Predictive Value of the PaO2/FIO2 Ratio for Mortality in Patients with Acute Respiratory Distress Syndrome: A Systematic Review and Meta-Analysis

PaO 2 /FIO2 (P/F) ratio has been used to define the severity of acute respiratory distress 42 syndrome (ARDS) despite the controversy of its clinical utility. This systematic review 43 and meta-analysis (SRMA) aimed to obtain summary estimates of predictive 44 performance of the P/F ratio for predicting mortality in ARDS patients.


Competing interests:
The authors declare that they have no competing interests.  Methods 46 We included a study wherein the study population comprised ARDS patients in any

53
Twenty-eight trials and 38270 patients were included in the meta-analysis. Most of the 54 study settings were in the intensive care units. The overall risk of bias was high. The 55 pooled sensitivity of the P/F ratio in all included studies for a P/F ratio of 100 was 43.6% 56 (95% CI, 36.9-50.5%) and the specificity was 71.1% (95% CI, 66.7-75.1%) and those for 57 a P/F ratio of 200 were 83.2% (95% CI, 78.2-87.2%) and 26.2% (95% CI, 21.2-31.9%).

58
Interpretation 59 The P/F ratio had high sensitivity and moderate specificity at a P/F ratio of 200 and 100 60 respectively., which supports the use of the P/F ratio for screening ARDS patients who 61 are at risk of deterioration. Acute respiratory distress syndrome (ARDS) is a condition of acute lung injury related 76 to inflammation and is characterized by high pulmonary vascular permeability and a 77 high amount of extrapulmonary water [1]. To evaluate ARDS severity, the PaO2/FIO2 78 (P/F) ratio is commonly used in the clinical setting according to the Berlin definition, 79 reported in 2012, and the P/F ratio is used to define ARDS severity [2]. Subsequently, 80 the association between the severity of Berlin definition and prognosis has been 81 evaluated in several studies [3][4][5] Although the clinical utility of the P/F ratio remains controversial in literature, its 91 predictive accuracy for mortality has not been systematically reviewed to the best of our 92 knowledge. Determining the integrated prognostic accuracy between the P/F ratio and 93 prognosis in ARDS patients may be useful in stratifying patients and allocating 94 appropriate medical resources in emergency medicine and intensive care settings.

95
The primary objective of the current study was to obtain summary estimates of predictive 96 performance, including sensitivity and specificity, across studies of P/F ratio for the 97 prediction of any type of mortality in patients diagnosed with ARDS. Study eligibility criteria 110 We included a study wherein the study population comprised ARDS patients in any 111 setting, including emergency departments, general hospital wards, and intensive care 112 units (ICUs). The index test was the P/F ratio or oxygenation index. We included studies 113 that involved adult patients (aged ≥18 years) and evaluated mortality. The reference 114 standard for this study was the mortality rates reported in each study. We included all 115 English-language abstracts and full-text articles describing retrospective and prospective 116 observational studies as well as randomized and quasi-randomized controlled trials. We 117 excluded case reports, case series, animal studies, and pediatric studies. We included 118 only the study with more patients for studies that used the same database.  Data extraction and quality assessment 142 We used a pre-defined data collection form for study characteristics and outcome data, characteristics from the included studies. We extracted information on the following study 146 characteristics: author information, year of publication, study design, eligibility criteria, 147 number of patients included, mean or median age, threshold used for patient stratification 148 by the P/F ratio, and mortality. 149 We selected the shortest outcome if the study reported several outcomes. We also    Table   184 1. The P/F ratio was evaluated on the day of ARDS diagnosis in 24 studies. The other  The pooled sensitivity of the P/F ratio across all included studies for a P/F ratio of 100 211 was 43.6% (95% CI, 36.9-50.5%) and the specificity was 71.1% (95% CI, 66.7-75.1%).

212
The pooled sensitivity of the P/F ratio across all included studies for a P/F ratio of 200  Sensitivity analysis 219 We conducted a sensitivity analysis by publication year for the studies published after the 220 Berlin definition was established. In this analysis, the sensitivity for a P/F ratio of 100 221 was 43.1% (95% CI, 36.0-50.5%) and the specificity was 71.7% (95% CI, 66.8-76.1%).

222
The sensitivity for a P/F ratio of 200 was 82.0% (95% CI, 76.5-86.4%) and the specificity 223 was 27.9% (95% CI, 22.1-34.5%). We also conducted subgroup analysis for mortality as 224 the short-term outcome: ICU, in-hospital, 28-day, and 30-day mortality. In this analysis, 225 the sensitivity for a P/F ratio of 100 was 40.4% (95% CI: 33.1-48.2%) and the specificity 226 was 72.6% (95% CI: 67.1-77.5%). The sensitivity for a P/F ratio of 200 was 82.0% (95% 227 CI: 75.9-86.8%) and the specificity was 28.9% (95% CI: 23.0-35.7%). In addition, the 228 sensitivity analysis was performed for mortality as the short-term outcome: ICU, in-  Key observation 239 We conducted a SRMA to evaluate the prognostic value of the P/F ratio for predicting 240 mortality in adult patients with ARDS. The risk of bias in the included studies was high.

241
With a P/F ratio of 100 as the cutoff, the sensitivity for predicting mortality was low and 242 the specificity was moderate. Of note, with a P/F ratio of 200 as the cutoff, the sensitivity 243 for predicting mortality was high, while the specificity was low. Cochrane databases did not reveal any existing systematic reviews and meta-analyses on 249 the prognostic accuracy of the P/F ratio. To the best of our knowledge, this is the first 250 SRMA of the prognostic value of the P/F ratio in ARDS patients. Another strength of our 251 study is that we were able to collect and analyze a comprehensive set of studies in which 252 the P/F ratio and outcome could be evaluated, and we were able to show the integrated 253 sensitivity and specificity of the P/F ratio rather than the single diagnostic performance 254 of the P/F ratio in individual studies. As for the stratification of ARDS patients by the P/F 255 ratio, which is also used in the Berlin definition, the sensitivity of a P/F ratio of 200 is 256 relatively high, which may be useful for screening ARDS patients at high risk of death.

257
The number of ICU beds, ventilators, and medical staff is limited, and appropriate 258 allocation of medical resources may become possible based on the P/F ratio. In addition, 259 the specificity of a P/F ratio of 100 was low, and prognosis should not be considered based 260 on the P/F ratio alone.

262
Future direction area of study 263 The varying background characteristics and heterogeneity of ARDS patients in this study 264 may have influenced the results; therefore to evaluate heterogeneity of these backgrounds 265 for the results will be mandated in the future study. In the present study, a P/F ratio of 100 266 as the cutoff for death was neither sensitive nor highly specific, which was the same as 267 that in the subgroup analysis that focused on mortality as the short-term outcome and in   In conclusion, our SRMA found that the P/F ratio had poor sensitivity and moderate 292 specificity at a P/F ratio of 100, whereas at a P/F ratio of 200, it was sensitive but poorly 293 specific for mortality. Our findings support the continued use of a P/F ratio of 200 for 294 screening patients with ARDS who are at risk of deterioration.  Competing interests 316 The authors declare that they have no competing interests.               Forest plots showing the sensitivity and speci city of the P/F ratio (cutoff 100) in the 20 included studies.

Figure 5
Summary of the receiver-operating characteristic (SROC) curve with summary point estimates of sensitivity and speci city with 95% con dence intervals (CIs) for a P/F ratio of 100.

Figure 6
Summary of the receiver-operating characteristic (SROC) curve with summary point estimates of sensitivity and speci city with 95% con dence intervals (CIs) for a P/F ratio of 200.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.