Part X · Metastatic disease: principles and cross-cutting management · Chapter 54
General principles of metastatic management
Why the first metastatic question is which dependency is failing rather than which agent is next.
1 · De novo and recurrent metastatic disease as different populations
Two patients can present with identical imaging and be different problems. One has metastatic disease at first diagnosis, while the other relapsed after adjuvant treatment. The stage is the same and the prior history is not.
De novo disease has never been exposed to selection. Nothing has been removed from the tumour population by treatment, so the sensitivity profile is closer to the untreated state. Recurrent disease is what survived adjuvant therapy, which makes it a selected population by definition. What selected it is written down in the chart.
Two facts from the adjuvant record therefore carry information. The first is which agents were given. The second is how long the disease-free interval was. Relapse during adjuvant endocrine therapy and relapse eight years after completing it are different statements about dependence. The annual hazard of recurrence in hormone receptor positive disease stays low and persists for well over a decade1. Late relapse is common enough that a long interval should not be read as proof of a dormant, indolent biology.
Registry comparisons showing longer survival for de novo than for recurrent disease are not comparing treatments. They are comparing a treatment-naive population with a selected one, over different lead times, with different competing exposures. The difference is real and it does not support any inference about what to give.
2 · Disease burden, tempo, and symptom-driven decisions
Three variables govern the urgency of a metastatic decision and they are routinely collapsed into one. Burden is how much disease is present. Tempo is how fast it is changing. Symptoms are what the patient feels. They are independent, and they can point in different directions.
Large burden with slow tempo and no symptoms does not require an urgent regimen. Small burden with rapid tempo and organ dysfunction does. The common error is to let a bulky scan drive the choice when the rate of change is the variable that predicts what happens over the next three months.
Tempo is the one that is least often recorded. It needs two measurements and a stated interval, which a single baseline scan cannot supply. Where an outside scan exists, obtaining it is usually worth more than ordering a new modality.
Where tempo is genuinely fast, the question becomes whether the response has to be cytotoxic. That is a question about how quickly each option works rather than about drug class, and it is argued with its evidence in Liver and lung disease.
In practice
Before choosing a first metastatic regimen, write down four things.
Where the disease is, and which sites are doing the work of making the patient unwell.
How fast the burden changed between the last two assessments, with the interval named.
What the patient reports, separately from what the imaging shows.
What organ function is threatened now, as against what may be threatened in a year.
If tempo and burden disagree, tempo usually determines the next three months. Burden determines what the disease will look like when the first regimen fails.
3 · Biopsy and molecular profiling at recurrence and at progression
A biopsy at recurrence answers three separate questions, and confusing them produces most of the argument about whether it is worth doing. The first is whether the lesion is breast cancer. The second is what its receptor status is now. The third is what actionable alterations it carries.
Tissue confirmation of metastatic disease changes management in a proportion of patients, mainly through receptor results that differ from the primary4. Whether a difference is true conversion or assay variation is the subject of Receptor discordance and conversion and is not settled by repeating the test.
Plasma adds a different kind of information, because it samples what is shedding rather than what is accessible to a needle. That makes it better at breadth and worse at certainty. Acquired ESR1 mutations are the clearest example, since they appear in patients previously exposed to an aromatase inhibitor and predict failure of further oestrogen deprivation5. Two technical requirements apply. Matched white blood cell sequencing keeps clonal haematopoiesis out of the result6. And a negative plasma test in low-volume disease is uninformative rather than reassuring, for the shedding reasons in Liquid biopsy and longitudinal monitoring.
Whether plasma can stand in for tissue has been asked as a trial question rather than a technical one. The story below states what those studies were built to find out.
Advanced disease · HR+/HER2- · later
Molecular exceptions to the endocrine sequence
Can these alterations be found in the clinic without a fresh biopsy at every progression?
Rationale
Each exception here is rare. Most of the alterations arise under treatment. The archival block from the original operation is therefore the wrong sample to interrogate. None of them is reported by the immunohistochemistry that assigns receptor status. The practical limit on all of this is the test rather than the drug.
Experiment
A platform trial registered patients on circulating tumour DNA, checked digital PCR against targeted sequencing and both against tissue sequencing, then routed patients into treatment cohorts by the mutation found. One of those cohorts matched a HER2 mutation to neratinib, which is the population the mutation-selected trials above recruited from tissue. Three of those trials carried a plasma measurement of their own alongside the tissue that decided eligibility: acquired HER2 and PIK3CA alterations at progression on the neratinib triplet, repair deficiency scores in plasma as well as in tumour on talazoparib, and serial tumour fraction during treatment in the burden-selected study.
Finding
Plasma agreed closely with tissue where the biopsy was contemporaneous. Routing on the plasma result produced responses in the matched cohorts, at rates that differed between cohorts rather than describing the strategy as a whole. The plasma measurements carried alongside the mutation-selected trials tracked what happened on treatment, with a fall or no change in tumour fraction marking benefit. Plasma is workable as the sample that nominates these patients.
Limitation
Agreement with tissue was measured where the biopsy was contemporaneous, which is the favourable case. Sensitivity falls when the tissue is older. Older tissue is the situation the plasma test exists to solve. The cohorts were single-arm. A matched response rate says who to treat next, not whether the match beat the alternative. Plasma reports what a tumour sheds. A negative result in low-shedding or bone-only disease is not evidence that the alteration is absent. None of the mutation-selected trials in this story required the alteration to be found in plasma. The eligibility their results describe is a tissue result.
Next question
A patient can carry more than one of these alterations. Each trial that nominated one made no provision for the others. Which target is acted on first when two are found, and does the order change what the second one is worth?
plasmaMATCHSUMMITTalazoparib Beyond BRCANIMBUS
Four studies stand behind that question. Not one of them randomised a patient between a plasma result and a tissue result. Read as a class they agree on a single point. A molecular result can nominate a patient for a drug. The drug then does something in the group so nominated. What they do not share is the job the assay was given. That difference is one of design rather than of biology.
plasmaMATCH put the test itself on trial. Patients were registered on a circulating tumour DNA result. Digital PCR was checked against targeted sequencing, and both were checked against tissue sequencing. The plasma call then decided which treatment cohort a patient entered7. It agreed closely with tissue sequencing where the biopsy was contemporaneous. Patients routed on it responded, at rates that differed between the cohorts rather than describing the strategy as a whole. The other three accepted a tissue result at entry without requiring one. SUMMIT also accepted a circulating tumour DNA result, NIMBUS a blood-based panel and Talazoparib Beyond BRCA a germline test, and each carried a plasma measurement alongside the entry result. SUMMIT matched an activating HER2 mutation to neratinib and tracked acquired HER2 and PIK3CA alterations as the triplet failed8. Talazoparib Beyond BRCA scored repair deficiency in plasma as well as in tumour9. NIMBUS followed tumour fraction serially through treatment10. Only in the first did every patient enter on a plasma result, which is why the practice rests on that trial and not on the other three.
The markers themselves also differ in kind, and the difference is one of measurement rather than of biology. An activating point mutation is a present-or-absent call, and a laboratory can be asked how often it calls correctly against tissue. A repair deficiency score and a mutational burden are continuous quantities, and neither selects anyone until it is cut somewhere. A cut nominated in a small correlative subgroup describes that subgroup before it describes the assay. Neither of those two trials was built to validate one.
What none of this settles is whether routing on plasma beats routing on what the clinic would otherwise do. The plasma-routed cohorts were single-arm, so a matched response rate says whom to treat next and not whether the match earned the slot. SUMMIT is the one exception and a partial one. It randomised a small substudy against fulvestrant with and without trastuzumab, which ranks the drugs given to a mutation-selected group. It does not rank the sample the mutation was found in. That agreement with tissue was also measured where the biopsy was contemporaneous, which is the favourable case rather than the case the plasma test exists to solve. Sensitivity falls as the archival block ages. A negative result in bone-only or low-shedding disease reports what the tumour sheds and not what it carries Liquid biopsy and longitudinal monitoring. Apart from plasmaMATCH, none of the mutation-selected trials required the alteration to be present in plasma. The eligibility their response rates describe is not a plasma result alone, which matters on the day a plasma-only finding is used to start one of these drugs.
One gap here is not partly answered but wholly unstudied. A patient can carry two of these alterations at once, and each trial nominated one marker and made no provision for the rest. Which target to act on first has never been asked. Nor has what the second is worth once the first has been spent.
The organising question is not which drug is best. It is which dependency is currently load-bearing, and which of the available agents attacks it.
That reframing has a practical consequence for how this part of the book is organised. Treatment is written as a set of branch points rather than as a numbered sequence. The branches that actually change the answer are ESR1 status, PIK3CA, AKT1 and PTEN status, HER2 expression level, and germline BRCA status. A new agent slots into a branch rather than renumbering a list.
Three rules follow. Each is worth stating plainly.
Match the agent to the failure, not to the line. Progression on endocrine therapy after four years and progression after four months are different failures, and only one of them suggests the class was never right. The distinction is developed in A working taxonomy, intrinsic, adaptive, and acquired resistance.
Do not assume cross-resistance from a shared indication, because two drugs that treat the same disease can fail for unrelated reasons. Where dependence is genuinely shared, cross-resistance follows. Where it is not, progression on one predicts little about the other Cross-resistance between classes and what it means for sequencing.
Spend the effective agents while the patient can tolerate them. Sequencing plans that reserve the best option for later assume a later that many patients do not reach.
That third rule is the one a randomised trial has tested directly. The test was run on the class where reserving looked most defensible. SONIA randomised the order of CDK4/6 inhibition rather than the drug, giving it first with an aromatase inhibitor or holding it for second line with fulvestrant11. Its endpoint was progression across two lines rather than one, which is the only endpoint that can adjudicate an order. Overall survival was reported later and showed no difference between the two orders12.
Read that as a statement about what a sequencing trial can measure. Deferring the drug was not shown to cost survival. Giving it first bought a much longer total exposure to the inhibitor and a higher count of grade 3 or worse adverse events. The trial also assumed that a second line would be reached. Neither finding transfers to a patient whose disease is unlikely to allow a second line at all, which is the patient the third rule is written for.
5 · Response assessment and patterns of progression
Response criteria were built to compare arms in trials. They are used in clinic to decide whether one person should change treatment. Those are different jobs and the tool fits the first better than the second. The lineage runs from the World Health Organization's reporting recommendations13 to the revised RECIST 1.1 criteria, which rest on a data warehouse of more than 6,500 patients14.
Bone-predominant disease is the standard difficulty, and it fails in both directions. Sclerotic change in a healing lytic lesion can appear as a new finding on computed tomography. Increased tracer uptake on bone scintigraphy reflects osteoblastic activity rather than tumour cell number. Both are pictures of response that read as progression Bone metastases and skeletal-related events.
Immunotherapy adds a second problem, because tumour burden can increase before it falls. That is why the response framework for these agents requires confirmation on a later scan rather than action on a single assessment15.
The pattern of progression matters as much as the fact of it. Progression in one site while everything else is controlled is a different event from progression everywhere. The first is addressed in Oligometastatic and oligoprogressive disease. The second means the regimen has failed.
The two terms describe different situations and the shared prefix hides that.
Oligometastatic disease means few deposits at presentation or at relapse. The definition rests on a count, and a count is a property of the imaging used to make it. Sensitive imaging moves patients out of the category without changing their biology. That is a measurement statement, not a biological one, and it should be said out loud before anyone is offered ablative therapy on the strength of a number of lesions.
Oligoprogression means most sites are controlled and one or a few are not. Here the inference is about clonal structure rather than about total burden. A single progressing deposit while the rest responds suggests a resistant subclone in one place, which is exactly the picture developed in Temporal heterogeneity and clonal evolution. Treating that site locally and continuing the systemic regimen is a reasonable response to that inference.
The honest position on the evidence is that it is thin. Two randomised tests are usually quoted, and they are usually read as pointing in opposite directions.
The randomised evidence for treating metastatic disease locally with the intention of changing its course.
Read the population and endpoint columns before the result column.
de novo metastatic breast cancer with an intact primary tumour and no progression after 4 to 8 months of systemic therapy; 390 enrolled and 256 randomised
overall survival
3-year overall survival 68.4% with early locoregional therapy against 67.9% without, hazard ratio 1.11 (90% CI 0.82-1.52), p=0.57. Locoregional progression at 3 years was 16.3% with locoregional therapy against 39.8% without
Patients with a controlled primary malignancy and one to five metastatic lesions, all amenable to stereotactic ablative radiotherapy. The commonest primaries were breast, lung, colorectal and prostate, with 18 of the 99 patients entered with breast cancer.
Overall survival, tested at an alpha of 0.20 in a randomised phase 2 screening design.
Five-year overall survival was 42.3% with added ablative radiotherapy against 17.7% with standard care alone, stratified log-rank P = .006. Five-year progression-free survival was 17.3% against 3.2% at four years, P = .001. No new grade 2 to 5 adverse events occurred.
They do not point in opposite directions, because they were not aimed at the same target. They separate on which site was treated, and on what each was built to detect there.
EA2108 treated the intact primary and left the metastatic deposits to systemic therapy16. SABR-COMET did the reverse, ablating every metastatic lesion in patients whose primary malignancy was already controlled17. A trial of the breast and a trial of the deposits are not two answers to one question, but one answer each to two different ones.
Three further differences make the contrast a matter of design rather than of radiotherapy. The breast cancer trial randomised only patients whose disease had not progressed on months of systemic therapy. Its population is therefore selected for systemic sensitivity. Its negative result answers a narrower question than the one usually asked of it. SABR-COMET enrolled four common primary tumours together. Most of its patients did not have breast cancer, so a breast-specific estimate from it is a subgroup result in a trial that was never powered for one. SABR-COMET also used a screening design, whose looser significance threshold exists to decide whether a phase 3 trial is worth running rather than to license practice. It is quoted as though it had been that phase 3 trial.
The case for operating on the primary came from registries. In a national database, women with stage IV disease whose primary was excised with free margins had a hazard ratio for death of 0.61 against those who had no operation18. A Geneva registry found a 40% lower risk of breast cancer death after complete excision19, and institutional series associated surgery with longer progression-free survival20,21. In another cohort, surgery remained independently associated with survival after adjustment for the more favourable disease of the women who received it22. Editorials at the time warned that the women chosen for surgery were those who would have done well anyway23,24.
Two earlier randomised trials then disagreed with each other. In Mumbai, 350 women whose disease had responded to chemotherapy were randomised, and median overall survival was 19.2 months with locoregional treatment and 20.5 months without it25. MF07-01 randomised 274 treatment-naive women at diagnosis26. The hazard of death was 34% lower with surgery, but the difference appeared only with longer follow-up, and its subgroup analyses were unplanned. A prospective registry of de novo disease in the same era found the 21-gene recurrence score independently prognostic for time to progression and two-year survival in ER-positive, HER2-negative disease27. Biology, in other words, was sorting these patients before any operation did.
The decision therefore rests on EA2108, and only for the decision it randomised. A patient presenting de novo, responding to systemic therapy, with the primary still in place, is exactly the patient in that trial. Operating on her buys local control rather than time. That is a defensible reason to operate when the primary is ulcerating or threatening the chest wall. It is not a reason to operate on a scan.
What the table cannot show is the question this section is actually about. Nobody randomised oligoprogression. No trial compared ablating the single progressing deposit and continuing the systemic regimen against changing the regimen. That is the commoner decision, and the one with a clonal rationale behind it. No column anywhere records the imaging used to count the lesions, so the entry criterion that defines the whole category is invisible in the row that reports it. The three rationales for treating locally are set out with their evidence in Local intervention in the metastatic setting.
7 · Treatment-free intervals and cumulative toxicity
Metastatic treatment is usually continuous, and the assumption that it must be goes mostly unexamined. Two things make it worth examining.
The first is that toxicity classes differ in whether they accumulate. Endocrine agents and targeted oral drugs produce toxicity that is largely a function of current exposure. Taxane neuropathy, anthracycline cardiac injury and marrow reserve are cumulative. Stopping the first group returns the patient close to baseline. Stopping the second group stops the accumulation and does not reverse it.
The second is that the goal is time at acceptable function, not maximal tumour suppression at every moment. A planned interval off cytotoxic therapy in a patient with controlled, low-tempo disease buys recovery that a dose reduction does not.
Ask at each cycle which toxicities are cumulative and which are not. Only the cumulative ones argue for a break rather than a reduction.
State a plan for the interval before starting it. Duration, what will be monitored, and what would end it early.
Record the reason a dose was reduced. A reduction made for an episode that resolved is a candidate for reversal, and it will not be reversed unless someone wrote down why it happened.
8 · Clinical trials and early integration of palliative care
Trial enrolment is a line of therapy, not a consolation. It is best discussed at the start of a line rather than after the last option has failed, because eligibility criteria are strictest at the point where they are hardest to meet. Prior therapy limits, organ function thresholds and washout requirements all tighten with each additional line.
Early palliative care is an intervention with randomised evidence behind it. In patients with metastatic non-small-cell lung cancer, early integrated palliative care improved quality of life and reduced depressive symptoms at 12 weeks. Fewer patients received aggressive care at the end of life. Median survival was longer in the integrated arm28. The population was lung cancer, and the survival finding should not be carried across to breast cancer as a claim about this disease.
What transfers is the design point, which is that referral was triggered by diagnosis and timing rather than by prognosis or by the exhaustion of treatment. That trigger is the part of the model that can be copied into a breast cancer clinic. End-of-life care is developed in Palliative and end-of-life care.
Interplay
Every principle in this chapter follows from the tumour being a population rather than a clone. De novo and recurrent disease differ because one population has been selected and the other has not. Oligoprogression is a claim about where a resistant subclone sits, not about how much disease there is. Matching the agent to the failure rather than to the line is the same argument at the bedside. What failed was a dependency in part of the population, not in the whole of it. The clonal account that licenses each of those readings is in Clonal evolution and treatment sequencing, and the reason a single biopsy can misrepresent the population it came from is in Spatial heterogeneity, biopsy selection, and biomarker interpretation.
Saphner T, Tormey DC, Gray R. Annual hazard rates of recurrence for breast cancer after primary therapy. J Clin Oncol 1996 14:2738-2746. PMID 8874335
Kennecke H, Yerushalmi R, Woods R, et al. Metastatic behavior of breast cancer subtypes. J Clin Oncol 2010 28:3271-3277. PMID 20498394
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
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
Schiavon G, Hrebien S, Garcia-Murillas I, et al. Analysis of ESR1 mutation in circulating tumor DNA demonstrates evolution during therapy for metastatic breast cancer. Sci Transl Med 2015 7:313ra182. PMID 26560360
Razavi P, Li BT, Brown DN, et al. High-intensity sequencing reveals the sources of plasma circulating cell-free DNA variants. Nat Med 2019 25:1928-1937. PMID 31768066
Turner NC, Kingston B, Kilburn LS, et al. Circulating tumour DNA analysis to direct therapy in advanced breast cancer (plasmaMATCH): a multicentre, multicohort, phase 2a, platform trial. Lancet Oncol 2020 21:1296-1308. PMID 32919527
Jhaveri K, Eli LD, Wildiers H, et al. Neratinib + fulvestrant + trastuzumab for HR-positive, HER2-negative, HER2-mutant metastatic breast cancer: outcomes and biomarker analysis from the SUMMIT trial. Ann Oncol 2023 34:885-898. PMID 37597578
Gruber JJ, Afghahi A, Timms K, et al. A phase II study of talazoparib monotherapy in patients with wild-type BRCA1 and BRCA2 with a mutation in other homologous recombination genes. Nat Cancer 2022 3:1181-1191. PMID 36253484
Barroso-Sousa R, Zanudo JGT, Li T, et al. Nivolumab plus low-dose ipilimumab in hypermutated HER2-negative metastatic breast cancer: a phase II trial (NIMBUS). Nat Commun 2025 16:4430. PMID 40360544
Sonke GS, van Ommen-Nijhof A, Wortelboer N, et al. Early versus deferred use of CDK4/6 inhibitors in advanced breast cancer. Nature 2024 636:474-480. PMID 39604725
Wortelboer N, van Ommen-Nijhof A, Konings IR, et al. Overall survival with first-line vs second-line CDK4/6 inhibitor use in advanced breast cancer: a randomized clinical trial. JAMA Oncol 2026 12:375-383. PMID 41712229
Miller AB, Hoogstraten B, Staquet M, Winkler A. Reporting results of cancer treatment. Cancer 1981 47:207-14. PMID 7459811
Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009 45:228-47. PMID 19097774
Seymour L, Bogaerts J, Perrone A, et al. iRECIST: guidelines for response criteria for use in trials testing immunotherapeutics. Lancet Oncol 2017 18:e143-e152. PMID 28271869
Khan SA, Zhao F, Goldstein LJ, et al. Early local therapy for the primary site in de novo stage IV breast cancer: results of a randomized clinical trial (EA2108). J Clin Oncol 2022 40:978-987. PMID 34995128
Palma DA, Olson R, Harrow S, et al. Stereotactic ablative radiotherapy for the comprehensive treatment of oligometastatic cancers: long-term results of the SABR-COMET phase II randomized trial. J Clin Oncol 2020 38:2830-2838. PMID 32484754
Khan SA, Stewart AK, Morrow M. Does aggressive local therapy improve survival in metastatic breast cancer?. Surgery 2002 132:620-6; discussion 626-7. PMID 12407345
Rapiti E, Verkooijen HM, Vlastos G, et al. Complete excision of primary breast tumor improves survival of patients with metastatic breast cancer at diagnosis. J Clin Oncol 2006 24:2743-9. PMID 16702580
Babiera GV, Rao R, Feng L, et al. Effect of primary tumor extirpation in breast cancer patients who present with stage IV disease and an intact primary tumor. Ann Surg Oncol 2006 13:776-82. PMID 16614878
Rao R, Feng L, Kuerer HM, et al. Timing of surgical intervention for the intact primary in stage IV breast cancer patients. Ann Surg Oncol 2008 15:1696-702. PMID 18357493
Blanchard DK, Shetty PB, Hilsenbeck SG, Elledge RM. Association of surgery with improved survival in stage IV breast cancer patients. Ann Surg 2008 247:732-8. PMID 18438108
Morrow M, Goldstein L. Surgery of the primary tumor in metastatic breast cancer: closing the barn door after the horse has bolted?. J Clin Oncol 2006 24:2694-6. PMID 16702578
Olson JA, Marcom PK. Benefit or bias? The role of surgery to remove the primary tumor in patients with metastatic breast cancer. Ann Surg 2008 247:739-40. PMID 18438109
Badwe R, Hawaldar R, Nair N, et al. Locoregional treatment versus no treatment of the primary tumour in metastatic breast cancer: an open-label randomised controlled trial. Lancet Oncol 2015 16:1380-8. PMID 26363985
Soran A, Ozmen V, Ozbas S, et al. Randomized Trial Comparing Resection of Primary Tumor with No Surgery in Stage IV Breast Cancer at Presentation: Protocol MF07-01. Ann Surg Oncol 2018 25:3141-3149. PMID 29777404
King TA, Lyman JP, Gonen M, et al. Prognostic Impact of 21-Gene Recurrence Score in Patients With Stage IV Breast Cancer: TBCRC 013. J Clin Oncol 2016 34:2359-65. PMID 27001590
Temel JS, Greer JA, Muzikansky A, et al. Early palliative care for patients with metastatic non-small-cell lung cancer. N Engl J Med 2010 363:733-742. PMID 20818875