The Body Hunters (20 page)

Sepsis felled hundreds of thousands of Americans ever year,
⁴⁶
including 70 percent of those who died in intensive care units.
⁴⁷
It is a mysterious ailment in which the normal inflammation that comes in response to a bout of infection fails to subside.
⁴⁸
The consequences can be dire. Blood coagulates into vessel-blocking clots. Fever, shakes, and vomiting may follow. The organs may fail. (When it gets this bad clinicians call the condition “severe sepsis.”
⁴⁹
) One-third with severe cases die within a month.
⁵⁰
Worse, nobody knows for sure why sepsis occurs, or can predict who might fall prey. With the medical establishment shelling out $17 billion every year treating septic patients, interest in developing new treatments runs high.
⁵¹

But it isn't easy. In the early 1990s the biotech company Centocor had attempted to launch a sepsis drug called Centoxin. The drug, a genetically engineered version of a human protein thought to be deficient in patients with sepsis, had protected animals from septic shock in some early lab testing, but follow-up studies had failed to replicate the promising results. Centoxin's performance in the 543 septic patients in the company's first major human trial was lackluster: as many patients died on the drug as did on placebo. But with $200 million already invested in Centoxin, the company was loath to let go.
⁵²
In a 1991
New England Journal of Medicine
paper the company re-analyzed its data, separating the gray results into their black and white components. While overall death rates hadn't changed in the drug and placebo groups, in certain sepsis patients—those with pathogenic bacteria in their blood, a condition known as gram-negative bacteremia—the death rates between the group on the drug and the group on placebo had indeed diverged, from 49
percent on placebo to 30 percent with the aid of the drug, the company explained.
⁵³

USA Today
enthused that Centoxin could save tens of thousands of patients every year. “The impact of this is like . . . introducing a drug like penicillin,” the paper quoted one awestruck physician.
⁵⁴
Centocor stock soared to nearly $60 a share.
⁵⁵
But if some patients had improved on the drug while overall the fate of the entire group of patients hadn't changed compared to placebo, critics pointed out, then clearly some patients—those who didn't have gram-negative bacteremia—had been actively killed by the med. Figuring out which patients would benefit and which might be harmed wasn't straightforward. The tests that confirmed a case of gram-negative bacteremia could take up to forty-eight hours.
⁵⁶
If docs felt pressured to administer Centoxin before they could figure out whether a patient might benefit from it, sepsis patients who did not have gram-negative bacteremia might start dropping like flies.
⁵⁷

When the FDA found other anomalies in Centocor's data, they ordered the company to start a new trial. In the new trial the benefits seen in the earlier one vanished: death rates among patients with gram-negative bacteremia were no different in the drug group than among those on placebo.
⁵⁸
What's more, over three hundred subjects given the drug died, twenty-eight more deaths than would have been expected had they been given no drug at all.
⁵⁹
Stock fell to less than $7 a share and the company laid plans to dump its aging stockpile of Centoxin.
⁶⁰

With Centoxin's high-profile crash and burn—the
New York Times
ran a lengthy front-page story on the drug's spectacular failure in 1993
⁶¹
—Eli Lilly would face a skeptical medical elite with its sepsis contender, Xigris. Like Centoxin, Xigris was a biotech version of a naturally occurring molecule, activated protein C, also thought to be deficient in patients with sepsis.
⁶²
Like Centocor, Lilly planned to charge an arm and a leg for its drug—$7,000 for a single course of Xigris. After all, unlike most areas of medicine, critical-care medicine hadn't been subjected to cost-cutting
pressures. “The feeling has always been that if you're sick enough to be in the ICU, you're sick enough to let the doctors do what they need to do,” Brown University critical care specialist Nicholas Ward said. “In critical-care medicine, we've pretty much had a Gold Card to do what we want.”
⁶³

Such carte blanche wouldn't have held had a cheap, effective therapy been available, of course. And a slow trickle of papers had appeared over the course of the 1990s showing that, at least anecdotally, low doses of corticosteroids were highly effective for sepsis. Corticosteroids were old drugs with well-understood properties, and a course of generics might run to about $50 or so.
⁶⁴
But when steroid researchers approached various drug companies to sponsor large-scale controlled trials to prove the drug's efficacy beyond a shadow of a doubt, they were roundly rebuffed. Patents on steroids had long expired and so the drugs were widely available for a pittance. No jackpot of extra sales income awaited anyone who proved how such drugs worked or whether they saved lives.
⁶⁵

Xigris, on the other hand, had one of the world's biggest drug companies behind it, and Lilly wouldn't cut corners. In the company's first trial of the drug, known as the PROWESS trial (Recombinant Human Activated Protein C Worldwide Evaluation in Severe Sepsis), they roped in nearly 1,700 patients at 164 centers in eleven countries for a massive placebo-controlled trial.
⁶⁶
In contrast, the first placebo-controlled study of steroid therapy had enrolled just thirty-one patients.
⁶⁷
The results were impressive—low-dose steroids had saved sixteen septic patients, while more than half of those given placebos had perished—but even wowed critical care experts had to admit that the trial was minuscule. The lead researcher for the PROWESS trial was Vanderbilt University's Gordon Bernard, MD, a researcher who had debunked high doses of steroids as a treatment for sepsis in 1987. Bernard had precious little truck for its low-dose cousin regimen, calling research supporting it weak and the investigator who led it stupid at a conference in Chicago.
⁶⁸

By June 2000, not even two years into the PROWESS trial, Xigris had proven so effective that the experiment was summarily discontinued. It would have been “unethical” to continue enrolling placebo patients, Lilly officials said, and deny them the lifesaving properties of Xigris. The drug had dropped the mortality rate by six percentage points, from 30 percent on placebo to 24.1 percent on the drug. The company analyzed the data every which way and yet the positive outcomes persisted. The effect was “consistent” across all the various subgroups of patients, whether they were only mildly ill or on death's door, elderly or youthful. The main adverse effect was serious—sometimes fatal hemorrhaging—but occurring in just 3.5 percent of patients, neatly balanced by the drug's benefits.
⁶⁹

Eight months later the results appeared in an early release edition of the
New England Journal of Medicine
, a privilege reserved for papers with urgent public health implications.
⁷⁰
The rapid notice wouldn't have made much difference to public health—the FDA hadn't even approved the drug yet so it wasn't available to physicians—but it was key to Lilly's efforts to build a buzz about the drug before it hit the market. The same month that the paper appeared the company arranged for Bernard to cohost a thirty-minute Fox cable television special explaining the dangers of sepsis.
⁷¹
Lilly enlisted leading sepsis experts to start spreading the word about Xigris.
⁷²

But when the FDA analyzed the PROWESS data, they found that Xigris hadn't worked for all patients with severe sepsis all the time, as the company's
New England Journal
paper had said. It had only worked for half the patients—those who were most sick—and even for them it had worked only
after
the company revised the protocol midway through the study.

In the trial, in order to ensure that the septic patients who received the drug were roughly similar in age, severity of disease, and other indicators compared to those who received the placebo, Lilly had evaluated them according to a scoring system, the Acute Physiology and Chronic Health Evaluation, or APACHE II. The scoring system runs from 0 to 71, with higher scores signifying greater
likelihood of death. When the FDA looked closely at the results it found that patients' APACHE II scores correlated with their responsiveness to Xigris. For the sickest patients, those with APACHE II scores over 25, the death rate on Xigris had dropped by 12 percent compared to placebo. But for those with scores between 20 and 24 the drug had dropped the death rate by only 4 percent. For patients given Xigris who had APACHE scores under 20, the drug had actually
increased
the death rate by 3 percent. Worse, these patients suffered from the drug's side effects—dangerous hemorrhaging—at the same rate.
⁷³
The company had also altered the trial protocol midway through the experiment; in the first half Xigris had shown no efficacy whatsoever. Xigris only started to work after Lilly restricted the study to “better quality patients,” as a Lilly researcher explained to a dumbfounded FDA advisory committee.
⁷⁴

The committee was split, ten in favor and ten against, but the FDA approved Xigris in November 2001 anyway.
⁷⁵
The agency added a single caveat: until Lilly provided more data on how the drug worked for patients with low APACHE II scores, the drug's use would be restricted to patients with APACHE scores over 25. It was a restriction doomed to fail, since the scoring system itself was a “moving window,” as a Lilly researcher admitted. If the patient qualified in the morning, could the drug be given in the afternoon? Who knew? Lilly would be banned from promoting any “off-label” uses, but doctors would be free to use Xigris, as with any other approved drug, however they saw fit. Although ignoring the restrictions could result in “very serious consequences for the patient,” as one FDA adviser acknowledged, many clinicians might sensibly decide to do just that. Besides posting a lengthy transcript of the advisory committee meeting on an obscure section of its Web site, the agency never publicly rebutted the company's rosy
New England Journal of Medicine
paper touting Xigris's efficacy for all sepsis patients.
⁷⁶

With the FDA stamp of approval, the marketing of the drug began in earnest. “New hope for taming deadly shock,” ran the
New York Times
headline on a long, approving article, featuring quotes
from a single sepsis expert—Gordon Bernard.
⁷⁷
Lilly dispatched over two hundred critical care specialists, at $1,000 a pop, to deliver promotional lectures about the new drug. It splurged on a special George Benson concert for docs likely to prescribe the drug.
⁷⁸
In an especially lucrative coup, the company convinced the Centers for Medicare and Medicaid to cover their patients' hefty Xigris bills.
⁷⁹

In September 2002, alarmed by the FDA's silence on the potential dangers of Xigris and Lilly's glowing marketing material on the drug, a clutch of critical care experts who had participated in the advisory committee meeting went public with their concerns in the
New England Journal of Medicine
. “We believe there is not sufficient evidence at present for it [Xigris] to become standard of care,” Massachusetts General Hospital's H. Shaw Warren, the National Institutes of Health's Anthony Suffredini and Peter Q. Eichacker, and University of Texas's Robert S. Munford wrote.
⁸⁰

Sales of the drug flagged. In order to realize the full sales potential of Xigris, the company would have to overcome the confusing restrictions the FDA had affixed to the label, industry analysts commented. But proving the drug worked in less sick patients could be “impractical and expensive beyond belief,” one sepsis expert said.
⁸¹
To prove any shred of efficacy, more subjects taking the drug would have to be shown to survive than those taking placebos, but if the FDA analysis of the PROWESS trial were borne out, it was unlikely that there would be much of a contrast between their fates. To show a difference, either huge numbers of patients would have to be enrolled or the trial would have to be conducted in places where such patients generally died faster and more frequently anyway. Then again, if the PROWESS trial results were borne out, the drug could actively kill such patients. Those most vulnerable were those who were the least sick to begin with, to boot. They had the best chances of survival without treatment.

Lilly, like other big drug companies, had decided to use India as a major center for its clinical trials in 2002.
⁸²
And so the company
looked to India to provide a significant number of subjects for its trial testing Xigris against placebo in sepsis patients with APACHE scores under 25, called the ADDRESS trial (Administration of Drotrecogin Alfa [Activated] in Early Severe Sepsis).
⁸³
In India the potential dangers of the drug for less ill patients weren't widely known, and in 2003 Lilly had even applied to the Indian drug controller's office, unsuccessfully, for permission to sell the drug for all patients with sepsis, not just those with high APACHE scores. As far as investigators like Kapadia were concerned, the main problem with Xigris wasn't its safety profile in less sick sepsis patients, but its cost. Kapadia had read the September 2002 critique of the PROWESS trial in the
New England Journal of Medicine
, but the controversy, according to him, revolved “around the price rather than the drug.”
⁸⁴
In the ADDRESS trial, the drug would be offered for free, a particularly appealing option given that it otherwise cost Indians between 600,000 and 1,000,000 rupees (U.S. $13,000 to $21,000) for a course. Poor Indians had been known to forego lifesaving surgeries on their children for a fraction of that price.
⁸⁵