Association of Dynamic Changes in the Neutrophil-to-Lymphocyte Ratio with Mortality in Patients on Maintenance Hemodialysis: A Retrospective Cohort Study

Introduction

Chronic kidney disease (CKD) represents one of the most significant public health challenges of the twenty-first century. In 2023, an estimated 788 million adults worldwide were living with CKD, a figure more than double the 378 million recorded in 1990, with a global age-standardised prevalence of 14.2%.¹ Projections indicate that CKD-related mortality will increase by over 30% from 2022, and CKD is expected to become the third leading cause of death in Western Europe by 2050.² A substantial proportion of these patients ultimately progress to end-stage kidney disease (ESKD), requiring maintenance hemodialysis (HD). The median prevalence of treated kidney failure currently stands at 823 cases per million population, with marked disparities between high- and low-income countries.³ Despite advances in dialysis technology and supportive care, mortality in patients on maintenance HD remains strikingly elevated. Cardiovascular disease and infection account for the majority of deaths in this population, underscoring the urgent need for reliable and clinically accessible risk stratification tools.⁴

Chronic inflammation occupies a central role in the pathophysiology of ESKD. Persistent immune activation drives endothelial dysfunction, accelerated atherosclerosis, and protein-energy wasting — a constellation of interrelated processes collectively described as the malnutrition–inflammation–atherosclerosis (MIA) syndrome.^5,6 While C-reactive protein (CRP) remains the most widely used biochemical marker of inflammation, its clinical applicability is limited by cost, availability in resource-constrained settings, and susceptibility to transient fluctuations unrelated to the underlying inflammatory burden.⁷

The neutrophil-to-lymphocyte ratio (NLR), derived from a routine complete blood count, has emerged as an inexpensive, widely available, and reproducible index of systemic inflammation and immune dysregulation.⁸ Elevated NLR has demonstrated prognostic value across diverse clinical contexts, including cardiovascular disease, solid malignancies, and chronic inflammatory conditions.^9,10 In the CKD and HD populations specifically, higher baseline NLR values have been independently associated with greater inflammatory burden, malnutrition, cardiovascular risk, and all-cause mortality.^11–13

However, a fundamental limitation shared by the majority of prior investigations is their reliance on a single baseline NLR measurement, implicitly treating the inflammatory state as static. Inflammation in patients on maintenance HD is, by nature, dynamic: it fluctuates in response to recurrent infections, changes in dialysis adequacy, vascular access complications, and evolving comorbidities.¹⁴ Consequently, a single cross-sectional NLR value may inadequately capture the true inflammatory trajectory of an individual patient. Longitudinal assessment of NLR change over time — herein referred to as ΔNLR — may therefore offer superior prognostic resolution compared with any single time-point measurement. To date, evidence examining the prognostic significance of ΔNLR in the HD population remains sparse.

The present retrospective cohort study was designed to address this knowledge gap. We investigated the association between change in NLR over a two-year follow-up period and all-cause mortality in patients on maintenance HD. The primary novelty of this study lies in its longitudinal design: rather than characterising inflammatory status at a single point in time, we evaluated the prognostic value of dynamic NLR trajectories. We hypothesised that patients exhibiting an increasing NLR during follow-up would carry a significantly higher risk of death compared with those demonstrating stable or declining values. If validated, ΔNLR may provide clinicians with a practical, cost-free, and universally available tool for ongoing risk stratification in routine HD practice.

Materials and Methods

Study Design and Population

This retrospective single-center cohort study included adult patients (≥18 years) on maintenance HD who had been receiving treatment for at least three months at baseline. Patients were identified through the electronic medical records of the dialysis center.

Patients with available complete blood count data at baseline (January 2024) and at 12 months were considered for inclusion in the analytical cohort. Patients who died before completing the 12-month follow-up period were excluded from the final analysis due to the absence of follow-up NLR measurements, which precluded the calculation of ΔNLR. A 12-month interval was selected for two reasons: first, comprehensive laboratory evaluations including complete blood count are routinely performed at baseline and at 12 months as part of standard clinical care at our centre; and second, this interval is consistent with prior longitudinal studies examining inflammatory biomarker trajectories in maintenance hemodialysis populations, thereby facilitating comparison with existing literature. Patients with missing laboratory data or active malignancy at baseline were also excluded.

Data Collection

Demographic characteristics, clinical data, dialysis vintage, and laboratory parameters were obtained retrospectively from hospital electronic records. Dialysis vintage was defined as the duration from initiation of maintenance HD to the baseline assessment in January 2024.

Laboratory data were collected at two predefined time points: baseline (January 2024) and 12 months after baseline.

Definition of ΔNLR

NLR was calculated by dividing the absolute neutrophil count by the absolute lymphocyte count obtained from routine complete blood count measurements. ΔNLR was defined as the difference between the NLR value measured at 12 months and the baseline NLR value. A positive ΔNLR indicated an increase in inflammatory burden over time, whereas a negative ΔNLR indicated a decrease. Patients were categorized according to ΔNLR direction for subgroup analyses.

Outcome and Follow-Up

The primary outcome of the study was all-cause mortality. Baseline assessment was performed in January 2024. Follow-up was conducted for 24 months from baseline. Survival time was defined as the interval between baseline assessment and death or completion of the 24-month follow-up period, whichever occurred first.

Ethics Approval and Consent to Participate

This study was approved by the Bartın University Clinical Research Ethics Committee (Approval No: 2026-SBB-0123). The study was conducted in accordance with the principles of the Declaration of Helsinki. Due to the retrospective design of the study and the use of anonymized clinical data, the requirement for informed consent was waived by the ethics committee.

Statistical Analysis

Statistical analyses were performed using the Jamovi program (version 2.6.19.0). Continuous variables were expressed as mean ± standard deviation and compared using the Student’s t-test. Categorical variables were compared using the Chi-square test. Comparisons between survivors and non-survivors were performed for both continuous and categorical variables. Changes in inflammatory parameters between baseline and 12 months were analyzed using paired comparisons. Survival analysis was performed using the Kaplan–Meier method, and differences between groups were evaluated using the Log rank test. Cox proportional hazards regression analysis was used to assess the association between ΔNLR and mortality. Due to the limited number of events, multivariable models were restricted to a small number of clinically relevant covariates to minimize overfitting. Although cardiovascular disease and vascular access type (catheter) were significantly associated with mortality in univariable analyses, they were not included in the multivariable model due to the limited number of outcome events (n=12), which precluded the inclusion of additional covariates without substantially increasing the risk of model overfitting. The proportional hazards assumption was evaluated using Schoenfeld residuals, and no significant violations were observed.

Results

Baseline Characteristics

A total of 64 patients were initially screened. All patients were treated at a single center in Turkey and were of Turkish ethnicity. After exclusion of five patients who died before the 12-month follow-up and had no available follow-up NLR measurements, 59 patients were included in the final analysis (Figure 1). During the subsequent follow-up period, 12 patients (20.3%) died and 47 (79.7%) survived. Baseline demographic and clinical characteristics are summarized in Table 1. The mean age of the study population was 58.12 ± 17.04 years. Female patients comprised 54.7% of the cohort. Diabetes mellitus was present in 61.0% and cardiovascular disease in 47.5% of patients. Arteriovenous fistula (AVF) was the predominant vascular access type (71.2%).

Table 1 Baseline Demographic and Clinical Characteristics (N=59)

Figure 1 Flowchart of patient selection and study population.

Patients who died were significantly older than survivors (67.69 ± 16.94 vs 55.47 ± 16.26 years, p = 0.007). Dialysis vintage was significantly shorter in the mortality group (50.67 ± 28.79 vs 97.21 ± 62.82 months, p = 0.004).

Serum albumin levels were significantly lower among non-survivors (3.46 ± 0.38 vs 3.97 ± 0.37 g/dL, p < 0.001). Hemoglobin levels were also significantly lower in patients who died (9.84 ± 0.94 vs 10.8 ± 1.37 g/dL, p = 0.017).

Baseline neutrophil count, lymphocyte count, baseline NLR, and CRP levels did not differ significantly between groups (p > 0.05 for all). However, at 12 months, lymphocyte counts were significantly lower (p = 0.005) and NLR values significantly higher (p = 0.006) in the mortality group.

ΔNLR was significantly higher in patients who died compared with survivors (2.01 ± 3.29 vs –0.42 ± 1.89, p = 0.021).

Categorical variables associated with mortality are shown in Table 2. Cardiovascular disease (p = 0.014) and vascular access type (p = 0.001) were significantly associated with mortality, whereas ΔNLR categorized as >0 vs ≤0 was not statistically significant (p = 0.602).

Table 2 Association Between Categorical Variables and Mortality

Longitudinal Changes in Inflammatory Parameters

Changes in inflammatory markers between baseline and 12 months are shown in Table 3.

Table 3 Changes in Inflammatory Parameters Between Baseline and 12 Months

No significant differences were observed in neutrophil counts (p = 0.100), lymphocyte counts (p = 0.183), or NLR values (p = 0.842) at the population level over the 12-month period.

Kaplan–Meier Survival Analysis

Kaplan–Meier survival analysis was performed to evaluate the association between ΔNLR categories and all-cause mortality. Patients were stratified into two groups according to ΔNLR (ΔNLR ≤ 0 vs ΔNLR > 0) (Figure 2).

Figure 2 Kaplan–Meier survival curves according to ΔNLR category in maintenance hemodialysis patients.

The mean survival time was 23.61 months in the ΔNLR ≤ 0 group and 22.00 months in the ΔNLR > 0 group.

Cox Proportional Hazards Regression Analysis

Cox regression analyses are summarized in Table 4.

Table 4 Cox Proportional Hazards Regression Analysis for All-Cause Mortality

In univariable analysis, each 1-unit increase in ΔNLR was associated with a 78% higher mortality risk (HR 1.78, 95% CI 1.31–2.41, p < 0.001).

In the fully adjusted multivariable model including age and serum albumin, ΔNLR remained independently associated with mortality (HR 1.58, 95% CI 1.16–2.16, p = 0.004). Serum albumin was identified as an independent protective factor (HR 0.10, 95% CI 0.02–0.50, p = 0.004), whereas age was not significantly associated with mortality (p = 0.340). The model demonstrated good discriminative ability (C-index = 0.861).

Discussion

Principal Findings and Contextualisation

In this retrospective cohort study of patients on maintenance HD, we demonstrated that ΔNLR — the change in NLR measured between baseline and a twelve-month follow-up assessment — was significantly associated with all-cause mortality when analysed as a continuous variable. Notably, baseline NLR at the time of enrolment was not independently associated with mortality, whereas its longitudinal trajectory was. This dissociation between cross-sectional and dynamic NLR measurements constitutes the central and novel finding of the present study, and directly addresses a critical gap in the existing literature: the predominant reliance on single time-point inflammatory assessment in a population where inflammation is inherently unstable and fluctuating.

NLR as a Marker of Systemic Inflammation in HD

Chronic low-grade inflammation is a hallmark of ESKD and contributes to adverse outcomes including cardiovascular events, protein-energy wasting, and immune dysfunction.¹⁵ The NLR, derived from a routine complete blood count, reflects the balance between innate immune activation (neutrophilia) and adaptive immune suppression (lymphocyte depletion) — both well-established features of the uremic milieu.¹⁶ Several studies have established that higher baseline NLR predicts mortality in dialysis populations. A large meta-analysis demonstrated that elevated NLR was significantly associated with both all-cause and cardiovascular mortality in CKD patients, including those on dialysis.¹⁷ Ouellet et al independently confirmed that higher NLR predicted all-cause mortality in chronic HD patients,¹⁸ and Li et al further showed that elevated NLR correlated with surrogate cardiovascular risk markers including pulse pressure, left ventricular mass index, and carotid intima-media thickness.¹⁹ Across different dialysis modalities, NLR has similarly demonstrated prognostic relevance, reinforcing the systemic nature of uremia-related immune dysregulation.²⁰

Why Dynamic NLR May Outperform Baseline Assessment

Despite these consistent findings, virtually all prior studies assessed NLR at a single time point. Most existing studies on inflammatory markers in HD patients are cross-sectional in nature, with a relative scarcity of clinical investigations evaluating longitudinal changes after dialysis initiation and their dynamic evolution before death.²¹ Inflammation in patients on maintenance HD is not static: it is driven by recurrent infectious episodes, fluctuations in dialysis adequacy, vascular access complications, and the cumulative burden of comorbidities.¹⁴ A single NLR value may therefore fail to capture the true inflammatory trajectory of an individual patient.

Our findings align with and extend this reasoning. The fact that baseline NLR was not predictive of mortality while ΔNLR was significant supports the hypothesis that it is the direction and magnitude of inflammatory change — rather than its absolute level at any given moment — that carries the most clinically meaningful prognostic information. This concept is further supported by longitudinal data from the HD literature: a study using linear mixed-effects models found that for each 1.0-unit increase in NLR over time, the fully adjusted all-cause mortality hazard ratio was 1.04 (95% CI 1.01–1.07, p = 0.006), confirming that longitudinal changes in NLR independently predict all-cause mortality risk in maintenance HD patients.²² The parallel with longitudinal nutritional indices is also instructive: dynamic increases in prognostic nutritional index and geriatric nutritional risk index scores over time corresponded to significant reductions in both all-cause and cardiovascular mortality risk in HD patients,²³ suggesting that serial reassessment of inflammatory and nutritional biomarkers may be broadly superior to single-point evaluation across multiple prognostic domains.

From a pathophysiological standpoint, a rising ΔNLR likely reflects progressive immune dysregulation characterised by sustained neutrophil-driven innate immune activation alongside relative lymphocyte depletion — a pattern consistent with cumulative uremic immune stress rather than transient inflammatory episodes. This distinction is clinically important: it suggests that monitoring the trend of NLR over time may allow identification of patients who appear stable at baseline but are undergoing ongoing inflammatory deterioration.

Positioning the Unique Contribution of This Study

The present study differs from existing NLR literature in two important respects. First, unlike studies that use NLR at a single post-initiation time point, we explicitly modelled the change in NLR over a defined interval and examined its independent prognostic value. Second, by demonstrating that ΔNLR retains prognostic significance as a continuous variable while categorical Kaplan–Meier analysis did not reach significance, this study highlights a methodological point with direct clinical implications: dichotomisation of continuous biomarkers reduces statistical power and conceals risk gradients, particularly in smaller cohorts.²⁴ Risk stratification using continuous ΔNLR trajectories may therefore offer more granular and clinically actionable prognostic information than threshold-based categorisation.

The finding that dialysis vintage — but not age — was associated with mortality in univariable analyses is also noteworthy. This may reflect the cumulative impact of dialysis-related morbidities (vascular access complications, recurrent hospitalisations, chronic fluid overload, and progressive cardiovascular remodelling) on long-term outcomes, independently of chronological ageing. It is also noteworthy that age was significantly associated with mortality in univariable analysis (p=0.007) but lost independent significance in the multivariable model (p=0.340). This attenuation likely reflects confounding between age and dialysis vintage, as older patients in this cohort tended to have shorter dialysis vintage — possibly due to later referral or more rapid disease progression — such that the prognostic signal attributed to age in univariable analysis was partially captured by dialysis vintage in the adjusted model. This observation merits further investigation in larger prospective studies.

Clinical Translation and Real-World Applicability

A key strength of ΔNLR as a prognostic tool is its near-zero marginal cost. The complete blood count is obtained at every HD session in routine clinical practice, and NLR can be calculated without additional laboratory tests, equipment, or specialised expertise. Unlike CRP or interleukin-6, NLR requires no additional financial outlay and is universally available even in resource-limited dialysis settings. The inexpensive nature and ready availability of NLR are recognised as practical strengths for everyday clinical practice in nephrology.²⁵

Translating ΔNLR into routine clinical workflow is therefore feasible without regulatory or cost barriers. Several practical implementation pathways warrant consideration. First, serial NLR measurements could be incorporated into standard six- or twelve-monthly laboratory reviews already mandated in most HD programmes, with ΔNLR calculated automatically from existing electronic health record data. Second, a consistent upward NLR trajectory over two consecutive review periods could serve as a flag for intensified clinical review — prompting evaluation of infection sources, dialysis adequacy, vascular access integrity, and nutritional status. Third, in settings where validated risk scores such as the Charlson Comorbidity Index or dialysis-specific mortality models are used, ΔNLR could be integrated as an additional, cost-free inflammatory input to augment predictive accuracy. Incorporating inflammatory biomarkers into traditional risk models substantially improves long-term mortality risk stratification in HD patients, offering a robust, clinically applicable tool to support individualised prognostic assessment and intervention planning.²⁶

Roadmap for Future Research

Although this study provides early evidence for the prognostic relevance of ΔNLR in HD patients, several important research questions remain unanswered and constitute a clear agenda for future investigation.

First, the optimal interval for NLR reassessment has not been established. While we used a twelve-month follow-up measurement, shorter reassessment intervals (three or six months) may capture clinically relevant inflammatory transitions more sensitively, particularly following acute events such as bacteraemia or access thrombosis. Prospective studies incorporating serial NLR measurements at multiple time points would allow the construction of inflammatory trajectory clusters and the identification of high-risk patterns.

Second, the threshold magnitude of ΔNLR that should prompt clinical intervention remains undefined. Future studies should aim to establish evidence-based cut-off values validated across diverse HD populations, stratified by dialysis vintage, vascular access type, and primary aetiology of ESKD.

Third, it remains unknown whether interventions targeting modifiable drivers of NLR increase — such as optimisation of dialysis adequacy, vascular access revision, antibiotic stewardship for recurrent infections, or anti-inflammatory nutritional supplementation — can attenuate ΔNLR and thereby improve outcomes. Randomised or quasi-experimental studies testing whether ΔNLR-guided clinical pathways reduce mortality would be of high clinical value.

Fourth, future studies should explore the additive predictive value of ΔNLR when combined with other dynamic biomarkers, including longitudinal albumin, C-reactive protein, or the geriatric nutritional risk index, within composite prognostic models.

Limitations

This study has several limitations. The retrospective single-centre design limits causal inference and generalisability. Patients who died before the twelve-month reassessment were excluded, introducing survivor bias that may have attenuated the observed association. The modest number of events restricted multivariable model complexity, precluding adjustment for all clinically relevant covariates. ΔNLR was calculated from only two time points and may not fully capture short-term inflammatory fluctuations. Finally, the absence of concurrent CRP data precluded direct validation against an established inflammatory reference standard.

Conclusions

This study demonstrated that ΔNLR over a twelve-month period was independently associated with all-cause mortality in maintenance HD patients, whereas baseline NLR was not. These findings suggest that the longitudinal trajectory of systemic inflammation carries greater prognostic relevance than a single cross-sectional measurement. Given that ΔNLR is derived entirely from routine complete blood count data at no additional cost, serial NLR monitoring represents a practical and widely accessible tool for risk stratification in routine HD care. Prospective validation in larger cohorts is warranted before clinical implementation.

Future Perspectives

Future studies should prospectively validate ΔNLR in larger, multicentre HD cohorts and define evidence-based thresholds for clinical use. Shorter reassessment intervals — at three or six months — may improve sensitivity for detecting early inflammatory deterioration. Whether ΔNLR-guided monitoring can inform timely clinical intervention, and whether it provides incremental prognostic value beyond established biomarkers such as albumin and CRP, merits dedicated investigation.