Randomized double-blinded placebo-controlled trial of hydroxychloroquine with or without azithromycin for virologic cure of non-severe Covid-19.

BACKGROUND: Hydroxychloroquine (HC) ± azithromycin (AZ) is widely used for Covid-19. The Qatar Prospective RCT of Expediting Coronavirus Tapering (Q-PROTECT) aimed to assess virologic cure rates of HC±AZ in cases of low-acuity Covid-19.
METHODS: Q-PROTECT employed a prospective, placebo-controlled design with blinded randomization to three parallel arms: placebo, oral HC (600 mg daily for one week), or oral HC plus oral AZ (500 mg day one, 250 mg daily on days two through five). At enrollment, non-hospitalized participants had mild or no symptoms and were within a day of Covid-19 positivity by polymerase chain reaction (PCR). After six days, intent-to-treat (ITT) analysis of the primary endpoint of virologic cure was assessed using binomial exact 95% confidence intervals (CIs) and χ2 testing. (ClinicalTrials.gov NCT04349592, trial status closed to new participants.).
FINDINGS: The study enrolled 456 participants (152 in each of three groups: HC+AZ, HC, placebo) between 13 April and 1 August 2020. HC+AZ, HC, and placebo groups had 6 (3·9%), 7 (4·6%), and 9 (5·9%) participants go off study medications before completing the medication course (p = 0·716). Day six PCR results were available for all 152 HC+AZ participants, 149/152 (98·0%) HC participants, and 147/152 (96·7%) placebo participants. Day six ITT analysis found no difference (p = 0·821) in groups' proportions achieving virologic cure: HC+AZ 16/152 (10·5%), HC 19/149 (12·8%), placebo 18/147 (12·2%). Day 14 assessment also showed no association (p = 0·072) between study group and viral cure: HC+AZ 30/149 (20·1%,), HC 42/146 (28·8%), placebo 45/143 (31·5%). There were no serious adverse events.
INTERPRETATION: HC±AZ does not facilitate virologic cure in patients with mild or asymptomatic Covid-19. FUNDING: The study was supported by internal institutional funds of the Hamad Medical Corporation (government health service of the State of Qatar).

Evidence before this study

Despite the worldwide scope of Covid-19 disease, as of August 2020 therapy is based upon inconclusive evidence. A variety of approaches have been studied, but there is no vaccine and optimal treatment remains elusive.

A search of PubMed-indexed Covid-19 evidence published (or available as accepted pre-publication manuscript) between 2019 and 1 August 2020 was executed. The evidence identified the antimalarial hydroxychloroquine (HC), often administered with the macrolide azithromycin (AZ), as one of the more commonly discussed Covid-19 therapies. In March 2020 these repurposed antimicrobials began to garner substantial worldwide attention to a degree that was arguably out of proportion to the quality of supporting data.

For Covid-19 patients with disease sufficiently severe to require hospitalization, HC's lack of efficacy is strongly suggested by an evolving literature that includes high-quality, large-scale randomized controlled trials (RCTs) in the United Kingdom (RECOVERY) and worldwide (Solidarity). However, as of end-August 2020 there are no double-blinded RCTs assessing HC±AZ facilitation of virologic cure in asymptomatic or mildly ill participants with laboratory-confirmed Covid-19.

Added value of this study

This is the first randomized, double-blind, placebo-controlled trial to report virologic outcomes of HC±AZ therapy in patients with asymptomatic or mild laboratory-confirmed Covid-19. Q-PROTECT adds to the existing evidence base in its firm demonstration, using objective laboratory measures, that neither HC nor HC+AZ has any role in eliminating or reducing SARS-CoV-2 in non-hospitalized cases. The study findings add to current evidence casting increasing doubt as to utility of HC±AZ anywhere in the Covid-19 acuity spectrum. Q-PROTECT fills a void in the evidence base in its addition of relatively precise RCT-based estimates of HC±AZ-effected virologic cure in ambulatory patients.

Implications of all the available evidence

HC, with or without AZ, is highly unlikely to result in meaningful benefit in patients with mild or asymptomatic Covid-19.

Alt-text: Unlabelled box

Introduction

In March 2020, the World Health Organization (WHO) acknowledged the pandemic status of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and its associated disease Covid-19. As of August 2020, there is neither a Covid-19 vaccine nor definitive therapy.

One therapeutic approach that was ultimately dropped from the UK's RECOVERY (in June 2020) and the WHO's Solidarity (in July 2020) is the repurposed antimalarial hydroxychloroquine (HC). Initial enthusiasm for HC, administered with or without the macrolide azithromycin (AZ), was fueled by a March 2020 report from a Marseille open-label (non-RCT) study [1] and a mid-April follow-up from the same group [2]. Covid-19 HC treatment attained a level of notoriety that was arguably disproportionate to the strength of supporting evidence.

In assessing the evidence in April 2020 – evidence that included a study from another French group [3] that failed to replicate Marseille's favorable results – reviewers concluded there was insufficient evidence to support HC as a standard of Covid-19 care [4]. Given the lack of certainty, in April 2020 a multidisciplinary group from the Qatari national healthcare system designed the Qatar Prospective RCT of Expediting Coronavirus Tapering (Q-PROTECT).

Pre-symptomatic transmission warrants focus as a target to protect a population from SARS-CoV-2. Q-PROTECT focuses on non-hospitalized low-acuity, emphasizing a dichotomous endpoint: presence of SARS-CoV-2 on real-time reverse transcriptase polymerase chain reaction (PCR). The primary aim was to determine whether, as compared to placebo, use of HC or HC+AZ was associated with higher rates of viral clearance after six days of therapy. There were two secondary study aims: assessment of day 14 virologic cure, and assessment of semi-quantitative change in viral load from baseline to day six.

Methods

Study design

Q-PROTECT was a parallel 1:1:1 allocation ratio RCT, with blinding of participants, study staff, treating clinicians, and analysts. The study occurred at two units of Qatar's national healthcare system, Hamad Medical Corporation (HMC). The first was the Emergency Department (ED) at HMC's tertiary hospital, Doha's Hamad General Hospital (HGH). The second unit was a 3500-bed quarantine facility 20 miles north of Doha, at Umm Qarn. Participants were enrolled at HGH ED or at Umm Qarn, and patient care and monitoring occurred at Umm Qarn.

The HMC ethics board reviewed and approved the trial protocol [5], which was registered at ClinicalTrials.gov (NCT04349592). The study followed CONSORT guidelines for trials.

Participants

The study's planned population consisted of SARS-CoV-2 PCR-positive males and females with mild or no symptoms. In practical application, as described in Appendix 1 Q-PROTECT sampling was composed of young, expatriate males.

The selection of the lower end of the acuity spectrum was driven by the need to comply with institutional ethics board requirements, which were in turn dictated by Q-PROTECT's inclusion of a placebo arm. At the time of study design, Qatar's national treatment criteria required antiviral therapy (e.g. HC, oseltamivir) in patients meeting any of the following criteria: hospitalization, tachypnoea (respirations >29/minute), or hypoxemia (pulse oximetry on room air <93%); treatment was also recommended for any patient with chest X-ray abnormality who had risk factors of older age (>60), immunocompromise, or co-morbidity (e.g. diabetes or hypertension). These preceding factors defining requirement for antiviral treatment also constituted exclusion criteria for Q-PROTECT.

Other inclusion and exclusion criteria were related either to logistics or to risks of study medications. Eligibility was restricted to adults (age at least 18) with positive SARS-CoV-2 PCR who were quarantined at Umm Qarn due to inability to self-quarantine.

Exclusion criteria based on documented or patient-reported past medical history were: retinal or macular disease; psoriasis; hepatic or renal disease; porphyria; glucose-6-phosphate-dehydrogenase (G6PD) deficiency; QT-interval prolongation; or hypersensitivity to HC or AZ.

Breastfeeding patients were ineligible. Pregnancy (as assessed by patient report) also constituted grounds for exclusion.

Medication-related exclusions based on drug safety were current therapy with tamoxifen, antimalarials, or dapsone. Exclusions for potential confounding of results were made in the case of recent (within one week) therapy with either of the study drugs or with any antivirals (e.g. oseltamivir).

Other exclusion criteria were dictated by initial laboratory and electrocardiography (ECG) results. Participants were excluded if laboratory assessment (within 24 h before study screening) revealed low levels of potassium or magnesium, or elevated creatinine or transaminases. ECG-based exclusion criteria related to QT prolongation risk followed American College of Cardiology (ACC) guidelines for HC+AZ therapy [6].

Participants were enrolled at two locations. During a run-in period, six participants were enrolled during an initial HGH ED visit and followed up at the Umm Qarn quarantine site. The study approach then changed such that all cases were both enrolled and followed at Umm Qarn. Further information on study enrollment site is provided in Appendix 1.

Potential participants were screened based on positive initial SARS-CoV-2 PCR testing that was ordered as part of routine care. Once PCR test results returned positive, study staff approached treating physicians for approval to discuss Q-PROTECT with patients. Written informed consent was obtained.

Randomization and masking

Once participants were enrolled, they were each given a pair of study medication bottles. The bottles’ medication contents were unknown to participants and staff. One bottle contained 21 tablets of either HC 200 mg (Sanofi-Aventis, Spain) or a similar-appearing placebo tablet; this bottle's label included an instruction to take the contents every eight hours for seven days. The other bottle contained six capsules of either AZ 250 mg (Pfizer, Italy) or a similar-appearing placebo. The second bottle's label included an instruction to take two capsules on the day of enrollment and one capsule each morning for the next four days.

Further details on Q-PROTECT randomization, allocation concealment, and triple-blinding procedures are provided in Appendix 1.

Procedures

The study's interventions were provision of self-administered study medications and execution of swabs for virologic testing. Further details on PCR testing days and sample storage are provided in Appendix 1.

Swabs were combination nasopharyngeal and oropharyngeal (Copan Diagnostics, Brescia, Italy). Swabs were executed by study-staff physicians, all of whom received training (in Covid-19 swab sample collection) from Qatar's national Communicable Disease Center. Samples were transported in universal transport medium to PCR testing equipment (see Appendix 1 for details on PCR testing).

For all PCR testing, the primary endpoint of day six virologic cure (as well as the secondary endpoint of day 14 cure) was defined as being met if the machine's cycle threshold (Ct) interpretation algorithm reported a result of negative. For participants who did not achieve day six virologic cure, the semi-quantitative secondary endpoint of Ct increase (i.e. drop in viral load) was assessed. Restriction of semi-quantitative endpoint assessment to non-cured participants was necessary due to the non-reporting of a Ct for cured participants (PCR assay did not extend beyond Ct of 40). The semi-quantitative endpoint assessment was based on median Ct value for all assessed targets (ranging from one for the Bioneer, to three for the Thermo Fisher TaqPath).

Monitoring procedures included daily ECGs for the first week, and daily in-person visits and physical examinations for the first two weeks. There were phone reassessments on days 15–20. A final in-person visit was executed on day 21 for most patients; this was changed to phone follow-up visit when the study protocol was modified to drop the day 21 swab execution (see Appendix 1).

The study was not focused on, nor was it powered to assess, clinical endpoints (including therapeutic risks). Symptom tracking focused on patient-reported fever and respiratory complaints (rhinitis, pharyngitis, cough, or chest pain). Adverse effects tracking was assessed with open-ended questioning and monitoring for events such as death, hospitalization, pneumonia development, or QT prolongation.

The study approach for QT monitoring was based on ACC recommendations [6] that recommend considering discontinuing therapy if QT is prolonged more than 30–60 msec. Q-PROTECT's initial protocol called for withdrawing participants for QT prolongation exceeding 30 msec, but the protocol was modified to increase the cut-off to 60 msec (see Appendix 1).

Primary and secondary outcomes

The study's primary outcome was achievement of virologic cure (PCR-negative status) as assessed on day six. The secondary outcomes were day 14 virologic cure and, for cases not achieving the primary outcome, virologic semi-quantitative analysis of Ct decreases from day one to day six. Additional exploratory endpoints are described in Appendix 2.

Statistical analysis

Sample-size calculations were estimated using the freeware STPLAN (Version 4.5, Department of Biomathematics, University of Texas M.D. Anderson Cancer Center, Houston, Texas, USA). Calculations were based on a best-estimate baseline (control group) virologic cure rate of 50%, as assessed at day six. Estimates were generated in March 2020 based on clinical experience in Qatar and on assessment of the relevant non-severe Covid-19 cases in the extant evidence base [7]. The minimum clinically important absolute effect difference was defined to be 10%. Given these assumptions, the study was designed to accrue 152 participants per group (total n = 456) to achieve 80–90% power (depending on drop-out rate). Details of sample-size and power calculations are provided in Appendix 1.

Study planning dictated intent-to-treat (ITT) analysis. Per-protocol analysis was also executed, but only for exploratory assessments. All analyses other than sample-size calculations were performed using Stata (version 16.1 MP, StataCorp, College Station, Texas, USA).

Interim analyses were conducted after accrual of 100 participants and 200 participants. Pre-specified O'Brien-Fleming levels were defined for the interim α (0.001 and 0.015) and final α (0.047) and this information was provided to Q-PROTECT's Data Safety Monitoring Board (DSMB). Details on interim analysis planning are provided in Appendix 1.

Q-PROTECT's primary dichotomous endpoint, virologic cure at day six, was assessed for each of the three study groups. This endpoint was reported as a proportion with 95% binomial exact confidence interval (CI). Comparison of the proportions was executed using χ2 testing. Pairwise absolute differences in between-groups proportions were calculated as the absolute risk difference with 95% CI.

Secondary endpoint analysis, conducted for participants who did not achieve the primary endpoint of virologic cure, assessed Ct changes from day one to day six. Ct increase (corresponding to decreased viral load) was assessed for normality (with Shapiro-Wilk testing identifying the data as non-normal). Ct was described using the median with its interquartile range (IQR) at baseline, and its 95% CI for day six and day 14. Kruskal-Wallis testing was used to assess for group-related differences in magnitude of Ct increase.

The trial is registered at ClinicalTrials.gov (NCT04349592).

Role of the funding source

Q-PROTECT was wholly funded and resourced by the study institution (the government healthcare entity of the State of Qatar). The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

Q-PROTECT commenced enrollment on 13 April 2020. Accrual of the n = 456 participants concluded on 1 August 2020 at the point of the trial's reaching its initially targeted accrual goal. Two interim analyses were conducted as planned, with results reviewed by the institutional ethics board and the DSMB; at each interim analysis the decision was made to proceed with Q-PROTECT accrual. At least one follow-up PCR (on day six or day 14) was available for all 152 HC+AZ participants, 150 HC participants (one of whom had only a day 14 PCR available, thus 149 HC participants were assessed for the primary endpoint), and 147 placebo participants. The participant trial profile is depicted in Fig. 1.

Fig. 1

Trial profile.

Baseline data showing demographic and clinical characteristics for each group are shown in Table 1.

Table 1

Baseline characteristics of the intent-to-treat population.

	Hydroxychloroquine  + Azithromycinn = 152	Hydroxychloroquine n = 152	Placebon = 152
Male sex	150 (98·7%)	149 (98·0%)	150 (98·7%)
Age (years)	42 (38–48)	40 (31–47)	41 (31–47)
Nationality
Bangladesh	31 (20·4%)	32 (21·1%)	29 (19·1%)
Egypt	0 (0%)	1 (0·7%)	0 (0%)
Ghana	5 (3·3%)	1 (0·7%)	1 (0·7%)
India	54 (35·5%)	63 (41·5%)	47 (30·9%)
Indonesia	0 (0%)	0 (0%)	1 (0·7%)
Kenya	0 (0%)	1 (0·7%)	1 (0·7%)
Nepal	45 (29·6%)	38 (25·0%)	47 (30·9%)
Pakistan	0 (0%)	1 (0·7%)	3 (2·0%)
Philippines	7 (4·6%)	6 (4·0%)	6 (4·0%)
Romania	0 (0%)	0 (0%)	1 (0·7%)
Somalia	0 (0%)	1 (0·7)	0 (0%)
Sri Lanka	10 (6·6%)	6 (4·0%)	14 (9·2%)
Sudan	0 (0%)	1 (0·7%)	1 (0·7%)
Uganda	0 (0%)	1 (0·7%)	1 (0·7%)
Baseline Ct*	22·0 (18·4–28·0)	23·4 (20·0–28·4)	22·2 (19·2–26·4)
Symptoms at enrollment
Patient-reported fever	46 (30·3%)	51 (33·6%)	52 (34·2%)
Respiratory symptoms⁎⁎	37 (24·3%)	36 (23·7%)	34 (22·4%)

Data are n (%) or median (IQR).

Ct=cycle threshold (median for all markers assessed); baseline Ct not available for two participants in hydroxychloroquine group and two participants in placebo group, thus data available for 452 of 456 randomized patients.

Respiratory symptoms: Chest pain, cough, pharyngitis, or rhinitis.

The primary outcome, virologic cure at day six, was assessed on an ITT basis and thus included participants who had discontinued study medications if those participants had day six PCR results. Day six virologic cure data were available for 152/152 HC+AZ participants, 149/152 (98·0%) HC participants, and 147/152 (96·7%) placebo participants. Results for the primary outcome are shown in Fig. 2 and Table 2. The between-groups differences in proportions achieving day six virologic cure were: placebo minus HC+AZ 1·7% (95% CI −5·5 to 8·9%), HC minus placebo 0·5% (95% CI −7·0 to 8·0%), HC minus HC+AZ 2·2% (95% CI −5·0 to 9·5%). The preceding results and those in Fig. 2 present absolute risk; differences in relative risk are presented in Appendix 2.

Fig. 2

Primary outcome of virologic cure at day six.

Table 2

Primary and secondary outcomes.

	Hydroxychloroquine + Azithromycinn = 152	Hydroxychloroquinen = 152	Placebon = 152	p
Virologic cure at day six*	16/152 (10·5%, 6·1–16·5%)	19/149 (12·8%, 7·9–19·2%)	18/147 (12·2%, 7·4–18·7%)	0·821
Increase in Ct from day one to day six*	7·2 (3·9–11·5; 6·1–8·8)	7·5 (3·4–11·5; 5·7–8·8)	8·0 (4·1–11·7; 7·3–9·0)	0·634
Virologic cure at day 14⁎⁎	30/149 (20·1%, 14·0–27·5%)	42/146 (28·8%, 21·6–36·8%)	45/143 (31·5%, 24·0–39·8%)	0·072

Data are n/N (%, 95% CI) or median (IQR; 95% CI for median). Ct=cycle threshold (lower values correspond to higher viral load).

Data not assessed for all randomized participants.

Data assessed for 394 of 403 cases not reaching day six virologic cure endpoint.

Table 2 also shows results for the secondary outcomes of day 14 virologic cure and change in Ct from baseline (day one) to day six. Day 14 virologic cure data were available for 149/152 (98·0%) HC+AZ participants, 146/152 (96·1%) HC participants, and 143/152 (94·1%) placebo participants. Fig. 3 depicts the day 14 virologic cure secondary endpoint achievement.

Fig. 3

Secondary outcome of virologic cure at day 14.

The study defined an additional secondary endpoint of Ct change from baseline to day six, to be calculated for participants not achieving virologic cure. In the HC+AZ group, after subtracting the 16 participants with day six virologic cure from the initial group (n = 152) there were 136 participants for comparison of day one and day six Ct; both baseline and day six Ct data were available for all 136 participants. In the HC group, subtraction of the 19 virologically cured participants left 133 remaining; both baseline and day six Ct data were available for 129/133 (97·0%). Subtraction of the 18 virologically cured participants from the placebo group left 134 remaining; both baseline and day six Ct data were available for 129/134 (96·3%).

There were no deaths or serious adverse events. There was no association (p = 0·708) between study group and development of pneumonia, which was diagnosed in seven participants (1·5%): three (2·0%) in the HC+AZ group, one (0·7%) in the HC group, and three (2·0%) in the placebo group. Pneumonia accounted for seven of the 11 all-cause hospitalizations, which were not associated (p = 1·00) with study group. (Further information on hospitalizations is provided in Appendix 2.)

No patient had palpitations, syncope, or other symptoms indicative of cardiac dysrhythmia. Torsade de pointes was not identified on any ECG. Further results regarding QT are provided in Appendix 2.

A total of 22 participants (4·8% of 456) withdrew from the study-medication portion of Q-PROTECT after receiving at least one day of study medication. The placebo group accounted for nine withdrawals (5·9% of 152), the HC+AZ group seven (4·6% of 152), and the HC group six (4·0% of 152). There was no association (p = 0·716 by Pearson χ2) between study group and participant withdrawal.

The most common reason for study withdrawal was the participant requesting withdrawal without giving a symptom-based reason; this occurred in 10 cases (three in HC+AC group, one in HC group, six in placebo group). Eight participants (five in HC group, three in placebo group) were withdrawn due to asymptomatic QT prolongation (see Appendix 2). Three withdrawals (two in the HC+AZ group and one in the HC group) were prompted by identification of lab abnormalities within 24 h of study enrollment. One participant in the HC+AZ group was withdrawn for (transient) diplopia.

Sensitivity analyses suggested that Q-PROTECT's primary and secondary endpoint findings were not influenced by the relatively few missing data. Similar lack of influence from withdrawals was suggested by per-protocol analyses. These analyses, as well as further exploratory analyses, are reported in Appendix 2.

Discussion

Q-PROTECT's main finding was a failure of HC±AZ to have any salutary effect in mild or asymptomatic SARS-CoV-2 infection. In a group of relatively young, healthy participants (virtually all males) enrolled within 24 h of testing positive for the virus, HC±AZ neither improved virologic cure rates nor reduced viral burden. The therapeutic failure of HC±AZ was clear at both day six and day 14.

A promising March 2020 report from Marseille [1] focused interest on HC±AZ to speed viral clearance and effect clinical improvement. Within a few weeks, though, neither of these benefits were found by another French group assessing a small (n = 11) series of hospitalized patients [3]. The ensuing months have seen substantial criticism of both the initial Marseille study and the overall evidence base regarding use of HC±AZ for Covid-19 [8].

HC's performance with respect to post-exposure Covid-19 prophylaxis was assessed by Boulware and colleagues in an RCT instituting therapy within four days of high-risk exposure [9]. Even with relatively high doses (1400 mg on day one, followed by 800 mg daily for days two through five) HC did not reduce rates of SARS-CoV-2 viral detection or (in the relatively large proportion of patients for whom there was no viral testing) development of Covid-19 symptoms.

With regard to HC treatment of mostly mild or asymptomatic cases, the Marseille group reported their ongoing open-label treatment results in May 2020; their non-controlled series (numbering 1061) continued to be favorable, with 92% having good clinical outcome and virologic cure [10]. A few months later, a July 2020 open-label RCT from Catalonia arrived at a different conclusion: in non-hospitalized cases HC improved neither symptoms nor viral load [11]. The same month, a double-blind North American RCT of adult outpatients with early Covid-19 diagnosis identified no clinical benefit with HC [12]

For hospitalized cases with mainly mild or moderate disease, two Chinese open-label trials reported in May 2020 that HC failed to speed clinical improvement or virologic cure [13,14]. The same month, though, an open-label study of intubated patients in Wuhan found that HC reduced mortality via attenuation of cytokine storm [15].

Three other May 2020 observational cohort studies, two from the USA [16,17] and the other from France [18], failed to identify HC benefit. Furthermore, the analysis from New York [17] suggested that the combination of HC+AZ was associated with increased risk of death from cardiac arrest. Editorial commentary began to emphasize need for careful consideration of risks and benefits when considering Covid-19 treatment with QT-prolonging drugs [19].

By June 2020, the balance of evidence supported a case against treatment of Covid-19 cases with HC±AZ. An international registry-based study of hospitalized Covid-19 cases identified increased risk of harm (due to ventricular dysrhythmia) in patients receiving HC or HC+AZ [20]. The RECOVERY investigators announced withdrawal of the HC arm from their large-scale adaptive RCT [21]. June also saw the USA's Food and Drug Administration (FDA) warning of unfavorable risk:benefit ratio with use of HC outside the hospital setting [22].

In July 2020, there emerged more conflicting evidence on HC use in hospitalized Covid-19 cases. Preliminary analysis by WHO's Solidarity investigators prompted HC's removal from their RCT [23], and an open-label Brazilian RCT concurred in finding no HC clinical benefits [24]. However, two other reports, one from the USA and one from Marseille, left open the possibility of a role for HC+AZ in Covid-19.

From the USA, Arshad and colleagues’ July 2020 observational cohort study reported that HC (and similarly, the combination HC+AZ) improved mortality in patients hospitalized with Covid-19 [25]. The Michigan group found that AZ alone did not improve outcome, and there was no statistically significant benefit to adding AZ to HC [25]. Favourable findings from the USA were echoed by the Marseille group, whose HC+AZ study population (now numbering 3119) spanned the acuity spectrum. The non-randomized Marseille series, which now included patients (n = 618) not treated with HC+AZ, found that HC+AZ cases had improved clinical outcomes and shorter duration of viral shedding [26].

As of 1 August 2020, the date of closure of enrollment in Q-PROTECT, existing data on HC±AZ use in hospitalized Covid-19 cases seems weighted toward the negative. However, as an early August commentary by Cohen states, the question of HC's potential utility in Covid-19 has not been definitively answered [27]. There is a void in the existing evidence base that is filled by the current study.

In its blinded RCT design, Q-PROTECT differs from all but two of the preceding studies (both of which were published by the same group of North American investigators) [9,12]. The preponderance of the HC Covid-19 evidence – even the largest, high-quality trials such as RECOVERY and Solidarity – comes from observational or open-label designs and addresses treatment in hospitalized patients.

It is noteworthy that one of the two double-blinded Covid-19 HC RCTs addressed post-exposure prophylaxis only [9]. The other [12] was limited by performance of SARS-CoV-2 testing in only 58% of participants. While both of the blinded trials undoubtedly advanced the state of Covid-19 knowledge, neither assessed the concrete endpoint of virologic cure in a PCR-positive Covid-19 population. Q-PROTECT is the first double-blinded RCT that assesses in virtually all of its participants, an objective virologic endpoint in cases (those with mild or no symptoms) in whom virologic clearance is critical to pandemic control.

It is not the case that virologic cure is more important than clinical outcomes, but Q-PROTECT's laboratory-assessed endpoints fill a gap in the blinded-RCT Covid-19 evidence base. Specifically, a blinded RCT could inform national decisions on utility of HC±AZ to expedite viral clearance and thus reduce transmission. When Q-PROTECT commenced in mid-April 2020, approximately 50,000 people in Qatar had been tested for Covid-19. There had been over 3000 positive SARS-CoV-2 results and seven deaths, with 252 new cases in the 24 h prior to study commencement. As enrollment closed in August 2020, the Covid-19 epidemic in Qatar was a few months past its peak but the daily PCR-positive n was still over 200. Approximately a half-million Covid-19 tests had been done in the country between March and August, with over 100,000 positive results and a death rate of roughly 1 in 1000. Q-PROTECT set out to determine whether HC±AZ could expedite viral clearance and thus likely reduce transmission.

When considered in context of the existing evidence, the current study contributes data that can help fill the final gaps in knowledge about HC±AZ utility in Covid-19. Q-PROTECT's main strengths are inherent in its blinded RCT design. There was no indication of flaws in either randomization or blinding, and the effect estimates are unbiased and reasonably precise (with acceptable CIs). The point estimates for the primary endpoint of day six virologic cure were nearly equal for HC and placebo (with HC+AZ's cure proportions lower). The point estimates for both secondary endpoints were actually more favourable for placebo than for either HC or HC+AZ. There is thus no indication that accruing a larger sample would change Q-PROTECT's results.

There are a number of study limitations that restrict the conclusions drawn from Q-PROTECT. Perhaps the most important is in the emphasis on a non-clinical endpoint rather than patient-centered outcomes (e.g. symptoms, immunity). The intent was to shed light on a public health outcome – transmissibility – via a surrogate of virologic testing. The assumption that PCR negativity on naso- and oropharyngeal swab samples is linked to lesser likelihood of Covid-19 transmission is rational, but unproven and potentially nuanced. It is likely, for example, that Ct is an oversimplified surrogate for transmission risk, and that factors such as respiratory symptoms (e.g. sneezing) may be important contributors [28].

Just as negative PCR may not always mean zero transmission risk, a positive PCR could simply reflect detection of inactive (non-infectious) viral remnants. There remains a small (but non-zero) chance that Q-PROTECT's non-identification of post-treatment PCR detection differences could obscure a clinically important infectivity difference. The study is limited by the failure to address the possibility that there could be inter-group differences in infectivity of whatever viral particles were present after treatment.

Even if the use of PCR is accepted as an indicator of transmission risk, there remain unanswered questions that translate into Q-PROTECT limitations. Selection of another Ct endpoint (e.g. Ct >30) may accurately classify patients at very low risk of transmission; if this is the case then the endpoint of negative PCR (i.e. Ct >40 on the equipment used in this study) would be too stringent. No post hoc analysis was executed on different Ct cut-offs.

Other study flaws constituted threats to both internal and external validity. The main internal validity problems included dropouts and other losses to follow-up. The most substantial external validity threats related to the medication regimen and the study population.

A potentially significant internal validity issue was failure to confirm medication compliance (e.g. by having staff administer medications or by assaying drug levels). Since unreported non-compliance with study therapy would likely be associated with an active-drug regimen (e.g. from gastrointestinal side effects), it is possible that differential medication compliance biased Q-PROTECT toward a null finding.

Even if internal validity questions are resolved, there were a number of study limitations that affect external validity. Among the most important are related to Q-PROTECT's study population and the study's specific medication regimen.

Q-PROTECT's participants were nearly all male, and relatively young. Viral clearance rates are likely similar in females and males, but older patients may clear Covid-19 more slowly [29]. Differential viral clearance in various races or ethnicities has not been well characterized. Q-PROTECT results are applicable only to patients similar to those enrolled in the current study.

The current study results applicability is also restricted in terms of medication regimen. While AZ use for asymptomatic or mildly symptomatic Covid-19 cases tends follow consistent dosing (500 mg on day one, 250 mg on days two through five), HC dosing varies widely across the Covid-19 evidence base. The Q-PROTECT regimen was selected in March 2020, to match the approach reported successful in Marseille [1]. However, some studies have utilized higher HC doses in the initial days of therapy, and many studies use different daily maintenance doses or durations of therapy. For example, as compared to the one-week Q-PROTECT regimen of 200 mg HC three times daily, Arshad and colleagues [25] and Mitja and colleagues [11] both utilized a day one 800 mg dose followed by 200 mg twice daily for less than a week. Tang and colleagues [13], while also focusing on mild or moderate disease, administered a higher initial dose (1200 mg daily for three days) and a higher maintenance dose (800 mg) for a longer time frame (two to three weeks). In their post-exposure prophylaxis study, Boulware and colleagues [9] also used a relatively high initial HC dose (1400 mg on the day of exposure).

In the Covid-19 evidence base, HC dosing levels do not invariably correlate with efficacy findings. This absence of definitive correlation does not exclude potential importance of dosing regimen. Expert reviewers have remarked that initial therapy with at least 800 mg may be necessary for viral clearance [30].

Physiologically based pharmacokinetic (PBPK) modeling has also utilized a day-one dosage of 800 mg, but overall PBPK recommendations are not substantially inconsistent with the dosing approach used in Q-PROTECT. Yao and colleagues [31] used PBPK models to assess multiple regimens of HC SARS-CoV-2 in an in vitro (Vero cell) model. Their recommendation for a first-day loading dose of 800 mg followed by four daily doses of 400 mg was aimed at balancing safety and efficacy, but they did not assess a day-one loading dose of less than 800 mg. The overall approach suggested by PBPK modeling was not markedly different from Q-PROTECT's regimen: the current study provided a smaller day-one dose (600 mg rather than 800 mg) but a larger subsequent daily dose (600 mg rather than 400 mg). It is possible that a dosing regimen different from that of Q-PROTECT could produce different results.

If it is the case that early exposure to HC is needed in order for the drug to effectively inhibit viral replication, then dosing issues may be overshadowed by the fact that the medication was given too late in the course of illness. This hypothesis, while not able to be tested in the current dataset (Q-PROTECT is underpowered for the assessment), seems an unlikely major confounder. Medication was instituted rapidly after PCR – within hours, and never more than 24 h – in patients with disease that was either mild or asymptomatic. However, it must be acknowledged that the negative findings of Q-PROTECT do not necessarily rule out HC benefit if the drug is given earlier in the course of infection.

Any benefit from HC must be weighed against drug-associated adverse effects. As used for Covid-19, HC's most common side effects are gastrointestinal and rarely severe [13 11]. Serious adverse events (defined as mortality or major non-transient morbidity) did not occur in Q-PROTECT and were also rare or absent in other HC RCTs [9,11,13,25] However, all reports acknowledge the most serious adverse effect of HC±AZ as QT prolongation with associated risk of dangerous dysrhythmias such as torsades de pointes (TdP).

Q-PROTECT was not powered to assess rare events such as TdP. Neither TdP nor any ventricular dysrhythmia was seen. However, in considering large-scale use of HC±AZ, even rare risks have important population-level implications [20]. Q-PROTECT's adverse-effect results should not be construed as confirming safety of HC±AZ. Exploration of rare but significant adverse effects remains the province of larger studies that are more focused on QT assessment.

The study's daily ECGs were judged to provide an acceptable safety margin for detection of significantly prolonged QT or concerning dysrhythmia. However, there remained important QT-related study limitations. The study methodology did not guarantee that for each participant, ECGs would be regularly timed (a few hours after medication dosing) [6], performed by the same machine, and undergo cross-validation (since machine algorithms can over- or underestimate QT) [32]. Study participants’ QT monitoring in the quarantine environment was characterized by use of different ECG machines, irregularly timed ECG execution, delayed availability of hardcopy ECG tracings, and lack of cross-checking of machine-reported QT. The study data are therefore not suitable for analysis of QT prolongation associated with HC±AZ. Other investigators using appropriately rigorous methodology have already quantified QT prolongation by HC and AZ, confirming the fact that whatever benefits HC±AZ may bring, come with attendant risk [33].

The lessons of Q-PROTECT must be considered in light of the trial strengths and weaknesses, the medication risks and benefits, and the existing evidence base. Taking all of these factors into account, the investigators conclude that HC±AZ shows no sign of usefulness in the population studied, and that there is low likelihood of undiscovered drug benefits outweighing therapeutic risks.

Data sharing

Deidentified Q-PROTECT study data (in spreadsheet form, with included data definitions) will be made available for sharing with publication, using Mendeley Data. Data sharing requests are participant to approval by the Q-PROTECT principal investigator. The Mendeley DOI and other information can be obtained by emailing: Sarah.Thomas19@Imperial.ac.uk.

Author contributions

Authors have made the following contributions consistent with ICJME recommendations for authorship:

All authors made meaningful contributions to manuscript preparation, review, and editing. In addition, the authors made contributions as follows:

- Stephen H. Thomas conceived the study, co-wrote the draft manuscript, and maintains overall responsibility for Q-PROTECT conduct and reporting.

- Ali S. Omrani made critical contributions/modifications to the study design and execution plans, and co-wrote the initial manuscript draft.

- Sameer A. Pathan led the execution stage of the study.

- Sarah A. Thomas planned the randomization and allocation concealment approaches, performed the literature search and initial-draft manuscript writing relevant to cell biology of putative drug action mechanisms, executed data-sharing processes, and assisted with revision of the manuscript during the editorial review process.

- Peter V. Coyle led the virology testing planning and execution and made key manuscript contributions in the arena of laboratory medicine.

- Naema Al Mawlawi and Reham Al Kahlout managed the lab specimens, executed all PCR analysis, and arranged virologic reporting data

- Tim R. E. Harris provided substantial manuscript editing, and coordinated physician resourcing for study participant accrual and data processing.

- Caroline E. Thomas led planning of, and manuscript writing describing, data entry spreadsheet/source-table design, preparing of participants’ data entry forms; she also executed clinical trials registration and assisted with revision of the manuscript during the editorial review process.

- Isma Qureshi and Zain Bhutta tracked study accrual, dispatched study staff to quarantine centers, managed data tabulation and recording, and executed initial screening to search for eligible participants.

- Hussam Al Soub and Muna Al Maslamani advised on study trial arms design, guided ethics-board communications regarding placebo utilization, and directed collaboration with national Covid-19-management authorities.

- Mohammed Abu Khattab provided infectious disease and public-health expertise during manuscript preparation, and arranged training for study staff.

- Mulham M. Saleh and Andrew Jeremijenko coordinated and oversaw accrual activities at Qatar's quarantine centers.

- Aftab Azad led interdepartmental collaboration in study planning and coordinated discussions with the national authorities responsible for access to both participants and study medications.

- Joann Daghfal assisted with participant identification, statistical planning, data analysis, and interpretation.

- Ashraf El Malik advised on pharmacology, executed design and preparation of placebos, and led the processes related to study-medication preparation and administration.

Funding

The study was supported by internal institutional funds of the Hamad Medical Corporation (government health service of the State of Qatar).

Declaration of Competing Interest

The authors have no financial or personal relationships with other people or organizations that could represent a conflict of interest.

Table A1

Power and sample-size calculations.

	Hypothesized % viral clearance	n per group (80% power)	n per group (90% power)
Best-estimate baseline endpoint achievement (50%); smaller effect size (10%)
Placebo	50%	116	152
Hydroxychloroquine	60%	116	152
Hydroxychloroquine + Azithromycin	70%	116	152
Best estimate baseline endpoint achievement (50%); larger effect size (20%)
Placebo	50%	25	33
Hydroxychloroquine	70%	25	33
Hydroxychloroquine + Azithromycin	90%	25	33
Very low endpoint achievement, smaller effect size
Placebo	10%	77	101
Hydroxychloroquine	20%	77	101
Hydroxychloroquine + Azithromycin	30%	77	101
Very low endpoint achievement, larger effect size
Placebo	10%	25	33
Hydroxychloroquine	30%	25	33
Hydroxychloroquine + Azithromycin	50%	25	33
Low endpoint achievement, smaller effect size
Placebo	20%	101	133
Hydroxychloroquine	30%	101	133
Hydroxychloroquine + Azithromycin	40%	101	133
Low endpoint achievement, larger effect size
Placebo	20%	29	38
Hydroxychloroquine	40%	29	38
Hydroxychloroquine + Azithromycin	60%	29	38
Fair endpoint achievement, smaller effect size
Placebo	30%	116	152
Hydroxychloroquine	40%	116	152
Hydroxychloroquine + Azithromycin	50%	116	152
Fair endpoint achievement, larger effect size
Placebo	30%	30	40
Hydroxychloroquine	50%	30	40
Hydroxychloroquine + Azithromycin	70%	30	40
Better-than expected endpoint achievement, smaller effect size
Placebo	60%	101	133
Hydroxychloroquine	70%	101	133
Hydroxychloroquine + Azithromycin	80%	101	133
Better-than expected endpoint achievement, larger effect size
Placebo	60%	21	27
Hydroxychloroquine	80%	21	27
Hydroxychloroquine + Azithromycin	90%	21	27

Table A2

Primary (day six) outcome: relative measures.

	Risk ratio (95% confidence interval)	Risk difference (95% confidence interval)
Hydroxychoroquine (12.8% cure rate) vs. placebo (12.2% cure rate)	1·04 (0·57–1·90)	0·01 (−0·07–0·08)
Hydroxychoroquine + Azithromycin (10.5% cure rate) vs. placebo (12.2% cure rate)	0·86 (0·46–1·62)	−0·02 (−0·09–0·05)
Hydroxychoroquine (12.8% cure rate) vs. Hydroxychloroquine + Azithromycin (10.5% cure rate)	1·21 (0·65–2·26)	0·02 (−0·05–0·09)

Table A3.

Symptom development or clearance.

	Hydroxychloroquine +  Azithromycinn = 152*	Hydroxychloroquinen = 152*	Placebon = 152*
Asymptomatic day one and:
Symptomatic day seven	8 of 78 (10·3%, 4·5–19·2%)	9 of 78 (11·5%, 5·4–20·1%)	13 of 85 (15·3%, 8·4–24·7%)
Symptomatic day 14	3 of 79 (3·8%, 0·8–10·7%)	3 of 77 (3·9%, 0·8–11·0%)	4 of 84 (4·8%, 1·3–11·7%)
Symptomatic day 21	4 of 78 (5·1%, 1·4–12·6%)	2 of 77 (2·6%, 0·3–9·1%)	2 of 85 (2·4%, 0·3–8·2%)
Symptomatic day one and:
Asymptomatic day seven	56 of 70 (80·0%, 68·7–88·6%)	55 of 69 (79·7%, 68·3–88·4%)	52 of 59 (88·1%, 77·1–95·1%)
Asymptomatic day 14	66 of 69 (95·7%, 87·8–99·1%)	64 of 69 (92·8%, 83·9–97·6%)	58 of 60 (96·7%, 88·5–99·6%)
Asymptomatic day 21	67 of 69 (97·1%, 89·9–99·6%)	68 of 69 (98·6%, 92·2–100·0%)	56 of 60 (93·3%, 83·8–98·2%)

Table A4.

QT prolongation.

	Hydroxychloroquine + Azithromycinn = 152	Hydroxychloroquinen = 152	Placebon = 148*	p
QT prolongation >30 msec	37 (24·3%; 17·8–32·0%)	31 (20·4%; 14·3–27·7%)	13 (8·8%; 4·8–14·6%)	0·001
QT prolongation >60 msec	5 (3·3%; 1·1–7·5%)	4 (2·6%; 0·7–6·6%)	2 (1·4%; 0·2–4·8%)	0·641
Maximum QT	418 (403–434; 411–422)	415 (403–434; 412–420)	406 (394–427; 404–414)	0·002
Maximum QT prolongation for cases with + QT prolongation	23 (15–31; 20–24)	20 (13–29; 16–23)	13 (9–22; 11–15)	<0·001

Data are n (%, 95% binomial exact confidence interval) or median (IQR; 95% confidence interval for median).

Data not available for all randomized patients.