Vitamin A Supplementation Was Associated with Reduced Mortality in Patients with Ebola Virus Disease during the West African Outbreak

ABSTRACT

Background

Micronutrient supplementation is recommended in Ebola virus disease (EVD); however, there are limited data on therapeutic impacts of specific micronutrients.

Objective

To evaluate the association between vitamin A supplementation and mortality in EVD.

Methods

This retrospective cohort included patients with EVD admitted to 5 International Medical Corps Ebola Treatment Units (ETUs) in 2 countries during 2014–2015. Protocolized treatments with micronutrients were used at all ETUs: however, because of resource constraints, only a subset of patients received vitamin A. Standardized data on demographics, clinical characteristics, malaria status, and Ebola viral loads (cycle threshold values) were collected. The outcome of interest was mortality between cases treated with 200,000 IU of vitamin A on care days 1 and/or 2, and those not. Propensity scores based on the first 48 h of care were derived using covariates of age, ETU duration, malaria status, cycle threshold values, and clinical symptoms. Patients were matched 1:1 using nearest neighbors with replacement. Mortality between cases treated and not treated with vitamin A was compared using generalized estimating equations to calculate RR with associated 95% CI.

Results

There were 424 cases analyzed, of which 330 (77.8%) were treated with vitamin A. The mean age was 30.5 y and 40.3% were men. The most common symptoms were diarrhea (85.6%), anorexia (80.7%), and abdominal pain (76.9%). Mortality proportions among cases treated and not treated with vitamin A were 55.0% and 71.9%, respectively. In the propensity-matched analysis, mortality was significantly lower among cases receiving vitamin A (RR = 0.77, 95% CI: 0.59, 0.99; P = 0.041). In a subgroup analysis of patients treated with multivitamins already containing vitamin A, additional vitamin A supplementation did not impact mortality.

Conclusion

Early vitamin A supplementation was associated with reduced mortality in patients with EVD, and should be further studied and considered for use in future epidemics.

Introduction

The Ebola virus disease (EVD) epidemic in West Africa which occurred during 2014–2016 resulted in >11,000 deaths and >28,000 infections, with Liberia and Sierra Leone having the largest burden of disease (1). During the epidemic, patients cared for in high-income countries had lower relative mortality, being approximately half that of the lowest mortality from any of the Ebola Treatment Units (ETUs) in Africa (2–7). In high-income countries, advanced critical care resources and supportive care strategies were available for treatment of patients with EVD that were not consistently accessible during the West African response (8). Although supportive care strategies in limited resource epidemic settings have potential to reduce mortality, the roles of specific components of supportive care have not been sufficiently studied.

Micronutrient supplementation is a primary part of supportive care of EVD (2, 3, 5, 9–15), and treatment guidelines for viral hemorrhagic fevers recommend supplementation when micronutrient deficiencies are present (16, 17). Micronutrient supplementation includes vitamin A, which has well-established immunological impacts in infectious disease states. Vitamin A is essential for both innate and adaptive immune functions and is integral in the maintenance of epithelial integrity, acute phase response, and function of several immune cell types (18–20). In vitro research has demonstrated that Ebola virus-specific proteins inhibit retinoic-acid-mediated immune responses, which may contribute to the virulence in EVD infection (21). As such, it is biologically plausible that vitamin A supplementation could yield clinical benefits among patients with EVD; however, to date, no data exist on the impact of vitamin A from outbreak populations in low-resource settings. Furthermore, given prior data demonstrating adverse effects with high-dose vitamin A supplementation in patients with acute infections, evaluation of vitamin A therapies is needed to inform care (22, 23).

During the West African EVD epidemic, International Medical Corps (IMC), working with local and international personnel, established and operated 5 ETUs in Liberia and Sierra Leone that cared for over 470 patients with confirmed EVD. These ETUs gathered data on clinical, epidemiologic, and laboratory variables which were used to develop a robust, multinational database able to research associations with specific forms of care provision on patient outcomes (24). This study evaluated the impact of early vitamin A supplementation on mortality among patients with EVD admitted to the 5 IMC ETUs in 2 countries in West Africa.

Methods

Study design, setting, and population

This retrospective multisite study used data collected at 5 ETUs operated by IMC from 15 September 2014 through 31 December 2015 in Liberia and Sierra Leone (24). The Sierra Leone Ethics and Scientific Review Committee and the University of Liberia and Rhode Island Hospital Institutional Review Boards all provided ethical approval for this study. All study activities complied with the Helsinki Declaration as revised in 1983.

All patients admitted to the 5 ETUs with a final diagnosis of EVD were eligible for inclusion. Patients who died before triage, were confirmed to be negative for EVD by real-time RT-PCR testing, or those who had missing data on vitamin A treatment, were excluded.

Clinical procedures

All patients were cared for by practitioners trained in ETU care using standardized guidelines, developed by IMC in consultation with local and international health authorities (Supplement 1) (16). The guidelines included supplementation with vitamin A; however, because of availability of resources, actual treatment provision varied. The vitamin A supplementation was administered orally using a capsule formulation containing retinyl palmitate 100,000 IU (Softgel Health Care India Ltd.). The IMC guidelines also recommended supplementation with ascorbic acid dosed 500 mg 3 times daily with ascorbic acid 250 mg tablets (Nanjing Baigingyu Pharmaceuticals Ltd.) and multivitamin once daily. Two formulations of multivitamins were used. The first sourced included vitamin A (2500 IU), thiamin (1 mg), riboflavin (0.5 mg), niacin (7.5 mg), ascorbic acid (15 mg), and cholecalciferol (300 IU) (Medopharm Private Ltd, Chennai, India), while the second included vitamin A (800 IU), thiamin (0.5 mg), riboflavin (0.5 mg), niacin (7.5 mg), and cholecalciferol (200 IU) (CSPC Ouyi Pharmaceutical Co). Although ETU clinical staff attempted to adhere to guidelines to the greatest extent possible, actual treatments across ETUs varied based on supply availability.

To be admitted to the EVD-confirmed ward, a positive laboratory confirmation by RT-PCR was required. Ebola virus RT-PCR cycle threshold values (which are inversely proportional to viral load) were obtained from the United States Naval Medical Research Center Mobile Laboratory for ETUs in Liberia and Public Health England and the Nigerian/European Mobile Laboratory for ETUs in Sierra Leone using previously published procedures (25). When performed, malaria testing was done with use of BinaxNow rapid diagnostic test (RDT), which identifies Plasmodium species of Falciparum, Malariae, Vivax, and Ovale.

Data collection

As previously described, trained providers collected data on demographics and baseline clinical signs/symptoms on standardized forms at the time of triage (24). Clinical data were recorded 1–6 times daily (median of 3 times/d) on individual patients, based on human resource availability using medical charts. Final disposition and diagnoses were recorded on standardized discharge forms, as described previously (24, 25). All data were digitized into a relational electronic database and data fidelity was assessed using Lot Quality Assurance Sampling, demonstrating 99% consistency with the source records (24, 26).

Statistical analysis

Descriptive analyses using frequencies with percentages, medians with IQR, or means with SDs were performed as appropriate. To allow for the performance of comparative analyses, clinical characteristics reported at any point during the day of analysis were considered present for the entire day. Statistical analyses were undertaken using R version 3.3.3 (27) and statistical significance was set at a value of P < 0.05.

The primary outcome was observed mortality during ETU care. Because of the high early mortality risks documented in the West African EVD outbreak and the specific study ETUs (25), and the resulting potential survivor bias, the primary predictor variable used was early vitamin A treatment. This was defined as treatment with 200,000 IU of vitamin A on care days 1 and/or 2 for a cumulative dose range of 200,000–400,000 IU per patient. Treatment exposure was coded as dichotomous, as received if patients were treated with vitamin A on either or both days 1 and 2 of ETU care or not received if they did not meet that treatment definition. Time to vitamin A initiation was summarized across the study cohort. Univariate analyses compared demographic and clinical characteristics between groups treated and not treated with vitamin A and also between patients suffering mortality and those who were discharged alive using Pearson chi-square or Fisher's exact tests for categorical variables and by Mann-Whitney or t tests for continuous variables, as appropriate.

Propensity score development and modeling

To control for confounding, a propensity score model (PSM) was used to match and compare patients treated and not treated with vitamin A during the first 48 h of ETU care (28). Variables included in the models were chosen based on their univariate relation to the treatment exposure or mortality in either the study cohort or based on prior literature (28). The included variables were age, bleeding, dysphagia, dyspnea, diarrhea, duration of ETU operation, triage cycle threshold value, and malaria RDT results. All variables were modeled as categorical except for age, where a cubic spline was used to control for the known quadratic relation between age and mortality in EVD (25). In the PSM, a common support interval approach to prevent overextrapolation was used in which all observations with a probability of receiving treatment less than the minimum in the treated cohort and greater than the maximum in the nontreated cohort were filtered out (29). After filtering for the common support interval, 30 patients were removed.

Vitamin A treated and untreated patients were then matched based on the nearest propensity score (PS) with replacement, and exactly matched on cycle threshold value and malaria testing results because of the high correlation of those factors with the outcome of interest (5, 25, 30, 31). Covariate balance was assessed before and after matching based on the standardized bias, to ensure balance was achieved between treatment and control cohorts. Following 1:1 matching, with 77 unique controls, the probability of observed mortality between cases treated and not treated with vitamin A was assessed. Mortality outcomes between cases treated and not treated with vitamin A were compared using propensity-matched generalized estimating equations to calculate RR with associated 95% CI.

Missing data and sensitivity analyses

Demographic and clinical data were available for all included patients, but there were some data missing for triage cycle threshold values and malaria testing results. Because patients who died were more likely to lack laboratory data, multiple imputations were not used. Instead, cycle threshold values were categorized as >22 (low viral load), ≤22 (high viral load), and missing, to prevent the introduction of bias (32). Similarly, results for malaria RDTs were coded as positive, negative, or missing.

Two sensitivity analyses were performed using propensity-matched generalized estimating equations as described in the primary analysis. Both analyses used the derived PSM and compared observed mortality outcomes to evaluate the addition of vitamin A to other micronutrient treatments. The first assessed differences between patients treated with ascorbic acid and no vitamin A compared with those treated with vitamin A and ascorbic acid. The second compared those treated with multivitamins and no vitamin A and those treated with multivitamins and vitamin A. In all sensitivity analyses, covariate balance was assessed before and after matching based on the standardized bias.

Results

Characteristics of the study population

There were 478 patients with EVD treated at the ETUs during the study period, 424 of whom met inclusion criteria for analysis (Figure 1). The mean age was 30.5 (SD 18.7) y and 40.3% were men. Median care duration across the 5 ETUs was 8 (IQR 5, 13) d. The proportion of patients who developed clinical signs and symptoms during ETU care is reported in Table 1. The most common symptoms were diarrhea (85.6%), anorexia (80.7%), and abdominal pain (76.9%). Triage cycle threshold values were obtained for 281 patients, of whom 159 (37.5%) had a high viral load (cycle threshold value ≤22) and 122 (28.8%) had a low viral load (cycle threshold value >22).

FIGURE 1.

Study population. EVD, Ebola virus disease.

TABLE 1.

Characteristics of the overall cohort of patients with Ebola virus disease

	Summary value1
Age, y	30.5 ± 18.7
Sex
Women	253 (59.7)
Men	171 (40.3)
Cycle threshold values2
High viral load	159 (37.5)
Low viral load	122 (28.8)
Missing results	143 (33.7)
Malaria results
Positive	40 (9.5)
Negative	203 (47.9)
Not tested	181 (42.7)
Clinical characteristics
Anorexia	342 (80.7)
Any bleeding	198 (46.9)
Coma	42 (9.9)
Confusion	57 (13.4)
Diarrhea	363 (85.6)
Dysphagia	249 (58.7)
Dyspnea	205 (48.3)
Fever	325 (76.7)
Abdominal pain	326 (76.9)
Vomiting	325 (76.7)

¹Values are presented as frequencies (percentages) or means ± SDs.

²Cycle threshold values categorized as: >22 (low viral load), ≤22 (high viral load).

Vitamin A treatment

During ETU care, 330 (77.8%) patients were treated with vitamin A within 48 h of admission. Frequency distributions for the time to initiation for cases receiving vitamin A are shown in Figure 2. In the treatment group, 186 (56.4%) patients received vitamin A on days 1 and 2, while 43 (13.0%) patients received vitamin A on day 1 only and 101 (30.6%) on day 2 only. Patients receiving vitamin A during the first 48 h were significantly less likely to have symptoms/signs of confusion and dyspnea during treatment, and less likely to have a positive RDT for malaria (Table 2). Among the 330 patients receiving vitamin A within 48 h, 325 (98.5%) also received ascorbic acid, while 230 (69.7%) patients also received a multivitamin.

FIGURE 2.

Initiation times of vitamin A treatment for patients.

TABLE 2.

Patient characteristics stratified by treatment status with vitamin A¹

	Vitamin A (−) (n = 94)	Vitamin A (+) (n = 330)	P
Age, y	27.9 ± 19	31.3 ± 18.6	0.131
Sex			0.338
Women	52 (55.3)	201 (60.9)
Men	42 (44.7)	129 (39.1)
Cycle threshold values2			0.850
High viral load	34 (36.2)	125 (37.9)
Low viral load	26 (27.7)	96 (29.1)
Missing results	34 (36.2)	109 (33.0)
Malaria results			0.002
Positive	8 (8.5)	32 (9.7)
Negative	55 (58.5)	126 (38.2)
Not tested	31 (33.0)	172 (52.1)
Clinical characteristics
Anorexia	62 (66)	221 (67)	0.856
Any bleeding	34 (37.2)	87 (26.4)	0.053
Coma	1 (2.1)	3 (1.2)	0.571
Confusion	12 (12.8)	15 (4.8)	0.032
Diarrhea	67 (71.3)	231 (70)	0.811
Dysphagia	45 (48.9)	125 (37.9)	0.060
Dyspnea	39 (41.5)	96 (29.1)	0.031
Fever	75 (79.8)	245 (74.5)	0.277
Abdominal pain	59 (63.8)	204 (62.1)	0.763
Vomiting	59 (63.8)	195 (59.1)	0.520

¹Values are presented as frequencies (percentages) or means ± SDs.

²Cycle threshold values categorized as: >22 (low viral load), ≤22 (high viral load).

Mortality outcomes

In the population studied, 244 (57.5%) patients infected with EVD died during care. As previously reported, there were no significant differences in mortality outcomes among the 5 ETU sites (25). Viral load on admission was strongly correlated with mortality: among cases who died, 44.7% had a cycle threshold value ≤22 while only 16.8% had a cycle threshold value >22 (P < 0.001). Nonsurviving patients were significantly more likely to have developed any diarrhea, dysphagia, and dyspnea. Nonsurvivors were also more likely to have positive RDT for malaria (Table 3). Among patients receiving vitamin A on care days 1 and 2 post admission, mortality occurred in 53.9% of cases, compared with 70.2% of cases not treated with vitamin A within 48 h. The relative observed mortality outcomes were maintained when the populations were stratified by country (Figure 3).

TABLE 3.

Patient characteristics stratified by mortality outcomes¹

	Survived (n = 180)	Died (n = 244)	P
Age, y	28.7 ± 15.3	31.8 ± 20.8	0.080
Sex			0.906
Women	108 (60)	145 (59.4)
Men	72 (40)	99 (40.6)
Cycle threshold values2			<0.001
High viral load	50 (27.8)	109 (44.7)
Low viral load	81 (45.0)	41 (16.8)
Missing results	49 (27.2)	94 (38.5)
Malaria results			0.005
Positive	10 (5.5)	30 (12.3)
Negative	101 (56.1)	102 (41.8)
Not tested	69 (38.3)	112 (45.9)
Clinical characteristics
Anorexia	86 (47.8)	119 (48.8)	0.840
Any bleeding	32 (17.8)	61 (25.4)	0.057
Coma	0 (0)	1 (0.8)	0.158
Confusion	5 (3.3)	13 (5.7)	0.231
Diarrhea	93 (51.7)	152 (62.3)	0.029
Dysphagia	43 (24.4)	90 (36.9)	0.006
Dyspnea	39 (21.7)	80 (32.8)	0.010
Fever	133 (73.9)	183 (75.4)	0.723
Abdominal pain	91 (50.6)	134 (55.3)	0.332
Vomiting	86 (48.3)	115 (47.5)	0.872

¹Values are presented as frequencies (percentages) or means ± SDs.

²Cycle threshold values categorized as: >22 (low viral load), ≤22 (high viral load).

FIGURE 3.

Observed mortality proportions for all patients. Sample size for each stratified exposure group annotated on bars.

Propensity matched analysis

Propensity matching achieved covariate balance among all predictors of interest associated with vitamin A administration, mortality, or both (Figure 4), allowing for appropriate intergroup comparisons. In the adjusted propensity-matched analysis, mortality was 55.0% among patients receiving vitamin A and 71.9% among patients not receiving vitamin A. The relative risk reduction of mortality with vitamin A treatment within 48 h was 0.77 (95% CI: 0.59, 0.99, P = 0.041).

FIGURE 4.

Standardized bias, before and after 1:1 matching of patients not treated with vitamin A within 48 h to patients treated with vitamin A within 48 h. Dashed line represents the commonly used threshold of balance treatment cohorts. ETU, Ebola Treatment Unit.

Sensitivity analyses

In adjusted propensity-matched analysis comparing patients receiving both vitamins A and ascorbic acid to patients receiving ascorbic acid and no early vitamin A supplementation, there was a nonsignificant reduced relative risk of mortality with the addition of vitamin A, similar to that observed in the primary analysis (RR = 0.80, 95% CI: 0.59, 1.07; P = 0.13). In the multivitamin sensitivity analysis, no significant difference in mortality risk was observed in patients treated with early multivitamins and vitamin A compared with those receiving multivitamins but no early vitamin A treatments (RR = 1.07, 95% CI: 0.62, 1.87; P = 0.80). In both the ascorbic acid and multivitamin PS, adequate balance across the covariates of interest was attained based on the predetermined standardized bias thresholds except for the variable of anorexia (data not shown).

Discussion

The current report is the only available patient-level data assessing vitamin A supplementation in a high-risk EVD outbreak population from a resource-limited African setting. The results demonstrate that oral vitamin A was associated with reduced mortality among patients infected with EVD receiving supplementation during the first 2 d of ETU care. This significant beneficial impact was maintained in adjusted propensity-matched models, controlling for confounding variables. Based on these findings, along with the minimal resource allocations required for the provision of vitamin A, this therapy should be further studied and considered for use in future outbreaks.

Micronutrient supplementation, inclusive of provision of vitamin A, is often used in supportive treatment approaches for EVD (2, 3, 5, 9–15). However, to date there is a paucity of evidence pertaining to the impact of micronutrients, and no prior data on the role of vitamin A supplementation as it relates to patient-centered outcomes. In the presented propensity-matched analysis, the observed mortality of patients treated early with 200,000–400,000 IU of vitamin A was significantly lower than patients who were not treated. Although this is the first ever study of the impacts of vitamin A in a population of EVD-infected patients, the importance of vitamin A in immunologic responses to infectious diseases is well established. Vitamin A plays a role in innate and adaptive immune functions, and supports epithelial integrity, and function of immune cells, including macrophages, natural killer cells, and lymphocytes (18–20, 33, 34). In addition, vitamin A supplementation has been found to reduce upper respiratory tract infections and decrease infant and child mortality (35, 36). Low serum retinol-binding protein, a predictor of hypovitaminosis A (37), has been inversely correlated with progression of viral infections (38). In relation to EVD, vitamin A is known to support gastrointestinal epithelial function and mucosal recovery after diarrheal illnesses (33, 39), which could be an important protective mechanism in EVD, where patients can suffer from copious gastrointestinal losses of over 5 L/d (40). Furthermore, vitamin A derivatives have shown activity against the Ebola virus in vitro (41, 42), and research has demonstrated that an Ebola virus–specific protein (viral protein 35) inhibits retinoic-acid type 1 interferon immune response, which may contribute to the virulence of EVD infection (21). Given the extensive roles of vitamin A in immune function and prior literature on infectious disease mechanisms and outcomes, it is biologically plausible that vitamin A supplementation could yield positive impacts in patients with EVD and further biochemical and prospective clinical study is warranted.

In sensitivity analyses assessing the addition of vitamin A supplementation to patients with EVD treated with oral multivitamins, no significant difference was identified in mortality risk. As the multivitamins used at the ETU contained vitamin A, with doses ranging from 800 to 2500 IU, the lack of benefit may stem from sufficient supplementation being delivered by the multivitamins themselves, and suggests that dose-response studies would be informative to the future use of appropriate vitamin A supplementation. In the subgroup of ascorbic acid–treated patients there was also no statistically significant difference in mortality risk found with the addition of vitamin A supplementation; however, the association and magnitude of effect was only minimally attenuated and consistent with that of the primary outcome. In that exploratory analysis, however, the relative risk of death was reduced for patients receiving early vitamin A treatment compared with those that did not, and it is possible that the sample size resulted in a type II error. Because of the inability to completely achieve adequate balance for all covariates and the likely limited power, the sensitivity analyses are solely hypothesis-generating, but do highlight the need for further research pertaining to vitamin A and broader micronutrient supplementation in EVD care.

Guidelines from the WHO on viral hemorrhagic fever state that patients do not need micronutrient supplementation if adequate fortified nutrition is administered (16). However, vitamin A is recommended for children aged <5 y who have not received supplementation in the past 6 mo (16). There is a high baseline prevalence of malnutrition and micronutrient deficiencies in sub-Saharan populations (43, 44), particularly in the West African populations that were most affected during the 2014–2016 outbreak, where the prevalence of vitamin A deficiency is over 15% in most countries (33, 45–48). A population-wide study from Sierra Leone found a substantially high prevalence of vitamin A deficiency of 17.4% (43), and a report from Liberia identified 2.2–13.2% of the population to be vitamin A-deficient (49). Research studies on vitamin supplementation have demonstrated the greatest benefits in populations with low baseline nutritional status (49), supporting the use of empiric treatment with vitamin A during EVD outbreaks involving populations likely to have high baseline nutritional deficiencies as in sub-Saharan Africa. However further research should be carried out pertaining to optimizing dosing strategies and further adjuvant treatments with synergistic impacts.

The study design means that the results must be interpreted with certain limitations. Vitamin A supplementation was not provided to patients in a randomized manner and, as such, there is potential for confounding within the results. However, it should be noted that nontreatment with vitamin A supplementation was based solely on supply limitations and not related to individual patient characteristics, as provision was a standard part of the clinical protocol at all 5 ETUs. Furthermore, propensity-matched analytical models were used to reduce the potential for confounding by indication, where more critically ill patients are more or less likely to receive treatment. Although rigorous data acquisition methods were used, there was some missing clinical information for laboratory data. To minimize bias from the missing data, all models were run with cases matched exactly on the status of those variables. Furthermore, although the PS used were able to produce well-matched comparison groups for study, the models were not able to take into account all supportive treatments provided across the ETUs, which precluded full assessment for synergistic treatment impacts. In addition, the present data do not allow for investigation of varying doses of vitamin A supplementation on patient outcomes and subsequent research on dose-response effects and adverse events would have utility. This will be particularly important given the previously observed adverse outcomes documented with high-dose supplementation in other acute infectious states (22, 23). Finally, all studied patients with EVD were cared for at IMC-operated facilities, which may reduce the generalizability of the results. However, as the clinical guidelines used at all sites were concordant with international guidelines in widespread use, and the resource constraints experienced were common at most ETUs during the 2014–2016 epidemic, these data are likely generalizable to future EVD outbreaks in similar low-resource settings.

In conclusion, stemming from the high mortality in EVD and the increasing frequency of outbreaks (50), it is apparent that EVD is a global health threat and that there are crucial needs for pragmatic and impactful care strategies. The present findings, which are derived from a large, multisite, epidemic population, are the first ever reported on the impact of vitamin A supplementation in a high-risk population in Africa and demonstrate that early treatment resulted in lower overall mortality in adjusted analyses. Given these findings, supportive therapy with vitamin A supplementation should be further studied and considered for use in EVD outbreaks occurring in populations with high burdens of nutritional deficiencies.

Notes

Funding for this study was provided by the NIH National Institute of Allergy and Infectious Diseases (R03AI132801).

Author disclosures: ARA, SMP, DY, SG, JLP, LA, DKC, SBK, MM, FS, SB, LL, TL, and ACL, no conflicts of interest.

The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the views of International Medical Corps or any governmental bodies or academic organizations.

A supplemental data file is available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: ETU, Ebola Treatment Unit; EVD, Ebola virus disease; IMC, International Medical Corps; PS, propensity score; PSM, propensity score model; RDT, rapid diagnostic test.