By Jake Rossen
By Jesse Marx
By Michelle LeBow
By Alleen Brown
By Maggie LaMaack
By CP Staff
By Jesse Marx
Meanwhile, the firm's international database of adverse events was building a private record of the factors that led to tendon rupture. By the end of the year, Johnson & Johnson quietly added the sentence to the Levaquin label noting that the risk might be elevated in people taking steroids, especially the elderly.
In May 2002, Great Britain's version of the FDA, the Medicines Control Agency, proposed changing the drug's label to say that it was more damaging to tendons than other drugs in its class. If that happened, no British doctor in her right mind would continue prescribing it.
Aventis convinced the regulators to hold off just a bit longer—long enough to see the results of the American study.
Two months later, Ingenix completed an interim study report and handed over the findings. They turned out to be quite fortuitous for companies selling the antibiotic, revealing no greater risk in taking it than with any other drug in its class. In fact, compared to the rest, Levaquin's parent drug—which was supposedly nearly identical to Levaquin—was the drug least likely to cause tendon problems, according to the report.
With this tentative conclusion in hand, Aventis got the European regulators off its back.
Almost a year later, Ortho-McNeil sent its near-final report to its partners for review. The response from Aventis was more shock than pleasure.
"I was surprised and puzzled by the results of this report and by some of the conclusions derived from the findings," wrote one of the Aventis reviewers.
Nonetheless, Tavanic had managed to avoid having to add a warning to its label. The final report of the Medicines Control Agency recommended "[n]o further action given the rarity and non-lethality of adverse reaction."
Triumphant, Johnson & Johnson prepared its research for publication in a lauded peer-reviewed journal. The company submitted an explanatory poster to a noted microbiology conference in the nation's capital, and it created a press guide to be sure journalists got the story straight.
"There is enormous commercial risk if these presentations are misunderstood," wrote David Grewcock, director of marketing for Levaquin, in an email to his colleagues. "We need to clearly explain everything from why the studies were conducted to what the results mean."
AT THE BOTTOM of a document printed on official letterhead, Drs. Arnold Widen and Babs Waldman penned their names.
It was May 2005, and the two physicians were urging the FDA to require a black box warning for drugs like Levaquin.
As medical directors for the Illinois Attorney General's Office, they knew they had a strong case against Levaquin. For months, they'd been hearing complaints from patients who'd suffered ruptured tendons while taking the drug.
"On behalf of the Office of the Illinois Attorney General, we are writing to you to request action by your agency in regard to a medication side effect, which although already recognized, is not well-known by practicing physicians and other health care professionals," the doctors wrote. "We are referring to the condition of fluoroquinolone-induced tendonopathy, including actual tendon ruptures, which can be multiple, can cause severe and protracted disability, and can require surgical correction."
The evidence against fluoroquinolones was stacking up. One case study described an 80-year-old man in France whose left Achilles, right rotator cuff, and left biceps tendons all ruptured after he took Tavanic. Another report confirmed the significant risk of rupture in patients taking drugs like Tavanic with steroids, especially the elderly.
But the best journals turned the paper down.
Finally, in June 2005, the less prestigious Pharmacoepidemiology and Drug Safety agreed to publish the Ingenix findings. But the editors told the authors to acknowledge what might be a major shortcoming in the study's design.
A common problem with using insurance databases for epidemiological studies is that the medical codes that doctors write in patient charts do not always reflect the final diagnosis. A physician might jot down the code for tendon rupture in the chart of a patient who has only a sprained ankle, for example, because the doctor wants to rule out tendon rupture with an X-ray.
The only way to be sure that the code mirrors the diagnosis is to review each chart by hand. But Ortho-McNeil didn't have time to do this.
Instead, the Ingenix team created an algorithm that was supposed to accurately predict which records in the database were real tendon ruptures, and which were false reports.
The algorithm was based on the company's review of a random sample of 328 of the medical records. For each chart, the researchers visited doctor's offices and pulled the hard copy, carefully confirming the diagnosis.
They found out that only 58 percent of the sample records were the kinds of Achilles tendon rupture they wanted to count. The rest were mostly rule-outs and ruptures due to trauma—weekend warriors wrenching their heels.
The results from the sample helped Ingenix craft an algorithm it then applied to the rest of the records. But as the journal editors noticed, this approach comes with a potential flaw: It could make the connection between tendon problems and drugs like Levaquin appear smaller than it actually is.