Cos., Gov't Seek to Keep Lid on Metadata

BOSTON Feb 3, 2006 (AP)— When the New England Journal of Medicine used a word-processing function to reveal that Merck & Co. had deleted study data about Vioxx and heart attacks, the pharmaceutical giant joined a long line of organizations bitten by information lurking in electronic files.
It's happened to no less than the White House, the Pentagon, the British prime minister's office and the United Nations.
Each time, making minor electronic adjustments to documents aired juicy details not meant for public disclosure such as the true author of a file or sensitive data hacked from a final draft.
The pitfalls of such hidden "metadata" have been long known in computer-savvy circles, but these high-profile leaks are driving new efforts to keep a lid on metadata.
So sensitive is the topic for the U.S. government that the National Security Agency released guidance in December on how agencies can properly redact reports.
For the corporate world, several companies are finding success in selling tools to automatically scan for and remove metadata.
Metadata is data about data. A word-processing document, for instance, has metadata on who authored it, when someone saved it and what that person did to it. Microsoft Corp.'s Word has a "track changes" feature that preserves a file's original text and shows another person's edits. All that is metadata.
This information is designed to stick around because it can help people organize their files and collaborate with one another.
But because it doesn't show up when a document is printed and doesn't appear on screen in normal settings, it's easy to forget about.
In the Merck case, the company said the Vioxx data that was uncovered by the New England Journal had been deleted merely because the heart attacks in question came after a cutoff date for collecting information in the study.
Fears about the hazards of metadata led Microsoft to pull back on a planned feature in Vista, the upcoming Windows operating system. Originally, Vista was going to let users drag and drop files into certain spots on the desktop in order to label documents with personalized categories. Currently, grouping files by category is more laborious.
But Vista testers told Microsoft it might become "a little too easy" to apply the categories and have them stick permanently to the document, said Mike Burk, a Windows product manager. That could get ugly if a file categorized in "projects I hate" were e-mailed, say, to a boss.
Meanwhile, the next generation of Microsoft's widely used Office software, which includes Word, Excel and PowerPoint, will make it simpler to strip metadata from files before they are disseminated.
Even so, Gartner Inc. analyst Michael Silver says the problem will remain metadata will exist in documents unless users make a point of getting rid of it.
"It's still a manual process. It's still something you have to remember to do," Silver said. "Any time you're relying on the user to remember something, there's a good chance that they'll forget."
And so it wasn't surprising that when the White House posted a policy paper about strategy in Iraq last year, a quick command revealed the author. What was significant was that it was penned outside the administration, by Duke University political scientist Peter Feaver, a National Security Council adviser.
Earlier, a United Nations report on the assassination of former Lebanese Prime Minster Rafik Hariri developed new layers of intrigue when it was revealed that damaging accusations about Syria's involvement had been removed before publication.
The NSA's December paper ("Redacting with Confidence: How to Safely Publish Sanitized Reports Converted from Word to PDF") could have helped the Pentagon prevent its hidden-data episode last May.
Before posting a report in Adobe Systems Inc.'s Portable Document Format about a U.S. soldier who had accidentally killed an Italian secret service agent in Iraq, officials covered up classified information with black bars. But there's a difference between covering and deleting information. Readers simply uncloaked the text by cutting it from under the black and pasting it elsewhere.
Automated tools to help protect against metadata releases have existed for a while, but they are beginning to see wider use.
For example, Workshare Inc. sells a product called Trace (a free version can be downloaded) that scans documents for metadata and ranks the findings by risk level. One way a high risk is assigned is when a document lists all its authors and storage locations, because such data can guide hackers.
For most of Workshare's six years in existence, the company's customers were primarily lawyers, who are particularly sensitive about client information escaping to the opposing side.
But in the past year, Workshare has seen business expand to 60 percent of the Fortune 1000, said CEO Joe Fantuzzi. Revenue has surpassed $25 million, and Fantuzzi believes metadata protection is on the verge of being a must-have for corporate technology buyers.
"I think it could just break out, just like antivirus," he said.
Unlike antivirus protection, however, the tools to strip metadata from documents rest in everyday users' hands. Microsoft has long made it possible for users to erase telling metadata before documents are disseminated. The problem is that the steps are often cumbersome or obscure.
Microsoft hopes to resolve that in part by adding a metadata scan as a "file" menu option in Office programs. And a tweak in Vista could help.
Despite having yanked the drag-and-drop metadata tool, Microsoft is adding new labeling functions to Vista, including an ability to slap comments and a rating of one to five stars on documents. The twist is that the box that will facilitate this will also include an option for erasing the information.
"We've worked really hard to find a middle ground," Burk said. "We feel like we've got a good kind of solution that gives customers the benefit of metadata without exposing them to too much risk."
Gartner's Silver suggests that Microsoft could go further, by upgrading its Exchange e-mail server software to let companies automatically strip metadata from documents before they are e-mailed or posted online a trick offered by third-party metadata protectors.
Burk said Microsoft might consider it.
Of course, wider use of metadata-scanning tools will reduce the juicy finds that have benefited journalists and others.
Richard M. Smith, a computer privacy expert at Boston Software Forensics, mined metadata to determine who in the British prime minister's office worked on a 2003 dossier on Iraq. He also has picked up clues about the provenance of video clips on al-Qaida-related Web sites.
Even in a world more attuned to the perils of metadata, however, Smith doesn't think the material will dry up.
"There are simply too many people who work in governments around the world and there is no way to educate them all about metadata," he wrote in an e-mail. "I expect to see a steady stream of slip-ups in the future."

Ex-Merck Exec to Testify at Vioxx Trial

NEW YORK - Former Merck & Co. Chairman and Chief Executive Raymond Gilmartin has agreed to a plaintiff lawyers‘ request that he testify at a Vioxx trial slated to begin in Atlantic City in late February, both sides said.
Gilmartin has given testimony in videotaped depositions, but Lanier said that it is always better to have a live witness than a tape. He called the depositions "canned speeches" and said it is harder for witnesses to give soliloquies or avoid questions in the presence of a judge.
A Merck official said the company believes Gilmartin will be a strong witness, noting his performance in front of a Senate committee hearing in late 2004, when he answered questions about the withdrawal of the pain reliever in September 2004. The drug was taken off the market after a study showed it doubled patients‘ risk of heart attacks and strokes when taken for longer than 18 months.
"Mr. Gilmartin has a good story to tell about Merck‘s responsible actions concerning Vioxx and the voluntary withdrawal of the product," said Kent Jarrell, a Merck spokesman.
Merck also announced it had sent aside an additional $295 million for litigation expenses bringing the total amount it has earmarked for legal costs to $970 million. However, the company said it spent $285 million on defense costs last year, leaving it with $685 million, which it believes will last through 2007.
The upcoming trial in New Jersey is slated to combine two plaintiffs, each a man who suffered a nonfatal heart attack after taking Vioxx for more than 18 months. Lanier represents one of the men.

Judge: Doctor Can't Testify in Vioxx Case

NEW ORLEANS — A physician called to testify that Vioxx caused a blood clot which killed a Florida man cannot do so, the judge in the first federal Vioxx lawsuit ruled Friday.
That appears to cut to the heart of Evelyn Irvin Plunkett's claim against manufacturer Merck & Co. in the death of her husband, Richard "Dickie" Irvin, who took the drug for a month.
Her attorneys had called Dr. Michael Graham as an expert to say that Vioxx could cause blood clots in that short a time and had caused the one that brought on Irvin's fatal heart attack in 2001.
But U.S. District Judge Eldon E. Fallon ruled late Friday that Graham wasn't qualified to do that. He wrote that the doctor has no pharmacology training, isn't qualified to explain how Vioxx might cause a blood clot, has never prescribed it, and got most of his knowledge about the drug in an eight-hour review of medical articles, expert reports and trial transcripts.
"Dr. Graham is qualified to testify as to the existence of a thrombus and its role in Mr. Irvin's death. He is just not qualified to testify that Vioxx can cause a thrombus and did cause Mr. Irvin's thrombus," Fallon wrote.
Attorneys in the case have been told not to talk to reporters about it.
Merck spokesman Kent Jarrell said Fallon's order speaks for itself.
Merck's motion seeking Graham's disqualification said Plunkett apparently has dropped the three medical experts who testified in the first trial, held in Houston because of Hurricane Katrina.
Plunkett claims that Vioxx caused the blood clot by tamping down an enzyme that would otherwise thin the blood and prevent clots.
Although Fallon has told attorneys he wants to try to work out a settlement after the first four federal Vioxx trials, he also has set up a plan that could bring more cases for early trial.
The next two are scheduled in May and June.
By Feb. 15, attorneys for Merck and for the plaintiffs' steering committee are to draw up two lists of five cases for early trial, Fallon wrote in an entry filed on Friday. One list will be of Louisiana cases, the other of cases filed in federal courts outside Louisiana. Each side can eliminate two cases from the other side's list.
Fallon will choose from the remaining cases.

Medical Journal's article is a wild card in federal Vioxx retrial

NEW ORLEANS -- A top medical journal's criticism of a study cited in Vioxx lawsuits is expected to play a part in the retrial of the first federal Vioxx lawsuit _ but it's not clear if it will be a legal hand grenade or a nuclear warhead. The disclosure that some negative data was omitted from the study could make manufacturer Merck & Co. look as if it's hiding something, legal experts say. That is what attorneys for Evelyn Irvin Plunkett, whose husband died after taking the drug for a month, say they can prove in the trial starting Monday.


The first federal trial _ held in Houston in the wake of Hurricane Katrina _ ended with a deadlock. Two jurors said the split was 8-1 in favor of Merck's contention that taking Vioxx had nothing to do with the death of Richard "Dickie" Irvin. The day those deliberations began, the New England Journal of Medicine published criticism _ one step short of retracting the study _ accusing its authors of withholding and deleting relevant data. The information alone is "damaging although not shattering," but its implications could be devastating, Northwestern University law professor Ronald Allen said. Vioxx was a $2.5 billion-a-year seller when Merck pulled it from the market in September 2004 because a study found that taking it for 18 months doubled the risk of heart attacks. An earlier study known as VIGOR, published by the journal in 2000, is being cited in allegations that Merck held back information about the risk of heart attacks. The journal's critique noted that its editors learned that the authors had deleted information about three additional heart attacks not reported in the original study, but reported to the Food and Drug Administration. That would have changed a section about side effects, showing that Vioxx patients had five times as many attacks as those on the pain reliever naproxen, rather than four times as many _ and that patients not considered high-risk were having heart attacks, it said. "Taken together, these inaccuracies and deletions call into question the integrity of the data," the journal said. The editors found a blank table labeled "CV events" _ heart attacks and other cardiovascular side effects _ on a first draft of the VIGOR study found on a computer diskette after the recall. It was deleted from later versions on the same disk. Dr. Gregory Curfman, the journal's executive editor, confirmed through a spokeswoman that software indicated that data was deleted from the table two days before the manuscript was submitted, and identified the deleting editor as "Merck." The company said that all that was taken from the table was data that was put into the body of the article, and it never included information about the three later heart attacks. But, he wrote, the Journal's editors didn't realize the significance of the deletion until Nov. 21, when Curfman was questioned by Plunkett's attorneys. They produced a memo showing that at least two authors knew about the heart attacks at least two weeks before submitting the first of two revisions and 4{ months before publication. Until then, the editors wrote, they had believed the authors didn't know about those heart attacks in time for publication. Merck says the heart attacks occurred after a reporting deadline. The article didn't mention any such deadline, and the journal would have expected updates anyway, Curfman wrote the lead author in an e-mail released by the court. Merck's attorneys, led by Phil Beck, plan to show that VIGOR was "done the right way," meeting all scientific standards, Beck said before the judge imposed a gag order on both sides. The main focus of Merck's arguments, as in the first trial, is that Vioxx had nothing to do with Irvin's death. He already had clogged arteries and a blood clot, they have said in court and in interviews. Plunkett's attorneys, led by Jere Beasley, claim Vioxx caused the clot by affecting an enzyme that thins the blood. Allen said the editorial's precise impact will depend on just how significant the data is considered _ and that, in turn, will help decide what use judges allow. It might simply be folded into the evidence, he said. But it might also be used as a possible indication that Merck was hiding something.
"If juries get a whiff of that, it's going to be devastating," he said. Howard Erichson, a Seton Hall Law School professor, called the editorial a stunning development. But, though he called it "terribly damaging" in a December interview, he now says it won't dramatically alter the case.


"It was about a few data points, not a fundamental rethinking of causation," Erichson said. "It's one piece of evidence that's going to undermine one of the Merck witnesses. "It's there, it's interesting, but it doesn't fundamentally change the course of the litigation." The case being retried starting Monday was the third among more than 9,600 state and federal lawsuits against Merck. Two state jury trials ended with a split decision: a $234 million verdict against Merck in Texas, and one clearing it of blame in New Jersey. Another trial is under way in Texas, and 10 more are scheduled in state and federal courts over the next six months. Merck has set aside $970 million for legal costs, and said it spent $285 million of that last year. The remaining $685 million should last through 2007, it said Tuesday.

Journal Article a Wild Card in Vioxx Trial

A top medical journal's criticism of a study cited in Vioxx lawsuits is expected to play a part in the retrial of the first federal Vioxx lawsuit - but it's not clear if it will be a legal hand grenade or a nuclear warhead. The disclosure that some negative data was omitted from the study could make manufacturer Merck & Co. look as if it's hiding something, legal experts say. That is what attorneys for Evelyn Irvin Plunkett, whose husband died after taking the drug for a month, say they can prove in the trial starting Monday. The first federal trial - held in Houston in the wake of Hurricane Katrina - ended with a deadlock. Two jurors said the split was 8-1 in favor of Merck's contention that taking Vioxx had nothing to do with the death of Richard "Dickie" Irvin. The day those deliberations began, the New England Journal of Medicine published criticism - one step short of retracting the study - accusing its authors of withholding and deleting relevant data. The information alone is "damaging although not shattering," but its implications could be devastating, Northwestern University law professor Ronald Allen said. Vioxx was a $2.5 billion-a-year seller when Merck pulled it from the market in September 2004 because a study found that taking it for 18 months doubled the risk of heart attacks. An earlier study known as VIGOR, published by the journal in 2000, is being cited in allegations that Merck held back information about the risk of heart attacks. The journal's critique noted that its editors learned that the authors had deleted information about three additional heart attacks not reported in the original study, but reported to the Food and Drug Administration. That would have changed a section about side effects, showing that Vioxx patients had five times as many attacks as those on the pain reliever naproxen, rather than four times as many - and that patients not considered high-risk were having heart attacks, it said. "Taken together, these inaccuracies and deletions call into question the integrity of the data," the journal said. The editors found a blank table labeled "CV events" - heart attacks and other cardiovascular side effects - on a first draft of the VIGOR study found on a computer diskette after the recall. It was deleted from later versions on the same disk. Dr. Gregory Curfman, the journal's executive editor, confirmed through a spokeswoman that software indicated that data was deleted from the table two days before the manuscript was submitted, and identified the deleting editor as "Merck." But, he wrote, the Journal's editors didn't realize the significance of the deletion until Nov. 21, when Curfman was questioned by Plunkett's attorneys. They produced a memo showing that at least two authors knew about the heart attacks at least two weeks before submitting the first of two revisions and 4 1/2 months before publication. Until then, the editors wrote, they had believed the authors didn't know about those heart attacks in time for publication. Merck says the heart attacks occurred after a reporting deadline. The article didn't mention any such deadline, and the journal would have expected updates anyway, Curfman wrote the lead author in an e-mail released by the court. Merck's attorneys, led by Phil Beck, plan to show that VIGOR was "done the right way," meeting all scientific standards, Beck said before the judge imposed a gag order on both sides. The main focus of Merck's arguments, as in the first trial, is that Vioxx had nothing to do with Irvin's death. He already had clogged arteries and a blood clot, they have said in court and in interviews. Plunkett's attorneys, led by Jere Beasley, claim Vioxx caused the clot by affecting an enzyme that thins the blood. Allen said the editorial's precise impact will depend on just how significant the data is considered - and that, in turn, will help decide what use judges allow. It might simply be folded into the evidence, he said. But it might also be used as a possible indication that Merck was hiding something. "If juries get a whiff of that, it's going to be devastating," he said. Howard Erichson, a Seton Hall Law School professor, called the editorial a stunning development. But, though he called it "terribly damaging" in a December interview, he now says it won't dramatically alter the case. "It was about a few data points, not a fundamental rethinking of causation," Erichson said. "It's one piece of evidence that's going to undermine one of the Merck witnesses. "It's there, it's interesting, but it doesn't fundamentally change the course of the litigation." The case being retried starting Monday was the third among more than 9,600 state and federal lawsuits against Merck. Two state jury trials ended with a split decision: a $234 million verdict against Merck in Texas, and one clearing it of blame in New Jersey. Another trial is under way in Texas, and 10 more are scheduled in state and federal courts over the next six months. Merck has set aside $970 million for legal costs, and said it spent $285 million of that last year. The remaining $685 million should last through 2007, it said Tuesday.