How machine learning can uncover hidden drivers behind treatment choices and support smarter pharmaceutical decision-making.

Introduction

Multiple myeloma (MM) is one of the most complex and rapidly evolving oncology / hematology markets. New treatment classes, diverse patient profiles, varying healthcare infrastructures, and increasingly individualized treatment pathways all contribute to a highly fragmented decision-making environment.

For pharmaceutical companies, understanding why one treatment is selected over another has become increasingly challenging. Traditional analytical approaches can identify individual factors that influence prescribing decisions, but they often struggle to capture how dozens of variables interact simultaneously.

This is where machine learning and explainable artificial intelligence (AI) can play a valuable role.

At the 2026 EphMRA Conference, Hilary Worton (SVP, Client Experience Optimization, APLUSA) explored how machine learning and explainable AI can help decode treatment decisions in multiple myeloma. Drawing on real-world patient-level data, the presentation demonstrated how advanced analytics can uncover decision drivers that may remain invisible through conventional analytical approaches or would require considerably more time to identify.

Why Multiple Myeloma Is an Ideal Use Case for AI

Few therapy areas illustrate healthcare complexity better than multiple myeloma.

Treatment decisions are influenced by a wide range of variables, including:

• Patient age and fitness
• Cytogenetic risk profile
• Prior treatment history
• Stem cell transplant eligibility
• Comorbidities
• Treatment setting and physician environment
• Access to innovative therapies

Each factor may influence prescribing decisions differently depending on the clinical context.

The challenge is that these variables rarely act independently. Instead, they interact in ways that can be difficult to identify using traditional analytical approaches.

Machine learning provides an opportunity to analyze these multi-dimensional relationships simultaneously and reveal patterns hidden within large, real-world datasets.

Moving Beyond Prediction to Understanding

For pharmaceutical decision-makers, understanding why HCPs choose a treatment over another is really important, especially in fields where there is a wide range of potential approaches and no clear established protocols.

This is why the study used explainable AI techniques to identify not only treatment patterns but also the factors influencing them.

Using patient-level data collected across multiple European markets, machine learning models were trained to evaluate treatment decisions and determine which characteristics had the greatest impact on therapy selection.

The result was a transparent framework capable of translating complex analytics into actionable business and clinical insights.

What the Analysis Revealed

Several findings confirmed predictable, or highly likely market dynamics.

For example, factors such as patient age, their fitness, and stem cell transplant eligibility emerged as major drivers of treatment choice across innovative multiple myeloma therapies.

However, some of the most valuable insights came from uncovering less obvious patterns.

The analysis identified specific clinical characteristics and cytogenetic profiles that appeared to influence treatment selection more strongly than expected. It also highlighted meaningful differences between countries, suggesting that healthcare infrastructure and local practice patterns may play a larger role in adoption decisions than is often assumed.

These findings demonstrate how AI can move beyond confirming existing assumptions and generate new hypotheses for further investigation.

Why This Matters for Pharmaceutical Companies

The value of machine learning does not lie in replacing human expertise.

On the contrary, the greatest impact comes from combining advanced analytics with therapeutic-area knowledge, market understanding, and knowledge about the data itself.

When applied appropriately, AI can help pharmaceutical teams:

• Identify the key drivers of treatment adoption
• Better understand patient segmentation opportunities
• Detect barriers to access and utilization
• Refine commercial and brand strategies
• Generate evidence-based hypotheses for future research
• Anticipate how evolving market conditions may influence treatment choices

In increasingly competitive oncology markets, these capabilities can provide a significant advantage when making strategic decisions.

From Static Analysis to Continuous Learning

Traditionally, treatment pathway analyses provide a snapshot of market behavior at a given moment in time.

AI-enabled approaches offer something different: the ability to continuously explore evolving datasets and identify emerging patterns as markets change.

As innovative therapies such as CAR-T cell therapies, bispecific antibodies, and antibody-drug conjugates (ADCs) continue to reshape the multiple myeloma landscape, this ability to uncover patterns in complex decision drivers becomes increasingly valuable.

The goal is no longer simply to describe what happened yesterday. It is to better understand and interpret and take actions for brands to impact what is likely to happen next.

Conclusion

As healthcare markets become more complex, the limitations of traditional analytical approaches become increasingly evident.

Machine learning and explainable AI offer a powerful complement to conventional market research by uncovering patterns that may otherwise remain hidden. When combined with robust real-world data and expert interpretation, these approaches can provide pharmaceutical companies with a deeper understanding of how treatment decisions are made in practice.

The future of healthcare insights is not about replacing human expertise with AI. It is about combining predictive models, therapeutic knowledge, and business understanding to make better-informed decisions in increasingly complex markets.