Preventing disease after exposure to COVID-19 using hydroxychloroquine: A summary of a protocol for exploratory re-analysis of age and time-nuanced effects.

BACKGROUND: A recently published randomized trial (Boulware et al., 2020, NCT04308668) of hydroxychloroquine (HCQ) for post-exposure prophylaxis found a reduction in Covid-19 of 17%. In the context of ambitious powering to detect a 50% reduction, this non-statistically significant finding could translate to a reduction of 22,000/130,828 cases (CDC 8/12/20) among US health care workers (HCW), impacting trajectory and resource utilization models that drive decisions on lockdowns and social distancing. Data found only in the appendix of Boulware et al. suggested greater differences in the effect of HCQ among sub-groups. There were reductions (36%) in younger (<35 years) and increases (110%) in older (>50 years) subjects. Our preliminary analysis revealed a significant negative correlation (slope -0.211, CI -0.328-0.094, p=0.016) between treatment lag and disease reduction, reaching 49% when initiated within one day (RR 0.51, CI 0.176-1.46, p=0.249). There were also differences in disease reduction by HCQ by type of exposure (HCW - 8% vs. household contacts - 31%; RR 0.691, CI 0.398-1.2). The definitions of exposure severity did not discriminate between the numbers or duration (> 10 minutes) of exposures. Differences between exposure types may result from younger HCW and higher risks in less trained household contacts with little access to advanced PPE. The ex-protocol use of zinc and ascorbic acid were likely confounders, as was the possibly active folate placebo. Exploratory reanalysis of the raw dataset may inform an age- and stage- nuanced approach to COVID-19 using HCQ testable by prospective studies and may provide insight into the various proposed mechanisms of HCQ. OBJECTIVES: To conduct an exploratory re-analysis of the de-identified raw dataset from a randomized study of the use of HCQ for post-exposure prophylaxis of COVID-19 with view to further defining: a) The time dependent effect of HCQ, b) The age dependent effect of HCQ; c) The sub-stratification of time- and age-dependent effects by exposure type and risk level, as well as by the use of zinc and ascorbic acid. d) The design of future clinical trials to test the hypotheses generated by this study. METHODS: Should granularity of data (by age, time-lag, level and type of exposure) be greater than that originally reported, Fisher Exact test will be used to compare the incidence of COVID-19 in HCQ- and control groups, for each sub-group stratification. Since the degree of loss of data granularity due to de-identification is yet unknown, exploratory analyses involving other demographic characteristics cannot be planned. Where sufficient data granularity exists, univariate regression analyses will be conducted to examine the effect of age- and time lag on any effect of HCQ. The possibility will be explored of conducting multivariate Cox regression analyses with propensity score matching to examine observational data relating to the use of zinc and ascorbic acid. This analysis will be expanded should a dataset from a similarly designed study (Mitja et al., 2020, NCT04304053), with directionally similar results, become available. This protocol was devised using the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) incorporating the WHO Trial Registration Data Set. Study Status: Protocol version 1.1, (August 19 2020) Protocol registered at: OSF Registries August 19 2020 Registration doi: https://doi.org/10.17605/OSF.IO/9RPYT