Respiratory Outcomes in Patients Following COVID-19-Related Hospitalization: A Meta-Analysis

Background: To determine the respiratory outcomes in patients following COVID-19-related hospitalization. Methods: Systematic review and meta-analysis of the literature. Results: Forced vital capacity (FVC, % of predicted): 0–3 months post discharge: 96.1, 95% CI [82.1–110.0]; 3–6 months post discharge: 99.9, 95% CI [84.8, 115.0]; >6 months post discharge: 97.4, 95% CI [76.8–118.0]. Diffusing capacity of the lungs for carbon monoxide (DLCO, % of predicted): 0–3 months post discharge: 83.9, 95% CI [68.9–98.9]; 3–6 months post discharge: 91.2, 95% CI [74.8–107.7]; >6 months post discharge: 97.3, 95% CI [76.7–117.9]. Percentage of patients with FVC less than 80% of predicted: 0–3 months post discharge: 10%, 95% CI [6–14%]; 3–6 months post discharge: 10%, 95% CI [2–18%]; >6 months post discharge: 13%, 95% CI [8–18%]. Percentage of patients with DLCO less than 80% of predicted: 0–3 months post discharge: 48%, 95% CI [41–56%]; 3–6 months post discharge: 33%, 95% CI [23–44%]; >6 months post discharge: 43%, 95% CI [22–65%]. Conclusion: The meta-analysis confirms a high prevalence of persistent lung diffusion impairment in patients following COVID-19-related hospitalization. Routine respiratory follow-up is thus strongly recommended.


INTRODUCTION
To date, over 200 million people worldwide have recovered from COVID-19 (https://www.worldometers. info/coronavirus/) (Worldometers (2020). Worl, 2020), but concern remains that some organs, including the lungs, might suffer long-term impairment following recovery from acute infections. Individual studies have shown that residual abnormalities of pulmonary function were observed in a subgroup of recovered COVID-19 patients, with the most common finding being a reduction in gas transfer as measured by diffusing capacity of the lungs for carbon monoxide (DLCO) Dhawan et al., 2021;Thomas et al., 2021). In this study, with meta-analysis, we aimed to determine the short (0-3 months), medium (3-6 months) and long (>6 months) respiratory outcomes in patients following COVID-19related hospitalisation. The findings will instruct appropriate interventions for subsequent increased healthcare utilisation post-COVID-19.

METHOD Criteria for Inclusion
We included randomised controlled trials (RCTs) and observational studies (cross-sectional, longitudinal, case-control and cohort) of patients with a confirmed diagnosis of COVID-19. The studies included aimed to determine the respiratory outcomes, in particular forced vital capacity (FVC) and diffusing capacity of the lungs for carbon monoxide (DLCO), in patients following COVID-19-related hospitalisation. The selected studies had to follow the ATS / ERS clinical guidelines. The included literatures should be published before May 15, 2021.

Criteria for Exclusion
Study's subjects who were not infected with COVID-19. Studies didn't report the time of hospital discharge or the time was calculated from diagnosis of COVID-19. Studies did not report FVC (% of predicted) or DLCO (% of predicted) or FVC <80% of predicted or DLCO <80% of predicted. Animal experiments,

Literature Retrieval and Selection
Firstly, according to the literature inclusion criteria, two researchers independently searched at Pubmed, ScienceDirect, Embase and Web of Science. Secondly, two researchers selected the literature and extracted the data independently in accordance with the standard data extraction table. When it came to divergences, a third researcher did the judgement. After the discussion, researchers reached a consensus. Finally, after the extraction and input of the data, two independent researchers did the subsequent analysis.

Extraction of Data
According to the inclusion criteria, we assessed the design of research, patients, and outcome indicators. First author, published year, number of cases, nationality, ages, body mass index (BMI), smoking status, respiratory comorbidities, time of assessment and, index quantity of FVC, % of predicted, DLCO, % of predicted; FVC <80% of  predicted and DLCO <80% of predicted were extracted from eligible studies.

Quality Assessment of Articles
The studies with randomised controlled trials were evaluated by Newcastle-Ottawa Scale (Bellan et al., 2021). As for no controlled trials, it includes the following aspects: 1) selection: Representativeness of the exposed cohort, selection of the non-exposed cohort, Ascertainment of exposure, Demonstration that outcome of interest was not present at start of study; 2) comparability: Research control matched important factors, but also controlled other important factors; and 3) outcome: assessment of outcome, follow-up long enough for outcomes to occur, adequacy of follow up of cohorts.

Synthesis and Analysis of Data
We used package "meta (version 4.18-0)" in R 4.0.1 and R studio to perform meta-analysis of the following pulmonary function tests (PFTs) indexes (1. FVC, % of predicted; 2. DLCO, % of predicted; 3. FVC <80% of predicted; 4. DLCO <80% of predicted.). Patients were divided into three groups: less than 3 months (0-3 months), more than or equal to 3 months and less than 6 months (3-6 months), and more than or equal to 6 months (≥6 months). We re-calculated the median (first quantile, third quantile) to mean ± standard deviation (SD) for FVC (% of predicted) and DLCO (% of predicted) in several studies. Statistical heterogeneity was measured through the I 2 statistic and classified as low (I 2 < 25%), moderate (I 2 25-50%), and high (I 2 > 50%) (Melsen et al., 2014). Subgroup analysis, according to the outcome assessment and severity, was performed. Sensitivity analysis was also employed to assess the change in pooled prevalence due to the selective exclusion of studies.

Literature Extraction
A total of 1,123 articles was retrieved from databases via the retrieval methods. Duplicate literatures were excluded through titles and abstracts. By reading the full text, we excluded 1,110 papers and conference abstracts with incomplete or no specific research method. Finally, 13 papers published in English were included (Liang et al., 2020; Huang et al., 2020;Venturelli et al., 2021;You et al., 2020;Lerum et al., 2021;Daher et al., 2020;Wu et al., 2021;Bellan et al., 2021;Li et al., 2020;van den Borst et al., 2020;Mo et al., 2020;Zhao et al., 2020;Huang et al., 2021), with a total of 3,455 patients. The evaluation of the quality of included studies by Newcastle-Ottawa Scale (NOS) (Stang et al., 2018) showed that two studies had a poor quality and the rest 11 studies passed the quality control. The basic characteristics of the included literatures were detailed in Table 1 and the procedure of literature retrieval and selection was shown in Figure 1.
Among the included studies, 10 studies reported FVC (% of predicted), eight studies reported DLCO (% of predicted), six reported FVC <80% of predicted, and nine reported DLCO <80% of predicted. Wu et al. (2021). reported all the indexes of the patients after the 3, 6 and 12 months following COVID-19-related hospitalisation ( Table 2). For those data reported in the form of median (first quantile, third quantile), we used R studio to re-calculate them into mean ± SD (Table 3).
Publication bias refers to the fact that research results with statistical significance are more likely to be reported and published than those without statistical significance and invalid results (DeVito and Goldacre, 2019). We examined the publication bias of meta-analysis of each indicator. There was no publication bias in FVC (% of predicted; p 0.93; Figure 2A), DLCO (% of predicted; p 0.54; Figure 2B) and DLCO (<80% of predicted; p 0.94; Figure 2C). For FVC <80% of predicted, less than 10 studies were included, so publication bias was not tested.
Comparison of longitudinal changes in the percentage of patients with FVC <80% of predicted.
These included six studies, which in total have eight groups of data showed the percentage of patients with FVC less than 80% of predicted. Based on the time of patients being discharged from hospital, we divided them into three groups: 0-3 months, 3-6 months and greater than 6 months. Meta-analysis showed that the percentage of patients with FVC less than 80% of predicted in 0-3 months, 3-6 months and ≥6 months post discharge was 10% (95% CI [6-14%]), 10% (95% CI [2-18%]) and 13% (95% CI [8-18%], respectively. The heterogeneity of 3-6 months was large, so the sensitivity analysis was carried out in this study. We removed the study from Wu et al. and got the meta-analysis result of this subgroup, which was 4% (95% CI [3-6%]) with I 2 6% ( Figure 5).
Comparison of longitudinal changes in the percentage of patients with DLCO <80% of predicted.

DISCUSSION
Post-acute COVID-19 syndrome, also known as long COVID, encompasses a wide range of physical and mental health symptoms that persist after recovery from acute SARS-CoV-2 infections (Nalbandian et al., 2021). Systematic studies of sequelae after recovery from acute COVID-19 are demanded to inform effective clinical management for patients suffered from long COVID. We recently reported the 3 months, 6 months, 9 months, and 12 months respiratory outcomes in patients following COVID-19-related hospitalisation from a relatively small prospective cohort (n 83) (Wu et al., 2021). In this study, we conducted meta-analysis to determine the short (0-3 months), medium (3-6 months) and long (>6 months) respiratory outcomes in patients following COVID-19-related hospitalisation. Significantly, we found a persistent reduction in DLCO over the study period, consistent with earlier reports (E et al., 2021). Low DLCO could be caused by interstitial changes or pulmonary vascular abnormalities following COVID-19 infections (Lang et al., 2020;Patel et al., 2020;Hanidziar and Robson, 2021). Our study has shown that up to a third of COVID patients still have evidence of defect DLCO 1 year after discharge (Wu et al., 2021), although longer term follow-up with a larger cohort will be required to confirm this observation.
In general, the heterogeneity of the studies included in the meta-analysis was low. However, the heterogeneity of DLCO less than 80% of predicted was higher, which may be caused by different ethnic groups, ages, disease severity, therapies and other factors. In general, the models we used were robust and reliable.
There are several limitations in this study. Firstly, age, sex ratio, nationality and disease severity of the patients included in the study are quite different, which may cause great heterogeneity and affect the final research results. Secondly, we only selected four indicators of lung function, so we cannot investigate the relationship between other indicators and discharge time. To be consistent and comparable with our earlier publication (Wu et al., 2021), we excluded those studies without data on FVC and/or DLCO values <80% of predicted. This might cause some false positive results considering the mean age of included patients is over 50 (van den Borst et al., 2020;Barisione and Brusasco, 2021;Milanese et al., 2021). In addition, pre-existing comorbidities for most COVID-19 patients are not known, which might cause certain bias of the results. Despite of these limitations, our findings in this meta-analysis are consistent with our previous report (Wu et al., 2021), confirming a high prevalence of persistent lung diffusion impairment in patients following COVID-19-related hospitalisation. Routine respiratory follow-up is thus strongly recommended.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.