Prolonged Noninvasive Respiratory Support Before Intubation Leads to Increased Duration of ECMO: More Evidence That Less Is More.

To the Editor: We read with enthusiasm the work by Giani et al., which addressed a work we previously published in the ASAIO journal.1,2 Though their findings were mixed, given the small sample size, we believe their letter supports the conclusion that prolonged respiratory support (RS) in patients with acute respiratory distress syndrome (ARDS) before extracorporeal membrane oxygenation (ECMO) can be injurious to the lungs and lead to worse outcomes. The prominent finding reported by Giani et al. was that prolonged noninvasive RS before intubation was associated with prolonged ECMO support. Duration of ECMO is a known predictor of ECMO outcomes; therefore, it can be reasonably inferred that this long time would result in worse overall outcomes.3 The adequate sample size is a challenging issue in ECMO research. Based on optimistic estimates derived from our original work (median ECMO duration of 12 days [RS <3 days], median follow-up of 24 days, and a hazard of ECMO decannulation of 2.0 (RS < 3 days vs. RS ≥ 3 days), the minimum sample size to provide 80% power to detect a difference in the rate of ECMO decannulation would be approximately N = 83.4 The sample described by Giani et al. is significantly smaller than this minimum, and therefore, the results reflect a high risk for type II statistical error. Relatedly, the cumulative incidence of successful ECMO liberation (Figure 1B from the study by Giani et al.) reported in the recent manuscript closely resembles our own, with near identical separation of the survival curves over the first 3 weeks of ECMO support. The authors did not include the number of patients at risk during this analysis. If few remained on ECMO towards the “tails” of this plot, the survival estimates for patients on ECMO the longest are likely imprecise. The high utilization of prone ventilation during venovenous (VV)-ECMO reported by Giani et al. is admirable. While not yet universally adopted, prone positioning during VV-ECMO may serve to hasten lung recovery. Hypothetically, there is little reason to suspect that the advantage of prone positioning for ARDS would differ based on if the patient is on ECMO or not. There were also differences in the type of noninvasive RS applied. The authors utilized helmet continuous positive airway pressure (CPAP), whereas we utilized facemask noninvasive ventilation and high flow nasal cannula. We know from prior studies that the major driver for patient self-induced lung injury (P-SILI) is increased lung volumes and transpulmonary pressure shifts.5 Moreover, PEEP can make the tidal volumes more homogenous and has been postulated to prevent P-SILI. Taken together, these differences could have influenced the onset and manifestation of P-SILI. To conclude, we believe that the study by Giani supports our hypothesis that prolonged RS before intubation may lead to worse outcomes in patients with COVID-19 supported with ECMO. Efforts should be made to minimize the duration of RS and consider early intubation in eligible patients. To mitigate the effects of prolonged RS, adherence to lung protective ventilation and prone ventilation, even while on ECMO, may help. We appreciate the efforts of Giani and await a larger meta-analysis or prospective trial to shed light on this topic.