The study by Azevedo et al14 offers three major opportunities to further improve outcomes in this high-risk population (Fig 1). First, although the difference was not statistically significant, patients who died spent more time in wards before ICU admission than patients who survived (2 median [0–8 interquartile range] vs 4 [0–13] days, P = .11). Similarly, earlier studies showed that delayed ICU admission was associated with higher mortality.7,22-24 Interventional studies on optimal ICU admission timing are warranted, not only in the overall population of patients with cancer but also in patients receiving chemotherapy and having a single mild organ dysfunction. Better delineation of the criteria that should prompt oncologists and hematologists to consider ICU admission, and intensivists to admit patients with cancer, is urgently needed. The second opportunity for improvement identified in the study by Azevedo et al14 pertains to the high mortality after NIV failure. NIV was recommended for first-line ventilatory support in immunocompromised patients at a time when MV was associated with 90% mortality.25 However, since then, the marked decrease in mortality and the concerns raised about NIV in hypoxemic patients have challenged the wisdom of this approach.18,26 We believe that NIV should not be used in patients with ARDS. In immunocompromised patients with acute respiratory failure but no criteria for ARDS, the evidence has to be confirmed. Thus, a trial of NIV is warranted to appraise the findings that were reported 15 years ago. Last, the study by Azevedo et al14 suggests a need for clearly defining the standard of care for critically ill patients with cancer. For instance, the finding that only one-third of patients with cancer admitted to the participating ICUs received MV and that among patients given MV only one-third received vasopressors and only 7% renal replacement therapy casts doubt on whether appropriate intensity of care was provided. Along this line, all the deaths occurred after treatment-limitation decisions. The use of intensive care must change in patients with cancer. Studies must provide survival rates separately for patients who receive full-code management, an ICU trial, or palliative ICU management. In addition to hospital mortality, these studies must provide data on long-term overall survival, event-free and disease-free survival, quality of life, and other markers of post-ICU burden.