Clinical signs and symptoms associated with WHO severe dengue classification: a systematic review and meta-analysis

ABSTRACT The World Health Organization (WHO) introduced the new dengue classification in 2009. We aimed to assess the association of clinical signs and symptoms with WHO severe dengue classification in clinical practice. A systematic literature search was performed using the databases of PubMed, Embase, and Scopus between 2009 and 2018 according to PRISMA guideline. Meta-analysis was performed with the RevMan software. A random or fixed-effect model was applied to pool odds ratios and 95% confidence intervals of important signs and symptoms across studies. Thirty nine articles from 1790 records were included in this review. In our meta-analysis, signs and symptoms associated with higher risk of severe dengue were comorbidity, vomiting, persistent vomiting, abdominal pain or tenderness, pleural effusion, ascites, epistaxis, gum bleeding, GI bleeding, skin bleeding, lethargy or restlessness, hepatomegaly (>2 cm), increased HCT with decreased platelets, shock, dyspnea, impaired consciousness, thrombocytopenia, elevated AST and ALT, gall bladder wall thickening and secondary infection. This review shows new factors comorbidity, epistaxis, GI and skin bleeding, dyspnea, gall bladder wall thickening and secondary infection may be useful to refine the 2009 classification to triage severe dengue patients.


Introduction
Dengue is the fastest spreading mosquito-borne viral disease globally, affecting 50 million individuals every year [1]. In the vast majority of individuals, dengue fever is a self-limiting disease that requires minimal supportive treatment. However, in less than 1% of patients, symptoms of severe dengue, including clinical fluid accumulation, shock, and multiple organ dysfunction could spell impending demise if left untreated. The new 2009 WHO classification for dengue was hence created to allow clinicians to triage patients easily according to their clinical presentations for more effective clinical management ( Figure 1) [1]. This new classification is intended to bring greater clarity on the severity of clinical presentations compared to the 1997 classification of dengue into undifferentiated fever, dengue fever [1] and dengue hemorrhagic fever.
The 1997 classification was proven to underestimate the severity of dengue infection [2]. Multiple studies had shown that plasma leakage causing clinical fluid accumulation, transaminitis and thrombocytopenia were more indicative of severe dengue instead of clinical manifestations of bleeding, as was prioritized in the old classification [3,4]. In febrile travelers returning from endemic regions, one study showed that a significant number of cases of severe dengue would have been missed if the WHO diagnostic criteria for dengue haemorrhagic fever would have been applied [3]. While many studies have effectively highlighted the shortcomings of the 1997 classification, there is a paucity of studies done today to ascertain if the clinical utility of the current 2009 classification has improved clinical diagnosis and management of dengue infections.
Previous review has reported that the new 2009 classification has a higher sensitivity and specificity compared with the 1997 classification [5]. However, there was a question for applicability in clinical practice and usefulness for triage using the revised dengue classification [6]. Several studies have assessed the association of clinical factors with severe dengue [7][8][9][10]. However, risk factors reported among severe dengue patients remained inconsistent [7][8][9][10]. The objective of this review was to synthesize the best of available evidence by conducting meta-analysis to assess the factors associated with severe dengue patients. This systematic review and meta-analysis therefore investigate the likelihood of new factors associated with severe dengue, which may be useful to further revise the existing dengue 2009 classifications for more accurate triaging of patients.

Search strategy
This review was conducted according to the standards outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [11]. No documented review protocol exists for this meta-analysis. The year 2009 was selected as the start date of searching articles as the introduction of new WHO dengue case classification in 2009 [1]. The search was performed in three databases: PubMed, Embase, and Scopus; covering literature between the period of January 2009 and December 2018. Manual search for reference lists of included studies was performed to check additional studies relevant to the topic. The keywords used in search are "dengue" OR "severe dengue" OR "dengue severity" AND "diagnosis" OR "clinical diagnosis" OR "warning signs." All the references were imported and removed duplicates by using bibliographical software package, EndNote version X7 (Thomas Reuters, New York, NY, USA). The studies were screened independently against the inclusion and exclusion criteria by two authors (TP and XZ), and a third author (PJ) resolved disagreement between the two reviewers regarding eligibility of a study.

Inclusion and exclusion criteria
Studies were included if they met the following criteria: (1) any type of studies (retrospective, prospective, or cohort, case-control, cross-sectional studies) reporting severe dengue (defined with 2009 WHO diagnosis criteria) compared with dengue fever; (2) studies that distinguished clinical signs and symptoms and/or laboratory features of severe dengue and dengue fever with or without warning signs; (3) studies that published on and after 2009; (4) studies that classified dengue severity according to new 2009 WHO classification; (5) studies that included either children or adults only or both children and adults. We excluded studies if they were narrative review, letters to editors, case reports and case series, incomplete information to extract data and not written in English.

Quality assessment
Two of the authors (TP and XZ) independently assessed the quality of each included study using the Newcastle-Ottawa Scale (NOS) [12]. NOS is the risk assessment tool developed to assess the quality of non-randomized studies used in systematic review and meta-analysis. It consists of three parameters of quality i.e. selection, comparability, and exposure with maximum of 4 points for selection of study groups, 2 points for comparability of groups and 3 points for exposures and outcomes. The NOS scores were divided into low quality (scores 1-3), intermediate quality (scores 4-6), and high-quality (scores 7-9) [13]. When any difference in opinion of quality assessment between the two authors happened, it was solved by a third author (PJ) via discussion and consensus.

Data extraction
The data were extracted from each study through structured data extraction forms. Items extracted for the characteristics of studies included the authors, year of publication, country, setting of study, study design, study population (children, adult or both), numbers of patients for dengue fever (with or without warning signs) and severe dengue, and diagnosis of dengue. Outcome data (clinical signs and symptoms and/or laboratory features) for severe dengue and dengue fever were extracted and compiled in the summary tables by one author (HTP), and cross-checked by another author (XZ) for accuracy and relevance.

Data analyses
Data were analyzed using RevMan software (Review Manager Version 5.3.5, The Nordic Cochrane Centre, Copenhagen). Dichotomous data was analysed using the Mantel-Haenszel (M-H) method; odds ratio (OR) with 95% confidence interval (CI) was calculated using either a fixed-effect or random-effect model with at least four or more studies though only 2 studies are needed for a meta-analysis theoretically. The test of overall effect was assessed using z-statistics at P < 0.05. Heterogeneity between studies was evaluated using the Cochrane Q (χ 2 test) and I 2 test. I 2 value considered to 0% as no, 25% as low, 50% as moderate and 75% as high heterogeneity [14]. The statistical significance for heterogeneity was set with a P value < 0.10. The fixedeffects model with weighting of the studies was used when there was a lack of significant heterogeneity (P > 0.10), while the random-effects model with weighting of the studies was used when there was heterogeneity between studies (P < 0.10) [14]. Sensitivity testing to identify the effect of the subgroups was performed by subgroup analysis based on study population. Subgroup analysis was performed to (1) explore the potential sources of heterogeneity among the studies and (2) evaluate the effect in a specific subgroup. The predefined subgroups were study population (children, adult, or both) and dengue severity (severe dengue or dengue fever with or without warning signs).

Results
Study characteristics and quality Figure 2 illustrates searching articles and the selection process. A total of 1790 records were identified, whereas a total of 478 duplicates were removed. The initial screening yielded 1312 articles, of which 246 articles were assessed for full text reading. A total of 207 full-text articles were excluded for the reasons mentioned in the study flow chart ( Figure 2). Finally, 39 articles [2,[7][8][9][10] were selected for inclusion in this meta-analysis according to the WHO classification for dengue, namely dengue without warning signs, dengue with warning signs and severe dengue, as well as unclassified signs or laboratory features. The date set for searching was 2009, all the studies were published after 2009. Table 1 provides a summary characteristic of prospective study (n = 16), retrospective study (n = 21) and case control study (n = 2). Sample sizes were varied among the studies, ranging from 8 to 2060 cases. This study included a population of children (n = 18), adult (n = 14) and both (n = 7) and they are from varying locations: Asia (n = 31), Brazil (n = 4), Germany (n = 1), Mexico (n = 2), and Spain (n = 1). Most studies were performed in hospital settings (n = 36) than healthcare network (n = 1), medical education and research institute (n = 1), tertiary care unit (n = 1). Comorbidities were reported in ten studies, the proportion of comorbidity varied from 0% to 100% in severe dengue and 13% to 55.7% in dengue fever with or without warning signs. Nineteen studies reported day of presentation of illness or fever, whereas median day of illness ranged from 3.5 to 5 days in severe dengue and 2-5 days in dengue fever with or without warning signs. Dengue infection was confirmed by clinically in two studies, whereas serology, ELISA, PCR, HIA, viral isolation, and nucleotide detection was used together with clinical diagnosis in 37 studies for confirmation of dengue infection. Assessing the quality of the studies by Newcastle-Ottawa Scale, 5 studies were high quality (scores 7-9), 33 studies were intermediate quality (scores 4-6) and only one study has low quality (scores 1-3).

Other signs and symptoms and laboratory features
Other symptoms of cough and diarrhoea in association with dengue infection were analysed but yielded non-significant results. Associated laboratory features of thrombocytopenia (OR: 2.70, CI: 1.60-4.55, z = 3.73, P < 0.001) was positively associated with severe dengue while hypoalbuminemia found no association with severe dengue (P > 0.05). The presence of a secondary dengue infection (a patient having a second or more dengue infection) was also statistically significant in the odds of being diagnosed with dengue infection (OR: 1.93, CI: 1.25-2.97, z = 2.96, P < 0.01).

Discussion
Our detailed meta-analysis comprises studies encompassing numerous countries globally suggests the current 2009 WHO clinical classification optimally identifies severe dengue infection.
Our study lies in its detailed meta-analysis of a wide range of studies encompassing numerous countries globally. We found that patients with comorbidity had 2-times higher risk of progression into severe dengue. This finding is in line with previous study indicating that pre-existing comorbidities were risk factors of severe organ involvement in dengue patients [49]. Digestive factors of vomiting, persistent vomiting, abdominal pain or tenderness were indicative of severe dengue in our study, which is consistent with previous study showing that vomiting and abdominal pain were most prevalent warning signs which occur prior to severe dengue [50]. Bleeding manifestations include mucosal bleeding (epistaxis, gum bleeding), GI bleeding (hematemesis and/or melena) and skin bleeding (petechiae, purpura, ecchymosis) were shown as valuable predictors of severe dengue in our study except for hematuria and vaginal bleeding. Consistent with previous meta-analyses, four kinds of bleeding: epistaxis, gum bleeding, hematemesis, and melena were related to the risk of development of patients with severe dengue [13]. Among bleeding factors, gastrointestinal bleeding proved highly indicative of severe dengue. A study also showed that patients with gastrointestinal bleeding had the highest risk of progressing into severe disease [23]. Notably, pleural effusion and ascites were significantly associated with severe dengue. Plasma leakage causing fluid accumulation, during which fever transitions into defervesence, was cited as a critical indicator of progression to severe dengue [33]. Concurrent increase in haemotocrit and rapid decrease in platelet count, vomiting and abdominal distention were significant in predicting the likelihood of severe plasma leakage as a warning sign of dengue [22]. In one Singaporean study, concurrent increase in haemotocrit and decrease in platelet count were found to be predictive of severe haemorrhage [22], which is consistent with our result.
Liver damage is a common complication of dengue, liver enzymes are valuable markers during dengue infection [51]. In our results, hepatomeagly (>2 cm), elevated AST and ALT were significantly different between severe dengue and dengue with or without warning signs. These findings are similar to the previous studies that liver enlargement and liver enzymes (AST and ALT) were significantly higher in severe dengue patient [48,52]. Interestingly, four articles highlighted the presence of gallbladder wall thickening as a clinical sign of dengue infection and which was found to be associated with severe dengue. In multiple studies, this was characteristic only for severe dengue [25,53]. One study showed that gallbladder thickening was present even before serological tests were positive [54] and as potential early predictors [53]. While thrombocytopenia was a significant predictor for severe dengue in many studies [52,55], our result revealed that platelet count less than 150000/mm 3 has value in ruling in dengue infection. However, one study surprisingly showed that it was unlikely to be a direct precipitant for clinical manifestations of bleeding [38]. Our analysis showed association of secondary dengue infection with severe dengue. As proven by other studies, patients presenting with a secondary dengue infection were associated with a higher risk of developing severe dengue [33,56], which suggests that the clinical presentation of severe dengue was affected by both host factors (secondary immune response and viral load) [57].
Our review has several limitations. Firstly, there was variability among the included studies in terms of study designs, study population, diagnoses, comorbidities and day of presentation of illness or fever, which weakens the comparison among different studies. The definitions and cutoff values of warning signs and severity were widely varied within the studies [58], which brought heterogenous application to rule out the cases. Some identified studies were performed on individuals of one demographic, such as being either from the paediatric or adult age group, which can lead to unaccounted variation in presenting signs or symptoms. Secondly, research conducted in regions endemic for dengue infection, especially countries near the equator, constituted an overwhelming majority in our study. Therefore, studies of dengue infection in less endemic countries could have been elided over, conferring selection bias for our study. Thirdly, in our meta-analysis for dengue without warning signs, it was unfortunate that a majority of the listed symptoms did not prove significant. Many listed symptoms that we studied also did not stem from an acceptable level of heterogeneity.
Therefore, these clinical signs and symptoms may be useful for triaging potential severe dengue in patients and may further guide further enhancement of the current WHO dengue severity classifications, though heterogenicity was considerably high. More large-scale multicenter studies may be carried on identifying the association of with severe dengue using standard definitions and classification.

Disclosure statement
No potential conflict of interest was reported by the author (s).