Researchers behind a multicenter study of respiratory management in critically ill patients with Pneumocystis jirovecii pneumonia (PjP) have published a formal response to methodological critiques, reporting that an additional competing-risk analysis preserved their central finding: high-flow nasal cannula (HFNC) remained independently associated with a reduced hazard of intubation compared with standard oxygen. The response, addressed to Wang and Cheng in the Annals of Intensive Care, also concedes that censoring patients who died without intubation may have overestimated the original “intubation-free” probability, and it details the statistical steps taken to address that concern.
The core of the new analysis is a Fine–Gray competing-risk model, a statistical framework that treats death before intubation as a competing event rather than as ordinary censoring. In standard survival analyses, patients who die without reaching the endpoint of interest are simply removed from the risk set at the time of death, which can inflate the estimated probability of remaining intubation-free. The Fine–Gray approach instead models the subdistribution hazard, explicitly accounting for the fact that death precludes intubation. In this reanalysis, HFNC was associated with a subdistribution hazard ratio of 0.46 for intubation compared with standard oxygen, with a 95% confidence interval of 0.28 to 0.78 and a p-value of 0.004. Non-invasive ventilation (NIV), by contrast, showed no significant effect, with a subdistribution hazard ratio of 0.84, a 95% confidence interval of 0.51 to 1.36, and a p-value of 0.47.
The authors also adjusted the model for two sources of heterogeneity they acknowledged in the original work: variability across centers and temporal changes in practice. The study cohort spanned three epochs, 2011 to 2014, 2015 to 2018, and 2019 to 2021, and critical care practice, including oxygenation strategy, has evolved considerably over that period. By incorporating both center and epoch into the competing-risk framework, the team sought to verify that the association between HFNC and lower intubation risk was not an artifact of practice patterns at particular hospitals or of shifts in care over time. They state that these findings confirm the robustness of their primary results after accounting for temporal and inter-center heterogeneity.
A second methodological question concerned patients who alternated between HFNC and NIV during their intensive care stay. The authors explain that these patients primarily received repeated NIV sessions, with HFNC used for comfort between sessions. They therefore classified such patients under NIV, arguing that this reflects bedside practice more accurately than a multi-state model would. In their view, a multi-state model would overemphasize the short HFNC intervals that separated NIV sessions, potentially misattributing time at risk. The authors are explicit that only a randomized trial could definitively disentangle the respective contributions of HFNC and NIV in this population.
The response also confronts the issue of residual confounding, a persistent limitation of retrospective observational research. Despite the use of inverse probability weighting and adjustment for center and epoch, the authors acknowledge that unmeasured differences between patients treated with different oxygenation strategies could still influence the results. A fully multivariable Fine–Gray model including all baseline covariates, they note, would require a dedicated analysis and exceeds the scope of a letter. Their stated goal was narrower: to verify that the association between HFNC and lower intubation risk persisted under competing-risk modeling and after addressing key sources of heterogeneity, an objective they argue these additional analyses achieve.
On the question of which patients benefit most, the authors address concerns about poor prognosis in specific subgroups, including patients with long-term corticosteroid use, solid tumors, or high SOFA scores. Exploratory subgroup analyses did not reveal significant interactions with HFNC, but the team is careful to state that the sample size was insufficient for firm conclusions. This is a common constraint in subgroup analysis: the statistical power to detect a genuine interaction is often far lower than the power to detect a main effect, and null findings in small subgroups should not be read as evidence of no difference.
Looking forward, the authors identify several promising avenues for improving outcomes in severe PjP. These include reinforcing pneumocystis prophylaxis and implementing standardized HFNC escalation strategies, with examples such as ROX index thresholds, early awake prone positioning, and predefined intubation criteria. The ROX index, a ratio incorporating respiratory rate and oxygenation, is used in other forms of respiratory failure to flag patients who may need escalation to intubation, and the authors suggest that such thresholds, along with structured escalation pathways, should be evaluated prospectively rather than assumed to work.
The bottom line, as the authors frame it, is that their study provides real-world evidence that HFNC is associated with a lower risk of intubation in severe PjP, although no survival benefit at day 90 was observed. The additional competing-risk analysis supports the robustness of this finding while underlining the inherent limitations of retrospective data. The authors emphasize that only a prospective randomized controlled trial will definitively establish the role of HFNC versus alternative oxygenation strategies in critically ill patients with PjP, and they position the current work as a step toward, not a substitute for, that trial.
For clinicians managing severe Pneumocystis pneumonia, a fungal-like infection that causes profound hypoxemia in immunocompromised patients, the practical implications remain cautious. The evidence base now includes a competing-risk analysis that accounts for death as a competing event and for practice variation across centers and years, strengthening the observational signal favoring HFNC over standard oxygen for avoiding intubation. Yet the same analysis found no significant effect for NIV, subgroup findings remain underpowered, residual confounding cannot be excluded, and mortality at day 90 was unchanged. The authors’ own conclusion is that definitive answers on oxygenation strategy in this vulnerable population must come from a prospective randomized controlled trial.
Subject of Research: A response letter presenting additional competing-risk analyses of respiratory oxygenation strategies and intubation risk in critically ill patients with severe Pneumocystis jirovecii pneumonia.
Article Title: Response to “high-flow nasal cannula and intubation risk in severe PjP: methodological and clinical perspectives”
Article References: Reizine, F., Stiegler, V., Tessoulin, B., Issa, N., Gaborit, B., Raffi, F., Boutoille, D., Biron, C., Lefebvre, M., Gaborit, B. J., Le Turnier, P., Deschanvres, C., Lecomte, R., Chauveau, M., Lécuyer, R., Asquier-Khati, A., Pineau, V., Prime, M., Delaunay, C., … Pouedras, P. (2025). Response to “high-flow nasal cannula and intubation risk in severe PjP: methodological and clinical perspectives”. Annals of Intensive Care, 15(1), Article 182. https://doi.org/10.1186/s13613-025-01573-6
Image Credits: AI Generated
DOI: 10.1186/s13613-025-01573-6
Keywords: high-flow nasal cannula, Pneumocystis jirovecii pneumonia, intubation risk, competing-risk analysis, Fine–Gray model, non-invasive ventilation, retrospective study, residual confounding, inverse probability weighting, randomized controlled trial, intensive care, HFNC escalation
News Source: Ophelia Keating. (October 30, 2026). Competing-risk analysis reinforces HFNC’s link to lower intubation risk in severe Pneumocystis pneumonia. Scienmag.
