A Look Inside Theranos’ Dysfunctional Corporate Culture

When a chemist raised concerns about the blood testing machines' high error rates, she was ignored. So she resigned.
Image may contain Head Human Person and Face
Elizabeth Holmes dropped out of Stanford and founded Theranos in 2003. The company imploded in 2015 after it was revealed that it had misrepresented its blood-testing technology. The SEC charged Holmes with fraud, stripped her of her controlling stake in the company, fined her $500,000, and barred her from being an officer or director of a public company for 10 years.Gilbert Carrasquillo/Getty Images

This story is adapted from Bad Blood: Secrets and Lies in a Silicon Valley Startup by John Carreyrou.

Alan Beam was sitting in his office reviewing lab reports when Theranos CEO and founder Elizabeth Holmes poked her head in and asked him to follow her. She wanted to show him something. They stepped outside the lab into an area of open office space where other employees had gathered. At her signal, a technician pricked a volunteer’s finger, then applied a transparent plastic implement shaped like a miniature rocket to the blood oozing from it. This was the Theranos sample collection device. Its tip collected the blood and transferred it to two little engines at the rocket’s base. The engines weren’t really engines: They were nanotainers. To complete the transfer, you pushed the nanotainers into the belly of the plastic rocket like a plunger. The movement created a vacuum that sucked the blood into them.

Or at least that was the idea. But in this instance, things didn’t go quite as planned. When the technician pushed the tiny twin tubes into the device, there was a loud pop and blood splattered everywhere. One of the nanotainers had just exploded.

Holmes looked unfazed. “OK, let’s try that again,” she said calmly.

Beam1 wasn’t sure what to make of the scene. He’d only been working at Theranos, the Silicon Valley company that promised to offer fast, cheap blood tests from a single drop of blood, for a few weeks and was still trying to get his bearings.

Excerpted from Bad Blood: Secrets and Lies in a Silicon Valley Startup by John Carreyrou, Buy on Amazon

Knopf

He knew the nanotainer was part of the company’s proprietary blood-testing system, but he’d never seen one in action before. He hoped this was just a small mishap that didn’t portend bigger problems.

The lanky pathologist’s circuitous route to Silicon Valley had started in South Africa, where he grew up. After majoring in English at the University of the Witwatersrand in Johannesburg (“Wits” to South Africans), he’d moved to the United States to take premed classes at Columbia University in New York City. The choice was guided by his conservative Jewish parents, who considered only a few professions acceptable for their son: law, business, and medicine.

Beam had stayed in New York for medical school, enrolling at the Mount Sinai School of Medicine on Manhattan’s Upper East Side, but he quickly realized that some aspects of being a doctor didn’t suit his temperament. Put off by the crazy hours and the sights and smells of the hospital ward, he gravitated toward the more sedate specialty of laboratory science, which led to postdoctoral studies in virology and a residency in clinical pathology at Brigham and Women’s Hospital in Boston.

In the summer of 2012, Beam was running the lab of a children’s hospital in Pittsburgh when he noticed a job posting on LinkedIn that dovetailed perfectly with his budding fascination with Silicon Valley: laboratory director at a Palo Alto biotech firm. He had just finished reading Walter Isaacson’s biography of Steve Jobs. The book, which he’d found hugely inspiring, had cemented his desire to move out to the San Francisco Bay Area.

After he applied for the job, Beam was asked to fly out for an interview scheduled for 6 pm on a Friday. The timing seemed odd but he was happy to oblige. He met with COO Sunny Balwani first and then with Holmes. There was something about Balwani that he found vaguely creepy, but that impression was more than offset by Holmes, who came off as very earnest in her determination to transform health care. Like many people who met her for the first time, Beam was taken aback by her deep voice. It was unlike anything he’d heard before.

At the time, Theranos was on the cusp of becoming a tech darling. Founded by the charismatic Stanford dropout in 2003, its promises to revolutionize blood-testing—and by extension, the vast industry of medical diagnostics—would be swallowed whole by most of the technology press, which would lavish Holmes with glowing coverage. (WIRED was not exempt). Only later—in October 2015—would the truth come out: Theranos was a fraud built on secrecy, deliberate fabrication, and hype. After I revealed that fraud, the company would begin an implosion that continues to this day.

Beam had no way of knowing any of this when he accepted Theranos’ job offer in August 2012. The lab he inherited was divided into two parts: a room on the building’s second floor that was filled with commercial diagnostic equipment, and a second room beneath it where research was being conducted. The upstairs room was the federally certified part of the lab, the one Beam was responsible for. Balwani and Holmes viewed its conventional machines as dinosaurs that would soon be rendered extinct by Theranos’s revolutionary technology, so they called it “Jurassic Park.” They called the downstairs room “Normandy” in reference to the D-day landings during World War II. The proprietary Theranos devices it contained would take the lab industry by storm, like the Allied troops who braved hails of machine-gun fire on Normandy’s beaches to liberate Europe from Nazi occupation.

In his eagerness and excitement, Beam initially bought into the bravado. But a conversation he had with Paul Patel shortly after the botched nanotainer demonstration raised questions in his mind about how far along the Theranos technology really was. Patel was the biochemist who led the development of blood tests for Theranos’s new device, which Beam knew only by its code name—“4S.” Patel let slip that his team was still developing its assays on lab plates on the bench. That surprised Beam, who had assumed the assays were already integrated into the 4S. When he asked why that wasn’t the case, Patel replied that the new Theranos box wasn’t working.

By the summer of 2013, as Chiat\Day scrambled to ready the Theranos website for the company’s commercial launch, the 4S, aka the miniLab, had been under development for more than two and a half years. But the device remained very much a work in progress. The list of its problems was lengthy.

The biggest problem of all was the dysfunctional corporate culture in which it was being developed. Holmes and Balwani regarded anyone who raised a concern or an objection as a cynic and a nay-sayer. Employees who persisted in doing so were usually marginalized or fired, while sycophants were promoted.

Employees were Balwani’s minions. He expected them to be at his disposal at all hours of the day or night and on weekends. He checked the security logs every morning to see when they badged in and out. Every evening, around 7:30, he made a flyby of the engineering department to make sure people were still at their desks working.

With time, some employees grew less afraid of him and devised ways to manage him, as it dawned on them that they were dealing with an erratic man-child of limited intellect and an even more limited attention span. Arnav Khannah1, a young mechanical engineer who worked on the miniLab, figured out a surefire way to get Balwani off his back: answer his emails with a reply longer than 500 words. That usually bought him several weeks of peace because Balwani simply didn’t have the patience to read long emails. Another strategy was to convene a biweekly meeting of his team and invite Balwani to attend. He might come to the first few, but he would eventually lose interest or forget to show up.

While Holmes was fast to catch on to engineering concepts, Balwani was often out of his depth during engineering discussions. To hide it, he had a habit of repeating technical terms he heard others using. During a meeting with Khannah’s team, he latched onto the term “end effector,” which signifies the claws at the end of a robotic arm. Except Balwani didn’t hear “end effector,” he heard “endofactor.” For the rest of the meeting, he kept referring to the fictional endofactors. At their next meeting with Balwani two weeks later, Khannah’s team brought a PowerPoint presentation titled “Endofactors Update.” As Khannah flashed it on a screen with a projector, the five members of his team stole furtive glances at one another, nervous that Balwani might become wise to the prank. But he didn’t bat an eye and the meeting proceeded without incident. After he left the room, they burst out laughing.

Khannah and his team also got Balwani to use the obscure engineering term “crazing.” It normally refers to a phenomenon that produces fine cracks on the surface of a material, but Khannah and his colleagues used it liberally and out of context to see if they could get Balwani to repeat it, which he did. Balwani’s knowledge of chemistry was no better. He thought the chemical symbol for potassium was P (it’s K; P is the symbol for phosphorus)—a mistake most high school chemistry students wouldn’t make.

Not all the setbacks encountered during the miniLab’s development could be laid at Balwani’s feet, however. Some were a consequence of Holmes’s unreasonable demands. For instance, she insisted that the miniLab cartridges remain a certain size but kept wanting to add more assays to them. Khannah didn’t see why the cartridges couldn’t grow by half an inch since consumers wouldn’t see them. Holmes had abandoned her plan of putting the Theranos devices in Walgreens stores and operating them remotely, to avoid problems with the FDA. Instead, blood pricked from patients’ fingers would be couriered to Theranos’s Palo Alto lab and tested there. But she remained stuck on the notion that the miniLab was a consumer device, like an iPhone or an iPad, and that its components needed to look small and pretty. She still nurtured the ambition of putting it in people’s homes someday, as she had promised early investors.

Another difficulty stemmed from Holmes’s insistence that the miniLab be capable of performing the four major classes of blood tests: immunoassays, general chemistry assays, hematology assays, and assays that relied on the amplification of DNA. The only known approach that would permit combining all of them in one desktop machine was to use robots wielding pipettes. But this approach had an inherent flaw: Over time, a pipette’s accuracy drifts. When the pipette is brand new, aspirating 5 microliters of blood might require the little motor that activates the pipette’s pump to rotate by a certain amount. But three months later, that exact same rotation of the motor might yield only 4.4 microliters of blood—a large enough difference to throw off the entire assay. While pipette drift was something that ailed all blood analyzers that relied on pipetting systems, the phenomenon was particularly pronounced on the miniLab. Its pipettes had to be recalibrated every two to three months, and the recalibration process put the device out of commission for five days.

Kyle Logan1, a young chemical engineer who’d joined Theranos right out of Stanford, had frequent debates with Sam Anekal about this issue. He thought the company should migrate to a more reliable system that didn’t involve pipettes, such as the one Abaxis used in its Piccolo Xpress analyzer. Anekal would reply that the Piccolo could perform only one class of blood test, general chemistry assays. (Unlike immunoassays, which measure a substance in the blood by using antibodies that bind to the substance, general chemistry assays rely on other chemical principles such as light absorbance or electrical signal changes.) Holmes wanted a machine that was more versatile, he’d remind Logan.

Compared to big commercial blood analyzers, another one of the miniLab’s glaring weaknesses was that it could process only one blood sample at a time. Commercial machines were bulky for a reason: They were designed to process hundreds of samples simultaneously. In industry jargon, this was known as having a “high throughput.” If the Theranos wellness centers attracted a lot of patients, the miniLab’s low throughput would result in long wait times and make a joke of the company’s promise of fast test results.

In an attempt to remedy this problem, someone had come up with the idea of stacking six miniLabs one on top of the other and having them share one cytometer to reduce the size and cost of the resulting contraption. This Frankenstein machine was called the “six-blade,” a term borrowed from the computer industry, where stacking servers on top of one another is common to save space and energy. In these modular stacking configurations, each server is referred to as a “blade.”

But no one had stopped to consider what implications this design would have with respect to one key variable: temperature. Each miniLab blade generated heat, and heat rises. When the six blades were processing samples at the same time, the temperature in the top blades reached a level that interfered with their assays. Logan, who was 22 and just out of college, couldn’t believe something that basic had been overlooked.

Aside from its cartridge, pipette, and temperature issues, many of the other technical snafus that plagued the miniLab could be chalked up to the fact that it remained at a very early prototype stage. Less than three years was not a lot of time to design and perfect a complex medical device. These problems ranged from the robots’ arms landing in the wrong places, causing pipettes to break, to the spectrophotometers being badly misaligned. At one point, the blood-spinning centrifuge in one of the miniLabs blew up. These were all things that could be fixed, but it would take time. The company was still several years away from having a viable product that could be used on patients.

However, as Holmes saw it, she didn’t have several years. Twelve months earlier, on June 5, 2012, she’d signed a new contract with Walgreens that committed Theranos to launching its blood-testing services in some of the pharmacy chain’s stores by February 1, 2013, in exchange for a $100 million “innovation fee” and an additional $40 million loan.

Theranos had missed that deadline—another postponement in what from Walgreens’s perspective had been three years of delays. Holmes was determined to launch in Walgreens stores by September.

Since the miniLab was in no state to be deployed, Holmes and Balwani decided to launch with an older device called the Edison. That, in turn, led to another fateful decision—the decision to cheat.

In June, Daniel Young, the brainy MIT PhD who headed Theranos’s biomath team, came to see Beam in Jurassic Park with a subordinate named Xinwei Gong in tow. In the five years since he’d joined Theranos, Young had risen up the ranks to become the company’s de facto number-three executive. He had Holmes and Balwani’s ear, and they often deferred to him to solve nettlesome technical problems.

In his first few years at Theranos, Young had seemed every bit the family man, leaving the office at six every evening to have dinner with his wife and kids. This routine had drawn snickers behind his back from some colleagues. But after being promoted to vice president, Young had become a different person. He worked longer hours and stayed at the office later. He got very drunk at company parties, which was jarring because he was always quiet and inscrutable at work.

Young told Beam that he and Gong were going to tinker with the ADVIA 1800, one of the lab’s commercial analyzers. The ADVIA was a hulking 1,320-pound machine the size of two large office copiers put together that was made by Siemens Healthcare, the German conglomerate’s medical-products subsidiary.

Over the next few weeks, Beam observed Gong spend hours opening the machine up and filming its innards with his iPhone camera. He was hacking into it to try to make it compatible with small finger-stick blood samples, Beam realized. It seemed like confirmation of what Patel had told him: the 4S must not be working, otherwise why resort to such desperate measures? Beam knew the Edison could only perform immunoassays, so it made sense that Young and Gong would choose the ADVIA, which specialized in general chemistry assays.

One of the panels of blood tests most commonly ordered by physicians was known as the “chem 18” panel. Its components, which ranged from tests to measure electrolytes such as sodium, potassium, and chloride to tests used to monitor patients’ kidney and liver function, were all general chemistry assays. Launching in Walgreens stores with a menu of blood tests that didn’t include these tests would have been pointless. They accounted for about two-thirds of doctors’ orders. But the ADVIA was designed to handle a larger quantity of blood than you could obtain by pricking a finger. So Young and Gong thought up a series of steps to adapt the Siemens analyzer to smaller samples. Chief among these was the use of a big robotic liquid handler called the Tecan to dilute the little blood samples collected in the nanotainers with a saline solution. Another was to transfer the diluted blood into custom-designed cups half the size of the ones that normally went into the ADVIA.

The combination of these two steps solved a problem known as “dead volume.” Like many commercial analyzers, the ADVIA featured a probe that dropped down into the blood sample and aspirated it. Although it aspirated most of the sample, there was always some unused liquid left at the bottom. Reducing the sample cup’s size brought its bottom closer to the probe’s tip and diluting the blood created more liquid to work with.

Beam had reservations about the dilution part. The Siemens analyzer already diluted blood samples when it performed its assays. The protocol Young and Gong had come up with meant that the blood would be diluted twice, once before it went into the machine and a second time inside it. Any lab director worth his salt knew that the more you tampered with a blood sample, the more room you introduced for error.

Moreover, this double dilution lowered the concentration of the analytes in the blood samples to levels that were below the ADVIA’s FDA-sanctioned analytic measurement range. In other words, it meant using the machine in a way that neither the manufacturer nor its regulator approved of. To get the final patient result, one had to multiply the diluted result by the same factor the blood had been diluted by, not knowing whether the diluted result was even reliable. Young and Gong were nonetheless proud of what they’d accomplished. At heart, both were engineers for whom patient care was an abstract concept. If their tinkering turned out to have adverse consequences, they weren’t the ones who would be held personally responsible.

As September 9, 2013, approached, the date Holmes had set for the launch, Beam grew worried that Theranos wasn’t ready. Two of the assays performed on the hacked Siemens analyzers were giving the lab particular trouble: sodium and potassium. Beam suspected the cause of the latter was a phenomenon known as “hemolysis,” which occurs when red blood cells burst and release extra potassium into the sample. Hemolysis was a known side effect of finger-stick collection. Milking blood from a finger put stress on red blood cells and could cause them to break apart.

Beam had noticed a piece of paper with a number on it taped to Holmes’s office window. It was her launch countdown. The sight of it made him panic. A few days before the launch, he went to see her and asked her to delay. Holmes wasn’t her usual confident self. Her voice was tremulous and she was visibly shaking as she tried to reassure him that everything would be OK. If necessary, they could fall back on regular venous draws, she told him. That briefly made Beam feel better, but his anxiety returned as soon as he left her office.

Anjali Laghari1, a chemist who headed the immunoassay group, was dismayed when she returned from her three-week vacation in India in late August. Her team had been trying for years to develop blood tests on Theranos’s older device, the Edison. Much to her frustration, the black-and-white machines’ error rate was still high for some tests. Holmes and Balwani had been promising her for a year that all would be well once the company introduced its next-generation device, the 4S. Except that day never seemed to arrive. That was fine as long as Theranos remained a research-and-development operation, which was still the case when Laghari had departed for India three weeks earlier. But now everyone was suddenly talking about “going live” and there were emails in her in-box referring to an imminent commercial launch.

Launch? With what? Laghari wondered with growing alarm.

In her absence, she learned, employees who were not authorized CLIA lab personnel had been let into the lab. She didn’t know why, but she did know the lab was under instructions to conceal whatever it was they were doing from Siemens representatives when they came by to service the German manufacturer’s machines.

Changes had also been made to the way samples were being processed on the Edisons. Under Balwani’s orders, they were now being prediluted with a Tecan liquid handler before being run through the device. This was to make up for the fact that the Edison could run at most three tests on one finger-stick sample. Prediluting the blood created more volume to run more tests. But if the device already had a high error rate under normal circumstances, an additional dilution step seemed likely only to make things worse.

Laghari also had concerns about the nanotainers. Blood would dry up in the little tubes and she and her colleagues often couldn’t extract enough from them. She tried to talk sense into Holmes and Young by emailing them Edison data from Theranos’s last study with a pharmaceutical company—Celgene—which dated back to 2010. In that study, Theranos had used the Edison to track inflammatory markers in the blood of patients who had asthma. The data had shown an unacceptably high error rate, causing Celgene to end the companies’ collaboration. Nothing had changed since that failed study, Laghari reminded them.

Neither Holmes nor Young acknowledged her email. After eight years at the company, Laghari felt she was at an ethical crossroads. To still be working out the kinks in the product was one thing when you were in R&D mode and testing blood volunteered by employees and their family members, but going live in Walgreens stores meant exposing the general population to what was essentially a big unauthorized research experiment. That was something she couldn’t live with. She decided to resign.

Tina Noyes, her deputy in the immunoassay group who had worked at Theranos for more than seven years, also quit.

The resignations infuriated Holmes and Balwani. The following day, they summoned the staff for an all-hands meeting in the cafeteria. Copies of The Alchemist, Paulo Coelho’s famous novel about an Andalusian shepherd boy who finds his destiny by going on a journey to Egypt, had been placed on every chair. Still visibly angry, Holmes told the gathered employees that she was building a religion. If there were any among them who didn’t believe, they should leave. Balwani put it more bluntly: Anyone not prepared to show complete devotion and unmitigated loyalty to the company should “get the fuck out.”

It may not be long before all of Theranos’ employees are out. In March the Securities and Exchange Commission charged Holmes and Balwani with fraud, stripping Holmes of her controlling stake in the company, fining her $500,000, and barring her from being an officer or director of a public company for 10 years. After laying off another 100 staffers, Holmes told investors last month that the company faces liquidation and may have to shut down as soon as July. Meanwhile, the U.S. attorney’s office in San Francisco is conducting a criminal investigation that could result in indictments of both Holmes and Balwani.

1These sources requested that I refer to them using pseudonyms, either because they feared retribution from the company, worried that they might be swept up in the Justice Department's ongoing criminal investigation, or wanted to guard their privacy.


Adapted from: Bad Blood: Secrets and Lies in a Silicon Valley Startup by John Carreyrou. Copyright © 2018 by John Carreyrou. Published by arrangement with Knopf, an imprint of The Knopf Doubleday Publishing Group, a division of Penguin Random House LLC.

(Note: When you buy something using the retail links in our stories, we may earn a small affiliate commission. Read more about how this works.)


More Great WIRED Stories