key: cord-0710830-jwzgc5t1
authors: Skipper, Caleb P.; Pastick, Katelyn A.; Engen, Nicole W.; Bangdiwala, Ananta S.; Abassi, Mahsa; Lofgren, Sarah M.; Williams, Darlisha A.; Okafor, Elizabeth C.; Pullen, Matthew F.; Nicol, Melanie R.; Nascene, Alanna A.; Hullsiek, Kathy H.; Cheng, Matthew P.; Luke, Darlette; Lother, Sylvain A.; MacKenzie, Lauren J.; Drobot, Glen; Kelly, Lauren E.; Schwartz, Ilan S.; Zarychanski, Ryan; McDonald, Emily G.; Lee, Todd C.; Rajasingham, Radha; Boulware, David R.
title: Hydroxychloroquine in Nonhospitalized Adults With Early COVID-19: A Randomized Trial
date: 2020-07-16
journal: Ann Intern Med
DOI: 10.7326/m20-4207
sha: a5ade2cfaf43ef3e312221f4e1e3352fdc31256c
doc_id: 710830
cord_uid: jwzgc5t1

BACKGROUND: No effective oral therapy exists for early coronavirus disease 2019 (COVID-19). OBJECTIVE: To investigate whether hydroxychloroquine could reduce COVID-19 severity in adult outpatients. DESIGN: Randomized, double-blind, placebo-controlled trial conducted from 22 March through 20 May 2020. (ClinicalTrials.gov: NCT04308668) SETTING: Internet-based trial across the United States and Canada (40 states and 3 provinces). PARTICIPANTS: Symptomatic, nonhospitalized adults with laboratory-confirmed COVID-19 or probable COVID-19 and high-risk exposure within 4 days of symptom onset. INTERVENTION: Oral hydroxychloroquine (800 mg once, followed by 600 mg in 6 to 8 hours, then 600 mg daily for 4 more days) or masked placebo. Measures: Symptoms and severity at baseline and then at days 3, 5, 10, and 14 using a 10-point visual analogue scale. The primary end point was change in overall symptom severity over 14 days. RESULTS: Of 491 patients randomly assigned to a group, 423 contributed primary end point data. Of these, 341 (81%) had laboratory-confirmed infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) or epidemiologically linked exposure to a person with laboratory-confirmed infection; 56% (236 of 423) were enrolled within 1 day of symptoms starting. Change in symptom severity over 14 days did not differ between the hydroxychloroquine and placebo groups (difference in symptom severity: relative, 12%; absolute, −0.27 points [95% CI, −0.61 to 0.07 points]; P = 0.117). At 14 days, 24% (49 of 201) of participants receiving hydroxychloroquine had ongoing symptoms compared with 30% (59 of 194) receiving placebo (P = 0.21). Medication adverse effects occurred in 43% (92 of 212) of participants receiving hydroxychloroquine versus 22% (46 of 211) receiving placebo (P < 0.001). With placebo, 10 hospitalizations occurred (2 non–COVID-19–related), including 1 hospitalized death. With hydroxychloroquine, 4 hospitalizations occurred plus 1 nonhospitalized death (P = 0.29). LIMITATIONS: Only 58% of participants received SARS-CoV-2 testing because of severe U.S. testing shortages. CONCLUSION: Hydroxychloroquine did not substantially reduce symptom severity in outpatients with early, mild COVID-19. PRIMARY FUNDING SOURCE: Private donors.

In outpatient or telehealth settings at least two of the following symptoms:

• fever (measured or subjective),

• chills, • rigors, • myalgia,

• headache, • sore throat,

• new olfactory and taste disorder(s) OR At least one of the following symptoms:

• cough,

• shortness of breath, or • difficulty breathing OR Severe respiratory illness with at least one of the following:

• Clinical or radiographic evidence of pneumonia, or • Acute respiratory distress syndrome (ARDS). AND No alternative more likely diagnosis

Clinically compatible symptoms with one or more of the following exposures in the 14 days before onset of symptoms.

Exposure: Close contact** with a person diagnosed with COVID-19; whereby close contact is defined as being within 6 feet for a period of 10 minutes to 30 minutes or more depending upon the exposure. In healthcare settings, this may be defined as exposures of greater than a few minutes or more. Data are insufficient to precisely define the duration of exposure that constitutes prolonged exposure and thus a close contact. (2) Suspect case A. A patient with acute respiratory illness (fever and at least one sign/symptom of respiratory disease, e.g., cough, shortness of breath), AND a history of travel to or residence in a location reporting community transmission of COVID-19 disease during the 14 days prior to symptom onset. OR B. A patient with any acute respiratory illness AND having been in contact with a confirmed or probable COVID-19 case (see definition of contact) in the last 14 days prior to symptom onset; OR C. A patient with severe acute respiratory illness (fever and at least one sign/symptom of respiratory disease, e.g. cough, shortness of breath; AND requiring hospitalization) AND in the absence of an alternative diagnosis that fully explains the clinical presentation.

The full list of exclusion criteria, included:

• Symptoms >4 days (per inclusion criteria)

• Age <18 years old • In Canada, additional exclusions mandated by regulatory authorities were: pregnancy, breastfeeding; severe diarrhea or vomiting; known cirrhosis with encephalopathy or ascites; known prolonged cardiac QT interval, ventricular arrhythmia, or history of sudden cardiac death; or QT-prolonging medicines (as below) (3)

• On April 20, 2020, additional U.S. exclusions were added for weight less than 50kg, structural or ischemic heart disease, personal or family history of cardiac QT prolongation, and QT-prolonging medications (as below).

Concomitant QT prolonging medications included current use of:

• QT prolonging medicines of: ○ Antimicrobials: azithromycin clarithromycin, erythromycin, ciprofloxacin, levofloxacin, moxifloxacin, ketoconazole, itraconazole, or mefloquine ○ Antidepressants: amitriptyline, citalopram, desipramine, escitalopram, imipramine, doxepin, fluoxetine, wellbutrin, or venlafaxine ○ Antipsychotic or mood stabilizers: haloperidol, droperidol, lithium, quetiapine, thioridazine, ziprasidone ○ Methadone ○ Sumatriptan, zolmitriptan

The prohibition of azithromycin and other QT prolonging medicines was at the express direction of the U.S. FDA as potentially unsafe in an outpatient clinical trial.

Symptom severity scores were recorded with the online participant surveys at baseline, days 3, 5, 10 and 14 for those who responded "yes" to a survey question of: "Are you experiencing COVID-19 symptoms?" i. Visual analog scale (0-10) for "overall symptom severity" was collected via a digital slider bar, which was marked with "0 = no symptoms"; 5 (placed in the middle); and "10 = severe symptoms"

ii. The exact wording of the question was as below: "Severity of overall symptoms?"

ii. For those who responded "no" to "Any symptoms experienced" the symptom severity score for that visit is assigned as zero.

i. For those hospitalized or with deaths, their symptom severity was scored as 10 if they did not respond to the visit survey.

ii. Individual symptoms were not scored, just the overall severity of symptoms.

The primary analysis cohort includes all participants who contributed at least one follow-up survey with symptom data, so that change in symptom severity score could be assessed. A longitudinal mixed model was used to estimate the overall treatment difference between the treatment and control groups. The longitudinal mixed model (SAS PROC Mixed) considered study ID and treatment group as class variables, study visit as a continuous variable (based on actual date of survey completion) and was further adjusted for baseline symptom severity score. Change in symptom severity was assessed for each survey day (SAS PROC GLM), adjusting for baseline symptom severity score.

Three sensitivity analyses were performed: a) Including only those with symptoms at baseline (thus excluding participants who were PCR positive at baseline with no reported symptoms); b) Those who were symptomatic at consent (i.e., those randomized through the symptomatic strata only); and c) Those who were asymptomatic at time of informed consent, randomized to the postexposure trial, and then developed symptoms prior to study medication start date on Day 1.

An additional sensitivity analysis was performed using overall symptom severity scores (rather than change in scores) and which included the 68 participants with no follow-up symptom data. We used a log-link gamma-errors generalized linear mixed model and used new_score= score+1 to account for the inability of the log-gamma distribution to handle zero values. With a log-gamma distribution, the estimate from the generalized linear mixed model is a difference of Hydroxychloroquine to placebo, so if there is no difference we would expect an estimate of 0. We generated 1000 estimates from simple random samples of n=400.

Median change in symptom score at each survey day is presented with interquartile range (IQR) as a sensitivity analysis with non-parametric statistics to assure that skewed data are not altering the result.

The primary outcome was also presented by subgroups (see Supplement Table 2 ) formed by COVID test results (confirmed PCR-positive vs. exposed to someone PCR-positive vs. other), age groups (18-35, 36-50, > 50), sex, and days from symptom onset to entry. These were the a priori subgroups defined in the Version 1.0 of the protocol.

Medication adherence was captured on study day 5. Another subgroup of interest is comparing the treatment groups for change in symptom severity score after day 5 by adherence reported at day 5 (<= 75% versus > 75%). For adherence, 75% equates to taking 15 or more of the 19 tablets of study medicine on time. Although not specified in the protocol, additional subgroup analyses were considered based on Zinc and Vitamin C use. These are presented due to the general public interest in zinc combined with hydroxychloroquine.

Note on the protocol:

The protocol is a combined protocol for both this trial, and our recently published companion trial using hydroxychloroquine as post-exposure prophylaxis (4) . Some endpoints and statistical analyses refer to data published in the companion trial and are not applicable to this treatment trial. Table 9 provides the comparison between those providing data for primary endpoint with follow up surveys versus not providing any follow up symptom severity data. Estimated in longitudinal mixed model adjusted for baseline severity score. P-value for trend of continuous variables are in parentheses. Medication adherence, zinc, and vitamin C use are post-hoc subgroups not specified in the protocol but identified by blinded investigators, prior to data analysis. The diagnostic testing results include results which returned after enrollment.

1 Duration of antecedent symptoms. As a continuous variable, there was not a significant interaction between randomized treatment group and days of antecedent symptoms (interaction P=0.93). Without considering any potential interaction or adjusting for multiple comparisons, there was a marginally statistical difference in those with 1-2 days of symptoms, with receiving hydroxychloroquine being superior to placebo. We did not appreciate a biologic plausibility that would explain such a response versus random variation within a small subgroup. While randomization is equal in the overall trial; for small subgroups randomization may not be equal.

Comparing those randomized to hydroxychloroquine to those randomized to placebo among those with 1-2 days of symptoms, hydroxychloroquine participants were more likely to have probable disease (73% vs. 56%) and less likely lab-confirmed disease (27% vs. 44%) (P=0.04); older by a median of 5 years (P=0.03), and had similar study medicine adherence (P=0.57) and biologic sex (P=0.13).

We would caution against over-interpretation of subgroup analyses. Notably, in the post-hoc subgroup analysis on medication adherence, those with >75% adherence randomized to hydroxychloroquine did statistically better than placebo . However, this group did not improve any more than the non-adherent hydroxychloroquine group nor the non-adherent placebo group. Another subgroup that could be over-interpreted is symptom duration. There was no difference in the treatment effect across the three groups (i.e. heterogeneity) of symptom duration (P=0.28). Thus, it would be generally inappropriate to select one subgroup, such as those with 1-2 days of symptoms, to claim a different effect based on a marginal p-value (5) . Both of these point to problems with subgroup analyses where small sample sizes, which may not be equally randomized within the subgroup, can be over-interpreted (5) . When performing multiple subgroup analyses, the probability of a false positive finding can be substantial (5, 6) .

Noteworthy for future research, the resolution of symptom severity slowed with increasing age, yet those >50 years of age had the largest, non-statistical relative difference with hydroxychloroquine use (23.6%). This was a small subgroup, but this may be a target population for future outpatient trials. There was no association between the presence of reported medication side effects and reported COVIDcompatible symptoms. 

Trial enrollment occurred in 40 U.S. States and the Canadian provinces of Quebec, Manitoba, and Alberta. The percentages are of the total trial enrollment of 491 participants.

The stacked bar graph distinguishes the relative proportions of those with presentation of cough, fever, or shortness of breath versus only other Covid-19 related symptoms (e.g. myalgia, fatigue, anosmia, sore throat, headache, chills, or rigors).

There is no statistical difference between the groups in the proportion of participants with symptoms at Day 14 (P=0.21). An alternative way to consider the magnitude of effect is for every 50 people treated with hydroxychloroquine, 1 additional person might be symptom free at day 5, 3 additional persons might be symptom free at day 14, and 20 persons would experience medication side effect(s) with 4 discontinuing hydroxychloroquine due to intolerability in order to achieve this possible effect.

The figure displays the mean (95%CI) change in symptom severity score from baseline. The change in symptom severity score was normally distributed at each time point (Supplement Figure 4 ).

Histograms demonstrate approximately normal distribution for the change in symptom severity score from baseline for each study survey day. Thus, using parametric statistics with the assumption of normal distribution is appropriate.

There was no impact on reducing secondary transmission. Of 319 participants with someone else living in their household, only 52 (16.3%) others became ill by day 14 (17% (29/168) hydroxychloroquine vs. 15% (23/151) placebo. The absolute difference is 2.0% (95%CI, -6.1% to 10.1%; P=0.65).

Interim-20-ID-01: Standardized surveillance case definition and national notification for 2019 novel coronavirus disease (COVID-19)

World Health Organization. Global Surveillance for human infection with coronavirus disease (COVID-19)

Postexposure prophylaxis or pre-emptive therapy for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2): study protocol for a pragmatic randomized-controlled trial

A Randomized Trial of Hydroxychloroquine as Postexposure Prophylaxis for Covid-19

Statistics in medicine--reporting of subgroup analyses in clinical trials

The challenge of subgroup analyses--reporting without distorting