Purpose: Mutations in the fibroblast growth factor receptor 3 (FGFR3) have been found in 70% of the low-grade non-muscle–invasive bladder cancer (NMI-BC) tumors. We aim to determine the potential of FGFR3 mutation analysis on voided urine to detect recurrences during surveillance of patients with low-grade NMI-BC.
Experimental Design: FGFR3 mutation status of the study inclusion tumor was determined from 200 low-grade NMI-BC patients. Patients with an FGFR3-mutant inclusion tumor were selected for analysis and monitored by cystoscopy, and voided urine samples were collected. FGFR3 mutation analysis was done on 463 prospectively collected urines. Sensitivity and predictive value of the assay were determined for detection of concomitant recurrences. Longitudinal and Cox time-to-event analyses were done to determine the predictive value for detection of future recurrences.
Results: Median follow-up was 3.5 years. The sensitivity of the assay for detection of concomitant recurrences was 26 of 45 (58%). Of the 105 positive urine samples, 85 (81%) were associated with a concomitant or a future recurrence. An FGFR3-positive urine was associated with a 3.8-fold (P < 0.0001) higher risk of having a recurrence in the Cox analysis. In contrast, only 41 of 358 (11%) FGFR3-negative urine samples were associated with a recurrence. Positive predictive value increased from 25% to 90% in patients having consecutive FGFR3-positive urine tests.
Conclusions: FGFR3 mutation analysis on voided urine is a simple and noninvasive diagnostic method for detection of recurrences during surveillance of patients presenting with a low-grade FGFR3-mutant NMI-BC tumor. Clin Cancer Res; 16(11); 3011–8. ©2010 AACR.
This article is featured in Highlights of This Issue, p. 2919
The high recurrence rate of low-grade non-muscle–invasive bladder tumors necessitates life-long frequent cystoscopic monitoring of patients, and this affects patient quality of life. Fibroblast growth factor receptor 3 (FGFR3) mutations are found in 70% of the low-grade tumors and associated with a good prognosis. Surveillance with FGFR3 mutation detection on voided urine could ultimately reduce the number of cystoscopies and improve patient quality of life. We investigated whether FGFR3 mutation analysis could detect concomitant recurrences and predict future recurrences. We determined the recurrence risk associated with an FGFR3-positive urine and the predictive value of consecutive FGFR3-positive urines in time. Our findings suggest that an FGFR3-based and individualized surveillance protocol could be possible and having an FGFR3-positive urine is associated with a four times higher recurrence risk.
Bladder cancer (BC) is the fifth most common malignancy in the world and can be divided into non-muscle–invasive BC (NMI-BC) and muscle-invasive BC tumors (MI-BC; ref. 1). Almost 80% of the NMI-BC patients will present with a G1 or G2, low-grade NMI tumor, making this a considerable burden on the urological practice. The prognosis of low-grade NMI-BC patients is good with a 5-year survival of 80% to 90% (2, 3). NMI bladder tumors are treated by transurethral resection, but 70% of the patients will develop at least one recurrence within 5 years and 10% to 20% will progress to MI disease. Currently, the surveillance protocol consists of cystoscopy every 3 to 6 months for the first 2 years, followed by less frequent observations when a patient stays recurrence-free (4). The frequent transurethral resections and follow-up cystoscopies make BC from diagnosis to death the most costly cancer in the world (5–8). The life-long follow-up of patients by invasive cystoscopy and associated high costs emphasize the need for a urinary biomarker, which can be used in a noninvasive assay for the detection of recurrences and to stratify patients with a high-risk profile, hereby creating the possibility to reduce the number of cystoscopies and thus improving quality of life.
About 70% of low-grade NMI tumors have a mutation in the fibroblast growth factor receptor 3 (FGFR3) gene (9–12). In addition, it was shown that patients with an FGFR3 mutation have a good prognosis (13–17). Because low-grade NMI-BC patients comprise the largest group in the BC urological practice, FGFR3 mutation analysis could serve as a biomarker for the detection of recurrences during follow-up and aid in identifying patients with a low risk of progression. Two studies using single-strand polymorphism analysis for FGFR3 mutations showed that it is possible to detect recurrent tumors in urine (12, 18). Because this technique is rather laborious, we developed a multiplex assay to identify the most common FGFR3 mutations and showed its potential on urine and tissue samples (19). In the present study, we optimized this assay and validated it in a longitudinal study that determines the recurrence risk associated with FGFR3-positive urine samples in a large patient cohort with FGFR3-mutant low-grade NMI-BC.
Materials and Methods
Patients and sample collection
Patients with available urine samples were selected from the participants of the “Cost-Effectiveness of Follow-up of patients with non-muscle invasive Bladder cancer trial” (CEFuB trial; ref. 20). Patients with stage pTa or pT1 and grade 1 to 2 tumors were included at transurethral resection. Patients with a history of carcinoma in situ were excluded from participation. A cystoscopic examination coincided with urine collection for molecular analysis, or urine was collected 1 month following the cystoscopy due to logistic reasons. Recurrence was defined as a histologically proven tumor. Progression was defined as progression to MI disease.
Tissue for DNA extraction was obtained by manual dissection from formalin-fixed, archival paraffin blocks containing tumor areas that were selected by pathologic examination of the corresponding histologic slides to contain a minimum amount of 70% tumor tissue. All tumors were reviewed by an expert pathologist (T.H.v.d.K.). Samples were first deparaffinized and DNA was extracted using the Qiagen DNeasy blood and tissue kit (Qiagen GmbH) according to the manufacturer's protocol.
Freshly voided urine (10-100 mL) was collected before cystoscopy and stored at 4°C until transportation to the Department of Pathology at Erasmus Medical Center. Cells were pelleted by centrifugation for 10 minutes at 3,000 rpm (1,500 × g). Cell pellets were washed twice with 10 mL PBS, resuspended in 1 mL PBS, transferred to an Eppendorf vial, and collected by centrifugation for 5 minutes at 6,000 rpm (3,000 × g). Supernatant was discarded and the cell pellet was stored at −20°C until DNA isolation. DNA was extracted using the QiAamp Viral RNA Mini kit (Qiagen) according to the manufacturer's protocol.
FGFR3 mutation analysis
A multiplex PCR of the three regions that contain the most frequent FGFR3 mutations was done as described by van Oers et al. (19). These regions comprise the following codon mutations: R248C and S249C (exon 7); G372C, S373C, Y375C, G382R, and A393E (exon 10); and K652M, K652T, K652E, and K652Q (exon 15). Primers used are depicted in Supplementary Data S1. PCR was done in 15 μL containing 1 to 5 ng of genomic DNA, 1× PCR buffer (Promega), 1.5 mmol/L MgCl2, 0.17 mmol/L deoxynucleotide triphosphates (Roche), 1.0 unit Taq polymerase (Promega), 5% glycerol (Fluka), 18 pmol of exon 7 primers, 7.5 pmol of exon 10 primers, and 10 pmol of exon 15 primers. Cycling conditions were as follows: 5 minutes at 95°C, 35 cycles at 95°C for 45 seconds, 60°C for 45 seconds, and 72°C for 45 seconds, followed by 10 minutes at 72°C. Total PCR product was treated with 3 units of shrimp alkaline phosphatase (Amersham Biosciences) and 2 units of exonuclease I (Amersham Biosciences) to remove excess primers and deoxynucleotide triphosphates. This was followed by a mutation analysis (ABI PRISM SNaPshot Multiplex kit, Applied Biosystems) using probes that anneal to the PCR product adjacent to the mutation site. Different lengths of polydeoxythymidylic acid tails were attached to the 5′-end to enable simultaneous detection on the sequencer. All probes are shown in Supplementary Data S2. The mutation detection reaction was done in 10 μL containing 1 μL of multiplex PCR product, 2.5 μL Ready reaction mix, 1× sequencing buffer, and probes as shown in Supplementary Data S2. Cycling conditions were 35 cycles of rapid thermal ramp (1.30) to 96°C, 96°C for 10 seconds, rapid thermal ramp (1.30) to 58.5°C, and 58.5 for 40 seconds, followed by treatment with 1 unit shrimp alkaline phosphatase. Separation was in a 30-minute run on 36-cm-long capillaries on an automatic sequencer (ABI PRISM 3130 XL Genetic Analyzer, Applied Biosystems), with the label indicating the presence or absence of a mutation.
The Statistical Package for the Social Sciences 11.5 (SPSS, Inc.) was used for data analysis. Sensitivity, specificity, and predictive value were determined for every follow-up moment with a urinary FGFR3 result with a concomitant cystoscopic examination. To take into account the fact that a patient could have more than one urine measurement and that these are likely to be correlated within one patient, we calculated the sensitivity, specificity, and the predictive value with random-effect regression models. These models account for within-patient correlation and are suitable for the analysis of repeated measurements within one patient, as in our study.
In the longitudinal analysis, we determined whether a urine sample is followed by a recurrence within 12 months or within the total duration of the study period. We analyzed the relationship between a FGFR3-positive urine and the associated recurrence risk with a Cox proportional hazard model. A tumor yes/no was taken as the outcome. The time from first urine measurement to tumor detection or end of study was taken as duration of follow-up. Also in this analysis, we accounted for the fact that one patient could have more than one urine measurement. Therefore, the consecutive urine measurements were taken as a time-dependent covariate in the Cox model. The hazard ratio from the model can then be interpreted as the effect of a single positive urine sample on the hazard of a tumor, regardless of the previous samples in the same patients. Additionally, the predictive value of subsequent FGFR3-positive urine tests was determined by Kaplan-Meier longitudinal analysis. This was done by analysis of one or more consecutive follow-up visits (series) with either FGFR3-positive urines (persistent FGFR3 positive) or FGFR3-negative series (persistent FGFR3 negative), meaning, that a series could be defined as one or more consecutive FGFR3-positive or FGFR3-negative outcomes ending at (a) a recurrence [including upper urinary tract (UUT)], (b) a change in urine FGFR3 status from positive to negative or negative to positive, or (c) at the end of follow-up. In case of (a) or (b), a new series would start. Curves of proportion recurrence were computed and stratified by mutational status using Kaplan-Meier analysis. The probability of recurrence development was compared using a log-rank statistic. Results were considered statistically significant at P < 0.05.
The FGFR3 and progression status of the tumor was known for 292 patients, of whom 193 patients had an FGFR3-mutant tumor. FGFR3-mutant tumors showed progression to MI disease in 10 of 193 (5.2%) cases compared with 15 of 99 (15.2%) in FGFR3 wild-type (WT) tumors (P = 0.004). This finding agrees with previous reports. Patients with an FGFR3-mutant tumor were then selected for analysis of urine samples for the detection of recurrences. From 134 patients with an FGFR3-mutant tumor, urine samples were available. Clinical and molecular characteristics of this population are shown in Table 1.
Performance of the FGFR3 mutation assay
We optimized the FGFR3 mutation assay by improving detection of the most common mutation, S249C, and adding two new mutations to the detection spectrum. We determined the sensitivity of the assay by diluting mutant tumor DNA with WT DNA. The results show that the two most common mutations, S249C and Y375C, can still be detected in a background of 40-fold normal DNA. For R248C and G372C, this was 25- and 10-fold, respectively. The results for the most common S249C mutation are shown in Supplementary Data S3. The FGFR3 assay was subsequently carried out successfully on all 463 urine samples, and mutations in FGFR3 could be detected in urinary-derived DNA (Fig. 1A-C). The analysis was done on urine samples from an age-matched patient cohort without BC (n = 100), and no mutations were detected.
Detection of concomitant recurrences
From the 134 patients included with an FGFR3-mutant tumor, 463 urine samples were analyzed for FGFR3 mutations and 45 concomitant histologically proven recurrences were found. A selection of patients is shown in Fig. 2 to illustrate the possible groups that patients could be divided into (full figure available online as Supplementary Data S4). We discern four types of patients: (I) patients with positive urine samples and a recurrence (FGFR3 mutation analysis predicts a recurrence), (II) patients with two or more FGFR3-positive urine samples in the absence of a recurrence (patients at risk for a recurrence), (III) patients with FGFR3-negative urine samples in the presence of a recurrence (FGFR3 analysis does not predict a recurrence), and (IV) patients without recurrences: the recurrence-free group [i.e., patients having consistently FGFR3-negative urine samples or a single FGFR3-positive urine sample (n = 7) within a series of negative samples]. The sensitivity for detection of a concomitant recurrence was 26 of 45 (58%), and the positive predictive value was 26 of 105 (25%). The urine assay detected three UUT recurrences that were not detected by cystoscopy, including one pT2G3. We observed that many urine tests were positive, whereas the concomitant cystoscopy did not reveal a tumor. Because FGFR3 mutations are tumor specific and do not occur in normal tissue or urine, we argued that these represent anticipatory positive results, and therefore, we did a longitudinal analysis on the occurrence of future recurrences.
FGFR3 mutation detection for identifying patients at risk for future BC recurrences: a longitudinal analysis
Because of the anticipatory effect, we extended the time of patient follow-up so that at least a year of clinical follow-up data was available after the last urine sample had been analyzed. This increased the number of recurrences to 79 (45 concomitant and 34 future recurrences). From the 105 positive urine samples, 58 (55%) were associated with a recurrence within a year after the positive urine test and 85 of 105 (81%) with a recurrence during the total study period (Fig. 3). The remaining 20 positive urines were not followed by a recurrence in the entire follow-up period. It should be noted that eight urine samples from this group were single FGFR3-positive urines with a low mutant signal, depicted in Fig. 2 as patient type IV. Of the other 12 urine samples with an FGFR3-positive signal, 11 are from three patients representing patient type II (Fig. 2) as at risk patients because all had multiple positive urine samples. These patients had insufficient follow-up. Interestingly, multiple urine samples from patient H6 also displayed loss of heterozygosity and this patient even had macroscopic hematuria. Despite this, an intravenous pyelogram did not reveal any upper tract disease. These findings suggest that type II patients might have minimal residual disease and warrant regular examinations.
Figure 3 shows that 358 of 463 urine samples were FGFR3 negative, of which 317 (89%) were from patients in whom no recurrence occurred. The other 41 (11%) negative samples were associated with a recurrence within 12 months after urine sample collection. However, 11 of these negative urines were within a series of multiple positive urine tests. Therefore, these samples could reflect the absence of tumor cells in the urine. In the other 30 FGFR3-negative urine samples that were associated with a recurrence, only a single urine sample was available before the recurrence in 14 cases.
In a Cox time-to-event analysis, determining the hazard risk of a recurrence with an FGFR3-positive test, it was shown that a single positive FGFR3 test was associated with a three times higher risk of a recurrence, regardless of the previous urine sample (hazard ratio, 3.8; P < 0.0001). Next, we determined the predictive value of multiple consecutive FGFR3-positive urine tests in time and recurrence development by Kaplan-Meier longitudinal analysis. The risk of developing a recurrence increases to 90% in patients with consecutive FGFR3-positive urine samples for 39 months (P < 0.0001, log-rank test; Fig. 4).
Clinicopathologic and molecular features of the detected and missed recurrences within 12 months following urine collection and analysis are shown in Table 2. Mostly pTaG1 and pTaG2 tumors were missed. It should be noted that one of two pT1G3 and one of two pT2G3 tumors were not detected, but in these cases, only one urine sample could be investigated before the recurrence. Three upper tract recurrences, one of which was pT2G3, were detected by the mutation analysis but not by cystoscopy. There were five patients with progression to MI disease from whom urine samples were available (Supplementary Data S4). In three of five cases, FGFR3 urine analysis predicted the recurrence.
In summary, an FGFR3-positive urine is associated with a three times higher recurrence risk, and having multiple consecutive FGFR3-positive urine tests is strongly associated with development of a future recurrence.
Mutations in FGFR3 are frequent in NMI bladder and UUT tumors. This is the first study that determines the potential of urinary surveillance by FGFR3 mutation detection on a large set of prospectively collected urine samples of low-grade NMI-BC patients presenting with an FGFR3-mutant tumor. In this study, 67% of the low-grade NMI tumors had a mutation in FGFR3, in agreement with previous studies (9 10, 12). Because these mutations are not found in urine of non-BC patients and normal urothelium, detection of an FGFR3 mutation is a strong indicator for the presence of tumor cells in the urinary tract (21). FGFR3 mutations could be detected in the presence of 5% tumor cells, making the assay highly sensitive for detection of the most common mutations. This is of great advantage because urine can also contain normal urothelial cells and lymphocytes. The assay is cheap with costs for consumables under $10 per sample, including DNA isolation, and can be done with only 1 ng DNA in a standard molecular diagnostics laboratory with a PCR machine and sequencer. This assay is also less labor intensive than sequence analysis.
FGFR3 analysis was carried out successfully on all urine samples in this study. The sensitivity of the FGFR3 assay for detection of concomitant recurrences was 58%. This low concomitant sensitivity has several possible reasons. Firstly, tumor cells are not always present in urine. This is illustrated in Fig. 2, where a subset of patients (bottom two, type I patients) is depicted in whom an FGFR3-negative urine sample is embedded between positive urines preceding a recurrence. Because this problem of urine sampling concerns all urinary tests that depend on the presence of tumor cells in the urine, this is an important issue and it would be of interest to determine the optimal time point and frequency of urine collection to improve the number of urinary tumor cells. Secondly, FGFR3 WT recurrences are sometimes found in patients with a primary FGFR3-mutant tumor (22). Whether this is the reason why some recurrences were missed here is not known at present because tumor material from the recurrences was not available for FGFR3 mutation analysis. We have previously shown that the stage and grade of these recurrences do not differ from mutant recurrences (22). Therefore, the absence of a mutation cannot be correlated with progression to a more severe phenotype. In most patients, WT recurrences are again followed by mutant recurrences. Thus, repeated urine testing should circumvent this potential problem. Thirdly, although cystoscopy is considered the golden standard, it is known that cystoscopy does not detect all tumors (22–25). Urine cytology was not routinely done in the participating centers due to the low sensitivity in tumors of low stage and grade (G1 tumors: range, 7-38%; G2 tumors: range, 18-46%), and this is considered a limitation of the study (26–29). When this work was under review, Miyake et al. published a peptide nucleic acid–mediated PCR clamping assay for detection of FGFR3 mutations in urine samples. They were able to detect seven of nine recurrences using this assay on DNA obtained from urine samples (sensitivity, 78%). Unfortunately, the assay could not be carried out on 27% of the urine samples due to low DNA concentrations. In our study, all samples were analyzed successfully. A very interesting finding in their study is the quantitative analysis of the number of tumor cells present in urine, shown to be a prognostic indicator of tumor recurrence (30).
For the clinician, it is important to know which decision to make when confronted with an FGFR3-positive or FGFR3-negative urine test. Our data show that ∼23% of the urine samples are positive for an FGFR3 mutation in patients included with an FGFR3-mutant tumor (Fig. 3) and that the majority of these positive urines (81%) are associated with a recurrence. We show that a positive FGFR3 urine is associated with a 3.8-fold higher risk to develop a recurrence, and these findings could provide a base to set up a new individualized surveillance protocol. Additionally, we show that multiple consecutive FGFR3-positive urines are significantly associated with a higher proportion of recurrences. In this study, patients in general had an equal number of cystoscopies, meaning, that our findings reflect a true effect of the mutation analysis and are not caused by differences in cystoscopy frequency between patients. Almost 70% of the patients presenting with a primary bladder tumor will have a mutation in FGFR3. We suggest that after the first transurethral resection, patients have a cystoscopy at 3-12-24 months and urinary FGFR3 mutation analysis at 6-9-15-18-21-27-30-33 months. In the far majority of cystoscopies, following a positive test, a recurrence will be found, especially when the urologist is aware of the positive urine test result (31). Hence, we suggest that a positive test should be followed by a cystoscopy. In case of a positive urine test and a negative cystoscopy, we suggest that UUT imaging is done. Frequency of imaging should be adjusted according to the clinicopathologic features of the tumor that have been associated with an increased risk of upper tract recurrences (e.g., tumor located in trigone, multiple tumors, and high-risk tumors; refs. 32, 33). If an FGFR3-mutant patient presents with a negative test result during follow-up, no cystoscopy is required if urine analysis is done after 3 and 6 months following the negative test. When FGFR3 urine analysis is implemented during surveillance, the number of cystoscopies will decrease in patients with negative urine tests and patients with positive urine tests who require more frequent cystoscopic monitoring are identified. Because ∼80% of the urine tests are negative (Fig. 3), the total number of cystoscopies can be reduced substantially, hereby lowering costs and improving patient quality of life.
At this moment, the standard cystoscopic follow-up protocol is indicated in patients presenting with an FGFR3 WT tumor because the risk of progression is higher in these patients. Detection of loss of heterozygosity by microsatellite analysis is the only evaluated molecular marker in the follow-up of patients with an FGFR3 WT tumor (20). Although results seem promising, currently, costs are too high to implement the microsatellite analysis test in a routine clinical setting. Food and Drug Administration–approved UroVysion fluorescence in situ hybridization could be an alternative for detection of FGFR3 WT high-risk recurrences but is not yet evaluated together with FGFR3 analysis.
We have also shown that not many tumors are missed when multiple urine samples were available. Moreover, three UUT tumors were detected, which can be considered an added confidence in the detection of BC recurrences. The finding of UUT having FGFR3 mutations agrees with a previous study by van Oers et al. (17). We also confirmed that patients with FGFR3-mutant tumors have a significantly lower rate of progression. NMI-BC tumors with an FGFR3 mutation differ genetically from NMI-BC tumors that are WT in that they have fewer genomic aberrations as was indicated by loss-of-heterozygosity analyses and comparative genomic hybridization (12, 34). This suggests that the larger number of aberrations in WT tumors will affect additional cancer genes, and this may be the cause of their more aggressive behavior. In this study, FGFR3 urine analysis predicted the recurrence in three of five cases with progression. In the other two cases where FGFR3 analysis did not predict the recurrence, only one urine sample was collected before the tumor, in contrast to the previous three cases where multiple urine samples were available. This suggests that collection of multiple urine samples could improve detection of recurrences.
In summary, patients with a low-grade FGFR3-mutant NMI bladder tumor represent >50% of all patients first diagnosed with BC (15). About 70% of these patients will develop recurrences, although their progression risk is low (11, 13 14, 22). Yet they need long-term and costly follow-up by repeated cystoscopic monitoring. The data presented here suggest that three monthly urine testing alternated by, for instance, yearly cystoscopies could serve as an alternative follow-up approach for these patients. Patient stratification by FGFR3 mutation status and such a surveillance schedule could substantially reduce the total costs, improve quality of life, and even lead to an earlier detection of recurrences, thereby preventing possible progression. Additionally, UUT recurrences are detected with this assay, which can be considered an added confidence in the diagnosis of BC. These findings are promising but preliminary due to the small number of patients and show the need for large, randomized, controlled trials to determine the feasibility of FGFR3 mutation analysis for the detection of recurrences during surveillance of patients presenting with low-grade FGFR3-mutant NMI bladder tumors.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
We thank all the urologists and pathologists who participated in this trial.
Grant Support: Koningin Wilhelmina Fonds grant 2006-3672.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received November 11, 2009.
- Revision received March 22, 2010.
- Accepted April 12, 2010.