RESEARCH DOSSIER
Pharmaceutical Spending and Specialty Drug Insurance Risk
Abstract. Specialty medicines create complex exposures for health insurers and public purchasing systems. Their economic effects arise through the eligibility and incidence of treated disease, treatment initiation and persistence, dosage and administration settings, gross prices and confidential concessions, insurance benefit design, and the timing and concentration of individual claims. This monograph investigates pharmaceutical expenditure as a contractually allocated risk, separating national spending accounts, pharmacy and medical benefit claims, patient cost sharing, manufacturer rebates, and the ultimate amount borne by the insurer. It synthesizes official international expenditure statistics, original observational evidence concerning treatment access, regulatory instruments, and quantitative insurance models. Its analysis distinguishes recurring specialty therapies from rare cell and gene therapies; examines how formulary and rebate incentives distribute costs and influence access; and derives transparent frequency-severity and reinsurance calculations. The findings establish the necessity of exposure-normalized measurement, source-specific legal and accounting definitions, tail-risk modeling, and evaluation of financial protection alongside insurer cost. Numerical scenarios are explicitly hypothetical, and published simulations are treated as conditional projections rather than observed expenditure. The monograph provides an integrated scientific explanation of pharmaceutical insurance risk without establishing a new theoretical system.
Keywords: Pharmaceutical spending; specialty pharmaceuticals; health insurance; biologics; cell and gene therapies; insurer loss cost; actuarial risk; drug indications; rebates; pharmacy benefit managers; formularies; patient cost sharing; catastrophic claims; reinsurance; solvency; payment reform; international comparison; source validation.
Introduction
Pharmaceutical insurance risk concerns the uncertain future financial obligation associated with providing access to medicines under defined health financing arrangements. Specialty therapies are often concentrated in small populations, administered through complex benefit channels, and subject to rapidly changing clinical indications and contracting practices. The structure of risk includes the arrival of an eligible patient, the clinical decision to initiate treatment, the contractual determination of coverage, the magnitude and timing of payment, and any subsequent rebate or reinsurance recovery. Each step changes the probability distribution of the insurer's eventual obligation.
A treatment can improve survival or prevent later health expenditure and still generate a near-term liquidity and capital shock for the payer financing it. Conversely, a high published price can misrepresent the ultimate insurer cost if the plan receives substantial contractual concessions. Research must therefore identify the payer, denominator, treatment setting, temporal horizon, and type of price before drawing an actuarial conclusion. Such identification is especially important for comparisons across insurance systems that use different pharmaceutical distribution channels.
The analysis is situated within Angela Bogdanova's Health Insurance, Insurance Pricing and Reserving, and Systemic and Emerging Risks. These established Science areas define the broader insurance context. This Research Dossier addresses the narrower research object: the effect of high-cost medicines and pharmaceutical innovations on expected claims, extreme losses, household financial protection, and institutional risk sharing.
Three questions structure the study. Which definitions and primary measurements establish pharmaceutical risk exposure? How do drug indications, utilization, rebates, and formularies affect insured spending and treatment access? How can insurance pricing, reserving, reinsurance, and solvency assessment incorporate catastrophic treatment events and uncertainty across markets? Answers derive from official expenditure and legal sources, peer-reviewed empirical studies, source-critical comparisons, and explicitly stated actuarial identities. No source is assigned evidentiary force beyond the population and method it actually covers.
Foundations and Source-Critical History
The scientific object and its measurement perimeter
Pharmaceutical spending becomes insurance risk when a financing arrangement commits to covering some uncertain portion of the cost of eligible medicines. It is essential to distinguish five transactions: the manufacturer or distributor's commercial sale, the pharmacy or provider's charge, the allowed amount recognized under the relevant benefit, the patient's payment obligation, and the insurer's ultimate retained payment after applicable recoveries. A sixth concept, total resource cost of the therapy episode, may include administration, clinical monitoring, hospitalization, and consequences outside drug benefits. The six values may differ significantly.
The United States' national health expenditure fact sheet reports $467.0 billion in retail prescription drug spending during 2024, an increase of 7.9% over 2023. The same report places 2024 national health expenditure at approximately $5.3 trillion. These statistics describe the specifically defined retail category and the national expenditure aggregate. They do not capture all physician-administered and hospital medicines as a single pharmaceutical category. A plan's specialty expenditure might accordingly rise faster than national retail spending, without any contradiction.
The OECD pharmaceutical expenditure chapter reports average 2023 retail pharmaceutical expenditure of $766 per capita across its reporting OECD countries after purchasing-power adjustment. The United States reported $1,713 and Germany $1,158 on that basis. These figures are measures of retail expenditure in purchasing-power-adjusted dollars, not insurer net claims, and they reflect differences in prescribing, retail distribution, product composition, and national reporting. The OECD's pharmaceutical spending indicator supplies accompanying methodological limits. Comparison without correcting the retail versus hospital-channel boundary can incorrectly turn differences in accounting into apparently different consumption.
A research dataset must record whether the price is a list price, manufacturer invoice price, provider reimbursement, allowed claim, patient plus plan gross payment, plan-paid payment, or truly rebate-adjusted retained cost. The point at which a rebate is accrued and actually received also matters. A rebate unavailable to the insurer cannot be treated as the insurer's saving. Each cost series must identify its reference period, covered population, currency conversion, benefit channel, and medical indication.
Development of the specialty insurance exposure
The commercial transition from conventional outpatient prescriptions to biologics, long-term targeted oncology medicines, and selected cellular and gene therapies modified the insurance loss process. Continuous therapies generate recurring claims in a stock of prevalent patients. One-time treatments can create a very large early payment while any benefits or future cost offsets unfold over years. Additional clinical indications change the eligible population, and improvements in survival can enlarge the treated population even when incidence is stable.
These structural changes make historical cost trends an incomplete forecasting instrument. An insurer estimating the consequences of a newly approved therapy must consider initiation and uptake, average duration, dose intensity, persistence, discontinuation, subsequent lines of therapy, and substitution. Changes in these variables can compound, whereas a model based exclusively on unit price inflation misses the relevant exposure shifts. At the same time, therapeutic substitution and future generic or biosimilar competition can lower the cost of certain treatments, subject to access, contracting, and clinical conditions.
International price research demonstrates a complementary distinction. RAND's international drug price comparison, based on 2022 data, measured substantial US manufacturer price differences relative to selected other countries, with branded and generic drugs behaving differently. Manufacturer price ratios do not translate directly into claim payment ratios. The analysis remains dependent on country and product samples, weighting, publicly observed prices, and estimation of confidential concessions. The insured economic price also depends on benefit design and on who captures the price concession.
Source classification and inferential rules
Evidence in this dossier belongs to four methodological categories. Administrative expenditure reports provide defined historical aggregates. Regulations and official documents establish obligations in particular jurisdictions and at specified effective dates. Clinical and health-economics studies provide population-specific associations or experimental findings, subject to their original designs. Quantitative modeling produces conditional relationships that hold under explicit stochastic and accounting assumptions.
These categories should be connected without being conflated. An official expenditure figure cannot demonstrate that a specific formulary caused the figure's growth. A retrospective association between cost sharing and treatment abandonment cannot be promoted to an experimentally identified universal elasticity. A published simulation of cell and gene therapy spending is conditional on its expected approvals, utilization, prices, and payer mix. Finally, a mathematical model does not claim that its assumptions have been observed in a real insurance population.
The relevant temporal distinction is equally decisive. A claim paid in one calendar year can produce a rebate in another. An individual covered in the year of treatment may leave the insurer before realizing longer-term savings. A medicine approved in one jurisdiction may remain unapproved or unreimbursed elsewhere. All cross-market arguments in the following chapters preserve these boundaries, using source-year qualifications and explicitly distinguishing illustrative calculations from historical observations.
Drug Utilization and Indications
The insured treatment pathway
The unit of pharmaceutical exposure must be defined in both population and clinical terms. A member-month describes the insurer's coverage exposure. An eligible patient-year describes the time during which a person meets specified treatment criteria. An administration identifies a delivered dose, while a prescription fill identifies a pharmacy transaction. None of these quantities is interchangeable. A single patient can generate multiple fills or administrations, and a paid prescription is not necessarily evidence of sustained use.
A defensible treatment pathway begins with disease incidence and prevalence, continues through diagnostic confirmation, eligibility under an approved indication, benefit authorization, prescribing, initial fill or administration, subsequent adherence or persistence, and eventual discontinuation or switching. Each stage is a conditional transition whose probability may change with clinical practice, price and coverage. Where data are available, a longitudinal cohort can estimate these transitions directly. Where clinical identifiers are absent, a claims-based proxy must be disclosed as a proxy rather than a validated clinical state.
Let E denote exposure in eligible person-years for a specified indication, q the probability of initiating eligible therapy, d average treatment duration in years, u the expected number of billable administrations or standardized dose units per treatment-year, and n the insurer's ultimate retained net cost per such unit. A first-order planning relation is:
Expected therapy cost = E × q × d × u × n.
The relation is dimensionally consistent when its inputs are measured according to these definitions. It is a simplification of a cohort process, not an assumption that initiation, duration, dose, and net price are independent. It also requires separate treatment of members continuing from prior years, prior authorization reversals, patients exiting coverage, and payments after the insured period. Where unit prices depend on cumulative volume or market-share thresholds, n itself depends on utilization.
For an insurer with M covered member-months, the corresponding average expected therapy PMPM is expected net therapy cost divided by M. This normalization makes differently sized portfolios more comparable but cannot remove differences in morbidity or benefit scope. When multiple therapies serve an indication, portfolio loss must also account for substitution and treatment sequence. When a patient receives more than one high-cost therapy, adverse dependence among their claims changes the portfolio's tail.
Indication expansion and dynamic prevalence
Drug utilization changes because population health, clinical eligibility and treatment access change. An indication expansion may increase the eligible population even if the price per unit is constant; improved detection can accelerate diagnosis; a new treatment can become preferred earlier in disease progression; and longer survival can increase the stock of continuing patients. An insurer therefore needs separate assumptions for newly eligible patients, current treated patients, treatment discontinuations and deaths, rather than treating the number of current claims as a fixed prevalence rate.
A useful dynamic cohort identity is: treated patients at the beginning of the next period equal surviving and continuing treated patients from the current period, plus new initiations and transfers in, minus discontinuations, deaths and transfers out. This accounting identity becomes clinically meaningful only after definitions of continuation, grace periods, and enrollment changes are fixed. Some therapies involve successive lines of treatment and repeat courses. The transition structure must reflect the drug's actual use and the insurer's contractual obligation.
For example, consider a purely hypothetical treatment with 1,000 eligible patient-years, a 20% initiation rate, twelve months of continuous treatment and $50,000 net annual therapy cost per full-year patient. Expected first-year cost is $10 million before partial-year adjustments. If improved diagnostics expand eligibility to 1,250 patient-years while other inputs remain unchanged, expected spending rises to $12.5 million. If penetration simultaneously rises from 20% to 25%, the count of treated full-year equivalents becomes 312.5 and spending reaches $15.625 million. This 56.25% increase results from multiplying two factors of 1.25. It does not establish any real incidence, treatment price or clinical outcome.
This decomposition also clarifies innovation forecasting. An apparent rise in insurer expenditures may result from increased access to beneficial therapy, a larger risk pool, a new labeled indication, higher prices, or some combination. Describing all growth as 'price inflation' obscures the mechanisms and can misdirect premium, procurement and clinical policy decisions.
Biologics and the limits of substitution
Biologic products require product-specific evaluation of manufacturing, regulatory status, administration, and clinical indication. The FDA's Purple Book background information distinguishes licensed biological products and pertinent biosimilar classifications within the US regulatory framework. The possibility of biosimilar competition creates a plausible price and bargaining channel, but authorization alone does not establish actual substitution, net insurer savings, physician adoption, or identical contracting treatment.
A biosimilar may enter a market where reference-product rebates are substantial. An insurer with access to a larger rebate on the reference product could face a smaller net cost difference than the quoted acquisition-price difference suggests. The calculation must reconcile patient share, plan payment, provider compensation and concessions actually received. Competitive effects also depend on therapeutic class and prescribing channel. International comparisons should not transfer one nation's rules for interchangeability, procurement or substitution into another jurisdiction without checking the relevant legal framework.
Specialty therapies can be delivered under pharmacy benefits, medical benefits, inpatient care, or integrated treatment episodes. Physician-administered biologics may be identified through procedure codes and provider billing rather than an ordinary retail pharmacy fill. A study restricted to retail dispensing can therefore miss much of the insured cost of a product and bias cross-country comparisons. An accurate observation frame must include administration and supportive-care claims when its defined objective is the full treatment episode.
Cellular and gene therapies
The FDA's approved cellular and gene therapy products document that advanced therapies are a clinically heterogeneous regulatory category. The mere presence of a product on an approval list reveals neither its reimbursement in a particular health plan nor its cost after manufacturer concessions. A treatment may additionally require genetic testing, specialist evaluation, collection and processing of cells, hospital administration, management of acute effects, and long-term follow-up.
A one-time therapy can generate a large near-term cash outflow while prospective health benefits accrue over many years. Insurer turnover matters: the payer purchasing the initial intervention may not be the payer benefiting from lower future treatment needs. A full social cost-effectiveness assessment therefore differs from the present value of recoverable savings to a plan with annually renewed coverage. Outcome-contingent agreements require an explicit clinical endpoint, observation period, information governance rules, and rules for payment when patients change coverage.
Conti and colleagues' cell and gene therapy financing simulation estimated average annual incremental US therapy spending of approximately $20.4 billion over 2023–2035 under its prospective assumptions. This is a forecast for a modeled therapy portfolio, not a realized amount paid annually during those years. The analysis also considers the concentration of expenditure across payer types and the charges associated with reinsurance. It is valuable evidence of a possible risk-transfer burden, and its conclusions must remain conditional on uptake, expected approvals, prices, and coverage.
Frequency-severity modeling
Let N denote the count of covered, eligible high-cost treatment events in an annual exposure period, and X_i the ultimate retained insurer cost of event i after member contributions and recoveries to which the insurer is entitled. Aggregate loss S is the sum of X_i over i = 1 to N. If N has a Poisson distribution with mean lambda and event severities are independent, identically distributed and independent of N, then E[S] = lambda E[X] and Var(S) = lambda E[X²]. These are mathematical consequences of the compound-Poisson assumptions, and the variance differs from lambda Var(X) because event-count uncertainty also contributes.
Consider a hypothetical health plan with 100,000 member-years, an expected two qualifying cases annually and a fixed ultimate retained payment of $2 million for each case. Expected annual loss is $4 million. With 1.2 million member-months, the expected pure treatment cost is $3.333 per member-month. Annual standard deviation is the square root of 2 multiplied by $2 million, approximately $2.828 million. For a Poisson count with mean 2, P(N ≤ 5) is approximately 0.9834, while P(N ≤ 6) is approximately 0.9955. Accordingly the 99th-percentile count is six, corresponding to $12 million under constant severity. This is an intentionally constructed illustration, not an estimate of gene therapy prevalence or required regulatory capital.
The example demonstrates that low expected PMPM and high annual aggregate volatility can coexist. Greater membership may diversify independent events, but changes in authorization, therapeutic indications, epidemic incidence, or technology uptake can correlate events across plans. More sophisticated analysis should test overdispersion, severity uncertainty and dependence, with comparison against observed event counts whenever sufficiently credible data exist.
Rebates and Formulary Design
Gross and net financial incidence
The insurer's ultimate loss is determined by contractual obligations and recoveries. Its relationship to the manufacturer's price is mediated by wholesalers, hospitals, pharmacies, pharmacy benefit managers, the patient's benefit design, and public reimbursement rules. A discount offered somewhere in this chain reduces the insurer's loss only when its contract or applicable program allocates the discount to the insurer. The phrase 'net drug price' is otherwise ambiguous because it may refer to the manufacturer, purchaser, plan sponsor, health program or entire supply chain.
A claim-level accounting relation can be written as follows:
Ultimate insurer benefit cost = plan-paid claim amount − manufacturer or intermediary concessions contractually received by the insurer + directly attributable adjustment obligations.
This formulation begins with plan-paid transactions after member cost sharing. If a dataset instead reports the allowed amount inclusive of member payments, the patient's portion must be subtracted before applying the insurer's own rebates. If rebates are already netted against the reported payment, subtracting them a second time causes undermeasurement. Pharmacy service fees, provider administration, contractual performance payments and benefit management charges may be recorded separately, so a full insurer loss-cost account must specify which items it includes.
The FTC second interim staff report provides an important primary institutional investigation of selected US specialty generic medicines. The report found that affiliated pharmacies of three large pharmacy benefit managers recorded more than $7.3 billion in dispensing revenue above estimated acquisition costs, using the National Average Drug Acquisition Cost benchmark, during 2017–2022 for the drugs examined. The same analysis estimated approximately $1.4 billion in spread-pricing income on those analyzed drugs. These are staff estimates based on the report's sample and measurement method; revenue in excess of a benchmark acquisition cost is not the same as accounting profit, and those totals cannot be extrapolated to all pharmaceuticals or every PBM.
The FTC findings nonetheless identify a general contract-governance question: does the intermediary's economic compensation move with the underlying plan's ultimate net acquisition cost, or can its incentives favor products and dispensing channels that are more profitable to the intermediary? The answer depends on the actual rebate-sharing terms, service-fee structures, reimbursement formulas and ownership relationships. A plan's pharmaceutical spending analysis that omits these relationships may confuse transfer payments with changes in resource cost.
Formulary design and access pathways
A formulary defines coverage conditions, preferred products, tiers and utilization restrictions. It can support the clinical use of cost-effective treatments, but it may also impose barriers when authorization is delayed or the benefit does not match patient need. Research should evaluate therapeutic value, utilization and patient access as connected outcomes. A decrease in an insurer's observed pharmacy spending after a restrictive policy does not by itself establish improved clinical efficiency; the policy may have shifted costs to patients, other care settings or later treatment periods.
Doshi and colleagues' original oral anticancer claims study investigated 38,111 patients with newly adjudicated prescriptions for 38 oral anticancer agents in 2014–2015. The investigators reported adjusted rates of prescription abandonment of 10.0% among patients whose out-of-pocket obligation was at most $10, compared with 49.4% among patients facing more than $2,000. This is a striking observed association under the study's definition of no fill of the index agent within 90 days after reversal. The retrospective design does not prove that the difference represents a universally causal cost-sharing effect. Patient characteristics, treatment alternatives and disease context can also affect the observed outcome.
A systematic review of specialty cost-sharing studies further supports attention to the association between patient payments and initiation, adherence or persistence, while its heterogeneous populations and methods limit direct transfer of any one estimate. The combination of the original oncology claims results and broader observational synthesis supports incorporating affordability and access indicators into insurer benefit evaluation. It does not authorize a fixed universal elasticity of utilization with respect to copayment.
A suitable formulary assessment reports medication initiation, time to first fill, adherence or persistence, clinically justified exceptions, denials and appeals, cost sharing by income or vulnerability where data lawfully permit, and downstream medical utilization. For treatment categories with substantially different risks or available alternatives, benefit rules should be assessed at the appropriate clinical and contractual level. A historical claim-level study should not be treated as an evaluation of a newly introduced product without an explicit bridge between populations.
Rebates, public rules, and price negotiation
Contractual rebates are especially important where formularies can confer preferred status on competing products. Under some arrangements, a larger manufacturer rebate can offset a higher gross price, while under others the member's cost sharing remains tied to a gross point-of-sale amount. Whether the insurer captures the concession, and whether the patient's out-of-pocket charge reflects it, therefore affects incentives and financial protection. Net costs must be measured from the perspective of the entity under study.
The US Medicaid Drug Rebate Program has a distinct statutory methodology. Its unit rebate calculation guidance describes the innovator-drug basic calculation, commonly the greater of a statutory fraction of average manufacturer price or average manufacturer price minus best price, with further provisions including inflation adjustments and product-category exceptions. This is not a universal discount formula for commercial insurance, Medicare Part B or all other countries. Any exact application must establish the current statutory product category, effective period and unit definitions.
The Medicare Drug Price Negotiation Program applies its negotiated maximum fair prices to specified selected drugs and implementation years. The initial selected cohort's maximum fair prices apply in 2026, but that rule is bounded by program criteria and does not create a global drug-price ceiling. Specialty insurance forecasts need to identify whether a specific product, payer and year actually falls within its scope.
The OECD's cross-country pharmaceutical statistics make another distinction: national expenditure series differ in the inclusion of hospital medicines, over-the-counter products and confidential price concessions. A country with extensive institutional dispensing may report comparatively low retail pharmaceutical spending even if hospitals finance considerable drug treatment. Policy comparisons must harmonize the location and method of treatment delivery before judging the generosity or efficiency of coverage.
Payment timing, rebate uncertainty and intermediary credit risk
A claim may be settled at the pharmacy or provider before the corresponding manufacturer rebate is determined, invoiced, disputed or received. Claims-incurred valuation, cash-flow liquidity and gross benefit expenditure can therefore differ for one treatment year. An actuarial projection should use an ultimate cost basis for pricing and reserves, while separately projecting the timing of expected payments and recoveries. It should avoid replacing actual contract-specific rebate receivables with a generalized 'industry discount.'
Consider an illustrative claim on which the insurer initially pays $100,000 and has an enforceable conditional right to receive $15,000. If receipt is certain, ultimate plan cost before other expenses is $85,000. If receipt occurs only when a contract condition is satisfied with probability 0.8, and full recovery is expected upon satisfaction, expected recovery is $12,000 and expected ultimate cost is $88,000. The immediate cash outflow remains $100,000. This example is mathematical and hypothetical, with no empirical claim about typical rebate probabilities.
Publicly accessible pharmaceutical datasets may not resolve these details. The Medicare Part D drug spending dataset reports defined drug expenditure and utilization measures that require reading its methodology, with gross drug cost not equivalent to net cost after all manufacturer rebates. Researchers must not infer confidential contractual payments from visible gross averages without stating and validating a separate estimation model.
Intermediary failure, renegotiation or concentrated rebate exposure may also create common counterparty risk. When a single PBM or supplier accounts for substantial recoveries, an adverse change can affect multiple products simultaneously. Tail scenarios should therefore include financial collectibility and settlement delay alongside clinical demand. Such scenarios may be material even when historic paid-claim frequencies are stable.
Catastrophic Spending Dynamics
Catastrophic loss as a distributional phenomenon
A therapy can impose catastrophic insurance risk without representing a large share of total national health spending. Catastrophe in this context refers to the size, concentration and unpredictability of financial obligations relative to the resources of the risk-bearing pool. Its definition must be tied to the insurer, benefit class, accounting period and chosen tail threshold. A $2 million treatment is material to a small employer pool differently from a national social insurance fund. The same therapy can be routine in clinical purpose while financially exceptional in the portfolio.
Three sources of concentration warrant distinction. Single-event severity concerns the expense of one member's treatment. Aggregate frequency concerns how many eligible events occur within a contract year. Systemic dependence concerns the possibility that adoption, diagnostic practice, clinical guidelines, reimbursement policy or supply conditions increase costs across many insurers simultaneously. The expected value of an event is insufficient to characterize any of these tails. Frequency-severity models should estimate not only expected payment but also dispersion, adverse quantiles, conditional tail loss and cash requirements before recoveries.
The prospective financing analysis of cell and gene therapies by Conti and colleagues models incremental annual costs and reinsurance for a future set of therapies. The projected $20.4 billion average annual spending over 2023–2035 is explicitly a scenario dependent on assumptions, and the study identifies uneven risk allocation across commercial insurance, Medicare and state Medicaid programs. Its value to actuarial science lies in the illustration of prospective capital and risk-transfer pressures. It is not a direct measurement of individual insurer claim frequency.
Correlation can defeat naive diversification. If adoption of a newly reimbursed therapy begins simultaneously throughout a country, many insurers may face more claims than their historical event counts suggested. If a new indication expands eligibility, the stochastic driver may affect a large cohort rather than independent isolated cases. Sensitivity models should permit correlated event frequencies, uncertainty in the future eligible population and changing episode severity. Scenario selection should be linked to documented scientific developments and contracting rules, not invented probability precision.
Liability recognition and reserving
Reserving for pharmaceutical benefits must distinguish incurred obligations from prospective events. When an eligible member has received a covered treatment before the financial reporting date but the provider has not yet submitted a complete claim, the insurer may face an incurred-but-not-reported liability. Claims that have been submitted and are under payment review create a different type of development uncertainty. Rebate receivables, payment reversals and retrospective adjustments also affect ultimate amounts. A possible therapy for a member who has not yet met the relevant insured event definition belongs to future exposure analysis, subject to the applicable insurance contract boundary.
A claims development analysis should consider reporting lag, processing lag, provider billing practice, administrative appeals, partial hospitalization episodes, and disputed reimbursement. Standard aggregate development factors may be unstable for rare treatment events because a single delayed claim can dominate the triangle. Event-level case reserves and a distinct large-loss development analysis may be more credible when counts are sparse. Exposure to emerging therapies also warrants expected-loss methods informed by the known treatment pathway where historical data are scarce.
The IFRS 17 insurance contracts standard addresses insurance contract measurement through requirements that include current estimates of cash flows and adjustments for nonfinancial risk within its specified framework. Its contract boundary, recognition and grouping rules must be applied to the actual reporting entity and contract. Accounting measurement cannot be equated automatically with internal pricing reserves or capital requirements. The IAIS Insurance Core Principles address prudential risk and capital adequacy, including technical provisions, under a supervisory framework whose legal adoption and implementation vary by jurisdiction.
An actuarial model should therefore maintain distinct documented outputs for expected pricing cost, best-estimate claims liability, the accounting measurement required by the applicable standard, liquidity, and solvency capital. These calculations may share assumptions and data while differing in time horizon, discounting, risk margins and recognition. Reporting them as if they were one number conceals the institutional purpose of each.
Reinsurance mechanisms and contract basis
Reinsurance can transfer specified excess severity or aggregate drug expenditure risk. Relevant arrangements include individual excess-of-loss or medical stop-loss protection, aggregate stop-loss, special gene therapy coverage, pooling arrangements and program-specific risk-sharing. A contract can provide financial protection only when its actual triggers include the therapy, the correct event definition, reimbursement basis and time period. Clinical labels or broadly worded marketing descriptions are insufficient evidence of coverage.
An individual excess layer attached at $1 million per qualifying event and capped at an additional $1 million would, under the earlier hypothetical fixed $2 million event severity, pay $1 million for each fully covered case. With an expected two cases, pure expected ceded loss is $2 million; pure expected insurer retention before reinsurance premium is another $2 million. The reinsurer's commercial premium also reflects uncertainty, expenses, financing, capacity and profit requirements. It need not equal the pure expected claim transfer. Exclusions, annual aggregates, patient mobility and timing of recovery can substantially change effective coverage.
It is important to keep the measurement basis consistent. A treaty may attach to billed expense, allowed expense, paid claims, total episode costs or retained claims after rebates. If gross payment is $2 million but the insurer is entitled to $400,000 in rebates, a net-of-rebate attachment calculation differs from a gross attachment calculation. Contract definitions must settle who receives rebates associated with a ceded claim and whether later settlements change reinsurance recoveries. Without this analysis, a purported capital relief estimate cannot be independently reproduced.
Reinsurance also reallocates risk rather than eliminates the system's aggregate medical expenditure. The insurer exchanges uncertain losses for a more predictable premium and a residual exposure to treaty terms and counterparty collectibility. A public risk pool may instead spread expenses through mandatory contributions or general taxation. The economically relevant comparison examines total financing cost, access, liquidity and risk-bearing capacity for the specified arrangement.
Insurance benefit design and household risk
Financial catastrophe has a second meaning at household level. A patient may experience medical debt, delayed treatment or catastrophic out-of-pocket health spending even when the insurer faces a manageable claim. Conversely, a comprehensive benefit with a strong member payment cap can protect the patient while increasing the payer's obligation. The two risk distributions must be assessed separately before drawing welfare conclusions.
For 2026, the official Medicare Part D redesign instructions set a $2,100 annual out-of-pocket threshold for covered Part D medicines, following $2,000 in 2025. The rule governs the legally specified Part D benefit and calculation of incurred costs; it does not cap every medical charge, every uncovered drug or all hospital expenditure. Annual protection can also differ from monthly payment smoothing, which changes the timing of what a member pays without necessarily changing the ultimate annual liability.
The WHO–World Bank UHC monitoring report provides a global perspective on effective service coverage and financial hardship. The WHO financial protection indicators define the relevant household expenditure measures. They are metrics of broad health financing and cannot be reinterpreted as direct measures of specialty pharmaceutical exposure. Nonetheless, they establish why insurer cost containment alone is an incomplete measure of insurance performance: an insurer can reduce its own observed drug payments if patients forgo treatment, while households or other public programs face different consequences.
The patient-level original evidence supports this distinction. The oral oncology abandonment study reported a strong association between higher out-of-pocket charges and abandonment of newly adjudicated prescriptions in its retrospective US cohort. A benefit policy that reduces a measured plan claim by delaying or preventing initiation can thus have important implications for access and downstream care. Determining causality and welfare requires evidence beyond the accounting reduction.
Long-term durability and mobility
Some therapies seek durable or curative effects, but uncertainties in real-world response, recurrence, later complications and follow-up create long-lived financial consequences. Insurers operate over coverage periods that may be shorter than the lifetime over which a therapy produces benefits. The patient can change employer, insurer, public program or country. A payment arrangement that credits long-run savings to the initial payer must therefore specify enforceable allocation rights, portability and a measurable outcome horizon.
Outcome-based contracts require particular caution. A refund based on a biomarker at one year is not equivalent to a guarantee of decade-long disease control. Refund probabilities, data validation, privacy rights, claims adjudication and counterparty solvency affect expected value. If an insurer cannot follow the patient after coverage termination, an outcome contract may need a trusted registry or explicit rules for missing follow-up. A theoretical promise to pay on durable results is not a financial hedge until its legal conditions and settlement mechanism are known.
Long-run cost offsets should be modeled as contingent counterfactual expenditures. If therapy prevents future care that would have been financed by other entities, those savings may be economically valuable but should not automatically be assigned to the initial insurer's balance sheet. The correct valuation states whether the objective is societal cost-effectiveness, a national health budget, the commercial payer's contractual cash flows, or household welfare.
Data, Measurement, and Validation
Data architecture and the claims research unit
The choice of dataset should follow the scientific question. National accounts are useful for measuring system expenditure under published statistical boundaries. Dispensing datasets measure drug transactions and quantities, but may omit rebates and physician-administered treatments. Plan administrative records can establish benefit claims and enrollment, but often lack complete clinical indications or long-run outcomes. Regulatory approval lists identify legally authorized products, not actual coverage and use. No one dataset answers all questions about specialty pharmaceutical insurance risk.
For an insurer's own risk analysis, a minimum longitudinal record should include: coverage dates and member-month exposure; age and relevant morbidity characteristics; clinical indication when reliably available; product, dose, quantity and treatment setting; initial authorization; claim status and reversals; allowed and plan-paid amounts; member liability; contractually allocated rebates; rebate accrual and settlement; reinsurance attachment and recovery; and dates of treatment, adjudication and cash payment. Privacy, data minimization and applicable consent restrictions control the use of patient information. If a required clinical characteristic is absent, the model must state what the claim code can and cannot identify.
The Medicaid State Drug Utilization Data and Medicare Part D Spending by Drug can support drug-specific spending and utilization research, but their reimbursement concepts are shaped by program reporting and do not constitute complete, plan-level net prices after confidential manufacturer rebates. Where high-cost medicines are billed through medical rather than pharmacy benefits, claims drawn only from these pharmacy sources are a subset of the risk object.
Data integration also requires a reproducible map of product codes and indication categories. National Drug Codes distinguish retail product packages in US data, while medical claims often use procedure or other billing codes. Product launch, formulation changes, code retirements, and provider reimbursement modifications can change the observed series without changing underlying disease incidence. A stable analytics pipeline therefore versions code maps and documents excluded claims.
Risk decomposition and mathematical consistency
For treatment class j and calendar period t, let M_t be covered member-months and U_jt the number of paid standardized therapy units. Let P_jt represent the average plan-paid cost per unit and R_jt the average insurer-entitled rebate per unit, with their measurement basis held constant. The net therapy PMPM can be written:
Net PMPM(t) = sum over j of U_jt × (P_jt − R_jt) / M_t.
This is an accounting identity when the variables are defined on the same ultimate-incurred basis. It is an approximation when rebates vary by market share, settlement period or tiers. Negative or implausible net results signal a possible mismatch of rebates to paid claims rather than automatic evidence of negative economic cost.
The decomposition separates exposure, treatment frequency and net unit severity. If a therapy's eligible population rises 25% and its treatment initiation fraction also rises 25%, expected treated volume can increase 56.25% before unit price or duration changes. More generally, for positive multiplicative factors in a consistently defined cohort, proportional spending growth is the product of their respective proportional changes. Correlations matter when estimating the factors, and the decomposition must explicitly add continuing treatments, migration and substitution where clinically relevant.
For annual aggregate cost S under an independent compound-Poisson model, expected loss equals expected event count times expected severity. Its variance equals event-count mean times the second raw severity moment. Actual portfolios may exhibit overdispersed frequency, mixed clinical severity and shared calendar shocks. Testing only the expected value is therefore an insufficient validation of extreme-risk decisions.
Validation design and identification
A credible prospective model should hold out future periods, with attention to launches and indication changes that could leak information into training samples. It should evaluate calibration and error across several levels: total PMPM, each major therapeutic class, member-level high-cost concentration, frequency of one-time treatment events, payment delays and the realization of rebates. Good total cost calibration can coexist with serious underprediction of large claims, especially when rare events are obscured by more numerous maintenance drugs.
The analyst should distinguish uncertainty caused by sampling from uncertainty caused by model structure and future policy. Bootstrap or Bayesian intervals can describe sampling variation where the dataset is adequate. Scenario tests should change incidence, eligible population, initiation, therapeutic duration, net price and rebate collections separately, then examine adverse combinations. A new indication can increase both count and severity if it changes the distribution of treated disease. Drug availability disruptions can force substitution into more expensive alternatives. Reinsurance recoveries depend on treaty conditions and the insurer's exact loss basis.
Evaluation of formulary policies requires a causal research design before claims of effectiveness are established. The Doshi and colleagues observational analysis controlled for observed characteristics while assessing the association of patient cost sharing with prescription abandonment. Residual confounding and treatment selection remain possible. A natural experiment or carefully evaluated policy transition can strengthen inference if its assumptions and alternative explanations are tested. A descriptive before-after comparison alone cannot establish that a spending decline represents an improvement in efficiency or patient welfare.
A modeling result should report its exposure denominator, study window, data exclusions, product and clinical cohorts, currency and inflation basis, gross/net convention, rebate settlement, feature availability, estimation procedure, validation metrics and scenario ranges. Exact reproducibility requires the code and data extracts where lawful, or sufficiently detailed aggregated calculations when restricted data prevent disclosure.
Independently checkable worked calculations
The following calculations are constructed solely for methodological demonstration; they are not empirical market estimates.
Example A: A portfolio has 60,000 members continuously insured for one year, yielding 720,000 member-months. Its plan-paid specialty claims total $12 million, with $2 million in rebates contractually owed and expected to be collected by that insurer. Gross plan-paid PMPM is $12,000,000 / 720,000 = $16.6667. Net PMPM is $10,000,000 / 720,000 = $13.8889. The expected benefit cost reduction is $2.7778 PMPM. If the insurer is entitled to only 75% of the rebates, expected recovery is $1.5 million and net PMPM is $14.5833. The difference illustrates contractual incidence, not a drug price change.
Example B: A therapy has an expected 20 treated episodes at $300,000 of insurer-retained ultimate cost each. The expected annual aggregate is $6 million. With 500,000 member-months the pure expected PMPM is $12. If incidence increases by 25% with unchanged severity, expected episodes increase to 25, expected cost becomes $7.5 million and PMPM becomes $15. If severity instead increases 25% with unchanged frequency, mean cost also becomes $7.5 million, but the new severity structure has different implications for event-level excess coverage. Simultaneous 25% increases in frequency and severity give $9.375 million, a 56.25% increase, and $18.75 PMPM.
Example C: In the hypothetical Poisson model developed previously, expected annual event count is two and fixed event severity is $2 million. Mean aggregate loss is $4 million and variance is 2 × ($2 million)² = $8 trillion in squared-dollar units. The standard deviation is approximately $2.828 million. A 99th-percentile event count of six corresponds to $12 million. This result follows mathematically from an independent fixed-severity Poisson assumption and changes with overdispersion, correlated adoption or random severities.
Example D: For an insurer making a $100,000 claim payment, with a potential $15,000 contractual rebate received with probability 0.8, expected rebate recovery is $12,000 and expected ultimate retained cost is $88,000 before other expenses. The immediate claim cash payment is still $100,000. If a separate excess treaty covers only claims after the rebate, its deductible and reinsurance recovery must be calculated using the treaty-defined net value. This is a valuation illustration, not a description of actual rebate contract frequency.
Source-critical limitations and uncertainty
The observational record is more complete for paid transactions than for medical indications, treatment efficacy and confidential discounts. The published empirical studies used in this monograph represent specific therapies, populations, years and data sources. Their observed associations should not be extrapolated to every specialty medicine. Likewise, official national accounts and OECD data have different boundaries from private insurer claims; mixing them introduces avoidable bias.
The report does not claim an original patient-level dataset, an exhaustive meta-analysis, independent clinical trial interpretation, or access to proprietary rebate contracts. Its original contribution is a transparent synthesis of published source-specific findings and the direct actuarial implications of clearly stated assumptions. It distinguishes uncertainty intrinsic to the observed evidence from genuine future uncertainty in launch, treatment uptake, patient mobility and policy change.
The scientific workflow of each future empirical study should preserve an audit table linking every quantitative claim to its source, original time period, level of aggregation, denominator, currency and exact economic concept. Where a parameter cannot be observed, it should be stated as a scenario input rather than presented as measured. This practice prevents manufactured precision in specialty drug risk assessments.
International Synthesis and Scientific Conclusions
United States: mixed insurance and multiple price conventions
The United States illustrates why national health spending cannot substitute for a retained insurer loss account. The 2024 national expenditure figures measure retail prescription spending across multiple financing sources. Commercial plan claims and Medicare or Medicaid reimbursement can be subject to separate statutory and contractual arrangements. Physician-administered treatments may be billed under medical benefits. The same clinical therapy can therefore enter distinct national reporting, plan accounting and reimbursement categories.
Private PBM arrangements form a further layer. The FTC specialty generic investigation studied selected drugs and corporate relationships, finding substantial dispensing revenue above a specified acquisition-cost benchmark. The inquiry supports close analysis of incentives and transfer payments but does not license a universal estimate of all PBM profits or an assumed discount for every insurer. Gross-to-net claims reconciliation requires the actual rights and liabilities of the sponsor.
Policy changes also move financial responsibility. In 2026, the Medicare Part D out-of-pocket threshold was set at $2,100 for covered Part D drugs. A reduction in patient cost exposure does not imply equal reduction in total therapy expenditure; expenditure can be transferred to plan sponsors, manufacturers or public financing under specified rules. Likewise, the Medicare negotiated-price program is product- and year-specific.
Observational evidence concerning US patients complements these legal and expenditure sources. The oral anticancer medication study links higher patient obligations with more abandonment of an index prescription in a defined claims cohort. The systematic review of utilization controls found substantial evidence concerning higher cost sharing and treatment initiation, while identifying heterogeneity in measures of adherence and the effects of prior authorization. A national financial forecast should not mechanically substitute these sample associations for a causal elasticity, but it should acknowledge the behavioral channels that they demonstrate.
Canada: public financing, expenditure forecasting and specialty diffusion
Canada provides a contrasting financing environment in which national expenditure reporting aggregates public and private sources and pharmaceutical coverage is distributed across different programs. The CIHI 2026 expenditure snapshot, released November 10, 2026, forecasts that drugs will represent 12.7% of Canadian health expenditure in 2026 and that drug spending will increase 3.7% that year. The snapshot separates finalized 2024 actual expenditure from preliminary and forecast estimates for 2025–2026. Its drug category is not identical to the US retail prescription drug measure or to a private insurer's net specialty claims.
CIHI identifies biologics and high-cost specialty medicines as continuing drivers of public program expenditure. This institutional observation has direct consequences for pricing and reserving: a public purchaser's cost forecast depends on its formulary population, reimbursement contracts, coordination with private benefits, and the distribution of new treatment eligibility. The snapshot also illustrates how new generic competition can change both net unit costs and treatment availability. Lower prices can encourage greater utilization, making the change in aggregate spending an empirical question rather than an arithmetic certainty.
Canada and the United States share exposure to new therapeutic indications and costly medicine innovation but differ in benefit arrangements and payer responsibilities. Comparing their expenditure totals without harmonizing covered populations, treatment settings, price adjustments, and forecast status would yield unsupported conclusions. Comparative scientific analysis begins from these institutions, rather than from an assumed common price.
OECD comparators and international accounting boundaries
The OECD Health at a Glance 2025 pharmaceutical comparison is useful precisely because it documents how retail pharmaceutical spending varies across countries and because its statistical conventions can be inspected. Differences in pharmacy dispensing, generic use, hospital delivery and purchasing arrangements alter reported retail totals. Purchasing-power-adjusted per capita amounts should not be combined uncritically with nominal insurer premiums, manufacturer list prices or exchange-rate-converted claims.
A research protocol that seeks comparability can instead define a common therapy or indication, specify the treatment episode and standardized product quantities, and calculate payer-specific net cost only for the data actually available. Missing confidential rebates should be treated as missing data with sensitivity scenarios, not as zeros by default. A country with lower observed retail spending may have more hospital-administered medicine use, better net procurement prices, lower uptake, or different underlying population need. The expenditure statistic alone cannot identify which explanation dominates.
Financial protection comparisons require the same discipline. The WHO–World Bank UHC monitoring framework examines service coverage and household financial difficulty at global scale. The WHO monitoring definitions address broad out-of-pocket health spending, not just specialty drug costs. Research that seeks to attribute catastrophic household expenditure to a particular medicine needs linked household and clinical expenditure evidence; national aggregate hardship indicators cannot provide that attribution on their own.
European Union: clinical assessment and national reimbursement
The European Union's joint health technology assessment arrangements introduce a valuable comparative distinction between common clinical appraisal and national pricing. The European Commission's joint clinical assessment program applies from January 12, 2025 initially to specified new oncology medicines and advanced therapy medicinal products. The governing EU legal structure concerns joint clinical assessment within its scope; it does not turn clinical appraisal into a uniform EU-wide reimbursement contract or a single net insurance price.
For insurers, the immediate consequences of a technology assessment depend on the nationally applicable coverage decision, legal benefit entitlements, contracting mechanisms, budget responsibilities, and permitted treatment settings. A therapy assessed through a shared European process can produce different ultimate financial exposures in different countries. This is an institutional source of variation, not a statistical defect that can be corrected by currency conversion alone.
The European case also strengthens the separation of efficacy, value and risk allocation. Positive clinical benefit may be compatible with substantial near-term financial risk. A conditional reimbursement agreement can redistribute that risk only when its outcome measures, payment contingencies and administrative responsibilities are enforceable. Comparison across countries should therefore examine both evidence-based clinical eligibility and the economic incidence of the payment mechanism.
Integrated scientific findings
The preceding evidence leads to a set of connected conclusions. First, pharmaceutical spending growth is a combined movement in eligible population, treatment initiation, persistence, dosage, product mix, gross prices and net concessions. It cannot be described adequately by a single national price index. Second, the party receiving a rebate matters. Economic benefit cost can differ from the gross price visible to patients, providers and public datasets. Third, formulary choices affect treatment access as well as claims; the available observational evidence makes it necessary to measure both outcomes while retaining appropriate limits on causal interpretation.
Fourth, one-time expensive therapies create a financing horizon different from their clinical benefit horizon. Small risk pools can incur very large adverse annual deviations even when their expected PMPM costs are moderate. Fifth, reserving and capital measurement depend on contract definitions and reporting regimes, and transferring loss to a reinsurer does not eliminate system-wide expenditure or counterparty risk. Sixth, international comparisons require harmonized drug and payer definitions, clinical cohorts, expenditure channels, effective policy dates, and currency conventions.
These findings support a practical agenda for further research. Longitudinal, privacy-preserving datasets should connect prescribing and treatment episodes with insurance claim settlement and reported clinical outcomes. Studies of formulary policies should use designs that can identify causal effects on treatment initiation and later health service use. Prospective financial studies should report scenario-specific uptake, eligibility and unit-cost assumptions; high-cost event models should publish frequency-severity calibration, tail sensitivity and reinsurance recoverability. Policymakers and purchasers should compare insurance savings with household burdens and health outcomes, using the same explicit counterfactual and time horizon.
Conclusions
Pharmaceutical Spending and Specialty Drug Insurance Risk identifies a coherent actuarial and institutional research object: uncertain expenditure from eligible medication treatment, allocated among insurers, patients, public programs and counterparties under defined contractual and regulatory rules. Its principal source-critical finding is that manufacturer prices, retail national accounts, provider invoices, gross paid claims, patient cost sharing and rebate-adjusted net insurer losses describe different economic quantities. Scientific comparison requires their boundaries to be declared before the figures are combined.
The analysis establishes that high-cost pharmaceutical risk emerges from the joint behavior of indication expansion, treatment initiation, dose and duration, product substitution, financing contracts and the arrival of rare exceptionally expensive events. A forecast based on unit price alone loses important information about frequency and clinical exposure. A gross-to-net financial model that ignores which institution receives rebates can misstate the insurer's ultimate cost. A claim-reduction assessment that disregards treatment access can also misinterpret a transfer of burden as an efficiency gain.
Catastrophic claims require more than an expected cost estimate. A complete actuarial decision considers tail variability, reserves for incurred obligations, liquidity before recoveries, reinsurance contract definitions, insurer turnover and uncertainty in long-term clinical effects. The mathematical examples demonstrate these principles while making no claim to estimate actual clinical prevalence or market prices. Accounting, prudential solvency and pricing decisions can use common data but require separate legally and economically defined outputs.
International source comparison confirms that health-financing institutions and reporting conventions shape observed expenditures. The United States' mixed purchasing and rebate mechanisms, Canada's public and private expenditure categories, the OECD's retail measurement conventions, and the European Union's distinction between clinical assessment and national reimbursement create genuinely different insurance risk contexts. The scientifically warranted approach is to compare common clinical objects only after specifying each financing boundary and year.
The original analytic synthesis is the integrated, source-linked explanation of exposure, financial incidence and claims concentration. The monograph establishes no new proprietary actuarial doctrine and claims no independently verified discovery of clinical efficacy. Its conclusions rest on the named primary documents, observed study populations, and fully disclosed mathematical assumptions.
References
Bogdanova, A. (2026a). Health Insurance. Insurance by Bogdanova. Published by Aisentica Development. https://insurancebybogdanova.com/science/health-insurance/
Bogdanova, A. (2026b). Insurance Pricing and Reserving. Insurance by Bogdanova. Published by Aisentica Development. https://insurancebybogdanova.com/science/insurance-pricing-and-reserving/
Bogdanova, A. (2026c). Systemic and Emerging Risks. Insurance by Bogdanova. Published by Aisentica Development. https://insurancebybogdanova.com/science/systemic-and-emerging-risks/
Canadian Institute for Health Information. (2026). National Health Expenditure Trends: 2026 Snapshot (published November 10, 2026). https://www.cihi.ca/en/national-health-expenditure-trends/nhex-trends-reports/nhex-trends-2026-snapshot
Centers for Medicare & Medicaid Services. (2025). National Health Expenditure Fact Sheet: Historical NHE, 2024. https://www.cms.gov/data-research/statistics-trends-and-reports/national-health-expenditure-data/nhe-fact-sheet
Centers for Medicare & Medicaid Services. (2025). Final Calendar Year 2026 Part D Redesign Program Instructions. https://www.cms.gov/newsroom/fact-sheets/final-cy-2026-part-d-redesign-program-instructions
Centers for Medicare & Medicaid Services. (2025–2026). Medicare Drug Price Negotiation Program: Selected Drugs and Negotiated Prices. https://www.cms.gov/initiatives/medicare-prescription-drug-affordability/overview/medicare-drug-price-negotiation-program/selected-drugs-negotiated-prices
Centers for Medicare & Medicaid Services. (2026). Medicare Part D Spending by Drug: Data and documentation. https://data.cms.gov/summary-statistics-on-use-and-payments/medicare-medicaid-spending-by-drug/medicare-part-d-spending-by-drug/
Centers for Medicare & Medicaid Services, Medicaid Drug Rebate Program. (n.d.). Unit Rebate Amount Calculation. https://www.medicaid.gov/medicaid/prescription-drugs/medicaid-drug-rebate-program/unit-rebate-amount-calculation
Centers for Medicare & Medicaid Services, Medicaid. (2025). State Drug Utilization Data: Dataset and methods. https://data.medicaid.gov/dataset/158a1baa-5506-400a-8ec3-97756f0b0536
Conti, R. M., DeMartino, P., Gruber, J., Lo, A. W., Sun, Y., & Wu, J. (2025). Innovative insurance to improve US patient access to cell and gene therapy. Milbank Quarterly, 103(1). https://doi.org/10.1111/1468-0009.12728
Doshi, J. A., Li, P., Huo, H., Pettit, A. R., & Armstrong, K. A. (2018). Association of patient out-of-pocket costs with prescription abandonment and delay in fills of novel oral anticancer agents. Journal of Clinical Oncology, 36(5), 476–482. https://pubmed.ncbi.nlm.nih.gov/29261440/
European Commission. (2025). Joint Clinical Assessments of Medicinal Products: Implementation under Regulation (EU) 2021/2282. https://health.ec.europa.eu/health-technology-assessment/implementation-regulation-health-technology-assessment/joint-clinical-assessments-medicinal-products_en
Federal Trade Commission. (2025). FTC Releases Second Interim Staff Report on Prescription Drug Middlemen: Specialty Generic Drugs (January 14, 2025). https://www.ftc.gov/news-events/news/press-releases/2025/01/ftc-releases-second-interim-staff-report-prescription-drug-middlemen
IFRS Foundation. (2026). IFRS 17 Insurance Contracts: Standard overview and guidance. https://www.ifrs.org/issued-standards/list-of-standards/ifrs-17-insurance-contracts/
International Association of Insurance Supervisors. (2026). Insurance Core Principles: ICP 17 and prudential risk standards. https://www.iais.org/icp-online-tool/
Ismail, W. W., Witry, M. J., & Urmie, J. M. (2023). The association between cost sharing, prior authorization, and specialty drug utilization: A systematic review. Journal of Managed Care & Specialty Pharmacy, 29(5), 449–463. https://pubmed.ncbi.nlm.nih.gov/37121255/
Mulcahy, A. W., Schwam, D., & Lovejoy, S. L. (2024). International Prescription Drug Price Comparisons: Estimates Using 2022 Data. RAND Corporation. https://doi.org/10.7249/RRA788-3
Organisation for Economic Co-operation and Development. (2025). Pharmaceutical expenditure. In Health at a Glance 2025: OECD Indicators. https://www.oecd.org/en/publications/health-at-a-glance-2025_8f9e3f98-en/full-report/pharmaceutical-expenditure_ab82eef9.html
Organisation for Economic Co-operation and Development. (2025). Pharmaceutical spending: Statistical indicator and methodology. https://www.oecd.org/en/data/indicators/pharmaceutical-spending.html
U.S. Food and Drug Administration. (n.d.). Background Information: List of Licensed Biological Products with Reference Product Exclusivity and Biosimilarity or Interchangeability Evaluations (Purple Book). https://www.fda.gov/drugs/biosimilars/background-information-list-licensed-biological-products-reference-product-exclusivity-and
U.S. Food and Drug Administration. (2026). Approved Cellular and Gene Therapy Products. https://www.fda.gov/vaccines-blood-biologics/cellular-gene-therapy-products/approved-cellular-and-gene-therapy-products
World Health Organization. (2025). Financial protection: Global Health Observatory definitions and indicators. https://www.who.int/data/gho/data/themes/topics/financial-protection
World Health Organization & World Bank. (2025). Tracking Universal Health Coverage: 2025 Global Monitoring Report. https://www.who.int/publications/i/item/9789240117808