The recurrence pattern following delayed breast reconstruction after mastectomy for breast cancer suggests a systemic effect of surgery on occult dormant micrometastases

To our knowledge, this is the first study investigating the dynamics of recurrences occurring after delayed breast reconstruction in breast cancer patients. The main result of our analysis is that when the time origin is set at the reconstruction date, the hazard rate for ensuing recurrence displays a first main peak in the 2nd year and a later minor peak at the 5th–6th year after reconstruction (Fig. 2, red line). The recurrence risk for the same patients, when the time origin is placed at mastectomy, is bimodal with peak positions at the same time points relative to mastectomy, after 2 and 5–6 years (Fig. 2, blue line), as expected [31]. Thus, when the time origin is moved for each reconstructed patient to the reconstruction date, the recurrence risk pattern is similar to that observed following primary mastectomy. Of note, time origin displacement reveals an increase of the early peak with a concomitant decrease of the late level of recurrence risk (Fig. 2), suggesting that recurrence redistribution is associated with the reconstruction maneuver which could be said to act as a wave breaker for recurrences. These findings suggest that mastectomy and reconstruction induce similar biological effects on subclinical preexisting metastases.

The effects of primary mastectomy have been investigated in both animals and humans during the past century [36]. An unintentional effect of surgery in breast cancer patients with clinically undetectable micrometastatic disease has been explained by a paradigm based on the concepts of tumor homeostasis, tumor dormancy, and surgery-related enhancement of metastasis growth [37]. The model assumes both cellular and micrometastatic tumor dormancy, with ordered transitions between these two quiescent states and subsequent development of overt metastasis and, subsequently, a transient phase of acceleration of metastatic growth. Preclinical studies have pointed to an angiogenic switch as a possible involved mechanism, in which the microenvironment is altered by tissue trauma to become proangiogenic with increased levels of VEGF and reduced levels of angiogenesis inhibitors such as TSP-1 [8]. Others have focused on the role of surgery-induced immunomodulation with demonstration of a stimulatory interaction between cells of the innate immune system and adjacent cancer cells [38]. The truth may lie in both models as the immune and angiogenic systems have multiple points of intersection [39]. Our findings support the concept that also delayed breast reconstruction may accelerate metastatic growth in subjects with dormant metastatic foci similar to the effects observed after primary surgery. This explanation is further supported by the finding that surgical approaches with different extents result in different recurrence risks, although with the same time rhythm (Fig. 3). Specifically, the more extensive reconstruction modalities DIEP/TRAM and bilateral surgical procedures give rise to a higher early peak in comparison with unilateral implant surgery. This difference is limited to the recurrence risk level, which is differently modulated within the same time cadence, while the two types of reconstruction do not affect long-term outcome differently (Fig. 4).

The proposed explanation assumes that the risk of clinical appearance of metastasis is dependent on the action of surgery-related factors on the subclinical metastatic state of the host. This assumption is confirmed by the marked influence of both tumor size and nodal status on the hazard level (Fig. 5). At diagnosis, these two well-known prognostic factors indicate the recurrence risk during the disease course. Thus, it is coherent that an additional triggering factor, such as delayed surgical reconstruction, may result in different outcomes when patients with different underlying recurrence risk are involved.

When studying the relation between outcomes of patients undergoing delayed reconstruction in comparison with patients undergoing mastectomy without reconstruction, crucial problems emerge. In addition to the retrospective nature of such studies, the reconstructed group is characterized by a selection event not yet occurred (and therefore unknown) at the mastectomy time and, moreover, occurring at varying patient-related times during follow-up, thus raising important issues in the statistical analysis. To overcome these drawbacks, we used two different approaches. In the first approach, which has been frequently adopted in this field in spite of its intrinsic naivety [28, 29], we performed a matched random choice of the control patients. For each reconstructed patient, we randomly identified a matched control patient with similar initial characteristics, selecting her among the not reconstructed patients who were disease free at the date of reconstruction of the considered patient. This matching modality resulted in a good balance between the two sets of patients (Table 1) and avoided drawbacks detectable in published reports, such as dissimilar patient characteristics (e.g., age) [40] or dissimilar prognostic factors [28, 29]. In addition, the recurrence dynamics of matched controls, when time origin was set at mastectomy (Fig. 6, blue line), was coherent with the analogous hazard rate pattern observable in similar patients from different databases [31]. It displays a first peak significantly higher than the corresponding peak of reconstructed patients (Fig. 6, red line) revealing that, in spite of the balanced initial prognostic factors, matched control patients display worse prognosis than reconstructed patients. This finding is consistent with a population-based registry study analyzing Danish women who received reconstruction with implants only [41].

In the second approach, we addressed the analysis in the framework of multiple timescales in multistate models [35]. The regression analysis involved all 868 control patients and 312 reconstructed patients. It accounted for the joint effect of prognostic factors (e.g., tumor size, nodal status, etc.) and, most importantly, the change in the hazard rate for recurrence resulting from specific events, such as reconstruction, occurring at varying time points. The multiple scale analysis provides evidence that following reconstruction women suffer a transitory, significant increase of recurrence risk during the first 2 years in comparison with not reconstructed patients (Fig. 8). This finding provides structural evidence for the enhancing effect of reconstructive surgery on subclinical metastases, which brings on the temporary raise of clinically evident recurrences. Such behavior is suggestive when it is considered in the light of the above-reported model. Indeed, the whole pattern would suggest that the early peak may be caused by an event that has been brought forward, and in the absence of reconstruction were to be expected at a later time. Taken together, the two analysis approaches provide evidence that reconstructed patients (a) suffer increased surgery-related recurrence rate following the usual bimodal pattern and (b) display disease-free survival that is not worse (maybe even better) than that of not reconstructed patients (Fig. 8).

Although not directly comparable, the recurrence pattern analysis apparently diverges somewhat from a previous investigation where trauma or intervening surgical procedures unrelated to cancer were not associated with an increased rate of breast cancer recurrence [42]. Still, the fact that the effect of delayed reconstruction on the recurrence dynamics does not translate into reduced recurrence-free or overall survival is in keeping with most reports on the same subject [28, 29, 40, 41, 43]. The reason underlying the similar or relatively better long-term outcome of reconstructed patients is presently undetermined. Patients opting for reconstruction tend to be younger and have less comorbidity. Most studies to date have focused on immediate reconstruction [43‐45]. We are inclined to ascribe this finding to selection bias of patients receiving reconstruction due to factors here unaccounted for, such as socioeconomic conditions and better general health. The former factor is related to the finding that patients who are of lower socioeconomic status are more likely to have a recurrence than women of higher social class [46]. Furthermore, current smokers or diabetic patients are not accepted for microvascular free (DIEP) or pedicle flap surgery (TRAM). In addition, body mass index, which is a prognostic factor in cancer [47, 48], would not be higher than 30 in patients offered advanced reconstruction. Most patients undergoing DIEP or TRAM procedures have a secondary or even tertiary surgery performed due to cosmetic purposes or complications. Thus, the observations are measurements of the effect from the first extensive reconstructive surgical procedure, independent of the duration of reconstructive surgery or the number of surgical events. Again, such subsequent procedures might represent a possible bias and the observed effects might therefore be diluted.

In conclusion, our study indicates that reconstructive breast cancer surgery constitutes an independent stimulating event on the growth of micrometastases leading to accelerated relapse rates. The effect is similar to that observed after primary breast cancer surgery. Importantly, this does not translate into worse long-term disease-free survival. Our results may provide indirect evidence that immediate reconstruction would be more beneficial than delayed as this obviates one possible growth stimulating event. Still, randomized trials assessing this question are not ethically or practically feasible. Further studies are ongoing and will shed more light on tumor biological mechanisms behind the observed phenomenon.