A twenty-eight-fold error is not a rounding discrepancy. It is not a minor data-entry artifact or an acceptable margin of analytical noise. Yet that is precisely the magnitude of pricing distortion that the National Council for Prescription Drug Programs has documented when billing units and CMS unit types fall out of alignment — a scenario that occurs more often than anyone in the pharmaceutical data business cares to admit.
The problem is structural, not incidental. Three of the most widely referenced drug pricing benchmarks in the United States — Wholesale Acquisition Cost, the National Average Drug Acquisition Cost
, and average sales price — each report per-unit drug costs. But “unit” means different things in each dataset. WAC prices are typically published in the manufacturer’s chosen packaging unit. NADAC reflects invoice prices reported by retail community pharmacies, normalized to dispensing quantities. ASP, meanwhile, captures a volume-weighted average net of concessions reported to CMS by manufacturers. The unit denominators across these systems are not interchangeable, and in many cases they are not even commensurable.
Consider budesonide. The drug treats conditions ranging from COPD to inflammatory bowel disease, and it exists in formulations that span inhalers, nebulization suspensions, oral capsules, and rectal preparations. Across NADAC, WAC, and ASP, the unit quantities for budesonide diverge so dramatically that reported per-unit prices produce wildly disparate cost figures for what is pharmacologically the same molecule. A nebulization suspension priced per milliliter, a capsule priced per each, and an inhaler priced per gram inhabit entirely separate numerical universes. No amount of cross-referencing resolves the discrepancy without first establishing a common denominator — and no federal standard mandates one.
This is not a niche concern. Drug pricing databases underpin formulary decisions, reimbursement negotiations, health economic modeling, legislative analysis, and investor valuations. Every downstream user inherits the unit ambiguity, often without knowing it.
The instinct is to normalize. Convert everything to a shared billing unit — NCPDP’s EA, ML, or GM designations — and then compare. But normalization assumes the upstream data maps cleanly onto those categories, which it frequently does not. A drug dispensed in a multi-dose inhaler does not reduce to milliliters in the same way a nebulization suspension does. The conversion factor is not a physical constant; it is a convention, and conventions vary by data publisher, by NDC, and sometimes by quarter.
Stoichiometric normalization — adjusting for the active pharmaceutical ingredient’s molecular weight and concentration to produce a standardized cost per milligram of API — has appeal as a theoretical framework. It offers a physics-level foundation that sidesteps the ambiguity of packaging units entirely. In practice, it introduces a different set of problems. Bioavailability differs across formulations. A milligram of budesonide delivered via dry powder inhaler does not produce the same therapeutic effect as a milligram delivered rectally. Pricing the molecule without pricing the delivery mechanism strips away precisely the clinical context that makes the comparison meaningful.
The industry’s workaround, such as it is, borrows from SSR Health, which has long been considered something close to a gold standard for net pricing analytics. SSR’s approach involves rolling unit-level data up into a “price per course of treatment” — a composite metric that bundles dosing, duration, and frequency into a single comparable figure. The method works, up to a point. It makes prices comparable across formulations by anchoring to clinical use rather than packaging geometry. But it also requires a dense layer of assumptions: maximum labeled dosing, a default treatment duration (often one year when the label is silent), and a singular indication profile for drugs that treat multiple conditions.
For a drug like budesonide, those assumptions fracture almost immediately. An eight-week induction course for eosinophilic esophagitis bears no resemblance to chronic maintenance therapy for Crohn’s disease. Pricing both at a one-year duration produces a number that describes neither use case accurately. The resulting “course of treatment” cost is a statistical centroid floating between clinical realities — useful for screening, perhaps, but unreliable for anything that demands precision.
SSR’s methodology is also opaque. The assumptions behind its treatment rollups are not publicly documented in detail. Researchers who rely on the data are, in effect, trusting a black box calibrated by a private vendor. The American Enterprise Institute has published analyses noting the limitations of SSR’s volume estimates, particularly for drugs distributed through specialty or non-traditional pharmacy channels, where dispensing patterns deviate from the retail norms that anchor the model. The recommendation in those analyses — restrict comparisons to retail drugs for highest accuracy — is a concession, not a solution. It means the methodology’s reliability is inversely proportional to the complexity of the drug’s distribution profile.
The deeper issue is that no authoritative body has built the reconciliation infrastructure. CMS publishes NADAC weekly but does not publish crosswalk tables mapping its units to WAC or ASP denominators. The FDA maintains the National Drug Code directory but does not enforce unit-level consistency across pricing datasets. Manufacturers report to multiple programs — ASP to CMS quarterly, AMP for Medicaid rebate calculations, WAC to pricing compendia — using whatever unit conventions each program requires, with no mandate for internal consistency.
What exists instead is a patchwork of ad hoc solutions. Analytics platforms like the one Daily Remedy is building attempt cross-referencing programmatically, flagging discrepancies for manual review. Payers maintain internal crosswalks that are proprietary and rarely shared. Academic researchers either restrict their analyses to drugs with clean unit alignment or disclose the limitation in footnotes that few readers parse.
The consequences are not abstract. A formulary committee comparing the cost-effectiveness of two budesonide formulations using different pricing databases may reach opposite conclusions depending on which unit convention each source applies. An investor modeling the gross-to-net spread for a manufacturer’s portfolio may overstate or understate net revenue by millions if unit mismatches contaminate the calculation. A state Medicaid program setting reimbursement benchmarks against NADAC may inadvertently over- or underpay pharmacies for specific NDCs where the unit definition diverges from the one embedded in their adjudication system.
The problem persists because fixing it requires coordination across entities with misaligned incentives. Manufacturers benefit from unit ambiguity that obscures true cost comparisons. PBMs and payers maintain proprietary crosswalks as competitive assets. CMS, which could mandate a universal unit standard, has historically prioritized other regulatory battles. And the data vendors who profit from the complexity have little commercial motivation to eliminate it.
There is no clean resolution to propose. Normalizing to NCPDP billing units is a reasonable first step but not a sufficient one — it standardizes the denominator without addressing the clinical non-equivalence of different formulations. Course-of-treatment rollups add clinical context but require assumptions that may not generalize. Stoichiometric normalization offers precision at the molecular level while sacrificing relevance at the therapeutic level.
What remains is a system in which the most basic question — what does this drug cost? — cannot be answered without first specifying which database, which unit convention, which formulation, and which set of unstated assumptions the asker is willing to accept. The unit problem is not a bug in the system. It is the system.
