In plain English: Theranos said it could run dozens, then hundreds, of tests from one drop. Its own machine ran one kind of test, one sample at a time, and the company quietly sent people for normal blood draws instead. Running many tests from one small sample is real, but it works by splitting the sample into many tiny reactions, or by instruments that read many molecules at once because the physics keeps their signals apart. A computer can translate any number of signals; it cannot create them.
1. The short answer
- Yes, Theranos claimed it, in numbers: 30 tests from one drop (Smithsonian, March 2014), "as many as 70 different tests from a single draw of 25 to 50 microliters" (Fortune, June 2014), a menu of 200-plus tests "ramping up to more than 1,000," and, in Holmes's pitch to Walgreens, every test "on one analyzer using chemicals from one consumable cartridge."
- No, it never had it. Its analyzer "was designed to perform only one method of testing, immunochemistries, and could process only one sample at a time," and at its height ran about 12 of the menu's 200-plus tests (SEC complaint). The company sent patients for venous draws from 2014, dropped tests from orders that fingerstick could not support, and stopped its proprietary methods for all but one test in September 2015, without ever announcing a retreat.
- Yes, other people have run many tests from one small sample, by several different physical routes, from a 14-analyte disc on a clinic counter to 5,400 proteins from 2 microliters in a research lab. None of them is a counter-top box returning dozens of cleared, quantitative results from one capillary drop in minutes.
- How Thimble would do it: the proven route, splitting one draw into two tubes and about thirty parallel reactions plus a few thousand imaged cells (about 40 assays and a CBC from about 170 µL), with the spectrum-based routes kept for the hub laboratory.
2. What Theranos claimed
| When | Claim | Source |
|---|---|---|
| Early 2014 | The website lists about 100 tests, rising to 214 by June, "all to be run from the same drop of blood" | Fortune, Smithsonian |
| March 2014 | "How to run 30 health tests on a single drop of blood": one nanotainer holding "about the amount of a raindrop," tested simultaneously for STDs, heart disease and diabetes | Smithsonian |
| June 2014 | "As many as 70 different tests from a single draw of 25 to 50 microliters"; "currently offers 200, and is ramping up to offer more than 1,000" tests; 1/100th to 1/1,000th of the usual blood volume | Fortune cover story |
| 2010 onward | To Walgreens: a device able to run any blood test from a few drops in real time, "on one analyzer using chemicals from one consumable cartridge," processing more than 1,000 CPT codes | SEC complaint |
| 2013 to 2015 | Marketing: "One tiny drop changes everything." "All the same tests. One tiny sample." | Fortune (2015) |
3. What it had, and the quiet retreat
| When | What was true |
|---|---|
| Design | The Edison "was designed to perform only one method of testing, immunochemistries, and could process only one sample at a time" (SEC). It never ran chemistry, blood counts or molecular tests. |
| Early 2014 | The website already said "a tiny finger-stick or a micro-sample from a venous draw," which admitted venipuncture while stressing tiny samples (Fortune, 2015). |
| May 2014 | Holmes told Fortune venipuncture was used "in addition to the micro-samples just to handle the volume." Customers at Wellness Centers were being told their ordered tests needed a venous draw. |
| 2014 | Holmes instructed employees to remove tests from an order when Theranos could not perform them by fingerstick (SEC). |
| December 2014 | By the company's later account 57 percent of orders were fingerstick; by a former employee's account only about 15 fingerstick tests ran on the proprietary device, the rest on modified Siemens analyzers with diluted samples. |
| May 2015 | Holmes disclosed, as if new, that third-party analyzers were used on venous blood for "esoteric" tests. |
| Mid-September 2015 | After an FDA inspection, the company stopped its proprietary methods for all but one test (the cleared HSV-1 IgG). |
| October 2015 | FDA called the nanotainer an uncleared Class II device; Theranos stopped using it except for that one test. |
| August 2016 | The miniLab was unveiled with venous data only; the device could not run all eleven presented assays on one 160 µL sample. |
The company never said "we can only do one method, one sample at a time." It kept the slogans, sent people to the needle, and pivoted to a new box. Its patents describe the full ambition: "Modular point-of-care devices" (US 9,121,851 and continuations) covering immunoassays, nucleic acid tests and chemistry in one device, "Point-of-care fluidic systems" (US 7,635,594), and "Systems and methods for sample use maximization" (US 9,464,981), which notes that the number of assays run in parallel "is often limited by the size of the device and the volume of the sample." That sentence, from Theranos's own patent, is the whole problem.
4. The three things every measurement needs
- A reaction or a physical signature that is specific to the target. An enzyme that only acts on glucose, an antibody that only binds TSH, an electrode that only responds to potassium, or a molecular mass or resonance that belongs to one compound.
- Enough sample for that reaction. A few microliters to a hundred, depending on the method and how dilute the target is. Hormones are present at picomolar to nanomolar levels; a reaction needs enough molecules to produce a signal above noise.
- A detector and a calibration. The detector reads light, current or mass; the calibration turns the reading into a number. This is the software part, and it scales almost for free.
Software cannot supply the first two. That is the precise sense in which the skeptic is right. The optimist is right about the third: once the physics keeps signals apart, one instrument can read hundreds of them from one sample, and the "translator" is ordinary calibration and spectral deconvolution. The art is in choosing physics that keeps the signals apart.
A fourth constraint cuts across all three: the tube. Cells are counted in EDTA, which binds calcium and (as the potassium salt) contaminates potassium results; chemistry and most immunoassays use lithium heparin or serum; clotting tests need citrate at an exact ratio. One undivided drop cannot serve every class, so a serious design splits the draw at collection.
5. Who has really done many tests from one small sample
Each row is a different physical route to "many from one."
| Route | Real example | Sample | Results per sample | Limit |
|---|---|---|---|---|
| Parallel micro-reactions, one reagent per chamber, read by a shared detector | Abaxis Piccolo: up to 14 analytes from 100 µL on one disc (1995); i-STAT CHEM8+: 8 analytes plus hematocrit on one cartridge; Randox Biochip: 23 to 48 immunoassays on one 9 x 9 mm chip from 6 to 150 µL; Truvian: 34 tests from one tube (venous, FDA-cleared 2025-2026) | 65 to 150 µL | 10 to 50 | Volume per reaction; one reagent per chamber; tube type |
| One spectrum, many molecules: mass spectrometry | Newborn screening by tandem MS: 20 to 60 metabolic disorders from one 3 mm punch of a dried heel-prick spot (about 3 µL of blood), routine since the late 1990s; forensic toxicology screens hundreds of drugs in one injection | 3 to 20 µL | Dozens to hundreds | A central-lab instrument; sample preparation; a standard for each analyte to quantify; miniature mass spectrometers are judged 5 to 10 years from clinical drug monitoring |
| One spectrum, many molecules: NMR | Nightingale Health: 249 measures per plasma sample on a CE-marked device; all 500,000 UK Biobank samples; 37 clinically validated biomarkers; a self-collected finger-prick consumer service in Finland with more than 250,000 results | About 85 µL plasma | 249 | 500 MHz magnets in a central lab; lipids and small metabolites, not proteins or cells |
| High-plex affinity arrays read by sequencing | Olink Explore HT: about 5,400 proteins from 2 µL plasma; SomaScan: 7,000 to 11,000 protein measurements from 55 µL; Alamar NULISA: 100 to 250 proteins from home-collected finger-prick dried blood (June 2026) | 2 to 55 µL | Hundreds to thousands | Research grade; days; costly; mostly relative quantitation; not cleared clinical results |
| Bead multiplex immunoassay | Luminex xMAP: up to 500 analytes per well in principle, 80 in commercial kits, from 12.5 to 50 µL | Tens of µL | Up to 80 | A laboratory instrument; narrow clinical use |
| Photonic sensor arrays | Genalyte Maverick: more than 100 silicon rings per chip, four per analyte, direct from whole blood; TSH and RNP cleared | Small whole-blood sample | About 25 per chip | Every analyte still needs its own antibody and its own clearance |
| Optical "scan everything" | Dxcover: infrared spectrum of 10 µL of dried serum plus a classifier, in a 2,200-patient brain cancer trial; Haemanthus: Raman plus AI, no product | 10 µL | A yes-or-no classifier, not a panel of numbers | Weak, overlapping signals; hemoglobin absorption and fluorescence |
| Point-of-care cartridges today | Commercial single-cartridge panels run 3 to 10 analytes; a reader such as Alere Triage offers about 20 single tests one at a time | One drop | 3 to 10 per cartridge | The honest state of the art at the counter |
Two of these are the strongest evidence for the optimist: newborn screening, which has read dozens of disorders from one dried drop for a quarter century, and Nightingale, which quantifies 249 measures from one plasma sample on a certified device. Both need a large instrument in a central laboratory. The strongest evidence for the skeptic is the last row: at the counter, where speed and simplicity rule, a cartridge does 3 to 10 things.
6. How Thimble would do it from one draw
- Now, at the counter: the parallel-reaction route. One draw of about 370 µL is split into an EDTA tube and a heparin tube. The Chem Disc runs 17 analytes from 100 µL in 24 reagent cuvettes read by one photometer; the Heme Card images a few thousand cells from 20 µL and reads hemoglobin and HbA1c on the same card; the Immuno Pack runs up to six strips from 15 µL each; the optional EC Card runs nine electrochemical measurements from 65 µL. About 40 assays plus a 19-parameter CBC from about 170 µL, in about 15 minutes. Every one of those methods has been cleared in someone else's product.
- At the hub: the spectrum routes. The same visit can send a serum microtube to the hub laboratory, where mass spectrometry and, in time, NMR metabolomics turn one small sample into dozens or hundreds of measurements that no counter-top device can do (drug levels, vitamin D by mass spectrometry, metabolic profiles). High-plex proteomics from finger-prick dried blood is a research product today and a candidate hub product later.
- Later, at the counter: sensor arrays. A photonic ring chip or a biochip array could raise the immunoassay count per cartridge from one to a few dozen. That is the version-2 immunoassay engine, after a freedom-to-operate review.
- Never: "scan the drop and read everything." The physics of blood does not cooperate, and no one has shown it quantitatively.
7. Who is right
The forensic scientist is right that a single sample type cannot serve every test and that software cannot invent chemistry. The engineer is right that one sample can yield many results when the signals are physically separable and the translator exists, and that the translator is the cheap part. The resolution is the design above: many tests from one draw, by splitting it into many small, specific reactions, with the "read everything at once" instruments where they belong, in a laboratory. Theranos's error was not the ambition; it was pretending that one secret box had already solved all three requirements, and hiding the needle.