Cancer Vaccines: Can They Improve Survival?

Because of their minimal toxicity, there has been long-standing interest in the potential benefit of therapeutic vaccines in the setting of metastatic cancer, especially melanoma.¹ Unfortunately, even if a vaccine can induce or enhance a response to one or more tumor-associated antigens (TAAs), most studies have failed to document a clinical benefit.² Notable failures have included allogeneic cell lines in melanoma,³ gp100 in melanoma,⁴ and anti-idiotype vaccines in lymphoma.⁵ However, potential vaccines vary in terms of their antigen sources, adjuvants, and routes and schedules of administration; so, one should not conclude that all therapeutic vaccines are ineffective.^2,6 The theoretical limitations of allogeneic, single-antigen, and oligo-antigen vaccines have been noted, and there has been increasing focus on autologous TAAs.^7,8 There is also the promising potential for combining vaccines with anticheckpoint monoclonal antibodies.⁹

However, what is the best endpoint to document clinical efficacy in trials of therapeutic vaccines and related immunotherapies? And a related question, can cancer vaccines prolong survival? In 1996, the author authored an article entitled “Why event-free survival is better than tumor response or other measures of survival as an endpoint in cancer trials.”¹⁰ The term event-free survival (EFS) from that era is synonymous with what is known today as progression-free survival (PFS), with endpoints defined by the date of death or the date of disease progression, whichever comes first. However, overall survival (OS), defined by the date of death from any cause, is usually considered the most meaningful endpoint and has continued to be the gold standard for efficacy in clinical trials.¹¹ In patients with metastatic cancer, therapeutic anticancer vaccines are rarely associated with objective antitumor responses; so, many investigators have focused on PFS as an endpoint for clinical trials. While PFS often is a surrogate for OS in cancer chemotherapy trials, it is not clear that PFS is a surrogate for OS when the mechanism of action of an immunotherapy product is long-term memory in the adaptive immune system.

Recent publications, related to four products that have been tested in prostate cancer and/or melanoma, were reviewed. Sipuleucel-T is an autologous dendritic cell (DC) product for prostate cancer that expresses a recombinant fusion protein that produces prostatic acid phosphatase (PAP) as a TAA and granulocyte–macrophage colony-stimulating factor (GM-CSF) as an adjuvant.¹² It is administered intravenously (i.v.) every 2 weeks for three doses. Prostvac-vf, previously known as prostate-specific antigen (PSA)-Tricom, is an antiprostate cancer vaccine that consists of two recombinant viral vectors encoding PSA as the TAA and three costimulatory molecules, B7.1, ICAM-1, and LFA-3.¹³ It is administered subcutaneously (s.c.) on days 1, 14, 28, 56, 84, 112, and 140, and 100 μg GM-CSF is injected nearby as an adjuvant on the day of each injection and for the following 3 days. Eltrapuldencel-T (formerly referred to as melapuldencel-T) is a patient-specific antimelanoma vaccine consisting of autologous DCs loaded with TAAs from irradiated autologous tumor cells (TCs) derived from a short-term autologous TC line.¹⁴ It is administered on weeks 1,2,3,8,12,16,20, and 24, and each dose of cells is suspended in 500 μg GM-CSF for each s.c. injection.

Ipilimumab is not considered a vaccine, but its primary mechanism of action is believed to be enhanced recognition of TAAs. It is a human monoclonal antibody extensively tested in metastatic melanoma that blocks the anticytotoxic T lymphocyte antigen-4 (CTLA-4) molecule that inhibits DC presentation of TAAs to host T cells that lead to memory and cytotoxic T cells.¹⁵ It is administered i.v. every 3 weeks for four doses. A recent publication showed that patients treated with ipilimumab, who experienced long-term survival, each had uniquely mutated antigens that share immunogenic tetrapeptide sequences that can elicit responses from T cells.¹⁶

The results of trials with the above products suggest that enhancing recognition of TAAs can be associated with improved OS, despite rare objective responses and no improvement in PFS. Given the perceived mechanism of action for these vaccine products, this lack of correlation between ORR, PFS, and OS should not be surprising. In each case, immune recognition of TAAs has to be induced, or enhanced, which takes time. Furthermore, in some, if not all instances, it may only be a subset of cells that express the most relevant TAAs. For instance, in the case of eltrapuldencel-T, the source of TAAs is self-renewing proliferating cells that share many characteristics with tumor stem cells.⁷ If such early tumor-initiating cells are being targeted preferentially, and many of their TAAs are not expressed on more differentiated TCs, then one would not expect any immediate effects on existing tumor lesions, including those that are not detectable, and these sites might increase after a few more proliferations of soon-to-be terminally differentiated TCs. However, if one is inducing or stimulating long-term memory T cells and an effective cellular and humoral response against TAAs on a subset of tumor stem cells and/or early progenitors, then one might get a long-term survival benefit without any suggestion of benefit based on ORR or PFS.

So, what about the specific immune responses that are being induced? It is known from numerous reports with various peptide vaccines that specific immune responses can be induced or enhanced in cancer patients by vaccination. Autologous vaccine products present a challenge in terms of measurement of immune response because of TAA heterogeneity among patients. Although there may be some common expression of certain TAAs, it is now known that every patient harbors unique neoantigens that result from their mutations. Tumors from melanoma patients express tens to hundreds of thousands of mutated proteins, many of which are nonsynonymous and therefore potentially antigenic.^39,40 In elegant work from the United States National Cancer Institute (NCI), complete exomic analyses and tumor-infiltrating lymphocytes have been used to prove that such unique antigens can be targeted by the patient's immune system and induce durable responses when billions of cells are infused after ablation of the host immune cells.^41,42 Recently, similar exomic and neoepitope analyses were used to show that the patients who derive long-term survival benefit from ipilimumab are also recognizing epitopes on uniquely mutated antigens.¹⁶

It should be noted that once upon a time PFS actually meant what it said, that is, it was defined only by the date of disease progression, and if patients died of another cause, they were censored from survival analysis as of that date. However, some trials, especially bone marrow transplant trials, came under attack for focusing on PFS rather than OS because so many patients were dying from the toxicity of therapy and being excluded in PFS analyses. Somewhere along the line, thought leaders and regulatory authorities decided to change the definition of PFS to include death,⁴³ thereby making it synonymous with the former EFS definition. If one is comparing clinical trials over time, it is important to be sure what is meant by PFS in any specific publication because, in older trials, it meant time to disease progression, whereas in the past decade, it increasingly has meant death or progressive disease.

In the 1996 article cited earlier,¹⁰ many of the problems associated with ORR were enumerated, most of which persist today, and an argument was made for the increased use of EFS (PFS today) rather than ORR or OS as the key endpoint, not only for phase III trials but also in phase II trials involving biologics. The use of OS as the key endpoint was considered less desirable because of the length of time needed in follow-up before appropriate statistical analysis could be performed, and because increased numbers of treatment options could confound cause and effect for any specific therapy or combinations of therapies. When that article was written, the focus was on the challenges of assessing emerging therapeutic vaccines by other than the traditional chemotherapy models. It was noted that this might be especially important for biological therapies, including immunotherapies, and a specific example was provided, namely the analysis of metastatic melanoma patients rendered free of disease by surgery. The evidence cited in the current article suggests that OS is a better endpoint than PFS for evaluation of such vaccines.

Although ORR and PFS are often considered as surrogates for OS, whether they can serve as surrogates for OS depends on the specific cancer type, the specific agents being tested, and the absence of other effective therapies. However, PFS has become very popular because when disease progression is the endpoint, a trial can be concluded much faster than if OS is the endpoint. Pitfalls associated with the use of PFS as an endpoint have been pointed out in other publications.^43,44 The same issues that confound PFS also confound ORR. Measurement of tumor size is imprecise; so, definition of time of date of disease progression and ORR are imprecise, and thresholds for defining progression are arbitrary. Second, if one wishes to use PFS as the endpoint, then it is essential that repeated measurements of the tumor be performed at frequent prescribed intervals to avoid various bias pitfalls that affect the timing of declaration of disease progression. This results in excessive testing that is costly and is associated with increased radiation exposure. Third, contrary to the original hypothesis for why PFS would be better than OS for immunotherapies, it turns out that ORR and PFS are not reliable surrogate endpoints for many emerging cancer immunotherapies because of their mechanism of action, and yet these agents are associated with improved long-term OS. Given the current understanding of the immune system and tumor immunology and memory in the adaptive immune system, for certain products, OS may be the only relevant efficacy endpoint and therefore remains the gold standard.