November
26, 2005
My Night With MAUDE
Ok. It's a movie reference and in this 1969
French film, "Ma
nuit chez Maud", the characters sit around and talk and
talk and make references to Pascal, the famous mathematician/philosopher.
They're talking about love; I'll be talking about money -- and the
uses of math in the pursuit thereof. And our lead character is MAUDE:
the Manufacturer and User Facility Device Experience database compiled
by the U.S. FDA.
In today's
Wall Street Journal, staff reporter Sylvia Pagan Westphal writes
of a new study commissioned by Boston Scientific -- it claims superiority
of its Taxus stent over the competing Cypher from Cordis/J&J. In
the article Boston COO Paul LaViolette states that his company hired
the Missouri-based CRO (Contract Research Organization) Certus
International to analyze the FDA's "adverse report" database
from January 2003 to June 2005. (I am assuming that would be the MAUDE
device database -- there is also the Adverse Event Reporting System
or AERS, but that's for drugs -- the article didn't specify which database
was used.) They came up with data showing the Cypher stent experienced
many more adverse events with regard to restenosis (reclosure) and
thrombosis (clotting). Here's a quick view of the results that Certus
certified:
Patients Who Experienced Adverse
Events
from Jan 2003 - June 2005 (source: FDA)
|
|
TAXUS
|
CYPHER
|
Restenosis |
110
|
580
|
Thrombosis |
331
|
589
|
Wow!!! The Cypher had five times more restenosis and
almost twice the thrombosis rate. Unbelievable!! Guess that puts all those
clinical trials to shame.
Unbelievable is the correct word, however. There
are so many things wrong with this data that it's hard to know where to
begin. Various Cordis officials complained in the WSJ that Boston Scientific
included a longer time period for the Cypher than for the Taxus (Taxus
went on sale in March 2004 -- the midpoint of the data set). Cordis also
said the data included Japan, where the Taxus is not yet approved for sale.
But dudes! You're all caught up in the details. Look
at the big picture. In the article Boston Scientific defends its sampling
method by stating that during the 30-month-period both companies sold a
similar number of stents, about 1.6 million each, so it's "even Stephen".
But in that very claim lies the flaw! Every clinical
trial shows and every cardiologist knows that the restenosis rate in all
drug-eluting stents is low -- somewhere around 5%. So do the math! (And
it's only 7th grade level math at that.) Stay with me here: 3.2 million
stents sold, maybe 2 stents per patient, that's 1.6 million patients, 5%
of all patients would mean there should be 80,000 patients who experienced
restenosis, not 690. That makes the FDA data off by a factor of over 11,000
percent! Do we have any way of determining which stent closed up in the
79,000 patients who are not in the MAUDE data? Not really. Is the MAUDE
data a valid statistical sample? The FDA doesn't think so (read on).
Ditto with thrombosis. The lowest rates are somewhere
in the 0.5-1.0% range (and that's in clinical trials -- Antonio Colombo's
group in Milan has shown that real-world
stent thrombosis rates are at least double or triple that). So even
at the very conservative 0.5%, there should have been 8,000 patients with
stent thrombosis, not 920.
The MAUDE
database is accessible to the public (after all, you paid for it).
Search it yourself. The MAUDE data is voluntary -- lots of adverse events
don't get reported for a number of reasons (and that's a subject for
another blog entry, perhaps a book).
What's the point of MAUDE then? Well, first off, it's
better than nothing. Secondly, it's quite instructive to read the stories
of device failures or problems, something all interventionalists should
do -- there's a wealth of information there about difficulties you might
encounter and ways to avoid or rectify them. Thirdly, if there is something
wrong with a device, one would hope the reports might spotlight a manufacturing
defect (Boston Scientific should know -- in 2004 reports
of balloon withdrawal problems got them to investigate and pinpoint
a laser weld problem, which was quickly fixed.)
My point is that while the FDA database of adverse events
may be useful for some things, it is not a comprehensive
registry or a randomized trial, and it is certainly not a
suitable well of data to use as fuel for the marketing spin machine. In
fact the FDA states right on the MAUDE
home page :
MAUDE data is not intended to be used either
to evaluate rates of adverse events or to compare adverse event
occurrence rates across devices.
Could it be clearer? For once the U.S. Government actually
writes in English.
Last
month I reported that Boston Scientific told Angioplasty.Org it had
formed a new internal "data messaging" unit to fight back against
the J&J propaganda that hurt Taxus sales since last spring's ACC
-- Cordis had used results from the REALITY trial to insinuate a safety
issue into the "Stent Wars" when in fact the numbers did not
support such a conclusion. But one would like to hope that the minds
at Boston Scientific, a company that has brought truly innovative and
life-saving technologies to the public, could and should do better than
spending their time and money trying to coax wildly inappropriate data
out of MAUDE.
Ah mon cheri, maybe, with the Liberté on
the horizon, they're just feeling French....
|