Prone Positioning in Children With ARDS: Title and subTitle BreakPositive Reflections on a Negative Clinical Trial

Acute respiratory distress syndrome (ARDS) is common, highly lethal, and not curable. Although important adult ARDS studies^{1 - 2} continue to undergo reappraisal,^{3 - 4} promising data have recently been reported for treatment of this entity in children. A recent randomized controlled trial,⁵ which described a beneficial effect of surfactant, was the first report of an effective treatment for ARDS.

In this issue of JAMA, Curley and colleagues⁶ from the same research group, the Pediatric Acute Lung Injury and Sepsis Investigators (PALISI), report on the effect of prone positioning in children with ARDS; oxygenation was improved, yet outcomes were unchanged. Clinicians will reflect on these results and ask how to incorporate the findings from this valuable study into providing the best possible care for their patients.

Why would a clinician wish to turn a critically ill child upside down? More than 30 years ago, Bryan⁷ reasoned that the shape of the diaphragm and the weight of the mediastinum would become better aligned when prone; consequently, lung regions poorly ventilated in the supine position would become functional. Bryan’s physiological reasoning has been repeatedly proven sound. In short, prone positioning consistently improves oxygenation.

If oxygenation is reliably improved, how could the study by Curley et al⁶ produce a negative result? Discrepancies between applying physiological principles and failure to improve clinical outcomes are not new; in certain instances, the application of simplistic but incomplete physiologic reasoning has been disastrous.^{8 - 9} In one study,⁹ unselective suppression of cardiac arrhythmias caused drug-induced cardiac deaths. In another study,⁸ simple correction of blood pressure in septic shock, without addressing the complexity of sepsis (or shock), increased mortality. The current study provides an opportunity to reassess the balance of physiologic expectation against achievable outcome.

Why would a treatment that improves oxygenation not help patients? Very low oxygen levels are lethal. In children with ARDS, the extent of impaired oxygenation predicts outcome: the worse the oxygenation, the greater the mortality. The primary supportive therapy in ARDS, mechanical ventilation, augments oxygenation as its primary goal. Mechanical ventilation, however, can also cause harm. The relationship between oxygenation and ARDS is complex—the use of higher tidal volumes in ventilated adults was associated both with better oxygenation and higher mortality in 1 trial.²

Although oxygenation is a central issue, hypoxia is not responsible for most of the deaths and simply improving oxygenation is not the answer. Therefore, improved oxygenation should not be a goal in itself but should be considered in the context of harm from the means used to attain it (eg, toxic supplemental oxygen, injurious ventilator pressures).

Focusing on oxygenation in the prone position may have been too simplistic. The most insightful result from the Italian prone-positioning study¹⁰ had little to do with oxygenation per se; instead, mortality was lower in adult patients who cleared carbon dioxide more efficiently when prone.¹¹ The ideal response to prone positioning may be a reduction in dead space reflecting an increased volume of perfused lung, not just any region of the lung. Although making perfect physiological sense, it is a recent observation, and one that has not been incorporated into the definition of a responder in the current study⁶ or any previous trial of prone positioning. Thus, defining a responder in terms of oxygenation response may simply “sell short” the physiology, and ultimately, the patients.

The study by Curley et al has several strengths. It was a multicenter trial, the intervention was comprehensive, and management was protocol-driven, although neither the protocols nor the adherence to them are described. Nevertheless, there were important differences between the groups at baseline; more patients in the prone group were already receiving high-frequency oscillatory ventilation, implying that they were sicker, or at least that the clinicians caring for them thought so.

Sample size is a major issue in any negative study. The authors provide a great deal of detail in explaining the statistical basis for their projections and the reasons for stopping the study on the grounds of futility. However, in the current study, the mortality was 8%; in such a population, more than 8000 patients (compared with 100 patients in the current study⁶ ) would be required to demonstrate a relative reduction in mortality of 20%. Investigators choosing to perform future randomized clinical trials will have to design larger studies that involve multiple centers, particularly when treatment effects are likely to be small as in pediatric critical care. Larger trials are now possible through consortia, such as the PALISI group or the Canadian Critical Care Trials Group.

Was the patient population in the study by Curley et al broadly representative? Probably not. Willson et al⁵ (the same consortium of investigators, but different study sites) reported mortality rates in the immunocompetent subgroups of 20% compared with 8% in the current study.⁶ There are several possible explanations, such as the care in the study by Curley et al was exemplary (8% control group mortality), the care in the previous study was deficient (20% control group mortality), or the present study population does not represent the true spectrum of ARDS in children. The latter likelihood is supported by 2 recent case series reporting mortality rates of between 22%¹² and 27%¹³ for unselected patients with pediatric ARDS. However, analysis of the previous adult study suggested that mortality was reduced only in the most critically ill patients.¹⁰ By excluding the sickest patients, the current investigators may have removed the ability for prone positioning to make its best impact. In addition, as long as mortality rates observed in the current study (8%) are anticipated, the incremental advantage of any conceivable future additional therapy (eg, surfactant) will be marginal at best, and would require the treatment of a very large number of children to save one life.

Although life-saving mechanical ventilation can clearly cause harm, children may be more resilient to this harm than adults. Laboratory data support such a contention,^{14 - 15} and a recent case series of children with ARDS¹² suggests that, within observed limits, increments in tidal volume might not impact outcome. Although ventilator settings must be monitored carefully in all patients, mechanical ventilation simply may be less injurious to children. If so, it may be far more difficult to demonstrate the impact of ventilation in children than in adults.

The goal of the study by Curley et al was to demonstrate improvement in a surrogate end point (ventilator-free days before day 28). The advantages of surrogate outcomes are that significant effects may be identified with smaller numbers of patients; this improves trial feasibility and reduces trial-associated patient deaths. Surrogate outcomes can be misleading, even though the one used in the current study is used in some ventilation trials. Consider, however, the logic of the patient whose death after 27 days of ventilation is statistically identical to a patient who is ventilated for 29 days, but goes on to make an excellent recovery. This paradox is accentuated by the earlier trial of pediatric ARDS⁵ in which mortality was decreased but no benefit in ventilator-free days was demonstrated; in that study, the surrogate outcome was less helpful than the mortality data. This incongruity clouds interpretation of these 2 studies^{5 - 6} and does little to reassure clinicians that this surrogate outcome is indeed meaningful in this population of children with ARDS.

So how should these findings be incorporated into practice? The study by Curley et al suggests that there are no major advantages to prone positioning in pediatric ARDS, if the anticipated mortality is in the range of 8%, and there also appears to be no associated harm. However, the study does not mean that prone positioning should not be used or that physicians do not have to teach or learn the underlying physiological principles. For instance, prone positioning should be tried in any child who has profound hypoxemia due to ARDS, provided reversible causes have been ruled out. Where prone positioning fits in the context of those patients with a higher anticipated mortality remains unclear; this is not addressed in the current study.⁶ Moreover, the incremental increase in current knowledge of physiology and trial design means that subsequent studies will build on reports such as this one,⁶ and use progressively more physiologically sound hypotheses and robust trial methodology.

Clinicians caring for critically ill children could interpret the current study to mean that the prone position should not be used in any child with ARDS. But the patients in this study population are very different from the usual children with ARDS. The clinician will not know how to apply this information for the more typical pediatric patient with ARDS. Such lack of certainty should not be disturbing, but rather should keep critical care physicians alert for better practice and searching for better understanding. Astute clinicians will continue to draw from their knowledge of physiology and from reports of clinical trials; in doing so, they will be guided by both and blinded by neither.