Models used by doctors to predict a patient’s chances of having a heart attack — including a new approach issued just 15 months ago — badly overestimate the number of “cardiovascular events” when compared with the total that actually occur, a team of researchers from Johns Hopkins University reported Monday.

Four of the “risk calculators” were off by 37 to 154 percent in men and by 8 to 67 percent in women, according to the researchers from the Johns Hopkins Ciccarone Center for the Prevention of Heart Disease, whose paper was published online in the journal Annals of Internal Medicine.

Their review included guidelines issued in November 2013 by the American Heart Association and the American College of Cardiology that recommended treatment with statins for anyone older than 40 who, according to that risk calculator, had a 7.5 percent chance of suffering a heart attack in the next 10 years.

Michael Blaha, director of clinical research at Ciccarone and one of the authors of the study, said the research shows that physicians should not rely too heavily on predictive models when making treatment decisions. Instead, he cautioned, the models should be used as a starting point for doctors, who should gather more information about each patient before responding with medication, procedures such as angioplasty or recommendations for lifestyle changes.

“What our study really does show is how limited these risk estimates can be,” he said in an interview.

Blaha acknowledged that the authors of the new guidelines had anticipated that their calculations would need validation and possibly changes. But the paper nevertheless brought a terse response from two of the officials who helped write the new American Heart Association guidelines. “The overall findings are not news at all,” Donald Lloyd-Jones, chairman of the department of preventive medicine at the Northwestern Feinberg School of Medicine and a member of the committee that developed the recommendations wrote on the Heart Association’s Web site. “We actually did much of this exercise ourselves in the guideline document.”

Another committee member, David Goff, said in an interview that the 2013 recommendations “represent such a big advance from the last ones” that physicians should use them even as questions are raised.

Goff, dean of the Colorado School of Public Health, also took issue with the study’s methodology, noting that it took conclusions based on data for whites and African Americans and applied it to Asian Americans and Hispanics as well as the other groups. And, he said, the guidelines were intended for patients who were not being treated; more than 80 percent of the population examined by the Johns Hopkins researchers had received treatment of some kind.

The annual numbers of heart attacks and strokes in the United States have declined as fewer people smoke and treatments for high blood pressure and cardiovascular diseases have been developed. Still heart disease remains the leading cause of death for Americans, with 720,000 people having heart attacks every year, the government says.

In the paper published Monday, the researchers compared the predictions of five risk calculators to the actual number of cardiovascular events suffered by 4,227 people ages 50 to 74 in the Multiethnic Study of Atherosclerosis. Three older models are probably inaccurate now because they relied on data from the landmark Framingham Heart Study, whose population was overwhelmingly white.

A fourth predictor, the Reynolds Risk Score, performed best, but was formulated using a more narrow, lower-risk population.

The new guidelines were off by 86 percent for men and 67 percent for women.

“Things are changing, the disease is changing, rates are changing,” Blaha said. “It’s very difficult, in my opinion, to use any historical data.”