Like cell proliferation, cell loss plays a significant role in the growth rate of tumours (21). Both factors contribute to a considerable inter-patient variation in the growth rate of morphologically similar tumours in the same site of the body. In the evaluation of response to therapy, monitoring tumour size via anatomical imaging (11) and molecular imaging, combining tumour size with its metabolism (22), are two frequently used methods.
Here, we evaluated the usefulness of a metric of cell loss, defined as the ratio between the concentration of TK1 in serum and tumor volume, for early prediction of the outcome of chemotherapy in patients with BC. An important finding was that this cell-loss metric, obtained prior to and 48 hours after the 2nd cycle of NACT, varied greatly between patients and, in addition, was significantly related to the pathological response established at surgery after six cycles of therapy. Thus, for a patient displaying a high cell-loss metric the pathologic response was more favourable. Further, in patients with remaining tumours, tumour size was inversely related to the early cell-loss metric.
These associations between cell-loss and pathologic response are notable not only in the clinical perspective but also because of their biological implications. Firstly, there were substantial inter-patient differences in tumour size prior to treatment, reflecting various stages of development. Also, the change in tumour volume after two cycles of therapy differed considerably between patients. In spite of the wide range of tumour size to which sTK1 was related, significant associations were found between the cell-loss metric and the presence or absence of tumour. Secondly, there was a time period of at least 4 months between establishment of the cell-loss metric and surgery. During this interval the patients were subjected to four further treatment cycles, with the addition of bevacizumab. The pathological response is the result of tumour cell loss, which is dependent on the fraction of proliferating cells exposed to varying concentrations of drugs. Tumours may also differ with respect to intrinsic resistance to chemotherapy or in the repopulation capacity of clonogenic cells between the treatment cycles (23). A poor pathologic response could be due to drug resistance as well as to efficient repopulation between treatments.
Thus, there are several factors that would have the potential of diffusing the association between an early cell-loss metric and the pathologic response. That the early cell-loss metric nevertheless showed a significant relationship with the pathologic response suggests that it represents an inherent tumour property - sensitivity to the cytotoxic substances - that can differ greatly between patients but is comparatively stable within patients, persisting through several cycles of chemotherapy. In fact, also the values of the cell-loss metric established before treatment showed a significant association with the pathologic outcome.
The present findings are also of relevance as regards the mechanisms for release of macromolecules into blood and suggest qualitative differences in cell death between tumours and normal tissues. Normal tissues with high cell turnover are tangibly affected by cytotoxic treatment. In any of the present patients the quantity of normal tissues with high fraction of proliferating cells is likely to have been many times greater than that of the tumour. For instance, the red bone marrow in a woman amounts to approximately 1200 grams, containing about 7.5x1011 nucleated cells (24), 14% being in S-phase (25). Therefore, if the pathway for removal of damaged cells had been the same in normal tissues and tumour, then the serum level of TK1 would not have been capable of reflecting a property of the tumour. In other words, whereas cell death in tumours is associated with a significant release of TK1, normal tissues must have functions preventing this release. It is generally assumed that the elimination of damaged normal cells follows the apoptotic pathway (26). Therefore, it seems likely to be a different pathway for tumour cell elimination, namely the necrotic pathway, and this would be responsible for the release of TK1 into blood. Leakage of macromolecules via the necrotic pathway is believed to be related to active phagocytosis (27). This makes it tempting to reflect upon certain new concepts of regulated immunity in oncology as well as the results of immunotherapy by blockade of the CTLA-4 protein (28) or PD-1 protein (29) on the surface of T-cells. Possibly, the success of such enhanced phagocytosis could be monitored via measurements of the concentration of TK1 in serum.
In 57 of the patients, the cell-loss metric could be established also prior to the 2nd treatment. Although the values 48h after treatment were approximately 50% greater, it appears that the relationship with pathologic response was higher for the pre-treatment values. An explanation for this could be that during treatment cell loss in normal tissues temporarily exceeds the capacity of the apoptotic pathway, resulting in a non-tumour specific release of TK1 into blood. Such a confounding factor would be less pronounced 2-3 weeks after treatment. As regards other tumour- or patient-related data, we did not find any factors which correlate with, or explain, the cell-loss metric. The values 48h after the 2nd treatment were independent of the baseline. In addition, the prediction of pathologic response could not be improved by combining the cell-loss metric with the histologic proliferation marker Ki67/Mib1.
It might appear remarkable that such a basic and well-established tumour property as the fraction of proliferating cells did not contribute to the predictive power of the cell-loss metric. Nevertheless, there is a reasonable explanation for this finding. Proliferation and cell loss are both complex phenomena. Proliferation may constitute a primary component in a network of processes whereby cytotoxic therapy results in cell loss. In other words, cell loss would be determined not only by the fraction of proliferating cells (as expressed by Ki67/Mib1) but also by a multitude of less well-known factors. If the cell-loss metric thus reflects a sum effect of several mechanisms, including the rate of proliferation, then, adding Ki67/Mib1 would not contribute to the predictive value of the metric. In the practical perspective, the cell-loss metric might be considered causally closer to the outcome of treatment.
The finding that a number of tumour properties did not differ between the quartile groups does not imply that they are clinically insignificant but that they are independent of the cell-loss metric. Therefore, it is logically possible that some of them would improve the prediction of pathologic response. This is the main theme of a following study (to be published), where it was found that combining the cell-loss metric with histopathologic markers, such as receptors for oestrogen and progesterone, improves the predictive power in terms of both sensitivity and specificity.
The clinical value of tumour biomarkers is to guide therapy. A distinction is made between prognostic markers, supposed to provide information about long-term outcome, and predictive markers, which reveal a tumour’s response to treatment. Ideally, the adequate choice of therapy would be based on tumour or patient characteristics established before treatment. For a defined type of tumour there is, nevertheless, always an inter-patient variability in the response to treatment. Therefore, predictive markers for early detection of the effects of treatment would be a valuable complement to tumour characteristics established at diagnosis. Among the most well-established tissue markers in oncology are the receptors for oestrogen, progesterone and growth-factor 2 (30). These are all used in the primary characterization of BC and constitute the targets in hormone therapy as well as in treatment with monoclonal antibodies. Molecular characterization of tumours has generated an increasing number of putative predictive biomarkers (9, 10). The manifold of such markers is in line with the demands of a more individualized treatment. In addition, the increasing sub-classification of tumours requires principles for exploring the usefulness of new biomarkers.
Nevertheless, there is a paucity of methods for the early evaluation of tumour response during treatment. Such methods would give a valuable contribution particularly in the management of patients for whom the statistically calculated benefit of a standard treatment is low and has to be balanced against unnecessary side effects. For instance, in low-grade, low-stage ER+/HER-2neu luminal-A tumours, pCR after cytotoxic treatment was achieved in less than 10% of patients and, in addition, pCR was not prognostic for long-term survival (1, 2). Early identification of individual patients with poor response would permit a switch to hormone therapy or motivate immediate surgery - and suffering due to unnecessary side effects could be avoided. In BC, clinical monitoring of tumour volume early during treatment have motivated shifts from anthracycline-based therapy to docetaxel (31) and from docetaxel-doxorubicin-cyclophosphamide to vinorelbine-capecitabine (32) in non-responding patients; and these shifts in treatment were associated with enhanced clinical and pathological remissions.
A few studies deal with the release of macromolecules early during chemotherapy and how such early response markers are associated with pathologic outcome or long-term survival. In patients with lung cancer a high activity of TK1 in serum after the first and second cycles of cytotoxic treatment was associated with a significantly longer survival (33). Analogously, in colon cancer a lack of increased TK1 activity during chemotherapy was related to a poor prognosis (34). Further, during chemotherapy for colon cancer, patients in whom the concentrations of cell-free mutated DNA had declined dramatically prior to the second treatment also displayed a substantial reduction in radiologic measures of tumour volume (35). In lung cancer, a rapid decrease in the serum concentration of mutated EGFR-DNA 14 days after initiating treatment with erlotinib (a tyrosine kinase inhibitor) was associated with tumour shrinkage 2 months later (36). Likewise, during the first week of chemotherapy for lung cancer, the levels of nucleosomes were substantially lower in patients who responded to treatment than in non-responders (37).
In BC, no significant changes in nucleosome levels have been found during the first two treatment cycles of NACT (38). However, an increased concentration of uncleaved cytokeratin-18, which is an indicator of necrotic cell death, early during the first cycle was associated with a favourable clinical response and improved survival (39). In triple-negative non-metastatic BC, the persistence of TP53 mutated DNA in serum before the 2nd cycle of anthracycline/taxane-based chemotherapy has been related to a shorter disease-free and overall survival. However, no association was found between ctDNA levels and pCR (40). In a pioneering study, patients with metastatic BC who displayed persistent high levels of circulating tumour cells after three weeks of cytotoxic therapy were subjected to a shift to another drug; there was, however, no improvement in survival (41).
To our knowledge there are no studies which address the clinical value of a measure that relates the levels of a macromolecule, released from disrupting tumour cells, to the volume of the tumour. The usefulness and predictive power of the TK1-based cell-loss metric have the potential of being improved in several ways. A limitation of the present study was that the patients were examined and treated in five different clinics. Methods for estimating tumour size included calliper measurement, mammography and ultrasonography, the accuracy of which ranges between 57 and 79 per cent (42). Methods may differ not only in accuracy but also with respect to the smallest tumour that can be detected. Thus, it might be considered whether in cases with small tumours a less sophisticated method would tend to yield values close to zero and, hence, a converse bias in the cell-loss metric. In the present study, the distribution of data does not suggest any bias of this kind. Nevertheless, although routine clinical management permits a variety of techniques for measuring tumour volume, new prognostic tools may motivate more standardized and accurate methods. Magnetic resonance imaging would have provided a higher accuracy and consistency in data, particularly in cases where tumours were small already prior to treatment. Another strategy for improving sensitivity and accuracy is to combine two different methods. At the Karolinska University Hospital, were the majority of the present material was handled, each patient was routinely examined with both mammography and ultrasonography.
Reactions of lymph nodes on therapy could not be assessed, but release of TK1 from metastatic lymph nodes cannot be excluded. Another issue is the time point for establishing the cell-loss metric. The precise time course for treatment-induced changes in sTK1 remains to be clarified, and it may, in addition, be dependent on the type of treatment. As already noted, the predictive value of the cell-loss metric appears to be higher prior to the 2nd treatment than 48 h after treatment. Advantages of the present study were the prospective layout of the original clinical trial and the absence of patients with distant metastases, which would have constituted sources of TK1 with unknown volumes. Prospective studies should be performed to confirm the present findings, to establish the optimal time points for the cell-loss metric during different treatments, and to define cut-off values for discriminating between responders and non-responders.