How health economics and outcomes research (HEOR) puts a value on steroid-toxicity

A microsimulation model presented at ISPOR Europe 2025 put a number on the lifetime clinical and economic burden of long-term steroid use in autoimmune disease: $24,915 to $41,368 USD per patient.[1] That range came from measuring steroid harm directly, and it is forcing a reckoning for pharmaceutical teams that have built cost-effectiveness cases on a proxy rather than on proof.

Dose reduction is the proxy - a default shorthand for steroid-sparing value: lower the milligrams, and the harm falls proportionately. It is intuitive and easy to model, and wrong often enough to matter. Individual susceptibility to steroid-toxicity varies, and attribution gets murky in patients on multiple treatments. Therefore, a model built on dose alone misses the complications and costs that follow a patient's actual exposure.

Cost-effective is not the same as cost-saving

Peter J. Neumann, ScD, Director of the Center for the Evaluation of Value and Risk in Health (CEVR) at Tufts Medical Center and Professor of Medicine at Tufts University School of Medicine, has spent four decades studying what healthcare spending buys. His summary is a warning for anyone pitching a steroid-sparing therapy on savings alone.

"Nothing in healthcare saves money. It is not quite true, but it is a good adage to keep in mind. We may have lots of cost-effective interventions that are very good value for money, but they do not necessarily save money.”

A therapy can be worth every dollar it costs and still add to net spending.

The microsimulation work he co-authored with John H. Stone, MD MPH, of Massachusetts General Hospital and Harvard Medical School, and Glenn Philips, PhD, of Argenx, makes that distinction concrete. Cutting steroid dose by half in the myasthenia gravis cohort modeled a gain of 0.56 quality-adjusted life years and $24,915 in avoidable adverse-event spending per patient. Eliminating steroid exposure entirely produced 1.41 QALYs and $41,368 in avoided spending. For drug developers, the economic case for a steroid-sparing therapeutic rests on QALYs gained and costs avoided, not on the disease indication alone.

Neumann is equally direct about why interventions disappoint once they scale beyond a trial population, a phenomenon he calls the woodwork effect.

"What happens when you have interventions is that they become widely used and diffuse into populations who may not benefit from them as much as the original analyses suggest. This is what is sometimes called the woodwork effect: people come out of the woodwork."

Broad, unstratified steroid-sparing claims tend to underperform their pilot data. Tools that identify which patients carry the heaviest toxicity burden deliver better value than blanket deployment, a targeting problem that requires measurement at the individual level, not the population average.

Why measurement has to apply to the individual

Microsimulation makes the population-level argument. It cannot tell a clinician or a payer what one patient's steroid exposure actually costs that patient. Clinical outcome assessments such as those in the GTI Family translate the overarching assumption into an individual measurement; how those instruments work is covered in Why steroid dose alone cannot define value.

Mark Kosinski, Senior Principal, Patient-Centered Solutions at IQVIA, has spent more than three decades building the infrastructure that turns patient-reported data into decisions, starting with the Medical Outcomes Study in 1991. His view is that steroid-toxicity belongs inside that discipline.

"Measurement matters, but only when it changes decisions."

A score with no reference range is just a number. Kosinski's team builds benchmark datasets, similar to the normal ranges on a blood glucose report, so a steroid-toxicity score outside the expected range for a patient's age, sex, and disease triggers a flag rather than sitting unread in a chart. Linked to claims and clinical data, that flag carries predictive weight for hospitalization and emergency care. Kosinski points to oncology patients given steroids ahead of toxic cancer therapies, where ignoring rising toxicity can undermine the primary treatment.

"If they are blind to the risk associated with toxicity, they're just going to keep on feeding steroids to patients. That is not going to have a good outcome."

The same architecture applies to population health management, as when a California managed care organization layered patient-reported outcomes over claims data to separate genuinely high-risk patients from heavy past users of services. Steroid-toxicity data plugs into the process in a way claims data alone cannot.

Measuring what a lab value cannot see

Lori McLeod, PhD, Vice President of Patient-Centered Outcomes Assessment at RTI Health Solutions, works on the psychometric side of the same problem: turning subjective experience into a number that holds up to scrutiny. Depression and anxiety, recognized effects of steroid-toxicity, are exactly the kind of construct her field measures.

"Lab values don't tell you about fatigue. Clinicians might miss mood shifts. Patients are the only ones who can report unobservable outcomes, and we need to ask the right questions to hear them properly."

Steroid dose is observable and easy to record. Steroid harm, however, is often observable only to the patient carrying it, so asking the right question is what allows that harm to enter the evidence base.

The proof point: measuring toxicity from data already on file

The clearest demonstration that direct measurement is practical, not aspirational, comes from a retrospective analysis of myasthenia gravis patients presented at ISPOR 2024.[2] Using the Optum Market Clarity electronic health record database, researchers examined 682 adult patients: 377 steroid initiators matched against 305 steroid-naive controls. Applying the GTI-MD across four domains - BMI, blood pressure, glucose tolerance, and lipid metabolism - the analysis found significantly elevated toxicity scores among steroid initiators, with BMI showing the largest differential worsening. This publication treats the analysis as the HEOR proof point; clinical detail is covered in the myasthenia gravis risk-benefit article.

Albert Whangbo, PhD, who leads the global Evidence Generation practice at ZS Associates and worked on that analysis, describes what it proved.

"Using the GTI-MD, we could turn what clinicians intuitively know about steroid-toxicity into quantifiable evidence, and we did it with data that were already sitting in EHRs."

The operational advantage over conventional trials is real: results in months, at a fraction of the cost, with no patient assessment burden. Whangbo is equally candid about where the evidence still falls short of what payers demand.

"Steroids are cheap and effective in the short term. Convincing payers that steroid-sparing therapies are worth the upfront cost is still a practical challenge. The GTI-MD is a bridge that gets us part of the way. Decision-makers still want the link to hard outcomes: future costs, hospitalizations, quality-of-life."

Why HEOR teams stopped waiting

Glenn Philips, PhD, Vice President of Health Economics and Outcomes Research at Argenx, frames the shift as necessity rather than preference.

Payers, Philips has found, tend to treat steroids as close to free. Short-term efficacy is well established, and the upfront cost of a novel steroid-sparing therapy is hard to justify against a decades-old standard of care. Without a validated way to quantify the harm steroids cause, the case collapses into a dose-reduction claim that never touches the complications actually driving spending: the downstream costs of diabetes, osteoporosis, and infection a patient develops after months on treatment.

Whangbo does not see the analytical gap as the limiting factor.

"There's an enormous opportunity to keep building the bridge from toxicity metrics to patient outcomes and economic consequences. The data exists. The analytical tools are here. What's needed now is collective will and storytelling."

The cost detail behind that storytelling is set out in the not-so-hidden costs of steroid-toxicity.

What this means for HEOR teams, payers, and drug developers

For HEOR teams, the myasthenia gravis retrospective settles the feasibility question. Direct toxicity measurement from data already in EHRs is possible now. At a fraction of the cost of a prospective study, the GTI-MD produces evidence a dose-reduction claim cannot. The strategic question is which teams build it first, a case made in full in the HEOR Measurement Gap article.

For payers, the incentive structure Neumann describes will not correct itself. Steroids remain cheap and effective at suppressing symptoms clinicians must manage, and that pressure will keep steering patients toward steroids unless the downstream cost is quantified against the acquisition-up-front cost of the steroid-alternative. Whangbo's bridge metaphor applies here: the GTI-MD gets the argument partway there, and closing the distance to hospitalizations, utilization, and quality-of-life outcomes requires ongoing work.

For drug developers, the lesson in Neumann's “woodwork effect” matters as much as the cost figures.

A steroid-sparing therapy aimed at a broad population will show weaker economic value than one targeted at patients carrying the highest measured toxicity burden, which depends on measurement at the individual level, not the population average.

References

1. Stone, J.H., Neumann, P., Narayanaswami, P., et al. Microsimulation Model to Estimate the Clinical and Cost Burden of Adverse Events Related to Long-Term Oral Corticosteroid Usage in Autoimmune Diseases in the United States. Poster presented at ISPOR Europe 2025, Glasgow, UK, November 9-12, 2025.

2. Phillips G, Stone JH, Qi CZ, Stone M, Gelinas D, Chamberas A, Amirthaganesan D, Kulkarni R, Whangbo A. Adaptation of the Glucocorticoid Toxicity Index-Metabolic Domains to Electronic Health Records to Evaluate Steroid Toxicity in Adults with Myasthenia Gravis in the United States. Value in Health, Volume 27, Issue 6, S1 (June 2024).