In the May, 1997, issue of Oncology News International , the updated eight-year results of NSABP protocol B-17 were reported as presented by Dr. Bernard Fisher in Paris. B-17 randomized patients with ductal carcinoma in situ (DCIS) into two groups: One group received excision only, the other excision plus postoperative radiation therapy.
The NSABP concluded that all subsets regardless of grade, size, or margin status were benefited by irradiation, and they recommended that all breast conservation patients with DCIS receive postoperative irradiation. However, although this was a prospective randomized trial, the pathologic features cited by the author to reach these conclusions were analyzed retrospectively.
Many of the significant prognostic features of DCIS that we have documented in our own studies (size, margin width, nuclear grade)[2,3] were not tabulated prospectively by the NSABP and were not available at the time of original analysis. In fact, at the time of retrospective analysis, 27% of the entire B-17 study population was unavailable for pathologic evaluation.
In the original study, all tissue was not submitted and sequentially processed. There were no guidelines regarding size measurement, specimen radiography, or mammographic-pathologic correlation. Margins were defined as free (clear) when tumor was not transected. In other words, in some cases, only a few fat cells separated DCIS from the inked margin.
In 40% of cases, no tumor size was provided by the initial pathologist, and, therefore, size could not be determined prospectively nor accurately analyzed retrospectively.
Possible differences in outcome (local recurrence) were also obscured by the pathologic definitions, which created minimally divergent comparison groups such as DCIS with one third or fewer of ducts exhibiting comedo necrosis versus one third or more, and which pooled nuclear grade (NG) I and II lesions with or without necrosis versus NG III DCIS.
Our database, including more than 440 conservatively treated DCIS patients[3,5-7], has led us to a diametrically opposite conclusion. Our DCIS cases were fully evaluated prospectively for grade, size, and margin status, and we have defined those features to accentuate, not diminish, the differences in the subgroups.
We found that irradiation provided no significant benefit in local control for low-grade DCIS (NG I with or without necrosis) and for all DCIS regardless of grade if the margins were greater than 10 mm or the re-excision was negative for residual DCIS (see figures).
Irradiation did provide a mean 13% benefit reduction in local recurrence rate in subsets of DCIS characterized by narrower margins, larger size, and/or higher grade (Van Nuys Prognostic Index scores 5,6,7). [See reference 3 for a description of the Van Nuys Prognostic Index.]
Like all other studies of irradiation for DCIS except B-17, half of the local recurrences were invasive regardless of radiation therapy. Longer follow-up shows that the differences in recurrence rates between irradiated and nonirradiated groups has begun to diminish, a feature suggested by prior studies of irradiation for DCIS.  Finally, in our patient population, invasive recurrences in the irradiated group were three times as large as in the nonirradiated group (35 mm versus 11 mm).
For patients most at risk of recurrence, those with NG III DCIS with any comedo necrosis, size greater than 40 mm, and margins less than 1 mm (Van Nuys Prognostic Index scores 8,9), radiation therapy made a major impact, reducing local recurrences from 100% at four years to 60% at six years, but clearly this is a pyrrhic achievement unacceptable in clinical practice.
Ours was not a randomized trial, but our methods of patient selection and clinical judgment are a more realistical reflection of actual clinical practice; moreover, we could not control the grade or the size of the DCIS, and patients frequently self-selected their mode of therapy regardless of recommendations. Our study does not compare treatments but, rather, prognostic factors over which we have no control.
We conclude that the morphologic heterogeneity of DCIS reflects a similar biologic heterogeneity and that therapies should be tailored to fit the actual risks of local recurrence. Although our methods must be validated, we have every expectation that they will be in programs now developing in other centers.