Predicting Successful Phase Advancement and Regulatory Approval in Multiple Myeloma From Phase I Overall Response Rates

Natural Language Processing for Patient Selection in Phase I or II Oncology Clinical Trials

PURPOSE

Early discontinuation affects more than one third of patients enrolled in early-phase oncology clinical trials. Early discontinuation is deleterious both for the patient and for the study, by inflating its duration and associated costs. We aimed at predicting the successful screening and dose-limiting toxicity period completion (SSD) from automatic analysis of consultation reports.

MATERIALS AND METHODS

We retrieved the consultation reports of patients included in phase I and/or phase II oncology trials for any tumor type at Gustave Roussy, France. We designed a preprocessing pipeline that transformed free text into numerical vectors and gathered them into semantic clusters. These document-based semantic vectors were then fed into a machine learning model that we trained to output a binary prediction of SSD status.

RESULTS

Between September 2012 and July 2020, 56,924 consultation reports were used to build the dictionary and 1,858 phase I or II inclusion reports were used to train (72%), validate (14%), and test (14%) a random forest model. Preprocessing could efficiently cluster words with semantic proximity. On the unseen test cohort of 264 consultation reports, the performances of the model reached: F1 score 0.80, recall 0.81, and area under the curve 0.88. Using this model, we could have reduced the screen fail rate (including dose-limiting toxicity period) from 39.8% to 12.8% (relative risk, 0.322; 95% CI, 0.209 to 0.498; P < .0001) within the test cohort. Most important semantic clusters for predictions comprised words related to hematologic malignancies, anatomopathologic features, and laboratory and imaging interpretation.

CONCLUSION

Machine learning with semantic conservation is a promising tool to assist physicians in selecting patients prone to achieve SSD in early-phase oncology clinical trials.

Daratumumab monotherapy for relapsed/refractory multiple myeloma, focussed on clinical trial-unfit patients and subsequent therapy

Real-world outcomes of daratumumab monotherapy (DM) for relapsed/refractory multiple myeloma (RRMM) have remained unclear. We conducted a multicentre retrospective study of 107 patients receiving DM for RRMM. The cohort included 64 trial-unfit patients whose characteristics could not meet inclusion criteria in two previous clinical trials (GEN501 and SIRIUS). The overall response rate (ORR), and median first and second progression-free survival (PFS1 and PFS2) and overall survival were 42·1%, and 3·6, 8·1 and 11·9 months, respectively. Refractoriness to carfilzomib and/or lenalidomide, and neutropenia (<1.0 × 10⁹ /l) resulted in poorer ORRs. An Eastern Cooperative Oncology Group Performance Status of ≥3, neutropenia (<1.0 × 10⁹ /l), thrombocytopenia (<75 × 10⁹ /l), and renal failure (glomerular filtration rate of <20 ml/min/1·73 m² ) were associated with poor PFS1 and PFS2 in respective univariate analysis. The modified trial-unfit group, based on the above factors, showed significantly negative impacts on PFS1 and PFS2 (hazard ratio 2·823 and 3·677, all P < 0·001) in multivariate analysis despite having a 34% ORR. Fatal infections occurred more often in the modified trial-unfit group than in the others (16·1% vs. 4·3%; P = 0·099). Despite failure of DM, subsequent therapy with pomalidomide-based therapy or carfilzomib-dexamethasone provided a 66·6% ORR. Real-world DM showed favourable efficacies for RRMM and, potentially, additional benefits with subsequent therapies. However, characteristics corresponding with trial-unfitness might offset the efficacy of DM.

Local Investigators Significantly Overestimate Overall Response Rates Compared to Blinded Independent Central Reviews in Phase 2 Oncology Trials

The overall response rate (ORR) is a largely adopted outcome measure in early-phase oncology trials. ORR is highly relevant in cancer drug development at the time of deciding whether to move to confirmatory phase 3 trials; moreover, ORR is gaining increasing relevance in fast-track registration procedures. No systematic analysis has been conducted so far to investigate whether a discrepancy exists between ORR assessed by local investigators and those assessed by blinded reviewers in phase 2 oncology trials. In this study, we carried out a search in the clinicaltrials.gov and EudraCT databases, looking at the trials reporting the results of both investigator-assessed and independently-assessed ORR. A discrepancy index was obtained by calculating the ratio of each investigator-assessed ORR on the corresponding independently assessed ORR, so that a discrepancy index >1 indicates that the investigator was "more optimistic," whereas a discrepancy index <1 indicates the opposite. We also analyzed different subgroups (by tumor type, by drug type, by year). Twenty trials met the search criteria; in some cases, >1 comparison was conducted in the trial, so that the total number of comparisons analyzed was 33. The estimated mean discrepancy index was 1.175 (95% confidence interval, 1.083-1.264; n = 33). In conclusion, local investigators significantly overestimate ORR compared to paired blinded reviewers in phase 2 oncology trials. This may represent a risk in drug development, when deciding whether to move to confirmatory, more expensive phase 3 trials. Blinded independent central review should be used in ORR assessment if a more conservative estimate of treatment efficacy is required, as in the case of fast-track drug developments leading to accelerated approvals of cancer therapies.