key: cord-283780-h4lwzpl9
authors: Zhang, John J Y; Lee, Keng Siang; Ang, Li Wei; Leo, Yee Sin; Young, Barnaby Edward
title: Risk Factors of Severe Disease and Efficacy of Treatment in Patients Infected with COVID-19: A Systematic Review, Meta-Analysis and Meta-Regression Analysis
date: 2020-05-14
journal: Clin Infect Dis
DOI: 10.1093/cid/ciaa576
sha: 
doc_id: 283780
cord_uid: h4lwzpl9

The coronavirus disease 2019 (COVID-19) pandemic spread globally in the beginning of 2020. At present, predictors of severe disease and the efficacy of different treatments are not well-understood. We conducted a systematic review and meta-analysis of all published studies up to March 15, 2020 which reported COVID-19 clinical features and/or treatment outcomes. 45 studies reporting 4203 patients were included. Pooled rates of intensive care unit (ICU) admission, mortality and acute respiratory distress syndrome (ARDS) were 10.9%, 4.3% and 18.4%, respectively. On meta-regression, ICU admission was predicted by raised leukocyte count (p<0.0001), raised alanine aminotransferase (p=0.024), raised aspartate transaminase (p=0.0040), elevated lactate dehydrogenase (LDH) (p<0.0001) and increased procalcitonin (p<0.0001). ARDS was predicted by elevated LDH (p<0.0001), while mortality was predicted by raised leukocyte count (p=0.0005) and elevated LDH (p<0.0001). Treatment with lopinavir-ritonavir showed no significant benefit in mortality and ARDS rates. Corticosteroids were associated with a higher rate of ARDS (p=0.0003).

A pandemic of coronavirus disease caused by the novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spread from Asia to the rest of the world in the first three months of 2020. The consequences for human health, the global economy and normal functioning of society have been unprecedented.

COVID-19 causes infection in any age group, though severe disease is more common in older adults [1] . The clinical spectrum of disease ranges from asymptomatic or subclinical infections to organ dysfunctionshock, acute respiratory distress syndrome (ARDS), acute cardiac injury, and acute kidney injury (AKI)and death [2] . As of May 2, 2020, there was a total of 3,421,226 confirmed cases globally. Of the 1,333,313 cases that had an outcome reached, 240,222 had resulted in mortality [3] .

The growth curve of COVID-19 academic literature since the first report of this outbreak from Wuhan, Hubei Province, China in December 2019 has been exponential [13, 131 publications found on the National Institutes of Health COVID-19 Portfolio and 8502 publications found on PubMed on May 2, 2020 using the search string 'coronavirus disease 2019 OR SARS-CoV-2']. However, systematic reviews which consolidate these findings remain scarce, with none focused on understanding the predictors for severe disease including the effects of different experimental antiviral and immune-modulatory treatments [4] .

To address this gap in the literature, we conducted a systematic review, meta-analysis and meta-regression to 1) investigate the predictive value of laboratory investigations for severe disease and adverse outcomes, and 2) evaluate the efficacy of antivirals and corticosteroids for COVID-19. M a n u s c r i p t 5

Reviews and Meta-Analyses (PRISMA) guidelines [5] . 

All titles and abstracts were screened independently by two reviewers (JJYZ and KSL) against a set of pre-defined eligibility criteria. Potentially eligible studies were selected for full-text analysis. Disagreements were resolved by consensus or appeal to a third senior reviewer (BEY). Agreement among the reviewers on study inclusion was evaluated using Cohen's kappa [6] .

All original studies reporting the clinical characteristics (symptoms and signs, laboratory investigations and radiological findings) and treatment outcomes of patients with COVID-19

were included in our meta-analysis. Case reports and series with a sample size of <5 were excluded per recommendations by the Cochrane Statistical Methods Group and in accordance with methodologies of previously published meta-analyses [7] [8] [9] . Other exclusion A c c e p t e d M a n u s c r i p t 6   criteria included non-English articles, non-original research papers, laboratory-based and   epidemiological studies with no clinical characteristics reported, as well as non-human   research subjects [see Supplementary Table 2] .

The quality of included studies was assessed using the Joanna Briggs Institute (JBI) checklist for prevalence studies and the JBI checklist for case series [10] . Full details are in Supplementary Tables 3 and 4 . In summary, these tools rated the quality of selection, measurement and comparability for all studies and gave a score for prevalence studies (maximum of 9) and case series (maximum of 10). Two researchers (JJYZ and KSL) assessed the quality of all included studies and discussed discrepancies until consensus was reached.

Data were extracted on the following variables: study details, sample size of study, method of diagnosis, age, gender, coexisting medical conditions, clinical symptoms, laboratory investigations, radiological findings, treatment details and patient outcomes.

Primary outcome measures were intensive care unit (ICU) admission rate, mortality rate and the event rate of ARDS. ICU admission was used as a surrogate marker for severe infection.

Secondary outcome measures included other morbidities such as respiratory failure, septic shock, coagulopathy, acute cardiac injury, AKI and secondary infection, as well as length of hospital stay (LOS) and discharge rate at the point of study completion.

A c c e p t e d M a n u s c r i p t 7

To account for intra-study and inter-study variance, random effects models were used for meta-analyses of variables and end points [11] . Pooled proportions were computed with the inverse variance method using the variance-stabilizing Freeman-Tukey double arcsine transformation [12] . Confidence intervals (CI) for individual studies were calculated using the Wilson Score confidence interval method with continuity correction. The I 2 statistic was used to present between-study heterogeneity, where I 2 ≤ 30%, between 30% and 50%, between 50% and 75%, and ≥ 75% were considered to indicate low, moderate, substantial, and considerable heterogeneity, respectively [13] . P values for the I 2 statistic were derived from the chi-square distribution of Cochran Q test. For pooling of means of numerical variables, we computed missing means and standard deviations (SDs) from medians, ranges (minimum to maximum) and interquartile ranges (IQRs) using the methods proposed by Hozo et al. and Wan et al [14, 15] .

Summary-level meta-regression was performed using the mixed-effects model after computation of the SD of Freeman-Tukey double arcsine transformed proportions.

Publication bias of studies was assessed using funnel plots, where an asymmetrical distribution of studies was suggestive of bias [16] . Quantitative analysis of funnel plot asymmetry was done using Egger's regression test, based on a weighted linear regression of the treatment effect (expressed as a Freeman-Tukey double arcsine transformed proportion) on its standard error [17] . The GRADE approach was used to evaluate the quality of evidence for each outcome [18] .

All statistical analyses were performed using R software version 3.4.3 (R Foundation for Statistical Computing, 2016), with the package meta [19] . P-values less than 0.05 were considered statistically significant. All included studies were non-randomized, retrospective observational studies. 42 studies reported data from China, with one each from Singapore, South Korea and Hong Kong.

Details of included studies are reported in Supplementary Table 5 . Of the 36 prevalence studies, 33 studies attained a full score of 9 on the JBI checklist for prevalence studies, two studies attained a score of 8 and one study attained a score of 7 [see Supplementary Table   3 ]. Of the nine case series, 7 studies attained a full score of 10, one study attained a score of 8 and one study attained a score of 7 [see Supplementary Table 4 ].

Of the total 4203 patients, 2797 were male (66.5%) and 1406 were female (33.5%). A c c e p t e d M a n u s c r i p t 9

The most common blood abnormalities observed were elevated C-reactive protein (CRP) (59.4%), decreased albumin (58.6%), elevated lactate dehydrogenase (LDH) (51.7%) and lymphopenia (47.7%). The most common radiological abnormalities seen on chest computed tomography (CT) scan were bilateral infiltrates (80.8%), ground glass opacities (73.0%), interlobular septal thickening (46.3%), subpleural lines (45.5%) and consolidation (41.6%). In terms of treatment, type of antivirals used included combinations of oseltamivir, ganciclovir, lopinavir, ritonavir, ribavirin and arbidol. Type of antibiotics used comprised moxifloxacin, ceftriaxone and azithromycin. Table 1 Funnel plots and Egger's regression test were done to assess for publication bias for ICU admission, mortality and ARDS rates. There was no evidence of publication bias for ICU admission (p = 0.42), mortality (p = 0.41) and ARDS (p = 0.14) [see Supplementary Figure   4 ].

At baseline, the quality of evidence derived from a review of COVID-19 studies was assessed as low, owing to their observational nature. The quality of evidence for respiratory failure was rated down to very low for imprecision, due to the large confidence interval range and the relatively small sample size analyzed. Despite considerable study heterogeneity demonstrated by the I 2 values for most outcome measures, there was no rating down due to inconsistency, as the heterogeneity could likely be explained by differences in patient demographics, diagnostic criteria, treatment methods and management protocols given that COVID-19 is a newly emergent disease.

Meta-regression was performed to identify risk factors of ICU admission, ARDS and mortality [ Table 2 ]. Fourteen studies with a total of 2153 patients reported ICU admission rates. 

Subgroup analysis was performed for studies with the use of corticosteroids reported.

Sixteen studies with a total of 2407 patients reported the use of corticosteroids. Pooled mortality rate in these patients was 7.2% (95% CI: 1.7 -15.4) and pooled ARDS rate was 22 .7% (95% CI: 9.9 -38.6). Meta-regression demonstrated a significant association between corticosteroids use and higher rate of ARDS (p = 0.0003) [Fig. 5 

A c c e p t e d M a n u s c r i p t 12

Our meta-analysis provides an in-depth analysis of the key epidemiological features, clinical characteristics, laboratory investigations, radiological findings, treatment details and outcomes of COVID-19 from published literature. We identified elevated LDH as a significant predictive marker of ARDS, and found that both elevated leukocyte count and elevated LDH predict mortality. Treatment with the anti-retroviral drug lopinavir-ritonavir was not associated with significant benefit, while corticosteroids were associated with possible harm.

Early recognition of severe infection may allow early intervention with supportive measures and therapeutics and improve outcomes [20] . Our meta-regression identified five significant markers of ICU admission: raised leukocyte count, raised ALT and AST, in addition to elevated LDH and finally increased procalcitonin. While 10.7% of patients had a raised leukocyte count in our meta-analysis, the degree of leukocytosis was modest (pooled mean leukocyte count was 6.0 x10 9 /L). Raised ALT and AST in severe COVID-19 disease may be a result of liver damage caused by the direct binding of SARS-CoV-2 to angiotensinconverting enzyme 2 positive cholangiocytes [21] . In our analysis, LDH was the only marker that significantly predicted all three measured outcomes: ICU admission, ARDS and mortality. LDH is released from cells upon damage to their cytoplasmic membrane, and is not only a metabolic but also an immune surveillance prognostic biomarker [22, 23] . LDH increases the production of lactate, which leads to enhancement of immune-suppressive cells and inhibition of cytolytic cells [24] . These changes could weaken the immune response mounted against the viral infection, resulting in more severe disease in patients with elevated LDH. Increased procalcitonin may have been a marker of bacterial co-infection, thereby resulting in complications of COVID-19 disease and hence a higher rate of ICU admission in these patients [25] . Interestingly, lymphopenia was not found to be a significant predictor of ICU admission, mortality and ARDS in our meta-analysis. A possible explanation may be that we analyzed lymphopenia as a dichotomous variable without taking into account A c c e p t e d M a n u s c r i p t 13 the degree of lymphopenia i.e. the numerical value of lymphocyte count, which lies on a spectrum and could affect disease severity among patients with lymphopenia.

The results of randomized clinical trials of COVID-19 interventions are of critical importance as only weak evidence supports the currently available repurposed and novel antivirals [26] .

Among the patients with antiviral use reported in our meta-analysis, overall rates of mortality, ICU admission and ARDS were 5.7%, 11.8% and 20.2%, respectively. We found no overall benefit from treatment with lopinavir-ritonavir, in line with a recent randomized controlled trial However, this trial demonstrated that lopinavir-ritonavir treatment granted a significant reduction in ICU length of stay in survivors. Further trials (NCT04252885 and NCT04307693) are in progress to assess the efficacy of both lopinavir and ritonavir in reducing the COVID-19 viral load, and we look forward to future developments to provide recommendations on the use of antiviral therapy [28, 29] .

Severe COVID-19 is associated with a dysregulated host inflammatory response, suggesting immune modulators as an attractive treatment modality [30] . Corticosteroids were used during the SARS-CoV outbreak, however, in a meta-analysis only four studies provided conclusive data, and all four indicated possible harm [31, 32] . These harms included risks of prolonged viremia, corticosteroid-induced diabetes, avascular necrosis and psychosis [31, 33, 34] . Our meta-analysis suggested that the use of corticosteroids is associated with disease severity (ICU admission) and higher ARDS rates. It is not clear if this effect is a consequence of corticosteroid treatment, or confounding by indication bias where sicker patients are more likely to receive corticosteroids. An RCT of corticosteroids in severe respiratory viral infections has long been called for, and at least one clinical trial in COVID-19

(NCT04244591) is ongoing [35] .

A c c e p t e d M a n u s c r i p t 14 SARS-CoV-2-induced pneumonia is marked by a cytokine stormhyperactivation of effector T cells and excessive production of inflammatory cytokines, particularly interleukin-6 (IL-6) [36] . Blockade of IL-6 function using tocilizumab, a specific monoclonal antibody against its receptor appears to be useful in alleviating hyperinflammation symptoms in severe cases [37, 38] . Selective Janus kinase-signal transducer and activator of transcription (JAK-STAT) inhibitors such as baricitinib may also be beneficial, though clinical trials are required and any benefit is likely to be greatest in combination with an effective antiviral [39] . To the best of our knowledge, this is the first systematic review and meta-analysis of COVID-19 to describe specific laboratory predictors of severe disease and adverse outcomes. Our study is also the first meta-analysis to evaluate the efficacy of antivirals and corticosteroids.

Careful attention should be given to the management of patients with raised leukocyte count, raised ALT and AST, elevated LDH, increased procalcitonin and raised leukocyte count as these factors predict ICU admission, mortality and ARDS. In terms of treatment efficacy, the use of corticosteroids in COVID-19 patients is significantly associated with higher rates of ARDS. Compared to other antivirals, the use of lopinavir and ritonavir is non-superior in terms of lowering mortality rate. Further prospective studies are vital to clarify our findings.

A c c e p t e d M a n u s c r i p t 16

No funding was used for the production of this work.

All authors have no potential conflicts of interest to disclose. M a n u s c r i p t 20 A c c e p t e d M a n u s c r i p t 28 Figure 5 sw

Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Prevalence of comorbidities in the novel Wuhan coronavirus (COVID-19) infection: a systematic review and metaanalysis

Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement

A coefficient of agreement for nominal scales

Extracorporeal Membrane Oxygenation in Pregnant and Postpartum Women With H1N1-Related Acute Respiratory Distress Syndrome: A Systematic Review and Meta-analysis

Extracorporeal Membrane Oxygenation in Pregnant and Postpartum Women: A Systematic Review and Meta-Regression Analysis

Excluding small studies from a systematic review or metaanalysis

The methodological quality assessment tools for preclinical and clinical studies, systematic review and meta-analysis, and clinical practice guideline: a systematic review

A basic introduction to fixed-effect and random-effects models for meta-analysis

Metaprop: a Stata command to perform metaanalysis of binomial data

Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0. The Cochrane Collaboration

Estimating the mean and variance from the median, range, and the size of a sample

Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range

Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials

Bias in meta-analysis detected by a simple, graphical test

GRADE: an emerging consensus on rating quality of evidence and strength of recommendations

How to perform a meta-analysis with R: a practical tutorial

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Specific ACE2 Expression in Cholangiocytes May Cause Liver Damage After 2019-nCoV Infection

Lactate dehydrogenase, a Risk Factor of Severe

Clinical characteristics and prognosis of community-acquired pneumonia in autoimmune disease-induced immunocompromised host: A retrospective observational study

Elevated lactate dehydrogenase (LDH) can be a marker of immune suppression in cancer: Interplay between hematologic and solid neoplastic clones and their microenvironments. Cancer biomarkers : section A of Disease markers

Procalcitonin in patients with severe coronavirus disease 2019 (COVID-19): a meta-analysis

Pharmacologic Treatments for Coronavirus Disease 2019 (COVID-19): A Review

A Trial of Lopinavir-Ritonavir in Adults Hospitalized with Severe Covid-19

The Efficacy of Lopinavir Plus Ritonavir and Arbidol Against Novel Coronavirus Infection (ELACOI)

Comparison of Lopinavir/Ritonavir or Hydroxychloroquine in Patients With Mild Coronavirus Disease (COVID-19)

The trinity of COVID-19: immunity, inflammation and intervention

SARS: systematic review of treatment effects

Corticosteroid Therapy for Critically Ill Patients with Middle East Respiratory Syndrome

Factors associated with psychosis among patients with severe acute respiratory syndrome: a case-control study

Effects of early corticosteroid treatment on plasma SARS-associated Coronavirus RNA concentrations in adult patients

Glucocorticoid Therapy for Novel CoronavirusCritically Ill Patients With Severe Acute Respiratory Failure (Steroids-SARI)

Clinical predictors of mortality due to COVID-19 based on an analysis of data of 150 patients from Wuhan, China

COVID-19: consider cytokine storm syndromes and immunosuppression

Tocilizumab treatment in COVID-19: A single center experience

COVID-19: combining antiviral and antiinflammatory treatments

A c c e p t e d M a n u s c r i p t 17 A c c e p t e d M a n u s c r i p t A c c e p t e d M a n u s c r i p t