July 2, 2026

By David Brunk, civil litigation attorney

When the Diagnosis Misses the Gene: Psychiatric Liability and the Shrinking Uninterpretable Defense

By David Brunk  ·  July 2, 2026  ·  Genetic Data

In 2012, a 21-year-old college student named Emilee Williams went to her physician with tremors, balance problems, difficulty concentrating, insomnia, and panic attacks. Her physician diagnosed her with fatigue and depression. Five months later, with symptoms worsening, the physician revised the diagnosis to anxiety. What Emilee actually had was Wilson's disease, a genetic copper metabolism disorder that is treatable when caught and catastrophic when missed. By the time the correct diagnosis arrived, she had suffered permanent brain damage. She now receives nutrition through a feeding tube.

A Missouri jury awarded her $28.9 million. The physician had not included neurological or genetic conditions in the differential diagnosis and had not ordered an MRI or a referral to a neurologist after the initial visit. Those were the steps that would have found Wilson's disease. They were not taken. Williams v. Mercy Clinic Springfield, No. 1531-CC00861-01 (Greene Co., Mo. 2017).

That case is nine years old, and it turns on a familiar principle: the duty to keep an open differential diagnosis when a patient's psychiatric symptoms could have an organic cause. What has shifted since is less the legal principle than the landscape around it, as more genetic information becomes interpretable and more testing becomes available.

The Interpretability Problem and Why It Is Shrinking

For years, a physician could defend a failure to order genetic testing partly on the ground that the results would not have been actionable. Clinical genetic testing regularly returned variants of uncertain significance, genetic findings with no established clinical meaning. A VUS in 2019 was genuinely ambiguous. Acting on it could cause harm. Not acting on it was defensible.

That defense is narrowing. AI-powered variant interpretation tools now classify variants automatically, applying machine learning to ACMG criteria to move VUS results toward pathogenic or benign determinations. QIAGEN acquired Genoox, maker of the Franklin variant-interpretation platform, for $70 million in 2025 specifically for this capability. VarSome, Genebe, and comparable tools apply the same automated approach. What was genuinely uninterpretable five years ago may now have a clear clinical interpretation available in a commercial database.

The legal implication is worth considering. A physician who ordered a panel in 2020, received a VUS result, and took no action was on solid ground. Whether that same physician stands differently in 2026, when the variant has since been reclassified as pathogenic by commercially available tools, is a harder question. The information existed, and the interpretation later became available, but whether the physician had a duty to learn of the reclassification and recontact the patient is unsettled law. Legal scholars are debating it: a 2022 Fordham Law Review analysis proposed assigning the duty to reinterpret variants to laboratories and the duty to recontact patients to ordering physicians, and the broader question has been discussed in the genetics and law literature since 2019. No court has yet imposed such a duty, and whether one should exist remains genuinely contested.

Beyond Drug Dosing: Genetic Workup as Diagnostic Standard

The pharmacogenomics conversation, focused on whether a prescriber should test for CYP2D6 or CYP2C19 before choosing a psychiatric medication, is well-established and worth having separately. But the liability question extends further.

Wilson's disease is one example. Anti-NMDA receptor encephalitis is another. This autoimmune condition, with genetic susceptibility factors, presents as acute psychiatric illness: psychosis, personality change, catatonia. It is routinely misdiagnosed as schizophrenia or bipolar disorder. When the correct diagnosis is established by antibody testing, patients have often spent months institutionalized and treated with antipsychotics that did not help and may have worsened the condition.

22q11.2 deletion syndrome is a third. Carriers of this chromosomal deletion have approximately a 30% lifetime risk of developing psychotic illness, plus cardiac defects, immune deficiency, and developmental abnormalities. For a young patient with first-episode psychosis, 22q11.2 is the kind of finding that can materially change the care plan: the treatment approach, the cardiac monitoring, the family counseling, and the prognosis discussion.

In pediatric psychiatry, the American College of Medical Genetics and the American Academy of Pediatrics both recommend chromosomal microarray analysis as first-tier testing for any child presenting with autism spectrum disorder or intellectual disability. Fragile X testing is essentially standard of care for males in that presentation. A behavioral diagnosis of autism that does not include that genetic workup is arguably incomplete under current professional guidelines.

The Pharmacogenomics Layer

For prescribing specifically, the framework is now substantial. The Clinical Pharmacogenomics Implementation Consortium has published gene-drug guidelines covering 34 genes and 164 drugs, with specific recommendations for antidepressants based on CYP2D6 and CYP2C19 metabolizer status. The FDA's Table of Pharmacogenetic Associations includes actionable markers for citalopram, pimozide, carbamazepine, and more than a dozen other commonly prescribed psychiatric drugs. The citalopram situation is the sharpest example: the FDA label already caps the standard maximum dose at half for poor CYP2C19 metabolizers, due to QTc prolongation and fatal arrhythmia risk.

The American Psychiatric Association's 2024 workgroup report on pharmacogenomic testing tools remains the primary defense-favorable authority in this space. It concluded that current evidence does not support routine combinatorial PGx testing for antidepressant selection. But two things about that report matter. First, the APA initially published it labeled as a "position statement" and subsequently corrected it to a "workgroup report," a distinction that narrows its authority as an official institutional shield. Second, the report addresses routine combinatorial testing. It does not address the specific gene-drug pairs that carry FDA label language, where the case for testing is stronger.

Commercial products are available, inexpensive, and marketed directly to psychiatric prescribers. GeneSight and Genomind's Mental Health Map analyze a swab sample, return results in days, and flag dozens of psychiatric medications by gene-drug interaction category. Automated variant interpretation is now being layered onto platforms like these. Arguments that testing was unavailable or impractical carry less weight than they once did.

Oregon's Loss-of-Chance Doctrine

Oregon's adoption of the loss-of-chance doctrine in Smith v. Providence Health & Services, 361 Or 456, 393 P3d 1106 (2017), is a consideration for practitioners in this state. A plaintiff need not prove that earlier testing would have produced a better outcome, only that the failure to test reduced the probability of one. A 33% lost chance was sufficient in Smith. In cases where a genetic condition went undiagnosed while a patient received other treatment, quantifying that lost probability would be a matter for expert testimony.

Where Things Stand

No Oregon court has held a psychiatrist liable for failure to order genetic testing, and whether genetic testing belongs in any particular psychiatric workup remains genuinely unsettled. Reasonable clinicians disagree, the professional guidance points in more than one direction, and the science is moving quickly. What can fairly be said is that several of the pieces such a claim would draw on are now in place: clinical guidelines for specific gene-drug pairs, commercially available testing, improving interpretation of results, an active scholarly debate about the duty to recontact, and, in Oregon, the loss-of-chance doctrine as a causation framework.

None of that establishes a standard of care, and this article does not propose one. The more modest point is that these are developments worth following. A clinician who understands where the science and the law are heading is better positioned to make deliberate choices about a diagnostic workflow, and to explain them, than one who first meets the question in the middle of a claim.

The allegations described here are taken from the filing and remain unproven; no responsive pleading is reflected in the source document.

David Brunk is a civil litigation attorney. He can be reached at david@newmanbrunk.com.