--- id: BC-830 title: "Neoadjuvant therapy" subtitle: "What can be taken out of a preoperative HER2 regimen, and what the specimen cannot license." status: drafted updated: 2026-09-26 ---
1 · Scope, and why this subtype made the neoadjuvant platform work
Early disease · HER2+ · neoadjuvant
How much chemotherapy a HER2-positive tumour still needs
Is a pathological complete response worth raising, or is it only a marker of the tumour the patient already had?
Rationale
Preoperative and postoperative chemotherapy were compared head to head in operable disease. The order changed nothing about survival. So the reason to treat first is not more cure. It is downstaging of the breast and the axilla. It is also information, because in HER2-positive disease the response is graded on the surgical specimen. That grade separates outcomes steeply.
Experiment
The order of the same chemotherapy was randomised in operable disease of every subtype. Within HER2-positive disease, trastuzumab was added to a neoadjuvant anthracycline and taxane backbone in locally advanced and inflammatory tumours, with event-free survival rather than a pathological endpoint as the primary measure. A platform trial with a shared chemotherapy control then followed patients by whether they reached a complete response. Separately, residual invasive disease found at surgery was used as the entry criterion for a trial that randomised what was given afterwards.
Finding
The order of treatment does not change survival. What the specimen shows does. Patients who reach a complete response have few events. Those left with residual disease have many. Acting on that residual disease with a conjugate lowers recurrence against continuing the antibody. Adding trastuzumab to neoadjuvant chemotherapy raised event-free survival against the same chemotherapy alone, which is the one trial here that measured cure rather than the specimen.
Limitation
The association between response and outcome is measured within patients rather than between treatments. A drug that raises the proportion reaching a complete response has not thereby been shown to lower recurrence. The clearest counterexample sits in this cell. The lapatinib and trastuzumab pair raised pathological complete response against trastuzumab before surgery, then failed to improve disease-free survival against trastuzumab when the same combination was tested adjuvantly. The trial that fixed the equivalence of the two orders predates HER2 testing and used chemotherapy nobody now gives. It licenses treating first in general rather than this backbone in particular. The trial that added trastuzumab before surgery enrolled locally advanced and inflammatory disease, which is not the operable population in which the order question was settled.
Next question
If the endpoint is worth raising at all, what raises it?
NSABP B-18NOAHI-SPY 2KATHERINENeoALTTOALTTO
Almost every methodological advantage of the neoadjuvant setting is larger here than anywhere else in breast cancer.
The tumour is in place, so response is measured rather than inferred. Response arrives in months rather than years, at rates high enough for a few hundred patients to separate two regimens. And response carries more prognostic information in this subtype than in any other. The pooled analysis of 12 trials and 11,955 patients found that association strongest in HER2-positive, hormone receptor-negative disease treated with trastuzumab. There the event-free survival hazard ratio, comparing patients who reached a complete response with those who did not, was 0.15. Across the whole pooled population it was 0.48, and for overall survival 0.361. Nowhere else in breast cancer does the specimen carry that much information about an individual.
That is why the questions asked in this cell are subtractive ones. Most patients here are cured by chemotherapy given with HER2 antibodies. Most of what the cure costs them comes from the chemotherapy rather than from the antibodies. Treating first does not itself add cure, because randomising the order of the same chemotherapy in operable disease changed neither disease-free nor overall survival2. That trial predates HER2 testing and used chemotherapy nobody now gives, so it licenses treating first in general rather than this backbone in particular. The reasons to treat first are downstaging of the breast and the axilla, and information.
The information is the part that changed practice. The specimen grades the response and the grade routes what follows, because residual invasive disease is the entry criterion for post-neoadjuvant intensification Post-neoadjuvant intensification for residual disease. Subtraction is therefore testable here in a way it is not after surgery, because a tumour that fails to respond announces itself before the operation. Chemotherapy-free and de-escalated neoadjuvant regimens records what happens when it announces itself too late.
Two consequences follow and they pull in opposite directions.
The first is productive. This is where HER2-directed drug development happens on a signal that is obtainable quickly and cheaply. Most of the agents in Targeted therapy and Antibody-drug conjugates and targeted delivery were first tested against this endpoint.
Each section below opens with the question its trials were built to answer, from the [Chapter Stories](stories.html) map, and closes with a table filtered on that same question. Result cells come from the publication named in the reference column. Cross-trial comparison is argued in prose, because the populations and the scoring conventions differ.
2 · Chemotherapy backbones with trastuzumab
How much chemotherapy a HER2-positive tumour still needs, continued
Can the anthracycline come out of the neoadjuvant backbone, and what should stand in its place?
Rationale
An anthracycline costs cardiac function and carries a small leukaemia risk. It is given alongside an antibody that itself depresses left ventricular function. The pivotal adjuvant trial that ran a trastuzumab regimen with and without an anthracycline had already found the two close. That made the omission worth testing where the readout is fast.
Experiment
A cardiac safety trial ran three dual-blockade regimens before surgery, one of them carrying no anthracycline, with the pathological endpoint secondary. A randomised trial then varied the anthracycline alone, on a fixed taxane, carboplatin and dual-blockade backbone, with response as the endpoint and event-free survival on follow-up. Separately, carboplatin was added to a taxane and anthracycline backbone in a trial enrolling both triple-negative and HER2-positive disease. A further trial set weekly nab-paclitaxel with dual blockade against docetaxel and carboplatin with dual blockade, neither arm carrying an anthracycline. A non-inferiority trial then kept a three-weekly taxane and dual blockade in both arms and randomised the carboplatin. The taxane formulation was isolated on its own in an all-comers trial where both arms kept the anthracycline.
Finding
Dropping the anthracycline did not cost event-free survival against keeping it, on a backbone that was otherwise identical. In the cardiac safety trial the anthracycline-free regimen produced responses in the same range as the two regimens containing one. Symptomatic left ventricular systolic dysfunction appeared there only in the arms that carried an anthracycline. Adding platinum raised response in triple-negative disease and added nothing in the HER2-positive stratum of the same trial, with a formal test of interaction separating the two. Removing carboplatin from an anthracycline-free taxane and dual-blockade regimen was non-inferior for complete response and caused severe toxicity less often. Among anthracycline-free regimens, weekly nab-paclitaxel with dual blockade raised response against docetaxel and carboplatin with dual blockade. Substituting nab-paclitaxel for solvent-based paclitaxel also raised response in the all-comers trial that changed nothing else.
Limitation
The trial that varied the anthracycline alone was sized on the pathological endpoint. Its event-free survival comparison is a follow-up analysis on few events. The absence of a difference there is not a demonstration of equivalence. The cardiac safety trial put response second and never compared its arms formally, so its three rates sit beside each other rather than against each other. The comparison that favoured nab-paclitaxel in HER2-positive disease changed two things at once, the taxane formulation and the carboplatin. It cannot say which change produced the gain. It also enrolled Asian patients only. The trial that isolates the formulation kept the anthracycline in both arms and enrolled every subtype, so it settles that question outside this cell rather than inside it. The platinum question was first answered in a stratum of a trial powered for the triple-negative population. The trial later powered for it in HER2-positive disease gave a higher docetaxel dose in the arm without carboplatin. Its tolerability gain therefore cannot be credited to the omission alone. It was sized on the specimen rather than on survival, and it enrolled patients in mainland China only.
Next question
If the anthracycline is dispensable, is the rest of the cytotoxic backbone dispensable too?
Of the six trials in this table, only one was built to test the anthracycline itself. One more was built and powered to test what the regimen that replaces it still needs.
Start with what is not in dispute. NOAH is the only row whose primary endpoint was survival rather than the specimen. Adding trastuzumab to an anthracycline and taxane backbone improved event-free survival against the same chemotherapy alone3. It enrolled locally advanced and inflammatory disease, where the event rate was high enough to show that. A pooled analysis of two randomised trials found that concomitant trastuzumab doubled the rate of complete response in breast and nodes, with a relative risk of 2.07, and lowered the risk of relapse, without a significant increase in cardiac events4. Every other row argues about what to put beside the antibody rather than about whether to give it.
The anthracycline is the contested part, and the rows disagree about it for reasons of design rather than of drug.
TRAIN-2 is load-bearing, because it exchanged three cycles of an anthracycline regimen for three more cycles of taxane and carboplatin, with the same dual blockade in both arms, then followed patients for event-free survival5. Its primary endpoint, pathological complete response, had already reported without a meaningful advantage for the anthracycline6. Read its interval rather than its point estimate. The hazard ratio for an event-free survival event ran from 0.50 to 1.63. That excludes neither a substantial benefit from omitting the anthracycline nor a substantial harm. No non-inferiority margin was set in advance. The trial was sized on the specimen, and the survival comparison is a follow-up analysis on few events.
So the case for omission does not rest on demonstrated equivalence. It rests on a harm measured in the same randomisation. An ejection fraction fall of 10 points or more to below 50% occurred in 7.7% of patients given an anthracycline and in 3.2% of those who were not5. Designing de-escalation trials and the non-inferiority margin problem sets out why a measured harm beside an unmeasured efficacy risk is a stronger argument than a null result would be.
TRYPHAENA is quoted alongside it and should not be. It was powered as a cardiac safety study. Its response rates were secondary and its three arms were never compared formally7. Three rates sitting beside each other are a compatibility result. Its cardiac readout is the part worth quoting, and it points the same way as the randomised cardiac difference in TRAIN-2.
The remaining three rows vary the cytotoxic partner rather than the anthracycline. HELEN-006 set weekly nab-paclitaxel with dual blockade against docetaxel and carboplatin with dual blockade, neither arm carrying an anthracycline. The nab-paclitaxel arm responded more often8. It changed the taxane formulation and the carboplatin at once, so it cannot say which change produced the gain. It also enrolled Asian women only, which fixes the comparator rather than widening it. GeparSixto added carboplatin to a taxane and anthracycline backbone in a trial enrolling triple-negative and HER2-positive disease together. Platinum raised response in the triple-negative population and added nothing in the HER2-positive stratum, with a formal interaction test separating the two9. That stratum was not what the trial was powered for, so the null there is weak evidence rather than none.
neoCARHP is the trial that was powered for the platinum question in this subtype. It kept a three-weekly taxane and dual blockade in both arms and randomised the carboplatin, against a prespecified non-inferiority margin for complete response10. Omission met that margin with less frequent severe toxicity than the platinum regimen caused. It therefore tests the carboplatin, one of the two changes HELEN-006 made at once, without the other. The separation is not complete, because the docetaxel dose was higher in the arm without carboplatin. Part of the tolerability gain may belong to the taxane dose rather than to the omission. The trial enrolled women in mainland China, most with stage II disease, and has not reported survival10.
What the table cannot show is a survival comparison for any of the three rows that varied the cytotoxic partner. HELEN-006 and neoCARHP have not reported one, and GeparSixto's survival result belongs to its triple-negative population rather than to the HER2-positive stratum beside it. Nothing here randomises the taxane schedule outside Asia. Nothing randomises the anthracycline inside heavy nodal disease, which is where an interval as wide as TRAIN-2's leaves the most room.
In practice
Under dual blockade, treat anthracycline omission as the default question rather than the exception. The toxicity evidence is randomised and the efficacy evidence shows no signal of loss.
Do not describe the omission as proven non-inferior. Describe it as a trade of an unmeasured and probably small efficacy risk against a measured cardiac one.
Keep the anthracycline in reach for very high-volume nodal disease. Nobody randomised the question inside that group, and events are common enough there for the width of TRAIN-2's interval to matter.
Neoadjuvant trials that vary the anthracycline or run without one.
Only TRAIN-2 randomises the anthracycline and carries a survival endpoint.
Women aged 18 or older with previously untreated stage II to III HER2-positive invasive breast cancer; metastatic, inflammatory and bilateral disease excluded.
Six 3-weekly cycles of investigator-selected taxane with trastuzumab and pertuzumab (THP) versus the same taxane with carboplatin AUC 6, trastuzumab and pertuzumab (TCbHP); docetaxel 100 mg/m2 without carboplatin, 75 mg/m2 with it.
Pathological complete response (ypT0/is ypN0) in the modified intention-to-treat population, non-inferiority margin -10%.
Pathological complete response 64.1% (245 of 382) with THP versus 65.9% (253 of 384) with TCbHP, absolute difference -1.8%, 95% CI -8.5 to 5.0, so the lower bound lies within the -10% non-inferiority margin.
Not yet reported. Overall and event-free survival are secondary endpoints, and survival follow-up is ongoing.
Asian patients aged 18 to 70 with untreated stage II to III invasive HER2-positive breast cancer and ECOG performance status 0 or 1.
Weekly nab-paclitaxel with trastuzumab and pertuzumab versus docetaxel plus carboplatin with trastuzumab and pertuzumab.
Pathological complete response, defined as ypT0/is ypN0.
Pathological complete response occurred in 66.3% of patients given nab-paclitaxel (220 of 332) versus 57.6% given docetaxel plus carboplatin (194 of 337), odds ratio 1.54, 95% CI 1.10 to 2.14, stratified p=0.011.
Not yet reported. Follow-up of the adjuvant phase is ongoing.
HER2+ operable, locally advanced or inflammatory breast cancer
three neoadjuvant regimens of pertuzumab and trastuzumab, two anthracycline-containing and one anthracycline-free
cardiac safety, with pathological complete response as a secondary endpoint
symptomatic left ventricular systolic dysfunction in 2 of 225 patients, both in an anthracycline-containing arm; pathological complete response 61.6%, 57.3% and 66.2% across the three arms
HER2+ locally advanced or inflammatory breast cancer, with a parallel HER2-negative cohort of 99 patients
neoadjuvant and adjuvant trastuzumab added to doxorubicin, paclitaxel, cyclophosphamide, methotrexate and fluorouracil vs the same chemotherapy alone
event-free survival
3-year event-free survival 71% with trastuzumab against 56% without, HR 0.59 (0.38-0.90), p=0.013
not separately reported at the primary analysis
3 · Dual blockade and what the pathological complete response gains bought
How much chemotherapy a HER2-positive tumour still needs, continued
Does adding a further drug to trastuzumab and a taxane, or replacing trastuzumab with a better-engineered antibody, raise pathological complete response?
Rationale
Trastuzumab binds one epitope. It does not prevent ligand-driven dimerisation. Pertuzumab blocks the dimerisation face. A small-molecule inhibitor blocks the kinase from inside the cell. Engineering the Fc region raises affinity for the activating receptor on immune effector cells. Releasing the brake on T cells asks the same question from the other side, which is whether one antibody leaves HER2 signalling, or immune killing, on the table.
Experiment
A second HER2-directed drug was added to trastuzumab on a taxane backbone in separate randomised trials of pertuzumab, of lapatinib and of pyrotinib, each reporting a pathological endpoint. One trial substituted an Fc-engineered antibody for trastuzumab alongside pertuzumab and paclitaxel. One added a checkpoint inhibitor instead of a second HER2-directed drug, with co-primary analyses in all comers and in the PD-L1-positive subset.
Finding
Pertuzumab added to trastuzumab and docetaxel raises the proportion reaching a complete response against trastuzumab and docetaxel, in the trial that established the pair and in the trial that repeated it. Pyrotinib added to that same backbone does the same against placebo. Lapatinib is not consistent. Added to trastuzumab on a taxane it raised response against trastuzumab in one trial and fell short of significance in another. The kinase inhibitors buy what they gain with diarrhoea. Improving the antibody rather than adding to it did not raise response against trastuzumab in the same regimen. Nor did adding a checkpoint inhibitor, in either of its co-primary populations.
Limitation
The pathological endpoint is not scored the same way across these trials. Some require only the breast to be clear. Others require the nodes as well. The rates therefore cannot be set side by side, even where the regimens nearly match. The backbones differ too, from docetaxel alone, to paclitaxel after an anthracycline, to a dose-dense anthracycline followed by paclitaxel. The trial that repeated the pertuzumab result enrolled an Asian population, which fixes the comparator rather than widening it. The third lapatinib trial, the one that gave a taxane with no anthracycline, has published long-term survival and gene expression rather than the pathological comparison it was designed around. It cannot be counted on either side of the lapatinib question. The Fc-engineering result exists only as conference and registry material, with no primary publication. Its unsuccessful comparison cannot be read at the level of detail the others allow.
Next question
If the antibodies are doing this much of the work, how much of the chemotherapy beside them is still earning its place?
Adding a second drug to trastuzumab on a taxane raises the proportion of patients reaching a pathological complete response. That is the class finding and most of these rows carry it. What the table cannot do is put the rates on one axis.
Compare the endpoint column before the result column. NeoSphere, NeoALTTO, NSABP B-41 and CALGB 40601 scored the breast alone. PEONY, PHEDRA and IMpassion050 required the nodes to be clear as well. The backbones differ by more than the endpoint does, from docetaxel alone, to paclitaxel after doxorubicin and cyclophosphamide, to a dose-dense anthracycline followed by paclitaxel, to six weeks of antibody before any chemotherapy at all. Three of the seven ran a placebo control. The rest were open-label.
The lapatinib rows show what that does to a comparison. Added to trastuzumab and a taxane, lapatinib raised breast response decisively in NeoALTTO, then missed significance in NSABP B-41 and again in CALGB 40601. The drug and the dose were not the variable. The control arms were. The trastuzumab control reached 29.5% in NeoALTTO, against 52.5% in NSABP B-41 and 46% in CALGB 40601. NeoALTTO gave six weeks of antibody alone and no anthracycline before surgery. B-41 gave four cycles of doxorubicin and cyclophosphamide first. A second HER2-directed drug has most room where the control arm leaves most residual disease, which is a statement about control arms rather than about lapatinib.
The row the pertuzumab decision rests on is NeoSphere, because it is the randomised comparison against trastuzumab and docetaxel, which is what the clinic would otherwise give. PEONY repeated that comparison with a stricter endpoint, independent central review and a placebo control11. It enrolled Asian women, so it fixes the comparator rather than widening the population. PHEDRA asked the same question with a kinase inhibitor and answered it the same way. The gain was bought with diarrhoea, grade 3 or worse in 44.4% of the pyrotinib group against 5.1% of the placebo group12.
IMpassion050 added something other than a HER2-directed drug and is the only unambiguous negative. Atezolizumab changed nothing in either co-primary population, and the PD-L1-positive analysis pointed the wrong way13. The five grade 5 events at the primary analysis all occurred in the atezolizumab arm. MARGOT / TBCRC 052 put an Fc-engineered antibody in place of trastuzumab beside pertuzumab and paclitaxel. It has no primary publication and no tabulated result, so it settles nothing in either direction.
Now follow the response gains to their outcomes, because that is the reason this section exists.
NeoSphere was followed for five years. Five-year progression-free survival was 86% in the pertuzumab, trastuzumab and docetaxel group and 81% in the trastuzumab and docetaxel group14. There were roughly 107 patients per arm, so the trial could never have measured an outcome difference of that size. The trial that generates a response difference is never the trial that can price it.
NeoALTTO is the cleaner case, because its drug pair was then tested adjuvantly at scale. ALTTO randomised more than eight thousand patients and missed its disease-free survival threshold against trastuzumab alone15. The same pair raised the specimen endpoint before surgery and did not raise disease-free survival after it.
CALGB 40601 is the same failure running the other way. Its primary response comparison did not reach significance16. Its survival comparison, from the same randomisation and at more than seven years, did, for relapse-free and for overall survival17. A reader who stopped at the specimen would have concluded the drug did nothing.
Put the three together and the endpoint is not biased in a fixed direction. It can overstate an outcome gain and it can understate one, which is what makes correcting for it impossible. Across 12 trials, the correlation between a treatment's increase in pathological complete response and its effect on event-free survival had a coefficient of determination of 0.03. For overall survival it was 0.24, and the authors judged neither sufficient to validate the endpoint1. Endpoints, surrogates, and the overall survival question sets out why that is the quantity a surrogate has to satisfy.
Caution
None of this argues against dual blockade, since pertuzumab produces a real adjuvant benefit in node-positive disease Dual blockade in the adjuvant setting and the neoadjuvant response gain pointed in the right direction. What fails is the inference from the size of a response gain to the size of an outcome gain. A regimen that raises pathological complete response has not thereby been shown to improve survival by any particular amount.
The overall survival column would settle these questions and it is almost entirely blank. No trial here randomised a second HER2-directed drug before surgery with survival as its primary endpoint, and none compared two second drugs against each other. A response improvement licenses a hypothesis about outcome. It licenses a change of regimen only where response itself changes the next treatment, which in this subtype it does Post-neoadjuvant intensification for residual disease.
Trials that add a further agent to trastuzumab on a taxane.
Compare the endpoint column before the result column, because breast-only and breast-and-node definitions both appear here.
Atezolizumab versus placebo, each added to dose-dense doxorubicin and cyclophosphamide followed by paclitaxel with pertuzumab and trastuzumab.
Coprimary pathologic complete response rate (ypT0/is ypN0) in the intention-to-treat and PD-L1-positive populations.
Pathologic complete response was 62.4% with atezolizumab versus 62.7% with placebo in the intention-to-treat population, a difference of -0.33% (95% CI -9.2 to 8.6, P = .9551). In PD-L1-positive tumours it was 64.2% versus 72.5%, a difference of -8.26% (95% CI -20.6 to 4.0, P = .1846). At the final analysis, 3-year event-free survival was 91.4% with atezolizumab versus 89.0% with placebo, stratified hazard ratio 0.90 (95% CI 0.50 to 1.59). Among the 434 patients who had surgery and postneoadjuvant therapy, 3-year disease-free survival was 92.9% versus 88.5%, unstratified hazard ratio 0.71 (95% CI 0.38 to 1.32).
Overall survival was not reported in this primary analysis.
Neoadjuvant pertuzumab with trastuzumab and docetaxel versus placebo with trastuzumab and docetaxel.
Total pathological complete response rate assessed by an independent review committee.
Total pathological complete response was 39.3%, 86 of 219 patients, with pertuzumab versus 21.8%, 24 of 110 patients, with placebo. The difference was 17.5% (95% CI 6.9-28.0, p=0.001).
Overall survival was a secondary endpoint of the final analysis at 62.9 months median follow-up, but no overall survival estimate is given in the abstract.
Weekly paclitaxel with trastuzumab plus lapatinib versus paclitaxel with trastuzumab versus paclitaxel with lapatinib.
Pathological complete response.
Pathological complete response in the breast was 56% (95% CI 47 to 65) with paclitaxel, trastuzumab and lapatinib against 46% (95% CI 37 to 55) with paclitaxel and trastuzumab, P = .13. Adding lapatinib raised response in hormone receptor-negative disease (P = .01) and not in hormone receptor-positive disease.
In the later long-term analysis, at more than 7 years of follow-up, overall survival favoured paclitaxel with trastuzumab and lapatinib over paclitaxel with trastuzumab, hazard ratio 0.34 (95% CI 0.12 to 0.94), P = .037, with relapse-free survival hazard ratio 0.32 (95% CI 0.14 to 0.71), P = .005.
Weekly paclitaxel with lapatinib, or with trastuzumab plus lapatinib, versus weekly paclitaxel with trastuzumab, all after doxorubicin and cyclophosphamide.
Pathological complete response in the breast.
Breast pathological complete response was 53.2 percent with lapatinib and 62.0 percent with trastuzumab plus lapatinib, versus 52.5 percent with trastuzumab (P=0.9852 and P=0.095).
Overall survival was not reported in this publication, which addressed pathological complete response at surgery.
neoadjuvant lapatinib with trastuzumab vs trastuzumab alone or lapatinib alone, each followed by the addition of weekly paclitaxel
pathological complete response
pathological complete response 51.3% with dual blockade against 29.5% with trastuzumab, a difference of 21.1 percentage points (9.1-34.2), p=0.0001; 24.7% with lapatinib alone
pertuzumab, trastuzumab and docetaxel vs trastuzumab and docetaxel, pertuzumab and trastuzumab, or pertuzumab and docetaxel
pathological complete response in the breast
pathological complete response in the breast 45.8% with pertuzumab, trastuzumab and docetaxel against 29.0% with trastuzumab and docetaxel, p=0.0141; 24.0% with pertuzumab and docetaxel and 16.8% with pertuzumab and trastuzumab alone
not reported at the primary analysis
4 · Chemotherapy-free and de-escalated neoadjuvant regimens
How much chemotherapy a HER2-positive tumour still needs, continued
Can conventional chemotherapy be left out altogether and the HER2 block given alone?
Rationale
Some HER2-positive tumours depend on the receptor to the point of addiction. In the trial that established the antibody pair, the arm given both antibodies without a taxane still produced complete responses in a minority of patients. If those tumours can be identified, the cytotoxic drug is being given to them for nothing.
Experiment
Chemotherapy-free dual blockade was given for a fixed window before surgery, with endocrine therapy added for oestrogen receptor-positive disease, across a series of single-arm and randomised phase II trials. One randomised the length of that window, a shorter course against a longer one. One asked whether the intrinsic subtype called from the diagnostic biopsy predicts who responds. One tested formally whether early responders given dual blockade alone could be spared the taxane.
Finding
Chemotherapy-free blockade produces complete responses in a minority. That minority is larger with the longer window than with the shorter one. It is larger in oestrogen receptor-negative than in receptor-positive disease. It is larger in tumours called HER2-enriched than in the rest. The trial that tested omission formally did not show non-inferiority against the taxane. On that test the taxane was not dispensable, even in early responders.
Limitation
Most of these trials have no chemotherapy arm and no survival endpoint. Nothing in them can support equivalence to chemotherapy. The duration trial combined a phase II design in its experimental arm with a pick-the-winner design and was explicitly not powered to compare the two windows. The gap between them is a numerical difference that was never tested. The trial that did test omission set early responders in the chemotherapy-free arm against every patient in the taxane arm, which compares a selected group with an unselected one. It also enrolled hormone receptor-negative disease only. The subtype trial had no comparator. A subtype that identifies responders to chemotherapy-free blockade has not thereby been shown to identify patients who can forgo chemotherapy. The single-arm windows and the duration trial all used lapatinib with trastuzumab rather than the antibody pair now given, so their rates do not transfer to the regimen a clinic would actually use.
Next question
If blanket omission fails, can the patients who can safely omit be picked out instead, either before treatment or from their response to it?
Three different things get called de-escalation here and they carry different risks. Removing the anthracycline, or the carboplatin that stands in its place, is Chemotherapy backbones with trastuzumab. Removing all conventional cytotoxic chemotherapy is this section. Removing treatment after a response has already been seen is Response-adapted strategies and imaging-guided escalation.
The idea is not unreasonable. In the trial that established the antibody pair, the arm given both antibodies without a taxane still produced complete responses in a minority20. If those tumours could be identified in advance, the cytotoxic drug would be going to them for nothing.
Seven rows have tested it and they agree on the shape of the effect. Chemotherapy-free blockade clears the breast in a minority of patients. That minority is larger with a longer window21, larger in oestrogen receptor-negative disease22, and larger in tumours called HER2-enriched by an intrinsic subtype call23.
They diverge on design rather than on regimen, and the divergence decides what each can claim. TBCRC 006 and PAMELA are single-arm. TBCRC 023 randomised the duration of blockade and not the blockade itself. PHERGain-2 is single-arm and adapts after surgery. PHERGain is chemotherapy-free only for patients whose tumours respond on imaging Response-adapted strategies and imaging-guided escalation. KRISTINE substituted a conjugate rather than omitting anything.
One row tests the actual question and it failed. WSG-ADAPT HER2+/HR− randomised twelve weeks of dual blockade with or without weekly paclitaxel and tested non-inferiority formally24. Read the design before the rate. The test set early responders in the chemotherapy-free arm against every patient in the taxane arm, which compares a selected group with an unselected one. It enrolled hormone receptor-negative disease only, the half of the subtype most likely to respond. The taxane still won. Five-year overall survival was 98% with the taxane and 94% without, on 134 patients, with a hazard ratio whose interval ran from 0.05 to 3.6325. That interval is the honest summary of what a trial this size can say about survival.
A drug difference runs across the rows and is easy to miss. TBCRC 006, TBCRC 023 and PAMELA gave lapatinib with trastuzumab. WSG-ADAPT, PHERGain and PHERGain-2 gave pertuzumab with trastuzumab. Rates from the first group do not transfer to the regimen a clinic now uses.
Duration is the only variable inside chemotherapy-free blockade that anyone randomised. TBCRC 023 doubled the window from twelve to twenty-four weeks and the rate rose, driven by the oestrogen receptor-positive subgroup21. It combined a phase II design in its experimental arm with a pick-the-winner design and was explicitly not powered to compare the two windows. The gap between them was never tested.
KRISTINE shows what removing chemotherapy costs when something is put in its place. Its three-year follow-up gave a hazard ratio for an event-free survival event of 2.61 against the conjugate, with an interval from 1.36 to 4.98. Locoregional progression before surgery occurred in the conjugate arm and not in the chemotherapy arm26. Once surgery had happened, invasive disease-free survival was similar between the arms.
That is a failure mode peculiar to the preoperative window. The tumour is still in the patient, so an underpowered regimen does not merely fail to shrink it. It permits local progression during the months the regimen runs. A de-escalation acceptable after surgery is not automatically acceptable before it.
One feature runs through every protocol here and changes what the whole table is evidence for. Where these trials say what follows surgery, a patient with residual disease went on to receive chemotherapy, a conjugate, or both24,27,28. What each of them tested is therefore deferral of chemotherapy in patients who turned out not to need it. None reports a non-responder left untreated. A reader who takes these rows as evidence that chemotherapy can be dropped has read a response-adapted strategy as a subtraction.
In practice
Anthracycline omission is the de-escalation with randomised survival evidence behind it5. It is a different proposition from this one and should be asked first.
Chemotherapy omission remains a trial question outside a trial. Where it is used off protocol, say that the supporting evidence is phase 2 and response-adapted.
Do not offer a chemotherapy-free strategy to a patient who would decline chemotherapy for residual disease. Every one of these strategies was built on giving it.
Caution
Eligibility does more work in this table than any drug does. TBCRC 006 enrolled stage II to III disease with a median tumour of 6 cm22. PHERGain-2 required immunohistochemistry 3+, node-negative disease and a tumour of 5 to 30 mm on magnetic resonance imaging28. Two single-arm rates from populations that far apart are not measurements of the same thing, and they are routinely quoted as though they were.
What the table cannot show is any comparison of chemotherapy-free blockade against chemotherapy with survival as the endpoint. Nobody randomised one. Nobody enrolled an assay-selected population prospectively and then randomised the taxane inside it, so the strongest observation here, that HER2-enriched tumours respond far more often, has never been used to assign treatment. And the one row that put omission to a formal test enrolled hormone receptor-negative disease only. The group in which the offer is most tempting and the response least likely has never faced that test Lower pathological complete response rates and what that does to the post-neoadjuvant decision.
Preoperative regimens that remove conventional cytotoxic chemotherapy.
Only the WSG-ADAPT row carries a formal test against a taxane, and it did not pass.
treatment-naive, HER2 immunohistochemistry 3+, node-negative disease with tumours of 5 to 30 mm on magnetic resonance imaging
single-arm neoadjuvant trastuzumab and pertuzumab, with adjuvant therapy assigned by pathological response and no randomised comparator
1-year decline in health-related quality of life and 3-year recurrence-free interval
pathological complete response in 236 of 396 patients, which is 59.6%; global quality of life fell by 10% or more in 42.8% (36.9-48.8); the recurrence-free interval endpoint is not yet reported
centrally confirmed HER2-positive stage I-IIIA with at least one PET-evaluable lesion; randomised 1:4, stratified by hormone receptor status
trastuzumab and pertuzumab with or without endocrine therapy, with chemotherapy added before surgery on failure to respond on imaging and after surgery for imaging responders without a pathological complete response, vs docetaxel, carboplatin, trastuzumab and pertuzumab
pathological complete response in PET responders, and 3-year invasive disease-free survival in the response-adapted group
3-year invasive disease-free survival 94.8% (91.4-97.1) in the response-adapted group, measured against a fixed threshold rather than a randomised comparator
women with HER2-positive breast tumours measuring 2 cm or more, median tumour size 5 cm, median age 51 years, 55% postmenopausal and 65% oestrogen receptor-positive
12 versus 24 weeks of lapatinib and trastuzumab without chemotherapy, randomised 1 to 2, with letrozole and ovarian suppression if premenopausal added for oestrogen receptor-positive tumours
in-breast pathological complete response in all evaluable patients
Among the 94 evaluable patients the rate of in-breast pathological complete response was 28% in the 24-week arm and numerically superior to the 12% seen in the 12-week arm. This was driven by increased pCR in the oestrogen receptor-positive subgroup, 33% versus 9%.
centrally confirmed HER2-positive stage II-III operable disease larger than 2 cm; randomisation stratified by hormone receptor status
trastuzumab emtansine plus pertuzumab vs docetaxel, carboplatin, trastuzumab and pertuzumab
pathological complete response
pathological complete response 44.4% with the conjugate against 55.7% with chemotherapy, a difference of -11.3 percentage points (-20.5 to -2.0), p=0.016; grade 3-4 adverse events 13% against 64%
HER2-positive, hormone receptor-negative early breast cancer without metastases; 160 screened and 134 randomised 5:2
trastuzumab plus pertuzumab for 12 weeks, with or without weekly paclitaxel
pathological complete response in early responders given dual blockade alone, tested for non-inferiority against every patient given the taxane
pathological complete response 90.5% with trastuzumab, pertuzumab and paclitaxel against 36.3% with dual blockade alone. Within the dual blockade arm early responders reached 44.7% and non-responders 8.3%, so non-inferiority was not shown and the taxane was not dispensable
Women aged at least 18 years with previously untreated, centrally confirmed HER2-positive stage I-IIIA invasive breast cancer of any hormone receptor status.
Single arm of lapatinib plus trastuzumab for 18 weeks, with letrozole or tamoxifen added for hormone receptor-positive disease.
The ability of the HER2-enriched intrinsic subtype to predict pathological complete response in the breast at surgery.
Pathological complete response occurred in 41% of 101 HER2-enriched tumours and in 10% of 50 non-HER2-enriched tumours. The odds ratio was 6.2 (95% CI 2.3-16.8, p=0.0004).
stage II-III HER2-positive, median tumour 6 cm; 64 of 66 evaluable; letrozole added where the oestrogen receptor was positive
lapatinib and trastuzumab for 12 weeks without chemotherapy, with letrozole in oestrogen receptor-positive disease, single group
in-breast pathological complete response
in-breast pCR 27% overall, 21% in oestrogen receptor-positive and 36% in receptor- negative disease; low-volume residual disease in a further 33% of receptor-positive cases
5 · Response-adapted strategies and imaging-guided escalation
How much chemotherapy a HER2-positive tumour still needs, continued
Can the response to the first weeks of treatment decide who needs chemotherapy, rather than the decision being made for everyone in advance?
Rationale
A baseline assay predicts. A short exposure tests. A tumour whose metabolic activity collapses under antibody alone has demonstrated the dependence an assay was trying to infer. The same logic applies at the other end of the course, because a complete response at surgery is the strongest evidence a tumour can give that what it received was enough.
Experiment
One trial randomised a response-adapted strategy against chemotherapy with dual blockade, giving antibodies alone and adding chemotherapy only to patients whose tumours failed to respond on imaging. One tested formally whether the metabolic change measured early in the first cycle predicts pathological complete response, against a prespecified accuracy threshold. Others adapted at the far end, using magnetic resonance imaging to decide when enough chemotherapy had been given, or withholding further chemotherapy from patients whose specimen showed a complete response.
Finding
Imaging separates responders from non-responders early. The adapted groups have done well on the endpoints reported so far. Adherence to withholding chemotherapy after a complete response was high where it was measured against a feasibility target. The formal test of predictive accuracy missed its prespecified threshold. The signal underneath it was real, since tumours without a substantial metabolic fall rarely reached a complete response.
Limitation
The randomised response-adapted trial reported its recurrence endpoint against a fixed threshold rather than against its own comparator arm. That arm was small by design. So the strategy was measured against an expectation rather than against chemotherapy. The trials that adapt on the pathological response are single-arm. Their recurrence and survival endpoints have since been reported and are excellent, and that changes nothing about the inference. With no comparator, an excellent outcome after a complete response describes the prognosis of a responder rather than the effect of stopping. The trial that tested metabolic prediction required oestrogen receptor-negative disease. It says nothing about the receptor-positive majority, where complete responses are least common and the decision is hardest.
Next question
If what is being avoided is systemic cytotoxic exposure, can the cytotoxic be carried to the target instead of given free?
Response-adapted designs differ in when they look, and the timing decides what they can claim.
Two of the five rows look while the tumour is still in the patient. PHERGain uses metabolic imaging after two cycles to decide whether chemotherapy is added before surgery. Imaging responders without a pathological complete response receive chemotherapy after surgery. TRAIN-3 uses magnetic resonance imaging to decide when enough chemotherapy has been given, sending the patient to surgery once the scan is clear31. The other three look at the specimen. DAPHNe, CompassHER2 pCR and PHERGain-2 all withhold further chemotherapy from patients whose surgery showed a pathological complete response.
Where they agree is unremarkable and worth saying once. Every adapted group in the table has done well on the endpoint reported so far, and adherence to withholding chemotherapy was high wherever it was measured against a feasibility target32. That is what a cohort selected by its own good response should look like.
Where they diverge is in what the adaptation is for. Early adaptation escalates, adding chemotherapy to patients who fail to respond. Late adaptation de-escalates, removing chemotherapy from patients who already have. The second inherits the strongest evidence in the subtype, because the same logic run in the other direction is what licenses post-neoadjuvant intensification Post-neoadjuvant intensification for residual disease. The first does not.
Early adaptation is also escalation on a surrogate of a surrogate. Metabolic response early in treatment is not pathological response at surgery, and the two are separate measurements with separate error. TBCRC026 tested that relation formally, scoring the metabolic change on day 15 of the first cycle against a prespecified accuracy threshold. The area under the curve was 0.72, which did not reject a null of 0.65, so the primary objective was not met33. The signal underneath it was real, since tumours without a substantial metabolic fall rarely reached a complete response. A real signal that misses a prespecified accuracy bar is not a licence to route treatment on it. That study required oestrogen receptor-negative disease, so it says nothing about the majority in whom the decision is hardest.
PHERGain is the only randomised row and the strategy currently rests on it. State what it established precisely, because it is usually overstated. Its three-year invasive disease-free survival was measured against a fixed threshold rather than against its randomised comparator, and that comparator arm held 71 patients by design29. Grade 3 or higher treatment-related adverse events occurred in 33% of the response-adapted group against 62% of the chemotherapy group. That is a real toxicity advantage attached to a strategy nobody has measured against chemotherapy.
PHERGain-2 is the only row that measured what de-escalation is supposed to buy. A fall of 10% or more in global quality of life at one year affected 42.8% of patients on a regimen carrying no conventional cytotoxic at all28. Whatever the omission is worth to a patient, it is neither nothing nor everything.
What the table cannot show is the counterfactual it exists to produce. Four of the five rows are single-arm, so a good outcome after a complete response is the prognosis of a responder rather than the effect of stopping. CompassHER2 pCR and PHERGain-2 have not reported their recurrence endpoints. And no design here can report what would have happened to the same patients had the withheld chemotherapy been given, because nobody was randomised to receive it.
In practice
Use response-adapted de-escalation inside a trial where one is open. Outside one, say that the supporting evidence is single-arm and that the omitted treatment has no comparator.
Distinguish the two adaptation points in conversation, since imaging-guided omission during treatment carries a progression risk that pathology-guided omission after surgery does not Response-adapted treatment.
clinical stage II-IIIa HER2+ breast cancer, 781 oestrogen receptor-negative and 1394 oestrogen receptor-positive
single-arm neoadjuvant taxane with trastuzumab and pertuzumab, with further chemotherapy omitted after a pathological complete response
3-year recurrence-free survival among patients achieving a pathological complete response
pathological complete response 43.8% of the 2141 patients who started treatment, 63.7% in oestrogen receptor-negative and 32.4% in oestrogen receptor-positive disease; the primary endpoint is not yet reported
treatment-naive, HER2 immunohistochemistry 3+, node-negative disease with tumours of 5 to 30 mm on magnetic resonance imaging
single-arm neoadjuvant trastuzumab and pertuzumab, with adjuvant therapy assigned by pathological response and no randomised comparator
1-year decline in health-related quality of life and 3-year recurrence-free interval
pathological complete response in 236 of 396 patients, which is 59.6%; global quality of life fell by 10% or more in 42.8% (36.9-48.8); the recurrence-free interval endpoint is not yet reported
Patients aged 18 years or older with stage II to III HER2-positive breast cancer and WHO performance status 0 or 1.
Single arm of paclitaxel, trastuzumab, carboplatin and pertuzumab with surgery once MRI showed a complete radiological response.
Three-year event-free survival, analysed separately in hormone receptor-negative and hormone receptor-positive patients.
Three-year event-free survival was 92.2% in hormone receptor-negative patients and 92.0% in hormone receptor-positive patients, with no comparator arm.
centrally confirmed HER2-positive stage I-IIIA with at least one PET-evaluable lesion; randomised 1:4, stratified by hormone receptor status
trastuzumab and pertuzumab with or without endocrine therapy, with chemotherapy added before surgery on failure to respond on imaging and after surgery for imaging responders without a pathological complete response, vs docetaxel, carboplatin, trastuzumab and pertuzumab
pathological complete response in PET responders, and 3-year invasive disease-free survival in the response-adapted group
3-year invasive disease-free survival 94.8% (91.4-97.1) in the response-adapted group, measured against a fixed threshold rather than a randomised comparator
Treatment-naive stage II to III HER2-positive breast cancer.
Single arm of neoadjuvant paclitaxel plus trastuzumab and pertuzumab, with adjuvant therapy guided by pathologic response.
Receipt of adjuvant non-HER2-directed cytotoxic chemotherapy, as a measure of feasibility of de-escalation.
Among patients with pathologic complete response, adherence to de-escalated adjuvant trastuzumab and pertuzumab without chemotherapy was 98.2 percent (95% CI 90.3 to 100), meeting the feasibility endpoint. There was no control arm.
6 · Conjugates in the neoadjuvant setting
How much chemotherapy a HER2-positive tumour still needs, continued
Can an antibody-drug conjugate replace conventional chemotherapy before surgery?
Rationale
A conjugate delivers a cytotoxic to cells that display the target. In a tumour selected for displaying that target, it ought to reproduce the cell kill of chemotherapy while sparing everything else much of the exposure. The neoadjuvant setting reads that claim out in months rather than years.
Experiment
The substitution was made randomly before surgery, conjugate with pertuzumab against docetaxel, carboplatin and dual blockade, with pathological complete response as the endpoint. In hormone receptor-positive disease the same conjugate was randomised against trastuzumab with endocrine therapy. The substitution was also tested after surgery, conjugate with pertuzumab against a taxane with dual blockade, both arms following an anthracycline. A single-arm study of conjugate with pertuzumab defined heterogeneous and non-heterogeneous tumours prospectively, from biopsies taken at more than one site in each tumour rather than from a single score. A newer conjugate was then tested not as a replacement for chemotherapy but as the opening of the regimen, against an anthracycline opening, with the same antibody and taxane triplet completing both.
Finding
Substituting the conjugate for chemotherapy before surgery lowered response against that chemotherapy, even though severe adverse events were far less frequent than on the chemotherapy it replaced. The same substitution after surgery did not improve freedom from invasive disease against the taxane with dual blockade. The heterogeneity study accounts for part of the shortfall, because tumours carrying regions without the target essentially did not respond, while tumours uniform for it responded well. The newer conjugate, placed before the taxane triplet rather than instead of it, raised response against the anthracycline opening.
Limitation
The heterogeneity comparison was planned prospectively. Its groups were still not randomised. Only a small minority of evaluable tumours were heterogeneous, so the response rate in that group rests on very few patients. The hormone receptor-positive conjugate trial used trastuzumab with endocrine therapy as its control, a comparator with no cytotoxic content at all. Beating it says nothing about replacing chemotherapy. The adjuvant comparison is a different question in a different setting, since both of its arms also received an anthracycline. The newer conjugate was tested in high-risk disease only. Its event-free survival is barely mature. Its winning arm still contains a taxane. These are different questions about conjugates rather than repeated confirmations of one.
Next question
If a target-positive report can conceal regions without the target, should HER2 be recorded as a proportion of the tumour rather than as a score on one block, and would routing on that proportion change who is offered a conjugate?
Three rows here, and the filter admitted one of them by a side door. TRAIN-3 appears because patients with residual disease went on to trastuzumab emtansine after surgery31. Its preoperative regimen contains no conjugate at all. It says nothing about giving a conjugate before surgery and belongs to Response-adapted strategies and imaging-guided escalation.
That leaves two trials that put a conjugate in front of the operation, and they disagree because they asked the conjugate to replace different things.
KRISTINE asked trastuzumab emtansine with pertuzumab to replace docetaxel, carboplatin and dual blockade outright. Response fell, and the three-year follow-up found the loss was not only in the specimen Chemotherapy-free and de-escalated neoadjuvant regimens. DESTINY-Breast11 asked trastuzumab deruxtecan to replace the anthracycline opening of a sequence that kept the taxane and both antibodies. Response rose against that anthracycline opening35. The same trial also gave the conjugate alone, without the taxane and antibodies. That arm closed early and responded least often.
Those arms make one statement between them. This conjugate earns its place in front of the taxane and antibody backbone rather than in place of it. The payload delivers cytotoxic chemistry to antigen-positive cells and to their immediate neighbours, and why that is not the same as systemic chemotherapy is in Antibody, linker, and payload design and Bystander killing and why it matters in heterogeneous tumors.
Part of the shortfall when a conjugate replaces chemotherapy has a mechanism, and it is not in the table. One prospective study biopsied more than one site in each tumour before giving trastuzumab emtansine with pertuzumab. Tumours uniform for the target responded well, while tumours carrying regions without it essentially did not respond36. Those groups were defined in advance rather than randomised. Heterogeneous tumours were a small minority of the evaluable cases, so the rate in that group rests on very few patients.
DESTINY-Breast11 enrolled high-risk disease only, so the trade it describes has not been priced in the smaller tumours where the anthracycline is already usually omitted.
The substitution was also tested after surgery. KAITLIN gave the conjugate with pertuzumab against a taxane with dual blockade, both arms following an anthracycline, and it did not improve freedom from invasive disease. That is a different question in a different setting Adjuvant and post-neoadjuvant therapy.
Caution
Event-free survival in DESTINY-Breast11 was 4.5% mature at the reported analysis, with a hazard ratio of 0.56 and an interval from 0.26 to 1.1735. Everything in Dual blockade and what the pathological complete response gains bought applies to that pairing, in both directions. The response difference is measured. The outcome difference remains a hypothesis with a wide interval, and it is the outcome that justifies exposing a curable patient to a conjugate.
Antibody-drug conjugates given before surgery in HER2-positive disease.
The arms column carries the argument. Each trial asked the conjugate to replace something different.
trastuzumab deruxtecan alone, or trastuzumab deruxtecan followed by paclitaxel with trastuzumab and pertuzumab, vs dose-dense doxorubicin and cyclophosphamide followed by the same triplet
pathological complete response
pathological complete response 67.3% with the conjugate sequence against 56.3% with the anthracycline sequence, a difference of 11.2 percentage points (4.0-18.3), p=0.003; 43.0% with the conjugate alone
not yet reported, with event-free survival 4.5% mature
Single arm of paclitaxel, trastuzumab, carboplatin and pertuzumab with surgery once MRI showed a complete radiological response.
Three-year event-free survival, analysed separately in hormone receptor-negative and hormone receptor-positive patients.
Three-year event-free survival was 92.2% in hormone receptor-negative patients and 92.0% in hormone receptor-positive patients, with no comparator arm.
trastuzumab emtansine plus pertuzumab vs docetaxel, carboplatin, trastuzumab and pertuzumab
pathological complete response
pathological complete response 44.4% with the conjugate against 55.7% with chemotherapy, a difference of -11.3 percentage points (-20.5 to -2.0), p=0.016; grade 3-4 adverse events 13% against 64%
not reported at the primary analysis
7 · Unresolved questions and what reads out next
Both awaited trials test omission rather than addition, and each will settle less than its size suggests. Both are single-arm by design, so neither can measure what the omitted treatment was worth.
CompassHER2 pCR will report recurrence-free survival in patients who reached a pathological complete response and then had chemotherapy omitted34. A low rate will make omission defensible without measuring the loss, because no arm kept the chemotherapy. PHERGain-2 will report a recurrence-free interval after a fully chemotherapy-free preoperative phase in a selected population, where the same limitation applies to a smaller sample28.
Five further questions have no trial designed to settle them.
Whether the response advantage in DESTINY-Breast11 becomes an event-free survival advantage. That needs follow-up rather than a new trial, and the current maturity is too low to support a decision Conjugates in the neoadjuvant setting.
Whether the Fc-engineered antibody comparison is ever published in full. It exists as conference and registry material with no primary publication, so it sits on neither side of the antibody question.
Whether any assay selects patients for chemotherapy omission better than clinical features do. Genomic response predictors are being reported alongside these trials34, and none has been validated prospectively as a tool for assigning treatment.
treatment-naive, HER2 immunohistochemistry 3+, node-negative disease with tumours of 5 to 30 mm on magnetic resonance imaging
single-arm neoadjuvant trastuzumab and pertuzumab, with adjuvant therapy assigned by pathological response and no randomised comparator
1-year decline in health-related quality of life and 3-year recurrence-free interval
Interplay
This chapter treats response as a property of a tumour and a drug. Part VI argues it is also a property of a population of cells with unequal antigen. A tumour scored HER2-positive on one core may contain a compartment that is not, and that compartment is what survives to become residual disease Consequences for pathologic complete response in the neoadjuvant setting. Read that way, the response rates in these tables are partly measurements of how uniform each enrolled population happened to be. That is not speculation here, because the one trial that biopsied several sites per tumour found response and antigen uniformity tracking each other36. It also explains why a bystander-capable payload changes the arithmetic in Conjugates in the neoadjuvant setting and a receptor-blocking antibody does not.
Cortazar P, Zhang L, Untch M, et al. Pathological complete response and long-term clinical benefit in breast cancer: the CTNeoBC pooled analysis. Lancet 2014 384:164-172. PMID 24529560
Fisher B, et al. Effect of preoperative chemotherapy on the outcome of women with operable breast cancer. J Clin Oncol 1998 16:2672-85. PMID 9704717
Gianni L, Eiermann W, Semiglazov V, et al. Neoadjuvant chemotherapy with trastuzumab followed by adjuvant trastuzumab versus neoadjuvant chemotherapy alone, in patients with HER2-positive locally advanced breast cancer (the NOAH trial): a randomised controlled superiority trial with a parallel HER2-negative cohort. Lancet 2010 375:377-384. PMID 20113825
Petrelli F, Borgonovo K, Cabiddu M, Ghilardi M, Barni S. Neoadjuvant chemotherapy and concomitant trastuzumab in breast cancer: a pooled analysis of two randomized trials. Anticancer Drugs 2011 22:128-35. PMID 21218604
van der Voort A, van Ramshorst MS, van Werkhoven ED, et al. Three-year follow-up of neoadjuvant chemotherapy with or without anthracyclines in the presence of dual ERBB2 blockade in patients with ERBB2-positive breast cancer: a secondary analysis of the TRAIN-2 randomized, phase 3 trial. JAMA Oncol 2021 7:978-984. PMID 34014249
van Ramshorst MS, et al. Neoadjuvant chemotherapy with or without anthracyclines in the presence of dual HER2 blockade for HER2-positive breast cancer (TRAIN-2): a multicentre, open-label, randomised, phase 3 trial. Lancet Oncol 2018 19:1630-1640. PMID 30413379
Schneeweiss A, Chia S, Hickish T, et al. Pertuzumab plus trastuzumab in combination with standard neoadjuvant anthracycline-containing and anthracycline-free chemotherapy regimens in patients with HER2-positive early breast cancer: a randomized phase II cardiac safety study (TRYPHAENA). Ann Oncol 2013 24:2278-2284. PMID 23704196
Chen XC, et al. De-escalated neoadjuvant weekly nab-paclitaxel with trastuzumab and pertuzumab versus docetaxel, carboplatin, trastuzumab, and pertuzumab in patients with HER2-positive early breast cancer (HELEN-006): a multicentre, randomised, phase 3 trial. Lancet Oncol 2024 26:27-36. PMID 39612919
von Minckwitz G, Schneeweiss A, Loibl S, et al. Neoadjuvant carboplatin in patients with triple-negative and HER2-positive early breast cancer (GeparSixto; GBG 66) - a randomised phase 2 trial. Lancet Oncol 2014 15:747-756. PMID 24794243
Gao HF, Ye GL, Lin Y, et al. Neoadjuvant Taxane Plus Trastuzumab and Pertuzumab With or Without Carboplatin in Human Epidermal Growth Factor Receptor 2-Positive Breast Cancer: The Randomized Noninferiority Phase III neoCARHP Trial. J Clin Oncol 2026 44:1587-1596. PMID 41576297
Shao Z, et al. Efficacy, Safety, and Tolerability of Pertuzumab, Trastuzumab, and Docetaxel for Patients With Early or Locally Advanced ERBB2-Positive Breast Cancer in Asia: The PEONY Phase 3 Randomized Clinical Trial. JAMA Oncol 2020 6:e193692. PMID 31647503
Wu J, et al. Neoadjuvant pyrotinib, trastuzumab, and docetaxel for HER2-positive breast cancer (PHEDRA): a double-blind, randomized phase 3 trial. BMC Med 2022 20:498. PMID 36575513
Huober J, et al. Atezolizumab With Neoadjuvant Anti-Human Epidermal Growth Factor Receptor 2 Therapy and Chemotherapy in Human Epidermal Growth Factor Receptor 2-Positive Early Breast Cancer: Primary Results of the Randomized Phase III IMpassion050 Trial. J Clin Oncol 2022 40:2946-2956. PMID 35763704
Gianni L, Pienkowski T, Im YH, et al. 5-year analysis of neoadjuvant pertuzumab and trastuzumab in patients with locally advanced, inflammatory, or early-stage HER2-positive breast cancer (NeoSphere): a multicentre, open-label, phase 2 randomised trial. Lancet Oncol 2016 17:791-800. PMID 27179402
Piccart-Gebhart M, Holmes E, Baselga J, et al. Adjuvant lapatinib and trastuzumab for early human epidermal growth factor receptor 2-positive breast cancer: results from the randomized phase III Adjuvant Lapatinib and/or Trastuzumab Treatment Optimization trial. J Clin Oncol 2016 34:1034-1042. PMID 26598744
Carey LA, Berry DA, Cirrincione CT, et al. Molecular Heterogeneity and Response to Neoadjuvant Human Epidermal Growth Factor Receptor 2 Targeting in CALGB 40601, a Randomized Phase III Trial of Paclitaxel Plus Trastuzumab With or Without Lapatinib. J Clin Oncol 2016 34:542-549. PMID 26527775
Fernandez-Martinez A, et al. Survival, Pathologic Response, and Genomics in CALGB 40601 (Alliance), a Neoadjuvant Phase III Trial of Paclitaxel-Trastuzumab With or Without Lapatinib in HER2-Positive Breast Cancer. J Clin Oncol 2020 38:4184-4193. PMID 33095682
Robidoux A, et al. Lapatinib as a component of neoadjuvant therapy for HER2-positive operable breast cancer (NSABP protocol B-41): an open-label, randomised phase 3 trial. Lancet Oncol 2013 14:1183-92. PMID 24095300
Baselga J, Bradbury I, Eidtmann H, et al. Lapatinib with trastuzumab for HER2-positive early breast cancer (NeoALTTO): a randomised, open-label, multicentre, phase 3 trial. Lancet 2012 379:633-640. PMID 22257673
Gianni L, Pienkowski T, Im YH, et al. Efficacy and safety of neoadjuvant pertuzumab and trastuzumab in women with locally advanced, inflammatory, or early HER2-positive breast cancer (NeoSphere): a randomised multicentre, open-label, phase 2 trial. Lancet Oncol 2012 13:25-32. PMID 22153890
Rimawi MF, Niravath P, Wang T, et al. TBCRC023: A Randomized Phase II Neoadjuvant Trial of Lapatinib Plus Trastuzumab Without Chemotherapy for 12 versus 24 Weeks in Patients with HER2-Positive Breast Cancer. Clin Cancer Res 2019 26:821-827. PMID 31662331
Rimawi MF, Mayer IA, Forero A, et al. Multicenter phase II study of neoadjuvant lapatinib and trastuzumab with hormonal therapy and without chemotherapy in patients with human epidermal growth factor receptor 2-overexpressing breast cancer: TBCRC 006. J Clin Oncol 2013 31:1726-1731. PMID 23569315
Llombart-Cussac A, Cortes J, Pare L, et al. HER2-enriched subtype as a predictor of pathological complete response following trastuzumab and lapatinib without chemotherapy in early-stage HER2-positive breast cancer (PAMELA). Lancet Oncol 2017 18:545-554. PMID 28238593
Nitz UA, et al. De-escalation strategies in HER2-positive early breast cancer (EBC): final analysis of the WSG-ADAPT HER2+/HR- phase II trial: efficacy, safety, and predictive markers for 12 weeks of neoadjuvant dual blockade with trastuzumab and pertuzumab ± weekly paclitaxel. Ann Oncol 2017 28:2768-2772. PMID 28945833
Nitz U, et al. De-escalated neoadjuvant pertuzumab plus trastuzumab therapy with or without weekly paclitaxel in HER2-positive, hormone receptor-negative, early breast cancer (WSG-ADAPT-HER2+/HR-): survival outcomes from a multicentre, open-label, randomised, phase 2 trial. Lancet Oncol 2022 23:625-635. PMID 35405088
Hurvitz SA, Martin M, Jung KH, et al. Neoadjuvant trastuzumab emtansine and pertuzumab in human epidermal growth factor receptor 2-positive breast cancer: three-year outcomes from the phase III KRISTINE study. J Clin Oncol 2019 37:2206-2216. PMID 31157583
Gluz O, Nitz UA, Christgen M, et al. Survival analysis of the WSG TP-II trial: neoadjuvant trastuzumab and pertuzumab plus endocrine therapy versus chemotherapy in HR+/HER2+ early breast cancer. J Clin Oncol 2026. PMID 42150145
Garrigós L, Ruiz-Borrego M, Pérez-García JM, et al. A chemotherapy-free, pathological response-adapted strategy using trastuzumab-pertuzumab and T-DM1 in HER2-positive early breast cancer: the PHERGain-2 study. Ann Oncol 2026 37:837-848. PMID 42097893
Pérez-García JM, Cortés J, Ruiz-Borrego M, et al. 3-year invasive disease-free survival with chemotherapy de-escalation using an 18F-FDG-PET-based, pathological complete response-adapted strategy in HER2-positive early breast cancer (PHERGain): a randomised, open-label, phase 2 trial. Lancet 2024 403:1649-1659. PMID 38582092
Hurvitz SA, Martin M, Symmans WF, et al. Neoadjuvant trastuzumab, pertuzumab, and chemotherapy versus trastuzumab emtansine plus pertuzumab in patients with HER2-positive breast cancer (KRISTINE): a randomised, open-label, multicentre, phase 3 trial. Lancet Oncol 2018 19:115-126. PMID 29175149
Louis FM, et al. MRI-based personalisation of neoadjuvant chemotherapy duration in HER2-positive early breast cancer (TRAIN-3): primary results from a multicentre, single-arm, phase 2 study. Lancet Oncol 2026 27:349-360. PMID 41713473
Waks AG, et al. A prospective trial of treatment de-escalation following neoadjuvant paclitaxel/trastuzumab/pertuzumab in HER2-positive breast cancer. NPJ Breast Cancer 2022 8:63. PMID 35538105
Connolly RM, Leal JP, Solnes L, et al. Updated Results of TBCRC026: Phase II Trial Correlating Standardized Uptake Value With Pathological Complete Response to Pertuzumab and Trastuzumab in Breast Cancer. J Clin Oncol 2021 39:2247-2256. PMID 33999652
Tung N, Zhao F, DeMichele A, et al. Pathologic complete response (pCR) rate and predictors of response to taxane, trastuzumab, and pertuzumab in HER2-positive breast cancer: secondary analyses of EA1181/CompassHER2 pCR. J Clin Oncol 2026. PMID 42623567
Harbeck N, Modi S, Pusztai L, et al. Neoadjuvant trastuzumab deruxtecan alone or followed by paclitaxel, trastuzumab, and pertuzumab for high-risk HER2-positive early breast cancer (DESTINY-Breast11): a randomised, open-label, multicentre, phase III trial. Ann Oncol 2026 37:166-179. PMID 41130363
Metzger Filho O, Viale G, Stein S, et al. Impact of HER2 heterogeneity on treatment response of early-stage HER2-positive breast cancer: phase II neoadjuvant clinical trial of T-DM1 combined with pertuzumab. Cancer Discov 2021 11:2474-2487. PMID 33941592