Abstract
Background
Distant metastasis (DM) remains the most commonly reported cause of death in patients with urothelial carcinoma of the bladder (UCB).
Objective
We aimed to develop a robust prognostic model to assess the risk of DM in patients with UCB.
Methods
We collected clinical data of 206 UCB patients treated with RC. Patients treated with RC between 2011–2015 that were enrolled as the training cohort (n = 105), while the patients between 2016–2019 were enrolled as the validation cohort (n = 101). Univariate and multivariate Cox regression models were used to identify independent risk factors associated with DM. We identified the variables by stepwise regression and established nomogram. We evaluated the nomograms using C-index, calibration and ROC curves. Decision curve analysis was performed to compare the net benefits between the nomogram and TNM staging. We divided the patients into high and low risk groups according to the nomogram and compared the DM between the groups.
Results
The neutrophil-lymphocyte ratio (NLR) was an independent predictor of DM. We established nomogram by T-stage, N-stage and NLR. The C-index of the nomogram was 0.766 and 0.739 respectively in the two cohorts. In the training cohort, AUC for the nomogram at 1, 2 and 3 years was 0.816, 0.812 and 0.812, respectively. In the validation cohort, the AUC for the nomogram at 1, 2 and 3 years was 0.751, 0.757 and 0.716, respectively. The calibration curve was satisfactory. The nomogram has a higher clinical benefit compared to the TNM staging system. Kaplan-Meier curves showed that patients from the high-risk group had a higher probability of DM than patients from the low-risk group.
Conclusions
Nomograms established by NLR, T-stage and N-stage can accurately predict distant metastases in patients with UCB.
Introduction
Bladder cancer (BLCA) is the tenth most common cancer in the world, whose incidence is increasing over the years. World statistics suggest that there were 573,000 new cases and 213,000 deaths in 2020 alone. 1 At present, radical cystectomy (RC) is considered the standard treatment for invasive high-risk bladder cancer. However, despite significant advancements in surgical techniques and perioperative management, about 50% of BLCA patients end up developing distant metastases (DM). 2 According to statistics, the 5-year survival rate for patients who develop DM is only 6.8%. 3 Therefore, accurate prediction of DM is important for developing personalized therapy for patients with BLCA.4,5
The TNM phase-staging system is important for determining treatment strategies and predicting prognosis of malignant tumors. However, the fact that the prognostic levels of BLCA patients in the same stage differed significantly suggests that it alone cannot effectively assess the prognosis of BLCA patients. Recent studies have found that circulating inflammatory cells are playing an essential role in promoting hematogenous metastasis of tumor cells. For example, macrophages and neutrophils promote intraductal survival, transportation, and vascular extravasation of CTCs through different mechanisms.6–9 Furthermore, a growing body of evidence has revealed the effectiveness of circulating inflammatory cell-related variables (e.g., neutrophil-lymphocyte ratio (NLR), lymphocyte-macrophage ratio (LMR) platelet-lymphocyte ratio (PLR), C-reactive protein-to-albumin ratio, albumin-to-globulin ratio, inflammation-based index and modified Glasgow prognostic score) as prognostic factors for malignant tumors.10–14 Therefore, we believe that inflammatory cell-related variables have potential value in predicting DM of bladder cancer.
In fact, DM is a complex multi-step process driven by both internal and external factors of tumors. Radical surgery does not clear circulating tumor cells (CTCs) which have spread to the bloodstream and dormant cancer cells which are in pre-metastatic stage.4,6 Not only that, with the participation of carcinogenic factors and inflammatory cells, the metastatic process of some postoperative patients will continue until it leads to death. Nevertheless, there remains the absence of models for predicting postoperative metastasis in BLCA, which hinders further individualized treatment of postoperative patients, and to some extent exposes the overtreatment or undertreatment of patients. Therefore, establishing a prognostic model related to DM has important clinical significance for guiding further postoperative treatment of BLCA patients.
The nomogram serves as a reliable and convenient statistical prediction tool which combines multiple independent risk factors to predict our endpoints of interest as well as has been widely applied in various solid tumors. 15 For BLCA, although there are many nomograms designed to predict treatment response, postoperative survival rate, there is still a lack of models to predict postoperative metastasis. The most common histological type of BLCA is urothelial carcinoma of bladder (UCB) (approximately 90–95%). In this study, we combined histopathology and systemic inflammatory cell-related variables to create a DM-related nomogram to recognize patients at high risk for DM and to help guide individualized treatment of BLCA patients after radical cystectomy.
Materials and methods
Study Design and Patients
This study included a total of 206 UCB patients who underwent RC surgery at Lanzhou University Second Hospital from March 2011 to October 2019. The following inclusion criteria were used for patients: (1) patients with UCB who underwent RC surgery and were confirmed by postoperative histopathology reports, (2) patients who were followed up for at least 1 month and had available clinical information, (3) no history of other malignancies in the past or the same period, (4) no invasive operation within 1 month before RC, (5) no radiotherapy before RC, (6) no immune system disease and recent history of infection. Among them, patients treated with RC between 2011–2015 were enrolled in the training cohort (n = 105), while the patients between 2016–2019 were enrolled in the validation cohort (n = 101). The entire study follows the principles of the Declaration of Helsinki. The study was approved by the Clinical Research Ethics Committee of Lanzhou University Second Hospital.
Follow-up
Due to the retrospective character of this study, there was no standardization of postoperative follow-up. At our institution, patients are generally recommended to be followed up each 6 months during the 3 years following RC and annually thereafter. All patients’ follow-up information was acquired from their latest medical review, including periodic clinical examinations and assessments such as CT scans, bone scans, PET-CT, and abdominal ultrasonography. Distant metastases were confirmed by the patient's imaging data or histopathology. The endpoint of the study was DM. The last follow-up date was February 29, 2022. Distant metastasis-free survival (DMFS) is defined as the time from the date of RC to the date of DM or the last follow-up visit.
Clinical Variables and their cut-off values
The following characteristics of patients were collected and analyzed: age at surgery, gender, T stage, N stage, TNM stage, tumor grade, history of transurethral resection of bladder tumor (TURBT), vascular invasion (tumor thrombus), surgical margin status, NLR, LMR, and PLR. All cases were staged according to the AJCC/UICC 8th edition TNM staging system. Cell counts (including neutrophils, lymphocytes, monocytes, and platelets) from the preoperative complete blood count data were collected to calculate NLR, LMR, and PLR values. Specifically, NLR/ PLR was calculated as neutrophil count/ platelet count divided by lymphocyte count, and LMR was calculated as lymphocyte count divided by monocyte count. To analyze the relationship with DM, we used X-tile software to determine the optimal thresholds for continuous-type variables based on the training dataset. The cut-off values for each continuous variable are as follows: age (57 years), NLR (2.17), LMR (3.52), PLR (118).
Establishment and evaluation of nomogram
Univariate Cox regression analysis was performed for each categorical variable in the training cohort. After that, we obtained the statistically significant variables and performed multivariate Cox regression analysis on these variables. The best predictors screened by backward stepwise regression model were used to develop a nomogram. Evaluation and validation of the nomogram were performed in both cohorts.
Statistical analysis
Statistical analysis was conducted with using R-4.1.0 software. First, categorical variables were classified according to clinical data, and continuous variables were transformed into dichotomous variables according to the best cut-off point from ROC curve. In the comparison of included variables between the two cohorts, categorical variables were presented by frequencies (percentages), and chisquare test was used to compare their differences. Then, univariate and multivariate Cox regression analyses were used in the training cohort to determine independent risk factors for DM in UCB patients. Backward stepwise regression model was used to identify the optimal predictive factors and established the nomogram. We used the concordance index (C-index), ROC curve, and calibration curve to evaluate the predictive and calibration power of nomogram. Decision curve analysis (DCA) was used to evaluate the clinical benefits of nomogram. In addition, we used the ROC curve to determine the cut-off of the nomogram to predict DMFS. Patients were classified into low-risk and high-risk groups based on the cut-off point of the nomogram, and Kaplan-Meier curve and log-rank test were used to compare the DMFS of the two groups. In this study, P < 0.05 was considered to be statistically significant.
Results
Patient characteristics
According to the inclusion criteria, the study recruited a total of 206 UCB patients. These patients included 173 men and 33 women. Their median age was 63 (28–82 years). There were 136 patients (66.0%) with a history of TURBT. Table 1 shows the pathologic and clinical characteristics of 105 patients in the training cohort and 101 patients in the validation cohort. By the end of the follow-up, 33 (31.4%) patients and 22 (21.8%) patients in the training cohort and validation cohort developed DM. The median date of DM in both cohorts was 11 months, and the percentage of patients with DM at 3 years was 87.9% and 95.5% of all patients who developed DM, respectively.
The characteristics of patients in the training and validation cohorts.
Independent risk factors
Univariate Cox regression analysis showed that T-stage (HR: 5.7, 95% CI: 2.82–11.5, P < 0.001), N-stage (HR: 3.59, 95% CI: 1.69–7.66, P < 0.001), TNM stage (HR: 4.53, 95% CI: 2.20–9.29, P < 0.001), vascular invasion (HR: 3.52, 95% CI: 1.66–7.46, P = 0.001), and NLR (HR: 2.30, 95% CI: 1.13–4.70, P = 0.022) were associated with postoperative metastasis in UCB patients in the training cohort. Subsequently, multivariate Cox regression analysis of these variables showed that T-stage (HR: 3.97, 95% CI: 1.77–8.89, P = 0.001), and NLR (HR: 2.44, 95% CI: 1.15–5.2, P = 0.02) were independent risk factors for DMFS. The results of univariate Cox and multivariate Cox regression analyses are listed by Table 2.
Univariate and multivariate Cox regression analysis in the training cohort.
Establishment of DMFS-related nomogram
Among the variables including “T-stage”, “N-stage”, “vascular invasion”, and “NLR”, we finally identified “T-stage”, “N-stage” and “NLR” to establish the nomogram by backward stepwise regression analysis (Figure 1). Each variable corresponds to a score. In the nomogram scoring system, “T3/4” has a score of 100, “N1/N2/N3 has a score of 50.7” and “NLR > 2.17” has a score of 55.6. Summing up the scores for each variable and locating them on the total score, which helped to calculate the probability of DMFS at 3- years for each patient.
A nomogram model to predict 3-year DMFS in UCB patients undergoing radical cystectomy
Evaluation of the nomogram
The C-index of the nomogram in the two cohorts were 0.766 and 0.739, respectively. The ROC curve of the nomogram scoring system is displayed in Figure 2. In the training cohort, the area under curve (AUC) of the nomogram at 1, 2, and 3 years were 0.816, 0.812, and 0.812, respectively. In the validation cohort, the AUC of the nomogram at 1, 2, and 3 years were 0.751, 0.757, and 0.716, respectively. In both cohorts, calibration curves showed good consistency between nomogram prediction and the actual occurrence of distant metastases (Figure 3). DCA for the training cohort showed that threshold probability between 0 and 1.0 (and between 0 and 0.5 in the external validation) was more beneficial for predicting distant metastasis or non-metastasis in patients with BLCA using the NLR-based nomogram (Figure 4). DCA also showed that the nomogram (cyan) compared to the TNM staging system (purple) had higher clinical benefits.
The ROC curves of the nomogram scoring system for predicting DMFS (a) In the training cohort, AUC for the nomogram at 1, 2 and 3 years was 0.816, 0.812 and 0.812, respectively. (b) In the validation cohort, the AUC for the nomogram at 1, 2 and 3 years was 0.751, 0.757 and 0.716, respectively. ROC：receiver operating characteristic, DMFS：distant metastasis-free survival, AUC：area under curve.
Calibration curves of the nomogram to predict 3- DMFS in the training cohort (a) and validation cohort (b).
The Decision Curves Analysis (DCA) for the nomogram of 3-year DMFS in the training cohort (a) and validation cohort (b).
Risk stratification performance of nomogram
We obtained the cut-off value (100 points) of the nomogram scoring system by analyzing the ROC curve in the training cohort. According to the cut-off value of 100 points, patients in the training cohort and validation cohort were divided into high-risk and low-risk groups, respectively. Kaplan-Meier curves for both cohorts showed that the high-risk group was more likely to develop DM compared to the low-risk group (Figure 5).
Kaplan–Meier curves of DMFS in low-risk or high-risk groups stratified by the nomogram in the training cohort (a) and validation cohort (b).
Discussion
An accurate assessment of the risk of metastases after RC is vital for decision making and improvement of prognosis for patients with UCB. Traditionally, the TNM staging system has been the golden standard for predicting recurrence after RC in BLCA patients. 16 But in fact, DM is a complex multi-step process driven by internal and external factors of the tumor. Therefore, the TNM staging system by itself is not sufficient to effectively assess postoperative metastasis risk in UCB patients. In the modern medical decision-making process, the nomogram can combine multiple variables to predict the probability of the event we are interested in, showing broad application prospects.
More and more evidence has shown that nomograms show higher predictive accuracy than TNM staging. For BLCA, there were many nomograms designed to predict treatment response, postoperative survival, or recurrence. 17 For example, Welty et al. designed the Cancer of the Bladder Risk Assessment (COBRA), a new simplified prognostic model for predicting cancer-specific survival (CSS). 18 The C-index of the model was 0.712 and 0.705 in the development and validation groups, respectively. Antonio et al. constructed the Cancer of the bladder risk assessment (CRAB) nomogram for predicting disease-free survival (DFS) with an accuracy of 0.71. 19 Moreover, the novel nomogram recently established by Siteng Chen et al. based on pathomics signature and NLR can be used as excellent predictors of OS to facilitate individualized treatment of BLCA patients. 20 However, the endpoint of these study was not distant metastases, which are the leading cause of death in bladder cancer patients. Di Chen et al. constructed the nomogram related to bladder cancer metastasis with good accuracy by analysing the SEER database. 21 But, their model has only been validated internally and all of its included samples were BLCA patients with DM at the time of diagnosis, not those who subsequently developed DM. Therefore, the development of a nomogram model to predict DM after radical cystectomy for BLCA was also one of the main objectives of this study.
As the understanding of tumor immunology has gradually increased, the use of inflammation-related biomarkers for predicting the survival prognosis of patients with cancer has received increasing attention in recent years. For example, clinical researches have shown that NLR is effective as a prognostic factor for various malignancies.22,23 Although some reports have also suggested that inflammatory cell-related markers such as NLR are independent prognostic biomarkers in BLCA patients, it remains unknown whether these markers can be used as predictors of DM.24–26 In the current study, we identified preoperative NLR was an independent risk factor of postoperative metastasis in UCB patients.
As a DM-related independent risk factor, high NLR reflects the attenuation of lymphocyte-mediated antitumor immune responses. Lymphocytes, a key element of the immune response, can selectively attack cancer cells, and the relative decrease in lymphocytes may result in a large number of CTCs not being recognized and eliminated. 27 On the other hand, high NLR also reflects enhanced neutrophil-dependent pro-metastatic effects. The traditional view of neutrophils as inert bystanders has now been replaced. In fact, stimulated and activated neutrophils can influence other cells, including cancer cells. 28 During tumor progression and metastasis, neutrophils can be classified into tumor-associated neutrophils, circulating neutrophils and metastasis-associated neutrophils according to their location. The overall role of neutrophils in cancer is mediated by a combination of these three types of neutrophils. 29 In fact, circulating neutrophils have recently been found to have multiple pro-metastatic roles. 6 For example, neutrophils can come into close contact with CTCs and form CTCs-neutrophil clusters, and this intercellular interaction contributes to improve the survival rate and metastatic ability of CTCs. 30 Moreover, circulating neutrophils can inhibit the function of T cells and natural killer cells, leading to significantly increased intraluminal survival time of tumor cells.31,32 Lartigue et al. showed that neutrophil-generated neutrophil extracellular traps (NETs) facilitate the adhesion of CTCs to capillaries and subsequent extravasation to target organs. 33 Moreover, NETs are likely to be necessary to awaken dormant cancer cells at metastatic sites. 34 In short, NLR reflects combined prognostic information for neutrophils and lymphocytes and is a better predictor than either one alone.
Well-designed predictive models can facilitate communication between physicians and patients and avoid clinical over- or under-treatment by identifying truly high-risk patients. The most common histological type of BLCA is UCB. We combined T stage, N stage and NLR to establish a nomogram related to postoperative metastasis in UCB patients. The nomogram showed satisfactory predictive and calibration power. Specifically, the C-index of the nomogram in the two cohorts was 0.766 and 0.739, respectively. DCA had a higher clinical benefit than the TNM staging system. Unlike some prognostic models that incorporate parameters that are difficult or costly to obtain, the nomogram in this study performed well in the prediction of postoperative metastasis in UCB, and all the parameters were easily accessible, cost-effective, and easy to assess, which facilitates the popularization of this predictive model. In the process of standardised treatment of UCB throughout the course of treatment, it is crucial to make accurate categorical judgements about the progression of the patient's condition. The model can be used in clinical work to predict the risk of metastasis in UCB patients preoperatively and thus facilitate individualised treatment strategies for UCB patients. In addition, we grouped UCB patients based on the cutoff point of the nomogram scoring system. The Kaplan-Meier curve demonstrated that the probability for DM was significantly higher in the high-risk group than in the low-risk group in both cohorts. The clinical application of this classification method would help clinicians more effectively recognize patients with UCB who are at high risk for DM and choose most suitable treatment. Therefore, we believe that the nomogram established in this study will facilitate individualized therapy in UCB.
However, our research has its limitations. Firstly, the study is retrospective, which inevitably suffers from selection bias, so a prospective cohort study is needed to verify the utility for this nomogram. Secondly, with the guidelines recommending the administration of neo-adjuvant chemotherapy (NAC) prior to RC, this treatment will be adopted by an increasing number of patients. 35 The predictive value of the predictive model we developed in patients treated with NAC needs to be further evaluated. Finally, this study performed a temporal validation of the nomogram rather than a true external validation. Therefore, future studies should externally validate the nomogram in other institutions.
Conclusions
Preoperative NLR is an independent risk factor for DM in UCB patients undergoing radical cystectomy. The nomogram established by NLR, T-stage, and N-stage can accurately predict postoperative DMFS in UCB patients. Nomogram-based risk stratification helps to recognize UCB patients who are at high risk for DM and choose most suitable treatment.
Footnotes
Acknowledgments
This work was supported by Special fund project for doctoral training program of Lanzhou University Second Hospital [Grant No.YJS-BD-14], CuiYing Science and Technology Innovation plan project of Lanzhou University Second Hospital [Grant No.CY2021-ZD81], Medical Innovation and Development Project of Lanzhou University [Grant No.lzuyxcx-2022-105].
Author contributions
Conception: S-PF
Interpretation or analysis of data: Z-B, H-Y and Z-XX
Preparation of the manuscript: Z-XX and D-YL
Revision for important intellectual content: Z-XX, D-YL, X-W and S-PF
Supervision: S-PF
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.