Validation of the preoperative controlling nutritional status score as an independent predictor in a large Chinese cohort of patients with upper tract urothelial carcinoma

Abstract Background Pretreatment controlling nutritional status (CONUT) score is a novel index which was used to predict outcomes in cancer patients. We aim to explore the prognostic significance of CONUT score in patients with upper tract urothelial carcinoma (UTUC) after radical nephroureterectomy (RNU). Patients and methods A total of 662 UTUC patients between 2004 and 2016 were retrospectively analyzed. Patients were categorized into three groups based on CONUT score (Normal: 0‐1; Light: 2‐4; Moderate/severe: 5‐12). Associations of CONUT score with oncological outcomes were analyzed using Logistic and Cox regression analysis. Harrell concordance index was used to assess the predictive accuracy of the multivariate models. Subgroup analyses were conducted according to tumor grade and stage. Results The median follow‐up duration was 41 months. Multivariate Logistic analysis showed that high CONUT score was independently associated with high‐grade disease, high pT stage, lymphovascular invasion, sessile carcinoma, variant histology, and positive surgical margins (each P < 0.05). Multivariate analysis demonstrated that CONUT score 5‐12 was an independent factor for worse cancer‐specific survival (CSS, hazard ratio [HR]:2.39, 95% confidence interval [CI] 1.55‐3.68, P < 0.0001), disease recurrence‐free‐survival (RFS, HR: 1.80, 95% CI 1.24‐2.60, P = 0.002), and overall survival (OS, HR: 2.26, 95% CI 1.53‐3.34, P < 0.0001). The estimated c‐index of the multivariate models for CSS, RFS, and OS increased from 0.755, 0.715 and 0.745 to 0.772, 0.723, and 0.756 when CONUT score supplemented. Subgroup analyses showed that especially in patients with high‐grade carcinoma and advanced stage (≥pT3), higher CONUT score predicts decreased CSS, RFS, and OS (all P < 0.05). Conclusion Preoperative CONUT score is a negative independent prognostic indicator for both pathologic and survival outcomes in UTUC, especially in those with high‐grade carcinoma and advanced stage. Adding this parameter into our clinical prediction model is appropriate so as to improve its predictive accuracy.


| INTRODUCTION
Upper tract urothelial carcinoma (UTUC) consists of ureteral carcinoma and renal pelvic carcinoma and occurs at a low frequency, accounting for roughly 5%-10% of urothelial carcinomas in western population. 1 Patients diagnosed with UTUC usually have worse prognosis in comparison of other genitourinary cancer (bladder cancer, prostate cancer, and renal carcinoma) despite undergoing radical nephroureterectomy (RNU) with bladder cuff resection, the standard procedure currently, with the 5-year survival rate <50% in muscle-invasive UTUC. 1 Therefore, identifying prognostic factors which might allow effective interventions before or after surgery is urgent so as to prolong patients' life expectancy.
In addition to the established pathologic prognostic indicators such as tumor stage and grade, recent interests have been switched to focus on exploring preoperative novel prognostic biomarkers in UTUC. Given the potential role of systematic inflammation and nutrition status in tumor progression and their associations with oncological outcomes, 2,3 several biomarkers that reflect the above two status have been explored in UTUC including neutrophil to lymphocyte ratio, 4 albumin to globulin ratio, 5 nutritional index, 6 and controlling nutritional status (CONUT) score. 7 Of them, CONUT score, a novel biomarker which is calculated from serum albumin, total lymphocyte counts, and total cholesterol concentrations, has shown its independent prognostic significance in various solid tumor types such as hepatocellular carcinoma, 8 gastric cancer, 9,10 and colorectal cancer. 11 It reflects individuals' nutrition status, immunological function, protein and lipids metabolism. Ishihara et al 7 first described the prognostic significance of pretreatment CONUT score in 107 UTUC sets of Japanese origin. They found that CONUT score was an independent predictor of cancer-specific survival (CSS) and overall survival (OS) in UTUC patients after RNU.
Thus, we sought to further explore and validate the prognostic value of CONUT score in a large Chinese cohort of patients with UTUC. We also tend to assess whether it would improve the predictive accuracy in multivariate models for UTUC prognosis.

| Study population
This retrospective study received the approval from the Ethical Committee of West China Hospital. A total of 801 patients diagnosed with UTUC who had undergone RNU at our center from January 2004 to December 2016 were retrospectively reviewed. We excluded 32 patients without available clinicopathological and laboratory information, 18 patients with autoimmune disease/hepatic disease, 11 patients with non-urothelial carcinoma, and 58 patients lost at the beginning of follow-up. No patients had received neo-adjuvant chemotherapies. The remaining 662 cases were included in our study for further analysis.
Open and laparoscopic RNU combined with open bladder cuff excision were performed at our department. Lymph node dissection was performed when suspected enlarged lymph nodes (identified by the preoperative radiology or intraoperative inspection).
Each patient's information was retrieved from their medical records. Pathologic specimens were reevaluated by two experienced pathologists. The World Health Organization with high-grade carcinoma and advanced stage. Adding this parameter into our clinical prediction model is appropriate so as to improve its predictive accuracy.

K E Y W O R D S
biomarker, controlling nutritional status, prognosis, radical nephroureterectomy, upper tract urothelial carcinoma T A B L E 1 CONUT

| CONUT score definitions
The laboratory information was obtained within 2 weeks before surgery. The following three indicators: serum albumin, total lymphocyte counts, and total cholesterol concentrations were used to calculate CONUT score (based on published reports, 8,11 see Table1). Patients were divided into three groups according to their CONUT score: normal group: score 0-1; light group: score 2-4; and moderate/severe group: score 5-12.

| Follow-up
Physical examination, urinary tests, and blood laboratory tests were routinely performed. Cystoscopy was done every 3 months for the first year after RNU, every 6 months for the next 2 years, and then once a year thereafter. Computed tomography or magnetic resonance imaging was performed every year or when suspected disease recurrence. Postoperative treatment regimens mainly included adjuvant chemotherapy (intravesical chemotherapy and/or systematic chemotherapy) and adjuvant radiotherapy, which were administrated according to tumor stage, nodal involvement, doctor's selection, patients' conditions, and desire. Intravesical chemotherapy drugs were mainly mitomycin C and pirarubicin. Systematic chemotherapy drugs were platinum-based regimens. And there was one patient who had received immunotherapy. Disease recurrence was defined as recurrence from the operating site, lymph nodes, and visceral metastasis (to clarify: intravesical recurrence was not classified into disease recurrence). CSS was defined as the time in months from the date of RNU to cancer-related death. Disease recurrence-free survival (RFS) was defined as the time in months from RNU to disease recurrence. OS was defined as the time in months from RNU to death from all cause.

| Statistical analysis
The Mann-Whitney U test and Chi-squared test were used to evaluate the continuous variables and dichotomous variables, respectively. Multivariate Logistic regression analysis was used to evaluate associations of CONUT score with adverse pathological outcomes after adjusting preoperative confounders. Probabilities of CSS, RFS, and OS were estimated by the Kaplan-Meier method (Log-rank tests were selected) and subgroup analyses were conducted according to tumor grade and tumor stage. T A B L E 2 (Continued) Cox proportional hazard regression models were used to identify the risk factors for CSS, RFS, and OS. Variables with a P value <0.05 in our univariate analysis were included in the multivariate model by using backward stepwise procedure (P < 0.2 for entry; P < 0.15 to remain). The c-index was calculated to assess the model's predictive accuracy. A two side P value <0.05 was considered statistically significant. Statistical analyses were carried out using R software (version 3.4.4) and IBM SPSS Statistics version 22.0 (IBM Corp., Armonk, NY, USA).

| Associations of CONUT score with pathological features
We have performed the multivariate logistic regression analysis to explore the associations of CONUT score with adverse pathological outcomes. Kaplan-Meier curves revealed that patients with higher CONUT score had significant worse survival compared with those who had normal CONUT score with respect to CSS, RFS, and OS (Log-rank test, each P < 0.0001, Figure 1A-C). Univariate and multivariate Cox regression analysis (backward stepwise method) were performed in our study (Table S1 and Table 4). Univariate analysis demonstrated that tumor grade, tumor stage, lymph node status, LVI, tumor size, PSM, tumor architecture, CVH, and CONUT score were significant predictors for CSS, RFS, and OS (each P < 0.01). In addition, hydronephrosis was a significant predictor for RFS, but not for CSS and OS. Surgical approach was a prognostic factor for CSS and OS while not for RFS. Factors with the P value <0.05 were included in the multivariate Cox regression models by adopting backward stepwise method. These seven factors were included in the final models including tumor grade, pT stage, lymph node status, tumor size, tumor architecture, CVH, and CONUT score. The results showed that CONUT score was an independent predictor for CSS ( The predictive accuracies of the multivariate models for survival outcomes considering whether CONUT score was supplemented were also assessed. In the base model, which included traditional variables of tumor grade, pT stage, lymph node status, tumor size, tumor architecture, and CVH, the predictive accuracy was 75.5%, 71.5%, and 74.5% for CSS, RFS, and OS, respectively. When the CONUT score was added, the predictive accuracy of the final model elevated to 77.2%, 72.3%, and 75.6% for CSS, RFS, and OS, respectively (Table 4).

| Subgroup analyses
As for tumor grade and stage are the two most important factors in the prognosis of UTUC, we therefore further examined the prognostic role of CONUT score in UTUC patients stratified by tumor grade and pathological T stage. K-M curves showed that patients with high CONUT score have significant lower CSS, RFS, and OS in those with high-grade UTUC and advanced pT stage (log-rank test, all P < 0.05, Figure 2D-F and 3D-F) while no significant differences were observed between groups in those with lowgrade UTUC and early stage (log-rank test, all P > 0.05, Figures 2A-C and 3A-C). We also sought to explore whether the independent value of CONUT score would exhibit when considering tumor grade and stage. Table 5 showed the subgroup analyses of associations of CONUT score with survival outcomes. Our results revealed that high CONUT score can independently predict worse CSS, RFS and OS in high-grade UTUC (HR: 2.65, 2.21, and 2.50, respectively) and in advanced UTUC (HR: 3.28, 2.40, and 2.79, respectively), while not in low-grade UTUC and early stage UTUC (all P > 0.05; Table 5).

| DISCUSSION
In this retrospective study, we enrolled 662 patients of Chinese origin with UTUC after RNU and we demonstrated that CONUT score was an independent predictor for worse oncological outcomes. High CONUT score had significant lower CSS, RFS, and OS rates compared with normal CONUT score, and this phenomenon was more pronounced in patients with high grade and advanced stage of UTUC. Moreover, the addition of CONUT score included in the multivariate model would improve its predictive accuracy. To the best of our knowledge, this study is the largest to address the relationships between CONUT score and oncological outcomes in UTUC.
Probing the prognostic value of CONUT score in cancer patients is never new. Actually, it has been proved that CONUT score was an independent factor for decreased survival in many cancer types, 8,9 and a recent meta-analysis which included four publications of 674 patients also found that high CONUT score independently predicted worse OS in patients with solid tumors. 14 In the context of the low incidence of UTUC, researches in this field are still limited. Our results were in line with these studies which support the CONUT score as an independent predictor for survival outcomes. To date, only one study which is conducted by Ishihara et al 7 was available in assessing the prognostic value in UTUC sets. Similarly, we both demonstrated that CONUT score is an independent factor for both CSS and OS. Nevertheless, owing to the limited number of cases (only 107 cases) included in their study, they failed to reveal the independent value of CONUT score for RFS (HR 2.26, 95% CI 0.97-4.94), which exhibited statistical difference in our study not only in the entire cohort (moderate/severe vs normal: HR 1.80, 95% CI 1.24-1.60), but in the high-grade UTUC (HR 2.21, 95% CI: 1.49-3.26) and advanced stage (HR 2.40, 95% CI: 1.53-3.75). Furthermore, our detailed subgroup analysis and the calculation of c-index enable us to get a better understanding of the role of CONUT score in UTUC.
CONUT score is calculated from serum albumin, total lymphocyte counts, and total cholesterol concentrations. F I G U R E 2 Kaplan-Meier curves for cancer-specific survival, disease recurrence-free survival and overall survival stratified by CONUT score in UTUC patients with low-grade carcinoma (A-C) and high-grade carcinoma (D-F) Albumin is a major component of serum total proteins and it is a reflection of both nutrition and inflammation status. 15,16 Hypoalbuminemia (low serum albumin level) could decrease individual's immunity and lead to poor oncological outcomes. 17 Researches also certified the negative prognostic role of hypoalbuminemia in UTUC patients, 18,19 and our previous study also showed the independent prognostic of albumin to globulin ratio in UTUC. 5 In addition, lymphocytes function in host immunity and are considered to have antitumor ability via affecting tumor cell growth, migration, apoptosis, and inducing cytotoxicity. 21 Previous study indicated that the infiltration of lymphocyte (CD4 + / CD8 + T-lymphocytes) might affect patients' response for neoadjuvant chemotherapy in urothelial carcinoma of the bladder, 22 and our previous work also revealed that low preoperative neutrophil to lymphocyte ratio was independently associated with worse survival outcomes. 4 In addition, low serum cholesterol levels was also associated with poor survival in colorectal cancer 23 and renal cell carcinoma, 24,25 but the specific mechanism remained to be elucidated. In all, CONUT, a combination of these three indictors above, is more powerful and can be applicated in current risk stratifications of UTUC.
Malnutrition is a common phenomenon which can be found in cancer patients and previous evidence support that malnutrition was associated with worse outcomes. 26 In the study conducted by Naito et al 27 they defined CONUT score 5-12 as malnutrition status. Our results showed that patients with CONUT score 5-12 had significant worse survival rate that those with normal CONUT score, especially in those with high grade and advanced stage, providing another strong evidence regarding the negative role of CONUT in UTUC. Most importantly, in comparison of the traditional pathological factors, CONUT score can be accessed easily and cost-effectively from which way it allowed the possibility of preoperative intervention to achieve the better tumor control effects.
Limitations should be noticed as well. First, as with all retrospective studies, the selection bias could not be avoided. Second, other potential inflammation indicators (eg, CRP) were not routinely measured before their hospitalization and their medication information was incomplete, which might affect our results. Third, although we have excluded patients with autoimmune status or hepatitis, other potential disease which was not detected preoperatively might affect the level of albumin, lymphocyte, and total cholesterol. Fourth, our F I G U R E 3 Kaplan-Meier curves for cancer-specific survival, disease recurrence-free survival and overall survival stratified by CONUT score in UTUC patients with early stage (A-C) and advanced stage (D-F) results also showed that the surgical procedure and adjuvant chemotherapy would not affect patients' survival in UTUC, they should be further validated in large-scale multicenter prospective clinical trials. Last, although our data showed that the addition of CONUT score would improve the models' predictive accuracy, the model validity (internal and external validation) is still warranted so as to assist clinical practice.

| CONCLUSIONS
In this study we included a large UTUC group of Chinese origin to validate the prognostic value of CONUT score in UTUC after RNU. We finally demonstrated CONUT score independently predicted worse survival in UTUC patients, especially in patients with high-grade disease and advanced stage. The addition of CONUT score would improve the predictive accuracy of the multivariate models for survival outcomes. We therefore suggest adding the CONUT score to the traditional predictive model and to the risk stratifications of patients with UTUC.