AI Prediction of Endocrine Treatment Resistance: Current Landscape and Clinical Implications

Abstract

Endocrine treatment resistance represents a significant barrier in the management of hormone receptor-positive malignancies, notably breast cancer. Recent advances in artificial intelligence (AI) have enabled the development of predictive models aimed at early identification of resistance, promising improved patient stratification and personalized treatment planning. This review synthesizes current scientific evidence on the epidemiology, pathophysiology, risk factors, diagnostic strategies, and management approaches to endocrine resistance, emphasizing the role of AI in transforming clinical practice. It also explores the latest guideline recommendations and the evolving landscape of AI-driven predictive analytics, with a focus on clinical applicability, future prospects, and the challenges that remain.

Introduction

Endocrine therapies are the mainstay of treatment for hormone receptor-positive cancers, with breast cancer being the prototypical example. While these therapies have significantly improved survival and quality of life, a substantial proportion of patients eventually develop resistance, leading to disease progression. Traditional approaches to predicting resistance rely on clinicopathologic and molecular features, but the integration of AI and machine learning algorithms has begun to revolutionize the predictive paradigm. This review provides a comprehensive overview of the current state of AI prediction models for endocrine treatment resistance, their underlying mechanisms, and clinical implications for healthcare professionals.

Epidemiology / Disease Burden

Hormone receptor-positive breast cancer accounts for approximately 70% of all breast cancer cases globally. Despite effective initial responses to endocrine therapy, up to 40% of patients develop resistance, contributing to relapse and mortality. The cumulative burden is considerable, with endocrine resistance responsible for a significant proportion of advanced disease cases and treatment failures. Epidemiological studies underscore the urgent need for predictive tools that can identify at-risk patients early in their treatment course, enabling timely intervention and improved outcomes.

Pathophysiology

Endocrine resistance arises from complex, multifactorial biological processes. Mechanisms include alterations in estrogen receptor (ER) signaling, mutations in ESR1 (estrogen receptor alpha gene), activation of alternative growth pathways (such as PI3K/AKT/mTOR), dysregulation of cell cycle checkpoints, and epigenetic modifications. Crosstalk between ER and growth factor pathways, such as HER2 and FGFR, further complicates the landscape. AI-driven analysis of multi-omics data—including genomics, transcriptomics, and proteomics—enables finer resolution of these mechanisms, allowing for the detection of resistance-associated molecular signatures before clinical progression becomes evident.

Risk Factors

Risk factors for endocrine resistance include intrinsic tumor characteristics (such as high tumor grade, low progesterone receptor expression, and HER2 positivity), specific genetic alterations (e.g., ESR1, PIK3CA, and PTEN mutations), prior exposure to endocrine agents, and incomplete suppression of estrogen signaling. Patient-related factors such as younger age, premenopausal status, and poor adherence to therapy also contribute. AI models enhance risk stratification by integrating heterogeneous data sources—clinical, pathological, and molecular—thus refining the identification of patients at greatest risk.

Clinical Features

Clinically, endocrine resistance is often detected retrospectively, when patients present with disease progression during or following endocrine therapy. Features may include new or enlarging metastatic lesions, rising tumor markers, or loss of hormone receptor positivity on repeated biopsy. Subclinical resistance may precede overt progression, and AI models are being trained to identify subtle data patterns that precede clinical manifestation, thus enabling earlier intervention.

Diagnosis

Conventional diagnosis of endocrine resistance relies on serial clinical assessment, imaging, and repeat biopsy to document disease progression and receptor status. Liquid biopsy techniques, analyzing circulating tumor DNA (ctDNA) for resistance-associated mutations, have gained prominence. AI algorithms applied to radiomic, genomic, and proteomic datasets have demonstrated promise in predicting resistance before clinical progression, with models achieving increasing accuracy and clinical relevance. Integration of AI-based predictive tools into diagnostic workflows is an area of active clinical research.

Treatment & Management

Management of endocrine resistance involves switching to alternative endocrine agents (e.g., from tamoxifen to aromatase inhibitors), combining endocrine therapy with targeted agents (such as CDK4/6 inhibitors or mTOR inhibitors), or transitioning to chemotherapy in advanced cases. AI-driven prediction of resistance can inform treatment sequencing and selection, potentially sparing patients from ineffective therapies and unnecessary toxicity. Personalized management, guided by AI predictions, represents a significant step toward precision oncology.

Recent Advances / Emerging Therapies

Recent years have witnessed remarkable advances in the application of AI to endocrine resistance prediction. Deep learning models analyzing electronic health records, imaging, and molecular data have achieved high sensitivity and specificity in identifying patients likely to fail endocrine therapy. Emerging therapies targeting resistance mechanisms—such as selective estrogen receptor degraders (SERDs), PI3K inhibitors, and novel CDK inhibitors—are being evaluated in conjunction with AI-guided patient selection strategies. Ongoing clinical trials are expected to validate the clinical utility of these approaches, with the ultimate goal of improving patient outcomes through earlier and more accurate prediction of resistance.

Guideline Recommendations

Current clinical guidelines emphasize the importance of regular reassessment of hormone receptor status and the use of molecular diagnostics to guide therapy modifications in the setting of resistance. While guidelines have yet to universally incorporate AI-based prediction algorithms, leading societies encourage participation in clinical trials evaluating these technologies. As evidence matures, it is anticipated that guideline panels will integrate AI-driven tools into standard practice, particularly for high-risk and refractory populations.

Conclusion

AI prediction of endocrine treatment resistance represents a paradigm shift in the management of hormone receptor-positive malignancies. By harnessing the power of machine learning and big data analytics, clinicians are increasingly able to identify patients at risk for resistance, personalize therapy, and improve clinical outcomes. While challenges remain—including the need for robust prospective validation, regulatory oversight, and integration into clinical workflows—the future of AI-driven predictive oncology is promising. Continued collaboration between clinicians, data scientists, and researchers will be essential to fully realize the potential of these transformative technologies in routine patient care.