© 1999 by Oxford University Press
Journal of the National Cancer Institute, Vol. 91, No. 12, 1080,
June 16, 1999
© 1999 Oxford University Press
CORRESPONDENCE |
Re: Chronically Depressed Mood and Cancer Risk in Older Persons
Affiliations of authors: D. Serraino, S. Franceschi (Epidemiology Unit), L. Fratino, V. Zagonel (Medical Oncology), IRCCS Centro di Riferimento Oncologico, Aviano, Italy; P. Pezzotti, Centro Operativo AIDS, Istituto Superiore di Sanità, Rome, Italy.
Correspondence to: Diego Serraino, M.D., Epidemiology Unit, IRCCS, Lazzaro Spallanzani, Via Portuense 292, 00149, Rome, Italy (e-mail: serrainod{at}ets.it).
A report by Penninx and colleagues (1) reported the results of a longitudinal study suggesting that chronically depressed elderly individuals have a nearly twofold increase in cancer risk. The role of depression as a risk factor for cancer has been widely debated, and the approach used by Penninx and colleagues to assess it among elderly patients contributes to the issue. We would like to stress some aspects of the study design and the statistical analysis that may have influenced the reported association.
From an original population of nearly 10 000 individuals interviewed in the first study year (i.e., 1982), only 4825 persons constituted the group from which the findings of the study were drawn. Of the approximately 5000 subjects not followed up, 1216 were excluded because they were diagnosed with cancer or were using anticancer drugs within 3 years before 1988—the year in which the cancer assessment started. Thus, the population initially evaluated for the presence of chronically depressed mood and the population evaluated for cancer substantially differed: Approximately one half of the patients examined for depression were not assessed for cancer risk.
The end point of the study (namely, cancer) was, in fact, a major criterion of exclusion from follow-up, and a differential length-biased sampling might have occurred (2). A similar bias has been described among prevalent cohorts of people with human immunodeficiency virus infection (3). By excluding those individuals enrolled in the study in 1982, and who developed cancer between 1985 and 1988, Penninx and colleagues might have depleted their three cohorts of the case subjects at highest risk for cancer. To what extent this depletion was associated with depression cannot be ascertained from the data presented in the paper. Hypothetically, one can assume that individuals identified as being not chronically depressed in 1982 were at a higher cancer risk than the depressed individuals, and, thus, they developed cancer before 1988. The inclusion of these cases in the follow-up discussion would have reversed the study results.
Two other issues are worth noting. First, the criteria adopted in the proportional hazard model to evaluate the assumption of proportionality of the risks between the two depression score groups might have been misleading, since the lack of interaction between age and sex does not seem to be sufficient to support the hypothesis of proportionality. To better evaluate such an assumption, we think that a time-dependent covariate transformation of the original variable should be added into the model (4). Second, the residual uncontrolled variability could hide an important risk factor associated with both cancer and depression. According to such a possibility [as already seen, for instance, in the association between smoking, human papillomavirus infection, and cervical cancer (5)], the effect of depression could actually be attributable to residual confounding.
In conclusion, we think that the important effort undertaken by Penninx and colleagues to identify older individuals with chronically depressed mood over a long time period may have been negatively counterbalanced by the exclusion of prevalent cancer cases. We are aware that this type of study among elderly patients presents several methodologic difficulties, but we would like to stress that an alternative analysis, in which time periods for the assessment of depression and of cancer are closer, may offer a further contribution to understanding the relationship between depression and cancer.
REFERENCES
1
Penninx BWJH, Guralnik JM, Pahor M, Ferruci L, Cerhan JR,
Wallace RB, et al. Chronically depressed mood and cancer risk in older persons. J Natl
Cancer Inst 1998;90:1888-93.
2 Sackett DL. Bias in analytic research. J Chronic Dis 1979;32:51-63.[CrossRef][Web of Science][Medline]
3
Alcabes P, Pezzotti P, Philipps AN, Rezza G, Vlahov D.
Long-term perspective on the prevalent-cohort biases in studies of human immunodeficiency
virus progression. Am J Epidemiol 1997;146:543-51.
4 Kalbfleish JD, Prentice RL. The statistical analysis of failure time data. New York (NY): John Wiley and Sons; 1980.
5 Phillips AN, Smith GD. Smoking and human papillomavirus infection. Causal link not proved.BMJ 1993;306:1268-9.cancerlit;93271810
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||