• breast carcinoma;
  • multifocal;
  • multicentric;
  • axillary lymph node status;
  • staging


  1. Top of page
  2. Abstract


For unifocal invasive breast carcinoma, increasing tumor diameter predictably correlates with a greater frequency of lymph node involvement, thereby facilitating treatment decisions. In invasive breast tumors presenting with multiple nodules, however, it is unclear whether tumor size correlates with lymph node dissemination in a similar manner.


The authors analyzed a series of 101 invasive breast carcinomas presenting with multiple macroscopically apparent lesions (2 foci: n = 77; 3: n = 20; 4: n = 4). Two different assessments of the tumor size (diameter of largest focus and combined diameter of all the foci) were then correlated with the status of axillary lymph nodes. For comparison with unifocal tumors, the authors used both external and internal control series (the latter consisting of 469 patients from their institution). The associations between lymph node status, tumor size, and multifocality were modeled using univariate and multivariate logistic regression, for each modality of tumor size assessment.


The logistic curves for multifocal and unifocal tumors were significantly different when the largest diameter was used as a tumor size estimate. Multifocal cases had higher frequencies of lymph node involvement than unifocal lesions of similar size category. In a multivariate logistic regression, the odds ratio of positive lymph node status in multifocal versus unifocal cases was 2.8 using largest diameter as a tumor size estimate (P < 0.0001). When the combined diameter assessment was used, however, the regression curve of multifocal cases was similar to that of unifocal cases, and the frequency of lymph node positivity was not significantly different in multifocal versus unifocal cases of the same size (odds ratio, 1.4; P = 0.13).


The authors' results show that, if aggregate diameters are used, unifocal and multifocal breast carcinomas are similar with respect to frequency of regional lymph node metastasis. Currently used algorithms, which use the diameter of the largest nodule, result in understaging of multifocal breast carcinomas due to underestimation of actual tumor size. Cancer 2002;94:1383–90. © 2002 American Cancer Society.

DOI 10.1002/cncr.10331

Tumor size exhibits a strong, direct correlation with axillary lymph node metastasis and an inverse correlation with survival in patients with breast carcinoma,1–8 in keeping with a consistent biologic relation between tumor burden and angiolymphatic dissemination. In cases of unifocal breast carcinoma, the largest tumor diameter is used as an approximation (or surrogate) of the tumor volume for staging purposes. For the staging of multifocal/multicentric breast carcinomas, the American Joint Committee on Cancer and the International Union Against Cancer (AJCC/UICC) recommend using the diameter of the largest tumor only.9 Thus, for the purpose of obtaining a simple and consistent measurement, the actual tumor burden is underestimated because secondary tumor foci, which are often sizable, are not included.

Multifocal/multicentric tumors are defined by the presence of two or more physically separate carcinomas in the same breast. In the currently accepted terminology, multifocality represents the presence of multiple foci of the same tumor, whereas multicentricity suggests different primary tumors in the same breast. For practical purposes, the distinction between multifocal and multicentric is based on topographic and histologic criteria with emphasis on the former (tumors with nodules located in different quadrants or in the same quadrant are considered to be multicentric or multifocal, respectively). Some authors have reported that multicentricity/multifocality adversely influences patient outcome10 and increases the incidence of lymph node positivity compared with unifocal tumors.11–15 All these previous studies have analyzed multifocal tumors using the diameter of the largest nodule only, as a tumor size estimate and did not attempt to describe the relation between aggregate estimates of tumor volume and prognosis/metastasis.

The current AJCC/UICC T classification criteria are based on the assumption that the behavior of multifocal/multicentric tumors is determined only by the size of the largest tumor. We hypothesize that the propensity of multifocal/multicentric tumors for metastasis is best described as a function of aggregate tumor size. This prompted us to explore the relation between tumor size and lymph node involvement in multifocal/multicentric tumors by using aggregate tumor size estimates. The findings were used to investigate whether the current staging criteria optimally reflect the metastatic behavior of multifocal breast carcinoma.


  1. Top of page
  2. Abstract

Patient Selection

From the files of Harper University Hospital, Detroit, Michigan, covering the period 1984–2000, we identified 101 patients having multifocal/multicentric invasive tumors. The criteria for inclusion in the study were identification of multiple, clearly separated, and macroscopically measurable tumor nodules in the same breast. We included in our study both multicentric (nodules in different breast quadrants) and multifocal tumors (nodules in the same quadrant). Only cases with invasive carcinoma were included in the study; cases having in situ tumors alone were not considered. The patients were treated with either modified radical mastectomy or lumpectomy with axillary lymph node dissection, and the pathologic lymph node status was available in all cases. We excluded all patients that had multifocal diffuse microscopic involvement of the breast by invasive tumor (i.e., without macroscopically identifiable nodules) or cases in which size could not be assessed accurately. Cases that were unifocal on macroscopic examination, but microscopically proved to have extensive minute tumor foci throughout the breast, also were excluded.

Data Collection

For each case, we recorded the diameter and location of all tumor nodules and also the number and status of axillary lymph nodes. The histologic type and the grade of all tumor nodules from each case were assessed by reviewing the archival hematoxylin and eosin–stained sections. For grading, we used the Scarff–Bloom–Richardson system (SBR),16–18 with the Nottingham modification.19, 20 The presence and extent of vascular invasion and ductal carcinoma in situ (DCIS) also were recorded. For each case, we recorded two different tumor size estimates: 1) the diameter of the largest tumor nodule (the estimate currently used for T classification of multifocal tumors); and 2) the sum of the diameters of all tumor nodules present in the breast.

Comparison with Unifocal Tumors

We compared the multifocal cases with a control series of 469 consecutive unifocal invasive tumors with known lymph node status, also obtained from our files between 1992 and 1999. To validate that our sample of unifocal tumors was representative (i.e., with respect to correlation between lymph node status and tumor diameter), we compared it with a published large series.1 For analysis, we used a multivariate linear regression model with the dependent variable being the frequency of lymph node positivity for each tumor size interval and the independent variables the tumor size intervals and the category of the study (either the published series or our internal control). We used the P value for the regression coefficient of “the category of the study” variable to assess any statistically significant differences between the two studies (cutoff for P = 0.05).

Data Analysis and Statistical Methods

Individual t-statistics and P values were obtained to assess the differences in mean tumor size between the unifocal series and each of the two tumor size estimates for the multifocal series. To determine statistical significance, we used a conservative Bonferroni correction to ensure that the combined type 1 error probability when performing multiple tests was 0.05.

Separate chi-square tests were performed to compare the frequency of axillary lymph node metastases in multifocal tumors. In the multifocal series, for each of the two diameter estimates, separate logistic fits21 were used to model the odds of axillary lymph node positivity as a function of the diameter. A similar fit was used to model the relation between lymph node status and tumor diameter in the unifocal series as well. The logit of lymph node status was seen to be a concave function of tumor diameter by visual inspection. Hence, we used the logarithm of the estimated diameter as our regressor in the fitted models.

To assess the influence of multicentricity on lymph node status, while controlling for tumor size (diameter), we used a multivariate logistic regression model with the logarithm of tumor diameter and multicentricity as the independent variables and the logit of lymph node status as our dependent variable. Again, for multicentric cases, we used two different diameter estimates.


  1. Top of page
  2. Abstract

Of the 101 cases in our multifocal/multicentric series, 77 (76.2%) had 2 tumor nodules, 20 (19.8%) had 3 nodules, and 4 (3.9%) had 4 nodules. The location of index (i.e., largest) nodules and additional tumor nodules were available in 79 cases (78.2%) and 81 cases (80.1%), respectively (Fig. 1). In 48 cases (47.5%), the tumor nodules were situated in different quadrants (i.e., multicentric tumors), in 30 cases (29.7%) nodules involved the same quadrant (i.e., multifocal tumors), and for 23 cases (22.8%) the location of tumor nodules was not specified.

thumbnail image

Figure 1. Location of the largest (index) nodule in multifocal tumors (A). Location of additional tumor nodules (B). uo: upper outer; lo: lower outer; li: lower inner; ui: upper inner; uc: upper central; lc: lower central; co: outer central; ci: inner central; c: central. numbers in each sector represent the percentage of cases with tumor in that sector. asterisk indicates that for 79 of 101 cases location of index tumor was available. double asterisk indicates that for 81 of 101 cases location of additional nodules was available.

Download figure to PowerPoint

Clinicopathologic characteristics of the multifocal cases are compared with our unifocal series in Table 1. Overall, there was a significant difference in axillary lymph node status between multifocal and unifocal cases (69.3% vs. 54.5%; P = 0.009). Mean age was not significantly different between the two groups. Distribution of histologic tumor types of the largest nodule in multifocal cases was similar to the histologic distribution among unifocal cases. The only statistically significant difference was the relative prevalence of mixed ductal and lobular carcinomas in multifocal versus unifocal cases (6.9% vs. 1.3%; P = 0.0005). Using a modified SBR grading system, we classified 8 multifocal cases (9.3%) as Grade 1, 43 (50%) as Grade 2, and 35 (40.7%) as Grade 3.

Table 1. Clinical and Histologic Characteristics of Patients with Multifocal and Unifocal Invasive Breast Carcinomas
CharacteristicMultifocal tumors (n = 101) (%)Unifocal tumors (n = 469) (%)P value
  • a

    A student t test was used for comparison.

  • b

    For multifocal cases, this represents the histology of index lesions.

  • c

    For unifocal series, histologic type was available in 466 cases.

  • d

    A chi-square test was used for comparison.

  • e

    Eighty-six multifocal cases were graded using a modified Scarff–Bloom–Richardson grading system. The grade was available for 274 unifocal tumors. Because grading was not performed uniformly for unifocal series, a comparison with multifocal cases was not possible.

Age (yrs)
 Invasive ductal85 (84.1)418 (89.6)0.16d
 Lobular3 (2.9)27 (5.7)0.25d
 Other types6 (5.9)15 (3.2)0.18d
  Medullary1 (0.9)4 (0.85)
  Mucinous1 (0.9)7 (1.5)
  Metaplastic2 (1.9)1 (0.2)
  Papillary1 (0.9)3 (0.64)
  Tubular1 (0.9)0 (0)
 Mixed ductal and lobular7 (6.9)6 (1.3)0.0005d
 I8 (9.3)
 II43 (50)
 III35 (40.7)
Axillary lymph node positivity70 (69.3)256 (54.5)0.009d

Distribution of histologic tumor types and grades in the additional foci of multifocal carcinomas is presented in Table 2. In 16 multifocal carcinomas (15.8%), the histologic type of primary and secondary tumor nodules were dissimilar, and in 5 cases (4.9%), the grade of tumor foci differed (in 4 cases the secondary foci were lower grade and in 1 case it was higher). Angiolymphatic invasion was identified in 28 cases (27.7%) and was unequivocal and extensive in 5 cases (4.9%). Foci of DCIS were identified in 68 cases (67.3%) and in 18 cases (17.8%) constituted greater than 30% of tumor volume.

Table 2. Histologic Types and Grades of Secondary Foci in Multicentric/Multifocal Invasive Breast Carcinomas
Histology and gradeMultifocal cases, additional nodules (n = 101) (%)
  • a

    Grading of additional foci was recorded in 85 cases of multifocal tumors.

 Ductal82 (81.1)
 Lobular10 (9.9)
 Other8 (7.9)
  Medullary2 (1.9)
  Mucinous2 (1.9)
  Metaplastic1 (0.9)
  Papillary1 (0.9)
  Tubular2 (1.9)
  Mixed ductal and lobular7 (6.9)
 I11 (12.9)
 II38 (44.7)
 III36 (42.4)

The comparison between our unifocal control group and the external control series1 is presented in Figure 2. There is no significant difference between the two groups in terms of the effect of tumor diameter on the proportion of axillary lymph node positivity as assessed by linear fits (P = 0.14), or a comparison within T classification groups (P = 0.2 for T1, P = 0.1 for T2, and P = 0.5 for T3). These data demonstrate that our sample of unifocal breast tumors is representative of the general population of unifocal breast carcinomas and thus constitutes a valid group for comparison with our multifocal/multicentric cases.

thumbnail image

Figure 2. Comparison between the current series of invasive unifocal tumors (469 cases) and a large published study (24,740 cases).1 (A) correlation between tumor diameter and lymph node status. a multivariate linear regression analysis showed no significant differences between our control series and the large study when controlling for tumor diameter (P = 0.14). (B) correlation between T classification groups and lymph node status. a chi-square test showed no significant differences between the two series for each t classification group.

Download figure to PowerPoint

The two different methods for estimating tumor size in multifocal/multicentric carcinomas (i.e., diameter of the largest nodule and combined diameters) resulted in statistically significant differences in both size and T classification distribution, when compared with unifocal cases. These data are summarized in Table 3. Using the diameter of the largest nodule as a size estimate produced a lower mean tumor size estimate for multifocal than for unifocal cases (2.53 vs. 3.47 cm, respectively; P = 0.026). Conversely, combining the diameters of multifocal/multicentric tumors resulted in a larger mean tumor size compared with unifocal lesions (4.2 vs. 3.47 cm, respectively; P = 0.003). Consequently, the multifocal tumors had a different distribution within T classifications depending on the method of tumor size estimation (Table 3).

Table 3. Mean Tumor Size Estimates and Tumor Classification Distribution in Multifocal versus Unifocal Invasive Tumors
Size and classificationMultifocal (n = 101)Unifocal (n = 469)P value
  • LD: diameter of the largest tumor nodule; CD: combined diameters of all nodules.

  • a

    Analysis of variance test was used for comparison.

  • b

    A chi-square test was used for comparison.

Mean tumor size estimate (cm)2.534.23.47< 0.0001a
T1 (%)49.5%17.8%34.5%< 0.0001b
T2 (%)45.5%53.5%47.6%
T3 (%)5%28.7%18%

When the currently recommended measurement was used for staging (i.e., diameter of the largest nodule), multifocal cases had a higher incidence of T1 tumors and a lower incidence of T2 and T3 tumors compared with unifocal cases (49.5%, 45.5%, and 5%, respectively, vs. 33.9%, 48%, and 18.1%; P < 0.0001). However, when using combined diameters as a tumor size estimate, 31.7% of the multifocal cases, which would be classified as T1 by using the diameter of largest nodule, were moved into higher T classifications.

The frequencies of positive axillary lymph nodes in each T classification group for multifocal cases and unifocal control cases are compared in Table 4. The currently used measurement for staging multifocal carcinomas (diameter of the largest nodule) resulted in a significantly higher incidence of positive lymph node status than unifocal control series within T1 and T2 classification groups. However, using the combined diameters as a size estimate resulted in frequencies of positive lymph nodes not significantly different from the unifocal control series within similar T classification groups. The most prominent change was in T1 classification, where the incidence of lymph node positivity for our multifocal and unifocal series became equal.

Table 4. Lymph Node Status and Tumor Classification in the Multifocal Series for the Two Tumor Size Estimates and in the Unifocal Control Series
Pos/Tot (%)Pos/Tot (%)P valueaPos/Tot (%)P valuea
  • LD: diameter of the largest nodule; CD: combined diameters, Pos/Tot: positive/total.

  • a

    A comparison was performed between multifocal and unifocal series using a chi-square test.

  • b

    Fisher exact test was performed because 1 cell had 0% frequency.

T156/159 (35.2)30/50 (60)0.0026/18 (33)0.8
T2134/225 (59.6)35/46 (76.1)0.0339/54 (72.2)0.08
T366/85 (77.6)5/5 (100)0.3b25/29 (86.2)0.3

A continuous model describing the relation between lymph node status and tumor diameter using logistic regression is presented in Figure 3. When largest diameter was used as a tumors size estimate, the curve for multifocal tumors showed a higher incidence of lymph node positivity than the unifocal tumors for any given tumor size. However, when utilizing the combined diameter, the curve for multifocal series was shifted downward becoming close to the curve for unifocal series and even superimposed on it for tumor size estimates < 2 cm. To verify whether statistically significant differences exist between multifocal and unifocal curves, we analyzed a multivariate logistic regression model assessing the impact of multifocality versus unifocality on lymph node status when controlling for tumor size. The results are presented in Table 5 and show that multifocality did not significantly influence lymph node status for the same tumors size when a combined diameter is used as a tumor size estimate. The odds ratio of having positive lymph nodes in multifocal versus unifocal tumors of the same size was 1.4 and was not statistically significant (P = 0.13). However, when size in multifocal tumors was reported as the largest diameter, the odds ratios of having positive lymph nodes in multifocal versus unifocal tumors were significantly higher (2.8; P = 0.0001).

thumbnail image

Figure 3. Correlation between lymph node status and tumor size in unifocal control series and multifocal series using two tumor size estimates. CD: combined diameters of all nodules; LD: largest tumor nodule diameter.

Download figure to PowerPoint

Table 5. Multivariate Analysis of Lymph Node Status Correlation with Tumor Size and Multifocality
AnalysisTumor size estimates in multifocal carcinomas
Combined diameterLargest diameter
  • CI: confidence interval; GOF: goodness of fit.

  • a

    Odds ratios of lymph node positivity in multifocal versus unifocal cases controlling for tumor size.

  • b

    Hosmer and Lemeshow goodness-of-fit test was used.

Odds ratioa1.42.8
95% CI of odds ratio0.9–2.31.7–4.6
P value for odds ratio0.130.0001
Overall GOFb0.30.6


  1. Top of page
  2. Abstract

Multicentricity/multifocality is not an uncommon feature of breast carcinoma. The reported incidence of such lesions ranges from 13% to 70%, depending mainly on the sampling method.10, 11, 13, 22–34 With the increasing utilization of breast conserving therapy and more accurate diagnostic imaging, multicentricity/multifocality is emerging as a practical issue in patient management. Detection and reliable staging of these lesions by pathologists thus is clinically important.

The biology of multifocal/multicentric breast carcinoma remains unsettled. From the clinician's perspective, the distinction between multifocal and multicentric tumors is settled on topographic criteria. However, molecular studies have indicated that multiple tumors present in the same breast are generally a manifestation of intramammary spread of one tumor (i.e., via the ductal system); that is, they are multifocal.35, 36 Occasionally, multiple synchronous primary tumors may occur (multicentric tumors).26, 27, 37, 38 The metastatic behavior of these subsets would, theoretically, be different. In the multifocal group, lymph node metastases likely would correlate with invasive tumor burden in a manner analogous to unicentric lesions. In the multicentric lesions, however, lymph node spread would reflect the size of the more aggressive primary lesion. Without molecular level tests, it is not possible to make the distinction in a reliable manner. We did not attempt to differentiate between multiple primary tumors and discontinuous intramammary spread, because it would not be practical to do so under ordinary circumstances, and moreover the staging schemes do not separate them. Therefore, we grouped all cases together under the designation multicentric/multifocal tumors. Because most molecular level studies identify a common progenitor for most clinically multifocal/multicentric lesions, we believe that basing T classification categorization on estimation of tumor bulk has clinical relevance.

In agreement with previous studies,11–15 our data indicate that multifocal/multicentric carcinomas as a group have a significantly higher incidence of positive lymph nodes than unifocal tumors. This propensity is not likely to be the result of a different histology, because we found that the distribution of histologic types is similar between multifocal and unifocal cases (Table 1). Also, the distribution of SBR tumor grades in our multifocal series (Grade 1: 9.3%; Grade 2: 55%; Grade 3: 34.7%) is within the limits reported in the literature for unifocal carcinomas: Grade 1: 5.1–33%; Grade 2: 31–57%; Grade 3: 20–63.6%.39–44 For angiolymphatic invasion, we found a frequency of 27.7% in our series that was again within the limits reported by other researchers in unifocal carcinomas, which varies between 21% and 56%.43, 45–49 The only significant difference between multifocal and unifocal tumors was the larger mean aggregate tumor diameter in the former (4.2 vs. 3.47 cm, respectively; P = 0.026). Similar results were reported by Fish et al.12 in a study including 37 patients with clinically multifocal tumors and 557 patients with unifocal tumors. These authors found that the mean size of unifocal tumors was smaller than multifocal cases when an aggregate measurement of foci was utilized (2.62 vs. 3.81 cm, respectively; P < 0.001). They also found differences in the number of patients who would be classified as having tumors less than 1 cm or between 1 and 2 cm depending on the method of tumor size calculation (largest diameter vs. combined diameters).

Note that the relation between lymph node status and tumor size in both multifocal and unifocal tumors is not linear but rather asymptotic. For small tumor sizes, there is a steep increase in lymph node positivity with relatively small increases in tumor size but as tumor size increases, reaching the asymptotic part of the curves, changes in size do not significantly affect lymph node status. This observation stresses the importance of precision when measuring the diameter of small tumors, as small errors in measurement will lead to fairly large errors in estimation of prognosis.

Our data suggest that the current TNM classification criteria (AJCC/UICC), which uses the diameter of the largest nodule for T classification of multifocal carcinomas, underestimates the total tumor volume and tends to understage these tumors. Brenin and Morrow have reached a similar conclusion in a study of 72 cases with additional tumor foci found on reexcision.15 They found that in a multivariate analysis, the group with positive tumor foci identified at reexcision (i.e., multifocal cases) had a higher risk of lymph node metastases than the cases with negative margins or no tumor at reexcision (odds ratio, 3.13; P = 0.0011) and concluded that current AJCC guidelines may result in understaging and undertreatment of these cases. In our study, when utilizing the standard tumor size estimate, multifocal/multicentric tumors have a significantly higher incidence of axillary lymph node metastases than unifocal tumors of similar size (odds ratio, 2.8; P < 0.0001). However, if we use combined diameters of all tumor nodules, the metastatic behavior of multifocal carcinomas is not significantly different from that of unifocal tumors (odds ratio, 1.4; P = 0.13).

One potential criticism of this study is that it uses tumor diameters when, more likely, the propensity for metastases is a function of tumor volume or surface area. As a consequence, even though the diameter of the largest focus underestimates total tumor volume, summing the diameter of all foci will overestimate actual tumor volume, because the volume is proportional to the third power of the diameter. However, the use of tumor diameter as a size estimate has a practical relevance because it is more readily measured, it is already used by the current staging system and also shows a good correlation with axillary lymph node status.

The conclusion that the propensity of multifocal tumors to metastasize is a reflection of total tumor volume is in contradiction to the assumption used by the current staging system (i.e., that the behavior of multifocal tumors depends only on the size of the largest nodule). We suggest that an alternative measurement scheme that takes into account additional nodules (such as combined diameters) represents a better predictor of the rate of axillary metastases. Alternatively, if the standard measurement (i.e., diameter of largest nodule) is used, multifocal tumors should be designated in a separate T category, to convey the increased risk for metastatic dissemination.


  1. Top of page
  2. Abstract