Our results should also consider the risk of bias of individual studies and GRADE quality ratings. After gathering additional unpublished information, the overall number of unclear items—across all items of the risk of bias—decreased from 63·5% to 35·2%. This reduction points to an urgent need for complete and open reporting in this research area. Additionally, the confidence of estimate for primary outcomes was low or very low in multiple comparisons, reducing the certainty of the findings. Most very low ratings were for indirect comparisons, suggesting the need for additional well designed head-to-head studies. Whereas previous pairwise16, 51 or network meta-analyses19 of ADHD medications rated all comparisons as low or very low quality, attributable in part to unpublished information that we gathered and a more nuanced assessment, we could rate some comparisons as high or moderate quality. Of note, these comparisons included the most commonly used drugs for ADHD (ie, methylphenidate and amphetamines). Additionally, our stringent criteria for the risk of bias (ie, a study was assessed at overall low risk only when all individual items were at low risk) could have contributed to downgrade the final GRADE ratings.
