New Techniques to Predict Response to and Select Patients for Immunotherapy

Liam Davenport


November 14, 2017

Clinicians reading only the headlines from oncology meetings of late might be forgiven for thinking that all of their patients with solid tumors should be given immunotherapy in one form or another.

After all, numerous trials presented at the European Society for Medical Oncology (ESMO) 2017 Congress set out the impressive strides that have been taken with checkpoint inhibitors and other immune-targeted drugs over the past 5 years. Not only that, but they also established that anti-programmed death (PD)-1 and anti-programmed death ligand 1 (PD-L1) drugs are effective in a whole range of solid tumors, and are even moving from the palliative, stage IV setting to locally advanced stage III disease.

For example, Medscape Medical News reported data from the PACIFIC trial showing that the PD-L1 inhibitor durvalumab (Imfinzi™, AstraZeneca) could be the new standard of care in patients with stage III, locally advanced non-small cell lung cancer in whom chemoradiotherapy has failed.

These trials also indicated that immunotherapies were typically effective across most, if not all, of the prespecified subgroups, including when stratifying by tumor antigen expression.

Yet it is clear from previous studies that the drugs do not work in all patients, and may even be harmful in a small proportion who progress rapidly. The next question therefore becomes: Can we identify those patients who will do well and target therapy at them, while singling out those who will not benefit, so that they can be spared any toxicities?

The French Connection

The answer may come from a southern suburb of Paris, where one of the leading oncology hospitals in Europe has a department dedicated to giving patients access to the latest drugs and accelerating the development of novel treatments.

Under the leadership of Jean-Charles Soria, MD, PhD, a professor of medicine and medical oncology at the Département d'Innovation Thérapeutique et d'Essais Précoces at the Gustave Roussy Cancer Campus in Villejuif, Paris, has worked with the pharmaceutical industry and biotech firms to not only dramatically increase the number of patients receiving innovative treatments but also rapidly expand its research program.

Focusing on phase 1 and 2 trials, the team has studied a number of potential compounds at various stages, including first-in-man, dose escalation, and cohort extension studies, to evaluate not only the molecules themselves but also immune modulators and epigenetic and metabolic factors that could affect outcomes.

Among other avenues, the team have studied immunotherapies in a number of cancers. The studies include the ongoing MOSCATO and MATCH-R trials, which are looking at molecular screening to optimize cancer treatment and the molecular mechanisms of acquired resistance to targeted therapies, respectively.

Building on the comprehensive patient and clinical data they have on tap from these two studies, the team at Gustave Roussy have conducted a series of research projects on the use of immunotherapies. This has culminated in several posters presented at the ESMO 2017 Congress, and a presentation by Aurélien Marabelle, MD, PhD, clinical director of the Cancer Immunotherapy Program at Gustave Roussy, during a session looking at the future of targeted therapies in oncology.

One of the poster presenters, Loic Verlingue, MD, PhD candidate, told Medscape that the overall aim is to improve the use of immunotherapy for patients whom they see in their clinic—"to have tools to say that this patient won't benefit from immunotherapy, or this patient will highly benefit from immunotherapy," he said.

"In Gustave Roussy, we have shown that there are detrimental effects from immunotherapy for a small portion of patients: 10%, for example, who have an increase in tumor growth under immunotherapy."

It's very important ... to select this treatment for the patient who we really believe will respond.

Dr Verlingue continued, "It's very important today to really select your patient for immunotherapy—not to harm the patient, not to give an ineffective treatment, but to select this treatment for the patient whom we really believe will respond."

However, that is currently easier said than done, as the advent of immunotherapy has turned our preconceptions of both treatment and treatment outcomes in oncology on their head in recent years. It is therefore necessary to understand those changes before being able to identify which patients will respond best to therapy.

A New Paradigm

Tackling this puzzle head-on, Dr Marabelle began his presentation by noting that we are in a new paradigm in oncology, in which "we design drugs to target the immune system rather than designing drugs to destroy the cancer cells," with the aim of helping "patients to get rid of their disease with their own immune system."

More than 30 different cancer types have been shown to be sensitive to PD-1 or PD-L1 drugs.

Pointing out that there are now more than 30 different cancer types that have been shown to be sensitive to PD-1 or PD-L1 drugs, he said that it makes sense to target the immune system.

These changes have also brought with them changes to the shape of survival curves for treatments, with benefits in progression-free survival (PFS) now translating into improvements in overall survival. However, the challenge is that whereas the relationship between PFS and overall survival was relatively straightforward with tumor-targeted therapies, "with immune targeted therapies, we have clearly different subsets of populations."

Dr Marabelle said that there are patients who do not respond to immunotherapy, "sometimes maybe even having hyperprogressive disease under immunotherapy," whereas others have long-term stable disease and yet more have durable remissions with the drugs.

"This is, I think, the explanation for the shape of the survival curves that we see now," he said.

Pointing to a study by Bellmunt and colleagues[1] of pembrolizumab (Keytruda®, Merck) as a second-line therapy in advanced urothelial cancer, Dr Marabelle said that the survival curves show that chemotherapy performed better than immunotherapy at the beginning of treatment, "and we can explain that by the fact that maybe we have rapidly progressive disease for some patients on anti–PD-1 and anti-PD-L1 [agents]."

Subsequently, the benefits in terms of survival with the drugs become apparent "thanks to the stable disease, and then we have the clear benefits in long-term overall survival for the patients with partial and complete responses."

He emphasized that it is important to be able to identify patients who will benefit from treatment and to avoid exposing patients unnecessarily to toxicities when there is little chance of success. This can be achieved through the identification of predictive biomarkers, which, in contrast to prognostic factors, are related to the patient and the disease and not to the treatment.[2]

However, Dr Marabelle believes that it may not be so easy to find predictive biomarkers for immunotherapies in oncology.

"When you talk with your colleagues in rheumatology, they say, 'Well, we've been working with immunotherapies and biologics for more than 20 years now, and we've not been able over the past 20 years to identify predictive biomarkers,'" he said. "I hope we will maybe be better at that."

Can a Durable Response Be Predicted

To get a better understanding of prolonged responses to therapy in oncology, another member of the Gustave Roussy team, Marie-Léa Gauci, MD, presented a study at the ESMO 2017 Congress drawing on data stored at their institute.[3]

Dr Gauci and colleagues identified 262 patients with 19 cancer types who were given anti–PD-L1 therapy between 2011 and 2017 and who had been evaluated using RECIST 1.1, or the immune-related response criteria. They found a highly significant relationship between the degree of response to immunotherapy and median overall survival (P < .001).

Specifically, patients who progressed during treatment had a median overall survival of 6 months, compared with 15 months in those with stable disease and 53 months in those who had a partial response. Median overall survival was not reached in those with a complete response to immunotherapy.

The team concluded that a complete response "might be a good short-term surrogate marker for overall survival benefits in immunotherapy trials" and called for clinical trials aiming at achieving a complete response in partial responders.

Talking to Medscape, Dr Gauci said that "it's important to put a patient in complete response to have a longer survival, because there were no deaths in this group of patients."

Singling Out Predictive Biomarkers

Returning to the quest to identify predictive markers, Dr Marabelle noted that the biomarkers that have been examined so far in oncology for anti–PD-L1 drugs can be divided into tumor markers and host markers.

The former include PD-L1 expression, CD8 levels, mutational load, mismatch repair deficiency (dMMR), and transcriptomics, whereas the latter comprise lactate dehydrogenase (LDH) levels, eosinophil and lymphocyte counts, and the microbiome.

Taking LDH as an example, he highlighted research by Weide and colleagues[4] in which patients with melanoma were stratified by LDH levels. This showed that patients with a ≤ 2.5-fold elevation in LDH level had a significantly better likelihood of survival with pembrolizumab than those with a greater increase.

However, undermining its utility as a predictive marker, Dercle and colleagues[5] demonstrated that baseline LDH levels are not related to tumor burden in melanoma. A similar pattern was seem with lymphopenia, with a relative lymphocyte count ≥ 17.5% linked to better survival[4]—but again, being unrelated to tumor burden.[6]

Dr Marabelle noted that PD-L1 expression itself offers some signals as to who will respond to immunotherapy, a phenomenon that has been seen "since the very beginning with these drugs."

However, data also show that expression does not correlate well with patient outcomes, suggesting that its expression in the tumor microenvironment was simply related to the activity of the drug, and not its effect on the tumor itself.

There have been studies, however, indicating that responses to immunotherapy are different between microsatellite-stable and microsatellite-instable (MSI) tumors.

The more mutations you have accumulating in the tumors, the more chances you have of generating proteins that could be recognized by the immune system as abnormal.

"You can explain that by saying that the more mutations you have accumulating in the tumors, the more chances you have of generating proteins that could be recognized by the immune system as abnormal," he said.

However, Dr Marabelle also said that recent research has indicated that "not all MSIs are equal."

For example, a study presented at the ESMO 2017 Congress by El Dakdouki and colleagues,[7] also from the Gustave Roussy group, looked at the impact of MSIs on survival and response to anti–PD-L1 therapy, retrospectively studying 43 patients with metastatic dMMR treated at their institute between 2014 and 2017.

The most common primary tumor was gastrointestinal, seen in 51% of patients; a further 26% of patients had gynecologic cancers, whereas genitourinary and central nervous system tumors accounted for 7% each. Lynch syndrome, which is an inherited form of dMMR, was detected in 28% of the patients.

Over a median follow-up of 5.6 months, the median overall survival was not reached, and the median PFS was 11.1 months.

When the researchers looked at specific subgroups, they found that patients with colorectal cancer had a significantly longer PFS than those with other primary tumors (P = .025) and those with Lynch syndrome had a significantly better PFS than those with sporadic MSIs (P = .028).

However, it was not possible to identify similar differences when looking at objective response rates, which echoes the difficulties previous studies have had in relating tumor-related factors, such as mutational status and tumor burden, to complete or partial response rates.

Signs of Hope

Nevertheless, there are indications that predictive factors could point to better overall survival. For example, Dr Gauci and colleagues found that, across tumor types, the neutrophil/lymphocyte ratio (NLR) was significantly associated with median overall survival.[3]

Patients with an NLR > 4 had a median overall survival of 9 months, compared with 19 months for those with an NLR < 4 (P = .0003). Moreover, patients with a score of 3 on the Royal Marsden Hospital prognostic score, which combines albumin and LDH levels with the number of tumor sites, had a particularly poor outcome, with a median PFS of just 1 month.

The strongest signal for a marker of outcome to come out of the Gustave Roussy Cancer Campus so far, however, came from a study led by Dr Verlingue[8] that was presented at the ESMO 2017 Congress.

He and his colleagues assessed whether real-time quantitative and functional assessment of immune infiltrate with RNA sequencing (RNAseq) could be used to determine the degree of clinical benefit in patients treated with immune checkpoint inhibitors.

They studied 1297 fresh-frozen biopsies from the MOSCATO and MATCH-R trials obtained in patients with a range of metastatic tumors, identifying 67 patients for whom RNAseq had been performed on biopsies obtained a mean of 41 days before treatment was started.

Dr Verlingue explained, "We analyzed RNAseq with different methods, different pipelines of analyses, and different enrichment score methods, and we've taken accessible immune signatures that any practitioner, any medic, can access in the literature and tried to evaluate the value of these signatures in relation to PFS and immunotherapy."

In all, they used 154 signatures related to immunity, alongside 50 hallmark signatures from the Molecular Signatures Database and 20 randomly selected signatures.

The results of the RNAseq analyses were available within a median of 42 days; the researchers believe that this underlines that the technique is "feasible in a clinically relevant" time frame.

RNAseq is the tool.

"What we found is that most of the signatures that are related to immune cell types and accessible in the literature are highly correlated with PFS and immunotherapy," Dr Verlingue told Medscape.

Specifically, they found that on the Fisher exact test, immune-related signals were associated with PFS at an odds ratio of 16.37 (P = 4.18E-16).

In comparison, the random and hallmark signatures were not correlated with PFS. The authors now aim to test their method in routine care, using an independent cohort.

Are We There Yet?

With all of these interlinked projects at Gustave Roussy beginning to bear fruit, does Dr Verlingue believe that we are at the point of being able to identify those patients who will benefit from immunotherapy, and those who will not?

"That's a big question, and we working hard on that," he said. "In fact, I'm quite confident in RNAseq, because it is the tool that can provide you with information about the tumor cell itself and about the microenvironment."

Dr Verlingue added, "That's really, really important to have a good predictive value for this treatment."


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: