Bridging the cardio-oncology care gap in India: Real-world guideline adherence and AI-driven workflow optimization in 2300 breast cancer patients at a tertiary cardiac center.

e12719 Background: Cardiotoxic cancer therapies necessitate guideline-recommended cardiac surveillance in breast cancer patients, yet real-world implementation and efficiency remain variable. Data from low- and middle-income settings are limited. We evaluated real-world adherence to ASCO cardio-oncology guidelines and the impact of AI-assisted echocardiography on monitoring efficiency and clinical response in a large breast cancer cohort. Methods: We conducted a retrospective real-world analysis of 2,300 breast cancer patients receiving potentially cardiotoxic therapy at a tertiary center. Guideline adherence was assessed using ASCO recommendations for baseline and surveillance echocardiography. AI-assisted echocardiographic LVEF measurements were compared with standard reporting. Primary outcomes included timeliness of cardiac imaging and reporting, detection of cancer therapy–related cardiac dysfunction, and subsequent clinical management. Results: A total of 2,300 breast cancer patients receiving potentially cardiotoxic systemic therapy were included (mean age 52 ± 11 years). Anthracycline-based regimens were administered in 1,426 patients (62%), and anti-HER2 therapy in 644 (28%). Cardiovascular comorbidities were common, including hypertension (38%), type 2 diabetes mellitus (31%), and obesity (24%). Baseline adherence to guideline-recommended echocardiography was 94% (95% CI, 93–95%). Prior to AI implementation, adherence to serial surveillance was 72% (95% CI, 70–74%), which significantly increased to 89% (95% CI, 88–91%) following AI-enabled workflow integration. Mean echocardiographic analysis time per study decreased by 74% (2.1 vs 8.2 minutes; p < 0.001). Cancer therapy–related cardiac dysfunction (CTRCD) occurred in 11.8% of patients (n = 271; 95% CI, 10.5–13.2%). All patients with CTRCD received guideline-directed cardioprotective therapy. During follow-up, mean LVEF improved by 7.25% (95% CI, 6.4–8.1%), and 95% of patients remained free from heart failure hospitalization. On multivariable analysis, diabetes mellitus (OR 2.8; 95% CI, 1.9–4.1) and age > 60 years (OR 2.1; 95% CI, 1.4–3.2) independently predicted CTRCD. AI-derived LVEF demonstrated strong agreement with expert assessment (accuracy 85% [95% CI, 83–87%]; precision 81%; recall 73%; F1 score 67%). Conclusions: Using ASCO recommendations for baseline and surveillance echocardiography, this real-world study identified a significant cardio-oncology care gap that was substantially reduced through AI-enabled workflow integration. Improved guideline adherence enabled timely detection and management of cardiotoxicity, resulting in meaningful LVEF recovery and low heart failure hospitalization rates in a resource-constrained setting.