Mild COVID-19 patients were hospitalized during March and April 2020 at the Bamrasnaradura Infectious Diseases Institute (BIDI), Department of Disease Control, Ministry of Public Health, Thailand. Patients were hospitalized when diagnosed as having mild COVID-19 confirmed with SARS-CoV-2 RT-PCR as per local regulation. Patients received either SOC (medication for alleviating symptoms), or a high loading dose of CQ (ldCQ), or CQ without loading dose plus LPV/r (CQLPV). Subjects who developed pneumonia were discontinued from the cohort to receive appropriate treatment. The ldCQ group is defined as receiving a 1,000 mg of CQ as a loading dose, then 500 mg twice daily for 4 days, and followed by 375 mg twice daily for 9 days or until RT-PCR negative. Subjects in the CQLPV group received the same CQ dosing schedule and a standard dose of LPV/r. Nasopharyngeal swabs were collected from the patients on the date of admission and every 2-3 days during hospitalization. A blood sample was collected from patients in the ldCQ and CQLPV groups 6 days after initiation of CQ treatment to verify the presence of physiological levels of CQ. We extracted baseline demographic data, clinical presentation and treatment outcome from medical records, when available. Nasopharyngeal swabs and blood samples were kept at -80C at BIDI and thawed once before sending to the Armed Forces Research Institute of Medical Sciences (AFRIMS) for further laboratory analyses, and a second time if viral isolation was necessary. All specimens were coded and anonymized prior to shipping.

Patients admission and discharge were at least dependent on a qualitative SARS-CoV-2 real-time RT-PCR test done at BIDI. Nasopharyngeal specimens were tested with either novel CoV 2019 real-time PCR (Jiangsu Bioperfectus Technologies, Taizhou, China) or real-time fluorescent RT-PCR (BGI technology, Shenzhen, China) for patients admitted on 31 Mar 2020, or earlier. Patients admitted after March 31, 2020 were tested with the Cobas SARS-CoV-2 qualitative assay using cobas 6800 systems (Roche Molecular Systems, Basel, Switzerland).

Quantitative real-time RT-PCR to measure SARS-CoV-2 genome equivalents (GE) was performed at the Department of Virology, AFRIMS, Bangkok, Thailand. The assay was modified from the US-CDC method, targeting the N gene (Lu et al., 2020). Briefly, three separate reactions were carried out for each extracted RNA sample, two for detection of SARS CoV-2 N1 and N2 targets, and one for human RNase P gene (RP) target using SuperScript III Platinum One-Step Quantitative RT-PCR System (Invitrogen, USA) and 2019-nCoV CDC qPCR RUO kit ^{Integrated DNA Technologies (IDT) USA}. Human RP gene detection was used as extraction control. Each 25-µl reaction contained 12.5 µl of 2x Reaction Buffer containing 0.4 mM of each dNTP and 6 mM MgSO₄, 1.5 µl of combined primer/probe mix for N1, N2, or RP (2019-nCoV CDC qPCR RUO kit), 0.5 µl of RNaseOUT (10 U/µl, Invitrogen, USA), 0.5 µl of 1:10 diluted ROX Reference Dye, 0.5 µl of SuperScript III RT/Platinum Taq, 4.5 µl of nuclease-free water, and 5 µl of extracted RNA sample. The single step RT-PCR consisting of a 30-min RT step at 50^oC, 2 min of Taq inhibitor inactivation at 95^oC, followed by 45 cycles of PCR at 95^oC for 15 sec and 55^oC for 30 sec was performed on an Applied Biosystems 7500 Fast Real-time PCR Instrument (Thermo Fisher Scientific, USA). The SARS-CoV-2 GE per sample was derived from standard curves of serial dilutions of in vitro SARS-CoV-2 RNA transcripts (IVTs).

SARS-CoV-2 isolation was done in a cell culture system at AFRIMS biosafety level 3 laboratory. The selected nasopharyngeal specimens were either specimens that were positive after a period of undetectable quantitative RT-PCR result, or specimens with at least a 2 log₁₀ viral load increase from the previous one. Virus isolation was performed using Vero E6 (monkey kidney epithelial cells) and A549 (human lung epithelial cells) cell lines with Eagle’s minimal essential medium (EMEM) containing 2-5% heat inactivated fetal bovine serum, 1% L-glutamine, 1% penicillin/streptomycin, 40 µg/ml gentamicin and 0.25 µg/ml fungizone. Cytopathic effect (CPE) was observed daily before harvesting or sub-culturing for 3 passages. Cell culture supernatant was harvested once the CPE was observed or day 5 at latest and tested using SARS-CoV-2 RT-PCR for confirmation.

Plasma levels of CQ were measured on day 6 post initiation of CQ treatment. Concentrations of CQ and desethyl-CQ were measured by Waters Acquity UPLC™ equipped with Xevo® G2-XS QToF (Waters Crop., MS, USA). 100 µl of plasma samples were extracted by adding 200 µl of iced-cold acetonitrile:methanol (25:75), incubated under BSC for 30 min, then 1-min vortexed and centrifuged at 10,000 rpm for 10 min. The supernatant was filtered through PTFE and transferred to UPLC vial. Chromatographic separation was performed on a Waters ACQUITY UPLC BEH C18 column (50x2.1 mm, 1.7 µm particle size) with pre-column of the same material (5x2.1 mm, 1.7 µm particle size) at a flow rate of 0.40 ml/min. A gradient elution composed of A) 5mM ammonium acetate pH 4.5 in water and B) 5 mM ammonium acetate pH 4.5 in acetonitrile: methanol (50:50 v/v). The column temperature was set at 40°C and injection volume was 5 µl. Compounds were ionized in positive ion mode with electrospray ionization source. The multiple reaction monitoring (MRM) was used to monitor the mass transition (m/z) of 320.33à142.22, 292.28à114.18, 336.20à247.23 and 379.29à271.22 for chloroquine, desethyl chloroquine, hydroxyl chloroquine and mefloquine (as internal standard), respectively. The MassLynx™ Software (Waters Crop., MS, USA) was used to operate the instruments, integration and quantification.

Of the 51 patients in this study, 25 received SOC (Table 1, Figure 1), 9 received high dose CQ plus LPV/r (CQLPV) and 17 received high dose CQ with loading dose (ldCQ). Most subjects were of working age and healthy. There were 3 patients with comorbidities: one pregnancy and one hypertension in the SOC group and one hypertension (a 63 years, male who later developed pneumonia) in the ldCQ group (Table 1). Common clinical features at admission were fever (n =14, 27%), cough (n =30, 59%), nasal congestion (n =30, 59%), rhinorrhea (n =19, 37%), myalgia (n =19, 37%), sore throat (n=17, 33%), and diarrhea (n =3, 6%).

Figure 1 shows the timelines of nasopharyngeal swabs for subjects in each group. Administration of ldCQ was sooner than CQLPV, 5 days versus 9 days PSO, respectively (Table 1). Median (interquartile range, IQR) duration of hospitalization was 12 (8 - 17) days. Three subjects were admitted for longer than 30 days, with the longest being hospitalized for 45 days. Of the 51 subjects, 39 (76%) were discharged after two consecutive negative RT-PCR results. The remaining 12 subjects (all in SOC) were discharged after resolution of symptoms and one negative PCR.

Detectable RNA were highest early on and declined over time (Figure 2). Individuals discharged with differing criteria had similar viral shedding durations. In subjects discharged after two consecutive negative RT-PCR (2^-PCR), the median number of days PSO with detectable RNA was 13 (95%IQR: 6.0, 54.5) with a long tail to the right driven by one individual (max: 60). This was comparable to the viral load kinetics in the subgroup discharged after only one negative PCR (1^-PCR), 14 days (95%IQR: 10.3, 32.6). The univariate hazard ratio of RNA dropping below detectable levels, 2^-PCR compared to 1^-PCR, was indifferent from one (95%CI: 0.81 to 4.16, p=0.13). Generalized additive models of log₁₀GE over time were close to linear in both groups (Figure 2) suggesting patterns of exponential decay in the kinetics. Regressing log₁₀GE on days PSO found the rate of decline to be 9% each day for the 2^-PCR group (95%CI: 4%, 12%) and 13% (95%CI: 9%, 16%) for the 1^-PCR group.

RNA remained detectable for a median of 18.0 days (95%IQR: 8.2, 49.4) in the CQLPV group and 16.0 days (95%IQR 5.0, 34.6) in the ldCQ group (Figure 1 and Table 1). These intervals overlapped with those observed in the SOC. We assessed the difference using a multivariate Cox-proportional hazard model to adjust for confounding factors. Following the stepwise removal of covariates, the hazard ratios suggested trending but not significant changes in viral shedding duration in the ldCQ (Figure 3). Adjusted for the shorter duration in individuals discharged with 2^-PCR and presence of fever at enrollment, the chance of RNA remaining detectable in the ldCQ group was 1.88-times the SOC (95%CI: 0.95 to 3.72) while the chance was 0.50-times for CQLPV (95%CI: 0.23 to 1.09).

The relationship between receiving CQLPV and the quantified genome equivalents (GE) in the log-linear regression was congruent with the inference from detectable RNA durations (insignificant effects on GE decline, 95%CI: -2%, 23%), Figure 4. In contrast, each day of ldCQ treatment improved the GE decline by 26% (95%CI: 17%, 35%). Each day of delay in CQ administration, hindered the effects by 23% (95%CI: 6%, 43%). When treated with SOC, GE declined 77% (95%CI: 16%, 95%) each day corresponding to RNA levels dropping below 0.1% of the levels at symptom onset by 3 to 40 days PSO. These effects were adjusted for variables retained in the multivariate regression after stepwise removal including fever at enrollment and intrinsic characteristics of individuals. Individuals with fever at enrollment had 16-times the GE of those without (95%: 4.17, 60.36) and showed 9% less in decline each day (95%CI: 2%, 15%). Except for males exhibiting 7% more decline each day than females (95%CI: 2%, 12%), other variations that achieved statistical significance had marginal effects on the decline.

To ensure that CQ was present at therapeutic levels in the CQ treatment groups, we measured the plasma level CQ and its metabolite, desethyl-CQ (dCQ) (Table 1). Median trough levels of CQ in the ldCQ and CQLPV groups were 626 ng/dL and 480 ng/dL, respectively. Of the 15 subjects who had serum CQ measurements, twelve (80%) had levels above 360 ng/dL, which are deemed adequate as per previous pharmacokinetic modelling study 29. There was no difference in viral kinetics between subjects who had adequate CQ level and those who did not.

To verify that the detected genome equivalents following viral load troughs were replication competent, we cultured specimens collected at the troughs and the subsequent resurgences.

Of 49 selected specimens, we successfully isolated SARS-CoV-2 from four samples acquired from three subjects between 12 to 24 days PSO. PCR’s cycle threshold (Ct) of these isolates collected on 12, 23, 18, and 24 days PSO were 26, 27, 36 and 36, respectively.

Adverse events in each treatment group are shown in Table 1. Common adverse events were nausea, and vomiting. Medications to relieve or prevent adverse events were available to all patients. The events were generally mild to moderate as per the DAIDS grading table 30, except for three subjects, which led to the termination of CQ treatments. Two subjects in the CQLPV group had grade 3 generalized skin rash after receiving antivirals starting for at least 10 days. Their skin biopsy showed acute generalized exanthematous pustulosis (AGEP). Lymphocyte transformation tests were negative with CQ and LPV/r. One subject in the ldCQ group had grade 2 nausea and vomiting. All three subjects were negative for SARS-CoV-2 by PCR prior to termination.