Efficacy of Immune Checkpoint Inhibitors against Advanced or Metastatic Neuroendocrine Neoplasms: A Systematic Review and Meta-Analysis

Simple Summary Neuroendocrine neoplasms (NENs) are relatively rare neoplasms, but their incidence is rising. Despite the recent understanding of and advances in the treatment of NENs, the clinical outcomes of patients with NENs, especially high-grade NENs, are poor. This meta-analysis was conducted to systematically evaluate the effectiveness of immune checkpoint inhibitors (ICIs) in patients with advanced or metastatic NENs. In 10 studies with 464 patients, the pooled overall response rate was 15.5% (95% confidence interval, 9.5–24.3%), but the response differed according to the primary tumor site, tumor differentiation, and drug regimen. Poorly differentiated NENs consistently showed a better overall response rate than well-differentiated NENs in multiple sensitivity analyses, suggesting that different expression of immune checkpoints and tumor mutational burden may influence the treatment efficacy of ICIs for advanced or metastatic NENs. The variation in treatment efficacy of ICIs according to tumor differentiation and drug regimen should be considered for patient-tailored management. Abstract We performed a systematic review and meta-analysis of the treatment efficacy of immune checkpoint inhibitors (ICIs) in advanced/metastatic neuroendocrine neoplasms (NENs). MEDLINE and EMBASE were searched to identify studies that provide data on treatment response and/or survival outcomes of advanced/metastatic NEN patients treated with ICIs. The overall response rate (ORR) was pooled using a random-effects model. Meta-regression was performed to explore factors influencing the ORR. Individual patient data (IPD) meta-analysis of survival was performed using stratified Cox regression. Ten studies (464 patients) were included. The overall pooled ORR was 15.5% (95% confidence interval (CI), 9.5–24.3%), and it varied according to the primary site (thoracic, 24.7%; gastro–entero–pancreatic, 9.5%), tumor differentiation (poorly differentiated, 22.7%; well-differentiated, 10.4%), and drug regimen (combination, 25.3%; monotherapy, 10.1%). All these variables significantly influenced the ORR. Tumor differentiation was associated with both overall survival and progression-free survival (hazard ratio of poorly differentiated tumors, 4.2 (95% CI, 2.0–8.7) and 2.6 (95% CI, 1.6–4.4), respectively). Thus, the treatment efficacy of ICIs for advanced/metastatic NENs varied according to primary site, tumor differentiation, and drug regimen. Poorly differentiated NENs showed a better ORR than well-differentiated NENs but had a negative impact on survival.


Introduction
Neuroendocrine neoplasms (NENs) represent a heterogeneous group of tumors originating from the neuroendocrine cells of various organs [1]. The behavior and prognosis of NENs are highly variable and depend on the site of origin, tumor grade, and differentiation [2]. Although NENs are relatively uncommon tumors, recent reports have shown that their incidence and prevalence are growing worldwide [3][4][5]. For patients with advanced/metastatic NENs, systemic therapy is the only available treatment option. The recommended regimens include sunitinib, everolimus, somatostatin analogue, peptide receptor radionuclide therapy, and cytotoxic chemotherapy [6][7][8]. Despite recent progress in the understanding of the pathophysiology and molecular biology, as well as advances in treatment strategies for NENs [9][10][11], the clinical outcomes for individuals with high-grade NENs are poor, with a reported median overall survival (OS) of 11 months and median progression-free survival (PFS) of 4 months for patients with poorly differentiated (PD) neuroendocrine carcinoma (NEC) [12,13].
Immune checkpoint inhibitors (ICIs) have made great breakthroughs in cancer therapy and have gained approval for the treatment of various types of cancer [8,14]. In the case of NENs, ICIs were approved for two distinct tumor types: Merkel cell carcinoma and small cell lung cancer [15][16][17][18], both of which have high tumor mutational burden (TMB) and environmental causes of immunogenicity [19]. For other NENs, several early phase trials and clinical studies have evaluated the efficacy of ICIs and given us a preliminary understanding of their role. To date, however, no attempt has been made to generate an evidence-based comprehensive summary of these agents from the scattered individual studies. We thus conducted this systematic review and meta-analysis of the antitumor activity and clinical outcomes of ICI-treated advanced or metastatic NEN patients using currently available published data, which may help further characterize and guide future studies on the role of ICIs in the treatment of NENs.

Search Strategy
The study was performed in accordance with the standard guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [20] and PRISMA for Individual Patient Data systematic reviews (PRISMA-IPD) [21]. This study was registered in OSF Registries (osf.io/3tgd4). A comprehensive search of MEDLINE and EMBASE was conducted by two reviewers (E.J.P. and H.J.P.) independently to identify relevant studies published before 1 April 2021. The detailed search strategy is shown in Table S1. The search was limited to human subjects and English-language studies.

Eligibility Criteria
The studies identified from the two databases were combined and duplicate studies were removed from the list. Titles and abstracts were screened for relevance. The relevant full texts were meticulously evaluated for compliance with the inclusion criteria listed below. To expand the search, the bibliographies were also screened for potentially suitable studies. Two reviewers (E.J.P. and H.J.P.) independently selected the studies, and disagreements, if any, were resolved by discussion. Based on the population, intervention, comparison, outcome, study design (PICOS) approach [22], studies fulfilling the following criteria were included: (a) population: patients diagnosed with NENs; (b) intervention: NEN patients treated with ICIs; (c) outcomes: treatment responses or survival endpoints (OS and/or PFS) assessed using Response Evaluation Criteria In Solid Tumors (RECIST) 1.1 [23]; and (d) study design: clinical trials and observational studies (prospective or retrospective). The exclusion criteria were: (a) case reports, reviews, letters, editorials, and conference abstracts; (b) no study of ICI treatment in NEN patients; (c) studies with insufficient response assessment data; (d) studies with cohort overlap. For studies with overlapping cohort, the study with the longest follow-up was selected.

Data Extraction
Two reviewers (E.J.P. and H.J.P.) independently extracted the relevant data from the included studies, and disagreements, if any, were resolved by discussion. The following data were extracted: (a) study characteristics, including authors, year of publication, and study design; (b) demographic and clinical characteristics of the patients, including sample size, types of administered ICI, primary tumor site, and tumor grade and/or differentiation according to the most recent World Health Organization (WHO) classification systems per site [10,[24][25][26]; (c) overall response rate (ORR) based on RECIST 1.1; (d) OS and PFS based on RECIST 1.1 [27,28]. The IPD for the OS and PFS was reconstructed by extracting the data from the given Kaplan-Meier curves using a digital software (WebPlotDigitizer v 4.5, https://automeris.io/WebPlotDigitizer, accessed on 12 October 2021) in accordance with the method proposed by Guyot et al. [29]. At all times, we used the endpoint definitions and analyses reported by authors and made no attempts to reclassify or recalculate the values published in the original reports.

Definition of Endpoints for Treatment Efficacy
Endpoints were defined in accordance with guidelines published by the United States Food and Drug Administration [28]. As a response-related endpoint, the ORR, a direct measure of the tumoricidal effects of a treatment, was defined as the proportion of patients showing a complete or partial response per RECIST 1.1. OS and PFS were used as survival endpoints. OS was defined as the time from treatment initiation to death from any cause. PFS was defined as the time from treatment initiation to progressive disease (according to RECIST 1.1) or to death from any cause. Patients were considered censored for a given endpoint if they did not experience this endpoint during their follow-up period.

Quality Assessment
Two reviewers (E.J.P. and H.J.P.) independently reviewed the risk of bias and methodologic quality of the included studies. The Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool [30] was used. Any discrepancy was resolved by discussion.

Statistical Analysis
For pooling the ORR, considering the characteristics of the study patients and the treatment differed between studies, we used the random-effects model with inverse-variance weighting to generate the summary estimate of the magnitude of effect. DerSimonian-Laird random-effect models [31] were constructed to synthesize the pooled ORR with 95% confidence intervals (CIs). The heterogeneity between studies was assessed with I 2 statistics and a Cochran Q test. An I 2 > 50% or p < 0.1 in the Q test indicated substantial heterogeneity [32]. The publication bias was estimated using funnel plots and Begg's test [33].
Subgroup analyses were performed to calculate the pooled ORR of each group of primary tumor site (i.e., thoracic [lung or thymus origin] or gastro-entero-pancreatic [GEP]), tumor differentiation (well-differentiated [WD] or PD), and drug regimen (ICI monotherapy or combinations of ICIs). Grade 3 neuroendocrine tumors (NETs) were included in the WD tumor group. Meta-regression analysis was performed to identify influencing factors for the ORR.
With regard to survival outcomes, using the reconstructed IPD, the OS and PFS were calculated using the Kaplan-Meier method. The influence of the primary tumor site, tumor differentiation, and drug regimen on the OS and PFS was assessed using one-stage stratified Cox proportional-hazard regression analysis to account for clustering at the level of each study [34]. Statistical analyses were performed with R, version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria).

Study Search and Characteristics
The study search process is outlined in Figure 1. The characteristics of the 10 included studies [19,[35][36][37][38][39][40][41][42][43], which involved a total of 464 NEN patients, are summarized in Table 1. Seven of these studies were nonrandomized clinical trials (two phase I studies and five phase II studies), and three were retrospective studies. Two studies included only largecell NECs (LCNECs) of the lung. The remaining eight studies included NENs of various primary sites, including the lung, pancreas, and gastrointestinal tract. All studies included patients with advanced or metastatic NENs, and six studies specified that the patients had progressed during previous systemic treatment(s). Two studies included only PD tumors (large-cell NEC of the lung), two included only WD tumors, and six included both. The ICI regimens varied between studies; six studies involved ICI monotherapy (three with pembrolizumab, one with nivolumab or pembrolizumab, one with toripalimab, and one with spartalizumab) only, two studies reported on an ICI combination therapy (ipilimumab and nivolumab) only, and two studies described both ICI monotherapy and combination therapy. According to the ROBINS-I tool, the risk of bias for overall risk of bias for the included studies was low in seven studies and moderate in three studies ( Figure S1).

Response-Related Endpoint
The pooled estimates of the ORR are shown in Figure 2. The pooled ORR of the included studies was 15.5% (95% CI, 9.5-24.3%) (Figure 2A). Visual inspection of the funnel plot revealed no asymmetry when pooling the ORR across all studies ( Figure S2), and no significant publication bias was evident according to Begg's test (p = 0.79). Heterogeneity was present in pooling the ORR from all studies (I 2 = 72%). When classified by primary tumor site ( Figure 2B), the pooled ORRs were 9.5% (95% CI, 4.5-19.2%) in the GEP group and 24.7% (95% CI, 16.1-36.1%) in the thoracic group. The pooled ORRs of pancreatic NENs and gastrointestinal (GI) NENs were 11.4% (95% CI, 4.6-25.6%) and 9.0% (95% CI, 4.4-17.6%), respectively. Based on tumor differentiation ( Figure 2C), the pooled ORR of the WD group was 10.4% (95% CI, 5.5-18.6%), while that of the PD group was higher (22.7%; 95% CI, 14.8-33.0%). The pooled ORR of the monotherapy group was 10.1% (95% CI, 5.0-19.3%), and that of the combination therapy group was 25.3% (95% CI, 17.1-35.7%) ( Figure 2D).   Although the heterogeneity between studies decreased when pooling the ORR of each subgroup, it still existed in the GEP (I 2 = 63%) and monotherapy group (I 2 = 70%), which may be attributable to the presence of both PD and WD tumors in the same group. In addition, Klein's study [36] reported an ORR of 40.0% for GEP tumors, far higher than any of the other studies (ranging from 4.8 to 15.6%). Patients included in Klein's study were treated with combination ICI therapy (nivolumab and ipilimumab), whereas patients in the other studies were mostly treated with monotherapy. In the monotherapy group, studies by Lu [37] and Shirasawa [41] reported higher values for the ORR (20.0 and 38.5%, respectively) than the others (less than 9.8%). In Lu's study, patients were eligible for inclusion if they had NEN with a high ki-67 index (≥10%). Shirasawa's study included LCNEC of the lung only and was of the retrospective design. Marginal heterogeneity was noted in the WD group (I 2 = 52%). There was no heterogeneity between the studies in the thoracic group, PD group, and the combination therapy group (I 2 = 0%).

Survival-Related Endpoints
The IPD for the OS and PFS was reconstructed in eight of the included studies [19,36,[38][39][40][41][42][43]. The IPD of 390 patients was reconstructed for the OS, and the median OS was 22.7 months (95% CI, 20.1-25.9 months). When divided by primary tumor site and tumor differentiation, the shortest median OS was noted for GEP PD tumors (6.8 months; 95% CI, 3.6-10.0 months), followed by GEP WD tumors (20.9 months; 95% CI, 20.1 months-not reached) and thoracic PD tumors (22.8 months; 95% CI, 12.3 months-not reached), and the longest median OS was noted for thoracic WD tumors (not reached) ( Figure S4). The median OS of pancreatic NENs was 20.9 months (20.1 months-not reached) and was not reached for GI NENs. The median OSs of WD tumors and PD tumors were 23.7 months (95% CI, 19.3-28.0 months) and 12.3 months (95% CI, 8.0-23.8 months), respectively. With respect to the drug regimen, the data were not separable according to the primary tumor site or tumor differentiation variables. Stratified Cox regression analysis, which included the primary tumor site, tumor differentiation, and drug regimen as covariates, indicated that tumor differentiation was the only influencing factor for OS. PD tumors had a hazard ratio (HR) of 4.2 (95% CI, 2.0-8.7) ( Table 3). For the PFS, the IPD of 388 patients was reconstructed (median PFS, 3.8 months; 95% CI, 3.5-4.1 months). When divided by primary tumor site and tumor differentiation, GEP PD tumors showed the shortest median PFS (1.8 months; 95% CI, 1.7-2.0 months). GEP WD tumors and thoracic PD tumors showed median PFSs of 3.7 months (95% CI, 3.5-4.0 months) and 3.8 months (95% CI, 2.5-5.2 months), respectively, and the longest median PFS was noted for thoracic WD tumors (5.6 months; 95% CI, 4.6-6.6 months). The median PFS of pancreatic NENs was 3.8 months (3.2-4.3 months), and that of GI NENs was 3.5 months (1.7-5.3 months). The median PFSs of patients with WD tumors and PD tumors were 4.2 months (95% CI, 3.4-5.0 months) and 2.8 months (95% CI, 2.0-3.6 months), respectively. From stratified Cox-regression analysis, as found for the OS, tumor differentiation was significantly associated with the PFS, while primary tumor site and drug regimen were not. PD tumors had an HR of 2.6 (95% CI, 1.6-4.4) ( Figure S5, Table 3).

Discussion
This study shows that, on average, ICIs show the ORR of 15.5% in patients with advanced or metastatic NENs. The primary tumor site, tumor differentiation, and drug regimen were found to affect the ORR. PD tumors showed a robustly higher response rate than WD tumors, although patients with PD tumors had poorer survival outcomes.
Given the large between-study heterogeneity in pooling the ORR from all 10 included studies, i.e., I 2 of 72%, it would be not appropriate to merely accept the meta-analytic summary estimate of 15.5%. We explored the source of the heterogeneity by performing subgroup analyses and meta regression. The primary tumor site, tumor differentiation, and drug regimen significantly affected the ORR and were thus revealed as the causes of heterogeneity.
Our study results suggested that ICIs may show substantial antitumor activity against thoracic NENs. The pooled ORR was 24.7% for all thoracic NENs, 24.2% for WD thoracic tumors, and 34.2% for PD thoracic (all LCNEC) tumors. In this regard, Sabari et al. [44] recently reported that LCNEC of the lung treated with ICIs had an ORR of 26.7% (4/15), and a retrospective study from six French centers [45] stated that a PR of 60% (6/10) was achieved for ICI-treated advanced LCNEC of the lung, which also suggested considerable antitumor activity of ICI treatment in PD thoracic tumors as observed in our investigation. Previous studies have reported an ORR for first-line chemotherapy in carcinoid and LCNEC of the lung of 23 and 34-47% [46][47][48][49], respectively, and for second-line chemotherapy in LCNEC of the lung of 17% [50]. Although direct comparisons are limited as patients included in our study were more heavily treated with prior systemic therapies than the above-mentioned studies, we speculated that ICI treatments have the potential to demonstrate better antitumor activity in advanced thoracic NENs than conventional systemic treatments. To confirm this, randomized trials comparing ICIs to standard treatment options, including chemotherapeutics, are necessary.
The current classification of NENs differs between organ systems and causes considerable confusion [51]. To address this, a common framework for NEN classification was proposed by the International Agency for Research on Cancer (IARC) and WHO expert consensus meeting [52] which suggested distinction between differentiated NETs and NECs, irrespective of their site of origin. Hence, we separated grade 3 NETs from PD NECs and included them in the WD tumor group. In our analyses, a different response to ICI treatment was noted in accordance with tumor differentiation. The ORR of the PD group was 22.7%, compared to 10.4% in the WD group, and this trend of higher ORR of PD tumors was robust in terms of the primary tumor site and drug regimen. In other solid tumors, PD-L1 expression and TMB have been identified as potential biomarkers of the tumor response to immunotherapy [53,54]. In the case of NENs, few studies have reported that the expression of PD-1 and PD-L1 is higher in grade 3 NETs and NECs than in lower grade tumors [55] and that TMB was low in grade 1-2 NETs compared to that in NECs [56], indicating that the different response to ICI treatment in WD and PD tumors may be partly attributable to the different expression of immune checkpoints and TMB. We also tried to separate WD tumors into grade 1-2 NETs and grade 3 NETs to the extent available, which was limited as the number numbers of studies and patients were small. Further studies are necessary to clarify the relationship between these potential biomarkers and the response of NENs to ICI treatment.
A prior meta-analysis performed by Bongiovanni et al. [57] investigated the activity of ICIs in NENs; however, their results were inconsistent from ours. In their study, the pooled ORR was 10%, and there was no significant difference in ORR between WD NENs (11%) and PD NENs (12%). Thus, we infer that the main source of discrepancy was the different study eligibility criteria used. We excluded studies with a sample size of <10 patients and conference abstracts [58] and only included the single most recently published study among the studies with overlapped cohorts. The significantly higher pooled ORR in PD NENs than WD NENs, one of our main results, shares the same context with prior studies, which showed higher expression of PD-L1 and TMB in higher-grade NENs than in lower-grade NENs [53,54], which may explain the higher response rate in PD groups compared to WD groups.
Combination ICI regimens showed a higher ORR (25.3%, pooled) than the monotherapy (10.1%, pooled) in our meta-analyses. Three of our included studies evaluated the efficacy of nivolumab plus ipilimumab [35,36,39] and consistently observed a high ORR of 24.1-27.3%. There is now growing evidence that ICI monotherapy has limited effectiveness against NENs, particularly PD tumors [59][60][61], which is consistent with our observations. Only the nivolumab plus ipilimumab combination ICI regimen was evaluable in our study, however, and exploration of the synergistic effect of various combinations of ICIs versus other immunotherapeutic and different anticancer treatments is required.
In our study, the median OS and PFS of all IPD-retrieved patients were 22.7 and 3.8 months, respectively. In one prior phase 2 trial [62], the median OS of 18 NEN patients (14 WD and 4 PD; 22% previously treated) treated with irinotecan plus cisplatin was 11.4 months. In our study, thoracic PD tumors (LCNECs) showed a median OS of 22.8 months. In comparison, prior studies of thoracic LCNEC patients who received firstline chemotherapy have reported a median OS ranging from 8.0-12.6 months [48,49]. In our study, GEP PD tumors had a median OS of 6.8 months and a median PFS of 1.8 months, whereas one previous study reported that patients receiving second-line chemotherapy for an extrapulmonary PD NEN (mostly GEP origin) showed a median OS of 6.2 months and median PFS of 2.3 months [63]. The characteristics of the heavily treated advanced or metastatic NEN patients in our included studies made it difficult to compare the survival outcomes between our present investigation and others. Although it seems that ICI treatment has the potential to provide survival benefit in patients with NENs, especially PD tumors, randomized trials comparing ICIs and standard chemotherapies are necessary to clarify this issue.
Our study has some limitations. First, the number of included studies (10 in total) was small. However, all of the relevant original articles published to date were included in our meta-analysis. A total of 464 NEN patients were analyzed, and the individual participant survival data of more than 380 patients was reconstructed. Second, in terms of the primary tumor site, only those of thoracic or GEP origin were available for separate subgroup analysis. Other sites of origin were not evaluable, as the number of cases was too small and the required information was not fully presented, which is an inherent limitation of a systematic review. However, considering that the GEP and lung are the most frequent sites of NENs, comprising approximately 70 and 20% of all NENs, respectively [2], we believe that our analysis is clinically meaningful. Third, the included studies were all single-arm investigations that did not use a control group for comparison. Therefore, we could not address whether ICIs have better antitumor activity or provide a survival benefit compared to current standard therapies. However, our current results still provide useful insights on the role of ICIs in treatment of NEN and could serve as a basis for designing future randomized controlled trials. Fourth, the role of various combination regimens of ICI was not explored, as there were only three studies which included the combination regimen, and the same single combination regimen (ipilimumab plus nivolumab) was used in these studies. Further studies would be anticipated to clarify the role of ICI combination with various regimens in the treatment of NENs.

Conclusions
In conclusion, our current meta-analysis provides evidence-based measures of the treatment efficacy of ICIs in patients with advanced or metastatic NENs. The overall ORR was 15.5%, which was found to be influenced by the primary tumor site, tumor differentiation status, and by the drug regimen. PD tumors showed a robustly higher response rate than WD tumors but were associated with poorer survival outcomes.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/cancers14030794/s1: Table S1: Detailed search queries; Figure S1: Risk of bias summary of included studies; Figure S2: Funnel plots for visual assessment of publication bias; Figure S3: Pooled ORR of grade 1-2 NETs; Figure S4: Overall survival based on individual patient data; Figure S5  Data Availability Statement: The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.