RETRACTED – Re: No benefit from hydroxychloroquine, with or without macrolide antibiotics in analysis of 96,000 patients
1 June 2020. Related: COVID-19: investigational drugs, COVID-19.
STOP PRESS: On 2 June the Lancet issued a Letter of Concern related to the paper reported below, specifically on needing to confirm the dataset used in the analysis. Similar issues have been raised in an open letter sent to the NEJM as did our own comments below the article. This article was left online but should not be relied on until these issues are further investigated. The WHO SOLIDARITY study has now reopened its HCQ arm. [8, 9, 10]
On 4 June, the Lancet published an update that three of the study authors retracted the study for publication due to lack of access to the databases needed for the investigation. 
After more than 1100 deaths, the UK RECOVERY study has now stopped the HQC arm. 
Simon Collins, HIV i-Base
On 22 May 2020, a large retrospective international meta-analysis published in the Lancet failed to show a benefit of either hydroxychloroquine (HCQ) or chloroquine (CQ) for treating COVID-19, with or without a macrolide antibiotic (generally azithromycin or clarithromycin) but did report increased risk of side effects. 
The results are important given the extensive ongoing studies using HCQ (largely based on variable results from small uncontrolled studies), especially since positive results have now been reported for remdesivir in a large placebo controlled study. 
Off-label use of HCQ has also been reported, including as prophylaxis for COVID-19, and stockpiling drugs have led to shortages for people with approved indications.
This analysis included results from more than 96,000 people hospitalised with PCR-confirmed COVID-19 between 20 December 2019 and 14 April 2020, and involved 671 hospitals in six continents.
The analysis included 14,888 people in four treatments (started within 48 hours): chloroquine alone (n=1868), chloroquine with a macrolide (n=3783), hydroxychloroquine alone (n=3016), or hydroxychloroquine with a macrolide (n=62210. This left 81,144 in the control group. Main outcomes included time in hospital for efficacy against COVID-19 and new ventricular arrhythmias as a safety measure. Starting treatment when on mechanical ventilation and use of remdesivir were exclusion criteria.
Baseline characteristics include mean age 53.8 years and 53.7% were men. Mean BMI was 27.6 kg/m2 (SD +/–5·5) and 30.7% were obese (BMI ≥30). The mean length of stay in hospital was 9.1 days (SD 6.4), with an overall in-hospital mortality of 11·1%.
Geographically, participants were from North America (65.9%), Europe (17.3%), Asia (7.9%), Africa (4.6%), South America (3.7%), and Australia (0.6%).
Overall, 10,698 (11.1%) people died in hospital. In multivariate analysis, controlling for age, sex, race or ethnicity, body-mass index, underlying cardiovascular disease and its risk factors, diabetes, underlying lung disease, smoking, immunosuppressed condition, and baseline disease severity) all four treatment groups had significantly higher rates of in-hospital mortality and new ventricular arrhythmias compared to the control group, see Table 1.
The commonly reported risk factors associated with poor outcomes to COVID-19 were significantly associated with higher risk of mortality, and with use of all four treatment groups, hence the importance of adjusted analysis.
Table 1: risk of in-hospital mortality with COVID-19
|Treatment arm||Mortality rate||HR
|New ventricular arrhythmias||HR
(1.223 to 1.457)
(1.935 to 2.900)
|HCQ + macrolide||23.8%||1·447
(1.368 to 1.531)
(4.106 to 5.983)
(1.218 to 1.531)
(2.760 to 4.596)
|CQ + macrolide||22·2%||1.368
(1.273 to 1.469)
(3.344 to 4.812)
Other independent predictors of higher rates of in-hospital mortality included:
- Black race HR: 1.344 (95%CI: 1.276 to 1.415).
- Hispanic race HR: 1.495 (95%CI: 1.400 to 1.597).
- Congestive heart failure HR: 1.756 (95%CI: 1.609 to 1.915).
- Arrhythmia HR: 1.626 (95%CI: 1.504 to 1.758).
- Oxygen saturation (SPO2) <94% HR: 1·664 (95%CI: 1.587 to 1.746).
Protective factors associated with a reduced risk included:
- Asian race HR: 0.717 (95%CI: 0.668 to 0.769).
- Use of an ACE inhibitor HR: 0.566 (95%CI: 0.514 to 0.624).
- Use of a statin HR: 0.793 (95%CI: 0.736 to 0.855) and
- qSOFA* <1 HR: 0.758 (95%CI: 0.726 to 0.792.
* Quick sepsis-related organ failure assessment
Independent predictors of ventricular arrythmia included:
- Coronary artery disease HR: 1.830 (95%CI: 1.613 to 2.076).
- Congestive heart failure HR: 3.914 (95%CI: 3.283 to 4.665).
- History of cardiac arrhythmia HR: 4.119 (95%CI: 3.525 to 4.812).
- COPD HR: 1.585 (95%CI: 1.256 to 2.001).
The discussion included a data review of other studies, largely reporting similar lack of benefit.
Although the investigators noted limitations from observational data they concluded that this large-scale, international, real-world analysis supports the absence of a clinical benefit of chloroquine and hydroxychloroquine and points to potential harm in hospitalised patients with COVID-19.
They suggested that these drug regimens should not be used outside of clinical trials and that urgent confirmation from randomised clinical trials is needed.
Although this article was peer reviewed and published in the Lancet, the study has also been criticised (note: before the retraction) for methodology relating to the dataset and for suggestions that finding such a large safety effect linked to treatment might be linked to unadjusted confounding.
For example, that in the context of patients hospitalised for COVID-19, HCQ or CQ could easily have been more readily prescribed for people with the most rapid deterioration when management of cardiovascular event might also have been less optimum. Also, while presented as a collaborative large international study, the paper is authored by four commercial researchers who are not directly connected to any of the datasets that are included.
However, its advantages include the size of the dataset and that it is representative of people in many different countries.
Also, a growing number of studies are now questioning the use of HCQ or CQ for COVID-19 based on unlikely efficacy or increased risk of toxicity.
Studies questioning efficacy include a recent article (ahead of peer review) that showed activity in in-vitro and animal studies but did not find data to support antiviral or clinical efficacy of HCQ as either treatment or PrEP. 
The large randomised UK RECOVERY study (with more than 10,000 participants) currently being run in the UK has published MHRA support to still continue the HCQ arm.  [Note: this decision has since been reversed]. 
The large international WHO SOLIDARITY trial (called DISCOVER in Europe) has just closed its HCQ arm. 
A paper focused on the risk of serious toxicity based on drug levels from intentional overdose studies reported that peak concentrations >13 umol/L (95%CI: 10 to 16) would be associated with >1% mortality. 
Use of an adult dose of 600 mg twice-daily for 10 days results in peak concentrations >10 umol/L in >60% of adults weighing 70 kg. It also notes that among more than 90 ongoing HCQ of CQ studies for COVID-19, only 0.2% adults weighing >70 kg in other high-dose studies would be expected to achieve peak drug levels >10 umol/L.
This paper reports that the high dose arm (600mg base chloroquine twice daily for ten days) of the Brazilian study that was recently stopped due to serious high toxicity, represented the standard malaria loading dose repeated 19 times at 12 hour intervals. 
It also suggests that there may have been confusion between salt and base weights. The Chinese guidelines on which the Brazilian study was based recommended 500 mg salt twice daily (two tablets of 250 mg, comprising 155 mg base each).
This report was first published on 23 May 2020 and updated on 2 June 2020 (over the Letter of Concern), on 4 June 2020 (after the retraction) and on 6 June following RECOVERY study announcement.
- Mehra MR et al. Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis. The Lancet. DOI: 10.1016/ S0140-6736(20)31180-6. (22 May 2020).
- Collins S. Remdesivir improves recovery time in early COVID-19 infection: first definitive results of benefit. HTB (22 May 2020).
- Maisonnasse P et al. Hydroxychloroquine in the treatment and prophylaxis of SARS-CoV-2 infection in non-human primates. NatureReseach. Pre-peer-review article. 10.21203/rs.3.rs-27223/v1. (6 May 2020).
- RECOVERY trial statement. Recruitment to the RECOVERY trial (including the Hydroxychloroquine arm) REMAINS OPEN. (22 May 2020).
- WHO SOLIDARITY trial (DISCOVERY trial in Europe).
- Watson JA et al. Concentration-dependent mortality of chloroquine in overdose. Before peer review. DOI: 10.1101/2020.04.24.20078303. (29 April 2020).
- Borba M et al. Chloroquine diphosphate in two different dosages as adjunctive therapy of hospitalized patients with severe respiratory syndrome in the context of coronavirus (SARS-CoV-2) infection: Preliminary safety results of a randomized, double-blinded, phase IIb clinical trial (CloroCovid-19 Study). medRxix doi: 10.1101/2020.04.07.20056424.
- Lancet Press Office. Expression of Concern – Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis. (2 June 2020).
journals/lancet/article/ PIIS0140-6736(20)31180-6/ fulltext
- Watson JA et al (on behalf of 174 signatories). An open letter to Mehra et al and The New England Journal of Medicine. (2 June 2020).
- Offerd C. Concerns build over Surgisphere’s COVID-19 dataset. The Scientist Daily
- Lancet retraction: “Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis” (4 June 2020).
- Collins S. Disastrous UK RECOVERY study stops hydroxychloroquine (HCQ) for COVID-19: more than 1100 deaths question ethics and safety overall. HTB (6 June 2020).