As part of our coverage of the American Society of Hematology (ASH) Annual Meeting & Exposition held December 9–12 in Atlanta, we are speaking with Matthew J. Maurer, MS, a statistician and assistant professor of medicine and biostatistics at the Mayo Clinic in Rochester, Minnesota. At the meeting, Matthew presented an analysis that he and his colleagues conducted on prospectively enrolled newly diagnosed diffuse large B-cell lymphoma (DLBCL) patients to try to understand the discrepancies seen in outcomes of single-arm nonrandomized trials, as well as those of larger randomized phase III trials in this patient group (abstract 4115).
Cancer Network: What is the issue that you and your colleagues saw in outcomes with these two different types of trials, and why focus on this specific tumor type?
Matthew J. Maurer: First, thanks for your interest in our work. I first started thinking about this issue at the ASH meeting in 2015 when John Leonard, MD, of Weill Cornell, was presenting data from a randomized trial of newly diagnosed DLBCL. He noted that the control arm did much better than expected. Sitting in the audience got me thinking about patient selection in the setting of newly diagnosed DLBCL, especially given the focus on molecular phenotyping or cell-of-origin–based designs, which often require central pathology review and a lengthier workup than the trials that historically have been done. Then a few months later, my colleague and the senior author on this study, Greg Nowakowski, MD, who is currently a primary investigator of a phase III randomized study in newly diagnosed DLBCL, expressed interest in looking into this topic as well. When we started looking into our data from our own Molecular Epidemiology Resource (MER) registry, it got very interesting.
Cancer Network: What was the study that you and your colleagues designed to analyze clinical outcomes among DLBCL patients on clinical trials, and how did the prior observational study you conducted help to inform this newer prospective one?
Matthew J. Maurer: We first looked at this concept in the MER of our Specialized Programs of Research Excellence grant that we have here at the Mayo Clinic with the University of Iowa to study lymphoma. The MER is a prospective, observational patient registry; we enroll patients with newly diagnosed lymphoma and then we follow them for outcomes. Through the MER we have been able to investigate a number of clinical topics in DLBCL, such as the utility of surveillance scans and clinical endpoints like event-free survival at 24 months. So, to examine the question of patient enrollment and the potential impact of delaying therapy for pathology workup to potentially go on a clinical trial, we took a look at the time between a patient’s diagnosis and their initiation of therapy, the diagnosis to treatment interval (DTI). We looked at this in about 1,000 patients with DLBCL enrolled on the MER and we presented these results at ASH last year. The results were quite striking, and so we then worked with our collaborators from the LYSA group in France to validate these findings in about 1,500 patients previously enrolled on their newly diagnosed DLBCL clinical trials.
Cancer Network: What were the important results of your study?
Matthew J. Maurer: What we identified in the MER was that patients who initiated therapy shortly after diagnosis had much more aggressive clinical features and worse outcomes compared with patients who had a longer period between their diagnosis and starting therapy. When we looked further, this association with outcome was independent of standard prognostic features that we use in DLBCL, such as the International Prognostic Index (IPI). This effect of time in between diagnosis and initiation of therapy was linear in fashion. So, in other words, the longer you can wait to start therapy, the less clinically aggressive the disease and the better the outcomes, even after controlling for standard prognostic features. This suggests that some of the good outcomes in control arms that have been observed on recent trials may be due to enrolling patients who can delay their initiation of therapy long enough to go through the workup of trial eligibility, which often includes central pathology review. Then we examined if this DTI effect remained in patients treated on clinical trials in the LYSA group. These trials were prior to molecular phenotyping or cell-of-origin–based designs. Indeed, we did see this effect even in patients enrolled on clinical trials. The validation in the LYSA cohort means that this effect has now been confirmed in both real-world data, as well as trial-based data from two independent international cohorts. This also suggests that the current inclusion/exclusion criteria in our newly diagnosed DLBCL trials may not be sufficient to provide a non-biased clinical trial selection.
Cancer Network: What do these results suggest about how to optimize clinical trial design in DLBCL, and do the conclusions also potentially extend to trial design for other types of patients?
Matthew J. Maurer: We know the majority of patients with DLBCL will be cured with R-CHOP, and clinical trials in newly diagnosed DLBCL are often targeted to patients with high-risk disease or patients with a higher than average risk of failing R-CHOP. There’s a concern that by prioritizing the assessment of the cell-of-origin and other clinical and pathologic criteria prior to beginning therapy that these trials may have under-enrolled the high-risk patients that are in need of the novel therapies being tested. From a statistical standpoint, trials failing to enroll these high-risk patients are in danger of being underpowered due to a better-than-expected event rate, which is what we have seen in some recent phase III clinical trials in this setting. In evaluating completed or ongoing clinical trials in newly diagnosed DLBCL, the time from diagnosis to treatment or the time from diagnosis to randomization should be reported for these trials. This will help us interpret the impact of the study design on the patients enrolled on the trial, as well as evaluate differences in outcomes not controlled by standard clinical prognostic features such as the IPI.
For randomized trials, this could help explain outcomes on the control arm and help us compare control arms from trial to trial. I think that we also need to be looking at the time from diagnosis to treatment in phase II trials and especially single-arm trials in which we are evaluating results and trying to interpret what is promising and worth moving forward. Phase III trials are long, difficult to conduct, very expensive, and also involve the lives of many patients, so it’s important for us to make proper inference on studies of regimens considered for moving into the phase III setting. I think going forward, future trials, both single-arm and randomized in newly diagnosed DLBCL, will need to consider modifying designs or allow rapid enrollment of patients to include those with clinically aggressive disease who are in need of more urgent therapy.
I think we can also potentially extend these results to the broad precision medicine setting. If we make it difficult and time-consuming to enroll on a trial, I think that there can be serious implications for patient selection on the trial. The danger is that the patients who enroll on the trial are not an accurate representation of the target patient population for the therapy. Many of these precision medicine trials are not randomized, and so I think there is concern of being too optimistic about single-arm trial results, simply due to the patient selection that may be due to unknown factors, such as what we found for the DTI in newly diagnosed DLBCL. These unknown factors can potentially influence outcome and that is something we are not currently controlling for.
Cancer Network: Thank you so much for joining us today, Matthew.
Matthew J. Maurer: Thank you, it’s my pleasure.