$4.5M Gene Therapy Claims Force Stop-Loss Pricing Overhaul

From reviewing stop-loss pricing across several renewal cycles, the standard large-claim frequency-severity framework has always assumed that catastrophic claims arrive randomly. An employer with 500 covered lives faces the same per-member probability of a $1M+ claim as one with 5,000 lives; the only difference is sample size. Cell and gene therapies violate that assumption in a fundamental way. A patient with sickle cell disease (ICD-10 D57) or metachromatic leukodystrophy (E75.25) is identifiable before enrollment. Their treatment cost is not a draw from a continuous severity distribution; it is a known quantity: $2.2 million for Casgevy, $3.1 million for Lyfgenia, $4.25 million for Lenmeldy. The pricing actuary's challenge is no longer estimating whether a large claim will occur, but modeling when an identifiable candidate will elect treatment and how that election correlates with plan enrollment timing.

Why Traditional Frequency-Severity Models Break Down

Standard specific stop-loss pricing relies on the excess loss factor (ELF): expected losses above a deductible d divided by total expected losses. The ELF is derived from a fitted claim size distribution, typically lognormal or spliced lognormal-GPD, and the severity of any individual claim is treated as a random draw from that distribution. The model works because claims above the specific attachment point are rare, unpredictable at the individual level, and approximately independently distributed across members.

Cell and gene therapy claims violate all three conditions. They are rare in aggregate (EBRI found just 9.2 users per 100,000 enrollees in 2022, accounting for fewer than 0.1% of plan members but 0.5% of total health spending), but they are not unpredictable at the individual level. A patient carrying a sickle cell diagnosis has been in the claims data for years before becoming a CGT candidate. Genetic testing and specialist referral patterns create a trail of prior authorization requests, ICD-10 codes, and pharmacy claims that make the future CGT claim foreseeable.

This foreseeability creates the core actuarial problem: anti-selection. An employer with a known CGT-eligible employee has an incentive to secure stop-loss coverage. The stop-loss carrier, pricing off a pooled frequency assumption that treats CGT claims as random, is unknowingly accepting a concentrated, correlated risk. The random-incidence assumption underlying the ELF calculation is violated, and the resulting PEPM is inadequate for the actual risk being transferred.

Building the CGT PEPM Load from First Principles

The prevalence-based expected cost model replaces the traditional ELF approach for the CGT component. For each FDA-approved therapy, the calculation proceeds in four steps:

Step 1: Eligible population. Estimate the prevalence of the target condition per 100,000 covered lives using ICD-10 diagnosis code frequency in claims data. Sickle cell disease (D57) affects roughly 100,000 Americans, concentrated in the commercial population at approximately 30 per 100,000 employer-plan lives. Metachromatic leukodystrophy (E75.25) is far rarer, at roughly 1 per 100,000 births. Each approved CGT has a distinct prevalence denominator.

Step 2: Treatment-election probability. Not every eligible patient will elect the therapy. Clinical referral data, specialist consultation rates, and observed uptake in the first two years since FDA approval provide the basis for this assumption. EBRI's observed utilization of 9.2 per 100,000 across all CGT categories implies an aggregate election rate well below the theoretically eligible population, reflecting access barriers, clinical eligibility screening, and patient choice.

Step 3: Per-treatment cost. Unlike traditional severity distributions, CGT costs are discrete and published. The pricing actuary works from a known cost schedule:

Therapy	Condition	WAC Price	FDA Approval
Casgevy	Sickle cell disease	$2.2M	Dec 2023
Lyfgenia	Sickle cell disease	$3.1M	Dec 2023
Lenmeldy	Metachromatic leukodystrophy	$4.25M	Mar 2024
Zynteglo	Beta-thalassemia	$2.8M	Aug 2022
Hemgenix	Hemophilia B	$3.5M	Nov 2022

Step 4: Anti-selection adjustment. This is the critical departure from traditional stop-loss pricing. The anti-selection multiplier estimates how much the observed claim probability in the insured population exceeds the population-average prevalence. The underwriter derives this by comparing enrollment timing patterns among employees carrying CGT-eligible diagnosis codes to baseline new-hire enrollment rates. If employees with a D57 diagnosis code enroll in the self-funded plan at 1.4 times the rate of the general population, the anti-selection multiplier is 1.4. Calibration uses the carrier's own claims experience, supplemented by EBRI's cross-employer utilization data.

The CGT-specific PEPM load then equals:

PEPM_CGT = Σ_therapies (prevalence_i × election_i × cost_i × anti-selection_i) / 12

Summed across all covered CGTs, this produces the pure premium PEPM before expense and risk loads. The EBRI data suggests the aggregate CGT component is currently modest relative to total stop-loss premium, which explains why observed market rates have not yet fully reflected the exposure. But with 10 to 20 new FDA approvals expected annually, the eligible population grows each year, and the PEPM load compounds.

Specific Deductible Stress Testing

The standard specific deductible stress test asks: for a group of N members with a specific attachment point S, what is the probability that at least one CGT claim exceeds S, and what is the expected excess?

P(at least one CGT claim exceeds S) = 1 - (1 - p)^N

Where p is the per-member annual probability of a CGT claim exceeding S. For a group of 300 lives and a CGT claim probability of 9.2 per 100,000 (EBRI's observed rate), the probability of at least one CGT claim in a given year is approximately 2.7%. The expected excess at a $250,000 specific deductible for a $4.25M Lenmeldy claim is $4.0 million, which for a 300-life group with $800 PEPM expected claims ($2.88M annual) exceeds the group's entire annual expected claims.

This concentration effect is what distinguishes CGT risk from traditional large-claim risk. A single CGT claim at $4.25M generates excess losses equal to 139% of a 300-life group's total annual expected claims. At a $500K specific deductible, the excess is still $3.75M, or 130% of expected. Even at a $1M specific, the excess is $3.25M. No traditional large-claim scenario produces this ratio of excess-to-expected for a group of that size.

For the stop-loss carrier, this means the expected excess loss per CGT claim dwarfs the premium collected from the group. A 300-life group paying $50 PEPM in specific stop-loss premium generates $180,000 in annual premium. One Lenmeldy claim at $4.25M produces $4.0M in excess above a $250K specific, creating a loss ratio exceeding 2,200% on that single group. The carrier's pooling mechanism must spread this risk across thousands of groups, and the pooled rate must be adequate to fund these extreme but foreseeable events.

Standalone vs. Wraparound Product Design

BCS Financial's entry into the standalone gene therapy stop-loss market in 2026 created a new product category that pricing actuaries must evaluate alongside traditional coverage. The two-tier structure separates large and small employer markets:

Stop Loss GT targets employers with 3,000 or more covered lives. These groups are large enough to self-insure most traditional stop-loss risk and may not purchase conventional specific coverage at all. The GT product provides a standalone layer covering only gene therapy ingredient costs, with a fixed PEPM that reflects the pooled CGT frequency across the carrier's book.

Stop Loss GTS serves groups of 101 or more employees and operates as a companion to an existing traditional stop-loss policy. It carves out CGT ingredient costs from the traditional policy, effectively removing the point-mass severity risk from the primary stop-loss carrier's exposure and transferring it to the specialized GT carrier.

The $4.05 PEPM quoted for 2026 provides a benchmark that pricing actuaries can decompose. At $4.05 PEPM across 1,000 covered lives, the annual premium collected is $48,600. The breakeven claim frequency at an average CGT cost of $3.0M per treatment is one claim per 61,728 member-months, or roughly one claim per 5,144 covered lives per year. That aligns with a frequency assumption slightly below EBRI's observed 9.2 per 100,000, reflecting the underwriting selection and coverage limitations embedded in the product design.

For the pricing actuary evaluating whether to recommend standalone GT coverage or retain CGT risk within traditional stop-loss, the comparison reduces to cost versus volatility. Traditional stop-loss at a $250K specific already covers CGT claims above the deductible, so the incremental value of standalone GT coverage depends on whether the traditional carrier has adequately priced the CGT tail. If the traditional PEPM reflects a stale frequency assumption that excludes the expanding CGT pipeline, the standalone product may be mispriced in the employer's favor.

Value-Based Contracts and Loss Emergence Timing

The USC Schaeffer Center's March 2026 report, "High-Stakes Medicine," documented financing mechanisms that shift the actuarial timing of CGT loss emergence. Value-based agreements (VBAs), in which manufacturers link rebates to clinical outcomes, are the most developed. Bluebird Bio offers up to 80% price refunds on Lyfgenia if hospitalization occurs post-treatment; BioMarin's warranty covers declining percentages over four years.

For the stop-loss carrier, VBAs transform a single-year catastrophic claim into a multi-year recovery stream. The initial claim hits the specific layer at full WAC cost in year one. Rebates arrive in years two through four, conditional on clinical milestones. This creates a timing mismatch: the stop-loss carrier pays the excess in the year of treatment, but the plan sponsor may not receive the VBA rebate until subsequent policy years, potentially under a different stop-loss carrier.

"Drug mortgage" models, in which a financial intermediary pays the manufacturer upfront and the plan sponsor amortizes the cost over three to five years, solve the timing problem differently. They convert the point-mass CGT claim into a level annual payment stream that may fall below the specific deductible entirely. A $4.25M Lenmeldy claim amortized over five years at $850,000 annually stays below a $1M specific deductible, eliminating the stop-loss recovery altogether and shifting the cost entirely to the plan sponsor's retained layer. The IFEBP identifies three program structures (stop-loss type, full carve-out, and installment plans), each with distinct implications for how CGT costs flow through the specific and aggregate layers.

Aggregate Corridor Distortion

The aggregate stop-loss problem is acute for small and mid-sized groups. A 200-life employer with $800 PEPM expected claims has $1.92M in annual expected claims. A single $4.25M CGT claim consumes 221% of annual expected claims before any specific recovery. Even after a $250K specific deductible recovery of $4.0M, the plan's total claims for the year include $250K in retained CGT cost plus all other medical claims. If those other claims run at expected ($1.92M), total retained claims are $2.17M, or 113% of expected, potentially breaching a 125% aggregate attachment in combination with normal claim volatility.

The structural issue is that the traditional aggregate attachment factor assumes a smooth claim distribution where no single member can consume a disproportionate share of total expected claims. CGT claims create a point mass in the retained loss distribution. For a 200-life group, a single member's retained CGT cost of $250K represents 13% of total expected claims, concentrated in one claimant. The aggregate corridor was not designed for this concentration.

Stop-loss carriers cede CGT tail risk through reinsurance layering, typically retaining the first $1M to $2M per claimant and ceding excess to a reinsurance panel. The retrocession market's appetite for CGT exposure has tightened as the number of approved therapies grows, and reinsurers increasingly demand CGT-specific treaty terms, including aggregate annual limits and per-therapy sublimits. This reinsurance cost flows through to the primary PEPM. If the retrocession market restricts CGT capacity, primary carriers face a choice between raising rates, tightening underwriting (through lasers on known CGT-eligible members), or accepting higher retained volatility.

Why This Matters for Pricing Actuaries

The gene therapy pricing problem is structural, not cyclical. Each new FDA approval expands the eligible population and adds a new point mass to the severity distribution. Three technical priorities follow:

Separate the CGT component from the traditional ELF. Gene therapy claims do not belong in the continuous severity distribution that drives the ELF calculation. They should be modeled as a discrete frequency-severity overlay, with prevalence-based frequency and a known cost schedule replacing the fitted tail distribution. Blending CGT claims into the traditional lognormal or GPD fit distorts the shape parameter and produces mispricing at every deductible level.

Build an anti-selection detection framework. The stop-loss underwriter needs a systematic method for identifying groups with elevated CGT probability. Diagnosis code screening of the employer's prior-year claims data, combined with enrollment timing analysis, provides the inputs for the anti-selection multiplier. Groups with identified CGT candidates should receive an explicit load or laser, not a pooled rate that subsidizes their risk across the book.

Model the aggregate interaction explicitly. For groups under 500 lives, the CGT claim's impact on the aggregate corridor must be modeled as a discrete scenario, not absorbed into the smooth distribution of net retained claims. A Monte Carlo simulation that includes CGT claim scenarios at their discrete cost levels, layered on top of the traditional retained claim distribution, produces a more accurate aggregate attachment factor than the analytical approach based on a continuous distribution.