Dear Bess. Thank you for writing this. I was trying, perhaps too obliquely, to agree with you about the need for an (inter)national trial database. I think that oncologists by and large do try and do the best for their patients. We can all reasonably be expected to be up to date on the latest standard of care. However, there’s a wider spectrum in how up to date we are with the latest research efforts.
Your point about it being reasonable to expect your oncologist to be au fait at least with the most relevant and exciting phase 3 trials in their area is very reasonable - and I think most specialist oncologists are, especially at large academic centres. Most of us do spend at least a few hours a month learning about the most recent papers in NEJM or the most recent presentations at ASCO or ESMO.
There is also the capacity to phone a friend - and even general oncologists working in rural or regional centres far away from, say, the National Cancer Institute (or state level equivalent) will often call or email a colleague at the relevant centre to ask about trials when faced with a case like your husband’s.
It’s just that the world is too big for all this to be done by humans! Of course there needs to be a database. Imagine if humans were still the only travel agents and sites like Flight Centre didn’t exist; you’d expect a person to know all possible flight paths between (say) New York and Dubai and which was the quickest vs the cheapest? Humans used to be able to do a reasonable job, before databases and algorithms, and a talented and plugged in human could probably still do OK - but clearly this is an area where a good search engine will win. Same with buying a dress at H and M, same with finding a good clinical trial, it’s just that in the latter case we’ve been too sandbagged by regulators to build it yet. And sadly it’s the patients and their families who suffer.
'Standard' clinical trials run very slowly, and mostly inefficiently. No one person is to blame, but solutions can be applied to fix this. Updating ct.gov is part of the fix but that demands downstream changes to how all component parts function.
In response to your article Bess, I wrote the following blog;
This whole problem just pattern-matched in my head to Patrick McKenzie’s long but riveting account of ~unfucking the covid vaccine rollout. https://worksinprogress.co/issue/the-story-of-vaccinateca/ And I’m not sure what to do with that. (Viscerally tbh I have an urge to Make Patrick See The Thing, but we can’t just throw Patrick at all the things? Hmmm)
1. Good luck with all of this. I went through this with my daughter when she was diagnosed with DIPG. It isn't easy, so I will pray for Jake and you.
2. When you think about standardizing all the information about clinical trials, you should research the UK's Electronic Health record system. They tried to put all the medical records in the UK online, and after $24 billion they failed. Years and Years of computer and technology system have been built up whose purpose is not Clinical trials. It isn't easy or cheap to reconfigure these systems for the purpose of helping create a better clinical trial database. When I look at these problems, I always try to remember Dunning-Kruger. It seems like it should be easy to me because I don't understand enough about the difficulty.
3. If you haven't been introduced to Dr. Azra Raza's "The First Cell" it is a good book to read on cancer research. Depending on your mind set, it can be an emotionally tough book to read. She talks first hand about her patients' cancer journey and her husbands.
4. After spending a bunch of years looking at these problems from the lens of pediatric cancer, I believe the fault lies mostly with the FDA and Congress. The rules for what new drugs can be approved, and what data can be used to prove the effectiveness of these drugs, isn't compatible with terminal diagnosis. Giving terminal children placebos is unethical. No member of Congress or the FDA would stand for it if it was their children, but the FDA requires it. The Scientific method is an iterative process. The FDA is not open to iterative testing.
5. The painful truth is you join a clinical trial with hopes of finding a cure, but in reality you are giving information to the doctors and researchers to help all future patients. The likely outcome is never good for the patient enrolling in the trials, but the long term benefits are unmeasurable.
I worked on the software that pulled together the data from a number of the UK EHR systems to make it available for clinical research. The component parts were a challenge, but ultimately it was regulations and red tape that put the brakes on any viable successful outcomes. The UK has some of the best coded EHR data in the world going back to 1989, but the restrictions on being able to effectively use the data are criminal.
I am writing a blog of my own on this topic but I am stalled on the ending.
I have a very different outlook than you and Jake on my approach to my cancer treatment, which is ironic since Jake and I are both influenced by Stoicism. I have declined all treatment and am focused on enjoying life to the full with my family. My calculation was that the benefits of chemo were not knowable and I didn't want to risk 9 months of feeling crappy and then dying at the end of it when I am still quite healthy for the moment. I think there's an overwhelming pressure to Fight! Fight! Fight! and I think that can cloud people's judgement when they make decisions about their treatment.
I don't mean to suggest that my approach is the right one or that yours is wrong. Jake and I are in very different places in our lives and in our disease progression after all. I just wanted to highlight that there other ways of thinking about this.
I was hoping we’d hear from you on this! Agreed. No one right way to do things which is why patient guided goals of care are so important. Also as you mentioned different cancers move differently. Jakes is particularly aggressive and not treating or delaying even a bit would have meant death, but that’s not always the case. I don’t think patient led care goals are something we do particularly well in medicine and need to improve on. But that also means that we need to get better at really explaining to people what’s potentially futile, what side effects are really like, what a treatment means in terms of benefit/distress. Too often we want to offer “somethjng” instead of being clear about what that “something” really means to a patient’s quality and quantity of life. All these conversations need to improve.
I do have a lot of sympathy for the oncologists and surgeons who have to have these conversations with patients who are dying. Except my neurosurgeon. He learned his bedside manner from Doctor House. No sympathy for him.
Oh yeah, there are definitely some ....unique approaches 🥶It’s also just really hard to do, and often times there’s no right way. In the ER we have a lot of these conversations that slipped through the cracks before, which is its own challenge, since we don’t have a preestablished patient relationship
I want to respond quickly to this and then share my own experience but as a patient with terminal cancer and as the CEO of a company that builds online communities for cancer patients.
> The political will required would be the willingness to hire programmers with UX experience.
It's definitely not a UX problem. It's a data problem. We spent several years building a trial search engine with the kind of dress-buying UX that you describe. It didn't work because the data is no good. It would be easy to add a handful of fields to the data in ct.gov that make the data relevant to patients and easy to search. AI will help a little — but not as much as fixing the data and fixing the data will be easy.
Smart Patients has over 100 communities for cancer patients and most have a focus on medical research and, especially, discussing clinical trials. The experience is very different for different kinds of cancer. Stage 4 renal cell carcinoma, for example, had practically zero treatments 15 years ago and the outlook was bleak. Once immunotherapies and targeted therapies started to come on line, there were suddenly dozens of trials and most of our members stay alive by surfing from trial to trial. But here's the thing: you don't need a trial search engine for RCC because there are only about 100 credible trials. Patients share spreadsheets with all the trials. By contrast, there are hundreds of trials of NSCLC and SCC so you need a way to search them.
By contrast in the other direction, there are almost no trials for low grade gliomas (LGG — my cancer) and very little research. All the research dollars go to glioblastomas. There have been only two new drugs approved for gliomas in 50 years*.
My moral in this story is that there is not just one story about clinical trials. Patients with different cancers will have different experiences. What works for SCC patients might not work for other cancers.
EPILOGUE
* Vorasidinib will likely be approved in the USA in the spring. Vorasidinib is the first new drug for gliomas since 2005. The trial showed a 50% improvement in progression free survival over the standard of care. Let's hope I live long enough to see it approved in the NHS!
Good point. Bad data is bad data. I don’t think building anything on the back of clinical trials.gov will solve the “ garbage in Garbage out problem”, but im sad to hear better ux didn’t help the experience a little. Trying to search something like clinical trials.gov that was never intended to be searched in that way seems like a Greek myth’s punishment. What would you find the most useful change to you, If you could be In charge of fixing the data?
2. Tidy up and standardize the names of drugs and conditions so you can search consistently.
3. Use a consistent format that's patient-friendly and machine-readable format for inclusion criteria.
I wrote a paper on inclusion criteria a while back. I'll see if I can find it. It's not a hard problem. It called a Domain-Specific Language (DSL) in the CS lingo.
At the moment, the inclusion criteria are all over the place. The same phrase can mean the opposite from trial to trial. I found 95 different ways of saying "the patient must not be pregnant."
Dear Bess. Thank you for writing this. I was trying, perhaps too obliquely, to agree with you about the need for an (inter)national trial database. I think that oncologists by and large do try and do the best for their patients. We can all reasonably be expected to be up to date on the latest standard of care. However, there’s a wider spectrum in how up to date we are with the latest research efforts.
Your point about it being reasonable to expect your oncologist to be au fait at least with the most relevant and exciting phase 3 trials in their area is very reasonable - and I think most specialist oncologists are, especially at large academic centres. Most of us do spend at least a few hours a month learning about the most recent papers in NEJM or the most recent presentations at ASCO or ESMO.
There is also the capacity to phone a friend - and even general oncologists working in rural or regional centres far away from, say, the National Cancer Institute (or state level equivalent) will often call or email a colleague at the relevant centre to ask about trials when faced with a case like your husband’s.
It’s just that the world is too big for all this to be done by humans! Of course there needs to be a database. Imagine if humans were still the only travel agents and sites like Flight Centre didn’t exist; you’d expect a person to know all possible flight paths between (say) New York and Dubai and which was the quickest vs the cheapest? Humans used to be able to do a reasonable job, before databases and algorithms, and a talented and plugged in human could probably still do OK - but clearly this is an area where a good search engine will win. Same with buying a dress at H and M, same with finding a good clinical trial, it’s just that in the latter case we’ve been too sandbagged by regulators to build it yet. And sadly it’s the patients and their families who suffer.
'Standard' clinical trials run very slowly, and mostly inefficiently. No one person is to blame, but solutions can be applied to fix this. Updating ct.gov is part of the fix but that demands downstream changes to how all component parts function.
In response to your article Bess, I wrote the following blog;
https://www.linkedin.com/pulse/investigator-sites-clinicaltrialsgov-failure-eclinical-doug-bain-lmu5e/
This whole problem just pattern-matched in my head to Patrick McKenzie’s long but riveting account of ~unfucking the covid vaccine rollout. https://worksinprogress.co/issue/the-story-of-vaccinateca/ And I’m not sure what to do with that. (Viscerally tbh I have an urge to Make Patrick See The Thing, but we can’t just throw Patrick at all the things? Hmmm)
Sure we can!
A couple of items.
1. Good luck with all of this. I went through this with my daughter when she was diagnosed with DIPG. It isn't easy, so I will pray for Jake and you.
2. When you think about standardizing all the information about clinical trials, you should research the UK's Electronic Health record system. They tried to put all the medical records in the UK online, and after $24 billion they failed. Years and Years of computer and technology system have been built up whose purpose is not Clinical trials. It isn't easy or cheap to reconfigure these systems for the purpose of helping create a better clinical trial database. When I look at these problems, I always try to remember Dunning-Kruger. It seems like it should be easy to me because I don't understand enough about the difficulty.
3. If you haven't been introduced to Dr. Azra Raza's "The First Cell" it is a good book to read on cancer research. Depending on your mind set, it can be an emotionally tough book to read. She talks first hand about her patients' cancer journey and her husbands.
4. After spending a bunch of years looking at these problems from the lens of pediatric cancer, I believe the fault lies mostly with the FDA and Congress. The rules for what new drugs can be approved, and what data can be used to prove the effectiveness of these drugs, isn't compatible with terminal diagnosis. Giving terminal children placebos is unethical. No member of Congress or the FDA would stand for it if it was their children, but the FDA requires it. The Scientific method is an iterative process. The FDA is not open to iterative testing.
5. The painful truth is you join a clinical trial with hopes of finding a cure, but in reality you are giving information to the doctors and researchers to help all future patients. The likely outcome is never good for the patient enrolling in the trials, but the long term benefits are unmeasurable.
I worked on the software that pulled together the data from a number of the UK EHR systems to make it available for clinical research. The component parts were a challenge, but ultimately it was regulations and red tape that put the brakes on any viable successful outcomes. The UK has some of the best coded EHR data in the world going back to 1989, but the restrictions on being able to effectively use the data are criminal.
I am writing a blog of my own on this topic but I am stalled on the ending.
I have a very different outlook than you and Jake on my approach to my cancer treatment, which is ironic since Jake and I are both influenced by Stoicism. I have declined all treatment and am focused on enjoying life to the full with my family. My calculation was that the benefits of chemo were not knowable and I didn't want to risk 9 months of feeling crappy and then dying at the end of it when I am still quite healthy for the moment. I think there's an overwhelming pressure to Fight! Fight! Fight! and I think that can cloud people's judgement when they make decisions about their treatment.
I don't mean to suggest that my approach is the right one or that yours is wrong. Jake and I are in very different places in our lives and in our disease progression after all. I just wanted to highlight that there other ways of thinking about this.
Good luck to us both!
I was hoping we’d hear from you on this! Agreed. No one right way to do things which is why patient guided goals of care are so important. Also as you mentioned different cancers move differently. Jakes is particularly aggressive and not treating or delaying even a bit would have meant death, but that’s not always the case. I don’t think patient led care goals are something we do particularly well in medicine and need to improve on. But that also means that we need to get better at really explaining to people what’s potentially futile, what side effects are really like, what a treatment means in terms of benefit/distress. Too often we want to offer “somethjng” instead of being clear about what that “something” really means to a patient’s quality and quantity of life. All these conversations need to improve.
I do have a lot of sympathy for the oncologists and surgeons who have to have these conversations with patients who are dying. Except my neurosurgeon. He learned his bedside manner from Doctor House. No sympathy for him.
Oh yeah, there are definitely some ....unique approaches 🥶It’s also just really hard to do, and often times there’s no right way. In the ER we have a lot of these conversations that slipped through the cracks before, which is its own challenge, since we don’t have a preestablished patient relationship
I want to respond quickly to this and then share my own experience but as a patient with terminal cancer and as the CEO of a company that builds online communities for cancer patients.
> The political will required would be the willingness to hire programmers with UX experience.
It's definitely not a UX problem. It's a data problem. We spent several years building a trial search engine with the kind of dress-buying UX that you describe. It didn't work because the data is no good. It would be easy to add a handful of fields to the data in ct.gov that make the data relevant to patients and easy to search. AI will help a little — but not as much as fixing the data and fixing the data will be easy.
Smart Patients has over 100 communities for cancer patients and most have a focus on medical research and, especially, discussing clinical trials. The experience is very different for different kinds of cancer. Stage 4 renal cell carcinoma, for example, had practically zero treatments 15 years ago and the outlook was bleak. Once immunotherapies and targeted therapies started to come on line, there were suddenly dozens of trials and most of our members stay alive by surfing from trial to trial. But here's the thing: you don't need a trial search engine for RCC because there are only about 100 credible trials. Patients share spreadsheets with all the trials. By contrast, there are hundreds of trials of NSCLC and SCC so you need a way to search them.
By contrast in the other direction, there are almost no trials for low grade gliomas (LGG — my cancer) and very little research. All the research dollars go to glioblastomas. There have been only two new drugs approved for gliomas in 50 years*.
My moral in this story is that there is not just one story about clinical trials. Patients with different cancers will have different experiences. What works for SCC patients might not work for other cancers.
EPILOGUE
* Vorasidinib will likely be approved in the USA in the spring. Vorasidinib is the first new drug for gliomas since 2005. The trial showed a 50% improvement in progression free survival over the standard of care. Let's hope I live long enough to see it approved in the NHS!
50% is amazing! Fingers crossed for you.
Good point. Bad data is bad data. I don’t think building anything on the back of clinical trials.gov will solve the “ garbage in Garbage out problem”, but im sad to hear better ux didn’t help the experience a little. Trying to search something like clinical trials.gov that was never intended to be searched in that way seems like a Greek myth’s punishment. What would you find the most useful change to you, If you could be In charge of fixing the data?
1. A patient-friendly summary would be good.
2. Tidy up and standardize the names of drugs and conditions so you can search consistently.
3. Use a consistent format that's patient-friendly and machine-readable format for inclusion criteria.
I wrote a paper on inclusion criteria a while back. I'll see if I can find it. It's not a hard problem. It called a Domain-Specific Language (DSL) in the CS lingo.
At the moment, the inclusion criteria are all over the place. The same phrase can mean the opposite from trial to trial. I found 95 different ways of saying "the patient must not be pregnant."
Only 95?
I uploaded my paper — Simplifying Eligibility Criteria here:
https://www.raggedclown.com/wp-content/uploads/2023/12/Simplifying-Eligibility-Criteria.pdf
Updated the post body to include your comments here. Checking out your paper now