Optimal Oxygenation: Exploring the Boundaries

Oxygen has been among us for billions of years. Since the beginning of human life, the concentration on earth has been virtually stable at 21%, meaning that our mitochondria and body are greatly adapted to this specific fraction of inspired oxygen. It characterizes a long-lived and trusted collaboration between a vital element and our eukaryotic cells. An extraordinary alliance that is uniformly embraced and that makes one reflect—what is there not to like about oxygen?

In our daily clinical practice, we are well aware of the detrimental effects of a serious deficiency in arterial oxygen levels (hypoxemia). In this context, oxygen therapy is life-saving and a cornerstone in the management of patients at risk. In recent years, the effects of hyperoxemia are also increasingly investigated and give rise to a more careful approach to supplemental oxygen. An excess of oxygen may lead to inflammation, oxidative stress with metabolic disturbances, hemodynamic changes with a reduction in cardiac output, and pulmonary restrictions including atelectasis and a ventilation-perfusion mismatch. However, we still do not know exactly at what boundaries arterial oxygen reaches critical levels and where the sweet spots are when targeting “normoxemia” for specific patients. Guidelines increasingly recommend patient-tailored and peripheral oxygen saturation (Sp_O2) guided approaches where both hypoxemia and hyperoxemia are actively prevented (1, 2).

Considering the accumulating evidence, we are on the eve of a deeper understanding of the extremes of oxygen levels. In this issue of Annals ATS, Tyagi and colleagues (pp. 1338–1345) explore the effects of the higher and lower end of the arterial oxygen spectrum (3). They do so following an observational approach in more than 2,000 intensive care unit (ICU) patients from a single center in Michigan, USA. Their findings confirm the harmful of effects of supranormal arterial oxygen concentrations (hyperoxemia) using a variety of (cumulative) exposure variables. The exposure to hyperoxemia was defined by a previously suggested threshold (Pa_O2 ⩾ 200 mm Hg) (4) and was found to be strongly associated with mortality for both the dose and duration. The results of this study build on previous observational work, yet identify new views on oxygen exposure as a risk factor for mortality. Major assets are the sensitivity analyses and the (alluvial plot) analyses in which the authors follow the day-by-day paths of oxygen exposure in the ICU. They provide more insight in the characteristics of the exposure and, thereby, offer providers a more detailed sense of oxygen therapy management and a firmer grip on how to limit the severity and duration of exposure.

Another important finding of this study is that the actual exposure to hyperoxemia is still far more common than we think and infer from surveys (5). One explanation could be that hyperoxic periods are not easily and attentively detected by clinicians. Another explanation could be that during the study period (2017–2018), the possible harm of excessive oxygen concentrations was not crystal clear. The impact of its harm is still not generally acknowledged, and there may be a good reason for that. First, results from previous cohort studies, but also from recent RCTs (randomized controlled trials) have not been unequivocal (6, 7). Second, oxygen is among the most widely used drugs in modern medicine and the effects of hypoxemia are usually more pertinent. Third, even when the risk of harm is substantiated, it remains uncertain after what duration and to what extent this may directly impact patient-centered outcomes. Fortunately, the authors address some of these issues in this cohort by suggesting that the relationship between duration of severe hyperoxemia and mortality is linear and has no “floor,” meaning that there is no minimal time of exposure that could be considered “safe.” Accordingly, any exposure to hyperoxemia, independent of duration, was associated with increased 30-day mortality.

The present study identifies meaningful methodological flaws in existing observational research but also exposes possible causes of discrepancy in (and limitations of) recent RCTs.

Such reasons may include overlapping oxygenation exposures between the intervention and control group. Comparing two groups targeting nearly contiguous oxygenation ranges that can both be considered normoxemia, may in turn be the key reason for underestimating the effects of hyperoxemia in interventional studies. In a research setting it proves to be very difficult to accomplish a sufficient contrast between the intervention and control group. To adequately differentiate between groups, both a statistical and a clinical significant difference is needed. Furthermore, the selection of specific patient categories may, depending on pathophysiology, rule out or strengthen the overall effects that are observed in large heterogeneous cohorts.

When interpreting the results of this study, a few limitations, such as the single institution data, should be considered. It is entirely possible that several unmeasured confounders could work together to either null out or reverse the observed effects of hyperoxemia. Residual bias, in specific informative censoring bias following death or rapid recovery, may have prevented included patients from longer exposure times. The authors made effort to minimize immortal time bias by only including patients with more than 24 hours of invasive ventilation, and by adjusting the models for mechanical ventilation time. However, the exact cause of death was not determined and can in general not be linked to oxygen directly. Predisposing lung injury and other co-morbidity may also impact the outcomes, even when the analyses are adjusted for illness severity or co-existing disease scores such as SOFA (sequential organ failure assessment) and Elixhauser.

The next steps in the field should be aimed at exploring the causal pathways, characterizing potential mechanisms, and identifying safe oxygen margins for individual patients. Until then, an attentive, tailored and goal-directed approach by titrating supplemental oxygen remains a safe and balanced strategy for oxygen therapy in the ICU.