There were 35 drug-related deaths in the 213 studies examined, but such deaths were much more frequent in the earlier experiments than in the more recent ones. There were 24 drug-linked deaths in studies from 1991 to 1994, 10 from 1995 to 1998 and just one from 1999 to 2002.
Deaths from other causes, including cancer, decreased, too - from 39 to 17 in the more recent studies.
Dr. Mace Rothenberg, a Vanderbilt University cancer research specialist who was not involved in the study, called the findings "clearly good news."
"It indicates that patients who agree to go on to these fairly risky studies are not placing themselves at an unacceptably high risk for severe toxicity or death," he said.
The studies involved a total of 6,474 patients and the overall death rate fell from 3 percent to 1 percent.
At the same time, there was also a decline in the percentage of tumors that shrank in response to the drugs being tested. But the researchers and other cancer experts said that does not mean the drugs were increasingly ineffective.
Some of the newest cancer drugs employ a more targeted approach than older medicines, attacking tumor cells while causing less damage to healthy tissue. These drugs often work by stopping but not necessarily reversing tumor growth, said Dr. Thomas Roberts Jr. of Massachusetts General Hospital, the study's lead author.
The researchers said the decrease in deaths may also be due to growing attention to patient safety regulations and increased oversight of human experiments during the past decade. For example, many institutions have developed educational programs for researchers and hired additional staff to oversee research, they said.
The researchers were encouraged to find not only fewer drug-linked deaths, but also a decline in cancer-related deaths.
"They are significant because this is something that really matters to patients I think when they are considering whether to proceed with standard treatments or experimental treatments," Roberts said.
His report appears in Wednesday's Journal of the American Medical Association.
The studies involved were Phase I clinical trials - the first time experimental drugs are tested in humans. These are small studies designed mainly to test the safety of an experimental treatment and to determine the maximum tolerable dose.
Patients who participate in Phase I cancer studies generally have advanced disease that has not responded to conventional treatments. While many hope participating will improve their health, that is not the primary goal in such studies.
The patients involved had various types of cancer, including colon, lung, kidney, breast and prostate tumors.
Rothenberg said the lower toxicity of newer, more targeted drugs may explain both the reduced deaths and reduced tumor shrinkage. While tumors do not always shrink in response to such drugs, they may "shut down" for long periods, allowing patients to live fairly normal lives, he said.
By Lindsey Tanner