Correspondence

Review of Methods in “Breast Augmentation: A Risk Factor for Breast Cancer?”

N Engl J Med 1994; 330:293January 27, 1994

Article

To the Editor:

We are writing to follow up on the article by Berkel et al. concerning breast implants and the risk of cancer (June 18, 1992, issue).1 The study was a population-based, noncurrent cohort-linkage study, which was performed under the auspices of the Alberta Cancer Board. Essentially, the study was designed to identify women who had undergone cosmetic breast augmentation between 1973 and 1986 and to link this cohort of women with the Alberta Cancer Registry to determine which women with implants subsequently had breast cancer. The actual breast-cancer rates were compared with the expected rates, which were calculated by applying age-specific incidence rates for breast cancer to the entire cohort's person-years at risk. The standardized incidence ratio reported was 47.6 percent, which was interpreted to mean that the rate of breast cancer in the implant group was significantly lower than expected.

After the publication of this report, some potential problems were identified, which led to a detailed review of the study methods and results. Although the final estimate of the revised standardized incidence ratio is not yet ready to be published, we wish to make Journal readers aware of some fundamental difficulties with the study that may well lead to a substantially different interpretation of the results.

An initial step in this study was the calculation of the number of person-years at risk for the cohort. The number of person-years at risk that was used for the study was 124,494. The investigators made no adjustment for potential out-migration but referred to a personal communication from an investigator from another study in the same area indicating that out-migration would be expected to be only 10 percent. However, that figure referred to a 5-year period, and the mean period during which out-migration could have occurred in this study was over 10 years. Estimates of out-migration for this particular cohort, based on readily available population statistics, would therefore result in substantially higher estimates. Another error arose from the consistent overestimation of person-years at risk by approximately one half-year per woman in the cohort. These adjustments, as well as a closer examination of some of the eligibility requirements, result in a reduction in the number of person-years at risk of approximately 25 percent. The impact on the expected number of cancers is greater, since the years lost were, in general, years in which the woman's risk of breast cancer was higher.

The linkage methods used in this study were complicated by incomplete data in some of the files. For example, maiden names were available in only 80 percent of the Alberta Cancer Registry files, and in none of the Alberta Health Care implantation records. In addition, the Alberta Health Care number is not an identifier that a woman keeps throughout her lifetime; approximately half the women in the implant cohort changed their Alberta Health Care numbers during the study period. As a result, some underlinkage would be expected. Using a reconstructed file that records previous Alberta Health Care numbers, we are identifying some files that were not reported as linkages in the initial study. We are currently in the process of reviewing the probable linkages to confirm their validity.

The problems identified above tended to introduce a bias resulting in an underestimate of the standardized incidence ratio. We expect that the true ratio is higher than 0.476, which is the ratio reported in the original paper. We therefore caution readers against placing too much reliance on the results. A further review is under way, which will address the problems noted and some others that have not been addressed in this letter. We intend to provide a revised estimate of the standardized incidence ratio, with a step-by-step justification for any changes, within the next few weeks.

Heather Bryant, M.D., Ph.D.
Penny M.A. Brasher, Ph.D.
Johan H. van de Sande, Ph.D.
Jean-Michel Turc, M.D.
Alberta Cancer Board, Edmonton, AB T5K 2L9, Canada

1 References

1. 1

   Berkel H, Birdsell DC, Jenkins H. Breast augmentation: a risk factor for breast cancer? N Engl J Med 1992;326:1649-1653
   Full Text | Web of Science | Medline

Citing Articles (3)

Citing Articles

1. 1

   LESLIE L. ROOS, RANDY WALLD, ANDRÉ WAJDA, RUTH BOND, KATHLEEN HARTFORD. (1996) Record Linkage Strategies, Outpatient Procedures, and Administrative Data. Medical Care 34:6, 570-582
   CrossRef

2. 2

   Bryant, Heather, Brasher, Penny, . (1995) Breast Implants and Breast Cancer — Reanalysis of a Linkage Study. New England Journal of Medicine 332:23, 1535-1539
   Full Text

3. 3

   Angell, Marcia, Kassirer, Jerome P., . (1994) Setting the Record Straight in the Breast-Cancer Trials. New England Journal of Medicine 330:20, 1448-1450
   Full Text