First Biosimilar Insulin Glargine Failed to Curb America’s Insulin Spending, Real-World Analysis Finds

When the first biosimilar version of insulin glargine arrived in the United States, policymakers and clinicians hoped it would do for insulin what generics had done for countless small-molecule drugs: flood the market with cheaper competition and drive prices down. A new exchange published in the Journal of General Internal Medicine suggests that hope has so far gone unfulfilled. In a letter to the editor, researchers Jonathan H. Watanabe and Michael W. Strand of the University of California San Francisco’s School of Pharmacy, together with statistician Weining Shen of the University of California, Irvine, respond to questions surrounding their real-world data analysis of insulin glargine utilization and spending before and after the introduction of the first biosimilar insulin glargine product. Their response, and the methodological care with which they defend it, offers a revealing window into why biosimilar competition in the insulin market has behaved so differently from the generic revolution that transformed other corners of American pharmacy.

Insulin glargine is a long-acting basal insulin analog that millions of people with type 1 and type 2 diabetes rely on to control blood glucose around the clock. For years, the market was dominated by a single originator product, making it one of the classic examples of a biologic drug with no direct competition. Biosimilars are copies of biologic medicines that are demonstrated to be highly similar to their reference products in quality, safety, and efficacy, yet without being clinically identical in the way generic small molecules are. Because insulin glargine is manufactured in living systems, its copies cannot be verified by a simple chemical assay; instead, regulators require extensive analytical, pharmacokinetic, and clinical comparative data. That regulatory architecture, designed to protect patients, has also slowed market entry and shaped how payers, pharmacies, and manufacturers respond once a biosimilar finally does arrive.

The clinical foundation for biosimilar insulin glargine was laid in randomized controlled trials conducted well before real-world uptake could be measured. The ELEMENT 2 study, published in Diabetes, Obesity and Metabolism, compared the investigational biosimilar LY2963016 with the originator insulin glargine in patients with type 2 diabetes who were either insulin-naive or already treated with glargine, and found similar efficacy and safety between the two products. Later, the phase III INSTRIDE 2 trial reported comparable results for a different glargine biosimilar candidate, MYL-1501D, versus insulin glargine over 24 weeks in patients with type 2 diabetes. These trials established that biosimilar glargine could, in principle, substitute for the originator without compromising glycemic control or safety. The question the California team pursued was whether that clinical equivalence translated into economic benefit in actual practice, outside the controlled conditions of a trial.

To answer it, the researchers turned to real-world data, examining patterns of insulin glargine utilization and spending across the period surrounding the biosimilar’s debut. Real-world data studies of drug utilization occupy a distinctive methodological niche. Unlike randomized trials, they cannot assign exposure; they must instead reconstruct a credible counterfactual: what would spending and use have looked like had the biosimilar never entered the market? The authors’ response letter makes clear that their analytical strategy leaned heavily on formal time-series methods, the statistical machinery developed for forecasting and intervention analysis in economics and engineering, adapted here to pharmaceutical claims data.

That machinery is not trivial. In their references, the authors cite the foundational work of George Box and Gwilym Jenkins on time series analysis, forecasting, and control, the framework that underlies modern autoregressive integrated moving average modeling. They also cite Rob Hyndman and Yeasmin Khandakar’s automatic time series forecasting tools for the R statistical environment, which allow analysts to fit and select forecasting models systematically rather than by hand. Most tellingly, they cite the 1987 Econometrica paper by Whitney Newey and Kenneth West on heteroskedasticity- and autocorrelation-consistent covariance matrices. That citation signals a careful attention to a problem that plagues interrupted time-series analyses of drug spending: the errors in monthly utilization and spending series are rarely independent, and their variance often shifts over time. Without Newey-West style corrections, standard errors can be badly underestimated, leading researchers to declare statistically significant changes that are really just noise in a serially correlated series.

The core finding that the authors defend in their response is sobering: the arrival of the first biosimilar insulin glargine did not produce the dramatic reduction in spending that generic entry typically delivers. Utilization and spending patterns for insulin glargine, analyzed with these rigorous time-series techniques, did not show the kind of sharp discontinuity that would indicate biosimilar competition reshaping the market. This result stands in stark contrast to the aggregate savings story told by the generic and biosimilar industry itself. The Association for Accessible Medicines, whose 2022 savings report the authors cite, has documented hundreds of billions of dollars in cumulative savings from generic and biosimilar medicines across the United States. The disconnect between that aggregate narrative and the insulin glargine case is precisely what makes the study, and the authors’ defense of it, so consequential.

Why would a biosimilar fail to move the needle where generics so often succeed? The economics of biologics differ in several structural ways. First, a single biosimilar entering a market dominated by one originator creates a duopoly, not the crowded field of many generic manufacturers that drives generic prices toward marginal cost. Second, rebate and formulary dynamics in the United States give pharmacy benefit managers and payers powerful reasons to keep preferred formulary positions with established products whose rebate structures are already negotiated; a biosimilar must often offer deep discounts simply to gain placement, and even then may capture only a fraction of volume. Third, insulin is frequently subject to patient cost-sharing arrangements, deductibles, and assistance programs that decouple the price paid by insurers from the price felt by patients, muting the demand-side response to list price changes. Fourth, switching friction matters: patients stable on a familiar insulin, and prescribers wary of any change in a life-sustaining therapy, may resist substitution even when regulators deem the products interchangeable or highly similar.

The authors’ response letter also addresses the inevitable methodological challenges that such an analysis invites. Real-world claims data are messy: utilization measured in units of volume must be reconciled with spending measured in dollars, and both are affected by changes in formulation sizes, packaging, rebates invisible to the analyst, and shifts in the insured population over time. Time-series intervention analysis requires specifying the date of the intervention, modeling pre-existing trends, and deciding whether to allow gradual or abrupt change. By grounding their approach in the Box-Jenkins tradition and using automatic forecasting procedures to establish a robust pre-intervention baseline, the authors can compare observed post-biosimilar trajectories against what the pre-existing trend predicted. The Newey-West correction then ensures that the statistical inference attached to any deviation is trustworthy. This is the kind of methodological hygiene that separates credible pharmacoeconomic research from casual before-and-after comparisons, and the authors’ willingness to defend it in print reflects the scrutiny that any claim about drug spending inevitably attracts.

The policy implications ripple outward. Insulin pricing has become one of the most visible flashpoints in American health care, prompting state-level copayment caps, federal action on insulin out-of-pocket costs for Medicare beneficiaries, and intense scrutiny of the rebate system. If biosimilar entry alone cannot discipline prices in this market, then the policy toolkit may need to focus on the intermediaries: the formulary and rebate negotiations that determine which products patients can actually access, and the cost-sharing design that determines what they pay at the counter. The insulin glargine experience suggests that simply adding a cheaper copy to the market is not sufficient; the structural incentives of the supply chain must be aligned so that savings reach patients rather than being absorbed in negotiations upstream.

There is also a forward-looking dimension. Insulin glargine was the first biosimilar in this therapeutic class, but it will not be the last biosimilar opportunity in diabetes care, and the lessons from this real-world analysis will inform expectations for every subsequent biologic that loses exclusivity. Pharmacists, health economists, and statisticians such as Watanabe, Strand, and Shen are building the evidentiary base that policymakers will need: careful, transparent, statistically defensible measurements of what actually happens when biosimilars enter complex markets. Their response letter, published in the Journal of General Internal Medicine, is a reminder that in pharmacoeconomics, as in clinical medicine, the details of method determine the credibility of the conclusion. If the United States wants biosimilars to deliver on their promise of affordable biologics, the insulin glargine experience shows that market entry is only the first step; the harder work lies in redesigning the incentives that decide whether a cheaper, clinically equivalent medicine actually changes what patients pay.

Subject of Research: Real-world utilization and spending outcomes for insulin glargine following the introduction of the first biosimilar insulin glargine in the United States

Article Title: Authors Respond: Insulin Glargine Utilization and Spending Before and After the First Biosimilar Insulin Glargine: A Real-World Data Study

Article References: Watanabe, J. H., Strand, M. W., & Shen, W. (2026). Authors Respond: Insulin Glargine Utilization and Spending Before and After the First Biosimilar Insulin Glargine: A Real-World Data Study. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10837-1

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10837-1

Keywords: insulin glargine, biosimilars, drug spending, pharmacoeconomics, type 2 diabetes, real-world data, time-series analysis, drug utilization, Journal of General Internal Medicine, health policy, insulin pricing, biologics