A survey of thousands of advanced cancers suggests a way to identify those most likely to respond to revolutionary therapies that unleash an immune response against tumours. But the results highlight how difficult it will be to translate such an approach into a reliable clinical test.
The findings1, published on 14 January in Nature Genetics, are the latest line of evidence to suggest that tumours with a large number of DNA mutations are more likely to respond to immunotherapies than are cancers with fewer mutations — and result in longer survival for people who receive treatment.
Researchers have long sought a way to select the people — generally a minority of patients — who are most likely to respond to immunotherapies and to spare others from the treatments' side effects, which can include kidney failure and lung problems.
Finding a threshold
But the immune system is complicated, and it has proved difficult to determine what makes one tumour vulnerable to treatment but allows another to escape unscathed. “The essence of pharmaceutical development is to try and find meaningful but relatively simple guideposts, which is often in defiance of the way that biology works,” says Chris Shibutani, a biotechnology analyst at the investment bank Cowen, in Boston, Massachusetts.
One hypothesis is that the more genetically different a tumour is from normal tissue, the more likely it is that the immune system will recognize and eliminate it. Luc Morris, a cancer researcher and physician at the Memorial Sloan Kettering Cancer Center in New York City, and his colleagues analysed DNA sequence data for advanced cancers from more than 1,600 people who had been treated with immunotherapies called checkpoint inhibitors. The team also analysed the sequences of advanced cancers from more than 5,300 people who hadn't been treated with checkpoint inhibitors. The study looked at ten different kinds of cancer, including melanoma and breast cancer.
In most of those cancers, the team found that a higher number of mutations was associated with a better chance of responding to checkpoint inhibitors1. This finding matches the results of other preliminary studies reported in recent years2. But the current study was the first to find improved survival in such a wide range of cancers, and in a population of people who had received a variety of previous treatments, says Morris.
The data also showed that the number of mutations that predicted a good response to immunotherapy varied from one type of cancer to another. This means that researchers would need to set a mutation threshold for each cancer type if they want to use this approach in clinical tests.
This is not an insurmountable challenge, but it could add to the complexity of an already complicated test. Counting up mutations in tumour genomes entails sequencing either the whole genome or just sections of it. Different DNA sequencing methods and algorithms for interpreting the resulting data can yield conflicting results. And it’s not clear whether all mutations should be valued equally, or whether some are more likely to prompt an immune response than others.
Despite these questions, some pharmaceutical companies have begun to include tests that measure the number of mutations in a tumour in clinical trials of their immunotherapies, says Shibutani. The results have been mixed, however. One company testing the approach — Bristol-Myers Squibb of New York City — found no survival advantage in selecting people using its mutation test.
But this could be due to a number of factors, such as where the company drew the line between high and low numbers of mutations, says Shibutani.
And Morris notes that survival results may have been affected by participants who switched from the control group to the immunotherapy arm of the trial. Checkpoint inhibitors were more likely to halt tumour growth in those with a higher number of mutations than in those with fewer mutations, he notes, even if there was no demonstrated difference in a person's survival rate.
Ultimately, Shibutani suspects that it will take a combination of characteristics, including mutation counts and information about the concentration of a given protein to select patients for immunotherapy. “This is the exact opposite of ‘keep it simple, stupid’,” he says. “Naturally, there’s more work to be done."