Two features of this subtype shape everything below. Survival is long enough that a patient may pass through many lines of treatment, so the order matters more here than in any other subtype. And the lines were established sequentially against the standard of their own day. The trial that defined second line enrolled patients who had never received what is now first line.
That second point is the one to hold onto. The evidence base is a stack of comparisons, each of them valid against its own control. The stack does not tell you what to do after a modern first line. Every table below therefore carries the population column, because the population is what dates a comparison and the result column does not.
The word line is doing less work in these trials than a reader expects. Most later-line trials set a minimum prior exposure and capped nothing above it, so a second-line label is a floor rather than a description of who was enrolled. Trials that admit a wide range of prior treatment tend to report a median number of prior regimens well above their stated minimum. Read that median, not the label. The line tables below file each trial once, under the lowest line its eligibility admitted, so a trial open from the second line sits in the second-line table even where most of its patients were further on.
The endpoint column dates a comparison as thoroughly as the population column does. The earliest trials here measured time to progression or response rate. The later ones measure progression-free survival by independent review. Those are not the same object, and a table sorted by year lines them up as though they were.
Read the tables as a record of what was asked rather than as a ranking of the drugs in them. Two rows may sit a generation apart and share no backbone at all. Nothing in the registry adjusts for that and nothing could. The ordering a clinic uses is a reading across the stack, and no trial inside it tested that ordering.
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 dual blockade with a taxane
Advanced disease · HER2+ · any
Layering HER2 blockade and choosing what follows it
Does blocking HER2 at more than one site do more than blocking it at one?
Rationale
HER2 has no ligand of its own. It signals by pairing. Its most active partner is HER3, which is ligand-driven. Trastuzumab binds the membrane-proximal domain and leaves the dimerisation arm exposed. Ligand-driven pairing therefore continues underneath it. A second antibody aimed at that arm, or a small molecule that shuts the kinase down from inside the cell, should close an escape the first antibody never covered.
Experiment
A sequence of first-line randomised trials, each adding one more layer to the layer below. Trastuzumab added to chemotherapy, in a large trial and again in a small one. Then pertuzumab added to trastuzumab and a taxane. Then a HER2 kinase inhibitor added to trastuzumab and a taxane in one trial, and added to trastuzumab with pertuzumab as maintenance in another. A large single-arm study ran alongside them, giving the same antibody pair with whichever taxane the investigator chose.
Finding
Adding the first antibody to chemotherapy delayed progression and lengthened overall survival, against the same chemotherapy given alone. Adding the second antibody did both again, against a single antibody given with the same taxane. Adding a kinase inhibitor delayed progression in both trials that tried it, each against the regimen it was added to. The kinase inhibitor added to the taxane doublet has since reported an overall survival advantage over that doublet. The kinase inhibitor added to the antibody pair as maintenance has not reported overall survival at all. The endpoint is not constant along the chain. The earliest trials measured time to progression or response rate rather than progression-free survival, so a longer interval is not the same measurement at both ends of it. The single-arm study reported tolerability for the antibody pair with each of the three taxanes, and had no comparator to rank them.
Limitation
These trials were run against the standard of their own year, so the additions are not stacked on one backbone. The first layer rests on one large trial with a time-to-progression endpoint and one small trial whose primary endpoint was response rate, so the two are not the same grade of evidence. CLEOPATRA's control arm had none of the drugs that now follow first-line progression. The survival difference it measured therefore belongs to a sequence rather than to pertuzumab alone. PHILA's control arm was trastuzumab and docetaxel without pertuzumab, so it says what a kinase inhibitor adds to a doublet rather than what it adds to the triplet a clinic now starts with. It also enrolled in one country. HER2CLIMB-05 randomised only patients who had not progressed during induction, so its population is selected for disease the backbone already controls. PERUSE had safety as its primary endpoint and no comparator at all, so its efficacy figures describe a cohort rather than test anything.
Next question
If a third layer of HER2 blockade beats two, what is the free taxane still doing in the regimen?
H0648gM77001CLEOPATRAPHILAHER2CLIMB-05PERUSE
Does the cytotoxic have to be given as free chemotherapy, or can it travel on the antibody?
Rationale
A conjugate uses the antibody as a delivery vehicle. The cytotoxic should concentrate where the antigen is displayed and spare tissue that does not display it. If that holds at the front of the disease course, the free taxane is doing nothing the conjugate does not do better. The stronger version of the idea drops the cytotoxic altogether, on the argument that a HER2-dependent tumour can be held by blockade alone.
Experiment
Two first-line randomised trials substituted a conjugate for the taxane. One compared trastuzumab emtansine, alone and with pertuzumab, against trastuzumab plus a taxane. The other compared trastuzumab deruxtecan plus pertuzumab against the taxane, trastuzumab and pertuzumab triplet, and carried a third arm of the conjugate alone. An early-phase programme is giving the newer conjugate alone and in combinations in the same line. The chemotherapy-free version of the question was asked twice and separately. A single-arm study gave lapatinib with trastuzumab across two cohorts defined by prior treatment. A randomised trial gave lapatinib with or without trastuzumab in disease progressing on trastuzumab.
Finding
The first-generation conjugate was noninferior to the taxane backbone and was not superior to it, with or without the second antibody. The newer conjugate with the second antibody delayed progression against the triplet. Chemotherapy-free dual blockade produced responses in a cohort untreated for metastatic disease and in a pretreated cohort, more often in the untreated one. Where blockade without chemotherapy was randomised against less blockade without chemotherapy, the pair delayed progression and lengthened overall survival against the single agent.
Limitation
Noninferiority is a claim about a prespecified margin rather than a demonstration that two regimens are the same. MARIANNE's primary publication reported no overall survival at all. DESTINY-Breast09's third arm, the conjugate without pertuzumab, was still blinded at the interim analysis, so nothing yet separates what the conjugate contributes from what the second antibody contributes. Its overall survival is not reported either. DESTINY-Breast07 has no publication on the record, so it names a programme rather than a result. TBCRC 003 is single-arm and its two cohorts were defined by prior treatment, so the gap between their response rates is a difference between populations rather than a comparison. A chemotherapy-free response rate from that era cannot stand as a comparator for a modern chemotherapy-containing regimen. EGF104900 did drop the cytotoxic, but it set its pair against a single kinase inhibitor rather than against any chemotherapy-containing regimen. It therefore says nothing about how chemotherapy-free blockade stands against the triplet a clinic starts with.
Next question
If the cytotoxic can be carried on the antibody or left out at the start, is the receptor still worth blocking once the disease has progressed on drugs aimed at it?
HER2-positive advanced or metastatic disease with no prior chemotherapy or HER2-directed therapy for metastatic disease; randomised 1:1:1, and 770 patients in the two unblinded groups are reported at this prespecified interim analysis, 383 given the conjugate with pertuzumab and 387 given the taxane triplet, with the third group still blinded
trastuzumab deruxtecan plus pertuzumab vs taxane, trastuzumab and pertuzumab
median progression-free survival 40.7 vs 26.9 months, HR 0.56, 95% CI 0.44 to 0.71
Patients with centrally confirmed HER2-positive metastatic breast cancer without progression after first-line induction therapy and with no or asymptomatic brain metastases.
Tucatinib plus trastuzumab and pertuzumab versus placebo plus trastuzumab and pertuzumab, as first-line maintenance.
Median progression-free survival was 24.9 months with tucatinib versus 16.3 months with placebo added to trastuzumab and pertuzumab, hazard ratio 0.641, 95% CI 0.514 to 0.799, P<.0001.
Not yet reported. Overall survival data remain immature.
Female patients in China with untreated HER2-positive metastatic breast cancer, median age 52 years.
Pyrotinib plus trastuzumab and docetaxel versus placebo plus trastuzumab and docetaxel.
Median investigator-assessed progression-free survival was 24.3 months with pyrotinib versus 10.4 months with placebo, a hazard ratio of 0.41 (95% CI 0.32 to 0.53).
In the long-term analysis at a median follow-up of about 35 months, overall survival favoured pyrotinib with a hazard ratio of 0.64 (95% CI 0.46 to 0.89), and median overall survival was not reached in either arm.
Inoperable HER2-positive locally recurrent or metastatic breast cancer with no prior systemic therapy for advanced disease except endocrine therapy.
Single arm of pertuzumab and trastuzumab with investigator-selected docetaxel, paclitaxel or nab-paclitaxel.
The primary endpoint was safety in a single arm. The most common grade 3 or worse adverse events among 1436 treated patients were neutropenia in 10% and diarrhoea in 8%, with no comparator arm.
Median overall survival was 65.3 months (95% CI 60.9 to 70.9) in the single arm.
Centrally confirmed HER2-positive advanced breast cancer with no prior chemotherapy for advanced disease; prior endocrine therapy for advanced disease was allowed.
T-DM1 alone or T-DM1 plus pertuzumab versus trastuzumab plus a taxane
Median progression-free survival was 13.7 months with trastuzumab plus taxane, 14.1 months with T-DM1, and 15.2 months with T-DM1 plus pertuzumab. Both T-DM1 arms were noninferior to trastuzumab plus taxane. Neither was superior.
Not yet reported. The primary publication gives no overall survival result.
Women with metastatic breast cancer that overexpressed HER2, receiving first-line chemotherapy.
Standard chemotherapy plus trastuzumab versus standard chemotherapy alone.
Median time to disease progression was 7.4 months with chemotherapy plus trastuzumab versus 4.6 months with chemotherapy alone, P<0.001.
Median survival was 25.1 months with chemotherapy plus trastuzumab versus 20.3 months with chemotherapy alone, P=0.01, a 20 percent reduction in the risk of death.
DESTINY-Breast071/2
not yet reported
Pertuzumab, trastuzumab and a taxane held first line for a decade. The reason is durability rather than response rate. Dual blockade covers the juxtamembrane domain and the dimerisation arm at once, which is the mechanism set out in HER family signaling.
Most randomised rows in that table add one more layer of HER2 blockade to the layer beneath it. Each of those delayed progression against the regimen its extra layer was added to. None of them failed. That is the class finding, and it is the only thing those rows say with one voice. MARIANNE and DESTINY-Breast09 are not layering trials. Each substitutes a conjugate for the free taxane, which is a different question.
What varies down the column is the measurement and the comparator rather than the drug. H0648g reported time to disease progression, the convention when trastuzumab was registered8. M77001 is a phase 2 trial whose primary endpoint was response rate7. The 2013 confirmatory and 2015 final prespecified survival analyses of CLEOPATRA quoted in HER2-directed antibodies are earlier data cuts of that trial rather than competing estimates9,10. Its primary report established the progression-free survival advantage11. From CLEOPATRA onwards the endpoint is progression-free survival, though not by the same reader. CLEOPATRA, MARIANNE and DESTINY-Breast09 used independent central review. PHILA and HER2CLIMB-05 used the investigator. An interval at the head of this table is therefore not the same object as an interval at its foot. The two cannot be read as one scale.
The control arms diverge further than the endpoints do. CLEOPATRA was run against a regimen carrying none of the drugs that now follow first-line progression, which places its survival difference in a whole sequence rather than in pertuzumab alone5. PHILA added a kinase inhibitor to a trastuzumab and docetaxel doublet rather than to the triplet a clinic now starts with3,12. It also enrolled in a single country. Part of its separation is therefore the contribution pertuzumab was never allowed to make in it, and the overall survival it has since reported inherits the same control arm. HER2CLIMB-05 randomised only patients whose disease had not progressed during induction2. Its population is selected for disease the backbone already controls. It describes what a third layer adds to a responder rather than what a three-drug start does for everyone.
The toxicity that decides between those two is different in kind as well as in size. Pyrotinib produced grade 3 or higher treatment-related events in 90% of patients against 76% on placebo3. Tucatinib produced diarrhoea in 72.7% of a maintenance population who were otherwise well2. A rate measured in a maintenance phase is paid by a patient with no symptoms to weigh it against.
DESTINY-Breast09 is the one comparison whose control arm is the taxane triplet a clinic starts with today, which is what makes it load-bearing rather than its effect size1. Its rationale is not that a conjugate is more active, which the later lines had already shown. It is that giving the agent before resistance is selected may convert responders into durable responders. That is the problem set out in Mechanisms of therapeutic resistance. Two features belong with its result. Overall survival was not the primary endpoint and was not mature at the reported interim analysis. And adjudicated drug-related interstitial lung disease or pneumonitis occurred in 12.1% of patients on the conjugate combination against 1.0% on the taxane triplet1.
MARIANNE is the reason to be careful with substitution arguments in general. The first-generation conjugate was noninferior to the taxane backbone and was not superior to it, with or without pertuzumab6. The easy reading of the two substitution trials is that trial conduct improved across the intervening decade, and it is the wrong one. MARIANNE ran against the weaker of the two available control arms, because its comparator carried no pertuzumab. A conjugate that could not beat the weaker comparator, set beside one that beats the stronger, separates the two drugs rather than the two designs. The chemistry behind that separation is in Antibody-drug conjugates and targeted delivery. Noninferiority is a claim about a prespecified margin rather than a demonstration that two regimens are the same Designing de-escalation trials and the non-inferiority margin problem. A failed superiority claim for one conjugate is not evidence about another. The stronger version of the substitution argument drops the cytotoxic rather than carrying it on the antibody, and that version is argued under Maintenance, de-escalation and treatment interruption.
Four things this table cannot show. It carries no overall survival for DESTINY-Breast09, whose third arm, the conjugate without pertuzumab, was still blinded at the interim analysis. Until that arm reads out, nothing separates what the conjugate contributes from what the second antibody contributes. DESTINY-Breast07 is the programme giving the newer conjugate alone and in combinations at this position, and it has no publication on the record. PERUSE gave the antibody pair with whichever taxane the investigator chose, with safety as its primary endpoint and no comparator4. Its preliminary report had already found paclitaxel a valid alternative backbone to docetaxel13. Its figures describe a cohort and cannot rank three taxanes against one another. And no trial here enrolled a patient progressing after a first-line deruxtecan conjugate, which is the population first-line practice is now creating. Central confirmation of HER2 was required in MARIANNE and HER2CLIMB-05. The trials of the first decade record no such step, so these populations differ by measurement as well as by era.
Other partners for the antibody pair were tested in single-arm studies, and none has a comparator. Weekly paclitaxel with trastuzumab and pertuzumab gave a six-month progression-free survival of 86% in 69 patients, with no febrile neutropenia14. Vinorelbine was tested in the two VELVET cohorts. In the first, 106 patients had a response rate of 74.2% and a median progression-free survival of 14.3 months15. In the second, where the two antibodies shared one infusion bag, the figures were 63.7% and 11.5 months16.
The taxane triplet has not become wrong. It remains the regimen with the longest follow-up, the widest availability and the best characterised late toxicity. Pertuzumab did not increase cardiac adverse events in CLEOPATRA. Cardiac adverse events occurred in 14.5% of the pertuzumab arm against 16.4% on placebo, and left ventricular systolic dysfunction in 4.4% against 8.3%17. Nor did it cost quality of life, with no difference in time to a 5-point decline in the FACT-B trial outcome index, at a hazard ratio of 0.9718. Where a deruxtecan conjugate is unavailable, or where baseline lung function makes cumulative pneumonitis risk hard to accept, the older triplet is a defensible first line rather than a compromise.
Caution
Interstitial lung disease is cumulative and does not respect line of therapy. Moving a deruxtecan conjugate earlier lengthens the exposure of patients who will live for years. The toxicity data in Antibody-drug conjugates and targeted delivery were generated in later-line populations with shorter exposure.
A chemotherapy-free first line is the obvious headline and it is not quite what the trial tested. The conjugate carries a cytotoxic payload, so what has changed is delivery rather than the presence of chemotherapy. What is genuinely lost is the finite induction model. The taxane triplet gives chemotherapy for a defined period and then stops, leaving antibody maintenance. The conjugate continues until progression. For a patient facing years of treatment, a gain in median progression-free survival and a continuous pneumonitis risk are being traded against a regimen with a known exit5,1.
3 · Second-line conjugate therapy
Layering HER2 blockade and choosing what follows it, continued
Is HER2 still worth blocking once the disease has progressed on HER2 blockade?
Rationale
Progression on trastuzumab admits two readings. Either the tumour has stopped depending on HER2, in which case the target should be abandoned. Or it has escaped one way of blocking a receptor it still needs, in which case the target should be attacked differently. The two readings imply opposite prescriptions. Only a trial that holds the chemotherapy fixed and varies the blockade can tell them apart.
Experiment
Randomised trials that held the chemotherapy fixed and varied only the HER2 blockade after progression on trastuzumab. Capecitabine with continued trastuzumab against capecitabine alone. Capecitabine with lapatinib against capecitabine alone. Lapatinib with or without trastuzumab in disease already progressing on trastuzumab, with no chemotherapy in either arm. Further down the course, and holding nothing fixed, a conjugate against whatever the physician would otherwise have chosen.
Finding
Both capecitabine trials delayed progression against capecitabine given alone, one by continuing the antibody through progression and one by adding a kinase inhibitor. Blocking the receptor at two sites delayed progression and lengthened overall survival against blocking it at one, in disease already progressing on one of them. Later in the course the conjugate delayed progression against whatever else would have been given. The target survives progression on the drugs aimed at it.
Limitation
These are separate trials with separate control arms. Two of those arms are weak by design. EGF104900 compared lapatinib plus trastuzumab against lapatinib alone, which is not a regimen a clinic would choose for anyone. TH3RESA's comparator was physician's choice in heavily pretreated disease, which is a floor rather than a standard, and it varied the chemotherapy as well as the blockade. GBG 26 randomised a small population against a time-to-progression endpoint. Its overall survival difference was not significant, so it shows that continuing the antibody delays progression rather than that it extends life. EGF100151 stopped at an interim analysis for superiority in time to progression, and its final overall survival did not separate. None of these trials required a repeat biopsy at progression, so the receptor's persistence is inferred from the fact that drugs against it still work rather than measured.
Next question
If the receptor stays worth blocking through successive lines, in what order should the drugs aimed at it be given?
GBG 26 / BIG 03-05EGF100151EGF104900TH3RESA
Trials open from the second line, each filed once under the lowest line its eligibility admitted.
HER2CLIMB, SOPHIA, DAISY and PATRICIA-SOLTI are argued in the next section. The two capecitabine trials at the foot of the table are where the principle of blocking HER2 through progression was established.
Patients with HER2-positive advanced breast cancer who had received two to four prior anti-HER2-based regimens; the randomised cohort C additionally required hormone receptor-positive disease and a PAM50 luminal A or B subtype.
In cohort C, palbociclib, trastuzumab and endocrine therapy versus trastuzumab-based treatment of physician's choice, randomised 1:1, with patients allocated to the physician's choice arm eligible for re-randomisation on progression if they still met the inclusion criteria; the earlier cohorts gave palbociclib plus trastuzumab in estrogen receptor-negative disease (cohort A) and in estrogen receptor-positive disease without or with letrozole (cohorts B1 and B2).
In cohort C the triplet was associated with significantly better progression-free survival than trastuzumab-based treatment of physician's choice, stratified hazard ratio 0.52 (95% CI, 0.29 to 0.95; two-sided P = 0.03). Progression-free survival rates after 24 months were 24.0% with the triplet and 4.3% with physician's choice.
Not reported. Neither the randomised cohort C publication nor the earlier single-arm cohort publication reports overall survival.
Patients with HER2-positive locally advanced or metastatic breast cancer previously treated with trastuzumab and a taxane, including those with brain metastases.
Tucatinib plus trastuzumab emtansine versus placebo plus trastuzumab emtansine.
Median progression-free survival was 9.5 months with tucatinib plus trastuzumab emtansine versus 7.4 months with trastuzumab emtansine alone, hazard ratio 0.76, 95% CI 0.61 to 0.95, P=0.0163.
Interim overall survival was immature. Median overall survival was not reached with tucatinib versus 38.0 months in the control arm, hazard ratio 1.23, 95% CI 0.87 to 1.74.
Adults with HER2-positive advanced or metastatic breast cancer after one or more prior anti-HER2 regimens, enrolled into a cohort with stable or active brain metastases (263 patients) and a cohort without brain metastases (241 patients).
Single arm of trastuzumab deruxtecan 5.4 mg/kg every three weeks, given in both cohorts.
progression-free survival at 12 months 61.6% in the brain metastasis cohort, with central nervous system progression-free survival 58.9%. Objective response was 62.7% in the cohort without brain metastases. There is no control arm to compare either against
metastatic breast cancer in three cohorts by HER2 immunohistochemistry, 72 overexpressing, 74 HER2-low and 40 non-expressing; 177 in the full analysis set
trastuzumab deruxtecan, single arm in each cohort
response 70.6% in HER2-overexpressing, 37.5% in HER2-low and 29.7% in non-expressing disease
Secondary endpoint. At a median follow-up of 14.1 months, median overall survival was not reached in the HER2-overexpressing or HER2-low cohorts and was 11.6 months (95% CI 8.3 to 17.3) in the non-expressing cohort. Single arm in each cohort, so there is no comparison.
Unresectable or metastatic HER2-positive breast cancer progressing after trastuzumab emtansine, age 18 or older, ECOG performance status 0 or 1.
Trastuzumab deruxtecan versus physician's choice of capecitabine plus trastuzumab or capecitabine plus lapatinib.
Median progression-free survival was 17.8 months with trastuzumab deruxtecan versus 6.9 months with treatment of physician's choice, a hazard ratio of 0.36 (95% CI 0.28 to 0.45) for progression or death, p less than 0.0001.
Overall survival favoured trastuzumab deruxtecan over treatment of physician's choice, but neither abstract gives the median values or the hazard ratio.
HER2-positive metastatic disease previously treated with trastuzumab and a taxane, all conjugate-naive
Trastuzumab deruxtecan versus trastuzumab emtansine, randomised 1:1 in a multicentre open-label trial.
progression-free survival at 12 months 75.8% against 34.1% with trastuzumab emtansine, hazard ratio 0.28. The updated analysis gave medians of 28.8 months against 6.8
median 52.6 months against 42.7 months, hazard ratio 0.73, in the long-term analysis
HER2-positive metastatic breast cancer aged 18 to 70 years, previously treated with trastuzumab and taxanes, with ECOG performance status 0 or 1.
Pyrotinib plus capecitabine versus lapatinib plus capecitabine.
Median progression-free survival was 12.5 months with pyrotinib plus capecitabine versus 6.8 months with lapatinib plus capecitabine, a hazard ratio of 0.39 (95% CI 0.27 to 0.56).
HER2-positive metastatic breast cancer previously treated with trastuzumab, pertuzumab and trastuzumab emtansine, with or without brain metastases; brain metastases could be untreated or progressing after local therapy.
Tucatinib versus placebo, each added to trastuzumab and capecitabine. The primary endpoint was tested in the first 480 patients randomised and the secondary endpoints in all 612.
median progression-free survival 7.8 months against 5.6 months with placebo added to trastuzumab and capecitabine, hazard ratio 0.54 (0.42-0.71). Among patients with brain metastases the medians were 7.6 months against 5.4, hazard ratio 0.48 (0.34-0.69)
median 21.9 months against 17.4 months, hazard ratio 0.66 (0.50-0.88)
Adults aged 18 years or older with performance status 0 or 1 and centrally confirmed, measurable, HER2-positive advanced breast cancer previously treated with trastuzumab and a taxane.
Trastuzumab emtansine 3.6 mg/kg plus atezolizumab 1200 mg versus trastuzumab emtansine plus placebo, randomly assigned 2:1 and given intravenously every 3 weeks.
Median progression-free survival was 8.2 months (95% CI 5.8-10.7) with atezolizumab versus 6.8 months (95% CI 4.0-11.1) with placebo, stratified hazard ratio 0.82 (95% CI 0.55-1.23; p=0.33). The primary endpoint was not met.
Not reported. Treatment assignment was unmasked early for futility and the primary publication gives progression-free survival and safety only.
Patients with HER2-positive breast cancer that progressed during treatment with trastuzumab.
Capecitabine plus continued trastuzumab versus capecitabine alone.
Median time to progression was 8.2 months with capecitabine plus trastuzumab versus 5.6 months with capecitabine alone, hazard ratio 0.69, 95% CI 0.48 to 0.97, P=.0338.
Median overall survival was 25.5 months with capecitabine plus trastuzumab versus 20.4 months with capecitabine alone, a difference that was not significant (P=.257).
Women with HER2-positive locally advanced or metastatic breast cancer progressing after regimens that included an anthracycline, a taxane and trastuzumab.
Lapatinib plus capecitabine versus capecitabine alone.
Median time to progression was 8.4 months with lapatinib plus capecitabine and 4.4 months with capecitabine alone, hazard ratio 0.49 (95% CI 0.34 to 0.71, P<0.001).
Median overall survival was 75.0 weeks with lapatinib plus capecitabine and 64.7 weeks with capecitabine alone, hazard ratio 0.87 (95% CI 0.71 to 1.08, P=.210).
The principle underneath this line is older than every drug in the table. GBG 26 / BIG 03-05 and EGF100151 each gave capecitabine in both arms and varied the HER2 blockade after progression on trastuzumab30,31. EGF100151 gave capecitabine at a lower dose in the lapatinib arm. One continued the antibody through progression. The other added a kinase inhibitor. Both delayed progression against capecitabine given alone. That is the evidence that the target survives progression on the drugs aimed at it. It is why the answer to HER2 failure is usually a different way of attacking the same receptor rather than a different target.
Neither of those trials lengthened life. The reason is a property of their designs rather than of the principle. Both used time to progression as the endpoint, which was the era's convention. GBG 26 / BIG 03-05 randomised a small population, and its survival difference was not significant. EGF100151 stopped at an interim analysis for superiority in time to progression, and its final survival analysis did not separate32. A trial stopped early on a progression endpoint is not built to answer a survival question.
The two conjugates in that table are not interchangeable and the difference is chemical. Trastuzumab emtansine carries a non-cleavable linker and a microtubule payload with no bystander effect. Trastuzumab deruxtecan carries a cleavable linker and a topoisomerase-I payload that crosses membranes, so it reaches cells that do not display the antigen. The consequences are developed in Antibody-drug conjugates and targeted delivery. That is the heterogeneity argument arriving as a treatment result, and the evidence for it across the expression range comes from DAISY22.
EMILIA and DESTINY-Breast03 are the same experiment run in two different eras. Each set a conjugate against the standard its own era had established, and each separated on progression and on survival29,24,33. DESTINY-Breast03 is the load-bearing row. Its comparator was the agent EMILIA itself had put in second line, and its population was conjugate-naive34. That lineage is what makes the ranking interpretable, because each step was tested against the thing it replaced.
Three rows are answering a different question from the rest of the table. PHOEBE set one kinase inhibitor against another on a shared capecitabine backbone, in China, so it ranks members of a class rather than asking whether the class belongs at this position26. Its result is a prespecified interim analysis, and the arm it beat has since been displaced from this line by the conjugates above it. A class ranking made against a comparator nobody now offers does not transfer. KATE2 appears here because its backbone is a conjugate. Its question is immunological, and it is argued under Third line and beyond with the other trials that added a mechanism from outside HER2. DESTINY-Breast12 is a single-arm study of the deruxtecan conjugate in patients with and without brain metastases, and it is argued in Central nervous system metastatic disease.
HER2CLIMB-02 and KATE2 share an identical control arm. Both gave trastuzumab emtansine to everyone, in patients previously treated with trastuzumab and a taxane. Adding a selective HER2 kinase inhibitor delayed progression. Adding a checkpoint inhibitor did not, and that trial was unmasked early for futility20,28. With design, backbone and population matched, the difference between them is mechanism rather than trial construction. Both were built on a backbone DESTINY-Breast03 has since displaced, so the question has not been asked again on a deruxtecan backbone.
HER2CLIMB-02 is the row that should slow a reader down. It improved progression-free survival against the conjugate alone, while its interim overall survival point estimate lay on the wrong side of one20. A progression endpoint moving one way and an interim survival estimate moving the other is not a shape on which to build a sequencing decision. Longer follow-up may resolve the discrepancy. Until it does, adding a kinase inhibitor at this position needs a reason beyond the curve it was licensed on.
What the table cannot show is the population second line is about to contain. DESTINY-Breast02 spans second and later lines alike, because entry required progression on trastuzumab emtansine and capped nothing before that23. Its overall survival favoured the conjugate, but the table records only the direction, without a median or hazard ratio. EMILIA carries the opposite problem. Its tabulated survival comes from the second interim analysis, while the 2017 final analysis gave medians of 29.9 against 25.9 months with a hazard ratio of 0.7535. That is one trial reported twice rather than two estimates in conflict. No trial above enrolled a patient who had already received a deruxtecan conjugate. If DESTINY-Breast09 moves that agent into first line, this section describes a population that is leaving the clinic.
The practical consequence is that second line now has two answers, and which one applies is set by what came first. After a taxane triplet, the answer is trastuzumab deruxtecan on direct randomised evidence. After first-line trastuzumab deruxtecan, the answer is inferred. The candidates are a tyrosine kinase inhibitor combination Third line and beyond, trastuzumab emtansine, or a non-cross-resistant conjugate. None has been tested in that position, and the inference rests on a class argument rather than on a trial.
None of these trials required a repeat biopsy at progression. The persistence of the receptor is inferred from the fact that drugs against it still work. It is nowhere measured.
What no trial answers is what a second conjugate is worth after the first has failed. The available evidence is retrospective. In a series of patients given both trastuzumab deruxtecan and sacituzumab govitecan in either order, progression-free survival on the second conjugate was usually shorter than on the first36. Activity was retained and the expected magnitude was smaller. The reasoning is developed in Cross-resistance between classes and what it means for sequencing.
Trastuzumab emtansine has not disappeared with all this. It retains a defined role in residual disease after neoadjuvant therapy, which is an early-disease question and belongs to Post-neoadjuvant intensification for residual disease. In the metastatic setting it is now a later option rather than a second-line standard.
In practice
When a patient progresses after a conjugate, the question is not which conjugate next but whether the resistance was to the target, the linker-payload, or the delivery. Antibody-drug conjugates and targeted delivery separates the three. Only the first is addressed by switching target.
Say which of the two arguments you are making. Either the lineage argument, which assumes the patient in front of you resembles the trial population, or a judgement about a patient who plainly does not.
Caution
Trial evidence attaches to a population, not to a line number. EMILIA put trastuzumab emtansine into the second line against a lapatinib doublet that no clinic would now offer. DESTINY-Breast03 then displaced it in a population that had never seen a conjugate. Reporting either as establishing a line number was accurate only while the line above it stood still. Whenever a regimen moves earlier, every downstream result loses the population it was measured in.
4 · Third line and beyond
Layering HER2 blockade and choosing what follows it, continued
In what order should the later-line HER2 drugs be given?
Rationale
After the first-line triplet the choices are a conjugate, a kinase inhibitor with an oral cytotoxic, or another antibody. Each was developed against whatever was standard when it entered trial. So the sequence a clinic uses was assembled from a chain of comparisons rather than chosen in any one of them.
Experiment
A conjugate against a kinase inhibitor with capecitabine, after trastuzumab and a taxane. A newer conjugate against that conjugate in the same line. The newer conjugate against physician's choice in disease that had already progressed on the older one. A kinase inhibitor added to the older conjugate. One kinase inhibitor against another with the same oral partner, early in the sequence and again late. A kinase inhibitor added to trastuzumab and capecitabine in heavily pretreated disease.
Finding
Every one of these comparisons favoured the newer agent, or the more heavily blocked arm, on its own progression endpoint. Overall survival did not follow everywhere. It separated for the conjugate against the kinase inhibitor doublet, for the newer conjugate against the older one, and for the kinase inhibitor added to trastuzumab and capecitabine. It did not separate for one kinase inhibitor against another with the same oral partner in the latest line. It ran the wrong way at the interim analysis for the kinase inhibitor added to the older conjugate. The single head-to-head of the two conjugates was decisive enough to move the newer one into second line. The order a clinic now uses is a reading across that chain rather than the result of any trial in it.
Limitation
No trial randomised the order. The controls are era-specific, and EMILIA's kinase inhibitor doublet is no longer what a clinic would put in a control arm. Physician's choice inside a conjugate trial is not the set of options a clinic would now reach for, so the separation measured against it overstates what the drug adds today. DESTINY-Breast02 spans second and later lines alike, because entry required only progression on trastuzumab emtansine and capped nothing before that. HER2CLIMB-02 improved progression-free survival against the conjugate alone while its interim overall survival ran the other way, which is not a shape a sequencing decision can be built on. PHOEBE's result is a prespecified interim analysis and its comparator has since been displaced. HER2CLIMB and NALA both admitted patients with brain metastases, so part of what they measured belongs to the central nervous system question rather than to this one.
Next question
If no trial randomised the order, can the next drug be chosen by measuring the target rather than by counting previous lines?
Can a mechanism other than HER2 blockade be added to extend what HER2 blockade can do?
Rationale
Part of trastuzumab's effect is immune. It engages Fc receptors on effector cells. The response to it tracks with the lymphocytic infiltrate in the tumour. Two routes follow from that. Re-engineer the Fc so it binds the low-affinity receptor allele better, or take the brake off the effector cells with a checkpoint antibody. A third route has nothing to do with the immune system. HER2 amplification is associated with replication stress, which raised the possibility that a PARP inhibitor would work here without a germline BRCA variant.
Experiment
An Fc-engineered anti-HER2 antibody against trastuzumab, each with chemotherapy, randomised in pretreated disease. Pembrolizumab added to trastuzumab in a single-arm study, with PD-L1 scored centrally on a metastatic biopsy and with parallel PD-L1-positive and PD-L1-negative cohorts. Atezolizumab or placebo added to trastuzumab emtansine, randomised. Niraparib added to trastuzumab with no germline selection.
Finding
The engineered antibody delayed progression against trastuzumab given with the same chemotherapy. Its final overall survival did not favour it. Checkpoint blockade added to trastuzumab produced responses only in PD-L1-positive tumours, and in a minority of those. Added to the conjugate in a randomised design, checkpoint blockade did not meet its primary endpoint. The PARP combination has produced no full report.
Limitation
PANACEA is single-arm, so its responses cannot be separated from what trastuzumab alone would have done in a PD-L1-selected group. Its PD-L1 scoring was central and required a metastatic biopsy, which is not how the marker is obtained in clinic. Its headline response applies to the PD-L1-positive cohort of the expansion stage rather than to everyone treated. KATE2 was unmasked early for futility, so follow-up is short on both arms. The PD-L1-positive subgroup its authors argue for is not reported as a subgroup result, so the case for pursuing it rests on an analysis nobody can read. TBCRC 050 sits on the record only as conference material, with nothing stated about enrolment, endpoint or result, so it names a strategy under test rather than a finding. SOPHIA separated on progression-free survival without a survival difference, which leaves the Fc hypothesis neither confirmed nor refuted. Its genotype analysis was exploratory and pointed one way for the low-affinity allele and the other way for the high-affinity one, so it generates the hypothesis it is often read as supporting.
Next question
Does the immune combination belong earlier in the course, before repeated HER2 blockade has thinned the infiltrate? And is the escape being chased here immune at all, rather than the luminal biology that the hormone receptor-positive half of this disease carries?
SOPHIAPANACEAKATE2TBCRC 050
Trials that required at least two prior lines for advanced disease.
Trials this section argues that were open from an earlier line sit in the previous table. Physician's choice in pretreated disease, where it is the comparator, is a floor rather than a standard.
Pathologically documented HER2-positive metastatic breast cancer previously treated with trastuzumab emtansine, with a median of six previous treatments.
Single arm of trastuzumab deruxtecan 5.4 mg/kg intravenously every 3 weeks.
Objective response was reported in 112 of 184 patients, 60.9 percent (95% CI 53.4 to 68.0), in this single-arm study with no control arm.
Median overall survival was 29.1 months (95% CI 24.6 to 36.1) at a median follow-up of 26.5 months in the updated analysis. Single arm, so there is no comparison.
Patients with HER2-positive metastatic breast cancer progressing during prior trastuzumab-based therapy, heavily pretreated.
Lapatinib plus trastuzumab versus lapatinib alone.
Lapatinib plus trastuzumab improved progression-free survival against lapatinib alone, hazard ratio 0.74 (95% CI 0.58 to 0.94, P=.011).
Median overall survival was 4.5 months longer with lapatinib plus trastuzumab than with lapatinib alone, hazard ratio 0.74 (95% CI 0.57 to 0.97, P=.026).
Beyond second line the evidence thins. The trials argued here were built to show different things, and three patterns run through them. Each pattern limits what can be read off the result column.
Every randomised comparison here favoured the newer agent, or the more heavily blocked arm, on its own progression endpoint. That is the whole of the agreement. Overall survival followed in some of them and not in others, and the split tracks the control arm rather than the drug.
Before the modern agents, trastuzumab was combined with most cytotoxic partners. Adding carboplatin to paclitaxel and trastuzumab improved response and progression-free survival41. Vinorelbine and taxane partners gave comparable efficacy42, and capecitabine was active and well tolerated43. In 27 patients pretreated with an anthracycline or a taxane, trastuzumab with capecitabine gave objective responses in 45% and a median progression-free survival of 6.7 months44. In CALGB 9840, weekly paclitaxel beat three-weekly paclitaxel, and trastuzumab added nothing when HER2 was not overexpressed45. The cardiac risk of the antibody was greatest with concurrent anthracyclines46.
HER2CLIMB is where the decision rests. It randomised tucatinib against placebo on a backbone of trastuzumab and capecitabine, which carries activity of its own, and it separated on progression and on survival27,47. A contrast against an active backbone is harder to achieve and easier to believe. EGF104900 also separated on survival, against lapatinib given alone40. Its primary report had shown the progression-free survival and clinical benefit advantage in trastuzumab-refractory disease48. No clinic would offer that control to anyone, so the size of the separation belongs to the comparator as much as to the combination. TH3RESA used physician's choice in heavily pretreated disease, which is a floor rather than a standard, and it varied the chemotherapy alongside the blockade39,49. NALA is the mirror image of those two. It set one kinase inhibitor against another on a shared capecitabine backbone, so both arms were active, and its survival comparison did not separate37. That is what an active control does to a survival endpoint. It is not a statement about neratinib. The backbone was shared in name rather than in dose. Capecitabine was given at a lower dose in the neratinib arm, which also carried mandatory loperamide prophylaxis. A head-to-head of two inhibitors that adjusts the partner drug between arms is not a clean comparison of the inhibitors.
Two trials have no control arm and should not be ranked against those that do. DESTINY-Breast01 established activity after trastuzumab emtansine in a single arm, in a heavily pretreated population, with response by independent review as its endpoint38. DAISY is not a line-of-therapy trial at all. It gave one conjugate across three cohorts defined by HER2 immunohistochemistry, to ask whether activity tracks the antigen22. Its HER2-overexpressing cohort is the one that belongs in this section, and the other two belong to the HER2-low argument. Both carried overall survival only as a secondary endpoint in a single arm, so neither can attribute survival to the conjugate.
Two trials rest on a biological hypothesis rather than on a position in the sequence, and they point in opposite directions about how well that works. SOLTI-1303 PATRICIA acted on its hypothesis at entry. It prescreened for a PAM50 luminal subtype before randomising palbociclib, trastuzumab and endocrine therapy against trastuzumab-based physician's choice19,50. Progression-free survival separated. The trial is small, its comparator is again a floor, and it reports no survival. SOPHIA did not select at entry, and it is the instructive negative. Margetuximab is engineered to bind CD16A more tightly, which should improve antibody-dependent cellular cytotoxicity. It separated on progression-free survival against trastuzumab given with the same chemotherapy, and its final overall survival did not favour it51,25. The CD16A genotype analysis was preplanned but exploratory, and it pointed one way for the low-affinity allele and the other way for the high-affinity one. It generates the hypothesis it is usually read as supporting. The standard for what an exploratory subgroup can carry is set in Clinical trials and evidence interpretation.
What the table cannot show is the thing a clinic most wants. No trial randomised the order. The controls are era-specific, and the kinase inhibitor doublet EMILIA displaced is no longer what anyone would put in a control arm. HER2CLIMB and NALA both admitted patients with brain metastases, so part of what they measured belongs to Central nervous system metastatic disease rather than here. And no trial argued here enrolled a patient who had received a deruxtecan conjugate earlier in the course.
Adding a mechanism other than HER2 blockade.
Two trials that asked a single question and one that has not reported.
Trial
Population
Experimental vs control
Primary endpoint
Result
Overall survival
TBCRC 0501/2 · awaited
Patients with HER2-positive metastatic breast cancer; the source document describes refractory disease and no requirement for a germline BRCA variant.
Niraparib plus trastuzumab. The located material describes no comparator.
Adults aged 18 years or older with performance status 0 or 1 and centrally confirmed, measurable, HER2-positive advanced breast cancer previously treated with trastuzumab and a taxane.
Trastuzumab emtansine 3.6 mg/kg plus atezolizumab 1200 mg versus trastuzumab emtansine plus placebo, randomly assigned 2:1 and given intravenously every 3 weeks.
Investigator-assessed progression-free survival in the intention-to-treat population.
Median progression-free survival was 8.2 months (95% CI 5.8-10.7) with atezolizumab versus 6.8 months (95% CI 4.0-11.1) with placebo, stratified hazard ratio 0.82 (95% CI 0.55-1.23; p=0.33). The primary endpoint was not met.
Not reported. Treatment assignment was unmasked early for futility and the primary publication gives progression-free survival and safety only.
Women aged 18 years or older with advanced, histologically confirmed HER2-positive breast cancer and documented progression during previous trastuzumab-based therapy, performance status 0 or 1, with a metastatic tumour biopsy available for central assessment of PD-L1 status.
Single arm. Phase 1b was a 3 + 3 dose escalation of pembrolizumab at 2 mg/kg and 10 mg/kg every 3 weeks with trastuzumab 6 mg/kg in PD-L1-positive tumours; phase 2 gave flat-dose pembrolizumab 200 mg every 3 weeks plus standard trastuzumab in parallel PD-L1-positive and PD-L1-negative cohorts.
The proportion of PD-L1-positive patients achieving an objective response, in phase 2; incidence of dose-limiting toxicity and the recommended phase 2 dose, in phase 1b.
Six (15%, 90% CI 7-29) of 40 patients with PD-L1-positive tumours achieved an objective response. There were no objective responders among the patients with PD-L1-negative tumours.
Not reported. The primary publication gives objective response, clinical benefit and safety in a single arm.
That second table is short because the field has tried this less often than the rationale deserves. Part of trastuzumab's effect is immune. It engages Fc receptors on effector cells, and the response to it tracks with the lymphocytic infiltrate in the tumour.
PANACEA found responses only in PD-L1-positive tumours, and in a minority of those52. Its PD-L1 was scored centrally on a metastatic biopsy, which is not how the marker is obtained in clinic. Its headline figure belongs to the PD-L1-positive cohort of the expansion stage rather than to everyone treated. Being single-arm, it cannot separate those responses from what trastuzumab alone would have done in the same selected group. KATE2 asked the randomised version of the question on a conjugate backbone, and it did not meet its primary endpoint28. It was unmasked early for futility, so follow-up is short on both arms. The PD-L1-positive subgroup its authors argue for is not reported as a subgroup result, which leaves that case resting on an analysis nobody can read. TBCRC 050 is testing a PARP inhibitor added to trastuzumab without germline selection, on the argument that HER2 amplification brings replication stress. It has not reported, so it names a strategy under test.
Neither reported trial closes the immune question. Both enrolled after HER2 blockade had already failed, which is where a lymphocytic infiltrate is least likely to be intact. The design that would settle it puts the checkpoint antibody at the front of the course. Nobody has run that trial in this subtype.
What all of this leaves for the clinic is a ranking assembled from unlike comparisons. An agent with a survival result against an active comparator has a stronger claim than one with a progression result against a floor. Brain disease, residual neuropathy, diarrhoea tolerance and hand-foot syndrome then decide between candidates of similar rank. Those are patient variables rather than trial variables, and no trial in either table helps weigh them.
5 · Central nervous system disease and the agents that reach it
The compartment now has its own chapter. Central nervous system metastatic disease carries the argument about whether the barrier explains what it is taken to explain. It also sets out how an intracranial response rate depends on who was admitted to the cohort, and what to do about leptomeningeal disease.
Two facts from that chapter bear on the choice of a systemic agent here, and only two. HER2CLIMB admitted patients whose brain metastases were untreated or progressing. Its progression-free survival in patients with brain metastases was a prespecified secondary endpoint, but its intracranial response and central nervous system progression-free survival analyses were exploratory27,53. Neratinib with capecitabine had shown intracranial activity in the phase II TBCRC 022 trial, with a composite central nervous system response rate of 49% in its main cohort54. DESTINY-Breast12 gave a full-length antibody conjugate in a single arm, to a cohort with stable or active brain metastases and to a cohort without21. It reported central nervous system progression-free survival in the first of those cohorts. There is no comparator, so it establishes activity in the compartment rather than a rank.
The practical consequence is narrow. Brain disease changes which HER2 agent you reach for first. It does not change whether the disease is still HER2-driven. It is not a reason to abandon systemic blockade in favour of local therapy alone. Everything else about the compartment belongs in Central nervous system metastatic disease and is not repeated here.
6 · Maintenance, de-escalation and treatment interruption
A patient in prolonged response faces a question the registration trials did not ask. The taxane is usually stopped for cumulative neuropathy while antibody therapy continues. That practice is near-universal and rests on very little randomised evidence. None of the first-line trials randomised the stopping point.
The chemotherapy-free version of the question has been asked twice. Neither answer transfers cleanly. TBCRC 003 gave dual blockade without chemotherapy in two cohorts split by prior treatment. The gap between those cohorts is a difference between populations rather than a comparison55. EGF104900 did randomise chemotherapy-free blockade, against a single kinase inhibitor rather than against any chemotherapy-containing regimen40. Neither of them speaks to removing the taxane from a modern triplet in a patient who is responding to it.
HER2CLIMB-05 is the closest thing this subtype has to a maintenance trial, because it randomised after induction rather than at diagnosis2. What it adds is a drug rather than permission to remove one. Nothing in the record tests subtraction directly.
Whether HER2-directed therapy itself can ever stop is unresolved. It is the same question Unresolved questions and future directions raises about defining cure in metastatic disease. No assay distinguishes eradicated disease from controlled disease, so the decision is made without the information that would settle it. Endocrine maintenance after induction does have randomised support in the hormone receptor-positive half of this subtype, and that evidence is in Metastatic triple-positive disease.
Interplay
Every de-escalation decision in this section is taken on a HER2 measurement made once, usually on the primary tumour, against a threshold chosen to select for trastuzumab. Continuing blockade through progression works because the antigen persists in most patients. The ones in whom it does not are not identified by anything in routine use, so they are treated as though they were the others. That is the ceiling argued in Where heterogeneity sets a ceiling on precision medicine appearing as a treatment-duration problem rather than a diagnostic one.
Choosing the next treatment after CDK4/6 inhibition fails
Does an activating HER2 mutation found at progression change what to give?
Rationale
A small fraction of these tumours carry an activating mutation in ERBB2. The fraction rises after endocrine therapy. The mutation switches the receptor on without raising its copy number. Immunohistochemistry and in situ hybridisation therefore both report the tumour as HER2-negative. A receptor activated by mutation should still be blockable at its kinase.
Experiment
Cohorts inside basket studies gave a HER2 kinase inhibitor to patients whose tumours carried an activating HER2 mutation without amplification. One gave an irreversible inhibitor with fulvestrant and trastuzumab, and carried comparison arms of fulvestrant alone and of fulvestrant with trastuzumab. Another gave the same irreversible inhibitor with fulvestrant, in cohorts split by whether fulvestrant had already been given. A third paired a selective HER2 kinase inhibitor with trastuzumab, adding fulvestrant for hormone receptor-positive disease.
Finding
Responses occurred in each of these cohorts, in a population defined by the mutation rather than by amplification. In the study that carried comparison arms, neither fulvestrant alone nor fulvestrant with the antibody produced a response. That places the activity in the kinase inhibitor rather than in the endocrine partner. Clinical benefit appeared whether or not fulvestrant had been given before.
Limitation
These are small single-arm cohorts inside basket studies. Each gives an estimate of activity and no comparison against what a clinic would otherwise give. One of them reports its intervals at a wider confidence convention than the rest of this registry uses. They pool mutations in different domains of the receptor, which do not all respond alike. The mutation was called by whatever assay the enrolling centre happened to run. One cohort enrolled hormone receptor-positive and hormone receptor-negative disease together and gave fulvestrant only to the first group. It reports response in each group, but the receptor-negative group is a handful of patients. Another defines its cohorts by previous fulvestrant exposure rather than by line, which leaves how heavily pretreated its patients were unrecorded.
Next question
If selecting on a genotype is what makes a small study work, what became of the resistance hypotheses that had no marker to select on?
SUMMITMutHERSGNTUC-019
Advanced disease · HR+/HER2- · later
Molecular exceptions to the endocrine sequence
Does blocking a mutant HER2 restore endocrine sensitivity, or does it work without the endocrine partner?
Rationale
HER2 mutations are enriched at relapse on endocrine treatment and in lobular disease. That pattern reads as a route out of oestrogen dependence rather than as a driver that arrived on its own. If it is a route out, blocking the receptor should return the tumour to the endocrine agent that had stopped working. The two drugs would then not be separable.
Experiment
One basket study randomised patients carrying a HER2 mutation between fulvestrant alone, fulvestrant with trastuzumab, and the triplet that adds neratinib, with crossover permitted at progression. Another asked the same question by exposure history instead of by randomisation, treating a fulvestrant-pretreated cohort separately from a fulvestrant-naive one. The tucatinib basket cohort gave fulvestrant to its hormone receptor-positive patients and withheld it from the rest.
Finding
Neither fulvestrant alone nor fulvestrant with trastuzumab produced a response. Patients who crossed over to the triplet did respond. Clinical benefit appeared in the fulvestrant-pretreated cohort as well as in the fulvestrant-naive one. Previous failure of fulvestrant therefore does not exclude benefit from a combination that contains it.
Limitation
The randomised substudy shows that the fulvestrant-based control arms were inert in this population. It cannot measure what neratinib adds, because no arm gave neratinib without fulvestrant. Neratinib alone was given only in an oestrogen receptor-negative cohort, which is a different population and the smallest of the three cohorts in that trial. The two oestrogen receptor-positive cohorts were never randomised against each other and differ in treatment history in every other respect as well. The tucatinib cohort is the one study that gave the endocrine partner to some patients and not to others. It reports response in each group, but receptor status and fulvestrant change together, and the group without fulvestrant is a handful of patients. The only within-trial comparison on offer therefore cannot be read as a test of the endocrine partner.
Next question
The shape here is general. A genomic test nominates a target. A drug is given. The response rate is read against nothing. Does that reasoning hold where the target is a DNA repair gene and a licensed drug class already exists for one version of the defect?
SUMMITMutHERSGNTUC-019
Activating HER2 mutations occur in tumours reported as HER2-negative by immunohistochemistry and in situ hybridisation. That is the whole point of the category, and the separation of biology from measurement inside it is exact. A HER2-negative report means the protein is not overexpressed and the gene is not amplified. Neither assay can see a point mutation. Finding one requires sequencing of tumour tissue or of circulating tumour DNA.
Trials entered on an activating HER2 mutation in disease the standard assays call HER2-negative.
Almost every patient in them was hormone receptor-positive, and every regimen pairs a HER2 inhibitor with endocrine therapy.
Heavily pretreated women with HER2-mutated metastatic breast cancer that was HER2-negative on local testing, enrolled as one cohort of an open-label basket study of HER2-altered solid tumours; those with hormone receptor-positive disease also received fulvestrant.
Single arm of tucatinib plus trastuzumab, with fulvestrant added for hormone receptor-positive disease.
Overall response rate.
Overall response rate was 41.9% (90% CI 26.9 to 58.2) in the 31 patients. The secondary endpoints of duration of response and progression-free survival were 12.6 months (90% CI 4.7 to not estimable) and 9.5 months (90% CI 5.4 to 13.8) respectively.
Secondary endpoint. Median overall survival was 20.1 months (90% CI 15.9 to not estimable) in this single-arm cohort, with 11 deaths, and the authors describe the data as immature.
Hormone receptor-positive, HER2-negative metastatic breast cancer with one or more activating HER2 mutations and previous CDK4/6 inhibitor therapy, enrolled as a cohort of the SUMMIT basket study. 71 patients were reported and 57 received the triplet.
Neratinib with fulvestrant and trastuzumab, against fulvestrant with trastuzumab and against fulvestrant alone. Only 21 patients were randomised, seven to each arm, and patients progressing on either fulvestrant arm could cross over to the triplet.
investigator-assessed objective response rate
objective response 39% (26-52) in the 57 patients given neratinib with fulvestrant and trastuzumab, and median progression-free survival 8.3 months (6.0-15.1). Neither fulvestrant alone nor fulvestrant with trastuzumab produced a response
Patients with HER2-mutated, HER2 non-amplified metastatic breast cancer, entered by oestrogen receptor status and previous fulvestrant exposure into a fulvestrant-treated cohort of 24, a fulvestrant-naive cohort of 11 and an exploratory oestrogen receptor-negative cohort of 5.
Single arm within each cohort, neratinib plus fulvestrant in the two oestrogen receptor-positive cohorts and neratinib monotherapy in the oestrogen receptor-negative cohort.
Clinical benefit rate in each cohort.
Clinical benefit rate was 38% (95% CI 18 to 62) in the fulvestrant-treated cohort, 30% (95% CI 7 to 65) in the fulvestrant-naive cohort and 25% (95% CI 1 to 81) in the oestrogen receptor-negative cohort.
not reported; clinical benefit rate in each cohort was the endpoint
Three features of that table matter more than any figure in it. Every regimen pairs a HER2 kinase inhibitor with an endocrine partner where the disease was hormone receptor-positive. The mutation reads as a route out of oestrogen dependence rather than as a driver that arrived on its own. Where the disease was not hormone receptor-positive, the endocrine partner was withheld. Every population was entered on a sequencing result rather than found inside a larger trial, so none of these is a subgroup. And every cohort reported activity, measured as objective response in two of the studies and as clinical benefit in the third. A receptor switched on by mutation is blockable at its kinase, and that is the class finding.
SUMMIT carries the only randomised comparison in the field, and it is small57. Read its denominators before its rate. Of the 71 patients with hormone receptor-positive, HER2-mutant disease who had progressed on a CDK4/6 inhibitor, 57 received neratinib with fulvestrant and trastuzumab. Their objective response rate was 39%, with a 95% confidence interval of 26% to 52%, and that figure and the median progression-free survival beside it are computed on the 57 rather than on the 71. Only 21 patients were randomised, seven to each arm, so the controlled part of the comparison is very small. Neither fulvestrant alone nor fulvestrant with trastuzumab produced a response, and patients crossing over to the triplet did respond. That places the activity in the kinase inhibitor rather than in the endocrine partner, which is the closest thing available to a demonstration of which component works. What the design cannot do is measure what neratinib adds, because no arm gave neratinib without fulvestrant in this population.
The three studies diverge on exposure history and on assay rather than on drug. MutHER defined its cohorts by whether fulvestrant had already been given rather than by line, which leaves how heavily pretreated its patients were unrecorded58. Clinical benefit appeared in both cohorts, so previous failure of fulvestrant does not exclude benefit from a combination containing it. SGNTUC-019 is the breast cohort of a basket study in HER2-altered solid tumours, and it gave fulvestrant to its hormone receptor-positive patients while withholding it from the rest56. It does report response by receptor group, 12 of 27 hormone receptor-positive patients given fulvestrant against 1 of 4 receptor-negative patients without it. Receptor status and fulvestrant change together, and four patients cannot carry a comparison, so the split is not a within-trial test of the endocrine partner. It also reports at 90% confidence rather than the 95% used elsewhere in this table. That is a reporting convention rather than greater precision, and it prints a narrower interval than the same data would give at 95%.
Longitudinal circulating tumour DNA in SUMMIT showed acquisition of further HER2 alterations and of PIK3CA mutations at progression57. This branch escapes the way the others do, rather than standing outside the argument in Temporal heterogeneity and clonal evolution.
Two limits apply to all three. The mutation was called by whatever assay the enrolling centre happened to run, so the entry criterion is not standardised across the table. And each study pools mutations in different domains of the receptor, which do not all respond alike. No overall survival comparison is reported anywhere in this table, and none of these designs could produce one.
Acquired HER2 mutation is also a described resistance mechanism to endocrine therapy, which connects this section to the evolution argument in Temporal heterogeneity and clonal evolution. The practical consequence is a testing decision rather than a treatment one. A patient with hormone receptor-positive disease progressing through successive endocrine lines is a patient in whom sequencing may find a target the original pathology report was never designed to look for. The mutation is also enriched in lobular histology, which makes sequencing worth requesting there on the histology alone Invasive lobular disease in the metastatic setting.
8 · Unresolved questions and what reads out next
Four questions matter most and none of them has an answer.
What follows first-line trastuzumab deruxtecan. If DESTINY-Breast09 moves that agent forward, every line below it rests on trials that enrolled patients for whom it was still available. Nobody has randomised anything in the post-conjugate setting. DESTINY-Breast07 does not fill that gap. It gives the conjugate alone and in combinations in first line, and it has no publication on the record.
Whether a mechanism other than HER2 blockade can extend what HER2 blockade does. TBCRC 050 is testing a PARP inhibitor without germline selection, on the argument that amplification brings replication stress. Nothing beyond conference material has been published. The immune route has been tried in PANACEA and KATE2 only after HER2 blockade had already failed, which is the least favourable place in the disease course to test it.
Whether HER2-directed therapy can be stopped in prolonged responders. No trial has asked the question in a randomised design. That absence reflects how hard such a trial is to run rather than how important the question is to the patients who would be in it. A patient in year six of continuous antibody therapy is being treated on an extrapolation from year one.
References
All 58 references
Tolaney SM, Jiang Z, Zhang Q, et al. Trastuzumab deruxtecan plus pertuzumab for HER2-positive metastatic breast cancer. N Engl J Med 2026 394:551-562. PMID 41160818
Dieras V, et al. HER2CLIMB-05: A Phase III Study of Tucatinib Versus Placebo in Combination With Trastuzumab and Pertuzumab as First-Line Maintenance Therapy for HER2+ Metastatic Breast Cancer. J Clin Oncol 2025 44:1597-1607. PMID 41369677
Ma F, et al. Pyrotinib versus placebo in combination with trastuzumab and docetaxel as first line treatment in patients with HER2 positive metastatic breast cancer (PHILA): randomised, double blind, multicentre, phase 3 trial. BMJ 2023 383:e076065. PMID 37907210
Miles D, et al. Final results from the PERUSE study of first-line pertuzumab plus trastuzumab plus a taxane for HER2-positive locally recurrent or metastatic breast cancer, with a multivariable approach to guide prognostication. Ann Oncol 2021 32:1245-1255. PMID 34224826
Swain SM, Miles D, Kim SB, et al. Pertuzumab, trastuzumab, and docetaxel for HER2-positive metastatic breast cancer (CLEOPATRA): end-of-study results from a double-blind, randomised, placebo-controlled, phase 3 study. Lancet Oncol 2020 21:519-530. PMID 32171426
Perez EA, et al. Trastuzumab Emtansine With or Without Pertuzumab Versus Trastuzumab Plus Taxane for Human Epidermal Growth Factor Receptor 2-Positive, Advanced Breast Cancer: Primary Results From the Phase III MARIANNE Study. J Clin Oncol 2016 35:141-148. PMID 28056202
Marty M, et al. Randomized phase II trial of the efficacy and safety of trastuzumab combined with docetaxel in patients with human epidermal growth factor receptor 2-positive metastatic breast cancer administered as first-line treatment: the M77001 study group. J Clin Oncol 2005 23:4265-74. PMID 15911866
Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med 2001 344:783-792. PMID 11248153
Swain SM, et al. Pertuzumab, trastuzumab, and docetaxel for HER2-positive metastatic breast cancer (CLEOPATRA study): overall survival results from a randomised, double-blind, placebo-controlled, phase 3 study. Lancet Oncol 2013 14:461-71. PMID 23602601
Swain SM, Baselga J, Kim SB, et al. Pertuzumab, trastuzumab, and docetaxel in HER2-positive metastatic breast cancer. N Engl J Med 2015 372:724-734. PMID 25693012
Baselga J, et al. Pertuzumab plus trastuzumab plus docetaxel for metastatic breast cancer. N Engl J Med 2011 366:109-19. PMID 22149875
Ma F, et al. Pyrotinib or placebo in combination with trastuzumab and docetaxel for HER2 positive metastatic breast cancer: long term survival results from randomised phase 3 PHILA trial. BMJ 2026 392:e087259. PMID 41839514
Bachelot T, Ciruelos E, Schneeweiss A, et al. Preliminary safety and efficacy of first-line pertuzumab combined with trastuzumab and taxane therapy for HER2-positive locally recurrent or metastatic breast cancer (PERUSE). Ann Oncol 2019 30:766-773. PMID 30796821
Dang C, Iyengar N, Datko F, et al. Phase II study of paclitaxel given once per week along with trastuzumab and pertuzumab in patients with human epidermal growth factor receptor 2-positive metastatic breast cancer. J Clin Oncol 2015 33:442-7. PMID 25547504
Perez EA, López-Vega JM, Petit T, et al. Safety and efficacy of vinorelbine in combination with pertuzumab and trastuzumab for first-line treatment of patients with HER2-positive locally advanced or metastatic breast cancer: VELVET Cohort 1 final results. Breast Cancer Res 2016 18:126. PMID 27955684
Andersson M, López-Vega JM, Petit T, et al. Efficacy and Safety of Pertuzumab and Trastuzumab Administered in a Single Infusion Bag, Followed by Vinorelbine: VELVET Cohort 2 Final Results. Oncologist 2017 22:1160-1168. PMID 28592618
Swain SM, Ewer MS, Cortés J, et al. Cardiac tolerability of pertuzumab plus trastuzumab plus docetaxel in patients with HER2-positive metastatic breast cancer in CLEOPATRA: a randomized, double-blind, placebo-controlled phase III study. Oncologist 2013 18:257-64. PMID 23475636
Cortés J, Baselga J, Im YH, et al. Health-related quality-of-life assessment in CLEOPATRA, a phase III study combining pertuzumab with trastuzumab and docetaxel in metastatic breast cancer. Ann Oncol 2013 24:2630-2635. PMID 23868905
Ciruelos E, et al. Palbociclib, trastuzumab and endocrine therapy in pretreated HER2-positive and PAM50 luminal advanced breast cancer (SOLTI-1303 PATRICIA cohort C). Clin Cancer Res 2026. PMID 41335375
Hurvitz SA, et al. Tucatinib and trastuzumab emtansine for patients with previously treated HER2-positive locally advanced and metastatic breast cancer: primary analysis of the randomized phase III trial HER2CLIMB-02. Ann Oncol 2025 37:341-352. PMID 41260264
Harbeck N, Ciruelos E, Jerusalem G, et al. Trastuzumab deruxtecan in HER2-positive advanced breast cancer with or without brain metastases: a phase 3b/4 trial. Nat Med 2024 30:3717-3727. PMID 39271844
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
André F, et al. Trastuzumab deruxtecan versus treatment of physician's choice in patients with HER2-positive metastatic breast cancer (DESTINY-Breast02): a randomised, open-label, multicentre, phase 3 trial. Lancet 2023 401:1773-1785. PMID 37086745
Cortes J, Kim SB, Chung WP, et al. Trastuzumab deruxtecan versus trastuzumab emtansine for breast cancer. N Engl J Med 2022 386:1143-1154. PMID 35320644
Rugo HS, Im SA, Cardoso F, et al. Margetuximab versus trastuzumab in patients with previously treated HER2-positive advanced breast cancer (SOPHIA): final overall survival results from a randomized phase 3 trial. J Clin Oncol 2022 41:198-205. PMID 36332179
Xu B, et al. Pyrotinib plus capecitabine versus lapatinib plus capecitabine for the treatment of HER2-positive metastatic breast cancer (PHOEBE): a multicentre, open-label, randomised, controlled, phase 3 trial. Lancet Oncol 2021 22:351-360. PMID 33581774
Murthy RK, Loi S, Okines A, et al. Tucatinib, trastuzumab, and capecitabine for HER2-positive metastatic breast cancer. N Engl J Med 2020 382:597-609. PMID 31825569
Emens LA, Esteva FJ, Beresford M, et al. Trastuzumab emtansine plus atezolizumab versus trastuzumab emtansine plus placebo in previously treated, HER2-positive advanced breast cancer (KATE2): a phase 2, multicentre, randomised, double-blind trial. Lancet Oncol 2020 21:1283-1295. PMID 33002436
Verma S, Miles D, Gianni L, et al. Trastuzumab emtansine for HER2-positive advanced breast cancer. N Engl J Med 2012 367:1783-1791. PMID 23020162
von Minckwitz G, et al. Trastuzumab beyond progression in human epidermal growth factor receptor 2-positive advanced breast cancer: a german breast group 26/breast international group 03-05 study. J Clin Oncol 2009 27:1999-2006. PMID 19289619
Geyer CE, et al. Lapatinib plus capecitabine for HER2-positive advanced breast cancer. N Engl J Med 2006 355:2733-43. PMID 17192538
Cameron D, et al. Lapatinib plus capecitabine in women with HER-2-positive advanced breast cancer: final survival analysis of a phase III randomized trial. Oncologist 2010 15:924-34. PMID 20736298
Cortés J, et al. Trastuzumab deruxtecan versus trastuzumab emtansine in HER2-positive metastatic breast cancer: long-term survival analysis of the DESTINY-Breast03 trial. Nat Med 2024 30:2208-2215. PMID 38825627
Hurvitz SA, Hegg R, Chung WP, et al. Trastuzumab deruxtecan versus trastuzumab emtansine in patients with HER2-positive metastatic breast cancer: updated results from DESTINY-Breast03, a randomised, open-label, phase 3 trial. Lancet 2023 401:105-117. PMID 36495879
Diéras V, et al. Trastuzumab emtansine versus capecitabine plus lapatinib in patients with previously treated HER2-positive advanced breast cancer (EMILIA): a descriptive analysis of final overall survival results from a randomised, open-label, phase 3 trial. Lancet Oncol 2017 18:732-742. PMID 28526536
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
Saura C, Oliveira M, Feng YH, et al. Neratinib plus capecitabine versus lapatinib plus capecitabine in HER2-positive metastatic breast cancer previously treated with two or more HER2-directed regimens: phase III NALA trial. J Clin Oncol 2020 38:3138-3149. PMID 32678716
Modi S, Saura C, Yamashita T, et al. Trastuzumab deruxtecan in previously treated HER2-positive breast cancer. N Engl J Med 2020 382:610-621. PMID 31825192
Krop IE, Kim SB, Gonzalez-Martin A, et al. Trastuzumab emtansine versus treatment of physician's choice for pretreated HER2-positive advanced breast cancer (TH3RESA): a randomised, open-label, phase 3 trial. Lancet Oncol 2014 15:689-699. PMID 24793816
Blackwell KL, et al. Overall survival benefit with lapatinib in combination with trastuzumab for patients with human epidermal growth factor receptor 2-positive metastatic breast cancer: final results from the EGF104900 Study. J Clin Oncol 2012 30:2585-92. PMID 22689807
Robert N, Leyland-Jones B, Asmar L, et al. Randomized phase III study of trastuzumab, paclitaxel, and carboplatin compared with trastuzumab and paclitaxel in women with HER-2-overexpressing metastatic breast cancer. J Clin Oncol 2006 24:2786-92. PMID 16782917
Burstein HJ, Keshaviah A, Baron AD, et al. Trastuzumab plus vinorelbine or taxane chemotherapy for HER2-overexpressing metastatic breast cancer: the trastuzumab and vinorelbine or taxane study. Cancer 2007 110:965-72. PMID 17614302
Yamamoto D, Iwase S, Kitamura K, Odagiri H, Yamamoto C, Nagumo Y. A phase II study of trastuzumab and capecitabine for patients with HER2-overexpressing metastatic breast cancer: Japan Breast Cancer Research Network (JBCRN) 00 Trial. Cancer Chemother Pharmacol 2008 61:509-14. PMID 17516068
Schaller G, Fuchs I, Gonsch T, et al. Phase II study of capecitabine plus trastuzumab in human epidermal growth factor receptor 2 overexpressing metastatic breast cancer pretreated with anthracyclines or taxanes. J Clin Oncol 2007 25:3246-50. PMID 17577021
Seidman AD, Berry D, Cirrincione C, et al. Randomized phase III trial of weekly compared with every-3-weeks paclitaxel for metastatic breast cancer, with trastuzumab for all HER-2 overexpressors and random assignment to trastuzumab or not in HER-2 nonoverexpressors: final results of Cancer and Leukemia Group B protocol 9840. J Clin Oncol 2008 26:1642-9. PMID 18375893
Seidman A, Hudis C, Pierri MK, et al. Cardiac dysfunction in the trastuzumab clinical trials experience. J Clin Oncol 2002 20:1215-21. PMID 11870163
Curigliano G, et al. Tucatinib versus placebo added to trastuzumab and capecitabine for patients with pretreated HER2+ metastatic breast cancer with and without brain metastases (HER2CLIMB): final overall survival analysis. Ann Oncol 2021 33:321-329. PMID 34954044
Blackwell KL, Burstein HJ, Storniolo AM, et al. Randomized study of Lapatinib alone or in combination with trastuzumab in women with ErbB2-positive, trastuzumab-refractory metastatic breast cancer. J Clin Oncol 2010 28:1124-30. PMID 20124187
Krop IE, et al. Trastuzumab emtansine versus treatment of physician's choice in patients with previously treated HER2-positive metastatic breast cancer (TH3RESA): final overall survival results from a randomised open-label phase 3 trial. Lancet Oncol 2017 18:743-754. PMID 28526538
Ciruelos E, Villagrasa P, Pascual T, et al. Palbociclib and Trastuzumab in HER2-Positive Advanced Breast Cancer: Results from the Phase II SOLTI-1303 PATRICIA Trial. Clin Cancer Res 2020 26:5820-5829. PMID 32938620
Rugo HS, et al. Efficacy of Margetuximab vs Trastuzumab in Patients With Pretreated ERBB2-Positive Advanced Breast Cancer: A Phase 3 Randomized Clinical Trial. JAMA Oncol 2021 7:573-584. PMID 33480963
Loi S, Giobbie-Hurder A, Gombos A, et al. Pembrolizumab plus trastuzumab in trastuzumab-resistant, advanced, HER2-positive breast cancer (PANACEA): a single-arm, multicentre, phase 1b-2 trial. Lancet Oncol 2019 20:371-382. PMID 30765258
Lin NU, Borges V, Anders C, et al. Intracranial efficacy and survival with tucatinib plus trastuzumab and capecitabine for previously treated HER2-positive breast cancer with brain metastases in the HER2CLIMB trial. J Clin Oncol 2020 38:2610-2619. PMID 32468955
Freedman RA, Gelman RS, Anders CK, et al. TBCRC 022: A Phase II Trial of Neratinib and Capecitabine for Patients With Human Epidermal Growth Factor Receptor 2-Positive Breast Cancer and Brain Metastases. J Clin Oncol 2019 37:1081-1089. PMID 30860945
Lin NU, Guo H, Yap JT, et al. Phase II Study of Lapatinib in Combination With Trastuzumab in Patients With Human Epidermal Growth Factor Receptor 2-Positive Metastatic Breast Cancer: Clinical Outcomes and Predictive Value of Early [18F]Fluorodeoxyglucose Positron Emission Tomography Imaging (TBCRC 003). J Clin Oncol 2015 33:2623-31. PMID 26169615
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
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
Ma CX, Luo J, Freedman RA, et al. The Phase II MutHER Study of Neratinib Alone and in Combination with Fulvestrant in HER2-Mutated, Non-amplified Metastatic Breast Cancer. Clin Cancer Res 2022 28:1258-1267. PMID 35046057