|Home | About | Journals | Submit | Contact Us | Français|
A previous epidemiologic study reported a 30% reduced risk of colorectal cancer among users of high doses of selective serotonin reuptake inhibitors (SSRIs). We assessed the association of colorectal cancer risk with SSRI and tricyclic antidepressant use in our hospital-based Case Control Surveillance Study.
For the SSRI analyses, we used data collected on 529 colorectal cancer cases and 1955 hospitalized controls collected from 1995 through 2008. For the tricyclic antidepressant analyses, we used data on 2889 cases and 7122 controls collected from 1976 through 2008. We used multivariable logistic regression analysis to evaluate the association of regular SSRI use and regular tricyclic antidepressant use (daily use for at least 3 continuous months) with colorectal cancer risk.
The odds ratio for regular SSRI use was 0.55 (95% CI 0.35–0.88) and it did not differ by duration of use. The odds ratio was 0.47 (95% CI 0.26–0.85) for colon cancer and 0.72 (95% CI 0.37–1.41) for rectal cancer. The odds ratio for regular use of tricyclic antidepressants was 0.77 (95% CI 0.52–1.16)
We found an association of reduced risk of colorectal cancer with regular use of SSRIs. In light of laboratory data indicating that SSRIs may inhibit colon cancer and one previous epidemiologic study that also observed a decreased risk, further investigation of the effect of SSRIs on the risk of colorectal cancer is warranted.
A case-control study nested in the Saskatchewan Health database reported a 30% reduced risk of colorectal cancer among people who had used high doses of selective serotonin reuptake inhibitors (SSRIs), and no association between tricyclic antidepressant use and colorectal cancer risk 1. There are no other epidemiologic data on the relation of antidepressant use to the risk of colorectal cancer. We assessed the association of colorectal cancer risk with use of SSRIs and tricyclic antidepressants in our hospital-based Case Control Surveillance Study, a study designed to assess multiple hypotheses regarding the effect of medication use on cancer risk.2
Nurses employed by the Case Control Surveillance study interviewed patients in collaborating hospitals starting in 1976. Patients were interviewed in Boston from 1976 to 1987, New York from 1978 to 1993, Baltimore from 1977 to 1986 and 1993 to 1996, and Philadelphia from 1977 to 2008. The population base for the study comprised people living within 50 miles of a participating hospital. Eligible patients were aged 18 to 79, under the care of a physician participating in the study, able to complete the interview (e.g., not deaf), and did not have certain excluded diagnoses (e.g., psychiatric).
Nurse-interviewers administered standard questionnaires to obtain information on demographic factors, medical and reproductive history, and lifestyle factors. Histories of medication use were elicited by asking about 43 indications and drug classes including depression and antidepressants. For each episode of use, the drug name and the duration, timing, and frequency of use were recorded. Details of the diagnosis were abstracted from discharge summaries and pathology reports. From 1976 through 1997, 95% of patients approached for the study participated, and from 1998 through 2008, 84% participated. The study was approved by the institutional review boards of all participating institutions and participants provided written informed consent.
For the analysis of SSRIs, which were first marketed in 1987, we used data on patients admitted to collaborating hospitals in Philadelphia from 1995 through 2008. (We did not include patients interviewed before 1995 due to very low prevalence of use in those years.) Cases comprised 529 men and women aged 20 to 79 diagnosed with colorectal cancer in the previous year, with no prior or concurrent cancers other than nonmelanoma cancer of the skin. Controls were frequency matched to the cases on 5-year age group, sex, and 2 categories of interview year, in a ratio of up to 4:1. Controls included 1955 men and women admitted to the hospital with diagnoses that we judged to be unrelated to SSRI use: hernia (15%), liver and kidney disorders (32%), orthopedic conditions (21%), benign neoplasms (12%), and other conditions (20%) (e.g., thyroid disorders, skin conditions). Controls were aged 19 to 79 and had no history of cancer other than nonmelanoma cancer of the skin.
We classified regular SSRI use as daily use for at least 3 continuous months. All other use was classified as sporadic use. SSRI use that began within the year prior to hospital admission is unlikely to have etiologic relevance and subjects whose only use began during the year before admission were kept in a separate category of recent initiators.
For the analysis of tricyclic antidepressants and colorectal cancer risk, we used data from patients admitted to participating hospitals in all study centers for the years 1976–2008. Cases met the same criteria as for the SSRI analysis and comprised 2,889 men and women aged 18 to 79. Controls (n=7,122) were aged 18 to 79, chosen from the same diagnoses as the controls for the SSRI analysis, and were matched on a ratio of up to 4:1 on 5-year age group, sex, and 6 categories of interview year. The controls were aged 18 to 79 and the distribution of control admission diagnoses was: hernia, 10%; liver and kidney disorders, 30%; orthopedic, 20%; benign neoplasms, 16%; and other conditions, 24%.
We used logistic regression models to estimate odds ratios (OR) and 95% confidence intervals (CI) for various categories of SSRI use and tricyclic antidepressant use relative to no use. We present odds ratios from a multivariable model adjusted for 5-year age categories, sex, interview year, use of nonsteroidal anti-inflammatory drugs (no use, recent initiator, sporadic use, regular use, unknown timing or duration of use), race (white, not white), and number of visits to a doctor two years before the hospital admission (0, 1–2, 3–5, 6+). The tricyclic antidepressant use analysis was also adjusted for study center.
The duration of regular SSRI use ranged from 3 months to 20 years, with a median of 3 years. The most commonly used SSRI was fluoxetine (33% of regular use), followed by sertraline (30%), paroxetine (22%), and escitalopram oxalate (10%). Most use (77%) continued into the year of hospital admission.
The odds ratio for regular SSRI use (excluding recently initiated use) was significantly reduced (OR=0.55, 95% CI 0.35–0.88), and odds ratios were similar for use of <3 years and 3 or more years (Table 1). The OR was =0.47 (95% CI 0.26–0.85) for colon cancer and 0.72 (95% CI 0.37–1.41) for rectal cancer. Odds ratios for regular SSRI use were similar when we compared the cases to the different subgroups of controls (data not shown).
The duration of tricyclic antidepressant use ranged from 3 months to 23 years, with a median of 2 years. The most commonly used drug was amitriptyline (47%), followed by imipramine (13%). The odds ratio for regular use of tricyclic antidepressants was 0.77 (95% CI 0.52–1.16), and the odds ratios did not differ by duration (Table 1). Although the odds ratio was lower for rectal than colon cancer, neither differed significantly from 1.0. Estimates from analyses using subsets of controls were similar (data not shown).
When we confined the analysis of tricyclic antidepressant use to Philadelphia, as was done for the SSRI analysis, the odds ratio for regular tricyclic use was 0.69 (95% CI 0.41–1.17) (based on 18 case and 83 controls users). There were too few users after 1995 to confine the analysis to the same years as the SSRI analysis.
We found a statistically significant 45% reduction in the risk of colorectal cancer for regular use of SSRIs. Reductions were similar among subjects who had used SSRIs for less than 3 and 3 or more years. We found a nonsignificant risk reduction for regular use of tricyclic antidepressants.
The one previous report on the topic of antidepressant use and colorectal cancer found a 30% reduced risk among subjects who had used a high doses of SSRIs (>6 × 10−6 mol per day) within 5 years of the index date; lower doses or use in the 6–10 years before the index date had no effect.1 The authors found no effect of tricyclic antidepressant use on risk.
Our study was limited by small numbers of users of SSRIs, which limited our duration analysis. We lacked information on dose which precluded a direct comparison of our results with those of the previous study. A useful comparison would have been to calculate the odds of colorectal cancer risk among users of SSRIs compared to users of tricyclics, that is, consider tricylic users as the nonexposed group. This was not possible since there was not enough overlap in the time period when the two drug classes were used: tricylics were most often used before 1995 and SSRIs after 1995.
Recall bias is a concern in any study where exposure assessment is based on self-report. However, in our study medication use was elicited by a long list of indications and the hypothesis under consideration was unknown to both interviewers and participants, which would have reduced biased reporting.
We controlled for several possible confounding factors but had no data on physical activity or diet.; the previous study of colorectal cancer and SSRIs controlled only for the use of other drugs (estrogen, oral contreceptives, and NSAIDs).1 Nevertheless control for confounders in observational studies can never be considered complete due to imperfect information and lack of information on unknown confounders.
There was a nonsignificant reduction in risk of colorectal cancer for regular use of tricyclic antidepressants, which might suggest that confounding by indication explained at least part of the risk reduction observed for SSRI use. This type of confounding could explain our results only if depression was less common among persons who went on to become colorectal cancer cases than among those who did not. However, there are no data to indicate that this is the case and two studies of depression and cancer found nonsignificant increases in the risk of colorectal cancer among participants with depressive symptoms.3,4 In addition, the reduced risk for tricyclic use suggests that some degree of selection bias may have occurred, such that people prescribed antidepressants of any sort were more likely to have physician contact and thus be diagnosed with the diagnoses that we chose as controls. However, most of the diagnoses that we chose as controls were serious conditions (e.g., liver and kidney disorders, benign neoplasms) that were likely to have come to diagnosis regardless of antidepressant use. The fact that estimates of the odds ratios were similar using different subgroups of the controls suggests that the selection of control diagnoses was appropriate.
Finally, we included only subjects living within 50 miles of the study hospitals, which was an effort to minimize bias resulting from selective admission of cases or controls. The similarity of our results on other drug-cancer associations to those of studies that used other methods suggests that the effort was successful.5,6
Serotonin plays a role in cell inflammation, proliferation, and repair at sites remote from the central nervous system7, 8, effects that may involve a cellular-uptake mechanism.9 Serotonin has also been found to promote angiogenesis in a mouse model of colon cancer.8 Various SSRIs have suppressed colorectal cancer cell proliferation in vitro10, 11 and in rat9 and mouse12 models of colorectal cancer. As regards the tricyclic antidepressants, while some laboratory data indicate they may enhance colon carcinogenesis13, more recent data suggest that they have cytotoxic effects against colon cancer cells.11, 14
We found a significant inverse association of risk of colorectal cancer with regular use of SSRIs. Like all observational studies, ours had several limitations and our results should be regarded as contributing positive albeit weak evidence to the hypothesis raised by the previous study. In light of laboratory data indicating that SSRIs may inhibit colon carcinogenesis, further investigation of the effect of SSRIs on the risk of colorectal cancer is warranted.
This study was funded by grant R01 CA45762 from the National Cancer Institute.
Patricia F. Coogan, Slone Epidemiology Center at Boston University, Boston, MA, USA.
Brian L. Strom, Center for Clinical Epidemiology and Biostatistics, Department of Biostatistics and Epidemiology, Center for Education and Research on Therapeutics, and Division of General Internal Medicine, Department of Medicine, University of Pennsylvania School of Medicine, Philadelphia, PA, USA.
Lynn Rosenberg, Slone Epidemiology Center at Boston University, Boston, MA, USA.