Copyright 1995 by Philip S. Wang. All rights reserved.
This report is distributed through the Drugs and Devices Information Line at the Harvard School of Public Health. If you have comments, please mail them to the Line at ddil@episun1.harvard.edu. We will compile readers' views periodically and append them to the article.
Galen observed that "melancholic" women suffered from cancer more frequently than did "sanguine" women (1). Since then, physicians and investigators have wondered whether there is a link between psychological factors, especially depression, and the development or progression of cancer. An entire body of scientific literature has developed around this question, with varied results and often conflictual conclusions.
Blumberg, West and Ellis (2) were one of the first groups to systematically examine these questions. They administered the Minnesota Multiphasic Personality Inventory (MMPI) to male cancer patients with a variety of malignancies. They then selected 25 patients, whose survival times from the date of first symptom were at or less than the 25th percentile expected for their cancer type, and 25 patients whose period of survival was at or beyond the 75th percentile. Patients in both groups were tested while in remission and had comparable ages, nationality, religion, socioeconomic status and I.Q. levels. The authors found that, at the time of testing, patients with rapidly progressing disease had had significantly higher levels of depression, anxiety, and an inability to relieve or reduce these through external expression.
Krasnoff (3) attempted to cross validate the research by Blumberg et al. but failed. He administered the MMPI, Rorschach test and Wechsler Intelligence Scale to 70 male and female patients with malignant melanoma. He then selected the scores of six patients who survived less than 20 months from the date of first symptoms and compared these to the scores of 16 patients who survived greater than six years. The only significant differences detected were that patients with rapidly progressing tumors had lower verbal intelligence quotient scores and lower socioeconomic status. Krasnoff points out that one important difference between his study and that done by Blumberg et al. was that all subjects in the latter knew they had diagnoses of inoperable cancers, while subjects in the Krasnoff study had widely varying knowledges of their disease status.
Klopfer (4) assembled the Rorschach projective test results of 24 patients, who subsequent to testing, were found to have slow and fast growing cancers. He interpreted their Rorschach projective tests blind to the patient's disease status. He reported that patients with slow growing tumors were characterized by significantly less subjective distress and a calm acceptance. He concluded that patients with high ego defensiveness do not have the vital capacity to fight off a cancer.
Shrifte (5) interpreted Rorschach tests from 22 indigent women with cervical cancer. Fifteen patients were in remission two years after testing and seven had died in that interval. The author initially found no distinguishing features between the two groups' test results. In a later re-interpretation, she found that the results of those in remission were characterized by a significantly greater ability to be influenced by and receive external influences. She concluded that patients who receive less from their environments have less resistance to cancer.
Stavraky et. al. (6) administered the MMPI, verbal Wechsler Intelligence Scale, and a projective test (the Differential Diagnostic Technique) to 204 cancer patients. The compared the scores of the 23 subjects with the longest survivals and the 30 with the shortest survivals, to stage matched controls. Their most striking finding was the significantly higher levels of underlying hostility and aggression, without a loss of emotional control, in the patients with the longest survivals. The authors hypothesized that hopelessness or a "giving-up" reaction carried a worse prognosis from cancer.
A study by Davies et. al. (7) seemed to confirm this psychological profile of the long survivor. Forty-six patients with metastatic cancer or hematologic cancers were given a semi-structured interview by a psychiatrist. They found that apathy and "giving up" were strongly correlated with decreased survival. However, it is important to note that patients with the highest levels of apathy at interview, also had the most advanced disease. The authors speculated that disease severity influenced the subjects psychological reaction (and survival) and not vice versa.
Derogatis et. al. (8) administered a psychological symptom check list, the Symptom Check List-90, to 23 cancer patients and their attending physicians. They found that patients with the shortest survivals were characterized by the greatest levels of depressive symptomatology. Again, these patients also had the most severe disease, prompting the authors to conclude that patients' worsened depressive symptoms were related to their more advanced disease.
Several reviews have described the methodologic limitations of these studies as well as some possible errors in the conclusions drawn from such studies (9)(10)(11). Perhaps greatest of the problems has been the fact that most of these studies employed retrospective designs. At the time of examination for depression and other psychological factors, the patients already were diagnosed with cancer. It has been impossible to establish causality--i.e., to distinguish whether the presence of the psychological factors caused worse disease progression or vice versa. Other limitations include the lack of adequate controls, small sample sizes, using invalidated measurements of depression, failing to control for possible confounders, using a variety of cancer sites and having inadequate controls for cancer staging.
A few epidemiological studies, done more recently, have employed a prospective design and larger numbers to examine whether depression affects cancer incidence, mortality and outcome. These studies have been better able to distinguish whether the depression preceded the diagnosis of cancer or not. However, they too have produced conflicting results.
Persky et. al. (12) in the Western Electric Study used the MMPI in a cohort of 2018 men whom they followed for 20 years. They found depression to be associated with increased cancer incidence at 10 years of follow-up (RR~ 2) and with increased mortality at 20 years follow-up (RR =1.9). The association did not seem stronger for one type of cancer versus another.
Linkins and Comstock (13), in a mental health study conducted in Washington County, Maryland, obtained scores on the Center for Epidemiologic Studies Depression Scale for 2264 participants who then remained cancer free for 2-4 years. Over a 12 year follow-up period, they found only a slight association of depressed mood with subsequent cancer; however, they did find strong effect modification of the relationship between depression and cancer by cigarette smoking.
Other studies have failed to show an association between depression and cancer. Hahn and Pettiti (14) examined MMPI scores of 8932 women who participated in the Walnut Creek Contraceptive Drug Study and found no significant subsequent increased risk of breast cancer among those with high depression scores.
Kaplan and Reynolds (15), in part of the Alameda County Study, did a 17 year follow-up of 6848 persons who were initially free of cancer. They found no link between baseline levels of depressive symptomatology and cancer incidence or mortality although they did find a significant association between high levels of depressive symptoms and non-cancer deaths.
Zonderman et. al. (16) used data from the National Health and Nutrition Examination Survey I Epidemiologic Follow-Up Study. They obtained 10 year and 15 year follow-ups on 6410 and 3814 individuals, respectively, who had completed the Center for Epidemiologic Studies Depression Scale or the General Well-Being Schedule depression subscale. They found that high scores on neither measure was a significant risk for cancer morbidity or mortality.
Edwards et al. (17) administered a questionaire to 1052 women prior to an examination for breast cancer--most women were there for routine examinations and none had any knowledge of their diagnostic status. The questionaire contained sections on life events, coping (Ways of Coping Checklist), Type A behavior pattern (Bortner Type A Scale) and the availability of social support. No significant relationships were found between coping, Type A behavior or the availability of social supports and developing breast cancer.
Two randomized controlled trials have examined whether psychological interventions in cancer patients can effect prognosis after cancer has occurred. Both appeared to show that improvements in a cancer patients psychological state prevent relapses.
In a study by Spiegle et al. (18), 86 patients with metastic breast cancer were randomized to receive either routine oncologic care, alone, or routine oncologic care plus one year of weekly supportive group therapy and self hypnosis. Survival from the time of randomization was significantly higher in the intervention group (36.6 months compared to 18.9 months).
Fawzy et al. (19) randomized 68 patients with malignant melanoma to a six week structured psychiatric group intervention and routine care or routine care alone. Subjects in the intervention group had a trend towards decreased recurrence rates and a significantly increased survival rate compared to controls.
Those who feel an association does exist between depression and cancer have proposed a variety of mechanisms to explain the relationship. Earlier theories, influenced by psychoanalytic thought, stressed a psychological basis for a predisposition to cancer. Denial, repression, inability to deal appropriately with anger and strong internalized control were felt to be defenses that created a personality at risk for cancer (20)(21)(22). The experience of profound losses such as loss of a parent early in life or the loss of a close relationship soon before the development of cancer were felt to be contributory towards the development of cancer as well (23).
More recently, endocrinologic and immunologic rationales for a susceptibility to cancer in depressed patients have been proposed. These theories suggest that stresses, such as depression, lead to excesses of hormones including adrenal corticosteroids (24)(25). It is through these corticosteroids that immunosuppression is mediated (26)(27)(28). Evidence for this in humans includes findings of increased levels of corticosteroids in depressed patients (29)(30)) and lymphocyte suppression in bereaved patients compared to non-depressed subjects (31)(32). Patients with compromised immune systems have been shown to have higher incidence rates of cancer (33)(34)(35).
Some have argued that indirect or confounding factors might explain any association between depression and cancer. Several potential confounders were adjusted for in the prospective studies outlined above. Persky et.al. (12) adjusted for age, number of cigarettes smoked, alcohol intake, occupational status, family history of cancer, body mass index and serum cholesterol . Their association between depression and cancer incidence and mortality remained significant. Zonderman et. al. (16) adjusted for age, sex, marital status, smoking, family history of cancer, hypertension and serum cholesterol. However, they failed to show a significant risk for cancer morbidity or mortality from depressive symptoms with or without these adjustments. Covey et. al. (36) reviewed the work by Linkins and Comstock (13) that demonstrated an interaction between depression and smoking on cancer risks. Covey pointed out that one confounder not considered was that depressed smokers may quit smoking at a significantly lower rate than non-depressed smokers--a hypothesis supported by their research and others (37).
The use of antidepressant medication has recently been considered as a factor related to depression and possibly linked to cancer. Laboratory studies by Brandes et al. (38) raised concerns that selective serotonin reuptake inhibitors (SSRIs), such as Prozac (fluoxetine), and tricyclic antidepressants (TCAs), such as Elavil (amitriptyline), may act as tumor promoters in rodents. Researchers had noted that SSRIs and TCAs are structurally similar to Tamoxifen-like compounds that bind to an anti-estrogen/intracellular histamine receptor. Binding to this site has been shown to stimulate tumor growth in vivo.
The experiments showed that carcinogen-treated rats, given Prozac or Elavil at doses comparable to therapeutic doses in humans, developed mammary tumors over twice as frequently and with a 30-40% reduced tumor latency, compared to controls. In another model, mice given mammary fibrosarcoma or melanoma cells and Prozac or Elavil also showed a 30-40% accelerated tumor growth rate compared to controls.
A group in Japan has explored the role of another tricyclic antidepressant, desipramine, as a tumor promoter (39). They showed that rats, given injections of azoxymethane and desipramine develop more colonic tumors than those given azoxymethane and saline.
Very few epidemiological studies have considered the possibility that antidepressant drugs themselves may have an effect on cancer incidence or outcome. Stoll (40) examined the use of psychotropic medications (including major tranquilizers, minor tranquilizers, hypnotics and antidepressants) in the year preceding diagnosis of breast cancer. He found a statistically significant increase in the rate of metastasis at the time of presentation or 12 months after diagnosis. The variety of medications used, small study size and retrospective design (introducing the potential for recall bias) make interpretation of these findings difficult.
Linkins and Comstock (41) have recently looked at the use of psychoactive medication (including major tranquilizers, minor tranquilizers, stimulants, and antidepressants) in the 2264 persons comprising their mental health survey. They found the adjusted relative risk of developing cancer with psychotropic medication use was increased although non-significantly. Again, however, the multiple medications used (they estimate only approximately 1.1% of their population were on antidepressants) and the small numbers in their study make their results difficult to interpret.
Friedman (42), using data from their National Cancer Institute screening study for carcinogenic pharmaceuticals did a 19 year follow-up of 1957 patients who used amitriptyline and 308 patients who used imipramine. They found a significantly elevated risk for liver cancer (only through 15 years of follow-up) and a significantly lower risk of pancreatic cancer for amitriptyline users. They observed no departure from expected numbers of cancers for imipramine users.
Clearly additional large prospective epidemiological studies are necessary to determine if depression predisposes patients to higher incidences, worse courses and outcomes from cancer. Future studies should be carefully designed to allow them to distinguish whether any differences in incidence rates and outcomes observed are due to the severity of depression or the presence of associated factors. Particular attention should be given to determining whether any increased risk from cancer may be introduced by the very medications used to treat depression.
Back to the DDIL Homepage.
Last updated: 4/18/95