You’ve been looking at adjuvant chemotherapy in bladder cancer and this is an interesting area because neoadjuvant chemotherapy has already been established but largely is not being widely adopted. Put me in the picture, some of the background to this.

There have been two large randomised trials showing that neoadjuvant chemotherapy is beneficial for muscle invasive bladder cancer and so that is the standard of care and should be considered the standard of care. But we know from population-based studies that a lot of patients don’t get neoadjuvant chemotherapy and they don’t get that treatment for a variety of reasons. If you look at population-based studies, patients actually more commonly get adjuvant chemotherapy. The problem is that the randomised trials exploring adjuvant chemotherapy have faced multiple challenges over the past several decades. They’ve utilised sub-optimal chemotherapy; they’ve been underpowered and the three most recent studies aiming to definitively address this question all closed early due to poor accrual, enrolling only 39% of the planned subjects. So we’ve been unable to answer this question definitively using randomised data.

So what did you do? You’ve got a sample of around 4,000 patients, haven’t you?

What we did to try and fill this knowledge gap, at least in part, was to use population based data and the database that we used is called the National Cancer Database. This is a database maintained by the American College of Surgeons and the American Cancer Society and represents data on approximately 70% of incident cases of cancer in the United States, so it’s quite large. What we did was take patients who had undergone cystectomy, who had pathologic T3, locally advanced bladder cancer and/or node positive bladder cancer without evidence of metastatic disease. We broke those patients into two cohorts: one cohort that was observed after surgery and another cohort that received adjuvant chemotherapy.

And what did you find?

We found that after using a statistical approach called propensity score methods to control for all of the potential confounders that you might think of, we found that adjuvant chemotherapy was associated with a survival benefit and the hazard ratio for overall survival favouring chemotherapy was approximately 0.72, similar to what’s been reported in randomised studies of neoadjuvant chemotherapy.

But there were differences between the two groups, those who received adjuvant chemotherapy and those who had only cystectomy?

The way that propensity score methods work is that the statistical approach allows one to develop a score, a probability score, which defines the probability that an individual patient would be assigned chemotherapy or observation based on their baseline characteristics. So you might think that a patient who is 60 years old with no comorbidities and with node positive disease might be very likely to receive adjuvant chemotherapy. And in fact your data might show that, they might have a high propensity score. One of the beauties about variability in practice, however, and one of the things that allows us to conduct these types of analyses is that there are other patients with the same propensity score who don’t get treated for various reasons. So now you have a group of patients that are matched, based on all of these confounders, and you can ask the question does treatment make a difference?

This was basically an observational study, how confident are you of that hazard ratio of about 0.7?

We used a few different techniques to make sure that the findings were robust to the statistical approach that we used and we did a couple of other things to feel more confident as well. We know that the techniques that we used are very useful in controlling for observed confounders but don’t do a very good job controlling for unobserved confounders like randomisation does. So what we did was take what we thought was the most important unobserved confounder, that is poor performance status, and build that into a model to see what percentage of patients on the observation arm would have had to have poor performance status for our results to no longer be significant. We found using that approach that the results were fairly robust and that it would have to be 40% or more of the patients on the observation arm that had to have poor performance status for our results to no longer be statistically significant. So it’s an observational study, it’s retrospective. The problem is we will likely never have prospective randomised data to answer this question and yet patients come into our clinics today facing this decision.

So in your view, what are the clinical implications?

This does make me feel much more confident in counselling patients who haven’t received neoadjuvant chemotherapy before surgery who have locally advanced or node positive bladder cancer that chemotherapy is providing a benefit to them and that it should be considered.