I’m a lab scientist by training, and early in my career I got lucky to work on a drug from molecule conception all the way through FDA clearance, which is a rarity in this field. Most pharma scientists are pigeonholed into their area of technical expertise, but because we were a small company with a small team (before we were acquired), everyone had to learn and do everything.

It was shortly after the completion of the human genome, which vastly expanded our understanding of disease biology. As an industry, we were starting to use genetic and genomic science much more in drug discovery. Before we had this understanding, drug discovery was largely about screening millions of chemical compounds against cells in a process called Hit-to-Lead, then trying to figure out the how and why of what worked after the fact. Getting from an active compound to first-in-human took years, and we didn’t always know what a drug would do once we put it into people (this is still somewhat true).

The Advent of Targeted Therapies (In All Diseases)

In the last 20 to 30 years, drug development has become incredibly intelligent. We know a lot about both human biology and the biological mechanisms of disease that cause symptoms and, in some cases, mortality. We’re not shooting in the dark; we are designing medicine for very specific targets of disease function. That doesn’t mean drug development is easy; some targets are hard to drug, some diseases are more complex than others, and there are factors when you take a chemical entity out of the lab and put it into people that can’t be anticipated in vitro. But the science we have now is extraordinary; we have never had more intelligent medicines in human history.

Most drugs are created by understanding the biological pathway driving disease (and creating symptoms) and developed to either block or modify that process. Take one of the largest selling drug classes in the world, TNF inhibitors. This class includes drugs like Humira and Enbrel, approved across several autoimmune indications including rheumatoid arthritis, psoriasis, and IBD. TNF inhibitors block signaling in the TNF (Tumor Necrosis Factor) pathway, originally discovered in cancer research before it was understood to drive inflammation in many autoimmune conditions. In rheumatoid arthritis, that inflammation can cause joint swelling and stiffness. In psoriasis, plaques, redness, and itchiness. For patients whose disease is driven by TNF inflammation, suppressing this signaling can provide substantial symptom relief.

But TNF inhibitors only work in patients whose disease is actually driven through TNF- between 30 and 40 percent of patients. It would make sense, then, to confirm that TNF is the source of inflammation before prescribing. But even though we have the technical ability to do that, we don’t determine that in patients before we prescribe this class of drugs. Across the indications in which TNF inhibitors are approved, the overall response rate is about 38%.

Why, when patients don’t want to take medications that don’t work for them, when it’s a waste of medicine to put someone on a drug that doesn’t target their disease biology, when for progressive diseases it means worsening symptoms patients often never recover from, with significantly decreased quality of life, why do we prescribe this way when we know better?

Reluctance to Limit the Market

More than 20 years ago, the FDA recognized this and convened a series of meetings with industry leaders to discuss a “companion diagnostic” strategy: requiring a diagnostic test to determine which patients should be prescribed which drugs as a condition of approval. The pushback from pharma was swift, and millions of dollars were spent lobbying Congress to argue this would slow innovation and prevent medicines from reaching patients. Part of what the FDA got wrong was the intention to require pharma to fund development of these tests, which doesn’t make any business sense, as these tests that would effectively shrink the market for their own drugs. It was well-intentioned, given the financial gap between how our system pays for drugs versus lab tests, but it didn’t align any stakeholder’s incentives or create a viable market path for the diagnostics. Instead of creating a payment mechanism for diagnostics to be paid in the market, separate from the drugs that would be dependent on them, it attempted to create a financial dependency for the diagnostic industry on pharma.

The market reality is that the blockbuster model has dominated pharma strategy for decades. Small companies do the hardest work of early R&D and Phase I clinical trials; big pharma is focused on drugs that drive sales into the billions. Despite decades of evidence that the majority of these drugs are effective in only a small percentage of patients, the trial-and-error model is widely accepted, mainly due to the financial upside for the pharmaceutical industry.

At the company I worked for, after seeing our initial data, the FDA suggested we might need a biomarker to triage which patients would be good candidates for our drug in clinical trials. The drug had a high severe adverse event (SAE) rate, and the FDA was telling us that combined with a response rate below 40%, we might need a biomarker to predict who would benefit in order to gain regulatory approval. That became my focus, developing a biomarker test to predict clinical benefit and identify risk for harmful side effects. Because we knew how the drug worked, this wasn’t hard. Within weeks, we had a working prototype that eventually went alongside our clinical trial. Over time, we learned we could very accurately predict both who would have a meaningful clinical response and who was at risk of a poor outcome, based on levels of certain proteins.

I was surprised when the FDA came back, after reviewing the cumulative data, to tell us they were going to approve the drug based on the response rate in a very challenging cancer — without requiring any biomarkers. By then, we had years of patient data showing we could predict both durable response and severe adverse events. At our launch party, I asked our CEO, whom I admired greatly, if we would use the test anyway, giving the drug only to patients likely to benefit and sparing those we knew would be harmed. He gently laughed at my naivety and told me we were going to give the drug to whoever the FDA would let us, because we had investors to repay. It was my first lesson in how our healthcare system really works.

The Reality of Low Response Rates

The average response rate for FDA-approved drugs is about 36%, across all diseases. The drug I worked on had a response rate of 40% and an SAE rate above 30%, both of which we knew at submission, and both of which we had diagnostic tools to predict. Many drugs have far lower response rates, often expressed as Number Needed to Treat (NNT): how many patients you have to treat with a drug before benefiting one. Across some of the most commonly prescribed drug classes, the NNTs are shockingly high.

Nature Graphic, 2015, Schork, N. Personalized Medicine: Time for one-person trials.

What this means for patients is that they are forced into a system of trial-and-error medicine, where their physician prescribes a drug without knowing whether it will work. Patients are often told it will take weeks or months to see if a drug works, and to come back and try something else if it doesn’t. With so many drugs across indications, and no tools used to predict response, we need a way to decide which patients should get which drugs, in what order. But how we determine who gets what drug in what order doesn’t end up being driven by clinical evidence.

How We Got the Order Wrong: A Brief Note on PBMs

I covered the mechanics of pharmacy benefit managers in detail in my last piece — the rebate doom loop, vertical integration with insurers, and how high list prices became good business for everyone in the chain except patients and employers. The short version, for context here: PBMs (not your physician) decide what drug is ordered based on their formularies, and that ordering is driven by the size of manufacturer rebates, not by clinical response rates or NNTs. Manufacturers raise list prices to fund those rebates; PBMs keep a slice; insurers offset premiums with the rest; and patients pay coinsurance calculated on the inflated list price, not the net price the insurer actually paid. The three largest PBMs now control roughly two-thirds of the national market and, with their parent insurers, account for about 22% of national health expenditures- but those numbers are declining as employers have started to recognize this and options like Mark Cuban’s Cost Plus Drugs have shown how much cheaper generic drugs can be outside of the PBM/insurance market.

The relevant point for this piece is what that ordering actually selects for. Formulary position is a function of rebate economics, not of response rate, NNT, or adverse event risk. A drug that works in 30% of patients can sit ahead of a drug that works in 60%, simply because it generates a bigger rebate. Physicians prescribing within those formularies aren’t choosing the most effective drug for the patient in front of them; they are required to use the drug the system has financially preferred- or need to spend their time and resources fighting for a different option for their patients. The majority of drugs prescribed through these mechanisms don’t work for the patient who receives them.

The Clinical Alternative: Diagnostics That Actually Predict Response

The right way to decide what drug a patient should receive is not the size of the rebate, but the biology of the patient’s disease. We have had the laboratory science to do this for two decades. Two distinct categories of testing already exist, are clinically validated, and in many cases are reimbursed, and yet they remain dramatically underused.

Pharmacogenomic (PGx) testing identifies genetic variants that change how a patient metabolizes or responds to a drug. These are not exotic tests. Several have been FDA-recommended or required on drug labels for years.

● DPYD testing before fluoropyrimidine chemotherapy (5-FU, capecitabine). Patients with DPYD variants metabolize these drugs poorly and can suffer fatal toxicity from standard doses. The UK’s NHS made DPYD testing mandatory before fluoropyrimidine treatment in 2020, following an MHRA safety update. The US has no such requirement, and patients continue to die from preventable toxicity.

● TPMT and NUDT15 testing before thiopurines (azathioprine, 6-MP) used in IBD, leukemia, and autoimmune disease. Patients with reduced enzyme activity face severe, sometimes fatal myelosuppression at standard doses. The FDA label has recommended testing since 2005, claims analysis shows that less than 20% of patients prescribed these drugs are tested first.

● HLA-B*57:01 testing before abacavir (HIV). One of the earliest PGx success stories, testing reduced hypersensitivity reactions to near zero and is now standard of care.

● CYP2C19 testing before clopidogrel (Plavix). Poor metabolizers convert the prodrug into its active form inefficiently and have measurably worse cardiovascular outcomes. FDA added a boxed warning in 2010 alerting prescribers that poor metabolizers may not benefit from the standard dose, but the agency stopped short of requiring testing and uptake has remained low.

● CYP2D6 and CYP2C19 testing in psychiatry, where SSRIs, SNRIs, and tricyclics are routinely prescribed by trial and error. Patients commonly cycle through three or four antidepressants over 12–18 months, a process that, for someone with major depression, is not benign.

Response-prediction diagnostics identify whether a patient’s specific disease biology matches the drug’s mechanism. Some are required as companion diagnostics (only in oncology); many more are clinically validated but not required, and therefore underused.

In oncology, we accept this principle. HER2 testing in breast and gastric cancer to predict trastuzumab response. EGFR mutation testing in NSCLC for tyrosine kinase inhibitors. BRAF V600E testing in melanoma and colorectal cancer. MMR/MSI testing to predict checkpoint inhibitor response. BRCA1/2 testing for PARP inhibitor candidacy. Oncotype DX, which identifies the roughly 70% of early-stage, hormone-positive breast cancer patients who can safely skip adjuvant chemotherapy, sparing them toxicity and cost without compromising outcomes. These tests are the reason oncology response rates in molecularly defined subgroups now look very different from the 36% population average.

In autoimmune disease, the gap is wider. Tests now exist to predict TNF inhibitor non-response before a patient cycles through 12 to 18 months of an ineffective biologic, measuring TNF pathway activity, anti-drug antibodies, or transcriptomic signatures of likely response. Tests exist to distinguish IBD patients whose disease is driven by IL-23 versus TNF versus integrin pathways, allowing first-line selection rather than sequential failure. None of these are routinely used, despite the fact that biologics are among the most expensive drugs in the system and the cost of a year of failed therapy, in dollars, in disease progression, in joint damage that doesn’t reverse, in reduction of quality of life, is enormous.

Cardiometabolic disease is the largest category with the greatest morbidity. We treat type 2 diabetes, hyperlipidemia, and hypertension almost entirely by trial and error with drug classes whose response varies substantially by patient biology. Lp(a) testing identifies patients with genetically driven cardiovascular risk that statins don’t address- a population for whom emerging Lp(a)-lowering therapies will matter and standard lipid management won’t. Polygenic risk scores for coronary artery disease are now clinically actionable for stratifying who benefits most from aggressive early prevention, even for patients without traditional risk factors. Continuous glucose monitoring data combined with metabolic phenotyping can identify which type 2 diabetes patients will respond to GLP-1s versus SGLT2s versus metformin alone, a question we currently answer with months of HbA1c trial and error while disease progresses.

The common thread: in every one of these examples, the test exists, the science is sound, and the cost is a small fraction of the drug it informs. What’s missing is a market that pays for the test instead of paying for failed prescriptions.

What’s Changing

While the FDA hasn’t exercised its authority to require diagnostics alongside the majority of approved targeted therapies, the diagnostic industry has built the tools anyway. Insurance companies put up barriers to paying for these tests because the tests upset the rebate economics, since fewer prescriptions of the highest-rebate drug means a smaller pool of rebate dollars to slice. But alternatives are emerging and patients are realizing they have better options.

Self-insured employers are starting to recognize that diagnostics let them sidestep the rebate game and reduce total spend, while getting employees onto the right medication faster instead of cycling through trial and error. That shift reduces sick time, disability time, and lost productivity, which are costs that don’t appear on a PBM rebate report but show up clearly on an employer’s balance sheet. The economics here aren’t hypothetical. When the entity paying for the drugs is also the entity paying for the absent employee, the math on a $300 test that prevents a year of $80,000 ineffective biologic looks very different than it does to a PBM optimizing for rebate yield.

The drugs aren’t the half of the problem we usually argue about. The other half is that we keep prescribing them to people they were never going to work for. We have the tools to stop. We just haven’t built a system that wants to use them.

Subscribe now

Sources and citations:

Spear BB, Heath-Chiozzi M, Huff J. Clinical application of pharmacogenetics. Trends in Molecular Medicine. 2001;7(5):201–204. PMID: 11325631. DOI: 10.1016/s1471-4914(01)01986-4.

This is the original citation for the response-rate-by-therapeutic-area table that has propagated through the precision medicine literature for 25 years. The Spear paper reports response rates ranging from ~25% (oncology) to ~80% (Cox-2 inhibitors), with a population-weighted average around 50% across the 14 indications they tabulated. The “~36% average” figure is broadly consistent with their data when weighted toward chronic disease therapeutic areas, and is also consistent with subsequent meta-analyses.

US National Cancer Institute. Personalized Medicine: Redefining Cancer Research. 2017. (Confirms the heterogeneity of response across cancer therapies.)

Schork NJ. Personalized medicine: time for one-person trials. Nature. 2015;520:609–611. (Reviews response heterogeneity across major drug classes.)

Singh JA, Hossain A, Tanjong Ghogomu E, et al. Biologics or tofacitinib for rheumatoid arthritis in incomplete responders to methotrexate or other traditional disease-modifying anti-rheumatic drugs. Cochrane Database of Systematic Reviews. 2016;CD012183. (Network meta-analysis showing ACR50 response rates of ~40–45% for TNF biologics + MTX.)

Hanauer SB, Sandborn WJ, Rutgeerts P, et al. Human anti-tumor necrosis factor monoclonal antibody (adalimumab) in Crohn’s disease: the CLASSIC-I trial. Gastroenterology. 2006;130(2):323–333. (Crohn’s disease clinical remission rate ~36% on adalimumab.)

Roda G, Jharap B, Neeraj N, Colombel JF. Loss of response to anti-TNFs: definition, epidemiology, and management. Clinical and Translational Gastroenterology. 2016;7(1):e135. (Reviews primary non-response and loss-of-response rates across IBD anti-TNF use.)

US Food and Drug Administration. Drug-Diagnostic Co-Development Concept Paper. April 2005. fda.gov/media/74735/download

This is the foundational FDA policy document for the co-development framework. It was never finalized as guidance, which is itself the story: the document was issued in draft, drew industry pushback, and was effectively shelved before being superseded by various individual guidances (most notably the 2014 In Vitro Companion Diagnostic Devices guidance).

Hayes DF, Allen J, Compton C, et al. Breaking a vicious cycle. Science Translational Medicine. 2013;5(196):196cm6. (History of the co-development framework and why it stalled.)

NHS England. Clinical Commissioning Urgent Policy Statement: Pharmacogenomic testing for DPYD polymorphisms with fluoropyrimidine therapies. November 2020. NHS England policy statement

Medicines and Healthcare Products Regulatory Agency. Drug Safety Update: 5-fluorouracil (intravenous), capecitabine, tegafur: DPD testing recommended before initiation. October 22, 2020.

Shaunak R, et al. Implementation of mass pharmacogenetic testing: Dihydropyrimidine dehydrogenase testing prior to fluoropyrimidine treatment. British Journal of Clinical Pharmacology. 2025. DOI: 10.1002/bcp.70057. (Detailed implementation review.)

Henricks LM, Lunenburg CATC, de Man FM, et al. DPYD genotype-guided dose individualization of fluoropyrimidine therapy in patients with cancer: a prospective safety analysis. Lancet Oncology. 2018;19(11):1459–1467. (The pivotal safety study.)

Sayani FA, Prosser C, Bailey RJ, et al. Thiopurine methyltransferase enzyme activity determination before treatment of inflammatory bowel disease with azathioprine: effect on cost and adverse events. Canadian Journal of Gastroenterology. 2005;19(3):147–151.

Stocco G, et al. Use of thiopurine testing in the management of inflammatory bowel diseases in clinical practice: a worldwide survey of experts. Inflammatory Bowel Diseases. 2011. PMID: 21351210. (The survey showing 27% of IBD physicians always wait for TPMT results before initiating azathioprine; 30% perform TPMT genotyping pre-treatment.)

Fargher EA, Tricker K, Newman W, et al. Current use of pharmacogenetic testing: a national survey of thiopurine methyltransferase testing prior to azathioprine prescription. Journal of Clinical Pharmacy and Therapeutics. 2007;32(2):187–195. (UK data.)

US FDA. Imuran (azathioprine) prescribing information; updates 2003–2018 incorporating TPMT testing recommendations.

US FDA. Purixan (mercaptopurine) prescribing information.

Mallal S, Phillips E, Carosi G, et al. HLA-B*5701 screening for hypersensitivity to abacavir. New England Journal of Medicine. 2008;358(6):568–579. (PREDICT-1 trial.)

US FDA. FDA Drug Safety Communication: Reduced effectiveness of Plavix (clopidogrel) in patients who are poor metabolizers of the drug. March 12, 2010. FDA Drug Safety Communication

Mega JL, Simon T, Collet JP, et al. Reduced-function CYP2C19 genotype and risk of adverse clinical outcomes among patients treated with clopidogrel predominantly for PCI: a meta-analysis. JAMA. 2010;304(16):1821–1830.

Pereira NL, Farkouh ME, So D, et al. Effect of genotype-guided oral P2Y12 inhibitor selection vs conventional clopidogrel therapy on ischemic outcomes after percutaneous coronary intervention: the TAILOR-PCI randomized clinical trial. JAMA. 2020;324(8):761–771.

Bousman CA, Stevenson JM, Ramsey LB, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for CYP2D6, CYP2C19, CYP2B6, SLC6A4, and HTR2A genotypes and serotonin reuptake inhibitor antidepressants. Clinical Pharmacology & Therapeutics. 2023;114(1):51–68.

Greden JF, Parikh SV, Rothschild AJ, et al. Impact of pharmacogenomics on clinical outcomes in major depressive disorder in the GUIDED trial. Journal of Psychiatric Research. 2019;111:59–67.

Sparano JA, Gray RJ, Makower DF, et al. Adjuvant chemotherapy guided by a 21-gene expression assay in breast cancer. New England Journal of Medicine. 2018;379(2):111–121. (TAILORx trial established that the ~70% of HR+ early breast cancer patients with low-to-intermediate scores can safely skip chemo.)

Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Trastuzumab for early-stage, HER2-positive breast cancer: a meta-analysis of 13,864 women in seven randomised trials. Lancet Oncology. 2021;22(8):1139–1150.

Bek S, Bojesen AB, Nielsen JV, et al. Systematic review: genetic biomarkers associated with anti-TNF treatment response in inflammatory bowel diseases. Pharmacogenomics Journal. 2017;17(5):403–414.

Mahil SK, Catapano M, Di Meglio P, et al. An analysis of IL-36 signature genes and individuals with IL1RL2 knockout mutations validates IL-36 as a psoriasis therapeutic target. Science Translational Medicine. 2017;9(411):eaan2514.

Vande Casteele N, Ferrante M, Van Assche G, et al. Trough concentrations of infliximab guide dosing for patients with inflammatory bowel disease. Gastroenterology. 2015;148(7):1320–1329. (For drug-level monitoring as a reactive marker.)

Tsimikas S, Karwatowska-Prokopczuk E, Gouni-Berthold I, et al. Lipoprotein(a) reduction in persons with cardiovascular disease. New England Journal of Medicine. 2020;382(3):244–255.

Kronenberg F, Mora S, Stroes ESG, et al. Lipoprotein(a) in atherosclerotic cardiovascular disease and aortic stenosis: a European Atherosclerosis Society consensus statement. European Heart Journal. 2022;43(39):3925–3946.

Chasman DI, Shiffman D, Zee RY, et al. Polymorphism in the apolipoprotein(a) gene, plasma lipoprotein(a), cardiovascular disease, and low-dose aspirin therapy. Atherosclerosis. 2009;203(2):371–376. (The WHS LPA analysis underlying the dramatic NNT difference.)

Khera AV, Chaffin M, Aragam KG, et al. Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations. Nature Genetics. 2018;50(9):1219–1224.

Klarin D, Natarajan P. Clinical utility of polygenic risk scores for coronary artery disease. Nature Reviews Cardiology. 2022;19(5):291–301.

American Medical Association. Competition in PBM Markets and Vertical Integration of Health Insurers with PBMs (2024 update). Policy Research Perspectives. 2024. AMA 2024 PBM analysis

Lalani HS, Sarpatwari A, Avorn J, Kesselheim AS, Rome BN. Estimated Medicare savings from a generic drug discount program. Annals of Internal Medicine. 2022;175(12):1788–1790. (The widely-cited analysis showing Medicare could have saved ~$3.6B in 2020 if it had paid Cost Plus Drug Company prices.)

Cost Plus Drug Company pricing methodology (publicly disclosed): costplusdrugs.com/about

Subscribe now

Every year, employers across the country receive a rebate check from their pharmacy benefit manager (aka PBM) and feel like they’ve won- who doesn’t love getting money back? Their broker smiles, points to the number, and calls it a success. What nobody tells them is that the rebate they’re celebrating is often a carefully engineered distraction, a small percentage of the value that has already been extracted from their plan, their members, and ultimately their bottom line. The prescription drug supply chain is intentionally one of the most opaque financial systems in American healthcare. This article is about pulling back the curtain on how PBMs actually make money, why the advisors employers trust to protect them often have a financial interest in keeping things murky, and why the rebate check sitting on your CFO’s desk may be costing you far more than it’s worth.

In 2024, Americans spent $467 billion on prescription drugs. Employer-sponsored health plans paid an estimated $130 billion or more of that total, flowing almost entirely through three PBMs that control 80% of the market. In that same year, drug manufacturers paid an estimated $334 billion in rebates and discounts, a figure that has since grown to $356 billion, to secure formulary access through those same PBMs. What percentage employers actually got back is, by design, nearly impossible to verify, and a much smaller fraction than what the PBMs retained. Here I am creating a map to understand where the money goes.

How PBMs Actually Make Money

The original genesis of the PBM was to help insurance companies or employers consolidate buying power and negotiate for lower drug prices with manufacturers. But in practice, PBMs have evolved into something far more complex, and far more profitable. The three largest PBMs (CVS Caremark, Express Scripts, and OptumRx) now control roughly 80% of the U.S. prescription drug market, and they make money at virtually every point in the supply chain- often more money than the manufacturers who actually make the drugs. (Note: there are other, smaller, emerging PBMs that are looking to grow market share and have different business models than these three.)

The PBMs charge employers an administrative fee to manage the benefit. They negotiate rebates from drug manufacturers in exchange for favorable formulary placement and then keep a substantial portion before passing anything back. They engage in “spread pricing,” which means charging the health plan more than they actually reimburse the pharmacy and pocketing the difference. They own mail-order and specialty pharmacies, steering volume to themselves. They charge retroactive “direct and indirect remuneration“ (DIR) fees, to pharmacies, often months after a prescription was dispensed, clawing back reimbursement in ways that are nearly impossible for plans to track. Combined together, this is the reason so many independent pharmacies have gone out of business in the past decade.

And PBMs sit on one of the most valuable datasets in healthcare: detailed prescription claims data that can be monetized in ways most plan sponsors never think to ask about. Each of these revenue streams, in isolation, may seem like a reasonable business practice. Together, they create a system where the entity specifically hired to save you money has a powerful incentive to do the opposite.

The companies that hold these assets are hidden in companies with different names and are domiciled to protect them from both litigation and taxes. The rebate aggregator entities for the three largest are Ascent Health Services (under Cigna/Express Scripts, domiciled in Switzerland), Zinc Health Services (CVS Caremark, Delaware), and Emisar Pharma Services (UnitedHealth/Optum RX, domiciled in Ireland). One more thing- the Big Three PBMs are actually subsidiaries of the insurance companies Aetna/CVS Health, Cigna, and UnitedHealth Group, each of which also owns hospitals, clinics, physician groups, and specialty pharmacies, meaning they control, and make money from, the entire continuum of care. If you wrote this into a movie script, it would be rejected as too farcical, but it’s how we pay for drugs and healthcare in the United States.

Then, there’s the broker.

When an employer hires a benefits broker or consultant, the assumption is that the broker works for them. But this is often wrong, or more accurately, the broker works for both employer clients and for the PBM, making money on both. A significant portion of the benefits consulting industry is compensated, in whole or in part, by the PBMs. PBMs pay brokers through a variety of arrangements: direct fees, “consulting” agreements, volume overrides tied to how many lives a broker steers their way, and a range of other payments that may not show up on a standard disclosure. The result is a structural conflict of interest that is similar to what the financial advisory industry looked like before the fiduciary rule forced fee transparency. A broker who receives meaningful compensation from a PBM has a financial incentive to recommend that PBM, to avoid scrutinizing their contract terms too aggressively, and to stay quiet when a competitor might deliver better value. The Consolidated Appropriations Act of 2021 introduced new broker compensation disclosure requirements that were intended to address this, but compliance has been inconsistent and many employers still don’t know how to ask the right questions or even what the answers mean. A 2023 Drug Channels analysis found that only about 60% of employers report receiving full rebate pass-through, even under contracts labeled ‘pass-through’, meaning that four in ten don’t believe they are getting what they were contractually promised, and even those who do may be missing revenue hidden through affiliated entities or other fee channels.

Rebates are not what they appear to be.

The rebate is the most powerful piece of misdirection in pharmacy benefit management, and it works because it contains just enough truth to be credible. Yes, drug manufacturers pay rebates. Yes, some portion of those rebates flows back to the plan. But formularies are often designed not to deliver the lowest net cost to the plan, but to maximize the volume of high-rebate drugs, which are typically high-list-price brand drugs, because that’s what generates the largest rebate pool for the PBM to draw from. A drug with a $200 list price and a $60 rebate passed through to the plan looks better on paper than a $90 drug with no rebate. But the net cost to the plan is higher, the net cost to the member at the point of sale is higher, and the PBM has retained a portion of the $60 before the employer ever saw it. Employers almost never have access to the data needed to verify what they’re actually paying net of all fees, rebates, and spreads, and PBM contracts are typically structured to make that kind of analysis difficult. The rebate check feels like a win because it’s real money in a real bank account. What’s invisible is the money that never made it in there.

As drug prices get higher, the price discrepancies get worse. Since there is no incentive to pursue lower cost generic drugs, and every incentive to pursue higher cost branded drugs, manufacturers raise their prices to get higher placement on the formulary, which means more rebates to the PBM. This also means more patients will have access to those more expensive drugs first, leading health plans and employers to pay much more for drugs than they have to.

The Imatinib/Gleevec Case Study

How much more could it really be? Perhaps the most enraging example to me (at the time of this writing) is the price differential for imatinib, the generic version of Gleevec. Gleevec is a drug that was approved in 2001 for the treatment of chronic myelogenous leukemia (CML), a devastating blood cancer. It was a true miracle of science; after the FDA approval of Gleevec, CML went from being a death sentence to a cancer with a greater than 90% survival rate. When Gleevec launched in 2001, its list price was an unprecedented $26,400 per year, reflecting both the years of R&D and the value of the new drug. By the time the first generic entered the market in February 2016, the list price for the generic had climbed to $11,800 per month ($141,600/year, which was more than 90% of the price of branded Gleevec at the time), not because the drug had changed, but because the system allowed it.

Most plan sponsors assumed that generic entry would solve the problem, but it did not. By 2023, pharmacy acquisition costs for a 30-pill supply of generic imatinib had fallen to just $59, a 98.8% decline from 2017. Yet the median Medicare point-of-sale price that same year was $1,602 a month, a markup of more than 2,600% over what the pharmacy actually paid for the drug; for commercial plans it was $2500. That gap, aka the “spread”, doesn’t vanish into thin air. The formulary dynamics around imatinib revealed yet another layer of the rebate illusion: by the time generic imatinib became available, over 60% of CML patients had already been shifted to nilotinib and dasatinib, newer branded drugs that have never demonstrated longer survival or better quality of life, but that generated higher rebate revenue and carried higher list prices. The generic arrived, the rebate check shrank, and somehow total plan spending on this drug class barely moved. That is the rebate illusion in its most complete form.

Then in 2022, Mark Cuban launched Cost Plus Drugs, an online pharmacy created to bypass PBMs and sell generic drugs at cost, plus a 15% markup, plus shipping. All of the costing information for drugs available are clearly displayed on the website- how much manufacturing costs, a markup calculation to run the business (15%), pharmacy labor to pull the drug, plus shipping. It seems so basic, but is also so radical because sponsors (and patients) have never has access to this information before. And the information is eye-opening.

One of the first drugs available, on the website, with transparent pricing for any customer or patient to see, was imatinib. The same generic version of the drug that had been approved in 2001, come off patent 10+ years ago, and which still has a 90%+ survival rate for cancers that were formerly death sentences, for a price of $13.18 a month.

Manufacturing: $7.12

15% markup: $1.02

Pharmacy labor: $5

Shipping: $5.25

This is how much it costs to make this drug: $13.18 with shipping included. The remaining $2,487 per refill that sponsors pay is not R&D, not innovation, not care delivery. It is just the cost of the middlemen. And remember, this drug is only for patients with cancer.

Why are plan sponsors paying thousands of dollars a month? Just to feel good about a small percentage coming back in a rebate at the end of the year? Because they have no idea.

You are not supposed to understand this. That’s the point.

This is a system that was designed to be hard to understand, to sell to customers who don’t have the time or the training to be able to delve into the details, who want to hand over these decisions to someone else. The uncomfortable truth is that the information asymmetry embedded in this system is not accidental, it is a feature, not a bug. PBM contracts run hundreds of pages and are written in language that even experienced healthcare attorneys find difficult to parse. The HR directors who manage benefits day-to-day are typically expert generalists juggling open enrollment, leave management, compliance, and a dozen other priorities; they rely on their (trusted) broker to be the subject matter expert and look out for their best interests.

CFOs see the rebate deposit, hear the words “cost containment,” and move on to the next agenda item. Plan members (employees and their dependents) have no visibility into any of this- they experience the system as a copay and a formulary which decides what drugs they will be prescribed (since often the PBMs have more control over this than the ordering physician does), with no idea of what’s driving either. Even self-funded employers who take the time to hire independent consultants and request detailed reporting often discover that their PBM contract contains audit rights that are narrowly defined, data that is aggregated in ways that obscure rather than illuminate, and carve-outs that exempt the most profitable activities from scrutiny. The people who do understand this system in its full complexity are experienced benefits actuaries, a small number of independent PBM auditors, and the PBMs themselves, and these resources are not equally distributed across the negotiating table.

The New Jersey Whistleblower Case

This New Jersey case makes clear what the data only implies. Christin (Chris) Deacon was the Assistant Director of Health Benefits for the State of New Jersey- not a confused HR director or an overwhelmed CFO, but someone whose literal job was to steward over $7 billion in public healthcare funds annually on behalf of 800,000 public employees. When she discovered that Horizon Blue Cross Blue Shield, the state’s third party administrator (TPA), was systematically overcharging the plan and sending plan members false documents that concealed the overcharges, she filed an internal complaint. It was reportedly suppressed. She left state government, filed a whistleblower lawsuit in 2021 alongside five others, and waited four years for a resolution.

In November 2025, Horizon settled for $100 million, which was New Jersey’s largest-ever non-Medicaid False Claims Act settlement. Deacon received none of the $12 million relator share. The lesson is not just that a sophisticated, well-resourced plan was defrauded for years without knowing the full extent of it. It’s that the person who figured it out, sounded the alarm, and spent years pursuing accountability had to do so as a private citizen, under seal, in federal court. That is how hard it is to see inside this system and how hard it is to fight back once you do. But some of us are trying.

Understanding what good looks like is the first step

There is a better way to structure pharmacy benefit management, and a growing number of employers are finding it, but it requires deliberate effort and a willingness to ask uncomfortable questions of advisors who probably prefer that you don’t. The first step is understanding your contract: transparent or “pass-through” PBM models exist, where the PBM charges a flat administrative fee and passes 100% of rebates and spread directly to the plan, making their compensation explicit and auditable.

Independent PBM auditing, which means hiring a third party with no financial relationship with the incumbent PBM to review claims, rebate reporting, and contract compliance, regularly surfaces discrepancies that plan sponsors didn’t know to look for. Some larger employers are pursuing direct manufacturer contracting or carving out their specialty pharmacy benefit entirely.

At a minimum, every plan sponsor should understand exactly how their broker is compensated, demand a full accounting of rebates generated versus rebates passed through, and ensure their contract includes meaningful audit rights with real data access. This requires knowing what the ‘data’ is, what data is relevant and important, and how to analyze data trends to understand your healthcare spending. The regulatory environment is moving slowly, but there is growing attention on the conflicts of interest, state PBM transparency laws are expanding, and federal disclosure rules are tightening. For now, the employers who get the best outcomes are the ones who treat their pharmacy benefit like what it actually is: a complex financial contract that requires active, highly informed management, not passive trust.

In October 2025, California did what Congress couldn’t and signed into law the most aggressive PBM reform in the country. SB 41 imposes an explicit fiduciary duty on PBMs to act in the best interests of the plans that hire them, bans spread pricing outright for new contracts, requires 100% rebate pass-through, and gives plan sponsors quarterly disclosure rights to see exactly where their money is going. It covers employers from the largest health systems to mid-size self-funded plans. Due to the financial implications of these changes, PBMs will almost certainly challenge parts of it in court. But the direction is clear: the practices described in this piece are no longer just bad business. In California, as of January 2026, several of them are illegal.

ERISA Implications- The Carrot and the Stick

The carrot here is saving money; the stick is… the implications of the legal exposure for plan sponsors. Most employers who sponsor a group health plan know the word “fiduciary” in the context of their 401(k) rather than in their health plan. Few understand that the same legal obligation (one that courts have called the highest duty known to law) applies equally to their health plan. The Employee Retirement Income Security Act (ERISA) of 1974 requires plan fiduciaries to act solely in the interest of plan participants and beneficiaries, and to do so with the care, skill, prudence, and diligence of a knowledgeable person familiar with the relevant industry. That is not a passive standard of checking a box and being done. It is not satisfied by signing a PBM contract a broker recommended, receiving a rebate check, and moving on. It requires active, documented, ongoing oversight of every vendor touching plan assets, including the PBM. And the litigation now working its way through federal courts suggests that a significant number of large employers have not been meeting it.

The cases are piling up. In February 2024, a class action complaint was filed against Johnson & Johnson alleging that it failed to meet its ERISA fiduciary obligations in its selection of a PBM and failed to negotiate more favorable pricing terms, resulting in increased costs including higher premiums, deductibles, and copayments for plan participants. The lawsuit alleged that J&J’s health plan paid its PBM for covered prescription drugs in excess of 200 percent, and in some cases 500 percent, of the cash price for the same drugs. In 2025, a similar suit was filed against JPMorgan Chase, alleging the plan paid excessively inflated prescription drug prices and that JPMorgan did not act prudently in overseeing its PBM, which caused participants to pay amounts for generic drugs at drastically higher prices than were available in the marketplace. These are not small companies with unsophisticated benefits teams. These are among the largest and most resource-rich employers in the country, and their employees sued them anyway for wasting their money.

The theory of liability is straightforward and, once you understand it, hard to look away from. Under ERISA, fiduciaries have an obligation to monitor plan service providers, including PBMs and TPAs. When considering service providers, fiduciaries must compare their services, fees, and expenses. Fiduciaries must understand the terms of any agreements entered with service providers and whether the fees charged to the plan are reasonable, and periodically review whether service providers are delivering services at a cost consistent with the agreements. Passive reliance on a broker’s recommendation is not a fiduciary process. Neither is renewing a PBM contract because it’s always been this way or because you like the broker or no one has time to check how the broker is being compensated. Courts have shown an increased willingness to let these cases proceed, reinforcing that ERISA requires active and ongoing oversight, not just passive reliance on third-party vendors.

In April 2025, the Supreme Court ruled unanimously in Cunningham v. Cornell University that employees don’t need to prove a fee arrangement was unreasonable just to get their lawsuit heard in court, the employer bears the burden of proving it was legitimate. This case was about a 403B retirement plan, not a health plan, but the precedent for liability is the same. For plan sponsors who have never audited their PBM contract or documented their fiduciary process, the ruling means the shift in legal burden is not theoretical. It is immediate.

The broker conflict of interest discussed earlier in this piece has now entered this legal frame directly. In late 2025, a new wave of class action lawsuits was filed against plan sponsors and their brokers alleging that employers allowed consultants and brokers to charge excessive fees and commissions, constituting a breach of ERISA fiduciary duty. It is important to note that these suits named both the plan sponsor and the broker as defendants, a departure from earlier cases that only sued the employer. Courts have consistently held that ERISA fiduciary duties are the highest known to law, requiring conduct that is both prudent and loyal and undertaken solely in the interest of plan participants. The plaintiffs in these cases allege that plan sponsors failed to test the marketplace, failed to benchmark pricing, and failed to scrutinize broker compensation structures that exceeded industry norms. The CAA broker disclosure requirements were supposed to create accountability here, but disclosure without understanding is not protection and most plan sponsors still don’t know what questions to ask with the information they’re receiving.

This is challenging for benefits managers because the stakes of non-compliance are concrete. If courts certify classes in these lawsuits, the scale of liability could grow rapidly, encompassing a wide participant group and resulting in significant financial exposure. Remedies under ERISA include monetary recovery for losses, disgorgement of profits, equitable relief, and removal of fiduciaries. And unlike most litigation risk, this one flows directly from inaction, from not auditing, not benchmarking, not demanding data, not reading the contract, and overall, not understanding how money (employee premiums) is flowing. The employer who says “we trust our broker” is not shielded. The employer who can document a rigorous, ongoing, well-reasoned fiduciary process, even if imperfect, is in a meaningfully different position than one who cannot. And with the ever-rising costs of healthcare, employees are paying attention to where their premiums are going.

What can you do?

The prescription drug supply chain will not fix itself, and the legal environment is no longer forgiving of employers who treat their pharmacy benefit as a passive line item. The case in New Jersey demonstrated what it takes for even the most informed insider to surface fraud and force accountability- years of sealed litigation, a whistleblower complaint that was reportedly suppressed, and a $100 million settlement that she herself acknowledged may not fully reflect the scale of what occurred. Most employers will never have a Chris Deacon on their team. That is precisely the problem. As plan fiduciaries under ERISA, plan sponsors have both the right and the responsibility to understand how plan assets are being used, whether their advisors are acting in the plan’s interest, and whether their PBM contract is structured to serve the plan or the middleman. The first step is deciding that “we get a rebate check” and “our broker handles it” are not acceptable answers to those questions.

Ask your broker for a full written disclosure of every dollar they receive in connection with your PBM.

Ask your PBM for a complete reconciliation of rebates generated versus rebates remitted, not a summary, not a PowerPoint deck, an itemized accounting.

Hire someone with no financial relationship to your existing vendors to read your contract. This is also a great opportunity to use AI to prescreen your contracts and start a question list for your prospective brokers. Document the process.

Expect questions from your employees about how their premiums are being spent, why certain drugs have placement on the formulary and others do not, and be ready to answer them.

Courts are now drawing a clear line between plan sponsors who treat fiduciary oversight as a real obligation and those who treat it as a formality, and the employers on the wrong side of that line are finding out the hard way. This is billions of dollars of employees’ premiums, and so the financial waste and legal exposure are real. The good news is, the information you need to protect your plan, your members, your premium dollars, and your organization already exists, it’s just buried in a contract your PBM drafted and a rebate report your broker summarized. It’s time to read the original. Line by line.

**********

I’ve worked in healthcare in some aspect for 25+ years, and the fact that PBMs exist and are allowed to function the way they do baffles me. As I transitioned my career away from the lab and started to learn about the business side of healthcare, I was incredulous that we have built these middleman functions into our healthcare system and allowed them to fester. The only way to change this system is with sunlight by making sure everyone understands how decisions are made, how the money flows and how it impacts their care, and what their rights are.

References and research:

[1] Deacon, Christin. “Testimony Before the U.S. Senate HELP Committee on Healthcare Price Transparency.” U.S. Senate HELP Committee, July 31, 2025. https://www.help.senate.gov/download/deacon-testimonypdf

[2] Chen, Christopher et al.. “Journey of Generic Imatinib: A Case Study in Oncology Drug Pricing.” JCO Oncology Practice, 2016. https://ascopubs.org/doi/10.1200/JOP.2016.019737

[3] Tran, Christine. “Medicare Part D Payments for Generic Imatinib From 2017 to 2023.” JAMA Internal Medicine, 2024. https://pmc.ncbi.nlm.nih.gov/articles/PMC10660232/

[4] Hill, Andrew et al.. “CostPlus and Implications for Generic Imatinib.” PMC / NIH, 2023. https://pmc.ncbi.nlm.nih.gov/articles/PMC9903877/

[5] Deacon, Christin. “Testimony for the Record — Lower Cost More Transparency Act.” U.S. House of Representatives, January 2024. https://www.congress.gov/118/meeting/house/116710/documents/HHRG-118-ED02-20240111-SD005.pdf

[6] NJ Office of Attorney General. “AG Platkin: Horizon Agrees to Settle False Claims Act Case for $100 Million.” NJ Attorney General Press Release, November 14, 2025. https://www.njoag.gov/ag-platkin-horizon-agrees-to-settle-false-claims-act-case-for-100-million/

[7] Davis Wright Tremaine. “California SB 41: Pharmacy Benefit Managers Now Held to Fiduciary and Transparency Standards.” DWT Employment Law Blog, October 2025. https://www.dwt.com/blogs/employment-labor-and-benefits/2025/10/california-pharmacy-benefit-managers-new-law-sb-41

[8] NFP Benefits Advisory. “J&J Lawsuit Impacts Health Plan Fiduciary Obligations.” NFP Insights, March 2024. https://www.nfp.com/insights/jj-lawsuit/

[9] Sequoia Benefits. “Employer Plan Sponsors’ Fiduciary Duties Under ERISA and the Rise in Prescription Drug Litigation.” Sequoia.com, May 2025. https://www.sequoia.com/2025/05/employer-plan-sponsors-fiduciary-duties-under-erisa-and-the-rise-in-prescription-drug-litigation/

[10] Frier Levitt. “Self-Funded Employee Health Benefit Plans and Consultants/Brokers Face String of ERISA Fiduciary Breach Lawsuits.” Frier Levitt Articles, February 2026. https://www.frierlevitt.com/articles/erisa-fiduciary-lawsuits-self-funded-health-plans-broker-fees/

[11] Congressional Research Service. “ERISA: Legal Framework and Recent Supreme Court Litigation.” Congress.gov, March 2025. https://www.congress.gov/crs-product/R48470

[12] Massachusetts Health Policy Commission. “DataPoints Issue 12: PBM Spread Pricing in Medicaid.” mass.gov, 2018. https://www.mass.gov/doc/datapoints-issue-12-printable-version/download

[13] Cancer Today Magazine. “Despite Generic Imatinib, Cost of Treating CML Remains High.” CancerTodayMag.org, 2020. https://www.cancertodaymag.org/cancer-talk/despite-generic-imatinib-cost-of-treating-cml-remains-high/

[14] Buchanan Ingersoll & Rooney. “New Wave of ERISA Class Action Lawsuits Name Brokers as Defendants.” BIPC.com, January 2026. https://www.bipc.com/new-wave-of-erisa-class-action-lawsuits-name-brokers-as-defendants

[15] McKool Smith. “McKool Smith Secures $100 Million Settlement Against Horizon Blue Cross.” PR Newswire via McKoolSmith.com, November 2025. https://www.mckoolsmith.com/newsroom-pressreleases-386

[16 AMCP. “Legislative Update: California Governor Signs PBM Reform Bill.” AMCP.org, October 2025. https://www.amcp.org/letters-statements-analysis/legislative-update-california-governor-signs-pbm-reform-bill

[17 Fein, Adam J.. “Latest CMS Data Reveal Six Trends Reshaping U.S. Drug Spending.” DrugChannels.net, February 3, 2026. https://www.drugchannels.net/2026/02/latest-cms-data-reveal-six-trends.html

[18] Commonwealth Fund. “What Pharmacy Benefit Managers Do, and How They Contribute to Drug Spending.” CommonwealthFund.org, March 2025. https://www.commonwealthfund.org/publications/explainer/2025/mar/what-pharmacy-benefit-managers-do-how-they-contribute-drug-spending

[19] Fein, Adam J. “PBM Power: The Gross-to-Net Bubble Reached $334 Billion in 2023.” Drug Channels Institute, July 16, 2024. https://www.drugchannels.net/2024/07/pbm-power-gross-to-net-bubble-reached.html

Subscribe now

We are living in an extraordinary era of medical science. We understand disease biology, genetics, and human physiology in ways that were unimaginable even a few decades ago.

We are not taking advantage of all this incredible science and advances in medicine because too often we don’t start with diagnosis. Too often, we treat symptoms instead of identifying root causes. Patients are placed on therapies based on symptoms, probability, pattern recognition, or trial-and-error, not objective diagnosis. And when we get it wrong, patients can spend years or sometimes decades being treated for the wrong disease.

How did we get here? How did we get to a point where we have so many diseases and conditions where we rush to treat without understanding what we are treating? As is often the case, let’s follow the money. In the US, coverage of diagnostics across all diseases is largely guided by Medicare policy first. Commercial payers, who have little incentive to cover diagnostic tools early, leverage Medicare policy in their decision making. Medicare has two main areas of data that they look at when they are considering whether to cover a new diagnostic, validity and utility. Broadly, validity means that you can demonstrate what you are detecting in a repeatable way and prove that what you are detecting has clinical meaning and is relevant in clinical decision making. Utility requires the demonstration that the physician will treat a patient differently once they have the output of the test, indicating that they will use the test to inform treatment decisions. On its face, both of those things seem very reasonable.

But we don’t always have treatments for diseases, and there is value in knowing a diagnosis even without treatment. (Difficulty in diagnosis certainly compounds the challenges of drug development.) Only in recent years have we had new drugs to treat Alzheimer’s disease, such as Lecanemab (Leqembi) and Donanemab (Kisunla). Before those drugs came to the market, diagnosing Alzheimer’s disease, and distinguishing it from vascular dementia or normal pressure hydrocephalus (NPH) or any of other many causes of cognitive decline had limited utility by the Medicare definition. Because there was no drug to specifically treat one type of dementia vs the other, there was no way to change the management of the patient. Therefore, Medicare did not cover any diagnostic test for Alzheimer’s, since none could demonstrate the required clinical utility data needed for coverage. Many tests have been clinically validated- meaning they have the data to show that they can diagnose Alzheimer’s disease or subtypes of Alzheimer’s disease- but none of them have been able to generate the clinical utility data, and so none were covered by Medicare.

In 2023 and 2024, when Leqembi and Kisunla were going through the FDA approval process, there were no Medicare-covered diagnostic tests to diagnose Alzheimer’s disease. Both drugs were approved with the requirement that amyloid pathology be confirmed before prescription, and yet, at the time of approval, Medicare did not cover the diagnostic tests needed to make that confirmation. Patients who could not afford a private amyloid PET scan (typically $3,000–$5,000 out of pocket) were effectively locked out of the information used to guide the most effective treatment choice. Or patients were prescribed the drug anyway, without confirmation they had Alzheimer’s disease, since there is no gating mechanism that requires confirmation of Alzheimer’s amyloid pathology before the drug leaves the pharmacy.

The situation is not unique to Alzheimer’s, by any means. We don’t have a diagnostic test for Multiple Sclerosis (MS); it is determined that a patient has MS ‘by exclusion’, essentially, by ruling out other diseases, such as Lyme or migraines. Clinical data shows that 50% of MS patients are misdiagnosed for the first 3 years of their symptoms, during which time they are often treated for other diseases, which they do not have. Since MS is a progressive disease, during this period of misdiagnosis and inappropriate treatment, MS patients often lose function they never regain. And of the people who have an MS diagnosis, 20% do not actually have the disease, even though they are often treated with expensive disease-modifying therapies for MS.

A disease that is often confounded with MS is Lyme disease, which people use as shorthand for a broader group of vector-borne illnesses- infections we acquire from ticks or fleas, from our pets or from being outdoors. Most people are aware of Lyme disease as a general category and know to be aware of a bullseye rash. Most people are not aware that Lyme disease is rarely just Lyme (a bacterial infection caused by Borrelia burgdorferi) but comes with co-infectious agents that are tested for far less frequently. People are also not aware that while symptoms can include joint pain and arthralgias, they also include heart disease, dementia, schizophrenia, and psychosis. Studies have found that a significant proportion of patients in inpatient psychiatric settings test seropositive for Borrelia burgdorferi, and multiple published case reports document resolution of psychotic symptoms following antibiotic treatment. A patient with psychosis being treated with anti-psychotic medication is unlikely to have their symptoms resolve if the underlying cause of the symptoms is infectious- because they weren’t correctly diagnosed. This is a huge opportunity to improve care, but we have to get the diagnosis right.

The mental health space impacts the most people and is where we have the least objective diagnostic tools used. One of the most troubling areas is the challenges of the differential diagnosis between a patient with Major Depressive Disorder, or unipolar depression, and a patient with Bipolar Disorder. The National Depressive and Manic-Depressive Association found that 69% of patients with bipolar disorder were initially misdiagnosed, and in more than one third of cases, 10 years or more passed before a correct diagnosis was made. Studies show that between 21% and 62% of patients with bipolar disorder are mistakenly diagnosed as having major depressive disorder.

The core problem here is that the two conditions look similar in their depressive phase symptoms and patients with bipolar disorder spend most of their time in a depressive episode, so when they present to a primary care doctor, depression is the natural diagnosis, based on symptoms. Some patients experience depressive episodes without a manic episode for five years or more, so for the first years of their illness, they are effectively “correctly” diagnosed with recurrent major depressive disorder, until mania eventually emerges. However, their biology tells a different story.

There are real consequences to this misdiagnosis and assumption of unipolar depression, because these are different diseases, with different medications that have been developed to specifically address chemical pathways for each one. To optimize the use of medication, we first must start with the right diagnosis, or risk significant consequences. A study of bipolar patients previously diagnosed with unipolar depression found that 55% developed mania and 23% became rapid cyclers after (incorrect) antidepressant treatment. Misdiagnosis is associated with poorer quality of life than either correctly diagnosed major depressive disorder or correctly diagnosed bipolar disorder, both of which already have a considerable impact on quality of life. Most tragically, bipolar disorder patients experience a significant increase in suicides when treated with medication for unipolar depression.

Women experience more consequences of the lack of diagnostic tools than men do, as most autoimmune disorders disproportionately affect women and most lack primary diagnostic tools. Women also experience symptoms of disease differently than men, but since most diagnostic criteria are built for men’s bodies, women are often missed. For example, the threshold for the use of troponin as a surrogate marker for a heart attack was set based on men’s bodies but not calibrated for the differences between men and women. Troponin is not a diagnostic test for a heart attack; troponin is a protein that is released into the bloodstream when heart muscles are damaged. During a heart attack, troponin levels rise, and the test detects whether those levels cross a threshold that signals significant cardiac damage. The problem is, this threshold was set without accounting for basic biological differences between men and women.

Cardiac troponin concentrations are persistently lower in women than in men at baseline, with a median concentration of 2.4 ng/L in women versus 3.7 ng/L in men. What this means in practice is that when a woman has a heart attack, her troponin levels rise from a lower starting point. The diagnostic implication is nearly half of women having a heart attack are missed at initial testing. A smaller heart muscle means less total myocardial mass, so even a proportionally significant injury releases less absolute troponin into the bloodstream. This is a known clinical problem, and guidelines now suggest the use of sex-specific thresholds for high-sensitivity troponin assays, but adoption has been uneven and slow. Despite identifying more women with myocardial injury following the adoption of sex-specific thresholds, women remain less likely to be investigated and treated for coronary artery disease than men- and heart disease is the #1 killer of women.

Women also face the consequences of a lack of diagnostic tools in conditions specific to their bodies. Endometriosis affects an estimated 10% of women of reproductive age (roughly 190 million people worldwide) although that number is likely an undercount, since there is no diagnostic test that does not require expensive and invasive surgery. The average time to the diagnosis of endometriosis is 9 years, and most commonly happens when a woman tries to conceive and is unable to. Endometriosis is the leading cause of female infertility, responsible for an estimated 40–50% of cases. The numbers are inperfect, because too many women are never diagnosed. Many girls and women who experience heavy and painful periods are treated with hormonal contraceptives to relieve their symptoms but are never told that those severe symptoms could be endometriosis. People with endometriosis have also higher rates of other autoimmune conditions (which also lack primary diagnostic tools) including lupus, MS, and IBD.

As a scientist, I can tell you we have the technical ability to create diagnostic tools for every disease; many diagnostics exist for the diseases I have discussed here and for many more. But we don’t have a healthcare system that values diagnosis or pays for the tools to diagnose. Across our three primary groups of payers- Medicare, commercial, and Medicaid- each has reasons for resisting paying for diagnostic tools. While scientists don’t like to think about payment- we all like to think if science is there then adoption will obviously follow- payment is the gating factor that determines whether patients have access to almost anything in healthcare.

Each of the three main payer groups has its own reasons for resisting coverage of diagnostic tools. Medicare requires demonstration of both clinical validity and utility; the former is reasonable, the latter creates a structural catch-22, since often utility can only be proven once treatments exist. Medicare also won’t pay for a diagnosis ‘just to diagnose,’ and because it’s focused on patients 65 and older, conditions that manifest earlier in life are deprioritized by design. Commercial payers follow Medicare’s lead on requirements, but their incentives are even more misaligned, since commercial insurance is not incentivized to reduce the total cost of care. (You can read about that here.) One growing exception: self-insured employers, who are increasingly motivated to get employees correctly diagnosed because misdiagnosis directly affects productivity as well as the total cost of healthcare. Medicaid, paradoxically, should be the most motivated to invest in diagnostics since its budgets are capitated, but it is also the most financially constrained and typically lacks the resources to evaluate new tools.

The last bucket is patient self-pay, mainly via direct-to-consumer channels. This is the smallest bucket, as it requires both the financial means to pay for diagnostic tests and the understanding of what to pay for. This segment is growing, as patients realize that more data empowers them to make more informed decisions about their health. The direct-to-patient diagnostics market is growing rapidly, but it’s still not delivering the right diagnostic tests to the patients who need them the most.

This isn’t a failure of science. It’s a failure of incentives.

We have the tools to diagnose disease more accurately than ever before. But we don’t use them because the system doesn’t reward knowing what’s wrong. Until that changes, patients will continue to be treated for their symptoms instead of for what is causing them. And that should be unacceptable.

So what can you do? Start by refusing to accept a diagnosis built entirely on symptoms with no objective test behind it. Ask your clinician directly: Is there a diagnostic test for this condition? Why hasn’t it been ordered? Is what you’re prescribing treating my disease, or managing my symptoms? These are not difficult questions, but they are ones the healthcare system has not trained us to ask, because a system that profits from ongoing symptom management has little incentive to encourage them.

Do your own research- this is genuinely one place where AI is a very useful tool. The diagnostic landscape is changing faster than most clinicians can track; there are validated tests for many conditions that go unordered simply because coverage is uncertain or awareness is low. Don’t assume your doctor has the full picture. Clinicians are often as frustrated as patients by these gaps, but they’re working within a system that doesn’t reward the time it takes to investigate.

What we lack is not science. It is a healthcare system that values knowing what is actually wrong with you before it starts treating you, and the collective willingness to demand that identification be the starting point. That is a policy failure, an economic failure, and ultimately a moral one.

So, push. Ask questions. Send me your questions. I may not always have the answer, but I will find the resources that do.

*****************************************************************************************************

Resources:

Alzheimer’s / Medicare diagnostic coverage

• Van Dyck CH, et al. “Lecanemab in Early Alzheimer’s Disease.” New England Journal of Medicine, 2023. (CLARITY AD trial for Leqembi)

• Sims JR, et al. “Donanemab in Early Symptomatic Alzheimer’s Disease.” JAMA, 2023. (TRAILBLAZER-ALZ 2 trial for Kisunla)

• Alzheimer’s Association. “Medicare Coverage of Lecanemab (Leqembi).” alz.org, 2023.

• Centers for Medicare & Medicaid Services. “CMS Expands Coverage of Amyloid PET Scans.” October 2023. cms.gov

Multiple Sclerosis misdiagnosis

• Solomon AJ, et al. “The contemporary spectrum of multiple sclerosis misdiagnosis: A multicenter study.” Neurology, 2016; 87(13):1393–1399. (Source for 50% misdiagnosed >3 years, 20% without disease, 70% on DMTs)

• Zürrer WE, et al. “Misdiagnosis and underdiagnosis of multiple sclerosis: A systematic review and meta-analysis.” Multiple Sclerosis Journal, 2024. (Pooled misdiagnosis 15%; delay 0.3–15.9 years)

Lyme disease and neuropsychiatric symptoms

· Fallon BA, Nields JA. “Lyme disease: a neuropsychiatric illness.” American Journal of Psychiatry, 1994; 151(11):1571–1583. (Up to 40% of Lyme patients develop neurologic involvement)

· Alvarado-Hernandez D, et al. “Lyme Borreliosis and Associations with Mental Disorders and Suicidal Behavior: A Nationwide Danish Cohort Study.” American Journal of Psychiatry, 2022. (28% higher risk of mental disorders; 2x suicide attempt risk)

· Hajek T, et al. “Higher prevalence of antibodies to Borrelia burgdorferi in psychiatric patients than in healthy subjects.” American Journal of Psychiatry, 2002; 159(2):297–301. (35% seropositivity in psychiatric inpatients — note: contested; seropositivity ≠ active infection)

Foundational seropositivity study in psychiatric inpatients: Hájek T, Pasková B, Janovská D, Bahbouh R, Hájek P, Libiger J, Höschl C. “Higher prevalence of antibodies to Borrelia burgdorferi in psychiatric patients than in healthy subjects.” American Journal of Psychiatry, 2002; 159(2):297–301. PMID: 11823274.

Nationwide cohort study — Lyme and mental disorders: Alvarado-Hernández D, et al. “Lyme Borreliosis and Associations With Mental Disorders and Suicidal Behavior: A Nationwide Danish Cohort Study.” American Journal of Psychiatry, 2022. doi: 10.1176/appi.ajp.2021.20091347.

The strongest epidemiological study to date. Patients diagnosed with Lyme borreliosis at hospital contact were 28% more likely to develop subsequent mental disorders, had a 42% higher rate of affective disorders, and were twice as likely to attempt suicide. Most concerning: 75% higher rate of completed suicide.

Neuropsychiatric review by lead researcher at Columbia: Fallon BA, Nields JA. “Lyme disease: a neuropsychiatric illness.” American Journal of Psychiatry, 1994; 151(11):1571–1583. PMID: 7943444.

Seminal paper establishing the full neuropsychiatric spectrum of Lyme, including paranoia, dementia, schizophrenia, bipolar disorder, panic attacks, and obsessive-compulsive disorder. Reported that up to 40% of Lyme patients develop neurologic CNS involvement.

Geographic correlation of schizophrenia to tick-endemic areas: Brown JS Jr. “Geographic correlation of schizophrenia to ticks and tick-borne encephalitis.” Schizophrenia Bulletin, 1994; 20(4):755–775. PMID: 7701281.

Early ecological study noting that regional schizophrenia rates correlate with tick density and Lyme endemicity, a correlation that has generated decades of follow-up research but has barely made into clinical practice.

Landmark meta-analysis (38 studies, 6,058 patients): Torrey EF, Bartko JJ, Yolken RH. “Toxoplasma gondii and Other Risk Factors for Schizophrenia: An Update.” Schizophrenia Bulletin, 2012; 38(3):642–647. PMC3329973.

Across 38 studies, individuals with schizophrenia had 2.73x higher odds of Toxoplasma seropositivity (OR 2.73, 95% CI 2.21–3.38), an association that exceeds the odds ratio for individual genetic risk factors. This is one of the most replicated findings in psychiatric epidemiology, and while not-vector borne, is another demonstration of infection related to schizophrenia.

Broader psychiatric meta-analysis: Sutterland AL, et al. “Beyond the association: Toxoplasma gondii in schizophrenia, bipolar disorder, and addiction: systematic review and meta-analysis.” Acta Psychiatrica Scandinavica, 2015; 132(3):161–179. PMID: 25877655.

Confirmed significant associations between T. gondii seropositivity and schizophrenia (OR 1.81), bipolar disorder (OR 1.52), OCD (OR 3.4), and addiction (OR 1.91), but not major depression.

Bipolar disorder / MDD misdiagnosis

Hirschfeld RM, et al. “Perceptions and impact of bipolar disorder: How far have we really come? Results of the national depressive and manic-depressive association 2000 survey.” Journal of Clinical Psychiatry, 2003. (69% initially misdiagnosed; 10+ years in >⅓ of cases)

Ghaemi SN, et al. “Diagnosing bipolar disorder and the effect of antidepressants: A naturalistic study.” Journal of Clinical Psychiatry, 2000; 61(10):804–808. (55% developed mania; 23% became rapid cyclers on antidepressants)

Wehr TA, et al. “Rapid cycling affective disorder.” Archives of General Psychiatry, 1988; 45(2):179–184. (51% of rapid cyclers associated with antidepressant use)

Shi L, et al. “Quality of Life Among Bipolar Disorder Patients Misdiagnosed With Major Depressive Disorder.” Psychiatric Services, 2004; 55(11). PMC1911165. (QoL worse in misdiagnosed patients)

Troponin and sex differences

Mills NL, et al. “Sex Differences in Cardiac Troponin Trajectories Over the Life Course.” Circulation, 2024. (Median troponin 2.4 ng/L women vs. 3.7 ng/L men)

Vogel B, et al. “Sex-specific differences in cardiac troponin: sensitivity of 56.1% in women vs. 70.1% in men.” International Journal of Cardiology: Heart & Vasculature, 2009.

Ferry AV, et al. “Prognostic Value of High-Sensitivity Cardiac Troponin in Women.” PMC, 2022.

Endometriosis

World Health Organization. “Endometriosis.” WHO Fact Sheet, March 2024. who.int (190 million affected; average 4–12 years to diagnosis)

Zondervan KT, et al. “Endometriosis.” New England Journal of Medicine, 2020; 382:1244–1256. (Leading peer-reviewed review of prevalence and diagnostic delay)

Davenport S, Smith D, Green DJ. Barriers to a Timely Diagnosis of Endometriosis: A Qualitative Systematic Review. Obstet Gynecol. 2023 Sep 1;142(3):571-583. doi: 10.1097/AOG.0000000000005255. Epub 2023 Jul 13. PMID: 37441792.

Tomassetti C, D’Hooghe T. Endometriosis and infertility: Insights into the causal link and management strategies. Best Pract Res Clin Obstet Gynaecol. 2018 Aug;51:25-33. doi: 10.1016/j.bpobgyn.2018.06.002. Epub 2018 Jun 18. PMID: 30245115.

La Rosa VL, Barra F, Chiofalo B, Platania A, Di Guardo F, Conway F, Di Angelo Antonio S, Lin LT. An overview on the relationship between endometriosis and infertility: the impact on sexuality and psychological well-being. J Psychosom Obstet Gynaecol. 2020 Jun;41(2):93-97. doi: 10.1080/0167482X.2019.1659775. Epub 2019 Aug 30. PMID: 31466493.

Subscribe now

I was inspired after listening to Mark Cuban talk about how he wants to f**k up healthcare in the US and how he keeps saying the quiet parts out loud. He’s a relative outsider and yet he has picked up on so many key points that insiders often don’t talk about publicly, for fear of being targeted. For years I’ve been told that not to talk too loudly about some of the financial misalignments in our healthcare system and how they impact patient care.

I’m a science nerd at heart; I love biology (just ask my poor high school sophomore son about my level of excitement in helping him study biology this year). I never thought 25 years into my career that I would be focused on health policy, payment, and ways to fundamentally break/change our healthcare system. But after spending so much time in the lab and seeing so little actually come to the clinic to impact patients- not because the data wasn’t good enough, but because the financial incentives weren’t aligned- I either had to shut up and stop complaining or do something to try to fix it.

I’m working to create a better understanding of how we got there and the financial incentives of our current healthcare system, because that is what drives decision making. I believe to fix something, you have to understand how it broke. Scientists and clinicians hate talking about this stuff, and so do I. Everyone wants to believe that the best science will shine through the BS and that everyone wants the best health outcomes for patients- despite almost no evidence of that happening at scale.

Healthcare costs in the US are almost 18% of GDP, that means almost 18% of money made in the US flows through our healthcare system in some way. That means there are a lot of people and institutions who are not financially incentivized to fix it, and when/if we really reduce our spending on healthcare, it will have to come from someone’s pocket, and no one wants it to be them.

First off, how did we get here?

A lot of the problems in US healthcare are rooted in how we pay for healthcare, so let’s start there. Right now we have a fragmented payment and payor system, based on your age, whether or not you work, and how much money you make. Roughly, this breaks down to:

Over 65 → Medicare (or Medicare Advantage)

Under 65, working or dependent on someone working → commercial health insurance through an employer, or insurance that is funded entirely by the employer

Income under 138% of the Federal Poverty Level → Medicaid

Commercial Health Insurance + Employers- how did this get linked?

This started all the way back in the 1940s, when the US was at war. The Stabilization Act of 1942 was a U.S. law that authorized the President to control wartime inflation by stabilizing prices, wages, and salaries at specific levels- which meant you couldn’t pay more or give bonuses to attract talent. The result was that for the first time, employers began providing health care benefits to attract workers when wages were frozen. This was when only about 10% of the population had any health insurance; people mostly just paid out of pocket when they visited their doctor or needed medications, costs were low and relatively simple.

Unions in particular latched on to this new way of legally increasing wages, and lobbied for it to explicitly not be subject to wage control. They were successful, and offering health insurance rapidly became something an employer could offer, at a cheap price, to attract talent. Then the unions lobbied for health insurance to be considerable non-taxable income and tax deductible for the business, to ensure that employers continued to offer this benefit. The IRS codified this in 1954, creating a significant tax benefit for employers to offer health insurance. (This tax incentive is why employers can’t just give employees the cash they would pay for premiums as a bonus, or supplement to buy whatever they want for healthcare without significant financial consequence.) The result was that by the 1960s, 80% of the population had health insurance. The people who were left out were older people who weren’t working, the sick, and the poor who worked in jobs that didn’t provide health insurance.

Medicare and Medicaid

In response to the growth in the commercial health insurance market and as part of the Social Security Act, Medicare was started in 1965, first only for people over 65, later adding on people who have specific long-term physical disabilities or end-stage renal disease. (Renal disease was made a specific Medicare “carve-out” (or dedicated entitlement) in 1972 because advancements in dialysis technology created life-saving treatment that was too expensive for most individuals, leading to a moral crisis where patients were denied care based on cost. Congress intervened, passing legislation allowing nearly all Americans with End-Stage Renal Disease (ESRD) to qualify for Medicare regardless of age. Of course, many conditions could be considered too-expensive now, but the carve-out has remained that same.)

Medicare is federally funded. Originally, it was just a government sponsored program, but in 1997 the Medicare + Choice program was started to give people options to choose commercial plans who would administer Medicare benefits for them (this is now called Medicare Advantage). In 2003, Part D was established to cover prescription drugs. More on Medicare Advantage later.

At the same time Medicare was started, Medicaid was also started as part of the Social Security Act to provide health insurance to people with low incomes. In 1997 it was expanded to include CHIP (Children’s Health Insurance Program), which covers children in families that make too much money for Medicaid but cannot afford commercial health insurance. Medicaid is funded at the state level, with some federal support. Because it’s funded at the state level, states have latitude to determine what they cover and so coverage varies widely from state to state. Medicaid covers roughly 50% of births and children under 5 and a disproportionate number of people in nursing homes or other intense care facilities, a smaller pot of money for the sickest population.

Medical records and coding

Along the way, we developed both a system of medical records, to track how patients are managed, and a coding system, to track how things are paid for. (If you really want to go deep into how much our coding system has created and enabled the financial misincentives of our healthcare system, I recommend listening to the Acquired podcast on Epic Systems.) The coding system enabled hospitals and physicians to bill insurance different amounts for different procedures and therapies. Over time, it evolved from enabling clinicians to get paid for the work they did to give control to the insurance company over what would be paid. Now, coverage dictates care. Physicians are reluctant to prescribe medications or perform procedures knowing the patient is going to be stuck with a large bill. They began to employ large teams to deal with ‘prior authorizations’ and denials, to manage ‘step therapy’ and often interfere with clinical decisions in a way that would have previously been unthinkable.

The Affordable Care Act (ACA)

As commercial insurance gained more control over the US healthcare system during the 1990s, health disparities grew. Patients who had genetic diseases were unable to obtain health insurance; patients who had a history of illness who reached their lifetime ‘cap’, effectively making them uninsurable for the rest of their lives. A baby who was born prematurely could reach their lifetime cap before they ever made it out of the hospital. The Clinton Administration attempted to pass bills for ‘universal healthcare coverage’, but they failed both as a result of lobbying from the insurance industry, who did not want to have to insure known sick people, and from a lack of a plan of how to balance the population.

In 2010, during the Obama Administration, the Affordable Care Act was signed into law.

The ACA had several genuinely significant accomplishments, which are easily taken for granted now. They include:

• Guaranteed issue and community rating ended the practice of denying coverage or charging higher premiums based on pre-existing conditions. This was transformative for people with chronic illness or health history who had previously been locked out of the individual market entirely.

• Elimination of lifetime and annual benefit caps protected people from the catastrophic situation of having insurance that simply ran out during a serious illness.

• Medicaid Expansion, raising the eligibility threshold to 138% of the federal poverty level and extending coverage to millions of low-income adults who previously fell into a coverage gap- too poor for marketplace subsidies, too “wealthy” for traditional Medicaid. States that expanded saw measurable drops in uninsured rates and improvements in access to preventive care.

• Dependent coverage to age 26 was one of the most popular and unambiguous wins — keeping young adults on parents’ plans during a period when they’re least likely to have employer coverage.

• Preventive care mandates required insurers to cover a range of preventive services without cost-sharing, including cancer screenings, vaccinations, and wellness visits. The theory was that upstream prevention reduces costly acute care downstream.

• The individual marketplaces created a structured way for people without employer coverage to shop for standardized plans, with income-based subsidies making coverage affordable for many for the first time.

Many of the ACA’s best outcomes, especially on coverage, were concentrated among specific populations (Medicaid-eligible adults, people with pre-existing conditions, young adults under 26). For people in the middle, particularly those with employer coverage or above the subsidy threshold in the individual market the ACA offered fewer direct benefits and in some cases contributed to higher premiums. This progress on access did not solve the deeper problems of cost and one clause in particular has resulted in an enormous, unprecedented increase in healthcare spending and cost- the Medical Loss Ratio (MLR).

What is the Medical Loss Ratio (MLR)?

If you’ve read this far, don’t leave now, because this is the crux of why healthcare spending is 18% of GDP. The ACA requires insurers to spend at least 80-85% of premium revenue on actual medical care (vs. administrative costs and profit). If they don’t, they must rebate the difference to customers. (How often do you think insurance companies want to pay back premiums? Not often.)

The idea was to cap insurer profits and ensure money flows to care. Since profit is capped as a percentage of premiums, the absolute dollar profit grows only as premiums grow. An insurer earning 15% margin on $1,000 in premiums makes $150. If premiums rise to $2,000, they make $300 - staying compliant. This gives insurers little reason to aggressively negotiate down prices with hospitals or drug companies, since higher spending means a higher premium base to earn margin on and report profits to Wall Street- since remember, most of these companies (United, Aetna, Humana, etc.) are publicly traded companies with a fiduciary responsibility to their shareholders.

Before the MLR, an efficient insurance company that reduced administrative costs dramatically could pocket those savings as profit or use it as a competitive tool to gain more market share. Under MLR rules, if you get too efficient on administration, you may fall below the 80% spending floor and be forced to rebate premiums. The regulation effectively penalizes operational efficiency and reduces incentives to lower the total cost of care. Utilization management (reviewing whether care is medically necessary, also known as prior authorization) is typically classified as an administrative cost. Since insurers must keep administrative costs low, the MLR creates pressure to reduce investment in utilization management, since that is the function that constrains unnecessary or fraudulent care.

The MLR rule treats healthcare spending as inherently good and administrative spending as inherently bad. But this isn’t true- some administrative spending (care coordination, fraud detection, utilization review) can reduce unnecessary medical costs and direct patients to better care. But by limiting the “bad” spending as a rule, it inadvertently inflated the “good” spending it was meant to ensure was used wisely.

How the MLR inflates pricing in healthcare

Health insurers make money on the spread between premiums and claims. Prior to the MLR, an insurance company had a strong incentive to drive down cost with hospitals, because every dollar saved in negotiated rates was a dollar that could become profit. Now, that incentive is largely gone.

If an insurer negotiates hospital prices down, then premiums follow (since premiums are priced to forecasted claims). A lower premium base means lower absolute profit, even at the same margin percentage. The insurer that fights hard for lower hospital rates is essentially working against its own revenue growth.

This creates a feedback loop with a powerful dynamic:

Hospitals raise prices (chargemaster rates and negotiated rates)

Insurers have incentive less resist, since higher claims support higher premiums

Higher premiums generate higher absolute profit within the MLR cap

Hospitals observe minimal resistance and raise prices further

Insurers blame the hospitals and clinicians for rising prices

This is a result of the financial misincentives of the US healthcare system. It isn’t a secret conspiracy or collusion; it is just the alignment of financial incentives based on the structure we have created. Because MLR compliance is calculated on gross premiums, increasing hospital costs feed directly into rising premiums, which feed the base on which insurers earn their capped margin.

The insurer’s profit mechanics looks like:

Profit = Premium Revenue × (1 - MLR floor)

Since maximizing profit means maximizing premium revenue, and the easiest way to do that passively is to allow medical costs to rise. Some economists would argue that before the MLR, insurers also had limited incentive to push back on large hospital systems due to market consolidation, and that the MLR didn’t change the market much. But before the ACA and the introduction of the MLR, United Healthcare’s share price was $30 (2010) and in 2025 it was $600, making it one of the best investments in healthcare (including in many 401ks).

Vertical integration compounds the effects of the MLR

But wait- there are even more misincentives that make this even worse. Specifically, vertical integration. In recent years, insurers have acquired hospital systems and physician groups, making payments to those owned entities count as medical spending toward the MLR. This means that:

The insurer can route dollars to its own hospital subsidiary, making money on the care and the premiums for coverage.

Those dollars count as “care spending” within the MLR, not as part of the allowed 15% profit (!).

The integrated entity extracts margin at the provider/hospital level instead of the insurer level, meaning they can charge much more and get paid for doing so.

The result? Hospital prices within the integrated system have zero downward negotiating pressure since both sides are the same company.

And those same groups have acquired pharmacies (controlling drug dispensing) and pharmacy benefit managers (PBMs, controlling drug tiering and rebates) to enable full profit control over every side of care. This is why there has been so much insurer-provider-PBM consolidation accelerated after the ACA.

Vertical integration is getting some scrutiny in DC and at the state level, but there is no regulation yet that will prohibit this from continuing.

In the US, we like to think of our market as the greatest capital market in the world. But capital markets require transparency, competition between sellers, and free market principles of supply and demand. In a true functioning capital market, insurers would compete for business by offering some narrow networks of lower-cost, higher-quality hospitals and passing savings to consumers through lower premiums. But the MLR removes any incentive to do this because lower-cost networks reduce premiums, shrinking the profit base, the administrative cost of building and managing tight networks counts against the MLR ratio, and there’s no reason for an insurer to do the hard work of selective contracting.

The Result of the MLR in the ACA

The MLR transformed insurers from astute price negotiators into passive premium collectors, with no incentive to reduce the total cost of care. Hospitals recognized that the entity on the other side of the negotiating table was financially incentivized by higher prices and responded accordingly. The resulting meteoric rise in prices in healthcare is the result of removing the insurer’s profit motive to push back on prices, which was a meaningful shift in the balance of power. And we are all paying the price.

****

This is the beginning; I will be writing about healthcare transformation and how we can get to the healthcare system we all want, lower cost of care, more precise care, more focus on prevention and diagnosis. Thanks for reading.

Subscribe now