In this study of 186,867 patients with HR positive, HER2-negative invasive breast cancer who underwent 21-gene RS testing in the National Cancer Database, we found that patients with ILC have higher rates of discordant clinical and genomic risk than those with IDC. This finding is consistent with our prior work, which showed higher rates of discordant clinical and genomic risk in ILC patients using the 70-gene signature . Additionally, we found that ILC patients with 1-3 positive nodes were significantly more likely to have RS ≤ 25 compared to those with IDC, in both those under age 50 years and those age 50 years or older.
Among those with RS ≤ 25, there was no difference in receipt of chemotherapy by histology regardless of nodal involvement. However, among those with RS >25, those with ILC were significantly less likely to receive chemotherapy than those with IDC among patients with negative nodes or 1-3 positive nodes. While these findings are consistent with our prior study, which showed lower rates of chemotherapy use in patients with high clinical risk ILC and high genomic risk as defined by 70-gene signature testing, they are nonetheless surprising given the general acceptance of chemotherapy in the setting of RS > 25 and did not appear to be driven by older age in the ILC group [5, 14]. These findings may reflect hesitation on the part of clinicians and patients to utilize chemotherapy in ILC, where multiple studies have shown less benefit in the neoadjuvant and adjuvant settings [22, 23]. Together, this illustrates the treatment dilemma that clinicians and patients face: while patients with ILC are more likely to have high clinical risk, which portends increased risk of recurrence without chemotherapy, genomic assays and reported series suggest decreased chemotherapy benefit .
We did not demonstrate any effect of chemotherapy on overall survival among patients with 1-3 positive nodes and RS ≤ 25 in our propensity score matched analyses, regardless of histologic subtype. Interestingly, we did find an association between chemotherapy and overall survival in unmatched analysis of IDC patients under age 50 with 1-3 positive nodes and RS ≤ 25, but not among those under age 50 with ILC. Given the retrospective nature of this analysis, differences in patient selection for receiving chemotherapy likely contribute to chemotherapy related outcomes. While we attempted to account for potential confounders in treatment selection by using propensity score matching by age at diagnosis, pathologic stage, and facility type, it is unlikely that we were able to fully adjust for potential bias between the chemotherapy and non-chemotherapy groups.
This bias is greater in our exploratory unmatched analysis in patients under age 50, which showed a significant improvement in overall survival among IDC patients who received chemotherapy compared to those who did not but no difference in those with ILC. There are several potential explanations for the findings of our exploratory analyses. One possibility is that among those under age 50 with 1-3 positive nodes and RS ≤ 25, the overall survival difference observed in the IDC group resulted from patient selection bias and not chemotherapy effect. Another explanation is that we were unable to detect an overall survival benefit in the ILC patients under age 50, with 1-3 positive nodes, and RS ≤ 25 because of small numbers in this subgroup. Lastly, it is possible that chemotherapy improves overall survival in IDC patients but not ILC patients under the age of 50 with 1-3 positive nodes and RS ≤ 25; for this reason, reporting of long-term and overall survival outcomes by histologic subtype from trials such as RxPONDER is needed.
While the RxPONDER trial showed a significant improvement in invasive disease-free survival in patients under age 50 who had 1-3 positive nodes and RS ≤ 25 who received chemotherapy, we cannot directly compare our results since the National Cancer Database lacks recurrence endpoints. It is important to note that for patients with HR positive HER2 negative tumors, and ILC especially, recurrence events and consequently impact on overall survival can happen at later timepoints, highlighting the need for longer term follow-up [25, 26]. Additionally, we lacked data on type of endocrine therapy which likely impacts outcomes.
This study has many strengths, including the use of a relatively large numbers of ILC patients, and is now the second study to demonstrate that patients with ILC have high rates of discordance between clinical and genomic risk based on widely used molecular assays. While many studies show that ILC tumors have lower response rates to chemotherapy in the neoadjuvant setting, and less benefit from chemotherapy in the adjuvant setting, there may still be a subset of chemotherapy sensitive ILC cases. Recent work has identified genomic signatures that identify subtypes within ILC, suggesting heterogeneity within this tumor type [27, 28]. The high incidence of high clinical risk among patients with ILC highlights the need for both more effective therapies, and potentially ILC specific prediction tools. More broadly, improving outcomes for these patients with ILC will require not only equitable enrollment of ILC patients into breast cancer clinical trials, but also histologic subtype specific reporting of trial results.