Background

The West African Ebola virus disease (EVD) epidemic of 2014 to 2016 resulted in at least 28,616 cases and at least 11,310 deaths [1]. Case management in prior Ebola virus outbreaks has generally focused on isolating patients with EVD, infection prevention and control procedures, and various degrees of basic supportive care [2]. During the West African outbreak, management of EVD involved progressively more advanced supportive care for patients [3,4,5]. At the onset of the outbreak, mortality was approximately 74% [6], but eventually fell to 31–37% [7, 8], perhaps due to improved early case finding and supportive care. Specific anti-Ebola virus treatment was only rarely available in West Africa. In contrast, among patients evacuated to European and American hospitals who received intensive care and anti-Ebola virus treatments, mortality was 18.5% [3, 9,10,11,12,13].

In prior outbreaks, there was little support to conduct trials, which meant there was limited evidence to guide clinical care during the recent West Africa outbreak. Several anti-Ebola virus treatments had been offered and administered to patients in West Africa, the United States and Europe. Most of these treatments had only recently been developed, not administered as part of a methodologically rigorous study or trial design, and therefore their net effects are unknown. To assess their potential benefit, we systematically reviewed the literature for all clinical studies that included specific anti-Ebola virus treatments administered to patients with EVD and reported on mortality, symptom duration, or adverse effects.

Methods

There is no published protocol for this review.

Literature search

With the assistance of a medical librarian, we searched MEDLINE, EMBASE, Global Health, Cochrane Library, African Index Medicus, and World Health Organization Global Index Medicus (WHOLIS) from inception to 9 April 2018. In MEDLINE and EMBASE, controlled vocabulary terms were combined with keywords for EVD and a broad range of study design terms, including but not limited to a sensitive search filter for randomized clinical trials (RCTs) of therapy [14, 15]. Full details of the searches are available in Additional file 1. We also searched Google, Google Scholar, and trial registries’ websites [16,17,18,19]. The reference lists of all relevant retrieved manuscripts were screened and hand-searched, and Ebola clinical care experts were consulted to identify any additional studies.

Selection criteria

We included studies with at least one patient with confirmed EVD of any age who received a therapy specifically targeting Ebola virus itself or its clinical consequences (including blood component-based strategies) and reporting on at least one outcome of interest (mortality, symptom duration after anti-Ebola treatment initiation, and adverse effects of the treatment). Eligible study designs included RCTs, non-randomized single-arm intervention studies (with or without a control group), prospective and retrospective cohort studies, and case reports and case series, without regard to publication language or date. We excluded studies of supportive care therapies that did not specifically target the Ebola virus (e.g. intravenous fluids, electrolyte and metabolic support, and organ-supportive care such as mechanical ventilation and renal replacement), vaccines for EVD primary prevention or post-exposure prophylaxis, and pre-clinical studies (e.g. involving animals or cell lines or computer models) of anti-Ebola treatments.

Two review authors (JSL and HYK), independently and in duplicate, screened titles and abstracts of retrieved citations and independently assessed full-text manuscripts of citations considered potentially eligible by either reviewer. Disagreements were resolved by consensus through discussion with two additional authors (NKJA and RAF).

Data extraction and quality appraisal

Two review authors (JSL and HYK) independently and in duplicate extracted data, including patient baseline characteristics (age, sex), study methods (design, eligibility criteria, and for RCTs, method of randomization, allocation concealment and blinding), study interventions and co-interventions, and clinical outcomes of interest (mortality at the latest time point available, adverse effects).

Three review authors (JSL, KT, RAF) assessed the risk of bias in the included studies using the Cochrane Collaboration’s Risk of Bias 2.0 tool for RCTs [20] and the ROBINS-I tool for non-randomized intervention and cohort studies [21]; a fourth author (NKJA) verified selected methodological details of these studies.

Statistical analysis

Where meta-analyses were not possible, we report the results as stated in the included studies. For each anti-Ebola virus treatment, we calculated the number and proportion of patients who died. If warranted on the basis of a sufficient number of trials (greater than or equal to 3) and sufficient clinical and methodological homogeneity, we planned meta-analyses of studies of the same intervention, using risk ratios and 95% confidence intervals (CI) to summarize dichotomous outcomes and weighted or standardized mean differences for continuous outcomes. We planned to assess between-study statistical heterogeneity using the I2 measure [22] and to use inverse-variance weighted random-effects models [23] for all meta-analyses to incorporate both between- and within-study variation. We considered P ≤ 0.05 (two sided) as statistically significant.

Results

Study selection

From 11,257 citations and studies listed in trial registries, 55 studies were selected for full-text review, of which 35 met eligibility criteria (Fig. 1): 1 RCT [24], 8 non-randomized intervention and cohort studies [25,26,27,28,29,30,31,32], 9 case series [11, 33,34,35,36,37,38,39,40], and 17 case reports published in 16 citations [10, 13, 41,42,43,44,45,46,47,48,49,50,51,52,53,54]. These 35 studies collectively examined 21 anti-Ebola agents, including 9 antivirals, 6 blood- or blood component-based therapies, 3 monoclonal antibody treatments, 2 vascular leak syndrome treatments (previously described for other indications [55, 56]), and 1 antimalarial drug, which we included because of the possibility of anti-Ebola virus effects [57]. Thirty-one eligible studies, published in 30 citations, were conducted during the 2014–2016 West African outbreak [10, 11, 13, 24,25,26,27,28,29,30,31,32, 36,37,38,39,40, 42,43,44,45,46,47,48,49,50,51,52,53,54] and reported on 4.8% (1377/28616) of all patients with a diagnosis of EVD.

Fig. 1
figure 1

Flow of studies through the systematic review

Full size image

RCT and non-randomized studies

Details of the setting, the intervention and control groups from one RCT and 8 non-randomized intervention studies of anti-Ebola therapies are reported in Table 1. The 8 non-randomized intervention studies include 3 single-arm intervention studies with concurrent controls, 3 single-arm intervention studies with historical controls, 1 single-arm uncontrolled intervention study, and 1 retrospective cohort study. The only RCT [24], which examined an anti-Ebola virus monoclonal antibody (ZMapp), enrolled 72 patients (0.25% of all patients with Ebola from the West African outbreak). Although the trial’s risk of bias is low (Table 2), the number of patients enrolled is small and the results are accordingly imprecise (Table 3). Mortality was 22% (8/36) in the ZMapp group and 37% (13/35) in the standard care group. Using Bayesian analysis, the posterior probability that ZMapp was superior to usual care was 91.2%, failing to meet the pre-specified threshold of 97.5%. Frequentist analysis showed a statistically non-significant risk difference of 15% favouring ZMapp (95% confidence interval [CI], − 36% [favouring ZMapp] to 7% [favouring usual care]).

Table 1 Randomized trial and non-randomized studies of anti-Ebola therapies
Full size table
Table 2 Risk of bias in a randomized trial and non-randomized studies of anti-Ebola therapies
Full size table
Table 3 Design and outcomes in a randomized trial and non-randomized studies of anti-Ebola therapies
Full size table

Risk of bias in all non-randomized single-arm intervention studies of convalescent plasma [28], whole blood [26], favipiravir [29], interferon-β-1a [27], artesunate-amodiaquine [32], TKM-130803 [25], and brincidofovir [31] was at least moderate (Table 2). Comparisons of mortality and adverse events between intervention and control arms were limited by non-randomized study designs (Table 3).

In the study of convalescent plasma, the intervention was associated with mortality of 31% (26/84) compared to 38% (158/418) in the historical control group, with an adjusted odds ratio of death of 0.88 (95% CI, 0.51 to 1.51) [28]. Limitations of this study include unknown level of neutralizing antibodies in plasma, small sample size in the intervention group, lack of data on the delivery of co-interventions and supportive care, and inclusion of historical controls. Among patients treated with convalescent whole blood [26], mortality was 28% (12/43) compared to 44% (11/25) in a concurrent standard care group. However, the intervention was not randomly assigned, leading to overall moderate risk of bias from confounding due to baseline and treatment-related differences, in addition to uncertainty in original patient selection criteria.

Favipiravir was investigated in a non-randomized single-arm study (n = 126) [29]. Among 99 adult and adolescent patients evaluated, favipiravir was generally well tolerated. However, the lack of a concurrent control group and uncertainty in patient selection criteria leads to moderate risk of bias. In another non-randomized single-arm study, mortality was 44% (17/39) in the favipiravir group and 65% (55/85) in the historical control group; however, the non-random assignment of the intervention, lack of a concurrent control group, and the potential for differential between-group treatments lead again to moderate risk of bias [30].

Interferon β-1a was examined in a non-randomized single-arm study and compared to controls, finding 21-day mortality of 33% (3/9) in the interferon β-1a group and 84% (32/38) in the expanded control group [27]. The non-random assignment of the intervention, lack of a priori sample size calculation, and potential for differential between-group treatments lead to moderate risk of bias.

In a retrospective cohort study of patients with EVD during a period when artemether-lumefantrine was used in an Ebola treatment unit for empiric anti-malarial treatment, mortality was 64% (125/194), compared to 50% (36/71) during a period of drug shortage when artesunate-amodiaquine was used instead [32]. However, there is moderate risk of bias due to the potential for unmeasured residual confounding; in addition, the biological plausibility of artesunate-amodiaquine as an anti-Ebola virus agent is uncertain.

In two non-randomized single-arm studies, TKM-130803 was associated with a 14-day mortality of 75% (9/12) compared 67% (2/3) mortality in the control group [25], and brincidofovir-associated 14-day mortality was 100% (4/4) [31]. The study designs lead to moderate risk of bias.

Given the small number of studies (often only one) of any single intervention and substantial heterogeneity in study design, we did not conduct any meta-analyses.

Case series and reports

Study details for 26 case series and case reports (reported in 25 publications) are presented in Additional file 1: Table S1, including the number of patients, country of infection, location of clinical care, anti-Ebola virus treatment given, critical care interventions, and clinical outcomes. Duplicate patient descriptions were noted among 13 manuscripts. One case series described patients (some previously reported in other studies) who received care in Europe or the USA [39], and another case series described patients (some previously reported in other studies) who received care in Sierra Leone [40]. Excluding these two cases series, a total of 32 patients receiving anti-Ebola virus therapies were described in case series and case reports, with a mortality of 22% (7/32). Some patients received > 1 anti-Ebola virus agent, including blood-based therapies (n = 25), monoclonal antibodies (n = 10), antivirals (n = 19), and therapies for vascular leak syndrome (n = 3).

Among these 26 case series and case reports, 3 case series described 11 patients who received anti-Ebola virus therapies (convalescent whole blood, Virustat [acyclovir], gamma globulin, Marburg convalescent plasma) in Africa prior to the West African EVD outbreak of 2014–2016. An additional report described a patient who received anti-Ebola virus therapies (convalescent plasma, human interferon) in the United Kingdom prior to the West African EVD outbreak. Four reports described 4 patients who received anti-Ebola virus treatment in West Africa (convalescent whole blood, ZMapp, convalescent leukocytes, GS-5743, favipiravir) during the 2014–2016 outbreak. Three case series and 1 case report described 4 patients who initially received anti-Ebola virus therapies in West Africa (convalescent whole blood, ZMapp, TKM-100802) before transfer to the USA. Finally, 3 case series and 1 case report described 5 patients who received anti-Ebola virus therapies solely in the USA, and 8 case reports described 7 patients who received anti-Ebola virus therapies solely in Europe.

Additional file 1: Table S2 presents the number of patients and mortality rates of 18 patients who were treated with an anti-Ebola virus therapy in addition to receiving intensive care. Mortality was 29% (2/7) in patients who received positive pressure ventilation, 25% (6/24) in those who received central venous access or hemodynamic support, 40% (2/5) in those who received renal replacement therapy, 15% (2/13) in those who received non-convalescent blood product transfusions, and 14% (1/7) in those who received parental nutrition. No information about critical care interventions was described in the remaining 14 patients

Additional file 1: Table S3 characterizes the setting, anti-Ebola virus therapy, primary outcome, and status of 9 registered trials of anti-Ebola treatments that have not yet been completed or published as of the search date of April 2018. Crude mortality is described in 9 case series and 17 case reports of anti-Ebola virus treatments (Additional file 1: Table S4)

Discussion

In this systematic review, we identified 1 RCT, 8 non-randomized single-arm intervention studies and cohort studies, 9 case series and 17 case reports evaluating 21 anti-Ebola virus therapies. The single RCT evaluated the monoclonal antibody ZMapp, had a low risk of bias, and found a statistically non-significant decrease in mortality. All interventions evaluated in non-randomized studies, including convalescent whole blood or plasma, interferon β-1a, favipiravir and the antimalarial artesunate-amodiaquine, were associated with an overall moderate risk of bias, and in some cases serious or critical risk of bias due to confounding, severely limiting inferences regarding treatment effects. The non-randomized evaluations of brincidofovir and TKM-130803 do not provide evidence to support future evaluation.

The strengths of this study include a comprehensive search of published and available non-published clinical literature, triplicate and independent assessment of risk of bias according to the Cochrane framework, and duplicate independent data abstraction. Our review is the first to summarize all such literature in humans with EVD. A previous systematic review included drug screening and pre-clinical studies and fewer clinical studies (n = 9) than we included, and focused on identifying existing drugs with potential therapeutic effect [58]. Other reviews provide further information on selected clinical studies of anti-Ebola virus therapies [59, 60], in addition to details of in vitro and animal studies [59].

However, there are many limitations of such a review. We are limited in inferences due to the moderate to serious risk of bias of the majority of studies and the small number of enrolled patients, leading to one RCT and many non-randomized studies all under-powered to detect differences in mortality. The small number of studies (often only one) of any single intervention and important heterogeneity in study design precluded meta-analyses. Several agents with promising pre-clinical findings or case report-based evaluations cannot yet be evaluated for effectiveness based on existing data and study designs. These include the nucleotide analogue prodrug remdesivir (GS-5734), the monoclonal antibody cocktail REGN3470–3471-3479, and the monoclonal antibody MAb114 [61], which in addition to ZMapp and favipiravir, have been evaluated by a World Health Organization-convened independent scientific committee for monitored emergency use of unregistered and investigational interventions, while awaiting additional evidence [62]. Finally, peer review and evaluations of additional therapies may be forthcoming; however, we believe that we are unlikely to have missed any publications of evaluable treatment effects.

Evaluating the effect of treatments on clinical outcomes of patients with EVD is challenging because of its uncommon, periodic, and lethal nature. There have been few pre-existing therapies with strong evidence of potential treatment effect, making prioritization difficult for clinicians, researchers, regulators and funders. Outbreaks have occurred in resource-challenged health systems in West Africa and most recently in DR Congo, Central Africa [63], often in remote areas with delayed recognition of the outbreak’s onset [64, 65]. Small outbreaks may end before any clinical or research response. A historically high mortality rate, including among healthcare workers, leads to diminished clinical capacity for care and research, but also a reluctance to consider using the methodologically strongest RCT design because of the implication that some patients will not receive a potentially beneficial investigational agent. The variable standard of supportive care contributes to a baseline high mortality rate but also makes estimation of treatment effects difficult, possibly leading to selection of patients who are poorly responsive to investigational anti-Ebola virus agents and an inability to compare therapies across studies [6, 66]. Uniform adoption of evidence-based supportive care guidelines [67] in future outbreaks may facilitate the evaluation of anti-Ebola virus therapies.

West African nations most affected in 2014–2016 had not previously experienced an Ebola outbreak, and there was limited pre-existing Ebola-specific clinical and research capacity. With overwhelmed national healthcare systems and a slow international response, there was little opportunity to evolve interventional research programs in parallel with outbreak care. Eventually, early diagnostic and descriptive studies gave rise to an appreciation of the potential impact of supportive and specific EVD therapy [39, 66]. However, among 28,616 infected patients, fewer than 5% had any therapies described or evaluated and only 0.25% participated in a RCT.

This review is a comprehensive summary of data collected on the effects of specific anti-Ebola therapies. Given Ebola’s high mortality rate, sparse treatment options, and high capacity for spread, it is imperative that an adequate research capacity in Ebola outbreak-prone regions be developed and well-supported by the international community. Rigorous prior knowledge synthesis is critical to plan relevant future research. At this stage, the small number of patients exposed to each intervention and design-related limitations preclude strong inferences on clinical effectiveness. However, a better understanding of the paucity of supportive evidence is valuable for various stakeholders. Decision-makers confronted with EVD outbreaks in the future may use these results to prioritize or avoid system-wide delivery of certain experimental interventions. Guideline developers might use the evidence summary to make graded recommendations regarding specific anti-Ebola virus therapies. Lastly, highlighting the insufficiencies of the existing body of evidence could help researchers to design future studies for implementation during an outbreak and to prioritize experimental therapies for future evaluation.

Conclusions

In this systematic review, we found only one RCT of anti-Ebola virus therapy that was associated with a low risk of bias and a signal of a treatment effect, suggesting that the monoclonal antibody ZMapp should be prioritized for further evaluation in another EVD outbreak. Moderate to serious risk of bias and small sample sizes preclude strong inferences regarding the clinical effects of convalescent whole blood or plasma, favipiravir, interferon β-1a, and the antimalarial artesunate-amodiaquine.