Final Report Summary - UROMOL (Prediction of bladder cancer disease course using risk scores that combine molecular and clinical risk factors)
Executive Summary:
The main objective of the UROMOL project was to predict bladder cancer disease course using risk scores that combine molecular and clinical risk factors. We tested new and highly promising molecular predictive markers in an optimized fashion including a prospective multicenter collection of tissues and urines, in a standardized set-up with the same standard operating procedures used in each clinical or testing centre.The objective was to identify molecular markers that will predict the likelihood of progression in patients with Ta or T1 bladder tumors. In those with minimal or high likelihood of progression, reduce the number of cystoscopies or increase surveillance and treatment, respectively. Furthermore, urine based biomarkers should be tested as diagnostic tools and as predictors for later tumor recurrence. We should develop a disease model and construct nomograms and algorithms combining the molecular biomarkers with known risk factors leading to a simple risk calculation that can be used in a clinical setting.
The platforms used were originally microarray based, but we decided to shift this to next generation sequencing as this would bring much more information. The UROMOL consortium has had six coordinating meetings in Barcelona, Aarhus, Rotterdam, Uppsala, and Vic, as well as numerous telephone conferences. The standard operating procedures needed for implementing tissue collection have been mutually agreed upon, and expanded with new technologies as they mature. A database for www application has been implemented in all centres that collect tissues, the staff has been trained at each center and clinical data has been entered.
Collection of tumor and blood samples was completed in august 2012, and we included a total of 3042 patients in the study. The patients have been followed during 8011 visits to the different clinical centres, and 1289 Ta/T1 tumors, 3359 urine samples and 2298 blood samples have been collected during the project period. This was far more patients than planned, but deemed necessary to obtain a strong power for later analysis.
We tested the heterogeneity of tumour samples in patients where two or more tumors were available. This prompted us to analyse an mRNA signature in ALL tumors from each patient. The mRNAs were selected among 96 RNAs with prognostic information, and 12 ended up as being ideal for use in single tumors, having strong predictive power. These have been tested on all 937 samples with good RNA quality.
The urine assays have been directed towards microsatellites, methylation markers and mutations occurring in urine. These have been tested on a retrospective material of 2000 urines and are currently being tested in the prospective 3000 urines set. The data from the retrospective set looks very promising as it seems possible to reduce the number of cystoscopies, and replace some of these with urine testing. The SNP variations in the genome that predict poor outcome have been identified using 1 mill SNPs and showed promising prognostic power in retrospective validation. They are currently being tested in prospective validation. Disease models have been built and algorithms developed enabling construction of a nomogram that can take the risk information connected to each variable into a simple web based tool for clinical usage.
Together with the urine based tests this should further reduce the general frequency of cystoscopies of the bladder. In general we can obtain models with a sensitivity and specificity above 80% which is promising for clinical exploitation. The project is continued by the consortium based on other external funding, allowing for a longer follow-up period, and for analysis of the very large dataset. A final recommendation for clinical inclusion of molecular parameters will be produced mid-2014.
Project Context and Objectives:
Clinicians need improved tools for selecting treatments and follow-up regimens for individual patients. A plethora of prognostic factors, including molecular alterations, have seen light in the last 10 years. However, very few such factors are used in clinical practice and those mainly as target identification. One reason for this is that the cohorts of patients examined are too heterogeneous with regard to stage and treatment to support therapeutically relevant conclusions. Another drawback is that most prognostic studies develop new markers but do not test pre-specified models using data from independent patient samples. This is in sharp contrast to those who carry out clinical drug trials, as these are generally prospective, with patient selection criteria, primary end point, hypotheses, and analysis plan specified in advance in a written protocol. The number of variables a clinician has to consider before planning a treatment and follow-up plan for a given patient is constantly increasing. With more treatment options available, knowledge of the genetic composition of the patient, presence or absence of predictive markers, and the classic knowledge of clinico-pathological variables, the considerations are becoming complex, and often have to be carried out under time constraints. This proposal therefore aims at establishing mathematical predictive algorithms or nomograms that can help guide the clinical selection of therapeutic regimens and follow-up plans. Entering variables into computer software that report back risk estimates, therapy suggestions, follow-up intervals etc. would be a step forward, and has already been established for the classic clinico-pathological parameters at leading US centers. This strategy has also been successfully applied for the past 15 years in the management of patients with testicular tumors. We now feel that the time is ready for clinical implementation of the markers we have pursued as well as those pursued by others. This is a very difficult part of the research process as it requires
validated classifiers with pre-specified cut-off limits, a well defined clinically relevant end-point, standardized protocols, and an extensive prospective multicenter testing. Furthermore, we believe it to be of importance to incorporate the molecular markers, together with clinico-pathological parameters, in mathematical algorithms for assigning a risk score to the patients, as well as for suggestions of treatment regimens, and follow-up strategies, also considering the enormous medical costs of bladder cancer to society.
The bladder cancer disease
Bladder cancer is, in terms of incidence, the fifth most common neoplasm in industrialised countries, accounting for about 5-7% of all new diagnosed malignancies in men, and about 2-2.5% in women. In addition to male gender, acknowledged risk factors today include high age, tobacco smoking and occupational exposure to carcinogens. The prevalence (persons alive with bladder cancer at any given time) is three to eight times higher than the incidence, making bladder cancer one of the most prevalent neoplasms, and hence, a major burden for all health care systems. The overall cause-specific five-year survival rate is about 65%. Bladder cancer outcomes are directly influenced by social deprivation. Thus people with high age and poor socioeconomic status are especially vulnerable regarding this disease. By far, the most common histological form in the industrialised world is urothelial cell carcinoma (UCC). In regions where Schistosoma infections are endemic, squamous cell carcinomas account for a significant fraction of bladder cancer; otherwise they are less frequent (app. 5%). Squamous cell carcinomas almost exclusively have an invasive course. UCCs, in contrast, tend to occur in two principal forms: papillary non infiltrating tumors and solid infiltrating cancers. Earlier the nonmuscle invasive tumors were collectively called superficial and treated similarly. Over the last decade, it has become increasingly clear that this grouping was inaccurate as the superficial tumors prognostically consisted of two separate groups, non-invasive (stage Ta) of usually low histological grade and lamina propria infiltrating tumors of high grade (T1). To the later category belonged also intraepithelial high grade lesions (stage Tis). The most frequent form of all newly diagnosed cancers are of the stage Ta category which constitute more than half of all newly diagnosed cases. The prognosis of these tumors is good as more than 90% never progress to higher stages with conservative (bladder-sparing) treatment (Holmang, Hedelin et al. 1995; Zieger, Wolf et al. 2000). However, more than 70% of the tumors recur in the bladder, which makes this tumor type responsible for the high prevalence rate. The frequency of recurrences has a significant impact on the patients’ quality of life. Consequently, after transurethral (endoscopic) resection of the tumors, an attempt to prevent recurrences is frequently made by intravesical instillations of Bacille Calmette-Guerin (BCG), which non-specifically stimulates the patient’s immune response against the tumor cells. Various chemotherapeutic agents may be used alternatively. Although the treatment is effective in most cases, long-term recurrences may occur (Witjes, Caris et al. 1998). The surgical removal of the whole bladder (cystectomy) may be considered in selected cases of patients with very frequent recurrences, high grade lesions and failure of BCG treatment. Invasion of the suburothelial connective tissue, the lamina propria (stage T1), is evident at diagnosis in about 20% of all cases (Zieger, Wolf et al. 2000; Holmang, Andius et al. 2001). These tumors have a worse prognosis and most will be fatal if not aggresively treated. Despite this, almost 30% will die of the disease due to extravesical disease extension. Those with high grade and solid phenotype have a worse prognosis (Schrier, Hollander et al. 2004). Carcinoma in situ (CIS, stage Tis) is seldom diagnosed as the primary lesion; concomitant CIS is more common, and may be found in up to 40% of stage T1 cancers (Zieger, Olsen et al. 2002) and in 50% of the muscle invasive stages (Wolf and Hojgaard 1983). CIS is treated by BCG instillations (if necessary repeated or sustained). The treatment is effective in about 50% of patients and may also be given in an adjuvant setting after transurethral resection of a stage T1 tumor (Hurle, Losa et al. 1999; Lebret, Bohin et al. 2000). In case of failure, cystectomy is the treatment of choice. About 30% of stage T1 tumors of the CIS-type will progress to muscle invasion under a conservative (bladder sparing) regimen (Zieger, Olsen et al. 2002); without BCG, about 50% will progress (Cookson, Herr et al. 1997). Large tumors and tumors with multiple recurrences or widespread CIS have a very high risk of progression, so many of these cases are primarily offered radical treatment (Oosterlinck 2001). However, for most patients it is impossible today to predict the disease course. Muscle invasive bladder cancer is a fatal disease if left untreated. Although endoscopic and partial bladder resections may cure the disease in selected cases, the standard treatment today is the complete removal of the bladder and adjacent organs (radical cystectomy) with pelvic lymph node dissection and urinary diversion. This is an extensive surgical procedure with high postoperative morbidity and some mortality. External beam radiotherapy is an alternative radical treatment option, especially in case of co-morbidity that speaks against extensive surgery. Metastatic disease is treated by cisplatin-based systemic chemotherapy. Five year cause specific survival rates are stage dependent and rank from 40% (stage T2) to less than 10% (stage T4). This application has its focus on non-muscle invasive tumors. To aid in the clinical management of the heterogenous TaT1 group, the EORTC has developed a scoring system and risk tables (Sylvester, van der Meijden et al. 2006). The basis for these tables is the EORTC database which provided individual patient data for 2,596 patients diagnosed with TaT1 tumours who were randomized in seven EORTC trials. The scoring system is based on the six most significant clinical and pathological factors. There are two major limitations to this instrument. First, it needs to be validated in a separate cohort. Secondly, it is mainly applicable to recurrent tumors not at first diagnosis. Thirdly, the scoring system had only modest discriminative abilities, and new, recently established markers need to be incorporated in risk assessment.
Modeling the disease process and the use of look-up tables and nomograms for risk prediction – the mathematical approach to clinical routine decision making. A new era of systems biology translation into the clinic.
Cancer is such a complex process, involving a large number of molecular alterations, that it should be recognized as a systems biology problem. The aim of systems biology is to integrate the available knowledge at different levels of biomolecular organization, providing a more global picture of the whole process. This integrated approach offers new prospects for understanding thecomplexity of cancer. However, all this genomic knowledge will not have an impact into the effective treatment of cancer unless it is translated into the clinic in a pragmatic and practical way. To that effect, prognostic models for the risk of disease recurrence and progression including the genomic information in addition to the usual clinical and pathological prognostic markers will help clinicians in their risk assessment of bladder cancer patients. Evidence-based decision support is then possible for different treatment options considering risks, expected benefits and medical costs. An appropriate modeling of the natural history of bladder cancer is essential for obtaining efficient prognostic models of disease. This modeling is usually addressed using traditional survival analytical techniques. The chronic and particular evolution of bladder cancer adds another level of complexity that can only be addressed through more advanced methods. In particular, two scenarios that should be properly addressed are competing risks and recurrent events. While the former deals with several causes for a unique time to failure, the latter considers a type of event which may occur repeatedly over time. Regarding the first point, one of the important questions that we will consider is why some patients experience a progression as a first event after diagnosis instead of a recurrence, 5% of non-invasive tumors in the Spanish series; the median time to develop a recurrence or a progression as first event is similar: 9.3 and 9.7 months, respectively suggesting a different aggressiveness/biology. Differences in behaviour do not seem to be explained only by classical clinical and pathological prognostic factors. Which prognosticators/biomarkers characterize and differentiate patients who progress from those who experience a recurrence? In this case, where the variable of interest is time until the first event, a model for competing risks is needed. Other multivariate survival methods for event history analysis could be considered; in particular, specific methodology to address recurrent competing risks data, that is, multiple failure times and their corresponding causes of failure, is needed. These methods will be contrasted to traditional methods for survival analyses, including Kaplan-Meier curves and Cox regression models considering one end point at a time, specifically recurrence and progression.
Recently, the use of nomograms based on the multivariate regression parameters, has been shown to improve and facilitate predictions at specific points after treatment. A nomogram is a graphic representation of a statistical model that incorporates multiple continuous variables to predict a patient’s risk of developing a specific endpoint (recurrence, survival, complications, etc.). Different studies have demonstrated the superiority of nomograms in providing improved accuracy compared with risk group assignment techniques (Kattan, Stapleton et al. 1997; Kattan, Eastham et al. 1998). Nomograms for bladder cancer have been proposed (Karakiewicz, Shariat et al. 2006; Karakiewicz, Shariat et al. 2006; Sylvester, van der Meijden et al. 2006) incorporating only clinical and pathologic data or including some specific molecular marker. Their predictive accuracy could be improved by including different genetic markers: gene expression signatures, germline genetic variation and specific genomic alterations.
The working hypotheses were:
1. We can predict the disease course of bladder cancer patients having non-muscle invasive tumors with a multigene expression signature in tumor tissue. This prediction becomes more exact when we take polymorphisms in inflammation genes into consideration.
2. We can diagnose bladder cancer recurrence using a urine based test that predicts if a tumor is present at a given visit to the clinic.
Towards this goal, the following objectives were pursued:
1. A prospective protocol for patient inclusion and handling of samples was constructed and standard operation procedure (SOP) for handling of samples in the laboratory and the analytical procedures was developed, as well as software for registration.
2. Prospective multicenter collection of tumor, blood and urine samples together with detailed follow-up information from patients with bladder cancer in Denmark, Sweden, The Netherlands and Spain, and later on including Germany and Serbia.
3. Development and validation of nomograms that weight the individual parameters and provide a continuous risk score with clearly interpretable end-points such as risk of progression or risk of recurrence.
4. Analysis of biological differences in synchronous and metachronous bladder tumors, for selection of analysis strategy.
5. Transfer of the bladder cancer outcome prediction signature to a RT-PCR platform and standardization against a training set of >200 tumors.
6. Validation of prognostic gene expression and miRNA signatures in prospectively collected bladder tumors using QPCR and the nolve and very recently implemented Next Generation Sequencing (NGS) technology.
7. Validation of SNP´s in inflammatory genes in prospectively collected bladder tumors using SNPlex platforms as well as high density SNP arrays.
8. Validation of urine biomarkers (microsatellites, FGFR3, PIK3CA, methylation) for longitudinal monitoring of recurrent UCC as well as for prognosis prediction and treatment selection. This was split into a retrospective part as well as a prospective part.
9. Development of a decision model that allows urologists and pathologists to stratify patients with non-invasive UCC for distinct follow-up regimes and in which a considerable percentage of follow-up cystoscopies are replaced by DNA (urine-derived) based assays.
10. Applying nomogram risk assessments, and other developed algorithms, on all patients included and linkage to clinical outcome. Development of a software program for implementation in the clinic and distribution of guidelines for use of the molecular bladder predictors.
Project Results:
WP2. Protocols for patient inclusion and for handling of samples and development of standard operation procedures (SOP) for handling of samples in the laboratory and the analytical procedures.
Standardization of all procedures is a very important issue in this project. Consequently, we have developed a protocol for patient inclusion and protocols for tissue, blood and urine sample handling and processing have been written and distributed, as well as implemented, by the consortium partners.
Standardization also includes the development of standard operating procedures (SOPs) for sample handling and processing and for analytical procedures in the laboratory. SOPs for sample handling and analytical procedures are being implemented when needed in the project. We have developed protocols for PCR based mutation detection of FGFR3, PIK3CA and RAS mutations in tumors and urine sediments. Furthermore, protocols for microsattelite analysis and methylation analysis in urine samples have been implemented. In addition, we generated protocols for RNA, miRNA, and DNA extraction together with a protocol for QC of RNA using the Bioanalyzer. Recently, we have developed a protocol for analysis of the 12-gene PCR signature, that we have decided to run on all samples, and are protocols for NGS in general and RNA-SEQ analysis in particular. The latter technology is new and did not exist at the time of writing the proposal. However, we believe it will add important angles to the project as it can replace the microarray originally planned (covers all transcripts), as well as provide non-coding RNA data in the same run. All protocols and SOPs are uploaded to our web site www.uromol.eu and available for all consortium members.
WP3. Collection of tumor, blood and urine samples from patients with bladder cancer in Denmark, Sweden, Holland and Spain
The objective of WP3 was to collect tumors, urine samples and blood samples from a total of 2000 patients enrolled in hospitals in Denmark, Sweden, The Netherlands and Spain. The infrastructure was established in the connected hospitals, ensuring that we could obtain tissue in a good quality and in a timely fashion, as well as blood and urine. The clinical data are being extracted from patient records and all data entered into our web-based database, as described in the previous report.
Collection of tumor and blood samples was completed (august 2012), and we have included a total of 3042 patients in the study. The patients have been followed during 8011 visits to the different clinical centres, and 1289 Ta/T1 tumors, 3359 urine samples and 2298 blood samples have been collected during the project period. The collection of urine samples ended in January 2011 as we had reached the number needed for the project.
In order to reach the desired number of patients for the study we have, based on an intensive effort, included more clinical centres in the study. The clinical centres encompass: Denmark (Aarhus, Copenhagen, Aalborg); Sweden (Uppsala); Holland (Rotterdam); Spain (Madrid, Barcelona); Germany (Munich, Jena, Erlangen); Serbia (Belgrade). Responsible persons in each center have been trained in database usage for the project. Furthermore, the protocols for tissue collection and sample handling have been implemented in all centers. Table 1 lists the inclusions of patients and collections of Ta and T1 tumors, blood and urine samples in each center. As shown we have successfully reached the aim of including more than 1200 Ta and T1 tumors in the project.
We are not aware of any other research driven diagnostic project in Europe that has such a large size and such a large power.
Finally, we have mailed a representative diagnostic microscope slide from all patients to prof. Ferran Algaba in Barcelona (partner 8), for a central review of stage and grade. This review has been completed for most centres (results from two centers are still missing, but will be available in the fall 2013).
WP4: Analyse biological differences in synchronous and metachronous bladder tumor
For this project it is important to know how stable our molecular progression classification is when patients present with multiple tumors at several visits prior to a potential disease progression event.
Earlier in the project we have analyzed metachrounous tumors from 17 patients using Affymetrix Exon arrays. A total of 39 tumors have been analyzed and the 88-gene progression signature genes were identified on the Exon array platform. First the probe sequences were re-annotated using BLAST (NCBI) against ref-seq sequences. Then the gene symbols were identified on the exon array platform from www.netaffex.com (Affymetrix web site for microarray annotation). Exon array data were normalized and gene expression measures generated using the interPlier method implemented in Genespring 10.0. The most specific and the probe sets with most probes were selected, when more than one probe set for a gene was present on the Exon array. Ten previously analyzed samples were re-analyzed on the exon array platform for re-generating and training the progression classifier using a script in R (www.r-project.org) previously used for the classification. All 39 samples were classified using the classifier. All of this was described in the previous report.
We have now analyzed the mutation status of FGFR3, PIK3CA and RAS genes in the 39 tumors in order to judge the consistency of the mutation status during the disease courses.
Earlier we reported a high consistency in the classification results when applying the 88-gene molecular classifier to metachronous tumors from the same patients. 71% of the patients showed similar tumor classification, indicating that the biological properties of the tumour may change over time. The mutation data obtained from analyzing the FGFR3, RAS and PIK3CA genes showed no difference in mutation status when comparing the mutation status in several tumors from the same patients.
Consequently, we conclude from this and previous work that it may be optimal to analyze all tumors from patients with bladder cancer as they occur during the disease course. However, due to the high cost of the microarray analyses/RNA-SEQ analyses we will initially only analyze one tumor per patient using the RNA-SEQ platform, but all tumours using the much cheaper targeted qRT-PCR platform. In this way we may be able to make better clinical guidelines on how the test should be used ultimately, which is one of the main purposes of the UROMOL project.
The results have been descried in the publication: Dyrskjot L, Reinert T, Novoradovsky A, Zuiverloon TC, Beukers W, Zwarthoff E, Malats N, Real FX, Segersten U, Malmstrom PU, Knowles M, Hurst C, Sorge J, Borre M, Orntoft TF. Analysis of molecular intra-patient variation and delineation of a prognostic 12-gene signature in non-muscle invasive bladder cancer; technology transfer from microarrays to PCR. Br J Cancer. 2012. Epub 2012/09/15.
WP5: Transfer of microarray based gene signature to QPCR platform and microarray and QPCR based classification of samples
In order to take advantage of the dynamic range of PCR based methods and for making clinical implementation easier, we have transferred several markers from the microarray based gene signature to a qRT-PCR platform. Consequently, all tumors collected in the project will be classified using both a high-throughput method and the qRT-PCR method.
Transfer to PCR platform: Initially we selected 96 genes associated with disease aggressiveness from published gene signatures (38 genes) and from reanalysis of microarray data (58 genes). Following primer design and initial validation of primer sensitivity and specificity we decreased the number of interesting genes to 35 in total. We performed qRT-PCR measurements of the 35 genes using 115 tumor samples from patients in Denmark, Sweden, UK and Spain. Previously, 102 of these tumors were used for validating our microarray based 88-gene signature. We selected the genes for inclusion in the optimal PCR signature by comparing normalized Ct values and clinical outcome using Pearson correlations, ROC analysis, Wilcoxon signed-rank tests and Cox regression analysis. Using these different statistical methods in a combined approach we selected 12 genes that showed significant correlation to outcome (FIG-A). Computation of risk scores utilized non-normalized Ct values based on the formula: average (Ct (genes down regulated in progressing tumors) - average (Ct (genes up regulated in progressing tumors)). Molecular prediction of outcome was carried out for 115 patients based on the generated PCR risk scores.
Classification of samples: All collected samples in the UROMOL project have been analysed using the 12-gene PCR signature risk score as described, and we have applied the two different cut-off values.
Because of technological advances in sequencing we have decided to shift from microarray analysis of the samples to RNA-SEQ analysis. This will give us additional information on SNPs, indels, translocations, splice variants, non-coding RNA etc. Furthermore, we do not have to run specific assays for non-coding RNAs as they are included, if of sufficient size.
We are currently performing the RNA-SEQ analysis of the samples (one tumor is analysed per patient if clinical information together with follow-up have been provided).
Transfer to PCR platform: An optimal cut-off value (0.79; 76% sensitivity and 86% specificity) was identified using ROC analysis (FIG-B) and the dichotomized scores showed significant correlation to progression free-survival (P<0.0001 FIG-C). In multivariate Cox regression analysis we found that the 12-gene signature was an independent prognostic variable (HR=7.4 [95% CI: 3.4-15.9] P<0.001) when adjusting for stage, grade and BCG/MMC treatment. When only including high-risk patients in the analysis (i.e. stage T1, or high grade, or concomitant CIS) we also found a significant correlation to progression free-survival (P<0.0001 log-rank test, FIG-C). Similarly, when including only the low risk subpopulation in the analysis we also found a significant correlation to progression free-survival (P=0.0002 log-rank test, FIG-D). We calculated the predictive accuracy using Harrell's C to 73% for clinical variables only, 75% for the 12-gene signature, and this number was raised to 82% when including clinical variables and the 12-gene signature in the model. Consequently, accurate prediction of progression is increased when including the molecular markers.
When changing the cut-off value to -0.17 to ensure a 90% sensitivity (and 61% specificity) of the test we also observed significant correlations to outcome for the entire patient cohort and for the high and low risk subgroups (FIGS- F-H), see Figure 1.
The results have been descried in the publication: Dyrskjot L, Reinert T, Novoradovsky A, Zuiverloon TC, Beukers W, Zwarthoff E, Malats N, Real FX, Segersten U, Malmstrom PU, Knowles M, Hurst C, Sorge J, Borre M, Orntoft TF. Analysis of molecular intra-patient variation and delineation of a prognostic 12-gene signature in non-muscle invasive bladder cancer; technology transfer from microarrays to PCR. Br J Cancer. 2012. Epub 2012/09/15.
Classification of samples:
We have extracted RNA from all tumors and measured carcinoma cell percentage and RNA quality (RIN). For the 12-gene QPCR assay we required a carcinoma cell percentage >10% and a RIN score >5. Furthermore, 500 ng of total RNA was required for the analysis. This resulted in a total of 937 samples that fulfilled the criteria for this analysis. All samples were analysed using an ABI7900 QPCR platform and the 12-gene risk scores were calculated for all samples as earlier described, and previously determined cut-off values were applied. Final follow-up for all patients, within the granting period, was available July 2013, and the data is currently being extracted from the database and finalized for analysis. Follow-up information from Spain and Erlangen is not available yet, but is expected to be available very soon. Therefore, we have instead correlated the molecular risk scores to clinical risk parameters to determine the potential prognostic value of the risk scores (clinical risk was defined as: stage T1 or high grade high risk, otherwise low risk). In the Figure 2 the samples with low molecular risk scores are listed to the right side of the red bar (optimal cut-off value), and the samples with high molecular risk to the left side. Furthermore, the samples were sorted according to risk score values. The black bars below the heat map denote samples from patients with clinical high risk. Importantly, we found a high correlation between clinical risk and molecular classifications (p=1.0E-22 Chi2 test), and we therefore expect that the 12-gene QPCR assay also will be significantly correlated with disease progression (high clinical risk is associated with disease progression).
When all follow-up data is available we will analyse all risk scores according to progression status and currently, we are performing RNA-Seq on one sample per patient with follow-up information available, where the carcinoma cell percentage is >50% and RIN >5. Furthermore, >500 ng total RNA is required. We have implemented a SciClone robot for faster and easier library construction, and we use ScriptSeq directional paired-end sequencing on the Illumina HiSeq 2000 instrument. We expect to sequence a total of approximately 500 samples and this work will be finished during the fall 2013. So far we have analysed some of the sequenced samples to determine if we can observe similar gene expression as observed using QPCR for the 12 test genes. Figure 3 shows the 12 genes from the QPCR risk assay using RPKM values obtained from the RNA-Seq data. Samples are sorted according to risk scores from the 12 gene QPCR assay. A clear correlation between the QPCR assays and the RNA-Seq data is observed, as the high risk genes are highly expressed in the high risk signature and vice versa.
WP6: Analyse microRNA signatures for prediction of the outcome in bladder cancer
The aim of WP6 is to validate if already identified miRNA molecules that correlate with clinical outcome do add additional predictive value to constructed nomograms.
Recently, other non-coding RNA molecules have been identified that have been documented to be involved in cancer development and progression of other cancers. In bladder cancer we are currently analysing large datasets of non-coding RNA expression from microarrays and RNA-SEQ analysis. Instead of just focusing on miRNA expression in this project we will instead focus on the broader group of non-coding RNAs. We are currently investigating the non-coding RNAs involved in bladder cancer, and these will be validated in the UROMOL material. This information on non-coding RNA expression in the UROMOL tumor material will be available from the RNA-SEQ experiments that are currently being performed on the collected samples. Validated non-coding RNAs will be imputed into constructed nomograms to determine the added clinical value of using non-coding RNAs for predicting outcome.
WP7: Analyse SNP’s in inflammatory genes associated with risk of disease
The initial aim of WP7 was to validate SNP´s in inflammatory genes in prospectively collected bladder tumors using SNPlex platform. In the meeting held in Aarhus in December 2008, the Consortium decided to carry out the first validation step with the retrospective series that each group has established previously (Rotterdam, Aarhus, Uppsala) in addition to the samples from the prognostic model development phase from Spain and two other groups who have gathered bladder cancer patients with high quality clinical follow-up and leukocyte DNA (Y. Allory, Paris; G. Steineck, Stockholm; P. Matullo, Torino). In all, we estimated that we would be able to analyze at least 1000 additional patients.
Partners 2 and 9 moved ahead in identifying additional SNPs predicting both recurrence and progression through different state-of-the-art strategies. These initiatives included a TagSNP-candidate pathway GoldenGate Illumina genotyping (768 SNPs in inflammatory genes and 768 SNPs in oncogene/tumour suppressor pathways involved in bladder cancer development) and a 1 million-probe Infinium Illumina chip.
Partner 4 has applied data-mining methods (Random Forest and MB-MDR) for the identification of gene-gene interactions and gene patterns associated with different curses of the disease.
A meta-analysis with the 150 most significant SNPs identified in the Spanish Bladder Cancer Study and in the phase I Houston MD Anderson Cancer Center (X. Wu) for each outcome of interest was first carried out. The Spanish series (n=836) and the US series (n=496) differed, as expected, in the distribution of cases according to stage and grade since the MDACC is a referral center. This cross-validation step allowed increasing the validity of the associations identified and producing robust data that should be validated in our prospective cohort (and possibly in other cohorts as well). However, it was felt that an additional replication step was required based on the tremendous heterogeneity of treatment and outcome of patients coming from the different centers participating in the UROMOL study, an issue that was identified during the review of all information from our retrospective analysis (see WP9). Therefore, we selected for the replication phase two patient series: 1) one, comprising patients from UROMOL partners (Rotterdam, Aarhus, Uppsala) and those from other collaborating EU colleagues (Y. Allory, Paris; G. Steineck, Stockholm; P. Matullo, Torino), and 2) a second set from MD Anderson Cancer Center. For the >1500 additional cases included in this phase we genotyped 111 SNPs using TaqMan assays and a combination of mid-throughput platforms. The results from these studies, summarized below, point to the crucial challenges that need to be met in order to develop strategies for outcome prediction in patients with non-muscle invasive bladder tumors. rs754799 was significantly associated with recurrence (combined HR=2.06 95% CI=1.55-2.75; P =7.45x10-7). rs4246835 was associated with progression (HR= 0.49 95% CI=0.37-0.64; P=1.77x10-7). Six additional SNPs had similar effects on progression in stage II data as stage I. rs754799 was also significantly associated with “any-event” (HR=1.89 95% CI=1.43-2.49 P=5.89x10-6). Two loci in chromosome region 6q22 and 12p12 also exhibited similar effect on the risk of “any-event” in stage II data as stage I.
Few, if any, genome wide prognosis association studies have been carried out with this depth. Our findings point to the need to perform several replication phases rapidly and in homogeneous clinical settings in order to achieve results that can be exploited in the clinical setting in an efficient manner and to expand the genetic analyses in order to increase the power to identify useful predictors. Because of these caveats, identified throughout the study, and thanks to the possibility of co-funding from other grants obtained by partners 2 and 9 who are mainly responsible for this WP, a decision was made to expand the genetic analyses to a larger set of SNPs with genome wide coverage. This decision has somewhat slowed the completion of the genotyping but it has not compromised in any way when the deliverables can be actually provided, because the final analyses require the ongoing completion of collection of follow-up data. On the other hand, this strategy will place the UROMOL consortium in a unique position to deliver information that cannot be provided by other consortia whose studies have significant biases in patient selection. Genotyping is ongoing and will be completed
WP8: Validation of urinary markers for diagnosis of recurrent UCC and development of a decision model for surveillance in which DNA tests replace cystoscopies
The retrospective urine study has been completed at month 24. All laboratory analyses for the prospective study (deliverable at month 48) have been completed. Data analysis is ongoing (see further below).
To determine a combination of markers with optimal sensitivity for the detection of recurrences in voided urine.
Patients with NMIBC (n=147) were included at trans-urethral resection of the primary tumor. At least three follow-up urine samples were required for patient selection, including a urine sample collected prior to resection of the primary tumor. DNA was extracted from formalin-fixed paraffin-embedded tissue of the primary tumor and the cell pellet of the collected urine samples. FGFR3, PIK3CA and RAS mutation analysis were performed on the tissue and urine DNA samples, followed by micro-satellite analysis (MA) and methylation analysis.
All molecular tests had a higher sensitivity for detection of primary tumors compared to detection of recurrent tumors; FGFR3: 75% vs. 68%; MA: 83% vs. 67%; methylation: 84% vs. 69% and cytology: 61% vs. 41%. Combining FGFR3 with PIK3CA and RAS assays improved the sensitivity for the detection of recurrences from 68% to 72%. FGFR3/MA increased the sensitivity from 68% to 79% and adding the methylation analysis to FGFR3 increased the sensitivity to 73%. All single and combined molecular tests had a higher sensitivity for the detection of primary and recurrent tumors than urine cytology. Conclusions: A combination of markers increases the percentage of patients eligible for urine-based follow-up, and increases the sensitivity of recurrence detection. In addition we find that urine cytology is not suitable for the follow-up of low stage and grade NMIBC patients.
We aimed to determine if FGFR3 mutation analysis on voided urine samples is cost-effective to partly replace cystoscopy in surveillance of patients. We analyzed data on surveillance with FGFR3 analysis on voided urine samples from 70 Dutch patients with FGFR3 positive primary tumors. Surveillance strategies were compared in a Markov decision analytic model that included estimates from the Dutch cohort and from other data sources. Modified surveillance consisted of FGFR3 mutation analysis on voided urine samples every three months and a cystoscopy at 3, 12 and 24 months, while standard surveillance consisted of cystoscopy every three months. Analyses were stratified for three different risk profiles (primary tumor, first to third recurrence, and fourth recurrence or more). Sensitivity analyses were performed to evaluate the impact of variation in costs, sensitivity and specificity of the FGFR3 mutation test and cystoscopy. The probability of being without recurrence (‘well’) after two years of surveillance was similar for all three risk profiles in the 2 surveillance arms. There was a slight advantage for the modified surveillance arm compared to standard surveillance.
Based on the retrospective study we decided not to use the MA analysis (requires too much DNA) nor the MLPA methylation assay (technology was found to be problematic for urine-derived DNA). Since the SnaPshot technology for FGFR3 mutation detection is in our opinion the best for analysis of urine-derived DNA, we decided to also use this technology for methylation markers. To obtain highest possible sensitivities for methylation, we selected 8 CpG islands from our genome-wide methylation study (Kandimalla, European Urology 2012) for detection of recurrent bladder tumors in voided urine. Sensitivity and specificity were determined on 100 urine DNA samples obtained before tumor resection (preTUR, test set) and 70 samples from age-matched non-cancer controls. Using logistic regression, the best combination of 3 markers was selected. A 3-plex methylation assay was then developed based on the SnaPshot technology and this assay was validated on an independent set of 103 preTUR urines (validation set). The three gene methylation panel OTX1, ONECUT2 and OSR1 identified recurrent bladder tumors in voided urine with a sensitivity of 73% at a specificity of 90% with an area under curve (AUC) of 0.85 (CI: 0.80-0.91 P< 0.0001). Combining the multiplex methylation assay with the FGFR3 mutation assay resulted in a sensitivity of 78% at a specificity of 90% with an AUC of 0.87 (CI: 0.82-0.93 P< 0.0001). The methylation assay is more sensitive than cytology and the FGFR3 assay both for the detection of low and high-grade recurrent tumors (Kandimalla Clinical Cancer Research 2013). Based on these results, we decided to combine the 3-plex methylation assay and the FGFR3 mutation assay for the analysis of the urines of the prospective study.
Early in 2013, point mutations in the TERT promoter were found by our collaborators (Real, Malats, Allory) to be frequent in bladder tumors. We then set up a SNaPshot assay for these mutations and analysed tumors and urine samples from Rotterdam and Spain. TERT mutations were observed in up to 80% of bladder tumors (n=468) regardless of stage and grade and not related to any clinical outcome. Mutation analysis on urines obtained prior to resection of a primary tumor detected the tumor in 62% and in 48% of the urines associated with a recurrence. Combination with FGFR3 mutation analysis increased sensitivity to 70 and 50%, respectively (Allory et al, European Urology 2013, in press). Based on these results we decided to include the TERT assay in our urine assay panel for analysis of the prospectively obtained urine samples using funds obtained from other sources.
Patient materials and urine analyses
The consortium has collected a total of 2728 urine samples from 1222 patients under surveillance for recurrences after resection of a primary NMIBC. The distribution of the urine samples with respect to stage/grade of the inclusion tumor and moment of collection is depicted in Figure 4. DNA has been isolated from all samples. FGFR3 and methylation analysis have been performed on all samples. We also performed mutation analysis of the PIK3CA and HRAS, NRAS and KRAS genes on a subset of urine samples obtained from patients with one of these mutations in the inclusion tumor. This stratification is needed since these mutations are relatively rare. In addition, we assayed the mutations in the TERT gene promoter as explained above on all samples from which sufficient DNA was left. Because of the wish to add the TERT mutation to our portfolio, the data analysis was postponed and is now underway. We expect this to be finished within 2013.
Urine diagnostic tests for monitoring of patients with a previous NMIBC should be validated on urines obtained from this patient category. Therefore, the upfront selection criteria were that the inclusion tumor should be Ta, T1 and G1 or 2. In addition, we did not want to include urine samples collected before resection of a primary tumor. The reason for this is that primary tumors are usually larger than recurrent tumors and that their stage and grade may be higher than G1/2. Hence sensitivity of tumor detection will be higher if such samples are included. As is evident form Figure 4, 1674 urine samples fall within our selection criteria. Obviously, we have also collected urine samples taken before removal of a primary tumor and from patients after resection of a high-grade primary tumor. We have analysed these latter samples as well. Thus we will also be able to provide sensitivity/specificity of our assays regarding detection of primary tumors, which may provide answers as to whether these assays are of use to identify bladder tumors in patients presenting with hematuria or for future population screening purposes. Data analysis of the assay outcomes in the group of urines taken from patients with a previous high grade primary tumor may provide insight into the potential use of urine diagnostics in the follow–up of these patients who are at risk of progression to MIBC. As a spin-off we determined whether the methylation markers together with FGFR3 could be of use in the identification of patients with bladder cancer when presenting with macroscopic or microscopic hematuria. Logistic regression analysis based on the five methylation markers, age, gender, type of hematuria resulted in an area under the curve (AUC) of 0.88 and an optimism corrected AUC of 0.84 after internal validation by bootstrapping. Using a cut-off value of 0.307 allowed stratification of patients in a low-risk and high-risk group, resulting in a sensitivity of 82% (44/54) and a specificity of 82% (94/115). Most aggressive tumors were found in patients in the high-risk group. The addition of cytology to the prediction model, improved the AUC from 0.88 to 0.89 with a sensitivity and specificity of 85% (39/46) and 87% (80/92), retrospectively.
This newly developed prediction model could be a helpful tool in risk stratification of patients presenting with painless hematuria. Accurate risk prediction might result in less extensive examination of low risk patients and thereby, reducing patient burden and costs.
WP9: Modelling the disease process: Develop algorithms that weight the individual clinical and molecular parameters and provide continuous risk scores with clearly interpretable end-points such as risk of progression or risk of recurrence
The objective of WP9 was to obtain predictive survival models including both clinical-pathological and molecular characteristics for the risk of tumor recurrence and progression for non-muscle invasive bladder cancer. It was first planned that the prognostic models for recurrence and progression should be developed using the data already collected in the retrospective series established by the participating partners from Aarhus (Partner 1), Rotterdam (Partner 5), Uppsala (Partner 7) and Spain (Partner 9). The phases of the modelling have considered: 1) the validation of the nomogram reported by Richard Sylvester (Eur Urol 2006); 2) the building of an extended model (basic-UROMODEL) fitting the consortium gathered data; 3) the inclusion of the biomarkers in this last model (marker-UROMODEL); and 4) the exploration of innovative survival analysis (multistage and competitive models). We have generated:
1. A database with the edited clinical and follow-up data from the retrospective series of Partners 1, 5, 7 and 9 was delivered to Partners 4 and 5 for analysis. Conducted by Partner 9.
2. A validation analysis of the nomograms reported by Richard Sylvester (Eur Urol 2006) and the CUETO Group (J Urol 2009) was conducted. Results were discussed in two teleconferences and in the annual meeting held in Madrid (January 2013). A manuscript describing these findings is under revision in Eur J Cancer. Conducted by Partner 5.
3. Data from the selected markers (111 SNPs - please see WP7 report - and FGFR3 and Ki67 expression from cases from Aarhus, Uppsala, Aarhus, and Spain) was delivered to Partner 4. Conducted by Partner 9.
4. Using the basic disease model (basic-UROMODEL) build with clinical and pathological parameters, an integrative model with biomarkers (marker-UROMODEL) has been accomplished by Partner 4.
Further actions:
1. The distribution of the 1,974 cases from the participating centres (Aarhus, Rotterdam, Uppsala, and Spain) according to the basic study variables (BasicJoint-DB) was provided in previous reports. This information was delivered to Partners 4 and 5 by Partner 9 following European standards of data privacy and confidentiality of personal data.
2. 1,915 patients with primary stage Ta or T1 NMIBC who underwent a transurethral resection in Spain (n=995), the Netherlands (n=639), and Denmark (a selected set of n=281) were considered in this validation study. We evaluated recurrence-free survival and progression-free survival according to the EORTC nomogram and the CUETO risk score for each patient using c-statistics to indicate discriminative ability. The 3 cohorts were comparable according to age and sex, but patients from Denmark had the highest stage and grade at start of follow-up (p<0.01). At least one recurrence occurred in 855 patients and 233 patients had a progression during a median follow-up of 77 months for Spain, 50 months for the Netherlands, and 64 months for Denmark. Patients from Denmark had the highest recurrence and progression rates (74% and 24%, respectively), whereas patients from Spain had the lowest rates (33% and 10%, respectively). The EORTC and CUETO risk scores both predicted progression better than recurrence, with c-statistics for progression ranging from 0.72 to 0.81 and for recurrence from 0.57 to 0.61. The EORTC and CUETO risk scores can reasonable predict the occurrence of progression, while prediction of recurrence is more difficult. New risk markers are needed to better predict recurrence of tumours in non-muscle invasive bladder cancer patients. A manuscript reporting these results is under revision in Eur J Cancer. This work has been completed by Partner 5.
3. Data from the selected markers (111 SNP genotypes, and FGFR3 and Ki67 expression) from cases from Aarhus, Uppsala, Aarhus, and Spain was completed and delivered to Partner 4 for analysis. This work has been coordinated by Partner 9.
4. Predictive models have been built by Partner 4 where biomarkers were added to the classic clinical and pathological parameters for the different outcomes of interest: event-free survival (EFS), recurrence free survival (RFS) and progression-free survival (PFS). The predictive models agreed to consider were:
a. “basic model” (including clinical and pathological parameters)
b. “basic model” + SNPs
c. “basic model” + Ki67
d. “basic model” + FGFR3
e. “basic model” + SNPs, FGFR3 and Ki67
The analysis was restricted to 10 years of follow-up. A stratified Cox model was considered accounting for the differences among the series. The effect of SNPs was included as a score variable corresponding to the different multiple genotypes. The results for the different outcomes of interest are summarized in Figure 5.
In summary, the considered biomarkers (11 SNPs and FGFR3 and Ki67 overexpression) only contributed significantly to the improvement of the classical clinico-pathological model for the outcome of progression. The heterogeneity and the retrospective nature of the data and follow-up warrants of a further validation using the prospective study of the UROMOL project. This new and large series of patients will also provide the option to assess the contribution of additional biomarkers and improve the models explored. The model of bladder cancer outcomes assessed by applying advanced methods of survival analysis will be applied to all UROMOL cases. This will allow us to up-date the model of dynamic prediction with the inclusion of the rest of the series.
WP10: Apply developed nomograms risk assessment on all patients included and compare to clinical outcome. Develop software for clinical implementation of nomograms
Progress Summary
The objective of WP10 was to apply the predictive survival models for the risk of tumour recurrence and progression for non-invasive bladder cancer as obtained in WP9. In WP9, these models were referred to as “basic-UROMODEL” and “marker-UROMODEL”. A software solution for patients and clinicians was developed for the basic-UROMODEL for progression-free survival, since this is the most promising risk prediction model. Progression is a more relevant clinical event than recurrence. This Work Package was planned for the fourth period, with 1 month per partner, and 11 months for partner 5 (Rotterdam).
Risk assessment for all patients and comparison to clinical outcome
The marker-UROMODEL (nomogram, deliverable 9.2 30 months) for the three considered outcomes (event-free survival, recurrence-free survival, and progression-free survival) was accomplished with the retrospective common series of patients. For detailed results, please see appendix 9.2 9.3 and 9.4. Further validation of these models is planned in the prospective data set gathered during the project.
Software solution for clinical implementation
For outcomes on risk prediction models for event-free, recurrence-free, and progression-free survival, please see appendix 9.2 9.3 and 9.4. Progression-free survival was predicted best and was elaborated on for an online application (see Figure 6).
We developed a prototype online risk calculator, using dedicated algorithms from Cleveland Clinic (www.r-calc.com) see Figure 7. For convenience, Stage+grade is presented as a drop-down menu and tumoursize ≥3cm and multiplicity are presented as checkboxes to avoid mistakes. Further testing of usability of the on-line calculator will be done, also with a more complete set of risk models after validation in the prospective data base.
Potential Impact:
Strategic impact
We translate genomic technologies into novel cancer biomarkers with the purpose of improving early diagnosis of bladder cancer, as well as predicting the prognosis of this, and selecting patients for increased surveillance and early radical therapy, as well as patients for diminished surveillance and early cessation of follow-up. We expect that we can predict the response to therapy, that is primary surgical removal of bladder tumors, as some patients will actually be cured after the first treatment (around 50%), and do not need follow-up for extended periods, whereas others will need not only repeated surgery but also intra-vesical BCG installation. Furthermore, we expand the expected impacts by the “easy to use software” for general use that will calculate the risk of progression in bladder cancer using both clinical and molecular parameters. The data and algorithms will be updated within a few months when the final follow-up data has been analyzed.
The UROMOL project has furthered our understanding of the basic mechanism underlying (bladder) cancer.
We have obtained new insight into the bladder cancer disease as we will analyze the largest and most comprehensively collected material ever seen in the bladder cancer area. As mentioned above a serious draw-back to most previous translational research has been the diminutive sample sizes used and the lack of a pre-defined protocol. We are not performing a fishing expedition, but rather we are able to draw firm conclusions on the relation between our molecular markers and the disease course, seen e.g. as single genes or mutations (FGFR3) in single genes. We will also explore completely new areas as we have an exploratory side-project that will analyze non-codingRNA´s in bladder cancer and relate these to other expression data we have obtained by switching to the NGS platform. In addition, we will analyze the SNP´s in the genome that may control the hosts response against bladder cancer. This is a novel approach for the prediction on the disease outcome and opens a complete new avenue for clinical utilization. Thanks to our sample size we will be able to pin-point the potential benefit of this biomarker.
- Novel drug targets. We are not aiming at the identification of novel drug targets. However, the molecules that we determine and that predict the outcome of the disease may form new drug targets. As such we will end with a profoundly validated set of potential targets. We know that some of the molecules we work with have a pharmaceutical interest and that some of the major players in that field are working with e.g. Survivin as a drug target. Our data will show if a preventive or prophylactic use of such a drug might work based on the targets expression early in the disease. A very important aspect is that new drugs for treating bladder cancer can be tested in populations where we can utilize our risk predictors. In that way the effect of drugs can be related to the risk score and it can be decided whether the drug should only be offered to high risk patients or low risk patients. Some drugs may only have a benefit in a subgroup, and as new drugs are very costly it is highly relevant to identify these patient groups.
- New treatment strategies. The whole idea of including molecular biomarkers in the risk calculation for bladder cancer patients is to be able to plan and carry out new treatment strategies. However, we have to be realistic and the sequence of events can only be so that we 1) identify and validate the markers that in a routine clinical setting can predict if a tumor is present, and predict the disease outcome; 2) Then we change the treatment strategy based on the biomarkers using our nomograms in a clinical intervention study carried out according to good standards for these.
- We know that this is a time consuming strategy as validation may take up to five years (this proposal) and then follows clinical intervention studies. However, it can in our opinion, only be made in this way, and the problem with most other translational studies today is that
- They are too small
- They are carried out within a too short time period to be conclusive
- They want to do both validations of markers and intervention at the same time – which makes it difficult to draw any conclusions. We have overcome these problems in the present project.
The interventions we aim at are the following:
- If urine based testing signalize a tumor being present or a Carcinoma in-situ being present a cystoscopy should be made.
- If urine testing (preferably carried out as sampling in the home of the patient and mailing sample in fixative to the laboratory) signalizes no tumor, there will be no cystoscopy.
- If a tumor is removed and the Nomogram (based on urine testing as well as the tumor genomic profile and the blood genomic profile together with clinical parameters) signalize a high risk disease course the surgeon will initiate.
- Short cystoscopy intervals
- Frequent urine analysis sampling
- Intravesical BCG treatment
- Removal of the bladder in cases with repeated high risk tumors and a good general health status
Validated biomarkers
This is the main aim of the project. As already mentioned a plethora of biomarkers exists, with narrow information on their clinical usefulness. We provided profound information on some of these due to our large cohort size.
Improve patient care
We believe that a rational basis for decision making utilizing nomograms that can handle complex sets of variables is the future in clinical cancer care. In that way this approach is indeed an patient care improvement. As a consequence we included not only classical clinico-pathological parameters in our decision making but also biological knowledge form the genomic area. Some patients may only need follow-up for a short period of time, say once after 12 months. Others may need cystoscopy at 6 months intervals. The former group is saved from invasive examination, and society is saving a lot on health costs (see below in the range of 30-50 mill Euros annually).
Urine based testing has obvious advantages such as benign non-invasive and sampling may be performed at home. We will exploit this target in the final phase of the project, and we do have a sample size that is large enough to draw conclusions.
Reducing the incidence and mortality
Our nomogram is intended for identification of patients with a high risk of disease progression. These will receive a closer follow-up and earlier radical treatment. This should reduce the incidence of muscle invasive tumors, as patients at high risk are devoted more attention. Regarding mortality we believe that will be reduced too, if cystostectomy is used earlier than today to the right subgroup of patients, namely those at high risk. Today we have disseminated disease in more than 25% of those that have their bladder removed. These should have been treated earlier with cystectoymy. With the present nomograms we expect to have a tool that will form the basis for selecting those patients hat should have their bladder removed, thereby reducing mortality.
Improving quality of life for patients
Giving the right treatment to the right group of patients is probably the best improvement of life one can offer a patient group. Of course they should decide themselves if they wish the treatment or not. Repeated cystoscopies year after year has a severe impact on quality of life of the patients. Can we removed 10% of these by using urine based testing to pin-point the time for a cystoscopy and using a nomogram to define a tight surveillance or more relaxed surveillance program, we have saved a lot of pain, anxiety and 40 mill Euros.
Public health impact
Bladder cancer is a frequent disease, and one of the two most prevalent cancer diseases among males in the EU. People live with the disease for many years and thus burden the health system for many years (median 7 years). There are around 460,000 Cystoscopies in the EU annually. If we can take the low risk patients out of follow-up with cystoscopy after 12 months, we may save 1-2 cystoscopies for 50 % of the incident population with stage Ta and T1 bladder cancers, which we conservatively estimate to at least 26,000 Cystoscopies saved annually in the EU. If we can also reduce the frequency of cystoscopies in the high risk group by selecting patients for BCG treatment earlier, (reduced amount of recurrences by a factor of at least 0.3) and by removing the bladder earlier in high risk patients, we may save further 20,000 cystoscopies, leading to a total reduction in cystoscopies of 46,000 - amounting to a cost saving of 46 mill Euros. As we expect mortality to be reduced and thereby more people to be active both socially and at the working place this also adds to the impact on health, but is difficult to estimate correctly.
List of Websites:
http://www.uromol.eu/
Aarhus University Hospital, Skejby, Denmark
Prof. Torben Falck Ørntoft
e-mail: orntoft@ki.au.dk
Assoc. prof. Lars Dyrskjøt Andersen
e-mail: lars@ki.au.dk
Academic ass. manager Christina Bak Pedersen
e-mail: cbak@ki.au.dk
Universitat Pompeu Fabra, Barcelona, Spain
Prof. Francisco X. Real
e-mail: freal@cnio.es
Centro Nacional de Investigaciones Oncológicas, Madrid, Spain
Prof. Núria Malats
e-mail: nmalats@cnio.es
Assoc. prof. Laia Palencia
Universidad de Vic, Vic, Spain
Prof. María-Luz Calle
e-mail: malu.calle@uvic.cat
PhD-student Núria Porta
Erasmus Medical Center, Rotterdam, The Netherlands
Prof: Ellen C. Zwarthoff
e-mail: e.zwarthoff@erasmusmc.nl
Prof: Ewout Steyerberg
e-mail: e.steyerberg@erasmusmc.nl
Uppsala University Hospital
Prof. Per-Uno Malmström
e-mail: per-uno.malmstrom@surgsci.uu.se
Post doc. Ulrika Segersten
e-mail: ulrika.segersten@surgsci.uu.se
Fundació Puigvert, Barcelona, Spain
Assoc. prof. Ferran Algaba
e-mail: falgaba@fundacio-puigvert.es