Levenson also examined outcomes stratifying trials by prescription of ICS at baseline, and by ICS use during the trial (either randomized ICS, or concomitant but not randomized ICS). The results did not show the clear benefit of ICS seen in the stratified analyses with mandatory randomized ICS noted above. The RD for the composite outcome among those with baseline ICS prescription was 2.56 (95% CI, 0.16 to 4.97) compared with 4.10 (95% CI, 1.67 to 6.53) among those without baseline ICS prescription. While this difference was not statistically significant, this result could reflect some evidence of ICS benefit if one considers confounding by severity, namely that those prescribed baseline ICS likely had more severe disease leading their physician to prescribe ICS. In the second of these analyses, examining concomitant (but not necessarily randomized) ICS use during the trial, a similar result was seen; RD for those using ICS during the trial was 2.80 (95% CI, 0.55 to 5.06) compared with 3.98 (95% CI, 1.42 to 6.55) for those not using concomitant ICS. The sharp contrast between these results and the Risk Difference for LABA versus non-LABA among patients using mandatory (randomized) ICS is only explicable if subjects not mandated by randomization to ICS were highly non-compliant with concomitant ICS therapy.