All Times EDT
Keywords: Bayesian meta-analysis, BEST-ITP model, longitudinal studies
The BEST-ITP model, developed by Ding et al, has the advantage of being able to estimate the treatment effect at a specific time-point of our interest, allowing all summary level data measured at different time-points for different studies. As a results, the model is being used in many meta analyses for mixed treatment effects across different longitudinal studies where observation values decrease over time by drug efficacy such as HbA1c or body weight. But when applying BEST-ITP model to the real data, various considerations need to be taken due to the diversity of the study. Our study shows that the precision and accuracy of estimated parameter by BEST-ITP model under multiple scenarios. We generated summary level data at each assumed study to have different sample size and number of time-points. In another scenario, short term studies were assumed in order to see the effect of study duration on the accuracy of extrapolation using the BEST-ITP model. Precision and accuracy were evaluated by Mean Square Error (MSE) and Bias, Standard Deviation through 1000 simulation runs. We found that the precision of the estimation is affected by the size of the studies which can be defined by both (1) having a large sample size at each time-point and (2) having a large number of time points to be used as a form of summary level data. We also found that unbalanced size of studies doesn't affect the estimation of treatment effect and applied BEST-ITP methods to the actual clinical trial results.