What's known on the subject? and What does the study add?
Upstaging to non-organ-confined (NOC) disease is frequent at the time of radical cystectomy for urothelial carcinoma of the bladder (UCB). Pre-surgical models that can accurately predict which patients are likely to have more extensive disease are sparse.
The present study developed an accurate nomogram for the prediction of NOC-UCB based on a cohort of patients with clinically organ-confined disease. Adoption of such a tool into daily clinical decision-making may lead to more appropriate integration of perioperative chemotherapy, thereby potentially improving survival in patients with UCB.
To create an accurate pre-cystectomy decision-making tool that allows for the accurate identification of patients with clinically organ-confined urothelial carcinoma of the bladder (UCB) who have non-organ-confined UCB (NOC-UCB) at cystectomy, as identification of patients with UCB most likely to benefit from neoadjuvant chemotherapy (NACTx) is hampered by inaccurate clinical staging.
Patients and Methods
A prospectively maintained single-institution database containing 201 patients who underwent cystectomy and pelvic lymph node (LN) dissection without NACTx for UCB was analysed.
Predictive variables for NOC-UCB included, among others, age, gender, transurethral resection of bladder tumour (TURBT) findings (stage, grade, histology, size, presence of carcinoma in situ, lymphovascular invasion [LVI], multifocality), history of intravesical therapy, time from TURBT to cystectomy, and cross-sectional imaging findings.
Clinical stage distribution was 19 patients with Ta, 15 with Tis, 67 with T1, and 100 with T2.
At the time of cystectomy, NOC-UCB and LN-positive disease were found in 71 (35%) and 38 (19%) of patients, respectively; 81 (40%) of patients had NOC-UCB (≥pT3/Nany or pTany/N+).
Tumour stage (P [trend] <0.001), presence of LVI (odds ratio [OR] 5.2; P = 0.02), and radiographic evidence of NOC-UCB or hydronephrosis (OR 3.2; P = 0.01) were independently associated with ≥pT3 Nany UCB.
Tumour stage (P [trend] < 0.001) and presence of LVI (OR 6.64; P = 0.01) were independently associated with (≥pT3/Nany or pTany/N+) UCB.
A nomogram to predict (≥pT3/Nany or pTany/N+) based on all three variables was highly accurate (area under the curve 0.828) and well calibrated, deviating <8% from ideal prediction. Decision curve analysis showed net benefit across all threshold probabilities.
NOC-UCB can be predicted with high accuracy by integrating standard clinicopathological factors with imaging information.
This model may help to identify patients with NOC-UCB who may benefit from NACTx.
Radical cystectomy with bilateral pelvic lymph-node dissection (PLND) is the mainstay therapy for non-metastatic high-risk non-muscle-invasive and muscle-invasive urothelial carcinoma of the bladder (UCB), resulting in durable oncological control and long-term survival for most patients with organ-confined UCB. However, 42–44% of patients without lymph node (LN) metastases but with disease extending to the perivesical fat and 65–67% of those with regional LN metastases have disease recurrence ≤5 years after cystectomy [1, 2]. Disease recurrence often manifests as distant metastases , suggesting a high prevalence of early systemic dissemination.
Level I evidence supports the use of neoadjuvant chemotherapy (NACTx) for clinical (c) T2–T4aN0M0 UCB . However, to date, NACTx has been seldom used in the general community and may indeed be less commonly used than adjuvant chemotherapy. A recent multi-institutional study reported that only 12% of patients with cT2–T4aN0M0 received NACTx, whereas 22% received adjuvant systemic chemotherapy . The exact reasons for these practice patterns remain unclear, but under-usage of NACTx may be due to the belief that cystectomy alone will cure a significant proportion of patients who would incur unnecessary and avoidable side-effects of overtreatment with NACTx.
The mainstay of clinical staging for UCB, an integration of pathological examination of the transurethral resection (TUR) specimen, bimanual physical examination, and cross-sectional imaging studies, is highly inaccurate . It is not infrequent to find that the clinical stage is discordant with the pathological stage at cystectomy, with pathological upstaging to ≥pT3 and/or N+ disease occurring in up to 60% of patients [6-10]. Current predictive tools, e.g. nomograms, which are commonly used for outcome prediction in prostate cancer, do not substantially improve prediction of non-organ-confined (NOC)-UCB in patients with clinically organ-confined disease [11, 12]. This is in part due to the fact that they rely only on clinical stage and grade.
More accurate clinical staging would allow for the selection of patients most likely to benefit from NACTx, thereby possibly resulting in wider adoption of this therapy by the urology community, potentially improving survival. Therefore, our primary objective was to develop a preoperative multivariable model, integrating clinical, pathological, and imaging variables to accurately identify patients with clinically organ-confined UCB who are at increased risk for pathologically NOC-UCB. We hypothesised that NOC-UCB could be predicted with reasonable accuracy.
Patients and Methods
After Institutional Review Board approval, we retrospectively reviewed all prospectively collected data for 296 consecutive patients with bladder cancer who were treated with radical cystectomy or partial cystectomy and bilateral PLND. We excluded patients who had inadequate TUR of bladder tumour (TURBT) data (11 patients), patients with non-urothelial pathology (16), those who underwent NACTx (58), patients with clinical T3–T4 stage disease (seven) based on TURBT and bimanual examination, and those with a concomitant diagnosis of high-grade upper tract urothelial cancer (three). The remaining 201 patients were the subject of the present analysis; all had clinically localised disease.
All surgical specimens were processed according to standard pathological procedures at our institution and were histologically confirmed to be UCB by a dedicated genitourinary pathologist. If the TURBT procedure was performed at another institution, representative slides were obtained and reviewed before proceeding to cystectomy. Tumours were staged according to the 2002 American Joint Committee on Cancer/Union Internationale Contre le Cancer TNM classification . Tumour grading was assessed according to the 1998 WHO/International Society of Urological Pathology consensus classification . lymphovascular invasion (LVI) was defined as the unequivocal presence of tumour cells within an endothelium-lined space, with no underlying muscular walls .
To identify predictors of pathological stage at the time of cystectomy, we created multivariable logistic regression models to predict (≥pT3/Nany) and (≥pT3/Nany or pTany/N+). Discrimination was measured by area under the curve (AUC). Regression coefficients were used to develop a nomogram.
Decision curve analysis  was used to explore the clinical value of the model predicting (≥pT3/Nany or pTany/N+) NOC-UCB. Because the value of a true positive (i.e. detection of NOC-UCB and subsequent treatment with NACTx) may differ from the disadvantages resulting from a false positive (i.e. potential over-treatment with systemic chemotherapy), the net benefit differentially weights true and false positives by using the threshold probability at which one would opt for NACTx. For example, if a provider (or patient) would opt for NACTx with a 30% risk of NOC-UCB but would forgo NACTx with only a 29% risk, then the threshold probability is 30%. Clinically valuable models show a net benefit to alternative treatment strategies throughout the clinically applicable range of threshold probabilities. Alternative strategies include treating all patients with NACTx or treating no one with NACTx. For the purposes of this analysis, we assume that the preoperative identification of NOC-UCB would lead to treatment with NACTx. All statistical tests were two-sided with significance set at P < 0.05. Statistical tests were performed with SPSS® 18 (SPSS Inc., IBM Corp., Somers, NY, USA) and R-statistics (the R foundation for Statistical Computing, version 2.1.13).
The descriptive variables of the 201 evaluable patients are shown in Table 1. The median (range) age at the time of cystectomy was 72.9 (41–92) years and 165 (82.1%) patients were male. Overall, 71 (35.4%) patients had (≥pT3/Nany) NOC-UCB and 38 (18.9%) had LN metastases (N+). In all, 81 (40.3%) patients had (≥pT3/Nany or pTany/N+) NOC-UCB.
Table 1. Pre- and postoperative clinical and pathological features of 201 patients who underwent cystectomy with PLND for UCB
*Evaluable, n = 193; †evaluable, n = 147; ‡evaluable, n = 181; §evaluable, n = 155; abnormal imaging defined by presence of hydronephrosis and/or suggestion of NOC-UCB.
Tables 2 and 3 show the univariable and multivariable regression models, respectively, for the prediction of NOC-UCB. In univariable analyses, TURBT tumour size, TURBT stage, TURBT LVI, intravesical therapy history, and an abnormal preoperative imaging study were associated with NOC-UCB in both models. In multivariable analyses TURBT stage, TURBT LVI, and abnormal preoperative imaging were independently associated with (≥pT3/Nany) NOC-UCB, and stage and LVI were independently associated with (pTany/N+ or ≥pT3/Nany) NOC-UCB. The AUC of these two models was 0.836 and 0.828, respectively. For both models, preoperative imaging improved the accuracy of LVI and TURBT stage combined (AUC 0.807 and 0.809, respectively) by a statistically significant margin (P < 0.001).
Table 2. Univariable preoperative logistic regression analyses predicting NOC-UCB in patients treated with cystectomy and PLND
≥pT3/Nany or pTany/N+
OR (95% CI)
OR (95% CI)
*P value for trend. †Abnormal imaging defined by presence of hydronephrosis and/or suggestion of NOC-UCB. IVT, intravesical therapy; CIS, carcinoma in situ.
Figure 1 shows the nomogram and calibration plot for predicting (≥pT3/Nany or pTany/N+) NOC-UCB. The nomogram was well-calibrated deviating maximally 7.7% from ideal predictions. In the decision curve analysis (Fig. 2), the model predicting (≥pT3/Nany or pTany/N+) NOC-UCB provided net benefit throughout the entire range of threshold probabilities as compared with the strategy of treating all patients with NACTx, or alternatively, treating no one. Table 4 shows the clinical effects of using the prediction model to guide the decision to use NACTx as compared with treating all patients, or no patients with NACTx. For example, only administering NACTx to patients with a nomogram-predicted probability of NOC-UCB of ≥50% would result in 81 patients in our cohort receiving this treatment. In all, 27 (33%) of the 81 patients would be treated with NACTx in the absence of NOC-UCB, representing the ‘over-treated’ group. The remaining 54 patients of the 81 predicted to have NOC-UCB, and thus treated with NACTx, had pathologically confirmed NOC-UCB. Therefore, 54 of the 65 (83%) patients with pathologically confirmed NOC-UCB would have been ‘appropriately’ treated with NACTx and 11 of 65 (17%), would have been ‘undertreated’.
Table 4. Number of patients with and without pathologically confirmed NOC-UCB (pT3–4/pNany or pTany/pN+) who would receive NACTx at various threshold probabilities based on nomogram prediction; all patients with clinically organ-confined UCB
Probability of NOC-UCB required in order to treat with NACTx (threshold probability), %
Patients with pathologically confirmed NOC- UCB (n = 65)
Would receive NACTx based on nomogram prediction of NOC-UCB
Would receive NACTx based on nomogram prediction of NOC-UCB
*Number over-treated = (number of patients predicted to have NOC-UCB) − (number of patients with pathologically confirmed NOC-UCB). % over-treated = number over-treated/ number of patients predicted to have NOC-UCB. For a risk threshold of 30%, the number over-treated is 90 − 56 = 34, and % over-treated is 34/90 × 100 = 38%. Assumption is that prediction of NOC-UCB drives the decision to use NACTx. †Column describes the ‘undertreated’ group; those patients with pathologically NOC-UCB who were not treated with NACTx. The percentage is based upon a total of 65 patients with pathologically confirmed NOC-UCB.
We developed a highly accurate nomogram (AUC 0.828) to predict pathologically NOC-UCB in patients diagnosed with clinically organ-confined UCB; these patients often do not receive NACTx due to the concern of possible over-treatment . In agreement with the literature , we found a significant rate of upstaging to NOC-UCB at the time of cystectomy, with 14% of clinical stage ≤T1 and 67% of clinical stage T2 patients upstaged to (≥pT3/Nany or pTany/N+). The model in the present study, which includes T-stage, LVI found at TURBT, and abnormal preoperative imaging, shows improved accuracy compared with a previous study where separate nomograms were developed to predict NOC-UCB (Accuracy 75.7%) and LN metastases (Accuracy 63.1%), respectively . The first model included age at cystectomy, grade, stage, and concomitant carcinoma in situ and the second model included only stage and grade.
We found that abnormal imaging was a strong independent predictor of NOC-UCB. These findings support those of previous investigators such as Stimson et al. , who reported that preoperative hydronephrosis was independently associated with NOC-UCB and LN metastases at the time of cystectomy. Additionally, in a recently analysed upper-tract urothelial cohort, local invasion on preoperative imaging (odds ratio [OR] 4.34; P < 0.001) as well as ureteroscopic biopsy grade (OR 3.86; P = 0.003) independently predicted non-organ-confined (≥pT3 or pN+) disease at the time of radical nephroureterectomy .
We also found that LVI found at TURBT independently predicted NOC-UCB at cystectomy. Previously, we and others have shown the strong association between LVI and the rate of metastases to regional LNs, supporting the hypothesis that lymphatic vessel invasion precedes or occurs concurrently with LN metastasis [15, 20-24]. Resnick et al.  reported decreased recurrence-free survival and decreased overall survival in patients whose TURBT specimens contained evidence of LVI. Lotan et al.  showed that LVI at the time of cystectomy was associated with UCB recurrence (OR 2.02, P < 0.001), and decreased overall (HR 1.84, P < 0.001) and cause-specific survival (OR 2.07, P = 0.001) in pN0 patients only. In an external multi-institution validation cohort, Shariat et al.  confirmed that the presence of LVI in the cystectomy specimen was independently associated with disease recurrence and decreased cause-specific survival. The impact of LVI was most pronounced in the subset of patients who did not have LN metastases and for those patients, when LVI was added to a base model, the predictive accuracy for recurrence-free-survival and disease-specific survival was improved by a prognostically significant margin. Indeed, infiltration of these microscopic lymphovascular spaces by tumour cells is probably the initial entry of neoplastic cells into the circulation, prior to the development of fulminant LN metastasis.
To examine the potential clinical impact that the present predictive model provides, we performed a decision curve analysis, a technique that evaluates the clinical consequences of using predictive models [16, 26]. The use of the present model to predict (≥pT3/Nany or pTany/N+) NOC-UCB for the purpose of guiding the use of NACTx, provided a net benefit relative to the two strategies of treating all patients with NACTx, or alternatively, treating no one (Fig. 2). This was true across almost the entire range of threshold probabilities.
A recent study reported that only 12% of patients with cT2–T4aN0M0 received NACTx  showing that the threshold probability for urologists to use NACTx may be very high. In other words, many urologists probably require a very high degree of suspicion that NOC-UCB is present before treating a patient with NACTx, showing a very low risk-tolerance for what they may perceive as overtreatment of organ-confined UCB. Using the present predictive model (Fig. 1) in our cohort, a highly risk-intolerant clinician (threshold probability 90%) would administer NACTx to 12 patients, all of whom would in fact have pathologically confirmed NOC-UCB (Table 4).
Lastly, the present nomogram predicts that for patients with clinically ≤T1 UCB and the presence of TURBT LVI and/or abnormal imaging, a significant number of patients will have NOC-UCB at cystectomy. Current European Association of Urology guidelines recommend NACTx for all patients with cT2–T4N0M0 UCB . However, there is no such recommendation for patients with clinically non-muscle-invasive UCB. While treating all cT2–T4N0M0 patients with NACTx may over-treat a significant number of patients, excluding all patients with <cT2 UCB from NACTx without consideration given to features such as LVI on TURBT or abnormal preoperative imaging, may deny some of these patients optimal therapy.
The present study is not without its limitations including, but not limited to the retrospective analysis and the small cohort. For example, there is inherent difficulty in determining the presence of LVI at the morphological level with significant differences between local pathologists and central pathology review . Additionally, repeat TUR was not done for every patient, but rather was performed at the discretion of the surgeon, which could have contributed to under-staging. For example, 13 of 67 patients with T1 UCB at TURBT had high-grade recurrent and/or BCG refractory UCB without muscularis propria in the specimen and were directed to cystectomy. Moreover, some data points that have been analysed by other authors were not included such as the frequency of variant UCB histology . Lastly, imaging protocols were not standardised, nor was the use of i.v. contrast for CT or MRI.
In conclusion, the current conventional clinical staging for patients with UCB before cystectomy results in a high rate of staging inaccuracy; therefore more accurate tools are needed, particularly for the purpose of efficaciously directing patients towards NACTx. The present study provides such a tool and reiterates that LVI found at TURBT and abnormal preoperative imaging findings can predict NOC-UCB at the time of cystectomy in patients with presumed clinically localised UCB. Decision curve analysis of the nomogram shows clinical benefit both for clinicians hesitant to use NACTx and for those who liberally recommend for its use. Adoption of such a tool into daily clinical decision-making may lead to more appropriate integration of perioperative chemotherapy, thereby potentially improving survival in patients with UCB. The clinical value of this model needs to be further assessed in external multi-institutional validation cohorts.
Source of Funding
Supported by Frederick J. and Theresa Dow Wallace fund of the New York Community Trust.