GLP-1s Are Rewriting Employer Trend Assumptions for 2027

Mercer projects average employer health benefit cost will exceed $18,500 per employee in 2026, a 6.7% jump it attributes partly to GLP-1-driven pharmacy growth (Mercer, November 2025). Aon's competing data point, drawn from 192,000 GLP-1 users, shows sustained use lowers medical cost growth by 6 percentage points at 30 months, but only for members with Type 2 diabetes (Aon, January 2026). Both figures are correct. They describe different populations, which is exactly the problem.

Having watched three consecutive employer survey cycles list GLP-1s as a top medical trend driver while the offset data lagged years behind, the real actuarial story by mid-2026 is not the drug class itself. It is the gap between what a defensible trend assumption requires and what the data actually supports. Mercer's 2026 National Survey of Employer-Sponsored Health Plans found prescription drug spending rose 9.4% among large employers, with 49% of those employers covering GLP-1s for weight loss, up from 44% in 2024 (Mercer, November 2025). Aon's January 2026 release, drawing on medical and pharmacy claims from more than 50 million commercial lives between July 2022 and March 2025, is the first large-sample data set with enough follow-up to say something about the offset side of the ledger, not just the drug cost side. Laura Rissover, Aon's VP of Health Analytics, framed the contribution directly: "By analyzing longitudinal medical and pharmacy claims for hundreds of thousands of GLP-1 users, we are able to observe real-world patterns and track how total medical costs evolve over time" (Aon, January 2026).

Two Curves Pointing in Opposite Directions

The near-term cost curve and the multi-year offset curve are not describing the same timeframe, and conflating them is how a trend assumption goes wrong. EBRI's October 2025 simulation, built on MarketScan claims data across a range of eligibility, adherence, and cost-sharing scenarios, put the first-year premium impact of expanding GLP-1 coverage at 5.3% to 13.8% annually, with the wider studies describing the range loosely as "6% to nearly 14%" depending on how broadly a plan defines eligibility and how consistently members stay on therapy (EBRI, October 2025). That is a same-year cost, showing up in the pharmacy line before any offset has had time to accrue.

Aon's offset curve runs on a different clock. Members with Type 2 diabetes who sustained GLP-1 use saw medical cost growth run 6 percentage points below non-users at 30 months, widening to 9 points for those with at least 80% adherence. For members using GLP-1s for weight loss without a diabetes diagnosis, the offset was smaller and slower: 3 points at 18 months, improving to 7 points with consistent use (Aon, January 2026). Thirty months is longer than the average employee tenure at a given firm, roughly four years including turnover and plan switching, which HRP's own cost-benefit modeling flags as the practical reason employers rarely capture the full offset before a member leaves the plan (HRP, 2026). A carrier or plan sponsor pricing a single renewal cycle is, by construction, looking at the front half of a curve whose payoff arrives on someone else's book.

That timing mismatch is why the same drug class can be simultaneously a 6.7% cost driver in Mercer's survey and a cost-reducing intervention in Aon's claims study. Neither number is wrong. The trend assumption an actuary builds depends on which slice of that curve the pricing period actually falls on, and for a 12-month group renewal, it is almost always the expensive slice.

Why Stop-Loss and Self-Insured Trend Assumptions Diverge for the Same Drug

The mechanism gets sharper once you separate how a stop-loss actuary treats GLP-1 claims from how a self-insured plan's own medical trend assumption treats the same drug class, because the two are pricing different tails of the same distribution. A self-insured plan's aggregate medical and pharmacy trend assumption is a population-average number: total expected claims across every enrolled life, smoothed by the law of large numbers across a group of hundreds or thousands. GLP-1 utilization enters that trend as an incremental PMPM add-on, typically modeled off pharmacy claims data showing GLP-1 costs climbing from roughly $1.50 PMPM in 2019 to $27.23 PMPM by the first quarter of 2025 across all covered lives, an 18-fold increase over six years (Willis Towers Watson Rx Collaborative, cited in actuary.info's stop-loss GLP-1 renewal analysis, 2026). That number gets built into the group's overall trend factor and spread across every premium dollar.

A stop-loss actuary is not pricing the average member. Specific stop-loss reimburses the tail: the small number of individual claimants who breach a specific deductible, commonly $100,000 to $250,000 for a mid-market group. GLP-1 claims rarely breach that threshold on their own, since net annual cost per patient after PBM rebates runs $4,800 to $8,400 depending on contract terms (industry pharmacy benefit data, cited in the same stop-loss analysis). What matters for specific-claims trend is not the PMPM average but the co-occurrence rate: how often a GLP-1 claimant also carries a comorbid diagnosis, cardiovascular monitoring, bariatric-adjacent care, or a second high-cost condition that pushes total individual claims over the attachment point. That is a frequency-of-breach question, not an average-cost question, and it requires claimant-level severity modeling that a PMPM trend factor cannot substitute for.

This is also where the site's own coverage understated the driver at the time it was written. The stop-loss carve-out and laser analysis published in May 2026 documented carriers responding to GLP-1 cost pressure with carve-outs, lasers, and raised attachment points, and cited the same Willis Towers Watson PMPM escalation used above, but it predates Aon's January 2026 offset finding entering wider trade circulation and had no visibility into how a diabetic-comorbid subpopulation's total cost trajectory might bend the specific-claims curve over a multi-year block. A stop-loss actuary setting a specific-claims trend assumption purely off rising GLP-1 utilization, without adjusting for the offset among the diabetic-comorbid subset most likely to breach an attachment point, risks overstating the tail-severity trend for exactly the claimants where the drug is most likely to prevent a cardiovascular or renal event that would otherwise generate the catastrophic claim.

The Utilization Math: How a Few Points Move the Renewal Number

The dollar impact of a GLP-1 trend assumption is unusually sensitive to utilization rate, more so than most drug classes, because the per-user net cost is high enough that small shifts in the denominator move real money. Using HRP's benchmark of $6,540 in average annual drug cost per GLP-1 user (HRP, 2026), a 100-life group at 5% utilization, five members on therapy, adds approximately $32,700 a year in gross pharmacy spend before any rebate netting or medical offset. A 150-life group moving between 4% and 6% utilization, six to nine members, adds a range of roughly $39,000 to $58,900 annually. That is a swing of nearly $20,000 driven by two percentage points of utilization on a group of 150 lives.

Group size	Utilization rate	Members on GLP-1	Approx. annual gross pharmacy add
100 lives	5%	5	~$32,700
150 lives	4%	6	~$39,200
150 lives	5%	7.5	~$49,050
150 lives	6%	9	~$58,900

Calculated at $6,540 per member per year (HRP, 2026); actual net cost varies with PBM rebate structure and member adherence.

Prior-authorization design is the lever that moves the utilization rate, and it moves it more than most benefit design changes because GLP-1 demand is uncorrelated with the plan's other utilization management tools. A plan with loose prior authorization, self-attestation of BMI, no step therapy requirement, no requirement to document a failed trial of lifestyle intervention, will see utilization climb toward the eligible population's full clinical prevalence, which HRP estimates at over 40% of privately insured adults meeting some clinical eligibility criterion (HRP, 2026). A plan requiring documented BMI thresholds, a failed six-month lifestyle intervention, and annual re-authorization tied to a minimum weight-loss percentage can hold utilization several points lower. Given the table above, the difference between a loosely administered prior authorization program and a tightly administered one is not a rounding error in the renewal; it is the difference between a $39,000 and a $59,000 pharmacy add on a 150-life group, and that gap compounds across every group in a book of business.

The Pooling Error: Diabetics Are Not the Weight-Management Population

The most consequential error a plan actuary can make with this data is pooling the diabetic and weight-management-only populations into a single trend assumption, because Aon's own results show those two cohorts do not behave the same way on either side of the ledger. The diabetic cohort's 6-to-9-point medical cost growth reduction at 30 months (9 points with high adherence) reflects a population with existing cardiometabolic risk, where GLP-1 therapy is displacing hospitalizations, ER visits, and complications that were already priced into that member's expected claims. The weight-management-only cohort's offset is smaller and slower to emerge, 3 points at 18 months, 7 points with consistent use, because that population's baseline medical cost trajectory did not carry the same near-term acute-event risk that the drug is displacing (Aon, January 2026).

The two populations also differ sharply on persistence, which determines whether either offset is ever realized. Nearly half of GLP-1 users discontinue treatment within 12 months, and multiple analyses find that discontinuation concentrates disproportionately among members using the drug for weight loss alone rather than for a diagnosed metabolic condition (HRP, 2026). A member who stops therapy at month nine never reaches the 18-month mark where even the smaller weight-management offset begins to show up. Pooling both populations into a single "GLP-1 members" bucket and applying a blended trend assumption systematically overstates the expected offset for the larger, faster-churning weight-management-only group and understates the true expected offset for the smaller, more persistent diabetic-comorbid group. A plan with 70% of its GLP-1 population in the weight-management-only category, roughly consistent with the ratio of total eligible adults to those with a diabetes diagnosis, is pricing a trend assumption that reflects almost none of the group actually driving the favorable Aon numbers.

The actuarial fix is not complicated in concept, though it requires data most plans do not currently capture cleanly: segment the trend assumption by indication code at the point of prescription, not by drug name, and apply the diabetic offset curve only to the diagnosed diabetic-comorbid subpopulation. Plans and stop-loss carriers that cannot segment claims this way, because prior authorization forms do not reliably capture ICD-10 indication or because pharmacy claims data lacks a clean crosswalk to medical diagnosis codes, are stuck defaulting to a blended assumption that is defensible only in the sense that better data does not yet exist, not in the sense that it produces an accurate trend.

Why a 2026-Vintage Assumption May Not Survive One Renewal Cycle

PBM rebate renegotiation and new market entrants are the two forces most likely to invalidate a trend assumption built on 2026 data before the 2027 renewal is even finalized. Net cost per patient already varies by more than 2x across carriers purely on PBM contract terms and formulary tier placement, from roughly $4,800 to $8,400 annually for the same molecules (industry pharmacy data). A single PBM renegotiation, or a formulary tier change moving a competing molecule to preferred status, can shift that per-user cost enough to change the utilization-sensitivity table above by a meaningful margin without any change in clinical utilization at all.

The supply side is moving even faster. Oral Wegovy launched in January 2026, removing the injection barrier that had constrained uptake among members reluctant to self-inject, and actuary.info's pipeline tracker counts nine additional obesity-drug candidates approaching launch, with Morningstar projecting a roughly 28% price decline as competition breaks the Novo Nordisk-Eli Lilly duopoly that has priced the category since 2023 (see actuary.info's GLP-1 pipeline and price-cliff analysis, 2026). A trend assumption built on 2026 net pricing, before that price decline materializes, will overstate the 2027 pharmacy cost line if competitive entry proceeds on schedule, and understate it if regulatory or manufacturing delays push the price decline into 2028. Either way, a single-point trend factor carried forward from 2026 vintage data without an explicit re-pricing checkpoint is a methodology error waiting to surface at the next renewal.

What This Means for 2027 ACA and Group Filings

The same pooling and timing problems that complicate a self-insured employer's trend assumption apply with less data and more regulatory scrutiny in the ACA individual and small-group market. Actuary.info's earlier analysis of the 2027 ACA credibility gap found that mid-size carriers with 80,000 individual-market lives typically have only 12 to 18 months of GLP-1-specific claims experience at meaningful scale, not enough to establish a stable trend rate under conventional credibility standards, and that net cost variance alone can shift the implied PMPM trend contribution by 3 to 6 percentage points depending on PBM terms (see actuary.info's ACA 2027 GLP-1 credibility gap analysis). Layering the diabetic-versus-weight-management pooling error onto that already-thin credibility base compounds the problem: an ACA risk pool with a different diabetic-to-weight-management mix than the employer group data underlying published benchmarks will not experience the same blended offset, and a carrier that imports an employer-market trend assumption wholesale into an individual-market filing is importing someone else's population mix along with it.

That risk-pool mismatch sits alongside the broader morbidity assumption questions already reshaping 2027 ACA rate development, documented in actuary.info's coverage of Wakely's 2027 morbidity data, and it compounds the medical trend benchmark PwC has already set for the group market. PwC's 2027 projection puts group medical cost trend near 9%, with GLP-1s cited as a named contributor alongside provider consolidation and site-of-care shifts (see actuary.info's PwC 2027 medical cost trend analysis). An actuary building a 2027 group filing off that 9% benchmark, without decomposing how much of it is a blended GLP-1 assumption versus other trend drivers, cannot tell whether their own book's diabetic-to-weight-management mix should push their filed trend above or below the industry number.

Why This Matters

The actuarial stakes here are not academic. A trend assumption that pools two populations with materially different offset curves and different persistence rates will misprice the specific group it is applied to in a predictable direction: it will understate cost for groups skewed toward weight-management-only utilization and overstate it for groups with a higher diabetic-comorbid concentration, since the blended assumption sits between the two true curves rather than at either one. For a self-insured plan sponsor, that means a renewal quote that is wrong in a knowable direction the moment the group's own claims mix is known. For a stop-loss carrier, it means specific-claims trend assumptions that misjudge exactly the claimants most likely to breach an attachment point. For an ACA carrier, it means importing an assumption built on someone else's risk pool into a filing that will be tested against the carrier's own emerging experience within a single plan year.

None of this argues for abandoning GLP-1 trend loading, which the cost data plainly requires. It argues for building that loading as two assumptions instead of one: a near-term utilization-driven cost curve that applies broadly, and a multi-year offset curve that applies narrowly, only to the diagnosed diabetic-comorbid subpopulation with the persistence to reach it. Carriers and plan actuaries who can segment their claims data by indication and track persistence separately from utilization will price 2027 renewals closer to the true expected cost of the population in front of them. Those who cannot will keep carrying forward a blended number that happens to be defensible only because the alternative, disaggregating a fast-moving drug class by indication and adherence, is harder to build.