Associated risk factors of severe dengue in Reunion Island: A prospective cohort study

Background Since 2018, a dengue epidemic has been raging annually in Reunion Island, which poses the major problem of its morbidity and mortality. However, there is no consensus in the literature on factors associated with severity of illness. The objective of this study was to identify the factors associated with the occurrence of severe dengue (SD) according to the criteria adopted in 2009 by the World Health Organization (WHO), during the 2019 epidemic. Methodology/Principal findings A total of 163 patients with RT-PCR-confirmed dengue were included in a multicenter prospective cohort study in Reunion Island between January and June 2019. Of these, 37 (23%) were classified as SD, which involves presentation dominated by at least one organ failure, and 126 (77%) classified as non-SD (of which 90 (71%) had warning signs). Confusion, dehydration, and relative hypovolemia were significantly associated with SD in bivariate analysis (p < 0.05). The factors associated with SD in multivariate analysis were a time from first symptom to hospital consultation over 2 days (OR: 2.46, CI: 1.42–4.27), a history of cardiovascular disease (OR: 2.75, 95%CI: 1.57–4.80) and being of Western European origin (OR: 17.60, CI: 4.15–74). Conclusions/Significance This study confirms that SD is a frequent cause of hospitalization during dengue epidemics in Reunion Island. It suggests that cardiovascular disease, Western European origin, and delay in diagnosis and management are risk factors associated with SD fever, and that restoration of blood volume and correction of dehydration must be performed early to be effective. Trial registration NCT01099852; clinicaltrials.gov


Introduction
Dengue is the most common mosquito-borne arbovirosis. In 2017, its incidence was estimated at more than 104 million people, of which 40,000 induced a death [1,2].
In 2019, in view of the re-emergence and rapid expansion of this arbovirosis, the World Health Organization (WHO) ranked dengue among the top ten global health threats [3]. Dengue is caused by four viral serotypes (DENV 1-4) with significant genotypic variability [4] and is characterized by a large clinical polymorphism with numerous signs of varying severity. In 2009, WHO proposed to distinguish between non-severe dengue, dengue with warning signs (WS) and severe dengue (SD), the latter being marked by major plasma leakage, severe bleeding and/or organ failure [5]. The case fatality can reach 20% in severe forms in the absence of early and appropriate medical management [6].
Reunion, a French island in the Indian Ocean with a population of 860,000, has been experiencing seasonal dengue epidemics since 2018. Nearly 70,978 confirmed cases, 2,672 hospitalizations, and 75 deaths have been reported up until May 2022 [7].
This re-emergence on the island has caused saturation of the healthcare system during epidemic peaks, necessitating a better estimation of each individual's risk of developing SD. The performance of prognostic factors for SD varies according to the studies and the definitions of organ failure used when defining SD [8]. The objective of this multicenter prospective cohort study, conducted in the university hospitals of Reunion Island, was to identify the clinical factors associated with the development of SD in our context.

Ethical approval
Ethical clearance was obtained by the French National Agency for the Safety of Medicines and Health Products (ANSM) (n˚IDRCB 2010-A00282-37) and by the committee for the  Table).

Study population
CARBO (Cohorte ARBOviroses) is a French multicentric prospective cohort study dedicated to the understanding of the acute stage of arboviral infections. It was proposed by the university hospital of Martinique and is registered on clinicaltrials.gov (NCT01099852). All patients suspected of dengue fever who were admitted to emergency departments of the University Hospital of Reunion from 1st of January to 30th of June 2019, were eligible. Patients were recruited if they had a positive blood sample for DENV by RT-PCR, and onset of symptoms �7 days (until 21 days for SD). They must have presented at least two symptoms among the following: fever (documented or declared), headache, rash, myalgia, arthralgia, abdominal pain or tenderness, bleedings and/or low platelets blood level below 150 G/l. Children younger than 6 years of age need to present fever (documented or declared by the family) with or without pain on an age-adapted visual analogue scale. The Tropical Fever Core multiplex RT-PCR (Fast Track Diagnostics) was used for on-site diagnosis. It was a conventional two step real-time RT-PCR, which could detect dengue (without differentiation between serotypes), chikungunya and West Nile viruses, Leptospira spp., Rickettsia spp. and Salmonella spp., and Plasmodium spp., which are pathogens that could circulate in the South West Indian Ocean region.
Serology was not routinely performed, but when there was a strong suspicion of dengue fever and a negative RT-PCR result. It was not systematically double checked three weeks later during the epidemic, or in ambulatory care, so the link between the data could not be made with confidence. For this reason, only cases with positive RT-PCR results were included in the study. Dengue reinfection was not systematically documented in 2019 because Reunionese people were considered naïve to dengue at this time, and serum neutralization was rarely performed.

Data collection and follow-up
The primary outcome was the determinants associated with severity by comparing characteristics of SD patients with non-SD patients. SD was identified according to the WHO 2009 classification. However, the definitions of organ failure were those of the various learned societies concerned (S1 Table). Sociodemographic and clinical data were collected by a clinician at each physical exam of the patient via a standardized questionnaire at Day 0 (D0), D3, D5-7, D8-10, D21, D45, D90, D180. Ethnicity was identified by patient self-determination, and divided into six categories: existence of a parent or an ancestor from Sub-Saharan Africa, India, West Europe, South-east Asia; being Native American; or from other origin(s). Biological data were collected only in the acute phase from day 1 to day 10 of symptoms. All biological exams were carried out in the laboratories of the University Hospital of Reunion. All collected data were reported in a case report form and are detailed in Table 1.

Data analyses
Quantitative variables were described with means and standard deviations while categorical variables were described as numbers and percentages.
A bivariate analysis was performed to compare clinical and laboratory features of DF patients according to severity of illness (defined by the WHO in 2009). Gaussian variables A multivariate logistic regression was performed to identify factors associated with SD. Variables with a p-value <0.2 in bivariate analysis were entered into a multivariate model. In the case of collinearity, detected by using the variance inflation factor (VIF), the most relevant variable was retained in the model. A backward stepwise automated procedure was used to select factors associated with p-value <0.05 in the final model. Adjusted odds ratios (OR) and their 95% confidence intervals (95%CI) were derived from the regression coefficients. Data were analyzed using SPSS software (IBM SPSS 23.0, IBM Corp. Armonk, NY, USA). All tests were two-tailed and the significance level was set at 0.05.

Characteristics of the study population
In total, of 172 eligible patients who consented to participate, 163 were included in the study (Fig 1). The average age was 52 years old, and there were more women (57%) than men. Indian origin was the most represented (13%). Cardiovascular comorbidities were the most reported medical background (46%), followed by hypertension (34%), diabetes mellitus (23%), and obesity (16%) ( Table 1). The most common serotype identified in 2019 was DENV-2.

Discussion
The prevalence of SD was 23% in this study. The warning signs had a low discriminatory value as their proportion was similar between the two groups. There was a co-circulation of two

Variables
Adjusted Odds Ratio (95%CI) P-value serotypes, dominated by DENV-2 [9]. This result is similar to that of a study carried out in French Guyana which identified 16% of SD during an epidemic mixing the same two serotypes [10]. Serotype 2 is the most virulent strain according to the literature [11]. The co-circulation of several serotypes in Reunion Island raises the threat of more serious future epidemics linked to the ADE (antibody-dependent enhancement) phenomenon observed during secondary infections [12]. A history of cardiovascular disease was associated with SD according to a recent meta-analysis, which reported that hypertension and diabetes were predictive factors for SD [13]. These comorbidities lead to dysfunction of the capillary endothelium, with increased vascular permeability and the occurrence of hypovolemic shock responsible for SD. Moreover, the co-existence of these comorbidities also favors organ decompensation [14,15]. The population of Reunion Island is particularly exposed to cardiovascular risk factors, which are potentially linked to genetic factors [16][17][18].
Our work suggests that Western European origin is also a risk factor for progression to SD. These results are similar to those of a study conducted in the Cuban population. In that study, the authors suggested that polymorphism in genes associated with the immune response may be responsible for the increased susceptibility of European ethnicity to ADE [19]. Risk factors associated with SD, such as cardiovascular disease and Western European origin, raise concerns that the recent spread of dengue to these new areas may increase the incidence of SD and its lethality [20,21].
Dehydration defined as the need for oral or intravenous rehydration was associated with SD. Rehydration is the main treatment for dengue with warning signs to prevent progression to a severe form. It may cause confusion or lipothymia, which are two warning signs present in one third of SD cases in our study. These elements suggest that dehydration plays a predominant role in the critical phase, just as it can cause complications in other arboviroses [22][23][24]. A late consultation, which delays intravenous rehydration, could also explain the increased risk of SD.
Indeed, delay in diagnosis and management was also associated with SD. Severe bleedings were the only severity component to be significantly associated with a consultation time greater than 2 days after onset. Although the date of illness onset is declarative, this is consistent with the literature which reports that delay in diagnosis is associated with severity including severe bleedings, complications, and mortality of dengue [6,[25][26][27].
No patient with chronic kidney disease presented with SD. The difficulty in applying the KDIGO definition could be the cause. Indeed, urine quantification was not routinely measured in conventional hospitalization, baseline creatinine was not always available, and isolated increase in blood creatinine was insufficient because of already high levels in patients with chronic kidney disease. We used Stage 3 of the KDIGO classification to define acute kidney failure. By choosing a severe and restrictive definition of kidney failure to target severe cases in an inpatient cohort, cases of moderate acute kidney failure were not selected. With a less restrictive definition, eighteen patients had acute kidney failure according to KDIGO Stage 2, including 3 patients with chronic kidney disease.
The strengths of this study were (i) the early inclusion and prospective follow-up of patients, (ii) systematic clinical examination by a clinical investigator at admission and up to day 10, and (iii) use of the definitions of each learned society to estimate organ failure. There were, however, limitations within this study: (i) it was a cohort not derived from a consecutive series, or randomly sampled among patients, (ii) the cohort represented 19% of patients who consulted the hospital with PCR-confirmed dengue in 2019 in Reunion Island according to epidemiological data [7] (these results were related to a time-consuming inclusion of patients in an epidemic context and a large number of patient refusals), (iii) the use of a uniform definition of SD for all age groups could lack reproducibility in the pediatric population, especially for the shock criterion and the biological variables of liver and kidney failure, and (iv) the viral serotype of dengue patients could not be identified in all patients during the epidemic. A serum library was established, which will allow for future studies.
This first prospective study conducted during a dengue epidemic in Reunion Island allowed us to describe the risk factors specific to the Reunionese population and the evolution of the disease among patients who attended the hospital. Our study highlights the poor performance of warning signs in predicting SD, as well as the variability of the criteria for admission to intensive care depending on the clinician. Besides a history of cardiovascular disease and Western European origin that were associated with SD in our setting, the time to diagnosis and management of dengue is crucial, and rehydration appears to play a central role in preventing progression to SD.
Our results will be used in other large-scale studies during future dengue epidemics in Reunion Island, which will allow the identification of predictive factors in the evolution towards a more severe form of the disease, as well as the drafting of referral protocols to specialized services adapted to the profile of the patient.