Billions of dollars’ worth of cancer drugs are discarded each year. Manufacturers must refund Medicare for some of this waste. A data-driven approach offers a practical path to greater efficiency.
Research estimates that up to 30% of oncology drugs are discarded each year, representing over $2 billion in annual costs in the United States alone.
For pharma companies, discarded drug volumes can reduce realized revenue and, under federal reimbursement policies, may require repayments to public payers, adding financial and operational pressure. For patients, the implications are tangible. Even with insurance coverage, billing is often based on the full vial rather than the amount administered, contributing to higher out-of-pocket expenses and insurance premiums.
This waste occurs because many oncology drugs are sold in fixed, single-dose vials, while patient doses vary based on factors such as weight or body surface area. In theory, sharing a vial across patients could reduce waste. In practice, this is often not feasible because patients with similar dosing needs are not reliably scheduled together, and many drugs have limited stability once a vial is opened.
Working at the intersection of data, technology and commercialization for the last 21 years—including at my current company, Indegene—I’ve seen firsthand how challenges like this shape decisions across the healthcare ecosystem. I view this issue as an opportunity to bring smarter forecasting and orchestration into the delivery of care.
CMS Rule Change Reshapes Manufacturer Accountability
In January 2023, the Centers for Medicare and Medicaid Services (CMS) implemented the Discarded Drug Refund Program. The policy requires manufacturers of certain single-dose or single-use package drugs covered under Medicare Part B to refund CMS for amounts discarded above a 10% threshold.
To enforce this policy, CMS issues reports identifying drugs subject to refunds, based on claims submitted by healthcare providers. The claims must include the exact amount of a single-dose container drug that was discarded (called the JW modifier) and attest that no amount from the container was discarded (the JZ modifier).
Manufacturers now face a delicate balancing act. Sales and utilization data often overstate true patient demand and complicate production planning and resource allocation, while refund obligations add a further financial dimension to this planning. Resizing vials may seem like a straightforward way to curb wastage, but in practice, the process is complex: regulatory approvals for such changes take time, manufacturing and distribution systems must be reconfigured, reimbursement and billing processes require updates, and pharmacy and clinical teams must be retrained to adapt to new workflows.
The Path Forward: Advanced Analytics
Given the complexity of resizing vials, a more practical solution lies in improving forecasting of oncology drug demand. Despite the scale of the issue, most forecasting practices remain rooted in traditional methods, relying heavily on historical sales data, static epidemiological models and long-range planning cycles. Real-world signals such as shifts from chemotherapy to targeted therapies, seasonal variations or trends in provider behavior are often excluded from forecasts.
This is where advanced analytics, powered by artificial intelligence (AI), can be transformative. By drawing on longitudinal claims data and electronic health/medical records (EHRs/EMRs), AI-driven forecasting models capture treatment complexity in ways traditional methods cannot.
The key lies in strengthening the data foundation. The model must understand the whole of treatment journeys and analyze dosing frequency and adherence patterns to more accurately estimate true utilization. Advanced statistical methods can help address missing or inconsistent dosing data. Together, these techniques create more robust forecasts that update as new therapies enter the market or treatment patterns shift.
Early pilots of AI-driven drug forecasting have shown significant improvements in predictive accuracy compared to traditional methods. In one such initiative, a leading global pharmaceutical company we collaborate with implemented a structured, data-driven approach to predict drug wastage in oncology therapies. The initiative began with mapping FDA-approved dosage cycles for selected drugs and incorporating patient parameters such as body weight and body surface area, both of which directly influence dosing. Multiple open and closed real-world data sources were triangulated to ensure comprehensive coverage, and only validated, complete patient journeys were included.
Using this integrated dataset, the team calculated drug wastage and benchmarked the results against historical CMS data. Our analysis revealed close alignment, with estimated wastage remaining within 2%–3% of CMS-reported values over a five-year period for therapies in multiple myeloma, mantle cell lymphoma, metastatic prostate cancer and metastatic small cell lung cancer. The results were input into forecasting models to enhance future decision-making.
From Smarter Forecasting to Positive Impact
When applied effectively, advanced analytics-driven forecasting reshapes how oncology drugs are managed across the system. It improves accuracy by reflecting real-world treatment changes, including therapy switches, patient drop-offs and site-specific practices rather than simply historical averages. This leads to more accurate demand prediction and avoids the pitfalls of over- or under-estimation that contribute to drug wastage.
It also helps manufacturers plan production more accurately and reduce the risk of stockpiles, and helps narrow the gap between supply and actual demand, particularly for temperature-sensitive and high-cost biologics for which timing is critical.
Over time, insights from predictive models could even inform packaging strategies. For instance, forecasting data may show that a large share of practices routinely discard residual volumes, making a strong case for introducing packaging formats better aligned with clinical usage patterns.
As forecasting models mature, the benefit will extend well beyond oncology. Real-time dashboards and intelligent planning cycles can enhance visibility of overall drug usage. By scaling these models thoughtfully and aligning them with diverse regulatory frameworks, data environments and organizational readiness, organizations can drive adoption across geographies and therapy areas, including in specialty biologics, immunotherapies and rare diseases.
For biopharma, the opportunity smart forecasting offers is not limited to reducing cost and improving efficiency. It offers the chance to build a more sustainable, adaptive and responsive healthcare system that better serves patients. The real question is not whether this transformation is possible, but how quickly stakeholders will choose to enact it.
