High flow nasal cannula versus conventional oxygen therapy in postoperative patients at high risk for pulmonary complications: A systematic review and meta-analysis

Gui-Ling Xiang,Qin-Han Wu,Liang Xie,Jie-Qiong Song,Xu Wu,Sheng-Yu Hao,Ming Zhong,Shan-Qun Li
DOI: https://doi.org/10.1111/ijcp.13828
IF: 2.6
2021-01-01
International Journal of Clinical Practice
Abstract:Background The effects of high flow nasal cannula (HFNC) on postoperative patients at high risk for pulmonary complications(PC) are controversial. We aimed to further determine the effectiveness of HFNC in postoperative patients at high risk for PC by comparison to conventional oxygen therapy (COT). Methods We performed a comprehensive search that compared HFNC with COT in postoperative patients at high risk for PC. The main outcomes were length of hospital stay (hospital LOS) and respiratory complications. Results Six trials with a total of 733 patients were pooled in our final studies. Except for Hospital LOS (I-2 = 53%, chi(2) = 8.51, P = .07) and rate of intubation or non-invasive ventilation (NIV) for respiratory failure (RF) (I-2 = 49%, chi(2) = 1.97, P = .16) between HFNC and COT, no significant heterogeneity was found in outcome measures. Compared with COT, HFNC was associated with a lower rate of intubation or NIV for RF (RR 0.23, 95% CI 0.08-0.66, P = .006) and rate of hypercapnia (RR 0.37, 95% CI 0.20-0.68, P = .002). As for the Hospital LOS, ICU LOS, rate of requirement of O2 after discontinuous and hypoxemia, HFNC did not show any advantage over COT. Trial Sequential Analysis (TSA) for Hospital LOS showed that monitoring boundaries were finally not surpassed and required information size (RIS) was not met. Conclusions The available randomised controlled trials (RCTs) suggest that, among the postoperative patients at high risk for PC, HFNC therapy compared with the COT significantly reduces rate of incubation or NIV for RF and rate of hypercapnia, meanwhile is safely administered. Further large-scale, multicenter, randomised and controlled studies are needed to confirm our results.
What problem does this paper attempt to address?