PIN136 MODELLING CD4 TRANSITIONS IN HIV PATIENTS: VALIDATING A MARKOV MODEL BASED APPROACH

Abstract

A key limitation of discrete-event simulation (DES) models in HIV is the requirement for repeated iterations to reach convergence. This is particularly problematic where clinical efficacy is informed by non-inferiority trials and small reported differences result in a prohibitively high number of iterations being required. Markov models offer an alternative to DES models that does not require multiple iterations to reach convergence. This study validates an alternative approach to estimating CD4 transitions within a Markov framework against data from a clinical study. A previously published cost-effectiveness model (CEM) was used to estimate the proportions of patients in each CD4 health state at the end of the study period. Transition probabilities between CD4 health states were estimated using data on baseline CD4 cell count and change in CD4 cell count over the study period. The proportions of patients in each CD4 health state estimated by the CEM were compared to those reported by a clinical study to validate the model data, with root mean squared error used to assess the level of agreement between the two. A high level of agreement was observed between proportions in each CD4 state as estimated by the CEM and as observed in a clinical study, successfully validating this method of modelling CD4 transitions within a Markov framework. The approach to modelling CD4 transitions presented here replicates CD4 profiles observed in clinical practice, without requiring large numbers of iterations to reach convergence. This makes it feasible to run the range of scenario and sensitivity analyses routinely required for health technology assessment in a reasonable timeframe, while maintaining robust and accurate estimates of health economic outcomes.