In this JCO Article Insights episode, Emily Zabor summarizes two original articles from the June 1st, 2023 Journal of Clinical Oncology issue: “A Clinical Genetic Risk Score for Predicting Cancer-Associated Venous Thromboembolism: A Development and Validation Study Involving Two Independent Prospective Cohorts”) by Muñoz et al, and “Derivation and Validation of Clinical Risk Assessment Model for Cancer Associated Thrombosis in Two Unique Healthcare Systems”) by Li et al, as well as the accompanying editorial “Fine Tuning Venous Thromboembolism Risk Prediction in Patients with Cancer”) by Jean Marie Connors. The original reports describe the development and validation of two new risk prediction models for venous thromboembolism in cancer patients and the editorial puts them into context of existing tools.
TRANCRIPT
The guest on this podcast episode has no disclosures to declare.*
Emily Zabor: Welcome to JCO Article Insights* for the June 1st, 2023 issue of JCO*. I'm your host, Emily Zabor, JCO* Biostatistics Editorial Fellow. Today, I will be providing summaries of three articles.
The first article, titled “A Clinical Genetic Risk Score for Predicting Cancer-Associated Venous Thromboembolism: A Development and Validation Study Involving Two Independent Prospective Cohorts”) by Andres Muñoz and colleagues, describes the development and validation of a risk score for venous thromboembolism in oncology patients based on both clinical and genetic features, called the ONCOTHROMB score. In developing this model, the authors sought to address the fact that venous thromboembolism is among the leading causes of death among patients with cancer. Whereas hospitalized cancer patients are typically treated with thromboprophylaxis, outpatient treatment with thromboprophylaxis is only suggested for patients at high risk for venous thromboembolism identified according to the Khorana score.
The authors sought to determine whether incorporation of known genetic risk factors along with clinical factors would result in improved predictive accuracy. The risk score was developed in a cohort of 364 patients and was validated in an external cohort of 263 patients. The primary outcome of interest was venous thromboembolism within six months of a cancer diagnosis. The authors used logistic regression with backward selection with a p-value threshold of 0.25 to first select the genetic variants to include in the genetic risk score and then to separately select the clinical features to incorporate. Then the genetic risk score and clinical features were combined into a single multivariable logistic regression model, and backward selection was performed again.
The final model included nine genetic variants, tumor site, TNM stage, and a BMI of greater than 25. In the validation data, the ONCOTHROMB score using a threshold selected with the Youden index resulted in an AUC of 0.686 as compared to an AUC of 0.577 for the Khorana score with a threshold of 3. The ONCOTHROMB score had statistically significantly higher sensitivity, whereas the Khorana score had statistically significantly higher specificity. The authors conclude that the ONCOTHROMB score demonstrated improved predictive ability and should be investigated further in clinical trials.
The second article, titled “Derivation and Validation of Clinical Risk Assessment Model for Cancer Associated Thrombosis in Two Unique Healthcare Systems”) by Ang Li and colleagues, describes the development and validation of a risk assessment model for venous thromboembolism, pulmonary embolism, and lower-extremity deep vein thrombosis in oncology patients undergoing systemic therapy. The authors developed this model to address the increased morbidity and mortality associated with venous thromboembolism among cancer patients and the fact that risk reduction for use of thromboprophylaxis as well as the efficacy-safety trade-off, and cost-effectiveness have been shown to be higher among patients selected as high risk for venous thromboembolism.
The primary outcome was venous thromboembolism within six months of treatment initiation. In the development data set, the authors used lasso-penalized logistic regression analysis to shrink some of the covariates to zero as a form of variable selection, then a multivariable logistic regression model was fit to the remaining covariates and those with an estimated odds ratio greater than 1.2 or less than 0.8 were retained in the final model. A linear risk score was created from the resulting beta coefficients. The risk assessment model was developed in a cohort of 9769 patients and validated in an external cohort of 79,517 patients. The final model included eleven factors: cancer subtype, pre-therapy BMI greater than or equal to 35, pre-therapy white blood cell count greater than 11, pre-therapy hemoglobin less than 10, pre-therapy platelet greater than or equal to 350, cancer staging 3 to 4, targeted or endocrine monotherapy, lifetime history of venous thromboembolism, history of paralysis and immobility in the last 12 months, recent hospitalization lasting greater than three days in the last three months, and Asian Pacific Islander race.
The final linear risk score was stratified into six categories and then dichotomized based on overall venous thromboembolism incidence of 7%, resulting in about half of the patient population being classified as high risk for venous thromboembolism. In the validation data, the model was associated with a c-statistic of 0.68 as compared to 0.6 for the Khorana score with a threshold of 2. Sensitivity analyses demonstrated that the model had similar discrimination in subgroups according to age, sex, and race-ethnicity. The authors conclude that their new risk assessment model has the potential to improve patient selection for thromboprophylaxis.
The third and final article titled “Fine Tuning Venous Thromboembolism Risk Prediction in Patients with Cancer”) by Jean Marie Connors, is an editorial accompanying the two previously summarized articles. Dr. Connors emphasizes that venous thromboembolism is the second leading cause of death in cancer patients, second only to cancer itself, and that while treatment with thromboprophylaxis has been shown to significantly reduce the occurrence of venous thromboembolism in randomized controlled trials, since the overall risk of venous thromboembolism is so low, treating all patients would result in significant overtreatment. Therefore, prediction tools are needed and many have been developed, with the Khorana score being the most widely used because it is the best validated and easy to calculate.
Dr. Connors observes that in the risk assessment model by Li and colleagues, more granularity is added to the Khorana score by stratifying gastrointestinal cancer subtypes and by adding aggressive NHL, myeloma, brain tumors, and sarcoma, and also incorporates additional cancer-specific and patient-specific risk factors. But while this risk assessment model reclassified approximately 25% of patients in both the development and validation sets, as compared to the Khorana score, there was only modest improvement in discrimination. With the c-index improving from 0.65 to 0.71 in the development set and from 0.6 to 0.68 in the validation set from the Khorana score to the new risk assessment model.
On the other hand, the ONCOTHROMB score developed by Muñoz and colleagues contains fewer clinical features as compared to the Khorana score but adds a genetic risk score based on eleven genetic variants. The Khorana score performed quite poorly in both the development and validation data in this study, with AUCs of just 0.58 and 0.56, so improvements to 0.781 and 0.720 were relatively large. Connors notes that prospective validation of both scores is still needed, especially of the ONCOTHROMB score in more diverse populations, as allele frequencies can vary widely and the studied populations were predominantly Western European. Connors cautions that there is still a need to demonstrate mortality benefit from venous thromboembolism prophylaxis and that the risk of bleeding and other complications must be properly weighed against the benefits.
That concludes this episode on the articles “A Clinical Genetic Risk Score for Predicting Cancer-Associated Venous Thromboembolism: A Development and Validation Study Involving Two Independent Prospective Cohorts” and “Derivation and Validation of Clinical Risk Assessment Model for Cancer Associated Thrombosis in Two Unique US Healthcare Systems,” and the associated editorial, “Fine Tuning Venous Thromboembolism Risk Prediction in Patients with Cancer.”
Thank you for listening and please tune in for the next issue of JCO Article Insights*.
The purpose of this podcast is to educate and to inform. This is not a substitute for professional medical care and is not intended for use in the diagnosis or treatment of individual conditions. * *
Guests on this podcast express their own opinions, experience, and conclusions. Guest statements on the podcast do not express the opinions of ASCO. The mention of any product, service, organization, activity, or therapy should not be construed as an ASCO endorsement.*
Like, share, and subscribe )so you never miss an episode and leave a rating or review.
Editorial:
Fine Tuning Venous Thromboembolism Risk Prediction in Patients With Cancer)
Derivation and Validation of a Clinical Risk Assessment Model for Cancer-Associated Thrombosis in Two Unique US Health Care Systems)
A Clinical-Genetic Risk Score for Predicting Cancer-Associated Venous Thromboembolism: A Development and Validation Study Involving Two Independent Prospective Cohorts)
Original Report:
Find more articles from the June 1 issue)