Development and external validation of a novel nomogram to predict intravesical recurrence after radical nephroureterectomy: a multicenter study

Abstract

Objective

This study aimed to establish and validate nomograms to predict the probability of intravesical recurrence (IVR) after radical nephroureterectomy (RNU) for upper urinary tract epithelial carcinoma (UTUC).

Methods

Clinical data of 528 patients with UTUC after RNU were collected from two medical centers between 2009 and 2020. We used the least absolute shrinkage and selection operator (LASSO) regression to select variables for multivariable Cox regression analysis in the training cohort and included independent risk factors into nomogram models predicting IVR-free survival (IVRFS). Another center was applied as the external cohort to validate the predictive accuracy and discriminative ability of the nomogram by performing area under the receiver operating curve (AUC), consistency index (C-index), and calibration curve.

Results

History of bladder cancer, tumor size, preoperative urine cytology, postoperative instillation, Ki-67, and platelet-to-lymphocyte ratio (PLR) were identified as independent risk factors for IVR. The prognosis model including these predictors demonstrated excellent discriminatory performance in both the training cohort (C-index, 0.814) and external validation cohort (C-index, 0.748). The calibration plots of the nomogram revealed good consistency in both cohorts. Finally, patients could be classified into two risk groups based on scores obtained from the nomogram, with significant differences in IVRFS.

Conclusion

Our study provided a reliable nomogram for predicting the probability of IVR in patients with UTUC after RNU. Risk stratification based on this model may assist urologists make optimal clinical decisions on the management of UTUC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00432-023-05016-2.

Introduction

Upper tract urothelial carcinoma (UTUC) is a malignant tumor that occurs in the urothelial cells of the renal pelvis or ureter. Its incidence rate is far lower than that of bladder cancer, accounting for 5–10% of all urothelial carcinomas (Rouprêt et al. 2021; Siegel et al. 2022). Compared with other urological tumors, UTUC has a more rapid clinical progression and a worse prognosis. Radical nephroureterectomy (RNU) with bladder sleeve resection is the standard surgical treatment for UTUC, especially for high-risk patients (Ito et al. 2013). However, it has been reported that intravesical recurrence (IVR) occurs in 15–50% of all patients with UTUC who underwent RNU, and with the majority occurring within 1 year after surgery (Raman and Park 2017). Once IVR occurs, patients have to undergo additional adjuvant treatment, such as transurethral bladder tumor resection surgery and intravesical instillation chemotherapy (Doeveren et al. 2018; Zhou et al. 2019). Therefore, IVR could cause a worse survival prognosis and significantly increase the psychological stress and financial burden of patients, severely affecting their quality of life (Zamboni et al. 2019).

Although the history of bladder cancer and tumor multifocality are the most commonly reported risk factors, factors that reliably predict IVR have not been identified (Azémar et al. 2011). Thus, how to leverage patient stratification tools to better identify patients who may have IVR at an early stage has become a timely issue to be addressed (Loizzo et al. 2022). Xylinas et al. constructed prediction models based on a larger population from Western countries; however, the discriminatory power of these models was not satisfactory in external validation and was only applicable to specific populations (Xylinas et al. 2014). For the Chinese population, there are significant differences in race, dietary habits, lifestyle choices, and exposure to common carcinogenic factors, especially aristolochic acid, compared with patients in Western countries (Chen et al. 2013; Singla et al. 2017). In this study, risk factors for IVR in patients with UTUC were identified. We developed a nomogram and conducted external validation, and hope to assist the urologist to optimize clinical decision-making.

Materials and methods

Patients selection

This retrospective study was approved by the Institutional Research Ethics Committee of China–Japan Friendship Hospital and Beijing Chaoyang Hospital. Informed consent was obtained from all eligible participants in advance. This work has been reported in line with the STROCSS guideline (Agha et al. 2019). The information of patients diagnosed with UTUC who received RNU treatment in the two hospitals from January 2009 to December 2020 were collected retrospectively, and all patient details have been de-identified. We included the patients who met the following criteria: (1) patients with UTUC confirmed pathologically; (2) patients with primary disease; (3) patients with unilateral onset; and (4) patients subject to RNU combined with cystic sleeve resection. We excluded patients according to the following criteria: (1) patients with bilateral UTUC; (2) patients subject to no RNU combined with cystectomy; and (3) patients with metastatic uroepithelial carcinoma.

Data collection

We obtained the following clinicopathological features of patients for analysis: sex, age at the first diagnosis, body mass index (BMI), tumor laterality, pathologic tumor stage (pT), lymph node status (pN0, pNx or pN+), tumor grade, tumor size, tumor multifocality, margin positivity, and lymphovascular invasion (LVI). Patients with previous history of hypertension, diabetes, and bladder cancer, expression of Ki-67, preoperative urine cytology, presence of hydronephrosis on the affected side, requiring postoperative instillation, and high serum creatinine level were also enrolled. We specially collected some blood inflammation biomarkers including neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and lymphocyte-to-monocyte ratio (LMR), which were obtained from routine blood examination performed 1 week before surgery. All resected tumor specimens were sent for pathological examination by senior pathologists. The pathologic tumor stage and grade were evaluated according to the 2009 International Union Against Cancer (UICC)/American Joint Committee on Cancer (AJCC) TNM classification system (Edge and Compton 2010) and the 2016 World Health Organization (WHO) grading system (Humphrey et al. 2016).

Follow-up

We monitored patients every 3 months during the first 2 years after surgery and every 6 months thereafter until 5 years. Follow-up data included blood tests, cystoscopic examination, urinary system ultrasound, chest and abdomen CT, urine exfoliated cytology, and urography. Selective bone scan, PET/CT or magnetic resonance (MRI) examination was performed if clinically indicated. Overall survival (OS) was defined as the time from the date of RNU to death from any cause. Intravesical recurrence-free survival (IVRFS) was defined as the time from the date of RNU to the date of the first IVR according to cystoscopic examination.

Nomogram development and validation

Patients of Beijing Chaoyang Hospital were enrolled in the training cohort and validated in the cohort of China–Japan Friendship Hospital. The least absolute shrinkage selection operator (LASSO) regression was utilized to screen variables and prevent overfitting. Multivariate Cox regression analyses were performed in the training cohort to determine the potential independent risk factors. Based on the prognostic model, we developed a new nomogram to predict IVR at 1-, 3-, and 5-year intervals in patients with UTUC who underwent RNU. Receiver operating characteristic (ROC) analyses were conducted in both cohorts, and the area under the ROC curve (AUC) and concordance index (C-index) were used to evaluate discrimination ability. To validate the accuracy of the model, calibration curves were drawn from 1000 bootstrap samples from the training and validation cohorts.

Statistical analysis

The categorical variables were expressed as the frequency (percentage). Continuous variables were presented as median values with ranges. We utilized the X-tile software (version 3.6.1) to determine the optimal cutoff for these biomarkers classified as low and high level (Camp et al. 2004). The cutoff values of NLR, PLR, and LMR were 5.0, 166.0, and 4.2. Pearson’s Chi-square test and non-parametric U test were used to compare the two groups. Kaplan–Meier survival curves were plotted and the log-rank test conducted to demonstrate the difference in different risk groups. Prognostic risk factors related to IVR were assessed by univariate and multivariate Cox regression analyses models, and the results were shown as hazard ratio (HR) with 95% confidence interval (CI). All p values were obtained from two-sided tests, and p < 0.05 indicated that the difference was statistically significant. R software (Version 4.1.2) and IBM SPSS Statistics (Version 24) were utilized to complete all statistical analyses and figures.

Results

Patient characteristics

A total of 528 patients matched the inclusion and exclusion criteria and included, with 308 patients in the training cohort and 220 patients in the external validation cohort. The demographic and clinicopathological variables are summarized in Table 1. 61 patients (19.8%) in the training cohort and 58 patients (26.4%) in the validation cohort developed IVR after RNU. There were no significant differences between the two cohorts except for the tumor location, pT stage, grade, tumor size, multifocality, postoperative instillation, PLR, and LMR. The median follow-up of the training cohort was 41 months (range 1–131 months), which was similar to the validation cohort (42 months, 2–143 months).

Table 1.

Characteristics of patients from two hospitals

Characteristic	Training cohort	Validation cohort	p value
	N = 308	N = 220
Age (years)	67 (33–88)	68 (38–86)	0.896
Sex			0.221
Male	155 (50.3%)	98 (44.5%)
Female	153 (49.7%)	122 (55.5%)
BMI	24.77 (14.54–34.69)	24.69 (16.44–33.13)	0.391
History of hypertension			0.988
Yes	154 (50.0%)	109 (49.5%)
No	154 (50.0%)	111 (50.5%)
History of diabetes			0.386
Yes	77 (25.0%)	47 (21.4%)
No	231 (75.0%)	173 (78.6%)
History of bladder cancer			0.743
Yes	15 (4.9%)	13 (5.9%)
No	293 (95.1%)	207 (94.1%)
Tumor location			0.004
Ureter	121 (39.3%)	110 (50.0%)
Renal pelvis	147 (47.7%)	98 (44.5%)
Both	40 (13.0%)	12 (5.5%)
Laterality			0.401
Left	154 (50.0%)	119 (54.1%)
Right	154 (50.0%)	101 (45.9%)
Pathologic stage			< 0.001
pT2 or less	183 (59.4%)	166 (75.5%)
pT3 or greater	125 (40.6%)	54 (24.5%)
Lymph node status			1.000
pN0/pNx	289 (93.8%)	207 (94.0%)
pN+	19 (6.2%)	13 (6.0%)
Tumor grade			< 0.001
Low	120 (39.0%)	22 (10.0%)
High	188 (61.0%)	198 (90.0%)
Tumor size			0.009
< 3 cm	154 (50.0%)	136 (61.8%)
≥ 3 cm	154 (50.0%)	84 (38.2%)
Tumor multifocality			0.001
Present	110 (35.7%)	49 (22.3%)
Absent	198 (64.3%)	171 (77.7%)
Hydronephrosis			0.258
Present	99 (32.1%)	82 (37.3%)
Absent	209 (67.9%)	138 (62.7%)
LVI			0.396
Present	62 (20.1%)	37 (16.8%)
Absent	246 (79.9%)	183 (83.2%)
Margin positive			0.540
Negative	289 (93.8%)	210 (95.5%)
Positive	19 (6.2%)	10 (4.5%)
Urine cytology			0.712
Normal	166 (53.9%)	123 (55.9%)
Abnormal	142 (46.1%)	97 (44.1%)
Postoperative instillation			0.001
Yes	105 (34.1%)	107 (48.6%)
No	203 (65.9%)	113 (51.4%)
Ki-67			0.276
< 20%	152 (49.4%)	120 (54.5%)
≥ 20%	156 (50.6%)	100 (45.5%)
Creatinine (mmol/L)	85.5 (34.9–1077.2)	90.9 (32.6–836.8)	0.090
NLR	2.49 (0.75–16.61)	2.29 (0.67–16.74)	0.600
PLR	133.35 (27.56–379.58)	100.57 (35.99–310.54)	< 0.001
LMR	4.04 (0.58–11.14)	4.03 (0.76–9.79)	0.015
Follow-up (months)	41 (1–131)	42 (2–143)	0.411

BMI body mass index, LVI lymphovascular invasion, NLR neutrophil-to-lymphocyte ratio, PLR platelet-to-lymphocyte ratio, LMR lymphocyte-monocyte ratio

Identification of risk factors for IVR

LASSO regression and multivariate analysis were applied to identify the prognostic risk factors related to IVR. In LASSO regression analysis, variables with non-zero coefficients at λmin were screened out in the training cohort (Fig. S1). Further multivariate Cox regression analysis revealed history of bladder cancer, tumor size, preoperative urine cytology, postoperative instillation, Ki-67, and PLR to be independent prognostic factors for IVR (Table S1).

Development and external validation of the nomogram for IVR

We constructed the nomogram based on the independent risk factors identified by multivariate analysis. The nomogram for predicting IVR is shown in Fig. 1. At the top of the nomogram, a scale provides a score for each prognostic variable. According to the sum of all scores axis at the bottom of the nomogram, probabilities of 1-, 3-, and 5-year IVRFS were able to be estimated. The median risk score of patients in the training cohort was determined as the cutoff score. All patients with UTUC were divided into two risk subgroups by the cutoff score: low risk (scores ≤ 83) and high risk (scores > 83). Kaplan–Meier curves showed shorter IVRFS in the high-risk group than in the low-risk group in both the training and validation cohorts (Fig. 2). To validate the accuracy and predictive ability of the nomograms, we used time-dependent ROC analysis and C-index. In the training group, the AUC values for 1-, 3-, and 5-year IVRFS were 0.820, 0.776, and 0.823, respectively, while in the validation group, they were 0.748, 0.766, and 0.732 (Fig. 3). In addition, the C-index of nomograms for IVR was 0.814 in the training group and 0.748 in the validation group. These results indicated that the nomogram provides a good discriminatory performance. Finally, the calibration curves for the training and validation group revealed that the IVR predicted by the nomogram was consistent with the actual observation results in Fig. 4.

Fig. 1.

Nomogram for predicting 1-,3-, and 5-year IVRFS

Fig. 2.

The Kaplan–Meier curves of nomogram in the training cohort (A) and validation cohort (B)

Fig. 3.

Nomogram ROC curves to predict 1-,3-, and 5-year IVRFS in the training cohort (A) and validation cohort (B)

Fig. 4.

The calibration curves of nomogram at 1,3, and 5 years in the training cohort (A) and validation cohort (B)

Discussion

The high incidence of IVR after RNU is a characteristic of UTUC, which has a significant impact on the survival outcome of patients (Rouprêt et al. 2021; Jiang et al. 2020). The modern management of UTUC patients should aim at customized methods to minimize the risk of tumor recurrence. Ishioka et al. built a risk stratification model for IVR based on patients with UTUC at 13 institutions in Japan (Ishioka et al. 2015). Higher tumor stage (≥ pT2), papillary tumor architecture, male sex, and LVI were identified as the risk factors and included in the model. According to the number of risk factors, patients were classified into different risk groups. Although the model is simple and practical, the discriminatory performance is poor with a C-index of 0.600. Some scholars thought it is not sufficiently effective for clinical decision-making (Laguna 2015). Besides, Xylinas et al. constructed nomograms to calculate the probabilities of IVR after RNU based on a larger population with 1839 patients (Xylinas et al. 2014). However, some predictors such as surgical access, the distal ureter management technique, age, and carcinoma in situ are not significantly associated with IVR in external validation, which reduced the calibration of nomograms. In this study, in addition to a history of bladder, cytology, chemotherapy, and tumor size, which have been reported to be risk factors for IVR (Ito et al. 2013; Azémar et al. 2011; Tanaka et al. 2014; Pieras et al. 2010), we also identified Ki-67 and PLR as independent predictors for IVR and added them to the nomogram, which acquired ideal discriminatory ability in both cohorts.

The pathophysiological mechanism of IVR is still unclear. There are currently two mainstream academic views, namely, ‘tumor seeding theory’ and ‘field-cancerization hypothesis’ (Habuchi et al. 1993; Li et al. 2020). Due to the complex pathogenesis of IVR, many scholars argued that the early IVR after surgery could be attributed to the dissemination and implantation of primary cancer cells, which is a true progression of disease recurrence. The late IVR may be a novel uroepithelial carcinoma induced by multiple carcinogens, unrelated to the primary tumor (Fang et al. 2014). The predictors in our nomogram echoed the two viewpoints. The positive urinary exfoliative cytology indicates a high risk of dissemination and implantation of tumor cells in the bladder (Raman et al. 2007). The history of bladder cancer could reflect the IVR is not caused by the primary tumor, but by independent gene alterations in different parts of the urothelium (Jones et al. 2005). Certainly, more research is required to reveal the detailed mechanism.

As an immunostaining marker of nuclear cell proliferation, Ki-67 was found to be frequently expressed in several malignancies such as breast, colon, and ovarian cancers (Krabbe et al. 2015; Wu et al. 2015). High expression rates of Ki-67 represents more aggressive and active proliferation of cancer cells, suggesting that primary tumors are more likely to progress and metastasize distantly (Yang et al. 2018). Krabbe et al. reported that the overexpression of Ki-67 was associated with poor prognosis in UTUC patients treated with RNU (Krabbe et al. 2015). However, the exact mechanism of Ki-67 in IVR remains controversial. Long et al. mentioned patients with low Ki-67 levels were more sensitive to intravesical instillations for IVR prevention after RNU (Long et al. 2016). By contrast, Lai et al. proposed that low-level Ki-67 expression was an independent predictor of IVR (Lai et al. 2021). In our study, we defined the Ki-67 overexpression as 20% after referring to relevant literature (Tsai et al. 2021; Fan et al. 2016). The results confirmed that the Ki-67 overexpression was closely related to IVR. We speculated that tumors with higher rate of Ki-67 overexpression are more prone to produce exfoliative cancer cell, thus increasing the risk of bladder implantation. However, the specific pathophysiological mechanisms remain to be further explored.

Recent research has demonstrated that inflammatory effects in the tumor microenvironment play an important role in the development, progression, and metastasis of human cancer (Mantovani et al. 2008; Hanahan and Weinberg 2011). The body’s immune response to malignancy may lead to changes in the levels of lymphocytes, neutrophils, monocytes, or platelets in peripheral blood. Our study showed that higher NLR is also associated with a higher risk of IVR. Platelets are generally considered to be mainly involved in the process of physiological hemostasis. However, during tumor cell migration, platelets can play a role in chemotactic integration, causing the immune escape of tumor cells and promoting the spread of cancer cells (Gresele et al. 2017). In addition, platelets can secrete protein aggregation factors and increase the adhesion between tumor cells and vascular endothelial cells, thus promoting the metastasis of tumor cells (Denis et al. 2005). In contrast, the level of lymphocytes reflects the cell-mediated immune response (Massagué and Obenauf 2016). Tumor-infiltrating lymphocytes play an important anti-tumor immune role in the tumor microenvironment. The study of Gooden et al. showed that a high level of tumor lymphocyte infiltration affects the prognosis of cancer patients (Gooden et al. 2011). The reduction of lymphocyte count in peripheral blood weakens the body's immune response ability, leading to immune tolerance and decreased host immune surveillance, which in turn lead to poor survival of tumor patients. Therefore, elevated PLR not only represents a stronger inflammatory response in patients with malignancy, but also reflects the impairment of cell-mediated immune function, thus providing a theoretical basis for PLR to predict the prognosis of tumor patients. The cutoff values in the present study were determined by X-tile software and are consistent with the ranges reported by previous studies (Shao et al. 2020). However, the results should be interpreted with more caution due to the differences in cutoff values.

The present study had several strengths as follows: first, the nomogram performed well in external validation, which increased credibility; in addition, the follow-up was relatively long and the sample size was relative large compared to previous studies; third, the PLR can be obtained before surgery compared to other risk factors, and thus guide further treatment reasonably. However, the current study had several limitations. First of all, an inherent bias was inevitable due to the design of the retrospective study. Moreover, due to the relatively long time span, the criteria related to blood tests, urine exfoliation cytology, and pathological examinations may have changed during this period, further affecting the reliability of the collected data. Additionally, this study involved two hospitals and the difference in surgical techniques, medical equipment, and the postoperative follow-up of patients should be considered. Finally, this nomogram was constructed based on the Chinese population and additional medical centers in other countries are needed to validate the generalizability of the model.

Conclusion

In summary, we developed a reliable nomogram based on clinical and pathological variables to predict the probability of IVR in patients with UTUC after RNN. Risk stratification based on the model may be an effective tool to assist clinicians in assessing individualized IVR rates and making optimal treatment decisions. Further prospective multicenter trials in different countries are required to validate our nomogram.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Study design: ZD, GZ. Data collection: TH, HZ, BJ. Data analysis and interpretation: WH, YB, ZL. Manuscript writing: ZL, BJ.

Funding

No funding was received.

Data availability statement

All the data were presented in the manuscript. No additional data are available.

Declarations

Conflict of interest

The authors declare that this study has received no financial support and there was no conflict of interest.

Statement of ethics

The studies involving human participants were reviewed and approved by the China–Japan Friendship Hospital and Beijing Chaoyang Hospital. The patients/participants provided their written informed consent to participate in this study.