How do you actually compute cost per milligram across brand, compounded and research-grade routes?

Every comparison I find online is either a marketing page or someone dividing one number by another with no statement of what the numerator and denominator actually are. I would like to do this properly for my own decision-making, and I keep getting stuck on what belongs in the cost.

The specific problem: a branded pen carton, a compounded vial and a research-grade lyophilised vial are three different objects, and dividing price by label milligrams treats them as though they were the same object. I do not think that is valid, but I cannot articulate exactly why beyond a vague sense that a metered pen dose and a powder mass are not the same thing.

What I want is the actual method: what to put in the numerator, what to put in the denominator, and which costs people systematically forget. Worked numbers would help more than a principle. I am in the US, no current coverage for this class, and I understand research-grade material is not approved for human use and that this is arithmetic rather than a plan.

Adding the cost lines that get left out, since the accepted answer's numerator is more complete than most and still misses several. In rough order of how often they are forgotten:

• Failed and lost consignments. If one parcel in eight does not arrive, that is a 14% surcharge on everything, not an anecdote. Multiply your per-mg figure by 1/(1 − failure rate). At a 12% failure rate the $4.52 above becomes 4.52 / 0.88 = $5.14.
• Discarded material. A vial that fails your own inspection, a reconstitution you abandon, a lot the assay comes back wrong on. Real, and it lands entirely in the numerator with nothing added to the denominator.
• Assay cost per lot, not per order. Testing one vial tells you about one lot. Buying across three lots to save on unit price means three assays, or two-thirds of your material untested.
• Storage. Trivial in money, non-trivial if you count a dedicated small refrigerator, and non-trivial in risk if you count what happens when a household appliance fails unnoticed.
• Payment friction. Network fees, exchange spread, and the effective cost of a payment method with no recourse. On a small order the spread alone can exceed the shipping.
• Your time. Sourcing, reading COAs, reconstituting, sending samples for testing, and tracking parcels is several hours per cycle. Price it at any rate you like; at zero the comparison is not honest.
• Monitoring you should be paying for anyway. Baseline and periodic labs cost money on every route, and on the unsupervised route nobody has ordered them. That is a cost that has been avoided rather than saved.

Applying only the consignment-failure and per-lot-assay corrections typically moves the research-grade figure into the $5 to $7 per delivered mg band. Which is still by far the cheapest row in the table — the point is not that the ranking changes, it is that the true ratio to the cash brand channel is around ten to one rather than the sixty to one that a naive division suggests, and a ten-to-one ratio is a different decision than a sixty-to-one one.

A methodological correction to how the compounded rows are usually computed, because there is a systematic error in them that nobody catches.

Compounded vials are sold on a concentration and volume basis, not on a milligram basis, and the two do not always agree. A vial marketed as "10 mg" is frequently 2 mL of a 5 mg/mL solution, and the extractable volume is not 2.00 mL — there is fill tolerance, there is unrecoverable residue, and some pharmacies deliberately overfill to guarantee the labelled extractable amount while others fill to nominal and let you discover the difference.

So for a compounded vial the correct denominator is extractable volume times assayed concentration, and you have neither number unless you ask. What you can do is measure: draw the entire vial into a syringe and read the volume. If a "2 mL" vial yields 1.88 mL, your effective cost per mg is 2.00 / 1.88 = 1.064, i.e. 6.4% higher than labelled, before any dead space. If it yields 2.15 mL, the pharmacy overfilled and you are better off than the label says.

Do this once per pharmacy and you learn something durable about how they fill. It also tells you which of the two philosophies you are dealing with, and that generalises to everything else about their quality system.

Second point on the same rows: a subscription bundle's stated monthly cost usually assumes a dose. If the price is flat across doses, then the cost per milligram falls as you titrate up and the early months are the expensive ones — which means comparing a bundle against a per-vial price at maintenance dose flatters the bundle if you are at 0.5 mg and flatters the vial if you are at 2.4 mg. State the dose whenever you quote a monthly figure, or the number is not interpretable.

One framing point that the numbers can obscure: for a large fraction of people reading this, the highest-value hours are not spent optimising the cheap rows of that table but on getting into the $2.60 row.

The spread between covered brand and cash brand in the accepted answer's table is a factor of twenty. No amount of sourcing skill on the research route beats a factor of twenty, and the covered row also comes with a licensed product, a clinician, and monitoring included. The reason people skip it is that the covered row requires an administrative fight with an uncertain outcome, while the cheap row requires a card and an evening — the effort is more predictable even though the payoff is much smaller.

Concretely, before treating an unapproved route as the economic answer, it is worth establishing three facts: whether your plan document excludes the category outright or merely applies criteria, whether a manufacturer cash channel exists at a price you can survive, and whether any patient-assistance or state programme applies to you. Those three checks take a couple of hours and they occasionally change the answer by an order of magnitude. If they all come back negative, at least the arithmetic above is then being applied to a real decision rather than to a hypothetical one.