Use PROSEVA study criteria not COVID-19 phenotype to guide proning treatment decisions

COVID-19 pneumonia is a new disease requiring urgent investigation to establish best practice. We fully support the development of novel hypotheses to test and guide the most appropriate and personalised management of critically ill patients with this condition. Most patients with critical respiratory illness due to COVID-19 pneumonia meet the Berlin Definition for Acute Respiratory Distress Syndrome, a straightforward set of characteristics with no links to the underlying biology. ¹ In several editorials and on the basis of a handful of patients, Gattinoni et al.^2–5 have proposed sub phenotypes of critical respiratory illness in COVID-19 pneumonia. The L-type features bilateral radiological opacifications, normal lung compliance (sometimes defined as greater than 40 ml/cmH₂O), low response to recruitment, and low response to prone positioning. H-type is characterised by low compliance and high response to recruitment and prone positioning. While there is more than one proposed discriminant separating these two phenotypes the most clinically accessible is compliance. Dynamic compliance is defined as the change in lung volume per unit pressure change in the presence of flow and, unlike static compliance, can be easily calculated from most ICU ventilators without interruptions to the respiratory cycle.

We queried the proposed lack of benefit from proning from more compliant lungs. At our institution all patients meeting criteria laid out in the PROSEVA study of prone positioning in ARDS undergo a trial of prone positioning. ⁶ We analysed the cohort of 47 patients invasively ventilated with COVID-19 pneumonia admitted to two general intensive care units in the city of Edinburgh, UK between 11 March and 19 April 2020. As a service evaluation of routinely available information, ethical approval was not required.

Among the 47 patients, the median age was 58 (range 34 to 77), median APACHE II was 15 (range 6 to 34), with females accounting for 15% of patients. The median nadir PF ratio in the first 24 h was 12.7 kPa (range 6.3 to 26.1 kPa). Eighteen patients (40%) met PROSEVA criteria for a trial of prone positioning and were proned. This cohort had median age 60 (range 43 to 76), median APACHE II 15 (range 6 to 30), median time from intubation to first proning of 14 h (range 2 to 128), with females accounting for 21%. Immediately prior to the first proning episode patients had median PF ratio of 13.3 kPa (range 8.5 to 25 kPa) and median dynamic compliance of 29 mL/cmH₂O (range 18 to 49 mL/cmH₂O). Two hours post proning the median PF ratio was 22.4 kPa (range 10.3 to 48.8 kPa), an increase of 9.7 kPa, and the median dynamic compliance was 29 mL/H₂O (range 19 to 49 mL/cmH₂O), a median change in compliance of <1 mL/cmH₂O. Figure 1 shows a positive relationship (r²= 0.14) between respiratory system dynamic compliance and oxygenation response to proning.

Figure 1.

The relationship between dynamic compliance before prone ventilation and change in PF ratio. All patients ventilated with volume control ventilation (6 mL/kg ideal body weight). Dynamic compliance calculated 1 h before prone ventilation. Change in PF ratio was calculated as PF ratio 2 h post prone ventilation minus PF ratio 1 h pre prone ventilation, n = 18. Best fit calculated using linear regression in SPSS version 22: y = −5 + 0.5x, r²= 0.13. The median increase in compliance after proning was 1 mL/cmH₂O (range −20 to +13).

The positive relationship between pre-proning dynamic lung compliance and change in PF ratio implies those with higher compliance will have a better improvement in oxygenation after proning. We cautiously conclude that the proposed separation into sub-phenotypes based on respiratory system compliance is not supported, especially with regard to effect on clinical practice. We acknowledge that PaO₂ response is a surrogate for clinical benefit from proning, and we also acknowledge that only one of our proned patients had compliance >40 cmH₂O putting our inferences about L-type at risk. However, we speculate that a high compliance severely hypoxic subpopulation may not exist, in keeping with other recent findings. ⁷ We also note that a recent case series in critical COVID-19 pneumonia that demonstrated no relationship between estimated lung weight and compliance, ⁸ a relationship central to the proposed L and H-type phenotypes. We suggest that until such time that data driven phenotypes with differing response to treatment can be established and prospectively confirmed clinicians should use the existing evidence base to determine the appropriateness for a trial of prone positioning, which does not include an assessment of compliance.