Using Bayesian hierarchical modeling for performance evaluation of clinical trial accrual for a cancer center.

Introduction: Slow patient accrual in cancer clinical trials is always a concern. In 2021, the University of Kansas Comprehensive Cancer Center (KUCC), an NCI-designated comprehensive cancer center, implemented the Curated Cancer Clinical Outcomes Database (C3OD) to perform trial feasibility analyses using real-time electronic medical record data. In this study, we proposed a Bayesian hierarchical model to evaluate annual cancer clinical trial accrual performance. Methods: The Bayesian hierarchical model uses Poisson models to describe the accrual performance of individual cancer clinical trials and a hierarchical component to describe the variation in performance across studies. Additionally, this model evaluates the impacts of the C3OD and the COVID-19 pandemic using posterior probabilities across evaluation years. The performance metric is the ratio of the observed accrual rate to the target accrual rate. Results: Posterior medians of the annual accrual performance at the KUCC from 2018 to 2023 are 0.233, 0.246, 0.197, 0.150, 0.254, and 0.340. The COVID-19 pandemic partly explains the drop in performance in 2020 and 2021. The posterior probability that annual accrual performance is better with C3OD in 2023 than pre-pandemic (2019) is 0.935. Conclusions: This study comprehensively evaluates the annual performance of clinical trial accrual at the KUCC, revealing a negative impact of COVID-19 and an ongoing positive impact of C3OD implementation. Two sensitivity analyses further validate the robustness of our model. Evaluating annual accrual performance across clinical trials is essential for a cancer center. The performance evaluation tools described in this paper are highly recommended for monitoring clinical trial accrual.

First-Line Immunotherapy with Check-Point Inhibitors: Prospective Assessment of Cognitive Function.

Approximately 40% of patients with cancer are eligible for check-point inhibitor (CPI) therapy. Little research has examined the potential cognitive impact of CPIs. First-line CPI therapy offers a unique research opportunity without chemotherapy-related confounders. The purpose of this prospective, observational pilot was to (1) demonstrate the feasibility of prospective recruitment, retention, and neurocognitive assessment for older adults receiving first-line CPI(s) and (2) provide preliminary evidence of changes in cognitive function associated with CPI(s). Patients receiving first-line CPI(s) (CPI Group) were assessed at baseline (n = 20) and 6 months (n = 13) for self-report of cognitive function and neurocognitive test performance. Results were compared to age-matched controls without cognitive impairment assessed annually by the Alzheimer's Disease Research Center (ADRC). Plasma biomarkers were measured at baseline and 6 months for the CPI Group. Estimated differences for CPI Group scores prior to initiating CPIs (baseline) trended to lower performance on the Montreal Cognitive Assessment-Blind (MOCA-Blind) test compared to the ADRC controls (p = 0.066). Controlling for age, the CPI Group's 6-months MOCA-Blind performance was lower than the ADRC control group's 12-months performance (p = 0.011). No significant differences in biomarkers were detected between baseline and 6 months, although significant correlations were noted for biomarker change and cognitive performance at 6 months. IFNγ, IL-1ß, IL-2, FGF2, and VEGF were inversely associated with Craft Story Recall performance (p < 0.05), e.g., higher levels correlated with poorer memory performance. Higher IGF-1 and VEGF correlated with better letter-number sequencing and digit-span backwards performance, respectively. Unexpected inverse correlation was noted between IL-1α and Oral Trail-Making Test B completion time. CPI(s) may have a negative impact on some neurocognitive domains and warrant further investigation. A multi-site study design may be crucial to fully powering prospective investigation of the cognitive impact of CPIs. Establishment of a multi-site observational registry from collaborating cancer centers and ADRCs is recommended.

Risk of SARS-CoV-2 Breakthrough Infection in Vaccinated Cancer Patients: A Retrospective Cohort Study.

Although messenger RNA (mRNA) vaccines have established efficacy for prevention of severe SARS-CoV2 infection in the general population, their effectiveness in patients with malignancy, especially those on anti-neoplastic therapies, remains an area of open research. In order to better understand the risk of developing breakthrough SARS-CoV-2 infection and the outcomes associated with breakthrough infection for cancer patients, individual patient data from a curated outcomes database at the University of Kansas were retrospectively reviewed to determine the rate of breakthrough infection during an 8-month period encompassing the height of the delta variant surge. Although the rate of breakthrough infection in cancer patients after two doses of an mRNA vaccine remained low at 1.1%, hospitalization and death rates were 27 and 5%, respectively. Patients with hematologic malignancies, especially multiple myeloma, and those on anti-neoplastic therapy at the time of vaccination were found to be at higher risk for developing breakthrough infection.

Estimating power for clinical trials with Patient Reported Outcomes - using Item Response Theory.

OBJECTIVES: Patient reported outcomes (PRO) are widely used in quality of life (QOL) studies, health outcomes research, and clinical trials. The importance of PRO has been advocated by health authorities. Patient Reported Outcomes Measurement Information System (PROMIS) is a collection of standardized measures of PROs using Item Response Theory (IRT). However, in clinical trials with PROs as endpoints, observed scores are routinely used for power estimation rather than IRT scores. This paper aims to fill this gap and estimate power in a two-arm clinical trials with PROMIS measures as endpoints with IRT model. STUDY DESIGN AND SETTING: We conducted a series of simulations to study the IRT power with validated PROMIS measures controlling factors including sample size, effect size, number of items, and missing data proportion. RESULTS: Our results showed that sample size, effect size, and number of items are important indicators of IRT based power estimation for PROMIS measures. When effect size is small and sample size is limited, IRT model provides higher power than the closed form formula. CONCLUSION: IRT based simulation should be used for power estimation in two-armed clinical, especially when there is small effect size or small sample size.

The need to study rural cancer outcome disparities at the local level: a retrospective cohort study in Kansas and Missouri.

BACKGROUND: Rural residence is commonly thought to be a risk factor for poor cancer outcomes. However, a number of studies have reported seemingly conflicting information regarding cancer outcome disparities with respect to rural residence, with some suggesting that the disparity is not present and others providing inconsistent evidence that either urban or rural residence is associated with poorer outcomes. We suggest a simple explanation for these seeming contradictions: namely that rural cancer outcome disparities are related to factors that occur differentially at a local level, such as environmental exposures, lack of access to care or screening, and socioeconomic factors, which differ by type of cancer. METHODS: We conducted a retrospective cohort study examining ten cancers treated at the University of Kansas Medical Center from 2011 to 2018, with individuals from either rural or urban residences. We defined urban residences as those in a county with a U.S. Department of Agriculture Urban Influence Code (UIC) of 1 or 2, with all other residences defines a rural. Inverse probability of treatment weighting was used to create a pseudo-sample balanced for covariates deemed likely to affect the outcomes modeled with cumulative link and weighted Cox-proportional hazards models. RESULTS: We found that rural residence is not a simple risk factor but rather appears to play a complex role in cancer outcome disparities. Specifically, rural residence is associated with higher stage at diagnosis and increased survival hazards for colon cancer but decreased risk for lung cancer compared to urban residence. CONCLUSION: Many cancers are affected by unique social and environmental factors that may vary between rural and urban residents, such as access to care, diet, and lifestyle. Our results show that rurality can increase or decrease risk, depending on cancer site, which suggests the need to consider the factors connected to rurality that influence this complex pattern. Thus, we argue that such disparities must be studied at the local level to identify and design appropriate interventions to improve cancer outcomes.

Optimizing Retrieval of Biospecimens Using the Curated Cancer Clinical Outcomes Database (C3OD).

To fully support their role in translational and personalized medicine, biorepositories and biobanks must continue to advance the annotation of their biospecimens with robust clinical and laboratory data. Translational research and personalized medicine require well-documented and up-to-date information, but the infrastructure used to support biorepositories and biobanks can easily be out of sync with the host institution. To assist researchers and provide them with accurate pathological, epidemiological, and bio-molecular data, the Biospecimen Repository Core Facility (BRCF) at the University of Kansas Medical Center (KUMC) merges data from medical records, the tumor registry, and pathology reports using the Curated Cancer Clinical Outcomes Database (C3OD). In this report, we describe the utilization of C3OD to optimally retrieve and dispense biospecimen samples using these 3 data sources and demonstrate how C3OD greatly increases the efficiency of obtaining biospecimen samples for the researchers.

Relevant Word Order Vectorization for Improved Natural Language Processing in Electronic Health Records.

Electronic health records (EHR) represent a rich resource for conducting observational studies, supporting clinical trials, and more. However, much of the data contains unstructured text, presenting an obstacle to automated extraction. Natural language processing (NLP) can structure and learn from text, but NLP algorithms were not designed for the unique characteristics of EHR. Here, we propose Relevant Word Order Vectorization (RWOV) to aid with structuring. RWOV is based on finding the positional relationship between the most relevant words to predicting the class of a text. This facilitates machine learning algorithms to use the interaction of not just keywords but positional dependencies (e.g. a relevant word occurs 5 relevant words before some term of interest). As a proof-of-concept, we attempted to classify the hormone receptor status of breast cancer patients treated at the University of Kansas Medical Center, comparing RWOV to other methods using the F1 score and AUC. RWOV performed as well as, or better than other methods in all but one case. For F1 score, RWOV had a clear edge on most tasks. AUC tended to be closer, but for HER2, RWOV was significantly better for most comparisons. These results suggest RWOV should be further developed for EHR-related NLP.