Associations between oxygenation status and prognosis in sepsis

To the editor,

We read with great interest the recent study [1] which investigated whether a relatively high arterial oxygen tension (PaO₂) is associated with improved survival rate in sepsis compared to conservative oxygenation targets. After propensity score matching (PSM), 2422 patients evenly distributed between the liberal and conservative oxygen groups. The findings indicated that a higher PaO₂ (≥ 80 mmHg) during the first three days in ICU was associated with lower 28-day mortality. While these results contribute significantly to the field of critical care medicine, several aspects require further clarification.

Mortality trends and potential bias

The study reported a significant reduction in 28-day mortality in the liberal group compared to the conservative group. However, according to the supplementary materials, we noted that the difference in mortality primarily occurred by day 7, with mortality rates of 14.3% in the liberal group compared to 6.9% in the conservative group—a reduction of approximately 7.4%. Interestingly, after day 7, the difference in mortality between the two groups appeared to stabilize, remaining at around 7% on day 14 (17.9% vs. 10.4%) and day 28 (22.5% vs. 14.8%), or even small on day 90 (31.3% vs. 27.0%). This pattern raises the question of whether the mortality benefit associated with higher oxygenation is predominantly a short-term effect or it still has more enduring implications. Also, it remains unclear whether this short-term mortality effect is directly attributable to differences in oxygenation status or whether it might be influenced by other confounding factors (e.g., patients who died within 7 days might be more likely due to disease severity), which highlighted the need for further analyses to better understand the reason for the observed mortality benefit and to explore its underlying mechanisms.

In addition, although PSM is a robust method for balancing measured covariates, it has its own limitations [2]. For instance, in this study, almost half of the patients were excluded during propensity score matching (PSM). By excluding a large proportion of study patients, PSM may reduce the representativeness of the sample and limit the generalizability of the findings. Moreover, while PSM adjusts for known confounders, it does not address potential unmeasured confounders, which may still significantly influence the observed outcomes [3]. For example, it remains unclear whether the observed differences in 7-day mortality rates between the liberal group and the conservative group are attributable to unmeasured confounders. Therefore, future analyses incorporating sensitivity analyses or analyses of the entire cohort may provide a more comprehensive evaluation of the robustness of these findings.

Subgroup analysis and interaction effects

In the subgroup analysis, authors reported that higher oxygenation levels were associated with a reduced 28-day mortality risk in the following populations: males, hospital-acquired sepsis, vasopressors, mechanical ventilation, acute respiratory distress syndrome, or lactate ≥ 4 mmol/L. However, the absence of reported p-values for interaction effects significantly limits the interpretability of these subgroup analyses. Without p-values for interaction, it is difficult to determine whether the differences observed across subgroups are statistically significant or merely reflect random sampling error. Furthermore, a substantial overlap in confidence intervals across subgroups suggests that the effects of higher oxygenation may not differ significantly among these groups. Reporting p-values for interaction effects would help clarify whether the mortality benefits of higher oxygenation are consistent across subgroups or if they are concentrated in specific patient subsets.

Competing risk analysis

The study employed a competing risk model to evaluate the relationship between oxygenation and ICU discharge, treating ICU discharge as the primary outcome and 28-day mortality as a competing event. While competing risk models are valuable tools in survival analysis, their application in this context warrants more careful consideration. The assumption underlying competing risk models is that the primary outcome and the competing event are mutually exclusive and independent, such as acute kidney injury and death. However, in critically ill patients, ICU discharge and 28-day mortality may not meet these criteria. Actually, when using ICU discharge as a positive outcome, the negative outcome is the inability to be discharged from the ICU, with death as the primary cause. This interdependence between ICU discharge and mortality challenges the validity of the model's assumptions and may lead to biased estimates of the effect of oxygenation on discharge outcomes.

Restricted cubic spline analysis

The study also utilized restricted cubic spline (RCS) regression to explore the relationship between PaO₂ levels and mortality. While RCS is an effective tool for modeling non-linear relationships, several points should be noted. In RCS analyses, the optimal reference point is typically chosen based on inflection points in the RCS graph, as these points offer the greatest clinical relevance and statistical interpretability. However, in this study, the selected reference point did not align with the inflection points, which increases the difficulty of interpreting the results. Additionally, the RCS analysis in this study was limited to the propensity score-matched cohort, excluding patients who were not included in the matching process. This limitation may hinder a comprehensive understanding of the relationship between PaO₂ and mortality. Expanding the analysis to include the full dataset could provide a more complete picture of the impact of oxygenation on all critically ill sepsis patients, thereby enhancing the generalizability of the findings.

Broader implications

Oxygen therapy is a cornerstone of ICU management, with optimal target still under debate [4]. The study’s results suggest that liberal oxygenation may confer a survival benefit, at least in the short term. However, the questions raised in this letter underscore the complexity of this issue and the need for further research to explore how oxygenation strategies affect outcomes in sepsis patients.

Conclusion

Dr. Hyun et al.’s study represents a valuable contribution by highlighting the potential benefits of liberal oxygenation in sepsis. We commend the authors for their rigorous investigation and encourage further studies to build upon these findings. Thank you for considering our comments on this compelling work.

Author contributions

Lihong Zhu wrote the letter, Juan Lin Raised the question. All authors have reviewed and approved the letter.

Funding

None.

Availability of data and materials

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

None.