In this study we found an inverse association between physical activity and lung cancer risk. We also found evidence for a positive association between lung cancer risk and white blood cell count, but not serum albumin. It has been hypothesized that physical activity may lower lung cancer risk by reducing chronic inflammation. However, we found no evidence that the associations of physical activity and white blood cell count with lung cancer risk were mediated through the same biological pathway.
Clearly smoking is a strong causal factor of lung cancer in both men and women, with a population attributable risk of approximately 75–90% in the United States (39
). Smoking prevention and cessation are the primary prevention strategies needed to reduce lung cancer incidence. However, the elucidation of other risk factors would aid in lung cancer prevention, particularly in never and former smokers, in whom about 50% of all new lung cancers are diagnosed (41
). This study adds additional evidence to the body of literature which suggests that physical activity is a protective factor against the development of lung cancer.
We observed an inverse association between physical activity and lung cancer at the upper end of the 10–40% range of risk reductions observed in the majority of past studies (14
). Given the strong relation between smoking and lung cancer risk, residual confounding of the relation between lung cancer risk and both physical activity and white blood cell count remains a concern. In models adjusted for sex, body mass index, alcohol, and education, but not smoking, the relations between lung cancer and physical activity and white blood cell count were stronger (HR = 0.43 and HR = 5.05, for third tertile vs. first tertile of total physical activity index and white blood cell count, respectively) than in models fully adjusted for smoking (HR = 0.55 and HR = 2.81, respectively). Thus it is possible that better measurement of smoking (e.g. more accurate reporting, biomarkers of smoking history) would further attenuate our findings. However, we were able to adjust for a number of prospectively-obtained self-reported smoking parameters, including smoking status, amount of smoking (pack-years) and time since smoking cessation. In analyses stratified by smoking status, physical activity appeared to be associated with reduced lung cancer risk among never and former smokers combined, though this did not reach statistical significance. Too few cases were observed among never smokers (N = 16) to examine this stratum separately. The relation between smoking and adenocarcinoma is weaker than for other cell types (42
). In our data, adenocarcinoma was associated with white blood cell count but not total physical activity index score. Though this was based on only 31 events, it suggests additional caution in interpreting the physical activity/lung cancer association.
Exercise is associated with reduced systemic inflammation (particularly C-reactive protein) both between persons in cross-sectional studies and within persons after the initiation of training regimes (21
). Inflammation has been proposed to promote carcinogenesis in a wide spectrum of cancers, including lung, through its effects on cell proliferation, survival, and migration (24
). Inflammatory lung conditions, such as chronic bronchitis and asthma, have previously been linked with increased lung cancer risk (43
). Furthermore, the use of aspirin and other non-steroidal anti-inflammatory drugs has been associated with reduced lung cancer risk (44
We investigated the relation between two inflammatory markers and lung cancer. White blood cell count is a widely-used nonspecific marker of systemic inflammation (26
). We observed reduced white blood cell counts in participants who reported higher physical activity levels, consistent with previous findings (19
). Notably, we found that this relation persisted after adjustment for self-reported smoking history. Three studies have reported positive associations between white blood cell count and lung cancer incidence or mortality after adjustment for smoking (30
). Similar to our study, Shankar et al. (46
) reported increased lung cancer mortality among subjects in the upper quartile of white blood cell count compared to those in the lowest quartile (RR = 2.58, 95% CI: 0.72–9.26 for quartile 4 vs. quartile 1). The results from our study (incidence HR = 2.81; 95% CI: 1.58–5.01; mortality HR = 3.75, 95% CI: 1.89–7.42) and Shankar et al. (46
) provide greater risk estimates than those for quartile 4 versus quartile 1 of white blood cell count in Erlinger et al. (47
) (mortality HR = 1.79, 95% CI: 0.88–3.62) and the recently reported results of the Women’s Health Initiative (30
) (incidence HR = 1.63, 95% CI: 1.35–1.97). The Women’s Health Initiative observed little difference between lung cancer incidence and mortality hazard ratios in relation to white blood cell count.
Serum albumin is a negative acute phase protein - its concentration in the blood is reduced in response to inflammation (49
). At least one study has reported an approximate 25% reduction in cancer mortality among middle aged men with a one standard deviation increase in serum albumin (51
). We observed little difference in serum albumin among participants according to physical activity level, and no association between serum albumin and lung cancer risk.
To investigate the hypothesis that physical activity lowers lung cancer risk by decreasing systemic inflammation, we further adjusted the regression model of physical activity and lung cancer risk for white blood cell count. In an adequately adjusted model, one would expect the association between physical activity and lung cancer risk to be attenuated if the relation was mediated at least in part by inflammation (represented by white blood cell count) (52
). However, we found that the associations between lung cancer risk and both physical activity and white blood cell count were practically unchanged after simultaneous adjustment. Thus the effect of physical activity on lung cancer risk does not appear to be mediated by inflammation, as represented by white blood cell count. Importantly, white cell blood count is only one marker of inflammation; it remains possible that other measures of inflammation may be more relevant to the relation of physical activity and lung cancer.
Physical activity has been proposed to lower lung cancer risk by a variety of other mechanisms. Physical activity might reduce the concentration of carcinogenic agents in the airways, reduce the duration of agent-airway interaction, and reduce particle deposition through increased ventilation and perfusion (53
). Physical activity also reduces insulin-like growth factor levels which stimulate cell proliferation (54
). Furthermore, physical activity may enhance immune function or endogenous antioxidant defenses (17
A number of limitations must be considered in the interpretation of this study. We utilized a simple assessment of physical activity. Although an increased heart rate is an objective measure associated with lack of physical activity (57
), heart rate is also modified by general health, stress, and other psychosocial factors. Questions regarding the number of blocks walked per week and flights of stairs climbed per day have previously been used in combination with data on recreational physical activity to measure the relation between physical activity and cancer risk in the Harvard Alumni Health Study (5
). We did not collect data on specific participation in recreational physical activities, but rather episodes of sweat-inducing activities. A moderate correlation (r = 0.54–0.62) has been reported between episodes of sweat-inducing activities and the Harvard Alumni Activity Survey scores (60
), including one study in a population of older women (62
). The association between sweat-inducing activities and physical fitness measured on a cycle ergometer, however, has been reported to be stronger in men (r = 0.46) than in women (r = 0.26) (60
). Our summary physical activity measure which combined blocks walked, stairs climbed, and sweat-inducing activities was more strongly related to lung cancer risk among men than in women (), though the test for effect modification did not reach statistical significance (Pinteraction
The limited scope of our physical activity assessment failed to capture variation in intensity and duration of sweat-inducing activities. To create our total physical activity index we assumed a typical duration of 30 minutes for sweat-inducing activities, with an intensity level equivalent to jogging (MET = 7). The results did not appear sensitive to variation in these assumptions: assuming a MET score of 5 for 30 minutes or a MET score of 9 for 1 hour for sweat-inducing activities both resulted in HR = 0.55 for the third tertile of total physical activity index compared to the first tertile.
Notably, our physical activity assessment also failed to capture past history of physical activity. Our failure to capture variation in duration, intensity, and past history of activity would be expected to attenuate the reductions in risk observed in our study. Much more sophisticated assessments of physical activity have been developed since the initiation of our study. Further studies are necessary, in particular, to evaluate lung cancer risk in relation to cumulative lifetime physical activity and to discriminate the effects of physical activity during different time periods in life.
Other unmeasured aspects of a healthy lifestyle may confound the relation between physical activity and lung cancer. A diet high in fruits and vegetables has been associated with reduced lung cancer risk (63
). Unfortunately we had limited information on diet and were unable to control for this in our analysis.
Strengths of this study included a population-based cohort of both sexes with excellent follow-up, the prospective assessment of physical activity and inflammatory markers, and the ability to control for a number of prospectively-obtained smoking parameters. It is possible that lower levels of physical activity among future cases might be expected at the baseline exam due to symptoms, such as pain or fatigue, related to undiagnosed lung cancer. To reduce the potential for this bias, we excluded all lung cancer cases who were diagnosed within 12 months of the baseline examination (N= 13). Other diseases, particularly of the lung, may also influence physical activity, inflammation, and lung cancer risk. However, we observed little change in the relations between lung cancer risk, physical activity, and white blood cell count after adjusting for self-reported emphysema and diabetes.
Lung cancer is both the most common cancer diagnosis in the world and the most common cause of death from cancer (64
). The global burden of smoking-related disease is overwhelming, with over 1.3 million new cases of lung cancer and approximately 1.2 million deaths in 2002 (64
). Smoking prevention and cessation are imperative in reducing the mortality associated with this disease. Continued study of physical activity in relation to lung cancer risk, particularly among never smokers, may further our understanding of this disease and reveal additional strategies for reducing its burden.