Search

Abstract Background Despite routine echocardiographic surveillance for childhood cancer survivors, the ability to predict cardiomyopathy risk in individual patients is limited. We explored the feasibility and optimal processes for machine learning-enhanced cardiomyopathy prediction in survivors using serial echocardiograms from five centers. Methods We designed a series of deep convolutional neural networks (DCNNs) for prediction of cardiomyopathy (shortening fraction ≤ 28% or ejection fraction ≤ 50% on two occasions) for at-risk survivors ≥ 1-year post initial cancer therapy. We built DCNNs with four subsets of echocardiographic data differing in timing relative to case (survivor who developed cardiomyopathy) index diagnosis and two input formats (montages) with differing image selections. We used holdout subsets in a 10-fold cross-validation framework and standard metrics to assess model performance (e.g., F1-score, area under the precision-recall curve [AUPRC]). Performance of the input formats was compared using a combined 5 × 2 cross-validation F-test. Results The dataset included 542 pairs of montages: 171 montage pairs from 45 cases at time of cardiomyopathy diagnosis or pre-diagnosis and 371 pairs from 70 at-risk survivors who didn’t develop cardiomyopathy during follow-up (non-case). The DCNN trained to distinguish between non-case and time of cardiomyopathy diagnosis or pre-diagnosis case montages achieved an AUROC of 0.89 ± 0.02, AUPRC 0.83 ± 0.03, and F1-score: 0.76 ± 0.04. When limited to smaller subsets of case data (e.g., ≥ 1 or 2 years pre-diagnosis), performance worsened. Model input format did not impact performance accuracy across models. Conclusions This methodology is a promising first step toward development of a DCNN capable of accurately differentiating pre-diagnosis versus non-case echocardiograms to predict survivors more likely to develop cardiomyopathy. Graphical Abstract

Abstract Young children receiving outpatient cancer care are vulnerable to undermanaged pain. App‐based solutions that provide pain treatment advice to parents in real‐time and in all environments may improve access to quality pain care. We used a parent co‐design approach involving iterative rounds of user testing and software modification to develop a usable Pain Caregiver Resource (PainCaRe) real‐time pediatric cancer pain management app. Parents of children (2–11 years) with cancer completed three standardized modules using a PainCaRe prototype. App usability and acceptability were evaluated using the validated System Usability Scale and a thematic analysis of app testing sessions and interviews. Iterative testing sessions were conducted until data saturation. Interview themes were synthesized into action items that guided revisions to PainCaRe and additional testing rounds were conducted as necessary. Twenty‐two parents participated in three testing cycles. Overall, parents described PainCaRe as an acceptable and potentially clinically useful pain management tool. Mean system usability scores were in the acceptable scale range during each testing cycle. Usability issues identified and resolved included those related to software malfunction, complicated app navigation logic, lack of clarity on pain assessment questions, and the need for pain management advice specifically tailored to child developmental stage. Using co‐design methods, the PainCaRe cancer pain management app was successfully refined for its acceptability and utility to parents. Next steps will include a PainCaRe pilot study before evaluating the impact of the app on younger children's pain outcomes in a randomized controlled trial.

Abstract Background Childhood cancer treatment while often curative, leads to elevated risks of morbidity and mortality. Survivors require lifelong periodic surveillance for late effects of treatment, yet adherence to guideline-recommended tests is suboptimal. We created ONLOOP to provide adult survivors of childhood cancer with detailed health information, including summaries of their childhood cancer treatment and recommended surveillance tests for early detection of cardiomyopathy, breast cancer, and/or colorectal cancer, with personalized reminders over time. Methods This is an individually randomized, registry-based pragmatic trial with an embedded process and economic evaluation to understand ONLOOP’s impact and whether it can be readily implemented at scale. All adult survivors of childhood cancer in Ontario overdue for guideline-recommended tests will be randomly assigned to one of two arms: (1) intervention or (2) delayed intervention. A letter of information and invitation will detail the ONLOOP program. Those who sign up will receive a personalized toolkit and a screening reminder 6 months later. With the participants’ consent, ONLOOP will also send their primary care clinician a letter detailing the recommended tests and a reminder 6 months later. The primary outcome will be the proportion of survivors who complete one or more of the guideline-recommended cardiac, breast, or colon surveillance tests during the 12 months after randomization. Data will be obtained from administrative databases. The intent-to-treat principle will be followed. Based on our analyses of administrative data, we anticipate allocating at least 862 individuals to each trial arm, providing 90% power to detect an absolute increase of 6% in targeted surveillance tests completed. We will interview childhood cancer survivors and family physicians in an embedded process evaluation to examine why and how ONLOOP achieved success or failed. A cost-effectiveness evaluation will be performed. Discussion The results of this study will determine if ONLOOP is effective at helping adult survivors of childhood cancer complete their recommended surveillance tests. This study will also inform ongoing provincial programs for this high-risk population. Trial registration ClinicalTrials.gov NCT05832138.