We identified 1378 high-risk UTUC patients, according to the European Association of Urology (EAU) guidelines, from a prospectively maintained database involving eight European referral centers. Surveillance protocol was based on cystoscopies and cross-sectional imaging, as per the EAU guidelines.
First, we evaluated the noncumulative risk of bladder and other-site recurrences (including distant metastasis and locoregional relapse) against the follow-up time points, as suggested by the current EAU guidelines. Second, in an effort to identify the time points when the risk of other-cause mortality (OCM) exceeded that of recurrence and follow-up might be discontinued, we relied on adjusted Weibull regression.
The median follow-up was 4 yr. A total of 427 and 951 patients with and without a prior BCa history, respectively, were considered. At 5-yr, the time point after which cystoscopies should be performed semiannually, the bladder recurrence risk was 10%; at 4 yr, the bladder recurrence risk was 13%. At 2 yr, the time point after which imaging should be obtained semiannually, the nonbladder recurrence risk was 42% in case of nonprior BCa and 47% in case of prior BCa; at 4 yr, the nonbladder recurrence risk was 23%. Among patients without a prior BCa history, individuals younger than 60 yr should continue both cystoscopies and imaging beyond 10 yr from RNU, 70–79-yr-old patients should continue only imaging beyond 10 yr, while patients older than 80 yr might discontinue oncologic surveillance because of an increased risk of OCM. Limitations include the fact that patients were treated and surveilled over a relatively long period of time.
We suggest intensifying the frequency of imaging to semiannual till the 4th year after RNU, the time point after which the risk of recurrence was almost halved. Cystoscopies could be obtained annually from the 4th year given a similar risk of recurrence at 4 and 5 yr after RNU. Oncologic surveillance could be discontinued in some cases in the absence of a prior BCa history.
In this study, we propose a revision of the current guidelines regarding surveillance protocols following radical nephroureterectomy. We also evaluated whether oncologic surveillance for high-risk upper tract urothelial carcinoma could be discontinued and, if so, in what circumstances.
Upper tract urothelial carcinoma (UTUC) is a relatively rare disease with an estimated annual incidence of two cases per 100 000 . Despite radical treatment, oncologic outcomes are still unsatisfactory . The risk of recurrence during follow-up, such as bladder, local, or distant recurrence can be as high as 47%, 18%, and 17%, respectively . Given these relatively high rates of recurrence, a stringent surveillance protocol based on cystoscopies and imaging is warranted in an effort to achieve timely detection of recurrence and treatment.
The European Association of Urology (EAU) classifies UTUC patients into two categories (low-risk and high-risk [HR] disease ) and suggests different schemes for oncologic surveillance following radical nephroureterectomy (RNU) for each group. Presently, sparse evidence exists regarding the appropriate surveillance scheme for HR UTUC patients in terms of frequency or duration, and conclusions from these studies are limited by their relatively small sample size or single-center nature . Additionally, in case of a lack of UTUC-specific evidence, recommendations are extrapolated from bladder cancer (BCa) surveillance protocols, leaving room for improvement. As an example, despite several studies recording an increased risk of tumor recurrence among UTUC patients with a prior history of BCa , surveillance protocols are not stratified according to a prior BCa history; yet, considering UTUC patients as a homogeneous population with an equal risk of recurrence regardless of the previous BCa history may not be appropriate . Furthermore, while following radical cystectomy (RC) for BCa, cross-sectional imaging should be obtained semiannually till the 3rd year and annually thereafter , the recommended frequency of imaging following RNU is semiannually till the 2nd year and annually thereafter. Given a higher rate of recurrence following RNU than that following RC , this schedule does not seem to be justified. Finally, economical costs should be taken into account, as well as radiation exposure related to follow-up protocols . In consequence, oncologic surveillance might be interrupted when the recurrence risk falls below a certain threshold without compromising oncologic outcomes.
Following these premises, we relied on a large multi-institutional database of patients treated with RNU and followed up according to the EAU guidelines. We aimed to evaluate (1) whether the current frequency of cystoscopies and imaging is appropriate or should be revised, (2) when oncologic surveillance can be deintensified or interrupted after RNU, and (3) whether patients with a prior BCa history should be followed up differently, given the higher risk of tumor recurrence than in patients without previous BCa . The ultimate goal of the study was to define an individualized risk-based surveillance protocol for HR UTUC patients.
After the approval from each institution’s review board, we relied on a prospectively maintained multi-institutional dataset, including 1378 HR UTUC patients treated with RNU with bladder cuff excision at eight European academic centers between 1992 and 2016. HR UTUC was defined, according to the EAU guidelines, as follows: hydronephrosis, clinical tumor size >2 cm, high-grade cytology or biopsy, multifocal disease, previous RC, or variant histology (VH) . All individuals were staged preoperatively with pelvic/abdominal computed tomography (CT) scan, chest x-ray, and bone scan (when indicated). Regional lymph node dissection was left at physician’s discretion . Neoadjuvant (NAC) and adjuvant AC) chemotherapy doses were administered according to disease characteristics, as well as patients’ and physicians’ preferences . Individuals’ health status was assessed according to the Charlson Comorbidity Index (CCI) , which was determined preoperatively without including the presence of a solid neoplasm in the score.
Follow-up consisted of cross-sectional imaging, outpatient cystoscopies, routine blood tests, and urinary cytology, according to the EAU guidelines for HR UTUC . Specifically, cystoscopies were performed every 3 mo till the 2nd year, semiannually till the 5th, and then annually thereafter. Imaging (including abdominal CT scan or magnetic resonance imaging) was obtained semiannually till the 2nd year and annually thereafter. In case of uncertain diagnosis of recurrence, abdominal/pelvic ultrasound and bone scan were additionally ordered on a case-to-case basis. Percutaneous, laparoscopic, or open biopsy confirmation of recurrence was pursued only in those patients for whom diagnostic accuracy could not be achieved with imaging alone.
Overall survival was defined as freedom from death from any cause. Deaths ascribable to UTUC were coded as cancer-specific mortality (CSM) events, whereas deaths attributable to other causes were coded as other-cause mortality (OCM) events. For the purpose of our study, tumor recurrence was categorized as either bladder recurrence or other-site recurrence. The former was defined as the presence of a suspicious tumor detected on cystoscopy or imaging, and then confirmed at transurethral resection of the bladder. In case of suspected urinary cytology and negative imaging or cystoscopy, a blue light cystoscopy was performed. Other-site recurrence was defined as local or distant recurrence; local recurrence included a recurrence in the regional lymph nodes and/or in the renal fossa. Time to recurrence and time to event, as either CSM or OCM, were evaluated from the time of RNU. Recurrences detected simultaneously were counted individually as site-specific events.
Descriptive statistics were obtained: frequencies and proportions were reported for categorical variables, while medians and interquartile ranges (IQRs) were reported for continuous variables. Differences among medians and frequencies were estimated with the Kruskal-Wallis test and the chi-square test, respectively. The Kaplan-Meier estimator was applied to evaluate the time to the event of interest, while the inverse Kaplan-Meier method was used to estimate the follow-up among survivors.
First, we evaluated the noncumulative distribution of recurrence against time from RNU in an effort to verify whether the current follow-up schedule is appropriate or should be revised. Second, we aimed to evaluate whether follow-up could be interrupted for certain patients. In order to predict recurrence and OCM during follow-up, we relied on the Weibull regression, which represents a parametric time-dependent model based on the Weibull distribution. Since the risk of recurrence following surgery usually decreases with time and the one of OCM usually increases, the use of a parametric survival model that allows for the hazard to change over time is ideal for this estimation, and it is an accepted method for evaluating the risk of dying from a competing event versus the one of succumbing to the disease over time .
In an effort to estimate the recurrence risk over time, patients were stratified according to pathologic stage (pTa/1/CISN0, pT2N0, pT3/4N0, and pTanyN+) as well as by recurrence site (bladder vs other sites). The risk of OCM was estimated by stratifying patients according to age at the time of surgery (60, 60–69, 70–79, and >80 yr), as well as by CCI (0–1 vs ≥2). The risk of the occurrence of the event of interest was estimated after adjusting for sex , VH , surgical margin status , and administration of NAC or AC. In order to partly account for the fact that patients were treated at different institutions, we incorporated center clustering in the regression using the cluster option in Stata (Stata Corp., College Station, TX, USA). All analyses were conducted separately for patients with and without a prior history of BCa.
The OCM and recurrence hazards obtained from the Weibull models were graphed against the time from RNU in order to identify the statistical interactions between the models, which marked the time points when the risk of OCM exceeded the risk of recurrence. From a practical standpoint, when the risk of OCM exceeds that of recurrence, it could be reasonable to deintensify the follow-up strategy for a specific anatomical region (bladder vs other sites) or discuss interrupting oncologic surveillance. To allow for a better use in clinical practice, the time points were rounded to the closest year or half year.
Overall, 1378 HR UTUC patients were identified: 427 individuals had a prior history of BCa, and among them 153 (36%) had undergone a previous RC (). Overall, the median age at RNU was 69 yr (IQR: 61–76). Among patients without a prior BCa history, 323 (34%), 123, (13%), 346 (36%), and 159 (17%) fell in the pTa/1/CISN0, pT2N0, pT3/4N0, and pTanyN+ categories, respectively. Among individuals with a prior BCa history, 186 (43%), 79, (19%), 126 (30%), and 36 (8%) fell in the pTa/1/CISN0, pT2N0, pT3/4N0, and pTanyN+ categories, respectively.
The median follow-up among survivors was 4 yr. Overall, among individuals without prior BCa, 73 died of OCM (25 with a CCI of 0–1 and 48 with CCI ≥2); 197 patients experienced bladder recurrence and 376 showed other-site recurrence. The 5-yr recurrence-free survival rates were 70% and 50% for bladder and other-site recurrences, respectively. The 5-yr OCM-free survival rates were 93% and 83% for patients with CCI 0–1 and CCI ≥2, respectively.
Overall, among individuals with prior BCa, 45 died of OCM (13 with a CCI of 0–1 and 32 with CCI ≥2); 127 and 190 individuals experienced bladder and other-site recurrences, respectively. The 5-yr recurrence rates were 59% and 43% for bladder and other-site recurrences, respectively. The 5-yr OCM-free survival rates were 90% and 76% for patients with CCI 0–1 and CCI ≥2, respectively. The sites of recurrence stratified according to pathologic stage and BCa history are summarized in
(A and C) Noncumulative risk of bladder recurrence and (B and D) other-site recurrence (resection bed, regional lymph nodes, and distant metastasis) among patients treated with radical nephroureterectomy for high-risk upper tract urothelial carcinoma, stratified according to prior bladder cancer history. Vertical lines correspond to the intensity of cystoscopies or cross-sectional imaging as per EAU guidelines. (A and B) Risk of recurrence at the time point corresponding to deintensification of cystoscopies (5 yr) or imaging (2 yr). (C and D) Hypothetical surveillance scheme with more intense cross-sectional imaging frequency, that is, semiannual imaging till the 4th year, and less intense cystoscopy scheme, that is, annually from the 4th year. Numbers at risk are from Kaplan-Meier analysis. BCa = bladder cancer; EAU = European Association of Urology; RNU = radical nephroureterectomy.
The noncumulative risks of bladder and other-site recurrences stratified according to the prior BCa history are depicted in . The vertical lines indicate the time of when cystoscopies (A) or cross-sectional imaging (B) should be performed, according to the EAU guidelines. For the analysis focusing on bladder recurrence, individuals with a history of BCa treated with RC ( = 153 [36%]) were excluded.
Overall, the risk of any recurrence was higher in case of prior BCa. After 2 yr, imaging should be deintensified and performed semiannually, as per EAU guidelines. At 2 yr, the risk of nonbladder recurrence was 42% in cases with nonprior BCa and 47% in cases with prior BCa, as shown in B. After 5 yr, cystoscopies should be deintensified and performed annually, as per the guidelines. At 5 yr, the risk of bladder recurrence at this time frame was 10% (A). At 4 yr, the risk of bladder recurrence was 13% and other-site recurrence was 23%. Furthermore, recurrence continued to decrease over time.
The hazard functions obtained from the Weibull models are depicted in . The age-, CCI-, stage-, and site-specific surveillance intervals deriving from the Weibull models’ interactions are presented in . Almost all patients younger than 70 yr exhibited a higher risk of recurrence than the risk of OCM (independently of CCI and recurrence site); hence, they should continue both cystoscopies and cross-sectional imaging beyond 10 yr from RNU. Conversely, among 70–79-yr-old patients, cystoscopies might be discontinued according to CCI and pathologic stage, whereas cross-sectional imaging for other-site recurrence should be continued beyond 10 yr. Finally, individuals older than 80 yr might discontinue oncologic surveillance before 10 yr because of exceeding the risk of OCM relative to the risk of any recurrence.
Weibull models illustrating the time points during oncologic surveillance at which the risk of other-cause mortality (OCM) exceeds the risk of recurrence among patients without a prior history of bladder cancer. From a practical standpoint, it would be plausible to interrupt oncologic surveillance or adopt a lighter follow-up strategy for a specific organ or system when the risk of OCM exceeds the risk of recurrence. (A) Bladder recurrence and (B) other-site recurrence stratified for patients with CCI 0–1. (C) Bladder recurrence and (D) other-site recurrence stratified for patients with CCI ≥2. CCI = Charlson Comorbidity index; RNU = radical nephroureterectomy.
Time points, in years, according to age, Charlson Comorbidity Index (CCI), stage, and relapse location when the risk of other-cause mortality exceeds the risk of urothelial cancer recurrence in the urinary bladder or in other sites for patients without a prior history of urothelial carcinoma.a
displays the results of the Weibull regression for patients with a prior BCa history. This analysis was not stratified by CCI, to avoid overfitting, and it demonstrates that oncologic surveillance should be continued beyond 10 yr in the majority of the cases, regardless of patients’ comorbidities, if they had prior BCa. A total of 153 (36%) patients who underwent RC for BCa were excluded from the bladder recurrence analysis; yet, they were included in the other-site recurrence analysis. Weibull regression models for patients with a prior history of BCa are summarized in . Regarding bladder recurrence, cystoscopies should be continued beyond 10 yr in almost all individuals with a prior BCa history. The only exception consists of pTa/1/CIS/pT2N0 patients older than 80 yr, who might interrupt cystoscopies at 9/10 yr from RNU. Considering the HR of other-site recurrence, cross-sectional imaging should not be discontinued even after 10 yr from surgery in all patients with a prior history of BCa.
Weibull models illustrating the time points during oncologic surveillance at which the risk of other-cause mortality (OCM) exceeds the risk of recurrence among patients with a prior history of bladder cancer. From a practical standpoint, it would be plausible to interrupt the oncologic surveillance or adopt a lighter follow-up strategy for a specific organ or system when the risk of OCM exceeds the risk of recurrence. (A) Bladder recurrence. (B) Other-site recurrence. RNU = radical nephroureterectomy.
Time points, in years, according to age, stage, and relapse location when the risk of other-cause mortality exceeds the risk of urothelial cancer recurrence in the urinary bladder or in other sites in patients with a prior history of urothelial carcinoma.a
 Consultation on UTUC, Stockholm 2018 aspects of risk stratification: long-term results and follow-up. The cornerstone of oncologic surveillance for surgically treated HR UTUC patients is represented by cystoscopies and cross-sectional imaging . Concerning the intensity of follow-up examinations, the current EAU follow-up scheme entails cystoscopies every 3 mo till the 2nd year, semiannually till the 5th year, and annually thereafter, while cross-sectional imaging should be obtained semiannually till the 2nd year and annually thereafter. Notably, all of them are “weak” and “expert opinion” recommendations. Despite the rate of other-site recurrence being higher than the risk of bladder recurrence itself, the frequency of cystoscopies, as per guidelines, is higher than the one of imaging. In the first part of our study, we evaluated the risk of recurrence over time against time from RNU. We found that, at the time points when imaging and cystoscopies should be deintensified, as per the EAU guidelines, the recurrence risks were approximately 45% and 10% for nonbladder and bladder recurrences, respectively. On these premises, we suggest intensifying the frequency of imaging to a semiannual one till the 4th year after RNU, the time point after which the risk of recurrence was almost halved. Cystoscopies could be obtained annually from the 4th year given a similar risk of recurrence at 4 and 5 yr after RNU. A hypothetical surveillance scheme with more intense imaging, that is, semiannual imaging till the 4th year, and a less intense cystoscopy scheme, that is, annually from the 4th year is proposed in C and 1D, respectively.
6 Presently, major guidelines do not provide recommendations on the deintensification or interruption of oncologic surveillance following RNU . In an effort to address this void, we created a personalized follow-up scheme for HR UTUC patients to help discussing follow-up deintensification or interruption based on patient’s risk of recurrence against the risk of dying of other causes. When weighing the risk of recurrence against the one of OCM, one could advise patients on deintensification or interruption of cystoscopies, while cross-sectional imaging should generally be continued for longer. From a practical standpoint, our tables on the follow-up scheme are easy to apply in clinical practice and can be used to ease patient counseling given the immediate visual construction of the tables.
Additionally, we also demonstrated that a scheme that is valid for patients without prior BCa should not be applied to individuals with a previous BCa history. In this patient category, oncologic surveillance should be continued in almost all cases beyond 10 yr. This information is pivotal at the health policy level when planning the health care investments regarding surveillance protocols for UTUC.
29 A novel risk-based approach simulating oncological surveillance after radical nephroureterectomy in patients with upper tract urothelial carcinoma. 30 Risk-stratified surveillance protocol improves cost-effectiveness after radical nephroureterectomy in patients with upper tract urothelial carcinoma.  A novel risk-based approach simulating oncological surveillance after radical nephroureterectomy in patients with upper tract urothelial carcinoma.  Risk-stratified surveillance protocol improves cost-effectiveness after radical nephroureterectomy in patients with upper tract urothelial carcinoma. Two prior studies evaluated surveillance protocols following RNU; yet, they are both limited by the fact that all patients, irrespective of tumor grade or risk group, were considered together . Shigeta et al  explored the risk of CSM and OCM in 714 UTUC patients and found that smokers had higher risks of CSM in the long term, for whom they proposed a different surveillance strategy. While their findings are compelling, oncologic surveillance is better shaped on the risk of recurrence and not on CSM. Similarly, in a population of 426 patients, Momota et al  investigated a risk-based surveillance approach and estimated a significant cost reduction when relying on risk groups for oncologic surveillance. Yet, they included patients with low- and high-grade disease together. While our study lacks data on smoking status, our findings stand out due to the relatively homogenous patient population that included patients with HR disease and the facts that we analyzed separately the bladder and other-site recurrence and the history of prior BCa.
 Certainly, the intensity and duration of follow-up schemes are associated with important health care expenditure . Although our proposed approach can help deintensifying or interrupting follow-up in selected categories, we proposed a more intense follow-up scheme until the 4th year after RNU. Yet, the potential early identification of individuals with metastatic disease might translate into not only a better patient prognosis, but also reduced overall health care costs. Indeed, benefits, also in terms of patients’ quality of life, of our proposed approach for oncologic surveillance need to be demonstrated in prospective series.
Our study is not devoid of limitations. First, it encompasses data from several centers treated over a relatively long period of time. Yet, given the rarity of the disease, a multi-institutional study represents the only way to overcome the limitations that would originate from a single-center series. Indeed, our approach leaves room to a certain degree of unaccounted variability, and prospective validation of our findings is warranted. Although all patients were followed up according to the EAU guidelines, physicians might have not been fully compliant with the guidelines on a case-to-case basis. Again, while this might introduce some degree of unaccounted variability, it is also representative of day-to-day practice. Finally, data on postoperative intravesical chemotherapy were not recorded in this multi-institutional project, and our analyses do not account for this.