Conference on meta-analysis in the design and monitoring of clinical trials by N. Geller, L. Freedman, Y. J. Lee and R. DerSimonian (eds), Statistics in Medicine, 15, 1233–1324 (1996)
✍ Scribed by Rebecca DerSimonian
- Publisher
- John Wiley and Sons
- Year
- 1999
- Tongue
- English
- Weight
- 44 KB
- Volume
- 18
- Category
- Article
- ISSN
- 0277-6715
No coin nor oath required. For personal study only.
✦ Synopsis
In June 1994, the Division of Epidemiology, Statistics, and Prevention Research (DESPR) of the National Institute of Child Health and Human Development organized a workshop on the role of meta-analysis in the design and monitoring of clinical trials. This workshop was motivated by concerns raised at the initial deliberations of the Data and Safety Monitoring Board (DSMB) for CPEP, a large trial conducted by DESPR to assess calcium efficacy in reducing the risk of pre-eclampsia in healthy nulliparous women. Several previous trials had been suggestive of a beneficial calcium effect for pre-eclampsia prevention and there was a disagreement among DSMB members as to whether the data from those trials should play a role in monitoring CPEP. In light of the strong opposing viewpoints expressed by some members of the DSMB on the issue, the Institute proposed this conference to review the usual practice and to identify the role of external data in the monitoring of CPEP. Conference participants included experts in the fields of meta-analysis, clinical trials and statistics. Without jeopardizing confidentiality of CPEP data, an introductory presentation described the existing trials and their relevance to CPEP design and monitoring. Other presentations focused on the pros and cons of using external data to monitor an ongoing trial in general terms. Some discussants cautioned against interfering with a well-designed trial regardless of what the external data indicate while others described the advantages of evaluating interim data in the context of external sources. A vague consensus seemed to be that while the informal use of external information is generally helpful in evaluating interim data, its formal incorporation into the monitoring rules is usually not warranted.
CPEP data are now complete and interim results are available. Over a three-year period, 4589, healthy nulliparous women were recruited early in their pregnancy into CPEP and were followed up until delivery. This sample size was large enough to yield more than 85 per cent power to detect a reduction in risk of about 40-50 per cent, the anticipated beneficial calcium effect based on previous trials. At study completion, 94)5 per cent of the total sample had complete follow-up to delivery, 65)5 per cent were compliant, the pre-eclampsia rate in the placebo group was 7)3 per cent, and the overall odds ratio estimate (95 per cent CI) was 0)94 (0)75-1)17). Thus, final CPEP data analysis indicated no beneficial calcium effect for pre-eclampsia prevention in healthy nulliparous pregnancies (p-value"0)6).
The group sequential boundaries originally adopted for CPEP were based on an O'Brien and Fleming (OBF) design with four interim analyses and 900 new pregnancies completed during each interval (Table I). An adds ratio estimate below the lower boundary would have prompted stopping the trial early in favour of a beneficial calcium effect. Similar to the final results, the observed odds ratios and the corresponding 95 per cent confidence limits at each interim analysis (Table I) remained above the OBF lower boundary and compatible with the null hypothesis. In other words, at no time during subject accrual or at study completion did CPEP data present the evidence required to reject the hypothesis of no effect.