Correspondence

Surgery for Early Breast Cancer

N Engl J Med 1996; 334:987-988April 11, 1996

Article

To the Editor:

Dr. Bailar (Nov. 30 issue)1 reassures physicians and patients that the lumpectomy study findings of the National Surgical Adjuvant Breast and Bowel Project (NSABP) are credible. However, he asserts that until a much broader analysis of the entire cohort from St. Luc Hospital alone is reported, all St. Luc data should be excluded from future interpretations of Protocol B-06. In fact, we did conduct extensive analyses to verify the integrity of the St. Luc data.2 Presenting findings from a single institution compromises the integrity and conduct of multicenter trials. It also increases the chances of finding false positive results, leads to the publication of inconclusive and biased results, and fails to demonstrate variability of findings among institutions. Only when the findings from a single center are shown within the context of those from other institutions is there validity to such a presentation. Consequently, we have compared outcomes from St. Luc with those from the eight other institutions reporting more than 20 deaths (Figure 1Figure 1Annual Event and Death Rates among Patients in Cohort B (of Fisher et al.2) at St. Luc Hospital, Eight Other Institutions, and All Institutions.). The annual death and event rates, along with 99 percent confidence intervals, indicate that the rates for St. Luc are similar to those for other institutions. Life tables showing survival or disease-free survival among St. Luc patients (Figure 2Figure 2Life Table Showing Overall Disease-free Survival and Survival among St. Luc Patients in Cohort B Who Were Treated by Mastectomy, Lumpectomy, or Lumpectomy and Breast Irradiation.) indicate no significant differences among the three treatment groups. Meticulous examination of original records for all St. Luc patients indicated that a single criterion required for enrollment was altered for only six patients.2 These alterations could not have biased the findings because they occurred before randomization. Eliminating data on hundreds of women who were properly randomized and followed and whose data were verified is both scientifically and ethically inappropriate.

Another of Bailar's criticisms relates to consent forms. To judge such forms obtained between 1976 and 1984 by today's standards is inappropriate. During that time, each participating institution determined the content and governance of its consent forms. The NSABP was among the first cooperative groups to collect, inspect, and archive consent forms centrally. Bailar's concern regarding “the failure of the physicians enrolling patients . . . to be meticulous about informed consent” is based on information in Table 6 of the National Cancer Institute (NCI) audit.3 These data are unreliable, because the auditors failed to seek this information with the same intensity used in obtaining “other missing or ambiguous items.”3

Finally, Bailar is concerned about patients whose consent was putatively obtained after surgery. Initially, in Protocol B-06, the biopsy established the diagnosis of invasive cancer. As the study progressed, surgical technique was altered so that biopsy was carried out as a lumpectomy.4 Consequently, some patients in the lumpectomy groups had no need for additional breast surgery subsequent to randomization. It might thus be interpreted that they consented to be in the trial after surgery. Obtaining consent after surgery in this circumstance represented appropriate medical and ethical conduct.

Bernard Fisher, M.D.
Stewart Anderson, Ph.D.
Carol K. Redmond, Sc.D.
University of Pittsburgh, Pittsburgh, PA 15261

4 References

1. 1

   Bailar JC III. Surgery for early breast cancer -- can less be more? N Engl J Med 1995;333:1496-1498
   Full Text | Web of Science | Medline

2. 2

   Fisher B, Anderson S, Redmond CK, Wolmark N, Wickerham DL, Cronin WM. Reanalysis and results after 12 years of follow-up in a randomized clinical trial comparing total mastectomy with lumpectomy with or without irradiation in the treatment of breast cancer. N Engl J Med 1995;333:1456-1461
   Full Text | Web of Science | Medline

3. 3

   Christian MC, McCabe MS, Korn EL, Abrams JS, Kaplan RS, Friedman MA. The National Cancer Institute audit of the National Surgical Adjuvant Breast and Bowel Project Protocol B-06. N Engl J Med 1995;333:1469-1474
   Full Text | Web of Science | Medline

4. 4

   Fisher B. Reappraisal of breast biopsy prompted by the use of lumpectomy: surgical strategy. JAMA 1985;253:3585-3588
   CrossRef | Web of Science | Medline

Author/Editor Response

Dr. Bailar replies:

To the Editor: The reply from Fisher and colleagues is incomplete. First, with respect to the St. Luc data, six proven alterations raise questions about all the other records. If it was important to have clear and specific criteria in the study protocol, it was more than a trivial offense for an investigator to falsify data to meet those criteria. Fisher et al. also offer no explanation for using 99 percent confidence intervals rather than the common 95 percent confidence intervals; the effect of this variation is to widen the intervals by about 31 percent and hence reduce the likelihood of finding a statistically significant difference. Their Figure 1, showing event rates, suggests that the risk was lower at St. Luc than anywhere else except at Institution 1, and they do not address directly the specific point that I raised in my editorial: that St. Luc patients, for unexplained reasons, seem to have had a more favorable prognosis than all the other patients. The statistical arguments against the study of subgroups of patients simply do not hold when the subgroups are selected because of specific questions about the integrity of the data.

Second, although patient consent is a long-term process rather than a single event, a critical step in the process is complete, accurate, and timely documentation. The NCI auditors did search for evidence of informed consent and noted the nature of the consent (written or oral) and the date relative to surgery. Although they did not attempt to clarify other consent issues, what they did uncover was disturbing. The claim that data from the NCI audit are “unreliable” is simply not acceptable in the absence of evidence that informed consent was in fact appropriately obtained, recorded, and filed but missed by the NCI.

Third, there is a critical difference between consent to undergo some medical intervention and consent to enter a research program to study that intervention. Here, the initial diagnostic procedure was changed from ordinary biopsy to segmental resection (lumpectomy) at about the time the prerandomization scheme was adopted. To call a segmental resection a biopsy is about as correct as to call a mastectomy a biopsy. Unless segmental resection was the standard diagnostic procedure for the patients at these hospitals who were not enrolled in the study, it was part of the experimental protocol, and the prerandomized patients had no effective means to decline one of the treatments (segmental resection) under evaluation.

John C. Bailar, III, M.D., Ph.D.
University of Chicago, Chicago, IL 60637-1470