Part VI · Heterogeneity, evolution, and metastatic biology · Chapter 29

Receptor discordance and conversion

A changed result has three possible causes, and only one of them is the tumour.

1 · Discordance rates for ER, PR, and HER2 between primary and metastasis

The largest synthesis is a meta-analysis of 39 studies comparing primary breast tumours with paired distant metastases1. Its results are reported in two directions, and the two have different denominators.

Among tumours positive in the primary, pooled conversion to negative in the metastasis was 22.5% for ER, with a 95% confidence interval of 16.4% to 30.0%. It was 49.4% for PR, with an interval of 40.5% to 58.2%. It was 21.3% for HER2, with an interval of 14.3% to 30.5%.

Among tumours negative in the primary, pooled conversion to positive was 21.5% for ER, 15.9% for PR and 9.5% for HER2. The corresponding intervals were 18.1% to 25.5%, 11.3% to 22.0% and 7.4% to 12.1%.

Those two sets of percentages are taken over different populations. Averaging them, or quoting one as the discordance rate, produces a number that means nothing.

A large single-region cohort gives the same picture from a different design. Among patients in Stockholm relapsing between 1997 and 2007, ER status changed between primary and relapse in 32.4% of 459 patients, PR in 40.7% of 430, and HER2 in 14.5% of 1042. The HER2 change was not statistically significant in that cohort. Across patients with two to six consecutive relapses assessed, the markers altered in 33.6% of 119 for ER, 32.0% of 116 for PR and 15.7% of 32 for HER2.

A prospective study reported lower figures, with discordance of 16% for ER, 40% for PR and 10% for HER2 among 121 women biopsied at suspected metastatic recurrence3.

Three observations hold across all three designs.

PR is the least stable marker everywhere. That is partly biology, because PR is an ER-regulated gene and therefore reports pathway activity rather than a fixed identity. It is partly assay, because PR staining is more sensitive to preanalytic handling, as set out in Preanalytics, fixation, and the invisible sources of error.

HER2 is the most stable. It is the only one of the three with a bimodal readout backed by a DNA measurement.

The between-study spread is itself a finding. The range in the meta-analysis was wide, and True biological conversion, assay variability, and sampling error is about where that range comes from.

2 · True biological conversion, assay variability, and sampling error

A changed receptor result has three possible causes. Separating them is the whole discipline of this chapter.

The first is assay variability. The two specimens were usually run in different laboratories, at different times, with different antibodies and different fixation. One concrete example is worth more than the general warning. Fragments of primary breast carcinoma were exposed to hydrochloric acid decalcification for one or five hours4. Mean ER, PR, HER2, Ki-67 and GATA-3 scores were all significantly lower than in formalin-fixed control fragments from the same tumours. Formic acid for one hour changed none of them significantly. A bone metastasis processed with hydrochloric acid can read as receptor-negative for a reason that has nothing to do with the tumour.

Thresholds contribute to the same cause. ER is positive at 1% to 100% of tumour nuclei staining and negative below 1%5. A separate ER Low Positive category covers 1% to 10%, because benefit at that level is poorly characterised. A specimen at 2% and a specimen at 0% differ by a small number of stained nuclei and by a category boundary.

The second cause is sampling error. The metastasis and the primary are different cells from different places. If the primary was spatially heterogeneous for that receptor, a changed result may mean only that a different compartment was sampled. Sampling limitations and what a single core can and cannot tell you sets out what a single core can and cannot establish, and Intratumoral variation as a source of discordant reports treats intratumoral variation as a source of discordant reports.

The third cause is true biological conversion. A population has genuinely lost or gained the receptor, usually because selection removed the compartment that expressed it. That mechanism is developed in Therapy as selective pressure, the central premise.

The order in which these can actually be addressed is the reverse of the order in which they are usually discussed.

Rule out the assay first, because it is the only one that can be tested directly. Ask what tissue was taken, whether it was decalcified and with what, and which laboratory ran each assay. Where it matters, request that the original block be rerun alongside the new specimen in one laboratory.

Then weigh direction and plausibility. Loss of ER after prolonged oestrogen deprivation is mechanistically expected. Gain of ER in a patient never exposed to endocrine therapy is less so, and should raise the assay question again.

Then weigh time and treatment. The synchronous nodal HER2 concordance of 95.28% in 148 patients is the closest available estimate of the floor, because almost nothing separated the two specimens except location6. The distance between that figure and the distant metachronous rates in Discordance rates for ER, PR, and HER2 between primary and metastasis is where time, treatment and site live.

Caution

A changed receptor result is routinely reported as conversion. Conversion is what remains after assay variability and sampling error have been excluded, and in most clinical reports neither has been excluded. Calling every change a conversion attributes to the tumour a difference that frequently belongs to the laboratory.

3 · Subtype switching, including luminal to HER2-enriched

Switching can also be assessed at the level of intrinsic subtype rather than individual markers. In 123 paired primary and metastatic tissues classified by a 50-gene assay, the rate of subtype conversion differed sharply by the subtype of the primary7. It was 0% for basal-like tumours, 23.1% for HER2-enriched, 30.0% for luminal B and 55.3% for luminal A. In 40.2% of cases a luminal A tumour converted to luminal B. In 14.3% of cases a luminal A or luminal B tumour converted to HER2-enriched.

Read the gradient rather than the individual figures. Basal-like tumours never switched. Luminal A switched most often. The direction of switching was consistently toward higher proliferation, and the metastases were enriched for proliferation-related and migration-related genes and depleted of luminal genes. That is not random drift.

Two interpretations fit the data and a paired sample cannot separate them. The metastasis-founding population may always have been the more proliferative compartment of the primary. Alternatively the phenotype may have shifted after seeding, under endocrine pressure or under the influence of the new site.

Subtype stability and clinically relevant change are also different questions. Among 20 matched primary tumours and resected brain metastases, 17 of the 20 metastases retained the subtype of the primary8. In the same 20 pairs, 17 showed expression changes of more than twofold in clinically actionable genes. ERBB2 expression rose more than twofold in 7 of the 20. Three of the 13 patients whose primary was HER2-negative had a paired brain metastasis scoring immunohistochemistry 3+, with amplification of the locus specific to the metastasis.

A metastasis can therefore keep its subtype label and still acquire a treatable alteration. The label was never the quantity that mattered.

4 · Discordance across metastatic sites within the same patient

Discordance is not uniform across sites, and the pattern has both biological and technical components that the published data cannot fully separate.

In the meta-analysis of paired primaries and distant metastases, ER discordance differed by site1. It was 20.8% in central nervous system metastases and 29.3% in bone, against 14.3% in liver. The respective 95% confidence intervals were 15.0% to 28.0%, 13.0% to 53.5% and 11.3% to 18.1%. PR discordance ran the other way by site. It was higher in bone at 42.7% and in liver at 47.0% than in central nervous system metastases at 23.3%.

Attributing that pattern to site biology would be premature. Bone specimens usually require decalcification, which suppresses receptor staining when acid is used4. Central nervous system specimens are typically resections rather than cores, are larger, and are obtained in a different clinical context. The site effect and the processing effect are confounded in every one of these series.

At least one site effect is not confounded in that way. In an expanded sequencing cohort, ERBB2 amplification or mutation was present in 13% of local disease against 24% of brain metastases8. That is a genomic measurement rather than a stain, and decalcification does not explain it.

The clonal basis sits underneath all of this. Where metastases descend from one founding population, one deposit represents the others reasonably well. Where they descend from separate subclones of the primary, it does not, and both patterns were seen in a rapid autopsy series9. Choosing the most accessible lesion to biopsy is therefore a decision with consequences, not a neutral convenience.

5 · Rebiopsy, when, which lesion, and when the result should change management

The prospective evidence is a single study and it is worth reading closely. Of 121 women undergoing biopsy of a suspected metastasis, receptor status could be analysed in 80%3. The treating oncologist recorded a treatment plan before the biopsy and again after. Management changed in 14% of women, with a 95% confidence interval of 8.4% to 21.5%. Fine needle aspiration and biopsy of bone both reduced the ability to analyse receptors. After a median follow-up of 12 months there was no trend toward an association between discordance and either time to treatment failure or overall survival.

Two readings of that last sentence are wrong. It does not show that acting on discordance is useless. It does not show that acting on it helps. A median follow-up of 12 months in 121 women cannot answer either question.

The practical decisions are which lesion and what to do with the answer.

In practice

Biopsy when the result could change the next decision. That is the recurrence itself, and a progression before a line of therapy whose selection is biomarker-defined. Biopsy and molecular profiling at recurrence and at progression places this in the wider profiling decision.

Prefer a lesion that can be processed without decalcification. Where bone is the only option, ask the laboratory for a formic acid or EDTA protocol rather than hydrochloric acid, and record which was used4.

Prefer a core to an aspirate. Receptor analysis failed more often after fine needle aspiration in the prospective series3.

Treat a gain and a loss asymmetrically. A negative-to-positive HER2 result opens an effective class, and the pooled base rate for that direction is 9.5%1. Acting on it is usually right.

Treat a loss more sceptically, particularly from a decalcified bone core in a patient whose primary was strongly positive. A false loss removes an effective and well-tolerated class. Rerun the original block alongside the new specimen before accepting it.

Do not reclassify on a single equivocal result near a threshold. A specimen at 2% ER and a specimen at 0% differ by a category and by very little else5.

6 · Prognostic meaning of conversion independent of any treatment change

Conversion is usually discussed as a trigger for a treatment change. The separate question is whether it carries prognostic information of its own.

In the Stockholm cohort, overall survival differed significantly according to the combination of primary and relapse receptor status2. Women whose ER-positive primary became ER-negative at relapse had a 48% higher risk of death than women whose tumours remained ER-positive. The hazard ratio was 1.48, with a 95% confidence interval of 1.08 to 2.05.

The inference has a limit that the study cannot remove. Women who lost ER also lost access to endocrine therapy. Part of the excess mortality is therefore the withdrawal of an effective treatment rather than a property of the tumour, and an observational cohort cannot separate the two contributions.

What can be said is narrower than it is usually stated. Conversion marks a population with worse outcome. Whether conversion causes the worse outcome, or whether both are consequences of the biology that produced the conversion, is not established.

The apparent conflict with the null prospective result is not a real conflict3. One study followed a large cohort to death. The other followed 121 women for a median of 12 months. They were not asking the same question at the same power.

Interplay

Receptor conversion is where the measurement problem and clonal evolution produce the same observation. A receptor result that has fallen is the tumour changing, a different part of the tumour being sampled, or the assay behaving differently. Those three explanations are the same three that govern HER2 loss in HER2 loss as an acquired resistance mechanism, antigen loss under conjugate exposure in Temporal heterogeneity and clonal evolution, and the discordance between blood and tissue in Liquid biopsy and longitudinal monitoring. Learning to rank them in one setting transfers directly to the others, which is why this chapter is placed before the management parts rather than inside them.

See Integrative biological interplay

7 · Reporting conventions and how to document a converted phenotype

Most of the confusion in this literature is recoverable at the point of reporting. A report that carries the following can be re-interrogated later. A report that does not cannot.

The specimen. Site, procedure, date and block identifier. Whether the tissue was decalcified, and with which agent and for how long4.

The assay. Antibody clone, platform, laboratory, and the scoring threshold applied. For ER staining between 0% and 10%, the status of the controls should be reported, and staining between 1% and 10% should carry the ER Low Positive category5.

The number, not only the category. A percentage of stained nuclei and an intensity, because a category conceals how far the result sits from the boundary.

Whether the prior specimen was retested alongside the new one, and where.

The prior result as it was originally recorded, with its own assay details rather than a summary of them.

The reason is arithmetic. The difference between 0% and 2% ER staining is one category boundary and a handful of nuclei. The difference between 2% and 90% is not. A record that says positive in both cases has erased the distinction that a later reader will need.

The convention worth adopting is to record conversion as a comparison between two named specimens assayed in named ways. A note that a liver core is ER-negative, having been ER-positive at 90% on the primary resection, is auditable by anyone who reads it later. A note that the patient converted is not, and it will outlive the evidence that produced it.

References

  1. Schrijver WAME, Suijkerbuijk KPM, van Gils CH, et al. Receptor conversion in distant breast cancer metastases: a systematic review and meta-analysis. J Natl Cancer Inst 2018 110:568-580. PMID 29315431
  2. Lindstrom LS, Karlsson E, Wilking UM, et al. Clinically used breast cancer markers such as estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 are unstable throughout tumor progression. J Clin Oncol 2012 30:2601-2608. PMID 22711854
  3. Amir E, Miller N, Geddie W, et al. Prospective study evaluating the impact of tissue confirmation of metastatic disease in patients with breast cancer. J Clin Oncol 2012 30:587-592. PMID 22124102
  4. Clark BZ, Yoest JM, Onisko A, Dabbs DJ. Effects of hydrochloric acid and formic acid decalcification on breast tumor biomarkers and HER2 fluorescence in situ hybridization. Appl Immunohistochem Mol Morphol 2019 27:223-230. PMID 28877070
  5. Allison KH, Hammond MEH, Dowsett M, et al. Estrogen and progesterone receptor testing in breast cancer: ASCO/CAP guideline update. J Clin Oncol 2020 38:1346-1366. PMID 31928404
  6. Ieni A, Barresi V, Caltabiano R, et al. Discordance rate of HER2 status in primary breast carcinomas versus synchronous axillary lymph node metastases: a multicenter retrospective investigation. Onco Targets Ther 2014 7:1267-1272. PMID 25050068
  7. Cejalvo JM, Martinez de Duenas E, Galvan P, et al. Intrinsic subtypes and gene expression profiles in primary and metastatic breast cancer. Cancer Res 2017 77:2213-2221. PMID 28249905
  8. Priedigkeit N, Hartmaier RJ, Chen Y, et al. Intrinsic subtype switching and acquired ERBB2/HER2 amplifications and mutations in breast cancer brain metastases. JAMA Oncol 2017 3:666-671. PMID 27926948
  9. Avigdor BE, Cimino-Mathews A, DeMarzo AM, et al. Mutational profiles of breast cancer metastases from a rapid autopsy series reveal multiple evolutionary trajectories. JCI Insight 2017 2:e96896. PMID 29263308