Capture-recapture method for assessing publication bias

Background: Publication bias is an important factor that may result in selection bias and lead to overestimation of the intervention effect. Capture recapture method is considered as a potentially useful procedure for investigating and estimating publication bias. Methods: We conducted a systematic review to estimate the duration of protection provided by hepatitis B vaccine by measuring the anamnestic immune response to booster doses of vaccine and retrieved studies from three separate sources, including a) electronic databases, b) reference lists of the studies, and c) conference databases as well as contact with experts and manufacturers. Capture recapture and some conventional methods such as funnel plot, Begg test, Egger test, and trim and fill method were employed for assessing publication bias. Results: Based on capture recapture method, completeness of the overall search results was 87.2% [95% CI: 84.6% to 89.0%] and log-linear model suggested 5 [95% CI: 4.2 to 6.2] missing studies. The funnel plot was asymmetric but Begg and Egger tests results were not statistically significant and trill and fill approach made no change in pooled effect. Conclusions: Capture recapture method may be a useful practical approach for estimating the number of missing studies which are not usually detected by search strategy, although assumptions of this method may limit its general application in systematic reviews. In addition, capture recapture method may be considered as an alternative approach for estimating the extent of publication bias based on overlapping information rather than mirror image of extreme values on funnel plot.