Key takeaways
- Coverage is not binary. The real decision is which subpopulation you cover, under what conditions, and for how long.
- Model net cost over three years, not one. Year-one spend is the least informative number you will look at.
- If you cannot describe your off-ramp, you have not designed a program — you have opened an entitlement.
The question is badly framed
Most benefits committees arrive at this decision framed as yes or no: do we cover GLP-1s for weight management, or do we exclude them? Framed that way, the answer is almost always driven by whichever number was presented most vividly — either a projected pharmacy spend increase that alarms finance, or an employee-relations argument that alarms HR.
Both framings are wrong because coverage is not binary. Every employer that covers GLP-1s covers them for *someone*, under *some* conditions, for *some* duration. The design space between blanket exclusion and unlimited coverage is where the entire outcome is decided, and it is where almost no committee spends its time.
A better opening question: which segment of our population would produce a defensible return on this therapy, and what would we need to be true to identify them reliably?
The three numbers that decide it
Eligible population share. Not the share of your census with a qualifying BMI — the share likely to actually enroll. These diverge enormously. Depending on demographics, communications, and friction, realistic first-year uptake among eligible members typically lands somewhere between 3% and 12%. A four-fold range on the single largest cost driver is why most first-year projections are wrong.
Net cost per member per month. Gross drug cost minus rebates, minus avoided spend, plus vendor fees. Buyers routinely model gross and get ambushed by the rebate timing, or model the vendor PMPM and forget the drug entirely. The drug is nearly always the dominant term.
Persistence at 12 and 24 months. This is the number that separates a program from a subsidy. Real-world discontinuation is high, and a member who stops at month five has generated close to full cost and close to zero durable benefit. Your program design should be judged primarily on whether it improves this number.
Why three-year modeling changes the answer
Year one is dominated by enrollment costs and produces almost no offsetting savings. Any analysis stopping at twelve months will make coverage look indefensible. That is not a finding; it is an artifact of the window.
Years two and three are where the model becomes informative: maintenance dosing at lower cost, some members transitioning off therapy entirely, and the first genuine avoided-cost signal in cardiovascular, orthopedic, and diabetes-related spend. Employers that model only year one systematically under-invest; employers that assume year-three savings arrive on schedule systematically over-invest.
Run three scenarios — conservative, expected, and high-uptake — and present all three. A committee shown a single point estimate will treat it as a forecast. A committee shown a range will ask better questions.
Design the off-ramp before you open the door
The single most common failure in employer GLP-1 programs is launching without a defined end state. Members start therapy, the pharmacy line grows, and there is no clinical or contractual mechanism for anyone to ever stop.
A credible off-ramp specifies: the clinical criteria that trigger a dose-reduction trial, who owns that decision, what behavioral and strength support wraps the taper, and what happens to coverage if a member declines. None of this is punitive — the evidence on lean-mass loss and weight regain makes a structured taper better medicine as well as better economics.
If your prospective vendor cannot describe their off-ramp protocol in specific terms, that is the most informative thing you will learn in the entire evaluation.
Put this to work
The RFP toolkit turns these guides into a scoring rubric and question bank you can send to vendors this week.