Hybrid-control arm construction using historical trial data for an early-phase, randomized controlled trial in metastatic colorectal cancer

Abstract

Background Treatment for metastatic colorectal cancer patients beyond the second line remains challenging, highlighting the need for early phase trials of combination therapies for patients who had disease progression during or following two prior lines of therapy. Leveraging hybrid control design in these trials may preserve the benefits of randomization while strengthening evidence by integrating historical trial data. Few examples have been established to assess the applicability of such design in supporting early phase metastatic colorectal cancer trials. Methods MORPHEUS-CRC is an umbrella, multicenter, open-label, phase Ib/II, randomized, controlled trial (NCT03555149), with active experimental arms ongoing. Patients enrolled were assigned to a control arm (regorafenib, 15 patients randomized and 13 analysed) or multiple experimental arms for immunotherapy-based treatment combinations. One experimental arm (atezolizumab + isatuximab, 15 patients randomized and analysed) was completed and included in the hybrid-control study, where the hybrid-control arm was constructed by integrating data from the IMblaze370 phase 3 trial (NCT02788279). To estimate treatment efficacy, Cox and logistic regression models were used in a frequentist framework with standardized mortality ratio weighting or in a Bayesian framework with commensurate priors. The primary endpoint is objective response rate, while disease control rate, progression-free survival, and overall survival were the outcomes assessed in the hybrid-control study. Results The experimental arm showed no efficacy signal, yet a well-tolerated safety profile in the MORPHEUS-CRC trial. Treatment effects estimated in hybrid control design were comparable to those in the MORPHEUS-CRC trial using either frequentist or Bayesian models. Conclusions Hybrid control provides comparable treatment-effect estimates with generally improved precision, and thus can be of value to inform early-phase clinical development in metastatic colorectal cancer.

Examining external control arms in oncology: A scoping review of applications to date

“…However, nearly 48% (11/23) of studies did not provide clear information on covariate adjustments—this concern has been underscored in the existing literature as it relates to the failure in accounting for baselines imbalances. 4 , 15 , 17 , 63 , 64 As for the data collection timeline, most studies aimed for a close temporal alignment between the ECAs and the treatment arm. However, temporal gaps in some instances indicate potential challenges, possibly related to data accessibility.…”

Using Bayesian Dynamic Borrowing to Maximize the Use of Existing Data: A Case-Study

“…This case-study has focused on the bridging of global datasets to a new region; alternative Bayesian approaches with informative priors could also be considered for this purpose [ 21 – 23 ]. Other applications of BDB could include extrapolation to pediatric populations from adult data [ 2 – 4 , 8 ] or supplementing a study with external control data [ 5 , 6 , 24 ].

Figure 6 Replacement of planned subjects with historical-data.

…”

“…One way to incorporate existing data are through a method known as Bayesian Dynamic Borrowing (BDB), an approach with increasing use in clinical development. The regulatory acceptability of BDB continues to evolve and experience with BDB varies between, and within, regulatory agencies [1][2][3][4][5][6][7][8].…”

Use of External Control Cohorts in Pediatric Brain Tumor Clinical Trials

“…If random assignment is feasible and a SOC treatment is available, most patients can be randomly assigned to the experimental treatment arm and a minority to the SOC therapy concurrent control arm with the intent to augment the control arm with external data (eg, hybrid design). [19][20][21] Once accrued, the patients in the concurrent control arm can be compared with the external control patients to verify consistency. One approach is to assess the similarity in the outcome distribution between the concurrent and external control patients.…”