Preoperative prognostic nutritional index is a powerful predictor of prognosis in patients with stage III ovarian cancer

Many established inflammation- and nutrition-related factors have been investigated as potential independent prognostic factors in various cancers, including the C-reactive protein/albumin ratio (CAR), lymphocyte/monocyte ratio (LMR), modified Glasgow prognostic score (mGPS), body mass index (BMI), and prognostic nutritional index (PNI). This study was performed to estimate the prognostic value of these factors in predicting survival and platinum resistance in ovarian cancer (OC), especially according to stage. Kaplan-Meier and multivariate analyses were performed to plot the survival curve and determine the independent prognostic factors. Additionally, the area under the receiver operating characteristic curve (AUC) was used to predict platinum resistance and prognosis by comparing the predictive ability of PNI and cancer antigen (CA)-125. In all patients, decreased PNI was significantly associated with platinum resistance and poor overall survival (OS) and progression-free survival (PFS). Regarding tumor stage, decreased PNI was significantly associated with poor PFS and OS only in stage III OC. Furthermore, the PNI also showed a significantly higher AUC value than CA-125 for predicting mortality and platinum resistance in all OC patients, but not in stage III patients. In conclusion, decreased PNI is a powerful predictor of a poor prognosis in OC, and especially for stage III cases.

C-reactive protein/albumin ratio (CAR), lymphocyte/monocyte ratio (LMR), albumin and lymphocyte count combined into the prognostic nutritional index (PNI), and CRP-and albumin-related factors of the modified Glasgow prognostic score (mGPS) [9][10][11][12] . As an efficient, simple, and convenient novel prognostic factor, the PNI is calculated according to the following formula: serum albumin value (g/L) + 0.005 × lymphocyte count (per mm 3 ) in peripheral blood 11 . Recently, PNI has been reported to be an independent prognostic factor for survival in different malignant carcinomas, including colorectal cancer, gastric cancer, lung cancer, and pancreatic cancer [13][14][15][16] . However, the prognostic importance of PNI for OC still needs to be elucidated, especially according to tumor stage. Although Miao et al. 17 reported that PNI was an independent prognostic factor in OC patients, they did not assess the combination of PNI with other established prognostic factors, such as CAR, LMR, and mGPS. Thus, it is meaningful to combine the PNI and other established nutrition-and inflammation-related prognostic factors to obtain optimal independent prognostic scores for predicting the chemoresistance and clinical outcomes of OC patients at different stages. Correlation between PNI and clinicopathological parameters. The relation between preoperative PNI and the clinicopathological characteristics of patients with OC is shown in Table 1. Decreased PNI was significantly associated with advanced FIGO tumor stage (P < 0.001), maximum residual tumor (P < 0.001), histological subtype (P = 0.001), malignant ascites (P < 0.001), cancer antigen (CA)-125 ≥ 35 U/ml (P < 0.001), platinum resistance (P < 0.001), lower LMR (P < 0.001), and higher CAR (P < 0.001) and mGPS (P < 0.001). However, there were no significant associations between PNI and age (P = 0.066), grade (P = 0.237), or body mass index (BMI) (P = 0.460). Among tumor stage III patients, decreased PNI was also significantly associated with residual tumor mass (P = 0.023), histological subtype (P = 0.005), malignant ascites (P < 0.001), CA-125 ≥ 35U/ ml (P = 0.006), lower LMR (P < 0.001), and higher CAR (P < 0.001) and mGPS (P < 0.001), but not with platinum resistance (P = 0.095).
When patients were stratified by FIGO tumor stage, high-PNI patients had significantly longer PFS than low-PNI patients only for cases of FIGO tumor stages III (P < 0.001) and IV (P = 0.005) (Fig. 2D,E). Similarly, high-PNI patients had significantly longer OS than low-PNI patients only in stages III (P < 0.001) and IV (P = 0.010) (Fig. 3D,E). However, the multivariate Cox regression model demonstrated that the PNI was an independent predictive factor of poor PFS (HR 1.815, 95% CI 1.113-2.958, P = 0.017) and OS (HR 1.699, 95% CI 1.035-2.789, P = 0.036) only in FIGO tumor stage III OC patients, as were residual tumor mass and chemosensitivity. All these findings show that the PNI is an independent risk factor for poor PFS and OS in OC patients, especially those at stage III.
Comparison of predictive ability. The receiver operating characteristic curve (ROC) and area under the receiver operating characteristic curve (AUC) values were used to compare the predictive ability among CA-125, PNI, and their combination for OS and platinum resistance ( Fig. 4 and Table 4). With respect to predicting mortality, the PNI had a significantly higher AUC value than the CA-125 (0.677 vs. 0.567, P = 0.044). The combination of the PNI and CA-125 had a higher AUC value than either alone, although the difference was not significant (P > 0.05). Regarding platinum resistance, the PNI showed a higher AUC value than CA-125 (0.699 vs. 0.560, P = 0.006) and their combination (0.699 vs. 0.692, P = 0.847). The combination of the PNI and CA-125 had a significantly higher AUC value than CA-125 (P = 0.007). However, the PNI did not have a significantly higher AUC than CA-125 with respect to OS (0.649 vs. 0.547, P = 0.388) or platinum resistance (0.618 vs. 0.520, P = 0.094) among FIGO tumor stage III patients. Furthermore, the combination of PNI and CA-125 also did not have significantly higher AUC value than either one alone.

Discussion
To date, no widespread nutrition-or inflammation-related factor has been found to index chemoresistance or prognosis in OC patients, especially according to tumor stage. Although the association between PNI and prognosis has been clarified in other cancers 18 , the impact of PNI on platinum resistance and clinical outcomes in OC, especially according to tumor stage, has not been clarified.
Laky et al. 19 showed that about 20% of newly diagnosed gynecologic cancer patients have malnutrition. More than 20% of cancer patients die from malnutrition rather than the cancer itself 20 . Due to the metabolic effects of tumor mass, malignant ascites, and small bowel obstruction, OC patients are more likely to present with malnutrition and cachexia 21 . Furthermore, the tumor is more prone to develop chemoresistance in malnourished OC patients 22 . Recently, Matassa et al. 23 also observed that oxidative metabolism drives inflammation-induced platinum resistance in OC. Lymphocytes were also reported to play a major role in immune responses by mediating the immunologic damage caused by various cancers 24 . As components of the PNI, both the albumin count and the lymphocyte count are closely related to inflammatory responses in cancer patients, which are independent predictors of long-term outcomes in OC 25,26 . According to the prognostic association between PNI and albumin and lymphocyte counts, it seems that PNI is a reflection of systemic inflammation, which may influence cancer growth and metastasis 17 . Thus, both inflammation-and malnutrition-related prognostic factors may induce chemotherapy resistance and predict the OS of OC patients.
Consistent with previous studies, our study demonstrated that FIGO tumor stage was an independent prognostic factor in OC patient 27 . To estimate the clinical outcomes of OC patients better, many inflammationand malnutrition-based markers, such as BMI, LMR, mGPS, and CAR, have been investigated as potentially important prognostic and predictive factors in OC patients [28][29][30] . Similar to the study by Miao et al. 17   demonstrated that the independent prognostic factor best predicting the OS of OC patients was PNI rather than BMI, CAR, LMR, or mGPS. The chi-square test determined that a PNI < 47.2 was not only associated with advanced FIGO tumor stage, maximum residual tumor, malignant ascites, platinum resistance, and lower LMR but also with higher CAR and mGPS. However, our study further showed that when patients were stratified by FIGO tumor stage, stage III patients showed the most significant association between PNI level and the outcome of the disease. Furthermore, ROC and AUC analyses showed that PNI was significantly superior to CA-125 in predicting mortality and platinum resistance in all-stage OC patients, but not in stage III cases. These results suggest that as an easily available laboratory hematological marker, PNI is superior to other nutrition-and inflammation-related prognostic factors in predicting survival in OC patients, especially for FIGO tumor stage III patients. Furthermore, PNI may also predict the platinum-based chemotherapeutic response of all-stage OC patients. This study provides further support for the proposition that elevated preoperative PNI is associated with a good prognosis in OC patients. A study by Liu et al. 31 showed that the CAR had superior prognostic ability compared to other established inflammation-related prognostic indices, such as the PNI, mGPS, neutrophil/ lymphocyte ratio (NLR), and platelet/lymphocyte ratio (PLR) in 200 OC patients. The reason for this difference may be that our study included LMR, BMI, and platinum resistance. In addition, the current study not only further assessed the correlation between PNI and tumor stage but also compared the predictive ability of CA-125, the PNI, and their combination with respect to OS and platinum resistance, according to ROC and AUC values. Nevertheless, both Liu et al. 31 and the present study used retrospective, single-center studies, and the number of patients was small in both. Therefore, more studies are needed to confirm these results. Furthermore, the mechanisms linking the PNI, poor prognosis and platinum resistance must be clarified.

Materials and Methods
Patients. In total, 237 newly diagnosed OC patients, treated with cytoreductive surgery and platinum-based chemotherapy between January 2007 and December 2015 at Nanfang Hospital of Southern Medical University, were identified. Pathological parameters, clinical data, and survival times were extracted from medical records. Patients who had active infection, coexisting hematologic malignancies, or other hematologic or autoimmune disorders were excluded. The primary endpoint of the study was PFS, which was calculated from the date of treatment to the date of recurrence or progression. OS was defined as the time from treatment to the date of death or last follow-up. All OC patients were followed up every 2-4 months for the first 2 years, and every 3-6 months thereafter until December 2016. At each visit, the patients were assessed by clinical and imaging examinations and the serum levels of CA-125 of patients were assessed. This study was approved by the medical ethics committee   Table 4. Comparison of the diagnostic performance in predicting mortality and chemoresistance. AUC, area under the receiver operating characteristic curve; PNI, prognostic nutritional index.
of Southern Medical University. All methods were performed in accordance with the relevant guidelines and regulations. Written informed consent was obtained from each patient. All of the following data were obtained from medical records: age, BMI, FIGO stage, massive ascites, surgery, residual tumor mass, tumor (histology, grade), chemosensitivity, and clinical characteristics (CA-125, CRP, albumin, lymphocyte, and monocyte levels). Based on previous studies, optimal debulking was defined as a maximum diameter of residual tumor after surgery of ≤2 cm 32,33 . Patients were defined as platinum-resistant if the disease progressed within 6 months after completing first-line platinum-based chemotherapy, while all other patients were defined as platinum-sensitive 34 . PNI was calculated according to the following formula: serum albumin (g/L) + 0.005 × lymphocyte count (per mm 3 ) in the peripheral blood 11 . CAR was calculated by CRP (mg/L)/ albumin (g/L) ratio 35 . LMR was defined as the absolute lymphocyte count/absolute monocyte ratio 36 . The mGPS encompassed both the CRP and albumin concentrations. Patients with both CRP > 10 mg/L and albumin < 35 g/L were allocated a score of 2. Patients with both CRP ≤ 10 mg/L and albumin ≥ 35 g/L were allocated a score of 0. Patients with only one of these abnormal levels were given a score of 1 12 . BMI, CAR, and LMR were categorized into two groups according to the cutoff values of ≥18.5 kg/m 2 , ≥0.5, and ≥3.82, respectively 9,37,38 .
Statistical analysis. Statistical analyses were performed with SPSS software (ver. 20.0; IBM Corp., Armonk, NY, USA). Comparisons between categorical variables were performed using the chi-square test. The optimal cutoff value for PNI was determined via a web-based application, programmed in R by Budczies et al. (http:// molpath.charite.de/cutoff/) 39 . Significant prognostic variables in univariate analyses were included in multivariate Cox regression models to determine independent prognostic factors, using a forward stepwise method. Differences in survival among classification groups were analyzed using Kaplan-Meier curves and log-rank tests. ROC curves were calculated for PNI and CA-125, alone and in combination. The AUC values were compared using MedCalc software (ver. 15.2.1; MedCalc Software bvba, Ostend, Belgium). A two-sided P value < 0.05 was considered statistically significant.