JEpidemiol 2009;19(1):24-27 doi:10.2188/jea.JE20080043 Original Article Physical Activity and All-cause Mortality in Japan: The Jichi Medical School (JMS) Cohort Study Shinya Hayasaka1, Yosuke Shibata1, Shizukiyo Ishikawa2, Kazunori Kayaba3, Tadao Gotoh2, Tatsuya Noda1, Chiyoe Murata1, Tomoyo Yamada1, Yasuaki Goto4, Yosikazu Nakamura5, and Toshiyuki Ojima1, for the Jichi Medical School Cohort Study Group 1DepartmentofCommunityHealthandPreventiveMedicine,HamamatsuUniversitySchoolofMedicine,Hamamatsu,Shizuoka,Japan 2DivisionofCommunityandFamilyMedicine,CenterforCommunityMedicine,JichiMedicalUniversity,Shimotsuke,Tochigi,Japan 3SchoolofHealthandSocialServices,SaitamaPrefecturalUniversity,Moroyama,Saitama,Japan 4JapanHealthandResearchInstitute,Tokyo,Japan 5DepartmentofPublicHealth,JichiMedicalUniversity,Shimotsuke,Tochigi,Japan ReceivedMay23,2008;acceptedAugust22,2008;releasedonlineJanuary22,2009 ABSTRACT Background: In April 2008, a new health check-up and health guidance system was introduced by the Japanese Government to promote increased physical activity. However, few studies have documented the health benefits of physicalactivityinAsianpopulations.Weexaminedtheassociationbetweenall-causemortalityandlevelofphysical activity in a Japanese multicommunity population-based study. Methods: TheJichiMedicalSchoolCohortStudyisamulticommunity,population-basedstudybasedin12districts inJapan.Baselinedatafrom4222menand6609women(meanage,54.8and55.0years,respectively)werecollected between April1992andJuly1995.The participantswerefollowedfor ameanduration of11.9years. Todetermine theassociationbetweenall-causemortalityandlevelofphysicalactivity,crudemortalityratesper1000person-years and hazard ratios (HRs) with 95% confidence intervals (CI) were determined using the Cox proportional hazards model. Physical activity was categorized by using physical activity index (PAI) quartiles. The lowest (first) PAI quartile was defined as the HR reference. Results: In men, the lowest mortality rate was observed in thethird quartile, with 95 deaths and a crude mortality rate of 7.6; the age- and area-adjusted HR was 0.59 (95% CI, 0.45–0.76), and the mortality curve had a reverse J shape. In women, the lowest mortality rate was observed in the highest PAI quartile, with 69 deaths and a crude mortality rate of 3.5; the HR was 0.81 (95% CI, 0.58–1.12). Conclusion: Our results suggest that increased physical activity lowers the risk for all-cause death in Japanese. Key words: all-cause mortality; cohort study; Japan; physical activity conducted in Western countries; only a few have been INTRODUCTION conducted within Asian populations. Furthermore, the few In April 2008, a new health check-up and health guidance studiesthathavebeenconductedinAsiawereperformedina system entitled, “Health Checkups and Healthcare Advice particular community or company; only rarely have multiple with Particular Focus on Metabolic Syndrome,” was communities been investigated. introduced by the Japanese Government. As part of the The purpose of this study was to examine the association system,allpeopleaged40to74arenowrequestedtoundergo between all-cause mortality and levels of physical activity in ahealthcheck-upandhealthguidanceaimedatdecreasingthe Japan using a multicommunity, population-based approach. incidence of metabolic syndrome and lifestyle-related diseases.1 As part of the guidance provided, emphasis is METHODS placed on the importance of increasing physical activity.1 In recent years, many epidemiological studies have The Jichi Medical School (JMS) Cohort Study is a confirmed the health benefits of physical activity,2–6 noting prospective, population-based study aimed at exploring an association between a high level of physical activity and the risk factors for cerebro-cardiovascular disease in 12 lower mortality. However, most of these studies have been communities in Japan.7–10 Enrollment into the JMS Cohort Addressforcorrespondence.ShinyaHayasaka,MD,PhDDepartmentofCommunityHealthandPreventiveMedicineHamamatsuUniversitySchoolofMedicine 1-20-1Handayama,Higashi-ku,Hamamatsu431-3192,Japan(e-mail:[email protected]). Copyright©2009bytheJapanEpidemiologicalAssociation 24 Hayasaka S,et al. 25 Study and baseline data collection were performed between JMS was responsible for ethical review of this research and April 1992 and November 1993. Details regarding the JMS approved this study. Cohort Study design and additional descriptive data are availableelsewhere.7–10Theresponserateforthecheck-upfor Statistical analyses all communities was 65%.8,10 A total of 12,490 participants We categorized participants into 4 groups according to their (4911 men and 7579 women) participated. Of these, we were PAIquartile.First,potentialconfoundersbyPAIquartilewere able to follow 12,393 participants, among whom physical expressed as means plus standard deviation (SD). P values activity data were available for 11,634 (4549 men and 7085 were calculated using one-way analysis of variance for women). We excluded 803 participants with a history of variables. Smoking status was tested using the chi-square stroke, cardiovascular disease, or carcinoma. Ultimately, data test. Next, crude mortality rates were calculated per 1000 from 10,831 participants (4222 men and 6609 women) were person-yearsandcategorizedbysexandPAI.Finally,theCox analyzed in the present study. proportionalhazardsmodelwasusedtocalculatehazardratios To ensure uniformity in data collection, we established a (HRs) and 95% confidence intervals (95% CI) for all-cause centralcommittee comprising thechief medical officers of all mortalityadjustedforageandarea(model1),andadjustedfor the participating areas. This committee developed a detailed age, area, smoking status, body mass index, systolic blood manual for data collection. Smoking habits were assessed pressure, and total cholesterol (model 2), in men and women. usingaquestionnairedevelopedforthestudy.7Smokerswere All statistical analyses were performed using the Statistical defined as participants who smoked at the time of the study. Package for Social Science® (SPSS) for Windows (SPSS Body mass index was defined as weight (kg) divided by Japan Inc., version 15.0, Tokyo, Japan). the square of body height (m). Systolic blood pressure was measured using an automated sphygmomanometer RESULTS (BP203RV-II, Nippon Colin, Komaki, Japan) on the right arm ofseated participants, after atleast 5 minutes seated rest. Participants were followed for an average of 11.9 years from Serum total cholesterol was measured by an enzymatic thetimeofbaselinemeasurementuntileitherdeathduetoany method (Wako, Osaka, Japan; interassay coefficient of cause, incident cardiovascular disease, relocation to another variation (CV), 1.5%). area,ortheendofthestudy;follow-upwascompletedfor99% Physical activity was assessed according to the criteria ofthecohort.Participantageatbaselinewas54.8±12.0years used in The Framingham Study. A questionnaire11 was (mean±SD) for men and 55.0±11.3 for women. During the administered in an interview conducted by a trained studyperiodtherewere834deaths:503menand331women. reviewer. Information obtained included average sleeping Tables 1 and 2 show the mean values and SD for potential time, working hours, and the kinds of activities conducted confounders, by PAI quartile, in men and women, during a typical workday and during leisure time in a normal respectively. There were significant differences between weekday. We classified activities into 5 groups according to quartiles in many potential confounders; however, the mean level of exertion. Each level of exertion was assigned a values for these potential confounders and for smoker rates coefficientbasedontheFraminghamStudy’sphysicalactivity did not vary considerably across physical activity groups. index weighting factors.11 The coefficients and time spent on Table 3 shows per 1000 person-years, categorized by sex an activity were then multiplied. We then summed the and PAI. In men, this rate was lowest in the third quartile multiplied values to produce the Physical Activity Index (Quartile 3), with 95 deaths and a crude mortality rate of 7.6. (PAI) over 24 hours. For example, “a person who sleeps ItwashighestinthelowestPAIquartile(Quartile1),with171 continuallywouldreceiveascoreof24.Anofficeworkerwith deathsandacrudemortalityrateof14.4.Inwomen,thecrude nooutsideexercisecouldhaveascoreof27(8hoursatabasal mortalityratewaslowestinthehighest PAIquartile(Quartile level,12hoursatasedentarylevel,and4hoursataslightlevel 4), with 69 deaths and a crude mortality rate of 3.5, and ofactivity).Alaborerwhoisinvolvedinheavyactivityinhis highestinQuartile1,with99deathsandacrudemortalityrate jobcouldhaveascoreof42(8hoursatabasallevel,8hours of 5.2. at a sedentary level, 2 hours at a slight level, 3 hours at a Table 4 shows the HRs and 95% CIs for model 1 and 2 moderate level, and 3 hours at a heavy level of activity)”.11 calculated using the Cox proportional hazards model for Information on deaths was collected using data from death groupscategorizedbyPAI;thelowestPAIgroupwasusedas certificates and the national vital statistics database with the the reference. In men, the HRs for the third quartile (Quartile permissionoftheAgencyofGeneralAffairsandtheMinistry 3)werethelowest:0.59(95%CI,0.45–0.76)forbothmodel1 of Health, Labour and Welfare, Japan. In addition, municipal and 2; the mortality curve had a reverse J shape. In women, governments obtained information annually on participants the HRs for the second PAI quartile (Quartile 2) were the who moved to other areas. Written informed consent was lowest:0.77(0.57–1.05)formodel1and0.79(0.58–1.08)for obtained individually from all participants at the mass model 2. In both sexes, the HRs of higher PAI groups tended screening examination. The Institutional Review Board of to be lower than those of the lowest PAI (reference) group. J Epidemiol 2009;19(1):24-27 26 Physicalactivity and all-causemortality in Japan Table 1. PotentialconfounderscategorizedbyPAI(men) Quartile1(≤28.8) Quartile2(28.9–34.3) Quartile3(34.4–38.4) Quartile4(≥38.5) Factors n Mean SD n Mean SD n Mean SD n Mean SD P Age (year) 1046 54.8 13.0 1046 57.0 12.0 1045 54.3 11.4 1059 53.1 11.1 <0.001 Bodymassindex (kg/m2) 1028 23.2 2.9 1032 23.0 2.9 1014 22.8 2.8 1041 22.8 2.8 0.006 Systolicbloodpressure (mmHg) 1037 131.2 20.4 1040 132.1 20.4 1027 131.4 20.1 1045 129.4 20.6 0.019 Totalcholesterol (mg/dl) 1037 187.8 35.6 1038 185.5 34.6 1029 183.4 33.1 1048 182.9 33.8 0.004 Smoker n(%) 545 (52.4%) 505 (48.4%) 554 (53.2%) 540 (51.2%) <0.001 Pvalueswerecalculatedusingthechi-squaretestforsmokersandone-wayanalysisofvarianceforothers. SD:standarddeviation Table 2. PotentialconfounderscategorizedbyPAI(women) Quartile1(≤28.0) Quartile2(28.1–30.2) Quartile3(30.3–33.9) Quartile4(≥34.0) Factors n Mean SD n Mean SD n Mean SD n Mean SD P Age (year) 1633 53.9 12.8 1630 54.7 11.7 1652 55.8 10.7 1642 55.8 9.6 <0.001 Bodymassindex (kg/m2) 1604 23.0 3.3 1591 23.2 3.2 1623 23.1 3.2 1614 23.1 3.1 0.509 Systolicbloodpressure (mmHg) 1622 126.9 21.3 1614 127.9 22.0 1638 128.4 20.6 1618 128.3 20.2 0.148 Totalcholesterol (mg/dl) 1620 197.1 35.7 1612 197.4 35.5 1644 198.6 34.5 1628 193.7 33.7 <0.001 Smoker n(%) 114 (7.1%) 107 (6.7%) 81 (5.0%) 63 (3.9%) <0.001 Pvalueswerecalculatedusingthechi-squaretestforsmokersandone-wayanalysisofvarianceforothers. SD:standarddeviation Table 3. Numberofall-causedeathsandcrudemortalityratesper1000person-yearscategorizedbyPAI Men Women Death Death PAIcategory Person-years n Rate PAIcategory Person-years n Rate Total 49,124 503 10.2 Total 77,693 331 4.3 Quartile1(≤28.8) 11,844 171 14.4 Quartile1(≤28.0) 18,891 99 5.2 Quartile2(28.9–34.3) 12,220 130 10.6 Quartile2(28.1–30.2) 19,197 76 4.0 Quartile3(34.4–38.4) 12,472 95 7.6 Quartile3(30.3–33.9) 19,634 87 4.4 Quartile4(≥38.5) 12,588 107 8.5 Quartile4(≥34.0) 19,971 69 3.5 DISCUSSION Our multicommunity, population-based study found that, among both men and women, groups with a higher PAI had Table 4. Adjustedhazardratiosfordeathfromanycause a lower risk of all-cause death than groups with a lower PAI. Men This was demonstrated by the fact that the groups with a Model1* Model2† higher PAI had lower crude mortality rates and lower HRs PAIQuartile HR 95%CI HR 95%CI than groups with a lower PAI (Tables 3 and 4). Quartile1(≤28.8) 1.00 1.00 Manypreviousstudieshaveexaminedthehealthbenefitsof Quartile2(28.9–34.3) 0.66 0.52–0.83 0.69 0.55–0.88 physical activity. Demonstrated benefits include control of Quartile3(34.4–38.4) 0.59 0.45–0.76 0.59 0.45–0.76 Quartile4(≥38.5) 0.76 0.59–0.98 0.75 0.55–1.04 bodyweight; improvement of hyperlipidemia, hypertension, osteoporosis, sleep disorders, and anxiety; and prevention of Women Model1* Model2† diseases associated with aging.2 Previous epidemiological HR 95%CI HR 95%CI studies have also revealed that physical activity is associated Quartile1(≤28.0) 1.00 1.00 with lower mortality.2–6 Our findings are compatible with Quartile2(28.1–30.2) 0.77 0.57–1.05 0.79 0.58–1.08 these previous studies. In addition, many previous studies Quartile3(30.3–33.9) 0.89 0.66–1.20 0.90 0.66–1.22 have shown a dose-response relationship between physical Quartile4(≥34.0) 0.81 0.58–1.12 0.83 0.59–1.16 activityandall-causemortality,andsomeofthereporteddata Model1*:adjustedforageandarea were compatible with a J-shaped, reverse J-shaped, or U- Model 2†: adjusted for age, area, smokingstatus, bodymass index, shaped mortality curve.3,4 Our study showed a reverse J- systolicbloodpressure,andtotalcholesterol HR:Hazardratio;95%CI:95%confidenceinterval shaped curve in men, suggesting that moderate physical J Epidemiol 2009;19(1):24-27 Hayasaka S,et al. 27 activity is indeed beneficial. The group with the highest grant from the Japan Health and Research Institute, Tokyo, PAI had a higher death rate than groups in the second and Japan. The authors thank Dr. Kenji Osuga of Osuga Hospital third quartiles, but a lower death rate than the group with the for his support. lowest PAI. In a comparison of highly active and moderately active individuals, Kiely DK et al observed that vigorous REFERENCES physical activity yielded no additional protective effect against stroke.12 Others have suggested that the risk for 1. MinistryofHealth,LabourandWelfare,Japan.Standardprogram ischemic heart disease increases with physical activity levels of health check up and health guidance. Tokyo: Ministry of beyond moderate.13,14 A previous cohort study in a Japanese Health, LabourandWelfare, Japan; 2007(inJapanese). 2. USdepartmentofHealthandHumanServicesPhysicalActivity provincial city also revealed a U-shaped relationship between physical activity index and incidence of all strokes.15,16 They andHealth.AreportoftheSurgeonGeneral.Atlanta,GA:U.S. Department of Health and Human Services, Center for Disease concluded that individuals performing heavy physical labor Control and Prevention, National Center for Chronic Disease. were at higher risk for cardiovascular disease. PreventionandHealth Promotion;1996. Inwomen,aninverserelationshipbetweenphysicalactivity 3. Lee IM, Skerrett PJ. Physical activity and all-cause mortality: and the all-cause death rate was observed; however, this what is the dose-response relation? Med Sci Sports Exerc findingwasnotstatisticallysignificant.Thislackofstatistical 2001;33(6):S459–71. significance may have been due to the fact that there were 4. Oguma Y, Sesso HD, Paffenbarger RS Jr, Lee IM. Physical fewer deaths among women than men. In addition, the activity and all-cause mortality in women: A review of the majority of women were classified as housewives, and it is evidence. BrJ Sports Med2002;36:162–72. possible that the level of physical activity associated with 5. Warburton DER, Nicol CW, Bredin SSD. Health benefits of housework was underestimated.12 Using data from 38 studies physical activity: the evidence. CMAJ2006;174:801–9. to estimate the average effect of physical activity and fitness 6. OgumaY.Physicalactivityandall-causemortality.ClinSports Med2006;23:189–98(inJapanese). on mortality, Oguma et al calculated that the median relative 7. NagoN,KayabaK,HiraokaJ,MatsuoH,GotoT,KarioM,et riskfordeathamonghighlyactiveindividualswas0.66.4Our al. Lipoprotein(a) levels in the Japanese population: influence HR results are in conformity with this previous result. of age and sex, and relation to atherosclerotic risk factors: In order to exclude the effects of participants with latent The Jichi Medical School Cohort Study. Am J Epidemiol diseases, we also calculated HRs for all-cause mortality after 1995;141:815–21. excluding deaths that occurred during the first 2 years of 8. Ishikawa S, Gotoh T, Nago N, Kayaba K. The Jichi Medical follow-up. However, the trends for HRs calculated with and School (JMS) Cohort Study: Design, baseline data and without this exclusion were similar. standardized mortality ratios. J Epidemiol 2002;12:408–17. Ourstudydoespossesssomelimitations.First,themeasure 9. Ishikawa S, Kayaba K, Gotoh T, Nakamura Y, Kajii E. of physical activity used was based on participant recall, and Metabolic syndrome and C-reactive protein in the general thus,thereislikelytohavebeenconsiderableimprecisionand population:JMSCohortStudy.Circ J 2007;71:26–31. some misclassification of participants. However, this blunt 10. Ishikawa S, Kayaba K, Gotoh T, Nago N, Nakamura Y, classification is likely to result in an underestimation of the TsutsumiA,etal.Incidenceoftotalstroke,strokesubtypes,and myocardial infarction in Japanese Population: The JMS Cohort associationbetweenphysicalactivityandmortality.11Second, Study.J Epidemiol 2008;18:144–50. we obtained information about physical activity only once, at 11. William BK, Paul S. Some health benefits of physical activity: baseline.Katzmarzyketalreported thatlessphysicallyactive TheFramingham Study.Arch Intern Med1979;139:857–61. people who increase their physical activity after initial data 12. Kiely DK, Wolf PA, Cupples LA, Beiser AS, Kannel WB. collectionhavealowermortalityratethanpeoplewhoremain Physicalactivityandstrokerisk:TheFraminghamStudy.AmJ less active.17 However, this second limitation is also likely to Epidemiol 1994;140:608–20. haveresultedinanunderestimationoftheassociationbetween 13. Shaper AG, Wannamethee G. Physical activity and ischaemic physical activity and mortality. heart disease in middle-aged British men. Br Heart J Inconclusion,ourmulticommunity,population-basedstudy 1991;66:384–94. foundthatahigherlevelofphysicalactivityisassociatedwith 14. Wannamethee G, Shaper AG. Physical activity and stroke in Britishmiddle agedmen. BMJ1992;304:597–601. a lower risk for all-cause death in men and women. These findings may encourage people to increase their level of 15. TsuchidaM,HayashiM,DateC,MuiK,KuriharaH,TanakaH. Aprospectivecohortstudyonrelationshipofphysicalactivityto physical activity in accordance with the aims of the new subsequentincidenceofstrokeandischemicheartdisease.Bull Japanese health check-up system. PhysFitness ResInst1987;65:118–28. 16. NakayamaT,DateC,YokoyamaT,YoshiikeN,YamaguchiM, ACKNOWLEDGEMENTS Tanaka H. A 15.5-year follow-up study of stroke in a Japanese Provincial City TheShibataStudy.Stroke1997;28:45–52. ThisstudywassupportedbyagrantfromTheFoundationfor 17. KatzmarzykPT,JanssenI,ArdernCI.Physicalinactivity,excess theDevelopmentoftheCommunity,Tochigi,Japan,andbya adiposityandpremature mortality.Obes Rev2003;4:257–90. J Epidemiol 2009;19(1):24-27