Part XIII · Triple-negative disease · Chapter 80

Metastatic therapy from first line onwards

What the three first-line assays actually select for, and what the conjugate record has and has not settled.

1 · Scope, and the biomarker triage at diagnosis

This chapter is the trial record for metastatic triple-negative disease, arranged by line. It says what was tested, what came back, and what each design cannot settle. How to approach the patient who carries this disease is in The biomarker triage, and what is left when it finds nothing, which covers risk, burden, symptoms and goals. Where the two chapters touch, The biomarker triage, and what is left when it finds nothing points forward to the tables here.

Triple-negative is a description by absence, and treating it as one disease is the error this chapter exists to prevent. The transcriptional and genomic diversity inside the label is set out in Triple-negative and other biologically diverse phenotypes¹.

Three tests decide first-line therapy. All three should be sent at metastatic diagnosis rather than at progression, because the first-line decision cannot wait for a result ordered later.

PD-L1, with the assay and the threshold named in the report. The antibody clone and the scored compartment differ between the trials that produced the indications, so a bare positive is not a statement about the patient.

Germline BRCA1 and BRCA2, extended to PALB2 and to somatic BRCA alterations. A patient never tested is a patient never eligible, which makes the access argument in Disparities across the breast cancer continuum a treatment argument here.

And HER2 immunohistochemistry, read specifically at the 0 against 1+ boundary. A tumour called triple-negative may be HER2-low, and therefore eligible for a conjugate approved on expression that the older classification called negative. The ultralow category below that boundary carries randomised evidence only in hormone receptor-positive disease, so it does not open the same route here HER2-low and HER2-ultralow disease.

That third point is the one most often missed. HER2-low is not a new biology. It is the same tumours under a threshold that moved, which is the measurement argument in Receptor assessment and the measurement problem and HER2-low and ultralow reframed as a heterogeneity problem. The order in which to send the three tests, and what to do while they are outstanding, is set out in Biomarker-driven treatment selection rather than here.

Each section below opens with the question its trials were built to answer, taken from the [Chapter Stories](stories.html) map, and closes with a table filtered on that same question. The story says what was being looked for and what the design cannot settle. The table says what else is in the cell. The prose argues the decision.

2 · First-line chemoimmunotherapy and the PD-L1 threshold

Four of those rows test one idea. Add a checkpoint inhibitor to a chemotherapy backbone in untreated disease, and compare it against the same backbone alone. TORCHLIGHT also admitted a small group after one chemotherapy line for metastatic disease. On the direction of selection they agree. Where overall survival was read in the whole enrolled population, it did not separate. Every survival signal that survives scrutiny sits inside an assay-positive population. That is the class finding, and it is narrower than the way the class is usually described.

There is no single PD-L1 test behind it. IMpassion130 used SP142 and scored staining on immune cells as a percentage of tumour area. KEYNOTE-355 used 22C3 and scored a combined positive score across tumour cells, lymphocytes and macrophages. A tumour positive by one is not necessarily positive by the other. The practical consequence is that the request has to name the assay the intended regimen was registered with. A report giving a score without a clone cannot be mapped onto either indication. The general form of this problem is in Immunotherapy and Predictive biomarkers and patient selection.

The trials diverge in three ways at once, and the antibody is the least important of them. The backbone differs: nab-paclitaxel in IMpassion130 and TORCHLIGHT, plain paclitaxel in IMpassion131, and a choice of three regimens in KEYNOTE-355. The assay differs, as above. The testing order differs, and that matters most. IMpassion130 required overall survival to reach significance in the intention-to-treat population before the PD-L1-positive subgroup could be tested formally. IMpassion131 tested progression-free survival in the PD-L1-positive population first. KEYNOTE-355 read overall survival at a score of 10 or higher, at 1 or higher, and in everyone.

Read that way, the two atezolizumab trials are not a disagreement about atezolizumab. They differ in the taxane, in the corticosteroid premedication a plain taxane requires, and in the population enrolled⁵. The comparison between them is across trials and was never randomised. The failure cannot be assigned to the backbone alone. TORCHLIGHT then reproduced the selection pattern with a third antibody on a nab-paclitaxel backbone³, which is the best available argument that the pattern belongs to the class rather than to one drug.

The first-line decision rests on KEYNOTE-355. It is the only trial here that carried overall survival as a primary endpoint to a prespecified score threshold and met it, against the chemotherapy the clinic would otherwise have given⁴. Three backbones were permitted, so the result does not belong to one taxane. Progression-free survival had already been reported separately⁷. The trial also shows where selection stops working, because the survival curves at a score of 1 or higher are not the curves at 10 or higher. That decay is itself the argument. A marker whose effect weakens as its threshold is lowered is behaving predictively, which is what makes this cut-point load-bearing rather than administrative.

IMpassion130 is the cautionary case rather than the load-bearing one, and it is a worked illustration of the hierarchical testing point in Clinical trials and evidence interpretation. Overall survival missed in the intention-to-treat population, so the hierarchy stopped there⁶. Every survival figure quoted for the assay-positive group descends from a test that was never formally performed⁸. The drug is used on that basis anyway, which is a defensible clinical judgement and not a statistical result. Progression-free survival there improved in the whole enrolled population as well as in the assay-positive subgroup, which is the one place the selection story is untidy.

ASCENT-04 renders in this table because it carries a checkpoint inhibitor in both arms. It is not a test of whether to add one. Both arms received pembrolizumab, and what was randomised was the cytotoxic partner against a conjugate, on a progression-free survival endpoint that favoured the conjugate². It belongs to the argument in Conjugates from refractory disease to the first line and appears here only because the filter is on modality. Its practical consequence belongs here, though. A positive PD-L1 result no longer determines the whole regimen. It determines only that a checkpoint inhibitor belongs in it.

What the table cannot show is the variable that decides eligibility. There is no column for the antibody clone, the scored compartment or the cut-off, so assays that select different patients print as one word. Nobody ran two assays on the same tumours inside a randomised trial and compared who they selected. Nobody randomised nab-paclitaxel against plain paclitaxel within a single checkpoint trial, which is the only design that could settle the backbone question. No first-line trial was powered to exclude a small benefit in the assay-negative population, which is not the same as showing there is none. Nobody randomised the band between the two thresholds as a stratum of its own, so the weak result at a score of 1 or higher is a diluted superset rather than a test of the patients between 1 and 9. And nobody randomised how long a checkpoint inhibitor should continue, or whether it should be carried past progression.

KEYNOTE-119 completes the picture from outside this table, because its line of therapy puts it elsewhere. Single-agent pembrolizumab after one or two prior treatments did not improve survival at either score threshold tested⁹. Checkpoint blockade in this disease needs a cytotoxic partner and needs to be given early. Nothing in the later-line record contradicts that.

3 · PARP inhibition for germline carriers

Who counts as a carrier is itself a measurement. In 8,085 unselected patients, 9.2% carried a pathogenic germline variant in a susceptibility gene, and BRCA1 or BRCA2 variants were commonest in triple-negative disease, at 11.2%¹⁷.

What those five rows agree on is narrow and firm. Where a germline BRCA1 or BRCA2 mutation is present, a PARP inhibitor delays progression better than the single agent a clinic would otherwise reach for. It also produces a markedly higher response rate. Neither registration trial converted that into longer survival on final analysis^18,19.

The rows differ in what the inhibitor was compared against rather than in the inhibitor. OlympiAD and EMBRACA gave it alone against the physician's choice of chemotherapy in carriers. BROCADE3 added it to carboplatin and paclitaxel, so its control arm already contained the platinum that the same repair defect selects for. That trial therefore asks the smaller question of what a PARP inhibitor adds to a regimen already aimed at the defect. DORA tested maintenance without chemotherapy after platinum, and judged each of its two arms against a historical control rather than against the other¹². It enrolled on stable disease or response to platinum rather than on germline status, and sustained benefit tracked neither germline BRCA nor PD-L1. TBCRC 048 is a single arm, and its question is which gene rather than which drug.

The decision rests on OlympiAD and EMBRACA together, because they are the only randomised comparisons against the treatment a carrier would otherwise receive at that point. They used different PARP inhibitors, different chemotherapy baskets and different randomisation ratios, and they returned the same shape of answer. Two trials reproducing one another with different drugs is the strongest evidence this section has.

The gap between the two endpoints is the interpretive problem, and it has a candidate explanation rather than a settled one. Subsequent PARP inhibitor and platinum use was commoner in the chemotherapy arm of EMBRACA than in the talazoparib arm. Adjusting for it moved the hazard ratio for death in the right direction, with a bootstrap interval still crossing one¹⁹. Crossover is a reason the survival signal may be diluted. It is not evidence that the signal exists.

TBCRC 048 widened the eligible set and, more usefully, narrowed it. Confirmed responses occurred only with germline PALB2 and with somatic BRCA1 or BRCA2 alterations. Nothing responded with ATM or CHEK2 alterations alone^14,20. That is a result about which gene rather than about how much repair deficiency, and it does not license a panel-wide indication. Its population was mostly oestrogen receptor-positive, so the row sits in this table on mechanism rather than on receptor status.

Two things the table cannot show follow from that. Nobody has randomised a PARP inhibitor against platinum, which is the other drug the same defect selects for, so the choice between them at first progression rests on no head-to-head evidence at all. And there is no column for what patients received afterwards, which is where the survival reading of both registration trials lives.

One result from outside this table belongs in the argument. In TNT, germline BRCA status raised the response rate to carboplatin and not to docetaxel²¹. The same defect that PARP inhibition exploits also selects for platinum, which is why the two decisions are made together rather than in sequence. It is also why they fail together. Secondary intragenic mutations that restore the BRCA2 reading frame confer resistance both to platinum and to PARP inhibition²². The mechanism is in DNA repair, replication stress, and genomic instability and the shared resistance in PARP inhibitor and platinum resistance, including restoration of repair.

4 · Conjugates from refractory disease to the first line

Four reported randomised comparisons, two conjugates, one direction. In every one of them the conjugate beat the chemotherapy it was built to replace, in refractory disease and in the first line, with a checkpoint partner and without one. No other class in this chapter has that record.

ASCENT established it in patients who had received at least two chemotherapy regimens, one of which could be (neo)adjuvant therapy followed by early relapse^25,27. The comparator arm is the part that changed practice, because it records the honest baseline this disease had before conjugates, and that baseline is short.

The class then moved forward in three reported trials that split the first line by whether checkpoint blockade was an option. ASCENT-03 took the patients for whom it was not²⁴. ASCENT-04 took the PD-L1-positive patients and held pembrolizumab constant in both arms². TROPION-Breast02 took a second conjugate into the immunotherapy-ineligible slot²³. TROPION-Breast05 carries datopotamab deruxtecan, with durvalumab, into PD-L1-high disease against chemotherapy with pembrolizumab, and has not yet reported²⁶.

Read the control arm column before the result column, because it explains the spread across the four reported rows. ASCENT was set against single-agent chemotherapy in heavily pretreated disease. ASCENT-03 and TROPION-Breast02 were set against first-line chemotherapy, which in ASCENT-03 could be a gemcitabine and carboplatin doublet. ASCENT-04 was set against chemotherapy with pembrolizumab, the strongest comparator in the table. The conjugate's margin narrows as the comparator strengthens, and none of that is a statement about the conjugates.

The divergence in that group is maturity and endpoint, not efficacy. TROPION-Breast02 carried overall survival as a dual primary endpoint and met it. ASCENT-03 reports progression-free survival with survival not yet mature, and a published methodological critique argues that crossover-adjusted analysis is needed before the first-line position is treated as settled²⁸. ASCENT-04 reports its survival data as immature at the analysis published. ASCENT-03 and TROPION-Breast02 were built for the same slot, immunotherapy-ineligible first-line disease. They are not equivalent in what has been demonstrated.

So the first line in immunotherapy-ineligible disease rests on TROPION-Breast02, which is currently the only first-line conjugate trial with a survival result against the chemotherapy a clinic would otherwise give. ASCENT-03 will become load-bearing when its survival reads out, and ranking the two before then is a statement about follow-up rather than about drugs. One detail of ASCENT-03 is worth carrying into the consent conversation. Confirmed response rates were close to identical between the arms while duration of response was not. What the conjugate buys there is the length of a response rather than the chance of one.

ASCENT-04 leaves its own question open. Its control arm was chemotherapy with pembrolizumab, not the conjugate alone, so the trial cannot say whether the checkpoint inhibitor is still needed once the conjugate is in the regimen. The single-arm cohorts of BEGONIA arms 7 and 8 pair the other conjugate with PD-L1 blockade and report the highest confirmed response rate recorded at this position²⁹. Neither cohort has a control arm, and they are exploratory rather than part of the record above. One enrolled all comers and the other only PD-L1-high tumours, which makes the two rates a difference in selection rather than a comparison.

Three things the table cannot show. No trial has randomised one TROP2 conjugate against the other, and none is planned, so the choice between them rests on indication and toxicity. No trial has randomised the order of conjugate and chemotherapy, so the case for giving the class first rather than keeping it in reserve is inferred from separate trials at separate lines. No trial selected patients on TROP2 expression, and the biomarker analysis of ASCENT found progression-free survival tracking an expression gradient without producing a threshold anyone could act on³⁰. And the population that established the class in ASCENT, heavily pretreated and conjugate-naive, has largely been abolished by moving the class forward.

DESTINY-Breast04 brought trastuzumab deruxtecan into a population the older classification called HER2-negative, and it improved both progression-free and overall survival against the physician's choice of chemotherapy^32,33. Read it with its population in view. The great majority of those enrolled were hormone receptor-positive, and the primary endpoint was read in that cohort. The triple-negative patients are a minority inside an all-patient secondary analysis. The row that opens the HER2-low route for this chapter is therefore mostly evidence from another chapter's population.

DESTINY-Breast06 is in the table on its expression criterion alone. It enrolled only hormone receptor-positive disease after endocrine therapy, so none of its patients had triple-negative disease²⁷.

DAISY argues that the boundary is arbitrary in the other direction, because confirmed responses occurred in immunohistochemistry 0 disease at a rate well above zero³¹. It is a single-arm phase II trial in cohorts defined by staining, so it establishes that such tumours can respond and not how often they will. The apparent disagreement between the two rows is about measurement rather than biology. Pathologist concordance is poorest exactly at the 0 against 1+ boundary Where heterogeneity sets a ceiling on precision medicine, which means a proportion of the patients in each cohort would have been assigned to the other by a second reader.

TRADE-DXd sits in this table without a result, because it is still enrolling. Its contribution is discussed in Unresolved questions and what reads out next, where the sequencing question it was built for belongs.

The comparison this table most needs does not exist. No trial has randomised a HER2-directed conjugate against a TROP2 conjugate in a tumour eligible for both, which is now the commonest choice a clinic faces in HER2-low triple-negative disease. Nor is there randomised evidence restricted to HER2-low tumours without hormone receptor expression.

Toxicity rather than efficacy separates the three conjugates in practice, and the dominant toxicity differs by linker and payload handling rather than by target Toxicity mechanisms, interstitial lung disease, ocular, and mucosal. Sacituzumab govitecan is limited by neutropenia and diarrhoea, and UGT1A1 genotype is worth knowing before the first dose in a patient with a suggestive history. Trastuzumab deruxtecan carries the interstitial lung disease risk that should shape ordering in anyone with poor baseline reserve or prior thoracic radiotherapy. Datopotamab deruxtecan is dominated by mucosal and ocular events instead, which are more tolerable in a frail patient and more disruptive in a working one.

5 · Chemotherapy sequencing when no target is present

A substantial fraction of patients have no PD-L1 positivity above threshold, no germline or somatic repair alteration and no HER2 expression. For them the sequence is chemotherapy. Platinum doublets were active in single-arm studies^34,35. In tnAcity, designed to find a new first-line doublet for this population³⁶, nab-paclitaxel with carboplatin gave longer progression-free survival than nab-paclitaxel with gemcitabine or gemcitabine with carboplatin³⁷. Adding iniparib to gemcitabine and carboplatin failed its coprimary endpoints³⁸. The randomised record behind it is older and thinner than the record behind everything above.

One comparison in that table beat a taxane, and it did so in a population that does not generalise. Platinum was not worse than a taxane anywhere in it. That is the whole of the agreement, and it is enough to license either choice at the start.

The apparent disagreement inside it is partly an endpoint difference. TNT read objective response rate as its primary endpoint and found none between carboplatin and docetaxel in unselected advanced disease²¹. A difference in how long benefit lasted could not have shown in that endpoint, and progression-free survival, read as a secondary endpoint, showed none either. CBCSG006 read progression-free survival, tested it for non-inferiority against a prespecified margin and then for superiority, and met both for the platinum doublet⁴⁰. The populations differ as well. One enrolled advanced disease broadly, the other untreated disease at Chinese centres only, with gemcitabine held constant in both arms. Neither result is a general statement about platinum against a taxane at any line.

CALGB 40502 / NCCTG N063H is in this table as a taxane comparison rather than as a triple-negative trial. It enrolled all subtypes and gave bevacizumab in every arm. Neither nab-paclitaxel nor ixabepilone beat weekly paclitaxel, and the nab-paclitaxel arm was the more toxic³⁹. The newer agent was not the better one.

The decision rests on TNT. It is the randomised comparison between the two drugs a clinic actually chooses between at this point, and it is the trial that carried the marker question. Its germline BRCA finding was a prespecified interaction rather than a randomisation, and it stood alone. BRCA1 methylation, low BRCA1 mRNA and a commercial homologous recombination deficiency score all failed to select.

TBCRC 009 is the exploratory counterpart and is deliberately outside the table. It gave single-agent platinum with a gene expression predictor written in as a coprimary endpoint, and the predictor failed outright⁴¹. The choice between cisplatin and carboplatin was left to the treating physician, so the trial does not compare the two agents. Its germline carriers were a subgroup and there were very few of them.

Survival does not settle the platinum question. Neither TNT nor CBCSG006 made overall survival its primary endpoint. TNT read it as a secondary endpoint, with crossover to the other drug offered at progression, and found no difference between carboplatin and docetaxel²¹. The updated CBCSG006 analysis found no significant survival difference either⁴². Nor has anyone randomised platinum against a taxane using germline carrier status as an entry criterion, which is the trial the TNT interaction asked for.

Those two rows carry the later-line record, and both enrolled metastatic breast cancer of any subtype. EMBRACE showed that a single cytotoxic agent can extend survival after two to five prior regimens⁴⁴. Study 301 then compared eribulin with capecitabine directly and found neither superior on either co-primary endpoint⁴³.

The difference between those two outcomes is the comparator. A basket of the physician's choice is easier to beat than one named active drug, and the second trial replaced the first comparator with the second. Within the class there is no evidence of a best agent. Neither trial was restricted to triple-negative disease, so both rows describe this population by inference from a subgroup rather than by design.

What those rows are most useful for is not choosing an agent. It is setting an expectation, and the honest source is the control arm column of the conjugate table in Conjugates from refractory disease to the first line. Read the ASCENT, ASCENT-03 and TROPION-Breast02 control arms together. Single-agent chemotherapy in the refractory setting delays progression by a matter of weeks. In the first line without immunotherapy it delays progression by several months. The difference between them is the effect of line rather than of agent, and it is the most useful single thing to hold before a fourth-line conversation. Response duration shortens with each line here more steeply than it does in hormone receptor-positive disease. Expectation setting is a clinical act, and it is done with absolute medians rather than with hazard ratios.

Three principles follow. All three are frequently violated. Sequential single agents are preferred to combinations outside a situation demanding immediate response, because combinations raise response rate and toxicity without a consistent survival gain. Treat to progression or intolerance rather than to a fixed number of cycles. And choose the agent by prior exposure, residual neuropathy, marrow reserve and what the patient will tolerate, because the comparative trials show no efficacy difference to choose on. Agent detail is in Cytotoxic chemotherapy.

6 · Antigen-defined and alteration-defined options

The search for something inside this disease to aim at has produced one instructive failure and a handful of small positives. The failure is the largest and is worth reading closely. Only the four AKT trials are randomised enough to sit in the table below. Everything else named in this section is exploratory or supporting weight, and is held out of the table by design rather than by oversight.

The pattern in that table is clean and it is the reason the axis is closed. Both randomised phase II trials favoured the AKT inhibitor on progression-free survival. Both phase III trials missed their primary endpoint. Two drugs, four trials, one paclitaxel backbone throughout.

The divergence is not the drug. It is that neither phase III trial repeated the phase II trial that sent it forward. LOTUS read progression-free survival in the intention-to-treat population and in a prespecified PTEN-low population, and separated nothing in the selected group⁴⁸. Selection had therefore already failed once inside the phase II work. IPATunity130 cohort A kept progression-free survival as its endpoint but admitted only tumours carrying a PIK3CA, AKT1 or PTEN alteration, so it never tested the unselected population its signal came from⁴⁶. PAKT read progression-free survival against a one-sided threshold⁴⁷. CAPItello-290 kept the unselected population but moved the primary endpoint to overall survival, with alterations assessed by retrospective central testing⁴⁵. One programme changed the population between phases and the other changed the endpoint.

The verdict rests on CAPItello-290. It is the largest trial in the table, it enrolled the population the phase II signals came from, and it read the endpoint that decides. It also closed the escape route, because the pathway-altered tumours inside it did no better than the rest. A negative result in an unselected population can always be blamed on dilution. A negative result in the selected subgroup of that same population cannot.

What the table cannot show is the gap inside CAPItello-290. Progression-free survival still favoured the combination there, in the whole population and in altered tumours, while overall survival did not. The result column carries the primary endpoint and not that discrepancy. An agent that shifts one endpoint without shifting the other is the general warning in Clinical trials and evidence interpretation, and this programme is its cleanest example in the chapter. Nothing in the table tests an AKT inhibitor with a partner other than paclitaxel, and nothing randomised altered against unaltered tumours inside one protocol.

Earlier candidate dependencies failed in ways their designs could not separate. TBCRC 001 added EGFR blockade to carboplatin with protocol crossover, so the same patients supply more than one response group⁴⁹. TBCRC 019 set its primary endpoint as the response rate within each arm rather than the difference between arms⁵⁰. TBCRC 011 gave bicalutamide to androgen receptor-positive tumours defined at a 10% nuclear staining threshold, and asked for oestrogen and progesterone receptor negativity without specifying HER2⁵¹. Its enrolled population is therefore wider than the cell it is filed under. TBCRC 032 added a PI3K inhibitor to enzalutamide and reported a clinical benefit rate for the combination with none for the antagonist alone, so the record carries no comparison⁵². Four trials, four designs that cannot attribute what they saw.

Four current options are defined by an alteration or a histology rather than by receptors, and all four are single arms reporting response rate. SGNTUC-019 treated activating HER2 mutations in conventionally HER2-negative disease and responded most often of the four⁵³. Talazoparib Beyond BRCA treated repair gene mutations outside BRCA1 and BRCA2, and its regressions were confined to a single gene⁵⁴. NIMBUS selected on tumour mutational burden within HER2-negative disease⁵⁵. DART / SWOG S1609 cohort 36 entered patients on metaplastic histology alone, and its few responses were still running when it reported⁵⁶.

None of these is a trial of triple-negative disease. Each enrolled on a criterion that crosses the label, and not all of them report a hormone receptor or HER2 split. Each cohort is small enough that one more responder would move the rate, so none of them is a number to plan around. Match the patient to the entry criterion rather than to the cell in the table. The alteration-defined logic and its trial designs are in Expression and alteration-defined treatment across subtypes, and the case for broad sequencing here is in Data science and precision oncology.

7 · Unresolved questions and what reads out next

What to do after a conjugate is the dominant unanswered question. It is also now the commonest clinical situation in later-line disease, created by the first-line results in Conjugates from refractory disease to the first line. The only clinical evidence is a retrospective series of patients treated with two conjugates in sequence⁵⁷. It is consistent with partial rather than complete cross-resistance.

TRADE-DXd is the one prospective study built around the question, and the single row is the point. It treats patients whose disease has progressed on a previous conjugate, in cohorts defined by which conjugate came first. It admits hormone receptor-positive and receptor-negative disease, and immunohistochemistry 0 tumours alongside HER2-low ones. Its endpoint is response rate and its cohorts are non-comparative, so it will not rank sequences and it cannot be read as a randomised comparison. It will produce the first prospective estimate of how much activity remains after a conjugate fails. Nothing else in the registry is designed to answer this, which is why the table has one row rather than several.

Whether responsiveness to checkpoint blockade can be manufactured where the assay says it is absent remains open. Each of the attempts produces disease control in a minority, and they disagree about who that minority is. TOPACIO/KEYNOTE-162 breast cohort combined a PARP inhibitor with PD-1 blockade irrespective of BRCA status, and its responses concentrated in tumours carrying a BRCA mutation⁵⁸. DORA gave chemotherapy-free maintenance after platinum and found durable control that did not track germline BRCA status¹². Those two are not in conflict about the drug. They differ in line, in selection and in endpoint, one reading response in advanced disease and the other reading progression-free survival in platinum-sensitive disease against a historical control.

The attempts share one design problem. TONIC allocated patients between priming cohorts and never compared them statistically, and only its first stage is reported⁵⁹. Pembrolizumab plus radiotherapy in metastatic TNBC had no control arm, so a shrinking unirradiated lesion cannot be credited to an abscopal effect⁶⁰. MEDIOLA breast cohort is a single arm combining PARP and checkpoint inhibition, so neither drug's contribution can be isolated⁶¹. Eligibility drifts as well. MEDIOLA breast cohort asked for a germline BRCA mutation in HER2-negative disease rather than in triple-negative disease, and DORA defined triple-negative at a receptor staining threshold rather than at no expression. Hormone receptor-positive tumours sit inside both.

Only TBCRC 043 tested the immune agent itself in a randomised comparison, adding PD-L1 blockade to carboplatin, and it favoured the combination on both progression-free and overall survival⁶². Its benefit did not follow PD-L1 status, which is the reverse of the selection the first-line trials were built on. It is an exploratory phase II trial and is not confirmatory evidence. It is also the only design in this group that could have attributed anything to the immune agent, which is the argument for repeating it at size rather than for acting on it.

Whether PD-L1 is the right marker at all is the question underneath the whole first line. It selects imperfectly, the assays disagree with each other, and the biology in Host and tumor interactions suggests tumour-infiltrating lymphocytes carry information the receptor stain does not. A better marker would change first-line practice more than a better drug would.

And the whole first line now rests on a conjugate for almost everybody, with no randomised evidence on what should follow it. Until a conjugate arrives carrying a payload from a different class, the choices after failure will continue to be made on antigen expression, which answers one question out of six Mechanisms of conjugate resistance.

References

All 62 references

1. Lehmann BD, Bauer JA, Chen X, et al. Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies. J Clin Invest 2011 121:2750-2767. PMID 21633166
2. Tolaney SM, de Azambuja E, Kalinsky K, et al. Sacituzumab govitecan plus pembrolizumab for advanced triple-negative breast cancer. N Engl J Med 2026 394:354-366. PMID 41564397
3. Jiang Z, et al. Toripalimab plus nab-paclitaxel in metastatic or recurrent triple-negative breast cancer: a randomized phase 3 trial. Nat Med 2024 30:249-256. PMID 38191615
4. Cortes J, Rugo HS, Cescon DW, et al. Pembrolizumab plus chemotherapy in advanced triple-negative breast cancer. N Engl J Med 2022 387:217-226. PMID 35857659
5. Miles D, et al. Primary results from IMpassion131, a double-blind, placebo-controlled, randomised phase III trial of first-line paclitaxel with or without atezolizumab for unresectable locally advanced/metastatic triple-negative breast cancer. Ann Oncol 2021 32:994-1004. PMID 34219000
6. Schmid P, Adams S, Rugo HS, et al. Atezolizumab and nab-paclitaxel in advanced triple-negative breast cancer. N Engl J Med 2018 379:2108-2121. PMID 30345906
7. Cortes J, et al. Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial. Lancet 2020 396:1817-1828. PMID 33278935
8. Emens LA, et al. First-line atezolizumab plus nab-paclitaxel for unresectable, locally advanced, or metastatic triple-negative breast cancer: IMpassion130 final overall survival analysis. Ann Oncol 2021 32:983-993. PMID 34272041
9. Winer EP, Lipatov O, Im SA, et al. Pembrolizumab versus investigator-choice chemotherapy for metastatic triple-negative breast cancer (KEYNOTE-119): a randomised, open-label, phase 3 trial. Lancet Oncol 2021 22:499-511. PMID 33676601
10. Schmid P, Cortes J, Pusztai L, et al. Pembrolizumab for early triple-negative breast cancer. N Engl J Med 2020 382:810-821. PMID 32101663
11. Schmid P, Cortes J, Dent R, et al. Overall survival with pembrolizumab in early-stage triple-negative breast cancer. N Engl J Med 2024 391:1981-1991. PMID 39282906
12. Tan TJ, Sammons S, Im YH, et al. Phase II DORA Study of Olaparib with or without Durvalumab as a Chemotherapy-Free Maintenance Strategy in Platinum-Pretreated Advanced Triple-Negative Breast Cancer. Clin Cancer Res 2024 30:1240-1247. PMID 38236575
13. Diéras V, et al. Veliparib with carboplatin and paclitaxel in BRCA-mutated advanced breast cancer (BROCADE3): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol 2020 21:1269-1282. PMID 32861273
14. Tung NM, Robson ME, Ventz S, et al. TBCRC 048: phase II study of olaparib for metastatic breast cancer and mutations in homologous recombination-related genes. J Clin Oncol 2020 38:4274-4282. PMID 33119476
15. Litton JK, Rugo HS, Ettl J, et al. Talazoparib in patients with advanced breast cancer and a germline BRCA mutation. N Engl J Med 2018 379:753-763. PMID 30110579
16. Robson M, Im SA, Senkus E, et al. Olaparib for metastatic breast cancer in patients with a germline BRCA mutation. N Engl J Med 2017 377:523-533. PMID 28578601
17. Sun J, Meng H, Yao L, et al. Germline Mutations in Cancer Susceptibility Genes in a Large Series of Unselected Breast Cancer Patients. Clin Cancer Res 2017 23:6113-6119. PMID 28724667
18. Robson ME, et al. OlympiAD final overall survival and tolerability results: Olaparib versus chemotherapy treatment of physician's choice in patients with a germline BRCA mutation and HER2-negative metastatic breast cancer. Ann Oncol 2019 30:558-566. PMID 30689707
19. Litton JK, et al. Talazoparib versus chemotherapy in patients with germline BRCA1/2-mutated HER2-negative advanced breast cancer: final overall survival results from the EMBRACA trial. Ann Oncol 2020 31:1526-1535. PMID 32828825
20. Tung NM, et al. TBCRC 048 (Olaparib Expanded) Expansion Cohorts: Phase II Study of Olaparib Monotherapy for Patients With Metastatic Breast Cancer With Germline Mutations inor Somatic Mutations inor. J Clin Oncol 2026 44:653-661. PMID 41604601
21. Tutt A, Tovey H, Cheang MCU, et al. Carboplatin in BRCA1/2-mutated and triple-negative breast cancer BRCAness subgroups: the TNT trial. Nat Med 2018 24:628-637. PMID 29713086
22. Sakai W, Swisher EM, Karlan BY, et al. Secondary mutations as a mechanism of cisplatin resistance in BRCA2-mutated cancers. Nature 2008 451:1116-1120. PMID 18264087
23. Dent R, Shao Z, Schmid P, et al. Datopotamab deruxtecan in patients with untreated, advanced triple-negative breast cancer (TROPION-Breast02): a randomised, open-label, international, phase III trial. Ann Oncol 2026 37:1066-1080. PMID 41937088
24. Cortes J, Punie K, Barrios C, et al. Sacituzumab govitecan in untreated, advanced triple-negative breast cancer. N Engl J Med 2025 393:1912-1925. PMID 41124233
25. Bardia A, Hurvitz SA, Tolaney SM, et al. Sacituzumab govitecan in metastatic triple-negative breast cancer. N Engl J Med 2021 384:1529-1541. PMID 33882206
26. Schmid P, Oliveira M, O'Shaughnessy J, et al. TROPION-Breast05: a randomized phase III study of Dato-DXd with or without durvalumab versus chemotherapy plus pembrolizumab in patients with PD-L1-high locally recurrent inoperable or metastatic triple-negative breast cancer. Ther Adv Med Oncol 2025 17:17588359251327992. PMID 40297626
27. Bardia A, Hu X, Dent R, et al. Trastuzumab deruxtecan after endocrine therapy in metastatic breast cancer. N Engl J Med 2024 391:2110-2122. PMID 39282896
28. Poggio F, Generali D. Methodological considerations in the interpretation of ASCENT-03. Eur J Cancer 2026 245:116962. PMID 42520594
29. Schmid P, Wang HC, Lynce F, et al. Datopotamab deruxtecan (Dato-DXd) in combination with durvalumab as first-line treatment for unresectable locally advanced or metastatic triple-negative breast cancer: results from arms 7 and 8 of the phase Ib/II BEGONIA study. Ann Oncol 2026 37:1081-1092. PMID 42167437
30. Bardia A, Tolaney SM, Punie K, et al. Biomarker analyses in the phase III ASCENT study of sacituzumab govitecan versus chemotherapy in patients with metastatic triple-negative breast cancer. Ann Oncol 2021 32:1148-1156. PMID 34116144
31. Mosele F, Deluche E, Lusque A, et al. Trastuzumab deruxtecan in metastatic breast cancer with variable HER2 expression: the phase 2 DAISY trial. Nat Med 2023 29:2110-2120. PMID 37488289
32. Modi S, Jacot W, Yamashita T, et al. Trastuzumab deruxtecan in previously treated HER2-low advanced breast cancer. N Engl J Med 2022 387:9-20. PMID 35665782
33. Modi S, et al. Trastuzumab deruxtecan in HER2-low metastatic breast cancer: long-term survival analysis of the randomized, phase 3 DESTINY-Breast04 trial. Nat Med 2025 31:4205-4213. PMID 41062831
34. Perez EA, Hillman DW, Stella PJ, et al. A phase II study of paclitaxel plus carboplatin as first-line chemotherapy for women with metastatic breast carcinoma. Cancer 2000 88:124-31. PMID 10618614
35. Nelli F, Moscetti L, Natoli G, et al. Gemcitabine and carboplatin for pretreated metastatic breast cancer: the predictive value of immunohistochemically defined subtypes. Int J Clin Oncol 2013 18:343-9. PMID 22350024
36. Yardley DA, Brufsky A, Coleman RE, et al. Phase II/III weekly nab-paclitaxel plus gemcitabine or carboplatin versus gemcitabine/carboplatin as first-line treatment of patients with metastatic triple-negative breast cancer (the tnAcity study): study protocol for a randomized controlled trial. Trials 2015 16:575. PMID 26673577
37. Yardley DA, Coleman R, Conte P, et al. nab-Paclitaxel plus carboplatin or gemcitabine versus gemcitabine plus carboplatin as first-line treatment of patients with triple-negative metastatic breast cancer: results from the tnAcity trial. Ann Oncol 2018 29:1763-1770. PMID 29878040
38. O'Shaughnessy J, Schwartzberg L, Danso MA, et al. Phase III study of iniparib plus gemcitabine and carboplatin versus gemcitabine and carboplatin in patients with metastatic triple-negative breast cancer. J Clin Oncol 2014 32:3840-7. PMID 25349301
39. Rugo HS, et al. Randomized Phase III Trial of Paclitaxel Once Per Week Compared With Nanoparticle Albumin-Bound Nab-Paclitaxel Once Per Week or Ixabepilone With Bevacizumab As First-Line Chemotherapy for Locally Recurrent or Metastatic Breast Cancer: CALGB 40502/NCCTG N063H (Alliance). J Clin Oncol 2015 33:2361-9. PMID 26056183
40. Hu XC, et al. Cisplatin plus gemcitabine versus paclitaxel plus gemcitabine as first-line therapy for metastatic triple-negative breast cancer (CBCSG006): a randomised, open-label, multicentre, phase 3 trial. Lancet Oncol 2015 16:436-46. PMID 25795409
41. Isakoff SJ, Mayer EL, He L, et al. TBCRC009: A Multicenter Phase II Clinical Trial of Platinum Monotherapy With Biomarker Assessment in Metastatic Triple-Negative Breast Cancer. J Clin Oncol 2015 33:1902-9. PMID 25847936
42. Zhang J, Lin Y, Sun XJ, et al. Biomarker assessment of the CBCSG006 trial: a randomized phase III trial of cisplatin plus gemcitabine compared with paclitaxel plus gemcitabine as first-line therapy for patients with metastatic triple-negative breast cancer. Ann Oncol 2018 29:1741-1747. PMID 29905759
43. Kaufman PA, Awada A, Twelves C, et al. Phase III open-label randomized study of eribulin mesylate versus capecitabine in patients with locally advanced or metastatic breast cancer previously treated with an anthracycline and a taxane. J Clin Oncol 2015 33:594-601. PMID 25605862
44. Cortes J, O'Shaughnessy J, Loesch D, et al. Eribulin monotherapy versus treatment of physician's choice in patients with metastatic breast cancer (EMBRACE): a phase 3 open-label randomised study. Lancet 2011 377:914-923. PMID 21376385
45. Schmid P, et al. Capivasertib plus paclitaxel as first-line treatment for metastatic triple-negative breast cancer: results from the randomised, global phase III CAPItello-290 trial. Ann Oncol 2025 37:650-662. PMID 41422862
46. Dent RA, et al. Ipatasertib plus Paclitaxel for Patients with PIK3CA/AKT1/PTEN-Altered Locally Advanced Unresectable or Metastatic Triple-Negative Breast Cancer in the IPATunity130 Phase III Trial. Clin Cancer Res 2024 30:4329-4338. PMID 39058425
47. Schmid P, et al. Capivasertib Plus Paclitaxel Versus Placebo Plus Paclitaxel As First-Line Therapy for Metastatic Triple-Negative Breast Cancer: The PAKT Trial. J Clin Oncol 2019 38:423-433. PMID 31841354
48. Kim SB, et al. Ipatasertib plus paclitaxel versus placebo plus paclitaxel as first-line therapy for metastatic triple-negative breast cancer (LOTUS): a multicentre, randomised, double-blind, placebo-controlled, phase 2 trial. Lancet Oncol 2017 18:1360-1372. PMID 28800861
49. Carey LA, Rugo HS, Marcom PK, et al. TBCRC 001: randomized phase II study of cetuximab in combination with carboplatin in stage IV triple-negative breast cancer. J Clin Oncol 2012 30:2615-23. PMID 22665533
50. Forero-Torres A, Varley KE, Abramson VG, et al. TBCRC 019: A Phase II Trial of Nanoparticle Albumin-Bound Paclitaxel with or without the Anti-Death Receptor 5 Monoclonal Antibody Tigatuzumab in Patients with Triple-Negative Breast Cancer. Clin Cancer Res 2015 21:2722-9. PMID 25779953
51. Gucalp A, Tolaney S, Isakoff SJ, et al. Phase II trial of bicalutamide in patients with androgen receptor-positive, estrogen receptor-negative metastatic breast cancer. Clin Cancer Res 2013 19:5505-5512. PMID 23965901
52. Lehmann BD, Abramson VG, Sanders ME, et al. TBCRC 032 IB/II Multicenter Study: Molecular Insights to AR Antagonist and PI3K Inhibitor Efficacy in Patients with ARMetastatic Triple-Negative Breast Cancer. Clin Cancer Res 2019 26:2111-2123. PMID 31822498
53. Okines AFC, Curigliano G, Mizuno N, et al. Tucatinib and trastuzumab in HER2-mutated metastatic breast cancer: a phase 2 basket trial. Nat Med 2025 31:909-916. PMID 39825152
54. 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
55. 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
56. Adams S, Othus M, Patel SP, et al. A Multicenter Phase II Trial of Ipilimumab and Nivolumab in Unresectable or Metastatic Metaplastic Breast Cancer: Cohort 36 of Dual Anti-CTLA-4 and Anti-PD-1 Blockade in Rare Tumors (DART, SWOG S1609). Clin Cancer Res 2021 28:271-278. PMID 34716198
57. Mai N, Klar Lieberman MM, Ferraro E, et al. Sequential antibody-drug conjugate therapy in patients with metastatic breast cancer treated with sacituzumab govitecan and trastuzumab deruxtecan. JCO Precis Oncol 2025 9:e2400898. PMID 40440568
58. Vinayak S, Tolaney SM, Schwartzberg L, et al. Open-label Clinical Trial of Niraparib Combined With Pembrolizumab for Treatment of Advanced or Metastatic Triple-Negative Breast Cancer. JAMA Oncol 2019 5:1132-1140. PMID 31194225
59. Voorwerk L, Slagter M, Horlings HM, et al. Immune induction strategies in metastatic triple-negative breast cancer to enhance the sensitivity to PD-1 blockade: the TONIC trial. Nat Med 2019 25:920-928. PMID 31086347
60. Ho AY, et al. A phase 2 clinical trial assessing the efficacy and safety of pembrolizumab and radiotherapy in patients with metastatic triple-negative breast cancer. Cancer 2019 126:850-860. PMID 31747077
61. Domchek SM, Postel-Vinay S, Im SA, et al. Olaparib and durvalumab in patients with germline BRCA-mutated metastatic breast cancer (MEDIOLA): an open-label, multicentre, phase 1/2, basket study. Lancet Oncol 2020 21:1155-1164. PMID 32771088
62. Lehmann BD, Abramson VG, Dees EC, et al. Atezolizumab in Combination With Carboplatin and Survival Outcomes in Patients With Metastatic Triple-Negative Breast Cancer: The TBCRC 043 Phase 2 Randomized Clinical Trial. JAMA Oncol 2024 10:193-201. PMID 38095878