The metLN, tumor size, histologic grade, and hormonal receptor expression status are reported to be the main prognostic factors associated with breast cancer [11
]. As the metLN is the only LN-related prognostic factor recognized by AJCC [9
], it is reasonable to question whether it is the most accurate way to represent the status of metastatic cancer in the LNs.
The dimension of tumor involvement in positive LNs has been distinguished in stages N0 and N1 as isolated tumor cells, tumor clusters, and micrometastasis by AJCC LN staging system [9
]. Survival analyses of the breast cancer patients with isolated tumor cells and micrometastasis in the sentinel lymph nodes showed controversial results [22
]. There is no further distinction of tumor dimension above 2
mm in the LNs, whereas only the metLN is taken into account. In other words, an LN with just a small focus of subcapsular tumor invasion and an LN with a near total replacement by the tumor are both counted as one positive LN. In addition, there is the dilemma that counting matted LNs may render a source of variability in determining the pathologic N stage [10
]. Although the largest dimension of the positive LN was recorded in our staging summary and can be used to reflect the extent of tumor involvement in cases of the matted LNs, this measurement has never been evaluated as a variable in the TNM staging system. To address these issues, we initially hypothesized that measuring the CSCA of the LNs instead of counting the metLN would give a more accurate picture of the metastatic cancer volume in the axillary LNs. The present state of computer technology allows us to measure the CSCA accurately and objectively.
Previously, LN CSCA has been studied in a variety of cancers. For example, in melanomas, LN CSCA has been found to correlate with the Breslow thickness and the likelihood of further nodal involvement in completion of LN dissection [28
]. Percentage of CSCA in sentinel LNs in breast cancer has also been evaluated and reported to be the most important predictor of frequency of additional positive nonsentinel LNs in multivariate analysis [30
]. However, there is prognostic analysis based on CSCA in this study of breast cancer. Early study on esophageal cancer by Komori et al. has reported the size of the largest cancer nest in the LNs is one of the most important prognostic factors [31
]. Recently, growing evidence and a large body of literatures summarized by Petrelli et al. [32
] on esophageal and other gastrointestinal cancer have demonstrated that LNR is an independent powerful predictor of estimated survival.
In our study, 292 breast cancer patients diagnosed from 1998 to 2000 were included for the evaluation of prognostic factors of breast cancer, allowing a sufficient time frame for follow-up study. Univariate Kaplan-Meier analysis demonstrated LN CSCA was a better prognostic indicator than N stage where N2 and N3 overlapped and crossed (). This indicates that patients with greater than 4 positive LNs (N2 and N3) showed similar survival without significant difference. There are other studies that support this observation [5
]. We also evaluated different cutoff points of LN CSCA, but none of them appeared to be able to categorize patient's survival very distinctively. Multivariate Cox's regression model demonstrates that CSCA (but not metLN) is an independent prognostic indicator when present alone without other LN factors. However, when both CSCA and metLN are present in the model, neither turns out to be an independent predictor. Furthermore, we thoroughly compared the LN CSCA method with N staging method in terms of cancer definition including metLN, histologic grade, T stage, and ECE (data not shown). These two methods were similar in characterizing the breast cancer, and they are both dependent on the metLN. Our results indicate that the quantitative LN CSCA can serve as an alternative to the AJCC's current LN staging system.
LNR in breast cancer has been studied in recent years and proposed to be a promising prognostic factor that outperforms the currently used LN staging scheme. Vinh-Hung et al. reviewed and summarized the latest study on the prognostic value of LNR in 2009 [18
]. The variation of the LN dissection and the inconsistency of axillary LN evaluation are caused by a biological variation in patients as well as different techniques across institutions [33
]. In addition, some patients who were staged with only one sentinel LN underwent neoadjuvant chemotherapy without axillary dissection [2
]. Thus, the N stage represented by the absolute metLN may not be a universal indicator of prognosis in breast cancer. For example, the difference between 3/3 (3 positive out of 3 total LNs examined) and 3/20 LNs (3 positive out of 20 total LNs examined) could be due to technical variations from different surgical approaches. Given the small denominator, it is unclear whether it is possible to harvest more LNs in the first scenario (3/3), and how many LNs would turn out to be positive had they been harvested. Both cases are categorized as N1 tumor based on the metLN. In contrast, when the total number of LNs is considered in the calculation of LNR, the cases have LNRs 1.0 (3/3) and 0.15 (3/20), respectively, and will likely fall into different LNR categories depending on the cutoff points applied.
We added LNR into our group of LN-related factors for the prognostic comparison. We adapted the analytical method used by Vinh-Hung et al. for our data analysis [18
]. Log-rank test demonstrated significantly different survival among three LNR groups (<0.1, 0.1–0.65, >0.65) (). In our multivariate analysis, LNR was the only independent significant prognostic predictor for cancer mortality among all three LN-related factors ().
We further evaluated the lack of agreement among the three different categories in survival prediction by comparing metLN and total LN examined (). Interestingly, our observation revealed that the metLN in all methods increased within the groups, while the total number of LNs examined increased from N1 to N3, stayed the same in the LN CSCA groups, and decreased from low to high LNR groups. Thus, taking the ratio of positive over total LNs is a powerful method with the acknowledgment of total LNs as an “internal control,” balancing off the unevenness from counting only the absolute metLNs.
A comprehensive assessment of all aspects of breast cancer diagnosis is beyond the scope of this paper. The current study takes the advantage of single patient population treated by one surgical team using consistent surgical procedures within a 3-year time period. This allows a retrospective comparison focusing on the different LN-related factors with the minimal variability among other parameters. The automated programs made the cancer area measurement simple and quick. To the best of our knowledge, it is the first study that compared three LN-related methods (N stage, LNR, and LN CSCA) on the same population of breast cancer patients from one institution diagnosed in three consecutive years.
On the other hand, we acknowledge the limitations of our study. The relatively small sample size due to a single source of patient population can affect statistical power of the study. Our comparison did not account for the percentage replacement of LNs by CSCA and its association to the survival, which will serve as an objective in our future follow-up study. The information regarding local recurrence, metastatic disease, and chemoradiation therapy was not included in our study. However, previous studies have shown that LNR can be used in decision making when the aforementioned conditions are present [18
]. In the cases without axillary LN dissection, that is, patients who underwent chemotherapy with the biopsy of only one positive sentinel LN (1/1), it was difficult to evaluate the LNR without sufficient information of the denominator. However, we still included these patients in our study owing to the comparison to other LN-related parameters (N stage and LN CSCA).
In conclusion, LNR categorizes breast cancer patients based on the ratio of the number of positive to total axillary LNs examined and is demonstrated to be the most powerful LN-related prognostic factor. It outperforms the other LN-related methods, such as the current LN staging system recognized by AJCC and the LN CSCA, in predicting breast cancer survival in our study. However, LN CSCA is still found to be a promising survival predicator in our pilot study. More extensive analysis in a large-scale study is needed to further define its role in breast cancer management.