Comparing the cost of two drugs that treat the same condition sounds like a straightforward exercise. It is not. Drug A is a once-daily oral tablet priced at $15 per unit. Drug B is a biweekly injection priced at $3,200 per dose. Which costs more? The answer depends entirely on what you mean by “cost” — per unit, per day, per month, per course of treatment, per quality-adjusted life year — and each choice of denominator produces a different ranking. Course-of-treatment pricing exists to resolve this ambiguity by rolling unit costs up into a standardized measure of what a complete treatment regimen would cost. The concept is sound. The execution is where things get complicated.
The method requires three inputs: the drug’s unit price, the labeled dosing regimen, and the treatment duration. Multiply the daily dose by the number of treatment days, multiply by the per-unit cost, and you have a course-of-treatment price that can be compared across drugs, formulations, and delivery mechanisms. For a drug with a fixed dose, a clearly defined treatment duration, and a single indication, the calculation is mechanical and the result is informative.
Most drugs do not meet these criteria. Dosing varies by indication, patient weight, disease severity, and clinical response. Treatment duration varies by condition — some therapies are episodic, some are chronic, some are indefinite. Many drugs treat multiple conditions with different dosing schedules for each. The course-of-treatment methodology must impose uniformity on this variation, and the assumptions it uses to do so are where the analytical work — and the analytical risk — resides.
The convention that has emerged as an industry default, used by SSR Health and replicated by other analytics platforms, is to assume maximum labeled dosing and, where the label does not specify a treatment duration, a one-year default. These choices are defensible on narrow grounds: maximum labeled dosing represents the upper bound of approved use, and one year provides a standardized time horizon that enables cross-drug comparison. They are also, for many drugs, poor approximations of actual clinical practice.
Consider a corticosteroid like budesonide. For eosinophilic esophagitis, the typical treatment course is eight to twelve weeks of induction therapy, potentially followed by a lower maintenance dose. For Crohn’s disease, treatment duration varies from several weeks to several months, with taper schedules that reduce the dose over time. For COPD, maintenance therapy may be indefinite but at doses and formulations that differ entirely from the gastrointestinal indications. A one-year course-of-treatment price at maximum labeled dosing describes a scenario that few prescribers would recommend for any of these conditions. The number is standardized. It is not realistic.
The maximum-dosing assumption introduces a systematic upward bias. Most patients are not prescribed the maximum labeled dose. Dose titration, step-down protocols, and individualized dosing are standard practice across therapeutic areas. A course-of-treatment price calculated at maximum dose will overstate the cost experienced by the majority of patients, potentially by a significant margin. For drugs with wide dose ranges — oncology agents dosed by body surface area, for instance, or biologics with weight-based dosing — the gap between maximum labeled dose and median administered dose can be substantial.
The one-year duration assumption introduces a different kind of distortion. For chronic therapies where patients remain on treatment for years or decades, a one-year time horizon understates total treatment cost but may be useful for annual budgeting. For episodic therapies — antibiotics, short-course anti-inflammatories, acute pain management — a one-year duration massively overstates actual use. The assumption treats a two-week antibiotic course as if the patient took the drug for fifty-two weeks. This is obviously wrong, but it is wrong in a way that enables comparison with chronic therapies that would otherwise occupy a different analytical category entirely.
The result is a metric that is most accurate for drugs with chronic, stable, single-indication use — and least accurate for drugs with variable dosing, multiple indications, or episodic treatment patterns. This is not a random distribution of error. It is a systematic bias that overstates the cost of drugs used episodically and understates the cost of drugs used chronically relative to what patients actually experience. Any formulary decision, cost-effectiveness analysis, or policy recommendation that relies on course-of-treatment pricing without acknowledging this bias inherits the distortion.
More sophisticated approaches exist but are not widely adopted. Indication-specific dosing models weight the course-of-treatment price by the distribution of actual indications, using claims data to estimate how many patients use the drug for each approved condition and at what dose. This method is more accurate but requires granular utilization data that is expensive to obtain and difficult to maintain. It also requires assumptions about indication attribution — determining why a given prescription was written — that are themselves imprecise because diagnosis codes on pharmacy claims are often unreliable.
Real-world evidence studies using electronic health record data can produce empirical estimates of actual treatment duration and average daily dose, but these studies are time-consuming, population-specific, and not available for most drugs at the time pricing decisions are made. The practical reality is that course-of-treatment pricing is calculated using label assumptions because label assumptions are available, standardized, and cheap. The alternatives are better but slower, more expensive, and harder to maintain at scale.
The tension is inherent: comparability requires standardization, and standardization requires assumptions that sacrifice accuracy for individual drugs. The question is not whether to use course-of-treatment pricing but whether the users of the metric understand what the assumptions are, where they break down, and how much the resulting prices might differ from reality for the specific drugs they are analyzing. Transparency about assumptions is not the same as solving the underlying problem, but it is a necessary first step toward ensuring that the numbers are interpreted with the appropriate degree of caution.
The AEI analysis of SSR Health’s methodology makes this point implicitly: when assumptions about dosing and duration are opaque, the conclusions drawn from the resulting prices inherit an unquantified uncertainty. That uncertainty does not appear in the spreadsheet. It does not carry an error bar. It sits inside the number, invisible until someone asks what “one course of treatment” actually means — and discovers that the answer is, quite often, a convenient fiction.
