Inclusion of zero total event trials in meta-analyses maintains analytic consistency and incorporates all available data

Abstract

Background: Meta-analysis handles randomized trials with no outcome events in both treatment and control arms inconsistently, including them when risk difference (RD) is the effect measure but excluding them when relative risk (RR) or odds ratio (OR) are used. This study examined the influence of such trials on pooled treatment effects.

Methods: Analysis with and without zero total event trials of three illustrative published meta-analyses with a range of proportions of zero total event trials, treatment effects, and heterogeneity using inverse variance weighting and random effects that incorporates between-study heterogeneity.

Results: Including zero total event trials in meta-analyses moves the pooled estimate of treatment effect closer to nil, decreases its confidence interval and decreases between-study heterogeneity. For RR and OR, inclusion of such trials causes small changes, even when they comprise the large majority of included trials. For RD, the changes are more substantial, and in extreme cases can eliminate a statistically significant effect estimate.

Conclusion: To include all relevant data regardless of effect measure chosen, reviewers should also include zero total event trials when calculating pooled estimates using OR and RR.

Figure 1

Meta-analysis investigating the use of low-dose dopamine to prevent or treat renal dysfunction [13]. Endpoint: mortality (total of 109 events in 1523 patients in the intervention group and 105 events in 1454 patients in the control group). The first three horizontal bar charts show the proportion of trials, patients, and outcome events in the meta-analysis of all "Trials with No Zeros" (trials with at least one event in both the intervention and control arms), "Trials with One Zero" (trials with no events in either the intervention or the control arm), and "Trials with Two Zeros" (trials with no events in both the intervention and control arms). The next three pairs of bar charts show the total weighting of all trials with no zeros, one zero, and two zeros, respectively, in the final pooled RD, OR, and RR. The first bar in each pair refers to a meta-analysis that includes only trials with at least one event in either the intervention or the control arm ("Event trials"). The second bar includes both trials with and without events ("All trials"). For each effect measure, a corresponding plot indicates the pooled treatment effect estimate with its 95% confidence interval (white circles when event trials only are included and black circles when all trials are included), and the I² heterogeneity measure. The RD is plotted on a linear scale, and OR and RR are plotted on the same logarithmic scale. Abbreviations: RD = risk difference, OR = odds ratio, RR = relative risk.

Figure 2

Meta-analysis investigating the use of antibiotics to treat sore throat [14]. Endpoint: acute rheumatic fever within 2 months of treatment (total of 37 events in 5656 patients in the intervention group and 74 events in 4445 patients in the control group). For a description of bar charts and plots see figure legend for Figure 1.

Figure 3

Meta-analysis investigating the use of prophylactic subcutaneous unfractionated or low-molecular weight heparin after hip fracture repair to prevent deep vein thrombosis [15]. Endpoint: non-fatal pulmonary embolism (total of 8 events in 245 patients in the intervention group and 0 events in 292 patients in the control group). For a description of bar charts and plots see figure legend for Figure 1.