Part BOutcome & teaching note

Technical Validation · 2003–2018

Theranos

At the cutoff, Theranos publicly rebutted accuracy criticism and remained in a continuing posture while Walgreens had reportedly paused expansion; the evidence set contains no verified cutoff board resolution.

This instructor-only report is bound to the outcome-blind October 28, 2015 packet. It tests the decision process against later evidence without rewriting what was knowable at the cutoff.

Outcome in brief

The frozen recommendation—to pause patient testing and retail expansion, preserve records and specimens, and require independent assay-by-assay and end-to-end workflow validation—was process-sound. It did not require a finding that every Theranos test was inaccurate or that anyone had committed fraud. It responded to asymmetric patient harm, a wide evidence gap, partner hesitation, and fresh quality-system observations while preserving the option to relaunch a validated subset. @judgment.theranos.outcome.process-quality

The public record after the cutoff developed along four dimensions:

  1. An independent black-box study of 60 healthy adults found significant interservice variability in 15 of 22 measurements and said Theranos flagged out-of-range results 1.6 times as often as the other services. The authors could not isolate collection, processing, instrumentation, timing, or their interaction as the source. @claim.theranos.outcome.jci-variability @src.theranos.outcome.jci-2016
  2. CMS determined that the Newark laboratory had condition-level deficiencies and that specified hematology practices posed immediate jeopardy. It later imposed sanctions after concluding the deficiencies had not been corrected. @claim.theranos.outcome.cms-immediate-jeopardy @claim.theranos.outcome.cms-sanctions
  3. Walgreens terminated the relationship and closed all 40 Theranos Wellness Centers in its Arizona stores. @claim.theranos.outcome.walgreens-terminated @src.theranos.outcome.walgreens-termination-2016
  4. By September 2018, Theranos said it intended an assignment for creditors and dissolution, reported at least USD 60 million of unsecured claims and about USD 5 million of estimated net remaining cash, and expected no shareholder distribution. @claim.theranos.outcome.wind-down

What the technical evidence establishes

The later independent comparison is strong evidence of service-level disparities, but it is not a controlled decomposition of the proprietary device. The blood-collection method, shipment to a central facility, sample preparation, timing, conventional versus new instrumentation, and absent Theranos technical replicates remained confounded. The correct conclusion is narrower than “all Theranos technology failed”: the commercial service lacked enough independently reproducible, workflow-specific evidence to justify broad patient use. @claim.theranos.outcome.jci-variability @conflict.theranos.outcome.technical-source-attribution

The FDA record at the cutoff also matters. One HSV-1 assay had a defined intended use and reported performance package. That evidence remains positive for that narrow assay; it does not validate hundreds of other assay-workflow combinations. Later appellate language about the number of small-sample assays and the subset run on Edison is relevant current record, but it must not be smuggled into the 2015 decision packet. @claim.theranos.cutoff.one-assay-clearance @claim.theranos.outcome.device-scope

CMS's immediate-jeopardy determination and sanctions are regulator actions about a laboratory and specified condition-level failures. They carry more adjudicative weight than a Form 483, but they still do not prove that every reported result was wrong or isolate one device as the sole cause. The durable analytical unit is the assay plus specimen, device, site, operator, process, and quality system—not a company-wide slogan.

Legal status is evidence, not decoration

The SEC complaint alleged that Theranos raised more than USD 700 million while making materially misleading claims and that its proprietary analyzer performed only about 12 tests from a menu exceeding 200. Those are complaint allegations in this corpus. Holmes's civil judgment was entered by consent without a general admission or denial and included specified remedies. Neither record is silently upgraded into litigated findings. @claim.theranos.outcome.sec-allegations @claim.theranos.outcome.sec-consent-remedies

The current appellate record likewise requires defendant-specific precision: Holmes was convicted on four investor-related counts and acquitted on patient-related counts, while Balwani was convicted on all submitted investor- and patient-related counts. The verdicts cannot be collapsed into one generalized finding about every claim, test, executive, or victim. @claim.theranos.outcome.verdict-scope @conflict.theranos.outcome.legal-scope

Arizona's consent terms, SEC remedies, criminal restitution, regulator sanctions, and creditor outcomes measure different legal and economic objects. The report preserves those distinctions instead of using the magnitude of later remedies as retroactive proof of a technical proposition.

Financial record: deliberately nonadditive

The deterministic table records six sourced amounts: the 2017 exchange-offer original issue price, Arizona consumer restitution, the Holmes SEC civil penalty, issuer-estimated remaining cash, issuer-reported minimum unsecured claims, and criminal restitution. @table.theranos.outcome.nonadditive-recorded-amounts

They must not be summed. The USD 582.2 million Form D figure was the aggregate original issue price of securities tendered by existing preferred holders in an exchange offer—not verified new cash. The USD 452 million restitution order is a criminal remedy involving identified victims, not enterprise value. Remaining cash and minimum claims were management estimates in a wind-down letter, not audited liquidation endpoints. @claim.theranos.outcome.exchange-not-cash @claim.theranos.outcome.restitution @claim.theranos.outcome.nonadditive-financial-record

No audited bridge in the corpus allocates value destruction among technical validation, laboratory operations, misrepresentation, financing, partner decisions, regulation, litigation, and later management. Warren therefore abstains from a target price, total enterprise-loss estimate, or causal percentage. @judgment.theranos.outcome.no-single-loss-bridge

Best-supported causal model

The primary hypothesis is a validation-governance gap:

flowchart LR
    A["Broad patient-facing claims"] --> B["Evidence and quality-governance gap"]
    B --> C["Unresolved technical and laboratory risk"]
    C --> D["Regulator and partner response"]
    D --> E["Trust, financing, and operating contraction"]
    E --> F["Creditor wind-down"]

This is a directional mechanism, not a quantified loss model. At the cutoff, one narrow clearance coexisted with scientific criticism, reported partner hesitation, and newly public FDA observations. Later independent, regulator, partner, and wind-down records are consistent with the risk that unsupported scope allowed technical and quality uncertainty to propagate through the enterprise. @hypothesis.theranos.validation-governance-gap

The rival hypothesis deserves real weight: collection, processing, conventional laboratory operations, financing, partner behavior, and later legal events may explain material portions of the outcome independently of the proprietary device. The black-box study's explicit limitations and the narrow HSV-1 evidence prevent a universal device-failure conclusion. @hypothesis.theranos.measurement-operations-and-other-causes

The evidence supports the primary hypothesis at moderate confidence, not a monocausal verdict. Board minutes, raw full-menu validation, device-to-result mapping, quality logs, partner contracts, audited financials, and capitalization records remain missing.

Counterfactual and signals

The feasible cutoff action was a controlled pause: preserve evidence, stop unsupported patient workflows, independently validate every offered combination, close quality-system issues, notify affected clinicians and patients where required, and relaunch only a passing subset. @counterfactual.theranos.independent-assay-gates

Four signals were available before the decision:

SignalLead timeFalse-positive boundary
One narrow clearance versus a broad menuAbout 3.5 monthsOther assays may have had nonpublic evidence or conventional workflows
Missing independent scientific transparencyMore than 4 monthsConfidential or unpublished data can still be valid
FDA quality and validation observationsHoursA Form 483 is not a final determination and may be remediated
Reported Walgreens expansion pause4 daysPartner diligence can be temporary and reversible

See @signal.theranos.clearance-menu-gap, @signal.theranos.scientific-transparency, @signal.theranos.quality-system-observations, and @signal.theranos.partner-pause.

Transferable rules

@rule.patient-facing-assay-validation-gate is a candidate safety-and-governance rule: when patient-facing claims outrun independent assay and workflow evidence, pause unsupported use and require predeclared validation and quality gates. It is not a fraud classifier, and passing the gate does not establish attractive unit economics or valuation.

@rule.preserve-legal-and-measurement-status is a candidate analytical rule: preserve whether a proposition is an observation, allegation, consent term, acquittal, conviction, appellate holding, company estimate, or audited fact. Aggregate monetary values only after deterministic reconciliation proves compatible scope and no overlap.

Most important lessons

  1. Match validation to the marketed claim. One assay, one specimen, or one device configuration cannot validate a heterogeneous service menu.
  2. Underwrite the full workflow. Collection, transport, preparation, instrumentation, software, operators, quality controls, and result correction all affect patient outcomes.
  3. Use gates before patient scale. Predeclared independent tests preserve option value while limiting asymmetric harm.
  4. Treat missing transparency as a diligence trigger. It supports a pause and evidence demand, not an automatic accusation.
  5. Keep regulator records in status. Inspectional observations are not final findings; later CMS determinations and sanctions have their own scope.
  6. Keep legal propositions defendant- and count-specific. Pleadings, consent judgments, acquittals, convictions, and appellate holdings are not interchangeable.
  7. Do not manufacture a loss bridge. Exchange values, remedies, creditor claims, cash estimates, and restitution are not additive.
  8. Separate process quality from hindsight. The pause was justified by cutoff uncertainty and patient-risk asymmetry, not by foreknowledge of later outcomes.
  9. Abstain when valuation inputs are missing. Technical conviction cannot substitute for verified cash, burn, liabilities, capitalization, and contracts.

Final boundary

This case does not establish that every Theranos assay was inaccurate, that every service difference came from Edison, that every public claim was adjudicated false, or that technical validation alone caused the enterprise failure. It establishes a narrower and more transferable principle: patient-facing scale should stop when marketed scope outruns independent workflow-specific evidence, and both legal status and financial measurement must survive intact from source to conclusion.

Observed after the cutoff

Outcome financials

1 tables

Later values do not backfill Part A. Definition changes, unknowns, and derived endpoints remain labeled.

Source-specific recorded amounts — different bases and not additiveAs Reported At Horizon · USD
MeasureRecorded amount
2017 exchange-offer original issue price582,200,3181
2017 Arizona consumer restitution4,650,0001
2018 SEC civil penalty500,0001
2018 issuer-estimated remaining cash5,000,0001
2018 issuer-reported minimum unsecured claims60,000,0001
2023 criminal restitution order affirmed in 2025452,000,0001
USD · USDReported values remain strings; no browser-side recalculation.

Transferable—but not universal

Candidate decision rules

2 hypotheses

These rules are case-derived hypotheses. Each retains “unless” conditions, kill criteria, counterexamples, and promotion gaps.

Candidatemoderate confidence

rule.patient-facing-assay-validation-gate

Pause the unsupported workflows, preserve records and specimens, and allow patient use only after independent blinded validation and quality-system gates clear for every offered combination.

Assay- and workflow-specific gates prevent narrow evidence from being extrapolated across heterogeneous tests and stop unresolved quality failures from propagating through patients, partners, regulators, and capital providers.

Use when

  • A product produces patient-facing or safety-critical decisions with asymmetric downside from false results.
  • Commercial claims or rollout scope exceed independently verified assay-, specimen-, device-, site-, operator-, or workflow-specific evidence.
  • A pause, ring-fence, or staged validation program remains operationally feasible.

Do not transfer when

  • A verified emergency-use pathway expressly permits the use and the residual risk is independently documented and disclosed.
  • The allegedly unsupported workflow can be proven outside the gate's scope with immutable, independently reproduced evidence.

Reverse or kill if

  • Relaunch only the combinations that clear predeclared analytical, clinical, specimen, device, site, operator, and quality-system criteria.
  • Kill the broader program if critical records are unavailable, remediation cannot be independently verified, or liquidity cannot support safe completion.
  • Narrow this rule if cross-case testing shows that mandatory full-workflow gating causes greater patient harm than bounded use under a verified emergency pathway.
Limitations and promotion gaps
  • A single failure case cannot identify optimal thresholds, duration, cost, or the value of a pause.
  • Full validation coverage is a safety gate, not evidence that a business model or valuation is attractive.
  • The Theranos public record cannot quantify how much failure arose from device performance versus collection, processing, laboratory operations, governance, financing, or misconduct.
Candidatemoderate confidence

rule.preserve-legal-and-measurement-status

Preserve each proposition's legal and epistemic status and prohibit additive financial or causal conclusions until an independently reconciled bridge proves compatibility and non-overlap.

Status and measurement separation prevents allegations from becoming findings, defendant-specific verdicts from becoming enterprise-wide conclusions, and heterogeneous monetary records from becoming a fabricated loss estimate.

Use when

  • A business-failure analysis combines regulator observations, complaints, consent judgments, verdicts, appellate holdings, company estimates, or monetary remedies.
  • The records concern different entities, dates, legal burdens, or measurement bases.

Do not transfer when

  • A controlling adjudication or audited reconciliation expressly supersedes the earlier status and maps every amount to compatible entities, dates, and concepts.

Reverse or kill if

  • Correct or withdraw any conclusion whose pleaded, consented, acquitted, convicted, affirmed, estimated, or audited status was misstated.
  • Permit aggregation only after deterministic reconciliation proves identical scope, compatible measurement bases, and no overlap.
Limitations and promotion gaps
  • Status discipline does not determine intent, liability, technical causation, or enterprise loss by itself.
  • Public dockets may omit sealed, superseded, or later-corrected records.

Lineage

Complete case source ledger

21 records

This list combines decision-cutoff and outcome evidence. Each report citation resolves to a source ID below. Third-party documents remain with their original publishers.

T1

src.theranos.cutoff.form-d-2010

Theranos Inc. Form D filed July 8, 2010

U.S. Securities and Exchange Commission · Jul 8, 2010

Regulatory FilingPrimaryContemporaneous

Used for: Cutoff-valid filed financing record · Historical funding-scale context

T3

src.theranos.cutoff.diamandis-2015

Theranos phenomenon: promises and fallacies

U.S. National Library of Medicine / PubMed · Jun 2, 2015

Academic ResearchSecondaryContemporaneous

Used for: Contemporaneous independent scientific criticism · Disconfirming evidence against unqualified company claims

T1

src.theranos.cutoff.fda-hsv1-2015

FDA decision summary for the Theranos HSV-1 IgG assay

U.S. Food and Drug Administration · Jul 8, 2015

Court Or Government RecordPrimaryContemporaneous

Used for: Cutoff-valid regulatory evidence for one assay and defined intended use · Method-comparison sample and reported performance context

T1

src.theranos.outcome.cms-letter-2016

CMS letter to Theranos concerning its Newark laboratory

Centers for Medicare & Medicaid Services; archived as a trial exhibit by the U.S. District Court for the Northern District of California · Sep 2, 2021

Court Or Government RecordPrimaryContemporaneous

Used for: Official condition-level and immediate-jeopardy determination · Survey scope and correction process

T1

src.theranos.outcome.cms-deficiencies-2016

CMS statement of deficiencies for the Theranos Newark laboratory

Centers for Medicare & Medicaid Services; archived as a trial exhibit by the U.S. District Court for the Northern District of California · Sep 2, 2021

Court Or Government RecordPrimaryContemporaneous

Used for: Detailed quality-control and patient-testing findings · Assay, device, personnel, and procedure scope

T1

src.theranos.outcome.cms-sanctions-2016

CMS imposition-of-sanctions letter to Theranos

Centers for Medicare & Medicaid Services; mirrored by The Wall Street Journal · Jul 8, 2016

Court Or Government RecordPrimaryContemporaneous

Used for: Imposed CLIA remedies and effective dates · CMS evaluation of Theranos correction submissions

T1

src.theranos.outcome.form-d-2017

Theranos Inc. Form D filed May 30, 2017

U.S. Securities and Exchange Commission · May 31, 2017

Regulatory FilingPrimaryContemporaneous

Used for: Later filed financing and capital-structure context · Exchange-offer basis disclosure

T1

src.theranos.outcome.arizona-ag-2017

Arizona Attorney General 2017 Annual Report

Arizona Attorney General's Office · Nov 1, 2024

Court Or Government RecordPrimary

Used for: Official retrospective summary of Arizona consent-judgment terms · Consumer testing, correction, restitution, penalty, and fee quantities

T1

src.theranos.outcome.sec-complaint-2018

SEC complaint against Theranos and Elizabeth Holmes

U.S. Securities and Exchange Commission · Mar 15, 2018

Court Or Government RecordPrimary

Used for: Regulator allegations about investor representations, device scope, and capital raised · Alleged mechanism and chronology for later causal analysis

T1

src.theranos.outcome.sec-judgment-2018

Final judgment as to Elizabeth Holmes in the SEC civil action

U.S. District Court for the Northern District of California · Mar 28, 2018

Court Or Government RecordPrimaryContemporaneous

Used for: Adjudicated remedies and consent posture · Civil penalty, officer-and-director bar, and share-return terms

T2

src.theranos.outcome.dissolution-letter-2018

Theranos stockholder letter concerning assignment and dissolution

Theranos, Inc.; mirrored by The Wall Street Journal · Sep 6, 2018

Issuer DisclosurePrimaryContemporaneous

Used for: Company account of default, creditor claims, estimated cash, assignment, and intended dissolution · Equity-recovery outcome