Skip to main content
. 2014 Dec 8;56(12):1235–1242. doi: 10.1097/JOM.0000000000000323

FIGURE 3.

FIGURE 3.

Article quality among included studies does not affect the risk estimate. Included studies were assessed and divided into two groups (one comprising seven articles and the other six) on the basis of quality scores. If two articles with an identical quality score need to allocate into two groups, then we arbitrarily allocated the newer article to the high-quality group, to balance the group sizes. The meta-analysis of articles (random-effects model) with higher quality scores is shown in the upper panel, and the meta-analysis of articles with lower scores is presented in the lower panel.