ARTICLES

Articles

Changes in physical fitness and changes in mortality

Gunnar Erikssen, Knut Liestl, Jrgen Bjrnholt, Erik Thaulow, Leiv Sandvik, Jan Erikssen

Summary
Background Point estimates of physical fitness give important information on the risk of death in healthy people, but there is little information available on effects of sequential changes in physical fitness on mortality. We studied this latter aspect in healthy middle-aged men over a total follow-up period of 22 years. Methods 2014 healthy men aged 4060 years had a bicycle exercise test and clinical examination, and completed a questionnaire in 197275 (survey 1). This was repeated for 1756 (91%) of 1932 men still alive by Dec 31, 1982 (survey 2). The exercise scores were adjusted for age. The change in exercise scores between surveys was divided into quartiles (Q1=least fit, Q4=fittest). An adjusted Coxs proportional hazards model was used to study the association between changes in physical fitness and mortality, with the Q1 men used as controls. Findings By Dec 31, 1994, 238 (17%) of the 1428 men had died, 120 from cardiovascular causes. There were 37 deaths in the Q4 group (19 cardiovascular); their relative risks of death were 045 (95% CI 029069) for any cause and 047 (026086) for cardiovascular causes. There was a graded, inverse relation between changes in physical fitness and mortality irrespective of physical fitness status at survey 1. Interpretation Change in physical fitness in healthy middleaged men is a strong predictor of mortality. Even small improvements in physical fitness are associated with a significantly lowered risk of death. If confirmed, these findings should be used to influence public health policy.

Introduction
Studies based on point estimates of physical fitness show that such estimates are a good long-term predictor of cardiovascular mortality and all-cause mortality in healthy people.17 However, these studies generally assume a homogenous decline in physical fitness with age throughout the population. Patterns of physical fitness may in fact vary between subgroups because of, for example, changes in physical activity and smoking habits over time: the Global Burden of Disease Study8 puts physical inactivity and smoking among the top ten risk factors that threaten global health. Information on changes in physical fitness and lifestyle may provide more information on long-term prognosis than can be obtained from a single estimate. To study the relations between physical fitness, changes in physical fitness, and mortality, we tested the physical fitness of a group of apparently healthy middleaged men. The men were tested on two occasions, with an interval of 7 years. Total follow-up time was 22 years.

Participants and methods

Participants
2014 men aged 4060 years took part in a baseline survey (survey 1) in 197275.9 Men were defined as apparently healthy if they had no evidence of heart disease, no diagnosed hypertension requiring drug treatment, and no diabetes mellitus, cancer, advanced pulmonary, renal, or liver disease, or other serious disorders, and were able to undertake a symptom-limited exercise test. 1756 (91%) of the men participated in a second identical survey in 198082 (survey 2), and our study uses these data. 258 of the participants in survey 1 did not participate in survey 2. Of these, 92 had died45 from cardiovascular causes, 27 from cancer, and 20 from other causes. Of the remaining 166 men who did not take part in survey 2, 47 were too ill to participate, 47 had too far to travel, and 72 gave no reasons for non-participation. Among the 1756 men who took part in the second survey, 328 were excluded for one or more of the following reasons: myocardial infarction (72), stroke (nine), angina pectoris (114), cancer (23), diabetes mellitus (32), pulmonary, renal, or liver diseases (41), hypertension requiring drug treatment (150), and inability to do the exercise test (21). Thus, data from 1428 men were included in our analysis. For these 1428 men, follow-up started at the date of survey 2 and ended by Dec 31, 1994. 1456 (86%) of the original cohort were surveyed for a third time in 198990 (survey 3), and some of these most recent data were used to check our models.

Lancet 1998; 352: 75962

Survey methods
Medical Department, Central Hospital of Akershus, N-1474 Nordbyhagen, Norway (G Erikssen MD, J Erikssen MD); Institute of Informatics, University of Oslo, Blindern, Norway (K Liestl PhD); Medical Department B, National Hospital, Oslo, Norway (J Bjrnholt MD, E Thaulow MD); and Medstat Research A/S, Lillestrom, Norway (L Sandvik PhD) Correspondence to: Dr Gunnar Erikssen

The methods of surveys 1 and 2 were the same.5,9,10,11 All participants had a clinical examination, with blood tests, spirometry, a chest radiograph, measurements of height and weight, and an exercise electrocardiographic (ECG) test. The exercise test used an electrically braked bicycle, which demanded a constant output of energy at each workload irrespective of pedalling rate. The test workload was set at 1405 kcal/min (100 W) to begin with, and was increased by 0703 kcal/min

THE LANCET Vol 352 September 5, 1998

759

ARTICLES

Characteristic

Mean (SD) or % in age-adjusted physical-fitness quartile (PF2) Q1 (n=357) Q2 (n=357) 565 (56) 1171 (0263) 39% 112% 64 (10) 131 (18) 4031 (780) 64 (12) 14 (10) 247 (27) 17% Q3 (n=357) 561 (50) 1519 (0293) 29% 151% 63 (9) 130 (17) 4268 (760) 64 (12) 14 (07) 256 (25) 16% Q4 (n=357) 571 (56) 2054 (0573) 16% 350% 60 (10) 129 (17) 4450 (781) 64 (11) 12 (06) 242 (22) 13% All (n=1428) 566 (54) 1377 (0600) 33%* 173%* 63 (10)* 131 (18)* 4110 (839)* 64 (12) 14 (09)* 247 (27)* 17%

Age (years) PF2 (KJ/kg) Smokers Physically active Resting heart rate (bpm) Systolic blood pressure (mm Hg) Vital capacity (mL) Total cholesterol (mmol/L) Triglycerides (mmol/L) Body-mass index (kg/m2) Positive-exercise ECG

567 (54) 0763 (0245) 47% 78% 67 (11) 133 (20) 3689 (836) 66 (13) 17 (11) 251 (33) 22%

Q1=least physically fit, Q4=fittest. *p for trend <0001. 10 mm ST-depression 008 s after J point.

Table 1: Characteristics of men in survey 2 according to quartile of physical fitness at survey 2 (PF2) (50 W) every 6 min. The men were encouraged to exercise to exhaustion or for as long as they were willing to continue.5 Total work capacity was calculated as the sum of all work done at all workloads, and physical fitness was defined as total work capacity divided by bodyweight. Physical activity codes were based on a questionnaire.12 We classified as physically active men who took physical exercise at least twice a week that caused sweating and shortness of breath, or took part in sporting activities.12 Physical-fitness scores in surveys 1 and 2 are referred to as PF1 and PF2 respectively.

Data collection
All mortality and morbidity data were collected after having been granted legal permission from the Norwegian Data Inspectorate and the Norwegian Board of Health to study death records in the database of Statistics Norway covering all deaths occurring in Norway, and all relevant patient records in Norwegian hospitals. Sudden deaths and deaths from ischaemic heart disease, stroke, and other arteriosclerotic diseases were classed as cardiovascular deaths. Mortality data are complete as of Dec 31, 1994.

Statistical methods
Since both physical fitness and mortality are related to age, we standardised our data accordingly. Standard mortality ratios are based on the mortality rates for 1990 reported by Statistics Norway. After log-transformation, both PF1 and PF2 decreased in a linear manner as a function of age, and their dispersion around regression lines was virtually independent of age. Ageadjusted quartiles of the distribution of PF1 and PF2 (Q1=least physically fit, Q4=fittest) were set around these regression lines. We used Coxs proportional-hazards model to study the association between physical fitness, cardiovascular mortality, and all-cause mortality, adjusted for the quartiles of physical fitness, and for age, systolic blood pressure, heart rate, blood concentrations of total cholesterol and triglycerides, body-mass index, amounts of physical activity, exercise results, and smoking status. Diagnostic plots of log S(t) versus log (t) showed that the assumptions of the proportional-hazards model were acceptable. Control runs with time-dependent covariates, which used information from survey 3, gave similar results to the standard Cox models. Our results are given as relative risks of death. For continuous variables, relative risks associated with increases of 1 SD are given. For discrete variables (physical fitness, exercise-test results, smoking status, amounts of physical activity) relative risks were calculated between groups. Pearsons 2 test was used to compare mortality between groups, and the Kendall rank test was used to assess any correlation between continuous variables or between ordinal variables. Statistical analyses used StatView (version 4.5) and S+ (version 3.4) computer programs.

findings. For all these variables, the highest quartile of physical fitness had the values indicating lowest risk. Total cholesterol concentration was not significantly associated with physical fitness. Between-group variations in physical fitness were associated with only small variations in body-mass index. By the end of 1994, 238 (167%) of the 1428 men who participated in survey 2 had died, from cardiovascular causes (120), cancer (75), and other causes (43). We found an inverse relation between all-cause mortality and physical fitness (table 2). With only physical fitness and age as predictors of all-cause mortality, the relative risk of death in quartiles Q2, Q3, and Q4 compared with the least fit quartile Q1 were 063, 037, and 031 respectively. With a larger set of risk factors included (table 3), the relations between all-cause mortality, cardiovascular mortality, and physical fitness remained similar. Of all the parameters included in the model, only smoking status, resting heart rate, systolic blood pressure, and vital capacity appeared as confounders of the predictive power of physical fitness. Another Cox model, which used total work capacity as the measure of physical fitness and bodyweight as a separate variable, gave similar results. Thus, physical fitness was a strong predictor of cardiovascular and total mortality even when other risk factors were taken into account. Although there was a good correlation between PF1 and PF2 (r=074), many of the men had changed their standard of physical fitness by survey 2. Table 4 groups participants according to quartiles of PF1 and according to quartiles of changes in physical fitness from PF1 to PF2. In five of the 16 subquartiles, PF2 was higher than PF1 (ratio>100). To illustrate what this change means in terms of exercise capacity, for the least fit quartile at survey 1 (PF1 Q1) those who increased physical fitness most (PF2/PF1 Q4), the PF2/PF1 ratio was 136% and the change in exercise capacity was only 2 min at bicycle load 2 (150W). Overall, our data show that there was a graded, inverse relation between changes in physical fitness and mortality irrespective of physical fitness at survey 1. Proportionalhazards analysis that included all risk factors, PF1, and
Cause Number (%) in age-adjusted physical fitness quartile (PF2) Q1 Q2 30 (46%) 23 (35%) 12 (19%) 65 Q3 21 (54%) 13 (33%) 5 (13%) 39 Q4 19 (51%) 12 (32%) 6 (14%) 37 120 (50%) 75 (32%) 43 (18%) 238 Total

Results
The results of survey 2 (table 1) show significant variation according to age-adjusted physical fitness in blood concentrations of triglycerides, heart rate, systolic blood pressure, body-mass index, amounts of physical activity, vital capacity, smoking status, and exercise-ECG 760

Cardiovascular Cancer Other causes All causes

50 (52%) 27 (28%) 20 (21%) 97

Table 2: Mortality after 13 years according to quartile of physical fitness at survey 2 (PF2)

THE LANCET Vol 352 September 5, 1998

ARTICLES

Characteristic

Relative risk (95% CI) of death From any cause (n=238) Cardiovascular cause (n=120) 252 (167377) 047 (026086) 050 (030086) 066 (042105) 100 176 (120260) 088 (052149) 112 (094133) 134 (114159) 092 (074115) 100 (083121) 102 (085122) 113 (097132) 137 (089209)

Age (increase of 10 years) Age-adjusted physical fitness (relative to Q1) Q4 Q3 Q2 Q1 Smoking status (smokers vs non-smokers) Physical activity (active vs non-active) Resting heart rate (increase of 101 bpm)* Systolic blood pressure (increase of 177 mm Hg)* Vital capacity (increase of 839 mL)* Total cholesterol (increase of 12 mmol/L)* Triglycerides (increase of 090 mmol/L)* Body-mass index (increase of 27 kg/m2)* Exercise ECG (positive vs negative)

273 (206364) 045 (029069) 048 (033071) 072 (052099) 100 163 (124215) 086 (059126) 116 (102131) 116 (102132) 089 (076104) 106 (093122) 097 (085111) 112 (100126) 110 (080152)

*Relative risk for an increase of 1 SD; value of 1 SD given in parentheses. p<0001; p<005; p<001.

Table 3: Relative risks of all-cause and cardiovascular mortality among 1428 apparently healthy men of survey 2 during 13 years follow-up

changes in physical fitness (log[PF2/PF1]), showed that changes in physical fitness have a highly significant effect on all-cause mortality (p<0001, table 5). PF1 had significant predictive power in proportional-hazards models also including PF2 (p<001, data not shown). Changes in physical fitness between PF1 and PF2 were associated with changes in other measured variables (details not shown). The group with the lowest PF2/PF1 ratio (Q1) became less physically active by survey 2 (148% physically active at survey 1, 109% physically active at survey 2). By contrast, men with the highest PF2/PF1 ratio (Q4) increased their physical activity over the same period (186% physically active at survey 1, 293% physically active at survey 2). We also found similar differences between groups Q1 and Q4 for sequential changes in systolic blood pressure, vital capacity, and resting heart rate. Moreover, in all PF1 quartiles, the groups with the lowest PF2/PF1 ratio were twice as likely to be smokers as those with the highest ratio.

Discussion
Our data from a group of healthy men show that not only fitness itself, but also the magnitude and direction of changes in fitness observed over a period of 7 years, give important information on the risk of death during a follow-up period of up to 15 years. Good physical fitness suggests the presence of normal cardiorespiratory
Characteristic and quartile of PF2/PF1 ratio Q1 (PF1) Q2 (PF1) Q3 (PF1) Q4 (PF1)

function, efficient oxygen transport and uptake, muscular and skeletal function, and psychological fitness. Physical fitness may therefore be a good surrogate measure of a number of important body functions, and therefore a good predictor of health outcome. Our data probably give a conservative estimate of the relation between physical fitness, health, and death. Many of the 328 men who were excluded from the present material for health reasons had low physical fitness at survey 1, as had most of those who had died, or who did not participate in survey 2 (data not shown). Participants were not randomly selected from the population at large, being apparently healthy men who were willing to participate in the study. However, the low attrition rate suggests that little additional selection bias appeared during the study period. Men who were unable to perform the exercise test effectively, or who were anxious about the test, would have achieved a lower physical-fitness score than they were in fact capable of. This effect would have lowered the mean physical score in these men and would have biased our results by tending to dilute true assocations. A low maximum heart rate may be due to lack of effort, but chronotropic insufficiency may also be associated with increased risk of coronary death.13,14 However, none of the men in our study were excluded because of poor chronotropic responses, and exclusion of those men with
Characteristic Age (increase of 10 years) Age-adjusted PF1 (relative to Q1 of PF1) Q4 (PF1) Q3 (PF1) Q2 (PF1) Q1 (PF2) Log (PF2/PF1) (increase of 0128)* Smoking status (smokers vs non-smokers) Physical activity (active vs non-active) Resting heart rate (increase of 101 bpm)* Systolic blood pressure (increase of 177 mm Hg)* Vital capacity (increase of 839 mL)* Total cholesterol (increase of 12 mmol/L)* Triglycerides (increase of 090 mmol/L)* Body-mass index (increase of 27 kg/m2) Exercise test (positive vs negative) Relative risk (95% CI) of death (n=238 events) 250 (188333) 060 (039091) 091 (064130) 096 (068135) 100 070 (062079) 160 (122211) 083 (057122) 117 (103132) 115 (101131) 090 (077105) 105 (092120) 097 (085112) 112 (100126) 111 (081153)

Mean change in physical fitness* Q1 (PF2/PF1) 61 Q2 (PF2/PF1) 85 Q3 (PF2/PF1) 103 Q4 (PF2/PF1) 136 13-year all-cause mortality Q1 (PF2/PF1) Q2 (PF2/PF1) Q3 (PF2/PF1) Q4 (PF2/PF1) Total 28/89 20/89 16/90 10/89 74/357

62 84 99 124 30/89 17/90 7/89 9/89 63/357 119 077 033 043

62 85 98 118 23/89 14/89 15/90 9/89 61/357 087 062 060 040

62 81 92 110 20/89 11/90 5/89 4/89 40/357 073 046 017 017

Standard (1990) all-cause mortality ratios Q1 (PF2/PF1) 122 Q2 (PF2/PF1) 080 Q3 (PF2/PF1) 072 Q4 (PF2/PF1) 047

*Ratio=(PF2/PF1) 100; values<100=decrease in physical fitness from survey 1 to survey 2; values >100=increase from survey 1 to survey 2.

*Relative risk for an increase of 1 SD; value of 1 SD given in parentheses. p<0001; p<005.

Table 4: Changes in physical fitness from survey 1 (PF1) to survey 2 (PF2) and changes in all-cause mortality

Table 5: Relative risk of all-cause mortality during 13 years follow-up associated with variables measured at survey 2, with both age-adjusted PF1 and log (PF2/PF1) in the model

THE LANCET Vol 352 September 5, 1998

761

ARTICLES

the lowest maximum heart rate (10% of the total study group) did not significantly affect our results. Adjustment for body size in our definition of physical fitness was of marginal significance, and changes in bodyweight over time did not affect our results for physical fitness or changes in fitness. Up to 30% of the men in survey 1 maintained or even increased their fitness by survey 2. Some men may have had undetected, intercurrent disorders at survey 1, which led to poor exercise-test results, but the overall increase in fitness was more likely to be linked to increases in leisuretime activity, giving up smoking, or both.15,16 It may be relevant that all participants in survey 1 received a written report of the results, and that those who were less fit received various recommendations for improving their health. Increased public awareness of the beneficial effects of giving up smoking and increased exercise may also have played a part in the observed increase in fitness during follow-up. Physical fitness is influenced by genetic factors (40%) and other factors (60%).17 Genetic causes of differences in physical fitness are difficult to measure. The effects of physical activity on fitness are more easily assessed. The most rapid declines in fitness between survey 1 and survey 2 were probably caused by changes in lifestyle, development of subclinical disease, or both. The beneficial effects on long-term physical fitness of giving up smoking, and the detrimental effects of continuing to smoke, are well known.16 A good standard of physical fitness has beneficial effects on serum lipid concentrations, fibrinolysis, glucose tolerance and insulin metabolism, platelet function, blood pressure, autonomic-system function, myocardial electric stability, dimensions of the coronary arteries,5 and the immune system.18 In our study, good physical fitness was associated with a favourable risk-factor profile, and improvements in physical fitness were associated with improvements in risk-factor profile. Followup studies based on single tests of physical fitness17 show similar results to ours, irrespective of selection procedures, methods, and followup, as do published data on physical activity.1926 To our knowledge, only one other study has used data on repeated tests of physical fitness,27 and the results accord with ours. Although our study was observational, the fitness data were corrected for several important confounders and the results seem biologically plausible. A large-scale, prospective, randomised study is needed to test these theories further, but such a study may not be possible. Many people rely on drug therapies and medical interventions to improve their health. We have shown that physical fitness is associated with a lowered risk of death, and that improvement in physical fitness over time also further reduces this risk. The men in our study who were and remained physically fit had the most favourable prognosis. However, according to these data, probably the most important suggestion is that moderate improvements in physical fitness, particularly among those who are least fit, bring substantial benefits to health.
Contributors
Gunnar Erikssen designed the study, analysed the data, and had the main responsibility for writing the paper. Knut Liestl provided scientific guidance and contributed to data analysis. Jan Erikssen initiated the first survey in 1972. Jrgen Bjrnholt undertook data collection. Erik Thaulow collected data and was involved in planning the study. Leiv Sandvik did

statistical analysis of the database. All investigators contributed to the writing of the paper.

Acknowledgment
We thank I Erikssen for valuable secretarial assistance.

References
1 Erikssen J. Physical fitness and coronary heart disease morbidity and mortality: a prospective study in apparently healthy, middleaged men. Acta Med Scand 1986; 711 (suppl): 18992. 2 Ekelund L-G, Haskell WL, Johnson JL, Whaley FS, Criqui MH, Sheps DS. Physical fitness as a predictor of cardiovascular mortality in asymptomatic North American men: the Lipid Research Clinics mortality follow-up study. N Engl J Med 1988; 319: 137984. 3 Slattery ML, Jacobs DR. Physical fitness and cardiovascular disease mortality: the US Railroad Study. Am J Epidemiol 1988; 127: 57180. 4 Blair SN, Kohl HW, Pfaffenbarger RS, Clark DG, Cooper KH, Gibbons LW. Physical fitness and all-cause mortality: a prospective study of healthy men and women. JAMA 1989; 262: 2395401. 5 Sandvik L, Erikssen J, Thaulow E, Erikssen G, Mundal R, Rodahl K. Physical fitness as a predictor of mortality among healthy, middle-aged Norwegian men. N Engl J Med 1993; 328: 53337. 6 Blair SN, Kampert JB, Kohl HW, et al. Influences of cardiorespiratory fitness and other precursors on cardiovascular disease and all-cause mortality in men and women. JAMA 1996; 276: 205210. 7 Barinaga M. How much pain for cardiac gain? Science 1997; 276: 132427. 8 Murray CJL, Lopez AD. Evidence-based health policylessons from the Global Burden of Disease Study. Science 1996; 274: 740-43. 9 Erikssen J. Aspects of latent coronary heart disease: a prevalence and methodological validation study of apparently healthy, working, middle-aged men. PhD thesis, University Hospital (Rikshospitalet) Medical Department B, Oslo, Norway: 1978. 10 Erikssen J, Skrede S. Serum lipids and latent coronary insufficiency. Scand J Clin Lab Invest 1977; 37: 24350. 11 Rose G, Blackburn H. Cardiovascular survey methods: WHO monograph series 56. Geneva: WHO, 1968. 12 Mundal R, Erikssen J, Rodahl K. Assessment of physical activity by questionnaire and personal interview with particular reference to fitness and coronary mortality. Eur J Appl Physiol 1987; 56: 14552. 13 Ellestad MH, Wan MKC. Predictive implications of stress testing. Circulation 1975; 51: 36369. 14 Sandvik L, Erikssen J, Ellestad MH, et al. Heart rate increase and maximal heart rate during exercise as predictors of cardiovascular mortality: a 16-year follow-up study of 1960 healthy men. Coron Artery Dis 1995; 6: 66778. 15 Hardman A, ed. Physical activity and cardiovascular risk: review in depth. J Cardiovasc Risk 1995; 2: 285329. 16 Sandvik L, Erikssen G, Thaulow E. Long-term effects of smoking on physical fitness and lung function: a longitudinal study of 1393 middle aged Norwegian men for seven years. BMJ 1995; 311: 71518. 17 Bouchard C, Lesage R, Lortie G, et al. Aerobic performance in brothers: dizygotic and monozygotic twins. Med Sci Sports Exerc 1986; 18: 63946. 18 Eichner ER, Calabrese LH. Immunology and exercise: physiology, pathophysiology and implications for HIV infection. Med Clin North Am 1994; 78: 37788. 19 Pfaffenbarger RS, Hyde RT, Wing AL, Hsieh CC. Physical activity, all-cause mortality, and longevity of college alumni. N Engl J Med 1986; 314: 60513. 20 Powell KE, Thompson PD, Caspersen CJ, Kendrick JS. Physical activity and the incidence of coronary heart disease. Annu Rev Public Health 1987; 8: 25387. 21 Leon AS, Connett J, Jacobs DR, Rauramaa R. Leisure-time physical activity levels and risk of coronary heart disease and death: the Multiple Risk Factor Intervention trial. JAMA 1987; 258: 238895. 22 Pfaffenbarger RS, Hyde RT, Wing AL, Lee IM, Jung DL, Kampert JB. The association of changes in physical-activity level and other lifestyle characteristics with mortality among men. N Engl J Med 1993; 328: 53845. 23 Thompson WG. Exercise and health: fact or hype? South Med J 1994; 87: 56774. 24 Lemaitre RN, Heckbert SR, Psaty BM, Siscovick DS. Leisure-time physical activity and the risk of nonfatal myocardial infarction in postmenopausal women. Arch Intern Med 1995; 155: 230108. 25 OConnor GT, Hennekens CH, Willett WC, et al. Physical exercise and reduced risk of nonfatal myocardial infarction. Am J Epidemiol 1995; 142: 114756. 26 Thune I, Brenn T, Lund E, Gaard M. Physical activity and the risk of breast cancer. N Engl J Med 1997; 336: 126975. 27 Blair SN, Kohl HW, Barlow CE, Pfaffenbarger RS, Gibbons LW, Macera CA. Changes in physical fitness and all-cause mortality. JAMA 1995; 273: 109398.

762

THE LANCET Vol 352 September 5, 1998