Central nervous system metastatic disease

When a limited number of brain metastases has already been treated focally, does adding whole-brain radiotherapy earn what it costs?

Rationale

Deposits exist beyond what MRI resolves. Irradiating the whole brain should therefore postpone the appearance of new lesions and the neurological death that follows them. The hippocampus and the rest of the normal brain take the same dose as the deposits, so whatever is bought in intracranial control is paid for in cognition. That trade has to be measured rather than assumed.

Experiment

Randomised comparisons that held the focal treatment constant and varied the whole-brain component. Whole-brain radiotherapy against observation after complete resection or radiosurgery. Radiosurgery alone against radiosurgery plus whole-brain treatment, with a cognitive primary endpoint. Cavity radiosurgery against whole-brain treatment after resection. Cavity radiosurgery against observation after resection. Whole-brain treatment with or without a radiosurgical boost, with survival as the primary endpoint. Alongside these, a prospective cohort asked whether the number of lesions treated with radiosurgery alone changes survival.

Finding

Whole-brain radiotherapy lowers relapse at the treated site and at new sites. It lengthens neither survival nor the time to loss of performance status. It worsens cognition against focal treatment alone, whether it follows resection or radiosurgery. Adding a radiosurgical boost to whole-brain treatment did not lengthen survival in the trial population as a whole. Survival after radiosurgery alone did not fall away as the number of metastases treated rose, across the range the cohort covered. Leaving the resection cavity untreated is a different question. It costs local control. The case is therefore against irradiating the whole brain rather than against irradiating anything.

Limitation

Every one of these studies enrolled brain metastases from solid tumours generally and was powered on that mixed population, so a breast-specific estimate is a subgroup at best. Cognitive deterioration is a scoring convention rather than a clinical event. One trial called it a decline beyond a set margin on any single test in a battery. Another used a reliable change index. Both scored only the patients still well enough to be tested, which is an increasingly selected group as the disease progresses. The survival advantage for a radiosurgical boost lives in a single-lesion subgroup on univariate analysis of a trial whose survival endpoint was not met in the population as a whole. The lesion-count comparison used to justify radiosurgery for several metastases comes from a prospective observational cohort compared against itself rather than from a randomisation.

Next question

If whole-brain radiotherapy remains the right answer for diffuse disease, can the part of its cognitive cost that comes from the dose be engineered away?

NCCTG N0574 EORTC 22952-26001 NCCTG N107C / CEC.3 Mahajan postoperative SRS trial RTOG 9508 JLGK0901

If whole-brain radiotherapy has to be given, can its cognitive cost be engineered down?

Rationale

The memory decline that follows whole-brain treatment is attributed to the dose delivered to the hippocampal stem cell niche, and to a vascular and glutamatergic injury that a receptor antagonist might blunt. Neither mechanism requires the dose to the metastases to change, so both are modifiable without giving up intracranial control.

Experiment

One randomised trial gave memantine or placebo through and after whole-brain radiotherapy, with delayed recall at a fixed timepoint as the primary measure. A later trial held memantine constant in both arms and varied the plan instead, conformally avoiding the hippocampi against conventional whole-brain treatment, with time to cognitive failure as the primary measure.

Finding

Avoiding the hippocampi lowers the risk of cognitive failure against conventional whole-brain treatment given on the same memantine backbone. Survival and intracranial progression were unchanged. The memantine trial favoured the drug on its own primary measure of delayed recall without crossing its threshold for significance. A secondary measure, the time to cognitive decline, did favour it.

Limitation

The memantine result is an attrition failure rather than a demonstration of no effect. Only a minority of those randomised were analysable at the primary timepoint, and the report states its own power as far below what the design called for. That matters for how the later trial reads. Its control arm already contained memantine on evidence that never met its own threshold, so the avoidance question was asked on a backbone the field had adopted rather than proved. Both trials scored cognition as a change on a test battery against a statistical rule, which is a measurement convention rather than a clinical event. Neither enrolled breast cancer specifically, and the hippocampal-avoidance trial further required that no metastasis sit in the region being spared.

Next question

Local treatment is being refined at its margins. Does anything given systemically alongside radiotherapy improve what happens inside the irradiated field?

RTOG 0614 NRG CC001

Does adding a systemic drug to brain radiotherapy improve intracranial control?

Rationale

Radiation and a drug act on the same lesion at the same time. A drug that reaches the brain at all should reach it best while the barrier is disrupted by the tumour and by the radiation itself. Radiosensitisation would also let the delivered dose do more without raising it.

Experiment

Randomised trials that held brain radiotherapy constant and added a drug. Concurrent lapatinib in HER2-positive disease. Concurrent temozolomide. Induction bevacizumab with etoposide and cisplatin given before whole-brain treatment rather than with it. Alongside those, a single-arm study paired radiosurgery with a checkpoint antibody to ask whether the combination was safe rather than whether it worked, and a protocol pairing a brain-penetrant HER2 inhibitor with radiosurgery sits on the record with no outcomes.

Finding

No randomised addition improved intracranial response. Concurrent lapatinib produced no complete responses at the radiological timepoint and no gain in objective response. Temozolomide did not raise response above whole-brain treatment alone. The induction arm separated from radiotherapy alone on brain-specific progression only against a significance level the trial had set unusually permissive. Its early brain response was no better than the control arm's.

Limitation

The induction trial declared its result against a prespecified alpha far looser than convention. Its interval for brain progression or death included no difference. It is therefore a signal-seeking result rather than a positive trial. The response endpoints in these trials were read on MRI at a fixed early timepoint, where radiation change and residual tumour are hard to separate, which is a measurement problem rather than a biological one. The radiosurgery and checkpoint study was small, single-arm and sized for safety, and the only report on its record is a long-term follow-up rather than a primary analysis, so its local control has no comparator. The tucatinib and radiosurgery study on the record is a protocol with no outcomes, and it asserts nothing about efficacy.

Next question

If adding a drug to radiation changes nothing inside the field, is the problem that the drugs used never reached the lesion?

NRG Oncology–KROG / RTOG 1119 WBRT with or without temozolomide A-PLUS SRS plus nivolumab TUTOR

Does giving a HER2-directed drug concurrently with brain radiotherapy improve intracranial control over radiotherapy alone?

Rationale

The compartment is being irradiated in any case. A HER2 inhibitor that reaches it could act as a radiosensitiser. That would buy local control without adding a modality and without waiting for the drug to work on its own.

Experiment

Whole brain radiotherapy or radiosurgery was randomised with or without concurrent lapatinib in HER2-positive brain metastases. Entry required at least one measurable, previously unirradiated lesion. Intracranial complete response at an early fixed timepoint was the primary endpoint. A dose-finding study is now testing the same idea with a more brain-penetrant inhibitor alongside radiosurgery.

Finding

Concurrent lapatinib did not improve intracranial complete response over radiotherapy alone. Objective response at the same timepoint was not improved either. The complete responses that did occur were in the arm given radiotherapy alone. Higher-grade toxicity was commoner with lapatinib added than with radiotherapy alone.

Limitation

A complete response rate at an early fixed timepoint is a radiographic surrogate chosen because it reads out quickly. It is not a measure of how long the compartment stays controlled. It is also the endpoint most exposed to post-radiotherapy imaging change, where treatment effect and residual tumour look alike. Lapatinib is the weakest brain-penetrant member of its class. The same class given without radiotherapy did change intracranial outcomes, which makes this evidence against one drug on one schedule rather than against concurrency. Entry required an unirradiated measurable lesion. The trial therefore says nothing about adding a drug to re-irradiation. The study that would test the concept with a better-penetrating drug is a dose-finding protocol with no outcomes on the record. A maximum tolerated dose cannot settle efficacy whatever it eventually reports.

Next question

If the concept failed with a drug that may never have reached tumour in useful concentration, is intracranial failure a delivery problem at all?

NRG Oncology–KROG / RTOG 1119 TUTOR HER2CLIMB

Can checkpoint blockade work inside the central nervous system?

Rationale

The brain was long described as immunologically privileged, which predicted that checkpoint blockade would fail there. Melanoma and lung cancer then produced durable intracranial responses. If the privilege argument is wrong, the compartment is open to an agent whose effect does not depend on the drug reaching every cell.

Experiment

Single-arm studies of pembrolizumab in brain metastases of mixed primary histologies, and in leptomeningeal dissemination from solid tumours where most patients had breast cancer. Separately, a checkpoint inhibitor combined with an antiangiogenic antibody and platinum in triple-negative disease with active brain metastases.

Finding

Both pembrolizumab studies met their primary endpoints, and the combination study reported a high intracranial response rate. None of the three had a comparator.

Limitation

Mixed-histology cohorts dilute what can be said about breast cancer. The leptomeningeal study is small enough that its confidence interval spans the range between useless and useful. The combination study cannot separate the contribution of the checkpoint inhibitor from the platinum and the antiangiogenic antibody, and the last of those alters the measurement.

Next question

Is the intracranial effect of checkpoint blockade a property of the compartment, or of the subtypes in which checkpoint blockade works anywhere?

Pembrolizumab in BM of diverse histologies Pembrolizumab in leptomeningeal carcinomatosis ABC study

Is exclusion by the blood-brain barrier the reason systemic drugs fail in the brain?

Rationale

The standard explanation for intracranial progression during effective systemic treatment is exposure, in that the drug holds disease everywhere it reaches and the brain is the place it does not. If exposure is the whole story, a molecule engineered to cross, or a drug chosen for its penetration, should convert into intracranial response.

Experiment

Three different ways of testing that claim. A window study gave an antibody-drug conjugate before craniotomy and measured the payload recovered from the resected metastasis, making delivery itself the endpoint. A peptide conjugate built to cross by receptor-mediated transcytosis was given to patients with recurrent brain metastases. Two further studies took drugs already chosen for brain penetration, a CDK4/6 inhibitor in hormone receptor-positive disease and an mTOR inhibitor added to HER2 blockade with chemotherapy, and asked for intracranial response.

Finding

Delivery and activity came apart. The window study met its concentration endpoint in the resected metastases. The CDK4/6 study measured unbound drug in resected brain metastases above the concentrations its two target kinases require, and returned almost no confirmed intracranial response in either of its cohorts. The transcytosis conjugate, the CDK4/6 inhibitor and the mTOR combination each missed the intracranial response rule they had set themselves in advance.

Limitation

A drug concentration in resected tissue is a pharmacological surrogate with no clinical outcome attached, and the tissue came only from patients selected for being operable. That holds for the window study and for the resection cohort of the CDK4/6 study alike. The transcytosis conjugate was read at two different response rates by investigators and by independent review, and those should be read separately rather than merged, with its preset rule assessed on the independent read. The CDK4/6 study ran its hormone receptor-positive HER2-negative and HER2-positive cohorts under separate two-stage designs, so it offers two small single-arm results rather than one. The mTOR study's response rests on the smaller group evaluable for efficacy rather than on everyone enrolled, and its correlative work found no pathway alteration that marked a responder, so no biomarker-defined subgroup survives in which the rationale still holds.

Next question

If exposure is not the whole explanation in the parenchyma, does the argument hold where exposure is lowest, and can a drug given systemically change the course of leptomeningeal disease?

Sacituzumab govitecan phase 0 window study ANG1005 JPBO abemaciclib CNS study LCCC 1025

Does engineering a drug for brain penetration turn intracranial failure into intracranial response?

Rationale

The standing explanation for central nervous system relapse is that drug does not reach the compartment. That explanation makes a prediction. Raising exposure, building a carrier that crosses, or choosing a molecule that crosses by design should convert patients who do not respond into patients who do.

Experiment

Four treatment studies came at that prediction from different directions. One measurement study tested its premise directly. Trastuzumab was given at high dose with pertuzumab in brain metastases progressing after radiotherapy, on the argument that exposure was the limiting variable. The same high-dose pair was given again with a checkpoint inhibitor added. Paclitaxel was conjugated to a peptide that uses a receptor-mediated transport route across the barrier. Everolimus, which crosses, was added to trastuzumab and vinorelbine in progressive brain metastases, under a staged design built to separate a useless intracranial response rate from a useful one. Separately, a window study measured a conjugate and its payload in brain metastasis tissue resected the day after dosing.

Finding

Drug is measurable in resected brain metastasis tissue, at the concentration that study set out to demonstrate. None of the four treatment studies delivered the intracranial activity its rationale predicted. The peptide conjugate missed its preset response rule on independent review. The everolimus study returned a single intracranial partial response, against a staged design built to find many more. The checkpoint combination fell below its prespecified efficacy threshold and was closed early. The high-dose antibody pair given without a checkpoint inhibitor confirmed central nervous system responses in only a minority of patients, while reporting a median overall survival among the longest in this cell. Reaching the compartment looks necessary. On this evidence it is not sufficient.

Limitation

The window study measured concentration and not benefit. Its lesions were operable and therefore large and accessible. It enrolled a glioblastoma cohort alongside the breast cancer one. Its drug is aimed at a different target from everything else under this question. No design here connects tissue concentration to response. The peptide conjugate reported an investigator-read and an independently-read intracranial response rate that differ from one another. They have to be read separately rather than averaged. That study also took all comers rather than restricting entry to HER2-positive disease. The everolimus study reported response on the patients evaluable for efficacy rather than on everyone enrolled. Its correlative claim about pathway mutations rests on almost no responders. The checkpoint combination stopped at a preset threshold, which is a judgement about that dose and that schedule rather than about the hypothesis. The high-dose antibody pair given alone had no preset response rule to fail. Its response result and its survival result point in different directions, from a single arm with nothing to compare against.

Next question

If penetration is not the limiting step in the parenchyma, is it the limiting step in the leptomeninges, where the barrier, the disease and the measurement are all different?

PATRICIA Atezolizumab plus pertuzumab and high-dose trastuzumab ANG1005 LCCC 1025 Sacituzumab govitecan phase 0 window study

Does adding a brain-penetrant HER2 tyrosine kinase inhibitor to systemic treatment control brain metastases that are still active?

Rationale

The compartment was treated as radiation's problem, on the argument that a full-length antibody does not cross an intact blood-brain barrier. That argument was never tested, because the trials excluded patients whose brain disease was untreated or progressing. An earlier single-arm study had already paired a tyrosine kinase inhibitor with capecitabine in brain metastases that had never been irradiated. It reported volumetric intracranial responses, which made a randomised version of the question worth asking.

Experiment

Tucatinib was randomised against placebo on a trastuzumab and capecitabine backbone in pretreated disease. Entry was opened to patients whose brain metastases were untreated or progressing after local therapy. Progression-free survival in that group was a prespecified secondary endpoint rather than a subgroup found afterwards. The intracranial analyses were planned alongside it. The same drug was then randomised on a conjugate backbone in second-line disease, and again as first-line maintenance in a population restricted to no brain metastases or asymptomatic ones.

Finding

Adding tucatinib lengthened progression-free survival in the trial as a whole and in the patients who entered with brain metastases. It lengthened overall survival in the trial as a whole. The prespecified central nervous system analyses run in the same direction as the trial rather than against it. On the conjugate backbone and in first-line maintenance, adding tucatinib lengthened progression-free survival as well.

Limitation

This is a group within a systemic trial rather than a trial whose entry required brain disease. The randomisation protects the drug comparison and not the choice of a central nervous system strategy. The control arm carried capecitabine, which has intracranial activity of its own. The contrast is therefore against an active backbone rather than against local therapy or against nothing. Entry required prior trastuzumab, pertuzumab and trastuzumab emtansine, which leaves the same question unanswered earlier in the disease. Neither of the two later tucatinib trials required brain metastases for entry. The maintenance trial admitted only asymptomatic ones. Neither carries a central nervous system population that stands on its own. Interim overall survival on the conjugate backbone was immature and did not favour the tucatinib arm.

Next question

If a small molecule reaches the compartment, does a conjugate whose antibody should not cross reach it as well?

LANDSCAPE HER2CLIMB HER2CLIMB-02 HER2CLIMB-05

Can an antibody-drug conjugate produce intracranial responses when its antibody is too large to cross an intact barrier?

Rationale

The size argument applies to an intact barrier. A metastasis large enough to measure on imaging has vasculature that is already disrupted. The question is therefore not whether the antibody crosses healthy brain but whether it reaches tumour. A conjugate also has only to deliver its payload. A membrane-permeable payload can act on cells the antibody never bound.

Experiment

Single-arm studies of conjugates in patients with brain metastases, entered by the state of the brain disease rather than by line of treatment. Trastuzumab deruxtecan was given in small cohorts defined as stable after local therapy, untreated and asymptomatic, or progressing after local therapy. It was then given in a large brain metastasis cohort studied alongside a cohort without brain disease. Trastuzumab emtansine was reported from the brain metastasis subgroup of a single-arm safety study, and was also given with neratinib in parallel cohorts split by prior local therapy and by prior exposure to the conjugate. A first-line module in active brain metastases is on the record as a design and nothing else.

Finding

Intracranial responses with the deruxtecan conjugate are routine rather than anecdotal. They occur in active disease as well as in disease already treated and stable. In the largest cohort, the central nervous system progression-free rate at the fixed timepoint was close to the overall progression-free rate in that same cohort. The emtansine conjugate produced intracranial responses too, at a lower rate than the deruxtecan cohorts report, in populations that are not otherwise comparable.

Limitation

None of these studies has a control arm. Each is measured against what its own cohort was expected to do. The central nervous system endpoints are not the same object: a progression-free rate at a fixed timepoint, an intracranial response by neuro-oncology criteria and an intracranial response by solid tumour criteria cannot be read off against one another. The fixed timepoints are not the same timepoint either. The largest of these studies performed no formal hypothesis testing at all, which makes its rates descriptive. The emtansine result is a post hoc analysis of the brain metastasis subgroup of a safety study rather than a prespecified endpoint. The conjugate module for active brain metastases in first or second line has only a design on the record. Nothing here speaks to a conjugate given before any HER2-directed therapy.

Next question

If response rates cannot be compared across these cohorts, what is actually driving the difference between them?

TUXEDO-1 DEBBRAH DESTINY-Breast12 KAMILLA TBCRC 022 DESTINY-Breast07 module 7

What systemic treatment is there when the disease in the brain is not HER2-positive?

Rationale

The intracranial record is built almost entirely on HER2-directed agents, because that is where the drugs were. A patient with hormone receptor-positive or triple-negative disease and active brain metastases has the same problem and almost none of the same evidence. Agents that cross the barrier for other reasons, or that are delivered differently, were tested on that argument.

Experiment

Single-arm studies of a cytotoxic agent combined with an antiangiogenic antibody in active brain metastases, and of a conjugate in HER2-negative brain disease. Separately, one randomised phase III comparison of a polymer-conjugated topoisomerase inhibitor against chemotherapy of the physician choice, in patients with treated and stable brain metastases.

Finding

The single-arm combinations produced intracranial responses in a substantial minority. The one randomised comparison in this group found no survival difference against ordinary chemotherapy.

Limitation

The studies that report activity have no control arm, and the study that had one enrolled stable treated brain metastases rather than active disease. Those are different populations and the second cannot refute the first. Antiangiogenic treatment also reduces contrast enhancement, which is what a response is measured on, so response in these regimens is harder to interpret than elsewhere.

Next question

Does an intracranial response in a single-arm study of a combination tell a clinic anything, when neither component has been tested alone in the same population?

U-BOMB SWOG S2007 ATTAIN

Does an intracranial response rate mean the same thing in a radiotherapy-naive cohort as in one that has progressed after radiotherapy?

Rationale

A response rate is read as a property of the drug. It is at least as much a property of who was admitted and of how a lesion was measured. A previously irradiated lesion has been selected for surviving radiation. It is also harder to measure, because treatment effect, necrosis and tumour resemble each other on imaging.

Experiment

Studies that split their population by central nervous system state on purpose and reported the strata separately. One regimen was given to a radiotherapy-naive cohort and to a post-radiotherapy cohort within the same trial. A conjugate study split its cohorts by whether brain disease was stable after local therapy, untreated or progressing. A kinase inhibitor given with a conjugate enrolled parallel cohorts of previously untreated brain metastases and of brain metastases progressing after local therapy. One study required progression after radiotherapy for entry. The earliest of them required that no brain radiotherapy had been given at all.

Finding

Within the pyrotinib regimen, intracranial response is much higher in the radiotherapy-naive cohort than in the cohort that progressed after radiotherapy. The deruxtecan cohorts point the same way, with the untreated cohort above the one progressing after local therapy. The neratinib and emtansine cohorts do not: their previously untreated cohort did not respond more often than the cohorts entered after local therapy. Across trials, the ranking of intracranial response rates tracks how permissive the entry criteria were at least as closely as it tracks which drug was given.

Limitation

Prior radiotherapy is not a clean variable. It marks lesions selected for surviving radiation. It also marks lesions that are harder to measure. No design here separates the two. None of these strata was randomised. Each contrast is therefore between cohorts that differ in more than their radiotherapy history. No trial randomised the order of radiotherapy and systemic treatment. What deferring radiotherapy costs the patient, if anything, is unmeasured. Reading these strata as a ranking of drugs is the error this grouping exists to prevent.

Next question

If the cohorts that look best are the ones that deferred radiotherapy, is there anything to be gained by giving the drug and the radiation together?

PERMEATE LANDSCAPE DEBBRAH PATRICIA TBCRC 022

Can systemic therapy alone change the course of leptomeningeal disease?

Rationale

Leptomeningeal involvement has been treated as a terminal event of the compartment, and the trials that established the modern systemic agents excluded it by protocol. If a drug reaches the meningeal deposits at all, survival counted in weeks might become survival counted in months without any intrathecal or radiation treatment at all.

Experiment

Two single-arm studies enrolled leptomeningeal disease specifically and set overall survival as the primary endpoint, each against a historical median named in advance. One gave tucatinib with trastuzumab and capecitabine in HER2-positive disease. The other gave trastuzumab deruxtecan in a cohort that admitted HER2-positive and HER2-low disease. A leptomeningeal group within the brain-penetrant conjugate study was reported alongside its parenchymal population.

Finding

Both cohorts ran past the historical median each had set as its comparator, and the deruxtecan cohort states that its prespecified endpoint was met. The leptomeningeal group within the conjugate study was reported as surviving longer than its own historical expectation, while the parenchymal primary endpoint of that same study was not met.

Limitation

A historical median is the weakest comparator available here, and the medians on offer were assembled before the systemic drugs these patients had already received existed, so the comparison contains an era effect as well as a treatment effect. Both cohorts are very small, and entry required fitness for a trial. Diagnosis rested on cytology or on MRI depending on the protocol, which is a case definition rather than a biological boundary. The conjugate's leptomeningeal group is a subset of an enrolled population rather than a separately powered cohort, so its survival claim is a subgroup measured against history.

Next question

If a drug given systemically can hold the leptomeninges, is there anything left for treatment aimed at the compartment itself?

TBCRC 049 DEBBRAH cohort 5 ANG1005

Does treating the leptomeningeal compartment directly add to treating the patient systemically?

Rationale

Cerebrospinal fluid is a space that can be dosed on its own, either by placing a drug in it or by irradiating the whole neuraxis. Both approaches take the exposure problem at face value. Both accept a toxicity confined to the compartment in exchange for an exposure that systemic dosing cannot reach.

Experiment

One randomised trial added intrathecal liposomal cytarabine to whatever systemic therapy the patient was already receiving, and left that systemic choice to the treating clinician. Three single-arm studies gave intrathecal trastuzumab in HER2-positive leptomeningeal disease, two of them dose-finding and one with a neurological progression endpoint. A randomised study compared proton craniospinal irradiation against photon radiotherapy to the involved field. A phase III successor to it sits on the record with nothing reported.

Finding

Adding intrathecal cytarabine lengthened leptomeningeal progression-free survival against systemic therapy alone. Intrathecal trastuzumab proved deliverable by both escalation routes. In the study with a neurological endpoint, most patients were free of neurological progression at the early timepoint. In the study that escalated to a twice-weekly schedule, response was confined to a minority. Craniospinal proton irradiation lengthened central nervous system progression-free survival against involved-field photon treatment.

Limitation

The randomised intrathecal trial measured progression in the compartment it was treating and allowed systemic therapy to vary from patient to patient, so the endpoint favours the intervention by construction and the background treatment is uncontrolled. The two intrathecal trastuzumab programmes arrived at different recommended doses and different schedules, which is a property of their escalation rules and their definitions of dose-limiting toxicity rather than of the drug. One of them admitted HER2-positive tumours of any origin rather than breast cancer alone. The proton trial randomised breast and lung cancer together, stopped early at a planned interim analysis, and compared treatment of the whole neuraxis against a field that by definition did not cover disease outside it, so part of the separation is the field size rather than the particle. Its confirmatory trial has no result on the record and asserts nothing yet.

Next question

Every leptomeningeal result here rests on a small cohort, a historical median or a comparator field that was never going to win. What would a trial have to randomise to say whether compartment-directed and systemic treatment are alternatives or additions?

DEPOSEIN Intrathecal trastuzumab, Bonneau phase I Intrathecal trastuzumab, Oberkampf phase II Intrathecal trastuzumab, Kumthekar phase I/II Proton CSI versus photon IFRT RADIATE-LM / NRG-BN014

In HER2-positive leptomeningeal disease, does the drug have to be delivered into the compartment?

Rationale

The leptomeninges are not the parenchyma. There is no bulky lesion with disrupted vasculature to exploit. Disease is read from cerebrospinal fluid cytology and from neurological examination rather than from target lesions. Survival is short enough that the choice of endpoint constrains the choice of design before any drug is chosen.

Experiment

Two routes, tested separately and never against each other. Systemically, tucatinib with trastuzumab and capecitabine was given as a single arm with overall survival against a prespecified historical control as its endpoint. Trastuzumab deruxtecan was given the same way in a leptomeningeal cohort. By the compartment-directed route, dose-finding and single-arm studies of intrathecal trastuzumab established a tolerated dose and reported cerebrospinal fluid concentrations, cytological and clinical responses and neurological progression-free survival. The only randomised test of the principle added intrathecal liposomal cytarabine to whatever systemic therapy the patient was already receiving, in breast cancer of any subtype.

Finding

Each systemic single arm reported survival longer than the historical control written into its own protocol. Intrathecal trastuzumab is tolerable at the doses taken forward. It reaches cerebrospinal fluid concentrations in the range sought. It produces cytological responses and neurological improvement. Those intrathecal studies carry no comparator, historical or randomised. The one randomised comparison lengthened leptomeningeal progression-free survival against systemic therapy alone. Overall survival in that trial did not differ significantly between the arms.

Limitation

A prespecified historical control is a fixed number taken from an era before brain-penetrant HER2 therapy. A cohort enrolled now beats it partly by being enrolled now. Both HER2-positive leptomeningeal single arms are very small, one of them extremely so. The intrathecal studies do not agree on a dose or on a schedule. One of them admitted HER2-positive tumours of any origin in its dose-finding portion rather than breast cancer alone. Their endpoints are different objects: a maximum tolerated dose, a cerebrospinal fluid response and a neurological progression-free survival cannot be pooled into a response rate. The randomised evidence for the intrathecal route comes from a drug that is not HER2-directed, in a population not selected by subtype. It supports the principle of compartment-directed treatment rather than the use of intrathecal trastuzumab. The peptide conjugate's leptomeningeal group is a subset of a parenchymal study, compared with historical survival and never selected for HER2. It sits beside these single arms rather than inside them.

Next question

If systemic HER2 therapy now reaches the leptomeninges well enough to move survival, which endpoint could show what the intrathecal route adds on top of it?

TBCRC 049 DEBBRAH cohort 5 Intrathecal trastuzumab, Bonneau phase I Intrathecal trastuzumab, Kumthekar phase I/II Intrathecal trastuzumab, Oberkampf phase II DEPOSEIN ANG1005 Radiotherapy followed by intrathecal trastuzumab plus pertuzumab