Objectives Poor cardiorespiratory fitness (CRF) is associated with death from cancer. If follow-up time is short, this association may be confounded by subclinical disease already present at the time of CRF assessment. This study investigates the association between CRF and death from cancer and any cause with 42 years and 44 years of follow-up, respectively.
Setting, participants and main outcome measures Middle-aged, employed and cancer-free Danish men from the prospective Copenhagen Male Study, enrolled in 1970–1971, were included. CRF (maximal oxygen consumption (VO2max)) was estimated using a bicycle ergometer test and analysed in multivariable Cox models including conventional risk factors, social class and self-reported physical activity. Death from cancer and all-cause mortality was assessed using Danish national registers. Follow-up was 100% complete.
Results In total, 5131 men were included, mean (SD) age 48.8 (5.4) years. During 44 years of follow-up, 4486 subjects died (87.4%), 1527 (29.8%) from cancer. In multivariable models, CRF was highly significantly inversely associated with death from cancer and all-cause mortality ((HR (95% CI)) 0.83 (0.77 to 0.90) and 0.89 (0.85 to 0.93) per 10 mL/kg/min increase in estimated VO2max, respectively). A similar association was seen across specific cancer groups, except death from prostate cancer (1.00 (0.82 to 1.2); p=0.97; n=231). The associations between CRF and outcomes remained essentially unchanged after excluding subjects dying within 10 years (n=377) and 20 years (n=1276) of inclusion.
Conclusions CRF is highly significantly inversely associated with death from cancer and all-cause mortality. The associations are robust for exclusion of subjects dying within 20 years of study inclusion, thereby suggesting a minimal influence of reverse causation.
- Physical fitness
Statistics from Altmetric.com
A well-known relationship between cardiorespiratory fitness (CRF) and cardiovascular disease1 exists. The relationship between CRF and cancer mortality, however, has been less studied and is more uncertain. CRF is an expression of maximal oxygen uptake per kilogram body weight and reflects maximal metabolic ability. The current knowledge concerning the relationship between physical activity and health has been generated from epidemiological studies from the 1950s up to the present day.2 ,3
During the last decades, there has been an increased interest in the relationship between CRF and the development of cancer. There appears to be a consistent relationship between a high CRF level and lower risk of cancer across cancer types,4–8 except for prostate cancer and certain other cancers.5 ,9 ,10 Many cancers often take decades to develop from subclinical states to manifest disease. Subclinical cancer may therefore be present at the time of CRF assessment and may lead to confounding of the findings if follow-up time is short: Thus, it may be that subjects with subclinical disease have a lower physical performance and subsequently are at increased risk of dying from cancer, resulting in the observed association. The relationship between CRF and death from cancer with a very long time of follow-up may be advantageous in order to account for this potential bias.11
In the present study, we hypothesised that CRF (maximal oxygen consumption (VO2max)), assessed from a bicycle ergometer test, was inversely associated with death from cancer and all-cause mortality among 5131 middle-aged men free from cancer at baseline and followed for 42 years for cancer mortality and 44 years for all-cause mortality. The long follow-up time allowed us furthermore to investigate the association between CRF and outcome after excluding subjects dying within 2 decades of study inclusion, thereby minimising the possibility that poor health might confound the findings between CRF and death from cancer.
The Copenhagen Male Study
The Copenhagen Male Study was established in 1970–1971 as a prospective study of middle-aged men employed at 14 large workplaces in Copenhagen.12 All men were interviewed by a physician (FG) and underwent estimation of physical fitness (VO2max). Physical fitness was determined using information on heart rate and work load from a bicycle ergometer and Åstrands nomogram. This method has been shown to have a correlation of 0.87 compared with measured VO2max and a SE of 5.7 mL/kg/min.13 Heart rate was measured during submaximal bicycle work in a steady state with the aid of a stopwatch and a stethoscope. The loads used were 100, 150 and 200 W. One, two or, in a few cases, three different loads were used. The load chosen in each case was determined from the weight and age of the subject or heart rate during the first minute of the test. The method used has previously been described in detail.12 The examination also consisted of a questionnaire, a short interview and measurements of height, weight and blood pressure. From the questionnaire, information on daily physical activity, lifestyle and general health was obtained. Self-reported physical activity was determined as light, moderate or high. The subjects classified themselves as never smoker, previous smoker or current smoker. Daily tobacco consumption was calculated from information on the number of cigarettes, cheroots, cigars or the weight of pipe tobacco smoked daily. As previously estimated by means of measurements of serum cotinine, the validity of tobacco reporting was high.14 The subjects were subdivided into three social classes, using a modification of a system by Svalastoga based on the level of education and job profile.15 Subjects in social class I were predominantly academics and well-skilled administrators or executives; subjects in social class II were primarily white-collar workers and skilled workers; men in social class III had a short education or were unskilled or semiskilled workers. Ethical Approval from an ethical committee was not possible since the ethics committee did not exist before 1980.
Overall, 6125 men were invited to participate of whom 5249 men (87%) agreed. Of these, a total of 5210 subjects underwent CRF testing. Seventy-nine of these subjects had a cancer diagnosis in the Danish Cancer Registry prior to their inclusion date and were therefore excluded leaving a total of 5131 subjects who were included in this study (figure 1).
End-points and follow-up
Vital status per 14 November 2014 was collected from the Danish national registers and was used as end of follow-up in the all-cause mortality analysis. Cause of death, used as end of follow-up for the any-cancer and cancer-specific mortality analyses, was collected from death certificates until 31 December 2012. A total of 60 individuals had two cancer diagnoses listed as cause of death and these individuals were included only once in the all-cause mortality and any cancer mortality analyses but were included in the cancer-specific analyses relevant to their cause of death. Death from cancer was determined as The International Classification of Diseases (ICD)-8: 140–209 or ICD-10: C00–97. In order to explore the relationship between CRF and specific cancers, but taking into account the limited number of end points, four cancer groups were defined; oral and digestive cancers (ICD-8: 140–159, ICD-10: C00–26), respiratory and thoracic cancers (ICD-8: 160–163, ICD-10: C30–39), male and urinary organs cancers (ICD-8: 185–189, ICD-10: C60–68) and other cancers (ICD-8: 170–184, 190–209, ICD-10: C40–58, C69–97). Follow-up was 100% complete. Seven individuals emigrated and one individual changed central person register number (CPR-no.) due to sex change. In these cases, date of emigration or change of CPR-no. was determined as date of censoring.
The subjects were not involved in the design of the present study.
All analyses were performed using Stata V.12.1 (StataCorp LP, Texas, USA). For baseline demographics, the population was divided into tertiles of estimated VO2max; categorical variables were analysed with the χ2 test and continuous variables with analysis of variance (ANOVA). The association between estimated VO2max and outcome was analysed in Cox proportional hazards models, and the proportional hazards assumption was tested using log-log Kaplan-Meier and Cox survival estimates plotted against time and was found to be met. Crude and multivariable analyses including possible confounding factors were performed. Clinical variables (table 1), included in the multivariate models were age, smoking (never, former, current), grams of tobacco per day, systolic and diastolic blood pressure, Body Mass Index (BMI), previous myocardial infarction (no/yes), diabetes (no/ yes), self-reported physical activity (light, moderate, high), alcohol (light: 0–2 units/day, moderate: 3–5 units/day, heavy: >5 units/day) and social group (low, middle, upper). The association between estimated VO2max and endpoints was analysed and reported in crude and multivariable models divided into tertiles of estimated VO2max, and as a continuous variable per 10 mL/kg/min increase in estimated VO2max. Also, to minimise the risk of reverse causation, sensitivity analyses were performed excluding individuals dying within 10 and 20 years of inclusion. A p value of <0.05 was considered statistically significant.
As shown in figure 1, a total of 5131 individuals were included. As of 14 November 2014, a total of 4486 (87.4%) deaths occurred during the 44.1 years of follow-up (mean 28.3 (SD 11.4)). Until 31 December 2012, a total of 1527 subjects (29.8%) had died of cancer, thereby allowing 42.2 years of follow-up (mean 28.0 (SD 11.0)) (table 1).
Baseline characteristics by tertiles of estimated VO2max are shown in table 1. With increasing tertiles of estimated VO2max, subjects were more likely to be younger, have a lower BMI, have lower systolic and diastolic blood pressure and have a higher self-reported physical activity level. Smoking was more prevalent in the higher tertiles of estimated VO2max, whereas high alcohol consumption was more likely in the lower tertiles of estimated VO2max. Social class was not related to estimated VO2max in the baseline demographics.
Cardiorespiratory fitness and death from any cancer
CRF was highly significantly associated with death from cancer. As shown in table 2 and figure 2A, there was a clear dose–response relationship between tertiles of estimated VO2max and cancer mortality. For every 10 mL/kg/min increase in estimated VO2max, risk of death from cancer decreased by (HR (95% CI)) 0.76 (0.71 to 0.82) in the crude model and 0.83 (0.77 to 0.90) after multivariable adjustments.
Cardiorespiratory fitness and all-cause mortality
CRF was significantly and inversely associated with all-cause mortality (table 2 and figure 2B). There was a dose–response relationship between estimated VO2max and all-cause mortality. For every 10 mL/kg/min increase in estimated VO2max, risk of death decreased by 0.73 (0.70 to 0.76) in the crude model and 0.89 (0.85 to 0.93) after multivariable adjustments.
Cardiorespiratory fitness and death from specific cancer groups
The relationship between CRF and death from groups of cancer is shown in table 3.
Oral and digestive cancers displayed the same reduction in risk of death with increasing estimated VO2max compared with the whole population and was borderline significant. In the crude model, mean age of death in the oral and digestive cancer group was 72.5 years compared with 75.9 years (p<0.001) for subjects dying from other causes. Respiratory and thoracic cancers were significantly and inversely associated with CRF. Cancer mortality in this group was associated with 4.1 years shorter life expectancy (71.8 vs 76.0 years; p<0.001) compared with subjects dying from other causes. The male and urinary organs cancer group was not significantly associated with estimated VO2max. To investigate this further, the male and urinary organs cancer group was stratified into prostate cancer (ICD-8: 185, ICD-10: C61, n=231) and other cancers (including bladder, kidney cancer and others, n=142). In these analyses, the HR in the full multivariate model for prostate cancer was 1.00 (0.82 to 1.2), p=0.97 per 10 mL/kg/min increase in estimated VO2max and, for the other cancers in the male and urinary organs group, the HR was 0.77 (0.59 to 1.01), p=0.056 and, thus, similar to the overall cancer mortality. Subject who died from prostate cancer died at a higher age compared with subjects dying from other causes (77.6 vs 75.5 years; p=0.002), whereas the other cancers in this group (bladder, kidney cancer and others) showed a 2.0 years (73.6 vs 75.6 years; p=0.02) lower age at death compared with subjects dying from other causes.
The ‘Other Cancer’ group, including cancers in the central nervous system, haematological cancers and others, showed a significant and similar reduction in risk of death with increasing estimated VO2max compared with the whole population after multivariable adjustments. In the crude model, a cancer in this group was associated with 4.7 years (74.0 vs 78.7 years; p=0.005) lower mean age at end of follow-up.
Cardiorespiratory fitness and mortality: self-reported physical activity level
As shown in table 1, there was a crude association between estimated VO2max and self-reported leisure-time physical activity level. This was explored in more detail. In the multivariable models including estimated VO2max and other variables, self-reported physical activity level was not independently associated with death from cancer (light vs moderate (HR (95% CI)) 1.07 (0.93 to 1.22), p=0.36; light versus high 0.99 (0.81 to 1.21), p=0.92). Furthermore, no interaction between estimated VO2max, any cancer mortality and self-reported physical activity was observed (p≥0.77).
Cardiorespiratory fitness and mortality: excluding subjects within 10 and 20 years of follow-up
To address whether poor health or subclinical cancer contributed to the association between CRF and mortality, the analyses were repeated excluding those who died within the first 10 years and first 20 years of follow-up, as displayed in table 2. As shown, excluding subjects up to 20 years of follow-up did not influence the findings between CRF and death from all causes or any cancer where risk estimates and CIs remained largely unchanged.
In the present study, the association between CRF and mortality was investigated in 5131 working, cancer-free men included in the Copenhagen Male Study in 1970–1971 and followed for 44 years, during which 87% of the population died. The main findings are as follows: First, CRF is inversely associated with risk of death from cancer. No association between CRF and death from prostate cancer was found. The association was, however, present in other groups of cancer. Second, CRF was inversely associated with all-cause and cancer mortality independent of self-reported physical activity level. This indicates that objectively measured CRF should be used in risk prediction of cancer rather than self-reported physical activity level, which was not independently associated with death from cancer in this study. Third, excluding subjects dying within 20 years of inclusion did not change the association between CRF and mortality, suggesting that findings from the present study and other studies are not explained by pre-existing subclinical cancer. In summation, the present findings suggest that a high CRF is protective against cancer mortality and is associated with increased longevity among middle-aged, employed men.
Globally, cancer prevalence is increasing and is a major threat to healthy ageing16 and there is no indication that this will change for many decades.17 ,18 Overwhelming evidence indicates that increased physical activity is related to longevity11 ,19 ,20 and the WHO has published global recommendations for physical activity based on this evidence. Currently, the WHO ranks physical inactivity as the fourth leading cause of death worldwide21 mediated through increased risk of developing obesity, diabetes, cardiovascular disease and possibly cancer.8 In the present study, we studied working middle-aged men and assessed physical activity level and CRF by means of a submaximal exercise test. In this study, men with higher CRF level lived longer and had a decreased risk of dying from cancer, as was shown with a clear dose–response relationship. An inverse relationship between CRF and mortality has previously been shown across several epidemiological studies.1 Our study is, however, exceptional, as it includes subjects followed for almost half a century with no subjects lost to follow-up. This is, to the best of our knowledge, the longest follow-up to date of an objectively measured assessment of CRF and death from cancer.
The relationship between CRF and cancer has been investigated previously19 and is generally in line with the present findings. We found that a third of the population, totalling 1527 individuals, died from cancer. The association between CRF and cancer has been found to be particularly strong for lung and colon cancer.22 Other cancers, such as lymphoma or prostate cancer, do not show a similar consistency.5 ,23 In the present study, we investigated main groups of cancers, thereby increased statistical power. The male and urinary organ cancer group was, in the initial analysis, not associated with CRF. When examining this group in detail, we found that this was explained by death from prostate cancer, which was not associated with CRF, whereas the rest of the cancers in the group were. Our findings are in contrast with some other studies that find a decreased10 or, surprisingly, an increased risk of incident prostate cancer5 ,9 with increases in CRF. This difference in findings may be due to whether the outcome of interest is incident prostate cancer where factors such as health awareness, time of screening and detection of cancer may play a role, or death from prostate cancer, as in the present study. As shown, subjects who died from prostate cancer in general lived longer than patients dying from other causes. When examining this further, subjects dying from prostate cancer in general smoked less tobacco (14.9 vs 12.5 g/day; p=0.002). Findings from other studies investigating CRF in relation to prostate cancer may therefore have been due to residual confounding not present in the current study. Overall, cancer mortality, which was used in the present study, may be a more robust outcome measure of the association between CRF and prostate cancer.
As many cancers have a long subclinical development, reverse causation may bias the results when studying CRF and health outcomes. In order to eliminate this possibility, we excluded subjects who died within 10 years and 20 years of inclusion. As shown, this did not change the results. The results of this analysis make it more likely that previous studies with shorter follow-up finding an inverse relationship between CRF and cancer are valid and not biased by reverse causation. This supports the hypothesis that a protective effect of a high CRF for cancer exists, and that this protective effect displays a dose–response relationship.
In addition to conventional risk factors, we included social class in the multivariable analysis which previously has been shown to be a predictor of health outcomes including cancer,24 ,25 possibly reducing the likelihood of residual confounding.
The inverse association between CRF and death from cancer observed in this and other studies may be biologically plausible as animal studies in mice have found that physical exercise is able to suppress tumour growth.26 Immunological mechanisms, increased insulin sensitivity, changes in endogenous hormone production, decreased gastrointestinal transit time and other factors have been suggested as pathophysiological mechanisms.11 ,27 Another explanation may be that genetic factors contribute substantially to the degree of CRF level, as suggested by twin studies, and constitutional factors may therefore play a role.28 Furthermore, CRF increases with physical training and activity in leisure time. Accordingly, the CRF measure used in this study is likely to reflect genetic and modifiable lifestyle factors.
A possible limitation in the present study is that CRF was estimated only once using an indirect method which is known to have a correlation of 0.87 compared with more precise laboratory measurements.13 Furthermore, it is likely that individual CRF levels changed during the course of follow-up. Imprecise VO2 estimation as well as a change in CRF status over time, however, would draw the findings towards the null hypothesis, and may thus be an unlikely explanation of the results. Also, causal relationships can be assessed using the method of instrumental variables; a suitable instrument was, however, not available in this population.
CRF is inversely associated with a risk of death from cancer and all-cause mortality in a dose–response dependent manner in a large group of employed middle-aged men free from cancer at inclusion followed for up to 44 years. This association was independent of conventional risk factors including self-reported physical activity and social class, and was robust for exclusion of participants dying with 10 and 20 years following baseline measurements suggesting that reverse causation did not influence the findings.
What are the findings?
In the present study, cancer-free male subjects underwent assessment of cardiorespiratory fitness (CRF) and were subsequently followed for more than 4 decades. There was a clear inverse dose–response-related relationship between level of CRF and risk of death from cancer, indicating that high fitness level is protective against death from cancer. The very long follow-up period allowed for studying the data where subjects dying within 20 years of inclusion were excluded, which did not change the association. It is therefore not likely that underlying poor health could confound the results. Altogether, these findings suggest that high CRF is protective against death from cancer later in life.
How might it impact on clinical practice in the future?
An improvement in physical fitness level should be promoted to prevent death from cancer and increase longevity. As longevity is closely related to fitness level, measurement of maximal oxygen consumption (VO2max) can in primary prevention be used to identify individuals at risk and thereby possibly used to encourage and inspire change in lifestyle factors.
We gratefully thank Dr Martina Chantal de Knegt for language editing, and proofreading the manuscript.
Contributors MTJ contributed substantially to the conception and design of the work, analysis and interpretation of data, as well as drafting the work, approved the final version of the manuscript to be published and agrees to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. AH contributed substantially to the analysis and interpretation of data, as well as revising it critically for important intellectual content, approved the final version of the manuscript to be published, and agrees to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. HB and AH contributed substantially to the analysis and interpretation of data, as well as revising it critically for important intellectual content, approved the final version of the manuscript to be published, and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. FG contributed substantially to the acquisition, analysis or interpretation of data, as well as revising it critically for important intellectual content, approved the final version of the manuscript to be published, and agrees to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All authors had full access to all of the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. MTJ affirms that the manuscript is an honest, accurate and transparent account of the study being reported, that no important aspects of the study have been omitted, and that any discrepancies from the study as planned have been explained.
Funding The Copenhagen Male Study was supported by grants from King Christian X Foundation, The Danish Medical Research Council, The Danish Heart Foundation and Else and Mogens Wedell-Wedellsborg Foundation.
Disclaimer The funders of the Copenhagen Male Study had no role in the present study.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Full data set is available from the corresponding author. Patient consent was not obtained but the presented data are anonymised and risk of identification is low.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.