Between January 2022 and December 2023, this study enrolled 199 patients from the Department of Obstetrics and Gynecology at the Second People’s Hospital Affiliated to Suzhou University, along with 120 patients from the Department of Obstetrics and Gynecology at Binhai County People’s Hospital in Yancheng City, who were diagnosed with ovarian cancer and required surgical intervention. The study received approval and oversight from the Ethics Committee of the Second People’s Hospital Affiliated of Soochow University (Approval No. JD-LK-2022-079-01) and Binhai County People’s Hospital in Yancheng City (Approval No. 2024-BYKYLL-041). The study was conducted in accordance with the revised 2013 Helsinki Declaration. Written informed consent was obtained from each participant or their legal representative.

The inclusion criteria were as follows: (1) age ≥18 years; (2) a confirmed diagnosis of ovarian cancer requiring surgical intervention, as verified by a physician; and (3) provision of signed informed consent by the patient.

The exclusion criteria were as follows: (1) incomplete laboratory data; (2) a diagnosis of ovarian cancer without surgical treatment; and (3) refusal to provide informed consent.

The study included a broad spectrum of variables: (1) Demographic data: age and body mass index (BMI); (2) Laboratory markers: TC, albumin, neutrophil, lymphocyte, and monocyte counts, carbohydrate antigen 125 (CA125), CA199, human epididymis secretory protein 4 (HE4), and carcinoembryonic antigen (CEA); (3) Comorbid conditions: hypertension and diabetes; (4) Tumor characteristics: tumor size, type of surgery (PDS or IDS), residual tumor post-surgery (yes or no), coronavirus disease 2019 (COVID-19) infection (yes or no), pathological type, tissue type, lymph node metastasis, International Federation of Gynecology and Obstetrics (FIGO) grade, differentiation status, history of chemotherapy, and targeted therapy; and (5) Length of hospital stay. Trained professionals collected laboratory markers and tumor characteristics. Biochemical analyses were conducted using a standardized automated system under the supervision of laboratory physicians, while pathologists determined tumor diagnoses and grading. Data on preexisting conditions were extracted from patients’ medical histories, and the hospital discharge system recorded the length of stay. Follow-up data were obtained through outpatient visits or telephone interviews. Survival outcomes were assessed at one, three, and five years post-surgery. The primary endpoints of the study were OS and CSS.

The study stratified variables according to criteria established by X-tile software ( Figure 1 ). Age was categorized into three groups: ≤59 years, 60–69 years, and ≥70 years. Tumor size was classified as <8 cm, 8–12 cm, and >12 cm.

The nutritional inflammatory markers, NLR, LMR, PNI, and NPS, were calculated using the following formulas: (1) NLR as the ratio of neutrophil to lymphocyte counts; (2) LMR as the ratio of lymphocyte to monocyte counts; (3) PNI as albumin (g/L) plus five times the lymphocyte count; and (4) NPS, determined based on serum albumin, TC, NLR, and LMR levels following established methodologies. Using X-tile software, PNI was stratified into three categories: <46.1, 46.1–51.2, and >51.2 ( Figure 1 ). NPS was assigned based on these criteria: (1) a score of 0 for serum albumin ≥40 g/L, TC >4.68 mmol/L, NLR <2.96, and LMR >4.44; (2) a score of 1 for serum albumin <40 g/L, TC ≤4.68 mmol/L, NLR ≥2.96, and LMR ≤4.44. The total NPS was calculated by summing the scores of these four parameters. Cumulative NPS values of 0–2 were categorized as low-NPS (L-NPS), while scores of 3–4 were defined as high-NPS (H-NPS).

The baseline characteristics of all enrolled patients were stratified according to survival outcomes. Non-normally distributed variables were expressed as quartile ranges and analyzed using the Wilcoxon rank-sum test. Categorical variables were presented as percentages and compared with chi-square tests. Univariate and multivariate Cox regression models were applied to calculate hazard ratios (HRs) and assess the independence of nutritional inflammatory indices as prognostic factors for OS and CSS in ovarian cancer patients. Kaplan-Meier curves were generated to illustrate OS and CSS across patient groups. To further examine the relationship between nutritional inflammatory markers and survival outcomes, restricted cubic spline (RCS) plots were constructed. A nomogram incorporating statistically significant markers from the multivariate Cox regression was developed. The nomogram’s diagnostic performance was evaluated using receiver operating characteristic (ROC) curves and calibration plots. ROC curves were also used to assess the prognostic value of different markers for ovarian cancer patients. External validation was performed concurrently. Finally, stratified analyses were conducted to explore the effects of various nutritional inflammatory indices on subpopulations of ovarian cancer patients. Statistical analyses were performed with R software (version 4.3.0) and STATA 17.0 (64-bit). A bilateral p-value <0.05 was considered statistically significant.

A cohort of 199 ovarian cancer patients who underwent surgical treatment at the Second Affiliated Hospital of Suzhou University was analyzed. Of these, 141 patients (70.85%) survived, while 58 patients (29.15%) succumbed to the disease. The demographic and clinical characteristics of both groups were summarized in Table 1 . Increased mortality risk was associated with advanced age, lower albumin levels, reduced lymphocyte counts, decreased PNI, elevated neutrophil counts, higher NLR, increased HE4, H-NPS, and postoperative tumor residue. Additional high-risk factors included high-grade histology, larger tumor size, FIGO stage IV, poor tumor differentiation, absence of targeted therapy, and a history of diabetes. Statistically significant differences between survivors and deceased patients were observed for all these variables (all p < 0.05). Furthermore, deceased patients had longer hospital stays compared to survivors. No significant differences were found regarding monocyte counts, CA125, CA199, CEA, tumor histological type, surgical procedure, COVID-19 infection, or lymphatic metastasis. The lack of statistical significance in postoperative survival and prognosis may be attributed to the small sample size of this study.

An analysis of 120 patients from Binhai County People’s Hospital ( Table 1 ) yielded results consistent with those from the Second Affiliated Hospital to Suzhou University. Factors including advanced age, elevated BMI, low albumin levels, reduced lymphocyte count, increased NLR, elevated HE4, low PNI, H-NPS, lymphocyte metastasis, poor differentiation, absence of radiotherapy or targeted therapy, residual tumor post-surgery, and comorbid diabetes were all significantly associated with a higher risk of mortality (all p<0.05). In contrast, neutrophil levels did not exhibit statistical significance in the Binhai County cohort, which may be attributed to the smaller sample size in our study.

Kaplan-Meier survival analysis revealed significant associations between PNI, NPS, and survival outcomes in ovarian cancer patients (p < 0.05), as depicted in Figure 2 . Patients with PNI > 51.2 exhibited the best OS, whereas those with PNI < 46.1 had the poorest prognosis ( Figures 2A-1 ). Likewise, survival outcomes were significantly better for patients with L-NPS compared to those with H-NPS ( Figure 2A-2 ). These patterns were similarly observed for CSS, as shown in Figures 2A-3, 2A-4 . Consistent results were obtained from the data of Binhai County People’s Hospital ( Figures 2B-1–4 ).

Univariate and multivariate Cox regression analyses were conducted to identify factors influencing OS and CSS. For OS, independent risk factors included age, PNI, NPS, tissue type, tumor size, targeted therapy status, and diabetes ( Table 2 ). Specifically, patients with PNI between 46.1 and 51.2 demonstrated improved survival outcomes compared to those with PNI <46.1 (HR = 0.461; 95% CI, 0.234–0.864; p = 0.036), while those with PNI >51.2 exhibited even more pronounced protective effects (HR = 0.357; 95% CI, 0.095–0.538; p = 0.025). In contrast, patients with H-NPS had significantly worse survival compared to those with L-NPS (HR = 1.576; 95% CI, 1.252–2.719; p = 0.013). For CSS, multivariate Cox regression analysis ( Table 3 ) identified age, NPS, tumor size, targeted therapy status, and diabetes as independent risk factors. Consistent with OS, patients with H-NPS experienced poorer outcomes compared to those with L-NPS (HR = 1.654; 95% CI, 1.278–3.138; p = 0.031).

An RCS diagram ( Figure 3 ) was constructed to explore the relationship between nutritional inflammatory markers and survival prognosis in postoperative ovarian cancer patients. OS analysis indicated a negative correlation between PNI and LMR with patient prognosis, suggesting that higher values of these indices were associated with improved survival outcomes ( Figures 3A, C ). In contrast, NLR exhibited a positive correlation with prognosis, with elevated NLR values linked to poorer survival ( Figure 3B ). The CSS analysis mirrored the findings from OS, revealing a negative correlation between both PNI and LMR with prognosis ( Figures 3D, F ), while NLR remained positively correlated, indicating worse outcomes as its value increased ( Figure 3E ). Additional analyses included markers such as albumin, TC, CA125, CA199, CEA, and HE4 ( Supplementary Figure S1 ).

Univariate and multivariate Cox regression analyses identified seven independent predictors of OS: age, PNI, NPS, tissue type, tumor size, targeted therapy use, and diabetes presence (p < 0.05; Table 2 ). Similarly, five variables were determined to be independent predictors of CSS (p < 0.05; Table 3 ). Nomograms incorporating these predictors were developed to estimate one-, three-, and five-year OS and CSS ( Figures 4A, D ). Risk scores for each factor in the nomogram, shown in Figure 4 , indicated that higher scores corresponded to increased mortality risk. The predictive accuracy of the nomograms was assessed, yielding a concordance (C-index) of 0.715 for the OS nomogram and 0.704 for the CSS nomogram. ROC curves for one-, three-, and five-year OS and CSS were plotted, with the area under the curve (AUC) values for OS being 0.722, 0.801, and 0.890, respectively ( Figure 4B ). For CSS, the AUC values were 0.708, 0.827, and 0.840, respectively ( Figure 4E ). Calibration curves for three- and five-year predictions closely aligned with the diagonal, indicating high consistency and reliable calibration, whereas the one-year calibration curves deviated, suggesting lower robustness in predicting one-year outcomes ( Figures 4C, F ).

External validation was performed using data from 120 patients at Binhai County People’s Hospital, applying the same characteristic variables to construct a nomogram for predicting 1-, 3-, and 5-year OS and CSS ( Figures 5A, B ). The nomogram’s risk scores for each factor were presented in Figure 5 , where higher scores correlated with an increased risk of mortality. Predictive accuracy was assessed by evaluating the C-index, which was 0.735 for the OS nomogram and 0.714 for the CSS nomogram. ROC curves for 1-year, 3-year, and 5-year OS and CSS were also generated. For OS, AUC values were 0.753, 0.782, and 0.724, respectively ( Figure 5A-2 ), while for CSS, AUC values were 0.743, 0.766, and 0.708 ( Figure 5B-2 ). Calibration curves for 3- and 5-year predictions closely approximated the diagonal, reflecting strong consistency and reliable calibration. In contrast, the 1-year calibration curve showed notable deviation from the diagonal, indicating reduced predictive accuracy for the 1-year outcome ( Figures 5A-3, 5B-3 ). A comparison with data from the Second People’s Hospital of Suzhou University revealed that while model performance at Binhai County People’s Hospital was somewhat lower, the evaluation metrics still confirmed the model’s overall effectiveness.

Multivariate Cox regression analysis was used to develop a diagnostic nomogram, with correlated variables plotted alongside the ROC curve to evaluate its predictive accuracy for survival and prognosis in postoperative ovarian cancer patients ( Figure 6 ). For OS, the model exhibited the highest predictive value, yielding an AUC of 0.910 ( Figure 6B-1 ). In comparison, the AUCs for individual variables were 0.678 for PNI, 0.627 for NPS, 0.424 for tissue type, 0.743 for tumor size, and 0.703 for targeted therapy ( Figure 6A-1 ). In the CSS analysis, the model again demonstrated the highest predictive value, achieving an AUC of 0.844 ( Figure 6B-2 ). The AUCs for other variables were 0.612 for NPS, 0.722 for tumor size, and 0.685 for targeted therapy ( Figure 6A-2 ). These results confirm the nomogram’s significant diagnostic utility, highlighting its capacity to accurately predict survival and prognosis in postoperative ovarian cancer patients.

Data analysis from Binhai County People’s Hospital revealed that the OS model exhibited the highest predictive value, with an AUC of 0.898 ( Figure 6B-3 ). In contrast, the AUC values for other variables were as follows: PNI, 0.703; NPS, 0.681; tissue type, 0.576; tumor size, 0.618; and targeted therapy, 0.669 ( Figure 6A-3 ). Likewise, for CSS, the model again demonstrated the highest predictive value, with an AUC of 0.848 ( Figure 6B-4 ). The AUC values for additional variables were: NPS, 0.662; tumor size, 0.604; and targeted therapy, 0.658 ( Figure 6A-4 ). These results further substantiate the diagnostic utility of the nomogram, reinforcing its robust predictive capability for survival and prognosis in postoperative ovarian cancer patients.

Subgroup analysis of OS and CSS in postoperative ovarian cancer patients was performed based on factors such as age, tumor size, histological type, and tissue characteristics to assess survival prognosis. OS analysis using the NPS ( Figures 7A–C ) revealed that patients aged 60–69 years with H-NPS faced a significantly higher mortality risk than those with L-NPS (HR = 3.534; 95% CI, 1.230–10.151; p = 0.019). Similar patterns were observed in patients with tumors >12 cm, endometrioid adenocarcinoma, FIGO grade I, as well as those undergoing surgery and chemotherapy (all p < 0.05). Additionally, among patients aged 60–69 years, those with tumors >12 cm, serous adenocarcinoma, low-grade histology, no lymphatic metastasis, absence of targeted therapy, and without hypertension or diabetes, demonstrated the lowest mortality when PNI >51.2 (p < 0.05) ( Figures 7D–F ). In terms of CSS, the NPS analysis yielded comparable results ( Figures 7G–I ), with higher mortality rates in patients with tumors >12 cm and endometrioid adenocarcinoma, regardless of hypertension, surgical procedure, or residual tumor presence (all p < 0.05). Further PNI analysis for CSS ( Figures 7J–L ) indicated that patients aged 60–69 years with tumors >12 cm, serous adenocarcinoma, high-grade histology, absence of lymphatic metastasis, receiving chemotherapy, without targeted therapy, and regardless of hypertension or diabetes, surgery type, or residual tumor showed the lowest mortality rates when PNI >51.2 (all p < 0.05).