PAGE ONE


DOCTOR'S DILEMMA

A Risk in Cholesterol Drugs
Is Detected, but Is It Real?
Data Crunching Hints
At Tie to Lou Gehrig's;
FDA Isn't Concerned

By AVERY JOHNSON
July 3, 2007; Page A1

As he examined data on a computer one day last fall, drug-safety reviewer Ralph Edwards saw something that concerned him: Of 172 people in his database who developed Lou Gehrig's disease or something similar while taking prescription medicines, 40 had been on statins, the huge-selling cholesterol drugs.

Dr. Edwards, director of the World Health Organization's drug-monitoring center, has amassed about four million reports of medical problems experienced by people taking prescription drugs. His job is to sift through these so-called adverse events, looking for "signals" of potential side effects.

The number of Lou Gehrig's cases associated with statins struck Dr. Edwards as high. He would have expected a number in the single digits, judging from how often other drugs in the database were linked to the disease. Still, the analysis didn't prove anything. Dr. Edwards hesitated to publicize his finding, wary of creating a drug scare and mindful that statins have been shown to reduce heart attacks significantly.

INTERVIEW



Read excerpts from Dr. Edwards's conversations with The Wall Street Journal.It's an increasingly common dilemma nowadays. Sophisticated software allows health authorities to troll through huge databases looking for possible drug dangers. The data mining can detect rare side effects that didn't show up in clinical trials. But it can also raise false alarms and force regulators to divert time and money from more pressing dangers.

"People reach different judgments on when to shout and when not to shout," says Robert Temple, medical director at the Food and Drug Administration division that evaluates drugs. "It's the hardest single thing -- the value and danger to screaming early."

The issue is likely to grow as regulators expand access to their databases. The FDA and other national drug regulators used to keep a tight lid on the adverse-event reports they collected, but now the WHO can freely disseminate them. Also, since 2004, the FDA has posted its quarterly data files on the Internet, and anyone can download them free of charge. Some companies have started offering software to crunch the data for side-effect patterns.

Adverse-event reports are a major way to identify side effects after a drug is on the market, but the data can be flawed or misleading. The reports can be turned in by anyone -- patients, doctors, even plaintiffs' lawyers. Media coverage or Internet postings claiming dangers in a particular drug may lead doctors and consumers to report more problems with that drug. Meanwhile, many problems never get reported.


Determining causation is another issue. If people taking a drug are older or more overweight than usual, an adverse-events database may show more links to heart attacks. That wouldn't necessarily point to any problem with the drug -- it would just reflect the less-healthy state of the people taking it.

Clinical trials in which patients randomly receive either a drug or a placebo are usually the best way to identify side effects, because any significant difference between the two groups is likely to be a result of the drug. However, trials are often too small or too short to identify unusual side effects, and they don't necessarily reflect how patients take a drug in the real world.

Drug Monitoring

The WHO, a United Nations body, set up its collaborating center for drug monitoring in 1968 after the thalidomide scandal, in which pregnant women who took the medicine for morning sickness gave birth to children with stunted limbs and other malformities. Now based in the Swedish university town of Uppsala, the center pools adverse events from 83 countries, including the U.S.

Dr. Edwards, a 64-year-old Briton with a background in internal medicine and clinical pharmacology, took the center's helm in 1990. His first experience with puzzling side effects came during a stint at the University of Zimbabwe almost 30 years ago. There, he noticed some patients with a rare skin condition, and tracked down some antibiotics they were taking as the cause.

In the mid-1990s, Dr. Edwards's team developed a software program to mine data with the help of artificial-intelligence Ph.D.'s from a local university. The software made it easier to analyze adverse events for one drug in the context of all event reports and fish out signals of problems.

At first, Dr. Edwards needed permission to publish from the national drug regulators on whose data he relied. Then, at a conference in Hong Kong in 2002, national drug agencies including the FDA loosened their grip, allowing the WHO to publish without permission. Dr. Edwards also received the right to share his data with anyone who requested it, he says.


He gets about 200,000 new adverse-event reports and pinpoints about 60 serious signals a year. Usually he notifies national health authorities and goes no further. Only rarely does he publish his concern. One such case involved babies getting withdrawal symptoms when their mothers took certain antidepressants. That report appeared in 2005 in the Lancet.

Last fall, Dr. Edwards decided to see what his database could tell him about statins and Lou Gehrig's disease, known medically as amyotrophic lateral sclerosis or ALS. The progressive neurodegenerative disease is almost always fatal. He had heard from an American doctor about an ALS-like case that was seemingly related to statins. Dr. Edwards himself had a friend who developed another serious nerve disease, called peripheral neuropathy, after taking a statin.

With a few clicks, Dr. Edwards came upon the 40 cases of ALS in people taking statins, and his software told him the number was much higher than expected.

DATA DEBATE


• The Issue: An analysis by a drug-monitoring group suggests a link between cholesterol-lowering statins and Lou Gehrig's disease, but U.S. regulators say they don't believe there's a risk.
• The Background: Scientists are increasingly using data-mining software to sift through reports of adverse events involving drugs.
• What's Next: More studies to figure out what's happening.Reviewing scientific literature, Dr. Edwards came up with a theory about the cases. All statins carry a warning on their label about a rare disorder that causes muscle pain or weakness, a long-known side effect of the drugs. More recently, studies have linked statins to peripheral neuropathy. Dr. Edwards speculated these diseases and ALS might be connected, since they involve some kind of neuromuscular degeneration, and he thought all might be triggered by statins in unusual instances.

Yet Dr. Edwards also learned that pharmaceutical companies and other researchers are studying statins as potential treatments for brain or nervous-system diseases such as Alzheimer's, multiple sclerosis and even ALS. Some believe the anti-inflammatory effects of statins may be useful in these diseases.

As Dr. Edwards reflected on whether to publish his findings, one point weighed heavily. ALS is rare -- it strikes about five in 100,000 people and affects some 350,000 people world-wide. By contrast, heart attacks are a leading cause of death, and statins have helped millions of people reduce their heart risk. Dr. Edwards didn't want to raise unwarranted fears about drugs whose value is well-documented.

"We were on the horns of a dilemma and it wasn't easy to resolve," Dr. Edwards says.

Behind the scenes at the FDA, officials were facing a similar dilemma. Ana Szarfman, an agency official who pioneered statistical analyses of adverse events, had noticed the ALS-statin signal in March 2006, about six months before it caught Dr. Edwards's attention. She was using the FDA's database, which overlaps with the WHO's but isn't identical. The FDA says that through the end of last year, about one-third of the ALS adverse-event reports in its database -- 99 of 298 -- involved people on statins.

Dr. Szarfman was known at the FDA for her sleuthing into unexpected side effects, such as the odd link between a Parkinson's-disease medicine, Mirapex, and isolated cases of compulsive gambling. She also helped confirm in 2001 that Baycol, a Bayer AG statin later withdrawn from the market, was linked to an especially high number of muscle-disease cases.

Dr. Szarfman took the ALS signal seriously and considered publishing an article about it. But she held off while the FDA took a closer look. The agency didn't want to jump to any conclusions about a life-saving class of drugs, officials say. Unlike the WHO, the FDA has the ability to press drug companies for more information. That's what it did in this case, asking Pfizer and other statin makers for all their clinical-trial data.

Statin Trials

Taken together, the long-term statin trials involved 120,000 patients. The number of ALS cases came to 20, spread equally between statin takers and those on placebo.

"The result of the analysis was very reassuring," says Dr. Temple, who takes a statin himself. "I personally was relieved because I like my statins."

Dr. Temple says a close look at the FDA's adverse-event data suggested some reports might be unreliable. Slightly more than half of the adverse-event reports in the FDA's database about statins and ALS were submitted by consumers, which is unusual because doctors usually account for the lion's share. Also, some of the ALS cases were reported soon after the patients had started taking statins, raising doubts about whether the drug could have been the cause. Officials noted that the incidence of ALS in the general population hasn't risen since statins became popular.

The FDA plans to publish an academic paper about its statistical findings soon, says an agency official. But based on the analysis of clinical trials, it decided it didn't need to issue any caution about statins.

Dr. Edwards came to a different conclusion. He talked the issue over with his staff and P. Murali Doraiswamy, an expert in neurodegenerative diseases at Duke University who has done research on statins and the brain. Dr. Doraiswamy agreed that the ALS signal was worrisome and should be investigated.

After months of hesitation, Dr. Edwards came across a study by Greek researchers that helped make up his mind. Typically, ALS leads to death within three to five years, usually because the lungs fail when the neurons that send instructions to the respiratory system stop working. But in some rare cases, ALS-like symptoms could be halted or even reversed, the study said. Dr. Edwards decided his finding might help some patients prevent their disease from deteriorating. He wrote a paper and submitted it to two prestigious journals.

Both journals, the British Medical Journal and the Lancet, rejected it, he says. Fiona Godlee, the BMJ's editor, says she can't comment on specific papers, but she calls adverse events hard to interpret. "You only have a record of the people who developed problems, without knowing how many people took the drug and didn't develop problems," she says. The Lancet declined to comment.

Dr. Edwards pitched his paper to Drug Safety, a little-known journal based in New Zealand, which rushed it into print last month.

The paper acknowledges that some Internet sites have been discussing an ALS-statin link for at least a year, raising the possibility of skewed reporting of adverse events. But it notes that the number of events was also high earlier in the decade, before the majority of the Internet postings. The paper calls on patients using statins to talk to their doctor about stopping if they experience severe neuromuscular symptoms.

Pfizer, which brings in nearly $13 billion a year from Lipitor, says it too noticed the ALS signal last year in adverse-event reports, but concluded there was nothing to it after examining all its data. The company says Dr. Edwards has the right to publish what he found, but signals derived from adverse events can unnecessarily alarm the public and create needless headaches for drug companies.

Manfred Hauben, a Pfizer medical director, has long specialized in analyzing adverse events, and several years ago started using sophisticated software adopted by Pfizer. A math junkie who jokes that he suffers from an impulse-control disorder and stays awake at night studying statistics, Dr. Hauben says his enthusiasm for data mining has cooled.

He says with so many drugs and diseases, mere coincidence will create some worrisome-looking links. "Especially when you're looking for rare events, most positive results are going to be false positives," he says. Also, data mining presents "a lot of opportunities to retrofit an analysis to pre-existing expectations," he says. "Two different vendors' software can give two different results."

Prompting Research

Dr. Edwards says his paper is merely intended to prompt more research into the matter -- not cause millions of heart patients on statins to panic and stop the drugs.

But he hopes it will be useful to some patients. "Suppose you started to get symptoms and your doctor said, 'Now you have two years to live,' " he says. "Wouldn't you want to know that there's some possibility that the disease is linked to the drug so you could stop taking the drug?"

Dr. Edwards isn't persuaded by the clinical-trial data collected by the FDA. The signal wouldn't necessarily show up there, he says, because ALS is so rare.

Sheila O'Donovan, a 62-year-old retired journalist who lives in Delray Beach, Fla., wishes the signal had been publicized sooner. Ms. O'Donovan was on Lipitor when she started losing control of her right thumb in 2005. Soon, she was limping and slurring her words. After going off the drug in April of last year, she suddenly felt "brighter," her muscle cramps stopped and she was able to swallow more easily, she wrote in an email. But the improvement was short-lived. Her neurologist diagnosed her with ALS last September and she's gone downhill since. Her voice is now almost gone and she needs a wheelchair to get around.

More evidence could be available in a year or two. Beatrice Golomb, an associate professor of medicine at the University of California, San Diego, is analyzing case reports of people who developed ALS-like symptoms after taking statins. And researchers at Stanford University are looking at patient records in Northern California kept by Kaiser Permanente, a big health-maintenance organization, to see whether people who developed ALS were more likely than control subjects to have used statins.

Write to Avery Johnson at avery.johnson@wsj.com
http://online.wsj.com/article/SB1183...googlenews_wsj

'This Is at Least a Signal'
WHO's Dr. Ralph Edwards
On Data Mining Over Drugs
July 2, 2007 11:13 p.m.
Dr. Ralph Edwards of the World Health Organization's collaborating center for international drug monitoring spoke with The Wall Street Journal's Avery Johnson about his decision to go public with a hypothesis linking a condition remarkably similar to amyotrophic lateral sclerosis, also known as ALS or Lou Gehrig's disease, to cholesterol drugs called statins.

The move was controversial because Dr. Edwards's group digs around in side-effect reports turned in at random by consumers and doctors -- not the same as scientifically rigorous clinical trials data. But Dr. Edwards stands by his choice, and thinks it offers a lesson to multinational drug companies and regulators. Below, some edited excerpts from a series of discussions. (See related article.)

* * *

On data mining and his methods:

A signal is a set of data constituting a hypothesis that's relevant for the safe and rational use of a drug. It consists of a hypothesis together with data and arguments.

Data mining is just a tool that's used to screen a huge amount of data. It doesn't remove the need for very careful judgment. Just because you come up with a number doesn't mean anything. It's just the start of the process.

There is the potential to misuse numerical data. [In the decision to open up his data to more people and for publication,] there was a real concern that the information would be misinterpreted. If this gets into unscrupulous hands, it could create problems.

Data mining is still really under discussion as a tool. Everyone's saying now, "Forget spontaneous reports [from patients or doctors]. They're unreliable." But if you go to the health-care database with complete patient records, how do you find anything? You have to use data mining. The industry uses the argument that the data is bad and you can't take any notice of it. They say all those reports are poor quality because you don't have the age and the duplication problem.

I've been in debates where you have hundreds of adverse-event reports but they're imperfect. Can you disregard 100 adverse-reaction reports just because of the way the information is transmitted? Someone must have made a link in order to send it in.

On his decision to publish his "signal":

What we have here is 40 case reports. I can't see any other reason why they're reported to us. Other neurodegenerative diseases are not reported in excess. We thought, "This is at least a signal."

All of our reactions were, "If this is true, this is a very important finding. If this is a cause-and-effect relationship, it needs working out as fast as possible."

I'm very aware that anyone who looks at the signal is going to say it's a difficult one. The disease is rare and deadly. Anyone who has it is going to look around and say, "My God, why me? Why should I have this?" And one thing they're going to think is, "Is it the drug?" They or their family are going to report it, and that could be an explanation [for the increased reports]. But if I have Alzheimer's, that could also happen. Why ALS? Why not Parkinson's or MS [multiple sclerosis]? Somehow, this has triggered off a number of reports.

For God's sake, if I had a really high lipid level, the benefit of statins is high and the theoretical risk is trivial. That's the kind of decision to make individually. Maybe you're someone with ALS in the family. If your lipid level is on the border, there's no risk short-term to stopping the drug.

This one is probably the most difficult decision I've had to make. The rarity and yet the severity of the adverse reaction and the possibility that if it turned into a drug scare millions would be concerned about taking the drug, and the risk of stopping it is immense.

On the drug industry and regulators:

I tend to believe that our problems in drug safety are more caused by attempts at secrecy until we have more certainty. The industry's problem is that nobody trusts them, and it's because they don't come out and say things earlier.

The drug industry least of anybody wants a drug scare. So their level of proof is likely to be higher than mine. People have threatened our positions, and tried to stop publications. It can get quite nasty.

The FDA's approach is that they need more information before they go public. One issue regulators point out is if we make data more available, what happens if we [the WHO] draw different conclusions or act sooner than they do? Are we making them look silly?

Anytime this comes up we agonize that it's going to create a drug scare. But whilst we've been sitting on information lots of people could've been damaged by the drug if the signal is correct. At what point should anyone go to the public and say there might be something wrong with this drug?

The bottom line is, what is it that health professionals and the public need to know? I think a lot of our thinking depends very much on what we think the public wants to know and needs to know. We take very little effort to have the public really be on board with this, to really educate them. We don't communicate, and that's upside-down. If we can't explain the problem, we have a problem ourselves.

We give the benefit of the doubt to the drug. Why shouldn't we give the benefit of the doubt to patients? Given the clear sense of concern on the Web, the FDA could've said, "If this is a real risk, it must be very rare." They should say what they're doing. After a plane crash, [the government] says what it's going to do.

On his role and the role of his WHO center:

Our function is to act as a failsafe on drug safety issues.

The dilemma for us in our position is somewhat different than the dilemma drug regulators have, in that a lot of their work is with the pharmaceutical industry. In most countries, any release of information on drug safety follows a negotiation between regulatory agencies and industry.

We have a slightly different philosophy. In this case, ALS is a very rare condition. I think if we can get this out there and say, "Be on the lookout and in individual cases stop the statin for a while" -- we may get more information that way than from setting up a study. It's a matter of very considerable debate and is an issue that I and many people find very troublesome indeed.

I'm a crusader for finding the best way to communicate information. I work very hard for those ideals. We are that little voice, we're not bound by the ties that bind them [regulators].

People should listen to what we have to say, look at the evidence we produce and question it. We're health professionals with a lot of experience in the field. Our knowledge of the data is immense.

On two prestigious journals' decisions not to print his paper:

[I found it] really irritating. The drug-safety sections of both [papers] are very limited -- the maximum is 800 words. I tried to compress 40 cases into 800 words and it didn't work. In both cases, they said this is a matter for a specialist journal.

I was surprised they didn't publish it. The journals have said they haven't changed. But they've had a lot of criticism for publishing the vaccines and autism signal. I suspect that the level of evidence of spontaneous reports is not good enough anymore. They used to like spontaneous reports.

On why he thinks clinical trials might not turn up an ALS signal:

ALS is rare. I think it just wouldn't show up, or wouldn't necessarily show up in clinical trials. [If they did see it,] they'd have no more than a half dozen patients with ALS by chance.

If they saw any patients at all, it could be due to chance. If the drug caused a 20% increase that would only be a couple more patients -- 14 in the drug arm and 12 in the control arm. Would you find that very convincing? No.

I'd imagine that data coming from various studies would also make it difficult. Meta-analyses compare apples to pears.

How were the adverse reactions recorded? You don't know how the patients are selected, whether their age is representative. [The trials are] relatively short term. Because ALS is a progressive disease, its onset might be rather slow.

On whether the signal he's identified will bear out:

I want more information to decide for myself.
http://online.wsj.com/article/SB1183...googlenews_wsj

Daily Health Policy Report

Prescription Drugs | Researchers Use Data Mining To Find Possible Rare Side Effects of Prescription Drugs
[Jul 03, 2007]
The Wall Street Journal on Wednesday examined prescription drug data mining, a process by which sophisticated software enables health officials to search through large databases looking for possible drug dangers. Data mining software allows health authorities to identify "rare side effects that didn't show up in clinical trials," the Journal reports. However, "it can also raise false alarms and force regulators to divert time and money from more pressing dangers," according to the Journal.

The Journal profiled the experience of World Health Organization Drug Monitoring Center Director Ralph Edwards. Edwards and his team in the mid-1990s developed software to mine drug data, and national drug agencies, including FDA, in 2002 allowed the center to publish and share data mining findings without permission. Edwards, who receives about 200,000 adverse-event reports and identifies about 60 serious signals annually, last year discovered a possible link between cholesterol-lowering statins and amyotrophic lateral sclerosis, or Lou Gehrig's disease.

However, the analysis based on data mining did not "prove anything," and Edwards was "wary of creating a drug scare and mindful that statins have been shown to reduce heart attacks significantly," the Journal reports. After "months of hesitation," Edwards published his findings in the journal Drug Safety and recommended that patients taking statins consult their physician if they experience any neuromuscular symptoms.

FDA also studied the connection between statins and ALS but determined that "it didn't need to issue any caution" about the drug, the Journal reports. Robert Temple, medical director at the FDA division that evaluates drugs, said, "People reach different judgments on when to shout and when not to shout. It's the hardest single thing -- the value and danger to screaming early."

Edwards said his paper is intended to prompt more research of the possible connection, adding that FDA's clinical trial data might not show the risk for ALS because it is so rare (Johnson, Wall Street Journal, 7/3).



http://www.kaisernetwork.org/daily_r...fm?DR_ID=46001

If you would, please send along an image of the Wall Street Journal page... for
'This Is at Least a Signal'
WHO's Dr. Ralph Edwards On Data Mining Over Drugs

I'm interested in seeing the same page on which this appeared in the online Wall Street Journal.

Cheers! and kind regards,
don saklad

email
dsaklad@zurich.csail.mit.edu