Original ContributionS see Changes in Physical Fitness and All-Cause Mortality A Prospective Study of Healthy and Unhealthy Men ‘Steven N. Blair, PED; Harold W. Koni Ill, PRO; Carolyn E. Barlow, MS; Ralph S. Paffenbarger, Jr, MD, DrPH; Larry W. Gibbons, MD, MPH; Caroline A. Macera, PhO Objective—To evaluate the relationship between changes in physical fitness and risk of mortality in men. Design.—Prospective study, with two clinical examinations (mean interval between examinations, 4.9 years) to assess change or lack of change in physical fitness as associated with risk of mortality during follow-up after the subsequent ‘examination (mean follow-up from subsequent examination, 5.1 years) Setting.—Preventive medicine clinic. Study Participants.—Participants were 9777 men given two preventive medi- cal examinations, each of which included assessment of physical fitness by max mai exercise tests and evaluation of health status. Main Outcome Measures.—All-cause (n=223) and cardiovascular disease (n=87) mortality. Results.—-The highest age-adjusted all-cause death rate was observed in men Who were unfit at both. examinations (122.0/10000 man-years); the lowest death rate was in men who were physically fit at both examinations (39,6/10000 man-years). Men who improved from unfit fo fit between the first and subsequent examinations had an age-adjusted death rate of 67.7/10 000 man-years. This is a reduction in mortality sisk of 44% (95% confidence interval, 25% to 50%) relative to men who remained unfit at both examinations. Improvement in fitness was as- sociated with lower death rates after adjusting for age, heath status, and other risk factors of premature morality. For each minute increase in maximal treadmill time between examinations, there was a corresponding 7.9% (P=.001) decrease in risk of mortalty. Similar results were seen when the group was stratified by health sta- tus, and for cardiovascular disease mortality. Conctusions.—Men who maintained or improved adequate physical fitness were less likely to die ftom all causes and from cardiovascular disease during follow-up than persistently unfit men. Physicians should encourage unfit men to improve their fitness by starting a physical activity program. (JAMA. 1996271081088) From the Cooper inate tor Rerebice Recoarch (Ors Blair and Koh and Me Brow) and Cooper Cine (©: Goobone), Dal, Tex: Stanford (Cal Unversity Senoo! of Werdiine (Ds Patienbargo) anc Schoo! of Pubic Hoa, Unversity of Seuh Carcina, Golrbia (@rMacor} pit requests to Division of Epidemiiogy, Ca0- per instute for Atri Research 12500 Preston Gases, 1675290 (Or Bia). Ash 2045-278 No. 14 \ LOW LEVELS of physical Stness are associated with an increased risk of all- cause and cardiovascular disease (CVD) mortality, with age-adjusted relative risks (RRs) of CVD up to eightfold greater for unfit groups than their ft counterparts." ‘The increased risk among unit individu- als is substantially greater than the risk reported for inactive participants in stu ies of physical activity and mortality, in which an approximate doubling of risk is seen when inactive groups are compared with active groups." Physical fitness (a condition) is measured more objectively and with greater reliability than physical activity (a behavior), which may account for the differences in the strength of the relationship between these tivo exposures and mortality, Studies on fitness and mortality have relied on a single baseline assessment of fitness, with subsequent. follow-up for mortality." With single exposure as- sessmnents, it is difficult to diseount the hypothesis that genetic factors or other confounding variables are an important cause ofthe relationship between fitness and mortality. Moreover, changes in the variable of interest during follow-up may affect the subsequent risk of mortality. One way to examine this issue more com- pletely is to evaluate the effect of changes in physica} fitness on risk of mortality. Extended follow-up in the Aerobies Cen- ter Longitudinal Study provides an op- portunity to evaluate the relationship of changes in physical fitness with mortal- ity in a cohort of men with two clinical ‘examinations and subsequent follow-up for mortality. We tested the hypothesis that changes in physical fitness, especially in men with an initial low level of Btness, produced changes in risk of mortality. SUBJECTS AND METHODS Subjects and Design Study participants were 9777 men, rranging in age from 20 to 82 years at Finess Changes and Morality Blair et al 1083baseline, who completed at least two preventive medical examinations at the Cooper Clinie in Dallas, Tex, from De- cember 1970 through December 1989. Although women are included in the co- hort there are too few women in the study with two examinations to perform the analyses reported herein.” All study subjects were residents of the United States, had complete examinations at both visits, and achieved at least. 85% of their age-predicted maximal heart rate (220-age [years)) during both treadmill tests. A total of 268 men were excluded because of failure to achieve the maxi- mal heart rate criterion; 140 were ex- cluded at baseline, and an additional 128 were excluded for failing to reach 85% or more of age-predicted maximal heart, rate at the subsequent examination, ‘Analyses were performed in the en- tire population and within health status subgroups. The 6819 men who had nor- mal resting and exercising electrocar- diograms, and no history or evidence of myocardial infarction, stroke, diabetes, or hypertension at both examinations, were defined as the apparently healthy subgroup. Men with one or more of these conditions at either or both of the ex- aminations were classified as unhealthy. ‘The 2958 unhealthy men had @ total of 3836 chronie conditions because some men had more than one condition. The distribution of conditions in this group ‘was as follows: myocardial infarction, 178; hypertension, 1733; diabetes mel- litus, 374; caneer, 18: electrocardiogram, 861; exercise electrocardiogram, 498. ‘The average (SD) interval between the two examinations was 4.9=4.1 years (range, 1 to 18 years). Follow-up for mor- tality after the subsequent examination was an average of 5.142 years (range, 1 to 18 years) (total follow-up, 47560 man-years). These intervals were com- arable for the health status, groups e interval between examinations was 4324.0yearsin healthy menand52+43 ‘years in unhealthy men, Follow-up was, 4.74.0 and5.24.4 years, respectively, in the healthy and unhealthy groups. Clinical Examination The study protocol was reviewed and approved annually by the Institutional Review Board. After patients gave their written informed consent and subse- quent registration in a follow-up study, examinations followed an overnight. fast, of at least 12 hours. Components of the examination included a elf-administered personal and family medical history, in- cluding demographic factors and health behavior, a physical examination by a physician, anthropometry, blood pres- sure measurement, and blood chemis- 1094 DAMA, Apri 12, 1995—Val 273, No. 14 try analyses. All patients completed a maximal treadmill exercise test, which included measurement of resting and exercising electrocardiograms and blood pressures, Trained laboratory techni- cians, with physician supervision, ad- ministered the assessments in accord with a standard manual of operations. Details of examination procedures have been published elsewhere. ‘The primary exposure variable for the analyses presented herein s the level of physical fitness. As used in this report, physical fitness refers to exercise test tolerance to a standard treadmill pro- tocol,* and treadmill test duration was the variable used in the present analy- ses. Test time with this protocol is highly correlated with measured maximal oxy- gen uptake (r=.92 in men"); thus, physi- cal fitness in this report is analogous to aerobic power. Blood pressures were measured by auscultatory techniques with amereury sphygmomanometer. Diastolie pressure was recorded as the disappearance of sound. Serum samples were analyzed by automated methods for lipids and glucose. Weight and height were mea- sured on a standard physicians’ balance beam scale and stadiometer. Body mass index was caleulated as weight in kilo- ‘grams divided by the square of height in meters. Cigarette smoking habits, aleo- hol intake, health behaviors, and family ‘medical histories were self-assessed by questionnaire at the time of the exami- nations, Data on medication use in the cohort are incomplete and are not in- cluded in the analyses presented herein, To evaluate the hypothesis that changes in fitness are caused by changes in self-reported activity, we computed change in treadmill time for self-reported physical activity habits during the month before the clinical examination in the subgroup of men who were examined after January 1, 1987. There were 1512 men available for analysis who had two complete examinations, A three-cat- egory physical activity index was com- puted for each man. Men who reported noleisure-time physical activity received an activity seore of 0, those who played sports or walked or ran up to 16.1 km (10 miles) week ascore of 1, and those who walked or ran more than 16.1 ka (>10 miles)a week a seore of 2. The index was computed for each of the two examina- tions, and a change in activity score was obtained for each subject. Average change (adjusted for age and follow-up interval) in treadmill test time from the first to the subsequent examination was 8 seconds for men who had no increase in activity, 41 seconds for men who in- creased their activity by one category ‘on the physical activity index, and 98 \ seconds for men who lad atwo-category change in the physiéal activity index. Each of these adjusted means was sig nificantly different from the others (P<.001), thereby supporting our as- sumption that change in fitness is an overall marker for change in activity. Mortality Surveillance We followed up participants for mor- tality from the second examination to the date of death or to December 31, 1989, ‘Mortality follow-up was accomplished by means of the National Death Index. Possible decedents were identified by the National Death Index, and this list was reviewed by three of us (S.N. HLW.K., and C.E.B.). This process iden’ tified individuals from 44 states for re- trieval of death certificates. We then contacted the department of vital sta- tisties in each of these states and re- quested copies of the official death cer- tifieates. We compared information on the death certificates with clinical rec- ords to confirm that the death certifi- cate pertained to that individual. The underlying cause of death was coded by a nosologist according to the Interna- tional Classification of Diseases, Ninth Edition, Revised, with CVD defined for the analyses presented herein as codes, 390 to 449.9, Data Analysis Previous follow-up studies of this co- hort. show that age-adjusted all-cause mortality is substantially higher in the first (least fit) quintile than in the see- ond.* Therefore, for categorie analyses, ‘unfit is defined as the least fit quintile and fit refers to all others (quintiles 2 through 5). Quintile cutoff points for each age group were established by means of measurements from the first examina- tion (data not shown). Treadmill time cutoff points and associated units of work metabolic rate divided by resting meta- bolic rate (METs) for the unfit category were less than 14.0 minutes (10.0 METS) for men aged 20 to 39 years, less than 12.0 minutes (92 METS) for men aged 40 to 49 years, less than 10.0 minutes (8.4 METS) for men aged 50 to 59 years, and less than 6.7 minutes (7.0 METTs) for men aged 60 years and older. We also analyzed change in fitness as a continu ous variable in proportional hazards ré- gression models, where change was cal- culated for each participant by subtract- {ing the treadmill time in minutes of the first test from that of the final test. Kaplan-Meier survival curves were generated to evaluate the survival e perience of initially unfit men who e- ther remained unfit or became fit. Mor tality rates were caleulated for men who Finess Changes and Moralty--Blair et a!‘able 1-—Basaline Charactensies From Fist Examination of 9777 Men, Aerobics Center Longiueinal 258231 either changed or remained unchanged in physical fitness status as deaths per 10000 man-years of observation. Rates were age adjusted by the direct method with the entire population used as the standard, Age groups used in the ad- justment were 20 to 39 years, 40 to 49 years, 50 to 59 years, and 60 or more years, The RR of death for fitness and Fitness change was calculated by means of the age-adjusted rates, with the men unfit at both examinations as the ref erence category. Multivariate statisti. cal models using proportional hazards regression were constructed to caleu- late adjusted RR" for both all-cause and CVD mortality. These models included age to the nearest year; interval be- tween examinations in years; baseline physical fitness level, change in fitness level; baseline values for weight, rest ing systolic blood pressure, serum cho- lesterol, fasting blood glucoge, cigarette smoking habit; and family history of eoro- nary heart disease. Furthermore, a bi- nary variable indicating health status (healthy or unhealthy) was added in analyses of the total population. Data on change in risk factors were examined in additional models. The 96% confidence intervals (Cis) round all point estimates also were calculated.” RESULTS Results reported herein are from 9777 men with at least two complete exami- nations during the study period. There ‘were 223 deaths (108in healthy men and 120inunhealthy men) during 47861 man- yearsof observation, and 87 deaths were caused by CVD. Baseline characteris- ties of survivors and decedents and for the healthy and unhealthy subgroups at the first examination are shown in Table 1. Because of the 8.4-year average age difference between survivors and dece- dents, age-adjusted data (not shown) also AMA, Apill 12, 1995~Vol 273, No. 14 1 SSE19 (91250) _S921.1 eRe249) _S4=14,R00=5H) S711 B04) 2092133 12752108 neaztt9 e393 wens ye7=8s 19248 sa1350 sa7247 se3:49 7220 weozts 119=20 164 ‘were calculated, This had little effect on the means and SDs, although the dif- ferences between survivors and dece- dentsnarrowed slightly for all variables. Dacedents were older and less physi- cally fit and generally had less favorable risk profiles than survivors. There were differences between the healthy and un- healthy cohorts for several clinical char- acteristics, ineluding weight, total cho- festerol, and blood pressure. Figure 1 shows survival curves for men who were unfit at both examina- tions and men who were unfit at the first examination but improved to be classified as fit at the subsequent ex- amination. The persistently unfit men had lower probability of survival. The survival curves diverge continuously throughout the 18 years of follow-up. ‘Age-adjusted all-cause and CVD death rates for fitness categories are shown in Table 2. Men who were unit at both vis- its had the highest death rate, men who were fit at both visits had the lowest death rate, and men who changed fitness status had intermediate rates. Overall, ‘men who were initially uniit and became St had a 44% lower age-adjusted risk of all-cause mortality (RR, 0.56; 95% CF, 0.42 to 0.75) and a 52% lower age-adjusted risk of CVD mortality (RR, 0.48; 954% CI, 031 to 0.74) than their peers who re. mained unfit, Beeause 75% to 82% of the ‘man-years of observation anda large pro. Portion of deaths were in the group of Ten who were fit at both examinations, we performed additional analyses to evaluate the possibility that there might, bbe a gradient of change in risk with the amount of fitness change in the initially fit group. Table 2 also shows the death rates for fit men who improved from fit- ness quintile 2 or 3 to quintile 4 or 5, and for men who were in quintile 4 ar 5 at both examinations. Men in quintile 2 or 3 at both examinations constitute the ref- Survival Probablity o24 Followup interval, y 6 8 1012 14 16 18 Figure 1.—Survval cures for change or lack of change in physical ftness nal men. Thelower line ‘represants survvat experience in men in the least fit quntie at both examinations; upper ine, men ‘who were unt at baseline but came ft (auintes. 2 thug 5) by the fal exarinaion, erence category. The men who were ini- tially fit and further improved their fit ness level had a 15% lower risk of death from all enuses than men in the reference category, whereas the men in quintiles 4 or 5 at both examinations were at the lowest risk (RR, 0.71; 95% Cl, 0.46 to 1.09). Corresponding ‘figures for CVD mortality were 28% lower risk for men ‘who improved their fitness, whereas per- sistently highly fit men had the lowest risk (RR, 0.48; 95% CI, 023 to 1.01). We examined the ‘relationship be- tween fitness change groups and all- cause mortality within specific age groups (Figure 2). Men who were i tially unfit and became fit by the sub- sequent examination had lower death rates than men who were unfit at both ‘examinations. Death rates for those who increased their fitness ranged from 33% lower than for those who remained unfit, inthe 40- to 49-year age group to more than 70% lower than for those who re- mained unfit in the 20- to 39-year and 50- to 59-year age groups. Men aged 60 years and older who improved their fit~ ness had death rates 50% lower than those of the persistently unfit men. For each age group, men who were fit at both examinations experienced the low- est death rates. Proportional hazards regression an- alyses showed that improved physical fitness was associated significantly with lower adjusted risks of all-cause and CVD mortality (Figure 3), Results are consistent for all-cause and CVD mor. ‘ality, and in healthy and unhealthy men, Changes in other risk factors did not add significantly to the prediction of mor- tality and thus were excluded from the final models presented herein. Aer ad- Justment for potential confounders, each ‘minute increase in treadmill time froma the first to the subsequent visit was Finess Changes and Moraty—Biair et of 1095\ ‘Table 2-—~Contiuities and Changes in Physical Fitness and Age-Adjusiad, Al-Cause, and Cardiovascular Disease Morality in 9777 Mop, Aerobics Center Lon. tudinal Stay, 1970 Through 1989 es Physical Fitness ‘Age-Adlusted 5% —e Manvonee Wo. ot woot ‘Goat Gate! Resetive Contidance {et Examination ‘and Examination of Follow-up Mon Becta 70000 MY. isk iInerval ‘A cause ve us 2007 Ea) 2 120 1.00 ore Fit 4054 50) 2% a7 056 Dara75 ee at 128 22 2 3 052, (0.98-0.70 fa Fi sages 355 7 998 233 020047 Ganaes ‘Gunaies 23 23 9150 1506 6 ae 1.00 a cy eras isi S as oss Bene ry as = ois 0 @ ma ort 046-109 “Cardiovaneclas Dawes vs 2428 "356 15 650, 1.00 Fi 2017 38 13 ata 48. o31076 Uni 1594 215, 3 279 08) (028-07 i mr Base = 142 oz 012099, ‘uitiee 23 ses? 1480 ~ 218 1.00 os col reat 2 158 O72 3 75560 “977 a 705 048 "Guile ot agespectic Rness dsirbuton + Ouinles'2 tough 8 of age-spoctic ress cstbuton g Al-Cause Death Fates/10000 Man-Years WW Unfit at Both Examinations: Unita Tet, Fit at and Examination I Fit at Both Examinations | 50-59 ‘Age Group, y Figura 2.—Al-causa death ates por 10000 man-yaars (og scala} in 9777 een by age groupe anc change, ar lack of charge in physical finess. Death rates are shown atop the bars andthe umber of death within the bars associated with a reduction in risk of mortality of 7.9% (P=.001) for all-cause, mortality and 8.6% (P=.027) for CVD mortality in the analyses for all men combined. A 4-minute inerease in tread- rill time with this protocol is equivalent ‘oan increase of approximately 2 METs or 10 mLikg per minute in maximum oxygen consumption, a degree of im- provement typically seen in exercise training programs. Anincrease in tread- mill time of 4 minutes in the current study was associated with an estimated reduction of approximately 30% in mor- tality risk after other variables in the model were taken into account. Although the proportional hazards analyses statistically adjusted for health 1096 JAMA, Apel 12, 1495—Vol 273, No. 14 status, subclinical disease at baseline po- ‘entially might confound these results by adversely affecting fitness and also increasing risk of mortality. Toevaluate this possibility further, we performed analyses onsubgroupsot the coat who died within years of follow-up and eom- pared them with men who died more than 3 years after the subsequent ex- amination. When compared with men who were unfit at both examinations, the men who improved from unfit to fit had an age-adjusted RR for all-cause mortality of 0.67 (95% CI, 0.61 to 0.74) in the first 3 years of follow-up and 0.38 (95% CI, 0.26 to 0.56) in the later follow- up interval. Additional analyses were done by constructing multivariable re- Pet62 Multivariate Adjusted % Reduction in Risk 2 ‘AlLCause Mortality CVD Mortality 3p Heatny Mon (Bl Uanoatiny Men Figure.—Adjusted percentage reduction indskfor alreause and cardiovascular iseaso (CVD) mor {aly rom proportional nazards rogression anly- ss, Tha bars shaw ceriction nak lareach minute impeoverant in treadmill ime From the fst 0 the subs@quent examination, Zara indicates tho ref ‘ence category of no change in treadmit ime, Fisk ware agjusted for Basel age, fnass, fami his tory of coronary heat sisease, systolic blood pres: Suto, total cholestrol level, fasting glucose love, ‘weight, smoking status, and inlorval Detwoe? ex: aminatons. Anatyses for alk en combined ware aso adjusted for heath status. The numoore of ‘eats por category for all men, healthy men, and ‘unhealthy men, respectively, were as flo cause morality, 220, 100, and 120; and CVO mor- tality, 85, 26, and 59, ‘gression models (controlling for poten- tial confounders deseribed above) for each of the two follow-up intervals. The results of these models were consistent with a reduced risk for all-cause mor- tality associated with improved fitness in both follow-up periods. Finess Changes and Mortaity—Bla talz 2 = z § s = 2 BM systolic BP (27.0 kgim) (140 mm Fg) Cholesterol Smoking Fitness (G2mmos (Any Amount} Leant Fit {240 moist) uintie) Figura 4.—Rolatve ks (acusted forage, tamly history of coronary heart tisease, health status, baseline valuss, arc changes frat variables tho figura, and inten in years bedwpon examinations) of al-cause ‘mortaliy by favorable cnangesin risk actors Dotwoen frst and aubsequont examinations. The analyees wore for men at riek on each particular varabo al the frst examination. Cutoff points Yasignting high risk are parenthticaly atthe botiom ofthe fgure, The numberof men a high ask end te number of dats) Aen characteristic wore as follows: body mass index (BMI), 2691 (66); systole load proesure (BP), 1013. (Gey ebolsioes 2212 7) opm scsng 160 (48) an pryseal as. 1018 Co) The proportional hazards regression analyses also demonstrated risk redue- tion for other favorable lifestyle or clini- cal status changes. Figure 4 presents results of proportional hazards analyses by showing adjusted RR estimates for losing weight, lowering blood pressure, reducing cholesterol, stopping smoking, and improving fitness in men at risk for of these factors at the first exami- nation, These RRs were adjusted for age in years, family history of coronary heart disease, interval between exami- nations, and baseline and change values of all variables in the figure. Standard definitions for identifying high-risk sta- tus at the baseline examination were used for overweight (body mass in- dex, 227.0 kg/m*), high blood pres- sure (resting systolic blood pressure, =140 mm Hg), and hypercholesterol- emja(=6.2 mmol/L [=240 mg/dL). With the exception of weight loss, all other presumed favorable changes were as- sociated with reduced risk of mortality, although the 95% CI included 1.0 for all variables except physical fitness. COMMENT We observed a reduction in mortality risk in the total population of men who maintained or improved physical fitness. This reduction was present in both healthy and unhealthy men, and the re- sults Were consistent across age groups. Changes in fitness, either an increase or decrease, were associated with hypoth esized and biologically piausibie changes in risk, irrespective of the baseline fits ness level. JAMIA, Apri 12, 1905 Multivariable statistical modeling showed that the findings were indepen- dent of confounding by other known risk factors. Both baseline status and changes in risk factors were considered in the analyses. Maximal exercise testing pro- vides an objective assessment of phy: cal fitness, and inereases in exercise test tolerance’ are caused by physiologic changes in the oxygen supply system. It is well established that physical fitness increases with exercise training," and although there are important individual differences in the response to training, only a minority of individuals are com- pletely unresponsive. We therefore as- sume that the changesin fitness reported herein are the result of a more physi- cally active way of life, and analyses in this cohort show significant increases in ‘treadmill test time in men who increase their activity. Heredity is a possible explanation for the higher death rates in unfit individu- als, There is a genetic contribution to aerobic power, estimated to be in the range of 25% to 40%." Data presented in the current study do not support the hypothesis that hereditary factors are solely responsible for the relationship between fitness and mortality. Men who were unfit at their initial examination but who became fit by the time of their subsequent examination had a 44% re~ duction inrisk of mortality during follow- up compared with similar unfit men who did not improve; this association is un- likely to resuit from heredity. In a pre- vious report, we showed an inverse gra~ dient of death rates across fitness cat- egories in men and women who had a family history of coronary heart disease, as well as those who did not have this characteristic, which suggests that the relationship between fitness and mor- tality is independent of parental history of coronary heart disease." Few studies on physical activity, St- ness, and health have been able toevalu- ate the effect of change in the exposure variablesto risk. A report from the Hi vard Alumni Study showed that an in- crease in physical activity is associated with a reduction in mortality risk dur- ing follow-up Healthy, middle-aged, inactive men at baseline who began to participate in moderately vigorous sports had a 28% lower risk of dying than men who remained sedentary. Pur- thermore, there was a graded associa tion of mortality risk across categories of increases and decreases in total en- ergy expended per week in physical ac- tivity" Data from the Alameda County ‘Sindy# on change in physical activity and mortality are consistent with the Harvard Alumni Study, and results in each of these studies remained after ad- justment for several potential confound- ing variables. ‘When coronary heart disease mortal- ity was considered in the Harvard Alumni Study, the size of the risk re- duction associated with increased physi- cal activity was equivalent to that as- sociated with other favorable changes, such as stopping smoking and avoiding obesity and hypertension. For example, ‘men who stopped smoking reduced their risk by 44% relative to men who con- tinued to smoke, and men who started moderately vigorous exercise program reduced their risk by 41% compared with the men who remained sedentary. Men in the Aerobics Center Longitudinal Study who improved their fitness had a 64% reduction in risk of death (Figure) compared with men who stayed unfit. This reduction in mortality risk was greater than for any of the other risk characteristics and was the only statis- tically significant reduction in multivar- inte analyses ‘The results of this investigation are consistent, with biological plausibility Increased levels of physical activity and fitness are hypothesized to act through beneficial fibrinolytic, hemodynamic, blood chemistry, blood pressure, and electrical changes to the cardiovascular system Presumably, such changes ‘were enhanced among those participants who improved their physical fitness. This study has limitations in that the population was predominantly white ‘men from middle to upper socioeconomic strata, and these factors limit the gen- eralizability of the findings. However, Fingss Changes and Mortaity—Biar et at 1097the internal validity of the study is not affected by the nature of the population. Strengths of the study include an ob- Jective assessment of physical fitness, land the thorough evaluations at the two clinical examinations provide extensive data on health status, health behaviors, and other risk factors, thus allowing for control of many potentially confounding variables. Results presented herein show the ef fect of changes in physical fitness on mor- tality and provide additional support for the hypothesis that an active and fit way References 1, Bruce RA, Hoseack KF, DeRouen TA, Hofer V. Enhanced risk angessment for primary coronary heart disease events by maximal exercise testing: 1 years’ experience of Seattle Heart Watch. J Am Coll Cardiol. 1988::506-57, 2. Bkeland L-G, Haskell WL, Johnaon JL, Whaley FS, Criqui MH, Sheps DS. Physical Stress aa a predictor of cardiovascular mortality in sayenptom- ‘Ue North American men: the Lipid Research Clin- ies Mortality Follow-up Study. N Engl J Bd. 1985 s1e:1379-1984, 4, Erikssen J Physical ftness and coronary heart disease morbidity and mortality: prospective study in apparently healthy, middle-aged men. Acta Med ‘Scand Suppl 1986771:189-192 4. Hein HO, Suadicani P, Gyntelberg P. Physial Sineas or phytical activity as a prediclor of is chaemie heart disease? 2 17-year follow-up in the Copenhagen Male Study. J intern Med, 1990282: amt, 5. Lie H, Mundal R, Eelkssen J. Coronary risk factors and incidence of coronary Geath in relation to physical Atness: seven-year follow-up study af mmiddleaged and elderly men. Bur Heart J. 1085; es7-180 6, Peters RK, Cady LD Jr, Bischoff DP, Bernstein L, Pike MC, Bhysial fitness and eubsequent myo- ral infarction in healthy workers. JAMA. 19635, 240:3052-9086. 1, Sandvik L, Brikasen J, Thaulow E, Erikaten 6, ‘MundalR, Rodahl K. Physical fitness aa predictor of mortality among healthy, middle-aged Norwe- ian men, N Bngl J Med. 1989:208595 637 1098 JAMA, April 12; 1095—Vol 273, No. 14 of ie improves health and delays death. ‘The most important new information from this study is the relationship between changes in fitness and risk of mortality. Prospective studies with aone- time assessment. of the exposure vari- able can be questioned because of con- cerns about the influences of hereditary factors, undetected preexisting disease, and uncertainty about the influences of making changes in activity and fitness. ‘This study and recent reports on physi- cal activity change alleviate some of ‘these concerns. Physicians can have con- 8 Sltery ML, Jacobs DR Jr Physical tess and ardovaszular disease mortality: the US Ralroad Study. Am J Rpideiol. 198832757150 8, Sobol J, Komitser M, De Backer Get a Protection against ischemic heart dveae inthe Folgian Fitness Study: phic ness rather than Pyscl asivity? Am J Bpidemiol 1067125801. fic, 10, ‘WieineenL, Bjure 5, Ehatrém Jol B, ta Nine year follow-up of aia eres tet in 4 random population sample of idle aged men. Cardiology, 1981;8teuppl 23°. 11 Powell KE, Thompson PD, Caspersen C3, Ken- dviok 8. Physical acvity and heinctence ofeare dary heart dnease Anni Rev Public Health. 867, 255-87. 12. Bai SN, Kohl HW II, Panenbarger RS Je, Clark DG, Cooper KH, Gibbons LW. Physi ness and al-cause mortality: a prospective toy of Healthy men and women, JAMA 1662022006. 240. 1, Bake B, Ware RW. An experimental study of “physi nes’ of Air Fore personnel US Armed Fores Mod J 195910875588 11 Paice tL, Bohannon RL, Cooper IH, tal. ‘comparative analysis of four protcla for maxk ‘nal treadmill tress tating, Am Heart 191692: 46 15, Kaplan EL, Meier P. Nonparametric estima: tion from incomplete observations. J Am Stat Aa. toe. 1osssa7-s1. 16. Cox DR. Regression models and fe tables. J Stat Soe. 19B;94187-200 fidence that their patients will reduce risk of mortality by inereasing physical activity and improving fitness, ‘This study was supported in part by Public Heath Servic eth grant AGOSHS bor he dora inttate on Agi, athena, Ma We Bale Kenneth I Cooper, MD, er eta lting the Aerobie Ceiter Longtaind Sty tnd the Cooper Cla psa nd todhicarg for anstance with “data colectomy Claude Boushard PhD, Wiliam Ls Halil PRD, Wile B° Kanne), MD, Kenneth, Powell MD, and Roy 4 Shephard, MD, fr entments onan eather rok tthe manasrp, lea Morrow Avan Laura Becker for manor brepraion; and Derothy Davi for ta analy 17, Kleinhaum DG, Kupper LL, Morgenstern H. Epidemiological Research: Principles and Quan titative Methods, Belmont, Calf: Lifetime Learn ing Publications; 192:200-307 18, American College of Sports Medicine. The re ommended quantity and quality of exerese for de- veloping and maintaining cardiorespratory and ‘muscular ness in healthy adults, Med Sei Sports Ezere 122266214 18. Bouchard C, Perusse L. Heredity, activitylevel, ‘Stness, and heaith. In: Bouchard C, Shephard RJ, Stephens T, eds. Physical Activity, Fitness, and “Health. Champaign, I: Human Kinetics; 1994:106- ue. 20, Patfenbarger RS Jr, Hyde RT, Wing AL, Lee Tt, Jung DL, Kampert JB. The association of ‘changes in physical-activity level anb other life ‘style characteristics with mortality among men. N Engl J Med 1988;528538-545. 21, Palfanbarger RS Jr, Kampert JB, Lee 1M, Hyde RT, Leung RW, Wing AL. Changes in phys. cal activity and other Ifeway patterns influencing longevity. Med Sei Sports Bzere. 199428:57 858, 22.'Kaplan GA, Haan MN. Ia thers role for pre- vention among the eldery?epidemiclogial evidence fromthe Alameda County Study. In:Ory MG, Bond K, eds. Aging and Health Care: Social Science and Policy Perapectives. London, England: Tavistock Publications ,198927-51. 28, Bouchard C, Shephard RJ, Stephens 7, ods Phyvical Activity, Fits, and Health Inlornt- tional Proceedings and Conveneus Statemest ‘Champaign, Il: Human Kinetis; 1904 ness Changes and Mortaity—Biair etal