In this investigation into the effects of previous lung diseases on lung cancer risk, we found associations with increased cancer risk for each of the diseases of interest. Comparisons among all histologic subgroups were consistent with increases in risk observed overall, with the exception of squamous cell carcinoma among persons with tuberculosis. Risk estimates were consistent across smoking subgroups; estimates were elevated in all subgroups, with the exception of chronic bronchitis. Our results among never smokers suggest an effect of previous lung diseases on lung cancer risk independent of tobacco smoking, probably acting through the inflammatory response and pathogenesis associated with the diseases.
The results of this pooled analysis corroborate the results of the previous meta-analysis suggesting that there was a large difference in the prevalence of COPD/emphysema among cases and controls (16
). This difference in prevalence among cases and controls may explain/confound the differential effects observed in genetic epidemiologic studies of lung cancer in which inconsistent effects have been observed among populations of similar genetic ancestry (41
) or may act as mediators in the associations between the variants and lung cancer risk (42
). Although chronic bronchitis and emphysema are commonly grouped together as COPD, we calculated detailed results for each condition separately in order to allow for differential effects of these two conditions, which have different pathologies and etiologies. Because we observed independent effects of both of these diseases when adjusting for the other in a fixed-effects analysis, we felt this to be a beneficial approach.
Reverse causality and the issue of temporality are paramount to the consideration of causality for these associations. It is certainly possible that some of the conditions were early manifestations or symptoms of lung cancer that were misdiagnosed, particularly for emphysema and chronic bronchitis. For pneumonia and tuberculosis, infections may have been the result of a weakened immune system due to lung cancer. In addition, tumors may have been interpreted as lesions from infections prior to cancer diagnosis. To address these issues, we conducted a latency analysis which found that diagnoses of the previous lung diseases more than 5 years and more than 10 years prior to cancer diagnosis were positively associated with lung cancer incidence. This suggests that reverse causality is not likely to fully explain these associations. For example, when the analysis was restricted to the conditions diagnosed 10 years prior to lung cancer, chronic bronchitis remained associated with an increased risk of lung cancer (RR = 1.45, 95% CI: 1.08, 1.95). Complete results of latency analyses are available in Web Table 3
. Note that in the cohort study included in this analysis (20
), both lung disease and smoking status were ascertained at baseline and the average follow-up time to diagnosis/censoring was approximately 7 years.
The use of self-reports for measuring previous lung diseases may have introduced misclassification bias into the studies included in the pooled analysis. Quantitative techniques for each of the previous lung diseases are presently available for improved diagnostic accuracy and disease classification; however, these were not employed in any of the component studies of the analysis. When effect estimates obtained using quantitative diagnostic tools for COPD (forced expiratory volume in 1 second, quantitative computed tomography, or radiographic evidence), pneumonia (microimmunofluorescence), and tuberculosis (radiography) were pooled in the previous meta-analysis (16
), the risk estimates derived using quantitative techniques were consistent with those derived using self-reported diagnoses. The similarity between effect estimates from the cohort study included in the analysis and the pooled case-control estimates (results not shown) suggests that potential bias due to misclassification of exposure, recall bias, and reverse causality may not explain the associations completely. Although none of the studies contained in this analysis validated self-reports with medical records, self-reported COPD has been shown to have a high level of agreement with spirometry results (43
). Despite the reports of these previous studies, misclassification of exposure may have produced underestimation of the burden due to the exposures, since several investigations have shown that COPD/emphysema is present in many lung cancer patients who do not report a history of COPD (45
For pneumonia, the question of persistence of inflammation arising from a condition with clinical transience should be addressed. Because this investigation did not contain information on the number of infections or the length and/or intensity of infection, it is difficult to conceptually include pneumonia with the other diseases in terms of persistence of inflammation. However, murine models have suggested that infection from Mycoplasma pneumoniae
can lead to long-term changes in peribronchial histopathology (48
), pulmonary airflow resistance, and elevated inflammatory biomarkers long after active infection clears (49
). This suggests that inflammation resulting from pneumonia may be more long-term in nature than clinical symptoms may suggest.
It is also possible that our results, particularly among never smokers, may have been confounded from exposure to other agents such as secondhand smoke or other occupational exposures. Secondhand smoke has been associated with increased risk of lung cancer (50
) and may be related to previous lung diseases (51
). However, it is unlikely to fully explain the large effects associated with several of the previous lung diseases. When we adjusted for secondhand smoke in our analysis among never smokers, the results remained, with risk estimates changing only slightly. For example, the relative risk associated with pneumonia among never smokers changed marginally from 1.35 to 1.45. In addition, occupational exposures may have acted as confounders in the associations tested, as they have been associated with lung cancer (52
). We examined the inclusion of restricted cubic splines to check for nonlinearity in both age and smoking (pack-years) as covariates in the association models. As was previously observed (54
), nonlinear components for age and smoking were significant in the models, suggesting a deviation from linear fit; however, the effect estimates for the lung diseases pooled across studies changed minimally. Therefore, we retained linear terms in the models to avoid overdispersion in small studies when examining the within-study effects.
For those instances where heterogeneity was observed in the overall estimates (emphysema, pneumonia, tuberculosis), removal of the outlying studies led to only slight differences in the pooled estimates. For emphysema and pneumonia, where more than one study was contributing to heterogeneity, meta-regression suggested that several sources, including continent, control type, and proportion of ever smokers, all accounted for some portion of the heterogeneity (results not shown). In subgroup analyses where more than 3 studies were included in a pooled estimate, the only major difference was seen for emphysema between Europe and North America (Web Table 1
). For emphysema, differences by continent of study may be a product of different diagnostic practices across populations, since diagnostic criteria for COPD differ across continents. More specifically, the diagnostic guidelines of the British Thoracic Society and the American Thoracic Society lead to marked differences in the prevalence of COPD when applied to the same population (55
). Diagnostic differences across study locations that are not discernable from questionnaires may also explain some portion of the heterogeneity. Although these differences in diagnostic practice should produce nondifferential variation in disease status classification across cases and controls, the potential of this to influence the results should not be precluded. Several studies included in this analysis displayed COPD (emphysema/chronic bronchitis) prevalence higher than that in the community at large (Web Table 2
), where it is often largely undiagnosed (56
). For emphysema, control source contributed significantly to heterogeneity, suggesting that the differences in diagnosis in population-based settings compared with hospital-based settings may affect the prevalence of disease reported and therefore the magnitude of estimates and associated population attribution measures.
Strengths of this investigation include the large sample size and the large number of exposed persons. The use of random-effect models, although it provides wider confidence intervals, reduces the likelihood of larger studies' overly affecting pooled estimates when combining data across studies by estimating both within- and across-study variance. The inclusion of prospective data is also a strength of this pooled analysis, although the number of cases collected prospectively was comparatively smaller, whereby the biases associated with case-control studies could be comparatively evaluated.
In conclusion, we observed elevated lung cancer risks associated with previous diagnoses of emphysema, chronic bronchitis, pneumonia, and tuberculosis in this pooled analysis of primary data. The observation of relatively consistent associations between several of the previous lung diseases and lung cancer risk across smoking groups, histologic subtypes, and study designs supports a direct association with lung cancer, reducing the likelihood of confounding by tobacco exposure. The most likely explanation for the increased risk associated with these diseases is the effect of the inflammatory response within lung tissue. Recent evidence has suggested that inflammation plays a pivotal role in the development of lung cancer (12
). Inflammation may increase the risk of cancer development as an initiator or promoter through 3 processes: increased genetic mutation, antiapoptotic signaling (59
), and angiogenesis (14
Whether acting as promoters in the causal pathway or as causative agents, these diseases appear to be markers of risk for the development of lung cancer that are clinically relevant. Most importantly, when considered as a group, the lung diseases examined in this pooled analysis affect large numbers of persons. In the United States, the prevalence of emphysema is 18.5 per 1,000 persons, and the prevalence of chronic bronchitis is 43.0 per 1,000 (60
). Although the incidence of pneumonia in the United States is unknown, there were approximately 1.4 million hospital discharges associated with pneumonia in 2005 (61
). While the incidence of tuberculosis in North America is low (4.8 per 100,000 population per year) (62
), in developing nations the disease affects millions. In Europe and Asia, these conditions collectively affect millions of persons, and thus the exposed population is large (63
). Therefore, the positive associations between the lung diseases examined and lung cancer risk are of substantial public health importance, and the consistent associations suggest that a nontrivial proportion of all lung cancer cases are attributable to these other lung diseases or their underlying pathologies.
The previous lung diseases examined in this investigation are significant for both public health and clinical practice. The development of lung cancer risk prediction models (54
) should continue to incorporate the lung diseases examined in this analysis for improved discriminatory ability among all patients, regardless of smoking history. The United Kingdom Lung Cancer Screening Trial, which uses computed tomography to screen for lung cancer, utilizes the lung cancer risk prediction model of the Liverpool Lung Project, which includes pneumonia as one of the factors (62
) for selection of high-risk individuals for the trial (66
). These diseases may be useful in determining who to monitor by providing a further resolution of risk stratification, particularly as new-era screening evaluations and initiatives advance (67