The authors stated that 17 studies were included in the review, but only 16 studies (n=4,061 participants) were listed in the table of overall results. Three studies were RCTs. The authors stated that nine of the 16 studies listed in the table of results had no control group; there was a discrepancy in this estimate, as the study by Chen appeared to have a control group.
The overall proportion of participants with leaks (with and without reinforcement) ranged from zero to 10%. In seven studies that reported any leaks (n=3,299), the population odds of leaks was 7.69% (95% CI 3.44 to 17.18). In three controlled studies (n=1,899), no reinforcement was associated with a marginally statistically significant decrease in the risk of leaks (RR 0.95, 95% CI 0.92 to 0.99) compared to reinforcement, with moderate to large heterogeneity. The funnel plot of the three controlled studies showed an asymmetric relationship.
Estimates in the forest plot that included three controlled studies did not correspond with the raw figures in the table. In the table, a higher rate of leaks was reported for one study (Jones) in the no reinforcement group in comparison to the reinforcement group and in another study (Chen) a lower rate of leaks was reported in the no reinforcement group in comparison to the reinforcement group. The forest plot displayed both estimates (Jones and Chen) to the right on the vertical no difference line, which suggested similar relative risks.