Aon's 50-Million-Person GLP-1 Dataset Prices the Employer Coverage Gap

Aon's analysis of more than 50 million commercial lives found GLP-1-treated Type 2 diabetics had 30-month medical cost growth 6 percentage points lower than untreated peers, widening to 9 points among members with 80% or greater adherence (Aon, January 2026). Yet IFEBP's 2026 survey shows employer coverage of GLP-1s for weight loss frozen at 36%, unchanged from 2025, leaving a quantifiable gap most plan actuaries have not yet priced.

The Aon Dataset: What "6 Points Lower" Means at the Group Level

Aon's second-phase release expanded its original May 2025 analysis to 192,000 GLP-1 users drawn from a de-identified pool of more than 50 million commercial medical and pharmacy claims running from July 2022 through March 2025 (Aon, January 2026). The scale matters less for its own sake than for what it buys: enough follow-up time to separate two populations that plan actuaries have mostly priced as one. Members with a Type 2 diabetes diagnosis on sustained GLP-1 therapy showed medical cost growth 6 percentage points lower than untreated diabetic peers at 30 months, widening to 9 points among those with 80% or greater adherence. Members using GLP-1s for weight loss without a diabetes diagnosis showed a smaller, slower-emerging offset: 3 points lower at 18 months, improving to 7 points with consistent use. Laura Rissover, Aon's vice president of health analytics, framed the contribution directly: "By analyzing longitudinal claims for hundreds of thousands of users, we observe real-world patterns and track medical cost evolution" (Aon, January 2026).

Those percentage points describe a growth rate differential, not an absolute dollar figure, which is exactly where most plan actuaries lose the thread. The American Diabetes Association puts the average annual excess medical expenditure for a diagnosed diabetic under 65 at $7,482, versus a demographically matched member without diabetes, with medical spending for a diabetic running 2.6 times what would otherwise be expected (ADA, 2024, analyzing 2022 claims data). Apply a representative 8% annual group medical trend to that excess-cost base and it compounds to roughly 21.7% growth over a 30-month window, an increase of about $1,623 per untreated diabetic member. A member on GLP-1 therapy with 80%-plus adherence, with growth running 9 points slower, would see that excess base climb closer to 12.7%, or about $952, a difference of roughly $671 per treated member over the same 30 months. At the more typical 6-point real-world offset, the avoided growth narrows to about $450 per member. Neither figure is Aon's own number; both are this publication's construction, applying Aon's percentage-point offset to the ADA's published excess-cost baseline, and every plan actuary building this model should substitute their own group's diabetic claims experience rather than borrowing this one.

Why the Near-Term Math Still Looks Expensive

Scale that per-member offset to a group and the near-term picture stays unfavorable, which is precisely why coverage decisions keep defaulting to no. Take a 5,000-life employer with a typical 10% diabetes prevalence, 500 diabetic members, of whom perhaps 30% take up GLP-1 therapy at high adherence, 150 members. The avoided cost-growth from the calculation above comes to roughly $100,650 over 30 months, about $40,000 a year, across that treated cohort. Net drug cost for those same 150 members runs $617 to $766 per month per person after manufacturer rebates (ICER, cited in EBRI's October 2025 simulation), or $1.1 million to $1.4 million annually. On a cost-growth-differential basis alone, current GLP-1 pricing outruns the diabetic offset by roughly an order of magnitude inside a 30-month window, which is the honest reason a one-year renewal actuary keeps landing on "expensive" even while reading Aon's favorable topline.

The cost-growth differential is not the whole offset, though, and this is the piece most trend memos skip. Aon's diabetic cohort also showed major adverse cardiovascular event reductions of 47% among women and 26% among men at high adherence, plus roughly 50% lower ovarian cancer incidence and 14% lower breast cancer incidence among female users (Aon, January 2026). A single avoided cardiovascular hospitalization carries a mean initial cost of $19,642 (Journal of Managed Care & Specialty Pharmacy, cited in a widely referenced commercial-claims cost study), a figure roughly 12 to 29 times the $671-to-$1,623 range of avoided cost growth calculated per treated diabetic member above. A group actuary who prices only the growth-rate differential and ignores avoided acute events is missing the larger, lumpier half of the offset, the half that shows up as a small number of very expensive claims that simply do not happen.

Building a Two-Cohort Morbidity Model

The framework a plan actuary needs is not a single blended GLP-1 trend load. It is two explicit cohorts, priced on two different clocks, with the near-term drug cost and the multi-year offset kept in separate columns until an explicit crossover is identified.

The gap in the second row is itself an actuarial finding worth stating plainly: there is no ADA-equivalent, publicly benchmarked excess-cost figure for obesity absent a diabetes diagnosis, which is a meaningful part of why most plan actuaries default to pricing the entire GLP-1 population as a single undifferentiated cost line rather than building the cohort split Aon's own data supports. A plan sponsor's own claims history, segmented by whether a GLP-1 user carries a diabetes diagnosis at initiation, is the only reliable substitute until a comparable public baseline exists.

Running the diabetic-comorbid cohort's dollars forward at the group level, using the $40,000-a-year avoided cost-growth figure from above, plus a conservative allowance for avoided acute events at a low single-digit annual incidence rate applied to the 150-member treated cohort, the crossover where cumulative avoided claims begin to approach cumulative drug spend typically lands in year three to year four of continuous, high-adherence therapy, not year one. That timeline sits close to typical employee tenure once turnover and plan switching are factored in, which is exactly why a plan sponsor pricing a single 12-month renewal cycle almost never captures the full inversion, and why the offset shows up more reliably in a carrier's book-level experience than in any single small or mid-size group's year-over-year trend.

Why 36% Has Not Moved

IFEBP's 2026 Pulse Survey, released in July and covering roughly 300 employer health plans among the organization's more than 33,000 member institutions, found combined diabetes-and-weight-loss GLP-1 coverage held at 36% for a second straight year, up from 34% in 2024 and 26% in 2023 (IFEBP, July 2026). Diabetes-only coverage climbed to 60% in 2026 from 55% in 2025, and 45% of plans now cover GLP-1s for other FDA-approved indications such as obstructive sleep apnea and cardiovascular disease. The class's cost footprint kept rising even as the weight-loss coverage line stalled: GLP-1 drugs represented 11.4% of annual claims in 2026, up from 6.9% in 2023 (IFEBP, July 2026). Carey Wooton, CEBS, IFEBP's associate vice president of education, summarized the posture bluntly: "Most employers are not covering GLP-1 drugs for weight loss but instead focusing on how to support the overall health of their workers" (IFEBP, July 2026).

Employers that decline full coverage are not doing nothing; they are substituting benefit-design workarounds for the coverage decision itself. IFEBP found 83% of employers excluding weight-loss GLP-1 coverage rely on medical-or-pharmacy plan carve-outs, 27% steer members toward direct-to-consumer cash platforms, and 21% promote FSA, HSA, or integrated HRA funding to shift the cost off the plan's own claims line (IFEBP, July 2026). Separately, 74% offer disease or case management, 61% provide nutritional counseling, and 61% cover bariatric surgery, cheaper and more familiar levers than an open-ended pharmacy benefit. These substitutions are a rational response to three constraints that Aon's dataset does nothing to resolve on its own: a one-year renewal accounting cycle that cannot credit a savings curve that pays off in year three or four; adverse selection risk, since a plan opening broad GLP-1 coverage draws disproportionately from members already carrying the highest morbidity; and the absence of any standardized actuarial certification or filed methodology for a GLP-1 coverage decision comparable to what governs a rate filing. Until one of those three constraints changes, a favorable longitudinal dataset does not by itself move a coverage rate that is set against a one-year lens.

Oral Pills Arrive at the Same Price, a Different Efficacy Curve

Two oral entrants complicate the adherence assumptions embedded in any GLP-1 morbidity model. Novo Nordisk's once-daily oral semaglutide tablet, approved December 22, 2025 and broadly available in the U.S. by early January 2026, produced 16.6% mean weight loss in the OASIS 4 trial among adherent patients, comparable to injectable Wegovy at the 2.4 mg dose. Eli Lilly's oral entrant, branded Foundayo, is a distinct non-peptide GLP-1 receptor agonist rather than an oral form of tirzepatide; its sales force activated April 17, 2026, and the drug had accumulated roughly 89,000 prescriptions with coverage confirmed at two of three major pharmacy benefit managers by mid-May, plus Medicare access through the CMS GLP-1 Bridge program beginning July 1, 2026. Foundayo's trial weight loss ran closer to 11% at its highest 17.2 mg dose over 72 weeks, well below injectable Zepbound's roughly 20.9% at its top 15 mg dose in the SURMOUNT-1 trial. Both oral drugs list near $1,050 to $1,060 per month, in the same range as injectable list pricing, so the pill format buys convenience and needle avoidance, not a lower cost basis.

The actuarial question the oral entrants raise is persistence, not price. A pill removes the injection-related discontinuation drivers that push roughly two in three GLP-1 users off therapy before 12 weeks in prior real-world data, but it also attracts a milder-severity population willing to accept a smaller efficacy result for convenience, a mix shift that could pull average adherence and average clinical benefit in opposite directions simultaneously. A plan actuary layering oral utilization into an existing GLP-1 morbidity model should not assume the injectable adherence curve or the injectable offset percentages carry over unchanged; both need their own tracking cohort once oral claims volume is large enough to be credible. Our companion piece on the oral launch works through the pharmacy-trend side of that reset in more depth.

The FDA Narrows the Lower-Cost Off-Ramp

Compounded semaglutide and tirzepatide had functioned as a lower-cost access valve for some employer plans and stop-loss carriers managing GLP-1 exposure without committing to full formulary coverage. On April 30, 2026, the FDA proposed excluding semaglutide, tirzepatide, and liraglutide from the Section 503B bulks list, finding no clinical need for outsourcing facilities to continue compounding them now that FDA-approved versions are broadly available; the public comment period closed June 29, 2026. FDA Commissioner Marty Makary framed the rationale around drug safety: "When FDA-approved drugs are available, outsourcing facilities cannot lawfully compound using bulk drug substances unless there is a clear clinical need." The agency cited more than 455 adverse event reports linked to compounded semaglutide and over 320 tied to compounded tirzepatide, many involving dosing errors from multidose vials. If finalized, the exclusion closes the compounded channel without reducing underlying demand, which pushes utilization in plans that cover at all toward full-price branded product and removes a cost-management lever some self-funded plans and their stop-loss carriers had been leaning on. Our deeper look at the compounding closure covers the reserving and repricing mechanics in full.

What Actuaries Should Formally Recommend

The renewal-cycle recommendation that follows from this data is a structured comparison, not a single trend load. Build near-term GLP-1 drug cost on a per-member-per-month basis against the net present value of the downstream cardiovascular and diabetes-related claim reduction the group's own diabetic-comorbid cohort can reasonably expect, discounted over a 3-to-5-year horizon rather than the renewal's own 12 months. Run that comparison separately for the diabetic-comorbid cohort and the weight-management-only cohort, since pooling them, as Aon's own results make clear, systematically overstates the offset for the larger, faster-churning weight-management population and understates it for the smaller, more persistent diabetic cohort. Layer in sensitivity analysis on two inputs that move the answer more than any other: adherence, since the gap between real-world and high-adherence offsets runs 3 to 4 points wide in Aon's data, and coverage penetration, since a broader eligibility definition pulls in members further from the acute-event risk that generates the largest dollar offsets. Present the result as a range with an explicit crossover year, not a point estimate, and update it annually as the group's own claims experience accumulates rather than relying indefinitely on a national dataset that was never built around any single plan's demographic mix.

Why This Matters

The IFEBP and Aon releases landing in the same year make the actuarial gap impossible to ignore: one data set shows employer behavior frozen at a one-year lens, the other shows a real, quantifiable, multi-year offset for a specific, identifiable subpopulation. Neither figure is wrong, and neither alone is sufficient to set a coverage recommendation. Actuaries who keep treating GLP-1 coverage as a single pharmacy-cost line item will keep recommending against it on a defensible 12-month basis, while leaving a genuine three-to-five-year savings opportunity unpriced for the diabetic-comorbid members who drive it. Building the two-cohort model, with its own crossover year, its own persistence assumptions, and its own sensitivity bands, is the difference between a renewal recommendation that reads as caution and one that reads as an actual actuarial judgment about where a specific group's risk sits on that curve.