Leslie Ballas: Hi, I'm Leslie Ballas. I am a radiation oncologist at Cedars-Sinai in Los Angeles, and I am honored to be joined today by David Fisher, who is a senior statistician and methodologist at the MRC Clinical Trials Unit at the University College London. He is going to talk to us today about the DADSPORT trial. David, thank you so much for joining me. I'm so excited to get to chat with you about this trial.

David Fisher: Thank you very much for inviting me to speak. Okay, so I'll just do a short presentation first of all to introduce the project to the listeners. So yeah, so this was a systematic review and meta-analysis that we entitled DADSPORT: Duration of Androgen Suppression with Postoperative Radiotherapy for Localized Prostate Cancer. And I was the lead statistician methodologist for this project and also the co-lead author with my colleague Sarah Burdett on our recent paper.

It was published earlier this year in European Urology, and the reference for that is on this slide. So the background for this is that in the postoperative setting, a number of trials have looked at whether using hormone therapy of either 6 or 24-month duration alongside radiotherapy could improve outcomes for high-risk localized prostate cancer patients. But various issues remained unclear and the most recent of these trials was the RADICALS-HD trial. And in anticipation of those results being available, we set out to perform a comprehensive review and meta-analysis of aggregate data in close collaboration with the trial investigators in order to maximize data quality and therefore the clinical impact. So we planned a systematic review and registered the protocol with all the methodology prior to the collection of data. And the important thing here is that our close collaborative approach meant that the protocol was designed in collaboration with the trial investigators and with our clinical experts. And we designed and collected a consistent set of summary results from each trial as well as any additional data we might need to resolve questions of sensitivity analyses or other additional questions.

The primary aim of this project was to assess the effect of hormone therapy in men who had received radiotherapy after surgery using the overall survival outcome. And we used standard fixed effects inverse variance meta-analysis methodology to do that. But we also wanted to look at evidence for differences using other outcomes which are listed here, both overall and by the duration of the hormone therapy as well as within patient subgroups and particular sensitivity analyses. We also used a separate analysis of a network analysis, which enabled us to make full use of the data from the RADICALS-HD trial design, which is a slightly more involved design than usual. So here are the primary results from the study. We had 5 trials in all, 6 comparisons because the RADICALS trial provided more than 1 comparison and over 4,000 participants in total, which represented 96% of all eligible. And using all this data, we showed that there was no clear evidence of any overall survival benefit on average from hormone therapy regardless of the duration. We found an average overall survival benefit of just 2% at 8 years. And we found evidence that even these small benefits may well be confined to a higher risk subgroup of patients. However, we did see significant improvements in metastasis-free survival and in prostate cancer-specific survival of 4% absolute benefit at 8 years. So in conclusion, by working collaboratively with the trialists, instead of just using reported results, the DADSPORT project has been able to provide reliable and detailed results just based on summary data.

And we've demonstrated that the average effect on overall survival is minimal, although benefits are seen for metastasis-free survival and prostate cancer-specific survival. And we also saw evidence of potentially larger benefits in higher risk subgroups. And therefore our take-home message really is that more work is needed going forward. So for example, with using more advanced imaging techniques to help identify the best patient subgroups where targeted therapy may be most effective.

Leslie Ballas: Thank you for that, David. I very much appreciate it. I have some questions. As a clinician, I think of a randomized controlled trial as the gold standard in providing the best evidence for treatment decision-making, especially well-done, large randomized controlled trials like RADICALS-HD. How do these aggregated data meta-analyses compare to a randomized controlled trial with 1,500 patients? Does it just provide the power for secondary analyses or would it actually benefit the primary outcome?

David Fisher: I think that very much depends on the results and the context. So if you have a large well-conducted randomized controlled trial, and if it definitively answers the primary question, there definitely is or definitely is not an effect, then maybe that is enough. And that is useful for clinical practice. But often it's not just the primary question that's of interest to clinicians. So we're looking at secondary analyses, patient subgroups as you mentioned. It also can be quite useful to widen the population that's being studied by looking at multiple trials because even with the best intentions, you may get trials done in different parts of the world, for instance, on different groups of patients can be useful to add context. And you may see a degree of heterogeneity in those results. In DADSPORT, the primary outcome of the overall survival, it's a small effect and that was known from the previous trials and therefore it wasn't definitively shown in previous trials. And therefore, yes, in this case, adding together the power of all those trials does give us more power to be able to hopefully make more definitive statements about the primary analysis, the primary outcome.

Leslie Ballas: So as you mentioned, when you have a meta-analysis, there's quite a bit of heterogeneity in patients. And so how do you account for that heterogeneity? I mean, obviously in a randomized controlled trial they take into account the differences in patients and try to smooth that out, but in this situation you don't have that. How does that affect outcomes?

David Fisher: I think that goes back to the point I made about looking at a wider patient set and the natural heterogeneity between the trial results. As long as that heterogeneity is not extreme, if it is extreme, then that points to specific differences between trials that needs to be accounted for. But even if there is a small amount of heterogeneity that is natural, it's part of the error, the natural error that you see between results. And if you look at multiple trials, you can see whether they, on average, are spread to one side or the other. And that can give you a better idea of what a sensible pooled estimate is likely to be. So for instance, particularly if that result is close to no effect as it is here, it's useful to see whether the spread includes more trials that are suggesting an effect on more trials that are not. So heterogeneity I would say is not necessarily a problem. Large heterogeneity is, but a small amount of heterogeneity can actually give a bit of clarity and nuance to the question.

Leslie Ballas: And so another just sort of basic question about the meta-analysis is you include a trial like RADICALS-HD with 1,500 patients, and then you include trials with much smaller numbers of patients with less follow-up like the GETUG study. And so it seems like obviously the RADICALS-HD will more heavily weigh and dominate the results. Why even include these smaller trials?

David Fisher: So from a theoretical perspective, when you are doing a systematic review, you want it to be comprehensive and you want to include all of the evidence that is eligible under the criteria that you set prior to carrying out the project. So that's the boring answer. Again, it's interesting to see the smaller trials with less follow-up or with other kind of nuances. It's really useful to see how they sit in relation to the other trials. As you say, because this particular trial was small, it doesn't have any undue influence on the results and we were careful to ensure that that was the case. But nevertheless, it's reassuring to see that it sits nicely next to the other trials and gives more context to the overall results. So it's important both from a theoretical aspect, but also I think really an interpretive aspect.

Leslie Ballas: Now to the meat of what the DADSPORT analysis presented, overall survival was not statistically significant. And oftentimes with prostate cancer, we don't see an overall survival benefit for quite a number of years. And so do you think that with time for each of these individual trials, the DADSPORT meta-analysis would also show overall survival benefit if you just gave it more time?

David Fisher: Yeah, so there are two parts to that to answer that question. So you can talk about whether you think the effect will be seen later or you can talk about whether the power will be there to detect a smaller effect if you have more data. I think I'm right in saying that there's no particular evidence to suggest that the effect is widening, that the curves are widening as you go on over time. And so I don't think there's any reason to suspect that the effect itself would become larger over time, but certainly I might be wrong on that.

But certainly if we wished to definitively have an estimate of the size of that effect, then more data would be necessary to say whether the effect was say a 2% effect at 8 years, for instance at a standard statistical significance level. We looked at what power we would have available from the studies included in DADSPORT, and we felt that the power we had available was sufficient to detect a clinically relevant effect size on OS. And therefore we feel that the detected effect that we've estimated here is smaller than what we consider clinically relevant. But yes, it's possible that with further follow-up, there may be additional nuances that can be seen in the data.

Leslie Ballas: The only reason that I sort of think about that is RTOG 9601, which was this trial looking at the addition of anti-androgen therapy in patients who had had prostatectomy. It was a randomized controlled trial done in the U.S. When you look at their overall survival curves, they do start to separate out after about eight or so years, if I'm remembering correctly. And in the end, overall survival did end up to be a statistically significant outcome. And so that's why I'm just curious about that effect.

David Fisher: So in terms of that particular trial, I think the patient population was slightly different in that trial to other trials included in DADSPORT. I think their baseline PSA levels were higher on average. So that speaks again to what we concluded from DADSPORT that we don't have enough evidence to say this, but it may possibly be that overall survival effects may be seen in those higher PSA subgroups. But we can't say that definitively because we only have data from that trial and from subgroups of other trials, including RADICALS. I think it's also important to place the overall survival effect, whether that be small or larger going forward with the quality of life issues associated with the hormone therapy. And I don't know whether clinicians ... How much of an effect they'd want to see and over what length of time before it wouldn't be worthwhile for patients.

Leslie Ballas: Absolutely. No, the quality of life component of hormone therapy is a major factor in deciding from the patient perspective whether or not they should go onto this treatment. And of course, whether or not a metastasis-free survival advantage is enough for a patient to take that detriment in quality of life. And so you bring up an extremely important point. The one last question that I have about that overall survival, that close overall survival potential benefit in the higher risk patients is, do you think that if you had more patients in the higher risk categories that that would be enough to sort of swing that P-value more either toward a statistical significance or away from one? Basically, my question is, is it a numbers issue?

David Fisher: I think ideally we would have more data on patients across the spectrum of PSA levels because I think what we'd like to do really is to have enough data to conclude that overall survival has a clinically relevant difference in this subset of patients, but not in that subset of patients. And I think without that, the question hasn't really been resolved.

Leslie Ballas: Well, David, thank you so much for taking the time to talk with us about DADSPORT. It's a really important analysis and provides us as clinicians really important information to provide patients, both because of what we were discussing about the quality of life and how patients would have to decide about whether a metastasis-free survival benefit was enough. But also just for us as clinicians looking at how do we evaluate those highest risk patients. So thank you very much. We really appreciate your time.

David Fisher: Thank you very much for having me.