The Solution to Cancer; Algorithms as Drugs
Christopher Gregg PhD, Associate Professor of Neurobiology and Human Genetics, Huntsman Cancer Institute Member, University of Utah; Co-founder Storyline Health Primordial AI and The Free Uncharted Cancer Patient Masterclass.
On a Saturday morning in December of 2018, I was alone in the house and received a call. My doctor was phoning my cell. He was upset and informed me that the back and hip pain I had been struggling with was due to cancer. The latest MRI found large tumors in my spine and hip. Worse, it was soon revealed that what was thought to be a benign growth removed from me 7 years earlier was actually cancer that had been misdiagnosed, inadequately treated, and left to grow unchecked. Having just published a paper on how elephants evolved solutions for cancer, I traumatically moved from researcher to cancer patient.
My wife and I waited until after Christmas to inform my daughter (11 years old at the time) and son (9) that I had male breast cancer - a rare form of cancer - and it was stage IV, which offered a median life expectancy of ~3 years. My oncologist at the University of Utah gently shared that I was at a dead end; there was no cure and no clinical trials available at that time for initial therapy of metastatic breast cancer in men. I would receive only palliative care to control the disease as long as possible and keep me comfortable while I died. But she offered a critical insight. After decades of practicing oncology and seeing waves of hype for different treatments come and go, she believed that the field needed to focus on “using the medicines we have in better ways”.
As a scientist, I understood I would have to take my fate into my own hands and find a solution outside of current standards that could offer me nothing. I've been working on the problem of how to use existing medicines in smarter ways for myself and others for the last 6 years - fate and heroic scientists and clinicians have helped me at every step, including Dr. Saundra Buys MD - my brave oncologist. We now know much more.
Cancer care is constrained by the tyranny of the small molecule patent. New candidate drugs are protected by small molecule patents that protect the unique molecular structure of the drug and provide companies exclusive access and use of this new drug. This is critical, because candidate treatments must succeed in clinical trials for oncologists and insurance companies to adopt them into the standard of care. Median clinical trial costs are ~$3.4 million for phase I, $8.6 million for phase II, and $21.4 million for phase III. Despite these significant investments, over 90% of new drugs fail to gain approval after entering clinical trials. Consequently, the financial returns from successful drugs must cover the costs of the numerous unsuccessful ones ($2.3B per approved drug), highlighting the immense economic risks and the need for substantial capital in drug development. Investors will only do that if the financial return is substantial. The return will only be substantial if the company can defend its intellectual property in the marketplace. Hence, the entire system is built on the extraordinary defensibility of a small molecule patent. This reality is stifling real advances in cancer treatment.
A successful new cancer drug only needs to extend life in a patient by >2.5 months. That is a billion-dollar win for pharma. Although these drugs may work initially, through Darwinian evolution and selection the cancer cells evolve and become resistant, rendering the drug useless. This is called cancer progression. While cancer has more approved drugs than any medical field, the disease always evolves resistance and the doctor (and patient) run out of options, lose control, and the patient dies (Fig. 1). One can see that the problem to solve is not simply making new and better drugs but developing solutions to prevent the development of treatment resistance. So why doesn’t pharma work on that?
Fig. 1. [Rough concept] Current standard of care paradigm for advanced cancer treatment always leads to cancer resistance and death.
A small group of pioneering scientists/doctors do focus on this, and I found them within months of being diagnosed. They are now my friends and I hope the world recognizes and embraces them - Drs. Robert Gatenby, Joel Brown, Sandy Anderson, Dawn Lemanne, and Carlo Maley. They developed mathematical models of cancer to reveal how to control it. These are algorithms, step-by-step procedures for combining and sequencing drugs at the right time and in the right dose to reduce the risk of having the cancer evolve resistance to the treatment, while still safely controlling the disease.
Critically, they recognized that farmers also have the problem that pests evolve resistance to pesticides, but the farming community has developed effective resistance management plans to combat this. In short, these plans (i) integrate non-chemical-based approaches for pest control to minimize chemical use, (ii) rotate between different pesticide chemical classes constantly so that the same chemical is never sprayed continuously, and (iii) minimize pesticide use by carefully monitoring pest populations and applying chemicals only when needed to control, not eliminate pests. None of these basic principles are used by oncologists. Instead, they spray the same chemical (drug) continuously at a maximal tolerable dose until resistance (Fig. 1).
A frustration is that these solutions for treatment resistance management have been known for over 15 years in the cancer field. Treatment resistance management in this setting refers to sequencing drugs from different chemical classes and switching drugs before resistance emerges, or giving a drug intermittently at key time points to control the disease with minimal drug use and without eliminating the drug-sensitive cells. There are also ways to impactfully integrate non-drug based interventions (eg. diet, exercise). These approaches, inspired by farming, can help cancer drugs to continue to work. They are expected to extend the life of stage IV cancer patients by years, not months, and perhaps even make advanced cancer a chronically manageable disease. Not only that, the quality of life for the patient goes up, because the patient may intermittently go off treatment to minimize drug use. So what happened?
Well, one must test these treatment paradigms in large clinical trials that can cost many millions of dollars, but the intellectual property is not there. You can’t patent simple rules of thumb for treatment resistance management. Moreover, if a metastatic cancer patient can control their disease with off-patent, low-cost drugs for years or decades, only the patient wins.
I am nearly 6 years out from my diagnosis and have rotated over 15 drug treatments from different chemical classes, failing on only one. In other words, my treatment resistance management plan is working so far. Like my heroic colleagues, I am desperate to change care and help patients by testing these approaches in trials. I have tried to build a free online masterclass to educate providers and patients, and frustratingly tried to raise money through grants, donors, foundations, and other approaches to accelerate this solution but, frankly, the only viable path that will sustainably cover the high costs of trials is a commercial solution.
“Algorithms as drugs” is the future of pharma and healthcare. I envision a shift in which the intellectual property strategy from the software tech industry is lifted into the pharma industry. Proprietary treatment algorithms that use existing medicines in smarter ways (Fig. 2) will be the defensible marketable products that are tested in clinical trials and delivered at massive scale to patients and caregivers through the web and smartphones. The intellectual property will be built on trade secrets and proprietary data and decision algorithms, rather than small molecule patents.
Algorithms as Drugs for cancer.
Fig. 2. Treatment algorithms for preventing drug resistance in cancer with the goal of cancer extinction.
This revolution will help enable scalable, low-cost precision medicine. Decision tools and care algorithms will integrate multiple different types of patient data and continuous monitoring technologies to personalize and adapt drug treatments. That means predicting patient responses and adapting the drug, dose, and timing to optimize the outcome for the patient based on symptom and disease metrics (not the genome). This science fiction is happening and real. The artificial intelligence revolution makes it feasible. But there are systemic problems.
First, the FDA regulatory process is structured for new drugs, not new algorithms. One can attempt to shoe-horn algorithms into the Software as a Medical Device (SaMD) bucket at the FDA, but it is not ready for sophisticated “Algorithms as Drugs” decision tools. The rules are emerging and in their infancy. A clear regulatory path is urgently needed.
Insurance needs to enable doctors to monitor patients more regularly and easily switch treatments for different care algorithms without the risk of losing coverage. Currently, if a patient switches to a new treatment, coverage for the previous treatment can be stopped. In this new Algorithms as Drugs world, insurance companies will benefit because patients will be on cheaper drugs for longer and take less drug to control the disease. They need to enable and incentivize this.
Personalized adaptive care algorithms are tricky to deliver and use reproducibly in a real-world clinical setting. Imagine the complexity of each patient in a hospital receiving vastly different care, including different drug treatment schedules, doses, drug combinations, lab tests, and procedures that must be precisely timed. Moreover, patients will need constant monitoring for this adaptive care. This means new, scalable, and remote clinical support tools are needed.
Transparency will be tricky. Companies will need to defend their trade secrets and data moats. Yet, doctors will be reluctant to deliver care algorithms that they don’t understand or have insight into. This means that the Algorithms as Drugs revolution will need educational infrastructure and training support, and regulations that engender confidence in clinical experts. In truth, we don’t understand how many drugs work, yet clinicians become familiar with response patterns, side effects, and risks. The same will be true for Algorithms.
This new algorithm-based system is predicated on large amounts of patient data. Patients who are desperate for improved treatments often do not realize that they are the limiting resource for finding and proving those treatments. Clinical trials compete for patients and every company who works on Algorithms as Drugs products will need to find ways to engage patients in donating their data to train models that predict optimal treatment algorithms for disease control. My own view is that someone took a placebo sugar pill in a previous clinical trial so that I could benefit from a new effective drug today, so I owe it to others to do my part for the system too and share my data. In exchange, companies should financially reimburse patients that donate data, empower them to have control over their data, acknowledge them, educate them, defend them, advocate for them, and protect their privacy. It is a partnership.
Finally, insurance will need to reimburse “Algorithms as Drugs” as robustly as it embraces the high costs of newly approved drugs. Care algorithms will be cheaper, lower risk, and much more scalable, but the market will demand some impressive financial returns in exchange for the risk and expense of funding the clinical trials. This will surely be cheaper than making new drugs and we are expecting these cancer care algorithms to support care for years, rather than just a few months. Thus, revenue can be spread out over time.
The need for the Algorithms As Drugs shift in cancer care is urgent. One in five people will die from cancer and the disease is striking younger people more frequently. The current precision genome medicine movement that makes new drugs for fewer people based on their specific genetic profiles will fail economically. The drugs still cost billions to make, yet the market is now smaller. Additionally, the successes are few and far between. We can do better. We can make advanced cancer a more manageable disease with the drugs we have by implementing new artificial intelligence solutions (with a little help from farmers). The lessons and technologies that solve this for advanced cancer will rapidly expand into other fields of medicine.
Metastatic patients frequently have billions of cancer cells in their bodies. One drug will never reliably cure most metastatic cancers. Individual cancer cells vary widely in terms of their properties and drug sensitivities. Small populations may be vulnerable to some drugs. Large populations vulnerable to others. Believe it or not, treatment sensitivity mostly needs to be worked out empirically in each patient - a great problem to optimize with new care algorithms! There are 574 approved cancer drugs. That is one hell of an arsenal. Drugs can be repurposed and combined in smarter ways with patient data in this new Algorithms as Drugs ecosystem and protected as intellectual property.
I have hard earned opinions. The NIH, which is the backbone of medical research, needs to prioritize a shift from genome medicine to adaptive care algorithms that use existing medicines in smarter ways. The FDA needs to create the regulatory pathways for this shift. Pharma needs to embrace a new algorithm-based intellectual property strategy that can help people more effectively and quickly at lower cost. Insurance companies, Medicare, and Medicaid need to support and reward effective alternatives to expensive new drugs. And patients need to get behind the Algorithms as Drugs movement by engaging, sharing data, informing their doctors, and expressing anger. If 1 in 5 people died from COVID, how angry would you be? Patients and their family members will drive this change in the world.
We are aware of the problems. We are aware of solutions. However, implementing these solutions requires substantial resources. Those who take the initiative and lead this change will reap rewards and patients like me are desperately rooting for you. Resources are needed for commercial efforts and technology development, to advance academic research into Algorithms as Drugs, support this new type of care in the clinic, provide patient and provider educational support, and lobby for clear regulatory pathways. It can’t come fast enough.
Our winning team!(That's me in the middle with my arms raised.)
Finally, insurance will need to reimburse “Algorithms as Drugs” as robustly as it embraces the high costs of newly approved drugs. Care algorithms will be cheaper, lower risk, and much more scalable, but the market will demand some impressive financial returns in exchange for the risk and expense of funding the clinical trials. This will surely be cheaper than making new drugs and we are expecting these cancer care algorithms to support care for years, rather than just a few months. Thus, revenue can be spread out over time.
The need for the Algorithms As Drugs shift in cancer care is urgent. One in five people will die from cancer and the disease is striking younger people more frequently. The current precision genome medicine movement that makes new drugs for fewer people based on their specific genetic profiles will fail economically. The drugs still cost billions to make, yet the market is now smaller. Additionally, the successes are few and far between. We can do better. We can make advanced cancer a more manageable disease with the drugs we have by implementing new artificial intelligence solutions (with a little help from farmers). The lessons and technologies that solve this for advanced cancer will rapidly expand into other fields of medicine.
Metastatic patients frequently have billions of cancer cells in their bodies. One drug will never reliably cure most metastatic cancers. Individual cancer cells vary widely in terms of their properties and drug sensitivities. Small populations may be vulnerable to some drugs. Large populations vulnerable to others. Believe it or not, treatment sensitivity mostly needs to be worked out empirically in each patient - a great problem to optimize with new care algorithms! There are 574 approved cancer drugs. That is one hell of an arsenal. Drugs can be repurposed and combined in smarter ways with patient data in this new Algorithms as Drugs ecosystem and protected as intellectual property.
I have hard earned opinions. The NIH, which is the backbone of medical research, needs to prioritize a shift from genome medicine to adaptive care algorithms that use existing medicines in smarter ways. The FDA needs to create the regulatory pathways for this shift. Pharma needs to embrace a new algorithm-based intellectual property strategy that can help people more effectively and quickly at lower cost. Insurance companies, Medicare, and Medicaid need to support and reward effective alternatives to expensive new drugs. And patients need to get behind the Algorithms as Drugs movement by engaging, sharing data, informing their doctors, and expressing anger. If 1 in 5 people died from COVID, how angry would you be? Patients and their family members will drive this change in the world.
We are aware of the problems. We are aware of solutions. However, implementing these solutions requires substantial resources. Those who take the initiative and lead this change will reap rewards and patients like me are desperately rooting for you. Resources are needed for commercial efforts and technology development, to advance academic research into Algorithms as Drugs, support this new type of care in the clinic, provide patient and provider educational support, and lobby for clear regulatory pathways. It can’t come fast enough.
Want to know more?
