If you look at the clinical trial landscape in oncology, you will see a nearly uncountable number of trials testing various immuno-oncology drugs (generally targeting PD-1, PD-L1, and/or CTLA-4) with almost every other therapy we know how to administer. The belief is that there are some of these combinations that will prove unusually effective, showing better results than one might expect from the activities of each part of the combination on its own. This sort of synergy is not a crazy hope – mechanistically, you can imagine that disrupting some biochemical pathways could potentiate the disruptive effects of hitting other ones, by disabling the compensatory effects that might otherwise slow them down. A classic example in chemotherapy is the idea that interfering with DNA repair pathways can make some forms of chemotherapy more effective, by making their damaging effects harder to overcome in rapidly dividing cells.
So are there such synergies with the immune checkpoint therapies? If you search for variations on that phrase, you’ll get an awful lot of hopeful literature references. A popular theme is the idea of making immunologically “cold” tumors into “hot” ones: that is, making a wider variety of tumor types susceptible to the checkpoint inhibitors by making them more reactive to the immune system than they usually are. There are many ideas about how to do that, often centered around the reaction of such tumors to existing therapies and their possibility of flipping that sort of switch. You’d probably only see a big effect if you were simultaneously dosing the right checkpoint inhibitor, and thus the large number of trials looking for something that stands out.
This new paper argues that we haven’t found it yet (here’s a look at it at Stat). It’s an analysis of thirteen Phase III trials combining various checkpoint inhibitors with each other and with other oncology drugs, then comparing the results to the closest-matched clinical data for the individual therapies in earlier trials. And the authors find no synergies, just additive effects of each therapy, one after the other. That’s not a bad thing (at least they aren’t interfering with each other and cancelling things out), but it’s not the good thing that everyone has been searching for, either. The authors note that this is in great constrast to many reports in the literature, but that these synergistic effects are invariably seen in mouse models rather than in the clinic. They place the blame on the differences between very homogeneous mouse models which have been selected over the years to show strong responses versus the far more varied human cancer patient population, and that has to be a big part of the answer. Oncology isn’t the only field that has suffered from this, either! There are patient-derived xenograft rodent models that are more realistic, but the problem is that those animals have to be immune-compromised, so checkpoint inhibitors aren’t going to have much effect in them to start with.
The paper goes into several other reasons to believe its results. The limitations in their analysis and in the underlying clinical data would be expected to make synergies more likely to appear even when they aren’t real. For example, some of the monotherapy comparisons are study arms in previously treated patients, where one would expect things to be less likely to work (and thus make the combination trial arms look better by comparison). So not seeing any evidence of them under these conditions is quite telling.
Another take-home from this study comes from remembering that these synergy-seeking trials aren’t just being done randomly. There’s a rationale for each of the combinations based on what we know about tumor biology, and the fact that we’re not seeing synergy leads you to the uncomfortable conclusion that we must not know as much about cancer biology as we think we do. That’s always been a safe bet, but this is a particuarly comprehensive way to have our noses rubbed in that fact. At the same time, it’s also true that giving combination of drugs is still the way to go in actual practice – that heterogeneity in the patient population means that you have a better chance of providing some benefit. It’s just that it’s not a better chance because we eagle-eyed biomedical types identified the perfect combo and zeroed in on it – because we are in the process of proving that we can’t do that yet!
