Audience at Leukemia and Lymphoma: Europe and the USA, Linking Knowledge and Practice
DUBROVNIK, CROATIA—An updated scoring system can more accurately identify lymphoma patients who may require thromboprophylaxis, according to researchers.
The revised scoring system, ThroLy, proved more effective than other systems for predicting thromboembolic events in lymphoma patients.
Researchers found the updated ThroLy had a positive predictive value of 22% to 25%, a negative predictive value of 96%, sensitivity of 56% to 57%, and specificity of 85% to 87%.
Darko Antić, MD, PhD, of the University of Belgrade in Serbia, presented these findings at Leukemia and Lymphoma: Europe and the USA, Linking Knowledge and Practice.
Dr. Antić said he and his colleagues developed ThroLy because other systems used to predict venous thromboembolism (VTE) are not quite right for lymphoma. He noted that the Padua score is not designed for cancer patients, and the Khorana score is predominantly for solid tumor malignancies.
“It’s good . . . , but it’s not specific for lymphoma patients,” Dr. Antić said.
With this in mind, he and his colleagues developed ThroLy. They based the scoring system on variables used in the Padua and Khorana systems as well as variables that are specific to lymphoma patients.
In a past study*, the researchers found several variables that were independently associated with risk for VTE in lymphoma:
- Previous VTE
- Previous acute myocardial infarction/stroke
- Mediastinal involvement
- Body mass index > 30 kg/m2
- Reduced mobility
- Extranodal localization
- Development of neutropenia
- Hemoglobin level < 100g/L.
Previous VTE, previous acute myocardial infarction/stroke, obesity, and mediastinal involvement were all worth 2 points, and the other factors were worth a single point.
Patients with scores of 0 to 1 were considered low-risk, patients with scores of 2 to 3 were considered intermediate-risk, and patients with scores of 4 or greater were considered high-risk.
Prospective validation
To validate and refine ThroLy, Dr. Antić and his colleagues used it to assess 1723 lymphoma patients treated at 8 institutions in Austria, Croatia, France, Jordan, Macedonia, Spain, Switzerland, and the United States.
Patients had indolent non-Hodgkin lymphoma (n=467), aggressive non-Hodgkin lymphoma (n=647), chronic lymphocytic leukemia/small lymphocytic lymphoma (n=235), and Hodgkin lymphoma (n=366). Most subjects (84%) were outpatients.
Nine percent of patients had thrombosis (n=142), with 7% having VTE (n=121).
ThroLy had a positive predictive value of 17%, compared to 11% with Khorana and 13% with Padua. The negative predictive value was 93%, 92%, and 95%, respectively.
The sensitivity was 51% with ThroLy, 42% with Khorana, and 70% with Padua. The specificity was 72%, 64%, and 52%, respectively.
“The positive predictive value was low [with ThroLy] but definitely higher than the positive predictive value of the other two [scoring systems],” Dr. Antić noted.
Updated models
To further improve ThroLy, the researchers updated the system, creating two new models.
Model 1 included the following variables:
- Type of lymphoma/clinical stage (aggressive/advanced)—1 point
- Previous VTE—5 points
- Reduced mobility—2 points
- Hemoglobin level < 100 g/L—1 point
- Presence of vascular devices—1 point.
Model 2 included all of the aforementioned variables as well as thrombophilic condition, which was worth 1 point.
With these models, patients were divided into two risk groups—low-risk (≤ 2 points) and high-risk (>2 points).
For Model 1, the positive predictive value was 22%, the negative predictive value was 96%, the sensitivity was 56%, and the specificity was 85%.
For Model 2, the positive predictive value was 25%, the negative predictive value was 96%, the sensitivity was 57%, and the specificity was 87%.
Dr. Antić said there were no major differences in model discrimination and calibration according to the country in which a patient was treated or whether patients were treated in inpatient or outpatient settings.
Dr. Antić did not report any conflicts of interest.
*Antić D et al. Am J Hematol. 2016 Oct;91(10):1014-9. doi: 10.1002/ajh.24466.