Colby Cosh: The drug that was approved on fake data

Health Canada has asked physicians to stop prescribing Tavneos (avacopan) to new patients as a result of “significant concerns” with the main clinical trial that initially supported approval of the medication for the treatment of a rare auto-immune condition. Photo by Unsplash

Article content

On Tuesday, Health Canada published a circular asking physicians to stop prescribing an obscure drug, Tavneos (avacopan), to new patients. We probably don’t have very many readers who are taking avacopan: it’s a drug approved recently (autumn 2021 in the U.S., spring of 2022 in Canada) for treatment of a rare and poorly understood autoimmune condition called “ANCA-associated vasculitis.” In western countries this disorder afflicts 10 or 20 persons per million a year, usually elderly ones. The symptoms, caused by out-of-control antibodies killing off interior blood vessels, range from “feeling poorly” to “choreographed and hasty organ failure.”

Article content

Article content

Health Canada’s terse instruction mentions “significant concerns” with the main clinical trial that initially supported international approval of avacopan. That study, which reported the results of a global trial involving 331 patients enrolled from 143 hospitals around the world, appeared in the New England Journal of Medicine (NEJM) in February 2021. It bore the names of four principal authors, two of them university professors and two of them employees of the avacopan patent-holder, ChemoCentryx. On June 29, NEJM announced the formal retraction of the study, noting that “the two academic authors” had asked for it to be withdrawn. The notice adds:

Article content

Article content

Article content

“… according to an ongoing Food and Drug Administration investigation conducted after publication, and without the knowledge of these two authors, the primary end-point assessments in nine patients were readjudicated after database lock and trial unblinding. This was not disclosed in the article and is inconsistent with proper research conduct.”

Article content

If you’re familiar with present-day scientific practice, you’re probably thinking that there’s a story behind this jargon, and it might be a story of marginal and essentially bureaucratic sloppiness, and then again it might be a tale of conscience-shocking malpractice. The U.S. Food and Drug Administration (FDA), it turns out, pulled back the curtain in April, sending out a warning that its drug-evaluation branch was thinking of taking avacopan off the market altogether.

Article content

Article content

In a letter to the regulatory-affairs office of Amgen, which bought ChemoCentryx after the drug was approved, the director of that FDA sub-agency outlines the whole mess. It was uncovered, in typical American fashion, when a securities-fraud lawsuit was filed against the company (before the Amgen acquisition) and an outside expert was hired to do a forensic investigation of the NEJM study. Even in the originally published version of the study, the numbers supporting avacopan’s effectiveness were hardly sensational, and some critics had raised purely statistical concerns about the study’s bona fides, leading to a bit of dialogue in the pages of NEJM.

Article content

Article content

What the full investigation revealed was that the data on the patients in the study had been officially “locked” in place, and the blinding of individual outcomes removed, on November 5 of 2019. At that point, two ChemoCentryx employees were allowed to peek at the full data; they quickly realized that the “superiority” effect of the drug hadn’t quite reached statistical significance.

Article content

According to emails quoted in the FDA notice, the unblinded researchers, whose company must have spent tens of millions of dollars testing avacopan, began to panic. “We cannot afford to miss a superiority outcome here,” one of them wrote, possibly violating what is now known as the Stringer Bell Rule. The ChemoCentryx employees broke the database “lock,” found a few patients whose ANCA disease progress could be “readjudicated” favourably, and were able to shift enough of them around to achieve significance and re-lock the data.