Berücksichtigung von Heterogenität in Meta-Analyse von Randomisierten Kontrollierten Studien

Abstract

Heterogeneity in meta-analysis can be caused by chance, methodological or clinical variations between the included primary studies. To identify a clinical variable as a cause of heterogeneity in meta-analysis, one should firstly investigate chance and methodological variations. Meta-analysis with individual patient data (MA-IPD) has a greater potential than that with aggregate patient data (MA-APD) to detect which subgroups of patients get less, more or no benefit from an intervention. In this thesis two systematic reviews and a MA-APD were undertaken. The first review searched systematically for empirical studies on the impact of bias in randomised controlled trials (RCTs) on the results of meta-analysis. 39 studies were identified and 134 empirical comparisons between trials of high and low methodological quality could be extracted and summarized using a random effects model. RCTs with low quality scores, with inadequate method of randomisation, with inadequate concealment of allocation and those without any type of blinding, on average, overestimated the treatment efficacy. However, most of the empirical studies did not consider clinical causes of heterogeneity. This fact may confound the relation between methodological quality and effect size. Simultaneous investigation of diverse sources of heterogeneity is required. The second review searched systematically for empirical comparisons between MA-IPD and MA-APD. 70 comparisons were extracted from 25 empirical studies. Two thirds of the comparisons showed a tendency to overestimate the effect size and to reduce its precision by MA-APD in comparison to MA-IPD. However, the differences between the point estimates of both types of meta-analysis were small in all comparisons but one, and the paired t-test found no significant difference between the two types of meta-analysis. Furthermore, only half of the studies reported the results of a heterogeneity test. The methodological quality of RCTs was assessed in both types of meta-analysis only in a quarter of the studies. Clinical heterogeneity was investigated only in one third of the studies, using both types of meta-analysis, with no consistent results. A published MA-IPD on the efficacy of statins was reanalysed using aggregate patient data. The summary effect sizes of all cause mortality in both meta-analyses were extremely similar. Although the Cochran test of heterogeneity was significant, this was neither reported nor taken into account in MA-IPD. MA-APD combined the RCTs using a random effects model. The methodological variations between the RCTs were not assessed in MA-IPD. MA-APD investigated possible methodological and clinical causes of variation between the RCTs. These were determined a priori and justified with external evidence. Subgroup analysis and meta-regression were used to explore the relation between the selected causes of heterogeneity and mortality. By using a multivariate regression model that included terms for the methodological quality and baseline low-density lipoprotein a significant negative relationship between the inclusion of women in RCTs and the extent of reduction in mortality was detected. This result suggests a lower efficacy of statins in women which deserves further investigation in future RCTs. It can be concluded that the investigation of various sources of heterogeneity in meta-analysis of RCTs is still rare. Careful exploration of chance and methodological variations should precede the investigation of clinical heterogeneity and confounding between the sources of heterogeneity should be taken into consideration.