Article
Agreement of intervention effects between classical randomized and registry based randomized trials
Search Medline for
Authors
-
Tim Mathes
- Institut für Forschung in der Operativen Medizin, Universität Witten/Herdecke
-
Pauline Klaßen
- Institut für Forschung in der Operativen Medizin, Universität Witten/Herdecke
-
Dawid Pieper
- Universität Witten/Herdecke
| Published: | March 6, 2018 |
|---|
Outline
Text
Background: Research has shown that registries are appropriate for collecting outcomes for pragmatic randomized trials. However, it is anticipated that the data quality of registry randomized trials (rRCTs) is lower compared to “conventional” RCTs (cRCTs), especially because of a higher amount of missing values and therefore higher risk of bias caused by differential missing.
The objective was assessing the difference in effect estimates of cRCTs and rRCTs
Methods: We searched Pubmed for rRCTs. We identified systematic reviews (SRs) that included one of the rRCTs (04/2017). We preferred Cochrane SR over non-Cochrane SR and included only the most up-to-date SR. We replicated the meta-analysis and compared the pooled odds ratio (OR) of cRCTs with the (pooled) ORs of the rRCT/s for mortality and diseases or complications incidence measures. To assess agreement between rRCT and cRCT, we counted the number of conflicting effect directions, overlapping 95% confidence intervals (CIs) and point estimates of rRCTs not included in the 95%CIs of the pooled effect of cRCTs. We extracted the study weights of the rRCT/s in the original meta-analysis to assess the impact on the body of evidence.
We will calculate the difference of the OR with 95% CIs, by subtracting the log OR of the rRCT, from the log OR of cRCTs and will compare the number of observed 95%/90% CIs overlap with the number of expected (assuming there is no difference) CIs overlap. We will present the results of these analyses at the conference.
Results: Overall we compared 15 and 14 effect estimates for mortality and incidence measures, respectively. Of this six included more than one rRCT. rRCT/s often had large weights in the meta-analysis (mortality: median 27%; incidence: median 33.1%) Conflicting effect directions were observed in 27% of mortality and 7.2% of other incidence measures. The point estimate of the rRCT was within the 95%CI of the cRCTs in 54% for mortality and 64% of incidence measures. 95%CI overlap was 100% for both outcomes.
Discussion: Our primarily results provide no indication, that effect estimates of rRCTs and cRCT differ systematically. The finding further supports the assumption that rRCTs are a promising tool for increasing external validity while controlling for internal validity.
In some research areas rRCT greatly contribute to the body of evidence (e.g., prevention). This suggests that in these areas using registries might be a good way to facilitate the conduct of RCTs.
