Year: 2014
Type: letter
Abstract: We would like to thank Voskoboynik and Arkenau 1 as well as Ballman 2 for their thoughtful commentaries on our work. 3Both raise important questions about how our nomogram would best be used in clinical practice.One concern is whether excluding high-risk patients for serious drug-related toxicity (SDRT) may make the phase I population more fit than the patients who participate in phase II/III studies.If this were the case, the toxicity observed in phase II/III studies might be greater than that predicted by phase I studies.We have an alternative view.Patients in phase I trials typically have advanced and heavily pretreated disease and are consequently at increased risk for adverse events that are related to both treatment and the underlying disease.As evidence of this, patients in our cohort had received a median of five previous agents, 47% had received previous radiotherapy, and 32% had four or more metastatic sites.Ballman acknowledges that many disease-related adverse events in these patients are likely to be misattributed as drug related, further exacerbating the detrimental effects of enrolling these patients.We believe that patients who are enrolled onto phase II/III trials are generally more fit than patients participating in phase I trials and would, on average, have lower scores on our nomogram.We acknowledge that this hypothesis needs to be tested in large and diverse phase II/III cohorts.Voskoboynik and Arkenau 1 comment that the use of our nomogram could slow recruitment and limit patient access to highly effective targeted therapies such as BRAF or anaplastic lymphoma kinase inhibitors.We support the concept of precision medicine in early drug development and agree that physician assessment must continue to play a critical role in the determination of who should participate in and is likely to benefit from phase I trials.This is especially true for studies involving a drug class that has previously shown high rates of clinical benefit in patients who harbor a predictive biomarker.However, we would emphasize that enrolling unfit patients onto the dose-escalation portion of a first-inhuman phase I study may cause other delays in the drug's development, and these competing risks must be weighed carefully.We agree with Ballman's 2 assertion that statistically robust phase I trial designs, such as the continual reassessment method, can yield gains in trial efficiency and dose estimation. 4,5We believe that patient selection also plays an important role in ensuring reproducible and efficient dose estimation.A phase I study that uses even the most robust design can yield unreliable results if inappropriate patients are enrolled.The studies used in our analysis were designed under the auspices of the National Cancer Institute and conducted by U01 phase I grantees at centers of early drug developmental excellence under the direct supervision of the Cancer Therapy Evaluation Program.Despite this, our analysis indicates that unfit patients were enrolled onto these studies with incompletely understood, but potentially detrimental, effects on the quality of the toxicity data generated.Therefore, we emphasize that both study design and patient selection are critical and complementary aspects of phase I studies, and both must be optimized.We cannot rely on sophisticated phase I study designs to compensate for suboptimal patient selection.
Source: Journal of Clinical Oncology
Language: Spanish
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: bronze
Funder: National Cancer Institute
Award: Cancer Center Support Grant
OpenAlex ID: W2026341926