RSClin Delivers More Precise Assessment of Distant Recurrence Risk and Chemo Benefit in Early Breast Cancer

Joseph A. Sparano, MD

The clinical tool RSClin™, which integrates the 21-gene expression assay and clinical pathologic features, was shown to provide more prognostic information than the 21-gene recurrence score or clinical pathologic features alone and a more precise prediction of absolute chemotherapy benefit in node-negative early breast cancer, according to findings from a study that were presented during the 2020 San Antonio Breast Cancer Symposium.

“RSClin is a new, validated clinical tool that integrates clinical and genomic risk to guide adjuvant chemotherapy use in node-negative early breast cancer with greater precision for patients with low or high recurrence score results. Recommendations for treatment based on TAILORx without RSClin is generally clear. For other scenarios that may not be as clear, RSClin is a new option that may be used to facilitate discussions and shared decision-making,” Joseph A. Sparano, MD, lead study author, and associate director for clinical research, Albert Einstein Cancer Center, professor of medicine and obstetrics, gynecology, and women’s health at the Albert Einstein College of Medicine, said in a virtual presentation of the data.

The 21-gene recurrence score provides prognostic and predictive information for distant recurrence and chemotherapy benefit, respectively, in estrogen receptor–positive, HER2-negative early breast cancer. However, clinical-pathologic features provide only prognostic information. The integration of genomic and clinical features provides an opportunity to direct adjuvant chemotherapy use with greater precision.

To that end, investigators sought to develop and validate a new tool, RSClin to inform prognosis and prediction of absolute chemotherapy benefit. For development, a patient-specific meta-analysis (PSMA) was performed using data from the NSABP B-14, TAILORx, and NSABP B-20 trials for individualized prognosis and individualized prediction of chemotherapy benefit.

Covariates in the model for individualized patients included recurrence score, tumor grade (low vs intermediate vs high), tumor size, and age. PSMA used Cox regressions of NSABP B-14 and TAILORx for log cumulative hazard estimate for prognosis of 10-year distant recurrence risk with endocrine therapy alone and the baseline hazard from TAILORx to represent current practice. PSMA also used Cox regressions of NSABP B-20 and TAILORx for log hazard ratio estimates for chemotherapy benefit, which when combined with this recurrence risk estimate for endocrine therapy alone provides estimates for 10-year absolute chemotherapy benefit.

External validation for prognosis was conducted by comparing RSClin risk estimates with observed distant recurrence rates in the real-world experience of the Clalit registry in Israel.

Regarding the prognosis model, the covariate effects of recurrence score (25 vs 10; HR, 1.80; HR, 3.38; HR, 2.58), tumor grade (2 vs 1; HR, 1.45; HR, 1.48; HR, 2.25), tumor grade (3 vs 1; HR, 3.53; HR, 2.56; HR, 2.35), tumor size (2.5 cm vs 1.5 cm; HR, 1.24; HR, 1.54; HR, 1.62), and age (65 vs 50; HR, 0.84; HR, 0.88; HR, 1.13) were generally comparable across NSABP B-14 (n = 577), arms A and B of TAILORx (n = 4854), and arms C and D of TAILORx (n = 4753), respectively, with the exception of age, which showed more variability, said Sparano, who is also associate chairman for Clinical Research in the Department of Oncology at Montefiore Medical Center.

Next, investigators assessed the prognostic information that is provided by RSClin and clinical-pathologic features alone with the likelihood ratio (LR) chi-square, which reflects the amount of information provided by a model, and the likelihood ratio test, which compares the amount of information provided by 2 models. When RSClin was compared with clinical-pathologic features alone, the likelihood ratio test favored RSClin in NSABP B-14 (Δ LR chi-square [degrees of freedom], 20.3[2]; P < .001), arms A and B of TAILORx (Δ, 25.2[2]; P < .001), and arms C and D of TAILORx (Δ, 40.7[2]; P < .001), reflecting a total Δ of 86.2([2]; P < .001).

Similarly, when RSClin was compared with recurrence score alone, the Δ favored RSClin across NSABP B-14 (Δ LR chi-square, 24.2[5]; P < .001), arms A and B of TAILORx (Δ, 40.2[5]; P < .001), and arms C and D of TAILORx (Δ, 58.1[5]; P < .001), reflecting a total Δ of 122.5([15]; P < .001).

“Therefore, likelihood ratio tests confirm that the RSClin model provides more prognostic information than either clinical-pathological variables, alone, or recurrence score alone,” said Sparano.

When the model was used to illustrate the estimated 10-year risk of distant recurrence, it was shown to provide prognostic information by recurrence score, irrespective of tumor grade, size, and age.

“With regard to prognosis for 10-year distant recurrence risk, RSClin provides individualized estimates that reflect current medical practice with endocrine therapy alone. It also provides more prognostic information than clinical-pathologic features or the 21-gene recurrence score used individually,” said Sparano.

Regarding the prediction model, the 10-year RSClin estimates for absolute chemotherapy benefit in a 55-year-old patient with a 1.5-cm tumor showed a positive correlation between chemotherapy benefit and recurrence score, irrespective of grade (1, 2, or 3). Additionally, higher-grade tumors derived greater absolute benefit from chemotherapy.

The same estimates in an intermediate (grade 2) tumor demonstrated a positive correlation between chemotherapy benefit and recurrence score, irrespective of size. Additionally, larger tumors derived greater absolute benefit from chemotherapy.

“With regard to prediction of absolute chemotherapy benefit, RSClin provided individualized estimates, and the individualized prediction of proportionally greater relative chemotherapy benefit with higher recurrence score contributed significantly to estimation of absolute chemotherapy benefit,” said Sparano.

“We also demonstrated external validation in the independent Clalit cohort, reflecting contemporary real-world practice,” said Sparano.

In terms of the external validation of RSClin, the characteristics of the Clalit cohort (n = 1098) were similar to those of the PSMA cohort (10,004). In comparing the average RSClin distant recurrence risk in each Clalit quintile with the Kaplan-Meier risk estimate using patient data in each Clalit quintile, investigators showed a Lin concordance correlation of 0.96. Moreover, the RSClin risk estimate showed a significant association with distant recurrence (standardized HR, 1.73; 95% CI, 1.40-2.15; P < .001).

When RSClin was used to determine the 10-year distant recurrence risk in clinical low-risk and clinical high-risk scenarios, the RSClin estimated absolute chemotherapy benefit varied significantly with recurrence scores ranging from 11 to 50 in both cases (low-risk model, 0%-15%; high-risk model, 1%-33%). Additionally, a smaller window of estimated absolute chemotherapy benefit was reported in the models, despite the same chemotherapy benefit hazard ratio, and varying 21-gene recurrence score (low-risk model, 2%-8%; high-risk model; 4%-17%).

“Yogi Berra once taught us that making predictions is hard. It is our expectation and hope that this new tool will make it easier to predict outcome and chemotherapy benefit in patients with early breast cancer who have the 21-gene recurrence score and with greater precision,” concluded Sparano.

Reference

Sparano JA, Crager MR, Tang G, et al. Development and validation of a tool integrating the 21-gene expression assay and clinical-pathologic features to individualize prognosis for distant recurrence and prediction of absolute chemotherapy benefit in early breast cancer. Presented at: 2020 San Antonio Breast Cancer Symposium; December 8-11; 2020; virtual. Abstract GS4-10. https://bit.ly/2IzUBQu