Table 3.

Design and outcomes in a randomized trial and non-randomized studies of anti-Ebola therapies

Agent Citation	Primary outcome Selected details of design	Dates of study Reason for termination	Mortality	Adverse events
Randomized clinical trial
ZMapp PREVAIL II Writing group, 2016 [24]	28-day mortality • Adaptive design: plan to update standard-of-care treatment with the investigational drug, if efficacious in interim analysis • All patients recruited in Guinea received favipiravir as standard of care • 6 randomization strata: baseline PCR CT value (≤22 vs. > 22) and location (Liberia/Sierra Leone vs. Guinea vs. USA); in analysis, location strata changed to Liberia/Sierra Leone/USA vs. Guinea • Planned maximum sample size, n = 200 (100 per group)	March–November 2015 Trial closed in January 2016 after affected countries declared nearly Ebola free	• 28-day mortality: 22% (8/36), intervention 37% (13/35), control • Bayesian RD − 14% [95% CrI, −34% to 6%] • Bayesian RR 0.62 [95% CrI, 0.29 to 1.24] • Posterior probability that intervention was superior to control, 91.2% (below pre-specified probability threshold of 97.5%)	• Serious adverse events: 31% (11/36), intervention 37% (13/35), control; p = 0.62 • One serious adverse event (hypertension) judged to be related to the ZMapp infusion
Non-randomized single-arm intervention study with concurrent controls
TKM-130803 Dunning et al., 2016 [25]	14-day survival, excluding deaths within 48 h of ETC admission • Concurrent observational cohort for patients who did not meet additional criteria for drug infusion • Plan for randomization of eligible patients to intervention vs. control arm if number of eligible patients exceeded available treatment beds; this scenario did not happen • Planned maximum sample size, n = 100	March–June 2015 Study closed after futility boundary reached	• 14-day mortality (intervention): 75% (9/12) 79% (11/14), if the additional 2 patients who died within 48 h are included • Probability of 14-day survival, given 48-h survival, 0.27 [95% CI, 0.06 to 0.58] • 2 of 3 patients in the observational cohort died	One patient had worsening tachypnea within 48 h of the second TKM-130803 infusion; event felt to be compatible with progression of EVD
Convalescent whole blood Sahr et al. 2017 [26]	Not stated; mortality and other outcomes reported • Patients who did not consent to intervention were recruited into control arm • No sample size calculation	December 2014–April 2015 Reason for stopping not stated	• Mortality: 28% (12/43), intervention 44% (11/25), control • One patient who received intervention dropped out and is excluded from the denominator • ORsurvival with intervention, 2.3 (95% CI, 0.8 to 6.5)	None
Interferon β-1a Konde et al. 2017 [27]	Clearance and/or reduction in viral RNA from day 1 to day 10, as determined by PCR and/or quantitative real time PCR • 21 control patients admitted to the same ETC during the same time period as the treated patients • 17 more control patients selected; they matched treated patients on specified criteria and received care in a Guinean ETC; time period of treatment of these additional controls not stated • ‘sample size of 30–50 chosen to assess feasibility’	March–June 2015 Reason for stopping not stated	• 21-day mortality: 33% (3/9), intervention 84% (32/38), all controls 81% (17/21), controls from same ETC log-rank p = 0.026 comparing intervention to 21 controls • Multiple regression models reported ORmortality with intervention, adjusted for CT, 0.13 (p = 0.022; CI not reported)	Not reported
Non-randomized, single-arm, intervention study with historical control
Convalescent plasma van Griensven et al., 2016 [28]	14-day mortality (including deaths from days 3 to 16 after PCR confirmation of EVD) • Control patients treated in the same ETC before the start of the study • Planned sample size, n = 260 (130 per group)	February–August 2015 Study closed in July 2015 due to low caseload	• Mortality 3–16 days after diagnosis: 31% (26/84), intervention 38% (158/418), control 34% (30/88), intervention, if 15 patients who also received it are included, of whom 4 died before day 3 • ORmortality with intervention, adjusted for age and CT, 0.88 (95% CI, 0.51 to 1.51)	• No serious adverse events • 8 patients had adverse reactions during or early after the infusion 5 increase in temperature 4 itching or skin rash 1 nausea 2 reactions requiring reduction in infusion rate
Favipiravir Sissoko et al., 2016 [29]	14-day mortality (changed to ‘on-trial mortality’ to include 1 patient who died at day 17) • Control patients (n = 540) from database of MSF ETCs in forested Guinea • Initial sample size, n = 180 (60 per group defined by age and time of treatment after symptom onset; definition of strata changed to include age and CT)	December 2014–April 2015 Study closed due to low caseload	Mortality: 54% (60/111), intervention 58% (315/540), control 51% (64/126), intervention, if 15 patients who also received it are included, of whom 4 died Adjusted analysis not reported	• Vomiting within 30 min of pill intake occurred in 30 instances (2%) in 21 patients. • No severe adverse events
Favipiravir Bai et al., 2016 [30]	Mortality (time not specified) • Intervention patients treated 1–10 November 2014 • Control patients treated in the same ETC, 10–30 October • Intravenous fluids limited in the study ETC	10 October-10 November 2014 Study closed when research team rotated out of ETC	Mortality: 44% (17/39), intervention 65% (55/85), control [unadjusted p = 0.027] Adjusted analysis not reported	Not reported
Non-randomized single-arm intervention study without controls
Brincidofovir Dunning et al. 2016 [31]	14-day mortality • No control group • Planned maximum sample size, n = 140	January 2015 Study closed because manufacturer stopped participation in all studies of brincidofovir for EVD	14-day mortality: 100% (4/4)	• No serious adverse reactions • No serious unexpected serious adverse reactions • Concern that intervention might have contributed to persistent diarrhea in 1 patient
Retrospective cohort study
Artesunate-amodiaquine Gignoux et al. 2016 [32]	Mortality (time not specified) Cohort study based on natural experiment: 71 patients prescribed artesunate–amodiaquine because of shortage artemether–lumefantrine (given to 194 patients)	June–October 2014 Study closure: not applicable	• Mortality: 50.7% (36/71), artesunate-amodiaquine; 64.4% (125/194), artemether-lumefantrine; 65.1% (41/63), no anti-malaria drugs • Adjusteda RRmortality 0.69 [95% CI 0.54 to 0.89]	Not described

Data are as reported in the primary studies. Abbreviations: CI confidence interval, CrI credible interval, CT cycle time, ETC Ebola Treatment Center, EVD Ebola virus disease, IV intravenous, MSF Médecins Sans Frontières, N/A Not Applicable, OR odds ratio, PCR polymerase chain reaction, RD risk difference, RNA ribonucleic acid, RR risk ratio

^aAdjusted for age, sex, CT value, time from symptom onset to admission, malaria test result, receipt or no receipt of IV fluids, and number of inpatients at the ETC on the day of patient admission