Milan I

The trial that arrived at the same answer from a different direction

In 1973, three years before Bernard Fisher opened enrolment in NSABP B-06 in Pittsburgh, Umberto Veronesi and his colleagues at the Istituto Nazionale dei Tumori in Milan began a trial that would travel to the same destination by a different road. That parallel journey is precisely what made the destination credible.

The Milan I trial asked whether a quadrantectomy — removing the tumour with a wide margin of surrounding tissue, the overlying skin, and the underlying fascia, plus axillary dissection and postoperative radiotherapy — produced the same survival as a Halsted radical mastectomy. It was a more conservative operation than the radical, but a more aggressive one than the lumpectomy Fisher was testing: the Italians removed considerably more tissue around the primary tumour, which reflected genuine scientific caution about where the acceptable margin lay. The two trials were therefore not duplicates; they bracketed the question of how little surgery was enough.

Eligibility was restricted to women with tumours no larger than two centimetres and clinically negative axillary nodes — a narrower entry criterion than B-06, which accepted tumours up to four centimetres. Roughly 700 women were randomly assigned to one of the two arms. The randomisation was straightforward and the follow-up methodical. What the trial had working in its favour, beyond Veronesi's organisational grip on a single institution, was exactly that institutional coherence: surgery, pathology, radiotherapy and data collection all happened in one place, reducing the variability that multi-centre trials must work to control.

The results, first published in the New England Journal of Medicine in 1981, showed no statistically significant difference in overall survival between the two groups. Local recurrence was somewhat higher in the conserving arm, as the investigators expected and reported honestly, but it did not translate into a survival disadvantage. Long-term follow-up, published over subsequent decades, confirmed the finding held at ten, fifteen and twenty years: survival curves for the two groups remained essentially superimposed.

The intellectual force of Milan I lay in its relationship to B-06. Fisher's trial operated across multiple American centres, used a broader patient population, and tested a less extensive resection. Veronesi's trial ran at a single European centre, used a more restricted population, and tested a wider excision. Both found equivalent survival. The convergence of two independent trials, built on different assumptions, in different countries, with different surgical procedures, arriving at the same answer, is what closed the argument that the radical mastectomy was necessary to control the disease — or at least what should have closed it. In practice, uptake of conserving surgery remained slow through the 1980s, and the 1990 National Institutes of Health consensus conference was still needed to formally endorse breast conservation as standard for eligible patients.

Milan I also fed directly into the Oxford overviews — the periodic meta-analyses run by the Early Breast Cancer Trialists' Collaborative Group, which pooled individual patient data from trials worldwide. When Veronesi's data entered that pool alongside Fisher's and the Scandinavian trials, the aggregate picture became statistically robust in a way that no single trial, however well-conducted, could achieve on its own.

The distinction between Milan I and B-06 carried a lasting lesson about surgical margins. Veronesi's quadrantectomy left local recurrence rates lower than Fisher's lumpectomy. The more tissue removed around the primary, the less likely the cancer was to return locally — a dose-response relationship within the spectrum of conserving surgery. That relationship later informed debates about the minimum acceptable margin width, a question that pathological assessment of the resected specimen would have to answer case by case.

Veronesi went on to lead further trials at the same institution, testing sentinel-node biopsy as a replacement for full axillary dissection, working from the same conviction that drove Milan I: that the right amount to remove is the minimum that does not compromise the outcome, and that only a properly randomised trial can find where that minimum lies.