Assessment of Heterogeneity in Heart Failure–Related Meta-Analyses

Background:

Assessment of heterogeneity in meta-analyses is critical to ensure the consistency of pooled results. Therefore, we sought to assess the evaluation and reporting of heterogeneity in heart failure (HF) meta-analyses.

Methods:

Study level meta-analyses pertaining to HF were selected from January 2009 to July 2019, published in 11 high impact factor journals. We tabulated the overall proportion of the meta-analyses reporting statistical heterogeneity and specific metrics and methods employed to quantify and explore heterogeneity.

Results:

Of 126 HF meta-analyses (612 outcomes), heterogeneity was reported for 422 outcomes (68.9 %) in 108 meta-analyses. Out of the 422 outcomes reporting statistical heterogeneity, 137 outcomes (32.5%) had no observable heterogeneity: (I²=0%), 40 outcomes (9.5%) had low heterogeneity (I²<25%), 76 outcomes (18%) had moderate heterogeneity (I²=25%–50%), and 169 outcomes (40%) had high heterogeneity (I²>50%). Reporting of statistical heterogeneity was not significantly associated with year of publication, funding source, disclosure information, or the type of studies pooled. Sensitivity analysis (n=68) was the most common statistical technique employed to evaluate the source of heterogeneity followed by subgroup analyses (n=59) and meta-regression (n=40).

Conclusions:

Despite being an essential component of meta-analyses, heterogeneity was not reported for nearly 30% of outcomes and variably handled in contemporary HF meta-analyses. As meta-analyses increase across HF science, interpreting and handling of heterogeneity should be standardized.