EHR Parameters Predict Multiple Myeloma Risk
New Predictive Model Offers hope for Earlier Multiple Myeloma Detection
Table of Contents
Multiple myeloma (MM), a cancer of plasma cells, often isn’t diagnosed until it’s reached an advanced stage, significantly impacting treatment outcomes. However, a new study published in british Journal of Haematology details the progress and simplification of a machine learning model capable of predicting MM risk up to five years in advance, offering a potential paradigm shift in early detection and treatment.
Identifying Individuals at Risk: A machine Learning Approach
Researchers from an Israeli health service organization analyzed data from 4256 patients aged 40-85 diagnosed wiht MM between 2002 and 2019, excluding those with unconfirmed diagnoses or precursor conditions. They then matched these patients with a 10:1 ratio of healthy controls, carefully considering age, sex, and geographic location.
The initial investigation involved a deep dive into the patients’ electronic health records (EHRs) from five years before their MM diagnosis, examining over 200 clinical and laboratory parameters. This data was fed into a complex machine learning model designed to identify patterns indicative of future MM development. While highly accurate, the initial model proved too resource-intensive for widespread implementation.
“Thus, in order to implement the predictive model on other platforms, we developed a simplified model that coudl be used by any community physician with limited computational resources,” the authors explained. This commitment to accessibility is a key strength of the research.
A Simplified Model for Broad Application
The refined model utilized just 20 variables,focusing on readily available clinical and laboratory data. Key indicators identified in patients who later developed MM included:
Elevated Erythrocyte Sedimentation Rate (ESR): A marker of inflammation in the body.
Lower Hemoglobin Levels: Indicating potential anemia.
Reduced Absolute Neutrophil Counts: Suggesting a weakened immune response.
Decreased Neutrophil/Lymphocyte Ratio: An imbalance in white blood cell populations.
Higher Globulin levels: Potentially signaling abnormal protein production.
Increased Ferritin Levels: Indicating iron storage abnormalities.
This simplified model demonstrated a promising predictive capability, achieving an area under the receiver operator characteristic curve (AUC) of 0.72. An AUC of 0.72 suggests the model can reasonably discriminate between those who will and will not develop MM.
Implications for Early Intervention and Treatment
The potential impact of this model extends beyond earlier diagnosis. Research, including a study published in The New England Journal of Medicine (Mateos et al.,2013),suggests that early intervention in high-risk smoldering multiple myeloma - a precursor condition – with treatments like lenalidomide plus dexamethasone can significantly delay disease progression compared to a “watch and wait” approach. This highlights the importance of identifying at-risk individuals before they develop symptomatic MM.However, the authors acknowledge the need for careful consideration when implementing the model. Setting the risk threshold is crucial. A lower threshold would increase detection rates but also lead to more false positives and unnecessary testing, driving up healthcare costs. Conversely, a higher threshold could miss genuine cases.
The Path Forward: validation and Implementation
the researchers emphasize that their model currently lacks external validation – testing its performance on datasets outside of the original study population.Further research is needed to confirm its accuracy and generalizability across diverse populations and healthcare settings.
Despite this limitation,the authors are optimistic about the model’s immediate utility. “[T]he models,especially the simplified ones,can already be implemented by every community physician,” they conclude,offering a practical tool for proactive risk assessment and potentially improving outcomes for individuals at risk of developing multiple myeloma. This accessible approach empowers clinicians to move towards a more preventative and personalized approach to MM management.
References
- Mittelman M, Israel A, Oster HS, et al. Can we identify individuals at risk to develop multiple myeloma? A machine learning-based predictive model. Br J Haematol. Published Online June 16, 2025.Doi: 10.1111/bjh.20136
- Mateos MV, Hernández MT, Giraldo P, et al. Lenalidomide Plus Dexamethasone for High-Risk Smoldering Multiple Myeloma. N engl J Med. 2013;369(5):438-447. doi:10.1056/NEJMoa1