After standard-dose chemotherapy (SDC), more than 50% of patients with HRPBC (defined as extensive axillary node involvement or inflammatory disease) experience relapse. Controversy has surrounded the use of HDC for HRPBC for over a decade. Current results from at least 15 randomized trials comparing diverse forms of HDC and SDC for HRPBC appear contradictory at this point [1]. In addition, some studies suggest that younger women might benefit selectively from HDC. An NCI-sponsored meta-analysis is in progress. Regardless of this unsettled issue, it is clear that a substantial percentage of HRPBC patients still relapse after HDC. Extensive prognostic studies are required to improve their outcome. Such investigations may allow us to identify adequate patient subsets and new therapeutic targets for trials of developmental therapeutics. Combinations of HDC, capable of ample cytoreduction, with novel agents that target specific tumor features responsible for post-transplant relapse may hold promise.

We analyzed clinical variables in HRPBC patients enrolled at the University of Colorado in clinical trials of HDC targeting 4–9+, ≥ 10+ nodes, or inflammatory disease. First, we developed a prognostic model among 176 patients treated from 1990 to 1997, and validated it in an external sample. Subsequently, the model was validated prospectively in a second cohort of 88 patients treated at Colorado since 1997.

We hypothesized that intrinsic biologic differences, insurmountable by HDC, existed between the two risk categories identified by the clinical model. Through immunohistochemical analyses of paraffin-embedded tumor blocks collected from the referring institutions, we studied a series of putative molecular candidates, related to signal transduction pathways or an angiogenic phenotype, which could be responsible, at least in part, for those differences.

At median follow up of more than 7 years, the relapse-free survival (RFS) and overall survival (OS) rates for the whole group of 264 patients treated at Colorado were 69.8% and 73%, respectively. The median time to relapse was 14 months (63.5% relapses within the first 2 years, 6.7% after the 5th year).

We identified three clinical variables independently associated with outcome: nodal ratio (number of involved nodes/number of dissected nodes), pathological tumor size, and hormone receptors [2]. A scoring system was constructed with those variables: score = (nodal ratio × 3.05) + (tumor size × 0.15) - (ER/PR × 1.15). In this formula, size is entered in cm, and ER/PR is assigned '1' if positive (estrogen receptor [ER] and/or progesterone receptor [PR] positive), or '0' if negative (both negative). A cutoff score of 2.41 yields the best sensitivity and specificity. Thus, patients with low (<2.41) and high (≥ 2.41) scores before transplant presented significant differences in outcome. This model was validated in an external sample of 225 HRPBC patients treated at Duke University with the same HDC. It was subsequently validated prospectively in our second patient cohort [3].

Overexpression of HER2, identified as an independent predictor of outcome, complemented the clinical model, establishing the following risk groups: low risk (low score, HER2-negative; 44% patients; 87% RFS), intermediate risk (low score, HER2-positive; 29% patients; 68% RFS), high risk (high score, any HER2; 27% patients; 49% RFS) [4].

We detected an independent prognostic effect of EGFR (epidermal growth factor receptor), particularly among HER2-positive patients [5], which suggests a synergistic effect through heterodimerization of both receptors. In contrast, we did not observe a prognostic effect of p53 status [4].

Tumor angiogenesis, assessed through CD31-stained microvessel count, was an independent adverse predictor of outcome [6]. In contrast, tumor VEGF (vascular endothelial growth factor) expression lacked prognostic significance in our population with locoregionally advanced tumors, in contrast to multiple prior observations in patients with earlier disease [6].

Finally, we observed that the presence of tumor cells contaminating the apheresis product, detected through immunocytochemistry for cyto-keratins, was independently associated with post-transplant relapse [7].

We can now predict which HRPBC patients are most likely to remain long-term disease free after HDC. Additionally, we identified important prognostic molecular markers that could constitute relevant targets for studies combining novel therapeutics with HDC in HRPBC.