Study highlights need for more accurate risk models in NMIBC

In this interview, Félix Guerrero-Ramos, MD, PhD, FEBU, shares key insights from the study, “Predicting Recurrence and Progression in Patients with Non-Muscle-Invasive Bladder Cancer: Systematic Review on the Performance of Risk Stratification Models,” for which he served as the lead author.¹ Guerrero-Ramos is the coordinator of the urologic oncology unit and the bladder cancer unit at Hospital Universitario 12 de Octubre, in Madrid, Spain.

This transcription has been edited for clarity.

Félix Guerrero-Ramos, MD, PhD, FEBU

Could you describe the background and rationale for this study?

We did this systematic review because we know that the identification of risk categories in our patients with non–muscle-invasive bladder cancer is of paramount importance. Depending on the risk category that you allocate your patient, they will receive a different therapy. For example, low-risk patients [may have] just follow-up, and high-risk patients and even some very high-risk patients [may receive] BCG or even radical cystectomy.

However, we know that the classifications have several limitations. That's why we performed a systematic review to assess the external validations of those classifications and to see the discrimination capability in the different risk groups. Just to set the scene, there are 3 stratification methods based on individual patient data with a statistic analysis, which are the EORTC, CUETO, and the EAU 2021 guidelines. There are 3 others based on expert consensus, which are the AUA guidelines, the NCCN guidelines, and the EAU guidelines in those classifications prior to 2021.

What were the key findings from this study?

We found that first of all, there was certain heterogenicity in the assessment of the outcomes and definitions. The definition of recurrence-free survival and progression-free survival were not homogeneous along all the validations. So, there's an important bias in interpreting the validations of these studies. We also found that the ability to discriminate the risk groups was quite poor, especially for recurrence, and was more accurate for progression. In this sense, we saw that most of the classifications overestimate recurrence, and some of them overestimate progression. There are some problems with this. There are some of these classifications, like EORTC, which is older, and patients were not treated with BCG; they did not receive a single instillation of mitomycin, or did not undergo a second TURBT when indicated. Same was for CUETO, which was designed for BCG-treated patients, but the BCG schedule in those patients is not the current one that we use. So, those patients received less BCG installations than we administer now to our patients. Also, those patients did not have a second TURBT.

One of the biggest issues that we found, and several other external validations have confirmed, is that the new EAU 2021 classification system is based on patients with a primary tumor (no recurrences) who were not treated with BCG. On the basis of that, we saw that they recommend an upfront radical cystectomy for those patients with very high-risk based on a risk of progression of around 59% at 10 years. There are some external validations assessing the value of BCG in this very high-risk group of patients. If you treat these patients with BCG, the progression rate at 10 years drops around 70%, so we might be overtreating a large percentage of our patients.

What are the implications of these findings?

One of the main conclusions that we can take from our study is that although the current classification systems are somewhat accurate, they are far from being perfect. We have to be aware of this in the clinic every day when we see our patients. Sometimes we might be including a patient in a risk category whose tumor is not going to behave like that category. For example, for all those high-risk and very high-risk patients where the guidelines recommend upfront radical cystectomy, we should individualize and take into account several factors. That way, some of the very high-risk patients could undergo BCG induction plus maintenance for 3 years and would avoid radical cystectomy. This is important when we make decisions.

Also, we have to take into account the limitations of all of the classifications. For example, if we're going to estimate the rate of recurrence and progression in a patient who we are going to treat with BCG, the EORTC classification system will not be valid because those patients were not treated with BCG. The same happens with the EAU 2021 stratification model, whose results give us more an estimation of the natural history of the disease rather than the real outcomes for patients when treated within the clinical setting.

Has there been any work done since the study, or is there any future work planned based on these findings?

There has been some work done before our study, and we speak about that. One of them is the publication of several molecular classifications. These molecular classifications have still several drawbacks. For example, most of them have not had external validations There's one molecular classification, where class 2a patients have a 20% risk of progression. That is not enough to recommend radical cystectomy for those patients, because you otherwise would be treating 80% of those patients with radical cystectomy, because 80% will not progress. Also, for molecular classifications, there are still certain heterogeneity between the classifications. The price of those techniques is expensive and we cannot perform all those tests in everyday practice.

In the future, these classifications may be implemented more. Currently, the only advantage I see for these classifications is the identification of several targets for targeted therapies. For example, if you identify mutations or fusions of FGFR, you could eventually use a drug for that. In the future, in our group, we're working with artificial intelligence and big data. We are gathering a large number of patients from many centers, and we're working with people from bioinformatics and computer science. They're helping us to develop more accurate models based on machine learning, with a great amount of data on many patients. We published some of this regarding the validation of the very high-risk population in the EAU 2021 classification, and we have more research in the process. In the future, I hope we can address some of the needs that we have in this part of the disease.

Reference

1. Guerrero-Ramos D, Subiela JD, Rodríguez-Faba Ó, et al. Predicting recurrence and progression in patients with non-muscle-invasive bladder cancer: Systematic review on the performance of risk stratification models. Bladder Cancer. 2022;8(4):339-357. doi:10.3233/BLC-220055