Pembrolizumab near the end of life in patients with metastatic pancreatic cancer: a multi-site consecutive series to examine survival and patient treatment burden

Abstract

Background

Pembrolizumab confers minimal benefit to most patients with pancreas cancer. We explored survival and patient treatment burden (for example, death within 14 days of therapy) in a subgroup who had early access to pembrolizumab .

Methods

This multisite study examined consecutive pancreas cancer patients, who received pembrolizumab from 2004 through 2022. Median overall survival of > 4 months was to be deemed favorable. Patient treatment burden and medical record quotations are presented descriptively.

Results

Forty-one patients (median age 66 years; range 36, 84) are included. Fifteen (37%) had dMMR, MSI-H, TMB-H, or Lynch syndrome; and 23 (56%) received concurrent therapy. The median overall survival was 7.2 months (95% confidence interval (CI): 5.2, 12.7 months); 29 were deceased at the time of reporting. Patients with dMMR, MSI-H, TMB-H, or Lynch syndrome had a lower risk of death: hazard ratio (HR): 0.29 (95% CI: 0.12, 0.72); p = 0.008. Medical record phrases (“brilliant response”) aligned with the above. One patient died within 14 days of therapy, and one was in an intensive care unit within 30 days of death. Fifteen patients enrolled in hospice; four of these died < 3 days later.

Conclusions

These unexpectedly favorable findings underscore the need for healthcare providers—including palliative care providers—to knowledgeably guide patients about cancer therapy even near the end of life.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00262-023-03397-4.

Introduction

Immunotherapy, or prescribing drugs to rally a patient’s own immune system to treat a cancer, has shown widespread efficacy in oncology [1–4]. Programmed cell death-ligand 1 (PD-L1) or programmed cell death protein-1 (PD-1) inhibitors are now the standard of care for multiple malignancies, including melanoma, lung cancer, kidney cancer, and refractory Hodgkin’s lymphoma [1–4].

However, non-endocrine metastatic pancreatic cancer is an exception. O’Reilly and others conducted a phase 2 trial that included 33 pancreas cancer patients who received monotherapy with the PD-L1 inhibitor durvalumab and reported a 0% response rate and a 6-month survival rate of only 35% [5]. This trial, as well as others, illustrates that pancreatic cancer is generally a non-immunogenic malignancy and that, for the most part, it is unresponsive to immunotherapy-based interventions [6]. Indeed, only 1–2% of patients with this malignancy carry genomic or somatic evidence of genetic instability, including high microsatellite instability (MSI-H), deficient mismatch repair (dMMR), or tumor mutational burden > 10 mutations per megabase (TMB-H), circumstances that suggest a patient could benefit from immunotherapy [7–9]. Based on the extant published literature, the message appears clear that in most patients with this cancer, PD-L1 inhibitors confer little, if any, benefit [10–12].

Interestingly, in 2014, the Food and Drug Administration (FDA) approved pembrolizumab as a second-line treatment for patients with melanoma. At that time, because this agent’s scope of use was only just beginning to be assessed in clinical trials, its manufacturer enabled clinicians to acquire off-label pembrolizumab for patients with relative ease and at no cost. However, although pembrolizumab was readily available at that time, one might question whether its use was truly of value, particularly as the patient treatment burden, or the negative aspects of cancer therapy, such as early death after its receipt had not been evaluated [13–16]. Indeed, the relatively easy access of this new agent back then raises questions relevant to future patients, who might be candidates for other novel drugs. In patients with a malignancy, which in retrospect is generally regarded as unresponsive to pembrolizumab, did these earlier treated patients acquire any benefit? Could patient treatment burden from pembrolizumab — or, specifically, death within 14 days of cancer treatment, death within 30 days of intensive care unit admission, and hospice enrollment within only 3 days of death — have overridden even modest life-prolonging benefits of pembrolizumab?

This descriptive study sought to begin to explore these questions with the goal of providing guidance to patients and healthcare providers as newer cancer drugs become available and as decisions are made about whether to acquire and prescribe such new agents in the absence of data which show likelihood of benefit. Such therapeutic decisions are particularly important to be aware of if they result in what appears to be minimal benefit and notable patient treatment burden.

Methods

Overview

The Mayo Clinic Institutional Review Board (IRB) approved the study protocol (22–005164). This multi-site study entailed medical record review from January 2004, when pembrolizumab first became clinically available, until July 2022.

Study design and acquisition and review of patients’ medical records

This is a retrospective mixed methods study that integrated quantitative as well as qualitative (direct quotations) data from the medical record to generate conclusions. A study protocol as well as data sheets, which together outlined the goals of the study, was written prior to the review of medical records.

The Mayo Clinic electronic medical record, which serves campuses in Minnesota, Arizona, Florida, Wisconsin, and Iowa, was used to identify patients who met the following criteria. All patients had to have received pembrolizumab and had to have had metastatic, non-endocrine pancreas cancer at the time of pembrolizumab therapy. All medical records underwent a rapid review by hand to assess whether these eligibility criteria were met. Each medical record that met criteria was then reviewed in detail for clinical demographics and outcomes as well as for clinical laboratory evidence of deficient mismatch repair, microsatellite instability (for example, instability of 2 or more loci), combined positivity score > / = 10, TMB-H (for example, tumor mutational burden > 10 mutations per megabase), or Lynch syndrome with such evidence defined by Olave and Graham [17]. Data relevant to the questions posed above were abstracted by one investigator (MHS); a second investigator undertook spot checks for accuracy (AJ).

Analyses

Data are presented descriptively with median values, percentages, ranges, hazard ratios (HR’s), and 95% confidence intervals (CI’s). Censoring was used to report survival data, and Kaplan–Meier curves were used to summarize these data. Cox regression models were constructed to generate exploratory comparisons between groups and to report HR’s. A priori, based on the NAPOLI-1 trial and a study from Smith and others, the study team decided that a median overall survival beyond 4 months would indicate that the administered therapy was potentially contributing to improved survival [18, 19]. Direct quotations from the medical record were used to provide clinical insight as to whether patients truly appeared or did not appear to be benefiting from cancer therapy.

Results

Baseline demographics

A total of 41 patients met the eligibility criteria (Fig. 1). The median age at pembrolizumab initiation was 66 years (range 36, 84). Most patients were women (Table 1). Twenty-three (56%) were receiving concurrent cancer therapy in addition to pembrolizumab.

Fig. 1.

The consort diagram shows how the final sample of 41 patients was arrived upon

Table 1.

Baseline Demographics; n = 41*

Age in years at pembrolizumab administration, median (range)	66 (36, 84)
Sex
Male Female	16 (39) 25 (61)
Cancer status at diagnosis
Resectable Locally advanced Unresectable/metastatic	11 (27) 6 (15) 24 (58)
Number of lines of therapy prior to pembrolizumab
0	9 (22)
1	11 (27)
2	12 (29)
3	7 (17)
4 or more	2 (5)
Was other therapy prescribed with pembrolizumab?
Yes	23 (56)
No	18 (44)
Indicators that pembrolizumab could be helpful (deficient mismatch repair, microsatellite instability, combined positivity score > / = 10, or Lynch syndrome)
Yes	15 (37)
No	26 (63)

^*Numbers in parentheses denote percentages unless otherwise denoted

Survival

The median overall survival of the group as a whole was 7.2 months (95% CI: 5.2, 12.7 months) with 29 deceased at reporting (Fig. 2).

Fig. 2.

With 29 deceased patients at the time of reporting, the median overall survival was 7.2 months (95% confidence interval: 5.2, 12.7 months)

When the median overall survival was assessed based on tumor or host indicators of potential benefit from pembrolizumab—specifically, dMMR, MSI-H, TMB-H, or Lynch syndrome — those patients with no indicators manifested a median overall survival of 7 months (95% CI: 1.5, 8.0 months), whereas those with such an indicator manifested an median overall survival that had not yet been reached: HR: 0.29 (95% CI: 0.12, 0.72); p = 0.008, in favor of longer overall survival with an indicator (Fig. 3). Although numbers are small, further exploratory analyses that examined not only dMMR, MSI-H, TMB-H, or Lynch syndrome but also whether patients were prescribed another drug with pembrolizumab did not suggest that the addition of another drug impacted survival.

Fig. 3.

Patients with no indicators of likely benefit from pembrolizumab (n = 26) manifested an median overall survival of 7 months (95% CI: 1.5, 8.0 months) (blue line), whereas those with such indicators (n = 15) had a median overall survival that had not yet been reached at the time of reporting (red line)

Burden of treatment for patients

Among the 29 deceased patients, only one died within 14 days of cancer therapy ( Supplementary Table 2). This patient manifested a progressive decline in activities of daily living and made the decision to enroll in hospice only 3 days before her death.

One patient was admitted to an intensive care unit within 30 days of death ( Supplementary Table 2). This patient was admitted to the hospital for nausea and vomiting after having received pembrolizumab 24 days prior. Low oxygen saturation levels and hypotension resulted in transfer to the intensive care unit, where the patient agreed to comfort measures and then died later that day.

Twenty-two patients (54%) visited an emergency department after receipt of pembrolizumab; the number of emergency department visits ranged from 1 to 14. Twenty-three patients (56%) were admitted to the hospital after receipt of pembrolizumab; the number of admissions ranged from 1 to 9.

Fifteen patients enrolled in hospice, and 14 died in hospice ( Supplementary Table 2). The median time to death after hospice enrollment was 2.2 weeks (range of 0, 10 weeks). Four patients died in hospice fewer than 3 days after enrollment. Of note, with respect to patient care burden in all patients, only one patient—a patient who died in hospice < 3 days after enrollment—had an indicator to suggest benefit from pembrolizumab.

Adverse events

The median number of administered doses of pembrolizumab was 3 (range 1, 45), and 11 patients (27%) experienced at least one related adverse event. Five suffered endocrine toxicity, four suffered skin toxicity, two had diarrhea, one had pneumonitis, and three others experienced poorly characterized side effects.

Qualitative medical record evidence of benefit or lack thereof

Selected quotations from the medical record illustrate lack of benefit from pembrolizumab, how benefits were sometimes not sustained, or how pembrolizumab was sometimes not tolerated (“CT imaging showing new and enlarging bilateral pulmonary metastasis…” and “Stopped… due to side effects…”). Similarly, as illustrated, pembrolizumab at times seemed to result in clear benefits (“tolerating pembrolizumab well”; “remission since 2020…”; and “brilliant response”). These quotations also show evidence of these different outcomes in the setting of indicators, such as MSI-high tumor status, and in the setting of multiagent therapy ( Supplementary Table 3).

Discussion

This study was undertaken to report on survival and patient treatment burden in pancreas cancer patients who had been prescribed pembrolizumab, an agent that is generally perceived as one that confers limited benefit for patients with this malignancy. Surprisingly, we observed that overall survival was more favorable than anticipated and that patient treatment burden appeared somewhat low. Furthermore, direct quotations from the medical record show that the administration of pembrolizumab appeared directly tied to these improved survival outcomes (for example, “brilliant response”). These findings indicate that, among highly select patients, the use of novel agents might be considered and that healthcare providers who specialize in palliative care should remain receptive to supporting these patients as they make thoughtful and informed end of life decisions that call for the administration of new antineoplastic drugs.

Two explanations might account for the relatively favorable overall survival. First, this favorable survival is likely the result of thoughtful patient sampling, as over a third of patients who received pembrolizumab had indicators to suggest potential benefit [8, 9]. In general, such indicators are found in only 1–2% of patients with pancreas cancer; the fact that a few hundred patients with this malignancy are seen annually at our institution and the fact that the final sample size for this study amounted to 41 patients speaks to such selective sampling. Although the conclusions from the current study cannot be widely applied to all patients with pancreas cancer and although they cannot be applied to all novel drugs, an argument can be made in support of considering the administration of novel agents to highly select groups of patients. Second, even in patients who did not have indicators to suggest that they would benefit from pembrolizumab, we observed a survival outcome that was better than anticipated. Importantly, 56% of patients in the current study received pembrolizumab in combination with other cancer agents. Typically, combinatorial antineoplastic therapy is applied when the possibility of synergy between agents seems possible or plausible. This issue of synergy between agents such as pembrolizumab and chemotherapy continues to be studied and might account for the favorable survival outcomes observed even in patients who did not, for example, have MSI-H tumors [20]. Our data underscore the importance of such ongoing investigation of synergy between checkpoint inhibitors and chemotherapy and might explain the relatively favorable survival among those patients who might have seemed less likely to derive benefit.

In addition, the patients in the current study also appeared to experience a relatively low patient treatment burden near the end of life. An earlier study focused on patients who had received an immune checkpoint inhibitor within 30 days of death and reported that 20% of these patients were admitted to the ICU within this period, 20% were admitted to the intensive care unit, and 25% died in an acute care setting (not in hospice) [21]. The patient treatment burden metrics reported in the current study are not nearly as negative as those reported earlier and, again, are likely reflective of careful patient selection. Indeed, these metrics of death within 14 days of cancer therapy, intensive care unit admission within 30 days of death, and hospice enrollment within fewer than 3 days of death have become indicative of the quality of cancer care that patients receive at the end of life and are important in defining the role of palliative care in cancer care [13–16]. The current study makes the point that carefully selected patients with aggressive malignancies can nonetheless receive cancer therapy and experience a seemingly low patient treatment burden — or, in effect, find cancer therapy seemingly both beneficial and tolerable.

Indeed, our study is somewhat unique in that it examined both survival and patient burden of treatment, not one or the other. If cancer therapy modestly prolongs life but causes extreme, intolerable suffering, such therapy might not be worth administering. Similarly, if a patient experiences an improvement in quality of life but at the cost of a markedly truncated survival, such a trade-off might not be acceptable to all. Our observation that pembrolizumab resulted in both favorable survival and favorable patient treatment burden again speaks to the importance of assessing both these outcomes in future research, especially in research that focuses on patients who are approaching the end of life.

This study has limitations. First, the patient population was a highly select group who received care at a large, quaternary medical center; who manifested heterogeneity with indicators for pembrolizumab efficacy (dMMR, MSH-I, TMB-H, or Lynch syndrome) and with an absence of such indicators; and who received not only single agent but also multiagent cancer therapy. These factors make generalization to other more specific populations challenging. Second, the retrospective design used here carries inherent limitations. Although we had been careful with data acquisition, it is possible that missing data—an inherent limitation of retrospective studies—might have led to an underestimation of patient treatment burden. Furthermore, such characteristics as patient performance score, which is a major determinant of whether to administer cancer therapy, were unavailable in the medical record and thus further interferes with our ability to provide definitive guidance on how to treat or palliate patients who are approaching the end of life. Lastly, the sample size is small, making it challenging to make broad conclusions. Nonetheless, to our knowledge, this study includes one of the largest populations of pancreas cancer patients who had indicators of potential benefit with pembrolizumab. Thus, these findings merit reporting.

In summary, this study examined pembrolizumab in select patients with non-endocrine metastatic pancreas cancer and observed relatively favorable survival outcomes and lower than anticipated patient treatment burden near the end of life. We conclude that healthcare providers—particularly those who specialize in palliative care — cannot be draconian in limiting the use of novel antineoplastic agents near the end of life and must work with patients and their families to make balanced decisions. With the emergence of newer, better-tolerated agents, healthcare providers will need to continually reassess the limits of when such therapy should be offered to patients or be strongly discouraged.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors contributed equally to this work.

Data availability

The data generated by this work are not available due to confidentiality.

Declarations

Conflict of interest

The authors have no relevant financial or non-financial interests to disclose.

Ethics approval

The Mayo Clinic Institutional Review Board (IRB) approved the study protocol (22–005164).

Consent to participate

Consent was not required for study, which integrated quantitative as well as qualitative (direct quotations) data from the medical record to generate conclusions.