A modified test for small-study effects in meta-analyses of controlled trials with binary endpoints

Publication bias and related bias in meta-analysis is often examined by visually checking for asymmetry in funnel plots of treatment effect against its standard error. Formal statistical tests of funnel plot asymmetry have been proposed, but when applied to binary outcome data these can give false-positive rates that are higher than the nominal level in some situations (large treatment effects, or few events per trial, or all trials of similar sizes). We develop a modified linear regression test for funnel plot asymmetry based on the efficient score and its variance, Fisher's information. The performance of this test is compared to the other proposed tests in simulation analyses based on the characteristics of published controlled trials. When there is little or no between-trial heterogeneity, this modified test has a false-positive rate close to the nominal level while maintaining similar power to the original linear regression test ('Egger' test). When the degree of between-trial heterogeneity is large, none of the tests that have been proposed has uniformly good properties.

Systematic review of the empirical evidence of study publication bias and outcome reporting bias

BACKGROUND: The increased use of meta-analysis in systematic reviews of healthcare interventions has highlighted several types of bias that can arise during the completion of a randomised controlled trial. Study publication bias has been recognised as a potential threat to the validity of meta-analysis and can make the readily available evidence unreliable for decision making. Until recently, outcome reporting bias has received less attention. METHODOLOGY/PRINCIPAL FINDINGS: We review and summarise the evidence from a series of cohort studies that have assessed study publication bias and outcome reporting bias in randomised controlled trials. Sixteen studies were eligible of which only two followed the cohort all the way through from protocol approval to information regarding publication of outcomes. Eleven of the studies investigated study publication bias and five investigated outcome reporting bias. Three studies have found that statistically significant outcomes had a higher odds of being fully reported compared to non-significant outcomes (range of odds ratios: 2.2 to 4.7). In comparing trial publications to protocols, we found that 40-62% of studies had at least one primary outcome that was changed, introduced, or omitted. We decided not to undertake meta-analysis due to the differences between studies. CONCLUSIONS: Recent work provides direct empirical evidence for the existence of study publication bias and outcome reporting bias. There is strong evidence of an association between significant results and publication; studies that report positive or significant results are more likely to be published and outcomes that are statistically significant have higher odds of being fully reported. Publications have been found to be inconsistent with their protocols. Researchers need to be aware of the problems of both types of bias and efforts should be concentrated on improving the reporting of trials.

The effects of excluding patients from the analysis in randomised controlled trials: meta-epidemiological study

OBJECTIVE: To examine whether excluding patients from the analysis of randomised trials are associated with biased estimates of treatment effects and higher heterogeneity between trials. DESIGN: Meta-epidemiological study based on a collection of meta-analyses of randomised trials. DATA SOURCES: 14 meta-analyses including 167 trials that compared therapeutic interventions with placebo or non-intervention control in patients with osteoarthritis of the hip or knee and used patient reported pain as an outcome. METHODS: Effect sizes were calculated from differences in means of pain intensity between groups at the end of follow-up, divided by the pooled standard deviation. Trials were combined by using random effects meta-analysis. Estimates of treatment effects were compared between trials with and trials without exclusions from the analysis, and the impact of restricting meta-analyses to trials without exclusions was assessed. RESULTS: 39 trials (23%) had included all patients in the analysis. In 128 trials (77%) some patients were excluded from the analysis. Effect sizes from trials with exclusions tended to be more beneficial than those from trials without exclusions (difference -0.13, 95% confidence interval -0.29 to 0.04). However, estimates of bias between individual meta-analyses varied considerably (tau(2)=0.07). Tests of interaction between exclusions from the analysis and estimates of treatment effects were positive in five meta-analyses. Stratified analyses indicated that differences in effect sizes between trials with and trials without exclusions were more pronounced in meta-analyses with high between trial heterogeneity, in meta-analyses with large estimated treatment benefits, and in meta-analyses of complementary medicine. Restriction of meta-analyses to trials without exclusions resulted in smaller estimated treatment benefits, larger P values, and considerable decreases in between trial heterogeneity. CONCLUSION: Excluding patients from the analysis in randomised trials often results in biased estimates of treatment effects, but the extent and direction of bias is unpredictable. Results from intention to treat analyses should always be described in reports of randomised trials. In systematic reviews, the influence of exclusions from the analysis on estimated treatment effects should routinely be assessed.

Disagreements in meta-analyses using outcomes measured on continuous or rating scales: observer agreement study

OBJECTIVE: To study the inter-observer variation related to extraction of continuous and numerical rating scale data from trial reports for use in meta-analyses. DESIGN: Observer agreement study. DATA SOURCES: A random sample of 10 Cochrane reviews that presented a result as a standardised mean difference (SMD), the protocols for the reviews and the trial reports (n=45) were retrieved. DATA EXTRACTION: Five experienced methodologists and five PhD students independently extracted data from the trial reports for calculation of the first SMD result in each review. The observers did not have access to the reviews but to the protocols, where the relevant outcome was highlighted. The agreement was analysed at both trial and meta-analysis level, pairing the observers in all possible ways (45 pairs, yielding 2025 pairs of trials and 450 pairs of meta-analyses). Agreement was defined as SMDs that differed less than 0.1 in their point estimates or confidence intervals. RESULTS: The agreement was 53% at trial level and 31% at meta-analysis level. Including all pairs, the median disagreement was SMD=0.22 (interquartile range 0.07-0.61). The experts agreed somewhat more than the PhD students at trial level (61% v 46%), but not at meta-analysis level. Important reasons for disagreement were differences in selection of time points, scales, control groups, and type of calculations; whether to include a trial in the meta-analysis; and data extraction errors made by the observers. In 14 out of the 100 SMDs calculated at the meta-analysis level, individual observers reached different conclusions than the originally published review. CONCLUSIONS: Disagreements were common and often larger than the effect of commonly used treatments. Meta-analyses using SMDs are prone to observer variation and should be interpreted with caution. The reliability of meta-analyses might be improved by having more detailed review protocols, more than one observer, and statistical expertise.

Combining follow-up and change data is valid in meta-analyses of continuous outcomes: a meta-epidemiological study

OBJECTIVE To investigate whether it is valid to combine follow-up and change data when conducting meta-analyses of continuous outcomes. STUDY DESIGN AND SETTING Meta-epidemiological study of randomized controlled trials in patients with osteoarthritis of the knee/hip, which assessed patient-reported pain. We calculated standardized mean differences (SMDs) based on follow-up and change data, and pooled within-trial differences in SMDs. We also derived pooled SMDs indicating the largest treatment effect within a trial (optimistic selection of SMDs) and derived pooled SMDs from the estimate indicating the smallest treatment effect within a trial (pessimistic selection of SMDs). RESULTS A total of 21 meta-analyses with 189 trials with 292 randomized comparisons in 41,256 patients were included. On average, SMDs were 0.04 standard deviation units more beneficial when follow-up values were used (difference in SMDs: -0.04; 95% confidence interval: -0.13, 0.06; P=0.44). In 13 meta-analyses (62%), there was a relevant difference in clinical and/or significance level between optimistic and pessimistic pooled SMDs. CONCLUSION On average, there is no relevant difference between follow-up and change data SMDs, and combining these estimates in meta-analysis is generally valid. Decision on which type of data to use when both follow-up and change data are available should be prespecified in the meta-analysis protocol.