Predictive markers for the early prognosis of dengue severity: A systematic review and meta-analysis

Background Predictive markers represent a solution for the proactive management of severe dengue. Despite the low mortality rate resulting from severe cases, dengue requires constant examination and round-the-clock nursing care due to the unpredictable progression of complications, posing a burden on clinical triage and material resources. Accordingly, identifying markers that allow for predicting disease prognosis from the initial diagnosis is needed. Given the improved pathogenesis understanding, myriad candidates have been proposed to be associated with severe dengue progression. Thus, we aim to review the relationship between the available biomarkers and severe dengue. Methodology We performed a systematic review and meta-analysis to compare the differences in host data collected within 72 hours of fever onset amongst the different disease severity levels. We searched nine bibliographic databases without restrictive criteria of language and publication date. We assessed risk of bias and graded robustness of evidence using NHLBI quality assessments and GRADE, respectively. This study protocol is registered in PROSPERO (CRD42018104495). Principal findings Of 4000 records found, 40 studies for qualitative synthesis, 19 for meta-analysis. We identified 108 host and viral markers collected within 72 hours of fever onset from 6160 laboratory-confirmed dengue cases, including hematopoietic parameters, biochemical substances, clinical symptoms, immune mediators, viral particles, and host genes. Overall, inconsistent case classifications explained substantial heterogeneity, and meta-analyses lacked statistical power. Still, moderate-certainty evidence indicated significantly lower platelet counts (SMD -0.65, 95% CI -0.97 to -0.32) and higher AST levels (SMD 0.87, 95% CI 0.36 to 1.38) in severe cases when compared to non-severe dengue during this time window. Conclusion The findings suggest that alterations of platelet count and AST level—in the first 72 hours of fever onset—are independent markers predicting the development of severe dengue.

• Embase, POPLINE, PubMed, Scopus, SIGLE: Dengue AND (marker* OR biomarker* OR factor*) AND (sever* OR DF OR DSS OR DHF OR shock) AND (predict* OR prognos* OR correlat* OR associat* OR relat* OR (logistic regression)) AND (early OR defervescence OR progress* OR develop*) Types of study to be included Cross-sectional and cohort observational (prospective and retrospective) studies.

Condition or domain being studied
Dengue infection is getting more common in recent years, affecting all age groups.(1) This mosquito-borne virus disease has rapidly spread throughout the tropics with local variations, influenced by rainfall, unplanned rapid urbanization, and temperature. (2) The up-to-date-minute figures point out nearly 2.5 billion persons and 40% of the world's population living in areas at risk of infection.(3) Approximately 50 to 100 million new dengue infections and 22,000 deaths are reported annually, predominantly in children.(4) Supportive treatment and close monitoring are now the first-line therapy given for dengue. There are four serotypes of dengue fever virus. Recovery from one serotype does not provide complete immunity against other serotypes. A subsequent infection by the other serotypes increases the risk of developing severe dengue.(2) As usual, resolution of the febrile response occurs around days 4-7, when most patients completely recover, but other individuals can experience complications.(5) The mechanism of this progression is not yet well understood. World Health Organization (WHO) has revised the new guideline for dengue diagnosis with the amendment of warning signs. Warning signs are to aid in recognizing patients on course to develop dengue hemorrhagic fever (DHF) or dengue shock syndrome (DSS), also known as severe dengue.(6) Although the revised scheme is more sensitive to the diagnosis of severe dengue, there remain impediment to the applicability. (7) In addition, many signs are subjective and present very late. (8) The foresight of DHF and DSS are unmet. Hence, there is an urgent need to identify the predictive factors of severe dengue. The corpus of this study is to find out the reliable predictors for severe dengue in terms of clinical signs, immunological, genetic and biochemical markers.

Participants/population
Inclusion criteria: • Publications reporting predictive markers in terms of clinical signs, immunology, biochemistry and gene associated with dengue severity comprising dengue fever (DF), dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS) The early reliable predictive markers in terms of clinical signs, immunology, biochemistry and gene associated with dengue severity comprising dengue fever (DF), dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS).

Secondary outcome(s)
Other predictive markers considered to be a candidate to predict dengue severity.

Data extraction (selection and coding)
Data extraction will be done by two independent reviewers and the conflicts amongst the reviewers will be Page: 2 / 4 PROSPERO International prospective register of systematic reviews consulted until the consensus is reached through the empiric counsellor. A data extraction form of the eligible articles is tabulated under Excel file by two independent reviewers so as to compile the data.

Risk of bias (quality) assessment
Two reviewers will independently assess the quality of the selected studies to evaluate the risk of bias by using Cochrane Collaboration's tool. Each of the studies will be classified as low risk, high risk or unclear risk of bias and any conflict will be resolved with discussion till a consensus is reached. If authors are not able to achieve agreement, senior authors/supervisors will be consulted.
Strategy for data synthesis Meta-analyses for respective factors will be performed separately using Comprehensive Meta-analysis software version 2.0 (http://www.meta-analysis.com) in which there is more than one study. The odds ratio (OR) will be computed together for both dichotomous and continuous variables when there are two groups of case and control. The I² index will be used to assess heterogeneity amongst studies. Random effects and fixed effect models will be used to calculate the mean effect size of studies with significant heterogeneity (I² >75%) and without significant heterogeneity (I² <75%), respectively. Pooled OR with the corresponding 95% confidence intervals will also be calculated. We use fixed-effect model with weighting of the studies when there is a lack of significant heterogeneity (p >0.10) and use random-effects model with weighting of the studies when there is heterogeneity between studies (p &lt; 0.10).