In a groundbreaking study published in BMC Cancer, researchers unveil the power of artificial intelligence to revolutionize the understanding and treatment of localized and locally advanced prostate cancer. Harnessing vast troves of electronic health records from eight major Spanish hospitals, this research offers unprecedented real-world insights into patient outcomes and clinical characteristics, reshaping the future landscape of prostate cancer care.
Prostate cancer remains one of the most prevalent malignancies affecting men worldwide, with localized and locally advanced stages presenting unique therapeutic challenges. To address these challenges, investigators employed an advanced natural language processing and machine learning platform known as EHRead®, allowing them to systematically extract and analyze unstructured clinical narratives from electronic health records dated 2014 to 2018. This approach circumvented traditional data limitations, offering a comprehensive view of patient journeys across multiple institutions.
Over 22,000 prostate cancer patients were initially identified, with 14,434 (65.1%) classified as having localized or locally advanced disease. From this cohort, 5,331 incident cases with robust clinical information were selected for detailed outcome analysis. These patients were monitored for a median duration of 2.3 years, enabling an evaluation of key endpoints such as real-world overall survival, metastasis-free survival, and event-free survival.
One of the most striking elements of the study was the nuanced risk stratification performed on the patient cohort. Patients were categorized into groups based on risk levels: low risk (7.3%), intermediate risk (36.5%), high risk (26.0%), locally advanced prostate cancer (5.9%), and an undefined risk group (24.2%). This stratification revealed a clear gradient in survival outcomes, with higher-risk groups exhibiting significantly worse prognoses, confirming clinical expectations and bolstering the importance of precise risk assessment.
Treatment modalities within the cohort reflected contemporary clinical practice. Radiotherapy emerged as the most common initial treatment (40.7%), followed closely by radical prostatectomy (37.1%). Lesser proportions underwent active surveillance or watchful waiting (6.4%), brachytherapy (4.2%), or androgen deprivation therapy (ADT) monotherapy (3.3%). The distribution of treatment choices sheds light on real-world clinical decision-making processes, balancing efficacy, side effects, and patient preferences.
Critically, patients who received ADT monotherapy prior to more aggressive interventions displayed the poorest baseline health characteristics and clinical outcomes. This finding underscores the limitations of ADT when used in isolation and highlights the necessity for more tailored, combination therapeutic strategies in managing advanced disease.
The temporal survival outcomes further illuminate the urgent need for early intervention and effective treatment algorithms. At 36 months, overall survival rates stood at 98% for low-risk LPC, 97% for intermediate risk, 93% for high-risk localized disease, and 91% for locally advanced prostate cancer. While these figures demonstrate encouraging short-term efficacy for curative treatments, they also emphasize the persistent risk of disease progression and oncologic events within a relatively short follow-up window.
The adoption of artificial intelligence technologies like EHRead® for retrospective observational studies heralds a transformative era in oncological research. This pioneering methodology enables researchers to transcend the constraints of structured databases, capturing rich clinical detail embedded in physician notes, pathology reports, and diagnostic imaging descriptions. Such granular data extraction enhances the accuracy of real-world evidence and facilitates personalized medicine initiatives.
Moreover, the study highlights the growing importance of interdisciplinary collaboration between data scientists, clinicians, and oncologists. By leveraging machine learning algorithms alongside expert clinical judgment, the research team was able to delineate complex patterns of disease progression and therapeutic response that would be challenging to detect through conventional means.
From an epidemiological perspective, the large sample size and multi-center nature of the cohort bolster the generalizability of findings, providing a compelling template for future multinational studies. The integration of diverse hospital datasets demonstrates the feasibility of interoperable data ecosystems that respect patient privacy while maximizing scientific insight.
Beyond survival statistics, the research points to several clinically relevant implications. The high incidence of oncologic events despite curative intent treatments necessitates constant vigilance and possibly the integration of novel therapeutics such as targeted agents or immunotherapies. Furthermore, the stratified survival rates highlight risk groups that may benefit from intensified surveillance or adjunctive therapies.
The study’s retrospective design does impose certain limitations, including potential biases related to data completeness and heterogeneity among institutions. Nevertheless, the innovative use of AI-driven data extraction mitigates many traditional challenges associated with retrospective analyses, providing a blueprint for leveraging electronic health records effectively in oncology research.
Ultimately, this work advocates for proactive clinical strategies grounded in sophisticated risk stratification and data-driven insights. Tailored treatment pathways, early intervention, and continuous monitoring may collectively improve patient outcomes, reducing metastasis rates and prolonging survival in localized and locally advanced prostate cancer.
As artificial intelligence continues to integrate into healthcare, the fusion of real-world data and machine learning promises to revolutionize cancer care paradigms. This study not only delivers vital evidence on prostate cancer outcomes but also exemplifies the transformative potential of AI-augmented research methodologies.
In summary, through meticulous analysis of electronic health records powered by advanced AI techniques, this research sheds new light on the clinical trajectories of localized and locally advanced prostate cancer patients. It highlights critical prognostic factors, treatment patterns, and survival outcomes, underscoring the urgent need for enhanced stratification and innovative care strategies in this prevalent malignancy.
Subject of Research: Real-world clinical characteristics and outcomes in localized and locally advanced prostate cancer using artificial intelligence-based analysis of electronic health records.
Article Title: Real-world evidence in localized and locally advanced prostate cancer: applying artificial intelligence to electronic health records
Article References: Maroto, J.P., Puente, J., Conde Moreno, A. et al. Real-world evidence in localized and locally advanced prostate cancer: applying artificial intelligence to electronic health records. BMC Cancer 25, 1618 (2025). https://doi.org/10.1186/s12885-025-14828-z
Image Credits: Scienmag.com
DOI: https://doi.org/10.1186/s12885-025-14828-z