We conducted an observational study of 235 early breast cancer patients who were diagnosed and treated at the IRCCS Regina Elena Cancer Institute in the time frame between July 2008 and September 2010. For all of them, blood samples were collected prior to any therapeutic procedures and assessed for circulating levels of biomarkers related to coagulation activation. One hundred fifty cancer free women comparable by age and recruitment period served as control group. Data analysis were planned and performed to allow the development of a tool instrumental to the assessment of risk categories based on features related to relevant clinicopathologic characteristics and coagulation profiling for the biomarkers of interest. Within our study population, results from multivariate analysis revealed a prognostic role for age, pT, circulating levels of FVIII and DD. These variables were then used to define risk categories, with cut off points obtained by applying the best “splitter” cut-off definition to our case-series. The so developed tool proved efficacy in distinguishing categories characterized by significantly different survival estimates. In more detail, the lowest risk of death from any cause was ascribed to patients aged 70 years or less, with a pT1 disease at diagnosis, circulating levels of FVIII within the normal range and low levels of DD. Age older than 70 years and FVIII levels greater than the pre-established cut off values where instead associated with an intermediate risk of death, while the highest risk of death was associated with a pT2 and DD levels beyond the threshold defined for our study population.
Among the clinically relevant patient- and disease-related determinants of interest, age and tumour size (pT) showed prognostic relevance in uni- and multi-variate analysis and were thus included in the model for risk stratification. This evidence is consistent with previous literature from the early setting [20, 21]. In strict regard to the patient age, our results provide several hints for discussion. In our case series, values of median age and best “splitter” cut-off for age were 60 and 70 years, respectively. In addition, the outcome considered for the overall case series was death from any cause as we lacked data on breast cancer specific mortality for the totality of our patients, as specifically pointed out in the paragraph concerning this study limitations. Thus, we addressed an outcome, i.e., death from any cause, which is definitely affected by aging. Indeed, the lack of specific data concerning the extent to which age, co-morbidities, and breast cancer have concurred to determine our patients’ death may be more appropriately exemplified by the use of a terminology distinguishing between “likely cause of death” vs “other causes of death”, particularly in light of the broad age range which characterizes our case-series, i.e., 29–39 years. The relation between aging and breast cancer is complex and the investigation of the underlying mechanisms animates intensely the inherent research area [22]. In reference to recently published and clinically focused evidence, Lodi et al. have evaluated relevant clinicopathologic features and breast cancer specific survival outcomes in a systematic review of women over 70 years with breast cancer. Sixty-three original studies published between 2006 and 2016 were considered. Consistently with our findings, the authors reported on the association between older age and significantly higher 5- and 10-year mortality [23]. Older age at breast cancer diagnosis should be considered not only in light of its prognostic role for the disease of interest, but also in reference to the role of DD and FVIII as biomarkers of aging, widely and consistently supported by the inherent literature in both non-cancer and cancer patients [24,25,26,27,28]. On this basis and in strict regard to our study population, we assessed the interaction between age and circulating levels of DD and FVIII in Cox models including an interaction term. In this specific cohort, we could not observed significant interaction between the variables tested (p = 0.20 and p = 0.94, for the interaction between age and DD and FVIII, respectively).
In this study population, we found no evidence supporting the prognostic relevance of the specific molecular subtype, i.e., luminal A, luminal B, HER2 enriched and triple negative breast cancer, on patient survival. Indeed, in multivariate analysis of OS, the related variables tested not significant (p: 0.74). This finding, i.e., lack of the impact of molecular subtype on the outcomes of interest in a breast cancer patient population from the early setting, is consistent with previous studies [14, 16] and in need of further assessment for clarification purposes. Indeed, in our case series, we exclusively observed some evidence of the prognostic relevance of ER expression, one of the main determinants of the specific molecular subgroups, in the subset of patients for whom DFS data were available (N: 62), with our results supporting a protective role of ER expression (p: 0.003). This same patient subgroup also offered the chance for evaluating our scoring tool in reference to DFS. Although somewhat limited by the restricted sample size, results from the analysis performed within this subset confirm the prognostic relevance of biomarkers related to coagulation disorders, and the need for including such evidence in risk assessment for early breast cancer patients.
An appropriate discussion of our results cannot exclude a referral to the existing evidence concerning the use of anticoagulants in cancer patients, which has been recently summarized in a Cochrane systematic review carried out by Kahale et al. In brief, the authors conducted a comprehensive literature search updated to December 2017. Of the identified 7668 unique citations, 16 manuscripts reporting on 7 randomized clinical trials (RCTs) fulfilled the inclusion criteria and were thus included. Overall, these trials enrolled 1486 participants. Results from the meta-analyses of the RCTs included do not rule out a mortality benefit from oral anticoagulation in people with cancer but suggest an increased risk for bleeding. In the attempt to interpret these findings correctly, the lack of data specifically referred to the site of cancer origin should be considered. Indeed, the need of further evidence specifically related to the cancer type and stage is acknowledged by the authors themselves when discussing their research implications [29].
The pathogenetic layout of the association between the activation of coagulation and cancer is multifactorial in nature. In addition, most of the actors involved play a pivotal role in several mechanistic pathways that sustain cancer-related biological processes with a notable degree of overlap. The previously mentioned role of FVIII and DD as factors involved both in cancer, thrombogenesis and aging may appropriately exemplify this latter assertion [24,25,26,27,28]. Cancer may provide an unusual and polyvalent frame within which patient- and disease-related features concur to outcome determinism, both for thromboembolic and cancer related events. The relationship between thrombosis and cancer is founded on the evidence that cancer promotes a prothrombotic switch of the host hemostatic system, and in turn, blood clotting activation is intimately tight to tumor growth and dissemination. The main mechanisms of cancer-related thrombosis encompass the expression of procoagulant factors at the tumour cell level, the release of microparticles, inflammatory cytokines e.g., tumor necrosis factor-alpha, interleukin-6, and proangiogenic factors, e.g., vascular endothelial growth factor, basic fibroblast growth factor by tumor and/or host cells, and the expression of adhesion molecules to bind platelets, endothelial cells, and leucocytes. These same properties are also involved in cancer progression [30, 31].
Our study has some limitations, which are mainly represented by the lack of data concerning cancer-specific survival for the overall case series. This is unfortunately common when working in the real word setting. Indeed, data collection and entering into dedicated databases has not stably entered the clinical practice. To mitigate such limitation, we have attempted to perform subgroup analysis in a subset of patients for whom cancer-specific survival data were available. Unfortunately, this subset was extremely limited in size (N: 62). This refrained us from conducting analysis beyond the multivariate models. However, also in this smaller subset, we could observe evidence supporting the prognostic relevance of both patient- and cancer-specific feature along with circulating levels of coagulation biomarkers. The lack of data on menopausal status should also be acknowledged, given the relevant differences in terms of risk factors, presentation at diagnosis, characteristics and management between pre- and post-menopausal breast cancer patients [32,33,34]. In the attempt to minimize such limitation, we codified a categorical variable with a 50-year cut off value and assumed that women aged less than 50 years (N: 70; 25.5%) were most likely premenopausal. However, in univariate models of OS and DFS, our surrogate variable of menopausal status did not test significant (p: 0.32 and p: 0.81, respectively).
Our study also has strengths of relevance. Among them, central management of biomarkers of coagulation activation is noteworthy. Plasma samples were collected and handled according to pre-specified and highly standardized operative procedures. Sample assessment was performed at the institutional laboratories, which are ISO-certified (ISO 9001 certification). This increases our confidence in the quality of the evidence stemming from our study. As cited in the “Results” section, the median follow up for the cohort of interest was 95 months, which is fairly acceptable in terms of length when assessing outcomes in a cohort of early breast cancer patients. However, this 10-year follow up window imposes considerations related to the remarkable advances achieved both in the loco-regional and systemic treatment [35,36,37,38,39,40] and invites caution in the generalization of our results to early breast cancer patients in current treatment. At the same time, this latter matter, along with the results from the work herein presented, encourages future investigation within this same research pipeline.