Dear Editor,

We read the recent article written by Rochwerg et al. in Intensive Care Medicine and greatly appreciate their efforts to assess the effect of high flow nasal cannula (HFNC) for acute hypoxemic respiratory failure in a systematic review and meta-analysis [1]. Nevertheless, some issues should be further discussed.

First, it is obvious that I2 = 0 in Fig. 3, why not apply the fixed effect model in meta-analysis and trial sequential analysis (TSA) [2, 3]? Second, they only performed TSA on invasive mechanical ventilation outcome and mortality outcome. But another major positive outcome of escalation of therapy was not evaluated by TSA. Third, \({\text{The}}\,{\text{relative}}\,{\text{risk}}\,{\text{reduction}}\) = \(\frac{{{\text{Incidence}}\,{\text{in}}\,{\text{control}}\,{\text{arm}}{-}{\text{incidence}}\,{\text{in}}\,{\text{intervention}}\,{\text{arm}}}}{{{\text{Incidence}}\,{\text{in}}\,{\text{control}}\,{\text{arm}}}}\) [4]. Therefore, the relative risk reduction of invasive mechanical ventilation outcome should be 16.58% \(\left( {\frac{{\frac{235}{805} - \frac{205}{842}}}{{\frac{235}{805}}}} \right)\), not 15%. While, the relative risk reduction of mortality outcome should be 4.60% \(\left( {\frac{{\frac{186}{685} - \frac{187}{722}}}{{\frac{186}{685}}}} \right)\), not 15%, too. Therefore, they cannot apply the same relative risk reduction (15%) to TSA of different outcomes. In addition, boundary required information size (RIS) on mortality outcome is ignored by Trial Sequential Analysis v.0.9.5.10 Beta due to little information used (0.92%). Therefore, TSA of mortality outcome cannot be drawn on principle.