Prognostic significance of uPA/PAI-1 level, HER2 status, and traditional histologic factors for survival in node-negative breast cancer patients

One of the greatest challenges in the treatment of node-negative breast cancer is deciding in which patients the benefit from adjuvant cytotoxic chemotherapy would outweigh its adverse effects. Traditionally, this decision has been based on clinical and histomorphologic prognostic factors, such as patient age, tumor size, tumor grade, presence of lymphovascular invasion, and steroid hormone receptor status. Human epidermal growth factor receptor 2 (HER2) status, first used as a prognostic marker, became an important factor for predicting response to anti-HER2 therapy and is now a crucial part of this decision-making.

The serine protease urokinase-type plasminogen activator (uPA) and its inhibitor plasminogen activator inhibitor-1 (PAI-1) are markers of tumor invasion and metastasis that have reached the highest level of evidence for clinical utility as prognostic factors in breast cancer.^1,2 Node-negative patients with high values of uPA and/or PAI-1 have been shown to benefit from adjuvant chemotherapy in a prospective randomized multicenter therapy trial.^3,4 In addition, independent prognostic value of uPA/PAI-1 was confirmed in a pooled analysis of 8377 breast cancer patients that showed high levels of uPA and PAI-1 to be the strongest predictors of relapse-free survival and overall survival apart from lymph node status.⁵ In spite of this evidence, these biomarkers are still not widely used in the clinic.¹

Unfortunately, none of the large trials investigating the prognostic value of uPA and PAI-1 included HER2 status in the survival analysis. Only limited information is available on the relative prognostic impact of these factors when considered along with traditional prognostic markers in the same group of breast cancer patients. Therefore, it is still uncertain whether uPA and PAI-1 can give additional clinically relevant prognostic information after traditional prognostic factors and HER2 status have been taken into account.

To address this issue, we undertook the present study with the aim of comparing the prognostic impact of HER2 status, uPA, PAI-1, and traditional prognostic factors tumor size, grade, histological subtype, lymphovascular invasion, steroid hormone receptor status, and patient age on diseasefree, overall, and breast cancer specific survival in node-negative breast cancer patients.

Our retrospective analysis included all patients with lymph node-negative invasive breast cancer without distant metastases who underwent primary surgical treatment in Maribor University Clinical Center, Slovenia, in the ten-year period between January 1 2000 and December 31 2009. Exclusion criteria were neoadjuvant chemotherapy and presence of another active malignancy during breast cancer treatment. Considering the Helsinki Declaration principles the Slovenian National Medical Ethics Committee approved this study (Approval No. 55/11/13).

Clinical information on diagnosis, treatment, and follow-up was obtained from patient medical records. Survival data were completed with updated information from Slovenian Cancer Registry. Data on tumor size, histological subtype, grade, lymph node status, steroid hormone receptor status, and HER2 immunohistochemistry were obtained from original histology reports from the primary surgery. HER2 gene amplification, uPA and PAI-1 levels were obtained from our institution’s Medical Genetics Laboratory. Due to economical limitations, uPA and PAI-1 could not be assessed in all patients. Some other histological data were missing in a small fraction of patients due to inconsistent hospital guidelines on histology reports in the past.

All patients underwent either modified radical mastectomy or breast conserving surgery and radiotherapy. Adjuvant systemic treatment was given according to the guidelines followed at our institution at the time and was not influenced by uPA and/or PAI-1 values. Patients who completed primary treatment were followed-up at our institution at regular intervals.

HER2 status was determined immunohistochemically and additional fluorescent in situ hybridization (FISH) with PathVysion HER-2 DNA Probe Kit (Abbott Molecular, Abbott Park, IL, USA) was performed in samples with an immunohistochemical result of 2+. Since the patients were diagnosed before the publication of new ASCO/ CAP guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors⁶ and human epidermal growth factor receptor 2 testing in breast cancer⁷ and treated accordingly, steroid hormone receptor status and HER2 status were determined using the old guideline recommendations in order to avoid cases with diagnostic-therapeutic mismatch.

uPA and PAI-1 were analyzed prospectively from representative pieces of tumor tissue that were frozen in liquid nitrogen after histologic evaluation. The frozen samples were pulverised using a micro-dismembrator, suspended in a buffer (pH 8.5) containing 0.02 M Tris-HCl, 0.125 M NaCl and 2% Triton X-100, and shaken for three hours at 4°C. The obtained suspension was centrifuged for 30 minutes at 100,000 x g. Protein content was determined with the Pierce^TM BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA), and uPA and PAI-1 content was determined with commercially available ELISA assay kits (American Diagnostica, Greenwich, CT, USA). uPA and PAI-1 content was expressed as nanograms of analyte per milligram of tissue protein.

The correlations between different variables were tested with Spearman’s rank correlation, Kruskal-Wallis test, Mann-Whitney U test and chisquare test, depending on the type of variables. The continuous variables uPA and PAI-1 were coded as binary variables using the previously optimized cutoffs of 3 ng uPA/mg protein and 14 ng PAI-1/mg protein to distinguish between low- and high-risk patients⁸, and combined into one variable (both low versus one or both high) as previously shown to be of greatest clinical relevance.⁹ Because adjuvant trastuzumab was not available for the whole cohort, the patients were divided into three groups based on HER2 status and adjuvant trastuzumab treatment for the survival analyses: HER2 negative, HER2 positive who were not treated with adjuvant trastuzumab, and HER2 positive who were treated with adjuvant trastuzumab. Disease-free survival (DFS) was calculated from the date of primary surgery until the date of disease recurrence or death from any cause, or the date of the last follow-up visit in case of no recurrence or death. Overall survival (OS) was calculated from the date of diagnosis until the date of death from any cause, or the date of the last follow-up visit. Breast cancer specific survival (BCSS) was calculated from the date of diagnosis until the date of death from breast cancer, or the date of the last follow-up visit or death from other causes for censored patients. Kaplan-Meier method was used to calculate survival curves and univariate Cox regression was used to assess the differences between the curves in univariate analysis. Multivariate analyses were performed by applying the multivariate Cox proportional hazards model. All variables regardless of univariate analysis results were initially included in the Cox model and the method used for the model was backward stepwise likelihood ratio (LR). Model if term removed was reported separately. All tests were performed at a significance level of p=0.05 and a confidence interval (CI) of 95%. All p values were two-sided. Statistical analysis was performed using the SPSS software package v. 21 (IBM, Armonk, NY, USA).

The results of our study indicate an important prognostic value of uPA/PAI-1 level in node-negative breast cancer patients. Even though uPA and PAI-1 values were associated with most of the prognostic factors currently in use for clinical decision-making in the adjuvant setting, multivariate analysis showed that uPA/PAI-1 level carries additional, independent prognostic information. This was particularly true for DFS analysis where uPA/PAI-1 level was the second most important variable in the survival model after age and was retained in the backward LR model after the final step along with age and estrogen receptor status. In OS analysis, the final model included only age and tumor grade, which is probably a reflection of the very strong association between higher uPA and PAI-1 values and less differentiated tumors. However, in the multivariate model that included all variables, the HR for death from any cause was 1.98 (95% CI 0.83–4.76) for patients with high uPA and/or PAI-1 compared to patients with both values low, which shows a substantial possibility of an important effect on OS that would have probably remained statistically significant in a larger sample. The same three variables were the most important in the multivariate model of BCSS as well. Here, the singular prognostic importance of tumor grade was even more evident. As opposed to DFS and OS which both included deaths from other causes as events in the analysis, younger age seemed associated with worse BCSS. It is important to emphasize that BCSS analysis must be interpreted with caution due to the small number of events. Considering the HRs and confidence intervals, it is probable that both age and uPA/PAI-1 level importantly influence BCSS and would have retained statistical significance in a larger sample. On the other hand, HER2 status only showed prognostic significance in univariate DFS and BCSS analyses when comparing HER2 negative to HER2 positive patients who were not treated with adjuvant trastuzumab. It did not remain an important determinant of DFS and BCSS in multivariate analyses and lost all prognostic value in OS analysis.

A subgroup analysis of DFS according to estrogen receptor and HER2 status showed that apart from the no longer relevant group of HER2 positive patients who did not receive adjuvant trastuzumab treatment, uPA/PAI-1 level was prognostically by far the most important in estrogen receptor positive tumors. In contrast, its prognostic value in multivariate analysis was practically null in HER2 positive patients treated with adjuvant trastuzumab and in triple negative patients. Conveniently, estrogen receptor positive patients are the ones with the highest uncertainty regarding the benefit of adjuvant chemotherapy, while the other two subgroups are generally recommended adjuvant chemotherapy ± anti-HER2 therapy regardless of other prognostic factors.

Our findings confirm those of numerous other studies that have reported uPA and PAI-1 to be statistically independent prognostic factors in lymph node-negative breast cancer patients.^10-18 However, these studies did not include HER2 status which is now an important part of the decision about adjuvant systemic therapy. Because of the observed associations between overexpression of the HER2/ neu protein and tumor proteolytic factors in breast and in other cancers^19-21, HER2/neu has been suggested to up-regulate the proteolytic enzymes, including uPA and PAI-1, and thus play a direct role in tumor invasion and metastasis.¹⁹ This possible association might be one of the additional reasons besides methodological difficulties why uPA and PAI-1 testing is still not that frequently used in the clinic in spite of the excellent evidence of its clinical utility.¹ Even so, the main reason for this inconsistency are practical issues. uPA and PAI-1 are determined using validated ELISA assay kits as described above. This measurement requires relatively large amounts of fresh or freshly frozen tumor tissue, which is not practical for clinical use, especially in needle or surgical biopsies and in small tumors. Besides, the results may not be available at the time of the histology report due to sample pooling. However, attempts to develop immunohistochemistry assays on formalin-fixed and paraffin-embedded tissues are ongoing.¹

Among the few studies that considered the clinical relevance of both HER2 and proteolytic enzymes for survival in breast cancer patients, the results are somewhat conflicting. Harbeck et al.²² showed that PAI-1 was the only independent prognostic factor for DFS and OS when considered along with tumor size, tumor grade, steroid hormone receptor status, uPA, HER2 status, MIB1, SPF, p53, and cathepsin D in 100 node-negative breast cancer patients. Similarly, Bouchet et al.²³ found no additional prognostic information of HER2/neu protein levels when evaluated in multivariate analysis together with uPA, PAI-1 and traditional histologic factors in the subgroup of 226 node-negative patients. They reported DFS to be independently influenced by PAI-1 and tumor size and OS by PAI-1 and uPA levels. Recently, Buta et al.²⁴ also reported superior DFS in 73 node-negative, postmenopausal, steroid hormone receptor positive breast cancer patients with smaller tumors and low PAI-1, independent of HER2 status. On the other hand, Konecny et al.¹⁹ reported that both uPA/PAI-1 level and HER2 status independently influenced DFS in addition to tumor size and nodal status in a group of 542 patients with a short follow-up not selected by nodal status. The same factors were found to be important in multivariate OS analysis as well, but this time HER2 status did not quite reach statistical significance. Interestingly, among 118 node-negative patients with long-term follow-up, Zemzoum et al.²⁵ found uPA/PAI-1 to be the only variable independently influencing DFS, and HER2 status to be the most important factor in the multivariate analysis of OS. uPA/PAI-1 and tumor grade were of borderline significance for OS. Complicating the matter even further, without reference to uPA and/ or PAI-1, Schmidt et al.²⁶ reported HER2 status to be prognostically significant in node-negative breast cancer patients only when determined by FISH and not when determined immunohistochemically. In our group of 273 node-negative patients with long-term follow-up who were available for multivariate analysis, the combination of uPA and PAI-1 levels clearly carried independent prognostic value for DFS and was important although not formally statistically significant in multivariate analysis of OS as well, while HER2 status determined with the usual combination of immunohistochemistry and FISH did not prove to be an important determinant of DFS or OS in multivariate analyses regardless of adjuvant anti-HER2 treatment.

Most of the studies confirming the prognostic value of uPA and PAI-1 in breast cancer patients have focused on patients who received no adjuvant systemic treatment. In fact, the prognostic information from these biomarkers seems diminished in patients who receive adjuvant treatment, particularly adjuvant hormone therapy^27,28, indicating a possible predictive role of uPA and PAI-1 for response to therapy. However, the prognostic importance of uPA/PAI-1 level was evident in our group of patients although the vast majority received adjuvant hormone therapy and a significant fraction were given adjuvant chemotherapy. Contrastingly, based on our results, HER2 status is clearly more important in its predictive than in its prognostic role.

Our study has several limitations, the principal one being its retrospective character and the associated possibility of bias. A major problem is missing data, particularly on uPA and PAI-1 values which could only be determined in about 40% of the patients because of our institution’s economical limitations and the fact that these markers still largely serve only academic purposes. Because of inconsistent histology reports in the past, some other information is missing in a fraction of patients. The total number of patients with available information was used for each univariate analysis and multivariate analyses were performed in the 273 patients in whom complete data were available. Even so, we are aware of only one study comparing the prognostic values of uPA, PAI-1, HER2, and traditional prognostic factors that included a slightly larger subgroup of 283 lymph node-negative patients.¹⁹ Another possible limitation is the comparatively high proportion of deaths from causes other than breast cancer, which may have somewhat obscured the results of both DFS and OS analysis and is probably also the cause of such a marked association of older age and worse prognosis, an assumption confirmed by the fact that older age was associated with improved breast cancer specific survival. Furthermore, our results must be considered in the light of the unavoidable multiple analyses and the possibility of a type I statistical error. To facilitate interpretation, 95% confidence intervals and not just the p values have been stated wherever possible. On the other hand, one of the main strengths of our study in addition to the combination of analyzed prognostic factors is the long follow-up which was in the range of 49-181 months, the median of 100 months being more than three times the median follow-up of the larger study by Konecny et al.¹⁹ Moreover, we believe that the scientifically sound and straightforward statistical analysis is a strength of our study as well.

Based on the results from our retrospective analysis of node-negative breast cancer patients with long-term follow-up, we conclude that the combined uPA/PAI-1 level carries important additional prognostic information, particularly for DFS, even after all traditional prognostic factors as well as HER2 status have been taken into account. We believe that routine use of uPA/PAI-1 level would further improve risk stratification and adjuvant therapy decisions in this setting, especially in estrogen receptor positive patients.