There are a number of limitations of the study. Although the inclusion criteria unquestionably selected patients who had severe respiratory failure, they did not preclude the use of NIV. Thus, it is not surprising that the authors significantly reduced the need for intubation in their NIV group, considering that the protocol assured a 100% intubation rate in the CMV group. Furthermore, it is disturbing that only 64 of 256 patients (25%) who met the inclusion criteria were eventually enrolled in the study, raising concerns about “cherry picking.” Did the attending physicians, who had “the final decision,” enroll only patients they thought likely to succeed with NIV, or did they withhold such patients so that they would not be randomly intubated? In addition, although the authors state that in their stratified analysis, NIV reduced endotracheal intubation independently of the type of ARF, the small number of patients and unequal distribution in some of the subgroups might have introduced bias. In the NIV group, 20 of 31 patients had either an exacerbation of COPD or cardiogenic APE, whereas in the CMV group 14 of 33 patients had received these diagnoses. It is well-known that COPD and APE patients respond favorably to NIV treatment,4 and this distribution could have favored the NIV group. Based on these concerns, the results should be generalized to other clinical settings with caution.