Report of the first seven agents in the I-SPY COVID trial: a phase 2, open label, adaptive platform randomised controlled trial

Abstract

Background: An urgent need exists to rapidly screen potential therapeutics for severe COVID-19 or other emerging pathogens associated with high morbidity and mortality.

Methods: Using an adaptive platform design created to rapidly evaluate investigational agents, hospitalised patients with severe COVID-19 requiring ≥6 L/min oxygen were randomised to either a backbone regimen of dexamethasone and remdesivir alone (controls) or backbone plus one open-label investigational agent. Patients were enrolled to the arms described between July 30, 2020 and June 11, 2021 in 20 medical centres in the United States. The platform contained up to four potentially available investigational agents and controls available for randomisation during a single time-period. The two primary endpoints were time-to-recovery (<6 L/min oxygen for two consecutive days) and mortality. Data were evaluated biweekly in comparison to pre-specified criteria for graduation (i.e., likely efficacy), futility, and safety, with an adaptive sample size of 40-125 individuals per agent and a Bayesian analytical approach. Criteria were designed to achieve rapid screening of agents and to identify large benefit signals. Concurrently enrolled controls were used for all analyses. https://clinicaltrials.gov/ct2/show/NCT04488081.

Findings: The first 7 agents evaluated were cenicriviroc (CCR2/5 antagonist; n = 92), icatibant (bradykinin antagonist; n = 96), apremilast (PDE4 inhibitor; n = 67), celecoxib/famotidine (COX2/histamine blockade; n = 30), IC14 (anti-CD14; n = 67), dornase alfa (inhaled DNase; n = 39) and razuprotafib (Tie2 agonist; n = 22). Razuprotafib was dropped from the trial due to feasibility issues. In the modified intention-to-treat analyses, no agent met pre-specified efficacy/graduation endpoints with posterior probabilities for the hazard ratios [HRs] for recovery ≤1.5 between 0.99 and 1.00. The data monitoring committee stopped Celecoxib/Famotidine for potential harm (median posterior HR for recovery 0.5, 95% credible interval [CrI] 0.28-0.90; median posterior HR for death 1.67, 95% CrI 0.79-3.58).

Interpretation: None of the first 7 agents to enter the trial met the prespecified criteria for a large efficacy signal. Celecoxib/Famotidine was stopped early for potential harm. Adaptive platform trials may provide a useful approach to rapidly screen multiple agents during a pandemic.

Funding: Quantum Leap Healthcare Collaborative is the trial sponsor. Funding for this trial has come from: the COVID R&D Consortium, Allergan, Amgen Inc., Takeda Pharmaceutical Company, Implicit Bioscience, Johnson & Johnson, Pfizer Inc., Roche/Genentech, Apotex Inc., FAST Grant from Emergent Venture George Mason University, The DoD Defense Threat Reduction Agency (DTRA), The Department of Health and Human ServicesBiomedical Advanced Research and Development Authority (BARDA), and The Grove Foundation. Effort sponsored by the U.S. Government under Other Transaction number W15QKN-16-9-1002 between the MCDC, and the Government.

Keywords: Acute lung injury; Clinical trial; Respiratory insufficiency; SARS-CoV-2.

Fig. 1

Consort flow chart of the ISPY COVID trial. Among the 172 participants who declined consent, the number and percentage of individuals who declined to sign consent after being randomised to a particular agent were: 7 (24%) for Razuprotafib, 16 (19.3%) for Apremilast, 10 (21.3%) for FamCox, 17 (14.9%) for CVC, 31 (30.4%) for IC14, 18 (15.7%) for Icatibant, 17 (29.8%) for Pulmozyme, and 42 (15.3%) for Control. The remaining 14 individuals who declined consent were randomised to other investigational arms that were started before the discontinuation of agents reported in this manuscript. Individuals that consented to Control arm were “shared” among investigational agent arms if they were randomised in a time epoch that included the active investigational agent.

Fig. 2

Time-to-Recovery, Modified Intention-To-Treat Population. Aalen-Johansen cumulative incidence curves showing the proportion of recovered patients in the mITT population up to 60 days of follow-up post randomisation in the presence of death prior to recovery as a competing risk. Numbers below each panel show patients at risk across time. Shaded areas show 95% pointwise confidence intervals. A. Apremilast arm, B. Cenicriviroc arm, C. Icatibant arm, D. IC14 arm, E. Celecoxib/Famotidine arm, F. Dornase arm.

Fig. 3

Survival Over Time, Modified Intention-To-Treat Population. Kaplan–Meier curves for overall survival in the mITT population according to treatment arm. The x-axis shows time in days. The numbers below each panel show patients at risk over time. Shaded areas show 95% pointwise confidence intervals. A. Apremilast arm, B. Cenicriviroc arm, C. Icatibant arm, D. IC14 arm, E. Celecoxib/Famotidine arm, F. Dornase arm.

Fig. 4

Progression to Mechanical Ventilation or Death in Patients Not Mechanically Ventilated at Baseline, By Agent, mITT Population. Kaplan–Meier curves for progression to mechanical ventilation or death in patients not mechanically ventilated at baseline in the mITT population according to treatment arm. The x-axis shows time in days. The numbers below each panel show patients at risk over time. Shaded areas show 95% pointwise confidence intervals.