I have to say, this new paper is good news on the oncology front. It's from researchers at Sloan-Kettering and BioNTech (among others) and presents the results of a small trial treating patients with pancreatic cancer, specifically pancreatic ductal adenocarcinoma (PDAC). This is another crack at the "personalized cancer therapy" idea and simultaneously at the "cancer vaccine" idea, and it's fair to say that over the years both of those (while scientifically appealing) have had many false starts and disappointments along the way. But recently that's been changing, and this trial is a clear example.

It only involved sixteen patients, and normally one wouldn't start drawing conclusions from such a small sample. But this is another example of effect size: PDAC is the most common form of pancreatic cancer, the third leading cause of cancer-driven death in the US, and is obviously a very, very bad diagnosis. Around 90% of the patients die within five years from this fast-moving disease, and there are really no effective treatments available. Now, if you run a clinical trial and get that number down to 87% or something, that's going to need a lot of statistical power to convince anyone that you have a real result. But in this case, 8 of the 16 patients have had no recurrance of disease over an 18-month monitoring period after treatment, and it's safe to say that that's unheard of in this disease. That's a huge effect size, and these results are being taken very seriously indeed. In addition, one of the patients showed (by imaging and tissue sampling) what appears to have been an attempted metastatic tumor in the liver, which showed up and then disappeared, apparently due to the treatment.

Here's how the trial worked: patients had tumor tissue removed surgically, and the BioNTech team used these samples to come up with a "neoantigen" profile for each individual patients. These are unusual proteins that are basically produced only in the tumor tissue and are thus strong candidates for raising an immune system attack, since you have such a good chance of not going after normal tissues. People have been trying this for many years now, and what we've learned so far is that (1) some tumor types produce a lot more such neoantigens than others and that (2) you're probably going to have to hit quite a few of them at once, given the variations in human immune systems. In this case, BioNTech produced mRNA vaccines that targeted up to 20 neoantigens per patient, each of them a custom dose. That, as you can readily imagine, is (at least for now) an expensive process - after years of work, BioNTech says that they can get it down to less than $100,000/dose. I'm sure it can be driven down further, but this is highly unlikely to ever be cheap. They also got an anti-PD-L1 antibody and a four-component chemotherapy cocktail (those last two have already been tried in PDAC therapy without the vaccine dose, to unimpressive effect). The vaccine was administered across nine doses.

These mRNA injections work exactly the way that mRNA vaccines work in general (as with the coronavirus vaccines from Pfizer/BioNTech and from Moderna): they are formulated in lipid nanoparticles that work their way into cells, causing them to produce the associated neoantigen proteins. These then set off an immune response, which can then get directed to the tumor tissue itself. Why, you may be asking yourself, do we have to go all the way around the barn like this? If these tumors are producing all these funky unnatural-looking proteins, why doesn't an immune response get mounted right there and then? That is a solid question, which we've gradually been answering: tumors, it turns out, can bring all sorts of mechanisms into play to keep that from happening. The environment right around the tumor tissue is extremely unfavorable to mounting an effective immune response; that's one of the ways a tumor become a tumor instead of a single rogue cell that is never heard from again. But if these neoantigen proteins are produced elsewhere in the body, they get responded to, and the resulting immune cell attack has a chance to really get going.

And that's what seems to have happened here. Very significantly, the eight patients who have shown benefit in this trial were the eight that mounted a vaccine-expanded T-cell reponse. It appears that in those patients up to 10% of their circulating T cells were in fact directed against the various neoantigens. (Even so, half of the responders only had T cells against one neoantigen, from what I can see). The other 8 patients did not show this sort of response, and their disease recurred after an average of about 13 months. It was a complex process overall: the team recruited 34 patients, and 28 of these were able to undergo the surgery to remove cancerous tissue (it should be noted that under the usual conditions, PDAC almost invariably comes roaring back after this sort of treatment). 19 of them were treated with the PD-L1 antibody, and 16 of them also got the personalized mRNA dose (one of the 19 had insufficient neoantigens, and no vaccine could be manufactured). 3 of the 16 did not complete the nine-dose schedule, due to disease progression, death, or inability to tolerate the accompanying chemotherapy regime.

So the obvious question is: what's the difference between the people who responded to this therapy and the people who didn't? It wasn't general immune competence - the team checked that by looking at the response to BioNTech's mRNA shot for coronavirus as a control, and all of the patients showed the expected immune profile after that. The clinicians tried to correlate all sorts of other factors to survival as well (response to the PD-L1 antibody, lymph node positivity, margin positivity, primary tumour size, the number of chemotherapy doses and density of intratumoural CD8+ T cells), and none of these tracked. They also showed equivalent numbers of all the major circulating immune cell types. The best theory at the moment is that the responding patients had neoantigens that were of higher immunologic quality (more likely to bring on a strong response), and that their tumors showed a better distribution of these across the various clonal lines in the tissue itself. Clearly a lot more work has to be done in this area - the authors say that they're working on further expanding the list of neoantigens and sorting them out by likely immune response, for example. 

This technique is going to be the subject of a larger trial, as it absoutely should be, and we'll see how many such refinements can be incorporated into that one. But even as it stands, this is an extremely promising result, which we can expect to see extended to other tumor types as well. Immuno-oncology marches on, and it's a glorious thing to watch.