The low level of proof of most studies considered, the diversity of definitions of early or late initiation, the heterogeneity of the study populations, unequal quality of available data, observed bias, including a strongly suspected publication bias, and the small number of patients in these studies mean that definitive conclusions cannot be drawn.
After exclusion of small studies, because of suspected publication bias, the pooled relative risk was 2.94 (95% CI: 1.89 4.59).
For statistically suspected publication bias, the trim and fill method was adopted to adjust the pooled risk (13).
The pooled effect size across studies in the high-exposure category was 2.94 (95% CI 1.89 4.59) after excluding the three studies with largest standard errors in order to make a crude adjustment for the suspected publication bias.
After exclusion of the three studies that had the highest standard error, because of suspected publication bias, the pooled effect sizes for studies in the high-exposure category was 2.94 (95% CI 1.89 4.59).
28 The overall quality was considered to be high when at least 75% of RCTs with no limitations of study design had consistent findings, direct and precise data and no known or suspected publication biases.
The major concern in judging the strength of evidence is a statistically suspected publication bias for the lower ACM risk associated with high PA, which may change the strength of the association.
The following criteria were used for assessment of the quality of evidence: High-quality evidence = there are consistent findings in at least 75% of RCTs with no limitations of the study design, consistency, directness, or precision and no known or suspected publication bias.
To qualify as high quality evidence, more than 75% of the RCTs within a comparison had to be judged to have no limitations of study design, have consistent findings among multiple studies, present direct (generalizable) and precise data, without known or suspected publication bias.
