TRENTON, N.J. — A doubled risk of heart attack, stroke and death persisted at least a year after people stopped taking withdrawn painkiller Vioxx, according to an analysis of long-term data from the study that led drugmaker Merck & Co. to stop selling the drug.

The analysis, published online today by the British medical journal The Lancet, also appears to show the higher risk started soon after patients began taking Vioxx, though the study notes a small sample size precludes a definitive finding on this issue.

Doctors critical of Merck and its reporting of Vioxx studies have long argued increased cardiovascular risks from the former blockbuster arthritis drug started after just a few months' use and persisted after use ended.

Merck continues to insist cardiac risks didn't increase until people took Vioxx for about 18 months — a cornerstone of its strategy to fight tens of thousands of lawsuits by people claiming harm from Vioxx.

The new findings should be interpreted cautiously because of the small number of patients who suffered heart attacks, strokes and related problems after participating in the three-year study, said Doug Watson, a cardiovascular epidemiologist and senior director at Merck Research Labs.

He said the study had some other limitations and noted the authors stated that "small numbers prohibit detailed conclusions about when the increased risk begins and ends."

"But our data are compatible with an early increase in risk that seems to persist for about 1 year after 3 years of treatment," the authors added.

Known by the acronym APPROVe, the study was intended to prove that Vioxx, heavily promoted as relieving pain with lower gastrointestinal risks than older anti-inflammatory drugs, could prevent recurrence of colon cancer. Merck stopped the study two months early and pulled the drug from the market on Sept. 30, 2004, when data showed roughly double the risk of cardiovascular complications and death in the group getting Vioxx, over those getting placebo.

Merck funded the new analysis, provided the data and commented on an early draft of the report but said it had no other involvement. The analysis was conducted by six scientists who worked on APPROVe, plus two statistics experts at the University of Wisconsin who were not involved then, the company said.