Use of Viremia to Evaluate the Baseline Case Fatality Ratio of Ebola Virus Disease and Inform Treatment Studies: A Retrospective Cohort Study

BACKGROUND The case fatality ratio (CFR) of Ebola virus disease (EVD) can vary over time and space for reasons that are not fully understood. This makes it difficult to define the baseline CFRs needed to evaluate treatments in the absence of randomized controls. Here, we investigate whether viremia in EVD patients may be used to evaluate baseline EVD CFRs. METHODS AND FINDINGS We analyzed the laboratory and epidemiological records of patients with EVD confirmed by reverse transcription PCR hospitalized in the Conakry area, Guinea, between 1 March 2014 and 28 February 2015. We used viremia and other variables to model the CFR. Data for 699 EVD patients were analyzed. In the week following symptom onset, mean viremia remained stable, and the CFR increased with viremia, V, from 21% (95% CI 16%-27%) for low viremia (V < 104.4 copies/ml) to 53% (95% CI 44%-61%) for intermediate viremia (104.4 ≤ V < 105.2 copies/ml) and 81% (95% CI 75%-87%) for high viremia (V ≥ 105.2 copies/ml). Compared to adults (15-44 y old [y.o.]), the CFR was larger in young children (0-4 y.o.) (odds ratio [OR]: 2.44; 95% CI 1.02-5.86) and older adults (≥ 45 y.o.) (OR: 2.84; 95% CI 1.81-4.46) but lower in children (5-14 y.o.) (OR: 0.46; 95% CI 0.24-0.86). An order of magnitude increase in mean viremia in cases after July 2014 compared to those before coincided with a 14% increase in the CFR. Our findings come from a large hospital-based study in Conakry and may not be generalizable to settings with different case profiles, such as with individuals who never sought care. CONCLUSIONS Viremia in EVD patients was a strong predictor of death that partly explained variations in CFR in the study population. This study provides baseline CFRs by viremia group, which allow appropriate adjustment when estimating efficacy in treatment studies. In randomized controlled trials, stratifying analysis on viremia groups could reduce sample size requirements by 25%. We hypothesize that monitoring the viremia of hospitalized patients may inform the ability of surveillance systems to detect EVD patients from the different severity strata.