Low-dose hydroxychloroquine therapy and mortality in hospitalised patients with COVID
Low-dose hydroxychloroquine therapy and mortality in hospitalised patients with COVID-19: a nationwide observational study of 8075 participants
Author links open overlay panelLucyCatteaua1NicolasDaubybcd1
MarionMontourcyaEmmanuelBottieaueJorisHautekietafElsGoetghebeurfSabrinavan IersselgElsDuysburghaHermanVan OyenahChloéWyndham-ThomasaDominiqueVan BeckhovenaBelgian Collaborative Group on COVID-19 Hospital Surveillance2
https://doi.org/10.1016/j.ijantimicag.2020.106144Get rights and content
Highlights
•Hydroxychloroquine (HCQ) 2400 mg over 5 days was used in Belgium for COVID-19.
•Impact of HCQ on mortality among 8075 patients with COVID-19 was assessed.
•Lower mortality in HCQ-treated patients as compared to supportive care.
•Lower mortality was irrespective of symptom duration.
ABSTRACT
Hydroxychloroquine (HCQ) has been largely used and investigated as therapy for COVID-19 across various settings at a total dose usually ranging from 2400 mg to 9600 mg. In Belgium, off-label use of low-dose HCQ (total 2400 mg over 5 days) was recommended for hospitalised patients with COVID-19. We conducted a retrospective analysis of in-hospital mortality in the Belgian national COVID-19 hospital surveillance data. Patients treated either with HCQ monotherapy and supportive care (HCQ group) were compared with patients treated with supportive care only (no-HCQ group) using a competing risks proportional hazards regression with discharge alive as competing risk, adjusted for demographic and clinical features with robust standard errors. Of 8075 patients with complete discharge data on 24 May 2020 and diagnosed before 1 May 2020, 4542 received HCQ in monotherapy and 3533 were in the no-HCQ group. Death was reported in 804/4542 (17.7%) and 957/3533 (27.1%), respectively. In the multivariable analysis, mortality was lower in the HCQ group compared with the no-HCQ group [adjusted hazard ratio (aHR) = 0.684, 95% confidence interval (CI) 0.617–0.758]. Compared with the no-HCQ group, mortality in the HCQ group was reduced both in patients diagnosed ≤5 days (n = 3975) and >5 days (n = 3487) after symptom onset [aHR = 0.701 (95% CI 0.617–0.796) and aHR = 0.647 (95% CI 0.525–0.797), respectively]. Compared with supportive care only, low-dose HCQ monotherapy was independently associated with lower mortality in hospitalised patients with COVID-19 diagnosed and treated early or later after symptom onset.
Previous article in issue
Next article in issue
Keywords
Hydroxychloroquine
COVID-19
SARS-CoV-2
Mortality
Observational study
1. Introduction
There is currently no robust antiviral or immunomodulatory treatment for infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of coronavirus disease 2019 (COVID-19). Chloroquine (CQ), an antimalarial drug, has been shown to have in vitro antiviral properties both against SARS-CoV and SARS-CoV-2 by different mechanisms [1], [2], [3], [4]. It has also been hypothesised that CQ could have a positive impact on COVID-19 outcome through immunomodulatory properties [5,6]. Hydroxychloroquine (HCQ), a derivative of CQ, has a long clinical track record as a treatment for malaria and inflammatory diseases such as systemic lupus erythematous and rheumatoid arthritis, with a favourable safety profile in acute and chronic use [7]. Both CQ and HCQ were selected by the World Health Organization (WHO) for potential repurposing for COVID-19. Early during the amplification phase of the epidemic in Belgium, and pending results of clinical trials, off-label administration of a ‘low-dose’ regimen of HCQ sulphate in monotherapy (400 mg twice on Day 1, followed by 200 mg twice a day from Days 2 to 5, i.e. a total dose of 2400 mg) was recommended as an acceptable immediate treatment option for hospitalised COVID-19 patients [8]. This guidance, officially released on 13 March 2020, was based on the following considerations: (i) HCQ was the only drug with demonstrated in vitro effect against SARS-CoV-2 available in Belgium at that time; (ii) HCQ exhibited a superior in vitro antiviral effect in comparison with CQ, likely explained by the higher accumulated intracellular drug concentrations [9]; (iii) limited pharmacokinetic data suggested that the selected dosage should have sufficient antiviral activity [10]; (iv) chronic administration of HCQ for rheumatological disorders has not been associated with major safety signals over decades of use; (v) restricting HCQ use to well-selected COVID-19 patients monitored at hospitals appeared as a reasonable risk/benefit compromise considering the well-known dose-dependent cardiotoxicity of the drug; and (vi) it was advised to Belgian hospitals to administer this off-label regimen whenever possible within clinical studies. Of note, azithromycin (AZM) and systemic use of corticosteroids were not recommended in the guidance [8]. Simultaneously, Sciensano, the Belgian Scientific Institute of Public Health, initiated a national surveillance of COVID-19 hospitalised patients that included treatments and outcomes among its variables, enabling the clinical surveillance of drug use and outcome.
So far, the impact of HCQ on the outcome of SARS-CoV-2 infection in humans remains undetermined. An increasing number of single-centre and multicentre retrospective studies using various HCQ dosages are being published with conflicting results [11], [12], [13], [14], [15], [16], [17], [18], [19]. Recently, the UK-based RECOVERY and WHO-led SOLIDARITY trials communicated that HCQ at the study dosage of 9200–9600 mg over 10 days provided no benefit in hospitalised patients with COVID-19 [20].
In the present study, we retrospectively assessed the association between HCQ monotherapy and in-hospital mortality in a nationwide registry of 8075 COVID-19 patients. Next, the impact of HCQ treatment on mortality was investigated according to the time between symptom onset and COVID-19 diagnosis.
2. Methods
2.1. Data collection
Sciensano's data collection of patients hospitalised with confirmed COVID-19 was initiated on 14 March 2020, 2 weeks after the first symptomatic case was reported in Belgium, and systematic registering was strongly encouraged by health authorities. Two independent online secured questionnaires in LimeSurvey (LimeSurvey GmbH, Hamburg, Germany) were made available: one with information after admission and the second after discharge. Information collected at admission included sociodemographic characteristics, clinical presentation, co-morbidities, chronic treatment with renin–angiotensin–aldosterone system inhibitors and diagnostic workup. Data collected at discharge included COVID-19 treatment details (antiviral and immunomodulatory drugs, including date of initiation and termination, mode of administration), clinical and laboratory markers of disease severity during hospital stay, admission to the intensive care unit (ICU) and final outcome at hospital discharge (dead or discharged alive).
2.2. Hydroxychloroquine treatment
On 13 March 2020, a task force (ND, SVI and EB, affiliated to the national reference institutions for emerging infections), co-ordinated by Sciensano, published a guidance for the management of patients hospitalised with COVID-19. Based on the above-described rationale, the ‘low-dose’ HCQ regimen (2400 mg in total over 5 days) was recommended as a reasonable emergency therapeutic option for hospitalised patients and was centrally provided for free [8]. A set of warnings were provided on its use, including corrected QT (QTc) determination in all admitted patients and close cardiac monitoring in case of baseline QTc exceeding 450 ms and in all conditions that could favour arrhythmia (underlying cardiopathy, congenital or acquired QTc prolongation, electrolytic disturbances, or use of other drugs prolonging the QTc interval) [8]. Treatment initiation was advised as soon as a diagnosis was made, with information to the patient about the off-label use. The final treatment decision was, however, left to the discretion of the treating physician.
2.3. Inclusion and exclusion criteria
We analysed all COVID-19 cases for whom both admission and discharge questionnaires were reported up to 24 May 2020. The analysis was restricted to those confirmed before 1 May 2020 by reverse transcription PCR (RT-PCR) and/or rapid antigen test on respiratory samples, with exclusion of those diagnosed by pulmonary computed tomography (CT) scan only. The SARS-CoV-2 rapid antigen test used in Belgium has a specificity of 99.5% compared with RT-PCR in respiratory samples [21]. Children aged <16 years, pregnant and post-partum women as well as patients who were discharged (either alive or dead) within 24 h after hospital admission or before diagnosis confirmation were excluded. In addition, we removed from this analysis all patients having started any COVID-19-related treatment before symptom onset, including for other clinical indications, as well as those having a missing date of diagnosis. To compare patients treated with HCQ monotherapy and supportive care with those receiving only supportive care, any patients treated with another COVID-19-related treatment (macrolides, tocilizumab, lopinavir/ritonavir, remdesivir, atazanavir or anakinra), whether prescribed with or without HCQ, were also excluded.
2.4. Statistical analyses
Participants meeting the inclusion criteria were divided into two groups: (i) COVID-19 patients treated with HCQ monotherapy in addition to supportive care (HCQ group); and (ii) those receiving supportive care alone (no-HCQ group). Demographic characteristics, pre-existing conditions, laboratory parameters, clinical features and outcome were described first by discharge status (survivors versus non-survivors) and second by treatment group (HCQ versus no-HCQ). The χ2 test for categorical variables and the Wilcoxon test for continuous variables were used to assess differences between groups. We considered a P-value of <0.05 to be statistically significant.
Missing data among important prognostic baseline covariates were assumed to be missing at random, i.e. independent of the underlying missing values given the observed data. This was handled by ten-fold multiple imputation performed in R software through the MICE package v.3.8.0 [22]. A competing risks proportional hazards regression with robust standard errors allowing for clustering within hospitals (R package SURVIVAL v.3.1-12) was then used to analyse in-hospital death competing with alive discharge from hospital. Hazards of this in-hospital death thus dropped to zero post discharge alive. Cause-specific hazards of treatment effect were adjusted for the baseline covariates age, sex, co-morbidities (cardiovascular disease, arterial hypertension, diabetes mellitus, chronic renal, liver and lung diseases, neurological and cognitive disorders, immunosuppressive conditions, malignancies, obesity and smoking status), clinical features [pneumonia diagnosis, acute respiratory distress syndrome (ARDS), admission to ICU within the 24 h following admission and time from symptom onset to diagnosis] and baseline laboratory parameters of disease severity consisting of lactate dehydrogenase (LDH) ≥ 350 IU/L, C-reactive protein (CRP) ≥ 150 mg/L and partial pressure of oxygen (paO2) < 60 mmHg. As HCQ prescription decreased over time, the calendar time of diagnosis was also included in the model.
The propensity of HCQ treatment was estimated from those same baseline covariates (R package IPW v.1.0-11[23]). An inverse propensity-weighted standardised cumulative incidence of in-hospital death for each treatment was derived using R package RISCA v.0.8.2 [24]. The competing risks analysis was repeated for patients treated within or beyond 5 days of onset of symptoms. Sensitivity analyses were performed (Supplementary material): they considered additional adjustments in the model, missing data impact and possible immortal time bias associated with delayed treatment receipt. Analyses were performed in SAS Enterprise Guide 7.1 and in R 3.6.3.
2.5. Ethical and privacy considerations
The hospital data collection is being performed by Sciensano, the Belgian Scientific Institute of Public Health, legally entitled for surveillance of infectious diseases in Belgium (Royal Decree of 21/03/2018). This COVID-19 hospital surveillance was authorised by an independent administrative authority protecting privacy and personal data and was approved by the ethical committee of Ghent University Hospital.
3. Results
As recorded on 24 May 2020, the Sciensano database contained a total of 15 544 case records of COVID-19 patients (Fig. 1), originating from 109 Belgian hospitals. Among those, both admission and discharge report forms were received for 10 920 patients [3311 (21.3%) discharge forms were missing for patients with admission data and 1313 (8.4%) admission forms were missing for patients with discharge data]. After having excluding patients not meeting the inclusion criteria (Fig. 1), 8910 cases were included for the descriptive analysis.
Fig. 1. Data flow for coronavirus disease 2019 (COVID-19) patient selection for the observational cohort study. CT, computed tomography; HCQ, hydroxychloroquine.
Approximately 60% of the hospitalised COVID-19 patients were aged ≥65 years (Table 1). In general, participants were severely ill with more than 80% having radiological pneumonia, large proportions presenting with laboratory parameters of severity, including pronounced hypoxaemia, and 5.5% requiring immediate admission to the ICU. The median time from symptom onset to COVID-19 diagnosis was 5 days. Patients with incomplete discharge data (n = 2332) were similar to the study population in terms of age and sex distribution as well as the frequency of pre-existing conditions, except for the proportion of active smokers (Supplementary Table S1). In the univariate analysis, compared with survivors, non-survivors were older and were more likely to be male and to suffer from pre-existing conditions (Table 1). In addition, non-survivors presented more often with laboratory markers of disease severity such as high levels of LDH (≥350 IU/L) and CRP (≥150 mg/L) and severe hypoxaemia (paO2 < 60 mmHg). Time from symptom onset to diagnosis was shorter in non-survivors (median 3 days vs. 6 days in survivors; P < 0.0001). Length of hospital stay was similar in both groups.
Table 1. Characteristics of coronavirus disease 2019 (COVID-19) patients by survival or non-survival status during hospitalisation
Characteristic |
No./total no. (%) |
P-value |
|
Total (n = 8910) |
Survivors (n = 6981) |
Non-survivors (n = 1929) |
|
Demographic characteristics |
|
|
|
|
Age (years) |
|
|
|
|
16–30 |
149/8906 (1.7) |
149/6979 (2.1) |
0/1927 (0.0) |
<0.0001* |
31–44 |
607/8906 (6.8) |
596/6979 (8.5) |
11/1927 (0.6) |
|
45–64 |
2685/8906 (30.2) |
2503/6979 (35.9) |
182/1927 (9.4) |
|
65–79 |
2655/8906 (29.8) |
2017/6979 (28.9) |
638/1927 (33.1) |
|
≥80 |
2810/8906 (31.6) |
1714/6979 (24.6) |
1096/1927 (56.9) |
|
Median (IQR) age (years) |
71 (57–82) |
66 (54–79) |
82 (73–87) |
|
Male sex |
4807/8819 (54.5) |
3711/6919 (53.6) |
1096/1900 (57.7) |
0.0017± |
Pre-existing conditions |
|
|
|
|
Cardiovascular disease |
3084/8910 (34.6) |
2093/6981 (30.0) |
991/1929 (51.4) |
<0.0001± |
Arterial hypertension |
3622/8910 (40.7) |
2641/6981 (37.8) |
981/1929 (50.9) |
<0.0001± |
Diabetes mellitus |
1985/8910 (22.3) |
1442/6981 (20.7) |
543/1929 (28.1) |
<0.0001± |
Chronic renal disease |
1166/8910 (13.1) |
733/6981 (10.5) |
433/1929 (22.4) |
<0.0001± |
Chronic liver disease |
237/8910 (2.7) |
160/6981 (2.3) |
77/1929 (4.0) |
<0.0001± |
Chronic lung disease |
1353/8910 (15.2) |
976/6981 (14.0) |
377/1929 (19.5) |
<0.0001± |
Neurological disorders |
832/8910 (9.3) |
555/6981 (8.0) |
277/1929 (14.4) |
<0.0001± |
Cognitive disorders a |
1001/8338 (12.0) |
627/6539 (9.6) |
374/1799 (20.8) |
<0.0001± |
Immunosuppressive conditions |
248/8910 (2.7) |
191/6981 (2.7) |
57/1929 (3.0) |
0.6049± |
Malignancy |
|
|
|
|
Solid |
730/8910 (8.2) |
507/6981 (7.3) |
223/1929 (11.6) |
<0.0001± |
Haematological |
174 /8910 (2.0) |
118/6981 (1.7) |
56/1929 (2.9) |
0.0007± |
Obesity a |
545/5457 (10.0) |
450/4313 (10.4) |
95/1144 (8.3) |
0.0327± |
Current smoker |
407/4757 (8.6) |
312/3793 (8.2) |
95/964 (9.9) |
0.1064± |
Medications |
|
|
|
|
ACE inhibitor |
1368/8907 (15.3) |
1030/6979 (14.8) |
338/1928 (17.5) |
0.0028± |
Angiotensin receptor blocker |
806/8907 (9.0) |
604/6979 (8.7) |
202/1928 (10.5) |
0.0135± |
COVID-19 treatments |
|
|
|
|
Supportive care only |
3533/8910 (39.6) |
2576/6981 (36.9) |
957/1929 (49.6) |
<0.0001± |
HCQ |
4542/8910 (51.0) |
3738/6981 (53.5) |
804/1929 (41.7) |
<0.0001± |
HCQ + macrolides |
761/8910 (8.5) |
617/6981 (8.5) |
144/1929 (7.5) |
0.0561± |
Lopinavir/ritonavir |
12/8910 (0.1) |
7/6981 (0.1) |
5/1929 (0.3) |
0.2358± |
HCQ + lopinavir/ritonavir |
18/8910 (0.2) |
10/6981 (0.1) |
8 /1929 (0.4) |
0.0504± |
HCQ + tocilizumab |
17/8910 (0.2) |
12/6981 (0.2) |
5/1929 (0.3) |
0.4367± |
HCQ + tocilizumab + macrolides |
7/8910 (0.1) |
5/6981 (0.1) |
2/1929 (0.1) |
0.6565± |
HCQ + remdesivir |
4/8910 (0.0) |
2/6981 (0.0) |
2/1929 (0.1) |
0.1685± |
Others |
16/8910 (0.2) |
14/6981 (0.2) |
2/1929 (0.1) |
0.3738± |
Laboratory parameters |
|
|
|
|
LDH (IU/L) (median (IQR) [no.]) |
343 (258–477) [7385] |
329 (251–459) [5909] |
394 (288–548) [1476] |
<0.0001* |
LDH ≥ 350 IU/L |
3563/7385 (48.2) |
2663/5909 (45.1) |
900/1476 (61.0) |
<0.0001± |
CRP (mg/L) (median (IQR) [no.]) |
62 (26–118) [8624] |
55.9 (21.8–108.2) [6802] |
91.2 (44.4–162) [1822] |
<0.0001* |
CRP ≥ 150 mg/L |
1487/8624 (17.2) |
973/6802 (14.3) |
514/1822 (28.2) |
<0.0001± |
paO2 (mmHg) (median (IQR) [no.]) |
66 (57–76) [6013] |
67 (70–77) [4713] |
61 (52–73) [1300] |
<0.0001* |
paO2 < 60 mmHg |
1834/6013 (30.5) |
1221/4713 (25.9) |
613/1300 (47.2) |
<0.0001± |
Clinical features |
|
|
|
|
Pneumonia b |
7184/8567 (83.9) |
5545/6710 (82.6) |
1639/1857 (88.2) |
<0.0001± |
ARDS |
1197/8423 (14.2) |
601/6710 (9.0) |
596/1713 (34.8) |
<0.0001± |
Invasive ventilation support |
736/8691 (8.5) |
367/6810 (5.4) |
369/1881 (19.6) |
<0.0001± |
Admission to ICU within 24 h after admission |
488/8900 (5.5) |
298/6974 (4.3) |
190/1926 (9.9) |
<0.0001± |
Time from symptom onset to diagnosis (days) (median (IQR) [no.]) |
5 (2–9) [8097] |
6 (2–9) [6393] |
3 (1–7) [1704] |
<0.0001* |
Length of hospital stay (days) (median (IQR) [no.]) |
9 (5–15) [8894] |
9 (5–15) [6970] |
9 (5–16) [1924] |
0.9320* |
IQR, interquartile range; ACE, angiotensin-converting enzyme; HCQ, hydroxychloroquine; LDH, lactate dehydrogenase; CRP, C-reactive protein; paO2, partial pressure of oxygen; ARDS, acute respiratory distress syndrome; ICU, intensive care unit.
NOTE: All of the pre-existing conditions and COVID-19 features were reported as assessed by the clinician.
aMissingness is due to later onset of data collection.
bDiagnosis by imaging [chest radiography and/or computed tomography (CT) scan].
⁎Wilcoxon test.
±χ2 test.
After further exclusion of patients who received alternative COVID-19 treatments either with (n = 818; including macrolides, n = 761) or without HCQ (n = 17), the comparative analysis was restricted to 8075 subjects: 4542 in the HCQ group and 3533 in the no-HCQ group (Fig. 1). Of the HCQ-treated patients, 78.2% initiated the treatment within 24 h after diagnosis.
As shown in Table 2, COVID-19 patients in the HCQ group were younger and male sex was predominant. Several co-morbidities were significantly less frequent in the HCQ group, including cardiovascular diseases, arterial hypertension, chronic renal disease, neurological and cognitive disorders, solid cancer and obesity, as well as the proportion of active smokers. On the other hand, at admission, patients in the HCQ group appeared to be sicker as reflected by a higher frequency of radiological pneumonia, ARDS, ICU transfer within the 24 h after admission and invasive ventilation support as well as a higher frequency of elevated LDH and CRP levels. The case fatality rate of the study population was 21.8% (1761 deaths/8075 patients) but was lower in the HCQ group (804/4542; 17.7%) than in the no-HCQ group (957/3533; 27.1%) (P < 0.001). Incidental use of steroids was very low in both groups, although it was slightly higher in the HCQ group (8.1% vs. 5.9%). On a side note, mortality in the 761 participants who received HCQ and AZM was 18.9%.
Table 2. Characteristics of coronavirus disease 2019 (COVID-19) patients (n = 8075) by treatment group.
Cont,
”The trouble with socialism is that you eventually run out of other people's money.” - Margaret Thatcher