Background
Dengue places huge burdens on the healthcare services and economies of dengue-endemic regions, where approximately 50% of the world’s population live [
1]. Complex interactions between host and viral factors make the clinical presentation highly variable. Although the majority of dengue virus (DENV) infections are asymptomatic or self-limiting, a small proportion of symptomatic patients develop features of severe disease, which include severe plasma leakage with hypovolaemic shock, significant bleeding and/or major organ impairment.
At present, guidelines for dengue treatment are limited to supportive care and careful fluid management [
2]; clinical trials investigating antivirals and immune modulators have been conducted, but none have demonstrated efficacy by virological or clinical endpoints [
3‐
6]. The clinical endpoints used in dengue research are strikingly variable, hampering reproducibility and comparability of findings. Seeking to correct this, an expert working group have recommended endpoints for use in early phase trials [
7,
8], including definitions for moderate and severe plasma leakage, bleeding, and organ involvement. Pending validation, the endpoints proposed by the above studies may be applicable to interventional studies involving treatment of patients before the onset of severe dengue (e.g., antiviral candidates in early infection, host-directed immune modulation in high-risk groups). However, patients presenting with established dengue shock already fulfil these categorical endpoints at baseline. Thus, more sensitive and dynamic markers of clinical progress with capacity to reflect both improvement and worsening over time are needed for patients with dengue shock. Further, Dengue is a multisystem disease and splitting endpoints into categories including vascular leakage, severe bleeding and organ impairment may fail to capture the whole clinical picture. As such, we believe that there is a role for using composite endpoints to assess dengue severity as a single entity, rather than scoring separate syndromes.
One possibility is the extension and/or modification of the Sequential Organ Failure Assessment (SOFA) score for dengue. The SOFA score was originally developed to quantify organ impairment in sepsis and takes into account functional impairment of six end-organ systems (respiratory, coagulation, cardiovascular, hepatic, renal and neurological) [
9]. The simpler quick SOFA (qSOFA) is a good screening tool for bacterial sepsis, and it has been found to perform well as predictor for mortality in non-bacterial causes of sepsis in low middle income countries, including dengue [
10]. However, qSOFA was not intended for use beyond triage, and is insufficiently sensitive to monitor clinical progress in patients with established sepsis. Recently, there has been growing consensus for the use of the full SOFA score and its derivatives as clinical endpoints in randomized controlled trials based in the intensive care unit (ICU) [
9,
11]. However, modification of the full SOFA score for dengue might be required; dengue has some unique pathophysiological phenomena including narrowed pulse pressure, which should be considered for inclusion in the cardiovascular component. Narrowing of the pulse pressure is one of the earliest manifestations of dengue shock and occurs prior to the development of hypotension [
2]. The pulse pressure is also used as a resuscitation target for fluid management and a prognostic factor in dengue shock [
12,
13]. In addition, arterial puncture for PaO
2 measurement is relatively contraindicated in thrombocytopaenia, and frequently unavailable in many dengue endemic settings.
We therefore set out to develop a modified SOFA score for dengue, determine how it performed to predict clinical outcomes in adults with dengue shock, to understand what relative contribution the modified and original SOFA parameters conferred on the prognostic ability of the score, and whether a change in SOFA score (delta SOFA) over the first 48 h of treatment might be a feasible composite endpoint for therapeutic trials in severe dengue.
Discussion
Here we have presented a modified SOFA score for dengue and shown the utility of both baseline mSOFA and delta mSOFA in predicting clinical outcomes in a dataset of 124 adults with dengue shock admitted to an ED/ICU in Ho Chi Minh City, Vietnam.
In this patient group, although requirement for organ support and mortality were rare events, mSOFA for dengue performed well to discriminate groups by clinical outcomes, including duration of ICU and hospital admission, requirement for organ support and death. Our findings support other studies showing that a one-time SOFA score at either admission, 24 or 48 h can predict mortality in adults with severe dengue [
18‐
24]. Aside from a single study in Brazil, all of these studies were conducted in India, and like the majority of derivation and validation cohorts for the original SOFA score, they all included patients with a higher median age, more comorbidities, higher frequency of organ failure and higher mortality than our own population. Studies from other dengue patient populations, including an elderly cohort in Taiwan, and patients with dengue induced haemophagocytic lymphohistiocytosis in Malaysia, also found that SOFA score is an effective discriminant between survivors and non-survivors [
25,
26]. Of note, co-infection was common among these cohorts: 23.4% of patients in the Taiwanese cohort had bacterial and/or fungal bloodstream co-infection [
27], and in a cohort of adults with severe dengue admitted to ICU in Brazil, 45% patients received antibiotics, and 22% were treated for septic shock [
24]. Co-infection may have had a significant but unmeasurable impact on SOFA score parameters, and might have contributed to the higher mortality reported in these cohorts, compared to our own population.
Dengue management in Vietnam is standardised using guidelines from the Vietnamese Ministry of Health [
28]. There may be some variations allowing for flexibility in clinical practice, however initial resuscitation should not vary widely between hospitals. We found the mSOFA and delta mSOFA scores performed similarly in the groups of patients transferred from other hopsital compared to those presenting directly to HTD, and between patients with and without comorbities. This suggests that differences in baseline characteristics may not impact the performance of the scoring system, further indicating its potential for pragmatic application.
Our study adds the following: first application of a modified delta SOFA with the addition of pulse pressure to reflect the unique pathophysiology of dengue, and the first time this score has been reported from Vietnam, where the population is significantly younger than the patient populations in the published literature. We have separately reported the individual component scores which make up the SOFA total, and would recommend this be applied in future studies, in order to improve the understanding and external validity of this composite endpoint [
9] and potentially improve future iterations of the scoring system.
We substituted PaO
2 for SpO
2 in our modified SOFA score to allow non-invasive assessment of oxygenation. We were unable to evaluate the performance of SpO
2 relative to PaO
2, but there is growing consensus that this modification is acceptable where arterial blood gas sampling is either unavailable or contraindicated. Haniffa et al. found that requirements for laboratory parameters as prognostic variables hampered the application of critical care scoring systems in low-to-middle-income countries (LMICs) and suggested that scores using readily available clinical data are preferable [
29]. Pandharipande et al. investigated substituting PaO
2/FiO
2 ratio for SpO
2/FiO
2 ratio, and found that the derived scores predicted outcomes with a similar performance to the original respiratory component [
30].
The main modification was the addition of the pulse pressure to the cardiovascular component of the orginial SOFA score. This aimed to capture the unique pathophysiology associated with dengue, whereby a slow vascular leak allows several homeostatic mechanisms to be upregulated, increasing the systemic vascular resistance in response to the gradually worsening hypovolemia [
31,
32]. Narrowing of the pulse pressure, characterised by increased diastolic blood pressure with maintained systolic blood pressure, reflects cardiovascular compensation and is not seen in other conditions like septic shock where hypovolaemic shock develops rapidly [
2,
33]. Mean blood pressure can overestimate the true perfusion pressure in some diseases with marked increases in systemic venous pressure like dengue [
34]. In the WHO management guidelines, pulse pressure is used as a diagnostic and resuscitation target instead of MAP in dengue shock [
12,
13]. Hence, our mSOFA score for dengue included narrowed pulse pressure in the cardiovascular score, in addition to mean arterial pressure and vasopressor requirement. The inclusion of pulse pressure in the mSOFA changed the cardiovascular score in 15 patients and did not significantly alter the overall performance of the score. As pulse pressure is an important diagnostic criteria for cardiovascular compromise and therapeutic target, particularly for paediatric dengue shock syndrome, we suggest this now needs further evaluation in cohorts of both adults and children with severe dengue. As pulse pressure is simple and routinely measured in all DSS cases ongoing prospective evaluation will not require extra data collection, add cost, or inconvenience the patients. However, our results show that although pulse pressure was associated with some clinical outcomes, mSOFA was a superior predictor of clinical endpoints in this population with established dengue shock.
Despite the accumulating evidence that the SOFA score, modified or otherwise, is useful in dengue prognostication, there is uncertainty about whether all components of the original score contribute useful information. In particular, the benefit of the coagulation component is unclear. In dengue, thrombocytopaenia is multifactorial in aetiology: likely a combination of bone marrow suppression and peripheral platelet consumption by both immune mechanisms and disseminated intravascular coagulation. Thus, thrombocytopaenia may reflect typical disease pathophysiology, rather than true end organ failure. Indeed, sub-analyses investigating the contribution of individual organ parameters to the prognostic capability of the SOFA score in dengue have indicated that respiratory, cardiovascular and neurological impairment have the greatest association with prognosis, with the coagulation component inflating the SOFA score, but adding little or no benefit in mortality prediction [
18,
22,
23,
25,
35]. Although this effect will not limit comparison between patients with dengue, it may misrepresent the severity of organ failure if directly compared with other diseases.
In this study, we evaluated for the first time whether using the dynamic delta modified SOFA score might be a feasible endpoint for future therapeutic trials in dengue. We found that delta mSOFA between baseline and 48 h was associated with requirement for organ support, duration of treatment and mortality. This supports a systematic review showing that delta SOFA, rather than other SOFA derivatives, was most significantly associated with mortality by meta-regression (slope 0.7, 95% CI 0.26:1.14, p = 0.004, I2 = 0%), explaining 32% of overall mortality. By contrast, fixed day SOFA was not significantly associated with mortality [
11]. Delta SOFA has demonstrated applicability as an endpoint for both pharmacological [
36] and non-pharmacological interventions [
37] in critical care randomized controlled trials. We plan to use the modified delta SOFA for dengue as an endpoint for the first time in an upcoming clinical trial evaluating host directed therapeutics in moderate-severe dengue [
38].
While this is the first effort to evaluate a modified SOFA score for dengue in adults with dengue shock, our study is not without limitations. First, our data only contains patients with a primary diagnosis of dengue shock; rather than severe dengue which encompasses organ impairment and bleeding, albeit several had co-existent liver or renal impairment. This is both a strength and limitation; allowing us to investigate the utility of the modified SOFA score for dengue in a population with a distinct clinical phenotype, while also limiting generalizability to patients with other types of severe dengue. The lack of bilirubin data at 48 h post enrolment hinders the ability to evaluate the liver derangement using the delta mSOFA score. A larger sample size is needed in order to evaluate uncommon complications in dengue, for example dengue primarily affecting the central nervous system. Going forward, we plan to apply this mSOFA score for dengue prospectively to adult patients with all phenotypes of severe dengue in the ICU using an established electronic data registry.
Secondly, we do not include children in this study. In hyper-endemic regions, children and adolescents represent the majority of severe dengue cases; we cannot simply extrapolate our findings to children because the normal range of clinical observations, presentation and outcomes of severe dengue differ between children and adults. Children present more frequently with pure dengue shock, and adults more frequently with mixed organ impairment [
39]. We plan to continue this work by assessing this mSOFA score in children with dengue shock and tailoring it to a specific pediatric score as necessary.
Ultimately, a balance must be struck between the external validity that using a generic score affords, while capturing the unusual pathophysiology of dengue shock, and maximizing applicability to the LMIC setting. Our findings from this study suggest that mSOFA for dengue is associated with meaningful clinical outcomes and could be useful for prognosticating patients admitted with dengue shock. We believe that delta mSOFA for dengue should now be evaluated in action as an endpoint for clinical trials in moderate-severe dengue.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.