Title : A study on the deviation of effect sizes assumed while designing Phase III Randomised Controlled Trials
Abstract:
Introduction: Effect size is, an important consideration in the sample size estimation while designing randomized controlled trials using a frequentist approach. This study aims to systematically review the RCTs published during 2018-2020 in leading oncology and medical journals to quantify the differences in the target and observed effect size. Methods: The journals considered for review were Journal of Clinical Oncology, Lancet Oncology, British Journal of Cancer, Annals of Oncology, and Journal of the National Cancer Institute. Trials were identified from the journal sites using keywords Phase III, Randomized Controlled Trials, Survival studies. A total of 85 RCTs published during the year 2018-2020 were assessed for eligibility. Trials were eligible for the review if they are Phase III superiority trials with time to event primary endpoint. Exclusion criteria were based on trials not providing enough detail about sample size calculation and published in a non-English language. The difference in the target and observed effect size was assessed using a standardized effect measure. Results: A total of 85 RCTs were included in the study. The most commonly reported source of effect size was literature-based and previous studies. 37.2% of trials in which the basis is not reported. The distribution of trials by cancer site was 80% of cancer with solid tumors. Also 78.80 % of patients with chemotherapy as a treatment modality and 10.60% with radiotherapy. It was observed that 53% of the study were non-significant. In which 61.2% of the significant and insignificant studies were don’t come under the target sample size. The Assumed HR for 20 significant studies were found which follows the 95% CI and 20 significant studies were do not follows the given 95% CI. In the case of non-significant study, the majority of the study (32 studied) were do not follow the given 95% CI only 13 non-significant studies followed the 95% CI Conclusion: The literature review concluded that from the previous studies target difference elicitation is advised, with multiple methods including, review of evidence and opinion-seeking as the better method for effect size quantification.
