More data flaws mar breast cancer trials.

New data irregularities have emerged in a continuing federal probe of a breast cancer reseach group. National Cancer Institute (NCI) audies of two Louisiana medical centers failed to find some required information on many patients enrolled in several key breast cancer studies.

These new reports adde to the growing concern about the quality of datra collected by the National Surgical Adjucant Breast and Bowel Project (NSABP), a group that includes 484 medical centers in the United States and Canada.

NSABP first came under fire after reports that a Montreal physician had falsified data underlying a study reported in 1985 that compared lumpectomy to mastectomy for the treatment of breast cancer (see p.282). Now, Tulane University and Louisiana State University (SLU) medical centers, both in New Orleans, have been drawn into the fray.

At Tulane, NCI auditors reviewed 75 charts for patients enrolled in a number of NSABP clinical trials between 1976 and 1994. In about 70 percent of the cases, they found insufficient documentation to determine the patient's eligibility for inclusion in the trials. Auditors could not find the required results of certain lab tests in some cases. In others, they found no documentation that patients had consented to participate in a trial.

LSU fared little better. The NCI squad reviewed 80 charts for patients enrolled in NSABP studies between 1971 and 1994 and failed to find sufficient documentation to confirm eligibility in about half the cases.

The missing data make it impossible to ensure that the patients enrolled by the Louisiana centers met the standards required by the various NSABP trials, says NCI's Michael A. Friedman. Both Tulane and LSU had trouble producing the original documentation to support data they submitted to NSABP's central office, located at the Universityh of Pittsburgh.

The auditors unearthed no sign of fraud at the Louisiana medical centers. However, Tulane and LSU certainly demonstrated sloppy data management, NCI charged. "The standards of record keeping and reporting at these institutions were unacceptable," said NCI.

At a meeting, this week of an NCI scientific advisory board, institution officials outlined their actions to improve the quality of the data collected by NSABP, including the monumental task of auditing the primary records of patients enrolled in NSAPB studies. NCI must go back and audit those records because lax procedures by NSABP's central office allowed such data problems to continue for years.

So far, NCI has audited roughly half of the patient charts that contributed to the study, known as B-06, comparing lumpectomy to mastectomy. Many of the patient records for that trial have been warehoused or stored on microfilm, making it difficult to check the originals against the data turned in to NSABP. So far, NCI has found no evidence of systematic manipulation of data or fraud in the B-06 trial other than a center in Montreal.

In a conference room on the National Institutes of Health's campus in Bethesda, Md., NCI's Bruce A. Chabner emphasized to the advisory board the importance of verifying the B-06 trial. Preliminary analysis of that trial suggests that many clinical investigators did not obtain proper informed consent from their patients. If those patients qualified for B-06 in every other way, the consent problem probably wouldn't undermine the study's scientific findings, Friedman says.

At the meeting, NCI officials outlined a number of measures to help maintain the accuracy of NSABP data in the future. For example, NCI wants audit schedules provided on a regular basis and immediate notification of any suspect data. In the end, the NCI advisory board approved a 1-year increase in funding so that NCI could step up its auditing of NSABP centers.

"The record simply must be set 0 straight," Friedman says.