The effect of study design on the reporting of mortality due to colorectal cancer in adults with mental illness in Nova Scotia.

OBJECTIVE Some studies suggest a higher case fatality from colorectal cancer (CRC) in psychiatric patients even though the incidence is no greater than in the general population. However, this finding is not universal and may be confounded by study design (for example, inception, compared with historical, cohort), mean cohort age, or delays in presentation with more advanced staging at diagnosis. We assessed how study design and cancer stage at presentation affected outcomes in psychiatric patients. METHODS A retrospective historical cohort of CRC presentations was used comparing psychiatric patients with control subjects. Psychiatric status was defined by either health service use or psychotropic prescription, and outcomes by logistic or Cox proportional hazards regression. RESULTS There were 3501 new presentations of CRC from 2001 to 2005. Psychiatric cases, as defined by in- or outpatient contacts, made up 15.9% (n = 558) of patients with CRC. There was no association between psychiatric status and stage at presentation, surgical intervention, or mortality after adjusting for confounders. There was a reduced likelihood of resection, an indicator of curative treatment, when on a selective serotonin reuptake inhibitor or serotonin noradrenergic reuptake inhibitor (n = 194; AOR 0.54; 95% CI 0.30 to 0.97) or antipsychotics (n = 28; AOR 0.27; 95% CI 0.08 to 0.92). CONCLUSIONS The association between psychiatric status and CRC may vary by study design and how psychiatric status is defined. These factors should be considered when interpreting any association between mental illness and cancer. As in studies of all-cause and cardiac mortality, historical cohorts of prevalent psychiatric cases may be less sensitive than inception cohorts of new patients, possibly because the excess of deaths is greatest within 7 years of initial presentation for a psychiatric disorder.