Table 2.
Meta-analysis on the efficacy of hydroxychloroquine on COVID-19 patients published in peer-reviewed journals.
| References | k/N | Main results | Heterogeneity analysis | Publication bias analysis |
|---|---|---|---|---|
| Ayele Mega et al. (27) | 20/6,782 | HCQ group did not differed on the rate of virologic cure (OR = 0.78; 95% CI [0.39–1.56]) or the risk of mortality (OR = 1.26; 95% CI [0.66–2.39]) compared to control. | Some analysis revealed high heterogeneity (up to I2 = 95%). Subgroup analysis of observational vs. RCTs studies. | Cochrane risk of bias tool for RCTs. Newcastle-Ottawa Quality Assessment scale (NOS) for observational studies. Subgroup analysis with low biased studies. |
| Bignardi et al. (28) | 12/7,629 | HCQ (with or without AZ) was not associated with mortality (RR = 1.09, [0.98–1.20]). | Moderate heterogeneity (I2 ≤ 54.6%). Subgroup analysis based on sensitivy analysis. | Egger test did not revealed sign of publication bias (p > 0.05). |
| Choudhuri et al. (29) | 14/12,455 | HCQ did not affect mortality compared to control group (RR = 1.003, [0.983–1.022]). | Low to high heterogeneity (up to I2 = 97.9%). No subgroup analysis. | Two authors independently evaluated within-study bias. |
| Das et al. (30) | 7/726 | HCQ did not affect the virological cure except after day 5 (OR = 9.33, [1.51–57.65]). | Null (I2 = 0%) to high heterogeneity (I2 = 96%) but small set of comparisons (k = 2). | Cochrane handbook to assess biased of RCTs (2 independent authors) and NOS for observational studies. ROBINS-I tool for non-randomised trials. |
| Ebina-Shibuya et al. (31) | 8/2,063 | HCQ was not associated with mortality (OR = 1.05, [0.53–2.09]). | I2 varies between 0 (adverse event), 31% (all cause death), 57% (time to viral clearance) up to 74% (for viral clearance at 7 days). Subgroup analysis on study design (RCTs vs. observational). | Cochrane Risk of Bias tool for RCTs. |
| Elavarasi et al. (32) | 15/10,659 | No significant reduction in mortality in HCQ group (RR = 0.98, [0.66–1.46]), fever duration (mean difference – 0.54 days) or clinical deterioration (RR = 0.90, [0.47–1.71]). | High heterogeneity for mortatliy and clinical deterioration (I2 = 87%), virological clearance (I2 = 80%), time to fever remission (I2 = 72%). Subgroup analysis of RCTs vs Cohort studies. | Cochrane Risk of Bias Tool for RCTs/Newcastle Ottawa Scale revealed significant bias without additional analysis. |
| Elsawah et al. (33) | 6/609 | No significant effect on viral clearance, clinical progressions, or mortality (p's > 0.10). Significant improvement on radiological progression (risk difference −0.20 [−0.36, −0.03]). | Low (I2 = 0%) to high heterogeneity (I2 = 94%) without subgroup analysis. | Cochrane Risk of Bias Tool (2 independent authors). Sensitivity analysis after removing the low-quality studies. |
| Kashour et al. (34) | 21/20,979 | No effect of HCQ on mortality (OR = 1.05, [0.96–1.15]) and small increased mortality with HCQ/AZ combination on a subset of studies (OR = 1.32, [1.00–1.75]). | No heterogeneity for the HCQ group and moderate for the HCQ/AZ comparison (I2 = 68.1%). Sensitivy analysis excluded studies with high risk of bias. | Neither funnel plot nor Egger's regression test revealed signs of publication bias (p = 0.276) |
| Fiolet et al. (26) | 17/11,932 | No difference in mortality for all studies or RCT (OR = 0.83 and 1.09). | I2 = 84% among non-RCTs with null heterogeneity for RCTs. | Funnel plot, Begg's and Egger's tests. |
| Ghazy et al. (35) | 14/12,821 | No difference between standard care en HCQ group (RR = 0.99, [0.61–1.59]). Mortality higher in HCQ/AZ comparison (RR = 1.8, [1.19–2.27]). | High heterogeneity was observed in different analysis (0% < I2 <98%). Subgroup analysis no revealed significant effect (e.g., mortality HCQ/AZ, RR = 2.23, [1.70–2.91]). | Publication bias assed by funnel plot. |
| Hussain et al. (36) | 6/381 | The risk of mortality in HCQ treated individuals is on average 2.5 times greater than in non-HCQ individuals (95% CI [1.07–6.03]). For moderate to mild symptoms, the rate of improvement was 1.2 higher compared to the control group (95% CI [0.77–1.89]). | These studies were perfectly homogeneous (I2/τ2 = 0). | Marginal asymmetry on funnel plot. |
| Hong et al. (37) | 14/24,780 | No effet of HCQ alone or in combination on mortality (OR = 0.95, [0.72–1.26]). | Substantial heterogeneity in all analysis (71% ≤ I2 ≤ 93%). Subgroup analysis of HCQ alone without effect (OR = 0.90, [0.60–1.34]). | Publication bias visible on funnel plot. Comparisons results with and without biased studies (no significant differences). |
| Lewis et al. (38) | 4/4,921 | HCQ group were not at fewer risks of developing COVID-19 (RR = 0.82, [0.65–1.04]), hospitalization (RR = 0.72, [0.34–1.50]) or mortality (RR = 3.26, [0.13–79.74]) compared to control but increased the risk of adverse events (RR = 2.76, [1.38–5.55]). | Statistical heterogeneity was assessed using the χ2 and I2 statistics (either 0 or 95%). Subgroup analysis based on (1) location contact with COVID-19, (2) dose of HCQ, and (3) pre- vs. post-exposure prophylaxis. No heterogeinty available for subgroups. | Funnel plot was not assessed giving the small number of studies. |
| Million et al. (8) | 20/105,040 | HCQ effective on cough, duration of fever clinical cure death and viral shedding (OR = 0.19, 0.11, 0.21, 0.32, and 0.43). | Q-test for the all set of studies (Q = 51.8, p < 0.001). I2 ≥ 75% and significant Q-test for subset of studies when k > 2 studies (except for deaths, I2 = 0%, p = 0.071). No subset analysis based on heterogeneity. | None. |
| Patel et al. (39) | 6/2,908 | No difference between HCQ and control group on mortality (OR = 1.25, [0.65, 2.38]). Higher mortality in HCQ/AZ group compared to control (OR = 2.34, [1.63–3.34]). | There was significant heterogeneity in mortality outcome (I2 = 80%) for HCQ. Subgroup analysis based on sensitivity analysis. For the HCQ/AZ groups, there was perfect homogeneity (I2 = 0%). | Funnel plot was asymmetrical. Subgroup analysis based on homogeneous studies. |
| Pathak et al. (40) | 7/4,984 | No difference in outcome with/without hydroxychloroquine (OR = 1.11, [0.72, 1.69]). | Moderate heterogeneity (32% ≤ I2 ≤ 44%). No subgroup analysis. | Funnel Plot and Egger regression asymmetry test (although not available in the paper). |
| Putman et al. (41) | 45/6693 | HCQ use was not significantly associated with mortality (HR = 1.41, [0.83, 2.42]). | Low heterogeneity (I2 = 0–32%) but small set of studies (k = 2 or 3). | Newcastle-Ottawa Scale for cohort studies and the Risk of Bias 2.0 tool for randomized controlled trials; case series assumed to be high risk by default. |
| Sarma et al. (9) | 7/1,358 | No differences on viral cure (OR = 2.37, [0.13–44.53]), death/clinical worsening (OR = 1.37, [1.37–21.97]) or safety (OR = 2.19, [0.59–8.18]). | Heterogeneity varies from null (for safety issues) to high (I2 = 72%, for virological cure). No subset analysis (except for the inclusion/exclusion of Gautret et al. [22]). | Cochrane/ROBINS-I/Newcastle Ottawa Scale (3 researchers). |
| Shamshirian et al. (42) | 37/45,913 | No difference on mortality in HCQ group (RR = 0.86, [0.71–1.03]) or HCQ/AZ comparison (RR = 1.28, [0.76–2.14]). | High heterogeneity (I2 = 87–90%). Meta-regressions indicated significant effect of age (p < 0.001). | Moderate publication bias for mortality based on Egger's test (p = 0.02). |
| Singh et al. (43) | 7/746 | No benefits of HCQ on viral clearance (RR, 1.05; 95% CI, 0.79 to 1.38; p = 0.74). Significantly more deaths in the HCQ group compared to the control group (RR, 2.17; 95% 1.32 to 3.57; p = 0.002). | Moderate heterogeneity in the clearance analysis (I2 = 61.7%, p = 0.07) and none in the death analysis (I2 = 0.0%, p = 0.43). No subset analysis based on heterogeneity. | Trim and fill adjustment, rank correlation, and Egger's tests. |
| Ullah et al. (44) | 12/3,912 | Higher mortality (OR = 2.23, [1.58–3.13]) and net adverse events (OR = 4.59, [1.73–12.20]) in HCQ group compared to control. | Moderate to high heterogeneity (I2 = 54–94%) without subgroup analysis. | Funnel plot revealed minimal publication bias. |
| Yang et al. (45) | 9/4,112 | HCQ-azithromycin combination increased mortality in COVID-19 patients (OR = 2.34;, [1.63–3.36]) though it was also associated with benefits on viral clearance in patients (OR = 27.18, [1.29–574.32]). HCQ-alone did not reveal significant changes in mortality rate, clinical progression, viral clearance, and cardiac QT prolongation. | Null to high heterogeneity (I2 = 84%). Subsequent subgroup analysis showed that HCQ treatment could recuded mortality and severe illness in severely infected COVID-19 patients (OR = 0.27, [0.13–0.58]). | Funnel plot analysis did not reveal obvious publication bias. Possible bias due to lack of demographic and clinical data. |
| Zang et al. (46) | 7/851 | No difference in illness duration between the HCQ group and the standard treatment group (RR = 0.66, [0.18–2.43]). Death was higher in HCQ group compared to standard (RR = 1.92, [1.26–2.93]). | Moderate heterogeneity was observed (41.2% ≤ I2 ≤ 72.1%) without subgroup analysis. | Cochrane Risk of Bias Tool for RCTs evaluated quality of studies (2 reviewers). Newcastle Ottawa Scale for observational studies and Egger test. |
To date, 24 meta-analyses were published on April 11th, 2021. This table only described peer-reviewed meta-analyses evaluating HCQ efficacy on COVID-19 patients. We reported the number of studies (k) and participants (N) after exclusion/inclusion criteria.