Refining normoxemia targets in acute burn care: Reply

In Reply:

We are grateful to Drs Tang, Han, and Liu for their insightful review and response¹ to our recent SAVE-O2 Burn randomized clinical trial investigating targeting normoxemia in adults with acute thermal burns admitted to a specialized burn unit.² We appreciate their constructive critiques regarding subgroup analyses, oxygenation variability, the technical limitations of peripheral oxygen saturation (SpO₂) monitoring in burn patients, and the merits of supplemental oxygen-free days (SOFD) as a primary outcome. We welcome this opportunity to address the issues they have raised.

As Drs Tang, Han, and Liu mention, subgroup analyses should always be interpreted with caution. The SAVE-O2 burn trial was cluster-randomized and stepped-wedge, powered for the primary outcome of SOFD, not for subgroup effects. While the subgroup analyses were prespecified in the statistical analysis plan, these secondary analyses were intended to investigate the potential for heterogeneity of treatment effect for hypothesis generation, but not to demonstrate causal effects within strata. Therefore, we were appropriately cautious about drawing any clinically meaningful conclusions from our subgroup analyses. Because these analyses were exploratory in nature, we did not apply post hoc Bonferroni or false discovery rate corrections.

Tang and colleagues also raise the possibility of confounding by specific comorbidities. With 1,437 patients from six sites and adjustment for multiple patient-level covariates, modeling each comorbidity separately would risk overfitting and unstable estimates. Therefore, we used the Elixhauser comorbidity index, a validated measure of overall comorbidity burden, to provide a parsimonious adjustment for baseline disease severity rather than modeling individual comorbidities separately, especially because we did not have specific hypotheses related to individual comorbidities.

Concerns about oxygenation variability and potential impacts on clinical outcomes are valid. We agree that fluctuations between episodes of hyperoxemia and hypoxemia can exacerbate oxidative stress and result in opposing effects on the pathophysiology of burn wound healing. However, we reported the patient's time spent in normoxemia (SpO₂ 90%–96%), hyperoxemia (SpO₂ > 96%), and hypoxemia (SpO₂ < 88%) during the first 72 hours.² Following the pragmatic trial intervention, time spent in normoxemia increased and time spent in hyperoxemia decreased, while < 1% of time was spent in hypoxemia across both study groups. These findings indicate that the intervention achieved separation between groups without increasing hypoxemia. While more granular longitudinal analysis of daily oxygen variability may be informative, it was outside the scope of the prespecified analytic plan in this pragmatic clinical trial.

We agree that SpO₂ monitoring can be challenging in patients with severe thermal burn injuries. However, we designed the SAVE-O2 trial as a pragmatic intervention implemented at the unit level in active burn units to reflect real-world practice. When pulse oximetry was unreliable or clinical concerns arose, arterial blood gases could be obtained at the discretion of the treating clinical team, and clinicians could override the trial protocol when clinically indicated (eg, suspected carbon monoxide toxicity). The trial population also included patients with inhalation injury (11%) and a range of burn severity. Thus, the study was intended to evaluate a targeted normoxemia strategy under real-world conditions where pulse oximetry may not always be ideal.

Finally, Tang and colleagues suggested that SOFD treats all supplemental oxygen equally, regardless of flow rate or fraction of inspired oxygen (FiO₂). We agree that accounting for differences in flow rate and FiO₂ could better reflect total oxygen exposure beyond the yes/no daily approach inherent to SOFD, though the tradeoff is interpretability to practicing clinicians resulting from a more complex statistical model. SOFD has the distinct advantage of having units of days alive and free of supplemental oxygen, which are clinically important and patient-centered; patients care more about liberation from supplemental oxygen than the specific flow rate of supplemental oxygen received.

Importantly, mortality was incorporated into the primary outcome by assigning an SOFD of −1 to patients who died in-hospital before day 28, to make death a worse outcome than being alive and 28 days on supplemental oxygen.^3,4 The lack of difference in SOFD, alongside the improvement in hospital-free days (HFDs) through day 90, likely reflects that these outcomes capture different constructs. We chose SOFD as our primary outcome because it closely aligned with an oxygen titration strategy and is directly related to oxygen use. Conversely, HFD captures overall hospital utilization and recovery, and many factors beyond oxygen management strategies can contribute to HFD through day 90. Therefore, it is plausible that our intervention could have little impact on oxygen-free days, yet also demonstrate improvement in HFD.

We once again thank Drs Tang, Han, and Liu for their reinforcement that clinicians should incorporate an assessment of the methodological approach when evaluating clinical trials and the relevance of the trial findings to their practice. The goal of the SAVE-O2 trial was to provide clinically meaningful results in the real-world practice setting of patients hospitalized with acute thermal burns.