Epidemiologic Approaches to Evaluating Clinical Outcomes of Drug-Drug Interactions in Electronic Healthcare Data
MetadataShow full item record
AbstractDrug-drug interactions (DDIs) are an increasingly important clinical and public health concern as individuals with multiple chronic conditions are living longer and drug regimens are becoming more complex. Pre-marketing screening for potential DDIs is a required step in the development of new medications; however, the impact of putative interactions on patient health outcomes is usually not quantified, leading to uncertainty in clinical practice. Lack of clinically relevant DDI data has been implicated as one of the major reasons behind the failure of healthcare systems to prevent DDI-related patient harm.
Electronic healthcare databases offer a valuable opportunity to evaluate the clinical consequences of potential DDIs and to identify interactions that are not detected during pre-marketing stages. This thesis examines approaches to pharmacoepidemiologic studies of drug-drug interactions in electronic healthcare data, along with methodological challenges and potential sources of bias that can arise in this setting.
In Chapter 1, we evaluated whether the clinical impact of interaction between clopidogrel and cytochrome P450 (CYP) 2C19-inhibiting selective serotonin reuptake inhibitors (SSRIs) differed based on how patients encountered the interaction. We found that initiating CYP2C19-inhibiting SSRIs later in clopidogrel therapy was associated with a decrease in the effectiveness of clopidogrel that was of similar magnitude to the association observed among patients who initiated clopidogrel while being treated with SSRIs, and combined the evidence using meta-analysis.
In Chapter 2, we evaluated several approaches to designing and analyzing case-crossover studies of drug-drug interactions based on two empirical DDI examples with prior evidence of harm. We found that in a case-crossover study of two drugs, a saturated model is a six-parameter model that differentiates all three ways patients can encounter an interaction. As compared to the traditional model with a product term, the saturated model can help identify heterogeneity across strata.
Finally, in Chapter 3, we developed a semi-automated, case-crossover-based screening approach for identifying clinically relevant interacting drug pairs in electronic healthcare data. The approach had high specificity and represents a promising option for generating the much-needed evidence on the relevance of drug-drug interactions in clinical practice, which was the primary motivation behind this dissertation.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:37945567