Association of neutrophil-to-high-density lipoprotein ratio with all-cause mortality among individuals with non-alcoholic fatty liver disease: the NHANES database prospective cohort study

Abstract

Background

Non-alcoholic fatty liver disease (NAFLD) is a global public health crisis linked to elevated all-cause and cardiovascular mortality. The neutrophil-to-high-density lipoprotein ratio (NHR) is an emerging inflammatory-lipid biomarker that integrates systemic inflammation and lipid dysregulation. However, its prognostic value for long-term mortality in NAFLD remains undetermined.

Methods

We conducted a population-based prospective cohort study using data from the National Health and Nutrition Examination Survey (NHANES) 1999–2018. A total of 5,320 participants diagnosed with NAFLD were enrolled and divided into four groups according to quartiles of NHR (Q1-Q4). Multivariate Cox regression, restricted cubic spline, threshold effect, and subgroup analyses were used to evaluate the association between NHR and all-cause mortality.

Results

During a mean follow-up of 9.2 years, 1,172 all-cause deaths and 392 cardiovascular disease (CVD) deaths were documented. In the fully adjusted Cox models, higher NHR was significantly associated with elevated risks of all-cause mortality (HR = 1.27, 95% CI: 1.02–1.57) and CVD mortality (HR = 2.15, 95% CI: 1.44–3.22) in a dose-response manner, but not with cancer mortality. Further adjustment for blood lipids and hepatic function markers did not materially alter the results, supporting the robustness of the observed associations. Threshold and spline analyses further revealed a significant nonlinear relationship between NHR and all-cause mortality, with an inflection point of 2.072, when NHR was equal to or above 2.072, each 1-unit increase was associated with an 8% elevation in all-cause mortality risk (HR = 1.08, 95% CI: 1.04–1.12, P < 0.001). Subgroup stratified analyses were performed as exploratory analyses. Although the association between elevated NHR and all-cause mortality reached statistical significance within several individual subgroups, no significant multiplicative interaction was detected for any stratifying factor, indicating no meaningful effect modification across these population strata.

Conclusion

Elevated NHR is independently associated with an increased risk of all-cause and cardiovascular mortality in NAFLD patients. NHR may serve as a simple, inexpensive, and readily available prognostic biomarker for risk stratification and clinical management in this population.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12876-026-05136-2.

Introduction

Non-alcoholic fatty liver disease (NAFLD) has become the most prevalent chronic liver disease worldwide, with a global prevalence of approximately 30% [1]. NAFLD encompasses a spectrum of conditions ranging from simple steatosis to non-alcoholic steatohepatitis (NASH), fibrosis, cirrhosis, and even hepatocellular carcinoma [2]. Beyond hepatic complications, NAFLD is recognized as a multisystem disorder strongly associated with cardiovascular disease, type 2 diabetes mellitus (T2DM), chronic kidney disease, and metabolic syndrome [3–5]. Accumulating evidence indicates that NAFLD significantly increases the risk of all-cause and cardiovascular mortality, highlighting the urgent need for reliable prognostic markers to identify high-risk individuals [5, 6]. Chronic low-grade systemic inflammation and disordered lipid metabolism are core pathological drivers of NAFLD progression and extrahepatic complications [7, 8]. Neutrophils, the most abundant circulating leukocytes, play a pivotal role in initiating and amplifying inflammatory responses. Their overactivation promotes hepatocyte injury, insulin resistance, and metabolic disturbance [9, 10]. Persistent neutrophil activation contributes to chronic low-grade inflammation, a key driver that links hepatic damage to extrahepatic complications and elevated mortality risk in NAFLD. High-density lipoprotein (HDL) exerts anti-inflammatory, antioxidant, and endothelial protective effects, and plays a critical role in maintaining lipid homeostasis and mitigating systemic inflammation [11, 12]. Impaired HDL function and decreased HDL levels are frequently observed in patients with NAFLD, and are closely associated with insulin resistance, hepatic steatosis, and disease progression [13]. Reduced HDL levels not only reflect disturbed lipid metabolism but also indicate weakened anti-inflammatory capacity, thereby contributing to the pathogenesis and adverse outcomes of NAFLD.

The neutrophil-to-high-density lipoprotein ratio (NHR) integrates inflammatory and lipid profiles, serving as a novel and easily accessible biomarker for evaluating systemic inflammatory status and metabolic risk [14–16]. Growing clinical evidence has confirmed that an elevated NHR is associated with poor prognosis in a variety of metabolic and cardiovascular disorders (CVD), including metabolic syndrome, T2DM, and atherosclerosis [17–19]. Nevertheless, the independent prognostic significance of NHR in patients with non-alcoholic fatty liver disease remains unclear. In particular, large-scale population-based evidence regarding the association between NHR and long-term all-cause and cause-specific mortality in NAFLD is still lacking.

Therefore, this large-scale prospective cohort study based on the publicly available data from the National Health and Nutrition Examination Survey (NHANES) database aimed to investigate the association between baseline NHR and all-cause, CVD, and cancer mortality in individuals with NAFLD. We further explored the non-linear relationship and threshold effect of NHR on mortality, as well as the stability of the association in different subgroups. The findings may provide a novel biomarker for prognostic risk stratification and clinical decision-making in NAFLD patients.

Methods

Study design and population

This prospective cohort study used NHANES 1999–2018 conducted by the Centers for Disease Control and Prevention (CDC). The NHANES employs a stratified, multistage probability sampling design to represent the non-institutionalized civilian population of the United States. Sample weights were applied to account for the complex sampling framework. All primary Cox regression models incorporated NHANES multi-cycle survey weights, strata and primary sampling units to account for complex sampling design. Unweighted Cox models were additionally performed as analyses to compare effect estimates with weighted main results.

We enrolled adult participants aged 20 years and older with NAFLD who underwent mobile examinations from NHANES 1999–2018, which provided comprehensive data for blood and physical measurements. NAFLD was defined by a US fatty liver index (USFLI) of 30 or higher, a well-established definition with an area under the receiver operating characteristic curve (AUROC) of 0.80 (95% CI: 0.77–0.83) in predicting ultrasound-con-firmed NAFLD [20]. The following groups of patients were excluded from the study: (1) excessive alcohol consumption (> 3 drinks per day for the male and > 2 drinks per day for the female); (2) participants with hepatitis B (positive hepatitis B surface antigen) or hepatitis C infection (positive hepatitis C antibody or HCV RNA). Moreover, we excluded participants with unknown exposure to NHR data and those without death follow-up data. As a result, a total of 5,320 participants were finally included in the analysis. The patient selection process is illustrated in Fig. 1.

Fig. 1.

Flowchart depicting the selection of participants

Ascertainment of NHR and outcome definitions

Neutrophil count and high-density lipoprotein cholesterol (HDL-C) concentration were measured using standardized laboratory methods. NHR was calculated as absolute neutrophil count (10⁹/L) divided by HDL-C concentration (mmol/L). Participants were categorized into four groups according to quartiles of NHR (Q1: lowest, Q4: highest; Q1: ≤2.72, Q2: 2.72–3.66, Q3: 3.66–4.88, Q4: > 4.88). We linked the National Death Index (NDI) from the National Center for Health Statistics (NCHS) to obtain the survival status of the participants. Moreover, we utilized the International Statistical Classification of Diseases, 10th Revision (ICD-10) to identify disease-specific deaths.

Assessment of covariates

Data were collected in cycles from 1999 to 2018, including information on demographic characteristics, lifestyle factors, medical history or comorbidities, laboratory measurements. (1) Demographic characteristics included age (less than 45, 45 to less than 65, and 65 or greater), gender, race (Mexican American, non-Hispanic white, non-Hispanic black, other), educational attainment (less than high school, high school, and greater than high school), body mass index (BMI) was used as the anthropometric indicator and classified into four groups based on standard clinical cutoffs: less than 18.5 kg/m², 18.5–24.9 kg/m², 25.0–29.9 kg/m², and 30 kg/m² or greater, and family poverty ratio (less than 1.3, 1.3 to less than 3.5, and 3.5 or greater). (2) Lifestyle factors included marital status (married or with partner, single (widowed or divorced or separated), never married), smoking status (never, former, now), alcohol consumption (none, light-moderate), and physical activity (low, high). (3) Medical history or comorbidities were defined by self-reported physician diagnosis. Medical history or comorbidities were defined by self-reported physician diagnosis. Definitions for each condition were specified as follows. Hypertension was diagnosed if participants had a self-reported history of hypertension, systolic blood pressure (SBP) ≥ 140 mmHg, or diastolic blood pressure (DBP) ≥ 90 mmHg; any single criterion was sufficient to confirm hypertension. Diabetes was defined based on any of the following criteria: self-reported history of diabetes, insulin treatment, oral anti-diabetic medication use, glycated hemoglobin (HbA1c) ≥ 6.5%, fasting blood glucose ≥ 126 mg/dL, or 2-hour postprandial blood glucose ≥ 200 mg/dL. Stroke was identified from an affirmative response to the survey question “Has a doctor or other health professional ever told you that you had a stroke?”. Cancer was defined by a positive answer to the question “Have you ever been told by a doctor or other health professional that you had cancer or a malignancy of any kind?”. CVD was assigned to participants who answered “yes” to any of the physician-diagnosed conditions including congestive heart failure, coronary heart disease, angina pectoris, myocardial infarction, or stroke. Chronic kidney disease (CKD) was defined as an estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m² or a urine albumin-to-creatinine ratio (UACR) ≥ 30 mg/g. (4) Laboratory measurements such as aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-glutamyl transferase (GGT), triglycerides (TG), total cholesterol (TC), low-density lipoprotein (LDL)-cholesterol, fasting blood glucose (FBG) and HbA1c, platelet (PLT) and albumin (ALB) were collected.

Statistical analysis

All statistical analyses were performed using SAS 9.4 and R software. To ensure data completeness and reliability, multiple imputation was performed for covariate data with missing values less than 20%. Specifically, predictive mean matching (PMM) was used for continuous variables, and logistic regression imputation was applied for categorical variables. For outliers in continuous data, values in normally distributed variables exceeding Mean ± 3 SD were truncated at Mean ± 3 SD. For skewed variables, values above the 99.9th percentile or below the 0.1th percentile were replaced with the 99.9th and 0.1th percentiles, respectively. Categorical and dichotomous data were presented as percentages, while continuous variables were reported as mean and standard error (SE). Between-group differences in clinical characteristics were evaluated using chi-square tests for categorical variables and weighted linear regression models for continuous variables. Cox proportional hazards regression models were constructed to estimate hazard ratios (HR) and 95% confidence intervals (CIs) for the association between NHR quartiles and mortality. Restricted cubic spline analysis was used to explore the relationship between NHR and all-cause mortality in greater detail. Threshold effect analysis was performed using two-piecewise linear regression to identify inflection points. Subgroup analyses were conducted applied across numerous stratified tests. Multiplicative interaction P-values were calculated to formally evaluate effect modification between NHR and each stratified variable. Moreover, for CVD-specific and cancer-specific mortality, we adopted conventional cause-specific Cox proportional hazards regression. This approach does not formally account for competing death events from other causes, which may introduce mild bias to cause-specific hazard estimates. Two adjustment frameworks were applied to distinguish the total and direct effects of NHR on all-cause mortality. Model 2 was minimally adjusted for demographic and lifestyle confounders to estimate the total association, while Model 3 further incorporated hypertension, diabetes, stroke, CKD, CVD, and cancer to calculate the direct independent effect of NHR beyond these metabolic intermediates. To quantify the mediating contributions of the three comorbidities, separate mediation analyses were performed for hypertension, diabetes and CVD, with only Model 2 covariates adjusted to avoid truncating causal pathways. A two-sided P < 0.05 was considered statistically significant.

Results

Baseline characteristics

The baseline characteristics of 5,320 NAFLD participants stratified by NHR quartiles are shown in Table 1. The mean age of participants gradually decreased from Q1 to Q4 (57.75 ± 0.51 vs. 51.03 ± 0.61 years, P < 0.001). The proportion of males, current smokers, and participants with BMI ≥ 30 kg/m², diabetes, and CVD increased significantly with elevated NHR (all P < 0.05). There were no significant differences in alcohol consumption, physical activity, hypertension, stroke, and cancer among the four groups (all P > 0.05). As presented in Table 2, triglycerides and platelet count increased with rising NHR (all P < 0.001), while total cholesterol, LDL-cholesterol, and albumin decreased gradually (all P < 0.001). No significant differences were observed in ALT, AST, and GGT levels among groups (all P > 0.05).

Table 1.

Weighted baseline characteristics of NAFLD participants from the 1999–2018 NHANES stratified by NHR

	NHR
	Q1	Q2	Q3	Q4	P value
Number of participants	1403	1376	1274	1267
Age (years)	57.75 ± 0.51	54.93 ± 0.55	54.26 ± 0.53	51.03 ± 0.61	< 0.001
Gender (%)					< 0.001
Male	46.51	55.33	60.39	63.51
Female	53.49	44.67	39.61	36.49
Race (%)					< 0.001
Mexican American	9.82	11.24	11.11	8.18
Non-Hispanic white	68.93	70.05	75.01	74.82
Non-Hispanic black	10.45	6.32	4.13	4.31
Other	10.79	12.38	9.76	12.69
Body Mass Index (kg/m²)					< 0.001
< 18.5	0.08	0.03	0.06	0.00
18.5- <25	6.18	5.39	4.38	2.85
25 - <30	31.40	24.86	24.82	21.38
≥ 30	62.34	69.72	70.75	75.77
Educational attainment (%)					0.038
< High school	21.22	21.61	21.66	23.43
High school	21.91	24.04	24.39	28.00
> High school	56.87	54.35	53.95	48.57
Family poverty ratio (%)					0.002
< 1.3	17.91	22.14	22.32	26.63
1.3 - <3.5	37.89	37.37	36.36	37.76
≥ 3.5	44.20	40.49	41.32	35.61
Physical activity (%)					0.508
Low	43.33	41.03	44.85	42.19
High	56.67	58.97	55.15	57.81
Marital status (%)					0.004
Married or with partner	70.95	69.73	69.94	67.12
Single	21.82	20.77	18.38	19.98
Never married	7.23	9.51	11.68	12.90
Alcohol status (%)					0.857
No	41.12	40.39	41.98	39.72
Light-moderate	58.88	59.61	58.02	60.28
Smoke (%)
Never	59.76	58.20	51.76	45.57	< 0.001
Former	33.34	33.88	34.11	29.64
Now	6.90	7.92	14.13	24.80
Stroke (%)	5.16	3.66	5.01	5.01	0.405
Hypertension (%)	58.09	57.67	59.57	59.85	0.763
Diabetes (%)	26.69	32.57	32.57	42.35	< 0.001
Cardiovascular disease (%)	14.89	15.51	17.63	20.46	0.017
Chronic kidney disease (%)	18.40	21.15	20.90	25.49	0.006
Cancer (%)	14.74	11.46	13.83	12.23	0.181

Other: other than non-Hispanic white, non-Hispanic black, and Hispanic, including multiracial

Marital status: Single-widowed or divorced or separated

Table 2.

Baseline levels of metabolic biomarkers in participants with NAFLD stratified by the NHR

	NHR
	Q1	Q2	Q3	Q4	P value
ALT, U/L	29.73 ± 0.61	31.04 ± 0.77	30.25 ± 0.57	31.05 ± 0.68	0.130
AST, U/L	27.42 ± 0.40	26.76 ± 0.44	26.72 ± 0.45	25.87 ± 0.45	0.130
GGT, U/L	42.75 ± 1.74	37.03 ± 1.33	36.55 ± 1.05	38.23 ± 1.40	0.080
Triglycerides, mg/dL	140.26 ± 2.85	162.37 ± 4.30	188.56 ± 5.39	220.66 ± 6.68	< 0.001
Cholesterol, mg/dL	203.36 ± 1.57	199.26 ± 1.62	193.05 ± 1.81	187.56 ± 1.58	< 0.001
LDL-cholesterol, mg/dL	119.98 ± 1.43	119.64 ± 1.42	113.90 ± 1.45	107.83 ± 1.30	< 0.001
Fasting blood glucose, mg/dL	116.51 ± 1.22	119.99 ± 1.25	118.83 ± 1.33	130.98 ± 1.79	< 0.001
Glycohemoglobin, %	5.89 ± 0.03	6.00 ± 0.04	5.99 ± 0.05	6.34 ± 0.05	< 0.001
Albumin, g/dL	4.22 ± 0.01	4.23 ± 0.01	4.21 ± 0.01	4.17 ± 0.01	< 0.001
Platelet, 1000cell/uL	239.04 ± 2.08	246.32 ± 2.09	252.44 ± 3.06	264.16 ± 2.61	< 0.001

Mean ± SE for continuous variables: P-value was calculated by weighted linear regression model

Association between NHR and mortality

During a mean follow-up period of 9.2 years, a total of 1,172 all-cause deaths, 392 CVD deaths, and 261 cancer deaths were documented. Table 3 presents the HRs and 95% CIs for all-cause, CVD, and cancer mortality across NHR quartiles, derived from three sequentially adjusted Cox regression models. For all-cause mortality, compared with Q1 in the unadjusted Model 1, Q4 was associated with a 38% higher risk (HR = 1.38, 95% CI: 1.12–1.71, P = 0.002), with a significant positive trend (P for trend = 0.009). After adjustment for sociodemographic and lifestyle factors (Model 2), Q3 and Q4 remained significantly associated with elevated risk (Q4: HR = 1.54, 95% CI: 1.26–1.88, P < 0.001; P for trend < 0.001). In the fully adjusted Model 3 (additionally adjusted for comorbidities), Q4 still had a 27% higher risk of all-cause mortality than Q1 (HR = 1.27, 95% CI:1.02–1.57, P = 0.029). A significant positive dose-response trend was observed across NHR quartiles (P for trend = 0.026), indicating that mortality risk increased incrementally with rising NHR levels. Given the relatively small number of CVD and cancer deaths, age was categorized into groups in these analyses to ensure the stability and convergence of Cox models. For CVD mortality, elevated NHR was strongly associated with an increased risk across all models. In Model 1, Q2, Q3, and Q4 all exhibited significantly higher risks relative to Q1, with Q4 showing a 132% elevated risk (HR = 2.32, 95% CI: 1.57–3.42, P < 0.001). The association remained robust in Model 2 and the fully adjusted Model 3, in which Q4 had a 115% higher risk (HR = 2.15, 95% CI: 1.44–3.22, P < 0.001) with a significant positive trend (P for trend = 0.010). For cancer mortality, no significant association was observed between NHR quartiles and cancer mortality in any of the three models (all P > 0.05; P for trend = 0.855 in Model 3). Further adjustment for blood lipids and hepatic function markers did not materially alter the results (Table 4). In the fully adjusted model with blood lipids and hepatic function, the HR for all-cause mortality in Q4 was 1.46 (95%CI: 1.12–1.90, P = 0.005). Similarly, further adjustment for blood lipids and hepatic function markers also did not materially change the association between NHR and CVD mortality (Supplementary Table 1). To eliminate confounding by alcohol consumption and validate the robustness of our core findings, we conducted a sensitivity analysis restricted exclusively to alcohol-free participants with NAFLD (Supplementary Table 2). For all-cause mortality, the highest NHR quartile (Q4) showed no significant association in the crude Model 1 (HR = 1.14, 95%CI: 0.77–1.70, P = 0.507). After adjustment for demographic, lifestyle and anthropometric covariates, participants in Q4 exhibited a significantly higher all-cause mortality risk relative to Q1 (HR = 1.60, 95%CI: 1.08–2.37, P = 0.018), with a significant linear trend across NHR quartiles (P for trend = 0.018). Further comprehensive adjustment for cardiometabolic comorbidities including hypertension, diabetes, stroke, CKD and CVD attenuated this positive association to non-significance (HR = 1.25, 95%CI: 0.84–1.87, P = 0.262). When evaluating cause-specific outcomes, the predictive value of elevated NHR for CVD mortality remained robust across all three adjusted models among non-drinkers. In the fully adjusted Model 3, individuals in NHR Q4 had a 3.31-fold higher risk of CVD death compared with Q1 (HR = 3.31, 95%CI: 1.43–7.65, P = 0.005), accompanied by a significant positive trend (P for trend = 0.030). By contrast, no statistically meaningful linear association was observed between NHR quartiles and cancer mortality in any adjustment model among alcohol-free NAFLD adults.

Table 3.

Hazard ratios (95% CIs) for all-cause and cause-specific mortality among NAFLD participants stratified by NHR

	NHR				P for trend
	Q1	Q2	Q3	Q4
All-cause mortality
Model 1 HR (95% CI) P	Reference	1.26 (1.02–1.55) 0.032	1.17 (0.94–1.46) 0.156	1.38 (1.12–1.71) 0.002	0.009
Model 2 HR (95% CI) P	Reference	1.15 (0.93–1.42) 0.192	1.24 (1.02–1.50) 0.028	1.54 (1.26–1.88) <0.001	< 0.001
Model 3 HR (95% CI) P	Reference	1.07 (0.86–1.32) 0.560	1.14 (0.93–1.38) 0.202	1.27 (1.02–1.57) 0.029	0.026
CVD mortality
Model 1 HR (95% CI) P	Reference	2.48 (1.66–3.70) <0.001	1.68 (1.13–2.48) 0.010	2.32 (1.57–3.42) <0.001	0.002
Model 2 HR (95% CI) P	Reference	2.36 (1.58–3.51) <0.001	1.88 (1.29–2.73) 0.001	2.78 (1.82–4.25) <0.001	< 0.001
Model 3 HR (95% CI) P	Reference	2.16 (1.45–3.21) <0.001	1.65 (1.11–2.45) 0.014	2.15 (1.44–3.22) <0.001	0.010
Cancer mortality
Model 1 HR (95% CI) P	Reference	0.88 (0.56–1.36) 0.556	1.05 (0.67–1.64) 0.824	0.98 (0.63–1.53) 0.941	0.873
Model 2 HR (95% CI) P	Reference	0.78 (0.50–1.22) 0.273	1.07 (0.68–1.70) 0.769	0.99 (0.62–1.56) 0.951	0.717
Model 3 HR (95% CI) P	Reference	0.77 (0.49–1.21) 0.253	1.03 (0.65–1.64) 0.893	0.95 (0.59–1.51) 0.816	0.855

Model 1: Non-adjusted model

Model 2: Adjusted for age, gender, race, BMI, educational attainment, family poverty ratio, physical activity, marital status, smoking status, and alcohol status

Model 3: Adjusted for age, gender, race, BMI, educational attainment, family poverty ratio, physical activity, marital status, smoking status, alcohol status, diabetes, hypertension, stroke, chronic kidney disease, cancer, and cardiovascular disease

Table 4.

Hazard ratios (95% CIs) of all-cause mortality according to NHR among NAFLD with further adjustment of blood lipids and hepatic function

	NHR				P for trend
	Q1	Q2	Q3	Q4
All-cause mortality
Model 2 + blood lipids HR (95% CI) P	Reference	1.18 (0.93–1.49) 0.174	1.28 (1.02–1.61) 0.035	1.59 (1.23–2.05) <0.001	< 0.001
Model 2 + hepatic function HR (95% CI) P	Reference	1.25 (1.02–1.53) 0.032	1.31 (1.08–1.58) 0.006	1.66 (1.36–2.01) <0.001	< 0.001
Model 2 + blood lipids+ hepatic function HR (95% CI) P	Reference	1.29 (1.03–1.63) 0.026	1.37 (1.09–1.73) 0.007	1.76 (1.36–2.28) <0.001	< 0.001
Model 3 + blood lipids HR (95% CI) P	Reference	1.09 (0.86–1.39) 0.466	1.18 (0.93–1.48) 0.171	1.33 (1.02–1.73) 0.035	0.027
Model 3 + hepatic function HR (95% CI) P	Reference	1.12 (0.91–1.38) 0.300	1.18 (0.97–1.43) 0.105	1.34 (1.09–1.65) 0.005	0.006
Model 3 + blood lipids+ hepatic function HR (95% CI) P	Reference	1.16 (0.92–1.47) 0.210	1.25 (0.99–1.58) 0.067	1.46 (1.12–1.90) 0.005	0.005

Model 2: Adjusted for age, gender, race, BMI, educational attainment, family poverty ratio, physical activity, marital status, smoking status, and alcohol status

Model 3: Adjusted for age, gender, race, BMI, educational attainment, family poverty ratio, physical activity, marital status, smoking status, alcohol status, diabetes, hypertension, stroke, chronic kidney disease, cancer, and cardiovascular disease

We performed independent mediation analyses to quantify the mediating roles of hypertension, diabetes and CVD in the NHR-all-cause mortality association (Supplementary Tables 3–5). Hypertension (mediating proportion = 4.48%, P = 0.040) and diabetes (mediating proportion = 15.76%, P < 0.001) were confirmed as significant partial mediators. CVD had no statistically significant mediating effect (P = 0.080). Collectively, mediation decomposition analyses demonstrated that hypertension, diabetes, and CVD partially mediated the total association between NHR and all-cause mortality. Adjustment for these intermediate conditions in fully adjusted Cox models therefore attenuates the total NHR effect and yields the direct prognostic effect of NHR independent of established cardiometabolic diseases. We built two model sets: minimally adjusted models without mediator variables to obtain the total NHR effect and prevent over-adjustment bias, and fully adjusted models adjusted for hypertension, diabetes and CVD to calculate the direct effect. We conducted unweighted Cox regression sensitivity analyses for all-cause, CVD and cancer mortality (Supplementary Table 6). The risk trends derived from unweighted models were largely consistent with weighted results in Table 3. Elevated NHR was significantly associated with higher CVD mortality, whereas no significant correlation between NHR and cancer mortality was observed regardless of weighting. In both weighted and unweighted models, hazard ratios of NHR mildly decreased after adjustment for hypertension, diabetes and CVD, consistent with partial mediation of these metabolic comorbidities. Minor discrepancies in hazard magnitudes between the two approaches are attributed to the unequal sampling scheme of NHANES.

Non-linear and threshold effect analyses

To further explore the nonlinear relationship between NHR and all-cause mortality among patients with NAFLD, we performed restricted cubic spline (RCS) analysis with 5 knots and threshold effect analysis using the segmented package. For improved model stability and convergence, cancer was excluded as a covariate and age was treated as a grouped variable in Model 3. RCS analysis demonstrated a significant positive nonlinear association between NHR and all-cause mortality (Fig. 2) (P for overall < 0.001; P for nonlinear = 0.041). The vertical dashed line in Fig. 2 represents the 2nd/3rd quartile cutoff of NHR (3.66). Meanwhile, threshold effect analysis (Table 5) identified a significant inflection point at an NHR of 2.072 for all-cause mortality (P for likelihood ratio test = 0.015). Overall, NHR was positively associated with all-cause mortality (HR = 1.06, 95% CI: 1.03–1.10). Specifically, when NHR was below 2.072, an inverse association was observed (HR = 0.53, 95% CI: 0.31–0.88, P = 0.015), whereas when NHR was equal to or above 2.072, each 1-unit increase was associated with an 8% elevation in mortality risk (HR = 1.08, 95% CI: 1.04–1.12, P < 0.001).

Fig. 2.

Restricted cubic spline curve of the association between NHR and all-cause mortality risk. The red solid line represents the hazard ratio (HR) of all-cause mortality, and the pink shaded area represents the 95% confidence interval (CI). The horizontal dashed line represents HR = 1 (no effect). The vertical dashed line represents the 2nd/3rd quartile cutoff of NHR (3.66)

Table 5.

Threshold effect analysis of NHR on all-cause mortality in NAFLD

Population	Cut-off value	Adjusted HR (95% CI), P	Log likelihood ratio
All-cause mortality	NHR < 2.072	0.53 (0.31–0.88) 0.015	0.015
	NHR ≥ 2.072	1.08 (1.04–1.12) <0.001
Male	NHR < 2.113	0.31 (0.16–0.60) <0.001	< 0.001
	NHR ≥ 2.113	1.07 (1.02–1.11) 0.006
Light-moderate Alcohol	NHR < 2.161	0.56 (0.31–1.03) 0.063	0.026
	NHR ≥ 2.161	1.08 (1.04–1.13) <0.001
With Hypertension	NHR < 2.049	0.45 (0.25–0.82) 0.009	0.006
	NHR ≥ 2.049	1.08 (1.04–1.12) <0.001
With Diabetes	NHR < 2.092	0.33 (0.12–0.87) 0.025	0.009
	NHR ≥ 2.092	1.07 (1.02–1.12) 0.004

Adjusted for age, gender, race, BMI, educational attainment, leisure-time physical activity, marital status, family poverty ratio, smoking status, alcohol consumption, diabetes, hypertension, stroke, chronic kidney diseases, and cardiovascular disease

Subgroup analyses revealed consistent threshold effects across populations. In male participants, the inflection point was 2.113. Above the threshold, each 1-unit increase was associated with a 7% elevated risk (HR = 1.07, 95% CI: 1.02–1.11, P = 0.006). In patients with light-to-moderate alcohol consumption, the inflection point was 2.161. No significant association was observed below the threshold (HR = 0.56, 95% CI: 0.31–1.03, P = 0.063), but above the threshold, each 1-unit increase in NHR was associated with an 8% higher risk of mortality (HR = 1.08, 95% CI: 1.04–1.13, P < 0.001). Similar threshold effects were observed in hypertensive and diabetic patients. Above their respective inflection points, higher NHR was linearly associated with an increased risk of all-cause mortality. The P for likelihood ratio test were 0.006 and 0.009 for the two groups, respectively.

Subgroup analyses

Subgroup analyses of adjusted model were conducted to determine the adverse effects of increased NHR on NAFLD patient survival, stratified by age, gender, race, smoke, alcohol consumption, physical activity, marital status, diabetes, hypertension, stroke, CKD, and CVD (Fig. 3). Our subgroup analyses revealed that the positive association between NHR and all-cause mortality was more pronounced in several specific populations. Specifically, female patients, older adults (≥ 65 years), current smokers, individuals with light-to-moderate alcohol consumption and low physical activity, Mexican Americans, Non-Hispanic whites, and patients with hypertension and CKD experienced a significantly increased risk of mortality with elevated NHR (all P < 0.05). Nevertheless, we found no statistically significant multiplicative interaction terms for all stratification variables, which suggests that the prognostic relationship of NHR with mortality did not differ substantially across distinct population subgroups.

Fig. 3.

Forest plot of exploratory subgroup analyses for the association between NHR and all-cause mortality. Multiplicative interaction P-values were adopted to assess effect modification, rather than relying solely on within-subgroup statistical significance. Multiple comparison correction was not performed for these stratified analyses

Discussion

Accumulating evidence has established correlations between NHR and NAFLD prevalence or hepatic fibrosis in cross-sectional cohorts [21, 22], and several population-based studies have reported that elevated NHR predicts cardiovascular events among general adults or patients with type 2 diabetes [17, 18, 23, 24]. However, few prospective investigations have systematically clarified the long-term prognostic performance of NHR specifically in NAFLD populations. The present study fills this critical research gap by utilizing a nationally representative US cohort to comprehensively assess linear, nonlinear, and threshold relationships between NHR and three distinct mortality endpoints. Furthermore, we validated the stability of these associations through multiple sensitivity analyses adjusted for liver function and lipid indicators, and conducted extensive exploratory subgroup analyses across demographic, lifestyle, and comorbidity strata. These comprehensive analyses collectively constitute the primary incremental contributions of the current work compared with prior literature.

NAFLD is characterized by chronic low-grade inflammation and disrupted lipid metabolism, two core pathological drivers that synergistically promote hepatic steatosis, hepatocyte damage, fibrosis progression, and life-threatening extrahepatic complications [20, 25]. Accumulating preclinical evidence indicates that neutrophils actively infiltrate the hepatic parenchyma in response to metabolic stress and free fatty acid overload. Within the liver, neutrophils release a cascade of pro-inflammatory cytokines including interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), and monocyte chemoattractant protein-1 (MCP-1), which further trigger hepatocyte biotoxicity, oxidative stress, mitochondrial dysfunction, and systemic insulin resistance [26–28]. In contrast, HDL exerts multiple protective effects by suppressing inflammatory cell recruitment, neutralizing oxidative stress, improving cholesterol efflux, and maintaining endothelial integrity. A marked reduction in HDL level or impairment in HDL function disrupts these protective pathways, aggravates hepatic and systemic inflammation, and accelerates metabolic deterioration and cardiovascular complications in NAFLD [29–31].

Beyond local hepatic inflammation driven by infiltrated neutrophils, the liver–spleen axis acts as a vital amplifier of systemic metabolic inflammation in NAFLD. In steatotic liver tissue, activated neutrophils secrete abundant IL-6, TNF-α and MCP-1, triggering a series of pathological events including hepatotoxic injury, excessive oxidative stress and mitochondrial dysfunction. Instead of being restricted within the liver, these pro-inflammatory cytokines travel via the portal venous system to the spleen, where they activate splenic macrophages and skew the splenic immune compartment toward a pro-inflammatory state. Subsequently, the spleen releases pre-activated monocytes and extra inflammatory mediators back to the liver, further exacerbating intrahepatic inflammation and accelerating the development of systemic insulin resistance, a core pathological feature of progressive NAFLD [32]. Such a reciprocal inflammatory loop between the liver and spleen converts isolated hepatic steatosis into persistent systemic immunometabolic disorder, which partially accounts for the strong prognostic performance of NHR integrating inflammatory burden and HDL protective capacity in our NAFLD cohort.

NHR is a composite biomarker integrating neutrophil-driven inflammation and HDL-mediated anti-inflammatory and cholesterol efflux functions. NHR captures the imbalance between excessive inflammatory activity and impaired lipid protection, a core pathological feature of metabolic syndrome and its hepatic and cardiovascular complications. Lu et al. reported linear and nonlinear associations between NHR and NAFLD prevalence and fibrosis [33]. Avcı and Ozturk further corroborated these findings by showing that NHR was significantly elevated in NAFLD patients compared with controls, positioning it as a clinically accessible inflammation marker in the context of hepatic metabolic dysfunction [34]. Importantly, Liang et al. extended this line of inquiry by demonstrating that a combined diagnostic model incorporating NHR alongside the systemic inflammation response index (SIRI) and ALT substantially improved NAFLD diagnostic accuracy, with NHR exhibiting a specificity of 85.82% among individual markers [16]. Beyond liver outcomes, NHR predicts cardiovascular events and mortality, the leading cause of death in NAFLD. Chuang et al. reported that elevated NHR independently predicted incident cardiovascular events and all-cause mortality in a large prospective cohort, highlighting its capacity to capture the atherosclerosis–inflammation interface that underpins CVD pathogenesis [17]. Further evidence demonstrated that NHR predicted adverse cardiovascular outcomes even among normoglycemic individuals with elevated fasting glucose, suggesting that the biomarker identifies subclinical inflammatory-metabolic risk states that precede overt metabolic disease [35]. Meanwhile, Mo et al. extended the metabolic relevance of NHR by demonstrating its strong association with insulin resistance in obese populations across both Chinese and US cohorts, reinforcing the notion that NHR captures a systemic immuno-metabolic derangement that bridges hepatic steatosis, metabolic deterioration, and cardiovascular risk [14]. Our results showed that higher NHR was associated with increased all-cause mortality. In fully adjusted models, the highest NHR quartile had a 27% higher risk of all-cause mortality and a 115% higher risk of CVD mortality than the lowest quartile. This suggests that NHR can effectively identify NAFLD patients at high risk of premature death, especially CVD-related death. Notably, NHR was not associated with cancer mortality, indicating its prognostic specificity for metabolic and cardiovascular-related deaths in NAFLD. The attenuated all-cause mortality association in fully adjusted models of the alcohol-free sensitivity analysis may be attributed to mediating effects of cardiometabolic disorders, which represent intermediate outcomes linking elevated NHR to long-term mortality risk. In addition, the reduced sample size after excluding all drinking participants lowered statistical power to detect modest associations after extensive covariate adjustment.

The nonlinear association between NHR and all-cause mortality carries important translational implications. A rise in NHR above the critical threshold corresponds to a gradual and discernible increase in mortality risk, thereby simplifying its bedside interpretation compared with biomarkers that exhibit U-shaped or J-shaped dose–response curves. This nonlinear pattern contrasts with other lipid-derived indices in NAFLD populations. For instance, the non-high-density lipoprotein cholesterol to high-density lipoprotein cholesterol ratio (NHHR) has been shown to exhibit a U-shaped relationship with all-cause mortality [36], while the stress hyperglycemia ratio (SHR) likewise displays a U-shaped association in which both extremely low and high values confer elevated risk [37]. The absence of such paradoxical low-value risk in NHR enhances the directness and clinical interpretability of this biomarker. The threshold effect analysis identified an NHR inflection point of 2.072, above which the mortality risk increased sharply and significantly. Notably, similar threshold effects were consistently observed across clinically relevant subgroups. In male patients, the risk cutoff was 2.113; in those with light-moderate alcohol consumption, 2.161; in hypertensive patients, 2.049; and in diabetic patients, 2.092. In all these subgroups, NHR values above the corresponding cutoff were independently associated with elevated all-cause mortality, whereas values below the cutoff conferred a lower mortality risk. Importantly, NHR is calculable from a standard complete neutrophil count and HDL-C, both of which are universally available in primary care and hepatology clinics-a pragmatic advantage over more complex composite scores. This accessibility aligns with the broader trend in NAFLD research toward identifying cost-effective, readily obtainable biomarkers. From a mechanistic perspective, the prognostic power of NHR likely stems from its dual capture of two convergent pathophysiological axes in NAFLD progression. As a ratio, NHR integrates neutrophil-mediated immune activation and hepatic inflammation with the anti-inflammatory, antioxidant functions of HDL-C. Lower HDL-C levels in NAFLD diminish these protective pathways, amplifying unchecked inflammation. Although the identified NHR inflection point of 2.072 revealed a clear nonlinear risk pattern in our cohort, this numerical threshold cannot be directly generalized to clinical practice currently. Multi-center external validation studies covering different ethnic groups and patient populations are essential to verify its clinical applicability before it can be adopted as a formal risk stratification tool.

Nevertheless, several limitations should be acknowledged when interpreting these findings. First, NAFLD was defined by USFLI instead of liver ultrasound or biopsy. Although it has favorable diagnostic performance and is widely used in relevant cohorts, the USFLI may introduce misclassification bias and dilute the true association between NHR and mortality. Moreover, it cannot distinguish simple steatosis from NASH or advanced fibrosis, so we were unable to perform stratified analyses by liver disease severity, resulting in unmeasured residual confounding. Second, although the nonlinear association and threshold effect are statistically robust, the identified NHR cutoff of 2.072 was derived from a specific study population and may require external validation across diverse ethnicities, healthcare settings, and NAFLD severity spectra before it can be recommended as a universal clinical decision threshold. Third, NHR, as a ratio of two routinely measured parameters, is inherently subject to biological and analytical variability-neutrophil counts can fluctuate with acute infections, medications (e.g., corticosteroids), and circadian rhythms, while HDL-C levels are influenced by diet, exercise, alcohol intake, and pharmacotherapy-all of which could introduce measurement noise and affect the reproducibility of the cutoff in individual patients. Fourth, a large number of exploratory subgroup comparisons were conducted without adjustment for multiple testing, which may increase the probability of false-positive findings. We relied on interaction tests to judge inter-group differences to mitigate this limitation. Fifth, although NHR may reflect inflammation and lipid dysregulation, we did not comprehensively evaluate endogenous antioxidant biomarkers such as bilirubin, or uric acid. Bilirubin, uric acid, and albumin are all biologically relevant endogenous antioxidant markers that may be associated with oxidative stress and mortality risk [38]. Therefore, the extent to which NHR captures oxidative stress-related pathways in NAFLD remains uncertain. Finally, due to unavailable core indicators for USFLI calculation, which may introduce selection bias and restrict the generalizability of our findings. The core conclusions of this study are mainly applicable to non-institutionalized U.S. adult NAFLD population with complete metabolic and liver function-related core indicators. Future prospective, intervention-oriented studies are warranted to determine whether NHR-guided risk stratification can meaningfully improve clinical decision-making and long-term outcomes in the management of NAFLD and its cardiometabolic complications.

Conclusion

Elevated NHR is independently associated with an increased risk of all-cause and cardiovascular mortality in NAFLD patients. NHR is a simple, low-cost, and widely available biomarker that can facilitate risk stratification and improve clinical management in this high-risk population. Further prospective studies are warranted to validate these findings and explore the underlying mechanisms.

Authors' contributions

MQ conceived and designed the study. YX completed the review of literature and wrote the first draft of the manuscript. JW and HC conducted data acquisition and performed the statistical analysis. MX and FD interpreted and revised the results. MQ reviews and grants final approval. All authors reviewed the manuscript.

Funding

This study was supported by the Jiangxi Provincial Natural Science Foundation (20252BAC200454) and the Clinical Training Program of the First Affiliated Hospital of Nanchang University (YFYLCYJPY202442).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent participate

NHANES was conducted with approval by the National Center for Health Statistics Ethics Review Board, and obtained informed written consent from all the individuals involved in the study.

Competing interests

The authors declare no competing interests.