Theranos: The Fraud That Bypassed Scientific Peer Review Entirely

In the summer of 2014, Theranos — a Palo Alto blood-testing startup founded eleven years earlier by a Stanford dropout named Elizabeth Holmes — was reported to be worth approximately $9 billion. Holmes herself, on paper, was a billionaire several times over, and was being covered in Fortune, Forbes, The New Yorker, and Wired as the youngest self-made female billionaire in the world and a generational successor to Steve Jobs. Her board of directors included two former U.S. Secretaries of State (George Shultz and Henry Kissinger), a former Secretary of Defense (William Perry), two former U.S. Senators (Sam Nunn and Bill Frist), a retired four-star general (James Mattis), and a former CEO of Wells Fargo (Dick Kovacevich). The company had announced retail partnerships with Walgreens and Safeway to roll out its blood-testing technology to consumers nationwide.

The central technical claim that justified all of this — the entire reason Theranos existed as a company — was that its proprietary “Edison” device could perform hundreds of laboratory blood tests from a single finger-prick drop of capillary blood, replacing the traditional venous-draw, multi-vial, central-laboratory workflow that had defined clinical chemistry for a century.

That central claim had never been published in a peer-reviewed journal. It had never been independently replicated by anyone outside the company. The actual performance of the Edison device was a “trade secret” that Holmes refused to disclose to journalists, regulators, or even, in many cases, to the partners and investors who were committing hundreds of millions of dollars on the assumption that the technology worked.

Eighteen months later, in October 2015, John Carreyrou’s reporting in The Wall Street Journal would document that the technology did not work as claimed and had not for years. Two and a half years after that, in March 2018, the U.S. Securities and Exchange Commission would charge Holmes and former Theranos president Ramesh “Sunny” Balwani with “massive fraud.” In January 2022, a federal jury in the Northern District of California would convict Holmes on four counts of wire fraud and conspiracy to commit wire fraud against Theranos investors. In November 2022, U.S. District Judge Edward Davila would sentence her to 135 months — eleven and a quarter years — in federal prison.

This is the story of how a company built a multi-billion-dollar valuation around scientific claims that had never been subjected to the most basic verification any laboratory science requires — and what every strategist evaluating biotech investments, healthcare technology, or any “proprietary algorithm” claim should learn from it.

Why This Case Belongs In A Replication Crisis Series

Theranos is not, strictly speaking, an academic-replication-crisis case. There were no retracted journal papers. There was no formal misconduct finding from a university research-integrity committee. The participants were not, primarily, academic researchers — they were Silicon Valley executives, venture investors, board members, and clinical-laboratory employees.

But the underlying epistemic failure is the same one the replication crisis exposed: the discovery that the verification infrastructure people assume exists, often does not exist. In academic psychology, the replication crisis exposed that prestige journals, peer review, and institutional reputation were not, in fact, the strong filter on truth that practitioners and consumers of the literature had assumed they were. In the Theranos case, the failure is the mirror image: a company managed to acquire all the social trappings of scientific legitimacy — Stanford pedigree, board prestige, mainstream press coverage, Fortune 500 partnerships, regulatory contacts — without ever submitting its core scientific claims to the verification processes those trappings are normally taken to imply.

The lesson generalizes far beyond medical devices. Any time a vendor’s central technical claim is described as “proprietary,” “trade secret,” or “competitively sensitive,” and the claim has not been independently verified by a party with the technical capacity and the structural incentive to verify it honestly, the strategist’s posture should be that the absence of verification is itself the most important piece of evidence available. That is what Theranos teaches.

Who Holmes Was And What She Claimed

Elizabeth Anne Holmes was nineteen years old in March 2004 when she dropped out of Stanford University’s chemical engineering program to found Real-Time Cures, a company she would shortly rename Theranos (a portmanteau of “therapy” and “diagnosis”). She had been a freshman in Channing Robertson’s chemical engineering lab the previous summer, working on patch-based drug-delivery and biosensor concepts. Her founding pitch was that the same kind of microfluidic chemistry that her lab work had touched could be miniaturized into a consumer-scale device that would perform a complete diagnostic blood panel from a few microliters of capillary blood.

The technical claim, restated in language a non-specialist can evaluate, is the following. A conventional clinical blood test requires drawing several milliliters of venous blood (a “tube of blood” from your arm), shipping it to a central laboratory, and running it through one or several large analyzers — instruments the size of refrigerators that use spectrometry, immunoassay, or hematology techniques that have been developed and validated over decades. Theranos claimed it could perform the same battery of tests — and ultimately a menu of more than two hundred different tests — from a single finger-prick drop of capillary blood, in a small countertop device that could sit in a Walgreens or a doctor’s office.

If true, this would have been a substantial advance in clinical chemistry. The volume scaling is genuinely difficult — many assays require a minimum analyte concentration that becomes harder to detect as sample volume shrinks. Capillary blood from a finger-prick has known compositional differences from venous blood (different ratios of plasma to red cells, contamination from tissue fluid, hemolysis from the squeeze required to express the drop) that can produce systematic measurement errors for many analytes. The traditional clinical-chemistry literature documents these issues at length. None of this is secret. It is the reason the field has standardized on venous draws despite the patient inconvenience.

What Theranos claimed it had done was solve these problems through proprietary microfluidic engineering. The specifics of how — what assay chemistry, what sample-preparation steps, what analyte-detection methods, what validation protocols — were the trade secret. The company would not describe them. Holmes, in interviews and on stage, would deflect technical questions by appealing to the proprietary nature of the technology and the standard Silicon Valley posture that revealing implementation details would compromise competitive position.

By 2014, Theranos had been operating in this posture for eleven years. The company had raised more than $400 million in private investment. It had received CLIA certification for its laboratory operations (a federal certification that the laboratory met certain procedural standards, but not a validation of the accuracy of any specific test). It had begun offering tests to consumers at Walgreens locations in Arizona and California. It had not, in those eleven years, published a single peer-reviewed paper describing the Edison device’s analytical performance.

The Peer-Review Evasion Mechanism

The absence of peer-reviewed publication is, on inspection, not an accident or a competitive choice. It was the central structural feature that allowed the fraud to continue.

In normal clinical chemistry, a new analytical method — a new way of measuring an analyte, or a new device that performs an established measurement — is validated by a sequence of steps that produce a public scientific record. Method-comparison studies compare the new method’s results against an established reference method on hundreds or thousands of samples spanning the clinically relevant analyte range, reporting bias (systematic offset) and precision (random variability) using standardized statistical procedures (typically following Clinical and Laboratory Standards Institute, CLSI, guidelines such as EP09). Interference studies test whether common substances in blood (lipids, hemoglobin from hemolysis, bilirubin, common drugs) cause measurement errors. Sample-stability studies establish how the analyte degrades between collection and measurement. Limit-of-detection and limit-of-quantitation studies establish the lowest concentrations the method can reliably measure. The complete validation package is typically submitted to peer-reviewed journals like Clinical Chemistry, Clinical Chemistry and Laboratory Medicine, Clinical Biochemistry, or Journal of Clinical Pathology, where reviewers with subject-matter expertise scrutinize the methodology before publication.

This is not optional. It is how the field establishes whether a new measurement method actually works. A clinical laboratory that is going to use a new analyzer to make patient-care decisions is supposed to read the published validation literature, ideally perform its own on-site verification using the published protocols, and only then deploy the method for patient use. The published validation is the substrate of the entire trust system.

Theranos’s central claim — that the Edison device could measure hundreds of analytes from a finger-prick drop — required exactly this kind of validation for each individual assay on the menu. None of it had been published. The company instead relied on a sequence of substitute legitimization mechanisms:

Selective disclosure to non-specialist press. Holmes gave extensive interviews to general-interest publications (The New Yorker’s December 2014 profile by Ken Auletta, Fortune’s June 2014 cover by Roger Parloff, Forbes’s “Bloody Amazing” cover, and many others) whose journalists were not, with the exception of a few science specialists, in a position to ask the kinds of technical questions a clinical chemist would have asked. The resulting coverage was overwhelmingly favorable and treated the technology’s existence as established fact rather than as a claim that required verification.

Recruitment of high-prestige non-specialist endorsers. The board of directors was structured around political and military prestige, not biomedical expertise. Shultz, Kissinger, Perry, Nunn, Frist (who was a physician, but a cardiothoracic surgeon, not a clinical chemist), Mattis, Kovacevich, and former Secretary of Defense William Perry brought enormous social legitimacy to the company. None of them had the technical capacity to evaluate whether the Edison device’s analytical claims were true. Their presence on the board was nevertheless treated by outside observers as a signal of credibility — the implicit reasoning being that people of this stature would not lend their reputations to a company whose technology did not work.

Lab-developed-test regulatory loophole. Under U.S. regulatory practice, clinical laboratories operating under CLIA certification have historically been permitted to develop and offer “laboratory-developed tests” (LDTs) without prior FDA approval of the specific assay, on the theory that the laboratory itself, operating under CLIA quality standards, is responsible for validating its own methods. The FDA in 2014–2015 began moving toward more active oversight of LDTs, in part in response to Theranos and similar cases, but for most of Theranos’s operating history, the company could legally offer tests to consumers without the kind of FDA premarket review (510(k) clearance, de novo classification, or premarket approval) that would have required submission of analytical-validation data to a federal reviewer.

Trade-secret framing of validation requests. When journalists, regulators, or laboratory professionals asked to see the validation data for specific assays, Theranos consistently invoked the proprietary, trade-secret nature of the technology. This framing — common and largely accepted in software and consumer-product industries — was applied to a domain (clinical-chemistry validation) where it had no historical precedent and where the substantive question (does this measurement method actually produce correct results?) cannot, in principle, be answered without disclosing enough of the methodology to allow scrutiny.

Internal suppression of scientific concerns. Carreyrou’s reporting and the subsequent trial evidence documented a pattern in which Theranos employees who raised internal concerns about the Edison device’s performance — including Ian Gibbons, the company’s chief scientist who had been with the company since near its founding, and a number of junior scientists and laboratory technicians — were marginalized, fired, threatened with litigation under non-disclosure agreements, or, in Gibbons’s case, drove him to suicide in May 2013 the night before he was scheduled to give a deposition in patent litigation. The pattern of internal suppression meant that the people best positioned to know that the technology did not work were precisely the people the company’s legal apparatus was most aggressively pressuring not to say so.

The composite effect of these mechanisms was that Theranos was able to acquire the appearance of scientific legitimacy without ever being subjected to the substance of scientific validation. The Edison device’s actual analytical performance — the bias and precision of each assay, the interference profile, the comparison against reference methods — was, on Carreyrou’s reporting and the subsequent CMS and trial evidence, much worse than the company’s marketing claimed, sometimes dramatically so, on a substantial fraction of the test menu. Patients were receiving results that did not match their actual analyte concentrations, with potential consequences for diagnosis and treatment.

The Diamandis 2015 Academic Critique

The first sustained scientific critique of Theranos in a peer-reviewed venue came not from American clinical chemists — most of whom had no access to the company’s validation data and therefore had nothing concrete to evaluate — but from a senior Canadian clinical biochemist who was prepared to write what amounted to a scientific position statement on the basis of the public record.

Eleftherios P. Diamandis, head of clinical biochemistry at Mount Sinai Hospital in Toronto and a professor at the University of Toronto, published an editorial in Clinical Chemistry and Laboratory Medicine in June 2015 — four months before the Carreyrou Wall Street Journal piece — titled “Theranos phenomenon: Promises and fallacies” (Diamandis, 2015, Clinical Chemistry and Laboratory Medicine, 53(7), 989–993, https://doi.org/10.1515/cclm-2015-0356).

Diamandis’s editorial laid out, in measured language, exactly the case a working clinical chemist would have made if a working clinical chemist had been listened to. He noted that capillary blood from finger-prick samples is compositionally different from venous blood in ways that affect a substantial number of common analytes, that the volume scaling claimed by Theranos was extremely difficult in principle for many assays, that the company had not published any of its method-comparison or analytical-performance data, and that the absence of such data made independent evaluation impossible. He explicitly compared the situation to the way scientific claims are normally vetted and noted that Theranos was operating outside the normal verification framework that the entire field of clinical chemistry depends on.

The editorial was not received with universal enthusiasm at the time. Theranos’s response was hostile. The general-interest press largely ignored it — Clinical Chemistry and Laboratory Medicine is a specialist journal that consumer business reporters do not follow. But within the clinical-chemistry community, the editorial circulated and was taken seriously. A follow-up editorial by Mario Plebani, head of laboratory medicine at the University of Padua and the editor-in-chief of the journal, was published in November 2015 — three weeks after Carreyrou’s first WSJ article — under the title “The Theranos saga and the consequences” (Plebani, 2015, Clinical Chemistry and Laboratory Medicine, 53(11), 1799–1800, https://doi.org/10.1515/cclm-2015-1118).

Plebani’s editorial made the structural point explicitly. The scientific community had been raising concerns about Theranos for months before the consumer press caught up. The mechanism by which the concerns were finally heard was investigative journalism, not peer-reviewed scientific exchange — because the company had successfully positioned itself outside the institutions where peer-reviewed scientific exchange would have constrained it. The lesson Plebani drew was that the field’s verification mechanisms had been bypassed not by failure of those mechanisms but by deliberate routing around them, and that the institutional reforms needed in response should focus on ensuring that companies offering clinical tests to patients cannot, in practice, evade the validation requirements that the science of clinical chemistry has developed for very good reasons.

These two editorials are the canonical academic-literature treatment of the case. They are short, accessible, written by senior figures in the relevant field, published in a serious peer-reviewed journal, and freely citable. A strategist asking “what did actual subject-matter experts think about Theranos at the time, on the basis of the public record alone” can read them in an hour and have a substantively better answer than the contemporary popular press provides.

The Carreyrou Reporting

The reporting that broke the case publicly is one of the most consequential pieces of business investigative journalism of the twenty-first century. John Carreyrou, then a senior reporter at The Wall Street Journal, began investigating Theranos in early 2015 after receiving a tip from Adam Clapper, a Missouri pathologist who had been reading the Theranos press coverage and was skeptical of the technical claims on the basis of his clinical-laboratory experience.

Carreyrou’s first major article, “Hot Startup Theranos Has Struggled With Its Blood-Test Technology,” ran in The Wall Street Journal on October 16, 2015 (Carreyrou, 2015, Wall Street Journal, October 16). The article documented, on the basis of interviews with current and former Theranos employees, internal documents, and outside laboratory professionals, that the Edison device was being used for only a small fraction of the tests Theranos offered, that the company was running most of its tests on conventional commercial analyzers (Siemens, Beckman Coulter, and others) that it had quietly purchased and installed in its laboratory, that proficiency-testing results for the Edison-run tests were poor, and that the company had been issuing patient test results based on technology whose accuracy the company’s own internal scientists had questioned.

The reporting continued through 2015 and 2016 with a series of follow-up articles documenting additional failures, regulatory actions, and corporate responses. Theranos initially mounted a vigorous public defense, denying the reporting, threatening litigation against sources, and attacking Carreyrou personally. The defense did not hold. The Centers for Medicare & Medicaid Services (CMS) inspected the Theranos Newark laboratory in late 2015 and in January 2016 issued findings that the laboratory’s operations posed “immediate jeopardy to patient health and safety.” In July 2016 CMS revoked the laboratory’s CLIA certification and barred Holmes from owning or operating a clinical laboratory for two years. Walgreens terminated the partnership and closed all Theranos blood-draw locations. The company laid off most of its staff in 2017 and 2018 and formally dissolved in September 2018.

Carreyrou’s book-length account, Bad Blood: Secrets and Lies in a Silicon Valley Startup (Carreyrou, 2018, Knopf), is the canonical narrative treatment. It is the single best entry point for understanding the full arc of the case and is the basis of nearly every subsequent popular treatment, including the Alex Gibney HBO documentary The Inventor: Out for Blood in Silicon Valley (2019), the Hulu series The Dropout (2022), and the ABC News podcast The Dropout (2019).

The book makes clear, in a way that the journal editorials could not, the human texture of how the fraud was sustained over more than a decade — the suppression of internal dissent, the use of non-disclosure agreements and litigation threats against employees, the deliberate isolation of subsystems within the company so that no one employee had a complete picture of how poorly the technology was working, the pattern of escalating commitments by board members and investors who had committed too much to back away. It is, beyond its specific subject, one of the best available studies of how a company can construct and maintain a parallel reality at the scale of hundreds of employees and hundreds of millions of dollars when the central technical claim of that reality is false.

The 2022 Conviction And 2025 Resolution

The federal criminal case, United States v. Elizabeth A. Holmes, was filed in the Northern District of California in 2018 (case number 18-cr-258). After multiple delays and a long pretrial process, Holmes’s jury trial began in August 2021 before Judge Edward Davila in the San Jose courthouse.

The trial ran for fifteen weeks. The government’s case, led by Assistant U.S. Attorneys Robert Leach, Jeffrey Schenk, Kelly Volkar, and John Bostic, presented testimony from former Theranos employees (including Erika Cheung, one of the original whistleblowers to Carreyrou; Adam Rosendorff, the lab director who had raised internal concerns; and Tyler Shultz, grandson of George Shultz and another Carreyrou source), former patients whose erroneous Theranos results had affected their medical care, and investors who testified to the misrepresentations on which they had committed funds. The defense, led by Kevin Downey and Lance Wade of Williams & Connolly, argued that Holmes had been a true believer in the technology, had been misled by Balwani (who had been her romantic partner during much of the relevant period), and lacked the criminal intent required for fraud conviction.

On January 3, 2022, the jury returned its verdict. Holmes was convicted on four counts: one count of conspiracy to commit wire fraud against Theranos investors, and three counts of substantive wire fraud against specific investors (Lakeshore Capital Management’s Brian Grossman, PFM Health Sciences, and Black Diamond Ventures). She was acquitted on four counts relating to patient-facing fraud and one investor count; the jury was unable to reach a verdict on three additional investor counts, which the government later declined to retry.

On November 18, 2022, Judge Davila sentenced Holmes to 135 months in federal prison and ordered $452 million in restitution. She reported to Federal Prison Camp Bryan in Texas on May 30, 2023. Balwani was tried separately and convicted in July 2022 on all twelve counts he faced; he was sentenced to 155 months and reported to prison in April 2023.

The legal record from the Holmes trial is the most authoritative public documentation of the substantive failures of the Edison technology. The trial transcripts, exhibits, and post-trial filings establish what the company knew internally about the device’s poor performance, when it knew, what it told investors and partners and patients, and how those representations diverged from the documented internal reality. For a strategist evaluating the case as a study in scientific-validation evasion, the trial record is the primary source and supersedes the popular-press accounts (which, while substantially accurate, lack the evidentiary precision of sworn testimony and authenticated documents).

Lessons For Healthcare And Biotech Investors

The Theranos case generates several lessons that are specific to the healthcare and biotech investment domain and that are directly actionable for someone evaluating an early-stage diagnostic, therapeutic, or medical-device company.

Independent analytical validation, published in a peer-reviewed venue, is the minimum acceptable evidence for a diagnostic or measurement claim. The fact that the company has CLIA certification, FDA clearance for a different product, or a CE mark in Europe does not substitute. The specific claim (“our device measures analyte X with accuracy Y at sample volume Z”) must be supported by a published validation study following standard CLSI or equivalent protocols. If no such study exists, the claim has not been validated, regardless of how confident the company appears. The investor’s posture should be that the absence of published validation is itself the most important data point.

Trade-secret framing is not an acceptable substitute for analytical validation. The published validation literature in clinical chemistry does not require disclosure of proprietary engineering implementation details. It requires disclosure of analytical performance characteristics — bias, precision, interference profile, linearity, limit of detection. A company that conflates these two things (refusing to disclose performance because performance disclosure would supposedly compromise engineering trade secrets) is either making a category error or is deliberately confusing the two to evade scrutiny. Either way, the conflation should trigger skepticism, not acceptance.

Prestige boards do not validate science. Theranos had Shultz, Kissinger, Perry, Nunn, Frist, Mattis, and Kovacevich. Their presence did not, in retrospect, mean the technology worked. It meant that the company had the social skills and resources to recruit prestigious people who lacked the technical capacity to evaluate the central claim. When you see a board structured around political or business prestige rather than subject-matter expertise in the company’s core technical domain, treat the board as a marketing asset, not an evidence asset.

The most informative voices are people with the technical capacity and the structural incentive to evaluate honestly. In the Theranos case, those voices were Diamandis (a senior clinical biochemist with no financial stake in Theranos), Plebani (a senior laboratory-medicine figure in the same position), Ian Gibbons (the company’s chief scientist, whose internal concerns were suppressed), the laboratory employees who would later become Carreyrou’s sources, and the working pathologists who responded to the early popular-press coverage with skepticism on technical grounds. The voices that were less informative were the consumer business press, the political and military prestige figures on the board, the venture investors who were committing capital on the basis of pattern-matching rather than diligence, and the Stanford engineering faculty who had endorsed Holmes early without independent evaluation. Investors evaluating future biotech opportunities should weight subject-matter expert opinion (especially when it is skeptical) above generalist coverage and prestige endorsement.

Pattern-match for fraud-favoring structural conditions, not for fraud itself. It is rarely possible for an outsider to directly verify whether a specific company’s claims are true. It is more often possible to evaluate whether the structural conditions that allow fraud to be sustained are present. The Theranos pattern — proprietary trade-secret framing of central scientific claims, absence of peer-reviewed publication, prestige board lacking subject-matter expertise, aggressive non-disclosure-agreement and litigation posture toward employees, founder cult-of-personality coverage, regulatory-loophole operation — recurs across cases. When a substantial number of these conditions are present, the probability of substantive verification failure is materially elevated, even if no individual condition is by itself disqualifying.

Generalizing Beyond Medical Devices: The “Proprietary Algorithm” Problem

The most important generalization from the Theranos case is to a much wider class of vendor claims that have nothing to do with medical devices specifically. The structural problem — a vendor offering a technical product whose central performance claim cannot be independently verified because the implementation is described as proprietary — recurs across many industries and is, in 2026, considerably more prevalent than it was when Theranos was operating.

Examples of the recurring pattern that strategists should treat as Theranos-adjacent:

Vendor “AI” and machine-learning systems whose performance claims rest on internal benchmarks. A vendor claims its model achieves performance X on task Y, but the benchmark is proprietary, the test set is undisclosed, and independent reproduction is impossible. The claim may be true. It may also be false. In the absence of an independent benchmark with public data, an independent test set, and reproducible methodology, the strategist has no evidential basis for distinguishing the two cases.

Algorithmic decision systems in employment, lending, or criminal-justice contexts whose accuracy and bias profile are protected as trade secret. The vendor sells a tool that makes consequential decisions, refuses to disclose how it works or how it has been validated, and asks the buyer to trust internal performance claims that no third party has audited. This is structurally identical to the Theranos posture and should be evaluated with the same skepticism.

Proprietary research methodologies sold as competitive intelligence, market research, or consumer-behavior insight. A consultancy or research firm offers a methodology whose key analytical steps are proprietary, whose validation against external benchmarks has not been published, and whose results cannot be independently reproduced. The claim that the methodology works rests on the firm’s reputation and the testimonials of past clients, not on a transparent evidence record.

Corporate internal research whose conclusions are used to justify strategic decisions but has never been subjected to peer review or independent replication. The firm’s analytics team produces a study showing that intervention X causes outcome Y, the methodology is internal, the data is internal, the analysis is internal, and the conclusion is treated as established. The same skepticism that a working clinical chemist would have applied to Theranos’s Edison claims should be applied here.

In each of these cases, the strategist’s posture should be the same as in the Theranos case. Absence of independent verification is the most important piece of evidence available. Trade-secret framing of validation requests is a red flag, not a competitive moat. Prestige and pedigree are not substitutes for verifiable performance. And the most useful diagnostic question is not “does the vendor seem confident?” or “is the team credentialed?” — it is “what would constitute disconfirming evidence, and what is the structural reason no party with the capacity to produce that evidence has done so?”

Strategist Takeaway

If you are using a vendor, an internal research finding, or a proprietary algorithm to justify a decision that materially matters, ask the following question and treat your inability to answer it as evidentially significant: What independent party, with the technical capacity to evaluate the central claim and no financial interest in the answer, has done so, and what did they find?

If the answer is “no one,” the central claim is not validated. It may still be true. But the rational treatment of an unvalidated claim is not to act as if it were validated. It is to size your reliance on the claim to reflect the strength of the underlying evidence, and to actively seek out the kind of independent verification that would justify higher reliance.

The Theranos case is the canonical demonstration of what happens when this discipline is not applied at scale — when a multi-billion-dollar company can be built on a central technical claim that no one with the relevant expertise and the relevant access has independently verified, and when the social trappings of legitimacy (prestige board, mainstream press coverage, Fortune 500 partnerships, regulatory contacts) are allowed to substitute for the substantive verification those trappings are normally taken to imply. The cost when the substitution fails — to patients, employees, investors, partners, and, in the largest cases, to the credibility of entire industries — is high enough that the discipline of demanding independent verification is, in the end, much cheaper than the alternative.

Read Diamandis 2015. Read Carreyrou 2018. Read the Holmes trial record. And then ask the question above about every vendor claim, every internal research finding, and every proprietary algorithm that is currently being used to justify a decision you would not want to be wrong about.

Sources

• Carreyrou, J. (2015, October 16). Hot startup Theranos has struggled with its blood-test technology. The Wall Street Journal. https://www.wsj.com/articles/theranos-has-struggled-with-blood-tests-1444881901
• Carreyrou, J. (2018). Bad Blood: Secrets and Lies in a Silicon Valley Startup. New York: Alfred A. Knopf.
• Diamandis, E. P. (2015). Theranos phenomenon: Promises and fallacies. Clinical Chemistry and Laboratory Medicine, 53(7), 989–993. https://doi.org/10.1515/cclm-2015-0356
• Plebani, M. (2015). The Theranos saga and the consequences. Clinical Chemistry and Laboratory Medicine, 53(11), 1799–1800. https://doi.org/10.1515/cclm-2015-1118
• United States v. Elizabeth A. Holmes, 18-cr-258 (N.D. Cal.). Verdict January 3, 2022; sentencing November 18, 2022 before Judge Edward J. Davila. Court documents available via PACER.
• United States v. Ramesh “Sunny” Balwani, 18-cr-258-2 (N.D. Cal.). Verdict July 7, 2022; sentencing December 7, 2022.
• U.S. Securities and Exchange Commission. (2018, March 14). Theranos, CEO Holmes, and former president Balwani charged with massive fraud. SEC Press Release 2018-41. https://www.sec.gov/news/press-release/2018-41
• Centers for Medicare & Medicaid Services. (2016, January 25). Statement of deficiencies and plan of correction, Theranos Inc., Newark, CA laboratory (CCN 05D2025714). CMS Form 2567.
• Auletta, K. (2014, December 15). Blood, simpler. The New Yorker. https://www.newyorker.com/magazine/2014/12/15/blood-simpler
• Parloff, R. (2014, June 12). This CEO is out for blood. Fortune. https://fortune.com/2014/06/12/theranos-blood-holmes/
• Gibney, A. (Director). (2019). The Inventor: Out for Blood in Silicon Valley [Film]. HBO Documentary Films.
• Clinical and Laboratory Standards Institute. (2018). EP09-A3: Measurement Procedure Comparison and Bias Estimation Using Patient Samples; Approved Guideline — Third Edition. CLSI.

Related

• The Replication Crisis hub — the full set of cases, methods, and decision frameworks for strategists evaluating “research-backed” claims about science, business, and human behavior.
• Diederik Stapel: The 58-Retraction Fraud That Reshaped Social Psychology — the canonical academic-fraud case in social psychology. Where Theranos evaded peer review entirely, Stapel exploited the trust structure within peer review to publish fabricated data in the field’s top journals for more than a decade.
• Michael LaCour And The Contact-Hypothesis Fraud That Was Caught By Replicators — a 2014 Science paper on attitude change toward marriage equality that turned out to be based on fabricated survey data, exposed when graduate students attempted to replicate the methodology and discovered the underlying survey company had never conducted the work.
• Brian Wansink And The Mindless Eating Lab — a Cornell food-science lab whose pattern of methodological misconduct resulted in 18 retractions and the researcher’s resignation. The Wansink case illustrates how p-hacking and HARKing can produce a sustained publication record from data that does not support the claims.
• Marc Hauser And The Harvard Cognition Lab — the 2010 Harvard misconduct finding against an endowed psychology professor, an early major modern fraud case that foreshadowed the replication crisis the field would more fully confront beginning in 2011.
• James Vicary And The Subliminal Advertising Fraud Of 1957 — the earliest case study in this hub of a commercial fraud built around a non-replicable scientific claim. Vicary’s claimed subliminal-advertising effects were never validated and the experimenter eventually admitted he had fabricated the original data.

FAQ

Was Theranos actually a fraud or just a failed startup?

It was a fraud, established to the criminal standard by a federal jury verdict. Failed startups happen routinely — a company can pursue a technology that turns out not to work, exhaust its capital, and dissolve without anyone having committed a crime, provided the company’s representations to investors, partners, and customers were honest at the time they were made. The Theranos case was different in two respects established by the trial evidence. First, the company’s representations to investors about its technology’s performance, regulatory status, and commercial traction were knowingly false at the time they were made, not aspirational predictions that did not pan out. Second, the company’s representations to patients about the accuracy of the blood-test results it was issuing were knowingly false in cases where the company’s internal data showed the results were unreliable. The jury convicted Holmes specifically on the investor-fraud counts. The patient-fraud counts were not upheld in her case (the jury was unable to reach unanimous verdicts on some and acquitted on others), but the underlying conduct is documented in the trial record and in the CMS findings.

Did anyone actually get hurt by Theranos blood tests?

Yes. The CMS findings, the trial testimony, and Carreyrou’s reporting all document patients who received erroneous Theranos test results that affected their medical care. Some examples that became part of the public record involved patients receiving false low-positive results for HIV (causing significant distress before corrected by retesting), false abnormal results for thyroid hormones (leading to inappropriate medication adjustments), false elevated PSA results (leading to unnecessary urological workup), and false abnormal results for vitamin D and other common analytes (leading to unnecessary treatment). The aggregate number of patients affected is in the hundreds of thousands range — Theranos voided or corrected approximately 1 million test results across 2014–2015 in the wake of the CMS findings. The reason the patient-fraud counts did not lead to conviction in the Holmes case was a question of legal proof of her specific knowledge and intent with respect to the patient-facing representations, not a question of whether patient harm occurred.

What is the difference between Theranos and a normal venture-backed startup with bold claims?

Many venture-backed startups make bold claims that do not, in the end, materialize. The line between aspiration and fraud is whether the claims are knowingly false at the time they are made. A startup that says “we believe our technology will eventually do X” is making a prediction. A startup that says “our technology does X today” when its internal data shows the technology does not do X is making a misrepresentation. Theranos fell on the misrepresentation side of that line, repeatedly, with respect to both investor-facing and partner-facing claims, on the basis of the documentary evidence in the trial record. The lesson for evaluators of venture-backed claims is not that bold claims are always suspect — they are not — but that there is an evidential distinction between aspirational claims about future capability (where the company should be honest about the gap between current and future state) and factual claims about current capability (where the company should be able to produce verification on demand). The diligence question is which category a given claim falls into and whether the supporting evidence matches the claim’s category.

Why did the board and investors not catch this earlier?

The trial record and Carreyrou’s reporting both document substantial evidence on this question. Several mechanisms contributed. The board was structured around political and military prestige rather than subject-matter expertise in clinical chemistry, so the directors who would have been in a position to ask the right technical questions (notably the physician members) were either positioned outside the operational decisions or were operating outside their primary clinical specialties. Sequential investors largely relied on prior investors’ diligence rather than performing their own — a pattern Carreyrou describes as “credibility lending” in which the presence of marquee names on the cap table substituted for substantive verification. The company’s aggressive use of non-disclosure agreements and litigation threats against current and former employees suppressed the flow of critical information that would normally have reached diligence-stage investors through reference calls and informal industry communication. And, more generally, the social pressure within Silicon Valley to participate in deals associated with celebrated founders and prestigious boards weighed against the kind of skeptical, technical diligence that would have surfaced the problems. None of these explanations excuse the failure — competent biotech investors with subject-matter expertise (Google Ventures was one of the better-known venture firms that declined to invest in Theranos after technical diligence raised concerns) did identify the issues — but they explain why the failure was as widespread as it was.

What changed in regulation or industry practice as a result of Theranos?

Several changes are partly attributable to the case. The FDA in 2014–2016 moved toward more active oversight of laboratory-developed tests, in part in response to Theranos and similar concerns, although the formal regulatory framework for LDTs continues to evolve as of 2026. Venture-capital due-diligence practice in healthtech and biotech has tightened, with more sophisticated technical-diligence frameworks becoming standard at major funds. Major laboratory-medicine professional societies (the American Association for Clinical Chemistry, the College of American Pathologists) have updated guidance on the analytical validation standards that should apply to direct-to-consumer testing. The Diamandis and Plebani editorials are taught in clinical-chemistry training programs as case studies. And the case is widely used in MBA-program ethics curricula as the canonical example of how startup hype, prestige board recruitment, and the lab-developed-test regulatory loophole combined to allow a major scientific-validation evasion. The substantive question of whether the underlying conditions have been adequately addressed — whether a future Theranos could be sustained for as long against tighter scrutiny — remains, in 2026, an open one.

Where should I start if I want to read the full case?

Read Carreyrou’s Bad Blood (2018) first. It is the single most accessible and comprehensive treatment of the full arc, well-sourced, and a fluent narrative. Then read the Diamandis 2015 and Plebani 2015 editorials in Clinical Chemistry and Laboratory Medicine — together they are about ten pages and they show what subject-matter experts were saying on the basis of the public record before the consumer press caught up. Then, for primary source depth, the Holmes trial transcripts and exhibits are publicly available through PACER (case 18-cr-258, N.D. Cal.) and the SEC complaint (Press Release 2018-41) provides the regulatory account. The HBO documentary The Inventor (Gibney, 2019) is a useful supplement that reproduces some of the most important video footage from the company’s marketing materials and from Holmes’s keynote appearances. The total reading-and-viewing time to develop a substantive grasp of the case is approximately fifteen to twenty hours, which is unusually high but commensurate with the case’s importance as a study object.