In this large, community-based, prospective cohort study, we found overall inverse associations of total mortality with intakes of total fat, SFA, MUFA, and PUFA in women, particularly among those aged ≥ 60 years, but not in men. In women, detailed analysis revealed a U-shaped association between total fat intake and total mortality; the lowest risk appeared within the second highest quintile of energy-adjusted fat intake. Regarding major causes of death in men, an upward trend in the HR for CVD mortality was detected with increasing PUFA intake and the HR for cancer mortality was elevated in the highest quintile of total fat intake. Among women, causes of death other than cancer and CVD contributed most to the decreases in HR associated with increasing fat intake.
In women, the lowest HR for total mortality appeared at the upper half of the fourth quintile of total fat intake, at about 23%E. Given the underestimation of the intake (%E) with the FFQ for the JACC Study (~16%), the lowest risk might be associated with a higher intake of about 28%E. This falls within the acceptable range recommended by the Food and Agriculture Organization of the United Nations (FAO) consultation  and the United States Department of Agriculture (USDA) guideline (20–35%E) , but may exceed the upper limit (25%E) proposed in dietary reference intakes for Japanese individuals of 30 years or older . If the lowest risk for female total mortality really is near 28%E of total fat intake, however, our findings may not be in line with those of Leosdottir et al.. They did not find any increase in risk among women who consumed total fat ≥ 35%E compared with those in the reference group with a mean of 30.8%E.
With regard to cancer, the World Cancer Research Fund (WCRF) and the American Institute for Cancer Research (AICR) extensively reviewed the literature and concluded that there is no convincing or probable evidence for significant effects of dietary fats on any type of cancer . In the present study, however, the HR for male cancer mortality was significantly increased by 17% in the top quintile of total fat intake (Table 2), although no linear trend was apparent. The elevated risk was predominantly observed for colorectal cancer; we did not present this finding in the Results section because detailed analyses by cancer site were outside of the scope of this study, and the relevant data from the JACC Study have yet to be published. Nevertheless, the systematic review by the WCRF and AICR  mentioned that there is a limited amount of fairly consistent evidence suggesting that consumption of foods containing animal fats is a contributing causal factor for colorectal cancer. The role of dietary fat in the development of colorectal cancer may thus help to explain the higher HR for total cancer mortality associated with higher intake of total fat.
For SFA, the lowest HR in women was found in the highest quintile of intake, which would correspond to 7.7%E considering the overestimation with the FFQ. Although whether the risk of total mortality rises with further increase in SFA intake was not clear, this percentage is consistent with saturated fat intake recommendations by the FAO consultation (<10%E) and USDA (<10 or 7%E). The dietary reference intake for Japanese (4.5–7%E) is lower. Both the FAO consultation and USDA guideline recommend replacement of excessive SFA with PUFA [1, 3]. This recommendation is mainly based on evidence that replacing dietary SFA with PUFA is associated with a decreased risk of CHD in both randomized controlled trials and cohort studies . In contrast, replacing SFA with carbohydrate or MUFA has been correlated with an increased risk of CHD in observational cohorts . Nevertheless, our findings regarding P/S ratio did not support the recommendation; in the present study, the risk for cardiovascular death increased at a higher P/S ratio in men.
In a detailed analysis specific to CVD from the JACC Study , SFA intake was inversely associated with mortality from stroke, but was not correlated with mortality from CHD, again consistent with substitution models replacing SFA with carbohydrate . The risk for overall CVD death was decreased with increasing intake of SFA in that study. The inverse correlation of SFA intake with CVD risk was much weaker in the current study compared with the previous analysis . However, this may be explicable principally by differences between the previous and present analyses in terms of the statistical analyses, including the covariates incorporated in the models. In particular, we did not include other fatty acids in the model, while the previous study included both n-3 and n-6 PUFA . This is because we did not intend to adjust for the effects of other fatty acid intakes correlated with SFA intake in the daily diet. When we applied a simple age- and sex-adjusted model as in the similar analysis in the previous study , an inverse association was clearly apparent; HRs for CVD across quintiles of SFA intake were 1.00, 0.89 (95% CI, 0.81–0.99), 0.91 (0.82–1.00), 0.77 (0.69–0.85), and 0.74 (0.67–0.83) (trend P < 0.001). This indicates that the SFA-CVD association was essentially comparable between the present and previous analyses. SFA intake tended to be associated, although not significantly, with a decreased risk of stroke in a meta-analysis of cohort studies . Regarding the role of SFA in relation to all-cause mortality when used to replace macronutrients, Nagata et al. showed an association between high SFA intake and higher total mortality among women in a model substituting SFA for carbohydrates .
In women, the inverse association between SFA intake and mortality was most evident for those causes of death other than cancer and CVD. Those causes mainly comprised injury (ICD-10: S00-T98) and pneumonia (ICD-10: J12-J18), accounting for 19.4% and 15.9% of all other causes of female deaths (n = 1,883), respectively. Low serum cholesterol levels have been associated with external-cause mortality such as suicide and accidents in prospective studies [32–34]. Such results may support the current findings, because the increased SFA intake raises serum cholesterol levels  and therefore might decrease the risk of external-cause mortality. In Japan, suicide and accidents accounted for almost all external causes of death (injury deaths), at 53.3% and 41.7%, respectively, in 1999, as the midpoint of the follow-up period for the present study . In addition to external causes of death, community-acquired pneumonia has been associated with low serum cholesterol in high-risk populations . Fat insufficiency may disturb the absorption of fat-soluble vitamins including vitamins A and E [1–3], which are required to maintain resistance to infectious diseases .
Only a few studies have directly correlated MUFA intake with all-cause mortality. In the Italian Longitudinal Study on Aging, higher MUFA intake was associated with decreased total mortality (HR for one standard deviation (SD) increment, 0.81; 95% CI, 0.66–0.99) , which may corroborate the inverse association among women in the current study. Alternatively, the Mediterranean diet has often been associated with decreased risk of all-cause death, cancer, and CVD . This may support our finding because the Mediterranean diet is rich in MUFA. However, other dietary factors in the Mediterranean diet, such as vegetables and fruit, have to be considered . Since dietary MUFA intake was very strongly correlated with SFA intake, examining the effect of dietary MUFA independently of SFA using multivariate models was difficult; the correlation coefficient for energy-adjusted intake was 0.86 in men and 0.85 in women in the analytic cohort.
PUFA intake correlated negatively with total mortality in women, which appeared to be due to a decreased risk of cancer death associated with a higher intake of n-3 PUFA and a decreased risk of death from other causes associated with a higher intake of n-6 PUFA. Marine n-3 PUFA was associated with a lower risk of breast cancer in a meta-analysis of cohort studies . Although no protective effect of n-3 fats against cancer was shown in a systematic review , typical fish intake is much higher in Asian populations, including Japanese, than in Western populations . One possibility, therefore, is that preventive effects of n-3 PUFA are suggested in studies from Japan. One study related linoleic acid intake to a lower risk of community-acquired pneumonia , which may partly explain the association between decreased risk of death from other causes and higher intake of n-6 PUFA.
We did not expect the elevated risk of male cardiovascular mortality associated with higher intake of n-6 PUFA, as several studies have shown inverse associations between n-6 PUFA intake and CVD . This finding could represent a chance phenomenon, since it was not observed in women. The role of dietary n-6 PUFA in the development of CVD, however, remains controversial [44, 45], and this study may provide additional data for discussion. Of note, the increased risk of male CVD death was principally due to stroke instead of CHD. Further investigations may therefore need to focus on stroke.
The JACC Study previously reported an inverse association between n-3 PUFA intake and CVD mortality  that was not clear in the current analysis. As mentioned in the discussion on SFA and CVD, this apparent inconsistency may be attributable to differences between the previous and current analyses in terms of the covariates included in multivariate models, especially dietary intakes of SFA and n-6 PUFA. When age- and sex-adjusted HRs were estimated as in the preceding study , we found an inverse (though not strong) association similar to that in the previous study; HRs for CVD across quintiles of n-3 PUFA intake were 1.00, 0.89 (95% CI, 0.80–1.00), 0.97 (0.87–1.09), 0.98 (0.88–1.09), and 0.89 (0.80–1.00) (trend P = 0.21). Regarding investigations on fat and cancer in the JACC Study, the higher serum level of n-3 PUFA was associated with decreased risk of male colorectal cancer  and higher dietary intakes of fish fat and long-chain n-3 fatty acids were correlated with a reduced risk of female breast cancer . Those studies, however, cannot be directly compared with our own because of the large difference in exposure assessment; they used serum fatty acids or specific fat/fatty acids (e.g., fish fat, long-chain n-3 fatty acids) instead of dietary total n-3 PUFA. Furthermore, those studies defined incidence of cancer as endpoints, while we adopted cancer deaths.
In Europe and the United States, high fat intake is associated with unhealthy living and unfavorable social class, represented as low intakes of fruit, vegetables and fiber, and increased smoking and alcohol intake on a population level . Conversely, a higher fat intake was associated with higher education, less smoking and male alcohol consumption, and higher consumption of fruit and vegetables in the current study (Table 1). One study based on nationwide surveys in Japan reported an increasing intake of fat (%E) with increasing household expenditure in men and women . Household expenditure also correlated positively with vitamin C and fiber intakes in that study. This suggests that individuals in a higher socioeconomic status in Japan consume not only more foods considered good for health (e.g., vegetables and fruit), but also more fat-rich foods, as seen in our study. Similar correlations between nutrient intakes and income or educational level have also been observed in Chinese women .
We therefore considered background characteristics and lifestyle factors including consumption of vegetables and fruit in participants as much as possible from the available data in our statistical analyses. Residual confounding, however, might have affected the results, and the differences in findings regarding fat and mortality among various populations may be partly attributable to differences in characteristics related to fat intake among populations.
No associations between fat intake and total mortality were apparent in men. This is inconsistent with findings from the Takayama Study among Japanese, which showed an inverse correlation with risk of all-cause mortality for total fat only in men . For comparison, we tried to further adjust for protein intake as performed in that study , but no significant correlation appeared in men.
The difference in alcohol consumption between sexes may be relevant at least in our cohort, and possibly among Japanese in general. Alcohol consumption contributes to energy intake and therefore decreases the percent of energy derived from fat, even if fat intake remains the same. Percentages of non-alcohol energy were not applied in the present study because dietary guidelines do not explicitly exclude alcohol energy from the denominator [1–3]. In the current cohort, women consumed only 1.7 g (12 kcal) per day of alcohol on average, whereas men drank 25.4 g (180 kcal) per day. We did not identify significant trends between fat intake and total mortality in men, even in the analysis limited to never-drinkers. Sex differences in our population are thus not well explained and should be examined in further investigations.
A tendency toward a somewhat reduced risk, however, was observed in the second highest quintile of total fat intake (HR, 0.92) among male never-drinkers, although this was not significant, partly due to the rather small sample size (n = 4,454). Larger studies of non-drinkers may also be warranted to evaluate effect modification or possible residual confounding by alcohol intake.
Similar sex differences were also found for SFA, MUFA, and PUFA in the present study; risk of total mortality was decreased with increasing intakes of SFA, MUFA and PUFA in women, whereas the risk was somewhat elevated with increasing PUFA intake in men. These findings, however, are not in line with preceding studies. Leosdottir et al.  reported a positive correlation between MUFA intake and total mortality, particularly in men, while Nagata et al.  found an inverse correlation of PUFA and a positive correlation of SFA with all-cause mortality in men and women, respectively. Further investigations are required before any conclusions can be drawn regarding differences in the effects of fatty-acid groups by sex.
The strengths of the current study include the prospective design, large sample size, long follow-up, and multivariate adjustments for potential confounding factors. The design enabled us to define death as an endpoint, and more than 11,000 events made detailed analysis possible. Moreover, the relatively lower level of fat intake among samples provided a unique opportunity to examine associations of fat intake with total mortality.
Some methodological limitations to our study warrant discussion. First, we used a simple FFQ with only 40 food items. Total energy intake may thus have been underestimated because of the foods not covered by the FFQ. However, the FFQ was validated by referring to dietary records, and the over- or underestimation of intakes (%E) of total fat and its components were adjusted using information from the validation study. We could not estimate the intake of trans-fatty acids with this FFQ. Although the mean intake in Japan [52, 53] is reported to be lower than the proposed upper limit (1% of total energy) , some Japanese, particularly women, might have a higher intake of trans-fatty acids . Second, we excluded substantial numbers of participants (n = 22,941) because of insufficient dietary information. The omitted participants were older and more likely to be men compared with those included in the analysis (mean age ± SD, 60.9 ± 9.8 vs. 56.1 ± 9.9 years; men, 45.7% vs. 39.4%). Other background characteristics, however, were quite similar between the 2 groups of participants, suggesting that these exclusions did not lead to any major selection bias. Third, fat intakes were assessed only once, at baseline. Fourth, the self-reporting of body weight and energy/nutrient intake might have led to bias. Underreporting of energy intake has been reported as more pronounced among women, overweight subjects and those with higher fat intake in European populations . In Asian populations, two studies reported that overweight or obese men and women (BMI ≥25 kg/m2) tended to underestimate their weight and/or BMI [55, 56]. Finally, because the present study was conducted among Japanese and the health effects of dietary fat may vary among populations with different genetic backgrounds , extrapolation to other populations must be made with caution.