Neutrophil-to-lymphocyte ratio for the clinical management of dengue: a systematic review and meta-analysis

1 Introduction

Dengue is the most widespread mosquito-borne viral disease, with an estimated 390 million annual infections globally (1, 2), and its incidence continues to rise, with approximately two-fifths of the world's population living in endemic regions (3). Transmitted primarily by Aedes aegypti and Aedes albopictus mosquitoes, the dengue virus (DENV) poses a significant threat in endemic countries, such as those in Asia, including India (4). Understanding the disease's clinical characteristics and laboratory parameters is critical for reducing associated mortality and morbidity (3), particularly given its progression through three distinct pathophysiological stages: febrile, critical, and convalescent phases (2).

The symptoms of dengue are similar to those of other viral infections, and clinical diagnosis based on symptoms and physical examination alone has limited accuracy. This underscores the need for accessible, cost-effective, and reliable diagnostic tools. Currently, the gold standard for diagnosing dengue is virus culture. Other tests, such as genomic detection, antigen detection, antibody detection, or biosensors, are effective tests that detect infection but have the major drawback of cost-effectiveness, specialized equipment, and the need for specialized laboratories (5–8). Rapid diagnostic methods have increased for prognostic evaluation because early detection of severe dengue could guide decision-making. Peptide-based biosensors, graphene-derived nanomaterial sensors, and immuno-magnetic agglutination tests have been evaluated to improve point-of-care detection for predicting complications by quantifying markers such as NS1, cytokines, etc., or specific serotypes (9–11). There is a growing importance of integrating diagnostic and prognostic tools to improve outcomes in dengue infections.

Early dengue infection is strongly associated with leukopenia, generally <5000 cells/mm3, in the critical part of the febrile phase. There is a greater proportional decrease in neutrophils than in lymphocytes, thereby changing the neutrophil-to-lymphocyte ratio (NLR), which could provide a prognosis of disease progression. Leukopenia is followed by a sharp drop in platelets, which often causes plasma leakage (5, 12). The use of NLR as a biomarker of infection is a low-cost and easily measured option and has been correlated with the severity of hemorrhagic manifestations and shock (13–15).

There is evidence that NLR is useful as a biomarker for other diseases, such as neuromyelitis optica spectrum disorder (NMOSD), which is increased in the patient group compared to the healthy control group, and is also associated with poor prognosis (15). NLR independently predicts adverse outcomes in patients hospitalized with community-acquired pneumonia (16), and there has been discussion about its potential as a promising marker for identifying post-surgical infections (17).

At present, there is no consensus on the usefulness of NLR as a biomarker and standardized use for dengue. Based on this background, this systematic review with meta-analysis seeks to summarize the studies conducted to date on the behavior of NLR and perform an in-depth analysis of the association of this marker with dengue.

2 Methods

This systematic review and meta-analysis follows the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement and the Cochrane Handbook for Systematic Reviews of Interventions (18). We have registered our protocol in PROSPERO with the identifier CRD42024538357.

2.1 Data sources and searches

A comprehensive systematic search was conducted in the databases PubMed, Embase, Scopus, Web of Science, and Google Scholar for studies published up to April 2024. The search in Google Scholar was specifically designed to identify relevant gray literature. We built the search strategy using both controlled vocabularies (MeSH, Emtree) and free-text terms relevant to “dengue” and “neutrophil-lymphocyte ratio” (Supplementary Material S1-Search strategy) following the Peer Review of Electronic Search Strategies Guidelines (19). We did not impose language or publication date restrictions. No deviations from the registered PROSPERO protocol occurred.

2.2 Eligibility criteria

We included studies that satisfied all these conditions: (1) participants with laboratory-confirmed dengue, (2) NLR measured in dengue patients, and (3) analytical observational designs (cohort, case-control, cross-sectional). We excluded narrative reviews, animal studies, case reports, conference abstracts, letters, and editorials. Additionally, we hand-checked reference lists of included studies and relevant reviews for further eligible articles.

2.3 Study selection

Two independent reviewers (GLP and CCB) imported search results into reference management software, EndNote X9, removed duplicates, and uploaded the remaining records into the screening tool Rayyan. They screened titles and abstracts independently and in duplicate and, in case of disagreement, resolved conflicts by discussion or with a third reviewer (MCL). Next, they independently and in duplicate reviewed full texts of selected articles and again reconciled any discrepancies. The complete list of all excluded articles is reported in (Supplementary Material S2- Excluded studies).

2.4 Outcomes

We aimed to evaluate NLR as a biomarker in patients with dengue across four primary outcomes: (1) DF vs. healthy controls; (2) DF vs. fever of unknown etiology (FUE), COVID-19, malaria, or scrub typhus; (3) classification of DF according to the World Health Organization (cases without warning signs, cases with warning signs and severe dengue) (20); and (4) mortality in dengue. A secondary outcome included its association with hemoglobin levels.

2.5 Data extraction

Pairs of reviewers (CCB and GLP) independently and in duplicate extracted data using a standardized form and adjudicated disagreements by consensus or via a third reviewer (MCL). Extracted information included study title, first author, year, design, country, main inclusion criteria, sample size for patients with dengue and controls, age, sex, timing of blood sampling, mean NLR with standard deviation according to sample stratification, sensitivity, specificity, outcomes, and time of follow-up for outcome measure. If data remained unclear or missing, we contacted the corresponding authors by email.

2.6 Risk of bias

The quality assessment of the included nonrandomized studies was performed independently by two authors (CQP and GLP) using the Newcastle–Ottawa scale (NOS) (21), assessing the domains of selection, comparability and ascertainment of exposure/outcome. A third reviewer (MCL) functioned as an adjudicator for any discrepancies in scores. Reviewers were blind to the study ascribed star-based score within the categories of each domain. Studies were classified as low risk of bias if they scored ≥6 stars, moderate risk if they scored 4-5 stars, and high risk if they scored <4 stars, applying the same absolute threshold across case-control and cohort studies (maximum 9 stars) and cross-sectional studies (maximum 8 stars, using the NOS adaptation for cross-sectional designs).

2.7 Data analysis

We performed meta-analyses using the R statistical package (version 4.3.2) with the metacont function of the meta package (version 7.0-0), using a random-effects inverse variance model to pool mean differences (MD) with 95% confidence intervals (CI). We estimated heterogeneity via I² (low: < 30%; moderate: 30%–60%; high: > 60%) and computed between-study variance (τ²) using the DerSimonian–Laird estimator. For diagnostic accuracy, we synthesized sensitivity and specificity using a bivariate random-effects model to derive summary ROC curves using the diagmeta package (version 0.5-1) while accommodating heterogeneity in cut-off points. Sensitivity analyses included leave-one-out. For studies reporting median and interquartile range (IQR), values were converted to mean and standard deviation using the method of Wan et al. (22).

2.8 Certainty of evidence

Main author (MCL) applied the Grading of Recommendation, Assessment, Development, and Evaluation (23) framework to rate certainty in our results, judging study limitations (risk of bias), imprecision (confidence interval width and sample size), indirectness (generalizability), inconsistency (heterogeneity across studies), and publication bias. We also considered effects of large magnitude, dose-response gradients, and plausible residual confounding under the guidelines of the GRADE handbook (24). Based on this framework, we categorized overall evidence certainty as high, moderate, low, or very low.

For diagnostic test accuracy outcomes (sensitivity, specificity, and AUC), certainty of evidence started at high, following standard GRADE guidance for diagnostic test accuracy studies. All other outcomes (mean differences, odds ratios, and beta coefficients) started at low, consistent with GRADE guidance for observational studies of association.

3 Results

3.1 Study selection

The systematic search retrieved 193 records. After removing duplicates, 78 studies remained for title and abstract screening, resulting in the inclusion of 15 eligible articles in the final analysis (3, 8, 14, 25–36) (Supplementary Material S3- PRISMA flowchart of included studies).

3.2 Characteristics of studies

We included 15 observational studies: 2 case–control, 2 cohort, and 11 cross-sectional, published from 2017 to 2024. Most studies were from Asia and Latin America, with a pooled sample of 6166 subjects. Participant age ranged from 7.6 ± 4.8 years in pediatrics to a median of 60 years in adults. NLR was reported as mean ± SD or median (IQR) across studies. Dengue infection was diagnosed by NS1 antigen, IgM/IgG serology, or WHO criteria. NLR was measured at admission or in the first few days of hospitalization. Full details are provided in Table 1.

Table 1. Characteristics of the included studies (n = 15).

3.3 Risk of bias assessment

The methodological quality of the 15 included studies was evaluated using NOS. Of these, two were case-control, two cohorts, and 11 cross-sectional studies. All case-control and cohort studies were rated having a low risk of bias, showing adequate ascertainment of exposure, although some lacked detailed reporting on control selection or follow-up adequacy. Among the cross-sectional studies, all demonstrated a low risk of bias with minor limitations regarding sample size justification, representativeness, or confounder control. Overall, the studies exhibited acceptable methodological quality, supporting the reliability of exposure and outcome assessments across the included evidence [Supplementary Material S4- The Newcastle-Ottawa Scale (NOS)].

3.4 DF vs. healthy controls

Bhattarai et al. (27) found that dengue patients had higher NLR than healthy controls, with an MD of 1.29 (95% CI: 0.87 to 1.71) (Table 2).

Table 2. Narrative synthesis of the included studies evaluating the neutrophil to lymphocyte ratio (NLR) in patients with dengue fever (DF).

Mean differences (MD) are calculated as dengue mean minus comparator mean (or poor-outcome mean minus good-outcome mean for the mortality comparison); positive values indicate higher NLR in dengue (or in the poor-outcome group).

SD, standard deviation; CI, confidence interval; NR, not reported; cβ, crude beta coefficient; RPR, relative probability ratio, aOR, adjusted odds ratio; aβ, adjusted beta coefficient.

*p < 0.05.

3.5 DF vs. FUE

Three studies and 948 participants (453 in the dengue group and 495 in the other febrile illness group) were analyzed. The overall analysis showed a non-significant difference in NLR between groups (MD: −0.22; 95% CI: −3.71 to 3.27, I² = 99.0%) (Figure 1). Sequential exclusion of individual studies did not resolve the extreme heterogeneity, with I² remaining near 99% throughout (Supplementary Material S5).

3.6 DF vs. COVID-19

Five studies, including 1,114 participants (398 with DF and 716 with COVID-19), were analyzed. Subgroup analysis by geographic region revealed a significant decrease in NLR with substantial heterogeneity (MD: −3.13; 95% CI: −5.99 to −0.27; I² = 99%) (Supplementary Material S6). Sequential exclusion of individual studies did not substantially reduce heterogeneity (Supplementary Material S7).

The summary of receiver operating characteristic (SROC) curve showed a sensitivity pool of 83.9% (95% CI: 69.5–92.2%) and a specificity of 83.9% (95% CI: 67.5–92.8%) (Figure 2). The area under the curve (AUC) was 0.907 (95% CI: 0.802–0.965), reflecting high overall diagnostic accuracy in distinguishing DF from COVID-19 infection.

3.7 DF vs. malaria

Kotepui et al. (30) evaluated NLR in patients with dengue vs. malaria and reported a significantly lower NLR in dengue compared with malaria, with an MD of −3.06 (95% CI: −3.47 to −2.65), a sensitivity of 71.5%, and a specificity of 64.9% (Table 2).

3.8 DF vs. scrub typhus

While Tran et al. (36) observed a higher mean NLR in dengue vs. scrub typhus, MD 0.44 (95% CI: 0.10 to 0.78), and an adjusted OR of 0.68 (95% CI: 0.57-0.81) (Table 2). Similarly, Chatterjee et al. (28) reported a relative percentage ratio (RPR) of 1.11 (95% CI: 0.27-4.58) for the same comparison, though this estimate was imprecise given the small sample size (n = 69).

3.9 DF severity

A total of three analyses explored NLR according to dengue severity. For DF with and without warning signs, three studies, including 541 patients (111 with warning signs and 430 without) showed a non-significant decrease in NLR with high heterogeneity (MD: −1.51; 95% CI: −3.43 to 0.41; I² = 94%) (Supplementary Material S8). Sequential exclusion of individual studies did not resolve the heterogeneity (Supplementary Material S9).

For severe DF vs. DF with warning signs, three studies, including 135 patients (24 severe and 111 with warning signs), showed a significant decrease in NLR with low heterogeneity (MD: −0.54; 95% CI: −0.91 to −0.17; I² = 0%) (Supplementary Material S10). Leave-one-out analysis confirmed stable pooled effects (Supplementary Material S11).

Similarly, for severe DF vs. DF without warning signs, three studies with 454 patients (24 severe and 430 without warning signs) revealed a non-significant NLR decrease with high heterogeneity (MD: −1.60; 95% CI: −3.65 to 0.45; I² = 87%), and sensitivity analyses confirmed the stability of the results (Supplementary Materials S12, S13).

3.10 Hemoglobin levels

Ashma et al. (14) observed an association between NLR and hemoglobin levels (adjusted β=1.782, p < 0.05; correlation 0.314) (Table 2).

3.11 Mortality

Agrawal et al. (3) reported that deceased patients had higher NLR compared with survivors, MD of 0.33 (95% CI: 0.10 to 0.56), with 60% sensitivity, 91.2% specificity, and an AUC of 0.909 (Table 2).

3.12 Certainty of evidence

Certainty of evidence started at high for diagnostic test accuracy outcomes (sensitivity, specificity, AUC) and at low for all other observational associations (mean differences, odds ratios, beta coefficients), following standard GRADE guidance. Certainty was subsequently downgraded for most outcomes due to imprecision (wide or unreported confidence intervals and fewer than 300 participants) and inconsistency (high heterogeneity across studies), and upgraded for one outcome due to a large effect size with a narrow confidence interval. Overall, the certainty of evidence ranged from moderate to very low, reflecting both methodological limitations and variability among the included studies (Table 3). This is consistent with GRADE guidance, in which certainty of evidence additionally reflects inconsistency and imprecision at the outcome level beyond individual-study risk of bias.

Table 3. Grade summary of findings.

CI, confidence interval; DF, dengue fever; FUE, fever of unknown etiology; MD, mean difference; aOR, adjusted odds ratio.

GRADE Working Group grades of evidence.High certainty: we are very confident that the true effect lies close to that of the estimate of the effect.Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

Magnitude of effect: Large effect with a narrow 95% CI. Therefore, we upgraded the evidence by one level.

Inconsistency: High heterogeneity between studies was detected in the meta-analysis. Therefore, we downgraded the evidence by one level.

Imprecision: The 95% CI crosses the line of no effect. Therefore, we downgraded the evidence by one level.

Imprecision: Wide 95% CI. Therefore, we downgraded the evidence by one level.

Imprecision: Fewer than 300 events leads to uncertainty in the effect estimate. Therefore, we downgraded the evidence by one level.

Imprecision: Confidence interval not reported, likely wide. Therefore, we downgraded the evidence by one level.

4 Discussion

4.1 Summary

We found that NLR is generally higher in dengue patients than in healthy cohorts, but not different compared to other febrile illnesses, so that it does not have standalone diagnostic value. In contrast, NLR was lower in dengue than in COVID-19 and lower than in malaria, but higher than in scrub typhus, denoting a variable neutrophilic pattern across comparator infections. No difference was found between mild and severe forms according to severity, although some studies found a prognostic role in relation to hemoconcentration and to mortality. Overall, the evidence found is limited and heterogeneous.

4.2 NLR as a diagnostic factor

In our meta-analysis, the NLR was higher in dengue than in healthy subjects, but it did not reliably discriminate against other undifferentiated fevers; however, it was lower in dengue than in COVID-19 and malaria, and higher than in scrub typhus. This suggests that NLR has limited usefulness as a single diagnostic marker but may be useful in differentiating dengue infections from infections that present with marked neutrophilia (such as COVID-19 or malaria). These results are consistent with the previous review by Thach et al. (37) that evaluated the potential of various markers to differentiate severe from non-severe dengue. Although this review did not find that inflammatory markers such as neutrophils or lymphocytes had any statistically significant utility, it did indicate significance in platelet count [standardized mean difference (SMD) −0.65; 95% CI: −0.97 to −0.32] and aspartate aminotransferase level (SMD 0.87; 95% CI: 0.36 to 1.38) in discriminating severe from non-severe cases. On the other hand, Fernández et al. (38) reaffirm the idea that combined hematological parameters (rather than a single ratio) tend to offer better discrimination when developing a multivariate model to differentiate dengue from other acute fevers that integrate clinical and hematological findings (such as leukocytes) and report an AUC of 0.65 for distinguishing dengue in their cohort, suggesting that diagnostic utility lies more in combined panels than in NLR alone. Physiopathologically, this is explained by the fact that dengue usually presents with leukopenia and reactive lymphocytic changes in the early stages (39), while infections such as COVID-19 (40) or malaria (41) cause sustained lymphopenia and marked neutrophilia, raising the NLR. Therefore, this marker would work better as a complementary (e.g., NS1, IgM/IgG, or CRP) marker to identify infections with neutrophilic predominance and as a dynamic indicator of inflammation, but its value as a single diagnostic test for dengue in settings with multiple febrile etiologies is limited.

4.3 NLR as a severity factor

In our meta-analysis, dengue severity showed a mixed pattern with NLR: comparisons between patients with and without warning signs, and between severe and non-severe forms, were not significant, with wide confidence intervals and high heterogeneity. However, the comparison between severe dengue and dengue with warning signs showed a small but statistically significant decrease in NLR with low heterogeneity, suggesting this specific contrast may carry some prognostic signal, while the broader severity comparisons remain inconclusive. This lack of significance is reflected in some studies in the literature: for example, Böer et al. (42) analyzed a cohort and found that although a monocyte-to-lymphocyte ratio > 0.13 was associated with the severity of dengue hemorrhagic fever, the NLR did not show a strong association on its own. On the other hand, Sadgir et al. (43) reported that patients with NLR < 2 had a higher probability of more severe disease, suggesting that a decrease in NLR could correspond to a particular immune response in severe cases. In addition, another study indicated that elevated NLR values on days 4–6 of the illness were associated with severe dengue, reinforcing the possibility that NLR dynamics throughout the course of the disease have prognostic value (44). Physiopathologically, these findings are supported by the fact that dengue is associated with relative neutropenia and reactive lymphocytosis in the critical phase (45), decreasing the NLR in patients who will progress to capillary leak syndrome, while in the late stages, systemic inflammation and metabolic stress can induce secondary neutrophilia (46), raising the NLR again in those who develop shock or multiple organ failure. Thus, both our results and external evidence suggest that NLR has prognostic value, but its interpretation must be closely linked to the clinical stage of the disease, which explains the variability and heterogeneity observed between studies.

4.4 Recommendations for future research

The marked heterogeneity presented in almost all of our outcomes is consistent with a previous review that also highlighted notable heterogeneity in cut-off values, the timing of NLR measurement, and the reference standards used in various studies to evaluate the usefulness of NLR in relation to dengue (47). In general, this heterogeneity in our results may be due to various causes, such as inconsistent case classifications and low statistical power in some studies, as reported in another review (37). Other reasons could be attributed to differences in populations (adults vs. pediatrics) and in the phase of the disease at which the measurement was taken. Certainty of evidence ranged from moderate to very low across diagnostic and prognostic outcomes, driven mainly by inconsistency and imprecision in the effect measures. All of this would hinder the direct extrapolation of NLR values to all clinical contexts and emphasize that our results should be interpreted with caution. These limitations underscore the need for future studies to standardize NLR cut-off points, define the timing of measurement in relation to the clinical phase of dengue, and use prospective designs with adequate samples to properly adjust for confounding factors such as co-infections, baseline haematological values and comorbidities. It would also be desirable for future studies to evaluate how the diagnostic or prognostic performance of NLR changes when combined with other emerging biomarkers (e.g., platelet count, hemoconcentration indices, or immunological markers) in multivariate models. Given the extreme heterogeneity observed in most pooled analyses (I² 87%–99%), pooled point estimates for these outcomes should be interpreted as indicative rather than definitive and should not serve as a basis for clinical decision-making in isolation.

4.5 Clinical applicability

From a clinical and public health perspective, our findings suggest that NLR could be integrated as an auxiliary tool within diagnostic and triage algorithms in resource-limited settings where access to confirmatory tests such as NS1, IgM/IgG, or CRP is insufficient. This is consistent with the guidelines of the Integrated Surveillance Strategy, which promotes the use of locally available basic tests (such as routine blood counts) for the initial assessment of febrile illnesses in settings with limited diagnostic resources (48), as well as the international guidelines of the Integrated Management of Childhood Illness (49) and the WHO essential diagnostics list (1), which recommend the use of basic hematological tests in the initial assessment of febrile illnesses in settings with a high endemic burden. Similarly, the APSED III strategy highlights the operational value of accessible hematological indices to strengthen syndromic surveillance in Asia-Pacific (50). Although the low or very low certainty of several of our outcomes limits the use of NLR as an isolated decision-making tool, its low cost, immediate availability, and relative usefulness in certain differential diagnoses support its potential incorporation into health policies aimed at optimizing initial triage and improving early detection in contexts where confirmatory tests are not widely available.

4.6 Limitations and strengths

This systematic review has some limitations. First, most studies were cross-sectional and retrospective, which may limit the ability to establish a clear causality between the effect of NLR and outcomes; furthermore, these results could also be subject to unmeasured confounding factors. Second, several sources of clinical and methodological heterogeneity likely explain the extremely high I² values observed across most comparisons (87%–99%): NLR cut-off values varied substantially between studies with no consensus threshold; the timing of blood sampling relative to symptom onset was inconsistently reported and likely differed within and between studies, which is particularly relevant given that NLR fluctuates across the febrile, critical, and convalescent phases of dengue; comparator definitions differed (for example, “healthy controls” vs. “other febrile illnesses” were not always clearly distinguished); and assay methods for complete blood counts were not standardized across laboratories and countries. Because the same biomarker behaves differently depending on disease phase, admission timing could not be reliably distinguished from later sampling in the severity analyses, which limits the interpretability of these comparisons. Third, additional subgroup analyses by population age (adult vs. paediatric), timing of blood sampling, and diagnostic criteria, as well as meta-regression, were considered but not performed for outcomes beyond DF vs. COVID-19, since most comparisons were informed by a single study (DF vs. healthy controls, malaria, and scrub typhus) or by three studies at most (severity comparisons), well below the minimum number of studies per covariate (typically ≥10) recommended by the Cochrane Handbook to avoid spurious or underpowered findings.

However, our study also has strengths. First, various sensitivity and certainty assessment analyses were performed to ensure both an adequate assessment of heterogeneity and a complete interpretation of our findings. In addition, several comparisons of the usefulness of NLR were performed to evaluate the use of this marker in various scenarios.

5 Conclusions

Our review shows that, owing to extreme heterogeneity between studies, variability in cut-off points, and certainty of evidence ranging from moderate to very low, the NLR should be regarded as an adjunctive tool at best rather than a stand-alone diagnostic or prognostic marker for dengue. Given its low cost and ready availability, however, it could have a complementary role in integrated clinical assessment algorithms, particularly in resource-poor settings, if the information it provides can be integrated with other clinical and laboratory parameters. The lack of standardization in the measurement phase, analytical methods and definition of severity also highlights the need for higher quality prospective studies to establish validated cut-off points and make a more accurate assessment of the true operational value of NLR in different epidemiological scenarios.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fepid.2026.1877646/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Shailendra Saxena, King George’s Medical University, India

Reviewed by: Amitabh Singh, Vardhman Mahavir Medical College & Safdarjung Hospital, India

Diksha Mohapatra, All India Institute of Medical Sciences Bhubaneswar, India