Theranos in Hindsight: The Warning Signs That Were Visible From the Outside

Almost everyone in diagnostics has an opinion on Theranos, and almost all of those opinions were formed after October 2015. The more useful question is not "how did they get away with it?" but "what could an attentive outsider have seen, with public information, before the story broke?" The answer is: quite a lot. And most of those signals can be checked in today's diagnostics startups too.

A word on fairness first. Hindsight makes every warning sign look obvious. Many legitimate startups share some of these traits, and none of them alone proves anything. What made Theranos stand out was that all of them appeared together, and that they reinforced each other.

Signal 1: No independent data, for a decade

Theranos was founded in 2003. By 2014 it was valued at around $9 billion; according to the SEC, it eventually raised more than $700 million from investors. In all that time it published essentially no peer-reviewed data showing that its technology worked.

That was noticed. In February 2015, months before the WSJ investigation, John Ioannidis published a viewpoint in JAMA titled "Stealth Research: Is Biomedical Innovation Happening Outside the Peer-Reviewed Literature?", with Theranos as the central example. A PubMed search at the time found no publications describing the company's technology, while coverage in business media was extensive. Clinical chemist Eleftherios Diamandis published a critical analysis of the company's claims in Clinical Chemistry and Laboratory Medicine the same year.

Why this matters for a diagnostic: A diagnostic test is a measurement. Measurements can be checked. Accuracy, precision, bias against a reference method, reportable range, interferences: all of these can be shown without revealing proprietary chemistry or engineering. A company that claims to measure blood better than everyone else but will not show a method comparison is not protecting its IP. It is avoiding the one comparison that matters.

Signal 2: Claims that conflicted with basic laboratory science

Theranos claimed it could run a broad menu of tests from a few drops of capillary blood. Experienced laboratory scientists were sceptical from the start, for concrete reasons: fingerstick blood is affected by tissue fluid and haemolysis, small sample volumes make low-concentration analytes very hard to measure precisely, and running many assay types (immunoassays, general chemistry, haematology) from a single microsample requires dilution that costs sensitivity.

None of this makes it impossible in principle. But it means the burden of proof is very high, and Signal 1 meant that burden was never met. When the WSJ published its investigation in October 2015, a former senior employee said that by the end of 2014, only 15 of the more than 240 tests Theranos offered were run on its proprietary Edison device. Most of the rest ran on conventional analysers bought from companies like Siemens.

Signal 3: A board of statesmen, not scientists

The Theranos board included former cabinet secretaries and other senior figures: George Shultz, Henry Kissinger, William Perry, Sam Nunn, James Mattis. Impressive names. What was missing was anyone with deep expertise in laboratory medicine, diagnostics regulation or clinical chemistry.

A board's job is to challenge management. A board that cannot ask a sharp question about coefficient of variation, proficiency testing or FDA clearance cannot challenge a diagnostics company on its core claim. The board was a credibility signal pointing outwards, not a governance mechanism pointing inwards.

Signal 4: Investors without diagnostics expertise

The large Theranos investments came largely from wealthy families and individuals, not from the specialised healthcare venture funds that normally lead diagnostics rounds. At least one specialist investor looked closely and passed. GV (Google Ventures) founder Bill Maris later said the firm declined in 2013: the pitch was full of "hand-waving," so GV sent someone from its life-science team to take a Theranos test at Walgreens. That person was asked for more blood than a finger prick. For a company whose entire promise was the finger prick, that was enough.

A missing specialist investor is not proof of a problem. But it is a question worth asking: if the funds with lab scientists on staff passed, what did they see?

Signal 5: A regulatory route that avoided independent review

Theranos ran its tests as laboratory-developed tests (LDTs) under CLIA in its own lab, a route that at the time did not require FDA review of the test's performance. That was legal. It also meant that no independent regulator had evaluated whether the core technology actually performed. When Theranos did go through the FDA in July 2015, it received 510(k) clearance and a CLIA waiver for a single test (herpes simplex virus 1 IgG), and used that milestone in its communications far more broadly than one assay justified.

Partners did not fill the gap. Walgreens' review relied largely on data Theranos supplied rather than on independent testing of the technology, and prosecutors later alleged that Holmes had shown Walgreens validation reports with pharmaceutical company logos that those companies had never authorised.

For a European context: the IVDR has largely closed this route. Under Article 5(5), in-house tests are restricted to health institutions under specific conditions, and commercial tests need a conformity assessment, for most risk classes involving a notified body. That does not make fraud impossible, but it puts an independent reviewer in the path much earlier.

Signal 6: Secrecy, turnover and legal pressure

Secrecy beyond what IP protection requires; departments isolated from each other; high turnover in key scientific positions; aggressive legal action against former employees and critics. Not all of this was public in 2014, but some of it was visible to anyone who looked at LinkedIn profiles or spoke to former employees.

Inside the company, people did raise alarms. Tyler Shultz and Erika Cheung became whistleblowers after their concerns were dismissed internally. Organisational culture that punishes inconvenient data is, in my view, the most reliable single warning sign there is, because it means the company can no longer correct itself.

Signal 7: Total founder control

Elizabeth Holmes held majority voting control; giving it up was part of her 2018 settlement with the SEC. Combined with a board that lacked domain expertise, there was no effective internal check on her decisions. The undisclosed personal relationship between Holmes and COO Ramesh "Sunny" Balwani removed the last potential counterweight in management.

How it ended

The WSJ published John Carreyrou's investigation on 15 October 2015. Regulatory inspections followed and found serious deficiencies; Theranos voided two years of Edison test results. In March 2018 the SEC charged Holmes and Balwani with "massive fraud." Holmes was convicted on four counts of investor fraud in January 2022 and sentenced to 11 years and 3 months in prison; Balwani was convicted on all twelve counts in July 2022 and sentenced to almost 13 years.

What investors and partners should check today

The lesson is not "be suspicious of ambitious founders." It is that diagnostics, unlike many other technologies, can be verified with a fairly standard set of questions. If a company cannot or will not answer them, that is the signal.

1. Show me the method comparison. Against which reference method, on how many samples, from which population, with what bias and imprecision?
2. Who has seen the data independently? A peer-reviewed paper, an external performance study, a notified body, a key opinion leader who actually ran samples.
3. What is the regulatory path, and when does an independent party review it? Under IVDR: which class, which notified body, which timeline?
4. Who on the board or among investors can challenge the technical core? If nobody can, add someone who can.
5. Can I see the lab? And speak to the people running it, not only to the founders.
6. How do people leave? Turnover in scientific and quality roles, and how departures are handled, says a lot about whether inconvenient data is heard.

Most diagnostics startups that struggle with these questions are not frauds. They are early, under-resourced, and have not yet built the evidence. That is fixable. The important thing is that investors, partners and founders treat these questions as normal, not as an accusation.

---

If you are an investor looking at a diagnostics company, or a founder preparing for a technical due diligence, I am happy to walk through these questions with you before someone else asks them. No theory for theory's sake.