Type of pre-existing chronic conditions and their associations with Merkel cell carcinoma (MCC) treatment: Prediction and interpretation using machine learning methods

Yves Paul Vincent Mbous, Zasim Azhar Siddiqui, Murtuza Bharmal, Traci LeMasters, Joanna Kolodney, George A Kelley, Khalid Kamal and Usha Sambamoorthi

PLOS ONE, 2025, vol. 20, issue 7, 1-22

Abstract: Objective: This study examined the prevalence of pre-existing chronic conditions and their association with the receipt of specific cancer-directed treatments among older adults with incident primary Merkel Cell Carcinoma (MCC) using novel predictive and interpretable machine learning methods. Methods: We adopted a retrospective cohort study design with data from linked Surveillance, Epidemiology, and End Results (SEER) registry and Medicare Fee-For-Service claims databases of older adults (≥ 66 years) diagnosed with primary incident MCC between 2008 and 2017. The study cohort consisted of 1,668 older adults with incident MCC and continuous fee-for-service Medicare enrollment for 24 months. Chronic conditions were identified during 12 months before cancer diagnosis date. Type of any MCC treatment (surgery-SRx, radiotherapy-RTx, chemotherapy-CTx, immunotherapy-ITx, and hormonal therapy-HTx) were derived for 12 months following cancer diagnosis. Receipt of any of these treatments and their associations with pre-existing chronic conditions were analyzed using separate eXtreme Gradient Boosting (XGBoost) predictive models and SHapley Additive exPlanations (SHAP) methods. Results: High cholesterol (75.5%), HIV (71.5%), hypertension (67.7%), arthritis (54.9%), coronary artery disease (47.1%), diabetes (43.5%), and hepatitis (37.1%) were some of the highly prevalent pre-existing chronic conditions. MCC treatment varied by type of chronic conditions and treatment modality. For example, a lower percentage of those with hypertension received ITx compared to those without hypertension (5.7% vs. 17.1%). A higher percentage of those with high cholesterol (13.9% vs 10.8%) received HTx compared to those without high cholesterol. XGBoost predictions revealed high predictive accuracy (area under the curve ranged from 0.72 (CTx) to 0.99 (ITx)). Hypertension (ITx), diabetes and thyroid disorders (HTx), congestive heart failure (RTx), and high cholesterol (CTx) were among the top ten predictors of MCC treatment. Congestive heart failure (RTx), hypertension (CTx), heart disease (ITx), thyroid disorders (HTx), and osteoporosis (HTx) positively predicted treatment, whereas high cholesterol (CTx), hypertension (ITx, HTx) and diabetes (ITx, HTx) negatively predicted treatment. Conclusions: Pre-existing conditions were highly prevalent among older MCC adults. Cardiovascular and metabolic diseases were the top 10 leading predictors of cancer treatment. However, the associations varied by type of treatment. In spite of the good performance of the model, especially for ITx and HTx, there is a need to replicate these findings using other data sources that provide access to larger population subgroups.

Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0327964 (text/html)
https://journals.plos.org/plosone/article/file?id= ... 27964&type=printable (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:plo:pone00:0327964

DOI: 10.1371/journal.pone.0327964

Access Statistics for this article

More articles in PLOS ONE from Public Library of Science
Bibliographic data for series maintained by plosone ().