Abstract and Introduction
Abstract
Depression has long been hypothesized to be associated with cancer incidence. However, there is evidence for a positive publication bias in this field. In the present study, we examined the association between various measures of depression and cancer incidence at several sites. A total of 14,203 members of the French GAZEL (Gaz et Electricité) cohort (10,506 men, 3,697 women) were followed up for diagnoses of primary cancers from January 1, 1994, to December 31, 2009. All medically certified sickness absences for depression recorded between January 1, 1990, and December 31, 1993, were compiled. Depressive symptoms were self-reported in 1993, 1996, and 1999 with the Center for Epidemiologic Studies Depression Scale. During a mean follow-up period of 15.2 years, 1,119 participants received a cancer diagnosis, excluding nonmelanoma skin cancer and in situ neoplasms. Considering 6 cancer sites (prostate, breast, colorectal, smoking-related, lymphoid and hematopoietic tissues, other sites) and 4 measures of depression, we found 1 positive association and 1 negative association. Overall, there was no compelling evidence for an association between depression and cancer incidence. Such null results should be considered when addressing concerns of cancer patients and their relatives about the role of depression in cancer onset.
Introduction
Depression has long been hypothesized to be associated with the risk of cancer. Early findings suggested that self-reported depressive symptoms might be associated with increased cancer incidence and mortality. However, subsequent well-designed, large-scale prospective studies failed to provide compelling evidence for this relationship, yielding null or even negative results. In a 1994 meta-analysis, McGee et al. concluded that there was a weak, nonsignificant trend regarding the association between depression and cancer incidence.
This topic attracted research attention again in 1998, when Penninx et al. found a nearly 2-fold increased cancer risk among persons who experienced chronic or recurrent depressive symptoms. Most epidemiologic studies in this field assess depressive symptoms only once. Instead, Penninx et al. repeatedly administered the Center for Epidemiologic Studies Depression Scale (CES-D) and defined chronic or recurrent depressive symptoms as 3 CES-D scores above the validated cutoff, with a 3-year interval between measures.
We are not aware of a formal replication of the study by Penninx et al. However, several epidemiologic studies with positive, null, or negative results were published afterwards, making the case for 2 independent meta-analyses. In the former, Oerlemans et al. found a weak, nonsignificant trend regarding the association between major depression and cancer incidence. In the latter, Chida et al. addressed several "stress-related psychosocial factors" and found depressive symptoms to be significantly associated with the incidence of cancer (odds ratio = 1.29, 95% confidence interval (CI): 1.14, 1.46). This result, however, should be tempered by evidence for a publication bias favoring positive results and by the methodological limitations of the studies included.
Given these limitations, the current literature does not allow us to either rule out or confirm a role for depression in cancer onset. The growing prevalence of both cancer and depression in the general population, as well as concerns regarding their relationship among health-care professionals, patients, and their relatives, warrants a more authoritative answer from epidemiologists. Therefore, we took advantage of data from a large-scale French prospective study, the GAZEL (Gaz et Electricité) cohort study, to examine the association between various measures of depression (repeated self-reports based on the CES-D and clinician-based diagnoses of depression) and incidence of cancer at multiple sites.