Denise Esserman

Objective: To develop and validate a supervised text-embedded transformer matching model to identify fall injuries in Medicare data, and evaluate the model’s performance -- alongside a validated rule-based algorithm-- against “ground truth” from an external reference standard (self-reported fall injuries leading to medical attention). Materials and Methods: Text embeddings of ICD-10-CM and CPT codes in Medicare claims/encounters from participants in the Strategies to Reduce Injuries and Develop Confidence in Elders (STRIDE) trial served as model inputs. Trained on annotated claims/encounters occurring within +/- one month of self-reported fall injuries leading to medical attention, the transformer model generated a continuous 0-1 probability that each claim/encounter was for a fall injury. The model was then applied to all claims/encounters in STRIDE and compared alongside the rule-based algorithm to the external reference standard. Results: The model achieved an area under the curve (AUC) of > 0.96 against annotated claims/encounters in 9 out of 10 holdout folds and 0.85 in the remaining fold. In the full STRIDE dataset, the model achieved a peak AUC of 0.86 (95% CI, 0.84-0.87) against the external reference standard, with results comparable to the rule-based algorithm. Discussion: Relative to rule-based approaches, which typically generate binary outcomes, the continuous event probability generated by the transformer model could support clinical endpoint adjudication, with high-probability predictions treated as events, moderate-probability predictions being adjudicated, and low-probability predictions treated as non-events. Conclusion: A text-embedded transformer model identified fall injuries with comparable accuracy to a rule-based algorithm, demonstrating “proof of concept” for use in endpoint adjudication.

Background: Several ovarian cancer studies have suggested that a body mass index (BMI) of 30 or higher is associated with lower compliance with National Comprehensive Cancer Network–recommended chemotherapy but primarily involved treatment before 2012, when dose capping was recommended for patients with higher body surface areas. Updated analyses in the contemporary treatment era are warranted. Methods: In a retrospective cohort of patients with newly diagnosed ovarian cancer receiving curative-intent carboplatin plus paclitaxel in the Yale-Smilow Cancer Network (2012-2022), we evaluated BMI at diagnosis in relation to relative dose intensity (RDI)—the ratio of completed chemotherapy dose intensity to the National Comprehensive Cancer Network–recommended dose intensity—which reflects dose modification both before and during treatment. We also assessed starting RDI (which reflects modifications before treatment) and received RDI (which reflects modifications during treatment). Data on hospitalizations and hematological chemotoxicities were collected. We examined the association between BMI (<25, 25-30, ≥30) and chemotherapy completion, hospitalizations, and toxicities using multivariable linear and logistic regressions. Results: Among 327 patients, the average RDI was 79.7%, and 44.3% had an RDI below 85%. Mean (SD) starting and received RDI were 97.9% (9.1%) and 81.8% (25.7%), respectively. Higher BMI was associated with higher RDI (Paggregate = .03) and received RDI (Paggregate = .04). Body mass index was not associated with starting RDI, dose reductions, delays, hospitalizations, or hematological toxicities. Conclusions: Among patients with ovarian cancer treated since 2012, the overall RDI was low. Relative dose intensity was higher among patients with a BMI of 25 or higher compared with a BMI below 25. Most dose modifications occurred during treatment and not before initiation. Studies with body composition data and interventions that maximize chemotherapy completion during treatment are warranted.

While the public health burden of SARS-CoV-2 infection has lessened due to natural and vaccine-acquired immunity, emergence of less virulent variants, and antiviral medications, COVID-19 continues to take a significant toll. There are thousands of new hospitalizations and hundreds of deaths per week in the United States, many of whom develop long COVID. Early identification of individuals at high risk of severe COVID-19 is key for monitoring and supporting respiratory status and improving outcomes. Therefore, precision tools for early detection of patients at high risk of severe disease can reduce morbidity and mortality. Here, we report an untargeted, longitudinal plasma metabolomics study of COVID-19 patients. One-carbon metabolism, a pathway previously shown as critical for viral propagation and disease progression, and a potential target for COVID-19 treatment, scored strongly as differentially abundant in patients with severe COVID-19. Targeted metabolite profiling revealed that one arm of the one-carbon metabolism pathway, the methionine cycle, is a major driver of the metabolic profile associated with disease severity. Further, genomic data from the profiled patients revealed a genetic contributor to methionine metabolism and identified the C677T allele of the MTHFR gene as a preexisting contributor to disease trajectory—patients that show aberrant one-carbon metabolite levels and that are homozygous for the MTHFR C677T, have higher incidence of severe COVID. Our results raise the possibility that MTHFR variant status may inform precision COVID-19 treatment strategies.

Composite endpoints are widely used in cardiovascular clinical trials to improve statistical efficiency while preserving clinical relevance. The Win Ratio (WR) measure and more general frameworks of Win Statistics have emerged as increasingly popular alternatives to traditional time-to-first-event analyses. Although analytic sample size formulas for WR have been developed, they rely on design parameters that are often not straightforward to specify. Consequently, sample size determination in clinical trials with WR as the primary analysis is most often based on simulations, which can be computationally intensive. Moreover, these simulations commonly assume independence among component endpoints, an assumption that may not hold in practice and can lead to misleading power estimates. To address this challenge, we derive refined formulas to calculate the proportions of wins, losses, and ties for multiple prioritized time-to-event endpoints. These formulas rely on familiar design inputs and become directly applicable when integrated with existing sample size methods. We conduct a comprehensive assessment of how correlation among endpoints affects sample size requirements across varying design features. We further demonstrate the role of correlations through two case studies based on the landmark SPRINT and STICH clinical trials to generate further insights.

Key Points: Of the 643 patients undergoing long-term hemodialysis enrolled in a clinical trial for chronic pain, 28% experienced at least one fall, an incidence of 0.68 per participant year. Accidents were the most frequent cause of falls, and it was rare for them to be related to the hemodialysis procedure or occur at the hemodialysis unit. In multivariable analyses, demographic characteristics, severity of pain symptoms, or medication use such as opioids were associated with the fall risk. Background: Falls are believed to be common in patients undergoing maintenance hemodialysis, but little is known about their frequency or outcomes. In this prospective study, we sought to increase our knowledge regarding the incidence, timing, circumstances, and outcomes of falls in this population. Methods: Between January 2021 and April 2023, adults undergoing maintenance hemodialysis from 103 US dialysis facilities were enrolled in the Heart Outcomes Prevention Evaluation Consortium trial, which randomized participants with moderate or severe chronic pain to a pain coping skills, cognitive behavioral therapy intervention, or usual care. Occurrence of falls was a prespecified trial outcome. The research team inquired about falls at each 4-week follow-up visit during the 36-week study. Multivariable regression was used to explore associations of demographic and clinical characteristics, including patient-reported symptoms, with fall risk. Results: Of 643 trial participants, 178 (28%) experienced 293 falls over a cumulative follow-up period of 429 participant-years for an overall rate of 0.68 falls per participant-year (95% confidence interval, 0.61 to 0.76). Accidents were the most frequent cause of falls (38%). It was rare for falls to be related to the hemodialysis treatment or to occur in the hemodialysis unit. Of the 293 falls, 36 (12%) were evaluated in the emergency department without subsequent hospitalization, 41 (14%) resulted in a hospital admission, and 19 (7%) led to a fracture. In multivariable analyses, demographic characteristics, severity of pain symptoms, or medication use such as opioids at enrollment were not associated with the fall risk. Conclusions: Falls were common in this cohort of maintenance hemodialysis patients with chronic pain, occurring in 28% of individuals during a planned follow-up of 36 weeks. Falls rarely occurred in the dialysis unit, with the vast majority occurring at participants' homes and due to accidental causes. There was no significant association between patient-reported symptoms or medication use and the risk of subsequent falls. Clinical Trial registry name and registration number: ClinicalTrials.gov, NCT04571619.

Enrolling patients to the standard of care (SOC) arm in randomized clinical trials, especially for rare diseases, can be very challenging due to the lack of resources, restricted patient population availability, and ethical considerations. As the therapeutic effect for the SOC is often well documented in historical trials, we propose a Bayesian platform trial design with hybrid control based on the multisource exchangeability modelling (MEM) framework to harness historical control data. The MEM approach provides a computationally efficient method to formally evaluate the exchangeability of study outcomes between different data sources and allows us to make better informed data borrowing decisions based on the exchangeability between historical and concurrent data. We conduct extensive simulation studies to evaluate the proposed hybrid design. We demonstrate the proposed design leads to significant sample size reduction for the internal control arm and borrows more information compared to competing Bayesian approaches when historical and internal data are compatible.