key: cord-0798252-47qkq7te authors: Gastine, Silke; Pang, Juanita; Boshier, Florencia A.T.; Carter, Simon J.; Lonsdale, Dagan O.; Cortina‐Borja, Mario; Hung, Ivan F.N.; Breuer, Judy; Kloprogge, Frank; Standing, Joseph F. title: Systematic review and patient‐level meta‐analysis of SARS‐CoV‐2 viral dynamics to model response to antiviral therapies date: 2021-02-28 journal: Clin Pharmacol Ther DOI: 10.1002/cpt.2223 sha: f858f072d28c2cc15c78377f1685e4f9dd8e2771 doc_id: 798252 cord_uid: 47qkq7te SARS‐CoV‐2 viral loads change rapidly following symptom onset so to assess antivirals it is important to understand the natural history and patient factors influencing this. We undertook an individual patient‐level meta‐analysis of SARS‐CoV‐2 viral dynamics in humans to describe viral dynamics and estimate the effects of antivirals used to‐date. This systematic review identified case reports, case series and clinical trial data from publications between 1/1/2020 and 31/5/2020 following PRISMA guidelines. A multivariable Cox proportional hazards regression model (Cox‐PH) of time to viral clearance was fitted to respiratory and stool samples. A simplified four parameter nonlinear mixed‐effects (NLME) model was fitted to viral load trajectories in all sampling sites and covariate modelling of respiratory viral dynamics was performed to quantify time dependent drug effects. Patient‐level data from 645 individuals (age 1 month‐100 years) with 6316 viral loads were extracted. Model‐based simulations of viral load trajectories in samples from the upper and lower respiratory tract, stool, blood, urine, ocular secretions and breast milk were generated. Cox‐PH modelling showed longer time to viral clearance in older patients, males and those with more severe disease. Remdesivir was associated with faster viral clearance (adjusted hazard ratio (AHR) = 9.19, p<0.001), as well as interferon, particularly when combined with ribavirin (AHR = 2.2, p=0.015; AHR = 6.04, p = 0.006). Combination therapy should be further investigated. A viral dynamic dataset and NLME model for designing and analysing antiviral trials has been established. Finding antivirals that target severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) will be crucial in managing 68 the ongoing pandemic. In addition to the development of novel agents, substantial efforts are underway to establish 69 whether currently available agents may be re-purposed 1 . A key biomarker for clinical antiviral activity is viral load in 70 bodily fluids and assessing a drug's or drug combination's ability to reduce viral load is an important first step in 71 identifying therapies that influence clinical outcome. 72 73 To correctly assess antiviral activity, it is first necessary to understand viral load natural history. As a rapidly progressing, 74 primarily respiratory viral infection, SARS-CoV-2 elimination from the body seems to be mainly driven by a combination 75 of innate immune response and exhaustion of target cells available for infection 2 . Observational cohort studies published 76 to date have shown that the rate of viral load decline seems slower in older patients, those with more severe disease and 77 those with comorbidities such as diabetes mellitus and immunosuppression 3, 4, 5, 6 . Interpreting these observational studies 78 requires caution because patients have often received antiviral therapies. Due to the time point of initial infection being 79 unknown, assessing viral load in response to treatment must account for time since symptom onset 7 . Since February 2020 case reports and case series of patient-level viral dynamics have been published, some of which 82 report dosing of antiviral drugs 8 . Clinical trials of antivirals and their association with viral load are also beginning to read load was reported in PCR Ct but no further information was available on the PCR assay. In this instance a conversion to 155 copies/mL was made using the mean slope and intercepts from all calibration curves. 156 157 Drug quality score 158 The second quality assessment on a 3-point scale related to reporting of the antiviral drug therapy administered: which 159 drug(s) and upon which days did patients receive the drug(s). The highest quality 1 was assigned when it was reported 160 which days each patient received each drug, or these data were provided by corresponding authors. If it was reported that 161 no antiviral was administered this was also assigned quality 1. Quality 2 was assigned when antiviral drug treatment was 162 reported, but ascertaining which days the patient had received the drugs was not possible. The lowest category, quality 3, 163 was assigned when it was not possible to determine whether or not antivirals had been administered. 164 165 166 Primary analysis of time to viral clearance using Cox proportional hazards modelling 167 The primary analysis was conducted on observed time to viral clearance, which was analysed fitting Cox proportional 168 hazards regression models with adjusted hazard ratios estimated for each covariate. We verified the assumptions of 169 proportional hazards using the Therneau-Grambsch test. 15 The data used for this analysis were limited to respiratory and 170 stool sampling sites only, as virus was found to be mostly undetectable at other sites. Furthermore, only data from patients 171 with known antiviral history (drug quality 1 and 2) were used. To assess the possible risk of bias in different drug and viral 172 load qualities, the analysis was repeated on two further subsets: Firstly, with only drug quality 1 and respiratory samples, 173 and secondly on assay quality 1 data only. Time to viral load dropping below the limit of detection was modelled with Cox proportional hazards regression in R 176 (version 3.6.3) 16 193 where β is the rate at which target cells become infected in the presence of virus, δ is the death rate of infected cells, ρ is 197 the rate of viral production from infected cells and c is the rate of clearance of free virus. This article is protected by copyright. All rights reserved This article is protected by copyright. All rights reserved 250 Results of the systematic search are given in Figure 1 , and details of included papers in Table S2 . Sixty-seven of these patients did not receive antivirals. The NLME model fits to the overall data, stratified by sampling site, are provided in Supplementary Table S3 This article is protected by copyright. All rights reserved 282 were most used and also most used in combination. Adding interaction terms for interferon plus lopinavir/ritonavir, 283 interferon plus ribavirin and lopinavir/ritonavir plus ribavirin in the Cox proportional hazard regression analysis showed a 284 trend towards synergy between interferons and ribavirin in the full dataset (AHR = 6.04, p=0.006 Figure 3 ), as well as in 285 the additional analysis taking in quality assessments to account for potential bias: respiratory data limited to drug quality 286 and in data limited to only viral load quality 1 data (Supplementary Figure S4 and Figure S5 ). Median sampling 287 frequency in the main survival dataset was 1 day but there was a skewed distribution of sampling frequencies with the 288 mean being 1.9 days and 4.8% of sampling frequencies being greater than 3 days. The main analysis was repeated 289 excluding events with sampling frequencies over 3 days to check for potential bias caused by interval censoring, but the 290 main effect sizes were similar ( Figure S6 ). Covariate relationships and drug effects were explored through NLME modelling with parameter estimates of the model 293 given in Supplementary Table S4 (Table S4) . Implementing an 300 additional synergy term, as detected in the Cox proportional hazard model, did not improve the NLME model. is detected in low levels in blood, urine, ocular secretions and breast milk ( Figure 2 ). In addition to simulating the 313 expected trajectory of viral load at each site, we were able to simulate the percentage of samples expected to be below a 314 typical detection limit of 10 copies/mL ( Figure 2 ). From this it can be seen, that from day 10 post symptom onset over a This article is protected by copyright. All rights reserved 315 quarter of upper respiratory samples have undetectable viral load. This emphasises the importance of early antiviral 316 therapy, and for Phase II trials using viral load as an endpoint to commence therapy in the first few days of symptom onset 317 in order to reliably differentiate antiviral effects from natural viral decline ( Figure 5 , Supplementary Table S5) . Although we followed PRISMA guidelines on individual patient-level meta-analysis methods, registered our review with 320 PROSPERO and prospectively published our analysis protocol prior to finalising our search, by including data from case 321 reports, case series and clinical trials it could be argued that the heterogeneous inclusion criteria of these data may bias the 322 treatment effects we estimated. We therefore repeated the primary analysis on subsets of the data based on sampling site, 323 data quality and sampling frequency (Supplementary Figures S4-6 S4 and S5 ). Remdesivir did not however significantly decrease δ in the NLME model, but this is likely due to 337 the low number of included patients. Cox proportional hazard analysis, consistent across data qualities and sampling site combinations, we found either a 347 significant or trend towards significant synergistic activity of interferon plus ribavirin (Figures 3, 4, S4 ). An extensive This article is protected by copyright. All rights reserved 348 body of literature exists to show both interferon alpha and beta are synergistic with ribavirin in vitro against both SARS-349 CoV-1 and MERS-CoV 25 . This synergy, however, was not confirmed when tested in the NLME analysis, indicating that 350 the detected synergistic effect from the time to viral clearance analysis might be confounded. Correlations in the timing for 351 start of drug treatment could be one confounder that is corrected for in the NLME approach. The time-dependant analysis 352 from the NLME model suggests and additive effect for interferon and ribavirin, rather than a synergistic effect. Thus, 353 combining interferons with a nucleoside analogue, possibly remdesivir or favipiravir as less toxic alternatives to ribavirin, 354 is a potentially promising combination for viral load suppression. In our secondary analysis, we included interferon plus 355 ribavirin in the NLME model and simulations show that virus should be suppressed 2-3 days faster compared to no 356 treatment ( Figure 5 ). However, it must be noted that recent evidence from the WHO SOLIDARITY trial shows that 357 interferons were associated with a trend to increased mortality 29 whereas an unpublished press release reports inhaled 358 interferon-β to be beneficial 30 . There is a clear need for a well-designed Phase II trial on interferons in early disease to 359 confirm or refute the signal seen in our data. In our secondary NLME analysis the simplified target cell limited model provided a good fit to data from each sampling 386 site. In many cases this approximated a mono-exponential decay, but in others, particularly in lower respiratory tract, there 387 was a pronounced peak in the first days following symptom onset. The model was stable with high inter-individual 388 variability on V 0 and β, reflecting the fact that relative changes in these parameters lead to the initial part of the curve 389 either rising then falling (in situations when V 0 ≈ -β) or approximately monoexponentially declining (when V(0) >> -β). In 390 addition, we found the model to be less sensitive to changes in γ, meaning it can take a wide range of values with little 391 influence on model fit, hence we did not estimate an inter-individual variability term on it. Increasing age was associated 392 with significantly slower δ, and there was a small effect of male sex also being associated with slower δ ( The major limitation of our work is the lack of clinical trial data and lack of data on potentially important re-purposing 405 agents such as favipiravir and nitazoxanide and that only one of the authors of a major clinical trial agreed to share their 406 data 9 . Through applying quality assessment criteria on drug history and assay reporting, pre-specifying our analysis in our 407 protocol and PROSPERO registration before undertaking Cox proportional hazards and NLME modelling we aimed to 408 reduce possible bias in the heterogenous data available. Whilst we were able to extract a limited common demographics 409 set, particularly in the high-quality data subset (age, sex, disease severity, antiviral drug histories), our data may be limited 410 by other non-antiviral medications that were not fully reported in the included papers. Furthermore, as many of our 411 included papers were on patients with mild or no symptoms and only contained data on one patient reported to have died, 412 we were unable to study associations of viral load and mortality. Viral load measured by PCR is not necessarily infectious 413 virus, and recently it has been shown that only in samples above 10 7 copies/mL can SARS-CoV-2 be cultured 29 . Accepted Article Therefore, our data should preferably be used to study viral trajectories in relation to antiviral therapy rather than to infer 415 probability of transmission. This article is protected by copyright. All rights reserved This article is protected by copyright. All rights reserved Prioritization of Anti-SARS-Cov-2 Drug Repurposing Opportunities Based on Plasma and Target Site Concentrations Derived from their Established Human Pharmacokinetics Temporal profiles of viral load in posterior oropharyngeal saliva samples and serum antibody responses during infection by SARS-CoV-2: an observational cohort study Temporal dynamics in viral shedding and transmissibility of COVID-19 Factors Leading to High Morbidity and Mortality of COVID-19 in Patients with Type 2 Diabetes Novel Coronavirus-19 (COVID-19) in the immunocompromised transplant recipient: #Flatteningthecurve Viral load dynamics and disease severity in patients infected with SARS-CoV-2 in Zhejiang province, China A drug-disease model describing the effect of oseltamivir neuraminidase inhibition on influenza virus progression Viral Load Kinetics of SARS-CoV-2 Infection in First Two Patients in Korea Accepted Article This article is protected by copyright. All rights reserved 9 Triple combination of interferon beta-1b, lopinavir-ritonavir, and ribavirin in the treatment of patients admitted to hospital with COVID-19: an open-label, randomised, phase 2 trial Statement from the Chief Investigators of the Randomised Evaluation of COVid-19 thERapY (RECOVERY) Trial on lopinavir-ritonavir Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data: the PRISMA-IPD Statement Analytical sensitivity and efficiency comparisons of SARS-COV-2 qRT-PCR primerprobe sets. medRxiv 2020 Molecular Diagnosis of a Novel Coronavirus (2019-nCoV) Causing an Outbreak of Pneumonia Proportional hazards tests and diagnostics based on weighted residuals A language and environment for statistical computing (version 3.6.3). R Foundation for Epidemiologic Features and Clinical Course of Patients Infected With SARS-CoV-2 in Singapore Treatment of 5 Critically Ill Patients With COVID-19 With Convalescent Plasma Evidence from two cases of asymptomatic infection with SARS-CoV-2: Are 14 days of isolation sufficient? Clinical and virological data of the first cases of COVID-19 in Europe: a case series Characteristics of pediatric SARS-CoV-2 infection and potential evidence for persistent fecal viral shedding Sequential analysis of viral load in a neonate and her mother infected with SARS-CoV-2 Saliva is more sensitive for SARS-CoV-2 detection in COVID-19 patients than nasopharyngeal swabs Lopinavir/ritonavir did not shorten the duration of SARS CoV-2 shedding in patients with mild pneumonia in Taiwan Accepted Article This article is protected by copyright. All rights reserved 56 Evaluating the accuracy of different respiratory specimens in the laboratory diagnosis and monitoring the viral shedding of 2019-nCoV infections Persistent viral RNA positivity during the recovery period of a patient with SARS-CoV-2 infection Clinical and virologic characteristics of the first 12 patients with coronavirus disease 2019 (COVID-19) in the United States Viral dynamics of SARS-CoV-2 across a spectrum of disease severity in COVID-19 A report of three COVID-19 cases with prolonged viral RNA detection in anal swabs Assessing Viral Shedding and Infectivity of Tears in Coronavirus Disease 2019 (COVID-19) Patients Prolonged virus shedding even after seroconversion in a patient with COVID-19 Prolonged viral shedding in feces of pediatric patients with coronavirus disease 2019 Accepted Article This article is protected by copyright. All rights reserved 64 Duration of SARS-CoV-2 viral shedding during COVID-19 infection Clinical Course and Outcomes of Patients with Severe Acute Respiratory Syndrome Coronavirus 2 Infection: a Preliminary Report of the First 28 Patients from the Korean Cohort Study on COVID-19 SARS-CoV-2 Isolation From Ocular Secretions of a Patient With COVID-19 in Italy With Prolonged Viral RNA Detection The index case of SARS-CoV-2 in Scotland SARS-CoV-2 Viral Load in Clinical Samples from Critically Ill Patients Viral dynamics in asymptomatic patients with COVID-19 Factors associated with the duration of viral shedding in adults with COVID-19 outside of Wuhan, China: a retrospective cohort study Viral loads in throat and anal swabs in children infected with SARS-CoV-2 Accepted Article This article is protected by copyright. All rights reserved 72 Chronological Changes of Viral Shedding in Adult Inpatients with COVID-19 in Wuhan, China Successful treatment of a centenarian with coronavirus disease 2019 (COVID-19) using convalescent plasma Detectable severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in human breast milk of a mildly symptomatic patient with coronavirus disease 2019 (COVID-19) Clinical Significance of a High SARS-CoV-2 Viral Load in the Saliva Clinical features, isolation, and complete genome sequence of severe acute respiratory syndrome coronavirus 2 from the first two patients in Vietnam Early administration of ritonavir-boosted lopinavir could prevent severe COVID-19 Detection of SARS-CoV-2 in human breastmilk This article is protected by copyright. All rights reserved