Three years ago, I learned of KD-019, an EGFR-inhibiting wonder TKI for which there just happened be a study enrolling, right at the mid-tier academic institution/portal to hell in which I was currently sitting. I nodded along appreciatively to phrases like better than Tykerb! while thinking, Kadmon. Kadmon. Oh, right, the one where the CEO just got out of prison.
My oncologist at the time sold it really hard, but I was unconvinced that calling it KD-019 instead of XL647 would do anything about that pesky 3% response rate in lung, so I didn’t enroll. Either no one else signed up, or they all did really badly, because the company only managed to eke out one abstract before the ClinicalTrials.gov status switched to “Active, Not Recruiting.” The abstract suggested that the drug was “well-tolerated” in, ahem, seven patients, six of whom actually took the drug. Insert derisive snort, but in general, nothing to see here. (And Kadmon was pretty intent on making sure you didn’t see it; any mention of the breast program was wiped from the website around the same time as the status changed, and the only thing that remains are the press releases.) The study finally switched to “Terminated” on August 28, 2017.
XL647 KD-019 tesevatinib isn’t dead yet. While Kadmon does seem to be shifting away from being an oncology company, the MDs on its payroll are still recruiting tesevatinib trials. The targets are brain patients now, specifically NSCLC cases with CNS mets and patients with recurrent GBM. Both trials are studying tesevatinib as a single agent, with the rationale for this prolonged charade hinging entirely on the fact that the drug passes the blood-brain barrier. Well, fine, but does it have any anti-cancer activity once it gets there?
When the failed phase II NSCLC study started enrolling in 2006, patients weren’t selected for EGFR mutations, and CNS mets weren’t exclusionary, but they had to be stable. So the new study is clearly a totally different thing, like completely different, despite about a quarter of NSCLC patients having EGFR-mutated disease, meaning they would have been represented in an all-comers cohort. Further, there was nothing remarkable noted for the 15% of patients who enrolled with CNS mets. There was a prior published attempt at optimizing patient selection, and it didn’t help much. This XL647 trial enrolled treatment-naive NSCLC patients on the basis of having an EGFR mutation or “being Asian, female, or having minimal or no smoking history” and found a 20% response rate overall and a 57% RR among patients with EGFR-sensitizing mutations. For context, fellow EGFR inhibitors gefitinib and erlotinib have shown phase III first-line response rates of 70-80% when the cohort was EGFR-enriched. Afatinib was approved on a RR of 56% (centrally assessed) and 69% (investigator assessed) in a study that reported a 90% disease control rate overall; this is a therapy that’s considered third-line. Osimertinib, the newest of the bunch, targets a rarer EGFR mutation (the more common ones affect exon 19 and 21) and demonstrated a 71% response rate in its phase III, and AZ recently teased a first-line PFS benefit for the drug compared to erlotinib and gefitinib.
Kadmon can’t catch up. So they’re elbowing their way into the crowded NSCLC market with a pivot to brain mets, untroubled by the mounting peer-reviewed publications in favor of better drugs, and clinging to rat evidence that tesevetinib penetrates the BBB more effectively than its predecessors (guys, come on, collect some CSF and at least try to replicate that finding in humans). But let’s take a closer look at how that latest EGFR-selected cohort is faring.
First, a review of the study design. There are three arms of 20 patients (total enrollment=60), each with their own primary endpoints:
- Arm A: Brain mets (BM) patients. Primary endpoint is RECIST 1.1 response in both brain and peripheral lesions.
- Arm B: Leptomeningeal disease (LM). Primary endpoint is symptom improvement. They’re using CTCAE 4, which is a generic encyclopedia of side effects and descriptions of their grades. It’s for AE reporting in general, as assessed by the investigator. Interesting. For background, LM is not a disease type that can be assessed quantitatively; lepto disease affects the membranes around the brain and spinal cord and is inherently non-measurable, so there’s no way to accurately evaluate it on imaging (you can see enhancement on MRI, but not discrete lesions you can track from timepoint to timepoint). Despite this, RECIST is equipped to deal with it; they’re just not using it here. N.B., patients with both brain and lepto mets are enrolled in the LM cohort.
- Arm C: added sometime in 2016; treatment-naive patients presenting with brain mets. Primary endpoint is RECIST 1.1 response in both brain and peripheral lesions.
And as a quick RECIST 1.1 recap, responses are categorized as one of the following: progressive disease (PD), which is worsening of measurable target lesions by 20% or more or the appearance of an unequivocal new lesion; stable disease (SD), which is neither sufficient reduction in the size of target disease to qualify as a partial response (PR), nor is it a sufficient increase in tumor burden to qualify as PD; PR, which is improvement of disease measurements by 30% or greater; and the unicorn, a complete response (CR), which is complete disappearance of disease. Onward.
Interim data were summarized in a poster at World Lung last year, which Kadman described in a barely coherent press release. They led with the bold claim that the vast majority of enrolled subjects experienced stable CNS disease for … an unspecified amount of time:
“Eleven of the first 13 enrolled patients treated with tesevatinib 300 mg QD in this ongoing study did not have central nervous system (CNS) progression on tesevatinib.”
We then glossed over the patients who got no meaningful benefit from the drug: “… five of the 12 pretreated patients had peripheral disease progression, while in four of those five patients, tesevatinib controlled CNS lesions.”
Useful. What good are imaging-stable CNS mets when your lungs and liver are being swallowed up by cancer? But let’s not dwell on that when there was – wait for it – a partial response. And not just any old PR. It was robust. RECIST responses are determined by mathematical calculation to preclude that kind of nonsense, but fine, robust:
“… one enrolled patient with no prior treatment showed a robust partial response in brain metastases in an MRI taken on Study Day 29 and showed a partial response in both brain metastases and peripheral disease at Study Day 57. This patient continues on tesevatinib as of Study Day 92.”
This study is intended to verify, by the way, Kadmon’s claim that “Currently approved EGFR inhibitors have poor brain penetration, limiting their ability to treat intracranial metastases.”
Let’s reconcile this mess of a press release with the poster. Here is a subject-level summary, including what one can guess from the poster to be best response. That includes the LM cohort, which is officially reporting a symptom-based endpoint; we’ll simplify things by using a similar analysis from an afatinib compassionate use program CNS cohort (n=31) as a model. The afatinib study used regular old RECIST 1.1 to evaluate CNS disease, where, unless patients had a CR/PR, best response of SD was defined as patients who received treatment for more than four months and PD as patients who received treatment for less than two months. This table will be a real shitshow if we require four months of treatment to qualify for SD, so we’re going to say three months on treatment is sufficient for SD. (It’s not going to make a difference; this is a demonstration, not a statistical analysis.) We’ll use the afatinib study’s two-month cutoff for PD, which is standard per the response confirmation rules of RECIST. The reasons for discontinuation are taken verbatim from the poster. Again, this stands alone from whatever Kadmon plans to present to the FDA; we’re just adapting what little data we have to afatinib’s example to gauge performance.
|Cohort (n=13)||Subject ID||Status on Trial||Reason for Discontinuation||Best Response|
|BM||045-002||Discontinued||CNS / Peripheral PD||PD|
|BM||045-003||Discontinued||CNS response / died due to aspiration||PD|
Bolded subjects benefited from the drug with either either stable disease or a partial response (3/13, or 23%). The additional italicized subject may be benefiting, pending sufficient follow-up to achieve SD (4/13, or 31%). For 31% to hold as our hypothetical clinical benefit rate (CBR), 036-001 would have to have hang onto their stable (or better) status for another 45 days; regardless of the fate of individual subjects, a CBR of 30% is disappointing; 50% is just about the lowest CBR you see for a targeted therapy.
Those 11 (of 13) subjects who had no CNS progression? This observation is purely academic, since most of these subjects are off-study anyway. Many had disease that can’t be followed on imaging, which could bias the sequence of calling PD toward peripheral versus CNS disease. One subject is dead, albeit from causes other than the drug that aren’t clearly attributable to progressive disease. When we look at patients who were stable (or better) for more than a few weeks and exclude the patients who quickly experienced progressive disease or “symptomatic deterioration”, which is way scarier than a bad scan, the scope of benefit narrows. No CNS progression for a couple of months doesn’t mean things turned out okay, and calling PD based on the disease that can be measured while the non-measurable CNS disease is “controlled” is not a credit to tesevatinib. It’s mostly expedient.
While it’s hard to top the audacity of raising a bunch of money for a new biotech venture the day after you’re released from prison, Kadmon does try. Their one treatment-naive responder is the only thing justifying continued recruitment for this study. This is clear from an abstract at the same World Lung, which is so sloppy that the PI’s name is spelled incorrectly and which mentions a hasty addition of a treatment-naive cohort in the face of a handful of patients who experienced CNS “symptom improvement” (a result that does not sing of FDA approval). SOC is to move from one EGFR inhibitor to the next on progression, ideally re-biopsying to check for new mutations (T790M is associated with acquired resistance to EGFR inhibitors; this is the mutation osimertinib targets). But Kadmon’s not playing that game. We’re not stopping at a 20-patient last-ditch cohort to buoy the results of a flawed study. One first-line responder, and suddenly we’re to believe that prior EGFR exposure is the lone Achilles heel for this drug. The company plans, per the World Lung press release, to “initiate a randomized, first-line study of tesevatinib monotherapy in patients with EGFR-mutant NSCLC who present with CNS metastases.” No word on what the control would be. It could take a while to find a suitably inferior agent.
Let’s pause and revisit the afatinib CNS study. In that analysis, there were no patients receiving afatinib first-line. They’d all received chemo and Iressa or Tarceva. Some of them received second-line EGFR inhibitors. There were patients who were fourth-line. The investigators measured an average time-to-failure of 3.6 months, which increased to 4.0 months in subjects with a confirmed EGFR mutation. In CNS disease, which is the market where tesevatinib claims to have an edge, there was a PR rate of 42% (13/31). When they excluded patients who had a only a systemic (“peripheral” in the Kadmon language) response, the overall response rate was 35%. The clinical benefit rate was 66%. This is apples-to-apples what tesevatinib is up against. And they want to go first-line? Their results are sufficient to leapfrog all these other agents that demonstrated efficacy with normal endpoints the first time out? Kadmon’s claim that the approved EGFR inhibitors are “limited” in treating CNS disease is just not true. This is a good SOC, with drugs that work.
Despite the attempt to menace patients (and investors) with yet another do-over, Kadmon does not appear to have advanced their randomized NSCLC death trap since the World Lung proclamation. Patients, if a first-line study ever does make it past Satan’s IRB and into your institution, know that you’ll get the wonder TKI pitch a lot, again and again. You get one chance at first-line therapy, at the dream of a complete response. CR aside, chemo and the marketed EGFRs have elicited durable responses, more than once. Many brain lesions are amenable to stereotactic radiotherapy. ALK mutations may not be mutually exclusive with EGFR mutations after all, so that’s worth a look, and alectinib is actually cooler than a PD-1/PD-L1, anyway. In other words, a lot needs to go wrong before tesevatinib should be considered viable first-line monotherapy.
As for the GBM, which is almost too heinous to discuss: the study has been open and enrolling since June 2016. With realistic expectations about outcomes in recurrent GBM, that’s enough time to get some interim data (and a similar timeline from the startup of the breast study to when we saw that “well-tolerated” press release). There doesn’t appear to be anything online, and we’ve seen the company’s willingness to cheer tesevatinib’s toxicity profile and vague cases of CNS non-progressors. Despite the prevalence of EGFR mutations in glioma (as many as 50% of cases), the use of EGFR inhibitors in GBM patients is associated with at best inconsistent and, realistically, minimal clinical benefit. This may be because EGFR inhibitors – including tesevatinib, if you revisit the molecular characteristics of the few patients who did respond – tend to work in patients with an exon 19 deletion or exon 21 point mutation, which are common in lung. According to this 2009 paper, those mutations have never been found in GBM. So how is this study ethical? How does it reflect a legitimate scientific question? The answer is not that we need more options for GBM. That catch-all defense for bullshit therapy is disrespectful to patients who run out the clock on doomed trials; of course we need more options for GBM. But we’re not going to identify those options wasting patients on tesevatinib.
After 10+ years of development, of stalled and failed and … ongoing trials, where are we at? Has there been any progress toward an FDA approval? I listed all the sponsor-led (including from the Exelixis days) tesevatinib studies I could find, and this is how it looks:
|Safety and Pharmacokinetics of XL647 Administered Orally to Subjects With Solid Tumors||Solid tumors||Completed||Yay!|
|Study of XL647 Administered Orally Daily to Patients With Solid Tumors||Solid tumors||Completed||Good so far!|
|Safety Study of XL647 and XL147 Administered in Combination Daily in Adults With Solid Tumors||Solid tumors||Withdrawn||Nevermind it’s fine it’s probably going to work great|
|Study of XL647 in Subjects With NSCLC Who Have Progressed After Responding to Treatment With Gefitinib or Erlotinib||NSLC, 1+ lines tx||Completed||3% RR; fail, do not pass go, sell drug to felon|
|KD019 Versus Erlotinib in Subjects With Stage IIIB/IV Non Small Cell Lung Cancer With Progression After First- or Second-Line Chemotherapy||NSCLC, 1+ lines tx||Terminated||LOL let’s just not|
|KD019 and Trastuzumab in Patients With Esophagus, Gastroesophageal Junction and Stomach Cancer||GI||Withdrawn||Nope|
|Study of the Combination of KD019 and Trastuzumab in Subjects With HER2-Positive Metastatic Breast Cancer||HER2+ MBC||Terminated||“Well-tolerated”, but clearly a fail|
|Study of Tesevatinib in Subjects With Non-Small Cell Lung Cancer, EGFR Activating Mutation, Prior Treatment With a Tyrosine Kinase Inhibitor, and Brain Metastases or Leptomeningeal Metastases||NSCLC w/ brain mets, prior tx w/ TKI||Recruiting||At interim: 1 responder, assumed CBR of 23%
Relying on new cohort of tx-naive pts to boost RR
|Study of Tesevatinib Monotherapy in Patients With Recurrent Glioblastoma||Recurrent GBM||Recruiting||Positive result will require miracle and/or creative data handling|
|TBD: First-line NSCLC w/ CNS mets||Previously untreated NSCLC w/ CNS mets||Idle threat||Sigh|
A couple hundred patients have been treated with this drug since it started with Exelixis; a handful have responded. What keeps this drug alive? Kadmon doesn’t want to be an oncology company; from their slides at Rodman last week, the onc program barely merited a mention. They had to publicly swap out their convicted felon of a CEO so they could list on the NYSE, so we know they’re immune to embarrassment. Tesevatinib does fit their we-care unmet-needs narrative, the flip side of which is the complacency it fosters, a tolerance for bad results; well, we tried. They’re going to die anyway. We gave them a chance. How many failed and failing trials can that excuse endure, when the drug has never proven clinically useful and has in fact demonstrated nothing but the nimbleness of its developers? None of this is sleight of hand; it’s happening in the open, with ClinicalTrials.gov and Pubmed providing a clear and accessible record. But investigators and sites get paid, they don’t have to tell patients they’ve exhausted their treatment options (which, depending on the patient’s status, can be either misery for both parties or a financial risk if the patient scuttles off to a more accommodating center) and patients get the comfort of having “more options”. Cancer centers are incentivized to be infinitely tolerant of questionable trials, and an open trial doesn’t convey support for a drug; it reflects nothing but the endorsement of the investors who fund it, which, if Kadmon’s opaque press releases are anything to go by, are themselves in the dark. With more on the line, potential subjects shouldn’t be so willing to look the other way.