The Oncotype DX recurrence score (RS) is a widely used test that provides prognostic information on the likelihood of disease recurrence and predictive information on the benefit of chemotherapy in early-stage, hormone receptor–positive breast cancer. Despite its widespread use, quality assurance of the RS does not receive the same level of scrutiny as other tests, such as human epidermal growth factor receptor 2 (HER2) immunohistochemistry.
To use modified Magee equations to calculate the Magee score (MS) as a quality check of RS.
The MS is an easily accessible prognostic model that uses histopathologic and immunohistochemical criteria. We identified cases where the RS and MS differed by 10 points or more or were in different risk categories. These instances were considered significant discordances. MS was presented along with RS at multidisciplinary tumor boards, and all discrepancies were discussed to determine clinical significance and appropriate next steps.
Twenty-five of 155 cases (16.1%) had discrepancies between RS and MS. Of these 25 cases, 3 (12%) had problems with either the RS or the histopathologic interpretation. Among the cases with concordant RS and MS, no RS or interpretive problems were identified.
Use of the MS as a quality control check for the RS can help ensure appropriate treatment decisions in breast cancer patients. Pathologists can play a key role in ensuring the quality of molecular-based prognostic scores by using histopathologic models to ensure accurate risk stratification and improve clinical outcomes.
Breast cancer is a prominent cause of mortality and morbidity worldwide, and the incidence of invasive breast cancer has increased by 0.5% annually since 2004.1 Adjuvant chemotherapy and hormonal treatment have reduced the 10-year risk of locoregionally recurrent and distant metastatic disease, both of which have poor prognoses regardless of prior chemotherapy treatment.2,3 It is therefore important to offer the appropriate adjuvant chemotherapy regimen as a treatment option when it is indicated. At the same time, however, systemic therapy should be used judiciously with consideration for toxicity, cost, and individual tumor biology.
Molecular technologies have significantly evolved to better classify and predict prognosis of individual tumors. This has paved the way for the development of tailored treatment regimens targeting specific molecular mutations within individual tumors. Efforts to stratify breast cancers into molecular subtypes began in the early 2000s with the creation of a classification system composed of the luminal A, luminal B, human epidermal growth factor receptor 2 (HER2) overexpressed, and basal-like phenotypes.4 Among these, contemporary molecular tools have had the most success predicting prognosis and therapeutic response with the luminal subtypes. Several US Food and Drug Administration (FDA)–approved tests are available to specifically characterize this subset of tumors and guide treating providers’ decisions to offer or withhold chemotherapy based on predicted tumor behavior. These tests include the 21-gene recurrence score (Oncotype DX), Prosigna Gene Signature (PAM50), MammaPrint, and EndoPrint.5
The Oncotype Dx Breast Recurrence Score (RS) (Exact Sciences Corporation, Madison, Wisconsin) was initially developed as a prognostic test to stratify the risk of recurrence in early stage, hormone receptor–positive, HER2-negative breast cancer.6,7 Subsequent retrospective work showed that the RS was predictive of the benefit of chemotherapy.8,9 More recently, randomized prospective studies10–12 have demonstrated that the RS can identify those patients who can benefit from chemotherapy and those who can forego chemotherapy. Of the multigene assays available, the Oncotype DX assay is currently the only assay with sufficient evidence to support its use in the American Joint Committee on Cancer (AJCC) staging criteria.11 The integration of the RS into the AJCC 8th edition makes it a clinically important biomarker for breast cancer. A low RS can influence treating providers to not recommend adjuvant chemotherapy, drastically changing the treatment for breast cancer patients. Ensuring that the RS is accurate and reflects the true recurrence risk for individual patients should be a primary concern for pathologists and clinicians. Currently, there is no quality assurance model commonly used by submitting pathologists for RS.
Because of the impact the RS has on patient care and treatment decisions,13–16 it is important that a quality assurance process is in place to ensure the accuracy of the score. Although the RS has been shown to accurately predict the benefit of adjuvant chemotherapy,6,10,11 the accuracy of RS for individual patients has not been addressed. There are several known causes of falsely elevated RSs that are commonly encountered, such as the presence of extensive inflammation or biopsy site changes.17–20 Conversely, there are also known causes of falsely lowered RSs, such as the selection of nonrepresentative tissue blocks for molecular testing. A quality assurance model would benefit clinical decision-making by ensuring the accuracy of the RS.
The modified Magee equations use estrogen receptor (ER) and progesterone receptor (PR) H-scores, Ki-67 proliferative index, HER2 status, tumor size, and histologic grade to generate a Magee recurrence score (MS).21 The use of predictive models using commonly ordered laboratory tests and histologic information to predict RS has the potential to greatly reduce the cost of breast cancer management. The current cost of the Oncotype DX assay is $3873 based on 2023 Medicare reimbursement rates,22,23 and it can be cost prohibitive in resource-limited institutions.24 Over the years, the modified Magee equations have been shown to adequately estimate the Oncotype RS.25 In 2015, Turner et al21 demonstrated that 100% of cases with an MS of less than 21.5 had an RS of less than 30. Similarly, using TAILORx trial cutoffs in 2021, Glasgow et al24 confirmed that 97% of patients with an MS of less than 18 had an RS of less than 25. Of patients with an MS greater than 30, 90% had an RS of greater than 25. Adjuvant chemotherapy is often foregone for an RS of 25 or less. Therefore, it is important that a low MS (less than 25) accurately estimates a low RS (less than 25) and that a high MS (greater than 25) accurately estimates a high RS (greater than 25), as is reflected in the Magee decision algorithm.
It is suggested that the MS, especially in low-resource settings, can be used to forego RS testing for patients with MSs of less than 18 or greater than 30 based on the high concordance rates in the extreme categories.24 However, since the Magee equations were last revised in 2013, the clinical algorithms used in the Oncotype DX assay have undergone several changes. RS now accounts for age and menopausal status. Additionally, the test is now ordered for patients with both node-negative and node-positive (up to 3 involved lymph nodes) disease.26 The MS, by contrast, is designed for node-negative disease without consideration of patient age or menopausal status, posing the question of whether the Magee decision algorithm is outdated. Still, recent evidence continues to strongly suggest that MS closely estimates the current iterations of the RS.22 The use of the MS in a quality assurance model for RS, therefore, continues to be a viable method for ensuring the accuracy of RS because of its ease of implementation and low cost. In this study, we assess a quality assurance model for breast cancer RS testing using the modified Magee equations.
METHODS
Oncologists requested RS on invasive breast cancer specimens that were ER positive and HER2 negative or equivocal. We calculated MS for the same set of specimens using the modified Magee equations,27 taking into account Nottingham score, tumor size, ER and PR H-scores, Ki-67 percentage, and HER2 status. We used the average of the 3 modified Magee equations for each case. Nottingham scores (ranging from 3 to 9) and tumor size (greatest dimension in centimeters) were extracted from pathology reports. Monoclonal antibodies for ER (clone SP1; FDA approved; Roche/Ventana Medical Systems, Tucson, Arizona), PR (clone 1E2; FDA approved; Roche/Ventana), HER2 (clone 4B5; FDA approved; Roche/Ventana), and Ki-67 (clone 30.9; Roche/Ventana) were applied to 4-μm-thick sections of paraffin-embedded tissue using the Ventana Benchmark Ultra automated immunohistochemistry (IHC) system. Semiquantitative ER and PR H-scores (0–300) were calculated as the products of stain intensity (0 = no staining; 1 = weak staining; 2 = moderate staining; 3 = strong staining) and percentage of tumor cells staining positively. The Ki-67 labeling index was recorded as the percentage of tumor cells with positive nuclear staining. HER2 status was obtained from the surgical pathology report, recorded as positive (IHC score 3+, or 2+ with positive fluorescence in situ hybridization [FISH]), negative (IHC score 0, 1+, or 2+ with negative FISH), or equivocal (IHC score 2+ with equivocal FISH) based on the 2013 and 2018 updated American Society of Clinical Oncology (ASCO) and College of American Pathologists (CAP) guidelines.28 For both RS and MS, we determined whether each case was high or low risk based on the cutoffs for RS outlined in the TAILORx trial.29 We categorized RS and MS into 2 score groups of 0 to 25 (low risk) and 26 to 100 (high risk). We calculated the proportion of all cases that were classified as high risk versus low risk. We additionally calculated a score difference for each case by subtracting RS from MS. We considered RS and MS discordant if (1) they fell into different risk categories (resulting in a risk mismatch that would result in different therapy recommendations) or (2) they differed by greater than or equal to 10 points. Any discordant scores prompted a histologic review of the case. We believe that this somewhat arbitrary threshold of 10 points for a “large” score difference represents a considerable discrepancy between the 2 scores because MS and RS range up to only 31 and 45, respectively. This review included confirmation of the Nottingham score, tumor size, and interpretation of ER, PR, and HER2 status. Additionally, we verified that a representative section of tumor was submitted for the Oncotype DX assay. We considered a representative section to be a slide that had the highest percentage of tumor content (when compared with other sections from the same case) with minimal necrosis, inflammation, and biopsy site changes. If the review identified an issue, the pathologist corrected the error, amended the final pathology report, and recalculated the MS. The RS and MS were reported together at all multidisciplinary tumor boards, with the MS being treated as a quality check on RS. Any case that had a significant discordance between MS and RS was discussed to ensure that the discrepancy was clinically significant and to decide next steps in care, including whether retesting RS was warranted. We identified the proportion of cases that represented either (1) acceptable agreement between MS and RS, (2) a risk mismatch, (3) a large score difference, or (4) a risk mismatch and a large score difference. Correlation between MS and RS was assessed by fitting a simple linear regression model. The coefficient of determination (R2) from the linear regression was used to quantify the proportion of variance in RS that can be explained by the linear relationship with MS. We performed all analyses in MATLAB R2022b.
For all cases, breast tissue was placed into formalin within 1 hour of removal from the body, and breast tissue was in contact with formalin for at least 6 hours, not to exceed 72 hours. Cold ischemic time and total time in formalin were recorded for each case during initial processing to ensure compliance with these time requirements.
RESULTS
From 2017 to 2023, 155 cases of invasive breast carcinoma had an RS and an MS. Patient age ranged from 28 to 89 years, with a median age of 60 years. RSs ranged from 1 to 45, with a mean and median of 16. MSs ranged from 7 to 31, with a mean and median of 16. We categorized RS and MS into 2 score groups of 0 to 25 (low risk) and 26 to 100 (high risk).29 One hundred thirty-nine cases (89.7%) had an RS of 0 to 25, and 16 cases (10.3%) had an RS greater than 25 (Figure, A). One hundred forty-seven cases (94.8%) had an MS of 0 to 25, and 8 cases (5.2%) had an MS greater than 25 (Figure, A). For 137 cases (88.4%), the MS and RS scores differed by less than 10, but MS was 10 or more points higher than RS for 11 cases (7.1%) and 10 or more points lower than RS for 7 cases (4.5%) (Figure, B). Of the 155 cases, 130 (83.9%) had concordant RS and MS values, 7 (4.5%) had a risk mismatch only, 13 (8.4%) had a score difference of 10 or more only, and 5 (3.2%) had both a large score difference and a risk mismatch (Figure, C). Twenty-five cases (16.1%) had an unacceptable discrepancy between MS and RS. MS and RS were positively correlated, with an R2 value of 0.37, indicating that MS explained approximately 37% of the variation in RS (Figure, D) prior to repeat RS testing or correction of histologic grade.
Review of the discordant cases revealed a subset of cases with errors in grading or the selection of blocks with known causes of falsely elevated RSs, such as prominent biopsy site changes and inflammation. Identification of these errors allowed for repeat Oncotype DX testing in 2 cases, both of which resulted in significantly lower RSs with potential changes in management decisions. Review of one case with an incorrect Nottingham score of 4 resulted in reclassification as a high-grade tumor. RS-MS discordance in the following 5 cases prompted review and subsequent retesting or risk reclassification.
The first case had an RS of 21 and an MS of 11. Histologic review confirmed low-grade morphology and ER and PR H-scores of 300. Sections submitted for RS testing had limited tumor and prominent biopsy site changes, a known cause of falsely elevated RS.17–20 Repeat testing using the previous core biopsy of the same lesion returned an RS of 9.
The second case had an RS of 38 and an MS of 22. Histologic review resulted in an amendment of the histologic grade with a change in the Nottingham score from 4 to 8. The MS was 27 using the higher Nottingham score in the calculation.
The third case had an RS of 26 and an MS of 16. The Oncotype DX assay was performed on the core biopsy specimen because of the patient’s age and the potential for neoadjuvant therapy. Histologic review of the tissue submitted for the assay revealed minimal tumor with abundant lymphocytic inflammation, a known cause of falsely elevated RS.20 Given that no additional tissue was available to repeat the Oncotype DX assay, the discrepancy was not resolved. The clinical decision was made to pursue neoadjuvant chemotherapy.
The fourth case had an RS of 22 and an MS of 14. Although considered low risk by the TAILORx cutoffs, this RS was classified as intermediate risk (between 18 and 31) at the time of diagnosis using the original risk classification system established in 2004, prompting further investigation into a risk category mismatch.12 Histologic examination of the tissue submitted for the Oncotype DX assay revealed minimal tumor with significant surrounding biopsy site changes and scar. Repeat testing using the previous core biopsy of the same lesion returned an RS of 12. In addition, review of the resection specimen by Genomic Health resulted in “failure” of the originally issued RS.
The fifth case had an RS of 2 and an MS of 22. Histologic examination confirmed the presence of representative tumor in the resection tissue sent for Oncotype DX testing. Repeat testing was performed on the core biopsy of the same lesion because of the large discrepancy between RS and MS. On repeat testing, the RS score was again 2. A review failed to identify an explanation for the discrepancy.
DISCUSSION
In this study, we have demonstrated that MS can be used for quality assurance to identify cases whose results inaccurately reflect tumor biology and thereby the risk of recurrence. One hundred fifty-four cases were examined for concordance of RS and MS values. Repeating Oncotype DX testing in 3 cases resulted in changes of the originally issued RS in 2 cases. Histologic review in one case resulted in an amendment of the histologic grade. Our findings suggest that the use of MS as a quality check for RS can help identify errors in prognostication, including grading and the selection of nonrepresentative tissue for molecular testing, ensuring accurate reporting of tumor morphology and RS. With the recent incorporation of Oncotype DX into the AJCC staging criteria and clinicians’ continued reliance on the results of this test, it will become increasingly important to have such quality assurance measures in place to ensure accurate reporting of RS.
It is well documented that the Magee equations accurately estimate RS,25 especially for tumors with an MS less than 18 or greater than 30. However, the prospective use of the MS as quality assurance for the RS has not yet been explored. In fact, the discussion of quality assurance for Oncotype DX is sparse, limited to a small handful of studies investigating the concordance between HER2 and ER/PR expression by in-house IHC or FISH as compared with the Oncotype DX real-time reverse transcription–polymerase chain reaction.30,31 None of these studies involved the systematic use of a quality check to ensure that each reported RS had not been unduly influenced by known preanalytical and analytical variables. These include the selection of nonrepresentative tissue for molecular testing, the presence of extensive inflammation,20,32 the inclusion of significant biopsy site changes,17,33 incorrect grading, and incorrect interpretation of hormone receptor status or Ki-67 percentage.
As the impact of RS on treatment course continues to expand,12,34,35 the importance of identifying and minimizing analytical errors cannot be overstated. There is little question that gene expression profiling has enhanced the ability to provide tailored treatment options to breast cancer patients. However, the benefit is only as good as the accuracy of the results. The Oncotype DX assay undergoes quality control and analytical validity assessments,36 but our institution’s experience highlights the need for an additional case-by-case, systematic quality check at the submitting institution to ensure the integrity of each RS. In each instance, there is a potential clinical implication, with the possibility of overtreating or undertreating the patient based on inaccurate RS results. Using the MS as a quality check allowed for early identification of these errors, better understanding of each patient’s tumor biology, expanded discussion at multidisciplinary tumor boards, and optimized treatment decisions.
Our R2 of 0.37 is somewhat lower than the value reported by the developers of the Magee equation (R2 of 0.44).21 This suggests that although MS is moderately informative of RS for the cases in our study, for many cases the 2 scores do not align closely. Of 155 cases, we identified 25 (16.1%, or nearly 1 in 6) that had a risk mismatch, a score difference of 10 or more, or both. Differences in risk category may be more important for patient outcomes than large score differences, as risk category differences result in different treatment. However, the relatively substantial number of cases in which there was a large score difference suggests that providers and pathologists should take care not to consider MS and RS as entirely analogous scores.
The potential impact of this quality assurance method in the United States can be quantified by extrapolating the Surveillance, Epidemiology, and End Results program and National Cancer Database Oncotype DX use statistics and invasive breast carcinoma incidence data. Review of nationally reported Oncotype DX use in 2017 (62 289),37 2016 (60 564), and 2015 (58 582) suggests a 3.1% increase year to year, which predicts Oncotype DX use of 72 561 in 2022. Alternatively, assuming 287 850 new cases of invasive breast cancer in female patients in 2022,1,38 a hormone receptor–positive, HER2-negative rate of 68% of invasive breast cancers,1 and an Oncotype DX use rate of 34% of all hormone receptor–positive, HER2-negative (by IHC or Oncotype DX reverse transcription–polymerase chain reaction) invasive breast cancers,39 Oncotype DX use is expected to be 66 551. In the present study, 2 of 155 cases (1.3%) were reclassified to lower risk groups after repeating RS testing. A reclassification rate of 1.3% because of routine Oncotype DX quality assurance review, therefore, could potentially impact between 865 and 943 patients’ treatment courses annually.
Our institution’s experience with Oncotype DX use highlights the need for a quality assurance model for molecular test results in breast cancer. Pathologists are best suited to weigh all considerations in the molecular investigation, including tumor heterogeneity, histologic correlation, block selection, and assay preference.40 Routine quality checks using histopathologic features and commonly available laboratory tests empower pathologists to have a positive impact on patient care and treatment planning discussions.
Since 2010, ASCO and CAP have partnered to provide guidelines for hormone receptor testing in breast cancer,41 to include recommendations for external quality assurance and an internal quality assurance model. Several updates have subsequently been released with strong recommendations for Oncotype DX testing in ER-positive, HER2-negative, node-negative patients since 2016.42 However, even in the most recent 2022 ASCO guideline update for breast carcinoma,43 the issue of quality assurance for any of the recommended molecular tests, including Oncotype DX, was not addressed. Fortunately, there is precedent for the recommendation of routine tissue quality assessment in molecular send-out tests.
In their 2017 guideline for molecular biomarker evaluation of colorectal carcinoma, ASCO, CAP, the Association for Molecular Pathology, and the American Society for Clinical Pathology jointly recommended routine evaluation of specimens for tissue quality, quantity, and tumor cell fraction.44 This recommendation, issued as an expert consensus opinion, suggested that pathologists submitting tissue for molecular examination should have knowledge of the tissue requirements for the requested assay, should consider the impact of tumor heterogeneity, and should be aware of the need to submit representative lesional tissue to have an accurate result. The committee also made a pointed recommendation to avoid submission of tissue with necrosis and degeneration, as these features are known to produce erroneous results. The current study demonstrates that similar problems are encountered with Oncotype DX. Equally pointed recommendations for routine quality assurance in breast cancer molecular testing would, therefore, be welcome and could greatly improve the care of select patients.
Although RS is considered more useful in terms of guiding treatment, MS should also be calculated. This is particularly important when RS is close to risk thresholds.21 Falsely high or falsely low RSs can result in unnecessary treatment or undertreatment, respectively. Therefore, using the MS as a quality check of the RS is an accessible way to ensure that appropriate treatments are being identified and offered to patients on a case-by-case basis. A limitation of this study is that the results are based on only one institution. Future studies should investigate how generalizable the results are to other institutions that may have different local practice characteristics.
In summary, use of the MS as a quality check for the RS can help ensure appropriate treatment decisions in breast cancer patients. Pathologists can play a key role in ensuring the quality of molecular-based prognostic scores by using histopathologic models to ensure accurate risk stratification and improve clinical outcomes. The variables that may lead to inaccurate prognostic predictions include the selection of nonrepresentative tissue for molecular testing, the presence of extensive inflammation, the inclusion of significant biopsy site changes, incorrect grading, and the incorrect interpretation of hormone receptor status or Ki-67 percentage.
Recurrence scores (RSs) and Magee scores (MSs). A, Proportion of cases that were classified as high risk (blue) or low risk (gray) using each score. B, Histogram of differences between the 2 scores, with large score differences highlighted in orange. C, Proportion of cases that showed an acceptable (gray) or unacceptable (colored) difference between MS and RS. D, Relationship between MS and RS prior to repeat RS testing or correction of histologic grade. Point colors indicate whether the difference between MS and RS is acceptable (gray) or unacceptable (colored). A regression line with 95% CIs is shown. The coefficient of determination (R2) from the simple linear regression model represents the proportion of variance in RS that can be explained by the linear relationship with MS.
Recurrence scores (RSs) and Magee scores (MSs). A, Proportion of cases that were classified as high risk (blue) or low risk (gray) using each score. B, Histogram of differences between the 2 scores, with large score differences highlighted in orange. C, Proportion of cases that showed an acceptable (gray) or unacceptable (colored) difference between MS and RS. D, Relationship between MS and RS prior to repeat RS testing or correction of histologic grade. Point colors indicate whether the difference between MS and RS is acceptable (gray) or unacceptable (colored). A regression line with 95% CIs is shown. The coefficient of determination (R2) from the simple linear regression model represents the proportion of variance in RS that can be explained by the linear relationship with MS.