Time toxicity in early-phase clinical trials.

Abstract

104 Background: Early-phase clinical trials (EP-CTs) are designed to determine optimal dosing, tolerability, and preliminary activity of novel cancer therapeutics. However, little is known about the time that patients spend interacting with the healthcare system (e.g., time toxicity) while participating in these studies. Methods: We retrospectively reviewed the electronic health records of consecutive patients enrolled in EP-CTs from 2017-2019 to obtain baseline characteristics and number of healthcare-associated days, defined as all inpatient and outpatient visits while on trial. Encounters were reviewed by a trained coding team and categorized based on intention (required by study protocol versus unplanned oncology visits for disease-related or study-related complications versus non-oncologic, such as primary care visits). We used univariate and multivariable analyses to identify predictors of increased time toxicity, defined as the proportion of healthcare-associated days among total days on trial. For ease of interpretation, we created a dichotomous variable, with high time toxicity defined as 20% or more healthcare-associated days during time on trial, and used regression models to evaluate relationships among time toxicity and clinical outcomes. Results: Among 408 EP-CT participants (median age=60.5 years, 56.5% female), patients spent a median of 21.8% days on trial interacting with the healthcare system. Time toxicity was highest during the first month on trial, with an average of 6.76 healthcare-associated days (5.38 days required by protocol, 1.19 unplanned days, and 0.18 days not related to the cancer diagnosis). Time toxicity decreased after the initial months on trial, with an average of 2.98 healthcare-associated days per month after six months on trial. Patients with gastrointestinal (B=0.07, p=0.002), head and neck (B=0.09, p=0.004), and breast (B=0.06, p=0.015) cancers, as well as those with worse performance status (B=0.04, p=0.017) and those on trials studying targeted therapies (B=0.04, p=0.01) experienced higher time toxicity. Patients with breast cancer (B=0.03, p=0.1) and those receiving targeted therapies (B=0.02, p=0.2) had no increased risk of unplanned time toxicity. High time toxicity was associated with decreased clinical benefit (OR=0.09, p<0.001) and objective response (OR=0.07, p<0.001) as well as worse progression-free (HR=2.10, p<0.001) and overall (HR=2.16, p<0.001) survival. Conclusions: In this large cohort of EP-CT participants, patients spent one-fifth of their time interacting with the healthcare system. We identified characteristics associated with higher time toxicity, including both overall and unplanned time toxicity. We found that high time toxicity correlated with worse clinical outcomes. We show that time toxicity is a readily measurable metric that could help inform patient-clinician discussions, guide future trial design, and identify at-risk patients.