Coronavirus-associated kidney outcomes in COVID-19, SARS, and MERS: a meta-analysis and systematic review

Abstract Objectives A meta-analysis and systematic review was conducted on kidney-related outcomes of three recent pandemics: SARS, MERS, and COVID-19, which were associated with potentially fatal acute respiratory distress syndrome (ARDS). Methods A search of all published studies until 16 June 2020 was performed. The incidence/prevalence and mortality risk of acute and chronic renal events were evaluated, virus prevalence, and mortality in preexisting hemodialysis patients was investigated. Results A total of 58 eligible studies involving 13452 hospitalized patients with three types of coronavirus infection were included. The reported incidence of new-onset acute kidney injury (AKI) was 12.5% (95% CI: 7.6%–18.3%). AKI significantly increased the mortality risk (OR = 5.75, 95% CI 3.75–8.77, p < 0.00001) in patients with coronavirus infection. The overall rate of urgent-start kidney replacement therapy (urgent-start KRT) use was 8.9% (95% CI: 5.0%–13.8%) and those who received urgent-start KRT had a higher risk of mortality (OR = 3.43, 95% CI 2.02–5.85, p < 0.00001). Patients with known chronic kidney disease (CKD) had a higher mortality than those without CKD (OR = 1.97, 95% CI 1.56–2.49, p < 0.00001). The incidence of coronavirus infection was 7.7% (95% CI: 4.9%–11.1%) in prevalent hemodialysis patients with an overall mortality rate of 26.2% (95% CI: 20.6%–32.6%). Conclusions Primary kidney involvement is common with coronavirus infection and is associated with significantly increased mortality. The recognition of AKI, CKD, and urgent-start KRT as major risk factors for mortality in coronavirus-infected patients are important steps in reducing future mortality and long-term morbidity in hospitalized patients with coronavirus infection.


Introduction
The novel coronavirus disease-2019 (COVID- 19), first reported in Wuhan, China, has become a worldwide pandemic and has caused over 28,918,900 confirmed cases of COVID-19 globally, including 922,252 deaths reported to the World Health Organization (WHO) as of 3:28 pm CEST, 14 September 2020 [1]. Apart from the rapid development of acute respiratory distress syndrome (ARDS), severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), as well as the previously identified severe acute respiratory syndrome coronavirus (SARS-CoV-1) and Middle East respiratory syndrome coronavirus (MERS-CoV), which are members of the coronavirus family [2], also have major associated but under-recognized extrapulmonary manifestations [3][4][5]. These three types of coronaviruses have caused catastrophic coronavirus pandemics in human history, namely SARS, MERS, and COVID-19. Among the organs affected, the kidneys are often involved due to the organ cross-talk between alveolar and tubular damage, i.e. the lung-kidney axis in ARDS [6]; the occurrence of kidney involvement usually indicates a worse prognosis [7,8]. Although the etiology of coronavirus-associated AKI is likely to be multifactorial, all three coronaviruses can directly invade renal cells through hijacking native surface receptors: angiotensin-converting enzyme 2 (ACE2) serves as a receptor for SARS-CoV-1 and SARS-CoV-2 [9,10], while MERS-CoV enters target cells via binding to dipeptidyl-peptidase 4 (DDP4) [11]. However, it is unclear how the virus causes cellular damage following the entry. If maintained during the course of infection, the kidney could function as a viral reservoir and urine become a potential source of viral transmission. Acute kidney injury (AKI) is the most frequent extrapulmonary organ dysfunction associated with ARDS and is an independent risk factor for mortality [12,13]. However, the reported prevalence and mortality of AKI for all three coronavirus infection differs between studies. All patients with chronic kidney disease (CKD), including those with end-stage kidney disease (ESKD) or on kidney replacement therapy (KRT), are immunosuppressed making them more susceptible to infection and potentially a more severe course [14][15][16]. The potential increased risk related to preexisting CKD and urgent-start KRT treatment is presently unclear. The influence of SARS-CoV-2/COVID-19 on sustained dialysis patients is also unknown.
In view of our previous experience with SARS, MERS, and more recent experiences of the COVID-19 outbreaks, we conducted a meta-analysis and systematic review to investigate the kidney involvement and patients' outcomes in hospitalized coronavirusinfected patients.

Data sources and search
This systematic review was performed following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. The registration of this review was published in PROSPERO (CRD42020200941). A search for published studies was performed using the PubMed database, EMBASE, and Cochrane library until 16 June 2020. Research articles on coronavirus (SARS-CoV, MERS-CoV, and SARS-CoV2) infected patients with information on kidney disease, AKI, dialysis, or kidney function were eligible and included. Keywords ('COVID-19 0 OR 'SARS-CoV2') or 'SARS-CoV' or 'MERS-CoV' and ('chronic kidney disease' or 'CKD' or 'kidney disease' or 'end-stage kidney disease' or 'ESKD') or ('acute kidney injury' or 'AKI') or ('kidney replacement therapy' or 'KRT' or 'blood purification') or ('dialysis' or 'hemodialysis' or 'blood purification') or ('mortality' or 'death') were combined to construct corresponding search formulas in databases. We used a combination of subject terms with free-text terms during the search, supplemented by a manual search and citation search. We also screened the latest relevant articles about COVID-2019 and met the inclusion criteria, through the "https://www.biorxiv.org/ search/covid-19" website.

Inclusion and exclusion criteria
Studies that met the following PICOS criteria were included: (1) articles were original reports including patients infected with coronavirus; (2) studies with outcomes of interest consisting of mortality or kidneyrelated outcomes, i.e. urgent-start KRT, AKI or dialysis; (3) types of articles were cohort studies, case series, and case-control studies. The exclusion criteria of the study included articles reporting patients infected with coronavirus other than SARS-CoV, MERS-CoV, and SARS-CoV2; articles without relevant outcomes of interest; commentaries or reviews; research articles with patient numbers below five.

Study selection and data extraction
Two reviewers (ZS and XC) independently screened the titles and abstracts and then checked the full text of all the articles that might be eligible. Differences were resolved through discussion or consultation with the third reviewer (XJ). The two researchers separately extracted data from the included studies, including first author of the article, year, study design, follow-up, number of reported cases, mortality, CKD, AKI, use of urgent-start KRT, ESKD, incidence or mortality of infected dialysis patients and related baseline characteristics. Mortality was defined as the death of patients during hospitalization.
The quality rating for each study was evaluated by the NOS (Newcastle-Ottawa scale). For the evaluation of case reports and case series included, we applied a generally recommended standard similar to NOS, based on the domains of selection, ascertainment, causality and reporting and provide signaling questions [17].

Patient and public involvement
No patient involved.

Certainty rating of evidence
The GRADE instrument (Grading of Recommendation, Assessment, Development, and Evaluation) was applied to rate the certainty of evidence and the strength of recommendations generated in our study [18][19][20]. The certainty of evidence was rated for kidney-related complications and underlying kidney diseases prevalence and associations with patients' prognosis. The five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) were taken into account to assess the confidence in effect estimates. Quality of evidence was characterized as high, moderate, low, or very low [20,21]. GRADE was assessed per http://gradepro.org.

Statistical analysis
Assessment of risk of bias was performed by two authors (XC and XJ) independently using the Newcastle-Ottawa Quality Assessment Scale (NOS) [22]. Studies were scored up to a maximum of 9 points by NOS. Study quality was classified into three categories: 0-3 (low), 4-6 (moderate), and 7-9 (high). Statistical analyses were performed using Revman software V5.4 (Cochrane). Sensitivity analysis on the results of pooled analysis was performed by method of excluding all the included preprinted literatures and method of one-byone elimination to verify the stability of the results. Rates for dichotomous data were analyzed using the Stata 16.0 (Stata Corp LP, TX, USA) Metaprop package. Since most studies were of retrospective design and heterogeneity between studies expected, the randomeffects model was chosen for data synthesis [23]. Odds Ratio (ORs) and 95% confidence intervals (CIs) were used for dichotomous variables as effect measures and were graphically visualized using Forest plots. Besides, T-statistic using Hartung-Knapp-Sidik-Jonkman method was performed for the degrees of freedom in the random-effects analysis, when the number of studies was < 10 by R version 4.03 (Metafor package) [24]. Heterogeneity across studies was evaluated using the Cochrane Q test and I 2 test (I 2 ¼ 100% ((Q-df)/Q). An I 2 value of 0-49%, 50-74% and >75% indicated low, moderate or high heterogeneity, respectively [25]. A twosided P value < 0.10 was considered statistically significant. Subgroup analysis was performed for each individual virus. If the number of studies was <9, publication bias was not investigated. Publication bias was evaluated with Begg's test, Egger's test and Funnel-plot. Meta-regression analysis was used to find potential heterogeneity. A two-sided P-value <0.05 except heterogeneity was considered statistically significant.

Quality rating
The quality rating for each study was evaluated by the NOS (Newcastle-Ottawa scale) (Supplementary Table 1). The average score of the included studies was 5.6 indicating moderate quality. The average score of studies for SARS, MERS, and COVID-19 were 5.8, 6.2, and 5.2, respectively. In domains of comparability, 52 studies with single arms did not fulfill the selection of the nonexposed cohort and only six studies [43,45,59,61,71,80] received scores. The GRADE tool was used to summarize pooled evidence for the main outcomes, as shown in Supplementary Table 2. Except several subgroup comparisons were rated as low, the majority of recommendations generated from this systematic review and meta-analysis were evaluated as very low.

Sensitivity analysis and meta-regression analysis
We further conducted sensitivity analysis to evaluate the influence of case series and preprinted literatures on the stability of results. First, the results maintained significance after excluding all the preprinted literatures included in the pooled analysis [27,28,38,39,41]. Second, the results also maintained stable by excluding the literatures included in the pooled analysis one by one. Moreover, too few studies were left in each subgroup after excluding all the case series, because this type of study occupied a relatively large proportion (about 50%). Thus, we kept the case series with number of patients reported equal or greater than 5 cases, and rated the quality of these literatures referring to a generally recommended standard [17]. Meta-regression analysis was used to find potential heterogeneity in primary results. However, different ethnicities and study types did not contribute significantly to the heterogeneity in four results (P all > 0.05).

Discussion
The COVID-19 pandemic that is currently raging around the world is causing major disruption to health systems [82]. As a member of the coronavirus family [2], COVID- 19 together with SARS and MERS lead to severe acute respiratory symptoms [83], as well as extrapulmonary disease [84]. Although the kidney is commonly affected, its contribution to patient mortality and morbidity is only belatedly being recognized. Compared with similar systematic reviews [85,86] that had been published so far, our research explored the impact of kidney-related events on the prognosis of patients in the face of coronavirus abuse from a more comprehensive and indepth perspective. We conducted this systematic review to investigate the incidence of AKI, the increased risk to patients with preexisting CKD, ESKD or urgentstart KRT and differences in kidney outcomes for all three recent coronavirus pandemics.
Our results indicate that AKI occurs in around onetenth of the infected study population with an overall mortality rate of 80.9%. The incidence of AKI was highest in MERS patients, while being similar between COVID-19 and SARS patients. The incidence in ICUtreated patients varied between 8.3% and 28.85% [81]. Compared to COVID-19 patients, the mortality rate was higher in SARS and MERS patients although fewer studies were reported for the SARS [56,57,60] and MERS [3,72] subgroups. AKI was associated with a significantly higher mortality in COVID-19 and MERS. In the SARS subgroup, this did not reach statistical significance possibly due to the small number of studies included. The incidence of urgent-start KRT use in coronavirus infected patients with AKI was 8.9% with an associated mortality of 80.7%. This probably reflects the fact that AKI patients requiring urgent-start KRT are generally more critically ill, likely to need ventilatory support or extracorporeal membrane oxygenation (ECMO) [87]. Many dialysis modalities [88][89][90], including CRRT, highvolume hemofiltration, plasma exchange, plasma adsorption and acute peritoneal dialysis have been reported, mostly as case reports or small case series. A consensus recommendation regarding the optimal dialysis modality, timing, dosage and duration for management of AKI in coronavirus diseases is urgently needed.
Our analysis also showed that the presence of CKD or ESKD was significantly associated with increased mortality. The overall prevalence of ESKD was higher than that of preexisting CKD, possibly due to different number of studies being enrolled for each analysis. Patients with MERS had the highest mortality in prevalent patients with CKD or ESKD. Several studies have reported on virus prevalence and mortality in patients on prevalent hemodialysis [30,38,42,44,[46][47][48][49][50][51]53,54,61,74,75]. The incidence of COVID-19 was 8.0% in routine hemodialysis patients, which was higher than SARS or MERS, but similar to the general population. The mortality rate for this subgroup was 25.7%, which was nearly the same as SARS but much less than with MERS (75%). Our analysis confirms that prevalent patients with CKD or on urgent-start KRT are at much higher risk of infection and of subsequent worse outcomes. Epidemic prevention measures must be strengthened especially in dialysis centers [91]. Specific measures that could be introduced include the setting up of isolation areas for dialysis centers, wearing personal protective equipment, tracking and isolating contacts and environmental disinfection. For infected patients, continuous bedside dialysis has been successfully deployed [92].
Our study had several limitations. First, we combined studies with a certain degree of heterogeneity, owing to the differences in the study design, sample size and population characteristics of the studies included. The specific reasons were as follows: (1) inclusion of case control study and case series can introduce bias to the result, which may lead to the high heterogeneity; (2) the number of infected patients enrolled in the included articles varied widely; (3) the inconsistent definitions of AKI or CKD or ESKD could have accounted for the variation of our results on AKI; (4) differences in the timing of outbreaks, geographical locations, ages, genders, habits, cares, and treatments may also contribute to the high heterogeneity; (5) the total sample size of SARS and MERS related studies was much smaller than COVID-19. Second, sampling bias may have contributed to part of our analysis when less than five cases were excluded. Third, renal function follow-up to assess renal recovery was not available. Fourth, several case series (about 8.62%) were included in our study, which could reduce the strength of the generated evidences. Then we tried to do the sensitivity analysis excluding all the case series, but found too few studies left. Also shown with the GRADE tool, the level of evidences generated in this study were low or very low; thus, more future high-quality researches are urged to confirm our results.
In conclusion, the kidney is commonly affectedly in patients with COVID-19, SARS and MERS. Renal events including AKI, preexisting CKD, and ESKD significantly increased the risk of mortality. Prevalent patients on urgent-start KRT also have an increased risk of infection and mortality. Routine hemodialysis patients were also at high risk of infection and mortality.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Author contributions
AO and ZM: act as guarantor for the validity of the study report. ZM and JX: study concept and design. JX and CX: acquisition of data. SZ and CX: extraction of data. JX and CX: checking of data. CX and BY: analysis and interpretation of data. XJ, CX, and SZ: drafting of the manuscript. AO and ZM: critical revision of the manuscript for important intellectual content.