Predictive Tools May Help Physicians Identify Patients at Risk for Multiple Myeloma

A predictive tool may help primary care physicians identify patients at high-risk of multiple myeloma (MM), according to a study published in the British Journal of General Practice.

The researchers retrospectively analyzed electronic health records data from the United Kingdom’s Clinical Practice Research Datalink from 2000 through 2014. In total, 1,281,926 eligible patients were evaluated. Within two years, 737 patients (0.06%) were diagnosed with MM.

Characteristics and symptoms independently associated with MM diagnosis included older age; male sex; back, chest, and rib pain; nosebleeds; low hemoglobin, platelets, and white blood cell count; and raised mean corpuscular volume, calcium, and erythrocyte sedimentation rate. The median time to MM diagnosis from index was 5.6 months.

Using this data, the researchers created two clinical prediction models to be used by primary care physicians. The full blood count (FBC) model included demographics, symptoms, and blood test components; and the all-test model also contained all tests used for myeloma diagnosis, including blood test components, calcium, creatinine, and inflammatory markers.

According to the investigators, both tests discriminated well between people with and without myeloma, with improved calibration in the FBC model. The area under the curve (AUC) was 0.84 for the FBC model (95% confidence interval [CI] 0.81–0.87) and 0.87 for the all-test model (95% CI 0.84–0.90). The FBC model showed a sensitivity and specificity of 62% and 90%, respectively, and 72% and 90%, respectively, for the all-test model. Positive predictive values were 0.34% and 0.40% for the FBC and all-test models.

The models classified 0.12% of patients in the top decile of predicted myeloma risk. Investigating patients in this risk threshold would result in fewer false alarms than testing based on individual symptoms or blood test abnormalities alone. The FBC model achieved a ratio of 270 false alarms per one MM case diagnosis, whereas testing based on presence of anemia, for example, results in 500 false alarms per MM diagnosis.

“This study presents the diagnostic accuracy of multiple thresholds of predicted myeloma risk to illustrate rule-in and rule-out approaches by maximising specificity or sensitivity,” wrote the researchers in conclusion. “The authors recommend selecting a threshold with a specificity >90%, such as the 90th percentile of the FBC model, leading to more true positives and fewer false positives compared with other approaches, such as acting on anemia alone.”