Prognostic value of the neutrophil-to-lymphocyte ratio in POEMS syndrome

Abstract

POEMS syndrome is a rare paraneoplastic disorder driven by clonal plasma cell proliferation and systemic inflammation, but existing prognostic models inadequately incorporate inflammation-related parameters. The neutrophil-to-lymphocyte ratio (NLR)—a simple, accessible marker of systemic inflammation and immune status—lacks systematic prognostic validation in POEMS syndrome. This single-center retrospective study enrolled 61 patients. The cohort was stratified into two eras (2010–2017 and 2018–2025) to assess temporal consistency. Receiver operating characteristic analysis was used to derive an exploratory NLR cut-off for overall survival. Kaplan–Meier curves and Cox proportional hazards regression were used to assess associations between NLR and progression-free survival (PFS), overall survival (OS), and independent prognostic significance. The effect modification by Li’s prognostic risk strata was explored using interaction and stratified analyses. Correlations were tested by Spearman’s coefficient between NLR and C-reactive protein (CRP), and AUCs were compared by DeLong’s test. ROC analysis showed moderate discrimination of baseline NLR for OS (AUC 0.736, 95% CI 0.613–0.860; P = 0.006), and the exploratory cut-off was 2.773. Patients with NLR > 2.773 had significantly shorter PFS (median 62 months vs not reached; P < 0.001) and OS (median 76 months vs not reached; P < 0.001) than those with NLR ≤ 2.773. Multivariable Cox regression confirmed NLR > 2.773 as an independent prognostic factor for inferior PFS (HR 3.547, 95% CI 1.086–11.586; P = 0.036). In Li’s high-risk subgroup, high NLR further separated survival curves, although the formal interaction for PFS was not significant (P for interaction = 0.270), and OS interaction modeling was limited by few death events. CRP showed prognostic relevance, NLR correlated moderately with CRP (Spearman’s ρ = 0.49, P < 0.001). A combined NLR + CRP model did not significantly improve OS discrimination over either marker alone (DeLong tests, all P > 0.05). Collectively, elevated baseline NLR is associated with inferior PFS and OS in POEMS syndrome and may reflect an inflammation-related risk phenotype. As a simple and widely available index, NLR may complement existing prognostic frameworks by incorporating an inflammation–immune dimension. External validation in independent multi-center cohorts is warranted.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10238-026-02111-9.

Introduction

POEMS syndrome is a rare paraneoplastic disorder driven by clonal plasma cell proliferation and characterized by demyelinating peripheral neuropathy and multisystem involvement [1]. Increasing evidence suggests that POEMS syndrome is essentially a cytokine-driven systemic inflammatory and vasculopathic syndrome, rather than a purely plasmacytic neoplasm [2, 3]. The pathological core lies in the abnormal expansion of clonal plasma cells and the sustained overproduction of cytokines, such as vascular endothelial growth factor (VEGF) and multiple proinflammatory cytokines [1, 4, 5]. These cytokines, secreted by clonal plasma cells and cells within the tissue microenvironment, promote angiogenesis, capillary leak, and multiorgan damage.

Inflammatory cytokines are closely linked to both the clinical manifestations and prognosis of POEMS syndrome. Previous studies have shown that serum levels of VEGF, IL-6, IL-12, and other cytokines correlate positively with disease activity [2, 3, 6]. More recently, a Mayo Clinic study demonstrated that elevated baseline IL-6 is significantly associated with inferior overall survival (OS), and proposed IL-6 as a serum biomarker with high predictive value for outcomes in POEMS syndrome [3], thereby underscoring the central role of the “inflammation–vascular” axis in disease progression [7, 8].

Currently, clinical risk stratification mainly relies on the prognostic model developed by Li et al. [9], which incorporates age, pleural effusion, pulmonary hypertension, and reduced estimated glomerular filtration rate to derive a risk score. This model effectively distinguishes low-, intermediate-, and high-risk patients, with 10-year OS rates of 98%, 75%, and 50%, respectively. However, it primarily reflects the burden of organ damage and does not capture systemic inflammatory load, potentially underestimating the contribution of inflammation to disease progression and prognosis.

The neutrophil-to-lymphocyte ratio (NLR) is a readily obtainable inflammation–immune marker derived from routine peripheral blood counts, reflecting the dynamic balance between a proinflammatory response (neutrophilia) and immunosuppression or immune exhaustion (lymphopenia). Numerous studies in solid tumors and hematologic malignancies have demonstrated that an elevated NLR is significantly associated with poor prognosis [10–13]. In multiple myeloma and malignant lymphoma, NLR has also been shown to serve as a valuable complement to traditional prognostic models such as the International Staging System (ISS) and the National Comprehensive Cancer Network–International Prognostic Index (NCCN-IPI) [14–16]. Nonetheless, whether NLR has prognostic relevance in POEMS syndrome, and whether it can compensate for the lack of inflammatory burden assessment in existing prognostic models, remains unexplored.

Therefore, we conducted a single-center retrospective study of 61 patients with POEMS syndrome to assess the prognostic association of baseline NLR with survival outcomes and to explore whether NLR may capture clinically relevant variation related to the inflammatory–immune state in this disease.

Methods

Study population and ethics approval

A total of 61 patients with POEMS syndrome who were diagnosed and treated at the First Affiliated Hospital of Nanjing Medical University between April 2010 and July 2025 were included. Clinical data were uniformly collected, collated and adjudicated by the same research team, and all these patients met the international consensus diagnostic criteria proposed by Dispenzieri et al. [4]. To mitigate potential bias arising from the long inclusion period, the cohort was stratified into two diagnostic eras (2010–2017 and 2018–2025). Clinical characteristics, laboratory assays, treatment strategies and outcomes were systematically recorded and compared between these eras. This retrospective cohort study was approved by the Ethics Committee of the First Affiliated Hospital of Nanjing Medical University (No. 2022-SR-448).

Laboratory measurements and NLR calculation

Baseline clinical and laboratory data were collected, including age, sex, initial and concomitant manifestations (peripheral neuropathy, skin changes, organomegaly, extravascular volume overload, endocrine abnormalities, etc.), laboratory parameters (complete blood counts, serum biochemistry, immunofixation electrophoresis, bone marrow morphology and immunophenotyping), and imaging findings. Serum C-reactive protein (CRP) was measured on the Olympus AU 600 analyzer (Olympus Mishima Co. Ltd., Shizuoka, Japan) using an automated high-sensitivity latex-enhanced immunoturbidimetric assay (Olympus reagents). Serum VEGF was measured using a human Quantikine ELISA Kit (R&D Systems, Minneapolis, MN, USA). All assays were performed following standardized laboratory operating procedures throughout the study period to minimize technical variability.

The NLR was calculated as the absolute neutrophil count (× 10⁹/L) divided by the absolute lymphocyte count (× 10⁹/L), both obtained from the first complete blood count at the time of diagnosis. According to the prognostic scoring model established in 2017 by Li et al. [9], patients were scored based on the presence of age ≥ 50 years, pulmonary hypertension, pleural effusion, and an estimated glomerular filtration rate (eGFR) < 30 mL·min⁻¹·1.73 m⁻², and were subsequently classified into a non-high-risk group (score < 2) and a high-risk group (score ≥ 2).

Response assessment and follow-up

Treatment response was evaluated according to the POEMS syndrome response criteria proposed by Dispenzieri et al. [4], integrating hematologic response, organ response, and neurological improvement. Progression-free survival (PFS) was defined as the time from diagnosis to disease progression, relapse, or death from any cause. Overall survival (OS) was defined as the time from diagnosis to death from any cause or last follow-up. Follow-up was continued until September 2025 through outpatient visits and telephone interviews. The median follow-up duration was 65 months (range, 1–136 months).

Statistical analysis

Statistical analyses were performed using SPSS 27.0.1, R 4.3.1 and GraphPad Prism 8.0, and figures and tables were generated with GraphPad Prism 8.0, R 4.3.1 and Microsoft Excel. A two-sided P value < 0.05 was considered statistically significant.

For continuous variables, normally distributed data were expressed as mean ± standard deviation and compared between groups using the t test, whereas non-normally distributed data were expressed as median (range) and compared using the Mann–Whitney U test. Categorical variables were presented as number (percentage) and compared using the Chi-square or Fisher’s exact test, as appropriate.

Receiver operating characteristic (ROC) curve analysis was used to evaluate the predictive performance of baseline NLR and CRP for OS events, with exploratory cut-off value determined according to the Youden index. Internal validation was performed using bootstrap resampling with 1,000 iterations, estimating the bootstrap distributions and 95% percentile confidence intervals (CIs) of sensitivity and specificity at the fixed cut-off. Univariable and multivariable logistic regression were used to evaluate the association between NLR > 2.773 and hematologic complete response (CR_H). Kaplan–Meier curves were generated for PFS and OS and compared using the log-rank test. Cox proportional hazards regression was used to assess potential prognostic factors for PFS and OS. Variables with P < 0.05 in univariable analyses were entered into a multivariable Cox model. The proportional hazards (PH) assumption was assessed using Schoenfeld residual tests. To explore whether the prognostic effect of NLR differed across risk strata defined by Li’s prognostic model, an interaction term (NLR group × high-risk status) was included in a Cox model, and stratified Cox analyses were performed. Spearman’s rank correlation coefficient (ρ) was used to assess the correlation between NLR and CRP. The predictive performances of NLR alone, CRP alone, and a combined model for OS were compared by calculating the area under the ROC curve (AUC), with pairwise comparisons performed using DeLong’s test.

Results

Baseline characteristics and determination of the NLR cut-off

A total of 61 patients with POEMS syndrome were enrolled. Given the long inclusion period, we evaluated potential temporal heterogeneity by stratifying the cohort into two periods: 2010–2017 (n = 26) and 2018–2025 (n = 35). As detailed in Supplementary Table 1, except for diabetes/impaired glucose tolerance, osteosclerotic bone lesions, treatment regimens, and treatment response, no statistically significant differences were observed between the two eras regarding age, gender, or key clinical features such as NLR level, distribution of Li’s prognostic risk groups and CRP level (all P > 0.05), supporting overall temporal comparability of the cohort.

ROC curve analysis showed that NLR had moderate discriminatory ability for predicting OS in this cohort (AUC 0.736, 95% CI: 0.613–0.860, P = 0.006). The exploratory cut-off value was 2.773 (Fig. 1), yielding a sensitivity of 93.3% and a specificity of 69.6%. Internal validation using 1,000 bootstrap resamples (fixed cut-off 2.773) showed a mean sensitivity of 93.8% (95% CI: 0.800–1.000) and a mean specificity of 69.8% (95% CI: 0.556–0.837; Supplementary Fig. 1). Based on this cut-off, patients were divided into a low-NLR group (NLR ≤ 2.773, n = 33) and a high-NLR group (NLR > 2.773, n = 28).

Fig. 1.

Receiver operating characteristic (ROC) curve of baseline neutrophil-to-lymphocyte ratio (NLR) for discriminating overall survival (OS) events in patients with POEMS syndrome

Baseline clinical and laboratory characteristics were summarized in Table 1. The groups were broadly comparable in terms of sex distribution, peripheral neuropathy, organomegaly, endocrine abnormalities, M-protein positivity, bone marrow plasma cell percentage, or Li’s risk stratification (all P > 0.05). However, patients in the high-NLR group were significantly older (57.86 ± 9.76 vs 51.64 ± 11.16 years, P = 0.025). Skin changes were more frequent in the low-NLR group (78.8% vs 39.3%, P = 0.002), whereas peripheral edema was more common in the high-NLR group (32.1% vs 6.1%, P = 0.008).

Table 1.

Baseline Characteristics of Patients with POEMS Syndrome Stratified by NLR

variable	NLR ≤ 2.773 (n = 33)	NLR > 2.773 (n = 28)	t/Z/X2 value	P value
Age (years), Mean ± SD	51.64 ± 11.16	57.86 ± 9.76	-2.297	0.025
Male, n (%)	20 (60.6)	15 (53.6)	0.307	0.580
Peripheral Neuropathy, n (%)	33 (100.0)	28 (100.0)	—	—
Organomegaly, n (%)
Hepatomegaly	5 (15.2)	5 (17.9)	0.000	1.000
Splenomegaly	16 (48.5)	11 (39.3)	0.520	0.471
Lymphadenopathy	19 (57.6)	14 (50.0)	0.350	0.554
Castleman disease*a	3 (30.0) (n = 10)	5 (50.0) (n = 10)	—	0.650
Skin Changes, n (%)	26 (78.8)	11 (39.3)	9.904	0.002
Extravascular volume overload, n (%)
Peripheral edema	2 (6.1)	9 (32.1)	6.972	0.008
Pleural effusion	7 (21.2)	9 (32.1)	0.935	0.333
Pericardial effusion	9 (27.3)	5 (17.9)	0.759	0.384
Ascites	7 (21.2)	7 (25.0)	0.123	0.726
Papilledema*	3 (13.0) (n = 23)	3 (17.6) (n = 17)	0.000	1.000
Endocrinopathy, n (%)
Thyroid abnormality	17 (51.5)	8 (28.6)	3.297	0.069
Gonadal abnormality	18 (54.5)	11 (39.3)	1.414	0.234
Diabetes/Impaired Glucose Tolerance	12 (36.4)	7 (25.0)	0.912	0.340
Adrenal axis abnormality	9 (27.3)	8 (28.6)	0.013	0.910
Thrombocytosis, n (%)	8 (24.2)	11 (39.3)	1.598	0.206
Osteosclerotic bone lesions, n (%)	14 (42.4)	7 (25.0)	2.037	0.153
Pulmonary hypertensionb*, n (%)	7 (28.0) (n = 25)	4 (20.0) (n = 20)	0.074	0.786
Elevated VEGF*, n (%)	29 (93.5) (n = 31)	25 (92.6) (n = 27)	0.000	1.000
Bone marrow plasma cells (%), median (range)	1.60 (0.00–7.20)	2.00 (0.40–8.40)	-0.018	0.986
M-protein Positive, n (%)	28 (84.8)	25 (89.3)	0.017	0.896
Immunofixation electrophoresis, n (%)			8.136	0.262
IgA-λ	17 (51.5)	9 (32.1)
IgG-λ	7 (21.2)	9 (32.1)
IgA-κ	1 (3.0)	0 (0.0)
IgG-κ	1 (3.0)	2 (7.1)
Light chain λ	0 (0.0)	3 (10.7)
IgA-λ + IgG-λ	1 (3.0)	2 (7.1)
IgG-κ + IgG-λ	1 (3.0)	0 (0.0)
Negative	5 (15.2)	3 (10.7)
Risk Stratification, n (%)			0.983	0.612
Low risk	6 (18.2)	3 (10.7)
Intermediate risk	15 (45.5)	12 (42.9)
High risk	12 (36.4)	13 (46.4)
Treatment Regimens, n (%)			7.967	0.158
Melphalan-Dexamethasone	2 (6.1)	1 (3.6)
Lenalidomide-Dexamethasone	19 (57.6)	10 (35.7)
Cyclophosphamide-Dexamethasone	0 (0.0)	3 (10.7)
Thalidomide-Dexamethasone	2 (6.1)	1 (3.6)
Bortezomib-Dexamethasone	5 (15.2)	8 (28.6)
Others	5 (15.2)	5 (17.9)
ASCT*, n (%)	1 (3.3) (n = 30)	3 (12.0) (n = 25)	0.506	0.477
Treatment Response*, n (%)
CRH	23 (74.2) (n = 31)	11 (47.8) (n = 23)	3.937	0.047
CRV	28 (90.3) (n = 31)	18 (75.0) (n = 24)	1.336	0.248
RN	25 (83.3) (n = 30)	15 (60.0) (n = 25)	3.743	0.053

Abbreviations: NLR: Neutrophil-to-Lymphocyte Ratio; SD: Standard Deviation; VEGF: Vascular Endothelial Growth Factor; ASCT: Autologous Stem Cell Transplantation; CR_H: Hematological Complete Response; CR_V: VEGF Complete Response; RN: Neurological Response

*Percentages were calculated based on the number of patients with available data, excluding missing values

^aProportions were calculated among patients who underwent lymph node biopsy. Note: Values in bold indicate p < 0.05

Clinical outcomes according to NLR level

In terms of treatment response, the CR_H rate was significantly higher in the low-NLR group (74.2% vs 47.8%, P = 0.047; Table 1), while no significant differences were observed between groups in CR_V or RN rates. In univariable logistic regression, NLR > 2.773 was associated with lower odds of achieving CR_H (OR = 0.319, 95% CI 0.101–1.004, P = 0.051). After adjustment for extravascular volume overload, renal insufficiency, high-risk category, and treatment regimen, the association remained in the same direction but did not reach statistical significance (adjusted OR = 0.260, 95% CI 0.060–1.129, P = 0.072), indicating that an independent association with CR_H could not be confirmed in this cohort.

Kaplan–Meier analysis showed that median PFS was not reached in the low-NLR group, whereas it was 62 months in the high-NLR group (HR = 0.221, 95% CI 0.093–0.525, P < 0.001; Fig. 2a). Similarly, median OS was not reached in the low-NLR group but was 76 months in the high-NLR group (HR = 0.165, 95% CI 0.059–0.460, P < 0.001; Fig. 2b), indicating that patients with low NLR had more favorable survival outcomes. Era-stratified analyses supported the consistency of the prognostic association across time periods (Supplementary Fig. 2). In the 2010–2017 subgroup, patients with high NLR (> 2.773) had inferior PFS (median 62 months vs not reached, HR = 0.336, 95% CI 0.119–0.951, P = 0.040) and OS (median 76 months vs not reached, HR = 0.291, 95% CI 0.094–0.909, P = 0.034). Similarly, in the 2018–2025 subgroup, high-NLR group remained associated with shorter PFS (median 53 months vs not reached, HR = 0.042, 95% CI 0.005–0.335, P = 0.003) and OS (P = 0.034), although OS medians were not reached; the HR for OS was not estimable because no deaths occurred in one NLR stratum.

Fig. 2.

Kaplan–Meier survival curves for a progression-free survival (PFS) and b overall survival (OS) according to neutrophil-to-lymphocyte ratio (NLR, ≤ 2.773 vs > 2.773)

Further subgroup analyses based on Li’s prognostic risk categories were presented in Fig. 3. In the non-high-risk subgroup, there were no statistically significant differences between the low- and high-NLR groups in median PFS (not reached vs 95 months, HR = 0.501, 95% CI 0.115–2.177, P = 0.357) or median OS (not reached vs 121 months, P = 0.155); the HR for OS was not estimable because no deaths occurred in one NLR stratum. In contrast, among high-risk patients, NLR provided clear prognostic discrimination: median PFS was not reached in the low-NLR group but was only 47 months in the high-NLR group (HR = 0.207, 95% CI 0.060–0.720, P = 0.013), and median OS was not reached in the low-NLR group versus 62 months in the high-NLR group (P = 0.005); the HR for OS was not estimable due to zero deaths in one stratum.

Fig. 3.

Kaplan–Meier survival curves according to baseline neutrophil-to-lymphocyte ratio (NLR) in subgroups defined by risk stratification: a progression-free survival (PFS) in the non-high-risk subgroup, b overall survival (OS) in the non-high-risk subgroup, c PFS in the high-risk subgroup, and d OS in the high-risk subgroup. Hazard ratios (HRs) were not estimable in panels (b) and (d) because no OS events occurred in one group

To evaluate effect modification by Li’s risk category, we fitted Cox models with an interaction term (NLR > 2.773 × high-risk), and additionally performed stratified Cox analyses to aid interpretation. For PFS, the interaction term was not statistically significant (HR for interaction = 4.40, 95% CI: 0.32–61.16; P for interaction = 0.270), indicating that effect modification could not be confirmed in this cohort. Stratified Cox analyses (Table 2) nevertheless suggested a stronger association of NLR with PFS in the high-risk stratum (HR = 8.77, 95% CI: 1.11–69.55; P = 0.040) than in the non-high-risk stratum (HR = 2.13, 95% CI: 0.41–10.99; P = 0.370), although confidence intervals were wide. Conversely, when stratifying by NLR, high-risk status was associated with inferior PFS within the high-NLR stratum (HR = 3.19, 95% CI: 1.09–9.30; P = 0.034), but not within the low-NLR stratum (HR = 0.78, 95% CI: 0.07–8.65; P = 0.837).

Table 2.

Stratified Cox regression analysis for PFS and OS

stratum type	stratification variable	endpoint	factor	HR (95% CI)	P value
Stratified by NLR	NLR ≤ 2.773	PFS	high-risk	0.777 (0.070–8.646)	0.837
	NLR > 2.773			3.191 (1.095–9.304)	0.034
	NLR ≤ 2.773	OS	high-risk	—	—
	NLR > 2.773			5.514 (1.456–20.885)	0.012
Stratified by risk level	non-high-risk	PFS	NLR > 2.773	2.130 (0.413–10.994)	0.366
	high-risk			8.775 (1.107–69.551)	0.039
	non-high-risk	OS	NLR > 2.773	—	—
	high-risk			—	—

Abbreviations: HR, hazard ratio; CI, confidence interval; PFS, progression-free survival; OS, overall survival; NLR, neutrophil-to-lymphocyte ratio. “—” indicated that the hazard ratio could not be reliably estimated (e.g., sparse/no events or model non-convergence)

For OS, the limited number of death events (n = 13) precluded reliable interaction modeling; therefore, interaction effects for OS were not reported. In stratified analyses, within the NLR > 2.773 subgroup, high-risk was associated with inferior OS (HR = 5.51, 95% CI: 1.46–20.88; P = 0.012), whereas no deaths occurred in the NLR ≤ 2.773 subgroup, precluding stable estimation. The stratified Cox analysis suggested that the prognostic value of NLR was mainly concentrated in the high-risk group. However, the interaction test indicated that this difference was not statistically significant, possibly due to limited sample size and an uneven distribution of events.

Univariate and multivariable analysis of prognostic factors

In univariable Cox regression (Table 3), several variables were associated with PFS. Patients with baseline NLR > 2.773 had a significantly increased risk of progression or relapse (HR = 5.458, 95% CI: 1.833–16.248; P = 0.002). Extravascular volume overload (HR = 6.867, 95% CI: 1.597–29.521; P = 0.010) and renal insufficiency (HR = 3.152, 95% CI: 1.151–8.628; P = 0.025) were also associated with shorter PFS, whereas sex, albumin < 35 g/L, thrombocytosis, osteosclerotic bone lesions, organomegaly, skin changes, and high-risk category were not statistically significant in univariate analyses (all P > 0.05). In the multivariable Cox model for PFS, both NLR > 2.773 (HR = 3.547, 95% CI: 1.086–11.586; P = 0.036) and extravascular volume overload (HR = 4.625, 95% CI: 1.036–20.658; P = 0.045) remained independently associated with inferior PFS.

Table 3.

Univariate and multivariate cox proportional hazards analysis for PFS and OS

prognostic factor	PFS				OS*
	univariate analysis		multivariate analysis		univariate analysis
	P value	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)
Male	0.342	1.554 (0.626, 3.862)			0.906	1.065 (0.378, 2.996)
Extravascular volume overload	0.010	6.867 (1.597, 29.521)	0.045	4.625 (1.036, 20.658)	0.034	8.991 (1.182, 68.412)
NLR > 2.773	0.002	5.458 (1.833, 16.248)	0.036	3.547 (1.086, 11.586)	0.009	14.837 (1.944, 113.269)
Albumin < 35 g/L	0.094	2.367 (0.863, 6.494)			0.593	1.341 (0.457, 3.933)
Renal insufficiency	0.025	3.152 (1.151, 8.628)	0.481	1.486 (0.494, 4.474)	0.264	1.851 (0.629, 5.445)
Thrombocytosis	0.084	0.339 (0.099, 1.155)			0.159	0.342 (0.077, 1.520)
Osteosclerotic bone lesions	0.778	0.852 (0.280, 2.596)			0.584	0.654 (0.142, 3.001)
Organomegaly	0.668	0.781 (0.251, 2.424)			0.583	0.694 (0.189, 2.553)
Skin changes	0.649	0.817 (0.343, 1.948)			0.454	0.677 (0.243, 1.881)
High risk	0.091	2.256 (0.878, 5.798)			0.017	5.006 (1.341, 18.690)

Abbreviations: PFS: Progression-Free Survival; OS: Overall Survival; HR: Hazard Ratio; CI: Confidence Interval; NLR: Neutrophil-to-Lymphocyte Ratio

*Multivariable modeling for OS was not reported due to limited events and model convergence issues; only univariable OS results were shown. Note: Values in bold indicate p < 0.05

For OS, univariate Cox regression showed that NLR > 2.773 (HR = 14.837, 95% CI: 1.944–113.269; P = 0.009), extravascular volume overload (HR = 8.991, 95% CI: 1.182–68.412; P = 0.034), and high-risk category (HR = 5.006, 95% CI: 1.341–18.690; P = 0.017) were associated with increased mortality. Multivariable modeling was attempted; however, due to the small number of death events, convergence warnings consistent with separation (i.e., potentially infinite coefficients) occurred.

PH assumptions were assessed using Schoenfeld residual tests (Supplementary Table 2). No violations were detected in the univariate Cox models for either PFS or OS (all P > 0.05). For the multivariable PFS model, the global Schoenfeld test was not significant (χ² = 1.37, P = 0.712), and covariate-specific tests were also non-significant (extravascular volume overload: P = 0.272; NLR > 2.773: P = 0.667; renal insufficiency: P = 0.910), supporting the validity of the Cox PH assumptions for the reported multivariable PFS model.

Correlation with inflammatory and disease burden markers

We further investigated whether NLR was related to systemic inflammatory status or tumor mass (Fig. 4). No significant correlations were observed between NLR and VEGF levels (657.13 ± 162.49 vs 725.72 ± 311.64 pg/mL, P = 0.559) or bone marrow plasma cells percentage [1.60 (0.00–7.20)% vs 2.00 (0.40–8.40)%, P = 0.986]. In contrast, CRP levels were significantly higher in the high-NLR group [18.15 (1.38–179.00) vs 7.25 (0.50–102.00) mg/L, P = 0.006]. At the individual level, baseline NLR showed a moderate positive monotonic correlation with CRP (Spearman’s ρ = 0.49, P < 0.001), suggesting that higher NLR may be associated with greater systemic inflammatory burden.

Fig. 4.

Comparison of a C-reactive protein (CRP), b bone marrow clonal plasma cell proportion, and c vascular endothelial growth factor (VEGF) levels between patients with low and high neutrophil-to-lymphocyte ratio (NLR, ≤ 2.773 vs > 2.773)

The prognostic performance of CRP was also evaluated. ROC analysis for predicting OS yielded an AUC of 0.765 (95% CI: 0.595–0.936; P = 0.002), with the exploratory cut-off of 33.4 mg/L (sensitivity 50.0%, specificity 97.1%; Supplementary Fig. 3). Kaplan–Meier analyses demonstrated that patients with CRP > 33.4 mg/L had significantly inferior outcomes compared with those with CRP ≤ 33.4 mg/L, including shorter PFS (median 62 vs 129 months, HR = 0.171, 95% CI 0.043–0.681, P = 0.012) and shorter OS (median 62 months vs not reached, HR = 0.073, 95% CI 0.0015–0.348, P = 0.001; Fig. 5). When modeled as a continuous variable in univariable Cox analysis, higher CRP was significantly associated with worse outcomes. For PFS, CRP was also significantly associated with an increased risk of progression (HR = 1.011, 95% CI: 1.001–1.020; P = 0.032). For OS, CRP showed a significant positive association with mortality risk (HR = 1.014, 95% CI: 1.003–1.025; P = 0.012). Schoenfeld residual tests did not suggest violation of proportional hazards for CRP in either PFS (P = 0.46) or OS models (P = 0.78).

Fig. 5.

Kaplan–Meier survival curves for a progression-free survival (PFS) and b overall survival (OS) according to C-reactive protein (CRP, ≤ 33.4 vs > 33.4)

Finally, to explore whether combining NLR with CRP improved prognostic discrimination, we compared ROC curves for OS prediction using NLR alone, CRP alone, and their combination (Fig. 6). The AUCs were 0.765 for CRP, 0.736 for NLR, and 0.757 for the combined model. DeLong tests did not show statistically significant differences (combined model vs CRP: P = 0.543; combined model vs NLR: P = 0.638; NLR vs CRP: P = 0.889), indicating that a statistically significant incremental gain from combining NLR and CRP could not be established in this cohort.

Fig. 6.

Receiver operating characteristic (ROC) curves for overall survival (OS) prediction by neutrophil-to-lymphocyte ratio (NLR) alone, C-reactive protein (CRP) alone, and their combination

Discussion

In this single-center cohort of 61 patients with POEMS syndrome, we systematically evaluated the prognostic impact of baseline NLR. Patients with NLR > 2.773 had markedly shorter PFS and OS than those with lower NLR, and NLR remained independently associated with PFS in multivariable analysis. When stratified by Li’s prognostic model, NLR appeared to further separate outcomes within the high-risk category, whereas survival differences were not evident within the non-high-risk subgroup. But the formal interaction test for PFS was not significant, and OS interaction modeling was not reliably estimable because of the small number of death events. Accordingly, subgroup findings should be viewed as exploratory signals that warrant confirmation in larger cohorts with adequate event numbers and formal interaction testing. Collectively, our findings support that systemic inflammatory burden may contribute to prognostic heterogeneity in POEMS syndrome and NLR, a simple and inexpensive parameter, merits further evaluation as a candidate biomarker.

Previous prognostic studies in POEMS syndrome have largely focused on baseline organ involvement and treatment response. Early work from the Mayo Clinic demonstrated that younger age, higher serum albumin levels, and achievement of hematologic remission were associated with better outcomes, and identified hypoalbuminemia as an important baseline risk factor for disease progression [17, 18]. Li et al.[9] subsequently developed the first clinical prognostic model incorporating age, pulmonary hypertension, pleural effusion, and eGFR, thereby underscoring the prognostic importance of organ damage burden and systemic reserve. However, these models did not include inflammatory markers. More recently, studies by Cook, Tomasso, and colleagues have shown at the cytokine level that IL-6 is a highly predictive biomarker for outcomes in POEMS syndrome and is closely correlated with VEGF levels [3, 19], thereby providing biological support for integrating inflammatory pathways into risk assessment. Our results extend this line of investigation by evaluating a routine blood-derived inflammatory index (NLR), providing an exploratory refinement of current risk stratification systems.

From a mechanistic standpoint, the potential prognostic value of inflammation-related indices is biologically plausible. Evidence from the multiple myeloma literature indicates that crosstalk between clonal plasma cells and the bone marrow microenvironment can foster an immunosuppressive, therapy-tolerant state, thereby supporting clonal persistence and influencing disease progression and treatment responsiveness [20–22]. This conceptual framework may also help contextualize the association between inflammatory burden and clinical outcomes in POEMS syndrome. An elevated NLR reflects an imbalance between proinflammatory activity and antitumor immune function. Neutrophils can amplify inflammation through the release of reactive oxygen species, proteases, and chemokines, leading to vascular endothelial injury and exacerbation of capillary leak [23, 24]. Consistent with this possibility, we observed a significantly higher incidence of peripheral edema in the high-NLR group than in the low-NLR group, and extravascular volume overload was independently associated with shorter PFS in multivariate Cox analysis. Conversely, relative lymphopenia may reflect impaired immune competence and reduced antitumor surveillance, which could plausibly compromise treatment responsiveness [25–27]. In line with this hypothesis, the CR_H rate was higher in the low-NLR group. In logistic regression analyses, NLR > 2.773 was associated with lower odds of achieving CR_H, and the association persisted after adjustment for key confounders but did not reach statistical significance. Thus, while NLR may have potential predictive value for suboptimal hematologic response, it cannot be considered an independent or reliable response predictor in current cohort and warrants validation in larger studies.

We also found that NLR tracked systemic inflammation as reflected by CRP. CRP itself showed prognostic relevance in our cohort: higher CRP was associated with inferior PFS and OS, and CRP remained significant when modeled as a continuous variable. However, combining NLR with CRP did not improve OS discrimination relative to either marker alone, suggesting that these indices may capture overlapping inflammatory information in this dataset. Accordingly, NLR should not be framed as a “superior” marker; rather, it may represent an integrative signal of inflammatory burden whose incremental value beyond established inflammatory surrogates remains uncertain without larger cohorts.

Although VEGF level and bone marrow clonal plasma cell burden are key disease-related parameters in POEMS syndrome, we did not detect an association between NLR and either VEGF levels or clonal plasma cell proportion. This negative finding should not be taken as evidence against a biological relationship. VEGF exhibits substantial interindividual variability, and clonal plasma cells in POEMS are often patchy and present at low frequencies, which can attenuate detectable correlations. We speculated that NLR may serve as an integrative biomarker capturing two distinct yet interconnected dimensions of the disease: the systemic inflammatory/immune response of the host and the intrinsic biology of the plasma cell clone.

An additional observation was that typical skin manifestations (hyperpigmentation, skin thickening, hemangiomas, etc.)[28] were less common in the high-NLR group despite poorer outcomes. First, retrospective ascertainment bias should be considered, as dermatologic findings can be under-recorded when severe systemic involvement dominates clinical attention. Second, patients with a lower inflammatory burden may experience a relatively indolent disease course, permitting the gradual emergence of characteristic cutaneous changes, whereas an inflammation-driven phenotype may be dominated by earlier visceral involvement, potentially preceding conspicuous skin manifestations. Third, pre-referral exposure to anti-inflammatory therapies (e.g., glucocorticoids) could theoretically increase neutrophil counts and reduce lymphocyte counts [29, 30], thereby elevating the NLR while potentially attenuating inflammatory skin findings, which could contribute to an inverse association. Our dataset cannot adjudicate among these hypotheses. Clinically, this observation underscores the need for heightened suspicion in patients with atypical or minimal skin findings but markedly elevated inflammatory markers. Future studies in larger cohorts are warranted to further clarify the clinical heterogeneity associated with different inflammatory phenotypes.

Several limitations should be acknowledged. First, the single-center retrospective design and modest sample size may introduce selection bias and missing-data constraints, which could reduce precision and yield unstable effect estimates. Second, because the study spanned a long period during which influential reports from the Mayo Clinic were published, both clinical awareness and diagnostic work-up evolved over time. Although we performed period-stratified analyses and assessed baseline comparability across eras, residual confounding from secular changes in clinical practice cannot be fully excluded. Third, the NLR cut-off value of 2.773 and its prognostic value require external validation in independent, multi-center cohorts and across broader patient populations. Fourth, routine measurement of key inflammatory cytokines (e.g., IL-6 and IL-12) was not consistently performed in the earlier years, limiting our ability to draw mechanistic inferences.

In summary, this exploratory study suggests that elevated baseline NLR is associated with less favorable survival outcomes in POEMS syndrome and may help identify patients with higher systemic inflammatory burden. As a simple and readily obtainable index, NLR could complement existing prognostic assessments by incorporating an inflammation–immune dimension, thereby refining risk stratification. If validated, the integration of NLR into clinical practice could translate into more nuanced management pathways. For instance, patients with a high NLR—particularly within higher-risk strata—might be monitored under a more intensive surveillance protocol, characterized by more frequent follow-up in the first year, earlier reassessment of treatment depth (e.g., hematologic response and VEGF/CRP levels), and a lower threshold for evaluating evolving organ dysfunction. From a therapeutic perspective, this stratification could aid in identifying a subset of patients for whom a more aggressive or earlier escalation of immunomodulatory therapy might be considered. It is crucial to emphasize, however, that any such application remains hypothetical and must await robust validation. Future large-scale, multi-center studies are warranted to validate these findings, to define operational thresholds, and to clarify the clinical utility of NLR, with the long-term goal of supporting more individualized management approaches in POEMS syndrome.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

T Z performed data analysis and contributed to manuscript writing. Q C performed data analysis and manuscript review. XX S, YY J, and N S participated in data collection and critical revision for important intellectual content. LJ C and J X approved the final version of the manuscript, supported the research, and agreed to be accountable for all aspects of the work.

Funding

This work was supported by grants from the National Natural Science Foundation of China (No. 82370205), Specialized Diseases Clinical Research Fund of Jiangsu Province Hospital (XB202405), Bethune Medical Research Innovation Program-Scientific Research Fund Project (No. 2024-YJ-156-J-047).

Data availability

The data supporting the conclusions of this research are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Consent to participate

Informed consent was obtained from all individual participants included in the study.

Consent to publish

Not Applicable.

Ethical

This retrospective cohort study was approved by the Ethics Committee of the First Affiliated Hospital of Nanjing Medical University (No. 2022-SR-448) and conducted in accordance with the World Medical Association Declaration of Helsinki.