Year: 1999
Type: article
Abstract: Serraino and colleagues raise a few questions about our study design and the statistical analysis used that we would like to address. First, it is not true that the study population initially evaluated for the presence of chronically depressed mood differed from the study population evaluated for cancer. The assessment of both depression and cancer started in 1982 for all persons included in our study. Although we had data about the hospital discharge diagnoses of cancer only for the 3 years before 1988, the assessment of prevalent cancer was further based on the self-report of cancer prior to 1982 and during any of the annual interviews between 1982 and 1988 and the use of anticancer drugs during these 6 years. In our report, we excluded persons who reported a history of cancer before 1982 and who had cancer between 1982 (and not 1985 as mentioned by Serraino et al.) and 1988, since a possible cause-effect relationship between depression and cancer cannot be determined in these persons. If we had not excluded persons with prevalent cancer from our study population, our study would have been biased by the possibility that the presence of cancer may cause depression. In persons with prevalent cancer, it can never be ascertained whether depression caused a higher risk of cancer or whether depression is the result of a cancer diagnosis. Thus, a further examination of depressed mood status between 1982 and 1988 in persons with cancer, diagnosed between 1982 and 1988 would not give any further information about the role of chronic depression in the onset of cancer. A statistical issue raised by Serraino et al. is the criterion adopted to evaluate the assumption of proportionality of hazards in the proportional hazard model. As written in the statistical analyses section (1), we did use the time-dependent covariate transformation of the chronic depressed mood variable to check the proportionality of hazards. In addition, we used a Kaplan-Meier analysis to check the association between chronically depressed mood and cancer incidence, since this analysis does not assume proportionality of hazards. This analysis confirmed the association between chronic depression and incidence of cancer. Finally, we agree with the comment of Serraino and colleagues that residual uncontrolled variability could hide an important risk factor associated with both cancer and depression. Therefore, we adjusted our analyses for important potential confounding variables, such as age, sex, ethnic origin, physical disability, cigarette smoking, and alcohol intake. Since an increased risk for cancer was found for both depressed smokers and depressed nonsmokers and the anatomic sites at which chronically depressed persons showed excess cancers were not predominantly tobacco related, it is not very likely that smoking habits explain the increased cancer risk among chronically depressed persons in our study. As in any observational study, we can never completely rule out the possibility that an unmeasured confounder explains the observed relationship.
Authors: B. W. J. H. Penninx, J. M. Guralnik
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2011008407