Background The global push for the use of hydroxychloroquine (HCQ) and chloroquine (CQ) against COVID-19 has resulted in an ongoing discussion about the effectivity and toxicity of these drugs. Recent studies report no effect of (H)CQ on 28-day mortality. We investigated the effect of HCQ and CQ in hospitalized patients on the non-ICU COVID-ward. Methods A nationwide, observational cohort study was performed in The Netherlands. Hospitals were given the opportunity to decide independently on the use of three different COVID-19 treatment strategies: HCQ, CQ, or no treatment. We compared the outcomes between these groups. The primary outcomes were 1) death on the COVID-19 ward, and 2) transfer to the intensive care unit (ICU). Results The analysis included 1064 patients from 14 hospitals: 566 patients received treatment with either HCQ (n = 189) or CQ (n = 377), and 498 patients received no treatment. In a multivariate propensity-matched weighted competing regression analysis, there was no significant effect of (H)CQ on mortality on the COVID ward. However, HCQ was associated with a significantly decreased risk of transfer to the ICU (hazard ratio (HR) = 0.47, 95% CI = 0.27–0.82, p = 0.008) when compared with controls. This effect was not found in the CQ group (HR = 0.80, 95% CI = 0.55–1.15, p = 0.207), and remained significant after competing risk analysis. Conclusion The results of this observational study demonstrate a lack of effect of (H)CQ on non-ICU mortality. However, we show that the use of HCQ — but not CQ — is associated with a 53% reduction in risk of transfer of COVID-19 patients from the regular ward to the ICU. Recent prospective studies have reported on 28-day, all-cause mortality only; therefore, additional prospective data on the early effects of HCQ in preventing transfer to the ICU are still needed.
After the emergence of SARS-CoV-2 in December 2019, the new coronavirus spread around the world, resulting in a pandemic. Unfortunately, there is still no proven effective drug or vaccine available against COVID-19, and hospitalized patients with COVID-19 are at high risk for admission to the ICU (10–20%), with 3–10% of patients requiring intubation, and 2–5% of patients dying
Among the drug candidates for treating COVID-19 are hydroxychloroquine (HCQ) and chloroquine (CQ) ). Insights into the underlying mechanisms of action of HCQ and CQ are still emerging. Both drugs have a large volume of distribution ). Their molecular structures are comparable, except that HCQ has an extra hydroxyl group. Both interfere with lysosomal activity and decrease membrane stability, reduce signaling pathways for Toll-like-receptors 7 and 9, and impact on transcriptional activity, inhibiting cytokine production.
There are only a few differences between the drugs, of which the most important is drug clearance
Some observational studies on the efficacy of (H)CQ report clinical benefits and antiviral effects (Gao et ). A few small, controlled trials have been inconclusive (Tang et al., 2020 , Chen et al., 2020 ). The Recovery study included 176 UK hospitals, comprising 1395 patients receiving high doses of HCQ (9200 mg cumulative dose), and reported no beneficial effects on all-cause mortality at 28 days (26.8% of treated patients versus 25% of controls) (Horby, 2020 ). The risk of admission to the ICU could not be calculated, since 17–60% of patients were already on (non-invasive) ventilation at randomization. A recent systematic review and meta-analysis, including 11 932 patients on HCQ, found that its use was not associated with reduced mortality (pooled relative risk of RCTs for HCQ use of 1.09)
Results of other prospective trials are not expected, since the European Discovery and the WHO Solidarity trials have discontinued their HCQ treatment arms because of lack of effect on mortality. Meanwhile, the US FDA and the Infectious Diseases Society of America (IDSA) advise against the use of (H)CQ outside the context of a clinical trial (Swank and McCarten, 2020 , Infectious Diseases Society of America Guidelines, 2020
Based on the available evidence present at the start of the outbreak, a Dutch treatment guideline was developed (RIVM, 2020 Off-label use of both HCQ and CQ was offered as a treatment option; however, the guidelines did not endorse either treatment in particular. Consequently, hospitals decided independently on a treatment protocol with either HCQ or CQ, or to give no treatment. This policy created a unique situation for comparing the efficacy of HCQ and CQ with no treatment in hospitalized non-ICU patients, with a reduction of potential bias by indication.
The study was designed as an observational, multicenter, cohort study of hospitalized COVID-19 patients. Before the first patients were admitted, Dutch hospitals independently implemented a treatment protocol with or without (H)CQ. As a consequence, Dutch patients were geographically allocated to their local hospital with or without the intention to treat with (H)CQ. Eligible patients were included retrospectively over the period from February 28 to April 1, 2020. Patients were followed up until they reached one of the clinical endpoints: (1) discharge for cured infection to home or rehabilitation center; (2) transfer from the COVID ward to the intensive care unit (ICU); or (3) death, either during their hospital stay on the ward (non-ICU) or following transfer to a hospice facility. Secondary outcomes were the effects of the use of azithromycin (AZM) and angiotensin-receptor blockers (ARB) on outcome.
All hospitals in The Netherlands were considered eligible to participate in the study, including academic hospitals as well as non-teaching hospitals. These hospitals were asked to participate early in the outbreak. All participating hospitals shared their data with the coordinating hospital (Isala, Zwolle), where the statistical analysis was performed. Data-sharing agreements were signed, and the Medical Ethics Review Committee (METC) of Isala approved a waiver for informed consent.
Inclusion and exclusion criteria were designed to select a study sample of hospitalized patients with moderate to severe COVID-19. New confirmed COVID-19 cases were included if they were aged > 18 years and if they were admitted to the emergency department (ED) and subsequently hospitalized on the non-ICU hospital COVID-19 ward. Exclusion criteria were age < 18 years, admission to the ICU, or death within 24 h after presentation at the ED. Patients transferred between Dutch hospitals, for example due to capacity issues, were also excluded. Confirmed COVID-19 infection was defined as either positive SARS-CoV-2 real-time reverse transcriptase polymerase chain reaction (PCR) on swab material, sputum, or bronchoalveolar lavage samples (Corman et al., 2020 ), or typical findings on chest computed tomography (CT). Typical CT findings were defined as CO-RAD 4–5, using the CO-RAD classification system (COVID-19 Reporting and Data System, developed by the Dutch Radiology Society to describe levels of suspicion for COVID-19 infection) (Prokop et al., 2020 ). Routine blood tests were carried out for hematological and biochemical analysis, according to standard hospital laboratory techniques. Since the use of (H)CQ for COVID-19 was off-label, patients were started on (H)CQ only after giving informed consent.
Data were extracted from Electronic Health Records (EHR) in all participating hospitals by medical students and/or infectious disease (ID) physicians. Data were collected on site using a standardized data-collection form on a secured website of the coordinating hospital. Patient data were immediately anonymized and encoded upon entry into the online research manager program. Collected data included patient characteristics, such as comorbidities, registered ICU-restrictive policy by treating physician, routine laboratory results, SARS-Cov2-PCR and chest CT-scan results, medical treatment before admission, and antibiotic treatment during hospitalization.
Differences between HCQ and CQ users (cases) and non-users (controls) were compared using χ2 statistics or the Fisher exact test for categorical variables, and the independent t-test or Mann-Whitney U test for continuous variables. The data were analyzed within a Cox proportional hazard regression framework. Follow-up commenced from the date of hospital admission and ended on the dates of death or ICU admission, and patients were censored at the time they were discharged from hospital. Hazard ratios were calculated for (H)CQ use in relation to the primary endpoints of death and ICU admission, or a combination of these endpoints denoted as a composite adverse endpoint. Death and ICU admission are competing risk events; therefore, competing risk regression analysis was conducted for these two endpoints according to the method developed by Fine and Gray, 1999 . Instead of KM survival curves, survival data were summarized using the cumulative incidence function (CIF) or cumulative risks of an event, which indicate the probability of the event at a given time. The proportional hazards assumption was confirmed by Schoenfeld’s global test and inspection of log (―log [survival]) curves. Propensity score (PS) matching was used for making causal inferences for the treatment on the clinical outcome. A set of pre-test covariates that were associated with the treatment was selected and PS scores were estimated using logistic regression, with treatment as the outcome measure. Separate PS-matched Cox regression models with and without adjustment for potential confounders were used (see Appendix), but only the results of the overall and inverse-probability-of-treatment-weighted (IPTW) Cox regression analysis are shown. Analyses were adjusted for gender, age, comorbidity CVA, comorbidity diabetes, comorbidity asthma/COPD, use of broad-spectrum antibiotics, therapeutic anticoagulation, prophylactic anticoagulation, first day at ED, and ICU restriction. The combined endpoint risk regression analyses were stratified by ICU restriction, because of the distinctive patient characteristics in this group. For PS estimation and matching the PS matching R package in SPSS and the PSMATCH2 package in Stata were used. All tests were two-sided and p < 0.05 was considered statistically significant. Statistical analyses were performed using SPSS version 24.0 and the STATA version 14 statistical package (Stata Corporation, College Station, TX).
Inclusion and baseline characteristics
Between February 28 and April 1, 2020, 1130 patients admitted to the 14 participating hospitals in The Netherlands met the inclusion criteria; 1106 patients were eligible for inclusion. After propensity score matching the analytic cohort consisted of 1064 patients, comprising 566 (53.2%) treated patients, both with HCQ (N = 189; 17.8%) and CQ (N = 377; 35.4%), and 498 (46.8%) untreated controls (see Figure 1).
This study demonstrated a new and clinically important finding: the use of HCQ on the COVID-19 ward is associated with a decreased risk of transfer to the ICU. After competing risk analysis, the risk of admission to the ICU was reduced by 53%. This finding suggests that starting early treatment with HCQ (within 1 day of admission) on the regular COVID ward might prevent progression to critical respiratory illness. This is consistent with the suggestion that HCQ treatment reduces the risk of disease progression more effectively earlier in the course of the disease . This holds true for many other viral infections, such as influenza and herpes simplex, where treatment must be initiated soon after onset of symptoms in order to confer benefit. However, treatment with HCQ before onset of symptoms did not prevent COVID-19, as was demonstrated in a randomized controlled trial investigating postexposure use of HCQ (Boulware et al., 2020
Second, we could not demonstrate a significant effect of treatment with HCQ or CQ on on-ward mortality. One of the strengths of our study was that we selected a clearly defined cohort of patients on the regular non-ICU COVID-ward, thus our results reflected mortality before transfer to the ICU only. In recent literature, evidence is accumulating that there is no beneficial effect of HCQ on mortality. Mortality numbers in systematic reviews and in prospective HCQ studies, such as the Recovery trial, are frequently reported in terms of 28-day all-cause mortality, and do not differentiate between on-ward mortality and mortality after transfer to the ICU (Horby, 2020 , Fiolet et al., 2020 ).
In our study, there was no significant difference in outcome between patients treated with AZM, nor in patients on ACEi or ARB therapy. It has been suggested that ARB therapy increases susceptibility to COVID-19, but other studies report conflicting results (Zhang et al., 2020 , Fosbøl et al., 2020 ). Our data confirm the lack of effect on outcome, both for pre-hospital use as well as for in-hospital continuation of ACEi or ARB therapy.
Surprisingly, we found a differential effect of HCQ and CQ on COVID-19, while in the literature these drugs are frequently reported in terms of a composite outcome. There are several possible explanations for this differential effect. The first explanation is a possible difference in pharmacokinetics between both drugs. There is a substantial difference in renal drug-clearance – 51% in CQ and 21% in HCQ (Schrezenmeier and Dörner, 2020 ). Furthermore, the distribution volumes of HCQ and CQ are different; HCQ has a volume of distribution of 5522 liters (whole blood), as compared with 14 000–56 000 liters for CQ (Drugbank, 2020a , Furst, 1996 , Drugbank, 2020b ). It is still a matter of debate whether the 4-aminoquinoline drugs have anti-viral activity or immunomodulating properties (Gautret et al., 2020 , Maisonnasse et al., 2020 ). The immunomodulating effect of HCQ in has been reported in rheumatology literature (Schrezenmeier and Dörner, 2020 ). In clinical practice, patients with rheumatoid disease are treated with HCQ but not CQ as anti-inflammatory therapy, according to clinical guidelines (Arayssi et al., 2018 ). It is conceivable that the beneficial effect of early HCQ in COVID-19 lies in the reduction of localized inflammation in the lung. This is supported by the results of a recent observational study, which indicated that the use of moderate-dose systemic corticosteroids on the general ward lowered the hazard of ICU transfer (Majmundar et al., 2020 ).
Another important strength of our study was the random distribution of patients between hospitals with different treatment protocols. Unintentionally, three groups of patients were created, almost as in prospective research. We were able to investigate the differences between patients on or off treatment with a reduced risk of bias by indication.
Credited to A.J.J. Lammers