Something quite remarkable happened in medicine today. Moderna and Merck announced positive Phase 3 results for their personalized cancer therapy, intismeran autogene, combined with Keytruda, in patients with high-risk melanoma. The 1,137-patient trial met its main recurrence-free survival endpoint and a key endpoint measuring distant metastasis. The companies call it the first positive Phase 3 readout for an individualized neoantigen therapy and an mRNA-based cancer therapy. The detailed results are still to come, but something else matters:how the medicine is made.
A sample of a patient's tumor is sequenced and mutations specific to that person's cancer are identified. From these, a set of neoantigens is selected and an mRNA therapy containing up to 34 of them is manufactured specifically for that patient. Merck and Moderna describe it as a treatment built around the unique mutational fingerprint of each patient's tumor.
Two people with melanoma may therefore receive two different medicines, not because their doctors chose different drugs, but because their cancers are biologically different. For most of modern medicine, information about a patient has helped us decide which medicine to give them. Here, information about the patient helps determine what medicine gets made. I think that is a much bigger shift than it first appears.
There is also an enormous amount of computation underneath this. Sequencing and computational biology help identify which mutations may matter to the immune system, and Personalis has provided genomic testing to Moderna's program. This may be what a lot of AI in medicine eventually looks like: intelligence embedded in the machinery through which we understand biology.
A person is more than their tumor
Today, this treatment is personalized primarily using molecular information from the tumor. But a person is more than their tumor. Genotype tells us what is encoded in our genes. Phenotype describes how that biology manifests in a person through laboratory values, illness, medications, imaging and treatment response. We have written before about why phenotypes may be a missing layer in personalized medicine.
And then there is time. A phenotype at one moment tells us something about a person then. A longitudinal phenotype tells us how that person has been changing. Two people can share a diagnosis and genetic characteristics while having very different treatment responses and disease trajectories. This is why I increasingly think personalized medicine is also a data problem: personalization needs context that persists across time.
The memory of the patient is still episodic
Yet this is precisely the kind of information healthcare is particularly bad at preserving. We are learning how to make medicines for individuals while healthcare still remembers individuals largely as collections of reports, prescriptions, scans and isolated encounters. The medicine is becoming personal. The memory of the patient is still episodic.
A lot of what we are building at Aether comes from trying to preserve a person's clinical history as something continuous rather than reconstructing it every time healthcare needs to understand them. Patient information has traditionally been a record of care that already happened. What happened today points towards something different: what we know about an individual may increasingly become an input into what medicine can do next.
We are beginning to learn how to make a medicine for one person. If that is where medicine is going, we will also need a deeper understanding of the person for whom that medicine is being made, preserved through their life rather than reconstructed one encounter at a time. We are learning how to make the medicine personal. Now we need to make the memory of the patient personal too.
References
- CNBC: Moderna, Merck cancer vaccine shows initial late-stage melanoma data
- Merck and Moderna: Phase 3 INTerpath-001 topline results
- ClinicalTrials.gov: INTerpath-001 (NCT05933577)
- Personalis and Moderna: genomic testing for the personalized mRNA cancer vaccine program
- National Human Genome Research Institute: Phenotype