Methods in pharmacoepidemiology : Four Studies, four Settings

Abstract: In the present thesis, four studies using four different settings and designs are presented in which associations between pharmacological treatments and possible adverse drug reactions are investigated; some methodological considerations identified in them are discussed. In a population-based case-control study, using information from telephone interviews, the association between use of acid-suppressing drugs and the development of acute pancreatitis was investigated. Statistically significant odds ratios were found, associating use with a duration of less than six months of proton-pump inhibitors (OR 3.2; 95% CI 1.4-7.4) or H2-antagonists (OR 4.9; 95% CI 1.6-15) with the outcome. However, confounding by indication was present to some extent, but the association was interpreted as probably causal. Using the UK General Practice Research Database (GPRD), a case-control study nested in a population of women with a diagnosis of menorrhagia, the relationship between use of tra-nexamic acid, mefenamic acid or norethisterone and the development of venous throm-boembolism (VTE) was evaluated. Use of tranexamic acid was too scarce to allow for certain conclusions; 3 cases and 4 controls exposed to the drug yielded an OR of 3.2 (95% CI 0.6-15.8). Use of mefenamic acid was associated with the outcome with an OR of 5.5 (95% CI 2.1-5.8), as was use of norethisterone, OR 2.4 (95 % CI 1.0-5.8). A diagnosis of anemia, indicating a more severe menorrhagia, was also associated with the outcome, OR 2.2 (95% CI 1.0-4.9); a fact which may indicate that confounding by indication was at hand, and that menorrhagia itself may constitute a risk-factor for VTE. The safety of antiretroviral treatment (ART) for HIV was analysed in a Swedish cohort-study of 1,072 patients. An association between ART and the development of midline- and inguinal hernias was identified. The hazard ratios (HR) were different depending on duration of ART containing a protease inhibitor (PI). The highest HRs were observed during the second to third year of treatment: HR 10.7 (95%CI 1.3-85.7) for midline hernia, and HR 4.4 (95% CI 1.1-16.6) for inguinal hernia. By linking a cohort of 5,756 patients treated with isotretinoin for severe acne with the Swe-dish national patient register of hospitalizations, the association between treatment with isotretinoin and hospitalizations for suicide attempt was analyzed. The incidence and Stan-dardized Incidence Ratios (SIR) of suicide attempts before, during and after treatment was estimated by use of a cohort crossover approach. The SIR for suicide attempts started to rise above one already two years before treatment. It was highest within six months after treatment, SIR 1.9 (95% CI 1.1-3.2) for first attempts; within three years after treatment, the observed number of suicide attempts was very close to the expected. It was concluded that a strong association between severe acne and suicide attempt was present, and that a significantly increased risk associated with use of isotretinoin could not be detected on the population level. However, the possibility could not be excluded that certain patients, pre-sumably sensitive to isotretinoin, made suicide attempts as a result of the treatment. A number of methodological aspects identified in the included studies are discussed in the thesis: confounding by indication was identified, challenged and estimated; the importance of dose-response and duration-response relationships in assessing causality was demonstrat-ed; reversed causality was observed and handled; assessing the presence of recall bias in an interview setting was described; and the effect of a presumed under-estimation of an important confounder was demonstrated using a simulation. Advantages and disadvantages of using the different data-sources, depending on type of treatment and character of adverse events, are discussed.