Open Access 01-12-2020 | Letter
Impact of HFOV in pARDS outcomes: questions remain
Published in: Critical Care | Issue 1/2020
Login to get accessExcerpt
We read with interest the analysis comparing outcomes in paediatric acute respiratory distress syndrome (pARDS) among those who required high-frequency oscillatory ventilation (HFOV) as ‘rescue therapy’, compared to those who did not [1]. The authors must be congratulated for using multiple matching techniques to account for some confounders. We also commend them for making the code used for analysis public [2]. We wish to clarify certain aspects of the study that may aid readers understand the context of the analysis better.
1.
The dataset used for genetic matching included at least 68 ‘matched-pairs’ of severe pARDS. However, the 68 ‘matches’ were derived from a pool of less than 39 ‘real’ patients with severe pARDS in the non-HFOV cohort. While multiple matching with replacement may improve matching, it would be helpful to understand how many multiple matches were made from the smaller pool of non-HFOV patients and whether statistical adjustments were made for the duplicates in addition to a pairwise comparison.
2.
The authors reported that the sensitivity analysis showed consistent findings with the primary analysis. Curiously, the odds ratios of mortality for mild, moderate and severe pARDS in the logistic regression analysis were 1.61, 1.02, and 1.00 respectively. This paradoxical trend may indicate an interaction between the use of HFOV and presence of moderate (or) severe pARDS, rather than the risk of HFOV per se. Was an interaction between the use of HFOV and severity of pARDS tested and accounted for in the logistic regression analysis?
3.
The intention to match patients in a propensity- or genetic-matched study is to account for confounders present, when an intervention was indicated, between those that received the intervention compared to those that did not. The use of day 1 OI, rather than the OI immediately prior to the initiation of HFOV, may have systematically underestimated the true severity of pARDS in the HFOV group, given that HFOV was used as a ‘rescue mode’ anytime within 7 days. An equivalent assignment of pARDS severity could be done using the ‘highest OI’ within the first 7 days in the non-HFOV group. Was the use of OI prior to initiation of HFOV, rather than day 1 OI, considered for analysis?