gms | German Medical Science

Introduction: Response rate in r/r AML heavily depends on already known molecular and clinical data ranging from below 10% in patients with complex karyotypes and increasing to at least 80% in patients with CEBPA-biallelic mutations. Thus, adjustment of trial results in comparison to historical controls may improve external validity in single arm phase-II trials. To this aim we used the biometrical trial design suggested by Edelmann et al. [1], a modified Simon’s two-stage design, to previously account for heterogeneity at the interim analysis and to recalculate sample size at this time point. During the phase II TEAM study (ClinicalTrials.gov Identifier: NCT04173585) efficacy of Bortezomib in combination with Gemtuzumab Ozogamicin (B-GA) is assessed in comparison to historical controls [2], [3]. Primary endpoint is the CR/CRi rate (response rate).

The design is implemented in the R-package hctrial.

Methods: The design used was introduced by Edelmann et al. [1]. For this design historical controls are used to fit a logistic regression model with different prognostic parameters (assessed at baseline). For this trial, two models are fitted – one for refractory patients and one for relapsed patients. With estimated (or in this case values from expert knowledge) and an optimal Simon’s two-stage design is planned:

=0.3, =0.5, α = 0.1, β = 0.1, n₁=22, r₁=7, n₂=24, r=17, n=46)

After recruitment of the first 22 patients, the response rates under the null and alternative of these patients are estimated using the fitted logistic regression model and r₁, r and n₂ are recalculated controlling type I and type II errors. Recalculation of n₂ results in an adjusted sample size for the second stage. For the interim analysis, the recruitment usually has to be suspended until recalculation of r₁ or until the primary endpoint is assessed for the recruited patients. If the number of responses are strictly greater than , another patients can be recruited, else the trial is stopped for futility. At the end of the trial, r' is recalculated using the information of the recruited patients. If the number of responses are strictly greater than r'', H₀ can be rejected.

Results: Currently, 20 patients are recruited and 9 patients already responded. To check, if recruitment has to be suspended for the interim analysis, the baseline information of these patients are used, together with 2 dummy patients with the best possible prognostic factors which leads to the highest possible recalculated critical value . In this case, increases to 8 thus 9 responses are needed.

Discussion: In the 20 recruited patients exactly 9 responses are observed so far, thus recruitment does not have to be suspended. When 22 patients are recruited, the other relevant values n₂ and r can be recalculated. This should be done promptly to avoid recruiting more patients than necessary in the second stage.

Conclusion: Though this design is newly-created, the usability of the package and the results are promising. This design offers a chance to include patient populations with very heterogeneous prognoses in one single-arm trial given that historical controls are available.

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.