Effect of trastuzumab

While ER/PgR loss has been observed following endocrine therapy^{30, 31 and 32}, there are minimal data supporting trastuzumab-driven HER2 loss. From four studies assessing trastuzumab effect on HER2 loss at relapse, no correlation was evident.^{9, 17, 19 and 23} This lack of correlation contrasts with findings from the neoadjuvant setting. A metaanalysis assessing receptor expression in breast cancer specimens pre- and post-neoadjuvant chemotherapy +/− trastuzumab found a significant decrease in HER2 expression in the surgical breast cancer specimens compared with pre-treatment biopsies in patients treated with trastuzumab.³³ However it is unknown whether this finding reflected temporary down-regulation of HER2 or a true change in tumor biology.

There are no data supporting trastuzumab as a driver of HER2 loss (Table 1), although evidence is limited, primarily as the majority of discordance studies included patients diagnosed with PBC prior to routine use of adjuvant trastuzumab. As the use of adjuvant trastuzumab has now been standard practice for several years, contemporary or future analyses may shed more light on this.

Effect of chemotherapy

In a retrospective study of 182 patients with HER2+ PBC and paired metastatic tumor biopsies, Niikura et al. reported a HER2 discordance rate of 24% (n = 43), with a significantly higher rate of HER2 loss among patients who received adjuvant chemotherapy compared with those who did not (35% vs. 10%, respectively, p = 0·02), irrespective of trastuzumab treatment.¹⁷ Liu et al. reported increased HER2 and ER/PgR discordance rates in 46 patients presenting with metachronous liver metastases after adjuvant chemotherapy, +/− trastuzumab, and +/− endocrine therapy (discordance rates of 30%, 54% and 11% for ER, PgR and HER2, respectively) compared with 12 patients with synchronously diagnosed PBC and liver metastases (rates of 0%, 33% and 8% for ER, PgR, and HER2, respectively).¹⁴ These findings suggest a potential impact on discordance from adjuvant treatment, though HER2 discordance after either trastuzumab or chemotherapy was not specifically reported. In contrast, five studies reported no correlation between adjuvant chemotherapy and HER2 discordance. ^{10, 12, 19 and 34}

From currently available data, chemotherapy does not clearly promote HER2 discordance (Table 1), though again further research is needed.

Other predictive factors

Individual tumor or patient characteristics predicting increased likelihood of discordance were not identified in any of the studies reviewed and have not, to our knowledge, been investigated.

Effect on prognosis

Interpretation of the prognostic impact of discordance is limited by the confounding influence of any treatment alteration based on the metastatic biopsy result. In particular, assessment of post relapse survival (PRS) is prone to this bias.

The majority of studies (n = 10) that included some assessment of discordance as a prognostic factor utilised PRS (Table 2), with an association with worse outcome in patients with discordance vs. concordance reported in several of these studies. ^{5, 8, 17, 22 and 35}

From Niikura et al.¹⁷, the outcomes of 168 patients with distant metastases were assessed, with overall survival (OS) being significantly longer in patients with HER2 concordance compared with HER2 loss, HR for OS = 0·47, p = 0·003. This difference was significant in both univariate and multivariate analyses. Of note, 16% (n = 7) of patients who lost HER2 expression continued to receive anti-HER2 therapy. Survival was not impacted by this treatment, implying that anti-HER2 therapy may have been ineffective in these patients, although this is based on a very low number of patients.

In an analysis of 151 breast cancer patients from Wilking et al.²², HER2 discordance was associated with increased risk of dying both from time of diagnosis of PBC and from time of relapse, on multivariate analysis, after adjustment for age, tumor stage, ER/PgR status, and year of diagnosis (OS:HR = 5.47, 95% CI.2.01–14.91; PRS: HR = 3.22, 95% CI 1.18–8.77).

In a study from Dieci et al.⁸ of 119 patients with recurrent breast cancer, PRS and OS were shorter in patients with loss of either ER or HER2. Median PRS for HER2 loss vs. HER2+ concordance was 16 vs. 88 months, p = 0·008, and median OS 40 vs. 108 months, p = 0·06 for HER2 loss vs. concordant HER2 positivity, respectively. Similarly, ER loss compared with concordant ER positivity was associated with poorer OS (median 59 vs. 130 months, p = 0·001). PgR loss did not affect survival⁸, consistent with other reports. ^{12, 30 and 36} Patients with HER2 loss at metastatic relapse continued anti-HER2 therapy, while it is not reported if patients with ER loss continued to receive endocrine therapy. Thus, poor outcome associated with HER2 loss cannot be attributed to withdrawal of anti-HER2 therapy in the setting of a false negative result.

Discordance based on tumor phenotype, classified as ER+/HER2−, ER+/HER2+, ER−/HER2+, and ER−/HER2− (triple negative, TN) utilising a cut-off for ER/PgR of 10%, was also reported, and was seen in 22% (n = 27), including nine patients with TN disease at recurrence. Both PRS and OS were poorest in patients with discordance that altered tumor phenotype to TN (PRS: HR 4.35, p < 0.0001; OS: HR 2.70, p < 0.007, for TN vs. other discordant cases, respectively).⁸ Liedke et al. similarly reported poor outcomes for discordant TN disease in their study of 528 paired primary and metastatic breast cancers; PRS was 45 months vs. 25 months and 16 months for receptor positive vs. concordant TN and discordant TN, respectively). For patients with distant metastases as first site of recurrence, discordance was associated with a significantly higher risk of a PRS event compared with concordant disease (HR = 1.92, p = 0.0014).³⁷ Worse survival in the setting of TNBC at relapse was reported in a study from Montagna et al., although the prognostic impact of discordance itself was not specifically addressed.¹⁶ In these latter two studies, the definition of triple negativity similarly used a cut-off of 10% for ER/PgR.

In one of the largest studies to date, Lindstrom et al. reported a retrospective analysis of 459 patients, finding that ER loss was associated with poorer OS compared with concordant ER + disease (HR = 1.48, 95% CI 1.08–2.05).¹³ However, data on HER2 status was only available for 104 patients, where a 17% discordance rate was seen, and no details were given on the prognostic impact of this HER2 discordance.

In contrast, a prospective trial from Amir et al.³, which included 121 patients with paired biopsies, reported HER2, ER and PgR discordance rates of 16%, 40%, and 10%, respectively. When considering patients with any receptor discordance (i.e. HER2, ER, or PgR) compared with patients with unaltered receptor status, post hoc analyses demonstrated no difference in either OS or time to treatment failure in patients with discordant, compared with concordant, HER2 expression (median OS 27.6 vs. 30.2 months for concordant vs. discordant groups, respectively, HR = 0.94, p = 0.85). With survival outcomes only reported for a combined group of ER/PgR and/or HER2 discordance, the relative impact of HER2 discordance on prognosis is uncertain.

Disease free survival (DFS), in contrast to PRS, is not influenced by treatment alterations based on the rebiopsy result. Four studies which included information of time to disease relapse reported no association between HER2 discordance rate and DFS^{9, 10, 18 and 21} (Table 3).

While some evidence suggests worsened PRS and OS with HER2 discordance, in particular HER2 loss, no firm conclusions are possible, as the vast majority of data are confounded by variable treatment alterations based on discordant biopsy results.

Clinical impact of discordance

Two aspects relating to the clinical impact of discordance were considered: firstly, whether a finding of discordance altered management of metastatic disease (Table 4), and secondly, whether altered management improved patient outcomes (Table 5).

Change in treatment due to rebiopsy

This clinical question is best assessed in a prospective trial, although several retrospective studies also report treatment alterations due to receptor discordance.

In a prospective pilot study from Simmons et al.³⁸ that enrolled 40 patients, 35 underwent biopsy, 29 of whom had suitable samples for analysis. Three biopsies revealed benign disease, one patient was diagnosed with follicular lymphoma, and the remaining 25 biopsies confirmed recurrent breast cancer. Two patients (8%) had gain of HER2, while three (12%) and seven (28%) patients had loss of ER and PgR, respectively. As receptor discordance was reported in a total of 10 patients (40%), two patients presumably had change in more than one receptor, although this is not specifically reported.

Alternations in treatment were at the discretion of the treating physician. Treated was altered in six patients (20% of those with evaluable biopsy, 17% of all patients with biopsy performed). Both patients with HER2 gain received trastuzumab or clinical trial therapy, while four patients with diagnosis of benign disease or second malignancy had their breast cancer management reverted back to the previous plan. Treatment alterations in patients with hormone receptor discordance (receptor expression loss in all cases) were not specifically reported. Reasons attributed to lack of change of management included factors such as tumor characteristics, patient wishes, and clinician acceptance of the biopsy result.

Following on from this pilot study, Amir et al.³ prospectively evaluated rebiopsy discordance rates and feasibility in a study of 121 patients with suspected recurrent breast cancer. Discordance of receptor status was seen in 61 of 121 patients, including 23 patients with alteration in either ER or HER2, the more clinically relevant receptors. Treatment decisions were again made by the treating physician, with management altered in 17 of 121 patients (14%), including six patients with HER2 gain all receiving trastuzumab. Two of four patients gaining ER received endocrine therapy, five of 11 patients with ER loss were treated with chemotherapy instead of endocrine therapy, and four patients with biopsies revealing an alternate diagnosis (benign disease n = 3, basal cell carcinoma n = 1) had management for breast cancer reverted back to follow-up. Treatment was not altered despite ER or HER2 discordance in ten patients, although reasons for this were not discussed. Additionally, feasibility of biopsy was assessed in terms of patient satisfaction, rebiopsy yield, safety and evaluation of time delays in seeking rebiopsy. From the 117 biopsies confirming recurrence breast cancer, 20% were unsuitable for IHC, while 29% were not analyzed by FISH. The median time delay for biopsy was 15 days (range 2–56 days). The only serious adverse event reported was a hospital admission for a bleed secondary to a punch biopsy, which was treated with conservative measures. Patient satisfaction was assessed by survey responses in 90 women pre- and post-procedure. Interestingly, despite a third experiencing anxiety pre-biopsy, and over one quarter experiencing pain of at least moderate severity associated with the procedure, the vast majority (88%) reported that they would recommend rebiopsy to others.

The prospective BRITS study²¹ enrolled 205 patients for rebiopsy at time of suspected breast cancer relapse. Two patients were excluded due to safety concerns, 35 were ineligible, and 18 were found not to have recurrent disease after rebiopsy. From 150 paired breast cancer biopsies insufficient PBC or metastatic tissue was available for nine and four patients, respectively. Receptor discordance rates in the remaining 137 cases were 10%, 25% and 3% for ER, PgR and HER2, respectively. Gain of HER2 was seen in three patients compared with one with HER2 loss. Alteration from ER negative to positive and ER positive to negative was seen in three and 11 patients, respectively. One patient had both loss of ER and HER. Changes in HER2 or hormone receptor status (incorporating expression of both ER and PgR) were deemed necessary to alter management. Thus, in the opinion of the accruing clinicians, 24 out of 137 patients (17%) had a clinically significant change in receptors. Alteration in treatment was planned for these 24 patients, although it is unclear whether all treatment changes were put into effect. Furthermore, five of these 24 patients appear to have had loss of PgR with persisting ER positivity and unchanged negative HER2 status, indicating that endocrine therapy may still have been appropriate treatment.

Several retrospective studies report alterations in treatment in some, but not all, patients with HER2 or ER/PgR discordance^{5, 7, 9, 10, 17, 22, 26 and 39} (Table 4), although as is the nature of retrospective studies, rationale why patients did or did not receive targeted therapy in the setting of discordance is unknown and potentially due to any number of factors, including patient fitness, extent of disease, availability of targeted therapy, or physician concern of an inaccurate rebiopsy result.

Importantly, in all studies, retrospective and prospective alike, treatment decisions have been made at the discretion of the treating clinician, confounding the results. Nonetheless, from the available data, receptor discordance appears not to always lead a change in management.

Response to targeted therapy

The effect of HER2 discordance on treatment response has been addressed in several studies, though findings are largely limited to case reports from retrospective cohorts.

Seven studies report data that indicates potential trastuzumab activity in the setting of HER2 gain,^{3, 5, 6, 10, 26, 35 and 39} (Table 5). From Amir et al.³, six patients had HER2 gain, all of whom received trastuzumab, while two of four patients with ER gain received ET. At 12 months follow-up no difference was seen in either time to treatment failure or OS between patients with or without ER/PgR or HER2 discordance. This may suggest an effect from alteration in treatment, although the true treatment effect cannot be verified due to small sample size and non-randomised treatment alterations.

Zidan et al.²⁶ retrospectively compared paired biopsies of PBC and metastases from 58 patients, demonstrating a HER2 discordance rate of 14%. One patient lost HER2 over-expression while seven patients gained HER2 positivity. Of these seven patients, three had died without receiving trastuzumab, while four patients were treated with trastuzumab-based therapy, with one complete response (trastuzumab + vinorelbine), one partial response (single-agent trastuzumab), one ‘minimal response’ (single-agent trastuzumab), and one patient with disease progression (regimen not reported).

In a retrospective study from Guarneri et al.³⁹ that included 75 women with paired primary and metastatic breast cancer biopsies, seven of 10 patients with HER2 gain received antiHER2 therapy, with median duration of therapy of 6 months (range 2–108 weeks), indicating that some patients at least derived clinical benefit from this therapy. Similarly, three of seven patients with ER gain received endocrine therapy, for a median duration of two years (range 48–128 weeks). Outcomes of untreated patients with receptor discordance were not reported.

Two studies suggest lack of trastuzumab response in patients with HER2 loss.^8 and 17 From Dieci et al.⁸, HER2 loss (n = 4) was associated with poorer outcomes despite on-going trastuzumab. Similarly from Niikura et al.¹⁷, no trastuzumab benefit was seen in the few patients (n = 7) with HER2 loss who continued this treatment.

In a study from Botteri et al.⁵ improved OS was evident in patients with receptor (HER2 and/or ER) gain compared with HER2 loss, suggesting a possible treatment effect. However whether patients with receptor loss continued to receive targeted therapy is not reported.

In a large retrospective study from Lower et al.³⁵, HER2 expression of paired biopsies was assessed in 382 patients with 90 patients losing HER2 expression, 37 patients gaining HER2, and 50 patients with concordant HER2+ disease. No patient had received adjuvant trastuzumab, while 125 patients were treated with trastuzumab in the metastatic setting, based on HER2 status of either primary or metastatic disease. Crucially, specifically which patients received trastuzumab was not reported. OS from time of primary diagnosis was significantly prolonged with HER2 gain, while no survival difference was seen when trastuzumab-treated patients were excluded, possibly reflecting effective trastuzumab treatment. However validity of this conclusion is uncertain given the limited details regarding allocation of trastuzumab treatment. Furthermore, methodological issues have likely contributed to high discordance rate (34%), with HER2 assessed by IHC only, and retrieved from case notes for both PBC and metastases.

Interestingly, two studies report an apparent differential treatment effect from trastuzumab in patients with gain compared with HER2+ concordant disease. In a study of 56 patients from Chang et al.⁶, HER2 status of PBC and relapsed disease demonstrated an overall discordance rate of 13%. Five of 41 patients converted from HER2− to HER2+, all of whom received trastuzumab-based therapy, two with partial response, two with stable disease, and one with progressive disease. OS for HER2+/+ disease was numerically longer than for HER2−/+, 61 vs. 52 months, respectively, though this was not statistically significant. For patients treated with trastuzumab, those with HER2+/+ disease had significantly better OS than those with HER2−/+ disease (median OS: 32 vs. 11 months, p = 0·023). A similar response pattern was seen in a study of 137 patients from Fabi et al..¹⁰ Discordance was seen in 10%; two patients with HER2 loss and 12 patients with HER2 gain. No patient received adjuvant trastuzumab, while trastuzumab was given to 18 of 35 patients with HER2+ metastases, including six of 12 patients with HER2 gain. Rationale for why some patients did not receive trastuzumab was not reported. There was a numerically greater, but not statistically significant, time to progression for trastuzumab-treated HER2+/+ patients compared with treated HER2−/+ patients (11 vs. 8 months, respectively), while untreated HER2+/+ patients appeared to do worse than untreated HER2−/+ patients (2 vs. 5 months, respectively, p = not significant). Due to very limited patient numbers, these findings can be considered hypothesis-generating only.

Discussion

Many studies assessing receptor discordance between primary and metastatic breast cancers have by now been published and, although issues of methodology may limit the majority, a relatively consistent pattern of ∼5–10% HER2 discordance rate is seen. Gain and loss of HER2 is relatively equally reported, in contrast with ER/PgR, where loss is more common than gain (Supplementary Table S1). To date, little emphasis has been placed on the clinical implications of discordance.

The prognostic significance of HER2 discordance is best assessed in prospective trials; however, the few prospective trials to date utilized endpoints relating to alterations in treatment, rather than survival outcomes. The available evidence suggests that loss of HER2 and/or ER may result in worse PRS and OS, although data were not consistent across trials and were often confounded by lack of treatment in the setting of receptor loss. Conversely, HER2 discordance was not associated with shorter DFS.

Furthermore, there was no evidence that indicated specific clinical factors, such as previous treatment, might predict likelihood of HER2 discordance at breast cancer relapse. Although not specifically assessed in any of the identified studies, one factor worth considering is whether HER2 discordance may drive alteration in ER or vice versa, given the known interaction between HER2 and ER⁴⁰, and HER2 attributed as a driver of tamoxifen resistance.^{41, 42 and 43} Comparing the rate of HER2 discordance rate with the corresponding rate of ER discordance in 29 studies that reported alterations in both receptors^{3, 4, 5, 7, 8, 9, 11, 12, 13, 14, 15, 16, 18, 19, 20, 21, 23, 28, 36, 37, 38, 39, 44, 45, 46, 47, 48, 49 and 50}, we found considerable variability (Fig. 1). Notably, any analytical or preanalytical errors leading to increased HER2 discordance will likely have also affected ER discordance equally, confounding this comparison.

A minority of studies (n = 9) also assessed alterations in phenotype; seven used combined assessment of ER and HER2 ^{5, 12, 13, 14, 18, 20 and 21}, one study used ER, HER2 and Ki67¹⁶ and one reported alterations in phenotype, but did not report ER and HER2 status of individual tumors.⁸ Definitions for ER positivity varied from IHC expression ⩾1% ^{5, 12, 18 and 20}, ⩾10%¹³, to Allred score >2 ^14 and 21, while the definition of HER2 positivity was similar across studies. From the seven studies assessing phenotype by ER and HER2 only, receptor status of 122 individual tumors were available. More tumors had alteration in ER and stable HER2 stable (n = 67), or stable ER and discordant HER2 (n = 42), than change in both receptors (n = 13) (Fig. 2). A similar pattern was seen when the additional 30 cases from the study of Montagna et al.¹⁶, which utilised Ki67 to define phenotypes, were included. No clear association between HER2 and ER discordance is evident from these analyses, although they are limited by small numbers, and the fact that data were extracted from diverse retrospective studies. Given the known interaction between HER and ER, further exploration of whether this interaction influences HER2 and ER discordance might be of interest.

Overall, this literature review demonstrated that data relating to the clinical impact of discordance are limited, inconclusive and conflicting; however, alternate endpoints may have relevance in clinical practice. From prospective studies, rebiopsy to assess for discordance appears feasible, with biopsy yield of over 80%, serious complication rate low, and patient satisfaction high.^{3, 21 and 38} Furthermore, treatment alterations based on the rebiopsy result were reported for 1 in every 6 to 7 patients undergoing the procedure^{3, 21 and 38}, a finding that has been proposed as rationale for recommending routine rebiopsy in patients with relapsed breast cancer. Importantly though, this endpoint is subject to selection bias, with patients more likely to be approached for enrolment if fitter, or if the diagnosis was uncertain, and physician bias, with physicians being more likely to recommend rebiopsy if they felt it would influence their management decisions. Thus, as the baseline threshold for altering treatment can vary between physicians, it is challenging to apply such results across all patients and practices.

Moreover, while rebiopsy can lead to change in management, this does not occur in all patients with receptor discordance. A contributing factor may be physician concern of a false negative or false positive result. From the trial from Simmons et al., lack of acceptance of the biopsy result was a reason attributed to not changing management despite discordant receptor status.³⁸ In a pooled analysis of the two largest prospective trials to date^3 and 21, Amir et al.⁵¹ further suggested that a change in treatment was more common in the setting of apparent receptor gain than loss. Data in support of this assertion are however limited. From the pooled analysis no statistically significant association was seen between the probability of change in therapy and the receptor profile of the primary tumor, while data relating to comparison of treatment change in the setting of receptor loss compared with gain were not entirely consistent with reported discordance rates.⁵¹ Nonetheless, it is reasonable to consider that the risks associated with continuing a targeted therapy despite apparent loss of receptor expression would be lower than the risk of ceasing a potentially effective therapy. Thus if targeted therapy will be trialled regardless of the rebiopsy result, receptor positive PBC could benefit less from rebiopsy.

Lastly, a small number of studies have reported that individual patients with receptor gain appear to benefit from addition of targeted treatment. Of interest are the, albeit limited, data suggesting a possible differential treatment effect of trastuzumab for HER2 gain, compared with HER2+ concordance.^6 and 10 One explanation for this could be heterogeneity of HER2 expression in differing metastatic sites; i.e. simultaneous HER2+ and HER2− metastases. Discordance between asynchronous metastatic biopsies has been reported in studies from Gancberg et al., Lindstrom et al., and Niikura et al., with discordance in HER2 expression in three of 16 patients (19%)⁵², five of 32 patients (16%)¹³, and two of 19 patients (11%)¹⁷, respectively. Although small numbers and methodological issues limit these findings, further investigation of potential tumor heterogeneity, simultaneous HER2+ and HER2− metastases, and the impact on treatment response would be of interest.

To biopsy or not to biopsy?

The decision to rebiopsy or not is a commonly encountered dilemma in breast cancer management, and one that must currently be made despite the paucity of evidence that routine rebiopsy improves patient outcomes. Furthermore, it is critical to recall that, while complication rate is relatively low, tissue biopsy is not a benign procedure. Considering these factors we have developed an algorithm (Fig. 3), based on findings from this literature review and practical issue such as site of recurrence, or availability of PBC results, which we are now using at our institution to guide rebiopsy decisions. This algorithm is not meant to be prescriptive; instead it suggests a pragmatic approach to rebiopsy that has been of use in the management of our breast cancer patients. For instance, such clinical scenarios where rebiopsy might be less feasible include positive PBC receptor status, where risks of a false negative result seem to outweigh the risks of continuing a potentially ineffective therapy, bone-only metastases, as the yield from bone biopsies is relatively low³, limited interventional radiology resources, and concern regarding delays in treatment commencement. Absent or poor quality PBC results do not necessarily imply need for rebiopsy, as reassessment of the PBC might provide adequate information for treatment decisions, especially if receptor expression is positive. Patient choice should also factor into the decision.

Finally, emerging techniques, such as receptor expression analysis of circulating tumor cells^53 and 54 are being developed and have shown early promise as a means of determining tumor receptor status based on a ‘liquid biopsy’. Considerable additional development and validation of these techniques are required before incorporation into routine clinical practice; however such techniques would vastly improve feasibility, safety, and rapidity of tumor receptor evaluation, thus removing some major barriers of rebiopsy that currently exist.

Conclusions

Currently, the best management approach for receptor discordance between primary and metastatic disease is unknown, and the very limited evidence of alteration in clinical outcomes based on rebiopsy does not seem strong enough to confirm rebiopsy as essential in every patient. We would suggest that before routine rebiopsy can be recommended and incorporated into practice guidelines, further research regarding the impact of HER2 discordance is needed, ideally through prospective trials.

Conflict of interest statement

The authors do not have any conflict of interest to declare. They do not have any financial or personal relationships with other people or organizations that could inappropriately influence or bias their work.

References