Application of publication bias detection methods in evaluating the effect of Albendazole in treating patients with Ascaris and Trichocephal

Meta analysis is a statistical method to combine the findings of a set of large number of published individual studies and re-analyse them. The use of meta-analysis methods in medical research has been increased, noticeably, in resent years. However, one of the major shortcomings in such analysis is that the researcher, could not access all conducted studies in the area of concern. This serious problem is called publication bias, and may lead to a misleading conclusion. In present study a meta analysis was carried out to evaluate the effect of albendazole in treating patients with ascaris and trichocephal and different publication bias detection methods were compared.

Data were gathered from 25 clinical trials, having two treatment groups and binary outcomes, published in international indexed journals during 1982-2000. The main objectives of all these studies were to compare the performance of single dose of albendazole to treat patients with ascaris and tricocephal parasites. Total number of participants, number of patients with ascaris and tricocephal and cure rates were recorded for each study. Log odds ratio and its variance were calculated for each study and then estimated for the overall researches using meta-analysis methods. Funnel plot, Begg correlation method, two regression methods introduced by Egger and two regression methods based on Funnel plot were used and compared in detecting publication bias. SAS software was employed for this mean.

Meta analysis showed that albendazole has 10.3 (95%CI: 8.35-12.65) times better performance in curing ascaris patients compared to those with tricocephal (log odds ratio 2.33). All the employed procedures showed similar and consistent results which confirmed non-significant publication bias. However, this consistency may not occur all the time.

Since this consistency may not occur all the time, it is recommended not to rely on results of only one method, but comparing and selecting a powerful method which fits the data structure.