Associations of dietary intakes of vitamins B1 and B3 with risk of mortality from CVD among Japanese men and women: the Japan Collaborative Cohort study

The evidence on the association between B vitamins and the risk of CVD is inconclusive. We aimed to examine the association of dietary vitamins B1 and B3 intakes with risk of CVD mortality among 58 302 Japanese men and women aged 40-79 years participated in the Japan Collaborative Cohort (JACC) study. The Cox proportional hazard model estimated the hazard ratios (HR) and 95% CI of CVD mortality across increasing energy-adjusted quintiles of dietary vitamins B1 and B3 intakes. During 960 225 person-years of follow-up, we documented a total of 3371 CVD deaths. After adjustment for age, sex, and other CVD risk factors, HR of mortality from ischemic heart disease, myocardial infarction, and heart failure in the highest v. lowest vitamin B1 intake quintiles were 0.57 (95 % CI 0·40, 0·80; Pfor trend < 0·01), 0.56 (95 % CI 0·37, 0·82; Pfor trend < 0·01), and 0.65 (95 % CI 0·45, 0·96; Pfor trend = 0·13). The multivariable HR of myocardial infarction mortality in the highest v. lowest vitamin B3 intake quintiles was 0.66 (95 % CI 0·48, 0·90; Pfor trend = 0·02). Atendency towards a reduced risk of haemorrhagic stroke mortality was observed with a higher dietary intake of vitamin B3 (HR: 0·74 (95 % CI 0·55, 1·01)) but not vitamin B1. In conclusion, higher dietary intakes of vitamins B1 and B3 were inversely associated with mortality from ischemic heart disease and a higher dietary intake of vitamin B1 was inversely associated with a reduced risk of mortality from heart failure among Japanese men and women.

was shown to decrease LDL-cholesterol, TAG and lipoprotein(a) levels, while increasing HDL-cholesterol level (7,8) . A meta-analysis of clinical trials suggested that vitamin B 3 supplements significantly reduced major coronary events, stroke and other cardiovascular events (9) .
Despite the abundant evidence on cardiovascular beneficial effects of vitamins B 1 and B 3 from animal studies (3,10,11) and supplemental clinical trials (12,13) , no human observational studies so far have investigated the associations of dietary vitamins B 1 and B 3 intakes with risk of CVD. Previous studies indicated that food sources rich in vitamins B 1 and B 3 such as fish/seafood and vegetables were associated with a reduced risk of mortality from CVD (14,15) . Yet, the evidence on dietary vitamin B complex/ CVD association was mainly directed to vitamins B 2 , B 6 , B 12 and folate, while the effects of dietary vitamins B 1 and B 3 intakes were not studied. Another issue is that most of the clinical trials on cardiovascular risk used high dosages of vitamin B 1 (200-300 mg/d) and B 3 supplement (1500-2000 mg/d), while the RDA of vitamin B 1 for Japanese men and women aged 50-69 years were 1·3 and 1·1 mg/d, respectively, and RDA of vitamin B 3 were 14 and 11 mg/d, respectively (16) . Among Japanese men aged 30-49 years, the estimated average requirement and RDA were 1·2 and 1·4 mg/d, respectively, for vitamin B 1 , and those for vitamin B 3 were 13 and 15 mg/d, respectively. Among Japanese women aged 30-49 years, the corresponding estimated average requirement was 0·9 and 1·1 mg/d, and RDA was 10 and 12 mg/d. The estimated average requirement and RDA of vitamins B 1 and B 3 in Japanese men and women aged over 70 years were even less than any other age groups (16) . Not all individuals prefer or can afford vitamin supplements for their health; thus, improving the dietary vitamins intakes is more achievable and acceptable by the general population. Therefore, after established effects of supplementary intakes of vitamins B 1 and B 3 have been determined, studying the cardiovascular impacts of dietary intakes of vitamins B 1 and B 3 is now warranted. Owing to the research gap in the field of epidemiology and similar food sources of vitamins B 1 and B 3 , their associations with CVD mortality were hypothesised in the present study. Therefore, we aimed to investigate the associations of dietary vitamins B 1 and B 3 intakes with risk of CVD mortality, which was considered as a proxy of CVD incidence risk, among Japanese men and women using the Japan Collaborative Cohort (JACC) study, a nationwide, communitybased prospective cohort study.

Study population and baseline data
Under the sponsorship of the Ministry of Education, Sports, and Science, the JACC study had the baseline survey (1988-1990) of 110 585 Japanese men (n 46 395) and women (n 64 190) aged 40-79 years from forty-five areas all over Japan. A detailed cohort profile of the JACC study was published previously (17) . Data on the baseline lifestyle and participants' characteristics, including demographic data, medical history of chronic diseases, diabetes mellitus, hypertension, smoking, alcohol consumption, exercise, diet and other items, were compiled via a self-administered questionnaire (online Supplementary Methods). The questionnaire included a validated forty-food item/FFQ which was distributed in thirty-two areas; therefore, we started with 86 401 subjects from those thirty-two areas. After the exclusion of non-respondents to FFQ (n 24 614), we further excluded those who reported a medical history of CVD or cancer (n 3142) and those who had implausible energy intakes defined as outliers of mean ± three standard deviations (n 343). Finally, a total of 58 302 individuals were eligible for the present study (22 989 men and 35 313 women) (online Supplementary Fig. S1). Written informed consent was acquired from community leaders or the individuals. The protocol of JACC study was approved by the Medical Ethical Committees of Nagoya University School of Medicine.

Dietary intake assessment
The participants were required to choose one from five frequency responses to describe the usual consumption frequency of forty food and beverage items over the past 12 months without specification of the portion size. The five responses were rarely, 1-2 times/month, 1-2 times/week, 3-4 times/week and almost every day. These frequencies were transformed into weekly consumption scores of 0, 0·38, 1·5, 3·5 and 7·0 per week, respectively (17,18) . A validation study among eighty-five individuals using four 3-d weighed dietary records over a 1-year period as a reference standard determined the portion size for each food and validated the FFQ intakes. The amount of nutrients in each food was calculated by multiplying the weekly consumption scores by the estimated portion size. The values of vitamins B 1 and B 3 and other nutrients from each food category were calculated according to the Standardised Tables of Food Composition, 5th revised version (19) which listed the nutrients content in 100 mg of different foods. Thus, the total vitamins B 1 and B 3 intakes were calculated by summing their intakes from all over the foods in the FFQ. The details of computation of nutrient intakes from FFQ (18) and the accuracy of food composition tables in Japan (20,21) were published previously. The Spearman rank correlation coefficients for vitamins B 1 and B 3 intakes between the FFQ and the four 3-d dietary records were 0·36 and 0·32, respectively, after energy adjustment (22) . The energy-adjusted mean ± standard deviation intakes in mg/d from weighed dietary record and FFQ were 1·08 (SD 0·20) and 0·71 (SD 0·20) for vitamin B 1 , but the respective values for vitamin B 3 were not reported (18) .

Mortality surveillance
The investigators annually or biannually confirmed the dates and causes of death in each area (17) . The International Classification of Diseases, 10th revision (ICD10) codes were applied to determine the underlying causes of death. In this study, our primary outcome was the total CVD mortality (ICD I01-I99). Cause-specific outcomes included mortalities from total stroke (ICD I60-I69), haemorrhagic stroke (ICD I60-I61), ischaemic stroke (ICD I63.0-I63.9), ischaemic heart disease (ICD I20-I25), myocardial infarction (ICD I20) and heart failure (ICD I50). This death certificate ascertainment was applied to all deaths within our cohort except for deaths that occurred outside of the original resident areas, which were treated as censored cases.

Statistical analysis
Energy-adjusted dietary intakes of vitamins B 1 and B 3 were categorised into five categorical groups (quintiles). The significance of differences in means or proportions of participants' characteristics and known risk factors of CVD in each quintile was tested by the ANCOVA and χ 2 test.
Person-years of follow-up were calculated from the baseline in 1988-1990 to their first endpoint in this follow-up as follows: death, moving out or the end of follow-up, whichever came first. The follow-up for mortality from CVD was conducted until 31 December 2009 in general; however, in four areas the followup was stopped until 31 December 1999, in another four areas until 31 December 2003 and in two areas until 31 December 2008 (17) . The Cox proportional hazard model was applied to calculate crude and multivariable-adjusted hazard ratios and 95 % CI for risk of mortality from CVD during the follow-up period ) across quintiles of dietary vitamins B 1 and B 3 intakes. We confirmed no violation of the Cox proportional hazard assumption because there were no significant interactions between the categorical rank variables of dietary vitamins B 1 and B 3 intakes and a function of survival time for all the tested outcomes. Multiplicative interactions of vitamins B 1 and B 3 with sex were tested to decide on presenting the data sex specifically or for combined men and women. The hypothesised confounders included age, sex, medical history of hypertension and diabetes, smoking status, ethanol intake, hours of sports, hours of walking, quintiles of BMI, years of education, perceived mental stress, daily utilisation of multivitamin supplementation, energyadjusted quintiles of Na and SFA intakes and quintiles of total energy intake. Details of these factors are given in online Supplementary Methods.
We assigned the median values to each quintile of vitamins B 1 and B 3 and tested their significance to calculate the trends across quintiles of vitamins B 1 and B 3 intakes. We further conducted a sensitivity analysis by excluding those who died within first 3 years of follow-up to avoid potential as-yet-undiagnosed diseases at baseline. All probability values for statistical test were two-tailed, and P < 0·05 was regarded as statistically significant. We applied the SAS statistical package (Version 9.4; SAS Institute Inc.) for statistical analysis. Table 1, participants in the highest quintile of both vitamins B 1 and B 3 intake were older, less educated, under less mental stress, had more walking time, had higher BMI and were less likely to be current smoker and to have a history of hypertension or diabetes. They also used multivitamin supplementation less frequently and consumed less alcohol but consumed more Na, SFA and total energy when compared with those in the lowest quintile. In this study, sources of vitamin B 1 were 31 % from pork, 17 % from vegetables, 10 % from fish and 7 % from potatoes, while sources of vitamin B 3 were 43 % from fish, 13 % from vegetables, 8 % from pork, 7 % from coffee and 6 % from green tea (data not shown in tables).

As shown in
Since no interaction with sex was observed for the association of vitamins B 1 and B 3 with CVD and specific endpoints, we combined the results of men and women in the main analyses. During 960 225 person-years of follow-up for 58 302 participants, we documented a total of 3371 deaths due to CVD, among whom there were 1504 deaths due to stroke (549 of which were due to haemorrhagic stroke and 816 of which were due to ischaemic stroke), 699 deaths were due to ischaemic heart disease (including 524 deaths due to myocardial infarction) and 564 deaths were due to heart failure.
There were 456 participants who died within the first 3 years of follow-up, and excluding those subjects yielded no substantial changes in the associations of vitamins B 1 and B 3 with mortality from ischaemic heart disease and myocardial infarction (online Supplementary Table S1).

Discussion
In this large community-based prospective cohort study of Japanese men and women, higher dietary intakes of vitamins B 1 and B 3 were associated with reduced risks of mortality from total CVD, ischaemic heart disease and myocardial infarction. Neither dietary vitamin B 1 nor vitamin B 3 intake was associated with the mortality risk of stroke, except for a tendency towards a reduced risk of haemorrhagic stroke with a higher dietary vitamin B 3 intake. Moreover, a higher dietary intake of vitamin B 1 was associated with a reduced risk of heart failure.
As far to our knowledge, the present study is the first to investigate associations of dietary vitamins B 1 and B 3 intakes with risk of CVD mortality despite the abundant evidence from animal studies and clinical trials on vitamins B 1 and B 3 supplements. Vitamins B 1 and B 3 in animal studies and human clinical trials showed protective effects against myocardial ischaemia.
One study on dogs showed that administration of vitamin B 1 decreased the metabolic needs of the heart, which was manifested as reduced myocardial oxygen consumption, mean peripheral pressure and left ventricular pressure up to 45, 25 and 10 % respectively (11) . A clinical trial on ten healthy adults and ten type 2 diabetes patients reported improvements in the brachial artery vasoactivity and the endothelium-dependent vasodilatation in both groups after a week of daily intravenous administration of 100 mg of vitamin B 1 (23) . Another randomised, cross-over and investigator-blinded trial on twenty adult healthy volunteers indicated the flow-mediated dilatation of the brachial artery was reduced by 50 % of its baseline diameter after smoking one cigarette, and the reduction in the flow-mediated vasodilatation with smoking one cigarette was only 25 % when 1050 mg/ d oral benfotiamine was administered for 3 d before the experiment (24) .
Another meta-analysis of thirteen trials of vitamin B 3 treatment demonstrated a significant reduced risk of non-fatal myocardial infarction (risk ratio: 0·85; 95 % CI 0·73, 1·00), a weak association with CVD mortality (risk ratio: 0·91; 95 % CI 0·81, 1·02) and no association with stroke (risk ratio: 0·89; 95 % CI: 0·72, 1·10) (27) . The mechanisms by which vitamins B 1 and B 3 might be protective against CVD mortality, especially those from ischaemic heart disease could be summarised here. Vitamin B 1 deficiency was highly prevalent in patients with type 2 diabetes (28) , which is considered as one of the risk factors for ischaemic heart disease. In addition, vitamin B 1 inhibits human infragenicular accelerated proliferation of arterial smooth muscle cells and mitigates atherosclerosis and endothelial dysfunction (29) . Another potential mechanism might be the protective effects of vitamin B 1 against ischaemic injury via reducing the metabolic needs of heart (5) . For vitamin B 3 , the reduced CVD risk may be involved in the favourable effects of vitamin B 3 on lipid metabolism (7,8) . Vitamin B 3 also has anti-inflammatory properties demonstrated by lowering C-reactive protein lipoprotein-associated phospholipase A2, inhibiting pro-atherogenic chemokines and enhancing serum levels of adiponectin (26) . Moreover, an antihypertensive effect of vitamin B 3 was also suggested (30) . We observed that a higher dietary vitamin B 1 intake was associated with reduced risk of mortality from heart failure. Vitamin B 1 deficiency was commonly considered to be correlated with a failing heart. In a meta-analysis of nine observational studies, the prevalence of vitamin B 1 deficiency was higher with an OR of 2·5 (95 % CI 1·7, 3·9) in heart failure group than in control (31) . Also known as wet beriberi or cardiac beriberi, vitamin B 1 deficiency was characterised by peripheral neuropathy and muscle weakness resulting in heart failure (3) . The vitamin B 1 deficiencyrelated heart failure was attributed to the vitamin B 1 role in energy metabolism (3) . Some studies reported that vitamin B 1 supplementation had beneficial effects on cardiac function (6) , but the evidence is still inconclusive (3) . Adjusted for age, sex, and socio-economic status (educational status). 2 Adjusted for age, sex, socio-economic status (educational status), and health behaviours (hours of walking, hours of sports, ethanol intake and smoking status). 3 A full mode with adjustment for age, sex, educational status, hours of walking, hours of sports, ethanol intake, smoking status, history of hypertension and diabetes, BMI, perceived mental stress, multivitamin supplementation, quintiles of energy-adjusted Na and SFA intakes and total energy intakes.

Vitamins B 1 and B 3 intakes and CVD mortality 1217
The dietary intake of vitamin B 3 in our study tended to associate with a lower risk of mortality from haemorrhagic stroke. In a case-control study including sixty-nine stroke cases and sixtynine controls, vitamin B 3 was found to be inversely correlated with risk of stroke (OR: 0·17; 95 % CI 0·04, 0·82) (32) . However, some meta-analyses concluded that vitamin B 3 had similar protective effects on both CHD and stroke outcomes (9,25) , while others failed to find protective effects against stroke (26,27) . The available studies and meta-analyses did not comment on the effect of vitamin B 3 on stroke subtypes. Vitamin B 3 was shown to reduce the blood pressure, an important risk factor for haemorrhagic stroke (30) . On the other hand, vitamin B 3 may promote vascular plasticity after an acute attack of stroke. In an animal experiment, Niaspan treatment of stroke in rats with diabetes promoted vascular remodelling and improved functional outcome (33) . Therefore, an impact of vitamin B 3 on risk of haemorrhagic stroke mortality could be plausible.
To the best of our knowledge, this is the first study to investigate the association of dietary intakes of vitamins B 1 and B 3 with the risk of mortality from CVD among Japanese population. The JACC study is a large, nationwide, community-based, prospective Japanese cohort study. The large sample size allowed us to investigate the associations of quintile categories of dietary vitamins B 1 and B 3 intakes with the risks of type-specific cardiovascular mortality as well as total CVD in Japanese population. Other strengths of this study included the prospective study design, the utilisation of a validated FFQ, the consistent endpoint determination and the exclusion of participants with CVD and cancer before the starting point of follow-up.
Limitations of this study mainly originate from the dietary assessment. The one-time measurement of dietary intakes cannot completely represent the consumption of nutrients during a long-term follow-up. The exclusion of 18 428 non-respondents to FFQ might result in a selection bias. Compared with 24 614 non-respondents to FFQ, the 61 787 respondents were more likely to be young and more educated (online Supplementary  Table S2). Several research of the JACC study reported the underestimation of nutrients intakes, which could be attributed to the limited number of food items in the used FFQ. Second, we had no data about the exact amounts or types of vitamin supplementation. To our knowledge, in the past century, vitamin supplementation was not common among Japanese population; thus, we believe that it would not affect the result substantially. In this study, approximately 88 % of participants did not use any vitamin supplementation and only 3 % used it on daily basis. The exclusion of daily supplementation uses did not change the result. Third, in Japan, the accuracy of heart failure death certificate diagnosis is a questionable issue. It is generally believed that heart failure death was overestimated before 1994, because most deaths of unknown origin such as cardiac arrest or arrhythmic death were more likely diagnosed as unspecific heart failure (34) . Therefore, approximately 27-50 % heart failure deaths were accounted for this misclassification (34) . Lastly, we did not have data on biomarkers of atherosclerosis and endothelial dysfunction, lipid metabolism and systematic inflammation such as C-reactive protein, lipoprotein-associated phospholipase A2, pro-atherogenic chemokines or serum levels of adiponectin and cannot determine all confounding effects from some other nutrients, lifestyles and socio-economic factors.

Conclusions
In this prospective cohort study, higher dietary intakes of vitamins B 1 and B 3 were inversely associated with a reduced risk of mortality from CVD among Japanese men and women. Dietary intakes of these vitamins from their food sources are available, accessible, affordable, safe and more acceptable by the general population than supplementary intakes. Therefore, dietary intakes of food rich in these vitamins could be encouraged for decreasing the risk of CVD mortality. However, our findings warrant further studies in different populations.