PublishedbyOxfordUniversityPress onbehalf of theInternational Epidemiological Association International Journal of Epidemiology2007;36:228235TheAuthor 2006; all rights reserved. AdvanceAccess publication19October 2006 doi:10.1093/ije/dyl224Gripstrength, bodycomposition,andmortalityCatharineRGale,*Christopher NMartyn, CyrusCooper andAvanAihieSayerAccepted 18September 2006Background Several studiesinolderpeoplehaveshownthatgripstrengthpredictsall-causemortality. The mechanisms are unclear. Muscle strength declines with age,accompaniedbyalossof musclemassandanincreaseinfat, buttherolethatbodycompositionplaysintheassociationbetweengripstrengthandmortalityhas beenlittle explored. We investigatedthe relationbetweengripstrength,body composition, and cause-specific and total mortality in 800 men and womenaged65andover.Methods During 197374the UKDepartment of HealthandSocial Security surveyedrandomsamplesof menandwomenaged65andoverlivingineight areasofBritain to assess the nutritional state of the elderly population. The surveyincludedaclinicalexaminationbyageriatricianwhoassessedgripstrengthandanthropometry.WeusedCoxproportionalhazardsmodelstoexaminemortalityover 24years of follow-up.Results Poorer gripstrengthwas associatedwithincreasedmortalityfromall-causes,fromcardiovascular disease, andfromcancer inmen, thoughnot inwomen.After adjustment for potential confoundingfactors, includingarmmuscleareaandBMI, therelativeriskof deathinmenwas0.81(95%CI 0.700.95)fromall-causes,0.73(95%CI0.600.89)fromcardiovasculardisease,and0.81(95%CI 0.660.98)fromcancer perSDincreaseingripstrength. Theseassociationsremainedstatisticallysignificantafterfurtheradjustmentforfat-freemassor%bodyfat.Conclusion Grip strength is a long-term predictor of mortality from all-causes, cardiovasculardisease, and cancer in men. Muscle size and other indicators of bodycompositiondidnot explaintheseassociations.Keywords gripstrength, mortality, bodycompositionSeveral studies have shownthat poor grip strengthpredictsincreased all-cause mortality in older people.16The underlyingmechanisms are poorly understood. The association persistsafter adjustment for body size and does not appear tobe explained by nutritional status, the presence of chronicdisease,ordegreeofphysicalactivity.Littleisknownabouttheinfluenceofgripstrengthonmortalityfromspecificcauses. Inthe only previous study into this relationin a cohort ofdisabled womenpoorer grip strength was linked withincreased mortalityfromcardiovascular andrespiratory disease,thoughnot fromcancer.6Musclestrengthisknowntodeclinewithage, accompaniedbyaloss of musclemass andanincrease infat.79Thereisevidence that body composition may influence mortality inolderpeople,10,11butwhetheritplaysapartintheassociationbetweengripstrengthandmortalityinolderpeoplehasbeenlittleexplored.During 197374 the Department of Health and SocialSecuritysurveyedrandomsamples of menandwomenaged65 and over living in eight areas of Britain to assess thenutritional state of the elderly population. The areas werechosensothat thesocioeconomic characteristics of thestudysample were representative of older people in Britain who wereliving at home. In addition to the assessment of their dietMRC Epidemiology Resource Centre, (University of Southampton),SouthamptonGeneral Hospital, SouthamptonSO166YD, UK.*Correspondingauthor. DrCatharineRGale, MRCEpidemiologyResourceCentre, Southampton General Hospital, Southampton SO16 6YD, UK.E-mail: crg@mrc.soton.ac.uk228 by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from and physical activity, the participants underwent a clinicalexaminationbyageriatrician, whichincludedanthropometryandmeasurement of gripstrength. Weexploredhowmusclestrengthandbodycompositioninfluencedcause-specific andtotal mortalityina 24year follow-upstudyof this nationalsample.MaterialsandmethodsParticipantsDuring1973and1974, 1775peopleineight areas of Britain(5inEngland, 2inScotland, and1inWales)wererandomlysampled from family practitioner committees lists of all patientsaged 65 years and over. The areas were Islington, Harrow,Hastings, Bristol, Salford, Rutherglen, Angus, and MerthyrTydfil.Stratifiedsamplingwasusedtoobtainequalnumbersofmenandwomenaged6574, and75andover.Of theparticipants selected, 1688werelivingat homeandwereinvitedtotakepart inthestudy; 1419(84%)agreed. Inall, 983 (69%) of those who participated inthe nutritionalsurveyagreedtobeexaminedbyageriatrician.Nutritional surveyParticipants kept a diary of every itemof food or drinkconsumedovera week.Theywereprovidedwithasetofscalestoweigheachitem. Aninterviewervisitedthemat least fourtimesduringtheweek. Iftheparticipantswereunabletocopewith the weighing procedure, the interviewer used thefood diary to quantify their consumption; food purchaseswereusedasacrosscheck. Nutrient intakewascalculatedbyusingafoodcompositiontablecompiledbytheDepartmentofHealthandSocial Security. Aspart of thesurvey, participantswere asked whether they had lost or gained weight intheprevious fewmonths. They were also asked whether theyengaged in any hobbies or activities away from home and, if so,howmuchphysical activitywas involved.Clinical examinationThe geriatricians measured height, weight, mid-armcircum-ference, and skinfold thickness at four sites (biceps, triceps,subscapular, and suprailiac), three times at each site. Theyquestioned participants about their smoking habits andmedicationstakenintheprevious6monthsandtookasampleof blood for biochemical and haematological analysis.Gripstrengthof theright andleft hands was measuredthreetimes using isometric dynometry. After the examination,the geriatricians recorded diagnoses of disease according totheInternational Classificationof Diseases categories.Mortalityfollow-upOfthe983participantsexaminedbyageriatrician, 921(95%)were traced through the NHS central register. We obtaineddeath certificates for those who had died and all causes of deathentered in parts I and II were coded according to theInternational Classificationof Diseases (ninthrevision). All caseswhere cardiovascular disease (codes 390459), cancer(140208), or respiratory disease (codes 462519) werementioned on the death certificate were counted as deathsfromthesecauses.Statistical analysisBody mass index (BMI) was calculated as weight (in kg)/height2(in metres). The averages of the triplicate skinfold thick-nessmeasurementsateachsiteweretakenandthe%bodyfatwas derived using the four average skinfold thicknessmeasurements in the formulae devised by Durnin andWormesley.12Fat mass was derived by multiplying bodyweightby%bodyfat.Fat-freemasswasderivedbysubtractingfat mass from body weight. Corrected arm muscle area,corrected for bone, was calculated from triceps skinfoldthickness andmid-armcircumferenceusingformulaedevisedbyHeymsfieldet al.13BasedoncriteriaoutlinedbytheWorldHealth Organization (WHO),14BMI was classified into thefollowinggroups: underweight (,18.5), normal (18.524.99),overweight (2529.99), andobese(>30). Thebest of thesixgripstrengthmeasurements was selectedfor useinanalysis.Nutrient intake variables were skewed and were transformed tonormalityusinglogarithms.Thecharacteristicsofthemenandwomeninthestudywerecomparedusingt-test or x2-test asappropriate. ACox proportional hazards model was used toexamine the associations between grip strength, bodycomposition measures and mortality over the 24 year follow-upperiodbasedondeathsthat occurredbeforeJanuary1, 1999.We analysed men and women separately as the relationbetweenBMI and%bodyfat andmortalitydifferedbetweenthe sexes (P for interaction terms ,0.01). The results arepresentedas relative risks (hazardratios) per SDincrease ingripstrengthandbodycompositionmeasures. Riskestimateswereadjustedforagein5yearstrata.ModelsincludingBMIor%bodyfat werealsofittedwithBMI squaredor %bodyfatsquaredinordertoassesswhethertherewereanysignificantJ-shapedor U-shapedassociations.The analyses that followare based on the 800 men andwomen who were examined by a geriatrician and had completedata onall anthropometric and body compositionvariables.Comparisonofthese800menandwomenwiththe436studyparticipants who declined to be examined by a geriatricianshowedthattherewasnodifferencebetweentheminagebutthat those whoagreed toa clinical examinationwere morelikelythanthosewhodeclinedtobemale(56.5%vs 41.1%,P ,0.001) and fromnon-manual social classes (36.3%vs28.1%, P50.02).ResultsSelectedbaselinecharacteristicsofthe800peopleinthestudy(452menand348women)areshowninTable1.Asexpected,thereweresignificantdifferencesbetweenthesexesinallbodycompositionvariablesandingripstrength. Menhadahigherdailycalorieintakethanwomen. Theyweremorelikelythanwomentobecurrentsmokersandtoreportthattheyengagedinhobbiesthat involvedhighlevelsof physical activity. Theyhad a slightly lower prevalence of disease, diagnosed at theclinical examination, but there was no difference betweenthemintheirassessmentofwhethertheirweighthadchangedin the previous 6 months. There was also no difference betweenthesexes insocial class distribution.Therewasastronginverseassociationbetweengripstrengthandage(r 5 0.43,P , 0.001).Partialcorrelationcoefficients,adjustedforageandsex, showedthat gripstrengthwasmoreGRIPSTRENGTH, BODYCOMPOSITION, ANDMORTALITY 229 by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from strongly correlated with height (r 5 0.31, P , 0.001),fat-freemass(r50.28, P,0.001), andcorrectedarmmusclearea (r 5 0.28, P , 0.001) than with BMI (r 5 0.11, P , 0.001)or %bodyfat (r 50.09, P50.008). Gripstrengthtendedtobepoorer inpeoplewithalower calorieintake(r 50.17,P,0.001). Meangripstrength, adjustedforageandsex, wasalsopoorer inpeoplewhoreportedthat theyhadlost weightcompared with those whose weight had remained stable,54.8 vs 58.6 kg (P 5 0.03), in smokers compared withnon-smokers, 56.7 vs 58.7 kg (P 50.04), in people withdiagnosed disease compared with those who were healthy,57.4vs62.7kg(P,0.001), inthosewhoreportedhavingnohobbies outside the home compared with those who hadhobbiesinvolvinghighlevelsofphysicalactivity,56.9vs60kg(P 5 0.04), and in those from manual compared withnon-manual social classes, 58.2 vs 60.4 (P 5 0.004). Alltheseassociations weresimilar inmenandwomen.Over the24year follow-upperiod, therewere756deaths,441inmenand315inwomen. Thecrudeall-causemortalityratewas102.7per1000person-years(124.8per1000person-years in men and 82.3 per 1000 person-years in women).Table2showstherelativerisksofdeathfromallcausesperSDincrease ingripstrengthandmeasures of bodycomposition.Risk estimates are shown for men and women separately,adjustedfor age, andthenfurther adjustedfor the potentialconfounding factors, height, smoking, social class, physicalactivity, diagnoseddiseaseat baseline, calorieintake, reportedweightloss, andthemeasuresofbody composition.Correlationcoefficients betweenBMI and%bodyfat were0.74inbothsexes,whilethosebetweenBMIandfat-freemasswere0.73inmenand0.79 inwomen. To avoidpotential problems withcollinearity,fat-freemass,%bodyfat,andBMIwereaddedtomultivariatemodels separately.Inbothmenandwomen,bettergripstrengthwasassociatedwithasignificantlyreducedriskofmortalityfromallcausesinage-adjustedanalyses. Inwomen, this relationwasweakenedby further adjustment for the potential confounding factorsand,separately,forfat-freemass,%bodyfat,orBMI.Inmen,adjustment for height, smoking, social class, physical activity,diagnoseddiseaseat baseline, calorieintake, reportedweightloss,andthemeasuresofbodycompositionhadlittleeffectontheassociationandit remainedstatisticallysignificant. Therewas no evidence of a statistical interaction between gripstrength and sex as regards mortality (P 5 0.71). Figure 1 showssurvival curves for the24year follow-upperiodfor all-causemortalityaccordingtothirdsofthedistributionofgripstrengthinmenandwomen.The relationbetween%body fat and BMI and mortalitydiffered between the sexes (p for interaction terms ,0.01).In men, there was a linear relation between these variables andmortality, withahigher%bodyfat orhigherBMI associatedwith a lower risk of death. BMI ceased to be a significantpredictor of all-cause mortality inmen, after adjustment forgrip strength and the potential confounding variables, buttherelationbetweenhigher%bodyfatandmortalitypersistedafter multivariate adjustment. Inwomen, there was a weakquadratic relation between %body fat and mortality thatwas weakened by multivariate adjustment. There was astatistically significant quadratic relationinwomenbetweenBMI, analysed as a continuous variable, and mortality thatpersisted in multivariate analysis, though when relativerisks were calculated according to WHOclassifications, theincreased risk of death in the underweight and obesegroups did not reach statistical significance. Compared withthose with a normal BMI, the multivariate-adjusted risk ofdeathwas 1.41 (95%CI 0.92.38) inthose witha BMI of,18.5kg/m2, 1.00(95%CI 0.741.34) inthosewithaBMIof 2529.9kg/m2, and1.14(95%0.761.71) inthosewithaBMI >30kg/m2.Over the follow-up period, there were 488 deaths wherecardiovascular disease was the underlying or a contributorycause.Thecrudemortalityratewas66.4per1000person-years(76.4inmenand57.2inwomen).Aswithall-causemortality,riskof deathfromcardiovasculardiseasewasreducedinmenandwomenwithbettergripstrength(Table3).Inwomen,thisrelationwasweakenedbymultivariateadjustment,butinmenthe association remained statistically significant after adjust-ment for potential confounders and for measures of bodycomposition. There were no significant associations betweenriskofdeathfromcardiovasculardiseaseandarmmuscleareaor fat-freemass ineither sexinmultivariateanalyses. Therewas a statistically significant quadratic association betweenBMI, analysedasacontinuousvariable, andriskof cardiovas-cular mortality in women that persisted after multivariateTable1Characteristics of thestudyparticipantsMen(n5452)Women(n5348)Age, y 74.7(5.8) 74.4(6.1)Height, m 1.67(0.72) 1.55(0.67)Weight, kg 67.6(11.9) 60.1(11.3)BMI, kg/m224.1(3.70) 25.1(4.48)%bodyfat 21.2(5.4) 32.0(4.4)Fat-freemass, kg 52.8(7.5) 40.6(6.2)Correctedarmmusclearea, cm244.9(10.8) 37.3(10.8)Gripstrength, kg 68.7(15.7) 46.1(10.6)Dailycalorieintake, kcala2106(1.29) 1549(1.28)Current smoker, n(%) 237(52.4) 65(18.7)Recentweightchange, n(%)No 370(81.9) 277(80.8)Lost weight 43(9.5) 30(8.7)Gainedweight 39(8.6) 36(8.7)Diseasediagnosedatclinical exam, n(%)390(86.3) 322(92.5)Activitiesawayfromhome, n(%)Noactivities 185(41.1) 155(44.8)Highphysical activity 49(10.9) 16(4.6)Lowphysical activity 138(30.7) 139(40.2)Bothhighandlowphysical activity65(8.2) 277(34.8)Social class, n(%)Non-manual 157(34.4) 133(38.4)Manual 299(65.6) 213(61.6)Values aremeans (SD) unless otherwiseindicated.aGeometricmean(SD).230 INTERNATIONALJOURNALOFEPIDEMIOLOGY by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from adjustment, though, as withall-cause mortality, this didnotreach statistical significance when risks were expressedaccording to WHOcategories: compared with those with anormalBMI,riskofcardiovascularmortalitywas1.55(95%CI0.852.8)inthosewithaBMI of ,18.5kg/m2, 1.04(95%CI0.71.5)inthosewithaBMIof2529.9kg/m2,and1.20(95%CI0.761.9)inthosewithaBMI >30kg/m2.Inmen,havingahigher BMI or a higher %bodyfat tendedtobe associatedwith a reduced risk of death fromcardiovascular disease,thoughaftermultivariateadjustmentstheserelationswerenotstatisticallysignificant.Over the follow-up period, there were 425 deaths wherecancer was anunderlying or contributory cause. The crudemortality rate was 58 per 1000 person-years (77.0 inmen and 40.4 in women). Risk of dying fromcancer wasreducedinmenwithbettergripstrength, andthisassociationremainedstatisticallysignificant after multivariateadjustmentfor potential confounding factors and for measures of bodycomposition(Table 4). Inwomen, the relationbetweengripstrengthandriskofdeathfromcancerceasedtobestatisticallysignificantinmultivariateanalyses.Inmen,ahigherBMIorahigher%bodyfatwasassociatedwithareducedriskofdyingfromcancer that persisted after multivariate adjustment. Inwomen, therewas aquadratic associationbetweenBMI andrisk of dying fromcancer in age-adjusted analysis, but thisceased tobe statistically significant after multivariate adjust-ment. Therewerenosignificant associations betweenriskofdeathfromcancer and armmuscle area or fat-free mass ineither sexinmultivariateanalyses.Respiratory disease was the third most frequent cause ofdeathmentionedondeathcertificates (n5314). Thecrudemortality rate was 43.1 per 1000 person-years (55.7 in men and31.5 in women). Better grip strength was associated withreduced mortality fromrespiratory disease in both sexes inage-adjustedanalyses, buttheserelationswerenotstatisticallysignificant once adjusted for potential confounding variablesand for measures of body composition (Table 5). Noneof the measures of body composition were significantpredictors of deathfromrespiratorydiseaseafter multivariateadjustment.We explored whether the associations described abovemightbeexplainedbyseriousillnessatthetimeofthesurveyby restricting the multivariate analyses to participants whosurvivedforatleast5yearsaftertheclinicalexamination.Theapparent protectiveeffect onall-causeor cancer mortalityinmenofahigher%bodyfatorahigherBMIdisappearedonceearlydeathswereexcluded.The reversedJ-shapedassociationinwomenbetweenBMI andriskof all-causeandcardiovas-cular mortality remained statistically significant and becamestronger once deaths in the first 5 years were excluded:compared with women with a normal BMI, women with a BMI,18.5kg/m2hadariskofall-causemortalityof2.03(95%CI1.153.59)andariskofcardiovascularmortalityof2.44(95%CI 1.254.77), after multivariate-adjustment, while thosewith a BMI >30 kg/m2had a multivariate-adjusted risk ofall-cause mortality of 1.41 (95% CI 1.502.07) and ofcardiovascular mortalityof 1.62(95%CI 1.032.55). HavingaBMIof2529kg/m2wasnotassociatedwithincreasedriskofeither all-cause (1.05, 95%CI 0.761.45) or cardiovascularmortality (1.10, 95%CI 0.731.63). The relations betweengrip strength and all-cause, cardiovascular, and cancermortality in men all remained statistically significant andbecame slightlystronger whendeaths inthe first 5years offollow-upwereexcluded. For 1SDincreaseingripstrength,themultivariate-adjustedrisksofall-cause,cardiovascular,andcancer mortality were 0.78 (95%CI 0.630.95), 0.73 (95%0.560.94), and0.73(95%CI 0.560.95), respectively.DiscussionInthis 24year follow-upstudyof elderlymenandwomen,poorer grip strengthwas associated withincreased mortalityfrom all causes, cardiovascular disease, and cancer inmen, thoughnot inwomen, afteradjustment forage, height,smoking, reported weight change, physical activity, calorieintake, and diagnosed disease at baseline. Grip strength atbaselinewasstronglypositivelycorrelatedwithcorrectedarmmuscleareaandfat-freemass,and,moreweakly,with%bodyMen0 5 10 15 20 25Time (years)0.00.20.40.60.81.0Survival probability74 kgWomen0 5 10 15 20 25Time (years)0.00.20.40.60.81.0Survival probability49 kgFigure1KaplanMeier survival curves for all-causemortalityaccordingtothirds of thedistributionof gripstrengthinmenandwomenGRIPSTRENGTH, BODYCOMPOSITION, ANDMORTALITY 231 by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from fat and BMI. These indicators of body composition did notexplain the association in men between grip strength andmortality.Fewprevious studies of therelationbetweengripstrengthandmortalityhaveexaminedtherolethatmusclesizeplaysintheassociation. Inafollow-upstudyof menintheBaltimoreLongitudinal Studyof Aging, gripstrengthat baselinewas asignificantpredictorofall-causemortalityinmenaged60andover,whereasrateofchangeingripstrengthwaspredictiveinmen aged under 60.4Men with a lower 24 h creatinineexcretion, used as an estimate of muscle mass, had an increasedriskof deathbut adjustment for this factor strengthenedtheassociation between grip strength, or rate of change in strength,andmortalitysuggestingthat therelationwasnot duetotheamount of muscleinthesemen. Morerecently, results fromtheHealth, AgingandBodyCompositionStudy, showedthatalthoughpoorer grip strengthwas associated withincreasedall-causemortality,lowmusclemass,measuredbyCTscanandDXA,wasnotstronglyrelatedtomortality.15Inourstudy,wehaddataoncorrectedarmmuscleareaandfat-freemass, butneither of these explainedthe increasedmortalityassociatedwithpoorer gripstrengthinmen. As regards mortalityfromTable2Relativerisks (95%CI) of all-causemortalityper SDincreaseingripstrengthandbodycompositionmeasuresAge-adjustedMultivariate-adjusted,amodel 1includesfat-freemassMultivariate-adjusted,amodel 2includes%bodyfatMultivariate-adjusted,amodel 3includesBMIMenGripstrength 0.77(0.680.87) 0.81(0.690.94) 0.81(0.700.94) 0.81(0.700.95)Armmusclearea 0.81(0.730.89) 0.84(0.720.98) 0.86(0.760.98) 0.89(0.761.03)Fat-freemass 0.85(0.750.97) 1.05(0.851.29) %bodyfat 0.79(0.690.90) 0.80(0.700.92) BMI 0.84(0.740.94) 0.94(0.811.09)WomenGripstrength 0.81(0.670.97) 1.11(0.861.43) 1.10(0.861.40) 1.04(0.811.32)Armmusclearea 0.90(0.801.02) 0.88(0.731.06) 0.94(0.811.08) 0.91(0.771.08)Fat-freemass 0.96(0.791.15) 1.09(0.821.45) %bodyfat 0.27(0.061.26) 0.41(0.053.13) %bodyfat squared 2.82(0.7910.0) 2.03(0.3810.9)BMI 0.25(0.110.59) 0.26(0.100.68)BMI squared 4.00(1.739.24) 3.83(1.499.84)aMultivariate models include age, height, social class, smoking, reported change in weight, daily calorie intake, physical activity, diagnosed disease at baseline,andothervariablesinthetable, thoughtoavoidpotential problemswithcollinearity, fat-freemass, %bodyfat, andBMIhavebeenanalysedinseparatemodels.Table3 Relativerisks (95%CI) of cardiovascular mortalityper SDincreaseingripstrengthandbodycompositionmeasuresAge-adjustedMultivariate-adjusted,amodel 1includesfat-freemassMultivariate-adjusted,amodel 2includes%bodyfatMultivariate-adjusted,amodel 3includesBMIMenGripstrength 0.70(0.600.82) 0.71(0.590.86) 0.72(0.600.87) 0.73(0.600.89)Armmusclearea 0.83(0.730.95) 0.86(0.701.04) 0.91(0.781.06) 0.91(0.751.10)Fat-freemass 0.86(0.731.01) 1.08(0.821.41) %bodyfat 0.84(0.710.99) 0.88(0.731.05) BMI 0.87(0.751.00) 0.97(0.811.17)WomenGripstrength 0.79(0.630.99) 1.03(0.761.38) 1.09(0.821.45) 1.02(0.771.36)Armmusclearea 0.98(0.841.13) 0.95(0.761.18) 0.99(0.831.17) 0.94(0.781.14)Fat-freemass 0.97(0.781.21) 1.05(0.751.48) %bodyfat 0.44(0.063.41) 1.10(0.713.9) %bodyfat squared 2.00(0.3810.6) 1.04(0.129.94) BMI 0.21(0.090.58) 0.17(0.60.54)BMI squared 4.87(1.8512.83) - - 5.74(1.9317.1)aMultivariate models include age, height, social class, smoking, reported change in weight, daily calorie intake, physical activity, diagnosed disease at baseline,andothervariablesinthetable, thoughtoavoidpotential problemswithcollinearity, fat-freemass, %bodyfat, andBMIhavebeenanalysedinseparatemodels.232 INTERNATIONALJOURNALOFEPIDEMIOLOGY by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from specific causes, grip strength was not significantly related to riskofdeathfromrespiratorydiseaseafteradjustmentforotherriskfactors,butitwasstronglypredictiveofmortalityinmenfromcardiovascular disease and cancer, and these associationsremained after control for corrected armmuscle area andother measures of body composition. These findings suggestthattheinfluenceofgripstrengthonsurvival mayhavemoretodowiththeeffectivenesswithwhichmusclefunctionsthanitssize.Aspoorergripstrengthhasbeenassociatedwithhigherfastinginsulinlevels,bothcross-sectionallyandlongitudinally,it seems likely that muscle weakness may precede thedevelopment of insulin resistance.16,17This may help toaccount for the link between poorer grip strength and increasedriskof deathfromcardiovasculardiseasefoundinmeninthecurrent studyandinacohort of elderlydisabledwomen.6Inthelatterstudy, gripstrengthwassignificantlyassociatedwithrisk of death from respiratory disease as well as fromcardiovascular disease, thoughnot fromcancer. The lack ofassociationbetweengripstrengthandrespiratorymortalityinour smaller studymaybeduetoreducedstatistical power.Thefactthatwefoundnostatisticallysignificantassociationsbetween grip strength and mortality in women may be areflectionof therelativelysmall numberof womenwhotookpart intheclinical examination. Therewasnoindicationof aTable5Relativerisks of respiratorymortalityper SDincreaseingripstrengthandbodycompositionmeasuresAge-adjustedMultivariate-adjusted,amodel 1includesfat-freemassMultivariate-adjusted,amodel 2includes%bodyfatMultivariate-adjusted,amodel 3includesBMIMenGripstrength 0.72(0.600.89) 0.80(0.631.02) 0.79(0.630.98) 0.80(0.631.01)Armmusclearea 0.76(0.650.89) 0.83(0.651.06) 0.86(0.711.03) 0.90(0.721.13)Fat-freemass 0.78(0.640.95) 1.06(0.761.46) %bodyfat 0.74(0.610.90) 0.79(0.630.99) BMI 0.84(0.710.99) 0.91(0.731.15)WomenGripstrength 0.72(0.530.99) 1.16(0.741.81) 1.04(0.671.59) 1.01(0.671.54)Armmusclearea 0.80(0.650.98) 0.77(0.561.07) 0.91(0.691.18) 0.80(0.611.06)Fat-freemass 0.96(0.701.31) 1.33(0.832.14) %bodyfat 0.87(0.651.17) 1.07(0.741.53) BMI 0.14(0.040.51) 0.22(0.051.01)BMI squared 6.77(1.9323.7) 4.88(1.0921.9)aMultivariate models include age, height, social class, smoking, reported change in weight, daily calorie intake, physical activity, diagnosed disease at baseline,andothervariablesinthetable, thoughtoavoidpotential problemswithcollinearity, fat-freemass, %bodyfat, andBMIhavebeenanalysedinseparatemodels.Table4Relativerisks of cancer mortalityper SDincreaseingripstrengthandbodycompositionmeasuresAge-adjustedMultivariate-adjusted,amodel 1includesfat-freemassMultivariate-adjusted,amodel 2includes%bodyfatMultivariate-adjusted,amodel 3includesBMIMenGripstrength 0.73(0.630.85) 0.81(0.660.99) 0.79(0.650.97) 0.81(0.660.98)Armmusclearea 0.77(0.680.88) 0.82(0.670.99) 0.82(0.700.97) 0.80(0.721.07)Fat-freemass 0.79(0.670.93) 1.04(0.791.37) %bodyfat 0.76(0.640.90) 0.77(0.650.92) BMI 0.83(0.710.95) 0.83(0.710.97)WomenGripstrength 0.79(0.630.99) 1.03(0.761.38) 1.28(0.891.83) 1.18(0.831.68)Armmusclearea 0.98(0.841.13) 0.95(0.761.18) 0.93(0.141.17) 0.80(0.631.01)Fat-freemass 0.97(0.781.21) 1.05(0.751.48) %bodyfat 0.36(0.043.34) 1.09(0.801.51) %bodyfat squared 2.23(0.3613.8) 1.66(0.1420.3) BMI 0.29(0.090.96) 0.38(0.101.52)BMI squared 3.53(1.1011.3) 2.58(0.6510.2)aMultivariate models include age, height, social class, smoking, reported change in weight, daily calorie intake, physical activity, diagnosed disease at baseline,andothervariablesinthetable, thoughtoavoidpotential problemswithcollinearity, fat-freemass, %bodyfat, andBMIhavebeenanalysedinseparatemodels.GRIPSTRENGTH, BODYCOMPOSITION, ANDMORTALITY 233 by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from statistical interaction between grip strength and sex. Gripstrength has been shown to predict mortality in previousstudies of women, but these studies were based on considerablylarger numbers.1,6Most of thestudies intotherelationbetweengripstrengthand mortality have been carried out in elderly people.Butfindingsinapopulationofhealthymenthatgripstrengthmeasured in middle age predicts long-term risk of deathsuggest that influences affecting muscle size and functionearlier inlife maybe important.18Greater strengthtends tobe associatedwithtallness soit mayinpart be anindicatorofbetternutritioninthefirstyearsoflife,thoughobservationsincohorts of middle-agedandelderlypeoplethat thosewhohad weighed more at birth had significantly better gripstrength, independentlyof current height andweight, impliesthat fetal development is a major determinant of musclestrength.19,20TheprognosticimportanceofincreasedBMIinolderpeoplehasbeencontroversial.21Arecentsystematicreviewconcludedthat mild to moderate overweight, defined as a BMI of2527kg/m2, wasnot linkedwithhighermortalityinelderlypeople, thoughtherewas evidencethat thosewithaBMI ofover27hadan increasedriskofdeathfromall causesandfromcardiovasculardisease.22Inthecurrentstudy,wetoofoundnoevidenceineither sexthat mildtomoderateoverweight wasassociated with increased mortality. There was a reverse-J-shaped association between BMI and mortality from all causesand cardiovascular disease, with the highest mortality occurringamongthosewithaBMI of less than18kg/m2, andslightlyincreasedmortalityinthosewithaBMI >30kg/m2but thisassociationwas onlypresent inwomen.Wefoundsomeevidencetolinkahigher%bodyfatorBMIinmenwitha lower mortality fromall causes and cancer.These associations persisted after adjustment for potentialconfounding factors, including weight loss, thoughnumbersreporting weight loss were lowand longer-termweight lossmight not havebeenidentified. Weight loss over aperiodofseveral years, whether assessed by decrease in BMI,23loss of fatmass, or loss of leantissue,11,24has beenlinkedtoincreasedmortalityinprevious studies of elderlypeople. Inthepresentstudy, theseassociations appearedtobeconcentratedamongpeoplewhodiedintheearlyyearsoffollow-up,astheyceasedtobestatisticallysignificantwhentheanalysiswasrestrictedtothose who survived for at least 5 years after the clinicalexamination.Ourstudyhassomelimitations. At thetimethenutritionalsurvey was conducted, in 197374, anthropometry was theusualmethodofassessingbodycomposition.Thedatacollectedmaybeless accuratethanthat obtainablewithmoremodernmethods such as dual energy X-ray absorptiometry.25Theresultsofthestudyarebasedon800participants. Thisis47%of thoseoriginallyinvitedtoparticipateintheDepartment ofHealthandSocial Securitys nutritional survey. Nodatawereavailable on the characteristics of those who declined toparticipate inthe survey, soit is not possible togauge howrepresentative these 800 people were of those originallyinvited, though there were indications that of those whoagreedtotakepartinthenutritionalsurvey,womenandthosefrommanualsocialclasseswerelesslikelytoagreetoaclinicalexamination. All comparisons, however, have been madeinternally, sounless therelationbetweengripstrength, bodycomposition,andmortalityisdifferentinnon-respondersorinthose we were unable to trace, no bias will have beenintroduced.Ourfindingsinthis24yearfollow-upofanationalcohortofpeople aged65andover that poorer gripstrengthpredictedincreasedmortalityinmenfromallcauses,fromcardiovasculardisease, and from cancer, after adjustment for potentialconfoundingfactors includingindicators of bodycomposition,provide further evidence that the effectiveness with whichmuscle functions may be a more important long-termdeterminant of survival thanits size. Theexplanationfor thegripstrength/mortalityassociationremains unclear.AcknowledgementsWethanktheDepartmentofHealthforallowingustousedatafromthe197374Department of HealthandSocial Securitysnutritional survey. The survey was coordinated by the lateProfessor ANExton-Smith.Conflict of interest: noneof theauthors has anyconflict ofinterest.Authorcontribution:CGaleanalysedthedataandwrotethefirstdraftofthereport.CMartyn,CCooper,andAAihieSayerwereinvolvedinplanningtheanalysisandcontributedtothefinal versionof thereport.KEYMESSAGES KEYMESSAGES KEYMESSAGES KEYMESSAGES KEYMESSAGES KEYMESSAGES Grip strength has been shown to predict all-cause mortality in older people, but less is known about its influenceonrisk of deathfromspecific causes and onthe role of body compositioninthe grip strength/mortalityassociation. Ina24yearfollow-upofanationalcohortaged65andover,gripstrengthwasastrong,long-termpredictorofmortalityfromall-causes, cardiovascular diseaseandcancer inmen, thoughnot inwomen. Indicators of bodycompositiondidnot explaintheseassociations. The effectiveness with which muscle functions may be a more important determinant of survival thanmusclesize.234 INTERNATIONALJOURNALOFEPIDEMIOLOGY by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from References1Al SnihS, Markides KS, Ray L, Ostir GV, GoodwinJS. HandgripstrengthandmortalityinolderMexicanAmericans.JAmGeriatrSoc2002;50:125056.2FujitaY, NakamuraY, HiraokaJ, SakataK, et al. Physical-strengthtests and mortality among visitors to health-promotioncenters inJapan. JClinEpidemiol 1995;48:134959.3Laukkanen P, Heikkinen E, Kauppinen M. Muscle strength andmobility as predictors of survival in 75-84-year-old people. Age Ageing1995;24:46873.4Metter EJ, Talbot LA, Schrager M, Conwit R. Skeletal musclestrength as a predictor of all-cause mortality in healthy men.JGerontol ABiol Sci MedSci 2002;57:B359B365.5PhillipsP. Gripstrength, mental performanceandnutritional statusas indicators of mortalityriskamongfemalegeriatric patients. AgeAgeing1986;15:5356.6Rantanen T, Volpato S, Ferrucci L, Heikkinen E, Fried LP,Guralnik JM. Handgrip strength and cause-specific and totalmortality in older disabled women: exploring the mechanism.JAmGeriatr Soc 2003;51:63641.7Beaufrere B, Morio B. Fat and protein redistribution withaging:metabolicconsiderations. EurJ Clin Nutr 2000;54(suppl. 3):S48S53.8LindleRS,MetterEJ,LynchNAet al.Ageandgendercomparisonsofmuscle strength in 654 women and men aged 2093 yr. J Appl Physiol1997;83:158117.9Nair KS. Agingmuscle. AmJClinNutr 2005;81:95363.10Roubenoff R, Parise H, Payette HAet al. Cytokines, insulin-likegrowth factor 1, sarcopenia, and mortality in very oldcommunity-dwelling men and women: the Framinghan heartstudy. AmJMed2003;115:42935.11WooJ, HoSC, ShamA. Longitudinal changes inbodymass indexand body composition over 3 years and relationship to healthoutcomesinHongKongChineseage70andolder. JAmGeriatrSoc2001;49:73746.12Durnin JV, Womersley J. Body fat assessed from total bodydensity and its estimationfromskinfold thickness: measurementson 481 men and women aged from16 to 72 years. Br J Nutr1974;32:7797.13Heymsfield SB, McManus C, Smith J, Stevens V, Nixon DW.Anthropometric measurement of muscle mass: revised equationsfor calculating bone-free arm muscle area. Am J Clin Nutr1982;36:68090.14WorldHealthOrganization. Obesity. Preventingandmanagingtheglobalepidemic.ReportofaWHOConsultationonObesity.Geneva,Switzerland: WorldHealthOrganization, 1998.15Newman AB, Kupelian V, Visser M et al. Strength, but notmusclemass, is associatedwithmortalityinthehealth, agingandbody composition study cohort. J Gerontol A Biol Sci Med Sci2006;61:7277.16Lazarus R, SparrowD, Weiss ST. Handgrip strength and insulinlevels: cross-sectional andprospectiveassociationsintheNormativeAgingStudy. Metabolism1997;46:126669.17Sayer AA, Dennison EM, Syddall HE, Gilbody HJ,PhillipsDI,Cooper C.Type 2diabetes, musclestrength,andimpairedphysical function: the tip of the iceberg? Diabetes Care 2005;28:254142.18RantanenT, Harris T, LeveilleSGet al. Musclestrengthandbodymass indexas long-termpredictors of mortalityininitiallyhealthymen. JGerontol ABiol Sci MedSci 2000;55:M168M173.19Kuh D, Bassey J, Hardy R, Aihie SA, Wadsworth M, Cooper C.Birth weight, childhood size, and muscle strength in adultlife: evidence from a birth cohort study. Am J Epidemiol2002;156:62733.20Sayer AA, Syddall HE, Gilbody HJ, Dennison EM, Cooper C.Does sarcopenia originate in early life? Findings from theHertfordshire cohort study. J Gerontol A Biol Sci Med Sci2004;59:M930M934.21ZamboniM,MazzaliG,ZoicoEetal.Healthconsequencesofobesityintheelderly:areviewoffourunresolvedquestions.IntJObesRelatMetabDisord2005;29:101129.22Heiat A, Vaccarino V, Krumholz HM. An evidence-basedassessment of federal guidelines for overweight and obesity astheyapplytoelderlypersons. ArchInternMed2001;161:1194203.23Keller HH, Ostbye T. Body Mass Index (BMI), BMI change andmortalityincommunity-dwellingseniors without dementia. JNutrHealthAging2005;9:31620.24WallaceJI, Schwartz RS, LaCroixAZ, UhlmannRF, PearlmanRA.Involuntaryweight loss inolder outpatients: incidenceandclinicalsignificance. JAmGeriatr Soc 1995;43:32937.25Fuller NJ, Jebb SA, Laskey MA, Coward WA, Elia M.Four-component model for the assessment of body compositioninhumans: comparisonwithalternative methods, and evaluationof the density and hydration of fat-free mass. Clin Sci (Lond)1992;82:68793.GRIPSTRENGTH, BODYCOMPOSITION, ANDMORTALITY 235 by guest on December 10, 2012http://ije.oxfordjournals.org/Downloaded from