Phenotypic Clustering And Longitudinal Complication Risk In Autologous Breast Reconstruction In 260 Patients: Evidence From NIH All Of Us

Le résumé fourni par la source

PURPOSE: Nationally representative longitudinal data on complication burden, flap selection, and individualized risk in autologous breast reconstruction remain limited. Most national databases, including NSQIP and SEER, lack extended follow-up, surgical granularity, or patient-reported outcomes, and often overrepresent urban, older, non-Hispanic white patients. The NIH All of Us Research Program, which includes over 849,000 participants and intentionally oversamples underrepresented groups, offers a unique opportunity to address these gaps using a diverse, population-scale cohort. This study represents one of the earliest applications of the All of Us dataset to the field of plastic and reconstructive surgery, providing new opportunities to examine complications and demographic trends for a critical reconstructive approach. METHODS: We identified 260 patients who underwent autologous breast reconstruction from 1995-2025 using Current Procedural Terminology codes within the All of Us Registered Tier Dataset (Version 8). Postoperative complications, such as flap failure, infection, wound dehiscence, hematoma, seroma, chronic pain, venous thromboembolism (VTE), hernia, and perioperative injury, were tracked at 30 days and 1 year. Logistic and multivariable regressions assessed complication predictors; Kaplan-Meier analysis evaluated complication emergence over time. Covariates included age, race, ethnicity, gender identity, BMI, and flap type. K-means clustering was used to identify phenotypic subgroups based on age and BMI. RESULTS: The mean age was 52.8 11.0 years. DIEP flap utilization increased steadily over time, particularly among younger patients. Free flaps (DIEP, fTRAM, SIEA, GAP) accounted for 41.9% of cases. One-year complication rates were highest among Black and Asian patients, though race was not an independent predictor in multivariable analysis. BMI >32.7 kg/m was associated with significantly increased 30-day complication risk (OR = 2.44, p = 0.007); age was predictive of 30-day, but not 1-year, complications. The most common 1-year complications were chronic pain, persistent pain, hernias, and muscle weakness. Free flap reconstructions showed the highest early complication burden, though flap failure was rare. Unsupervised machine learning through K-means clustering revealed three discrete phenotypic groups with differing complication risk: (1) older, low-BMI; (2) younger, low-BMI; and (3) mid-aged, high-BMI patients. Higher morbidity, including structural complications and chronic pain, was observed in the older and high-BMI clusters. Younger, low-BMI patients had the lowest overall complication burden. Despite differences in morbidity, flap success rates remained high across all phenotypes. CONCLUSIONS: Autologous breast reconstruction is broadly effective across diverse populations, with low flap failure rates and high success even in older or high-BMI patients. BMI, not age or race, emerged as the strongest predictor of early complications. Chronic pain represents a common and underrecognized long-term outcome, emphasizing the need for better surveillance and prevention strategies. K-means clustering based on continuous variables like age and BMI identifies high-risk phenotypes, offering a valuable framework for personalized risk stratification and preoperative planning. By integrating machine learning with a nationally representative cohort, this study demonstrates a scalable model for identifying high-risk phenotypes in real-world populations. These methods, and the diversity, longitudinal depth, and granularity of All of Us, offer distinct advantages over datasets like NSQIP, enabling more equitable and individualized surgical outcomes research.© 2026. Plastic Surgery Research Council | All rights reserved |*Source: https://ps-rc.org/meeting/Program/2026/OS17.cgi*

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.