21 ORIGINAL ARTICLE Basal phenotype: a powerful prognostic factor in small screen-detected invasive breast cancer with long-term follow-up A J Evans, E A Rakha, S E Pinder, A R Green, C Paish and I O Ellis... J Med Screen 27;14:21 214 See end of article for authors affiliations... Correspondence to: A J Evans, Breast Institute, Nottingham City Hospital, Hucknall Road, Nottingham NG5 1PB, UK; andrew.evans@nuh.nhs.uk Accepted for publication 1 March 27... Objective To assess the frequency and prognostic significance of a basal phenotype in a group of women with screen-detected invasive breast cancers with long-term follow-up and to focus particularly on women with small (o15 mm) breast cancers. Methods The study group was derived by finding women common to a consecutive series of 1944 invasive breast cancers diagnosed in Nottingham between 1986 and 1998 with a known basal phenotype status and a prospectively collected database of all screen-detected breast cancers. In total, 356 women constituted the study group. Pathological and radiological features were recorded. Basal cell markers used were CK5/6 (cloned5/16134) and CK14 (clone LL2). Tumours were classified as of basal phenotype if > or ¼ 1% staining was seen with either marker. Results Of all screen-detected lesions, 43 (12%) had a basal immunophenotype and 313 (88%) were non-basal. There were 15 (35%) and 4 (13%) breast cancer deaths in the basal group and nonbasal groups, respectively (P ¼.6). On univariate analysis, nodal stage, histological grade, lympho-vascular invasion (LVI) status, invasive size and basal phenotype had prognostic significance. On multivariate analysis, basal phenotype, LVI and nodal stage maintain prognostic significance. Of the 189 women with o15 mm lesions, eight of 2 (4%) of the basal group and eight of 169 (5%) of the non-basal group died of breast cancer (Po.1). On multivariate analysis, basal phenotype was the only factor to maintain independent prognostic significance. Conclusions Basal phenotype is a powerful prognostic factor for women with small screen-detected invasive breast cancer. INTRODUCTION The prognosis of small (o15 mm) invasive breast cancers is generally good. 1,2 A significant proportion of women with such tumours do, however, eventually die of breast cancer. Identifying those women in this group who are at a higher risk of breast cancer death would enable powerful adjuvant systemic therapies to be focused on this subgroup, while women with a very low risk of breast cancer deaths might be spared the morbidity of such adjuvant systemic therapy. The traditional three main breast cancer prognostic factors, namely size, lymph node stage 3,4 and histological grade, 5 are of less value in predicting the behaviour of small screen-detected lesions than in larger tumours. 6 Grade ceases to have prognostic importance at small sizes, 2 while lymph node positivity is present in less than half of the women in this group who die of breast cancer. 2,6 Further sub-division of the invasive tumour size of cancers o15 mm in extent also fails to give robust prognostic information. 2 However in recent years, lympho-vascular invasion (LVI) has been shown to add prognostic information, especially in women who are node negative. 7,8 In an attempt to improve prognostication in women with small screen-detected breast cancers, other novel approaches have been pursued. A number of studies have looked at whether mammographic appearance confers prognostic information. Some have suggested that the presence of mammographic comedo calcification indicates a poor prognosis, but others have not been able to confirm this finding. 1,2,9 11 A recent study has also claimed that mammographic spiculation is an independent good prognostic factor in this group. 12 Normal breast ducts consist of two layers, an inner layer of epithelial cells, the majority of which express luminal epithelial cytokeratins (e.g. cytokeratins 18 and 19), with lesser numbers of basal/myoepithelial cells with a basal immunophenotype (expressing basal cytokeratins such as 5, 6 and 14) present in normal ducts. These epithelial cells overlie the myoepithelial cell layer in normal breast parenchyma. Using immunohistochemistry, a number of groups have shown that between 2% and 18% 13 of breast cancers express basal markers and that these tumours have a poor prognosis. 14 16 It has been shown that the basal phenotype in invasive breast carcinomas is associated with larger size, higher histological grade, comedo necrosis and with nodal and oestrogen receptor negativity. However, a basal phenotype has also been shown to have independent prognostic significance on multivariate analysis. 14 The aim of this study was to assess the frequency and prognostic significance of the basal immunophenotype in a group of women with screen-detected invasive breast cancers with long-term follow-up and to focus particularly on the significance of this tumour phenotype in women with small (o15 mm) breast cancers.
Basal phenotype and screen-detected breast cancer 211 METHODS This study group was derived by combining two data-sets. The first is a consecutive series of 1944 invasive operable breast cancers diagnosed in Nottingham in women under the age of 7 years, between 1986 and 1998 and entered into the Nottingham Tenovus Primary Breast Carcinoma Series. A previous study had ascertained the basal phenotype of this group of patients. 14,15 This data-set was combined with a separate prospectively collected database that included the pathological and radiological findings of all screen-detected invasive breast cancers diagnosed in Nottingham since the inception of mammographic breast screening in 1987. (Screening during the study period was of women aged 5 64 years, every three years using two views at first screen and a single medio-lateral oblique view at subsequent screens.) The 356 patients common to both databases constituted the study group used in the present analyses. Pathological features recorded were tumour type, histological grade, nodal status, microscopic invasive size and LVI status (probable LVI was categorized as absent/negative, as is our usual clinical practice). The basal cell markers were determined by immunohistochemistry using antibodies targeted to CK5/6 (cloned5/16134) or CK14 (clone LL2) as previously described. 14 Tumours were classified as being of basal phenotype where 1% or more of tumour cells showed reactivity for CK5/6 and/or CK14. The proportion of patients in each group, which were triple negative (ER, PgR and HER2 negative), was recorded. Radiological features assessed in this study were the presence of mammographic comedo calcification (malignant calcification with any elongated linear or branching forms) and spiculation (the presence of either a spiculated mass or architectural distortion). Patient survival was ascertained from hospital records. When follow-up status or cause of death was not clear, the local cancer registry or the patient s family doctor was contacted for clarification. Statistical methods Analyses of two groups of patients were performed using StatView V. 5..1 on an Apple Macintosh computer. One included all women with screen-detected invasive cancer and the second analysis included only those women whose cancers were o15 mm in size. The histological correlates of basal- and non-basal-type screen-detected cancers were ascertained. The prognostic value of the pathological, immunohistochemical and radiological features were analysed in a univariate fashion with w 2 tests and using Kaplan Meier survival curves with log-rank tests. Multivariate analysis was performed, including all those features that were of significance in initial univariate analysis. The effect of the Nottingham Prognostic Index (NPI) as a prognostic factor for small basal tumours, and the prognostic effect of basal phenotype on small tumours within NPI prognostic groups was evaluated. RESULTS Analysis of all patients The main study group constituted 356 screen-detected cancers (194 prevalent round cancers and 162 incident round cancers). Forty-three (12%) were of basal phenotype and 313 (88%) were non-basal. Those women with basaltype tumours were significantly younger than those with non-basal carcinomas (152 of 313 (49%) 6 years or less versus 23% aged over 6 years, (1 of 43), P ¼.6). The histopathological findings in the basal and non-basal groups are shown in Table 1. Basal tumours were more likely to be high grade than non-basal tumours (P ¼.5). Basal and non-basal tumours had, however, similar lymph node status, tumour size distributions and LVI status. Basal tumours were more likely to be triple negative than nonbasal tumours (Po.1). Mean follow-up in those alive was 12 years in the basal group and 11 years in the non-basal group. In those patients with basal-type carcinomas, there were 15 (35%) breast cancer deaths and one (2%) woman died from other causes. In the non-basal group, there were 4 (13%) breast cancer deaths and 25 (8%) non-breast cancer deaths. The difference in breast cancer deaths was statistically significant (Po.1) between the two groups. Breast cancer-specific survival curves for all patients with basal- and non-basaltype tumours are shown in Figure 1. In univariate analysis, lymph node stage (Po.1), histological grade (Po.1), LVI status (Po.1), invasive tumour size (P ¼.3) and basal phenotype (Po.1) all had prognostic significance as predictors of breast cancer death. Mammographic comedo Table1 Pathological and survival status of 356 screen-detected invasive breast cancers according to basal phenotype Basal, n (%) Non-basal, n (%) Statistical significance Total 43 313 Breast cancer death 15 (35) 4 (13) P ¼.6 Alive 27 (63) 248 (79) Dead from other cause 1 (2) 25 (8) Grade1 1 (23) 111 (36) P ¼.5 Grade 2 12 (27) 122 (39) Grade 3 21 (49) 79 (25) Node not assessable 2 (6) P ¼.38 Node negative (>3 nodes retrieved) 3 (7) 217 (7) One to three nodes positive 1 (23) 61 (2) Four or more nodes positive 3 (7) 14 (4) Definite vascular invasion 5 (12) 57 (19) P ¼.28 o1 mm 7 (16) 75 (24) P ¼.73 1 15 mm 17 (4) 118 (38) 16 2 mm 11 (26) 66 (21) >2 mm 8 (19) 53 (17) Triple negative 16 (37) 14 (5) Po.1 www.jmedscreen.com Journal of Medical Screening 27 Volume 14 Number 4
212 Evans et al. calcification and mammographic spiculation were not significantly associated with outcome in this series. In multivariate analysis (Table 2), LVI status, lymph node stage and basal phenotype maintained prognostic significance; histological grade was of borderline significance when basal phenotype was also included. Analysis of small cancers In total, 189 screen-detected cancers were of invasive size of o15 mm. Of these, 2 (11%) were of basal type and 169 (89%) women had tumours of non-basal subtype..2 Chi-Square DF P-Value 5.121 1.236 () Event Times () (1) Event Times (1) 2.5 5 7.5 1 12.5 15 17.5 2 Time (Years) Figure 1 Breast cancer-specific survival for patients with invasive breast carcinomas, by basal phenotype The histopathological findings of the basal and non-basal tumours o15 mm in size are shown in Table 3. Basal tumours o15 mm in size were more likely to be high grade than non-basal tumours o15 mm in size (P ¼.7). There was, however, no difference in nodal status, size distribution or LVI status between the groups of small basal and small non-basal tumours. Small basal tumours were more likely to be triple negative than small non-basal tumours (Po,1). There were eight (4%) breast cancer deaths in the patients with basal-type carcinomas o15 mm in size as compared with eight (5%) breast cancer deaths in those with small non-basal-type tumours. This difference in breast cancer deaths between these two groups was statistically significant (Po.1). Breast cancer-specific survival curves for patients with small basal and non-basal tumours are shown in Figure 2. In univariate analysis, nodal stage (Po.1), LVI status (Po.1) and basal phenotype (P ¼.1) had prognostic significance as predictors of breast cancer death in women with small carcinomas. Histological grade, invasive size, mammographic comedo calcification and mammographic spiculation were not significantly associated with outcome in this group of small tumours. The NPI did not significantly predict survival of small basal tumours. Basal phenotype did significantly predict survival of small tumours within the NPI moderate prognostic group (Po.1) (Figure 3) but not the poor prognostic group. In multivariate analysis (Table 4), basal phenotype was the only factor to maintain prognostic significance in this group of patients with invasive breast carcinomas o15 mm in size. DISCUSSION Most women with small screen-detected breast cancers are unlikely to die as a result of their breast cancer. 1,2 Decisions concerning the benefit of adjuvant systemic therapy are Table 2 Multivariate analysis of prognostic factors for all screen-detected invasive breast carcinomas Degrees of freedom Coef Standard error Coef/SE w 2 P value Exp (coef) Grade 1.32.166 1.89 3.6.58 1.37 Nodal stage 1 62.183 3.6 13.1.3 1.94 Basal phenotype 1 8.34 2.242 5.27.25.56 Invasive size 1.12.12 1 2.31 1 LVI 1 1.27 3. 9.26.2 4 Coef, coefficient; SE, standard error; LVI, lympho-vascular invasion Table 3 Pathological and survival status of 189 screen-detected invasive breast cancers o15 mm in size according to basal phenotype Basal, n (%) Non-basal, n (%) Statistical significance Total 2 169 Breast cancer death 8 (4) 8 (5) Po.1 Alive 11 (55) 149 (88) Dead from other cause 1 (5) 12 (7) Grade1 6 (3) 79 (47) P ¼.7 Grade 2 6 (3) 68 (4) Grade 3 8 (4) 22 (13) Node not assessable 12 (7) P ¼.1 Node negative (>3 nodes retrieved) 16 (8) 13 (77) One to three nodes positive 2 (1) 24 (14) Four or more nodes positive 2 (1) 3 (2) Definite vascular invasion 3 (15) 22 (13) P ¼ o1 mm 7 (35) 75 (44) P ¼ 2 1 14 mm 13 (65) 94 (56) Triple negative 6 (33) 7 (5) Po.1
Basal phenotype and screen-detected breast cancer 213 therefore more difficult in this group than in women with large symptomatic tumours, where the benefits of such therapy are typically more clear cut. These decisions are also complicated by the number and cost of current adjuvant therapies available. Application of the traditional prognostic factors of size, nodal stage and histological grade are less useful in women with small tumours and these features enable identification of only a minority of those women who will eventually die of breast cancer. There is therefore a.2 Chi-Square DF P-Value 6.99 1.86 2.5 () Event Times () (1) Event Times (1) 5 7.5 1 12.5 15 17.5 2 Time (Years) Figure 2 Breast cancer-specific survival for patients with o15 mm invasive carcinomas, by basal phenotype.2. 5 Survival Functions NPI: MPG 1 OS 15 2 BASAL No Yes Figure 3 Survival of small invasive cancers in the moderate prognostic group (Nottingham Prognostic index) according to basal phenotype status 25 need for more accurate prognostication in women with small invasive breast cancer. The poor survival of women with tumours which express basal-type cytokeratins was first reported in 1987. 17 Subsequent studies have confirmed the early development of recurrences, metastasis and breast cancer death in these patients. 14,18 The reported correlates of tumours with a basal phenotype include large tumour size, high histological grade, negative expression of steroid hormone receptors, FHIT protein, MUC1 and luminal associated cytokeratins and positive expression of p53, epidermal growth factor receptor, BRCA1, P-cadherin and neuroendocrine markers. 14,15 Basal-type tumours have also been reported to have high rates of comedo-type necrosis and lymph node negativity. Some authors have also reported an association between expression of basal markers and HER2 negativity. 19 The correlation between histological comedo-type necrosis and basal phenotype is a possible explanation for studies which have shown an association between mammographic comedo calcification and poor outcome of small screendetected invasive cancer. 1 The current study did not however show an association between mammographic comedo calcification and outcome for either screen-detected cancers as a whole or small screen-detected cancers. This study found that an association between basal phenotype and high tumour grade was maintained in all women with screen-detected cancers but that no association between basal phenotype and large size and nodal negativity was found. The lack of association between invasive tumour size and basal phenotype in this group of screendetected carcinomas may be because large basal tumours present clinically as symptomatic cancers rather than being detected at mammographic screening. The failure of this study to show a correlation between basal phenotype and node negativity may be a function of the relatively small size of this study; there is only a modest association in previous studies. 14 It should be noted that while the strengths of the present study include the length of follow-up (1 11 years) and prospective data collection, these data include only a small number of o15 mm basal-type tumours in total and further confirmation of these findings is required. Nevertheless, we have found that a basal phenotype is a predictor of breast cancer death in women with screendetected carcinoma, and that a basal phenotype is a particularly powerful prognostic indicator in women with small o15 mm tumours. Multivariate analysis has confirmed the independence of this prognostic influence and, indeed, in small lesions a basal phenotype was the only factor to maintain independent prognostic significance in these small invasive breast cancers. In this series, we have used a simple definition of basal phenotype; tumours were included within this category if expression of either of two immunohistochemical markers (cytokeratin 5/6 or cytokeratin 14) showed positivity in 1% or more of tumour cells. We have previously found that using a 5% cut-off did not add to the prognostic potency of basal phenotype status but did reduce the Table 4 Multivariate analysis of prognostic factors for o15 mm invasive cancers Degrees of freedom Coef Standard error Coef/SE w 2 P value Exp (coef) Grade 1.214.267 1 5 2 1 Nodal stage 1.559.317 1.76 3.1.8 1.75 Basal phenotype 1 1.335 19 3.19 1.15.1.26 Invasive size 1.47.71 6 4.51 5 Coef, coefficient; SE, standard error www.jmedscreen.com Journal of Medical Screening 27 Volume 14 Number 4
214 Evans et al. number of tumours with a basal phenotype. Very few tumours show 1 9% staining, so 1% or more staining was taken to be an appropriate cut-off point. There is some ongoing debate regarding the inclusion criteria for the basal type of breast carcinoma, with some groups suggesting a more restrictive definition such as mandatory ER and HER2 negativity but with either basal markers or epidermal growth factor receptor positive immunohistochemistry. The benefits of a simple definition such as that applied in the present study include the availability of these basal cytokeratin markers in histopathology laboratories, the ease of application and the lower potential for problems with both quality assurance and reproducibility. In addition, if costs of assessment are kept as low as possible, this would aid the ease of undertaking routine basal phenotype status analysis on all invasive breast carcinomas. The size of the difference in breast cancer deaths between those women with small basal-type cancers and those with small non-basal-type lesions (4% and 5%) is large enough to influence adjuvant therapy decisions, if the findings of this study are confirmed by other groups. In the future, it might be possible to withhold adjuvant systemic therapy from a larger proportion of women with small non-basaltype screen-detected tumours while ensuring women with small basal-type cancers receive adjuvant systemic therapy, despite the absence of nodal involvement or LVI. In conclusion, we have found that a basal phenotype is a powerful independent prognostic factor in women with small screen-detected invasive breast cancers with longterm follow-up.... Authors affiliations A J Evans, Radiologist, Radiology, Breast Institute, Nottingham City Hospital, Nottingham, UK E A Rakha, Pathologist, Histopathology Department, Nottingham City Hospital, Nottingham, UK S E Pinder, Pathologist, Histopathology Department, Addenbrookes Hospital Cambridge, Cambridge, UK A R Green, Research Scientist, Division of Pathology, School of Molecular Medical Sciences, University of Nottingham, Nottingham, UK C Paish, Biomedical Scientist, Histopathology Department, Nottingham City Hospital, Nottingham, UK I O Ellis, Pathologist, Histopathology Department, Nottingham City Hospital, Nottingham, UK REFERENCES 1 Tabar L, Chen H-H, Duffy SW, et al. A novel method for prediction of longterm outcome of women with T1a, T1b, and 1 14 mm invasive cancers: a prospective study. Lancet 2;355:429 33 2 James JJ, Evans AJ, Pinder SE, McMillan RD, Wilson ARM, Ellis IO. Is the presence of mammographic comedo calcification really a prognostic factor for small screen detected invasive breast cancer? Clin Rad 23; 58:54 62 3 Carter GL, Allen C, Henson DE. Relation of tumour size, lymph node status and survival in 24,74 breast cancer cases. Cancer 1989;63:181 7 4 Todd JH, Dowle C, Williams MR, et al. Confirmation of a prognostic index in primary breast cancer. Cancer 1987;56:489 92 5 Elston CW, Ellis IO. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long term follow up. Histopathology 1991;19:43 1 6 Tabar L, Duffy SW, Vitak B, Chen H-H, Prevost TC. The natural history of breast cancer what have we learned from screening? Cancer 1999;86:449 62 7 Pinder SE, Ellis IO, Galea M, O Rourke S, Blamey RW, Elston CW. Pathological prognostic factors in breast cancer. III. Vascular invasion: relationship with recurrence and survival in a large study with long term follow-up. Histopathology 1994;24:41 7 8 Kato T, Kameoka S, Kimura T, Nishikawa T, Kobayashi M. Blood vessel invasion as a predictor of long term survival for Japanese women with breast cancer. Breast Cancer Res Treat 22;73:1 12 9 Tabar L, Tony Chen HH, Amy Yen MF, et al. Mammographic tumour features can predict long-term outcomes reliably in women with 1 14 mm invasive breast carcinoma. Cancer 24;11:1745 59 1 Peacock G, Given-Wilson RM, Duffy SW. Mammographic casting-type calcification associated with small screen-detected invasive breast cancer: is this a reliable prognostic indicator. Clin Rad 24;59:165 7 11 Thurfjell E, Thurfjell MG, Lindgren A. Mammographic findings as a predictor of survival in 1 9 mm invasive breast cancers. Worse prognosis for those cases presenting as calcification alone. Breast Cancer Res Treat 21;67:177 8 12 Evans A, Pinder S, James J, Ellis IO, Cornford E. Is mammographic spiculation an independent, good prognostic factor in screen detected invasive breast cancer? AJR 26;187:1 6 13 Nagle RB, Bocker W, Davis JR, et al. Characterisation of breast carcinomas by two monoclonal antibodies distinguishing myoepithelial from luminal epithelial cells. J Histochem Cytochem 1986;34:869 81 14 Rakha EA, Putti TC, Abd El-Rehim DM, et al. Morphological and immunophenotypic analysis of breast carcinomas with basal and myoepithelial differentiation. J Pathol 26;28:495 56 15 Rakha EA, Abd El-Rehim DM, Paish C, et al. Basal phenotype identifies a poor prognostic subgroup of breast cancer of clinical importance. Eur J Cancer 26;42:3149 56 16 Nielsen TO, Hsu FD, Jensen K, et al. Immunohistochemical and clinical characterisation of the basal like sub-type of invasive breast cancer. Clin Canc Res 24;1:5367 74 17 Dairkee SH, Mayall BH, Smith H, Havkett A. Monoclonal marker that predicts early recurrence in breast cancer. Lancet 1987;1:514 18 Rodriguez-Pinilla SM, Sarrio D, Honrado E, et al. Prognostic significance of basal-like phenotype and fascin expression in node-negative invasive breast carcinomas. Clin Cancer Res 26;12:1533 9 19 Jones C, Ford E, Gillett C, et al. Molecular cytogenetic identification of subgroups of grade III invasive ductal breast carcinomas with different clinical outcomes. Clin Cancer Res 24;1:5988 97