Safety profile of EZH2 inhibitors for cancer: a systematic review and meta-analysis

Search strategy and selection criteria

We conducted a systematic review and meta-analysis to identify published clinical trials that employed EZH2-targeted inhibitors and reported TRAEs. Comprehensive searches were conducted in PubMed, Embase, CENTRAL, and ClinicalTrials.gov before May 31, 2024. The search terms used included “EZH2”, “inhibitor”, and “clinical trial” (see Table S1 for detailed search details). To ensure thorough coverage, references of relevant reviews and articles were manually reviewed.

Selection was based on the following criteria: (1) studies reporting on prospective clinical trials before May 31, 2024, (2) clinical trials of EZH2 inhibitors as monotherapy or in combination in patients with solid or hematological tumors, (3) clinical trials that provided detailed incidence or data on TRAEs, and (4) studies published in English. Exclusion was based on the following criteria: (1) studies with fewer than ten participants in the EZH2 inhibitor group, (2) trials using sequential therapies, (3) studies without clear TRAEs data, (4) publications reporting duplicate populations. For duplicated publications, we only included the most recent study or the study reporting the most complete profile of TRAEs.

Literature searches and data extraction were independently conducted by two reviewers (ZZ and XC), with discrepancies resolved through discussion or consultation with a senior author (HS).

This study adhered to the PRISMA guidelines (Page et al., 2021), and a prospective protocol was registered with INPLASY (202330028).

Data extraction

Data extracted included trial name (or author details), publication year, trial phase, type of cancer, name of the EZH2 inhibitor, dose of EZH2 inhibitor, number of patients receiving the therapy, and the number of patients experiencing at least one TRAEs. Specific data on all-grade TRAEs (Common Terminology Criteria for Adverse Events (CTCAE) grades 1 to 5) (Shah, 2022), Grade 3+ TRAEs (CTCAE grades 3 to 5), severe TRAEs, treatment-related deaths, TRAEs leading to dose reduction, TRAEs leading to therapy interruption, TRAEs leading to therapy discontinuation, and TRAEs leading to death were recorded. Adverse event terminology was standardized according to the Medical Dictionary for Regulatory Activities (MedDRA) (MedDRA, 2018).

Outcomes

According to the prespecified protocol, the primary outcomes were the overall incidences of TRAEs (defined as the number of patients experiencing at least one adverse event divided by the total number of patients), the profiles of all grade TRAEs, severe TRAEs, grades 3 or higher TRAEs, and treatment-related death. Additional analyses included the incidence of TRAEs leading to dose reduction, therapy interruption, and therapy discontinuation.

Statistical analysis

We applied random-effects models with logit transformation for meta-analysis of TRAEs incidences and TRAEs profiles. The effect size of all pooled results was represented by 95% CI with an upper limit and a lower limit. Subgroup analyses were performed using incidence with 95% confidence intervals, calculated through random-effects models.

Heterogeneity among studies was assessed using I² statistics. All included studies (single-arm studies) were assessed by methodological index for non-randomized studies (MINORS) (Slim et al., 2003). The sensitivity analysis was performed by excluding each study one by one for the pooled results with high heterogeneity. Potential publication bias was evaluated through visual inspection of funnel plots. All statistical analyses were performed using STATA software, employing packages suitable for meta-analysis.

Protocol deviations

In the course of this systematic review and meta-analysis, several deviations from the initially registered protocol were necessary to accommodate new data and expand the scope of the analysis. Firstly, the cut-off time for including published clinical trials was extended to May 31, 2024. This adjustment was made to incorporate the most recent clinical trials involving EZH2 inhibitors, ensuring that our analysis reflects the latest developments and data in this rapidly evolving field. Secondly, we expanded our selection criteria to include combination therapies comprising three or more classes of therapeutic agents. This inclusion was driven by the growing clinical interest in more complex combination regimens that target multiple pathways, which may influence the safety profile and efficacy of EZH2-targeted treatments. These changes were made to enhance the comprehensiveness and relevance of our meta-analysis, allowing for a more detailed and current understanding of the TRAEs profiles associated with EZH2 inhibitors in cancer treatment.

Study selection and characteristics

Figure 1 shows a flowchart of the study selection procedure. The electronic searches yielded 351 potentially relevant records, of which 79 potentially eligible records were assessed (Table S2). After further screening, we identified 28 studies reporting adverse events data of EZH2 inhibitors in cancer treatment (including treatment-emergent adverse events (TEAEs) and TRAEs) (Table 1). Finally, 22 eligible studies (20 trials) involving 1,002 patients, were included in the quantitative analysis (Abida et al., 2021; Chawla et al., 2021; Chi et al., 2022, 2023; Eisai, 2022; Feng et al., 2023; Gounder et al., 2020; Harb et al., 2018; Hussain et al., 2024; Italiano et al., 2018; Izutsu et al., 2023; Morschhauser et al., 2020; Palmieri et al., 2023; Palomba et al., 2022; Puram et al., 2023; Salles et al., 2023; Sarkozy et al., 2020; Schweizer et al., 2022; Song et al., 2024, 2022; Yap et al., 2019; Zauderer et al., 2022).

Table S3 shows the baseline characteristics of the 22 studies (20 trials). Four trials (NCT04407741, NCT04224493, NCT04204941, and NCT04179864) are randomised controlled trials in phase II or phase III. However, the results of these phases have not yet been published. Cancer types tested in these studies included hematologic malignancies (n = 9), solid tumors (n = 7), and both (n = 6). Among the 1,002 patients included in the meta-analysis, 666 patients (15 studies) received tazemetostat, 174 patients (three studies) received SHR2554, 57 patients (one studies) received PF-06821497, 41 (one studies) received GSK2816126, 32 patients (one studies) received CPI-1205, and 25 patients (one studies) received valemetostat. For drug combinations, 744 patients (13 studies) received monotherapy, 33 patients (two studies) received tazemetostat with chemotherapy, 117 patients (four studies) received tazemetostat/SHR2554 with immunotherapy, 44 patients (one study) received tazemetostat with targeted therapy, 43 patients (one study) received tazemetostat with immunotherapy and targeted therapy, 14 patients (one study) received tazemetostat with endocrine therapy.

Quality assessment

Two independent authors (Z.Z. and X.C.) assessed the quality of each study included in the meta-analysis using MINORS. Evaluations were compared, and any inconsistencies between the review authors were discussed and resolved. The MINORS scores ranged from 12 to 16 out of a possible score of 16 (Table 2).

Table 2:

Methodological index for non-randomized studies (MINORS).

Trial name	Year	MINORS total score	1	2	3	4	5	6	7	8
NCT02395601	2018	15
NCT02082977	2019	15
NCT03460977	2023	14
NCT03603951	2022	13
NCT03603951	2024	15
NCT04407741	2023	14
NCT04224493	2023	14
NCT01897571	2020	15
NCT01897571	2017	15
NCT02601950	2020	16
NCT04204941	2021	12
NCT03213665	2023	14
NCT04179864	2021	12
NCT02860286	2022	15
NCT02889523	2020	14
NCT03456726	2022	15
NCT02601937	2021	14
NCT03854474	2024	13
NCT02220842	2022	15
NCT04705818	2023	12
NCT04624113	2023	13
NCT04102150	2023	15

DOI: 10.7717/peerj.18871/table-2

Note:

Items 1–12 represent: 1, a clearly stated aim; 2, inclusion of consecutive patients; 3, prospective collection of data; 4, endpoints appropriate to the aim of the study; 5, unbiased assessment of the study endpoint; 6, follow-up period appropriate to the aim of the study; and 7, loss to follow-up less than 5%; 8, prospective calculation of the study size. An item scored 0 means not mentioned, one means reported but inadequate, and two means reported and adequate. The total score was 16 for self-controlled studies. Use red for 0, yellow for 1 and green for 2.

Any-grade TRAEs

Any-grade TRAEs were reported in 418/487 patients of 10 studies. The pooled result of any-grade TRAEs was 86% (95% CI [79–94%]; I² = 89.5%). Subgroup analyses were performed based on drug, drug combination, tumor type, and drug dose. For Tazemetostat, the pooled rates were 85% (95% CI [77–94%]; I² = 68.5%) for all-grade TRAEs. In the case of SHR2554, the pooled rates were 82% (95% CI [64–100%]; I² = 96.3%). The rate of any-grade TRAEs in GSK2816126 and valemetostat each reported in only one study, were 90% and 96%, respectively (Fig. 2A). When restricted to monotherapy studies, the pooled rates for tazemetostat, SHR2554, and valemetostat were 80%, 99%, and 96%, respectively (Fig. S1A). Further limiting the analysis to monotherapy studies using active doses, the pooled rates for tazemetostat, SHR2554, and valemetostat were 82%, 100%, and 96%, respectively (Fig. S2A).

When subgroup analysis was performed based on drug combination, the probability of any-grade TRAEs was 90%, 39%, 86%, and 95% for monotherapy, combination with immunotherapy, combination with endocrine therapy, and combination with targeted and immunotherapy, respectively (Fig. S3A). When analyzed by cancer type, the rates of any-grade TRAEs were 94%, 86%, and 70% in studies that included only hematologic tumors, only solid tumors, and both, respectively (Fig. S3B).

Grade 3 or higher TRAEs

Grade 3 or higher TRAEs were reported in 177/602 patients of 13 studies. The pooled result of grade 3 or higher TRAEs was 33% (95% CI [21–44%]; I² = 93.5%). For tazemetostat, the pooled rates were 34% (95% CI [21–47%]; I² = 89.9%) for grade 3 or higher TRAEs. In the case of SHR2554, the pooled rates were 40% (95% CI [24–56%]; I² = 63.6%). The rate of grade 3 or higher TRAEs in CPI-1205, GSK2816126 and valemetostat each reported in only one study, were 22%, 0, and 56%, respectively (Fig. 2B). When restricted to monotherapy studies, the pooled rates for tazemetostat, SHR2554, and valemetostat were 23%, 40%, and 56%, respectively (Fig. S1B). Further limiting the analysis to monotherapy studies using active doses, the pooled rates for tazemetostat, SHR2554, and valemetostat were 27%, 50%, and 56%, respectively (Fig. S2B).

When subgroup analysis was performed based on drug combination, the pooled rates were 27%, 29%, 81%, and 49% for monotherapy, combination immunotherapy, combination chemotherapy, and combination targeted and immunotherapy, respectively (Fig. S4A). When analyzed by cancer type, the pooled rates were 41%, 35%, and 4% in studies that included only hematologic tumors, only solid tumors, and both, respectively (Fig. S4B).

Severe TRAEs

Severe TRAEs were reported in 87/586 patients of 11 studies. The pooled result of severe TRAEs was 15% (95% CI [9–22%]; I² = 87.5%). For tazemetostat, the pooled rates were 15% (95% CI [7–23%]; I² = 87.9%) for severe TRAEs. In the case of SHR2554, the pooled rates were 20% (95% CI [8–32%]; I² = 51.4%). The rate of severe TRAEs in GSK2816126 and valemetostat each reported in only one study, were 0% and 28%, respectively (Fig. 3A). When restricted to monotherapy studies, the pooled rates for tazemetostat, SHR2554, and valemetostat were 15%, 20%, and 28%, respectively (Fig. S1C). Further limiting the analysis to monotherapy studies using active doses, the pooled rates for tazemetostat, SHR2554, and valemetostat were 19%, 29%, and 28%, respectively (Fig. S2C).

When subgroup analysis was performed based on drug combination, the pooled rates were 15%, 12%, and 21% for monotherapy, combination chemotherapy, and combination targeted and immunotherapy, respectively (Fig. S5A). When analyzed by cancer type, the pooled rates were 20%, 18%, and 1% in studies that included only hematologic tumors, only solid tumors, and both, respectively (Fig. S5B).

Compliance

Compliance was analyzed from three aspects: TRAEs leading to discontinuation, dose reduction, and interruption. TRAEs leading to discontinuation was reported in 28/583 patients of 11 studies. The pooled result of discontinuation was 4% (95% CI [2–5%]; I² = 15.0%). For tazemetostat, the pooled rate was 4% (95% CI [1–7%]; I² = 45.5%). In the case of SHR2554, the pooled rate was 4% (95% CI [0–7%]; I² = 0). The rate in GSK2816126 and valemetostat each reported in only one study, were 2% and 8%, respectively (Fig. 3B).

TRAEs leading to dose reduction was reported in 53/587 patients of 11 studies. The pooled result of dose reduction was 8% (95% CI [4–13%]; I² = 80.4%). For tazemetostat, the pooled rates were 10% (95% CI [3–16%]; I² = 87.2%) for dose reduction. In the case of SHR2554, the pooled rates were 8% (95% CI [3–12%]; I² = 0). The rate in GSK2816126 and valemetostat each reported in only one study, were 5% and 8%, respectively (Fig. S6A). TRAEs leading to interruption was reported in 175/505 patients of nine studies. The pooled result of interruption was 34% (95% CI [30–38%]; I² = 88.8%). For tazemetostat, the pooled rate was 36% (95% CI [31–41%]; I² = 93.5%). In the case of SHR2554, the pooled rate was 36% (95% CI [28–44%]; I² = 0). The rate in GSK2816126 and valemetostat each reported in only one study, were 24% and 16%, respectively (Fig. S6B).

TRAEs leading to death

TRAEs leading to death was reported in 5/583 (0.9%) patients of 11 studies. Among them, three died due to infection, two died due to cardiopulmonary failure. In the study (NCT03603951) (Song et al., 2022), one patient died due to skin infection and toxic epidermal necrolysis and another one patient died because of respiratory failure. Both patients received SHR2554 350 mg twice daily. In the study (NCT02889523) (Sarkozy et al., 2020), one patient died because of infection-related death. The patient received tazemetostat (800 mg twice daily) with R-CHOP regimen every 21 days. In the study (NCT02220842), one died due to cardiac decompensation secondary to the event of hyponatremia, and one died due to septic shock. Both of them received tazemetostat (800 mg twice daily) every 21 days in combination with atezolizumab (1,200 mg) on day 1 of each cycle.

Specific any-grade TRAE spectrum

The most common laboratory any-grade TRAEs for EZH2 inhibitors were thrombocytopenia (139/466, 30%), anemia (158/597, 26%), neutropenia (98/450, 22%), ALT increased (30/456, 7%), lymphopenia (25/391, 5%), AST increased (21/403, 5%), and hypokalemia (14/277, 5%) (Fig. 4A). The most common non-laboratory any-grade TRAEs for EZH2 inhibitors were nausea (148/597, 25%), fatigue (139/584, 24%), decreased appetite (32/163, 20%), alopecia (27/144, 19%), muscle spasms (17/106, 16%), rash (21/121, 17%), vomiting (67/484, 14%), diarrhea (64/524, 12%), and pyrexia (17/279, 6%) (Fig. 4B).

For tazemetostat, the most common laboratory any-grade TRAEs were anemia (63/347, 18%), neutropenia (35/273, 13%), thrombocytopenia (31/257, 12%), lymphopenia (7/189, 4%), hypokalemia (3/125, 2%), ALT increased (4/251, 2%), and AST increased (4/251, 2%) (Fig. 4C). The most common non-laboratory any-grade TRAEs for tazemetostat were fatigue (83/322, 26%), nausea (80/335, 24%), decreased appetite (24/127, 19%), rash (21/121, 17%), alopecia (17/119, 14%), vomiting (32/247, 13%), diarrhea (34/274, 12%), and pyrexia (9/226, 4%) (Fig. 4D).

For SHR2554, the most common laboratory any-grade TRAEs were thrombocytopenia (88/173, 51%), anemia (71/173, 41%), neutropenia (52/141, 37%), hyperbilirubinemia (18/141, 37%), AST increased (17/141, 12%), ALT increased (16/141, 11%), lymphopenia (12/141, 9%), and hypokalemia (11/141, 8%) (Fig. 4E). The most common non-laboratory any-grade TRAEs for SHR2554 were nausea (28/141, 20%), fatigue (18/141, 13%), vomiting (18/173, 10%), and diarrhea (11/141, 8%) (Fig. 4F).

Specific grade 3 or higher TRAE spectrum

The most common laboratory grade 3 or higher TRAEs for EZH2 inhibitors were neutropenia (54/641, 8%), thrombocytopenia (52/622, 8%), anemia (41/641, 6%), lymphopenia (12/611, 2%), ALT increased (8/570, 1%), AST increased (4/570, 1%), and hypokalemia (3/436, 1%) (Fig. 5A). The most common grade 3 or higher non-laboratory TRAEs for EZH2 inhibitors were constipation (4/138, 3%), pyrexia (6/339, 6%), fatigue (7/591, 1%), nausea (7/611, 1%), decreased appetite (3/268, 1%), and vomiting (4/589, 1%) (Fig. 5B).

The most common laboratory grade 3 or higher TRAEs for tazemetostat were neutropenia (9/199, 5%), anemia (7/199, 4%), thrombocytopenia (5/199, 3%), lymphopenia (2/199, 1%), ALT increased (1/199, 1%), AST increased (1/199, 1%), and hypokalemia (1/199, 1%) (Fig. 5C). The incidence of grade 3 or higher non-laboratory TRAEs with tazemetostat is limited (all < 2%) (Fig. 5D).

For SHR2554, the most common laboratory grade 3 or higher TRAEs were thrombocytopenia (29/173, 17%), neutropenia (14/173, 8%), and anemia (12/173, 7%) (Fig. 5E). The incidence of grade 3 or higher non-laboratory TRAEs with SHR2554 was not reported.

Sensitivity analysis

Sensitivity analyses were conducted to evaluate the robustness of the pooled estimates for any-grade TRAEs (Fig. S7A), grade 3 or higher TRAEs (Fig. S7B), and TRAEs leading to discontinuation (Fig. S7C). The analyses involved sequentially omitting individual studies from the meta-analysis. The overall results remained stable and were not significantly altered by the exclusion of any single study, suggesting that our findings are robust and not unduly influenced by any particular study.

Risk of bias

To assess the presence of risk of bias in our meta-analysis, funnel plots were constructed for any-grade TRAEs (Fig. S8A), grade 3 or higher TRAEs (Fig. S8B), and TRAEs leading to discontinuation (Fig. S8C). The analysis of the funnel plots for any-grade TRAEs suggested potential publication bias in three studies: NCT01897571 (2017), NCT01897571 (2020), and NCT04407741 (2023). Similarly, the evaluation for grade 3 or higher TRAEs revealed potential biases in five studies: NCT01897571 (2017), NCT02601937-2 (2022), NCT02860286 (2022), NCT02395601 (2018), and NCT04204941 (2021). For TRAEs that led to therapy discontinuation, potential publication bias was observed in one study, NCT04224493 (2023). These findings indicate the need for cautious interpretation of the pooled results, particularly in the presence of indicated biases.