Hydroxychloroquine/chloroquine and the risk of acute kidney injury in COVID-19 patients: a systematic review and meta-analysis

Abstract Objectives Hydroxychloroquine/chloroquine has been widely used as part of the standard treatment for patients with coronavirus disease 2019 (COVID-19). We conducted a systematic review and meta-analysis to determine whether hydroxychloroquine/chloroquine increases the risk of acute kidney injury (AKI) in COVID-19 patients. Methods PubMed and Embase were searched for related publications from inception to Dec 31, 2021, including randomized controlled trials (RCTs) and non-randomized studies of interventions (NRSIs) comparing the risk of AKI and/or increased creatinine in COVID-19 patients receiving hydroxychloroquine/chloroquine and other controls (active treatment and placebo). We conducted separate meta-analyses for RCTs and NRSIs based on fixed-effect model, with odds ratios (ORs) being considered as effect sizes. Results We included 21 studies in the analysis, with 12 were RCTs. Based on the RCTs, compared to placebo, the OR was 1.19 (95% confidence interval [CI]: 0.86, 1.64; p = .30, n = 4, moderate quality) for AKI and 1.00 (95%CI: 0.64, 1.56; p = .99, n = 5, moderate quality) for increased creatinine for patients received hydroxychloroquine/chloroquine treatment; compared to active treatment, the odds was 1.28 (95%CI: 0.65, 2.53; p = .47, n = 2, low quality) for AKI and 0.64 (95%CI: 0.13, 3.20; p = .59, n = 1, low quality) for increased creatine. Evidence from NRSIs showed slightly increased odds of AKI, with low quality. Conclusion Based on current available studies which were graded as low to moderate quality, there is insufficient evidence to conclude that hydroxychloroquine/chloroquine use is associated with increased risk of AKI or raised creatinine. Abbreviations: AKI: acute kidney injury; COVID-19: Coronavirus Disease 2019; RCT: randomized controlled trials; NRSI: non-randomized studies of interventions; OR: odds ratios; ROBIS-I: Risk Of Bias In Non-randomized Studies – of Interventions


Introduction
The coronavirus disease 2019 (COVID-19) pneumonia is caused by a new type of coronavirus, posing a huge threat to human health. According to the World Health Organization Coronavirus Dashboard, there have been 230,418,451 confirmed cases and 4,724,876 deaths globally, as of September 24, 2021 [1]. The common symptoms of COVID-19 pneumonia include fever, cough, headache, and other influenza-like symptoms [2]. While increasing evidence suggests that many COVID-19 patients have complications in the urinary system, acute kidney injury (AKI) is one of the most severe conditions [3][4][5]. As estimated by a meta-analysis, about 11% of the COVID-19 patients would develop AKI over 2 to 28 days follow-up and 6.8% needed the renal replacement therapy [6].
AKI is defined as a sudden loss of excretory kidney function that occurs within a few hours or days, presenting as an increase in serum creatinine levels with or without reduced urine output [7]. Many factors can lead to AKI, such as decreased renal perfusion/arterial pressure by any other conditions (e.g., heart failure), damage to tubular cells by nephrotoxic agents, or acute inflammation of blood vessels and glomeruli by systemic illness [8,9]. For COVID-19 patients, it is still unclear if AKI is largely due to hemodynamic changes, nephrotoxic agents or if the virus also leads to direct cytotoxicity [10].
Hydroxychloroquine, a derivative of chloroquine that has both antimalarial and anti-inflammatory activities, has been considered as part of the standard care for COVID-19 patients in some countries, especially during the first pandemic phase [11,12]. Although existing well-conducted meta-analyses have denied the potential benefits of hydroxychloroquine/chloroquine treatment on the prognosis of COVID-19 patients [12][13][14][15], millions of patients have already received such treatment. Hydroxychloroquine was found could potentially induce or intensify AKI by increasing lysosomal pH and inhibiting autophagy, because it has been demonstrated that Hydroxychloroquine inhibits autophagic flux by impairing autophagosome-lysosome fusion [16]. However, epidemiological evidence suggested controversy findings that make it unclear whether the development of AKI was associated with the use of hydroxychloroquine/chloroquine.
In this study, we conducted a systematic review and meta-analysis of current studies to explore the relationship between hydroxychloroquine/chloroquine and the risk of AKI.

Methods
The current systematic review was registered in PROSPERO (CRD42021265663) and was conducted and reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses statement 2020 [17].

Inclusion criteria
We included original studies that investigated the relationship between hydroxychloroquine/chloroquine treatment and the risk of AKI in COVID-19 patients (both adults and children). The primary intervention was hydroxychloroquine or chloroquine, while the control was any other treatment meant for active treatments, supportive care, or placebo. The primary outcome was AKI reported by the original studies, and the secondary outcome was increased creatinine. For the study design, we considered randomized controlled trials (RCTs), cohort studies, and case-control studies. For simplicity, we referred to the latter two types of studies as non-randomized studies of intervention (NRSIs). For studies with few patients who may have had renal failure before the infection, we only considered new cases of AKI after treatment.

Literature search
We searched PubMed and Embase for studies on the treatment of COVID-19 patients with hydroxychloroquine/chloroquine published from inception to July 10, 2021 (LZ). An updated search on Dec 31, 2021 was conducted during the process of revision. We did not search for unpublished sources because the current largest preprint database for health science (medRxiv) with the topic of COVID-19 has already been indexed in PubMed. We did not restrict the outcome of the search strategy to avoid potential omissions for eligible studies. The following medical subject headings and/or keywords were used for hydroxychloroquine/chloroquine: 'Hydroxychloroquine', 'Oxychloroquine', 'Plaquenil', 'Chloroquine', 'Chlorochin', 'Chingamin', 'Khingamin', 'Nivaquine', 'Aralen', and 'Arechine'. Moreover, the following medical subject headings and/or keywords were used for COVID-19:

Literature screen
Two review authors (LZ, ZZ) independently conducted the literature review. We first screened the titles and abstracts of the publications obtained by the search, followed by a review of the full text of the remaining records. Any disagreements were resolved through discussion. For each potential study, we also checked ClinicalTrial.gov for the outcome information.

Data extraction
Two review authors (ZZ, LQ) conducted data extraction using a pre-specified Excel sheet template. The following information was extracted: name of the first author, country where the study was conducted, patient characteristics (including age, comorbidity, diagnostic methods), intervention agent and dosage, control agent and dosage, outcomes, sample size of the study, number of outcomes in each group, follow-up of the study for the outcomes, and type of studies. Any disagreements in the extracted information were resolved through discussion.

Risk of bias
As planned, we used the Cochrane ROB 2.0 tool to assess the RCT's risk of bias [18]. This tool considers five domains: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in the measurement of the outcome, and bias in the selection of the reported result. For the cohort and case-control studies, we primarily planned to use the Newcastle-Ottawa Scale checklist [19], but we found that it could not be presented through a risk of bias plot. Therefore, we used the Risk Of Bias In Non-randomized Studies -of Interventions (ROBIS-I), as suggested by the Cochrane handbook, to assess the risk of bias [20]. To reduce subjective judgment, all review authors assessed the risk of bias.

Statistical analysis
The meta-analyses were conducted using RevMan 5.3 (Cochrane; London, United Kingdom) and Stata 12.0 (StataCorp LLC; Texas, United States). Given that both the primary and secondary outcomes were binary outcomes, we used the odds ratio (OR) and its 95% confidence interval (CI) to measure the potentially harmful effects of hydroxychloroquine/chloroquine. Furthermore, we used the Mantel-Haenszel method to pool the study-specific effects [21], and studies with different designs (RCTs vs. NRSIs) were pooled separately. Statistical significance was set at p < .05. Considering that the events of AKI and increased creatinine could be rare in some of the studies, we neglected the potential heterogeneity among the included studies and used the fixed-effect model to increase the statistical power. This decision was based on the recommendations of the Cochrane Handbook [21]. Considering the potential zero events in some studies, we used risk difference based on the Mantel-Haenszel method as a sensitivity analysis for those with double-zero studies [22].
Heterogeneity was measured using I 2 . As suggested in the Cochrane handbook [21], we treated I 2 < 30% as no or slight heterogeneity; otherwise, there was moderate or substantial heterogeneity (I 2 30%). For metaanalyses with moderate or substantial heterogeneity, the evidence was downgraded and a sensitivity analysis using a random-effect model based on the inverse-variance method was conducted.
Subgroup analysis was performed by type of control to determine if there was a difference in the ORs for active and placebo controls. We primarily planned to use age and dosage of treatment for the subgroup analysis, but half of the included studies did not provide detailed age information (i.e., adults or children), and only a few studies used different doses, making the planned subgroup analyses unimplementable. A further sensitivity analysis was conducted by omitting one study at a time to determine if the results were stable.
Egger's method was used to detect potential publication bias. Finally, the Grading of Recommendations Assessment, Development and Evaluation tool was used to rate the global evidence of each meta-analysis to form the conclusions [23].

Baseline characteristics
The initial literature search in the two databases resulted in 4,830 records, and we identified 798 duplicates by author's name, title, and publication year. Therefore, 4,032 publications were screened by title and abstract. We further excluded 3,832 records that did not meet the criteria, and the remaining 200 publications were screened through a review of the full text. After excluding 134 with different outcomes, 33 with different interventions, ten with different populations, and three with different study types, we finally included 20 studies [24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43] for the current systematic review, with a kappa value of 0.85 between the two reviewers ( Figure S1, Supplementary Appendix). The updated search in 31, Dec 2021 (10 July, 2021 to 31, Dec 2021) resulted in 648 records, 72 were identified as duplicates, and 1 RCT [44] was identified as new studies that included in current meta-analysis. Table 1 presents the main characteristics of the included studies. Among the 21 included studies, 14 considered AKI outcomes, and seven analyzed increased creatinine levels. Of the 14 studies on AKI, six were RCTs, six were cohort studies, and two were nested case-control studies. For the seven studies on increased creatinine levels, six were RCTs, and one was a cohort study. For the 21 studies, 13 clearly specified the age range, with 12 focusing on adults and one focusing on children. The daily total dosage of hydroxychloroquine/chloroquine ranged from 200 mg to 1200 mg (available from 18 studies), with 13 studies providing 400 mg, two studies providing 200 mg, one study providing 600 mg, one study providing 800 mg, and one study providing 1200 mg. The follow-up for the safety outcomes ranged from 3 to 42 days, while three studies did not provide such information. Eight studies were conducted in Europe, seven in Asia, and six in America.   bias due to the deviation from the intended intervention, and one study [39] had a high risk of bias in the selection of the reported result. In addition, four studies [24,26,27,38] had a possible bias of deviation from the intended intervention, and five [29,[39][40][41][42] presented some concerns regarding the bias of missing outcomes. For the nine NRSIs, six studies [30,31,34,35,37,43] showed a critical bias, three [32,33,36] showed moderate bias due to confounding, and four studies [31,34,35,37] showed moderate bias due to the selection of participants. In addition, three [31,32,36] showed critical bias, and three [32,34,43] showed moderate bias due to the measurement of outcomes, and one [43] showed moderate bias due to the selection of the reported result.

Hydroxychloroquine vs. active treatment
Two RCTs compared the risk of AKI, and one RCT compared the risk of increased creatinine between hydroxychloroquine/chloroquine and active treatment ( Figure  3). Compared to active treatment, the meta-analysis showed that the OR of AKI with hydroxychloroquine/ chloroquine was 1.  Figure S3 (Supplementary Appendix).

Sensitivity analysis and publication bias
The results of the sensitivity analyses are presented in the Supplementary Appendix file ( Figure S4 to S7), which showed that the effects remained stable in the random-effect model based on the evidence from RCTs. However, for evidence of NRSIs, the removal of the studies by Grimaldi [31] or Schneider [33] caused the pooled OR to no longer be significant (p > .05) comparing hydroxychloroquine/chloroquine and active treatment. However, this did not alter the direction of the effects. Considering the limited number of studies on the two outcomes in the meta-analyses, we combined them to detect the potential publication bias, with our results suggesting absent bias (Egger's p ¼ .56 for metaanalysis based on RCTs for placebo; Egger's p ¼ .86 for meta-analysis based on NRSIs for active treatment).

Discussion
We conducted a systematic review and meta-analysis to investigate the risk of AKI and increased creatinine levels in COVID-19 patients treated with hydroxychloroquine/chloroquine. To the best of our knowledge, this was the first systematic review to investigate the harmful effects of hydroxychloroquine/chloroquine in COVID-19 patients. Our results suggest that there is currently no strong evidence that hydroxychloroquine/ chloroquine increases the risk of AKI in COVID-19 patients.
In our meta-analysis, the potential for increased risk of AKI was detected by evidence from cohort and casecontrol studies. However, the quality of the evidence was low, possibly due to the confounding factor that the majority of the NRSIs had a critical risk of bias from potential confounders. In addition, the unstable results in the post-hoc sensitivity analysis for the meta-analyses of NRSIs suggested some evidence of publication bias. We noticed that the sample size of Schneider's study was small, while the effect was extremely large. Further, the study had a substantial impact on the pooled effects, which is clear evidence of publication bias. However, Egger's test failed to confirm this.
Our study's direct evidence suggests that the routine administration of hydroxychloroquine/chloroquine is not associated with nephrotoxicity in COVID-19 patients. Meanwhile, the indirect evidence suggests that other active treatments may not also result in nephrotoxicity in COVID-19 patients. This further indicates that the AKI cases reported in COVID-19 patients were more likely due to factors other than drug toxicity. Hirsch et al. [45] found that patients who required mechanical ventilation had an almost four times higher risk of developing AKI than non-ventilated patients. Chen et al. found that in hospitalized COVID-19 patients, a higher level of serum phosphorus was associated with an increased risk of AKI [46]. Further research on the molecular biology mechanism is helpful in exploring the potential relationship between COVID-19 infection and AKI.
The current study had some strengths. First, the current study used a systematic review approach to summarize all available studies to ensure more representative and credible results. Second, we also maintained a standard and strict process during the conduct of the systematic review and meta-analysis to guarantee the quality of the current study. Furthermore, the quality of the included RCTs and the homogenous effects ensure a moderate quality of evidence, making the findings conclusive. However, this study had some limitations. First, although we included both RCTs and NRSIs, the quality of the evidence from NRSIs was proved to be low, contributing little contribution to the conclusion. Furthermore, the analysis of NRSIs takes a large amount of time, which largely prolonged the conduct of the current systematic review. Second, there were only four or five RCTs for each outcome, and most of the RCTs were based on an openlabel design, which may have led to some bias in the results. Therefore, we downgraded the evidence as moderate. Third, some RCTs reported limited information on the safety outcomes, excluding renal system adverse events, and therefore were not included in the current study. However, such RCTs may have collected the information, but the authors failed to report it, leading to a potential loss of evidence. These limitations should be addressed in updated meta-analyses.
In conclusion, based on currently available studies which were graded as low to moderate quality, there is insufficient evidence to conclude that hydroxychloroquine/chloroquine use is associated with increased risk of AKI or raised creatinine.

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

Funding
This study is supported by the Department of urology, Hubei Hospital of Traditional Chinese Medicine.