LAUSR

In this issue of Pediatric Critical Care Medicine, Numa et al (1) present a single-center retrospective study of the association between Pao 2 measured within the first hour of admission and risk-adjusted mortality of children admitted to a PICU between 2012 and 2017. A retrospective chart review identified patients with a Pao 2 measured within 60 minutes of admission to the PICU (28% of the total PICU patients). Cases were assigned to Pao 2 groups (1–5, 51–100, 101–150 mm Hg, etc, with a final group > 350 mm Hg), and mortality was adjusted for Pediatric Index of Mortality (PIM)–3. The authors found that the cases were older, more likely to be treated with invasive or noninvasive mechanical ventilatory support, and had a higher mortality compared with controls. Although admission diagnoses were similar in the entire cohort compared with the cases, postoperative cardiac surgery was significantly skewed toward the cases (n = 256 in the entire cohort, n = 245 in cases, leaving only n = 11 in the control group). The authors found a U-shaped relationship between mortality and admission Pao 2 , with the nadir at Pao 2 equals to 101–150 mm Hg and the highest in patients with Pao 2 greater than 350 mm Hg. When adjusted for PIM-3, the U-shaped relationship between Pao 2 and mortality was lost, but mortality increased sharply when Pao 2 was greater than or equal to 250 mm Hg. Given the skewed population of cardiac surgery patients in the cases, the subgroup analysis was repeated without these patients, and the authors found similar results. In order to determine whether Pao 2 , as an independent term, improved the accuracy of PIM-3 (Fio 2 :Pao 2 is already included), the authors explored Pao 2 as a dichotomous (Pao 2 ≤ or > 250 mm Hg) and as a continuous variable, finding that Pao 2 greater than 250 mm Hg marginally improved performance of PIM-3, providing a 2.66 odds ratio for hyperoxia predicting death (p = 0.047). The authors conclude that hyperoxia was associated with increased risk-adjusted mortality but identify the limitation of a selection bias of patients with a Pao 2 measured in the first hour. Other limitations such as use of vasoactive medications, ventilatory support, fluid management, as well as surgical and procedural interventions, were not controlled. Delivery of supplemental oxygen is, without debate, a potential life-saving therapy. However, investigation into oxygen toxicity in the lungs and systemically dates back to at least the 1930s (2). Intersections between regulation of inflammation, apoptosis, necrosis, and oxidative stress are plausible mechanisms by which hyperoxia contributes to systemic illness and multiple organ failure in critically ill patients (3–5). Laboratory studies suggest that patients with respiratory failure may be the most susceptible to the effects of oxygen toxicity due to interactions with mechanical stretch that worsen lung injury (6, 7). However, this does not appear to be the population that the authors have identified in the study by Numa et al (1), as those with lung injury are unlikely to have a Pao 2 greater than 250, even when Fio 2 equals to 1.0. Despite overwhelming evidence that hyperoxia contributes to pulmonary and systemic organ injury, animal and human data suggest that oxygen toxicity is not likely to be due to a single, discrete exposure (2, 8). Furthermore, otherwise healthy primates and humans are among the least susceptible to oxygen toxicity in the lungs, requiring prolonged exposures to even hyperbaric conditions to detect small impairments in pulmonary function (2). Therefore, there is concern for significant residual confounders in the study by Numa et al (1). The PIM-3 score is designed to be a parsimonious measure to adjust for mortality risk. The majority of data represented in PIM-3 is based on diagnostic factors, with only minimal directed adjustment for physiology variables (systolic blood pressure, Pao 2 :Fio 2 [PF] ratio, base excess). Patients who are treated with high degrees of supplemental oxygen at admission to the ICU may have significant physiologic derangements that are inadequately captured by elements in the PIM-3 score. To that end, we recently found in a retrospective observational study including over 1,000 pediatric acute respiratory distress syndrome (ARDS) patients that the cumulative average of Pao 2 on day 1 of mechanical ventilation for ARDS also had a U-shaped curve with mortality: Pao 2 less than 60 mm Hg 44% mortality, Pao 2 60–100 mm Hg 16% mortality, Pao 2 greater than 100 mm Hg 23% mortality. However, this mortality risk was no longer significant after adjusting for ventilator settings, pediatric ARDS triggers, Pediatric Risk of Mortality, and PF ratio (9). *See also p. 699.