Meta-analysis to Examine Effect of Amifostine on Survival in Radiotherapy-Treated Patients

August 1, 2004

The 14 reports in this special supplement discuss theuse of the cytoprotectant amifostine in patients withcancer of the head and neck, esophagus, lung, andcervix, as well as those with lymphoma and acutemyelogenous leukemia. Discussions focus on thepotential of this agent to both reduce radiation sideeffects such as xerostomia and permit doseescalation of chemotherapy and/or radiotherapy.Improvements in treatment outcome and quality oflife as a result of cytoprotection are examined.

VILLEJUIF, FRANCE-After more than 25 randomizedtrials, the effect of amifostine(Ethyol) on the survival of patientstreated with radiotherapy is stillunknown. This is because most ofthe trials were neither designednor powered to detect small survivaldifferences and had followupperiods that were either short(less than 2 years) or very short (afew months), according to JeanBourhis, MD, PhD, of the departmentsof biostatistics and radiotherapy,Institut Gustave-Roussy,Villejuif, France. To assess whetherthis drug is associated with asurvival benefit in radiotherapytreatedpatients, he and his col-leagues are conducting a metaanalysisof amifostine trials.Nonmetastatic Ca Only
Selection criteria include properlyrandomized trials performedbetween 1990 and June 2002, inwhich patients were randomizedto receive radiotherapy or chemoradiotherapyplus amifostine, orto radiotherapy or chemoradiotherapywithout amifostine. Patientshad nonmetastatic head andneck, lung, or pelvic carcinomas.Updated data are being collectedfor all randomized patients frompublished and unpublished trials,and extensive checking and validationwill be done to ensure theintegrity of randomization and folonilow-up. Dr. Bourhis noted "pseudorandomization"might happenin some clinical trials.Tumor Control an Endpoint
The primary endpoint of themeta-analysis is the effect of amifostineon survival. The secondaryendpoint is effect on tumor control.The meta-analysis will includemore than 2,000 patients from 8trials of non-small-cell lung cancer,13 of head and neck cancer,and 7 of pelvic cancer. The investigatorsexpect to report results inabout 1 year.Dr. Bourhis's decision to includeunpublished trials evoked somequestions during the discussionperiod. "We are trying to avoidpublication bias in favor of positivetrials," he said. "We have to becareful not to avoid trials that havenot been published becausethe results were not exactly asexpected."