Risk of hematologic toxicities with programmed cell death-1 inhibitors in cancer patients: a meta-analysis of current studies.

BACKGROUND: Programmed cell death-1 (PD-1) inhibitor-related hematologic toxicities are a category of rare but clinically serious and potentially life-threatening adverse events; however, little is known about their risks across different treatment regimens and tumor types. The objective of this study was to compare the incidences of PD-1 inhibitor-related hematologic toxicities among different therapeutic regimens and tumor types.
METHODS: Twenty-six original articles on PD-1 inhibitor trials were identified based on a PubMed search completed on September 26, 2017. The incidences of hematologic toxicities were collected.
RESULTS: A total of 26 studies containing 5,088 patients were included in the meta-analysis. PD-1 inhibitor monotherapy was associated with an increased risk of all-grade anemia in cancer patients (5%, 95% CI 4%-6%), particularly in patients with renal cell carcinoma (RCC) (8%, 95% CI 6%-12%), compared with all-grade thrombocytopenia (2%, 95% CI 1%-5%), leukopenia (2%, 95% CI 1%-3%), and neutropenia (1%, 95% CI 0-1%). However, low incidences of high-grade hematologic toxicities were observed in cancer patients treated with PD-1 inhibitor monotherapy. The use of PD-1 inhibitors in combination with ipilimumab, peptide vaccines, or chemotherapy had significantly higher risks than PD-1 inhibitor monotherapy for all-grade anemia (13%, 95% CI 5%-31%), thrombocytopenia (6%, 95% CI 2%-18%), leukopenia (5%, 95% CI 1%-35%), neutropenia (4%, 95% CI 1%-26%), and only high-grade thrombocytopenia (4%, 95% CI 1%-15%). In addition, all-grade and high-grade hematologic toxicities in chemotherapy and everolimus treatment arms were more frequent than in PD-1 inhibitor monotherapy arms.
CONCLUSION: The risks of PD-1 inhibitor-related hematologic toxicities were higher in RCC than in other cancers, and during combination therapy. These results may contribute toward enhancing awareness among clinicians about frequent clinical monitoring when managing PD-1 inhibitors.

Introduction

In recent years, programmed cell death-1 (PD-1) inhibitors have shown remarkable efficacy in the clinic, thus accelerating US Food and Drug Administration (FDA) approval of these agents for cancer therapy. Nivolumab and pembrolizumab are the human PD-1-blocking antibodies that have been approved for treatment of non-small-cell lung cancer (NSCLC),1–9 melanoma,10–13 renal cell carcinoma (RCC),14,15 urothelial carcinoma,16,17 and head and neck squamous cell carcinoma (HNSCC).18 Furthermore, anti-PD-1 antibodies in combination with other drugs have also shown significant effects in many refractory cancers.19–23

As a result of widespread prescribing of PD-1 inhibitors in the clinic, the profile of immune-related adverse events (irAEs) started to raise many concerns. These irAEs are characterized by T-cell response hyperactivation (overproduction of CD4+ T-helper-cell cytokines and abnormal migration of cytolytic CD8+ T cells) leading to normal tissue damage or organ system failure. Among them, hematologic toxicity during PD-1 inhibitor therapy is of particular interest owing to its causing potentially life-threatening events (severe hypoxia, infection, and bleeding) if not promptly recognized and adequately treated.

Although many clinical trials have reported hematologic toxicities associated with PD-1 inhibitors, these are a series of relatively rare adverse events, and hence the knowledge based on the individual cohort data from each trial is limited. Given the increasing number of published PD-1 inhibitor trials, a systematic assessment may provide important knowledge on these categories of rare but clinically significant and potentially serious irAEs. In this study, we conducted a meta-analysis to address the incidences of PD-1 inhibitor-related hematologic toxicities among cohorts with different treatment regimens and tumor types.

Methods

Search methods and study selection

The literature published in the English language up to September 26, 2017, that included the results of prospective trials of PD-1 inhibitor therapy in cancer patients using nivolumab or pembrolizumab, including monotherapy and combination therapy, was identified by a PubMed search and by examining the references of published trials, review articles, editorials, and other relevant articles. The following keywords or corresponding Medical Subject Heading terms were used: “nivolumab”, “pembrolizumab”, and “PD-1 inhibitor”. Articles that were published online “ahead of print” were included. Meeting abstracts without published full-text original articles were excluded. The search focused on trials of nivolumab and pembrolizumab because of their having being approved by the US FDA as well as being widely available on prescription in the clinic.

Data extraction

The data were extracted by one primary investigator (J-D Sui) and then reviewed independently by 2 secondary investigators (Y Wang and Y Wan). The following information was collected from the eligible articles: principal author’s name; year of issuance; type of treatments; type of tumors; trial phase; clinical trial information; number of enrolled patients; and number of events with all-grade and high-grade (grades 3–5) anemia, thrombocytopenia, leucopenia, and neutropenia. Hematologic toxicities in each trial were defined according to the Common Terminology Criteria for Adverse Events (CTCAE) of the National Cancer Institute.

Statistical analysis

Review Manager version 5.3 was used for statistical analysis. The data type “dichotomous” was selected to compare the incidences of hematologic toxicities between PD-1 inhibitor monotherapy and chemotherapy or everolimus monotherapy. A value of p<0.05 was considered statistically significant.

The data type “generic inverse variance” was selected for the outcome of incidence; if the p-value was close to neither 0 nor 1, and both n×p and n×(1 − p) were more than 5, p could be calculated by n/X, in which n and X refer to the number of patients with hematologic toxicities and the total number of patients in treatment arm, respectively. SE was calculated using the formula (p(1 − p)/n)1/2, in which p and n have the same meanings as above. On the other hand, if these conditions could not be met, p and SE were calculated by the following formulas: p = ln(odds) = ln[X/(n − X)], SE(p) = SE[ln(odds)] = [1/X + 1/(n − X)]1/2. Significantly, this is the method used for categorical data; thus, the calculations for ORs should be transformed using the following formula: Pf = OR/(1 + OR), lower limit (LL) of 95% CI = LLOR/(1 + LLOR), upper limit (UL) = ULOR/(1 + ULOR). In our study, the latter method was adopted owing to the low incidences of adverse effects.

Heterogeneity was evaluated by the Cochran chi-square test and the I2 test. Heterogeneity was considered statistically significant if p<0.05. If substantial heterogeneity was not observed, the pooled estimate was calculated based on the fixed effects model. If substantial heterogeneity existed, data were analyzed using a random effects model. I2<30% represented a slight level of heterogeneity, 30%–60% was moderate while I2>60% meant that the heterogeneity was high. Publication bias was assessed using funnel plots showing the relationship between OR and SE (log[OR]). Publication bias existed if the funnel plots lacked symmetry.

Results

Eligible studies and characteristics

Based on our search strategy, a total of 485 records were identified for screening. Exclusion criteria are presented in Figure 1. Accordingly, a total of 26 full-text articles were considered eligible for our analysis, including 9 phase III trials,1–3,7,8,11,13,17,18 5 phase II trials,9,10,15,16,23 2 phase I/II trials19,24 and 10 phase I trials.4–6,12,14,20–22,25,26 Of these included studies, 10 trials evaluated PD-1 inhibitor monotherapy (6 nivolumab5,6,9,12,14,24 and 4 pembrolizumab4,16,25,26), 10 trials evaluated PD-1 inhibitor monotherapy versus chemotherapy control (6 nivolumab vs chemotherapy1–3,11,13,18 and 4 pembrolizumab vs chemotherapy7,8,10,17), 1 trial evaluated nivolumab monotherapy versus everolimus,15 3 trials evaluated nivolumab/ipilimumab combinations,19,21,23 1 trial evaluated nivolumab/chemotherapy combination,22 and 1 trial evaluated nivolumab/peptide vaccine combination.20 Tumor types tested in these studies included NSCLC (n=11),1–9,21,22 melanoma (n=6),10–13,20,23 RCC (n=2),14,15 urothelial carcinoma (n=2),16,17 small-cell lung cancer (SCLC) (n=1),19 hepatocellular carcinoma (n=1),24 HNSCC (n=1),18 triple-negative breast cancer (n=1),26 and malignant pleural mesothelioma (n=1).25 The characteristics of all the included trials are summarized in Table 1.

Figure 1

Flow diagram of study inclusion.

Abbreviations: mAbs, monoclonal antibodies; PD-1, programmed cell death-1.

Table 1

Characteristics of all included studies

Study	Drug	Tumor type	Phase	No of treated patientsa	Median age (years)	Clinical trial information
Alley et al (2017)25	Pembrolizumab	Malignant pleural mesothelioma	1b	25	65	NCT02054806
Antonia et al (2016)19	Nivolumab and ipilimumab	SCLC	1/2	115	63	NCT01928394
Balar et al (2017)16	Pembrolizumab	Urothelial carcinoma	2	370	74	NCT02335424
Bellmunt et al (2017)17	Pembrolizumab	Urothelial carcinoma	3	266	66	NCT02256436
Borghaei et al (2015)1	Nivolumab	NSCLC	3	287	62	NCT01673867
Brahmer et al (2015)2	Nivolumab	NSCLC	3	131	63	NCT01642004
Carbone et al (2017)3	Nivolumab	NSCLC	3	267	64	NCT02041533
El-Khoueiry et al (2017)24	Nivolumab	Hepatocellular carcinoma	1/2	48	62	NCT01658878
Ferris et al (2016)18	Nivolumab	HNSCC	3	236	60	NCT02105636
Garon et al (2015)4	Pembrolizumab	NSCLC	1	495	64	NCT01295827
Gettinger et al (2015)6	Nivolumab	NSCLC	1	129	65	NCT00730639
Gettinger et al (2016)5	Nivolumab	NSCLC	1	52	67	NCT01454102
Gibney et al (2015)20	Nivolumab and peptide vaccine	Melanoma	1	33	47	–
Hellmann et al (2017)21	Nivolumab and ipilimumab	NSCLC	1	77	65	NCT01454102
Herbst et al (2016)7	Pembrolizumab	NSCLC	3	682	63	NCT01905657
McDermott et al (2015)14	Nivolumab	RCC	1	34	58	NCT0730639
Motzer et al (2015)15	Nivolumab	RCC	2	406	62	NCT01668784
Nanda et al (2016)26	Pembrolizumab	Triple-negative breast cancer	1b	32	50.5	NCT01848834
Reck et al (2016)8	Pembrolizumab	NSCLC	3	154	64.5	NCT02142738
Ribas et al (2015)10	Pembrolizumab	Melanoma	2	357	62	NCT01704287
Rizvi et al (2015)9	Nivolumab	NSCLC	2	117	65	NCT01721759
Rizvi et al (2016)22	Nivolumab and chemotherapy	NSCLC	1	56	64	NCT01454102
Robert et al (2015)11	Nivolumab	Melanoma	3	206	65	NCT01721772
Topalian et al (2014)12	Nivolumab	Melanoma	1	107	61	NCT00730639
Weber et al (2015)13	Nivolumab	Melanoma	3	268	59	NCT01721746
Weber et al (2016)23	Nivolumab and ipilimumab	Melanoma	2	138	62	NCT01783938

Note:

Includes the number of patients treated in PD-1 inhibitor arms but does not include patients treated in the control arms without PD-1 inhibitors.

Abbreviations: HNSCC, head and neck squamous cell carcinoma; NSCLC, non-small cell lung cancer; PD-1, programmed cell death-1; RCC, renal cell carcinoma; SCLC, small cell lung cancer.

Overall incidence of hematologic toxicities

For the incidences of hematologic toxicities, all PD-1 inhibitor arms were considered. The summary incidences of all-grade anemia, thrombocytopenia, leukopenia, and neutropenia were 5% (95% CI 4%–6%), 2% (95% CI 1%–5%), 2% (95% CI 1%–3%), and 1% (95% CI 0–1%), respectively (Figure 2). The test for heterogeneity was significant for anemia and thrombocytopenia, so the random effects model was used to minimize the influence of heterogeneity. Regarding high-grade toxicities, there were relatively low incidences of anemia, thrombocytopenia, leukopenia, and neutropenia (Figure S1).

Figure 2

Forest plots of incidences of all-grade hematologic toxicities during PD-1 inhibitor monotherapy: (A) anemia, (B) thrombocytopenia, (C) leukopenia, and (D) neutropenia.

Abbreviations: PD-1, programmed cell death-1; IV, inverse variance.

Incidence of hematologic toxicities in PD-1 inhibitor monotherapy versus combination therapy

The incidences of hematologic toxicities during PD-1 inhibitor monotherapy versus combination therapy were compared in the studies of NSCLC, melanoma, and SCLC because the included studies of combination therapy were conducted in patients with these 3 tumor types.19–23 Combination regimens included nivolumab and ipilimumab given concurrently or sequentially,19,21,23 and nivolumab plus peptide vaccines20 or chemotherapy.22 The incidences were significantly higher in the combination therapy group compared with the monotherapy group for all-grade toxicities (13% vs 5% for anemia; 6% vs 2% for thrombocytopenia; 5% vs 2% for leukopenia; 4% vs 1% for neutropenia) (Figures 2 and 3) and high-grade thrombocytopenia (4% vs 1%) (Figures 4 and S1). However, there was no significant difference between the monotherapy group and combination therapy group for high-grade anemia, leukopenia, and neutropenia (Figures 4 and S1).

Figure 3

Forest plots of incidences of all-grade hematologic toxicities during PD-1 inhibitor combination therapy: (A) anemia, (B) thrombocytopenia, (C) leukopenia, and (D) neutropenia.

Abbreviations: PD-1, programmed cell death-1; IV, inverse variance.

Figure 4

Forest plots of incidences of high-grade hematologic toxicities during PD-1 inhibitor combination therapy: (A) anemia, (B) thrombocytopenia, (C) leukopenia, and (D) neutropenia.

Abbreviations: PD-1, programmed cell death-1; IV, inverse variance.

Incidence of hematologic toxicities in PD-1 inhibitor monotherapy versus chemotherapy or everolimus control

The relative risks (RRs) of hematologic toxicities were calculated by comparing the development of toxicities from the PD-1 inhibitor arm to those from the control arm in the same trial. The RRs of all-grade anemia, thrombocytopenia, leukopenia, and neutropenia were 0.19 (95% CI 0.13–0.29; p<0.00001), 0.05 (95% CI 0.02–0.12; p<0.00001), 0.05 (95% CI 0.02–0.12; p<0.00001), and 0.02 (95% CI 0.01–0.04; p<0.00001), respectively (Figure 5). The random effects model was used for calculating the RR of all-grade anemia owing to the remarkable heterogeneity. Moreover, the RRs of high-grade toxicities were 0.12 for anemia (95% CI 0.08–0.18; p<0.00001), 0.06 for thrombocytopenia (95% CI 0.02–0.19; p<0.00001), 0.06 for leukopenia (95% CI 0.02–0.18; p<0.00001), and 0.02 for neutropenia (95% CI 0.01–0.04; p<0.00001) (Figure 6). The combined results demonstrated that the use of PD-1 inhibitor monotherapy is associated with a decreased risk of developing all-grade and high-grade hematologic toxicities.

Figure 5

Forest plots of relative risk of all-grade hematologic toxicities associated with PD-1 inhibitor monotherapy (experimental) versus chemotherapy or everolimus monotherapy (control): (A) anemia, (B) thrombocytopenia, (C) leukopenia, and (D) neutropenia.

Abbreviations: PD-1, programmed cell death-1; M–H, Mantel–Haenszel (dichotomous).

Figure 6

Forest plots of relative risk of high-grade hematologic toxicities associated with PD-1 inhibitor monotherapy (experimental) versus chemotherapy or everolimus monotherapy (control): (A) anemia, (B) thrombocytopenia, (C) leukopenia, and (D) neutropenia.

Abbreviations: PD-1, programmed cell death-1; M–H, Mantel–Haenszel (dichotomous).

Incidence of hematologic toxicities during PD-1 inhibitor monotherapy in diverse tumor types

Subgroup analysis was conducted to investigate the incidences of hematologic toxicities by distinct tumor types. It is notable that the incidence of all-grade anemia appeared to occur somewhat more frequently in RCC (8%, 95% CI 6%–12%) compared with NSCLC (4%, 95% CI 3%–6%), melanoma (4%, 95% CI 3%–6%), urothelial carcinoma (3%, 95% CI 2%–5%), and HNSCC (5%, 95% CI 3%–10%), but the incidences of high-grade anemia among these tumor types could not be compared with each other because of the rare incidences (Figures 7 and S2). Moreover, no significant differences were noted between NSCLC and melanoma for all-grade leukopenia, neutropenia, and thrombocytopenia (Figure S3). However, this conclusion requires more trials for it to be further verified.

Figure 7

Forest plots of incidence of all-grade anemia during PD-1 inhibitor monotherapy by tumor type: (A) NSCLC, (B) melanoma, (C) RCC, (D) urothelial carcinoma, and (E) HNSCC.

Abbreviations: PD-1, programmed cell death-1; NSCLC, non-small cell lung cancer; RCC, renal cell carcinoma; HNSCC, head and neck squamous cell carcinoma; IV, inverse variance.

Publication bias

The heterogeneity of some studies was indicated by the assessment of all-grade anemia in the monotherapy data (I2=46%), which appeared to be concentrated in the studies on NSCLC (I2=55%). Funnel plots for the monotherapy data were shown to visually demonstrate publication bias (Figure S4).

Discussion

To date, the risks of developing hematologic toxicities from PD-1 inhibitors have not adequately assessed. Herein, we report the results of a meta-analysis including data from 26 clinical trials and 5,088 cancer patients, focused on PD-1 inhibitor-related hematologic toxicities. Our results demonstrated that the overall risk of all-grade anemia is significantly higher than the risk of thrombocytopenia, leukopenia, and neutropenia during PD-1 inhibitor monotherapy (Figure 2). Combination therapy was associated with a significantly increased risk of all-grade hematologic toxicities and only high-grade thrombocytopenia compared with monotherapy (Figures 2–4 and S1). Our analysis also revealed that compared with chemotherapy or everolimus control arms, the use of PD-1 inhibitor monotherapy was associated with a significantly lower risk of developing all-grade and high-grade hematologic toxicities (Figures 5 and 6). Furthermore, there was a nearly double incidence of all-grade anemia in RCC compared with NSCLC, melanoma, urothelial carcinoma, and HNSCC (Figure 7), implying a potential tumor-specific relationship that needs to be further clarified. To our knowledge, this is the first meta-analysis report to focus on PD-1 inhibitor-related hematologic toxicities in cancer patients that compares the risks among different therapeutic regimens and tumor types.

The incidence of hematologic toxicities was significantly higher in the combination therapy group compared with the monotherapy group for all-grade hematologic toxicities but only high-grade thrombocytopenia, indicating the additive effects of 2 agents in developing myelosuppression, particularly reducing platelet counts and potentially leading to severe bleeding. Of the agents used in combination regimens, ipilimumab is another immune-checkpoint inhibitor which acts via the cytotoxic T-lymphocyte antigen-4 pathway and is known to be associated with a variety of irAEs.27–30 Although less clinically apparent, and thus less recognized than PD-1 inhibitor-related hematologic toxicities, ipilimumab-associated myelosuppression has been previously reported in a clinical trial.27 Although the detailed role and effects of ipilimumab on the development and severity of hematologic toxicities when used in combination with PD-1 inhibitors remain to be further investigated, clinicians should be alerted to the significantly higher incidences of hematologic toxicities and perform regular hematologic monitoring in cancer patients during combination therapy.

A higher incidence of anemia among patients with RCC compared with patients with other tumor types may be due to these patients being more prone to developing drug-related erythrocytopenia because of underlying kidney conditions, including frequency of hematuria and incomplete deficiency of erythropoietin. Existing tumor burden in the kidney may also impair the capability for renal elimination of metabolites from the blood, leading to the accumulation of toxic metabolites, some of which are potent inhibitors of erythropoiesis. Although the underlying reasons remain to be further investigated, these observations emphasize a need for careful monitoring of hemoglobin concentration and red blood cell count in patients with RCC during PD-1 inhibitor therapy, because of the possibility of anemia. However, our results indicated that there is no different susceptibility to developing PD-1 inhibitor-related leukopenia, neutropenia, and thrombocytopenia among patients with different tumor types; this is probably due to the limited data from current trials.

Our meta-analysis has some limitations. First, the analysis was performed at the study level rather than analyzing data on the individual patient, meaning that potential variables at the patient level, such as age, prior chemotherapy, and palliative irradiation, were not included in the analysis. Second, different doses of therapeutic agents, frequencies of administration of agents, and types of malignancies may be sources of heterogeneity among the included studies. Third, we did not include hepatocellular carcinoma, triple-negative breast cancer, or malignant pleural mesothelioma for subgroup analysis because there was only one published report on each of these, with relatively small sample sizes. Likewise, other tumor types, such as lymphoma or ovarian cancer, were not included because of the paucity of published reports at the time of data collection. Finally, publication bias was present in the monotherapy data for all-grade anemia, which was particularly concentrated in NSCLC studies, probably reflecting underreporting of small, negative, or non-significant data in the published literature.

Conclusion

Our meta-analysis has demonstrated that PD-1 inhibitor monotherapy increases the risk of anemia in cancer patients, particularly in patients with RCC, compared with thrombocytopenia, leukopenia, and neutropenia. However, the risks of hematologic toxicities are lower in patients treated with PD-1 inhibitor monotherapy compared with combination therapy and chemotherapy or everolimus treatment. Moreover, the use of PD-1 inhibitors in combination with ipilimumab, peptide vaccines, or chemotherapy is associated with an increased risk of developing severe bleeding that requires platelet transfusions. The motive of this meta-analysis was to identify what is known and what remains uncertain based on systematic investigations of the published clinical data. We believe that with more data sharing from multiple studies in the future, we will move closer toward the united goal of maximizing the benefits of cancer immunotherapy.