Multiple recurrences and risk of disease progression in patients with primary low-grade (TaG1) non–muscle-invasive bladder cancer and with low and intermediate EORTC-risk score

Aim To assess the prognostic value of multiple recurrences on the risk of progression in a large cohort of TaG1 bladder cancer of low and intermediate risk based on the EORTC score and to evaluate prognostic factors of multiple recurrences. Materials and methods We retrospectively analyzed a French cohort of 470 patients with primary TaG1 bladder cancer diagnosed between 1986 and 2010 and followed until 2012. They were classified at low and intermediate risk using the EORTC risk score. Associations between the number of recurrences and the risk of progression to high grade Ta/T1 bladder cancer and progression to muscle-invasive disease were assessed. The characteristics of recurrences, as occurrence time or localization, and risk of other recurrences were evaluated. Results Out of 470 patients, 251 had recurrence, 34 progressed to high grade Ta/T1 and 17 to muscle-invasive disease, including 4 who had non muscle-invasive progression first. The median follow-up was 7.2 years (interquartile range: 4.2–10.9). In half the progressions, no previous recurrence was observed. No association between the number of recurrences and the risk of progression was detected. Even after 5 years free of event, patients had a 15% risk of recurrence. History of two or more recurrences increased by 4.5 the risk of subsequent recurrence. Time between two recurrences inferior to six months and multifocal localization increased the risk of recurrence. Conclusion Surveillance of patients with TaG1 should be continued beyond 5 years of follow-up. However, cystoscopy exams could be spaced after 5 years. Multiple TaG1 recurrences did not appear to be prognostic for disease progression, but increased significantly the risk of subsequent recurrences. Short time between two recurrences and multifocal localization may serve to adapt monitoring of patients with TaG1 Bladder cancer.


Introduction
Bladder cancer (BC) is the 9 th most commonly diagnosed cancer in the world [1]. The worldwide age standardized incidence rate (per 100,000 person-years) was 9.0 for men and 2.2 for women in 2012 [1]. In France, the incidence is increasing by about 1% per year [2][3]. Approximately 75-85% of BC are non-muscle-invasive. Of these, 70% are stage Ta, 20% are T1 and 10% are in situ (Cis) [4][5][6][7]. Ta papillary tumors with histopathological grade (G) 1 (or low grade) are those most frequently found at diagnosis. The natural history of superficial bladder cancer is characterized by a high rate of recurrence (between 30% and 78% at 5 years) in the same stage/grade but a limited five-year risk of progression to muscle invasive disease of stages T2 to T4 (7% to 40%) [8][9][10][11]. These recurrent events alter the patient's quality of life and might increase the risks of progression either to a High-grade Ta, T1, CIS, or to T2-T4 pathological stage, which in turns have a major impact on the therapeutic management.
American and European studies have identified baseline prognostic factors of recurrence and/or progression [7][8][9][10][11][12][13]. Tumor size (> 30 mm), pathological stage T1, high grade, presence of Cis, number of lesions (multiple lesions), history of recurrence during the first year after diagnosis, and advanced age appeared to increase the risk of recurrence and/or progression. The European Organization for Research and Treatment of Cancer (EORTC) developed in 2006 a scoring system for Ta-T1 BC to predict the risk of progressions to muscle-invasive disease based on data from 7 clinical trials [8]. The European Association of Urology (EAU) proposed to categorize patients with non muscle invasive BC into three risk groups identified from this score [4], to facilitate treatment recommendations. Based on the 2018 EAU guidelines, regular monitoring via cystoscopy is recommended in low-risk TaG1 patients. Immediate post-operative intravesical instillation of chemotherapy (IPOIC) may be added at all risk levels. For the other risk groups, a medical treatment of several months should be discussed: intravesical Mitomycin C or BCG instillations for intermediate-risk patients, and BCG instillations for high-risk patients [4].
The associations between recurrence and progressions to higher stage / grade or to muscleinvasive disease are poorly documented in the low grade (TaG1) population. Furthermore, recurrence rate is most often measured at baseline and is not updated with the occurrence of new recurrences [7][8][9][10][12][13]; the link between tumor characteristics and the risk of subsequent recurrences, taking into account the whole patient history of recurrences after the first transurethral resection of a bladder tumor (TURB), has been poorly investigated so far. Finally, tumor evolution in homogeneous cohorts of patients with primary TaG1 and long-term follow-up are lacking.
The objectives of this study were to investigate the association between the number of previous recurrences and progression to high grade Ta/T1 and to muscle invasive disease in the  TaG1 population at low or intermediate risk, to validate prognostic factors of recurrence and  to investigate the inter-relationship between repeated recurrences, time to recurrence, or multifocal localization, after TURB diagnosis. New insights on the natural history of TaG1 papillary tumors may improve our ability to adapt the follow-up of those patients who are considered at low-risk by the EAU classification.

Study population
This is a French retrospective observational cohort of primary consecutive TaG1 pure urothelial BC diagnosed and treated between 1986 and 2010 in Foch hospital [9]. The Gustave Roussy Cancer Center ethic committee (CSET) approved this research. No signed informed consent was requested as the study was purely observational. Individual data of patients, who underwent a first TURB and a macroscopic radical resection (possibly after two interventions), were collected from medical and anesthesiologist records. A cut-off date was set at 15/01/2012.
Tumor stage and grade were determined by two uropathologists with a central review according to the TNM and grade classification (1973 World Health Organization classification) to include solely initial diagnosis of TaG1 BC. Patients with an upper urinary tract tumor or primary concomitant Cis were excluded [14,15]. The tumor grade of events that occurred during follow-up was reviewed by two uropathologists (VM, CR) and reclassified using the 2004 WHO classification: G1 and G3 were classified in low and high grade respectively. All G2 cases were classified in low or high-grade, taking into account clinical context when necessary. Cis status was systematically reclassified in high-grade disease. Baseline was taken as the date of TURB. Patients were followed according to the French Association of Urology guidelines: Regular monitoring included cytology and flexible cystoscopy performed three months after initial TURB, then every trimester for two years, every semester thereafter until five years, and then yearly. Computed tomographic scan examination were done every other year for upperurinary tract check-up [9]. All patients who had a suspicion of disease recurrence underwent an additional TURB: detrusor in the specimen were observed for all T1 or T2 recurrences as standard practice included some muscle resection but no fulguration was performed; recurrences were confirmed by a uropathologist and treatments were collected. No patient received any adjuvant treatment after the initial TaG1 diagnosis prior to the first recurrence (no IPOIC, BCG, or mitomycin C instillations). For patients who discontinued the follow-up at Foch hospital, the new treating urologists were contacted to update patients' clinical status; however recurrences or progression were then not centrally reviewed. The management of tumor recurrence followed the EAU recommendations [4].

Endpoints
Time to progression to higher grade Ta / T1 was defined as the time from diagnosis to the date of first increase to T1 stage or high grade/G3 tumors or Cis. Time to progression to muscleinvasive disease was defined as the time from diagnosis to the date of first increase to T2 stage or higher. Patients still alive and without disease progression at the cut-off date were censored. The primary endpoint was "all types of progression" i.e. high-grade Ta or T1 to T4 progression whichever came first. The time to first recurrence was defined as a new TaG1 or Ta low grade/ G2 tumor in the same way. Deaths without tumor progression were censored at the date of death and not considered as an event. When investigating the risk of the first recurrence, we ignored progression before recurrence, as it rarely occurred.

Prognostic factors of progression and/or recurrence
Data on size and number of tumors was collected from TURB reports. Tumor site (including multifocal status) and the type of care (surveillance, type of adjuvant treatment or surgery) were available at baseline and during follow-up though standardized case report forms filled in by surgeons. Smoking status (smoker or non-smoker) was collected from anesthesiologist medical records at baseline [16].

Longitudinal statistical analysis
The median follow-up was estimated using the inverse Kaplan Meier method [17]. Log-rank test and Cox model analysis were performed to identify prognostic factors of recurrence and progressions. The cut-off values for the categorical variables came from the univariate analyses or literature data. No model selection was performed and all multivariate models were adjusted on: age, sex, smoking status, treatment received during follow-up (yes or no), and tumor characteristics such as number of tumors (1, 2-7, >7), size (�30 mm or >30 mm). To analyze the risk of progressions to higher grade Ta or to muscle invasive diseases, we accounted for multiple recurrences occurring in the same patient (explanatory variable) using multivariate time dependent Cox model. To analyze the effect of the number of recurrences on the risk of subsequent recurrence, landmark was used in order to correct for the "selection bias" [18,19]: In this analysis, baseline time was postponed to 3 years at which 50% of the recurrences had occurred, and at later timepoints. Finally, to investigate the impact of the localization of recurrence number k and the risk of recurrence k+1 as well as the delay between two recurrences, we used multivariate frailty models, which produce hazard ratios and confidence intervals as association measures [20,21]. As a sensitivity analysis, we excluded recurrences that occurred less than 3 months after TURB, which could be the result of an incomplete resection of the primary tumor. All statistical analyses were performed using SAS 9.4.

Study population
Between April 1986 and December 2010, 504 consecutive patients underwent a TURB for newly diagnosed TaG1 BC. Thirty-four patients were subsequently excluded because of uppertract tumor localization (n = 6) and treatment at baseline (n = 28) leading to a study population of 470 TaG1 (Fig 1). Mean age of patients was 66 years. Sex-ratio (M/F) was 4.3:1. Overall, 70% of the population were smokers and 59% were at low-risk of progression to muscle-invasive disease according to the EORTC score ( Table 1). The median follow-up was 7.20 years (interquartile range: 4.20-10.88 years). Follow-up was discontinued after 10 years by standard practice and 65 were lost-to follow-up, i.e. no data could be collected for their last planned visit. Recurrence occurred in 251 patients. In total, the median number of recurrences per patient was 1 (range: 0-19). Progression to T1 or TaG3 cancer occurred in 34 patients and progression to muscle-invasive disease occurred in 17 patients, including 4 who had non muscleinvasive progression first. Respectively 17 and 8 of them occurred despite the absence of previous TaG1 recurrence events (Fig 1). The characteristics of the 47 patients who progressed are listed in Text A in S1 File. During the follow-up, 201 patients had no relapse or progression events and 73 patients died (including 59 patients who never experienced disease progression).

Prognostic factors of progression
At seven years, the cumulative incidence of progression to high grade TaT1 disease and to muscle-invasive disease were 8.2% (95%CI: 4.8-10.5) and 1.8% (95%CI: 0-2.7) respectively (Fig 2 and Figure A Panels A.1, A.2 and A.3 in S1 File). Whatever the type of first progression considered no statistically significant baseline prognostic factor was found. Prognostic value of recurrences that occurred over time is summarized in Table 2 (Tables A and B in S1 File). The number of previous recurrences was not a significant prognostic factor for the outcome "all types of progression" even after multiple recurrences. Similar results were obtained when considering only progression to high grade Ta/T1 disease. Results were similar after the exclusion of the 20 patients who recurred within 3 months after the initial TURB (See Table C in S1 File). . Tumors were more often multifocal at baseline than at recurrences (17.9% vs. 6.1%). Therapeutic management of recurrences was IPOP (14%), mitomycin (7%) and BCG (12%). Surveillance was maintained after 63% of the recurrences. As shown on Fig 3, the median recurrence-free time Recurrences and progression in patients with primary TaG1 non-muscle-invasive bladder cancer was 3.2 years (95%CI: 2.3-4.4) and incidence peaked before three years (Figure A Panel A.1 in S1 File). The cumulative incidence of first recurrence reached 60% at seven years. When we focus on patients who were free of recurrence at 2 years, the cumulative risk of recurrence over the 4 following years was 30% and patients with no recurrence at 5 years had still a 15% risk of developing a first recurrence between years 5 and 9 (Table B in S1 File).

Prognostic factors of recurrence
Baseline prognostic factors associated with the risk of first recurrence were: the size of the tumor (p<0.001) and the number of lesions (p <0.001). After adjustment, patients relapsing once in the three first year after TURB were more likely to have subsequent recurrence compared to those who did not (HR = 2.11; 95%CI: 1.28-3.49) as were patients with 2 or more recurrences (HR = 4.72; 95%CI: 2.77-8.05) ( Table 3 and Table C in S1 File). History of recurrences in the last 6 months increased by 33% the risk of subsequent recurrence (Table 4 and Table D in S1 File). Low intra-patient correlation was found between repeated recurrences, which suggested that the times before recurrence were relatively independent of each other. Finally, the multifocal localization (HR = 1.47; 95%CI: 1.19-1.80) was significantly associated with the recurrence-free time (Table 3 and Table D in S1 File).

Discussion
In this large single-center cohort of 470 primary TaG1 low and intermediate risk BC with long-term follow-up, we found that half of the patients with disease progression did not experience any prior tumor recurrence resulting in the absence of association between the two outcomes, even for patients with multiple (>4) recurrences. However, history of 2 or more recurrences increased by 4.5 the risk of subsequent recurrence. Even 5 years free of recurrence after the initial TURB, patients had a 15% risk of developing a new recurrence, suggesting that follow-up should be maintained on a long-term period. Indeed, we also found that previous recurrence within the last 6 months was prognostic of new recurrence. Finally, we confirmed that multifocal localization at diagnosis was a risk factor of subsequent recurrence [7][8][9][10][11][12][13].
Overall, the frequent cystoscopic examinations entail a heavy cost to the society, and BC induces the highest cost per patient from diagnosis to death among cancer types [22]. The possibility to reassure TaG1 patients with multiple recurrences regarding the risk of progression and to increase the interval between two cystoscopy exams in low risk patients both motivate the identification of prognostic factors. American and European studies that focused on baseline prognostic factors of recurrence and/or progression [7][8][9][10][11][12][13] reported that prior recurrence rate (<1/year, > 1/year) or the history of recurrence (>1 vs 0, without precision of time of occurrence) were prognostic of recurrence and of progression. However, inclusion criteria allowed a prior history of BC; the presence of T1 pathological substage, or adjuvant treatment after the initial TURB. All these parameters increase the heterogeneity in study population and strongly impact the clinical outcome of the patients. Few studies have investigated the impact of repeated recurrences on the risk of progression. Lujan et al. accounted for multiple recurrences [11] and concluded to a mild association between these two risks but this study presented the same heterogeneity regarding the inclusion criteria as discussed previously.  Recently, Golabesk (2017) reported that among a single-institution retrospective cohort of 704 patients including 548 Ta patients (median follow-up = 5 years), only 3.6% of the initial tumors relapsed after 5 years of follow-up, which was lower than the predicted values from the EORTC score [23]. In Balan (2018), long-term survival of a homogeneous cohort of 164 Ta low grade patients was similar between G1 and G2 tumors but they did not investigate timing of events [24]. Although EORTC risk tables provide a prognostic tool, no molecular biomarkers accurately predict disease progression. In recent years, the hypothesis of "distinct biological pathway" has  been proposed to explain the different evolution between NMIBC [5,[25][26][27]. This hypothesis involved mutation of tumor suppressor genes, like P53, or oncogene deletion like FGFR3 that would characterize two distinct genetic pathways associated either to multiple recurrences or to progression. However, mutation patterns have not yet been identified to consistently discriminate between patients with low or high-risk of disease progression. In addition, no molecular biomarker has been implemented in the clinical practice. Recently, Hedegaard et al performed a comprehensive transcriptional analysis in a large cohort of 460 primary Ta/T1 bladder cancers and found that they could be subgrouped into 3 major molecular classes with basal and luminal-like characteristics, with widely different outcomes [28]. Large prospective cohort studies, collecting repeated epidemiological, clinical, pathological, and molecular data, such as the COBLANCE study will be key to explore the relationships between epidemiological, clinical, pathological, and molecular data.
The major strengths of our study include a long median follow up period of more than seven years. Number of events was substantial and provided an important power for our statistical analysis of risk of recurrence. Moreover, the study population was homogeneous in terms of treatment management, which reduces potential confounding factors. Our study is one of the first that took into account the entire clinical history of the patients after TURB.
The principal limitations of this cohort are related to its retrospective nature and the risk of selection bias. The follow-up and the clinical management corresponded to daily clinical practice. Therefore, modifications of the evaluation of patients may have occurred. This heterogeneity is somehow reduced by the fact that the same team staged and treated all the patients, which improved the standardization of the treatment of non-muscle invasive recurrences. Despite we adjusted analyses on the received treatment for recurrence, variations over years might have affected the risk of progression. Furthermore, despite a cut-off date at which status of all patients was queried via their treating urologist, there was still 14% of lost-to-follow up patients. For the patients who were followed-up outside of the Foch hospital, no imaging data was collected to confirm recurrences or progression reported by the treating physician. A second important limitation is the insufficient statistical power to investigate the progression to muscle-invasive disease and the modification of this risk over time. Finally we lack epidemiological data. For instance, smoking status was reported as a binary (yes/no) variable but no data was available regarding quantity and duration of smoking which may increase the risk of recurrence and progression [16]. Another potentially important missing variable was the Ki 67, a marker of cell proliferation that has been associated with shorter time to recurrence and time to progression [29]. A meta-analysis of TaG1 cohorts should be carried out to have enough power to investigate the prognostic value of recurrences on the risk of progression to muscle invasive diseases.

Conclusion
The number of low-grade (TaG1) recurrences did not appear to be prognostic for disease progression, but increased significantly the risk of subsequent recurrences as a short time to recurrence and multifocal localization did. Therefore, surveillance of patients with TaG1 tumors at low / intermediate risk based on the EORTC score might be adapted based on those simple variables. It should be continued beyond 5 years of follow-up as 17 of the 47 progressions occurred 5 years after initial TURB and the hazard of any progressions remained relatively high but yearly cystoscopy could be recommended instead of 6-monthly exams to account for the decreasing risk of recurrence.
Supporting information S1 File. Supporting information. Figure A: Hazard functions for recurrence (A.1), progression to Ta/T1 stage (A.2) or progression to muscle invasive bladder cancer (A.3). Fig A represents the hazard function for the first recurrence, which is the rate at which new events occur in the at-risk population; it was maximal before one year after diagnosis and then decreased progressively over time; at three years, the cumulative incidence of recurrence was close to 0.7. Fig B represents the hazard function for the progression to Ta/T1 stage bladder cancer or nonmuscle invasive bladder cancer (NMIBC); it decreased continuously over time. Fig C represents the hazard function for the progression to muscle invasive bladder cancer (MIBC). The hazard risk of muscle invasive bladder cancer progression was low in the first year after diagnosis and then increased thereafter. Table A: Sensitivity analysis of the risk of progression after exclusion of the 20 patients who relapsed within the 3 months after the initial TURB.  Table B: Multivariate Hazard Ratios (95% Confidence Intervals) of the effect of the number of recurrences on NMIBC progression-free time with details on covariates-Number of recurrence is a time-dependant variable in the cox model. Table C: Univariate and multivariate analyzes of the effect of the number of recurrences on progression-free survival. Table B: Cumulative incidence of first recurrence over the 4 years in patients free of recurrence at increasing timepoints.