Prediction of Axillary Lymph Node Involvement of Women with Invasive Breast Carcinoma

Reviewers: John Han-Chih Chang, MD
Source: Cancer 1998; Volume 83 (Number 5): pages 948 - 955

Background

Axillary lymph nodes (ALN) are the most commonly involved siteof disease in breast cancer that has spread outside the primary lesion. Clinical examination is a relatively poor predictor of ALN status inbreast cancer. Because the prognostic implications of ALN involvementhave a bearing on treatment decisions, ALN dissections (ALND) have beena mainstay of work-up in breast malignancies. Chronic lymphedema (armswelling) from ALND is a significant sequelae that occurs in aproportion of patients getting this surgery. The goal of this studywas to determine subsets of patients in whom ALND could be avoided, iftheir risk of involvement was either too low or very high to warrantthis diagnostic (and some would argue "therapeutic")procedure.

Materials and Methods

This article details the attempt of this retrospective review tocreate a model of accurately predicting the ALN positivity. Women withinvasive breast cancer were identified from the Breast Cancer OutcomesDatabase (BCOD), which is the breast tumor registry of patientsreferred to the British Columbia Cancer Agency (BCCA) in Vancouver,British Columbia, Canada. Patients were gathered from BCOD who werediagnosed between January 1993 to December 1996. At least one lymphnode had to have been examined. There was exclusion of patients withT4 disease (involvement of skin, chest wall or inflammatory breastcancer) or N2 disease (fixed nodes) or metastatic disease. Alsoexcluded patients with prior chemotherapy (ChT) or radiation therapy(RT) before ALND. Four thousand six hundred and sixty patients meteligibility to be included in this review.

Univariate, bivariate and multivariate logistic regression analyseswere performed on all study patients. Clinical factors, such as age atdiagnosis and at menopause, menopausal status, family history ofunilateral, bilateral breast cancer, number of relatives, palpabilityof primary tumor and axillary lymph nodes, were evaluated. Thepathological factors included tumor grade, histological classification,size of primary tumor, estrogen receptor status, margin status,presence or absence of lymph vascular space involvement (LVI) orperi-neural invasion.

Results

Median age of the patients was 58 years (range of 22 =96 89). Median number of nodes dissected was 10 (range 1 =96 53). Univariateand bivariate analysis yielded the results seen on Table 1. Thepercentage of ALN positivity associated with each variable of all thecategories are listed. Table 2 has listed the percentage of patientshaving the a certain number of lymph nodes positive. Back to table 1,no significant associations were seen in family history, number ofrelatives with breast cancer, estrogen receptor status in relation tonodal involvement.

Multivariate logistic regression analysis was utilized to constructa model predicting nodal involvement in the entire sample. Initially,a six category model was utilized and is depicted in the Table 3. The"B" value represents difference between the risk of nodal involvement for that category and the average for all categories of that variable (a negative B means a lower than average risk of ALN metastases). The model, the authors contend, correctly predicted the ALN status in 76.63% of thecases. They felt that the model was too cumbersome to be of clinicalsignificance, since there are too many possible combinations for which to account. Thus, they developed a 3 category model utilizingpalpability of primary tumor and ALN=92s, tumor size and presence orabsence of LVI. They found this to be of comparable predictionaccuracy in ALN status (76.56%). Table 5 displays the data and therisk of ALN metastases based on their model. Based on this, theydivided the patients into 4 risk groups for ALN involvement: Very Low[< 10% risk], Low [10-14% risk], Intermediate [15-50% risk] and High[> 50%]. In table 5, very low risk patients hadnonpalpable disease of 5 mm or less without LVI [dashed line outline]. Low risk patients were those that had palpable disease of 6-10 mmor were nonpalpable without LVI [single line outline]. High riskpatients had clinically palpable ALN's or LVI in palpable tumors 2 cmor greater or nonpalpable tumors 3 cm or greater [double line outline]. Intermediate risk patients cover everyone in between.

Table 6 gives the data on the percentage of ALN positivity in eachrisk group. Because of the prognostic implication of greater than 3ALN's involvement for local control, this was also included in thedata presented in that table. Table 7 yields the data on the usage ofany systemic therapy (Tamoxifen/ChT/Both) according to risk groups andALN status. This article details a clinical model proposed by thisCanadian Group to hopefully guide us to an age of non pathologicalstaging for ALN.

Discussion / Critique

I feel that the biggest criticism of this article has to be thepredictive value of 75% in the model that was proposed. Mostclinicians would find that very unreliable [25% chance of error]. Iwould agree there is a subset of patients that have a very low risk ofALN mets that one may consider foregoing the full LND since the risk ofchronic lymphedema is not insignificant (5-15%). Since thesentinel lymph node biopsy is fairly minimal in its morbidity and hasan accuracy of 95% or greater, I truly see this model as having alimited role in predicting those that could forego any ALN procedure.At the other end of the spectrum, the authors claim that ALND areover-treatment for those who have locally advanced disease or acombination of bad prognostic factors, that they would get systemictherapy anyway. Why put them through a full ALND when it does notchange your management? To answer that question, I look to historicaldata that has demonstrated that greater than 3 ALN's involvedportends a worse local control and must be addressed adequately withRT. Also in the past, those that have greater than 9 ALN's involved would be considered for high dose ChT with a bone marrow transplant. That issue is a very controversial one, since most of the data coming to the forefront disputes the utility of bone marrow transplant foradvanced or metastatic breast cancer. In summary, the proposed modelis a nice way predict ALN involvement. However, due to its accuracy andthe advent of sentinel lymph node biopsies, it is not likely toeliminate surgical sampling of ALN's as part of breast cancerwork-up.