Effect of age, gender, race, and type of opioid on opioid side effects: Evaluation of tools to analyze observational studies

Opioids are the pillar of pain management, but little is known about factors that increase the risk of developing side effects. In observational studies it is likely that the groups compared are different, and it is crucial to control for these differences to obtain valid results. There are several tools that are used for this purpose. However, it is not always clear what is the preferable strategy. The specific goal of this dissertation was to improve knowledge about opioid side effects and test new methods to accomplish that. The aims were: First, to compare matching with a fixed number of controls to matching with a variable number of controls. Second, to compare logistic regression to propensity score. Third, in light of steps one and two, to use the appropriate technique to evaluate the effect of type of opioid, age, gender, and race on opioid side effects. For the first two aims, we performed Monte Carlo simulations and carried out comparisons of the techniques in terms of bias, precision, robustness, and type I and II errors. For the third aim, we performed a secondary analysis of a post-marketing surveillance study. For the analysis of this aim, we used logistic regression because we demonstrated that it was a better approach for this purpose than matching or the propensity scores. We found that matching with a variable number of controls removed more bias than matching with a fixed number of controls with little loss in precision. We also found that propensity scores were a good alternative to control for imbalances when there were seven or fewer events per confounder. However, in these circumstances the propensity scores had low power. Logistic regression was a better choice when there were at least eight events per confounder. Finally, we found that meperidine produced less nausea/vomiting and less respiratory depression than morphine. The risk of respiratory depression increased with age. Women exhibited nausea/vomiting more than men. Caucasians exhibited nausea/vomiting more than African Americans. The results of this dissertation should be of interest to researchers involved in observational studies and to providers who care for patients receiving opioids.