6 PUFA in blood. Evidence from clinical trials also supported a protective effect of soybean oil with a higher ratio of n-6 to n-3 PUFAs on CHD risk. 28,29 Contrary to our findings, a recent re-analysis of the Sydney Diet Heart Study, a secondary prevention trial, reported a significant increase in total mortality in participants assigned to higher n-6 PUFA intervention. 30 However, this study was very small (n =221), of short duration (39 months), only among individuals with existing CVD, and the intervention likely reduced n-3 PUFA intake. In addition, the results may have been confounded by trans fat in the special margarines used for intervention that were high in linoleic acid 6.
We observed a modest inverse association between marine n-3 PUFA intake and total mortality. Previously, prospective cohorts in generally healthy populations yielded mixed results 31-35, while most RCTs found non-significant effects of fish oil supplementation on total mortality 9,36. The significant inverse association between intake of n-3 PUFAs and death due to neurodegenerative diseases has not, to our knowledge, been previously reported.
We found a significant inverse association between MUFA intake and total mortality. In contrast, previous studies generally reported non-significant or even positive associations with MUFA 2,37,38 This discordance might be due to the strong correlations between MUFA and SFA because animal fats are major sources of both types of fats in most Western diets, and between MUFA and TFA because partial hydrogenation produces both. In our two cohorts the correlation between MUFA and SFA has decreased during the follow-up and the major food sources of MUFA have shifted from animal-sourced to plant-sourced foods over time (Table S18), thus we had greater power to differentiate the association for MUFA with mortality. Consistent with our analysis, the major source of MUFA in Mediterranean populations, olive oil, has been associated with a substantially lower total mortality, 37 Important benefits of MUFA from plant sources have also been supported by the PREDIMED trial, in which the addition of olive oil and nuts, also high in MUFA, reduced incidence of cardiovascular disease and diabetes 39,40.
Compared to overall carbohydrates, higher SFA intake was associated with a slight increase in total mortality, but not significantly associated with CVD mortality. The lack of the association with CVD is expected because the major sources of carbohydrate in typical Western diet are highly processed foods with large amounts of refined starch and sugar, providing a high glycemic load that can increase cardiovascular disease risk independent of SFA. 7,25,41,42 In our substitution analyses, replacing SFA with unsaturated fatty acids was associated with substantially lower risk of total and CVD mortality. These findings were generally consistent with evidence from previous studies. 2,43-45 However, a recent meta-analysis concluded that specific types of fat including saturated, monounsaturated or polyunsaturated fats had no significant impact on risk of CHD. 5 Another meta-analysis also reported nonsignificant associations of SFA intake with total and CVD mortality and CHD incidence. 46 However, most studies included in these meta-analyses did not explicitly model the effects of macronutrient substitution and did not specify the comparison source of energy for the type of fat under scrutiny, which limits the interpretations of these findings. 6,43 Our analyses provide strong evidence that using PUFAs and/or MUFAs as the replacement nutrients for SFAs can confer substantial health benefits, while replacing SFAs with carbohydrates has little impact on CVD mortality. However, the effects of replacement by carbohydrates may depend in part on the quality of carbohydrates.
Data on specific types of dietary fat and non-CVD mortality are sparse. We observed no major effects of most types of dietary fat on cancer mortality, although a modest inverse association with linoleic acid was observed. Our findings of an inverse association between higher intake of PUFAs, primarily α-linolenic acid, and lower mortality due to neurodegenerative diseases are consistent with limited evidence on incidence of major neurodegenerative diseases, including Alzheimer disease 47, amyotrophic lateral sclerosis 15, and Parkinson's disease 48 in prospective cohorts. Finally, we observed a positive association between saturated and trans fat intake and respiratory disease mortality, but an inverse association with PUFAs. These findings are novel and therefore require confirmation in further studies.
Our results have several limitations. First, reverse causation is a possible explanation for our findings, because people with chronic disease and poor health might change their habitual diet. However, we excluded participants with known major chronic diseases at baseline. Also, it is likely that those concerned about a serious illness would change toward a diet generally perceived to be healthier, which would not explain our findings. In addition, our findings remain largely unchanged when we excluded the first 4 years of follow-up or added a 4 year lag period between dietary assessment and each follow-up period. Secondly, because our study was observational in nature, causality cannot be established. However, our results were largely consistent with those from results from existing observational studies and randomized clinical trials on diet and CVD-related outcomes. Thirdly, even though we adjusted for many potential confounders, residual confounding could not be ruled out. Fourthly, measurement errors are inevitable in estimates of food and nutrient intakes. However, our adjustment for energy intake and use of prospectively collected, cumulatively averaged intake using many repeated dietary assessments reduced the impact of measurement errors. 19 The strengths of the current study included a large sample size, high follow-up, and repeated assessments of dietary and lifestyle variables during a long-time period.
In summary, we found that different types of dietary fats have divergent associations with total and cause-specific mortality. Replacement of saturated fats with unsaturated fats can confer substantial health benefits and should continue to be a key message in dietary recommendations. These findings also support the elimination of partially hydrogenated of vegetable oils, the primary source of trans fatty acids.
Funding/Support: This study was supported by research grants UM1 CA186107, P01 CA87969, R01 HL034594, R01 HL088521, UM1 CA167552, R01 HL35464, R01 HL60712 and P30 DK46200 from the National Institutes of Health.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Willett and Hu had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Design and conduct of the study: Wang, Willett, and Hu
Collection, management, analysis, and interpretation of the data: All authors
Preparation, review, or approval of the manuscript: All authors
Decision to submit the manuscript for publication: All authors
Statistical analysis: Wang and Hu
Conflict of interest disclosure: Hu has received research support from California Walnut Commission and Metagenics.
Additional Contributions: We are indebted to the participants in the Nurses’ Health Study and the Health Professionals Follow-up Study for their continuing outstanding level of cooperation; to the state cancer registries for their kind help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY; to the staff of the Nurses’ Health Study and the Health Professionals Follow-up Study for their valuable contributions; to Drs. Alberto Ascherio and Kathryn C. Fitzgerald for their helpful comments; to Lisa Li for her valuable programming assistance. The authors assume full responsibility for analyses and interpretation of these data.