Please be dying, but not too quickly, part 1: a clinical trial story

A three-part, very deep dive into the insanity that is the "modern" clinical trial system. Buckle up.

You’re reading part 1 right now; Part 2: The patient’s perspective is at the link, and Part 3: The end of the journey is also available. If you need help finding a clinical trial or you’d like to get in touch to talk more about how we can make this process better please reach out to DrBStillman@gmail.com. I’d love to talk.

My husband Jake is dying of recurrent / metastatic squamous cell carcinoma (R / M HNSCC), and he’s exhausted conventional treatments, so his only chance at survival is a miracle clinical trial drug. But what clinical trial might save him? How do we find it? Those are the question we’ve been trying to answer since learning on July 21 that he has eight new tumors in his neck and lungs. Finding the right clinical trial is literally a life, or more likely death, issue for Jake. Despite the stakes, from the patient’s perspective, the clinical trial process is impressively broken, obtuse and confusing, and one that I gather no one likes: patients don’t, their families don’t, hospitals and oncologists who run the clinical trials don’t, drug companies must not, and the people who die while waiting to get into a trial probably don’t. I guess the dead don’t have preferences, but I’m going to assume they’d rather be living. Per Judy Seward, head of clinical trial experience at Pfizer, only 8% of adult cancer patients take part in trials. Given our recent experience, I’m amazed it’s that many.

The median patient with R / M HNSCC lives about twelve months after diagnosis. If you count Jake’s diagnosis as happening in April 2023, you can do the math on his probable survival time. Jake’s cancer, however, is uncommonly fast and aggressive, and so, in the absence of a clinical trial drug that works, he might die at any time. The path from “a clinical trial drug is our only hope” to having a drug injected into his arm has been Herculean. Although I’m a doctor, I’ve been stymied by the clinical trial process. I love Jake, probably more than it’s advisable to love anyone, but love isn’t a treatment for R / M HNSCC (If correlation were causation, my love might actually be clinically deleterious).

I’m writing this partly as a guide for other people in our position: people who want to take their best shot at prolonging their own lives, or the lives of their loved ones. But this guide is for a wider audience, too: policymakers or others who might have the power to revise the process and make it more sane than it is now, doctors (especially oncologists), and anyone who wants to understand why we seem perpetually on the brink of medical breakthroughs that are painfully slow to actually be implemented. The FDA’s invisible graveyard is filled with people who could be benefitting from trials and their results. We should all care that the primary process for bringing new and necessary treatments for serious diseases from the lab to the exam room is broken. I wish I could make this guide shorter, but the process is complex and doing wrong increases the likelihood of fatality, so I’ve erred on the side of thoroughness.

Starting the search: Canvassing Physicians

The process of finding a clinical trial started July 21, the evening we learned about Jake’s cancer recurrence. We sent a MyChart message to his sluggish Mayo oncologist, and he called us.

“It’s not what we’d hoped,” he said, stating the stunningly obvious—in his defense, though, “fatal cancer recurrence” is one of these situations where words are inadequate. “But you’re scheduled for chemo on Monday, and so we have a plan.”

We knew, from reading the R / M HNSCC literature, that chemo is only palliative and would at best delay death by a few months.

That was a plan?

He went on: “If you want, we can discuss clinical trials at a later date. I’ll have my PA see what’s available and get back to you.”

What a strange thing to say to someone who rapidly grew eight new tumors over the course of a month: that there will be time to discuss later. When was later? “Later” was what we didn’t have. For weeks, that was the extent of the conversation.

This wasn’t our first experience with the oncologist’s lackadaisical approach to cancer care, so I’d already been in touch via desperate Facebook messages to online physician communities, and I’d begun sending emails to other oncologists who had been recommended to me in those conversations. A few of them had previously discussed post-surgical treatment options with us. Although the average patient doesn’t have this ability, one benefit of being a physician is casual access to other physicians. Most professions, including mine, have a kind of favor bank, and I know that one day I’m going to get a message from an oncologist who needs something from an ER doc, or the ER doc network.

Even with this vast network, making progress was and is incredibly difficult. If we can hardly manage it, I don’t know how a regular patient does. My sense is that, largely, lacking outside guidance, they don’t. Since the day of Jake’s initial recurrence, I’ve been frantically begging for the second opinions that would have taken weeks to otherwise schedule, or plane flights to our-of-state research centers. For most people, I think they run out of time and energy to pursue possible clinical trials and thus don’t wind up in trials, or wind up in sub-optimal ones.

The online responses from physicians in doctor groups were helpful, but often vague; I’d hoped that a head and neck oncologist would recommend a specific trial, preferably one that they could refer us to. But most docs didn’t know of a specific trial, so they recommended large research institutions they thought had a good reputation for hard cases. Others took Jake’s clinical information (which I’d posted with his permission: this was an emergency) and started a clinical trial search on clinicaltrial.gov, and they posted the top five findings. There were maybe ten such responses, and, ominously, none listed the same trials.

How is it that ten doctors can put in the same basic, relevant clinical data into an engine meant to list and search for all existing clinical trials, only for no two to surface the same study?

If you can answer this question, then you’ve interacted with ClinicalTrials.gov. If you haven’t, it’s a government-based website, so perhaps it shouldn’t be so surprising that it’s clunky, inefficient, and frustrating.

Continuing the search: Clinicaltrials.gov

As of the writing of this essay, if you go to Clinicaltrials.gov and search for Jake’s condition, you’ll get something like this:

2,500 studies! That isn’t helpful. We can narrow the results somewhat:

Around 751 studies is still too hard. Shoot. And if you change the wording of the search subtly, you’ll get different outcomes: searches for “Squamous cell carcinoma of the head and neck,” “Head and Neck Squamous Cell Carcinoma,” “Squamous Cell Carcinoma of the tongue,” “Tongue Squamous Cell Cancer” all yield different results. There’s no apparent standardization.

Some example “condition or disease” searches and their results (with no other additional search terms):

• Head and neck cancer: 6,895 studies

• Cancer of the Head and Neck: 4,283 studies

• HNSCC: 1,532 studies

• Squamous Cell Carcinoma of the Tongue: 106 studies

• Tongue Cancers: 152 studies

• Oral Cavity SCC: 661 studies

• SCC of the Oral Cavity: 31 studies

There’s no apparent (the word “apparent” will appear a lot) correlation between studies that would be best for Jake and the order in which they’re listed. Adding more search terms will narrow the number of studies, but, because there’s no standardized way for researchers to choose multiple keywords for their studies, I’m narrowing the field to studies where the person filling in the study information online randomly decided to list multiple keywords.

Then, when I did select a handful of studies, how was I supposed to evaluate if nivolumab, RAPA-101, BCA 101 or MCLA-158 was ideal? Oh, and the name of the drug and the acronym assigned to it in the trial were frequently different, confounding further Google Scholar searches. For example, “MCLA-158” is the same thing as “Petosemtamab.” Of course! Why wouldn’t it be?

I have a background in biochemistry, so I understood the difference between a bispecific antibody, an immunotherapy, and a targeted molecule—but not what those differences meant for Jake’s cancer, or all the relevant up-to-date literature on the matter.

Each study has different eligibility criteria, and eligibility criteria—like keywords—aren’t standardized so that they can be easily searched across studies. Instead, the information page for each study outlines unique inclusion and exclusion criteria. Many have overlapping inclusion and exclusion criteria, but there can be long lists of additional criteria for each arm of a study, and it’s up to the patient or their doctor to read through them line by line to see if prior medications, current medications, certain genomic sequencing findings, numbers of lines of therapy, etc. makes the trial relevant.

It’s easier to find a dress to my exact specifications out of thousands on H&M,com than it is to find a clinical trial. I can open a search bar, click “dress,” select my material from another click box (Which allows me to select from the  same options the people listing the garments chose from), then click on the boxes for my desired color, dry clean or machine wash, fabric, finish, closure, and any other number of pre-selected categories before adding additional search keywords. I find a handful of options all relevant to my desires within a matter of minutes. And a dress isn’t life or death.

I quickly tested my theory: A search for “dress” showed 2,156 options.

Filters (finally, checkboxes!) were chosen: party dress, gold, sequin, long sleeve, a-line cut. Here are my 2 matches fitting my exact (if somewhat weird) specifications:

Someone get their web developer a government contract…

The most important decision of our lives seemed impossible to adequately evaluate based on the primary method we had to research options. I went back online hoping that the physicians who had given us lists of potential studies for Jake knew more about how to optimize the search than I did.

Each study on clinicaltrials.gov has a unique “NCT” number, which is sometimes called the ClinicalTrials.gov Identifier. I opened an Excel spreadsheet, made a list of 50 NCT codes offered by other doctors, and, 20 hours later, had read through each trial, compared Jake’s medical data to the inclusion and exclusion criteria and searched google scholar and PubMed for any information about the trial drug, so that I could make some basic judgments about whether or not the trial was a good fit for Jake. Or even a potential fit.

Fifty down, 700 to go. Continuing at this rate, I would complete a comprehensive review of options within 108.3 days, blowing past Jake’s estimated life expectancy. In a not-very-research-appropriate fit of pique, I cried at the keyboard. I imagined myself at Jake’s funeral, still sifting through NCTs, the sound of his oncologist’s voice in my head asking if we wanted to discuss options, or if we’d already decided “what we wanted to do.”

What I wanted to do was to keep having a meltdown, but there wasn’t time. I called an oncologist we’d seen earlier for a second opinion, hoping a real-time conversation would get us somewhere. “I’d go to M.D. Anderson in Houston, because they do a lot of trials,” he said. Which trials? He didn’t know, but he repeated that M.D. Anderson does a lot of trials, so they would have options he didn’t at Banner Hospital (a local, Arizona outpost of MD Anderson, which operates independently from the Houston mothership). Were M.D. Anderson’s trials the right options? He told us that the oncologists at M.D. Anderson Houston wouldn’t review the trial options with us until we had an official in-person visit, which could take weeks to schedule, and he couldn’t tell us what was available, because he didn’t know and didn’t have access to Houston’s study databases.

I called another doctor we’d seen, named Katherine Price, who is at Mayo Rochester. She reviewed Jake’s case and was (and is) incredibly helpful—one of the heroes of this story. She said: “I know UC—San Diego Medical Center and Memorial-Sloan Kettering [MSK] in New York both do a lot of trials. I’ll reach out and see what they have.” She promised to call back with more information. She said she’d check with her contacts at M.D. Anderson and see if, among their many trials, anything looked promising. Maybe, I thought, specialists there could refer us to the right trial if M.D. Anderson wasn’t hosting it?

Online, more doctors chimed in with the names of hospitals known for HNSCC, suggesting we see what trials they hosted. “Hospital first, trial second” seemed to be the generally recommended way.

While I appreciated the help and the referrals, how were we supposed to know if the right trial for Jake was hosted at any of these sites? How were we supposed to know where to go if we didn’t know what we were going for?

I opened Excel back up and reviewed my list of 50 studies. I prepared to look at the next 50 and felt despair.

Then I met Eileen.

Find the right trial, then get into the trial (simple, right?)

Eileen Faucher is a biochemist with a business sideline helping patients like Jake navigate the clinical trial process. Bingo. If one of my former colleagues hadn’t referred me to her, I’m not sure we could have found the right trial. I first emailed Eileen on July 23. Her views on the trial process were congruent with mine: The key is to find the right trial, and then get into that trial at whatever site might be available. Trial first, hospital second.

All trials aren’t equal, and a hierarchy of likely efficaciousness could be guessed at, if not promised—they’re called “trials,” not guarantees. Eileen charged a lot per hour but her fees were worth every penny: she was efficient, and without her I’d have been lost. She was clear about her own limitations and the system’s limitations: she could help identify promising trials, but a promising trial may not work out. Think of the “promising” persons you may have dated who, when you got to know them, weren’t so promising.

ClinicalTrials.gov didn’t make Eileen break out in a cold sweat or give her gastrointestinal distress. Having spent years consulting for the pharmaceutical companies, and with a background in cancer research, Eileen was familiar with medication acronyms and where the money was going. She could evaluate a world that was, to her, familiar.

In ten hours, Eileen produced a spreadsheet of 500 studies, organized into viability, and preference based on drug targets, and study type. The first 150 were loosely organized from most to least promising based on their specificity to HNSCC, prior published data, study phase, and supporting research. I don’t know how a normal person can navigate the clinical trial landscape without connections to the medical system along with someone like Eileen, but I may have found my next career. Watch this space for further announcements: People need this service, and it seems that drug companies, oncologists, and hospitals can’t or won’t make the process easier for patients. So maybe I should.

Important categories used to rank studies include:

•  Specificity to cancer type: Did the trial focus on all solid tumors, or had the drug already shown promise or a mechanism of action specific to HNSCC?     

• Phase: There are three relevant clinical phases, and, while there are a bunch of nuances to actual real-world practice and FDA requirements, I’m going to hit the highlights for a patient looking to get into the right trial:

Phase 1: A phase 1a study brings a study drug out of the lab and into humans for the first time. They are “single ascending dose” studies. Usually, one to three patients at a time are given a dose, then there’s a waiting period, then the dose is increased. This process is repeated until researchers determine the maximum dose a typical person can tolerate. For the patient, a phase one study is a little like throwing spaghetti at the wall and hoping it sticks: you might benefit from the medication, but most trials don’t advance to the next phase.[1]

Following a successful phase 1a study of dosing is, typically, a phase 1b “expansion” study. In 1b, the dose (or doses) that will be used for future phase II studies are further narrowed down, and additional tolerability and safety studies are performed.

Early efficacy “endpoints,” might also be evaluated, such as “what was the average time to response?” “how much did the tumors shrink?” and “what were the patient’s overall survival rates?”  Patients are also often further grouped into “cohorts” based on their tumor type, which helps to determine if a study drug behaves differently between cancer types.

Phase 2: A phase 2 study has already established tolerability and dose. It’s looking for responses to the treatment. It’s asking the question: “does this drug work?” Phase 2 treatments are more appealing because their endpoint is efficacy, but phase 1b expansion arms are similar, much more common, and, arguably, close to the same thing as phase 2 trials. If a phase 1b study drug appears more promising than a phase 2 drug, I’d prefer to be a patient in the 1b study. Drug companies, however, might not publish enough data for a prospective patient to know whether Joe Pharma’s 1b study is better than Jane Drug Company’s phase 2 study, or vice-versa.

Phase 3: None were available. Phase 3 studies are the large-scale randomized trials usually pit the standard of care against a new treatment regimen with published efficacy data that may replace the previous standard of care. A phase 3 trial is typically your best bet, if one is available, but phase 3 studies often aren’t available.

• Preliminary Data: Did early phase 1 trials show promising outcomes?

• Research & Development (R&D):

• Based on research history, was this drug similar to another drug type/category that had positive results? Is the drug a new iteration of a repeated failure?

• Is the drug company pouring money into the drug, which suggests unpublished positive data or insider knowledge? “Follow the money” is usually good advice across fields.

I started reviewing each study on Eileen’s list. As I mentioned, Eileen understood the oncology drug space, how to compare and evaluate different molecular targets and immunotherapies, study design and pharmaceutical cash flow. But the second half of determining whether a study is “right” involves understanding the patient’s clinical picture and determining if they’re medically eligible.  

Evaluating for clinical relevance is where my expertise came into play. Did prior surgeries, pathology results, treatments, or tests disqualify Jake? Were the genetic tumor markers being targeted—Jake’s tumor was tested by both CARIS and Mayo in-house sequencing—relevant for his disease? Did he meet other medical requirements? Together, Eileen and I narrowed the first 150 studies down to a top 50, then reviewed the data again and made a list of what we hoped were the 35 most promising trials.

Unfortunately, even an expert search like Eileen’s can be scuppered by the poor digital management of ClinicalTrials.com. Jake had read about a promising Moderna study for mRNA-2752, which involves “an injection of the study drug directly into the tumor.” Eileen reviewed the study and thought it looked particularly promising, but mRNA-2752 wasn’t on her list because whoever posted it to ClinicalTrials.gov listed it under “all solid tumors,” instead of selecting each of the multiple tumor types that the study was willing to include. Because that person never chose “HNSCC (or a variant)” as a keyword, it remained invisible to Eileen. Jake had set up a bunch of Google news alerts related to mRNA cancer vaccines, immunotherapies, and other technologies, so he was coming up with stuff that we weren’t seeing on clinicaltrials.gov. Jake also found Moderna’s mRNA-4359 trial, which is an immune checkpoint inhibitor in solid tumors.

If just this part of the process sounds frustrating, that’s because it is. But don’t worry, later parts will be equally—if not more frustrating—in unique ways. And I will repeat that Jake’s life is on the line. Your life, or the life of someone you love, may be on the line too. A number of oncologists suggested that we find a local clinical trial in Arizona because sifting through trials and then moving would be too hard. Wouldn’t it be nicer to sleep in your own bed? That is an opinion but it isn’t my opinion, or Jake’s. We are fortunate enough to be in a position to move to whichever city might host the most promising trial, and we believed (and still do) that the potential benefits between trials was vast I need Jake to have the best possible shot at living.

Eileen found five top trials:

• NCT04815720: Pepinemab in Combination With Pembrolizumab in Recurrent or Metastatic Squamous Cell Carcinoma of the Head and Neck (KEYNOTE-B84)

• NCT03526835: A Study of Bispecific Antibody MCLA-158 in Patients With Advanced Solid Tumors.

• NCT05743270: Study of RP3 in Combination With Nivolumab and Other Therapy in Patients With Locoregionally Advanced or Recurrent SCCHN.”

• NCT03485209: Efficacy and Safety Study of Tisotumab Vedotin for Patients With Solid Tumors (innovaTV 207)

• NCT05094336: AMG 193, Methylthioadenosine (MTA) Cooperative Protein Arginine Methyltransferase 5 (PRMT5) Inhibitor, Alone and in Combination With Docetaxel in Advanced Methylthioadenosine Phosphorylase (MTAP)-Null Solid Tumors (MTAP).

Plus, the Moderna studies Jake had found. Based on the names alone, you can see why it would be difficult if not impossible for someone without some expertise in cancer oncology to evaluate trials. And yet it seems that many oncologists have only a vague sense of what’s happening regarding clinical trials in their field! I’m surprised there aren’t more people like Eileen, offering to sell guidance to the clinical trial process. When I visited Spain, it felt like there was a guide on every corner willing to offer historical or wine tours, or, better still, historical wine tours (it’s hard to get bad wine in Spain, which I learned from my extensive research on the subject). The clinical trial decision is much more important and yet it seems most patients are on our own.

AI Companies: Not there yet

This section may rapidly become out of date, but a billion “AI” companies tout products that will match patients to trials. Jake’s essays on his impending death and the FDA’s complicity in letting him die led to a bunch of AI companies contacting us. Plus, we were referred to MassiveBio and XCures, which boasted both improved clinical trial matching through “AI,” as well as clinical navigators (Massive Bio) and a PhD with clinical research experience reviewing your data (XCures). Their algorithms interact with the same frustrating ClinicalTrials.gov platform the plebs do, but I wondered if the AI companies might be able to do a more thorough job. Their systems probably weren’t programmed to get frustrated and need to take frequent breaks to scream into a pillow.

Massive Bio states they’ll take your comprehensive medical history—generated from the electronic medical records of the various hospitals and clinics visited—and provide a “care summary” (I’d written one for Jake). But I never received that summary, and as of checking their website five minutes ago, there isn’t one uploaded. We did, however, receive a “SYNERGY-AI” clinical trial matching results form that listed five trials across Arizona, California and New Mexico. These results weren’t great.

Only five trials were listed: one was for radiotherapy, which Jake wouldn’t qualify for, since he’d already had radiation and it’s unwise to re-irradiate the same place within such a short time. Another was only for patients who had not yet received first-line therapy for their recurrence (Jake was on immunotherapy and chemo), and the other was a phase 1a dose-escalation study that, especially when compared to the top 35 studies that Eileen and I had found, seemed only tangentially relevant to Jake. Our #12 study was in the list. Our highest ranked study in Arizona, which was actively recruiting, wasn’t listed at all.

As a side note, Massive Bio is also performing a meta-study called “SYNGERGY-AI” to evaluate if their clinical-trial matching algorithm leads to improved outcomes. I like this idea a lot, and think it’s a valuable question to eventually ask, but I can save them a lot of time and money with the answer based on their current system: No. But call me, guys, if you want to discuss where things might have gone wrong.

Edit: Massive Bio’s response in the comments. It seems fair to foreground what they say.

The XCures website says:

The XCures platform transforms complex unstructured medical data directly from the patient’s medical records into structured data suitable for analysis and review. Our A.I. engine can actively cross-reference this data against the vast digital library of oncology data to match patients with potential treatments. This empowers patients and their oncologists to make more informed and effective treatment decisions.

From XCures, we were referred to cancercommons.com, where we were matched with Emma Shtivelman, PhD, who is a “clinical scientist” and reviewed Jake’s records. She expanded the search past AZ and CA, but noted she had “doubts about a number of trials in terms of eligibility.” She wanted to include them in case they were relevant. She identified that bispecific antibodies might be a good target, which Eileen and the literature also support, but, although some supporting data was attached to the five recommendations that followed, most of the studies were first-in-human phase 1a dose escalation trials. None of the five studies listed were on our top thirty list, though there were multiple bispecific antibody studies on our top thirty list. Some recommendations regarding potential off-label drug targets were rescinded in a later e-mail.

Edit: XCures responded on LinkedIn. It again seems fair to foreground what they say.

These AI methods are slick and seem well-supported, but they may be suffering from “garbage in, garbage out.” The companies have or appear to have VC money, and they’re funneling data into larger research studies about the use of AI in the trial space. I’m sure they mean well and want to help patients. The ideas sound good. We’re not the first people to observe the craziness of clinicaltrials.gov. Based on what we’ve seen so far, however, the “AI” tech isn’t there yet and the people reviewing the results also don’t seem to have the expertise or the time to identify that the searches aren’t very useful or comprehensive;  a clinical liaison is also useless without some guidance on eligibility, which wasn’t really evaluated for the trials Jake was “matched” to.  

Patients shouldn’t be making life-or-death decisions based on these results, which missed a large number of trials for which Jake was eligible. Look, I understand that scaling companies don’t have the time to do what Eileen and I did for Jake. She and I have medical and research backgrounds, and she spent 15 hours while I spent weeks on the project for one patient. One day, I hope these efforts will be scalable, but as the field currently stands, the time constraints needed to be thorough are prohibitive for both scalable businesses and for a patient’s personal oncologist.

Without feeling comfortable learning to search clinicaltrials.gov themselves, or without having access to help like Eileen, I can see how patients would choose to focus on the AI-discovered trials and go no further, simply because they didn’t have the resources to figure out what’s really going on. In this way, the AI searches—especially since they not only miss the optimal trials, but also the best-matched trials in a patient’s local area—seem much less useful than making appointments at the various hospital systems in your area and meeting with their local oncologist to ask what they happen to have available in terms of trials.

That being said, I’d like to see a scaleable AI solution, but as long as the data the AI trains on (what’s currently available on ClinicalTrials.gov) is incomplete and poorly structured, I don’t see it happening. You can’t out-program a bad initial data set. There will need to be a multi-prong approach before the AI folks can get it right and I hope they do!

If you’ve gotten this far, consider the Go Fund Me that’s funding Jake’s ongoing care.

You just finished part 1. Part 2: The patient’s perspective is at the link, and Part 3: The end of the journey is also available.

If you’re in need of clinical trial assistance, you can reach out at DrBStillman@gmail.com

---

[1] Interestingly, in the past few years, there have been a number of positive responses to phase 1 drugs in the immunotherapy and targeted molecule space. Keytruda, a game changer for melanoma, and, unfortunately, less of one for head and neck cancers, was approved for use directly from phase 1, a previously unheard of achievement in the clinical trial space. Still, first-in-human trials, and early phase ones—unless they have a particularly compelling and unique mechanism of action—are very hard to compare because they lack preliminary data and are therefore unlikely to be worth traveling for.