Biomedical subjects
Publications and source records attributed to Mark E Sale.
We describe a general and robust method for identification of an optimal non-linear mixed effects model. This includes structural, inter-individual random effects, covariate effects and residual error models using machine learning. This method is based on combinatorial optimization using genetic algorithm.
A dosing algorithm was used to guide the conversion of 77 patients with epilepsy (16 years of age) from valproate monotherapy through lamotrigine adjunctive therapy at 200mg daily to lamotrigine monotherapy at 500 mg daily in an open-label study comprising a lamotrigine escalation phase (8 weeks), a valproate withdrawal phase (6 weeks), and a lamotrigine monotherapy phase (4 weeks). The algorithm was designed to maintain stable trough concentrations of lamotrigine during valproate withdrawal to minimize seizure risk. Of the 77 patients, 11 prematurely withdrew for administrative reasons (noncompliance, consent withdrawn, loss to follow-up, protocol violation). Of the remaining 66 patients, 18 withdrew because of adverse events, and 48 completed the study. Among the 67 patients with at least one lamotrigine trough serum concentration after initiation of therapy (pharmacokinetic population), mean trough serum concentrations at the end of the lamotrigine escalation phase at a lamotrigine daily dose of 200 mg (7.9 microg/mL, SD = 3.3) did not differ significantly from values during the valproate withdrawal phase (8.7 microg/mL, SD = 3.5) and the lamotrigine monotherapy phase at a lamotrigine dose of 500 mg daily (7.2 microg/mL, SD = 3.3). A similar pattern of results was observed among the 34 patients who completed the study on lamotrigine (completer population). No serious rashes were reported, and the incidence of withdrawals because of decreased seizure control was low (n = 2, or 3%; both incidents were judged to be related to noncompliance). By employment of a four-step dosing algorithm, lamotrigine serum concentrations can be maintained at a stable level with favorable tolerability during a transition from lamotrigine 200mg daily as adjunctive therapy with valproate to lamotrigine monotherapy 500 mg daily.
Drug development at different stages may require assessment of similarity of pharmacokinetics (PK). The common approach for such assessment when the difference is drug formulation is bioequivalence (BE), which employs a hypothesis test based on the evaluation of a 90% confidence interval for the ratio of average pharmacokinetic (PK) parameters. The role of formulation effect in BE assessment is replaced by subject population in PK similarity assessment. The traditional approach for BE requires that the PK parameters, primarily AUC and Cmax, be obtained from every individual. Unfortunately in many clinical circumstances, some or even all of the individuals may be sparsely sampled, making the individual evaluation difficult. In such cases, using models, particularly population models, becomes appealing. However, conducting an appropriate statistical test based on population modeling in a form consistent, at least in principle, with traditional 90% confidence interval approach is not so straightforward as it may appear. This manuscript proposes one such approach that can be applied to sparse sampling situations. The approach aims to maintain, as much as possible, the appropriateness of the hypothesis test. It is applied to data from clinical studies to address a need in drug development for assessment of PK similarity in different populations.
Subject withdrawal from a study (also called dropout, or right censoring), is common in late phase clinical trials. A number of methods dealing with dropouts have been used in practice, the most common being "last observation carried forward" (LOCF). Many of these methods, including LOCF, can result in biased estimates of the efficacy or potency of the drug, especially in the modeling context. If the likelihood of dropout is correlated to the underlying unobserved data, the dropout is informative and should not be ignored in the modeling process. The topic of informative dropout in the context of longitudinal data has received much attention in the statistical literature, in the setting of linear and generalized linear models. We extend the approach to nonlinear models. The dropout hazard, as well as the longitudinal data, is modeled parametrically. Parameters are estimated by maximizing the approximate joint likelihood as implemented in the software NONMEM. Using data from actual clinical trials, we explore the impact of the dropout model on the ability of the joint model to predict observed longitudinal data patterns.