OUP user menu

Healthy lifestyle behaviours and cardiovascular mortality among Japanese men and women: the Japan collaborative cohort study

Eri Eguchi, Hiroyasu Iso, Naohito Tanabe, Yasuhiko Wada, Hiroshi Yatsuya, Shogo Kikuchi, Yutaka Inaba, Akiko Tamakoshi
DOI: http://dx.doi.org/10.1093/eurheartj/ehr429 467-477 First published online: 14 February 2012


Aims To examine the combined impacts of healthy lifestyle behaviours on cardiovascular disease (CVD) in Asians.

Methods and results A total of 18 747 men and 24 263 women aged 40–79 without a history of stroke or coronary heart disease (CHD) at baseline in 1988–90 were followed up until 2006. Participants scored one point for each following lifestyle behaviour: consumption of fruits ≥1 intake per day, fish ≥1 intake per day, milk almost every day, exercise ≥5 h per week and/or walking ≥1 h per day, body mass index (BMI) of 21–25 kg/m2, alcohol intake <46.0 g per day, non-smoking, and sleep duration of 5.5–7.5 h per day. During 16.5 years of follow-up, there were 1907 deaths from total CVDs including 849 strokes and 402 CHDs. For both genders, persons with the highest scores had the lowest CVD mortality. The multivariable hazard ratios (95% confidence interval, population-attributable fraction) for the highest (7–8) vs. lowest (0–2) score categories were 0.35 (0.25–0.49, 52.3%) in men, and 0.24 (0.16–0.36, 44.6%) in women. Similar associations were found for stroke: 0.36 (0.22–0.58, 45.0%) in men and 0.28 (0.15–0.53, 43.4%) in women, and for CHD: 0.19 (0.08–0.50, 76.2%) and 0.20 (0.09–0.47, 34.5%), respectively.

Conclusion Mortality from stroke, CHD, and CVD in the highest healthy lifestyle score category was one-third in men and one-fourth in women of those in the lowest scores, suggesting that a large fraction of CVD could be prevented through lifestyle modification.

  • Mortality
  • Stroke
  • Coronary heart disease
  • Cardiovascular disease
  • Lifestyle and risk factor

See page 428 for the editorial comment on this article (doi:10.1093/eurheartj/ehr446)


Lifestyle behaviours such as the consumption of fruits,1 fish2 and milk3, moderate physical activity,4 having a normal body mass index (BMI),5 moderate alcohol intake,6 not currently smoking,7 and moderate sleep duration8 were associated with lower mortality from cardiovascular disease (CVD) in the Japan Collaborative Cohort (JACC) study. We have also documented that the associations between a combination of these healthy lifestyle behaviours lead to a lower mortality from total death,9 and a longer life expectancy.10 While many studies have shown inverse associations between these lifestyle behaviours and CVD mortality,1118 few studies have investigated the impact of the combination of these behaviours on CVD risk until now. These results emphasize the importance of public health campaigns targeting lifestyle modification in CVD prevention.

Stampfer et al.19 showed that a combination of five healthy lifestyle behaviours was associated with a 83% reduction in the incidence of coronary heart disease (CHD), and a 75% reduction in the incidence of CVD among American women. Similarly, Chiuve et al.20 have shown a 69–79% reduction in the incidence of stroke among American men and women; and Myint et al.21 demonstrated that a combination of four bad lifestyle behaviours was associated with a 45–240% increase in the incidence of stroke among European men and women.

To our knowledge, no prospective study has reported a combined effect of healthy lifestyle behaviours among CVD in Asian populations until now. In this large prospective cohort study, we developed a healthy lifestyle score based on eight healthy lifestyle behaviours, and examined the potential magnitude of the combined impact of them on mortality from stroke, CHD, and total CVD among Japanese men and women aged 40–79.


Study population

A baseline survey of the JACC study was conducted in 1988–90. A total of 110 792 subjects (46 465 men and 64 327 women) aged 40–79 years in 45 areas across Japan completed self-administered questionnaires, including lifestyle information and medical histories pertaining to CVD and cancer. The details of the study procedure have been described previously.18,2225 Of all subjects, 2576 men and 3288 women were excluded due to positive histories of stroke, CHD or cancer, as were another 25 142 and 36 776, respectively, due to missing information for calculating the healthy lifestyle score. A total of 43 010 men and women (18 747 and 24 263, respectively) were eligible for the study. There was no substantial difference in CVD risk factors between individuals with complete healthy lifestyle scores and those with missing.

Mortality surveillance

The cause and date of death among participants were identified by reviewing all death certificates in each area. According to the International Classification of Diseases, the 10th revision, cause-specific mortality was determined in terms of stroke (I60–I69), CHD (I20–I25), and total CVD (I01–I99). By 31 December 2006, except for several areas where follow-up was terminated at the end of 1999 (six areas) and 2003 (five areas), a total of 6644 subjects had been censored because of death, and 2158, because of moving out. The median follow-up period was 16.5 (inter-quartiles: 12.8–17.8) years. This study was approved by the Ethics Committees of the Nagoya University School of Medicine, the University of Tsukuba, and Osaka University.

Healthy lifestyle score

Consumption of fruits, fish and milk

Possible responses were ‘rarely, 1–2 days a month, 1–2 days a week, 3–4 days a week and almost every day’ during the preceding year. For fruit (citrus fruits, other fruits, and fresh fruit juice) and fish, each frequency weight was set at 0, 0.05 (1.5/30), 0.214 (1.5/7), and 0.5 (3.5/7) with 1.0 signifying the response ‘almost every day’ and divided by three. We allocated 1 point each for fruits ≥1 intake/day (≥7 intakes weekly), fish ≥1 intake/day (≥7 intakes weekly), and milk almost every day according to the evidence in a previous report.13 A moderate reliability and validity for these measures have been reported elsewhere.22

Walking and sports participation

For walking, possible responses were: ‘almost never, 0.5, 0.5–1, and 1 h or more’ on average on a daily basis. For sport, possible answers were: ‘almost never, 1–2, 3–4, and 5 h or more’ on average on weekly basis. We allocated 1 point for those who walk ‘>0.5 to <1 h’, and ‘1.0 h and more’ per day, and/or those who participate sports ‘5 h or more’ per week according to the evidence in previous reports 4. The reliability and validity for these measures have been mentioned elsewhere.4,23

Body mass index

Self-reported weight (kg) was divided by the square of self-reported height (m).

According to the evidence in previous reports,5 one point was allocated to the category of BMI 21.0 to <25.0 kg/m2.

Smoking status

Categories were smokers, non-smokers including past smokers. One point was allocated to non-smoker.

Alcohol consumption

Categories were non-drinkers, ex-drinkers, or current drinkers. Drinkers also reported the frequency as ‘less than once, 1–2 times, 3–4 times, or almost every day’ of alcohol consumption per week and the amount per occasion during the previous year. The daily amount of ethanol consumption was calculated. We allocated 1 point for all of non-drinkers, ex-drinkers, and current drinkers of 1–46.0 g ethanol/day according to the evidence in previous reports.6

Sleep duration

Average sleep duration on weekdays during the preceding year was classified into seven categories: <4.5, 4.5–5.4, 5.5–6.4, 6.5–7.4, 7.5–8.4, 8.5–9.4, and ≥9.5 h per day.

We allocated 1 point for sleeping duration of 5.5–7.4 h/day according to the evidence in previous reports.8

The allocated points of lifestyle behaviours were totalled for a healthy lifestyle score ranging from 0 to 8. We showed the components of the score in Supplementary material online, Table S1. The scores were grouped into six categories (0–2, 3, 4, 5, 6, 7–8 points) for analyses to keep the number both in men and women balanced in each category. Also, we ran a sensitivity analysis combining both men and women without categorization of the score.

For consideration of the magnitude of the impact of individual lifestyle behaviour, we also constructed a weighted score according to the magnitude of individual hazard ratios (HRs).

Statistical analysis

We first calculated sex-specific age-adjusted mean values and prevalence of risk factors by analysis of covariance, and examined the linear trend by a regression model for adjusted mean values and by a multiple logistic regression model for adjusted prevalence using the median values of healthy lifestyle score as representative values of the score categories.

To investigate sex-specific effects on CVD, we conducted the analyses among men and women separately. Age- and multivariate-adjusted individual impacts and population-attributable fraction (PAF) of the components of the score on mortality from CVD were calculated by Cox proportional hazard models and standard PAF calculation methods26. Covariates for multivariable-adjusted analysis were: age (years), histories of hypertension and diabetes mellitus (yes or no), education level (attended school until 13, 13–15, 16–18, or beyond 19 years old), perceived stress (high, medium, or low), and regular employment (yes or no) and components of healthy lifestyle score other than specified variable. Since the effect of smoking is the largest among men, we repeated the analysis stratified by smoking status: current smoker or non-smoker; and examined the effect modification of each lifestyle behaviour and smoking status using cross-product terms of each lifestyle variable and smoking status in the Cox proportion hazard model.

We calculated age- and multivariable-adjusted HRs and 95% confidence intervals (CI) to determine the sex-specific and smoking status-stratified associations between combination of the lifestyle behaviours and the risk of mortality from stroke, CHD, and total CVD during the follow-up period. For multivariable HR, we used the same covariates listed above. Linear trends were examined by a Cox proportional hazard model. We also calculated sex-specific age-adjusted average annual mortality rate for CVD according to score categories using the direct standardization method using the age distribution of national model population in 1990. The smoking status-stratified analysis was conducted using seven healthy lifestyle behaviours, after excluding smoking status. The reference categories were 0–2 for sex-stratified, or 0–1 for smoking status-stratified analysis, respectively. The age-adjusted mortality rate from CVD was also constructed. Effect modifications by sex or smoking status were tested separately by an interaction terms of either sex or smoking status by healthy lifestyle score.

PAF was calculated by the following method9,26: Embedded Image where Pi is the prevalence of score i at baseline, and HRi is the age-adjusted HR for the score i compared with a category of 7–8. In addition, PAF+1 was calculated as Embedded Image where HRi and HRi+ 1 are the age-adjusted HRs of an category for score i and score i+ 1, respectively, compared with a category of 7–8.

We repeated the same analysis using a weighted score27. Each was calculated by: Embedded Image where HR is individual multivariable HRs of CVD mortality according to each lifestyle behaviour. Weights were summed up as total weighted healthy lifestyle score for men and women, and divided into six categories.

Also, we constructed sex- and smoking status-stratified Kaplan–Meier's survival curves for men and women. To examine the possibility of reverse causation, the same analyses were repeated after excluding early deaths that occurred within 5 years from the baseline.

We used the 9.1.3 version of SAS for all statistical analyses. All probability values for statistical tests were two-tailed, and P< 0.05 were regarded as statistically significant.


The mean age in years of the participants was 55.6 for men and 56.1 for women. During the 16.5-year follow-up period, there were 849 deaths from stroke (441 men and 408 women), 402 from CHD (240 and 162), and 1907 from total CVD (1012 and 895). In total, men had a 20% higher mortality from CVD than women, with an age-adjusted annual mortality rate of 5.13 deaths per 1000 in men and 4.16 in women. The rate was highest in the healthy lifestyle score category of 0–2 points for both genders (6.47 and 9.19 among men and women), and was lowest in the score category of 7–8 points (2.91 and 2.28 deaths, respectively).

Table 1 shows sex-specific age-adjusted mean values or prevalence of risk factors at baseline by healthy lifestyle score categories. The respective percentages of 0–2, 3, 4, 5, 6, and 7–8 points of the score were 13, 19, 25, 22, 14, and 6% in men and 1, 6, 17, 27, 28, and 20% in women. Compared with people of the reference 0–2 category, those of higher categories were more likely to be older, hypertensive, higher educated, with higher perceived mental stress, and regular job among men, and, more likely to be younger, hypertensive, and higher educated among women.

View this table:
Table 1

Mean age and age-adjusted prevalence of cardiovascular risk factors by healthy lifestyle score

Healthy lifestyle score (points)
0–234567–8P for trend
 No. at risk (%)2506 (13)3566 (19)4664 (25)4209 (22)2611 (14)1191 (6)
  Age (years)545555565758<0.0001
  Fruits ≥1/day (%)133453718495
  Fish ≥1/day (%)132535476080
  Milk almost everyday (%)92036516987
  Habitual exercise or walking (%)456271798594
  Body mass index 21–25 kg/m2 (%)233850617388
  Ethanol intake <46.0 g/day (%)355768788895
  Non-smoker (%)132741567489
  Sleep 5.5–7.4 h/day (%)193647566787
  History of hypertension (%)2119181818170.01
  History of diabetes (%)6666660.88
  College or higher education (%)161619192229<0.0001
  High perceived mental stress (%)2625252526290.01
  Regular employment (%)7577777779790.002
 No. at risk (%)325 (1)1534 (6)4041 (17)6552 (27)6886 (28)4925 (20)
  Age (years)575857565655<0.0001
  Fruits ≥1/day (%)72648708697
  Fish ≥1/day (%)3716314877
  Milk almost everyday (%)3920375884
  Habitual exercise or walking (%)184058708293
  Body mass index 21–25 kg/m2 (%)71426405782
  Ethanol intake <46.0 g/day (%)87979999100100
  Non-smoker (%)5782929698100
  Sleep 5.5–7.4 h/day (%)112542567188
  History of hypertension (%)2120211919180.03
  History of diabetes (%)5443330.26
  College or higher education (%)889101213<0.0001
  High perceived mental stress (%)2222212121210.93
  Regular employment (%)3633333434360.09
  • —, Statistical testing was not conducted because of a component of the healthy lifestyle score.

Table 2 shows sex-specific and smoking-stratified HRs and 95% CIs and prevalence for individual health behaviours. In both sexes, people with each healthy lifestyle behaviour had the lower risk of mortality from total CVD compared with those without it. Multivariable HR was lowest for non-smoker in men and ethanol intake <46.0 g/day, and non-smoker in women. The PAFs and weights calculated according to the corresponding HRs were 3.3% and 1.0 for fruit ≥1/day, 4.4% and 1.0 for fish ≥1/day, 7.5% and 1.8 for milk almost every day, 5.2% and 2.5 habitual exercise or walking, 11.9% and 3.6 for BMI 21–25 kg/m2, 5.3% and 2.3 for ethanol intake <46.0 g/day, 2.5% and 5.1 for non-smoker, 8.6% and 2.5 for sleep 5.5–7.4 h/day for men, 5.7, 8.1, 2.2, 4.1, 15.9, 0.7, 2.3 and 8.7% for PAFs, and 20.0, 2.9, 2.0, 0.6, 2.2, 4.5, 8.9, 5.8, 3.2 for weights for women, respectively. The total score of weights were 20.0 for men and 30.1 for women.

View this table:
Table 2

Sex-specific and smoking status-stratified hazard ratios and 95% confidence intervals of cardiovascular disease mortality for individual health behaviours

Prevalence of healthy behaviour (%)TotalSmoking statusP for interaction
 Person-years270 105146 534123 572
  Fruits ≥1/day vs. others, n55.2560/452304/262256/190
   Age-adjusted HR (95% CI)0.86 (0.77–0.97)0.97 (0.82–1.14)0.77 (0.64–0.94)0.03
   Multivariable HR (95% CI)0.93 (0.82–1.06)0.98 (0.83–1.17)0.87 (0.72–1.06)0.39
  Fish ≥1/day vs. others, n39.2416/596239/327177/269
   Age-adjusted HR (95% CI)0.91 (0.80–1.03)0.99 (0.84–1.17)0.82 (0.68–0.99)0.03
   Multivariable HR (95% CI)0.93 (0.82–1.05)1.00 (0.84–1.18)0.82 (0.68–1.00)0.15
  Milk almost every day vs. others, n40.8406/606218/348188/258
   Age-adjusted HR (95% CI)0.81 (0.72–0.92)0.97 (0.82–1.15)0.70 (0.58–0.85)0.06
   Multivariable HR (95% CI)0.88 (0.77–1.00)0.98 (0.83–1.17)0.76 (0.63–0.92)0.04
  Habitual exercise or walking vs. others, n71.0726/286403/163323/123
   Age-adjusted HR (95% CI)0.83 (0.72–0.95)0.78 (0.65–0.94)0.87 (0.71–1.07)0.07
   Multivariable HR (95% CI)0.84 (0.73–0.97)0.79 (0.66–0.96)0.89 (0.72–1.10)0.24
  Body mass index 21–25 kg/m2 vs. others, n52.2444/568247/319197/249
   Age-adjusted HR (95% CI)0.77 (0.68–0.87)0.81 (0.68–0.95)0.75 (0.62–0.91)0.12
   Multivariable HR (95% CI)0.78 (0.69–0.88)0.80 (0.67–0.94)0.75 (0.62–0.90)0.51
  Ethanol intake <46.0 g/day vs. others, n68.3712/300365/201347/99
   Age-adjusted HR (95% CI)0.77 (0.67–0.89)0.82 (0.69–0.98)0.80 (0.63–1.00)0.22
   Multivariable HR (95% CI)0.85 (0.74–0.97)0.85 (0.71–1.01)0.86 (0.68–1.08)0.57
  Sleep 5.5–7.4 h/day vs. others, n48.5363/649204/362159/287
   Age-adjusted HR (95% CI)0.78 (0.69–0.89)0.82 (0.69–0.98)0.77 (0.63–0.93)0.07
   Multivariable HR (95% CI)0.84 (0.73–0.95)0.85 (0.71–1.01)0.81 (0.67–0.99)0.37
  Non-smoker vs. others, n45.6446/566566446
   Age-adjusted HR (95% CI)0.69 (0.61–0.78)
   Multivariable HR (95% CI)0.70 (0.62–0.80)
 Person-years359 19716 222342 974
  Fruits ≥1/day vs. others, n72.7601/29436/28565/266
   Age-adjusted HR (95% CI)0.76 (0.66–0.88)0.77 (0.47–1.26)0.77 (0.67–0.89)0.97
   Multivariable HR (95% CI)0.82 (0.71–0.95)0.75 (0.45–1.26)0.83 (0.71–0.96)0.79
  Fish ≥1/day vs. others, n41.0330/56521/43309/522
   Age-adjusted HR (95% CI)0.81 (0.71–0.93)1.04 (0.61–1.75)0.81 (0.70–0.93)0.27
   Multivariable HR (95% CI)0.87 (0.76–1.00)0.98 (0.57–1.71)0.86 (0.74–0.99)0.40
  Milk almost every day vs. others, n47.2393/50226/38367/464
   Age-adjusted HR (95% CI)0.88 (0.77–1.01)1.25 (0.76–2.05)0.87 (0.76–1.00)0.20
   Multivariable HR (95% CI)0.96 (0.84–1.10)1.22 (0.72–2.06)0.95 (0.82–1.09)0.52
  Habitual exercise or walking vs. others, n73,6637/25839/25598/233
   Age-adjusted HR (95% CI)0.84 (0.73–0.98)0.53 (0.32–0.87)0.88 (0.76–1.03)0.04
   Multivariable HR (95% CI)0.86 (0.74–0.99)0.51 (0.30–0.86)0.90 (0.77–1.05)0.02
  Body mass index 21–25 kg/m2 vs. others, n48.7315/58018/46297/534
   Age-adjusted HR (95% CI)0.71 (0.62–0.81)0.77 (0.44–1.34)0.71 (0.62–0.82)0.50
   Multivariable HR (95% CI)0.73 (0.64–0.84)0.65 (0.37–1.17)0.73 (0.63–0.84)0.72
  Ethanol intake <46.0 g/day vs. others, n99.2886/960/4826/5
   Age-adjusted HR (95% CI)0.41 (0.21–0.78)0.49 (0.17–1.38)0.54 (0.22–1.30)0.53
   Multivariable HR (95% CI)0.54 (0.28–1.05)0.50 (0.17–1.48)0.63 (0.26–1.52)0.64
  Sleep 5.5–7.4 h/day vs. others, n61.9358/53724/40334/497
   Age-adjusted HR (95% CI)0.76 (0.66–0.87)0.66 (0.39–1.11)0.77 (0.67–0.89)0.96
   Multivariable HR (95% CI)0.80 (0.69–0.91)0.58 (0.33–0.99)0.81 (0.71–0.94)0.65
  Non-smoker vs. others, n95.3831/6464831
   Age-adjusted HR (95% CI)0.57 (0.44–0.74)
   Multivariable HR (95% CI)0.67 (0.52–0.87)
  • Multivariable adjustment: age, history of hypertension, history of diabetes, education level, regular employment, perceived mental stress, and seven health behaviours other than specified variable.

In the smoking status-stratified analysis, the protective effects of each health behaviour were generally larger among non-smokers than smokers in men. Also, for women, most of the associations remained statistically significant among non-smokers. The smoking interactions between each healthy lifestyle behaviour with total CVD mortality were statistically significant for fruit and fish consumptions, and borderline significance for milk consumption, habitual exercise or walking, and sleep duration among men.

Table 3 shows sex-specific age- and multivariable-adjusted HRs of mortality and crude mortality rates from stroke, CHD, and total CVD according to healthy lifestyle score. In both men and women, age-adjusted risks were lower with higher healthy lifestyle score in a graded fashion. These inverse associations remained statistically significant after adjusting for other risk factors. The multivariable HRs (95% CI) of mortality from stroke, CHD, and total CVD for 7–8 point category compared with the 0–2 were 0.36 (0.22–0.58), 0.19 (0.08–0.50), and 0.35 (0.25–0.49), respectively, for men, and were 0.28 (0.15–0.53), 0.20 (0.09–0.47), and 0.24 (0.16–0.36), respectively, for women. The results of sensitivity analysis including both men and women without categorization were not substantially different from the results above. The PAF of stroke, CHD, and total CVD for the 7–8 category was 45.0, 76.2, and 52.3% in men, and 43.4, 34.5, and 44.6% in women. The respective PAF+1 towards the 7–8 point category was 25.8, 41.6, 26.4% in men and 25.2, 22.3, 25.9% in women.

View this table:
Table 3

Sex-specific and smoking status-stratified hazard ratios and 95% confidence intervals for cardiovascular disease mortality by healthy lifestyle score

Healthy lifestyle score (points)P for trend
 Person-years35 05451 10066 54961 27938 42917 695
   Age-adjusted mortality rate3.
   Age-adjusted HR (95% CI)1.000.85 (0.63–1.14)0.63 (0.47–0.85)0.45 (0.33–0.61)0.44 (0.31–0.62)0.35 (0.22–0.56)<0.0001
   Multivariable HR (95% CI)1.000.85 (0.63–1.14)0.64 (0.48–0.86)0.45 (0.33–0.62)0.46 (0.32–0.66)0.36 (0.22–0.58)<0.0001
  Coronary heart disease
   Age-adjusted mortality rate1.331.121.441.351.070.31
   Age-adjusted HR (95% CI)1.000.79 (0.50–1.26)0.93 (0.62–1.41)0.83 (0.54–1.26)0.60 (0.37–0.99)0.19 (0.07–0.49)0.0008
   Multivariable HR (95% CI)1.000.80 (0.51–1.27)0.94 (0.62–1.43)0.84 (0.55–1.28)0.62 (0.38–1.02)0.19 (0.08–0.50)0.0012
  Total cardiovascular disease
   Age-adjusted mortality rate6.476.255.424.624.112.91
   Age-adjusted HR (95% CI)1.000.88 (0.71–1.08)0.75 (0.62–0.92)0.59 (0.48–0.72)0.51 (0.40–0.64)0.34 (0.25–0.48)<0.0001
   Multivariable HR (95% CI)1.000.88 (0.71–1.08)0.76 (0.62–0.93)0.59 (0.48–0.73)0.52 (0.41–0.67)0.35 (0.25–0.49)<0.0001
Healthy lifestyle score (except for smoking status)
0–123456–7P for trend
Total cardiovascular disease
  Person-years865922 00137 66139 61226 59212 009
   Age-adjusted mortality rate7.185.996.165.895.203.78
   Multivariable HR (95% CI)1.000.88 (0.61–1.28)0.84 (0.60–1.20)0.81 (0.57–1.14)0.68 (0.47–0.98)0.54 (0.34–0.85)0.001
  Person-years439413 43826 93734 68828 20315 912
   Age-adjusted mortality rate7.656.154.644.084.002.98
   Multivariable HR (95% CI)1.05 (0.64–1.71)0.77 (0.52–1.15)0.59 (0.41–0.86)0.48 (0.34–0.69)0.47 (0.32–0.67)0.33 (0.21–0.51)<0.0001
 Person-years446021 46358 20796 888103 24674 934
   Age-adjusted mortality rate3.812.662.211.841.951.01
   Age-adjusted HR (95% CI)1.000.63 (0.35–1.13)0.52 (0.30–0.92)0.43 (0.25–0.76)0.42 (0.24–0.74)0.24 (0.13–0.44)<0.0001
   Multivariable HR (95% CI)1.000.68 (0.37–1.23)0.57 (0.32–1.00)0.48 (0.27–0.84)0.48 (0.27–0.84)0.28 (0.15–0.53)<0.0001
  Coronary heart disease
   Age-adjusted mortality rate2.
   Age-adjusted HR (95% CI)1.000.41 (0.18–0.93)0.41 (0.19–0.87)0.28 (0.13–0.59)0.24 (0.11–0.51)0.19 (0.08–0.44)<0.0001
   Multivariable HR (95% CI)1.000.42 (0.18–0.97)0.42 (0.19–0.89)0.29 (0.14–0.63)0.25 (0.11–0.54)0.20 (0.09–0.47)<0.0001
  Total cardiovascular disease
   Age-adjusted mortality rate9.195.574.854.174.002.28
   Age-adjusted HR (95% CI)1.000.55 (0.37–0.81)0.48 (0.34–0.70)0.42 (0.29–0.60)0.39 (0.27–0.56)0.22 (0.14–0.32)<0.0001
   Multivariable HR (95% CI)1.000.58 (0.39–0.86)0.51 (0.35–0.74)0.45 (0.31–0.65)0.42 (0.29–0.60)0.24 (0.16–0.36)<0.0001
Healthy lifestyle score (except for smoking status)
0–123456–7P for trend
Total cardiovascular disease
   Age-adjusted mortality rate26.3110.
   Multivariable HR (95% CI)1.000.27 (0.09–0.78)0.12 (0.04–0.34)0.15 (0.06–0.40)0.13 (0.04–0.39)0.20 (0.06–0.69)0.02
  Person-years265017 66353 40893 176101 37874 699
   Age-adjusted mortality rate9.195.574.854.174.002.28
   Multivariable HR (95% CI)0.18 (0.07–0.49)0.14 (0.06–0.35)0.12 (0.05–0.29)0.11 (0.04–0.26)0.10 (0.04–0.24)0.06 (0.02–0.14)<0.0001
  • Multivariable adjustment: age, history of hypertension, history of diabetes, education level, regular employment, and perceived mental stress.

For the smoking status-stratified analysis, both crude mortality and HRs were generally higher in smokers (Table 3). The multivariable HRs (95% CI) of mortality from total CVD for the highest category to the lowest was 0.54 (0.34–0.85) among smoking men and was 0.33 (0.21–0.51) among non-smoking men. The respective HRs among women were 0.20 (0.06–0.69) and 0.06 (0.02–0.14), respectively.

Sex-specific survival curves of mortality from total CVD are illustrated in Figure 1. The larger decline in survival rate was observed for the lower healthy lifestyle score categories compared with the higher among both men and women. Women showed more divergent curves of mortality from total CVD than did men. However, there were no significant interaction between sex and healthy lifestyle score with stroke, CHD, and total CVD: P-value for interaction = 0.58, 0.13, and 0.64, respectively.

Figure 1

Kaplan–Meier survival curves of mortality from total cardiovascular disease according to the healthy lifestyle score among men (A) and among women (B). The purple line shows the highest score category, and the black line shows the lowest.

Figure 2 illustrates the survival curves of mortality from CVD by smoking status among men. There was not substantial difference between smoker and non-smoker for first 10 years. However, there were significant interaction between the score and smoking status (P for interaction = 0.03). Supplementary material online, Figure S1 illustrates the result for women. The decline of the survival curves seemed more steep in smoking women, but there was no significant difference between smokers and non-smokers (P for interaction = 0.80).

Figure 2

Kaplan–Meier survival curves of mortality from total cardiovascular disease according to the healthy lifestyle score in men for smokers (A) and for non-smokers (B). The purple line shows the highest score category, and the black line shows the lowest.

After the exclusion of deaths within 5 years from baseline, the associations between the score and mortality did not change materially. The multivariable-adjusted HRs (95% CI) for healthy lifestyle scores of the highest compared with the lowest were 0.37 (0.22–0.63) for stroke, 0.21 (0.08–0.55) for CHD, and 0.38 (0.27–0.55) for total CVD among men, and 0.30 (0.15–0.60), 0.23 (0.10–0.56), and 0.27 (0.17–0.43), respectively, among women.

Lastly, the repeated analysis of association with CVD mortality conducted according to weighted healthy lifestyle score is shown in Supplementary material online, Table S2. The magnitude of impact on the highest score category compared with the lowest was smaller for all stroke, CHD, and CVD, but similar dose response trends with narrower CIs were observed.


Findings from this large-scale prospective study of Japanese men and women aged 40–79 years revealed graded inverse associations between the number of healthy lifestyle behaviours and mortality from stroke, CHD, and total CVD.

These associations have been documented previously among Western populations. In the Nurses’ Health study, 84 129 women were followed up for 14 years, and the HRs for CHD and CVD incidence were 0.17 (0.07–0.41) and 0.25 (0.14–0.44), respectively, for women who had five healthy lifestyle behaviours compared those with none of them.19 In the Health Professional Follow-up study, 42 847 men were followed up for 16 years, and the HR of CHD incidence was 0.13 (0.09–0.19) for men who had five healthy lifestyle behaviours when compared with those with none of them.27 Similarly, in the Nutrition Potsdam study, 23 153 men and women aged 35–65 years were followed up for 8–12 years, and the HRs were 0.81 (0.47–0.93) for myocardial infarction incidence, and 0.36 (0.05–0.57) for stroke for men and women together who had four (no score for alcohol consumption) healthy lifestyle behaviours.28 The components of the score of lifestyle behaviours of those three studies were more or less similar to those of our study, i.e. (Diet, exercise, BMI, alcohol consumption, and smoking status). Scoring was similar for the Nurse Health study and HPFS as follows, i.e. scoring with in the top 40% of healthy diet, moderate to vigorous exercise for ≥30 min per day, BMI of <25, alcohol consumption 5–45 g or 5–30 g per day, and non-smoking.19,27 For the Nutrition-Potsdam study, healthy behaviours were high consumptions of fruits, vegetables, whole grain bread, and low consumption of meat, physical activity >3.5 h per week, a BMI<30 kg/m2, and never smoking.28 However, in our study, we expanded the lifestyle behaviour to include fish and milk consumption and moderate sleep duration because of the emerging evidences on their CVD effects.2,3,8,12,13,18 The score for diet as a healthy lifestyle behaviour, of many of the previous studies19,21,27,28 was so widely expanded to include higher consumption of fruit, vegetable, nuts, soy, whole-grain bread, cereal fibre, multivitamin, marine n-3 fatty acids and folate, and polyunsaturated to saturated fat ratio, and lower intakes of meat, trans fat, glycaemic load and allocated one point for those who scored high. However, we allocated one point each for each of the three components of dietary intake, i.e. (fish, milk, and fruits) for practical purposes to be easily applied by the public. The healthy lifestyle score was categorized to 0–2, 3, 4, 5, 6, 7–8 for this study to balance the number in men and women in each category. However, the analysis without categorization did not change the result and the categorization did not seem to alter the results. Also, the results of the relationship between weighted healthy lifestyle score and CVD mortality did not make substantial difference from that of non-weighted score.

Our results showed that each lifestyle behaviours had impacts on CVD mortality as previous studies showed.18,1118 According to the estimation of PAF for each lifestyle, moderate BMI had the biggest impact on cardiovascular health in the present study.

Although several studies19,27,28 have focused on the risk for CVD among men only, women only, or men and women combined, few studies have addressed a sex difference in the impacts of lifestyle behaviours. In the study of 43 685 men and 71 243 women from the Health Professional Follow-up study and the Nurses’ Health study, the multivariable-adjusted HRs of stroke among people with the highest lifestyle score compared with the lowest were 0.31 (0.19–0.53) in men and 0.21 (0.12–0.36) in women.20 Moreover, in the EPIC Norfolk study, 20 040 men and women aged 40–79 years followed up for 10–14 years also show a graded inverse association between four lifestyle behaviours (plasma concentration of vitamin C ≥50 µmol/L, regular physical activity, moderate alcohol intake 1–112 g per week, and non-smoking status) and stroke mortality. The HRs were 1.48 (0.62–3.53) for men and 3.49 (1.71–7.12) for women in lowest lifestyle score categories compared with the highest.21 Both of these two studies20,21 demonstrated larger magnitude of risk reduction among women compared with men in Western countries. In our study, the HRs of total CVD mortality in each healthy lifestyle score category were 1.2–1.5-fold higher among men than among women. The survival curves of total CVD mortality were also less divergent across the healthy lifestyle score categories among men compared with women. Also, they were less divergent among smokers than non-smokers based on both relative risk and absolute risk analyses among men while smoking was generally associated with the higher crude mortality rate in each of the lifestyle score categories. These findings suggest that the protective effect of healthy lifestyle behaviours on morbidity from CVD was masked by smoking. It has been reported that smoking reduces the beneficial effect of physical activity29 and that vitamin C is effective to reduce hs-CRP levels, an emerging cardiovascular risk marker, only among non-smokers.30

In previous studies that have compared individuals in the low-risk category with the rest of population, the reported PAF has been 35% in men and 47% in women for stroke,20 62%27 in men and 82%19 in women for CHD, and 74%19 in women for total CVD. Although a simple comparison of PAF is not satisfactory, because of the different scoring methods of lifestyles among studies, all investigations including our study indicated a large potential for the prevention of CVD by the implementation of healthy lifestyle behaviours. Moreover, in the present study, the result of PAF+1 indicated that improving just one of these behaviours may translate into benefits equating to a one-fourth decrease in the mortality from total CVD for both men and women.

The strong points of our study were: first, the large-scale cohort with subjects from all over Japan which included >1900 deaths from total CVD; Second, the long follow-up period with a 16.5 year median; Third, the collection of a number of variables at baseline and the adjustment for potential confounding variables; and fourth, the sensitivity analysis with different categorization of the score, for the use of weighted score, and the exclusion of the first 5 years of death from the baseline. These advantages allowed us to estimate the impacts of healthy lifestyle behaviours on CVD mortality. The healthy lifestyle score we adopted was practical to understand and to calculate, and corresponded to possible lifestyle improvements. Thus, it motivates both individuals and health promoters for lifestyle improvement.

The followings are our study limitations. First, the baseline data were obtained at one time point only. Because of the possible changes in lifestyle over the follow-up period, non-differential measurement errors would have attenuated the observed associations, and the real associations might be stronger. Second, we used mortality data rather than incidence as the endpoint; episodes of CVD and/or other diseases may induce lifestyle changes and consequently mortality risk in some individuals. To reduce this effect at most, we excluded those with a history of stroke, CHD, and cancer at baseline, and excluded early deaths at baseline for the secondary analysis. Third, there might be a concern about selection bias as we excluded 61 918 participants who had missing information. The exclusion of these individuals, however, was unlikely to influence the relationships, since these baseline characteristics did not differ substantially from the rest of the cohort.

In conclusion, the number of baseline healthy lifestyle behaviours was inversely associated with the risk of mortality from CVD. Mortality risk from stroke, CHD, and total CVD in the highest healthy lifestyle score category was one-third among men and one-fourth among women, compared with those of the lowest score category. The inverse association among men was more evident for non-smokers than for smokers. These results suggest a large role of lifestyle modification in the prevention of CVD.


This work was supported by grants-in-aid for scientific research from the Ministry of Education, Science, Sports and Culture of Japan (Monbusho); 61010076, 62010074, 63010074, 1010068, 2151065, 3151064, 4151063, 5151069, 6279102, 11181101, 17015022, 18014011, and 20014026.

Conflict of interest: none declared.


The authors sincerely express their appreciation to Dr Kunio Aoki, Professor Emeritus, Nagoya University School of Medicine and the former chairman of the JACC study, and Dr Haruo Sugano, the former Director, Cancer Institute, Tokyo, who greatly contributed to the initiation of the JACC study, as well as Dr Yoshiyuki Ohno, Professor Emeritus, Nagoya University School of Medicine, who was the past chairman of the study. The authors also wish to thank Dr Tomoyuki Kitagawa, Cancer Institute of the Japanese Foundation for Cancer Research and the former chairman of Grant-in-Aid for Scientific Research on Priority Area ‘Cancer’ and Dr Kazao Tajima, Aichi Cancer Center Research Institute and the former chairman of Grant-in Aid for Scientific Research on Priority Area of Cancer Epidemiology for their full support of this study.


View Abstract