Your mileage may vary

I have been involved in no less than five GLP-1 conversations in the past week. Four were out in the community. Someone excited about how much weight they had lost, someone happy with how they look, someone else glad about how they feel. Often I'm also asked about insider secrets on how to buy it for less, since most employers don't cover it. The other two conversations were benefits meetings.

Overall, I cannot think of another drug class that has moved from the formulary meeting to the family gathering this fast. GLP-1s are different because the people taking them talk about it, and everyone sees it. The people who want the results talk about it, and the people paying for it want to talk about it too.

There's a problem. For all the data, advertising, and conversation surrounding GLP-1s, every number published about whether this coverage pays off belongs to somebody else's plan. Potentially worse, the range between those numbers is now wide enough that citing any one of them tells you almost nothing.

Remember, there are two classifications of GLP-1 utilization. One for diabetics, another for weight loss. This conversation centers on the latter, and the distinction matters more than it sounds. Diabetes coverage is close to universal and generally is not argued about at renewal. Weight management is the one that lands on the agenda, and everything below applies only to it.

Let's break this down, before you run to the calculator to test your numbers and see your math. Emphasis on your.

The two languages in the room

Sit in on the conversation and you will hear two people who are not speaking the same language, and generally do not notice.

HR goes first. "We have people asking. Some of them are asking twice. One of them cried in my office. They're saying their doctors are recommending it." Then the part that matters most and gets heard least: "If we say no, they might leave."

The CFO answers in a different register entirely. "What is the impact?" "What does it do to the stop loss?" "What does that do to my renewal?"

Both of those people are being responsible. Neither of them is wrong. They are just measuring different things, and the meeting generally ends with the CFO's numbers on a page and HR's argument in the room, which is not a fair fight even when the CFO is trying to be fair.

Here is why that matters. The CFO's question can be modeled and the HR question cannot, so the CFO's question wins by default. Not because it is the better question. Because it carries known risks. HR's risk, retention, is unknown. Will people REALLY leave? Or will they be resourceful?

What everybody already agrees on

The drug is expensive. Nobody argues about this, including the people taking it, who tend to be the most price aware people in the conversation because many of them have priced the cash option themselves.

Employer guidance commonly models $600 to $900 per member per month net.¹ The emphasis is on net cost, which is after rebates. The total cost, without rebates, is much higher. This is an important distinction, and the calculator supports it.

A carrier book analysis put the annual increase in pharmacy cost per treated member at roughly $6,540.² The International Foundation of Employee Benefit Plans found GLP-1s for weight loss accounted for 10.5% of total annual claims in 2025, up from 8.9% in 2024 and 6.9% in 2023.³ Ninety one percent of large employers say they are worried about the long term cost.⁴

Coverage is spreading anyway, at least for now, and in pockets. KFF put it at 19% of employers with 200 or more workers and 43% of those with 5,000 or more.⁵ So the answer to "why doesn't my plan cover this" is often just headcount, which is a deeply unsatisfying thing to tell somebody at a cookout.

But that question deserves better than a shrug, because the real answer is buried in three places at once. It is buried in finance, where a plan under 500 lives can have its entire renewal moved by a handful of treated members. It is buried in philosophy, where some sponsors believe a plan exists to insure risk and others believe it exists to buy health, and those two beliefs produce different formularies. And it is buried in timing, because the plan that pays for the therapy is frequently not the plan that would ever see potential benefit.

The retention argument, and how to test it

The most common argument I hear for offering it has nothing to do with claims. It is retention. NFP's 2026 benefits trend report found 29% of employees would switch employers to access a GLP-1 benefit.⁶

That is a real finding and I would not wave it off. It is also the argument that gets made loudest by the people who have already decided, which is generally a sign to slow down. Twenty-nine percent say they would switch. Saying and doing are different, and a benefit that attracts someone is not automatically a benefit that keeps them.

Given that, the honest version of the retention argument is testable, and the test is uncomfortable. If retention is the reason, then the value of the coverage rises with how long people stay. Run your own turnover through it. At 15% annual turnover, a bit more than half your treated population is still with you in three years. At 25%, it is closer to a third.

In other words, the retention argument works best at employers who already have good retention, which are the employers who need it least, at least for this purpose.

Longevity, adherence, and the thing nobody wants to say

Here is the part that gets skipped.

The benefit requires continuous therapy, or at a minimum lifestyle change alongside it. Real world persistence in commercial claims has run between roughly a third and two thirds at one year depending on when the cohort started, improving as shortages eased.⁷ Prime Therapeutics found persistence around 15% at two years and 8.1% at three.⁸ Randomized withdrawal trials show rapid weight regain after discontinuation, and the literature increasingly frames that as disease recurrence rather than treatment failure.⁹

If you add the lifestyle change component, and you cover it, what is the cost? Do you make additional benefits available, such as gym memberships, counseling, and coaching? Many would argue that without it, continuous therapy stops being a plan design choice and becomes the only outcome available.

Let's face it. The three year persistence data we have covers people who started during a shortage, when access itself was the reason many of them stopped. Anyone telling you they know the ten year cost curve on GLP-1s is telling you about a forecast, not a measurement.

What the drug might actually cost

This is where the ground is moving fastest, and where employer modeling is most out of date.

The cash market has repriced. Lilly sells Zepbound direct at $299 to $449 a month depending on dose. Novo lists Wegovy self pay from $199, and oral semaglutide has been quoted as low as $149 for a starter dose. TrumpRx sits around $350.¹⁰ Semaglutide's negotiated Medicare price for 2027 is roughly $274 per 30 day supply, about 71% below its 2024 list.¹¹

Hold those numbers next to the $600 to $900 per member per month net cost, after rebates, that employer guidance still models.

That gap is not anybody's misconduct. It is what happens when a product is sold through several channels at once and the channels price differently, which is ordinary in pharmacy and always has been. But it does mean the most expensive way for a covered employee to obtain this drug may currently be through the plan that covers it, and that is a procurement fact rather than a clinical one.

This is also the part that never shows up in a plan design comparison. A benchmarking spreadsheet lays out deductibles, coinsurance, copay tiers, and out of pocket maximums side by side, and every one of those numbers is real and worth having. Not one of them tells you whether your plan captures a rebate on a drug class now running at better than 10% of claims, or what net price you are actually paying for it. I made this argument at more length in From snout to tail. The cost shares are the front of the store, they are what everyone compares, and they mostly decide who finances the spend rather than how much of it there is. The net versus list question is a tail cost wearing a pharmacy invoice, and it does not fit in the column headings.

Note what is not coming, and when it might. Semaglutide and tirzepatide have US patent protections that run into the early 2030s.¹² Whatever relief arrives before then arrives through contracting and procurement. I am discounting the compounded market, which is under heavy scrutiny since the shortages were declared resolved.

There is a real counterpoint here, and it is worth holding. Once those patents lapse, ordinary generics follow, and generic entry has historically taken prices down more than 75% in other categories.¹⁷ If that happens here, the arithmetic in this piece stops being a problem to solve and becomes a waiting game.

The difficulty is that a plan sponsor decides in the fall for the following January. A price cliff in 2031 does not help a renewal you are pricing now, and every year you cover it between now and then is paid at today's number, not at the one you are waiting for.

The question that moves the answer more than the drug does

Before any of the arithmetic matters, there is a question that moves the outcome further than persistence, turnover, and trend combined. Does your plan actually capture the rebate?

If you are fully insured, generally you do not. The rebate exists and it is negotiated, it just settles somewhere upstream of your renewal, and what reaches you is a rate rather than a rebate. If you are self funded, it depends entirely on how the pharmacy contract is written, which definitions it uses, what is excluded, and when the money actually lands. Some plans see most of it. Some see a portion net of fees. Some see it a year later, which is its own problem when you are pricing a renewal now.

Here is what that single question does to the same 500 employee plan.¹⁶

Look at what moves and what does not.

The target price does not move. It is $2,438 either way, because that number falls out of your plan spend, your expected reduction, and how long people stay, and none of those care what you are currently paying. What changes is the distance to it.

The other two doors move a great deal, and both move the wrong way. The reduction you would need doubles, from a figure that is already above anything the published evidence supports to one that is roughly double that. The number of people you could treat halves, from eleven to six out of roughly 700 covered adults, which is no longer a coverage policy so much as an exception process.

In other words, a plan that does not capture rebates is not facing a slightly worse version of this decision. It is facing one where two of the three doors are effectively shut and only procurement remains open.

That is worth knowing before the meeting rather than during it, and it is knowable. Ask where the rebate goes, ask what it is calculated on, and ask when it arrives. Those three answers change the arithmetic more than any assumption you will argue about for the next hour.

Count the adults, not the employees

One more thing before the arithmetic, because it is the mistake I see most often and it runs in the expensive direction.

A 500 employee group is not a 500 life plan. Depending on your enrollment mix it is commonly 1,000 or more covered lives, of which perhaps 700 are adults. Spouses enroll, spouses have BMIs, and spouses fill prescriptions. Children largely do not qualify here, since the weight management labels are adult indications with narrow adolescent exceptions.

The fix is a single multiplier a CFO can pull off an enrollment report in about a minute.¹⁶ Take employee headcount, multiply by 1 plus the share of enrolled employees who cover a spouse or adult partner, and use that as your adult population. Cover a spouse on 40% of contracts and you multiply by 1.40. Everything downstream follows from that one number, and it moves the answer more than most of the assumptions people spend the meeting arguing about.

Model this off employee headcount alone and you will understate your treated population by roughly 40%, which understates cost by half and flatters every number that follows. The calculator asks for employees and then asks how many adults you cover per employee, so the two populations stay separate. Your enrollment census beats any assumption on that one.

The arithmetic that actually decides it

So here is the question I would put in front of the CFO and HR both, because it is the only one they can answer together.

Not "does this work." Not "will people stay on it." Not "will trend come down enough." Those are all unanswerable at the plan level right now, and pretending otherwise is how these meetings go sideways.

The answerable question is: at what price does this stop being an argument?

Take a plan with 500 covered employees and, at a 1.40 multiplier, roughly 700 covered adults. Take 40% of those adults as clinically eligible under the FDA label, which sets weight management eligibility at a BMI of 30 or above, or 27 with a weight related condition. Take 12% of the eligible group as starting therapy. That is 34 people. At $7,200 a year net, the plan spends $1,209,600 over five years. Assume the therapy reduces total plan spend by 2%, credit that benefit only to people who are still employed and still on therapy a year later, and the plan keeps about $410,000 of it.

Run it backwards and the number that falls out is $2,438 per member per year. Roughly $203 a month.

That is the price at which this stops being a debate and becomes a purchase. And it sits inside the range the cash market is already quoting.

The theory

Every published analysis of this question produces a break even price, and they do not agree. One puts it at $7 a month for a broad population and $177 under ideal conditions.¹³ Actuarial modeling has GLP-1s falling well short of standard cost effectiveness benchmarks at $700 to $800.¹⁴ A sell side analysis argues the opposite, that net prices of $7,500 to $9,000 a year already sit below a value based price of $12,202 to $16,765, meaning manufacturers are capturing roughly half the value they create.¹⁵ My arithmetic above lands at $203.

The simple argument over net cost and list price ecosystems shows how challenging comparing any report to your own plan could be.

Those numbers span two orders of magnitude. They are not in conflict because somebody made an error. They differ because each one answers a different question, over a different horizon, with a different definition of benefit, for a different population.

They are nothing more than EPA mileage estimates: useful, standardized, produced in good faith, and almost never what you get. The sticker says 32 highway and 24 city. You get 19 because you live where you live and drive how you drive.

Every one of those published figures assumes a workforce. Yours has an actual one, with an actual turnover rate, an actual age curve, an actual eligible share, an actual dependent mix, and an actual contract with an actual net price on it. It is a theory because you need to look at your data.

So I built the calculator. Put in your headcount, your adults per employee multiplier, your eligibility, your turnover, your persistence if you have it, your plan spend, and whatever reduction you are willing to defend. It returns the price you would have to source at for the arithmetic to close, and two other doors if procurement is not the one you can move.

That number is yours. It is not transferable, and neither is anybody else's.

Where this actually lands

Every employer carries benefits as a load on labor. You may call it fringe, your controller may call it burden, and the number may live in a different place on your statements, but the mechanic is the same. You take what the plan costs, you spread it across the people it covers, and that becomes part of what an employee costs to employ. It shows up in headcount planning, in what you can afford to pay someone, and in what you have to charge for their time.

That is what makes this different from most pharmacy decisions. It is not a line item inside drug spend. It is a change to the cost of labor, and other decisions are built on top of that number.

How much that matters depends on how locked in your rate is. A firm that reprices annually can absorb a surprise and adjust at renewal. A nonprofit working under a negotiated rate, a professional services firm quoting loaded hourly rates, a manufacturer costing jobs off a burden rate, and a government contractor bidding a wrap rate for a performance period years out are carrying the same decision with progressively less room to change their mind.

The contractor case is the sharpest version, where a coverage decision made this fall gets priced into work performed in 2029 and you do not get to reprice because utilization ran hotter than you modeled.