Thromboprophylaxis remains a critical concern in cancer patients initiating chemotherapy due to their heightened risk of venous thromboembolism (VTE). Although the Khorana score (KS) has been widely used as a risk assessment tool, recent studies have highlighted its limitations in accurately predicting VTE incidence [1]. Moreover, existing risk models have often been derived from predominantly white or European populations, limiting their applicability to more racially and ethnically diverse cohorts.
A new risk assessment model
Fortunately, recent advancements in clinical informatics, including enterprise data warehousing (EDW), machine learning, and natural language processing (NLP), offer promising opportunities to develop improved risk assessment models. In this context, this commentary discusses a study that utilized a cancer registry and EDW-linked data sets to derive and externally validate a simplified VTE risk assessment model (RAM) in patients with newly diagnosed cancer receiving systemic therapy [2].
By addressing the shortcomings of previous models and incorporating a more diverse patient population, this new RAM holds the potential to enhance thromboprophylaxis strategies and improve patient outcomes.
The model was created using data from racially and ethnically diverse populations in the United States. The RAM includes 11 risk predictors, such as cancer subtype, advanced stage, targeted/endocrine therapy, history of PE/DVT, history of paralysis/immobility, recent hospitalization, and Asian/Pacific Islander ethnicity [2].
The new RAM stratifies patients into high- and low-risk groups with 6-month VTE risks of 8–10% and 3%, respectively. The model outperformed existing clinical risk scores, offering better discrimination and coverage of potentially preventable VTE cases in the high-risk group. The RAM relies solely on clinical predictors, making it easily implementable without specialized biomarkers or external tools [2].
Advantages
The strength of the new RAM lies in its simplicity and intuitive design. By incorporating both cancer-specific and patient-specific predictors, the model achieves a stratification of approximately 50% of cancer patients undergoing modern systemic therapy into high-risk and low-risk groups. Identifying patients with a higher likelihood of developing VTE allows for targeted preventive measures, potentially reducing the burden of this serious complication [2].
One notable advantage of the improved RAM is its modification of the original KS components, clarifying cancer subtype definitions and expanding eligibility to encompass all cancer and therapy types. This adaptation leads to more accurate risk group assignments, increasing discrimination and coverage in the validation cohort. Additionally, the reliance on clinical predictors, without needing specialized biomarkers, facilitates real-time implementation and reduces the complexity associated with other risk models [2].
Furthermore, including Asian/Pacific Islander ethnicity as a predictor addresses an important disparity in VTE incidence among different racial/ethnic groups. While the authors chose not to include race/ethnicity subgroups with imprecise definitions, the new model demonstrates consistent performance across diverse populations, providing an equitable approach to risk assessment [2].
An independent external validation of this model has been conducted in a comprehensive cancer center.
External validation
The validation study encompassed adult patients with incident cancer diagnoses from the MD Anderson Cancer Center Tumor Registry between January 2017 and January 2021. The analysis included data from 21,142 cancer patients with a median follow-up of 8.1 months, among whom 5.7% experienced VTE within six months after systemic therapy. The distribution of the novel risk score revealed a gradient of risk, with higher scores associated with increasing 6-month VTE incidence rates. Compared to the Khorana score, the novel risk score demonstrated superior discrimination and calibration, with a c statistic of 0.71 [3].
The risk score has shown consistent performance across 110,428 patients from various cancer cohorts, with a c statistic ranging from 0.68 to 0.71. It demonstrates a modest improvement over the Khorana score, capturing 25% more VTE events in the high-risk group. Moreover, the model calibration remains stable, even when applied to patient populations with significant demographic variations. This indicates that the simple risk score offers flexibility and precision in estimating VTE incidence across diverse cancer patient populations [3].
Despite the performance of the novel risk score, the authors acknowledge the continued relevance of the Khorana score, especially in healthcare systems without integrated EHR data. The current risk score complements the Khorana score by incorporating additional risk predictors and clinical judgment from EHR data. However, further data on bleeding outcomes are needed to facilitate the clinical implementation of risk prediction scores for VTE in cancer patients [3].
Potential and limitations
Notably, the new risk score successfully reclassified a significant portion (20%) of patients previously assessed by the Khorana score, leading to a 25% increment in VTEs captured in the high-risk group. This result highlights the potential of the novel risk score to identify patients who may benefit from intensified thromboprophylaxis and closer VTE monitoring during cancer treatment [2, 3].
Despite its potential, the study has some limitations, including being retrospective in nature and requiring confirmation in a large prospective study. Furthermore, the model’s utility in patients with hematologic malignancy remains uncertain due to limited data in this category. Additionally, its role as a predictive biomarker for pharmacologic thromboprophylaxis has not been tested yet [2].
Conclusion
Overall, the new RAM represents an important step in improving thromboprophylaxis and patient selection for cancer-associated thrombosis, applicable to diverse cancer patient populations. However, further research and validation are necessary to establish its widespread clinical use.
The consistent discrimination and calibration of the novel risk score across cohorts with heterogenous demographics offer promising implications for its broader applicability in clinical practice. If validated in further studies, this risk score may be a valuable tool for selecting high-risk populations for clinical trials and guiding personalized VTE monitoring in cancer patients undergoing systemic therapy. Adopting such a risk score could pave the way for improved clinical outcomes and a more targeted approach to thromboprophylaxis in this vulnerable patient population.
References
- Khorana AA, Kuderer NM, Culakova E, et al. Development and validation of a predictive model for chemotherapy-associated thrombosis. Blood 2008;111:4902-4907.
- Li A, La J, May SB, et al. Derivation and Validation of a clinical risk assessment model for cancer-associated thrombosis in two unique US health care systems. J Clin Oncol. 2023;41(16):2926-2938. doi:10.1200/JCO.22.01542
- Li A, De Las Pozas G, Andersen CR, et al. External validation of a novel electronic risk score for cancer-associated thrombosis in a comprehensive cancer center. Am J Hematol. 2023;98(7):1052-1057. doi:10.1002/ajh.26928