Neuroendocrine neoplasms (NENs) are a heterogeneous group of malignancies originating from the neuroendocrine cells of various organs. Most NENs are well-differentiated neuroendocrine tumors (WD-NETs), and more than 50% of NETs arise in the gastrointestinal tract and pancreas (1). For unresectable or metastatic NENs, systemic therapeutic options have been proven to delay disease progression and improve survival, which include somatostatin analogs (SSAs), inhibitors of the mammalian target of rapamycin, receptor tyrosine kinase inhibitors, peptide receptor radionuclide therapy (PRRT), liver targeted therapy, and cytotoxic agents (2). Belzutifan, a selective small-molecule inhibitor of hypoxia-inducible factor 2a, has demonstrated a high response rate in VHL-associated pancreatic neuroendocrine tumors and also recently approved by the US Food and Drug Administration (3). Given the high heterogeneity of WD-NETs and the lack of robust predictive biomarkers to guide treatment selection, no clear recommendations exist regarding therapeutic sequences or in combination.

Currently, the role of cytotoxic chemotherapy in advanced WD-NETs remains debated and predominantly considered in advanced NETs with high Ki-67 proliferation index and tumor burden, in a rapidly progressive disease, after failure of the other therapies, and/or when tumor size reduction by cytotoxic intervention is a therapeutic goal (4–6). So far, alkylating agents alone, such as streptozotocin, temozolomide, dacarbazine, or in combination with 5-fluorouracil or capecitabine, are widely used (7). However, alkylating agent-induced DNA damage can be repaired by O⁶-methylguanine-DNA methyltransferase (MGMT), and MGMT positivity was shown to be a relevant biomarker for poor efficacy (8, 9). The wider molecular analysis revealed that no particular mutational or transcriptional profile was associated with TMZ response in contrast to MGMT promoter methylation (10).

Oxaliplatin-based chemotherapy (OX-CT) has also shown promising antitumor activity across multiple treatment lines, with response rates ranging from 17% to 30% and progression-free survival ranging between 7 and 14 months (4–6), which was not influenced by the MGMT status (11–13). Moreover, the addition of bevacizumab to FOLFOX yielded a higher ORR of 52.6% in pancreatic NETs and 56.5% in grade 3 NETs (14). These evidence-based findings have been incorporated into the updated ENETS guidance papers in digestive neuroendocrine tumors, especially in high Ki-67 index or grade 3 differentiation (15–17). Despite the fact that more and more prospective and retrospective evidence have emerged, no meta-analysis has been done to address oxaliplatin-based chemotherapy in advanced WD-NETs and potential factors to predict response. Moreover, clinical data in Chinese cohort is still lacking.

To our knowledge, this meta-analysis was the first comprehensive overview of the available data for oxaliplatin-based chemotherapy in well-differentiated advanced NETs. Albeit with a small size and different baseline clinicopathological characteristics from systemic review, our single-center cohort study in Chinese WD-NETs aligned with results of the meta-analysis and confirmed that oxaliplatin-based chemotherapy was an effective and safe therapeutic option. There were three major findings from both the meta-analysis and our study.

Firstly, oxaliplatin-based chemotherapy showed promising activity in WD-NETs. Notably, this therapeutic activity appears independent of both MGMT status and chemotherapy regimens (GEMOX or FOLFOX/CAPOX). With the exception of MGMT-NET trial, there is a paucity of direct comparative studies evaluating oxaliplatin-based regimens against the others, such as CAPTEM- and STZ-based chemotherapy. There was ongoing controversy regarding the optimal choice of treatment regimens.

The MGMT-NET study was the first randomized trial comparing alkylating agents and oxaliplatin-based chemotherapy in WD-NETs (13). In the total population with both dMGMT and pMGMT, oxaliplatin-based chemotherapy had a comparable activity to alkylating agents, with the best ORR of 30.2% vs. 35.5%, PFS of 12.6 vs. 12.1 m, and OS of 48.8 vs. 50.2 m. However, in proficient MGMT NETs, the best ORR was higher in oxaliplatin-based chemotherapy than alkylating agents (38.9% vs. 11.5%), albeit the PFS was not significantly different between chemotherapy arms. Of note, CAPTEM was the most commonly used alkylating regimen (50%) in the MGMT-NET trial, while STZ-fluorouracil accounted for only 4.8%. As representative of alkylating agents, most studies on STZ-based regimens focus on well-differentiated G1-G2 pancreatic NETs, with a median ORR of 33% and PFS of 14.5 m. However, data on G3 NETs and ep-NETs were inconclusive (35, 36). By far, streptozotocin was still unavailable in China.

In contrast with alkylating agents, the efficacy of oxaliplatin-based chemotherapy was independent of MGMT status (13). In the oxaliplatin arm of the MGMT-NET trial with MGMT status assayed by IHC (n = 31, 20 dMGMT and 11 pMGMT) (13), the 3-month ORR (35.0% vs. 27.3%) and PFS (12.2 and 12.4 months in dMGMT and pMGMT) were quite similar. Although there were more G3 (66.7%) and pMGMT (66.7%) NETs in our cohort, ORR of 29.6% was comparable to that of the oxaliplatin group in MGMT-NET.

Of note, the technique to determine the MGMT status remains an issue in clinical practice. In the two prospectively randomized trials (MGMT-NET and E2211), the concordance between methylation test (using pyrosequencing) and IHC was only 58.3% and 36%, respectively. However, both methodologies showed comparable predictive value for response to temozolomide. In addition, a low MGMT score (≤50) was found as a good cutoff to predict tumor response (18). Besides that, IHC was very convenient and cost-effective. Thus, in our study, the MGMT status was assessed using IHC, which aligns with the methodology described by Cros et al. (18) and in the MGMT-NET trial (13).

With regard to the impact of different oxaliplatin-based regimens, no significant difference in ORR and PFS, respectively, was found in patients who received GEMOX or fluoropyrimidine-based chemotherapy (FOLFOX/CAPOX) in this meta-analysis. Although there was a much higher percentage of patients receiving GEMOX in the oxaliplatin group of the MGMT-NET study than this meta-analysis (81.4% vs. 17.5%), the ORR was comparable (27.9% vs. 23.9%). This consistency suggested that oxaliplatin-based chemotherapy had a similar activity whether combined with gemcitabine or fluoropyrimidine. GEMOX was often used in thoracic NETs. Of note, 35.7% (15/42) in the MGMT-NET study and 36.1% (48/133) in the GEMOX group of this meta-analysis were of thoracic origin. Even when examining the subgroup of thoracic NETs, GEMOX maintained comparable efficacy to fluoropyrimidine-based chemotherapy in terms of both ORR and PFS.

Besides that, the growing use of the FOLFIRINOX (leucovorin, 5-FU, and oxaliplatin, irinotecan) schedule was recommended in NCCN guidelines (37) in NET G3 with unfavorable biology. Two cases of G3 NETs were included in the study by Borghesani et al. (38), with the promising ORR of 50% (1/2) to mFOLFIRINOX and PFS of 15.4 m, which appeared to confer a PFS advantage over our cohort and this meta-analysis. Future prospective trials are warranted to validate the potential superiority of mFOLFIRINOX in G3 NETs and to determine the optimal Ki-67 index cutoff for predicting treatment response.

Secondly, chemotherapy appears to elicit a greater antitumor activity in patients with pNETs in both our cohort and meta-analysis. The high sensitivity to chemotherapy for pNET was not only specific to oxaliplatin-based chemotherapy but also the same to alkylating agents (6, 11, 39–41). Although the significance of higher chemotherapy response in pNET was lost when low-quality studies were excluded in a sensitivity analysis, the effect was still suggestive (OR = 0.45, 95%CI: 0.19–1.07) and approaching the threshold for marginal significance (p = 0.07) (6). Of note, there were no thoracic NETs in non-pNETs in this systemic review, and all patients received non-oxaliplatin-based chemotherapy. Besides that, only 125 non-pNETs and 111 pNETs were included in the review (6). By contrast, a larger population (521 non-pNETs and 441 pNETs) was included in this meta-analysis. A significant lower ORR in the non-pNET patients when compared to pNETs was shown, which was independent of chemotherapy regimens (GEMOX vs. FOLFOX/CAPOX). The significance also remained after a sensitivity analysis excluding those studies with low quality.

By far, the reasons why well-differentiated pNETs (mainly G1-G2) were more sensitive to cytotoxic agents than epNETs were still unclear. MGMT methylation was ever reported to be significantly higher in pancreatic NET (50%) than extra-pancreatic NET (0%–15%) (18, 42–44), which may partly explain the high efficacy of CAPTEM. However, it is hard to be correlated with response to oxaliplatin. Actually, no difference was seen in MGMT expression between pNETs (66.7%) than epNETs (66.7%) in our study. Although pNETs have more frequent mutations in the MEN1 and DAXX/ATRX genes which were up to 44% and 40%, respectively (45, 46), these molecular alterations failed to predictive of chemotherapy response (10, 47). Considering the very heterogeneous epNETs, there should be a certain proportion of epNET patients for whom chemotherapy may be very active. Thus, specific patient selection criteria (yet to be identified) other than the primary location of the NET may be more important in therapeutic decision-making.

Thirdly, tumor grade and Ki-67 had an important impact on the response rate to oxaliplatin-based chemotherapy. In the meta-analysis, the ORR of OX-CT in G3 NET was very promising and significantly higher than that in G1–2 NET (45.5% vs. 24.7%). However, PFS was worse than G1–2 NET (8.9 vs. 17 m), reflecting a generally better prognosis in low proliferative tumors. By far, the prognosis of NET G3 patients is poor, and management is still challenging due to the lack of prospective data. Chemotherapy is still most commonly used in clinical practice. Among the different chemotherapy regimens, platinum-etoposide (PE) chemotherapy, which is the standard treatment in advanced neuroendocrine carcinoma, has shown to be less active in G3 NETs (24 cases) with ORR of 24% and PFS of only 5 months (48). In one retrospective analysis with NET G3 from three German cancer centers (27), the median PFS for PE chemotherapy (6.9 months) was much lower than non-PE (9.0 m). As for the optimal non-PE regimens, both alkylating agents and oxaliplatin-based chemotherapy were effective options. When used as first line, FOLFOX was shown to have a higher response rate than CAPTEM (56.4% vs. 27.3%), which might result from a higher Ki-67 index in the FOLFOX arm. However, when looking only at G3 NETs with pancreatic origin, the PFS for FOLFOX and CAPTEM were 8.5 and15.2 m, respectively. For extra-pancreatic G3 NETs, the PFS for CAPTEM and FOLFOX were 1.6 and 6.9 (27). All of the retrospective results indicated that both primary location and grade should be taken into consideration when making decision on chemotherapy regimens.

Apart from grade, our study also analyzed the specific Ki-67 range to predict response to oxaliplatin-based chemotherapy. We found that a Ki-67 cutoff of 10% can discriminate outcomes in G2-G3 NET. No objective response has been observed with FOLFOX/CAPOX in tumors with Ki-67 less than 10%, while patients with Ki-67 more than 10% had a response as high as 36.4%. Of note, in the 10 pancreatic NETs with Ki-67 ≥40%, the ORR was 40% (data not shown). By contrast, in the study by Lamberti, none of the 34 patients with Ki-67 ≥40% who received FOLFOX/CAPOX (34 cases) had a response (49). Such a discrepancy on Ki-67 as a predictive marker was also seen in CAPTEM. One retrospective multicenter study by Wang et al. indicated that patients with a Ki-67 range of 10%–40% were the most responsive to the CAPTEM regimen (50). In one study by Jeong et al. for patients with Ki-67 >20% and <30% versus ≥30% and <55% versus ≥55%, the ORR values of CAPTEM were 18.2% versus 50.0% versus 0% (p = 0.079), respectively (51). MGMT was reported to be less frequently inactivated in NET G3 than G1/G2 (13, 52, 53), which might underlie the different Ki-67 range to predict response to alkylating agents. In our study, more deficient MGMT pts were in G2 NETs (43.4%) than G3 (27.8%), and only one patient (1/9) with Ki-67 more than 40% was MGMT-negative. Thus, for the heterogenous G3 NETs, distinct subgroups might benefit from alkylating agents and oxaliplatin-based chemotherapy, respectively.

In our cohort study, hepatic tumor burden was also predictive to the efficacy of oxaliplatin-based chemotherapy. The median PFS in patients with hepatic tumor burden ≤25% (10.2 months) was longer than those with >25% (4.7 m). The response rate in patients with low HTB also had a trend of higher ORR than those with high HTB. Hepatic tumor burden was known as a prognostic factor in neuroendocrine neoplasms. Low HTB was associated with prolonged survival and favorable treatment response (54–56).

Several limitations of our study deserve comments. Firstly, most studies in systemic review were of retrospective design, small sample size and mixed population, different treatment cycles, and both progressive or stable disease at enrollment. Hence, selection bias was inevitable, and statistical power for subgroup analyses was limited, so did our study in a Chinese cohort. Future prospective studies are warranted to extend these findings. Secondly, the relevance of ORR has been challenged in the field of NET for a long time. The response rate often rose as Ki-67 grade increased, while overall survival decreased. Similarly, the better ORR in G3 than G1–G2 to oxaliplatin-based regimens in this review did not transform significantly to the advantages of disease control, PFS, and OS. Increasingly, many trials involving NETs have evolved from the overall response rate to use PFS (ECOG-ACRIN E2211, NCT04919226, NCT03351296, and NCT02595424) or OS (NCT04365023) as the primary endpoint. Thirdly, the included studies were published between 2007 and 2024. The WHO classification of NETs was updated during this period, leaving the analysis of G3 NET inaccurate in some studies. Although 18 (66.7%) patients were classified as grade 3 in our cohort, the small size needs expansion to evaluate OX-CT efficacy in G3 NETs in the future. Fourthly, data concerning adverse events were not reported in detail in many included studies, so did our cohort. The toxicities of oxaliplatin-based chemotherapy in NETs were consistent with the previously published safety profiles (57). A drug holiday in patients with stable or responding disease after envisaged or actual cycles might be suitable. Two patients in our cohort had a treatment-free interval of 9.3 and 19.7 m, respectively.

Oxaliplatin-based chemotherapy was shown to be an effective and well-tolerated therapeutic option in advanced well differentiated neuroendocrine tumors, as evidenced by both the systemic review and our small Chinese cohort ( Figure 6 ). Notably, the efficacy was independent of MGMT status and specific chemotherapy regimens (GEMOX or fluoropyrimidine-based chemotherapy). A favorable response was seen in NETs of pancreatic origin, those with high Ki-67 more than 10% and low hepatic tumor burden (≤25%). In the future, the personalized approach to oxaliplatin-based chemotherapy will be largely based on both clinical and molecular-driven thinking within a NEN-dedicated multidisciplinary team.