Cost per month is the wrong denominator if a third of the people paying it never reach the thing they are paying for. A UK analysis reworked the arithmetic, dividing annual pharmacy cost by the proportion of patients hitting a composite target to produce a cost per patient treated to target [1]. It is the same instinct behind asking what a GLP-1 is worth against what it does, applied to a head-to-head rather than to a single drug.
The targets combined glycemic control at two thresholds with weight loss at 5%, 10% or 15%, and required avoiding hypoglycemia. Response rates came from a post-hoc cut of the SURPASS-2 trial (n = 1,845), and costs came from the prices NICE appraised in TA924. At the strictest combination, HbA1c at or below 6.5% with weight loss of at least 15% and no hypoglycemia, tirzepatide cost £5,650, £8,665 and £9,462 less per patient reaching it at the 5, 10 and 15 mg doses respectively.
The pattern inside that result is more interesting than the headline. Differences grew as targets got stricter, which is what happens when one drug produces more large responses: the denominator shrinks faster for the weaker performer. And the pattern reversed at the loosest target, where HbA1c below 7.0% with 5% weight loss and no hypoglycemia was cheaper per responder on semaglutide. A reader whose goal is modest is being pointed at the other drug by this analysis.
Two further limits are worth holding. These are UK pharmacy costs in pounds at appraised NHS prices, which are not US prices and are a long way from the cash prices this desk collects, so nothing here converts into what a reader will be quoted. And a post-hoc analysis of one manufacturer's trial, reported as favorable to that manufacturer's drug, is a category of evidence to read at arm's length regardless of whether the arithmetic checks out.
The underlying dose problem recurs often enough that this site tracks it separately in comparisons run at mismatched doses. For the longer-horizon version of this question see the lifetime cost model, and for the standard health-economics framing that regulators actually use, see cost per quality-adjusted life year.