Studies of Counselings' Impact on Survival Challenged

June 1, 1997

SEATTLE--The two studies most often cited as proof that psychological support can improve survival in cancer patients came under challenge at the annual meeting of the American Association for the Advancement of Science.

SEATTLE--The two studies most often cited as proof that psychologicalsupport can improve survival in cancer patients came under challenge atthe annual meeting of the American Association for the Advancement of Science.

Wallace L. Sampson, MD, associate chief of hematology and oncology,Santa Clara Valley Medical Center, San Jose, Calif, and clinical professorof medicine, Stanford, argued that flaws in their methodology account forthe studies' positive outcomes.

The two studies were conducted by psychiatrists. One at Stanford University,led by David Spiegel, MD, involved a group of women with disseminated breastcancer. The second, carried out by Fawzy Fawzy, MD, and his colleaguesat UCLA, looked at a group of patients with stage I melanoma.

Both trials randomized patients to groups that provided psychologicalsupport in addition to their medical treatment, or to control groups thatreceived only medical therapy.

One problem in the Stanford study, in addition to having small numbers,is that the researchers randomized significantly more patients to the treatmentgroup than to the control group, in order to have enough patients for therapygroups, Dr. Sampson said. Equal numbers or an excess of controls is moreusual.

He noted that the paper did not specify how randomization was accomplishedor when assignments took place after acceptance into the study, althoughtherapy and control groups seemed to be similar in retrospect.

"If one violates the initial randomization, one cannot rely onthe results, especially if the randomization is broken for a specific reason,"he said.

Dr. Sampson noted a more disquieting problem when he compared the survivalfindings of the Stanford study with survival statistics in national andregional data bases, particularly the National Cancer Institute's SEERdata, and in studies of the natural history of breast cancer.

"The therapy group in the Spiegel/Stanford study lived as longas expected from national, regional, and other data," he said. "Thecontrol group was aberrant, in that all patients were dead by four years,whereas every other study shows that in patients at this disease stage(after metastasis has been diagnosed), about 10% are still alive at 10years."

He cited another methodological problem: The study was originally designedto look at quality of life issues--survival was not an endpoint--and theoutcome data were obtained retrospectively. "In my mind," Dr.Sampson said, "that study is simply not interpretable and should neverhave received the amount of publicity it got." Other studies haveshown no effect of support groups on survival, he added.

The UCLA melanoma study by Dr. Fawzy and his team was prospective anddid find a statistically significant difference in the survival of thetherapy group compared to the control group. However, here again Dr. Sampsonfaulted the study because its survival data for the control group deviatedwidely from other data available on such patients. "When you comparetheir data to national data, you find that their patients who receivedthe intervention did no better and no worse than these patients in general,"he said.

SEER data show that at five years, about 90% to 95% of stage I melanomapatients are still alive and well, "and that is what their therapygroup did," he said. While the therapy group followed the norm, thecontrol group did not. Only 74% of controls were alive at six years, andthis created the same validity problem for the UCLA study as this dichotomydid in the Stanford trial.

"In neither of the two studies does one find consistency with allother data reported," Dr. Sampson said. "I simply suspect theseare aberrant results."