Effect of prone position on respiratory parameters, intubation and death rate in COVID-19 patients: systematic review and meta-analysis

Prone position (PP) is known to improve oxygenation and reduce mortality in COVID-19 patients. This systematic review and meta-analysis aimed to determine the effects of PP on respiratory parameters and outcomes. PubMed, EMBASE, ProQuest, SCOPUS, Web of Sciences, Cochrane library, and Google Scholar were searched up to 1st January 2021. Twenty-eight studies were included. The Cochran's Q-test and I2 statistic were assessed heterogeneity, the random-effects model was estimated the pooled mean difference (PMD), and a meta-regression method has utilized the factors affecting heterogeneity between studies. PMD with 95% confidence interval (CI) of PaO2/FIO2 Ratio in before–after design, quasi-experimental design and in overall was 55.74, 56.38, and 56.20 mmHg. These values for Spo2 (Sao2) were 3.38, 17.03, and 7.58. PP in COVID-19 patients lead to significantly decrease of the Paco2 (PMD: − 8.69; 95% CI − 14.69 to − 2.69 mmHg) but significantly increase the PaO2 (PMD: 37.74; 95% CI 7.16–68.33 mmHg). PP has no significant effect on the respiratory rate. Based on meta-regression, the study design has a significant effect on the heterogeneity of Spo2 (Sao2) (Coefficient: 12.80; p < 0.001). No significant associations were observed for other respiratory parameters with sample size and study design. The pooled estimate for death rate and intubation rates were 19.03 (8.19–32.61) and 30.68 (21.39–40.75). The prone positioning was associated with improved oxygenation parameters and reduced mortality and intubation rate in COVID-19 related respiratory failure.


Materials and methods
In accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for designing and implementing systematic review studies, the following steps were taken: a systematic literature search, organization of documents for the review, abstracting and quality assessment of each study, synthesizing data, and writing the report 21 . The protocol of the study was registered in the International Prospective Register Of Systematic Reviews (PROSPERO) at the National Institute For Health Research. Registration number in PROSPERO is CRD42021257619.
Search strategy. According to the PICO framework, the systematic literature search was conducted on PubMed, EMBASE, ProQuest, SCOPUS, Web of Sciences, Cochrane library, and Google Scholar databases. MeSH Keywords were connected with AND, OR and NOT prone position and respiratory parameters, and their suggested entry terms were the main keywords in the search strategy.  www.nature.com/scientificreports/ five categories: qualitative, randomized controlled trial, nonrandomized, quantitative descriptive, and mixed methods studies 22 . This tool consists of five items for each category, each of which could be marked as Yes, No, or cannot tell. Based on the scoring system, score one is assigned to Yes and score 0 to all other answers. In other words, the total score would be the percentage of affirmative responses. To evaluate the final scores qualitatively, the scores above half (more than 50%) were considered high quality.
Data extraction. Data  , in the before-after design, we calculated the change score MD (mean after prone position minus mean before prone position), and in Quasi-Experimental design, we calculated MD (mean in supine position minus mean in prone position). Then Standard deviation in Before-After design and Quasi-Experimental design was calculated based on formulas (1) and (2): where SD before , SD after , and Corr is the standard deviation in before prone position, standard deviation after prone position, and correlation coefficient between before and after where SD prone position , SD supine position , n 1 , and n 2 is the standard deviation in prone position group, the standard deviation in supine position group, the sample size in the prone position and supine position groups. Then pooled MD (PMD) was calculated by the "Metan" command 23,24 . Heterogeneity was determined using Cochran's Q test of heterogeneity, and the I 2 index was used to quantify heterogeneity. In accordance with the Higgins classification approach, I 2 values above 0.7 were considered as high heterogeneity. To estimate the PMD for respiratory parameters and subgroup analysis (study design and ventilation), the fixed-effect model was used, and when the heterogeneity was greater than 0.7, the random-effects model was used. The meta-regression analysis was used to examine the effect of study design, sample size, BMI, age and prone position (PP) duration as factors affecting heterogeneity among studies. The "Meta bias" command was used to check for publication bias, and if there was any publication bias, the PMD was adjusted with the "Metatrim" command using the trim-and-fill method. In all analyses, a significance level of 0.05 was considered.

Result
Overall, 1970 studies were found through databases. After excluding redundant papers, 855 studies remained. After reading abstracts, 775 studies were excluded from the list. Then, the full text of the remaining 80 studies was reviewed, and 52 studies were excluded. Finally, 28 studies included in qualitative analysis and 26 studies with a total sample size of 1272 participants were included in the quantitative analysis. The flowchart of this selection process is shown in Fig. 1. Studies were published during 2020-2021, most studies were done in the UK, China, and Spain with three studies and range participants age were 17-83 years old (Tables 1 and 2). Supplementary 2 shows risk of bias assessment for included studies. All studies were high quality (more than 50% scores).
Pooled mean difference of respiratory parameters in total and based on subgroups. Figure 2 showed the forest plot for MD of PaO 2 /FIO 2 Ratio in included studies.  Figure 3 and Table 3 showed the PMD of other respiratory parameters in included studies. The PMD of SPO 2 (Sao 2 ) in the study with before-after design, quasi-experimental design, and in total was 3.38 (95% CI 1.68-5.09), 17.03 (95% CI 12.19-21.88), and 7.58 (95% CI 4.93-10.23); respectively. This means that the prone position in COVID-19 patients leads to significant improvement corresponding to Spo 2 (Sao 2 ). Also the PMD of Paco 2 in COVID-19 patients was significantly decreased in quasi-experimental design (PMD: − 18.49; 95% CI − 34.50 to − 2.47 mmHg) and in total (PMD: − 8.69; 95% CI − 14.69 to − 2.69 mmHg). No significant change was observed for PMD of PaCo 2 in the before-after design. The PMD of other respiratory parameters showed in Table 3 and  4). Figure 5 showed PMD of respiratory parameters based on ventilation status. PMD of Spo 2 (Sao 2 ) in Intubation and Non-intubation subgroup was 10.56 (95% CI − 18.15 to 39.26) and 8.57 (95% CI 3.47-13.67); respectively. This means that the prone position in COVID-19 patients with non-intubation leads to significant improvement corresponding to Spo 2 (Sao 2 ) but Intubation have no effects on Spo 2 (Sao 2 ) improvement. Also PMD of PaO 2 /FIO 2 in Intubation and non-intubation subgroup was 65.03 (95% CI 6.06-123.99) and 49.56 (95% CI 26.56-72.56); respectively. This means that the prone position in COVID-19 patients leads to significant improvement of PaO 2 /FIO 2 Ratio, but this value for Intubated patients was higher than non-intubated groups. Situation of other parameter was showed in Fig. 5 Heterogeneity and meta-regression results. According to Cochran's Q test of heterogeneity, there was significant heterogeneity among studies (p < 0.001). Except for PaCo 2 in the before-after design, the heterogeneity amount was more than 85% based on the I 2 index, which indicates high heterogeneity. Table 4 presents the results of the univariate meta-regression; there are significant associations between study, results with study design corresponding to SPO 2 (Sao 2 ) percent (Coefficient: 12.80; p < 0.001). No significant associations were observed for other respiratory parameters with sample size, study design, BMI, age and PP duration (Table 4).

Discussion
This systematic review analyzed the effects of prone position on respiratory parameters, intubation, and death rate. We found that prone position initiation leads to improved oxygenation parameters (PaO 2 /FiO 2 ratio, SpO 2 , PaO 2 , and PaCO 2 ) in patients with mild to severe respiratory failure due to confirmed COVID-19. However, the prone position did not change the respiratory rate in patients with hypoxemic respiratory failure suffering from COVID-19.
Most of the studies (18/28 studies) demonstrated significant improvement in PaO 2 /FiO 2 ratio after prone positioning. Moreover, the improvement of SpO 2 (SaO 2 ) and PaO 2 has been shown in 15 and 7 studies, respectively. Although the effect of prone position after resupination has declined in five studies 1,5,8,16,29 , early prone positioning should be considered as first-line therapy in ARDS patients 43 . Initiation of prone position in ARDS patients by reducing shunt, and V/Q mismatch, brings about an increase in the recruitment of non-aerated areas of the lungs, secretion clearance, improvement work of breathing (WOB) and oxygenation, and reduction of mortality compared with the supine position [44][45][46] . Prone position by enhancement in PaO 2 /FiO 2 ratio not only leads to a decrease in the classification of respiratory failure but also prevents further complications due to ARDS, such as multi-organ failure (MOF), which is the most common cause of mortality in this devastating condition 47 .
The efficacy of prone positioning may be affected by various protocols, such as different settings (ICU or emergency department), the timing of initiation (early or late), duration (prolonged or short sessions), positioning (prone position with or without lateral position), respiratory support in intubated or non-intubated patients (mechanical ventilation, NIV, nasal cannula, helmet, face mask) and the severity of ARDS 48 . Even though in this study PaO 2 /FiO 2 ratio was significantly higher in the prone-positioning group with mild to severe ARDS, a further meta-analysis need to assess the impact of prone position in a different classification of ARDS with mild condition. In this systematic review, the prone position time varied from less than 1 to 16 h in a day. In eight studies, the prone positioning has been implemented for about 16 h a day. The prolonged prone positioning (no less than 10-12 h and ideally for 16-20 h) leads to improved oxygenation and a significant reduction in mortality in patients with severe ARDS. On the other hand, reducing the number of turning in patients with critical conditions can decrease the risk of more complications 48 . Although PaCO 2 did not demonstrate a difference in five studies 5,16,25,29,40 , the PMD of PaCO 2 in COVID-19 patients significantly decreased totally. The prone position by increasing the dorsal recruitment, PaCO2 clearance, and decreasing the dead space can also lead to better ventilation. Moreover, a higher PaCO 2 clearance due to the prone position is related to a significant decrease in       54 . In terms of respiratory rate, in few studies, the respiratory rate reduction was significant, but we found that respiratory rate did not change during the prone positioning in the overall analysis. Our systematic review and meta-analysis demonstrated that prone positioning leads to a lower mortality rate in confirmed COVID-19 patients. Although in this systematic review and meta-analysis, many studies have assessed the impact of prone position on the short term (28 days) mortality, where they benefit from prone positioning protocols, the effect of prone positioning in the long-term (3 months or more) mortality is unclear. Therefore, further studies will be needed to demonstrate the relationship between prone positioning in COVID-19 patients and long-term mortality. Furthermore, this study confirmed that the improvement of oxygenation parameters due to the prone position might be associated with a lower intubation rate in COVID-19 patients.

Conclusions
In our systematic review of 28 studies, prone positioning has been compared with supine positioning in hypoxic adult patients with COVID-19. We found prone position by optimizing lung recruitment, and the V/Q mismatch can improve oxygenation parameters such as PaO 2 /FIO 2 Ratio, Spo 2 (Sao 2 ), PaO 2 , PaCO 2 . Nevertheless, the prone position did not change their respiratory rate. Moreover, the initiation of prone position might be associated with a lower mortality and intubation rate. Since most patients demonstrated improved oxygenation and lower mortality and intubation rate, we recommend the prone position in patients COVID-19.Similar to other studies, our research had some limitations. (1) Some studies did not report values of the respiratory parameters in different groups and just reported significantly parameter (like that p-value); which we have to exclude this studies from the quantitative analysis that this limitation was not be resolved even by data requesting from corresponding authors. We would like to perform the gender-specific estimation, but it was not possible due to insufficient data in the primary studies; (2) also we tend to estimate the pooled MD in different geographical regions or country-specific estimation based on available methods 50 , since the infrequent studies number, this estimation will not be robust. Table 3. Result of meta-analysis for calculation of pooled mean difference of respiratory parameters; publication bias and fill and trim method. CI confidence interval, N number of study, PMD pooled mean difference, Pao 2 partial pressure of oxygen, FIO 2 fractional inspiratory oxygen, Sao 2 oxygen saturation (arterial blood), RR respiratory rate.    Figure 6. Association between sample size with mean difference (MD) of PaO 2 /FIO 2 Ratio (mmHg) (A) and Spo 2 (Sao 2 ) (B) using meta-regression. Size of the circles indicates sample magnitude. There was no significant association between sample size with MD of PaO 2 /FIO 2 Ratio and Spo 2 (Sao 2 ).