TL;DR
This study leverages a massive, nationwide electronic health record (EHR) dataset to provide the most comprehensive epidemiological assessment of cancer-associated venous thromboembolism (VTE) in the US to date. It reveals that VTE risk remains high, particularly for certain cancer types and stages, and importantly, that VTE within the first year of cancer diagnosis significantly increases mortality.
What Problem Does This Paper Address?
Cancer is a major risk factor for developing blood clots, specifically venous thromboembolism (VTE). These clots can disrupt cancer treatment, leading to significant morbidity and mortality. While the incidence of VTE in cancer patients is known to be high (up to 20%), previous studies have been limited by small sample sizes, homogenous patient populations (e.g., specific demographics or geographic regions), or reliance on older data. This leaves a gap in understanding the current epidemiology of cancer-associated VTE (CAT) across the diverse US population, including how it varies by cancer type, treatment, and patient characteristics, and its impact on survival.
Methods
The researchers employed a retrospective cohort study design using a large, longitudinal EHR dataset. They developed a robust VTE phenotyping algorithm using ICD-10-CM codes and medication data, validated for accuracy in identifying VTE events. To identify risk factors and associations, they utilized multivariable Cox regression models, incorporating time-varying covariates for factors like systemic therapy. For feature selection in their models, they employed the Absolute Shrinkage and Selection Operator (LASSO) to identify statistically robust and clinically relevant predictors.
Datasets
The primary dataset used was Epic Cosmos, a collaborative dataset of Epic electronic health record systems. This dataset is described as longitudinal, comprehensive, and representative of the US census in terms of age, race, ethnicity, and insurance coverage, encompassing patients from all 50 states. The study included data from 2018 to 2023. No new benchmarks were introduced; rather, existing data was leveraged for analysis.
Key Insights and Ideas
- VTE is a persistent threat: Despite advances in cancer treatment, the incidence of CAT has not decreased, highlighting the ongoing challenge.
- Not all cancers are equal: Certain cancer types (pancreatic, bile/gallbladder, acute lymphoblastic leukemia) and advanced stages are significantly associated with a higher risk of VTE.
- Treatment matters, but complexity exists: Systemic therapies, particularly chemotherapy, increase VTE risk, but the impact can vary within treatment categories.
- VTE is a significant mortality driver: The occurrence of VTE within the first year of cancer diagnosis is an independent predictor of increased mortality, underscoring its clinical importance.
- Anticoagulant trends are shifting: There's a notable increase in the use of apixaban and a decrease in low molecular weight heparin for VTE treatment in cancer patients.
Key Results
The study identified 1,628,626 patients with a new cancer diagnosis between 2018 and 2023.
- The cumulative incidence of VTE at 6 months was 2.7%, and at 12 months was 3.7%.
- For patients receiving systemic therapy within 90 days of diagnosis, the 12-month VTE incidence was higher at 5.68%.
- Major risk factors for VTE included prior VTE, cancer type (e.g., pancreatic cancer had a HR of 3.3), advanced stage (Stage IV had a HR of 3.23), and chemotherapy-based treatment regimens (HR 1.93).
- VTE within the first year of cancer diagnosis was independently associated with increased mortality (e.g., PE onset HR 2.4, LE-DVT onset HR 2.1, UE-DVT onset HR 2.3).
- Pancreatic cancer (HR 7.61) and Stage IV/metastatic cancer (HR 6.33) were associated with the highest mortality risk when VTE occurred.
Limitations and Future Work
The authors acknowledge several limitations:
- Data Source: Cosmos only includes participating Epic health systems, meaning care outside these systems might be missed. However, they mitigated this with strict inclusion criteria and found comparable survival rates to SEER data.
- Misclassification: Reliance on administrative codes for outcome ascertainment can lead to misclassification, though a validated VTE phenotyping algorithm was used.
- Cancer Diagnosis/Stage Timing: EHR data might not perfectly reflect true clinical dates for diagnosis or stage, potentially introducing left truncation or misclassification.
- Representativeness: While broadly representative, the cohort overrepresented White patients and underrepresented Asian and Hispanic individuals.
- Anticoagulation Data: Distinguishing between treatment-dose anticoagulation and prophylaxis was challenging due to data limitations.
Future work could involve:
- More granular classification of anti-cancer treatments to better understand their specific impact on VTE risk.
- Further research into refining targeted prophylactic strategies for high-risk subgroups.
- Integrating EHR-based risk tools into routine practice to improve VTE prevention.
- Exploring the complex interplay of social determinants of health and VTE risk in cancer patients.
Summary updated