This case series demonstrates that academic breast oncologists may be reasonably accurate at using standard prognostic criteria to distinguish high-risk patients from low- or intermediate-risk patients as determined by the RS. High agreement was observed among oncologists in predicting the Oncotype DX® RS, especially when designating an RS of low or intermediate versus high risk. There still remains significant inability to discern low- from intermediate-risk cases, as evidenced by the large number of discrepancies that occurred.
In this analysis, most oncologists did not recommend chemotherapy for predicted low RS patients, whereas they always recommended chemotherapy for predicted high RS patients. Predicted intermediate scores led to wide variability in treatment recommendations, but most oncologists did not recommend chemotherapy in patients whom they felt to be at intermediate risk. These practice patterns were also seen in a retrospective review of 285 patients in which RS and chemotherapy decisions were recorded in a multicenter, community-based, health care system [16
]. In the current study, the added information of the actual RS changed the chemotherapy recommendation in 19% of cases, resulting in an overall decrease of five clinical scenarios for which chemotherapy would be recommended. Cases with alternative pathology reads were associated with changes in predicted RS risk in over one third of scenarios and were associated with changes in the chemotherapy recommendation in >10% of cases.
Others have performed similar analyses. Acs et al. [17
] recently presented their findings on 154 patients with early-stage, ER+
breast cancer and an available RS in which surgical oncologists, medical oncologists, and pathologists were asked to estimate the risk for recurrence. Risk estimates agreed with the RS 54% of the time, with the most common discrepancy being overestimation (32%). Using the assumption that patients with a low- or intermediate-risk RS do not benefit from chemotherapy, and are thus “overtreated,” they concluded that 82% and 69% of patients would unnecessarily receive chemotherapy without and with the use of the RS, respectively (p
= .03). Lo et al. [18
] recently reported the results of RS predictions and treatment recommendations before and after the RS was obtained from 17 medical oncologists in 89 patients. Oncologists changed the treatment recommendation in 32% of cases after knowing the RS results. Similar to the current study findings, the greatest change was from a pretreatment recommendation of chemotherapy to a post-test recommendation of hormonal therapy alone. The authors allowed an answer of “equipoise,” reflecting neither superiority nor inferiority for either treatment choice. If one assumes that an answer change to equipoise reflects ambiguous conclusions of superiority but no actual treatment change, then the percentage of treatment recommendations changed by the RS may be 26%, close to our findings.
How much variation in treatment recommendations can be attributed to chance alone is not well understood. Studies have shown that the practice patterns of medical oncologists in managing patients with early-stage breast cancer vary and are often discordant with national guidelines [19
]. Based on repeated measures GEE analysis of chemotherapy recommendations, the effect of being given the RS on an oncologist's chemotherapy recommendation was not statistically significant (i.e., the change within a given oncologist in the chemotherapy recommendation was not larger than that attributed to random fluctuation in an oncologist's recommendation).
These data lead to several important observations. First, conventional prognostic criteria are useful to identify the risk for distant recurrence as assessed by the RS. As expected, the trend to recommend against chemotherapy in predicted scenarios of low RS risk and to recommend chemotherapy in predicted scenarios of high RS risk did not significantly change after being given an actual low and high RS. However, differences emerged in the frequency of chemotherapy recommendation when not given the RS versus being given the RS in the intermediate-risk group (44% versus 25%; p = .035). Thus, up to 19% of patients would have been misclassified and not received chemotherapy, with clinically important implications.
A recent economic analysis of the Oncotype
DX® test concluded that the RS can actually augment the classification made by the NCCN system. In that study [21
], 28% of patients stratified by the NCCN criteria as low risk were reclassified by the RS as intermediate or high risk. Conversely, almost 50% of patients classified as high risk were reclassified by the RS as low risk. Assuming that these newly classified patients would not receive chemotherapy, RS testing would increase the quality-adjusted life-years by 8.6 years and produce a 5% decline in overall health care–associated costs. These results suggest that the clinical utility of the RS is in reducing costs, morbidity, and mortality of chemotherapy without changing outcome. Our data show similar results in the number of cases predicted to be low risk reclassified to intermediate or high risk (20%), but no predicted high-risk cases were truly low risk by the Oncotype
There are several limitations to this study. The relatively low number of cases studied is a reflection of the lack of national consensus on the use of the RS at the time of the study. The issues of experience and expertise are important to note. The ability to generalize the current findings to settings outside tertiary-care cancer centers with dedicated breast oncologists and highly experienced breast pathologists is likely limited. Even though two well-experienced breast pathologists, blinded to outcomes, read the cases, this study demonstrated that even slight differences in the quantification of hormone receptor expression may change the predicted RS and chemotherapy recommendation. To a lesser extent, this may also be true of small differences in histologic grade. But even considering this, minor changes in pathologic reads were associated with changes in the predicted RS risk in 54% of cases and resulted in a change in the chemotherapy recommendation in 13% of cases. This underscores the issue that even minor discrepancies in pathologic reads can change clinical decisions in an important number of clinical scenarios.
Finally, it should be noted that the RS is weighted using information regarding HER-2 and proliferation genes (e.g., Ki-67). Although the oncologists in this study were provided information regarding HER-2 (standard immunohistochemistry [IHC] and fluorescence in situ hybridization [FISH] if appropriate), they were not provided with an immunhistochemical marker of molecular grade (Ki-67). Therefore, it is possible that the addition of this information may have further improved the ability of academic oncologists to discriminate among RS categories. Notably, only two cases were considered HER-2+ by standard IHC and/or FISH criteria, and the current NCCN guidelines do not recommend using the RS for tumors that are HER-2+.