dragged on, news surfaced that Holmes would be hosting a fund-raiser
for Hillary Clinton’s presidential campaign at Theranos’s headquarters
in Palo Alto. She had long cultivated a relationship with the Clintons,
appearing at several Clinton Foundation events and forging a
friendship with their daughter. The fund-raiser was later relocated to
the home of a tech entrepreneur in San Francisco, but a photo from
the event showed Holmes holding a microphone and speaking to the
assembled guests with Chelsea Clinton at her side. With the election
eight months away and Clinton considered the front-runner, it was a
reminder of how politically connected Holmes was. Enough to make
her regulatory problems go away? Anything seemed possible.
I went back to my source and, this time, cajoled him into leaking the
whole inspection report to me. Running 121 pages long, the document
was as damning as one could expect. For one thing, it proved that
Holmes had lied at the Journal’s tech conference the previous fall: the
proprietary devices Theranos had used in the lab were indeed called
“Edison,” and the report showed it had used them for only twelve of
the 250 tests on its menu. Every other test had been run on
commercial analyzers.
More important, the inspection report showed, citing the lab’s own
data, that the Edisons produced wildly erratic results. During one
month, they had failed quality-control checks nearly a third of the
time. One of the blood tests run on the Edisons, a test to measure a
hormone that affects testosterone levels, had failed quality control an
astounding 87 percent of the time. Another test, to help detect prostate
cancer, had failed 22 percent of its quality-control checks. In
comparison runs using the same blood samples, the Edisons had
produced results that differed from those of conventional machines by
as much as 146 percent. And just as Tyler Shultz had contended, the
devices couldn’t reproduce their own results. An Edison test to
measure vitamin B 12 had a coefficient of variation that ranged from 34
to 48 percent, far exceeding the 2 or 3 percent common for the test at
most labs.
As for the lab itself, it was a mess: the company had allowed
unqualified personnel to handle patient samples, it had stored blood at