Susanne Oksbjerg Dalton, 1 Lene Mellemkjær, 1 Jørgen H. Olsen, 1 Preben B. Mortensen, 2 and Christoffer Johansen 1

American Journal of Epidemiology Copyright 2002 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 155, No. 12 Printed in U.S.A. Depression and Cancer Risk Dalton et al. Depression and Cancer Risk: A Register-based Study of Patients Hospitalized with Affective Disorders, Denmark, 1969 1993 Susanne Oksbjerg Dalton, 1 Lene Mellemkjær, 1 Jørgen H. Olsen, 1 Preben B. Mortensen, 2 and Christoffer Johansen 1 The authors investigated the cancer risk of patients hospitalized for depression in a nationwide Danish cohort study. All 89,491 adults in Denmark who had been admitted to a hospital with depression, as defined in the International Classification of Diseases, Eighth Revision, between 1969 and 1993 were identified. There were 1,117,006 person-years of follow-up. Incidence rates of all cancers and of site-specific cancers were compared with national incidence rates for first primary cancers, with data being adjusted for sex, age, and calendar time. A total of 9,922 cases of cancer were diagnosed in the cohort, with 9,434.6 having been expected; this yielded a standardized incidence ratio of (95% confidence interval (CI):, 1.07). The risk of cancer was increased for the first year after hospital admission, with brain cancer especially occurring more frequently than expected. When the first year of follow-up was excluded, the increase was attributable mainly to an increased risk of tobacco-related cancers: Standardized incidence ratios for non-tobacco-related cancers were (95% CI: 0.97, ) after 1 9 years of follow-up and (95% CI: 0.95, ) after 10 or more years of follow-up. These data provide no support for the hypothesis that depression independently increases risk of cancer, but they emphasize the deleterious effect that depression can have on lifestyle factors. Am J Epidemiol 2002;155: 1088 95. cohort studies; depression; neoplasms; registries Received for publication September 12, 2001, and accepted for publication February 24, 2002. Abbreviations: CI, confidence interval; ICD-8, International Classification of Diseases, Eighth Revision; SIR, standardized incidence ratio. 1 Danish Cancer Society, Institute of Cancer Epidemiology, DK- 2 Copenhagen, Denmark. 2 National Center for Register-based Research, Aarhus University, DK-8000 Aarhus, Denmark. Reprint requests to Dr. Susanne Oksbjerg Dalton, Danish Cancer Society, Institute of Cancer Epidemiology, Strandboulevarden 49, DK-2 Copenhagen, Denmark (e-mail: sanne@cancer.dk). The idea that depression is an etiologic factor in the development of cancer is widespread in the general population and among health professionals (1, 2). The prevailing hypothesis for such an association is that depression impairs immune function, which in turn predisposes a person to the initiation or progression of neoplastic disease (3). Increased incidences of cancer, especially of non-hodgkin s lymphoma, have been observed among patients with welldefined immunodeficiency disorders (4 6). The observed alterations in both numerical and functional measures of immunity in depressed persons (7) suggest the existence of a link between depression and cancer. Furthermore, because depression affects behavior, lifestyle factors such as diet, smoking (8, 9), and alcohol use (10) could mediate a relation with cancer. A number of longitudinal epidemiologic studies have been conducted to investigate a possible association. Some of these studies have shown evidence of an increased incidence of cancer in depressed persons (11 17), and some have shown no association ( 22). The evidence is difficult to interpret, since diverse psychometric scales have been used to define depression, and only some of the cohort studies in the literature have been large enough to evaluate the risk of site-specific cancers. In a nationwide, populationbased, long-term follow-up study in Denmark, we investigated the risks of all cancers and of site-specific cancers among patients with depression defined as persons admitted to a hospital with a diagnosis of a psychiatric affective disorder using register-based data for a 25-year period. MATERIALS AND METHODS The Danish Psychiatric Central Register has compiled computerized data on admissions to psychiatric hospitals and to psychiatric wards of general hospitals in Denmark since 1969, with coverage of 95 percent (23). The information transferred to the Register includes the patient s unique personal identification number (which incorporates sex and date of birth), the date of admission, the date of discharge, one main discharge diagnosis, and up to three auxiliary discharge diagnoses. Use of the 10-digit personal identification number, which is assigned to all residents of Denmark shortly after birth by the Central Population Register, ensured that a complete history of psychiatric hospitalization could be established for each patient, and it also permitted accurate linkage of information between registries. Discharge diagnoses in the Psychiatric Central Register were classified 1088

Depression and Cancer Risk 1089 according to the International Classification of Diseases, Eighth Revision (ICD-8) (24), through 1993, after which the Tenth Revision of the International Classification of Diseases (25) was used. We identified 447,475 hospital admissions for 95,4 persons with diagnoses of affective disorders between April 1, 1969, and December 31, 1993. The diagnoses included in the present study were categorized into four ordinal levels (table 1) according to type of psychopathology. This classification was used in a previous study, but we modified it to include only affective disorders (22). Levels 1, 2, and 3 represent psychotic disorders, and level 4 covers neurotic disorders. Of the 95,4 persons with diagnoses of depression, 699 (0.7 percent) were excluded because they were under 15 years of age at the time of admission or discharge, their personal identification number was coded incorrectly, or they died before or on the date of the first recorded admission. A diagnosis of schizophrenia prior to the first admission for depression led to the exclusion of an additional 1,662 persons (1.7 percent), on the assumption that such patients could not be considered to be suffering from an affective disorder. Data on the psychiatric patients were linked to the death certificate files of the National Board of Health to obtain information on date of death. Information on cancer occurrence was obtained through linkage to the files of the Danish Cancer Registry, which has collected information on all individuals in Denmark with a diagnosis of cancer since 1943, including benign brain tumors and urinary bladder papillomas. The cancers were classified according to the modified Danish version of the International Classification of Diseases, Seventh Revision (26). In all, 3,292 (3.5 percent) individuals in whom cancer (other than nonmelanoma skin cancer) had been diagnosed before the first admission for depression were excluded from follow-up. This left 355,3 hospital admission records for 89,491 persons for the analysis. Follow-up for cancer began on the date of the first recorded admission for depression and ended on the date of cancer (other than nonmelanoma skin cancer) diagnosis (n 8,512), death (n 29,208), schizophrenia diagnosis (n 1,948), or December 31, 1995 (n 49,823), whichever came first. The first registered diagnoses were allocated to one of the four ordinal levels (table 1), and persons with these diagnoses contributed person-years to these levels until they were readmitted with a diagnosis allocated to a lower level; they contributed person-years at the new level from that date onward. However, if a patient was readmitted with the same diagnosis or a diagnosis allocated to a higher level, he or she remained at the initial diagnostic level. Patients readmitted with a diagnosis of schizophrenia were censored for that diagnosis from the date of readmission onward. The numbers of cases of cancer observed among individuals hospitalized with depression were compared with those expected on the basis of the age-, sex-, and calendar yearspecific incidence rates of first primary cancers in Denmark. Nonmelanoma skin cancer was not considered to be a first primary cancer, and any cancer occurring subsequently was counted as the first primary cancer. The standardized incidence ratio (SIR), taken as the ratio of the observed number of cancer cases to the expected number, was used as a measure of relative risk, and 95 percent confidence intervals were calculated assuming a Poisson distribution of the observed number of cancers (27). SIRs were calculated for all types of cancer combined, for tobacco-related cancers combined (cancers of the buccal cavity, esophagus, pancreas, larynx, lung, kidney, and urinary bladder (28)), and for a number of site-specific cancers. Furthermore, data for the first year of follow-up were analyzed separately to avoid problems in ascertaining the temporality of disease. RESULTS The 89,491 individuals hospitalized with depression accrued 1,117,006 person-years of follow-up, yielding a mean length of follow-up of 12.5 years (range: 0 26.7 years). For the majority of patients, the diagnosis at the time of first admission was unipolar depression (table 2). The overall female:male ratio was 1.78, and the mean age at first admission was 48.7 years (47.0 years for men and 49.7 for TABLE 1. Four ordinal levels of diagnoses of affective disorders used in a Danish cohort study of depression and cancer risk, 1969 1993 Diagnostic level 1. Biploar illness Diagnosis Manic-depressive psychosis, manic type Manic-depressive psychosis, circular type Reactive excitative psychosis Modified ICD-8* code 296.19 296.39 298.19 2. Unipolar psychosis 3. Reactive psychosis 4. Dysthymia Involuntional melancholia Manic-depressive psychosis, depressed type Manic-depressive psychosis, other Manic-depressive psychosis, unspecified Reactive depressive psychosis Personality disorder, affective type Neurotic depression 296.09 296.29 296.89 296.99 298.09 30 300.49 * ICD-8, International Classification of Diseases, Eighth Revision (Danish version).

1090 Dalton et al. TABLE 2. Descriptive characteristics of 89,491 persons hospitalized with affective disorders in Denmark, 1969 1993 Diagnostic level* at time of study entry Characteristic All levels Bipolar psychosis Unipolar psychosis Reactive depression Dysthymia No. % No. % No. % No. % No. % Total 89,491 9,876 36,299 21,283 22,033 Sex Male Female 32,171 57,320 36 64 4,129 5,747 42 58 12,2 24,157 33 67 8,588 12,695 40 60 7,312,721 33 67 Mean duration of follow-up (years) 12.5 11.3 10.5 11.1 12.9 Mean age (years) 48.7 44.6 53.6 47.0 44.2 Year of first registration 1969 1973 1974 1978 1979 1983 1984 1988 1989 1993 28,203 23,017 16,075 11,975 10,221 32 26 13 11 2,952 2,137 1,825 1,6 1,348 30 22 16 9,662 9,404 6,879 5,348 5,006 27 26 19 15 6,130 5,430 4,248 2,913 2,562 29 26 20 12 9,459 6,046 3,123 2, 1,305 43 27 10 6 No. of hospital admissions for affective disorders 1 2 3 5 6 48,263 16,857 15,609 8,762 54 19 17 10 3,355 1,682 2,378 2,461 34 17 24 25 17,379 7,332 7,407 4,1 48 20 20 12 13,577 3,905 2,793 1,008 64 13 5 13,952 3,938 3,031 1,112 63 5 Alcohol-related diagnosis (ever) 15,494 17 1,270 13 3,616 10 4,534 21 6,074 28 * Bipolar psychosis: ICD-8 codes 296.39, 296.19, and 298.19; unipolar psychosis: ICD-8 codes 296.09, 296.29, 296.89, and 296.99; reactive depression: ICD-8 code 298.09; dysthymia: ICD-8 codes 300.49 and 30. ICD-8, International Classification of Diseases, Eighth Revision (Danish version). Number of affective admissions from time of study entry to death, cancer diagnosis, or end of follow-up. Alcohol-related diagnosis: ICD-8 code 303. women). Of the 89,491 individuals identified, 39,977 (45 percent) had had another admission for another psychiatric disorder prior to their first admission for depression. More than half of the cohort members (54 percent) were admitted to a hospital only once for depression during the study period; for a further 36 percent, the admission diagnosis was stable and the level of psychopathology did not change at subsequent admissions. Alcohol-related psychiatric disorders (ICD-8 code 303) were diagnosed in 15,494 (17 percent) of the patients during the study period. Overall, 9,922 cases of cancer were diagnosed in the cohort, with 9,434.6 having been expected; this yielded an SIR of (95 percent confidence interval (CI):, 1.07; table 3). The overall relative risk was highest during the first year of follow-up (SIR, 95 percent CI: 1.11, 1.29), was close to unity during the period 1 9 years of follow-up, and was significantly increased by 7 percent during the later period of follow-up ( 10 years). Most of the excess risk could be attributed to an increased risk of tobacco-related cancers, except during the first year of follow-up, when the risk of non-tobacco-related cancers combined was increased (SIR 1., 95 percent CI:, 1.25). Similar patterns of increased risk of tobacco-related cancers and changes with length of follow-up were seen for men and women (data not shown). Patients admitted to a hospital with bipolar illness or unipolar manic-depressive psychosis did not have an increased risk of cancer overall (table 3). However, patients in the two diagnostic levels of reactive depression and dysthymia had significantly increased risks of all cancers during the first year of follow-up. Most of the excess risk observed in the later periods of follow-up was for tobaccorelated cancers. All four groups had a substantially increased risk of brain cancer during the first year of follow-up, with SIRs ranging from 2.5 for dysthymic patients to 4.5 for patients with reactive depression. The risk of brain cancer among patients with bipolar psychosis was increased nearly three times, but not significantly so. None of the groups showed an increased risk of brain cancer after more than 1 year of follow-up. Tobacco-associated cancers and brain cancer accounted for approximately 44 percent of the 106 excess cases of cancer observed in the cohort during the first year of follow-up. Table 4 shows cancer risks after more than 1 year of follow-up, by site, for the combined diagnostic groups of bipolar or unipolar psychosis and reactive depression or dysthymia, since the cancer patterns were similar in the combined entities. Of the tobacco-associated cancers, the risk of kidney cancer did not reach significance in patients with reactive depression or dysthymia (SIR, 95 percent CI: 0.97, 1.45). No deviation from unity was observed for non-hodgkin s lymphoma in any of the diagnostic groups. A decreased risk of rectal cancer was found for both

TABLE 3. Standardized incidence ratios for all types of cancer combined and for tobacco-related and non-tobacco-related cancers, by diagnostic group, in patients hospitalized with an affective disorder in Denmark, 1969 1993 Diagnosis Total cohort Total First year of follow-up 1 9 years of follow-up 10 years of follow-up Obs* SIR*, 95% CI* Obs SIR 95% CI Obs 9,922 2,813 7,109 277 6,832 1.21 1., 1.07 1.16, 1.25,, 1.32 0.97, 654 2 472 46 426 1.37 1. 3.27 1.06 Portion of follow-up period 1.11, 1.29 1., 1.59, 1.25 2.39, 4.36 0.97, 1.17 4,655 1,224 3,431 2 3,289 SIR 95% CI Obs SIR 95% CI 1.09 1.24,, 1.16 0.97,, 1.46, 4,613 1,407 3,206 89 3,117 1.07 1.30 0.84, 1.10 1.23, 1.37 0.95, 0.67,, Diagnostic level Bipolar psychosis 1,217 292 925 26 899 0.92 0.82 0.93, 0.82, 0.93, 1.06 0.53, 1.20, 1.07 62 44 4 40 1.16 1.30 1.11 2.72 0.89, 1.49 0.77, 2.06 0.80, 1.48 0.73, 6.97 0.75, 1.42 557 133 424 12 412 0.81 0.92, 1.09 0.78, 1.11 0.93, 1.12 0.42, 1.42 0.93, 1.13 598 1 457 10 447 0.88 0.97 0.64 0.87, 0.74, 0.88, 1.06 0.31, 1.17 0.89, 1.07 Unipolar psychosis 4,345 1,4 3,201 119 3,082 0.95, 1.01, 1.11 0.93,, 1.43 0.92, 290 76 2 21 193 1.07 3.10 0.91 0.89, 1.12 0.84, 1.34 0.85, 1.12 1.92, 4.75 0.79, 2,176 538 1,638 58 1,580 0.95 1.11 0.90, 0.87, 0.90, 0.84, 1.43 0.89, 1,879 530 1,349 40 1,309 1., 1.08 1.08, 1.29 0.93, 0.70, 1.34 0.93, Reactive depression 2,075 663 1,412 59 1,353 1.13 1.41 1.20 1.08, 1. 1.30, 1.52, 1.09 0.92, 1.55 0.97, 1.08 4 58 126 112 1.62 2.03 1.48 4.55 1.36 1.39, 1.87 1.54, 2.62 1.23, 1.76 2.48, 7.63 1.12, 1.64 997 297 700 29 671 1.12 1.32 1.06 1.21, 1.17, 1.48, 1. 0.81, 1.74 0.97, 1.13 894 308 586 16 570 1.07 1.43 0.95 0.73 0.95, 1. 1.27, 1.59 0.87, 0.42, 1. 0.88, Dysthymia 2,285 7 1,571 73 1,498 1. 1.56 1.06 1.34 1.13, 1.23 1.45, 1.68 1.01, 1.12, 1.68, 1.11 1 30 88 7 81 1.32 1.58 1.25 2.54 1.09, 1.58 1.07, 2.26, 1.54, 5.24 0.95, 1.48 925 256 669 43 626 1.15 1.40 1.07 1.80 1.08, 1.23 1.24, 1.59, 1.16 1.30, 2.43 0.97, 1.13 1,242 428 8 23 791 1.68 0.82 1.13, 1.26 1.52, 1.84 0.97, 1.11 0.52, 1.23 0.97, 1.12 * Obs, observed; SIR, standardized incidence ratio; CI, confidence interval; ICD-8, International Classification of Diseases, Eighth Revision. Observed number of cases/expected number of cases. The expected number of cases was the number of cancer cases expected on the basis of age-, sex-, and calendar-yearspecific incidence rates of first primary cancers in Denmark. Cancers of the buccal cavity, larynx, lung, esophagus, pancreas, kidney, and urinary bladder. Bipolar psychosis: ICD-8* codes 296.39, 296.19, and 298.19; unipolar psychosis: ICD-8 codes 296.09, 296.29, 296.89, and 296.99; reactive depression: ICD-8 code 298.09; dysthymia: ICD-8 codes 300.49 and 30. Depression and Cancer Risk 1091

1092 Dalton et al. TABLE 4. Standardized incidence ratios for cancer in patients hospitalized with an affective disorder and followed for 1 or more years after first admission, Denmark, 1969 1993 Diagnostic level (combined categories) Site of cancer (modified ICD-7* code) Bipolar or unipolar psychosis Reactive depression or dysthymia Obs* SIR*, 95% CI* Obs SIR 95% CI Total 5,210 0.95, 1.01 4,058 1. 1.10, 1.17 Tobacco-related cancers Buccal cavity (0 8) Esophagus (150) Pancreas (157) Larynx (161) Lung (162) Kidney (0) Bladder (1) 1,342 83 50 160 39 602 136 272 0.92 1.15 0.95, 1.07 0.73, 1. 0.85, 1.51 0.89, 1.22 0.68, 1.30 0.93, 1.09 0.87, 1.23 0.92, 1.17 1,289 101 40 138 48 653 103 206 1.47 1.54 1.43 1.52 1.54 1.59 1.23 1.39, 1.55 1.26, 1.88, 1.95 1.27, 1.79 1.13, 2.04 1.47, 1.72 0.97, 1.45 1.07, 1.41 Stomach (151) Colon (153) Rectum (154) Liver (155) Breast (170) Cervix uteri (171) Corpus uteri (172) Prostate (177) Melanoma (190) Nonmelanoma skin cancer (191) Brain (193) Thyroid (194) Non-Hodgkin s lymphoma (200 and 202) Leukemia (204) and unspecified sites 3,868 135 417 5 46 790 95 172 190 112 739 120 21 102 110 634 0.97 0.87 0.93 0.79 0.70 0.92 0.97, 0.73, 0.85, 0.68, 0.92 0.73, 1.32 0.97, 1.11 0.56, 0.85 0.79, 1.07 0.83, 1.11 0.79, 1.15, 1.11 0.81, 1.16 0.64, 1.59 0.79, 1.17 0.81, 1. 0.92, 1.08 2,769 109 304 120 36 601 93 113 111 67 520 111 12 73 86 413 1.20 1.16 0.83 1.27 1.06 0.81 0.87 0.70 1.10 1. 0.81 1.22, 1.07, 1.45, 1.30 0.69, 0.89, 1.76, 1.15 0.65, 0.72, 0.84, 1.23 0.54, 0.89 1.01, 1.20 0.93, 1.37 0.42, 1.42 0.80, 1.28 0.97, 1.50 0.89, 1.08 * ICD-7, International Classification of Diseases, Seventh Revision; Obs, observed; SIR, standardized incidence ratio; CI, confidence interval; ICD-8, International Classification of Diseases, Eighth Revision. Bipolar psychosis: ICD-8* codes 296.39, 296.19, and 298.19; unipolar psychosis: ICD-8 codes 296.09, 296.29, 296.89, and 296.99; reactive depression: ICD-8 code 298.09; dysthymia: ICD-8 codes 300.49 and 30. Observed number of cases/expected number of cases. The expected number of cases was the number of cancer cases expected on the basis of age-, sex-, and calendar-year-specific incidence rates of first primary cancers in Denmark. combined groups (SIR 0.79 for bi- or unipolar psychosis and SIR 0.83 for reactive depression or dysthymia). A similar pattern was found for cervical cancer, with a reduction in risk of 19 30 percent. Risks were increased for colon cancer (SIR 1.16) and nonmelanoma skin cancer (SIR 1.10) among patients with reactive depression and dysthymia, who also had a significantly reduced risk of malignant melanoma (SIR 0.70). DISCUSSION In this study of cancer risk among persons hospitalized for depression, we found no increase in the risk of cancers at all sites when we excluded the first year of follow-up and subtracted smoking-related cancers. The risk of non- Hodgkin s lymphoma was not increased in any of the diagnostic groups, while the risks of cervical and rectal cancer were generally reduced. Our study had several advantages. It was a populationbased, nationwide cohort study with more than 1 million person-years of follow-up. The size of the study is reflected in the relatively narrow confidence intervals for the risks of site-specific cancers. Reporting of cases to the Danish Psychiatric Central Register is nationwide, and the register covers all inpatient psychiatric facilities in Denmark. Use of the unique 10-digit personal identification number assigned to everyone in Denmark made it possible to eliminate loss to follow-up and incorrect linkage of information. The diagnostic criteria for affective disorders remained unchanged and were considered to be valid throughout the study period (29). Our cohort did not include all Danes with depression, since it excluded persons with unrecognized disease and those being treated in general medical practices or as outpatients. In addition, the number of beds in the psychiatric hospital system was reduced by more than 50 percent during the study period; accordingly, the proportion of patients (especially dysthymic patients) admitted to psychiatric hospitals decreased considerably. Of course, this could have led to attenuation of any effect of depression on cancer risk, since all persons with depression, regardless of whether they were

Depression and Cancer Risk 1093 admitted to a hospital, were included in the background rates. In a previous study, we observed no increased risk of cancer (except for non-hodgkin s lymphoma) in a large population-based cohort of users of antidepressive medication when we excluded the first year of follow-up (30). Only 13 percent of that cohort had ever been admitted to a hospital for psychiatric inpatient treatment. This supports the conclusion that treatment of depression outside of a hospital did not explain our general finding of no association between affective disorders and cancer in the present study. We included bipolar and unipolar psychoses in our definition of depression, although these disorders have not previously been associated with cancer risk, in order to cover the full spectrum of affective disorders. Our definition of depression probably covered only more severe cases of depression with symptoms that exceeded the threshold for hospitalization. This would have limited our ability to generalize our findings to all persons with depression, but it secured a certain exposure level. The Danish Cancer Registry is nationwide and population-based; practically all cases of cancer occurring in Denmark since 1943 have been reported to the registry (31). To avoid selective inclusion of patients with symptoms resembling depression that were actually due to unrecognized cancer, we separately analyzed data for the first year of follow-up after first admission for an affective disorder. The finding of an increased risk of brain cancer among patients with all levels of psychopathology strongly supports our suspicion that such selection occurred. The increased risk of cancer in general and of brain cancer in particular at first admission for depression underlines the importance of considering physical disease as a possible differential diagnosis for patients with depressive symptoms and points to the challenge of distinguishing between depression and the behavioral effects of organic pathology. Further selection would occur if persons with a previous diagnosis of cancer and an increased risk of developing a second cancer also had a greater risk of being admitted to a hospital with depression. Such selection was avoided in this study, since we excluded all patients with a previous cancer from the cohort and excluded prevalent cases from the background population when calculating the SIRs. To the best of our knowledge, only one previous cohort study has included patients with a psychiatric diagnosis of depression in assessing overall cancer risk. Among 923 depressed inpatients, overall cancer risk was significantly increased when compared with that of a cohort of 3,573 persons followed for up to 19 years, when the first 2 years of follow-up were excluded from the analysis (n 63; SIR 1.38, 95 percent CI: 1.06, 1.76) (). Most patients in that study had a diagnosis of neurotic depression (83 percent); however, analyses of site-specific cancers revealed no increased risk of tobacco-associated cancers, and most of the excess risk was found to be due to cancers of the breast, endometrium, or skin. Thus, even though the findings are in line with our observation of an overall increased risk in the group of patients with dysthymia, which is equivalent to neurotic depression, the distribution of sites for which there was an increased risk is different. The authors of other prospective studies assessed depression psychometrically. In a cohort of 2,0 middle-aged men followed for 20 years, depressed mood at baseline was associated with an increased incidence of all cancers during the first 10 years of follow-up (11). Six subsequent cohort studies of population-based samples comprising 1,529 6,913 subjects followed for 10 17 years did not confirm an association with cancer in general (12, 13, 16,, 20, 21). Only some of these studies provided information on site-specific cancers, including excess risks of lung cancer (12, ) and of smoking-related cancers combined (13). However, a study of 4,825 persons aged 71 years or older who had been assessed as having depressed mood three times over a 6-year period before the start of follow-up (average, 3.8 years) found an adjusted hazard ratio for all cancers of 1.88 (95 percent CI: 1.13, 3.) (15). In that study, the analyses of site-specific cancers revealed no predominant association with tobaccorelated cancers. The short follow-up period, the inclusion of data for the first year in the analyses, and the very limited age range are obvious limitations of the study. Therefore, the evidence for an increased risk of all cancers, unaffected by lack of statistical strength, confounding, or bias, must be considered weak. Our findings indicate an excess risk of sitespecific cancers, mostly those with a known association with smoking, rather than a generalized effect on cancer risk among depressed patients. We found an increased risk of smoking-associated cancers among patients hospitalized with reactive depression and dysthymic disorders that remained significant when data were stratified for latency. Two cohort studies in which the risk of smoking-associated cancers was estimated among persons psychometrically found to be depressed suggested that depression modifies the effect of smoking on cancer risk (12, 13); however, since smoking is a strong risk factor for cancer and adjustment was made using broad categories of smoking, the possibility of residual confounding due to inadequate control for smoking remains in both of those studies. Although a causal relation between depression and smokingassociated cancers cannot be excluded on the basis of our results, the magnitude of the association more likely reflects an increased prevalence of smoking among depressed persons (8, 9). Alcohol is an established risk factor for some of the smoking-associated cancers (cancers of the buccal cavity, larynx, and esophagus) (32). In our study, as many as 17 percent of patients had been admitted with an alcohol-related psychiatric disorder, and the elevated risk observed in all diagnostic categories for esophageal cancer supports the notion that an increased alcohol intake might have contributed to the observed increased risk of tobacco-associated cancers. This conclusion is further supported by the insignificantly increased risk of liver cancer among patients with reactive depression and dysthymia. Both tobacco smoking and alcohol intake can be regarded as confounding factors in the possible association between depression and cancer. Depression may cause an individual to smoke more and/or to drink more alcohol and thereby indirectly lead to an increased risk of cancer. Our data do not answer this question; however, our findings do call attention to a possible link between depression, smoking, and alcohol intake.

1094 Dalton et al. No significantly increased risks of tobacco-associated cancers were seen in the groups of patients with unipolar or bipolar manic depression, despite the fact that patients with bipolar disorders have higher rates of smoking than patients with major affective disorders, anxiety disorders, and personality disorders (8). This observation may be due to the difficulties of diagnosing physical illness in psychotic patients, who have paramount mental symptoms and altered behavior during acute phases of their illness. An alternative explanation that has been proposed for the low incidence of cancer in schizophrenic patients is that neuroleptic medications have an antineoplastic effect (33). Many hospital patients admitted with bipolar disorders are treated with neuroleptic agents, and these compounds may affect their risks of smoking-associated cancers. In studies of cancer incidence among immunesuppressed subjects, non-hodgkin s lymphoma, cervical cancer, and various types of skin cancers have been found at increased rates (4 6). In the present study, which was restricted to hospitalized patients, we found no deviation in the risk of non-hodgkin s lymphoma in any diagnostic groups. Antidepressant medications may also affect cancer risk by acting on the immune system, thus representing a possible confounder in the proposed relation between depression and cancer (34). In a population-based cohort study, an increased risk of non-hodgkin s lymphoma was observed in long-term users of tricyclic and tetracyclic antidepressants (SIR 2.5, 95 percent CI: 1.4, 4.2) (30), which supports the possibility of an etiologic role of antidepressants independent of depression. In the present study, we had no information on patients use of antidepressive medication; however, our findings do not indicate an excess risk of immune-related cancers among depressed persons. There seemed to be a generally decreased risk of cervical cancer in this study. This might be explained by a reduced risk of human papillomavirus infection (35) due to the negative effect of depression on sexual activity. To our knowledge, the general tendency toward a reduced risk of rectal cancer in all groups has not been described previously. This finding is not easily explained by current knowledge of the risk factors for this cancer, and the possibility of decreased surveillance in this cohort should be considered. An increased risk of colon cancer was seen in the reactive depressed patients and an increased risk of nonmelanoma skin cancer was seen in the dysthymic patients, whereas the risk of cutaneous melanoma was reduced in the latter group of patients. These results, observed in single diagnostic groups, could be chance findings in view of the large number of statistical analyses performed. In conclusion, our study provides no support for the hypothesis that depression increases risk of cancer. The increased risks observed in our study can probably be ascribed to behavioral factors, such as increased tobacco smoking and alcohol consumption by patients suffering from affective disorders. In addition, the increased risk of cancer, especially brain cancer, during the first year of follow-up might have been due mainly to reverse causality. The lack of an association between depression and cancer is important and encouraging, but our results call attention to the health effects of the unfortunate distribution of lifestyle factors among these patients. ACKNOWLEDGMENTS This study received financial support from the Psychosocial Research Foundation (grant 96 225 54) of the Danish Cancer Society. Dr. Preben B. Mortensen was supported by the Theodore and Vada Stanley Foundation. The authors thank statistician Lars Thomassen for computing assistance. REFERENCES 1. Fox BH. Depressive symptoms and risk of cancer. (Editorial). JAMA 1989;262:1231. 2. Croyle RT. Depression as a risk factor for cancer: renewing a debate on the psychobiology of disease. J Natl Cancer Inst 1998;90:56 7. 3. Ader R, Cohen N, Felten D. Psychoneuroimmunology: interaction between the nervous system and the immune system. Lancet 1995;345:99 103. 4. Kinlen LJ. Immunosuppressive therapy and acquired immunological disorders. Cancer Res 1992;52:5474 6. 5. Johnson CC, Wilcosky T, Kvale P, et al. Cancer incidence among an HIV-infected cohort. Am J Epidemiol 1997;6: 470 5. 6. Birkeland SA, Storm HH, Lamm LU, et al. Cancer risk after renal transplantation in the Nordic countries, 1964 1986. Int J Cancer 1995;60:3 9. 7. Herbert TB, Cohen S. Stress and immunity in humans: a metaanalytic review. Psychosom Med 1993;55:364 79. 8. Hughes JR, Hatsukami DK, Mitchell JE, et al. The prevalence of smoking among psychiatric outpatients. Am J Psychiatry 1986;3:993 7. 9. Glassman AH, Helzer JE, Covey LS, et al. Smoking, smoking cessation, and major depression. JAMA 1990;264:1546 9. 10. Schuckit MA. Alcohol and depression: a clinical perspective. Acta Psychiatr Scand Suppl 1994;377:28 32. 11. Persky VW, Kempthorne-Rawson J, Shekelle RB. Personality and risk of cancer: 20-year follow-up of the Western Electric Study. Psychosom Med 1987;49:435 49. 12. Linkins RW, Comstock GW. Depressed mood and development of cancer. Am J Epidemiol 1990;132:962 72. 13. Knekt P, Raitasalo R, Heliovaara M, et al. Elevated lung cancer risk among persons with depressed mood. Am J Epidemiol 1996;4:1096 103.. Friedman GD. Psychiatrically-diagnosed depression and subsequent cancer. Cancer Epidemiol Biomarkers Prev 1994;3: 11 13. 15. Penninx BW, Guralnik JM, Pahor M, et al. Chronically depressed mood and cancer risk in older persons. J Natl Cancer Inst 1998;90:88 93. 16. Gallo JJ, Armenian HK, Ford DE, et al. Major depression and cancer: the 13-year follow-up of the Baltimore Epidemiologic Catchment Area sample (United States). Cancer Causes Control 2000;11:751 8. 17. Jacobs JR, Bovasso GB. Early and chronic stress and their relation to breast cancer. Psychol Med 2000;30:669 78.. Kaplan GA, Reynolds P. Depression and cancer mortality and morbidity: prospective evidence from the Alameda County Study. J Behav Med 1988;11:1 13. 19. Hahn RC, Petitti DB. Minnesota Multiphasic Personality

Depression and Cancer Risk 1095 Inventory-rated depression and the incidence of breast cancer. Cancer 1988;61:845 8. 20. Zonderman AB, Costa PT, McCrae RR. Depression as a risk for cancer morbidity and mortality in a nationally representative sample. JAMA 1989;262:1191 5. 21. Vogt T, Pope C, Mullooly J, et al. Mental health status as a predictor of morbidity and mortality: a 15-year follow-up of members of a health maintenance organization. Am J Public Health 1994;84:227 31. 22. Hjerl K, Andersen EW, Keiding N, et al. Breast cancer risk among women with psychiatric admission with affective or neurotic disorders: a nationwide cohort study in Denmark. Br J Cancer 1999;81:907 11. 23. Munk JP, Mortensen PB. The Danish Psychiatric Central Register. Dan Med Bull 1997;44:82 4. 24. World Health Organization. International classification of diseases. Manual of the international statistical classification of diseases, injuries, and causes of death. Eighth Revision. Geneva, Switzerland: World Health Organization, 1967. 25. World Health Organization. ICD-10: international statistical classification of diseases and related health problems. Geneva, Switzerland: World Health Organization, 1992. 26. Storm HH, Pihl J, Michelsen E. Cancer incidence in Denmark, 1993. Copenhagen, Denmark: Danish Cancer Society, 1996. 27. Rothman KJ, Boice JD Jr. Epidemiological analysis with a programmable calculator. (NIH publication no. 79 1649). Washington, DC: US GPO, 1979. 28. Dreyer L, Winther JF, Pukkala E, et al. Tobacco smoking. APMIS 1997;105:9 47. 29. Kessing L. Validity of diagnoses and other clinical register data in patients with affective disorder. Eur Psychiatr 1998;13: 392 8. 30. Dalton SO, Johansen C, Mellemkjær L, et al. Antidepressant medications and risk for cancer. Epidemiology 2000;11:171 6. 31. Storm HH, Michelsen E, Clemmensen IH, et al. The Danish Cancer Registry: history, content, quality and use. Dan Med Bull 1997;44:535 9. 32. Dreyer L, Winther JF, Andersen A, et al. Avoidable cancers in the Nordic countries: alcohol consumption. APMIS 1997;105: 48 67. 33. Mortensen PB. The occurrence of cancer in first admitted schizophrenic patients. Schizophr Res 1994;12:5 94. 34. Steingart AB, Cotterchio M. Do antidepressants cause, promote, or inhibit cancers? J Clin Epidemiol 1995;48:07 12. 35. Kjær SK. Risk factors for cervical neoplasia in Denmark. APMIS 1998;106(suppl 80):5 42.