Abstracts

Rationale: For drugs with a narrow therapeutic index (NTI), small differences in concentrations may lead to serious toxicities or therapeutic failures. NTI classification of drugs without a defined therapeutic index has been challenging. We sought to develop a systematic approach to NTI drug classification that integrates clinical practice data from the public literature and electronic medical records with pharmacokinetic (PK)/pharmacodynamic (PD) modeling. We evaluated this approach for lamotrigine (LTG), a potential NTI drug lacking a well-defined therapeutic index.Methods: We conducted a systematic PubMed and Embase literature search (1985–2013) to identify studies reporting LTG PK, PD, and safety data. We extracted relevant data into a drug database and estimated the LTG therapeutic index. We also extracted pre-specified medical record data (LTG levels, dosing, safety events, seizures) from adult patients admitted to Duke Hospital (2011–2013) with at least one available LTG level. We selected a LTG population PK model from the literature search that fit our medical record data by evaluating model fit with predictive performance measures and normalized prediction distribution errors (NPDE). We then simulated LTG exposures (trough, maximum, and average concentrations) on the day of identified therapeutic failures/safety events and explored the exposure-response relationship using the simulated exposures and observed PD/safety data.Results: We extracted data from 77 relevant papers, and the estimated LTG therapeutic index was 1.3–20. We identified 45 Duke patients (median [range] age 43 [20–74] years; weight 78.5 [46–175] kg; 1 [1–2] LTG concentrations/patient); 26% had a seizure, and 20% had an adverse event. A literature population PK model predicted our data well with a ratio of observed/predicted concentrations of 0.92 (95% CI: 0.75–1.14) and a mean NPDE of -0.076 (p-value = 0.77). However, there was no evident relationship between simulated LTG exposure and safety events or therapeutic failures.Conclusions: The use of clinical practice data to aid in classification of drugs with NTI is a promising approach. However, before universal implementation of this methodology, limitations in sample size and accurate characterization of the concentration-response relationship need to be overcome.