Diagnostic accuracy of peripheral venous lactate and the 2009 WHO warning signs for identifying severe dengue in Thai adults: a prospective observational study

Background Dengue is the most common mosquito-borne viral disease in humans. However, the sensitivities of warning signs (WSs) for identifying severe dengue in adults are low, and the utility of lactate levels for identifying severe dengue in adults has not been verified. Therefore, we aimed to evaluate the diagnostic accuracy of using peripheral venous lactate levels (PVL), as well as WSs established by the World Health Organization, for identifying severe dengue. Methods We prospectively evaluated individuals hospitalized for dengue who were admitted to the Hospital for Tropical Diseases in Thailand between May 2013 and January 2015. Blood samples to evaluate PVL levels were collected at admission and every 24 h until the patient exhibited a body temperature of <37.8 °C for at least 24 h. Data were recorded on a pre-defined case report form, including baseline characteristics, clinical parameters, and laboratory findings. Results Among 125 patients with confirmed dengue, 105 (84.0 %) patients had non-severe dengue, and 20 (16.0 %) patients had severe dengue. The presence of clinical fluid accumulation as a WS provided high sensitivity (75.0 %, 95 % confidence interval [CI]: 50.9–91.3 %) and specificity (90.5 %, 95 % CI: 83.2–95.3 %). The PVL level at admission was used to evaluate its diagnostic value, and receiver operating characteristic curve analysis revealed an area under the curve of 0.84 for identifying severe dengue. At the optimal cutoff value (PVL: 2.5 mmol/L), the sensitivity and specificity were 65.0 % (95 % CI: 40.8–84.6 %) and 96.2 % (95 % CI: 90.5–99.0 %), respectively. A combined biomarker comprising clinical fluid accumulation and/or PVL of ≥2.5 mmol/L provided the maximum diagnostic accuracy for identifying severe dengue, with a sensitivity of 90.0 % (95 % CI: 68.3–98.8 %) and a specificity of 87.6 % (95 % CI: 79.8–93.2 %). Conclusions Clinical fluid accumulation and/or PVL may be used as a diagnostic biomarker of severe dengue among adults. This biomarker may facilitate early recognition and timely treatment of patients with severe dengue, which may reduce dengue-related mortality and hospital burden.


Background
Dengue is the most common global mosquito-borne viral disease in humans [1]. Dengue is currently endemic in more than 125 countries, and Central and South America, South-East Asia, and the Western Pacific region are the most seriously affected [2,3]. In their 2020 goals for dengue, the World Health Organization (WHO) called for a 25 % reduction in morbidity and a 50 % reduction in mortality [1]. The key to reducing mortality from severe dengue is early recognition and timely treatment before the patient develops dengue shock and organ dysfunction [3].
Previous reports have demonstrated that there has been an epidemic shift from children to adults regarding those who experience more severe dengue [2,4,5]. In Thailand, a previous study reported that the number of adults with dengue has increased dramatically during recent decades [4]; more so, approximately 27.9 % of all hospitalized adults with dengue develop severe dengue, according to the 2009 WHO definitions [6]. A recent systematic review regarding the application of the 2009 WHO warning signs (WSs) for identifying severe dengue revealed a wide range of sensitivities (59-98 %) and specificities (41-99 %) [7]. In adults with dengue, the WSs have low sensitivities for identifying severe dengue, as they were developed based on findings that were limited by retrospective designs and small sample sizes [8,9]. Unfortunately, using WSs to identify severe dengue increases the hospital burden and work load of healthcare personnel as much as two fold. This is especially true for resource-limited countries and areas with endemic dengue, particularly Thailand [10,11].
Following dengue virus infection, the complications of dengue may be caused by an immunologic mechanism that results in increased capillary permeability, which leads to altered microcirculatory blood flow and ultimately to decreased tissue perfusion during the critical phase of dengue [12]. Under anaerobic conditions or in tissues with poor perfusion, lactate is formed by the reduction of pyruvate via the enzyme lactate dehydrogenase in order to produce ATP for cellular energy. In this context, serum lactate has been demonstrated to be a prognostic marker in critically ill patients [13,14]. Several reports have also demonstrated that serum lactate in critically ill patients was correlated with in-hospital mortality and organ dysfunction [15][16][17][18][19][20][21]. In a previous study, we showed that serum lactate is an independent factor for identifying severe dengue in adults [22]. Thus, we hypothesized that elevated lactate levels might serve as a diagnostic biomarker for identifying tissue hypoperfusion among patients with dengue before clinical manifestations of severe dengue develop. To investigate this hypothesis, we performed this prospective observational study among patients with dengue who were admitted to the Hospital for Tropical Diseases in Bangkok, Thailand. Using data from these patients, we evaluated the diagnostic accuracy of peripheral venous lactate (PVL) and WSs for identifying severe dengue.

Ethical considerations
The study's design was approved by the Ethics Committee of the Faculty of Tropical Medicine, Mahidol University (MUTM 2013-008-02) and the Ethics Committee of Phramongkutklao Hospital, Bangkok, Thailand (IRBRTA 1571/2013). The Standards for Reporting of Diagnostic Accuracy (STARD) were followed in this study [23]. Prior to participating in the study, written informed consent was obtained from all patients, (and/or the patient's guardians if the patient was 15-18 years old).

Study design
This prospective observational study was performed among patients who were admitted to the Hospital for Tropical Diseases (Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand) between May 2013 and January 2015. Individuals ≥15 years old were eligible for inclusion if they presented with clinical indicators of dengue, which were defined as acute fever and ≥2 of the following symptoms: 1) headache, 2) retro-orbital pain or ocular pain, 3) myalgia, 4) arthralgia, 5) rash, 6) a positive tourniquet test (≥20 petechiae per 1 square inch), or 7) leukopenia (defined as a white blood cell count <5.0 × 10 3 cells/μL). In addition, to be eligible for inclusion, participants had to have the presence of dengue viral infection confirmed via reverse-transcriptase polymerase chain reaction (RT-PCR) from an admission serum sample or dengue-specific immunoglobulin M (IgM) and immunoglobulin G (IgG) using enzymelinked immunosorbent assays (ELISA) of paired serum samples (taken upon admission and 2 weeks after admission). Patients with a history of underlying medical illness, mixed infection, or who were currently pregnant were excluded from this study.
Laboratory tests at admission included a complete blood count, blood chemistry analysis, and assessment of PVL levels. Blood samples for PVL were subsequently collected every 24 h until the patient exhibited a body temperature of <37.8°C for at least 24 h. To exclude the presence of other infections, hospitalized patients provided two blood cultures for microbiology analysis and underwent a urinalysis and chest radiography at admission. Other diagnostic tests for infectious diseases were also performed when clinical findings were suspect. The treating physicians were blinded to the results of the PVL tests. All patients with dengue received standard treatment according to the 2009 WHO guidelines for dengue [24]. All patient data, including baseline characteristics, clinical parameters, and laboratory findings, were recorded on a pre-defined case report form.

Case definitions for dengue
Using the 2009 WHO dengue case definitions [24], we classified patients as having either non-severe or severe dengue on the basis of their clinical and laboratory data. Patients with non-severe dengue were sub-categorized according to the presence or absence of WSs. Non-severe dengue without WSs was defined as an acute fever with ≥2 of the following symptoms: 1) nausea, 2) vomiting, 3) rash, 4) myalgia, 5) arthralgia, 6) a positive tourniquet test, or 7) leukopenia. The WSs included 1) abdominal pain; 2) persistent vomiting (defined as vomiting with signs of dehydration during physical examination); 3) clinical fluid accumulation that manifested as pleural effusion, ascites, or a serum albumin level <3.5 g/dL; 4) lethargy; 5) a liver span of >15 cm; 6) bleeding from a mucosal area, including the nose, gums, gastrointestinal tract, or vagina; 7) an increase in hematocrit of 2 % above the sex-specific reference range for a healthy Thai adult; and 8) a platelet count of ≤100 × 10 3 /μL. Patients were diagnosed with severe dengue if they met all of the criteria for non-severe dengue, as well as at least one of the following: 1) severe plasma leakage that caused shock or respiratory distress, 2) severe clinical bleeding (i.e., bleeding in the vital organs or spontaneous bleeding from a mucosal area that required a blood transfusion), or 3) severe organ involvement (evidenced by an aspartate aminotransferase [AST] level >1,000 U/L, an alanine aminotransferase [ALT] level >1,000 U/L, a serum creatinine level ≥3 times above baseline, myocarditis, and/or encephalitis). Plasma leakage was defined as a ≥20 % increase in hematocrit (above the reference range) or clinical evidence of fluid accumulation. Shock was defined as a systolic blood pressure of <90 mmHg with a <0.5 mL/kg/h decrease in urine output, impaired consciousness, AST level >1,000 IU/L, or ALT level >1,000 IU/L. Respiratory distress was defined as a respiratory rate of ≥24 breaths/min with <95 % oxygen saturation on room air and/or the need for oxygen therapy.

RT-PCR
Dengue viral RNA was extracted from each patient's serum using a 2-step PCR method, as described by Lanciotti et al. [25] and modified using the methods outlined by Reynes et al. [26]. The RNA was extracted from the acute serum samples using the PureLink® Viral RNA/DNA Mini Kit (Invitrogen™, USA), according to the manufacturer's instructions.

Serology for dengue viral infection
All sera were tested using 4 separate captures in ELISA assays for IgM and IgG against the dengue virus and the Japanese encephalitis virus, as described by Innis et al. [27]. To differentiate between dengue and other flavivirus infections, we determined the ratio of dengue IgM to Japanese encephalitis virus IgM, with a ratio of ≥1.0 indicating dengue virus infection and a ratio of <1.0 indicating other flavivirus infection. The ratio of dengue IgM to dengue IgG was also calculated, with a ratio of ≥1.8 indicating a primary dengue infection and a ratio of <1.8 indicating a secondary dengue infection.

Measurement of peripheral venous lactate
Blood samples were collected from a vein in an upper extremity without the use of a tourniquet. The 2 mL blood sample was placed in a vacutainer that contained sodium fluoride, and the sample was immediately placed on ice, sent to the laboratory center, and analyzed for lactate within 10 min of being drawn. A colorimetric assay with an enzymatic reaction that converted lactate to pyruvate was performed using an auto-analyzer (Roche/Hitachi Cobas C Systems, USA), according to the manufacturers' protocols. In this assay, L-lactate is oxidized to pyruvate and hydrogen peroxide by lactate oxidase, and the hydrogen peroxide reacts with peroxidase to generate a colored dye. The intensity of the color is directly proportional to the L-lactate concentration, as measured by the increase in absorbance. Lactate levels are expressed as mmol/L. The laboratory personnel were blinded to the sample sources, and the coefficient of variation for this assay in our laboratory was 1.1 %.

Sample size calculation
A previous retrospective study at the Hospital for Tropical Diseases indicated that the incidence of severe dengue was 27.9 % among adults who were hospitalized with dengue according to the 2009 WHO definitions [6]. However, it has also been estimated that the incidence of severe dengue would be approximately 20 % among dengue patients if they received close observation throughout treatment and care during their hospitalization [22]. Based on this information, we calculated that a sample size of 120 patients was needed for this study, using a specificity of 90 % with a confidence interval of ±6 %.

Statistical analysis
All data were analyzed using SPSS software (version 18.0; SPSS Inc., Chicago, IL). Numerical variables were tested for normality using the Kolmogorov-Smirnov test. Variables with non-normal distribution were summarized as a median and inter-quartile range (IQR) and were compared using the Mann-Whitney U test for 2group comparison. Categorical variables are expressed as frequencies and percentages, and were analyzed using the Chi-square test or Fisher's exact test, as appropriate. The optimal lactate cutoff value for identifying severe dengue was determined using receiver operating characteristic (ROC) curves. Diagnostic parameters were evaluated using 2 × 2 tables and 95 % confidence intervals (CI) to determine their sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), positive likelihood ratio (LR+), and negative likelihood ratio (LR-). All tests of significance were 2-sided, with a p-value <0.05 indicating statistical significance.
Daily changes in PVL and daily fluid intake in patients with severe and non-severe dengue We analyzed the daily changes in PVL levels during hospitalization among patients with severe and nonsevere dengue. Patients with severe dengue had significantly higher PVL levels (compared to patients with non-severe dengue) at admission through day 3 of hospitalization. However, they exhibited a decrease in median PVL levels to <2.0 mmol/L on days 3 and thereafter (Fig. 3a). To determine the amount of fluid intake during hospitalization, we evaluated daily oral and intravenous fluid intake to determine the total fluid received among patients with severe and nonsevere dengue. The amount of fluid intake was significantly lower among patients with severe dengue than those with non-severe dengue during day 1 through day 6 of hospitalization. In addition, the amount of intravenous fluid administered was significantly higher among patients with severe dengue than among those with non-severe dengue during day 1 and day 2 of hospitalization. However, the total fluid received was similar in both groups (Fig. 3b). Regarding the type of intravenous fluid, all patients with non-severe dengue received 0.9 % sodium chloride with or without dextrose as fluid therapy. Of 20 patients with severe dengue, 17 (85.0 %) patients received 0.9 % sodium chloride with or without dextrose and 3 (15.0 %) patients received Ringer's acetate with or without dextrose as initial fluid resuscitation. However, 5 (25.0 %) patients received dextran and 4 (20.0 %) patients received 5 % albumin for additional fluid resuscitation.

Discussion
A patient's lactate level is a biomarker for the severity of systemic hypoperfusion independent of organ failure and shock [20]. In addition, elevated arterial or central venous lactate can predict in-hospital mortality in many conditions and diseases, including infection, sepsis, liver disease, trauma, and cardiac arrest [14]. Moreover, our previous studies revealed that hospitalized adults with dengue who had PVL levels ≥2.0 mmol/L at admission were at risk for the development of severe dengue (odd ratio, 7.340) [22]. Nevertheless, the utility of lactate levels for identifying severe dengue in adults has not been verified thus far, and we hypothesized that elevated lactate levels might be a diagnostic biomarker for patients with dengue and tissue hypoperfusion before they develop clinical symptoms of severe dengue. Among the 125 hospitalized individuals with confirmed dengue viral infection, 105 (84.0 %) patients had non-severe dengue and 20 (16.0 %) patients had severe dengue. Interestingly, this incidence of severe dengue was lower than the incidence found in our previous retrospective study (27.9 %) [6]. This discrepancy may be due to the close observation and management of patients who participated in this prospective study. Our results also revealed that severe plasma leakage (75.0 %) was the most common complication among patients with severe dengue, which validates findings from a previous study [28]. Our analyses also revealed that most WSs (and the number of WSs) provided low specificities and sensitivities for identifying severe dengue. The one exception was the presence of clinical fluid accumulation (sensitivity: 75.0 %, specificity: 90.5 %, PPV: 60.0 %, NPV: 95.0 %, LR+: 7.9, and LR-: 0.3); previous studies have reported similar findings [8,9].
When we evaluated the diagnostic value of PVL levels at admission, we found that the optimal PVL cutoff value of 2.5 mmol/L yielded a sensitivity of 65.0 %, specificity of 96.2 %, PPV of 76.5 %, NPV of 93.5 %, LR+ of 17.1, and LR-of 0.4. Similarly, a previous systematic review of lactate levels among patients with suspected infection revealed that patients with lactate levels of 2.0-3.9 mmol/L had an increased risk of progressive organ dysfunction and in-hospital mortality, including patients without hypotension [29,30]. The combined biomarker of clinical fluid accumulation and/or PVL level of ≥2.5 mmol/L provided the maximum diagnostic accuracy (sensitivity: 90.0 %, specificity: 87.6 %, PPV: 58.1 %, NPV: 97.9 %, LR+: 7.3, and LR-: 0.1) for identifying severe dengue. These findings might be due to plasma leakage resulting from increased microvascular permeability that then leads to elevated hematocrit levels and/or clinical fluid accumulation; this may subsequently cause impaired microvascular perfusion, resulting in elevated lactate levels and eventually shock with organ dysfunction [31,32]. Therefore, the combined biomarker of clinical fluid accumulation and/or PVL levels ≥2.5 mmol/L could provide the maximum diagnostic accuracy for identifying severe dengue.
Regarding fluid therapy for patients with dengue included in this study, the type and amount of fluid therapy were adjusted on the basis of the WHO South East Asia Regional Organization (SEARO) guidelines [2]. The majority of patients received 0.9 % sodium chloride with or without dextrose as initial fluid therapy. In patients with severe dengue, colloids including dextran and 5 % albumin were used as additional fluid resuscitation. Patients with severe dengue received a significantly lower amount of oral fluid intake per day during day 1 and day 6 of hospitalization, but the amount of intravenous fluid administered per day was significantly higher during day 1 and day 2 of hospitalization. However, the total fluid intake per day was similar in both groups. After fluid therapy, the PVL levels decreased, but the PVL levels were significantly higher among patients with severe dengue than among those with non-severe dengue during day 1 and day 3 of hospitalization.
Our study was designed as a prospective observational study, which could minimize the risk of missing data and reduce bias. Blood samples were regularly collected under blinded conditions for treating physicians and investigators in order to demonstrate, without bias, the dynamic changes of PVL throughout the febrile phase as well as the critical and recovery periods. However, this study had some limitations. First, we recruited only patients with dengue at a single center for tropical diseases in Thailand, which is likely not representative of the global population of patients with dengue. Second, we enrolled only patients who were hospitalized with dengue.  Third, patients were admitted to the hospital at different times after fever onset, but all patients with dengue were admitted during the febrile phase.

Conclusions
The presence of clinical fluid accumulation and/or PVL of ≥2.5 mmol/L could be used as a diagnostic biomarker for severe dengue among patients hospitalized with dengue. This biomarker might facilitate early recognition and timely treatment of patients with severe dengue, which can reduce dengue-related mortality, hospital burden, and unnecessary hospitalizations.