The FDA estimates probably are low. Doctors aren't required to report malfunctions even if they suspect a patient death was due to one and not to the underlying heart disease, a sometimes difficult distinction.

About 20 of every 1,000 defibrillators implanted are malfunctioning, and defects led to 31 deaths between 1990 and 2002, concluded research sponsored by the Food and Drug Administration. The deaths represent only a fraction of the more than 400,000 defibrillators implanted during those years, but a troubling number nonetheless as regulators grapple with how to better ensure the safety of hard-to-replace devices.

Despite the defects, defibrillators "have saved many lives and the benefits of the devices clearly outweigh the risks," said Dr. Scott Gottlieb, an FDA deputy commissioner.

But, the study "does alert FDA that there is a trend that needs to be addressed, and points out the need for our agency to improve the way it regulates these products, and we are doing just that," said Dr. Daniel Schultz, FDA's medical device chief.

Tough Balancing Act

The study came as the FDA and cardiac specialists opened a daylong meeting Friday to discuss safety problems with implanted defibrillators and pacemakers, and how to decide which frightened patients need to have the devices surgically replaced when manufacturers learn that some parts might fail.

That can be a tough balancing act: Failures are rare, and ways to predict whose implant will fail don't always exist, while surgery to remove and replace the implant itself is risky.

But it's of growing public concern, because since January, the nation's three leading defibrillator makers -- Guidant, Medtronic and St. Jude Medical --- have issued recalls or safety warnings affecting more than 200,000 defibrillators. Guidant in particular has come under fire after acknowledging that it waited three years before alerting doctors and some 24,000 patients about an electrical system defect with one of its recalled models ultimately linked to two deaths.

FDA is considering changes to the way it monitors such devices, Schultz said.

Device Hardware Problems

The study, conducted for FDA by Dr. William Maisel, a Harvard medical professor, found that from 1990 to 2002, there were 2.25 million pacemakers and almost 416,000 implanted cardiac defibrillators, or ICDs, implanted in the United States. More than 17,000 of them -- 8,834 pacemakers and 8,489 ICDs -- had to be surgically removed because of confirmed malfunctions, Maisel found.

These were not simple problems like battery depletion. Eighty percent were device hardware problems, such as with electrical connections.

More disconcerting than the actual number, FDA said, was the trend: In the mid-1990s, some 7.9 ICDs per 1,000 implants were replaced because of malfunctions. That rose to a high of 38.6 per 1,000 implants in 2001 before dropping slightly to a rate of 20.7 the following year.

More than half of the malfunctions occurred during the study's last three years.

In contrast, the rate of replacement for malfunctioning pacemakers steadily dropped during the study, to a low of 1.4 per 1,000 implants. There were 30 confirmed deaths due to malfunctioning pacemakers during the 12-year period, a lower rate than for defibrillators because so many more pacemakers were implanted.

Likely Low Estimates

Pacemakers use mild electrical current to speed a slow heartbeat. Defibrillators use a bigger jolt of electricity on hearts that go into life-threatening irregular beats.

Pacemakers have been around for decades, but only since the mid-1990s have ICDs become fairly common -- and these devices are becoming increasingly complex, as manufacturers make them smaller and add pacemaker features.

These likely are low estimates, FDA officials said Friday. The study tracked only surgically removed devices for which the manufacturer later confirmed a malfunction. Moreover, doctors aren't required to report malfunctions even if they suspect a patient death was due to one and not to the underlying heart disease, a sometimes difficult distinction.