I am not particularly surprised that the SELUTION DeNovo paper has not yet been published if this is the interpretation being promoted by the authors.
The noninferiority margin was 2.44%. In the intention-to-treat analysis, the upper bound of the risk difference for TLF was 2.38%, therefore technically meeting the criterion for noninferiority. However, in a noninferiority trial the per-protocol analysis carries greater weight, and in that analysis the upper bound of the risk difference is 2.63%, meaning that noninferiority is not achieved. To be precise, in noninferiority trials both intention-to-treat and per-protocol analyses are required, and concordant results are generally expected; when they diverge, the evidence for noninferiority is considered weak.
The authors now state that “the results of the per-protocol analysis did not confirm noninferiority, but they are similar to the intention-to-treat analysis in both direction and magnitude.” Well, where I come from we would say: “If my grandfather had wheels, he would have been a cart.”