Nevin Manimala Statistics

Treatment Effect Estimates from Pilot Trials Are Unreliable

J Pain Symptom Manage. 2023 Sep 2:S0885-3924(23)00661-9. doi: 10.1016/j.jpainsymman.2023.08.020. Online ahead of print.


The CONSORT guideline defines a pilot trial as a small-scale version of a desired future efficacy trial that is intended to answer the key questions of whether and how a larger study should be done. For example, a pilot trial might evaluate willingness of participants to undergo randomization or test different instruments for measuring an outcome. In addition to assessing feasibility of a future study, many investigators use pilot trials to obtain preliminary estimates of efficacy that might be used to gauge the promise of a new intervention, or to use in grant proposals to justify research funding. However, pilot trials do not provide reliable information about intervention efficacy due to the statistical phenomenon called sampling variability. In this brief tutorial we use computer simulation to demonstrate the influence of sampling variability on estimates of efficacy from pilot trials. The simulations show how pilot trial results can lead to erroneous decision making using a survival endpoint in a parallel arm clinical trial as an example. No background in programming or mathematics is required to understand the example. However, R code is provided for interested readers.

PMID:37666368 | DOI:10.1016/j.jpainsymman.2023.08.020

By Nevin Manimala

Portfolio Website for Nevin Manimala