Treatment for gastrointestinal and pancreatic neuroendocrine tumours: a network meta‐analysis

Martin A Walter; Cédric Nesti; Marko Spanjol; Attila Kollár; Lukas Bütikofer; Viktoria L Gloy; Rebecca A Dumont; Christian A Seiler; Emanuel R Christ; Piotr Radojewski; Matthias Briel; Reto M Kaderli

doi:10.1002/14651858.CD013700.pub2

. 2021 Nov 25;2021(11):CD013700. doi: 10.1002/14651858.CD013700.pub2

Treatment for gastrointestinal and pancreatic neuroendocrine tumours: a network meta‐analysis

Martin A Walter ¹, Cédric Nesti ², Marko Spanjol ¹, Attila Kollár ³, Lukas Bütikofer ⁴, Viktoria L Gloy ⁵, Rebecca A Dumont ¹, Christian A Seiler ², Emanuel R Christ ⁶, Piotr Radojewski ¹, Matthias Briel ⁵, Reto M Kaderli ^2,^✉

Editor: Cochrane Gynaecological, Neuro-oncology and Orphan Cancer Group

PMCID: PMC8614639 PMID: 34822169

Abstract

Background

Several available therapies for neuroendocrine tumours (NETs) have demonstrated efficacy in randomised controlled trials. However, translation of these results into improved care faces several challenges, as a direct comparison of the most pertinent therapies is incomplete.

Objectives

To evaluate the safety and efficacy of therapies for NETs, to guide clinical decision‐making, and to provide estimates of relative efficiency of the different treatment options (including placebo) and rank the treatments according to their efficiency based on a network meta‐analysis.

Search methods

We identified studies through systematic searches of the following bibliographic databases: the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library; MEDLINE (Ovid); and Embase from January 1947 to December 2020. In addition, we checked trial registries for ongoing or unpublished eligible trials and manually searched for abstracts from scientific and clinical meetings.

Selection criteria

We evaluated randomised controlled trials (RCTs) comparing two or more therapies in people with NETs (primarily gastrointestinal and pancreatic).

Data collection and analysis

Two review authors independently selected studies and extracted data to a pre‐designed data extraction form. Multi‐arm studies were included in the network meta‐analysis using the R‐package netmeta. We separately analysed two different outcomes (disease control and progression‐free survival) and two types of NET (gastrointestinal and pancreatic NET) in four network meta‐analyses. A frequentist approach was used to compare the efficacy of therapies.

Main results

We identified 55 studies in 90 records in the qualitative analysis, reporting 39 primary RCTs and 16 subgroup analyses. We included 22 RCTs, with 4299 participants, that reported disease control and/or progression‐free survival in the network meta‐analysis. Precision‐of‐treatment estimates and estimated heterogeneity were limited, although the risk of bias was predominantly low.

The network meta‐analysis of progression‐free survival found nine therapies for pancreatic NETs: everolimus (hazard ratio [HR], 0.36 [95% CI, 0.28 to 0.46]), interferon plus somatostatin analogue (HR, 0.34 [95% CI, 0.14 to 0.80]), everolimus plus somatostatin analogue (HR, 0.38 [95% CI, 0.26 to 0.57]), bevacizumab plus somatostatin analogue (HR, 0.36 [95% CI, 0.15 to 0.89]), interferon (HR, 0.41 [95% CI, 0.18 to 0.94]), sunitinib (HR, 0.42 [95% CI, 0.26 to 0.67]), everolimus plus bevacizumab plus somatostatin analogue (HR, 0.48 [95% CI, 0.28 to 0.83]), surufatinib (HR, 0.49 [95% CI, 0.32 to 0.76]), and somatostatin analogue (HR, 0.51 [95% CI, 0.34 to 0.77]); and six therapies for gastrointestinal NETs: 177‐Lu‐DOTATATE plus somatostatin analogue (HR, 0.07 [95% CI, 0.02 to 0.26]), everolimus plus somatostatin analogue (HR, 0.12 [95%CI, 0.03 to 0.54]), bevacizumab plus somatostatin analogue (HR, 0.18 [95% CI, 0.04 to 0.94]), interferon plus somatostatin analogue (HR, 0.23 [95% CI, 0.06 to 0.93]), surufatinib (HR, 0.33 [95%CI, 0.12 to 0.88]), and somatostatin analogue (HR, 0.34 [95% CI, 0.16 to 0.76]), with higher efficacy than placebo. Besides everolimus for pancreatic NETs, the results suggested an overall superiority of combination therapies, including somatostatin analogues.

The results indicate that NET therapies have a broad range of risk for adverse events and effects on quality of life, but these were reported inconsistently.

Evidence from this network meta‐analysis (and underlying RCTs) does not support any particular therapy (or combinations of therapies) with respect to patient‐centred outcomes (e.g. overall survival and quality of life).

Authors' conclusions

The findings from this study suggest that a range of efficient therapies with different safety profiles is available for people with NETs.

Plain language summary

Treatment options for neuroendocrine tumours

Review question

We reviewed the evidence on safety and efficacy of therapies for neuroendocrine tumours (NETs) in the gastrointestinal tract and the pancreas to provide a ranking of these treatment options.

Background

NETs are a varied group of rare cancers, which can occur anywhere in the body. However, most neuroendocrine tumours derive from the gastrointestinal tract or the pancreas. There are many types of NETs with different growth rates and symptoms. While some NETs produce excess hormones, others do not release hormones, or not enough to cause symptoms. The treatment options, as well as their combinations and sequencing, depend on the type of tumour, its location, aggressiveness, and whether it produces excess hormones.

Until now, no clear recommendations could be given about which NET therapies were the most effective and caused the fewest adverse events. We used statistical methods to compare all therapies with each other based on the available information.

Study characteristics

We included 22 randomised controlled trials (studies in which participants are randomly assigned to treatment groups), published before 11 December 2020, with a total of 4299 people. There were differences in tumour location (gastrointestinal and pancreatic), tumour type, sample size, treatments, and quality of the research between the studies.

Key results

This analysis suggests, in general, a superiority of combination therapies, including somatostatin‐like medications, in both gastrointestinal and pancreatic NETs. However, in pancreatic NETs, everolimus was the most effective therapy with the highest certainty of evidence compared to the other treatments. Furthermore, the results indicate that NET therapies have a broad range of risk for adverse events and effects on quality of life. Because disease is often advanced at presentation and treatment is often given with the intent to control and shrink disease, rather than be ultimately curative, treatment adverse events and quality of life are key considerations.

Quality of evidence

We rated the certainty of the evidence as high to low for the different therapies. An overall ranking of the treatments (and combinations) was not possible. In order to make an informed decision, advantages and disadvantages of each therapy, including its risks for adverse events and effects on quality of life, have to be balanced against each other. Evidence from this network meta‐analysis (and underlying RCTs) does not support any particular therapy (or combinations of therapies) with respect to patient‐centred outcomes (e.g. overall survival and quality of life).

Summary of findings

Background

Description of the condition

Neuroendocrine tumours (NETs), sometimes referred to as carcinoid tumours, are a heterogenous group of malignancies (cancers) that arise from cells of the endocrine (hormonal) and neurological systems. They have an estimated overall 20‐year limited‐duration prevalence (number of people alive on a certain day who were diagnosed with a NET within the previous 20‐year period) of 171,321 and a yearly age‐adjusted incidence of 6.98 cases per 100,000 according to the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) 18 registry (Dasari 2017). A population‐based study found a 6.4‐fold increase in incidence between 1973 and 2012 (Dasari 2017). NETs are more common at higher age, with an incidence among people 65 years or older of 25 per 100,000. About 61.0% of NETs derive from the gastrointestinal tract or the pancreas (Lawrence 2011), and accordingly these tumours are called gastroenteropancreatic NET (GEP‐NET). Other sites for primary NET include lungs, thyroid, ovaries, cervix, pituitary, and adrenal glands (Hallet 2015).

The relative frequency and annual incidence rate per 100,000 of GEP‐NETs differ site by site and, in some cases, change over time and are different between countries and continents (Fraenkel 2014). NETs of the rectum are the most common in east Asia and the USA, while small intestinal NETs are the most common in males, and appendiceal NETs the most common in females in the UK (Fraenkel 2012; Fraenkel 2014). Racial discrepancies have been found in the US SEER registry, with small intestinal NETs being found more often in African‐Americans than in the white population (DePalo 2019).

Most GEP‐NETs are sporadic, but approximately 5% arise in the context of cancer predisposition syndromes (Clift 2020). Neuroendocrine tumours, especially those of the pancreas (pNET), may be associated with familial syndromes. Multiple endocrine neoplasia type 1 (MEN 1) is the most common familial syndrome associated with NET, while Von Hippel‐Lindau syndrome, neurofibromatosis type‐1 and tuberous sclerosis are rarer.

Depending on localisation and stage of the disease, they present with a broad clinical spectrum, from asymptomatic people with an incidental discovery on imaging to florid endocrinopathy. Up to 30% to 40% of GEP‐NETs may be secretory (i.e. 'functional'), releasing a variety of hormones and hormone‐like substances (Clift 2020). Serotonin‐secreting small bowel NETs may lead to cardiac valve fibrosis (carcinoid heart disease) as a consequence of hormone hyper‐secretion.

The diagnosis of GEP‐NETs is usually based on a histopathology that demonstrates neuroendocrine features, such as positive immunohistochemical staining for synaptophysin and chromogranin A. The grading of GEP‐NETs, on the other hand, is based on the mitotic index using Ki‐67 immunohistochemistry (which estimates how many cells are dividing within a tumour and how quickly it might grow). The World Health Organization (WHO) classification divides NETs according to their proliferative activity into grade 1 (Ki‐67 index ≤ 2%) and grade 2 (Ki‐67 index 3% to 20%). Based on their morphological characteristics, grade 3 tumours are subdivided into well differentiated NET and poorly differentiated neuroendocrine carcinomas, both with Ki‐67 index > 20% (Klimstra 2019). The grading aids in the prognostication of survival: the five‐year survival rates of grade 1, 2 and 3 NETs are 96%, 73% and 28%, respectively (Ramage 2012).

Description of the intervention

Tumour growth, treatment and outcome vary considerably with the location of the primary lesions, as well as with their grade, extension, and stage (Lawrence 2011; Modlin 2008; Yao 2008 (2)). A broad spectrum of therapeutic options permits staged disease management with various treatment combinations and sequencing. This approach, however, requires a highly interdisciplinary and dynamic approach, which typically involves physicians of various specialties who work in concert to manage these often‐complex cases and select a treatment strategy from an array of available options.

Management strategies depend on primary tumour, locoregional and distant metastases, differentiation, tumour‐related symptoms, syndromes and presence of carcinoid heart disease. Depending on primary tumour size and site, NETs are treated surgically whenever feasible, as this is the only potentially curative treatment (Yao 2008 (2)). In metastatic, well differentiated NETs, somatostatin analogues (SSA), and interferon alpha (IFN) as a possible second‐line therapy, are a cornerstone in the palliative setting, as effective means of improving quality of life (QoL) and delaying disease progression (Cives 2014; Clift 2020). More recently, molecularly targeted drugs like the mTOR‐inhibitor everolimus, the multi‐targeted receptor tyrosine kinase inhibitor sunitinib, and the vascular endothelial growth factor (VEGF) antibody bevacizumab have been introduced into the clinical setting following trials demonstrating efficacy in people with progressive NET (Kunz 2013; Pavel 2016; Yao 2017). The radiolabelled somatostatin receptor ligand lutetium‐177‐DOTATATE also recently demonstrated a benefit over treatment with somatostatin analogues alone in people with progressive NET (Strosberg 2017). Liver‐directed therapies further broaden the therapeutic landscape (Pavel 2016). In advanced grade 3 pNET and advanced symptomatic or progressive grade 1 or 2 pNET, systemic chemotherapy with streptozocin‐ or temozolomide‐based regimens is the first choice of treatment. In grade 3 NEC, platinum‐based chemotherapy is recommended as a first‐line therapy (Pavel 2016).

Why it is important to do this review

Several available therapies have demonstrated efficacy in terms of disease control and/or progression‐free survival in randomised controlled trials (RCTs). However, translation of these results into improved care faces several challenges, as several therapies were compared with placebo only and a direct comparison of the most pertinent therapies is incomplete (Kaderli 2019). In a previous systematic review and network meta‐analysis on pNETs and neuroendocrine tumours of the gastrointestinal tract (GI‐NETs), we found several monotherapies that were superior to placebo, including everolimus, interferon, and sunitinib in pNETs and somatostatin analogues in pNETs and GI‐NETs (Kaderli 2019). Furthermore, the results suggested a superiority of combination therapies, especially those including somatostatin analogues. On the other hand, NET therapies have a broad range of risk for adverse events and effects on QoL, which need to be considered while choosing the appropriate treatment. A systematic comparison of benefits and harms of all currently available therapeutic modalities will allow informed clinical decision‐making for clinicians, patients and policy makers.

Furthermore, there is ongoing research in the treatment of NETs. Surufatinib has demonstrated a higher progression‐free survival in GI‐NETs in the SANET‐ep trial (Xu 2020 (ep)) and in pNET in the SANET‐p trial (Xu 2020 (p)). New results for axitinib and somatostatin analogue are expected in GI‐NET (AXINET trial, NCT01744249), for everolimus and streptozocin plus fluorouracil in pNET (SEQTOR trial, NCT02246127), and for lutetium‐177 (¹⁷⁷Lu)‐DOTATATE and everolimus both in GI‐NET and pNET (COMPETE trial, NCT03049189). It is, therefore, vital to provide a regularly updated systematic review and network meta‐analysis for clinical decision‐making based on the best available and most recent evidence.

Objectives

To evaluate the safety and efficiency of therapies for NETs, to guide clinical decision‐making, and to provide estimates of relative efficiency of the different treatment options (including placebo) and rank the treatments according to their efficiency based on a network meta‐analysis.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs), including randomised controlled cross‐over trials.

If a post hoc subgroup analysis was available and reported disease control after 12 months and/or progression‐free survival for either pNET or GEP‐NET only, the subgroup analysis was used for the network meta‐analysis instead of the main study including more than one type of NETs.

Types of participants

People of any age with any type and any stage of GEP‐NETs.

Types of interventions

We included RCTs comparing at least two treatments of any kind (including usual care or placebo) in NETs, administered in any way.

Examples of treatments include the mechanistic target of rapamycin inhibitor everolimus (Yao 2016), the multi‐targeted receptor tyrosine kinase inhibitor sunitinib (Raymond 2011), the vascular endothelial growth factor (VEGF) antibody bevacizumab (Yao 2017), the radiolabelled somatostatin analogue lutetium‐177 (177Lu)‐dotatate (Strosberg 2017), and new combinations of previously established therapies (Pavel 2011). Several therapies were compared only with placebo, while others were directly compared.

Every individual drug or drug combination, as well as placebo, represent individual nodes in the network meta‐analysis. Due to the low number of included studies, we grouped together all different somatostatin analogues, as well as all different intervention doses, modalities, and administration frequencies.

Types of outcome measures

Primary outcomes

Disease control after 12 months
Progression‐free survival

Secondary outcomes

Overall survival
Occurrence of adverse events according to the treatment applied (grades 3 to 4, any grade)
Quality of life (QoL)

Disease control is defined as the sum of complete response, partial response and stable disease, or as the total minus the number disease progressions. Progression‐free survival is the length of time during and after the treatment, that a patient lives with the disease, but it does not grow. We used unblinded, investigator‐assessed progression‐free survival outcomes. Adverse events were classified according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI 2010): Grade 1 corresponds to mild, grade 2 to moderate, grade 3 to severe or medically significant, and grade 4 to life‐threatening adverse events. Effects on QoL were quantified based on the QoL Questionnaire C30 of the European Organization for Research and Treatment of Cancer (EORTC QLQ‐30) (Aaronson 1993).

Search methods for identification of studies

Electronic searches

We identified trials through systematic searches of the following bibliographic databases on 11 December 2020:

Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 12) in the Cochrane Library;
MEDLINE via Ovid (January 1947 to 11 December 2020);
Embase.com (January 1947 to 11 December 2020).

In addition, we checked trial registries (ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform Search Portal [apps.who.int/trialsearch/]) for ongoing or unpublished eligible trials and manually searched for abstracts from scientific and clinical meetings related to NETs in 2019 and 2020 (annual ENETS conference and neuroendocrine tumour symposium of the NANETS).

We searched all databases from 1 January 1947, until present, and imposed no restriction on language of publication (Appendix 1; Appendix 2; Appendix 3).

Searching other resources

We scanned the reference lists of the included RCT reports and relevant review articles for additional references.

Data collection and analysis

Selection of studies

With two review authors working in duplicate, we independently screened all abstracts and obtained the full‐text report of potentially relevant studies. Subsequently, we screened all potentially relevant studies in the same way. Any discordance was resolved by a third review author.

Data extraction and management

We used a data collection form for study characteristics and outcome data which has been piloted in our previous systematic review and network meta‐analysis on therapeutic options for neuroendocrine tumours (Kaderli 2019). One of the review authors extracted study characteristics from included studies, and a second review author verified the extractions. We extracted the following study characteristics.

Characteristics of included trials: first author, year of publication, study origin, type of treatments, median duration and median follow‐up of each treatment, percentage of people with complete follow‐up, availability of a sample size calculation, and number of participants randomised for each treatment.
Participant data: separately for each treatment: primary tumour site, tumour grading, presence of metastases and functional tumours, percentage of female participants and the participants' median/mean age; main primary tumour (pNET and/or GI‐NET) for all treatments.
Clinical outcomes: complete response, partial response, stable disease, disease control, disease progression, investigator‐assessed progression‐free survival, median overall survival, occurrence of adverse events (grade 3 to 4, any grade), and QoL.

Any discordance was resolved by a third review author. Data were entered into Review Manager software (RevMan 2014) and checked by a second review author for accuracy.

Due to the well‐defined patient characteristics, we did not expect significant effect modifiers and, due to the low number of included studies, we could not systematically analyse effect modifiers.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias for each RCT, using the Cochrane risk of bias tool (Higgins 2011), which utilises the following domains.

Random sequence generation
Allocation concealment
Blinding of participants and personnel
Blinding of outcome assessment
Completeness of outcome data
Selectivity of reporting
Other bias (including baseline imbalance, protocol deviations, inappropriate influence of funders)

We provided a summary risk of bias assessment for each study using the method outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Each domain was rated as low (bias is unlikely to seriously alter the results), high (bias is likely to seriously weaken confidence in results), or unclear risk of bias. All discordance was resolved by a third review author.

Measures of treatment effect

We used odds ratios as effect measures for disease control after 12 months and hazard ratios as effect measures for progression‐free survival, both accompanied by 95% confidence intervals (95% CIs). We applied a continuity correction for studies with a zero cell count by adding 0.5 to all cell frequencies. We summarised all results using forest plots with combined effect estimates and size of squares proportional to the inverse of the standard errors. Due to the low number of included studies and the heterogeneity of secondary outcomes, we presented these outcomes for each intervention (if available) using descriptive statistics — i.e. number and percentage of adverse events, and mean and standard deviation of the change of QoL.

We ranked treatments based on P scores, measuring the extent of certainty that a treatment is better than another one, averaged over all competing treatments (Rücker 2015).

Unit of analysis issues

The analysis was made at the individual allocation level.

Multi‐arm trials were included in the network meta‐analysis. The correlation of treatment effects on different comparisons was accounted for by re‐weighting all comparisons of each multi‐arm study (Rücker 2012; Rücker 2014).

We included cross‐over trials in the qualitative analysis. However, they were excluded from the network meta‐analysis due to the inappropriateness of the study design: including only the first intervention period of a cross‐over trial discards more than half of the information in the study.

Dealing with missing data

We contacted authors of included RCT reports for information on unreported outcomes and missing outcome data in their studies.

If a RCT report did not report hazard ratios and further data could not be obtained by contacting authors, we estimated the hazard ratios from reconstructed Kaplan‐Meier curves (if available) by using a Cox proportional hazard model.

Assessment of heterogeneity

We assessed heterogeneity using all pairwise comparisons available from more than one trial. We calculated the between‐study variance Ƭ², the within‐design component of Cochran's Q (i.e. the weighted sum of squared differences between pairwise comparisons from multiple trials) and the associated I² (percentage of variation across studies due to heterogeneity rather than chance). If quantification of heterogeneity was not possible (i.e. if there was no comparison done in more than one trial), we fitted fixed‐effect models; otherwise, we used random‐effects models.

We assessed homogeneity and transitivity based on the distribution of neuroendocrine tumour types, and the differences in doses and application route, especially for somatostatin analogues.

We assessed inconsistency using closed loops within the network (if any) and calculated the between‐design component of Cochran's Q and the associated I². In addition, we performed a netsplit analysis and compared direct and indirect estimates via a ratio of odds or hazard ratios.

We calculated the total Cochran's Q as the sum of between‐ and within‐designs component and the associated I².

Assessment of reporting biases

To assess the risk for reporting bias, we first searched for a protocol for each of the included studies. For this, we went through the reference lists of corresponding published articles. If there was no reference to a protocol, we searched PubMed, Embase, and the internet for a protocol. If a protocol was available, we compared the mentioned outcomes and planned statistical analyses in the protocol with those in the published report. If no protocol was available, we used information from a corresponding registry entry of the included study to compare planned outcomes and analyses with those in the published report. If neither a protocol, nor a registry entry was available, we compared the outcomes and described analyses in the methods section of the published report with those reported in the results section. Any unexplained differences between the protocol, registry entry, or methods section and the reported results provided evidence for an increased risk of reporting bias of an included study.

If there were 10 or more included studies for individual pairwise meta‐analyses, we created funnel plots for visual inspection to detect potential asymmetry.

Data synthesis

We separately analysed two different outcomes (disease control and progression‐free survival) and two types of NET (pNET and GI‐NET) in four network meta‐analyses. The NET types were distinguished to ensure that the selected studies were similar except for the interventions being compared. If a study included several NET types, we included the respective subgroup analyses (if available): for pNET, one subgroup analysis was included for the analysis of progression‐free survival (Phan 2015 (2) instead of Caplin 2014) and, for GI‐NET, one subgroup analysis was included for the analysis of disease control and progression‐free survival (Castellano 2013 instead of Pavel 2011) and two subgroup analyses were included for the analysis of progression‐free survival (Dasari 2015 instead of Caplin 2014 and Singh 2018 (1) instead of Yao 2016). Otherwise, we relied on expert opinion whether or not to include the study and used sensitivity analyses to assess the effect of the decision.

Before including an intervention in the network meta‐analysis, we assessed the respective study populations critically in terms of the transitivity assumption. Interventions only given to a subset of participants (i.e. those critically ill) were not included in a sensitivity analysis. However, since the network is currently very sparse, the benefit of additional studies might outweigh a certain risk of violation of the transitivity assumption. The comparison among all interventions (including placebo) were of interest and we would not define a decision and a supplementary set. However, if more data become available, we might focus on a specific set of interventions.

Because the network is sparse, we merged similar interventions, i.e. different doses, administration intervals and routes of application of the same compound. When more data become available, we will consider splitting nodes if the effects are suspected to be different.

We performed the network meta‐analyses with a frequentist approach using R‐package (R Core 2019) netmeta (Rücker 2021). If quantification of heterogeneity was possible, i.e. if there were pairwise comparisons included in more than one trial, we used random‐effects models. Otherwise, we used fixed‐effect models. Validity of the network in terms of consistency was assessed quantitatively by comparing direct and indirect estimates for each loop of the network and qualitatively using GRADE (as described in section Assessment of heterogeneity).

Subgroup analysis and investigation of heterogeneity

In view of the small number of RCTs included in this review, we refrained from any subgroup analysis, including subgroup analysis based on tumour grading, since the separate analysis for each treatment included in a RCT was frequently missing.

If there was evidence for heterogeneity, we assessed participant and trial characteristics for a potential source of the heterogeneity.

Sensitivity analysis

Currently, the network is very sparse and we were not able to undertake sensitivity analyses. If sufficient trials would have been identified, we would have considered several sensitivity analyses for the primary outcomes. We would, for example, only use low risk of bias trials (trials without a high risk for selection, performance, detection, attrition, reporting or other biases), exclude trials with a mixture of different types of NETs and use alternative or no merging of nodes. We would have also considered different analytical approaches, such as fixed‐effect only, or a Bayesian instead of the specified frequentist approach (e.g. using R package BUGSnet (Béliveau 2019)).

Summary of findings and assessment of the certainty of the evidence

We used the GRADE approach to assess confidence in estimates of effect (certainty of evidence) associated with specific comparisons, including estimates from direct, indirect, and final network meta‐analysis (Brignardello‐Petersen 2018; Puhan 2014; Salanti 2014). Our confidence assessment addressed risk of bias (limitations in study design and execution), inconsistency (heterogeneity of estimates of effects across trials), indirectness (differences in population, interventions, or outcomes to the target of the network meta‐analysis) and imprecision (e.g. wide 95% confidence intervals including or close to the null effect). Limitations in any of these domains resulted in a decrease of the certainty of evidence from high to moderate, low, or very low‐certainty by ‐1 (serious concern) or ‐2 (very serious concern). We based indirect evidence on the most dominant loops (i.e. the shortest path between two treatments) and potentially rated it down for intransitivity (differences in study characteristics that may modify treatment effect in the direct comparisons along the path). We obtained the final network meta‐analysis confidence rating from the higher of the direct and indirect rating excluding imprecision and we rated it down for imprecision and incoherence (difference between direct and indirect estimates).

All studies and study arms used for the network meta‐analyses had included adult people with advanced GEP‐NET that were in need of and eligible for systematic therapies, supporting the transitivity assumption of the network meta‐analyses.

In the summary of findings tables, we included estimates of effects, ranking and certainty of evidence for different treatment options compared with placebo for disease control and progression‐free survival in pNET and GI‐NET.

Results

Description of studies

See: Characteristics of included studies and Characteristics of excluded studies.

Results of the search

In our previous systematic review and network meta‐analysis on pNETs and GI‐NETs with the same search methods, we included 38 studies in the qualitative synthesis (30 primary studies and 8 subgroup analyses) (Kaderli 2019). The previously published searches on 27 November 2015 and 2 March 2018 led to the identification of 7243 records (Kaderli 2019). Following de‐duplication across the databases, the combined total yield of the updated search on 11 December 2020 was 1058 records:

CENTRAL: 255 records
MEDLINE (Ovid): 546 records
Embase: 257 records

Two additional records were added through scanning the reference lists of included RCT reports. After reading the abstracts, we excluded 991 records because they did not match the inclusion criteria. After assessing the full text, we excluded 23 records. In all, we included 55 studies in the qualitative analysis (39 primary RCTs and 16 subgroup analyses). A total of 22 studies reported disease control and/or progression‐free survival and were included in the network meta‐analyses (see Figure 1).

Included studies

We included 55 studies in 90 records in the qualitative analysis, reporting 39 primary RCTs (Arnold 2005; Bergsland 2020; Caplin 2014; Elf 2018; Faiss 2003; Jacobsen 1995; Kölby 2003; Kulke 2016; Kulke 2017 (1); Kulke 2017 (2); Lange 1992; Lepage 2020; Liu 2020; Maire 2012; Meyer 2014; Moertel 1980; Moertel 1992; O'Toole 2000; Öberg 1989; Pavel 2011; Pavel 2018 (1); Pavlakis 2020; Raymond 2011 (1); Rinke 2009; Sakata 2006; Salazar 2018; Saslow 1998; Soulen 2020; Strosberg 2017; Van Der Zwan 2018; Vinik 2016; Wolin 2015; Xu 2020 (ep); Xu 2020 (p); Yao 2008 (1); Yao 2011; Yao 2016; Yao 2017; Zhang 2020) and 16 subgroup analyses (Anthony 2012; Castellano 2013; Dasari 2015; Di Gialleonardo 2020; Fisher 2016; Ito 2012; Lombard‐Bohas 2015; Phan 2015 (1); Phan 2015 (2); Pusceddu 2018; Raymond 2011 (2); Singh 2018 (1); Strosberg 2011; Strosberg 2020; Wolin 2016; Yao 2019) (see Characteristics of included studies for details. Overall, 4654 patients were recruited and 26 different therapies were evaluated, including biotherapies, chemotherapies, targeted drugs, locoregional therapies, surgical treatment, and targeted radiopeptide therapy.

A total of 22 RCTs, which included 4299 patients, reported disease control and/or progression‐free survival and were included in the network meta‐analysis (Arnold 2005; Caplin 2014; Castellano 2013; Dasari 2015; Faiss 2003; Kölby 2003; Kulke 2016; Kulke 2017 (1); Öberg 1989; Pavel 2011; Phan 2015 (2); Raymond 2011 (1); Rinke 2009; Salazar 2018; Singh 2018 (1); Strosberg 2017; Xu 2020 (ep); Xu 2020 (p); Yao 2008 (1); Yao 2011; Yao 2016; Yao 2017).

Eighteen of 22 RCTs included in the network meta‐analysis were industry‐sponsored (Arnold 2005; Caplin 2014; Castellano 2013; Dasari 2015; Faiss 2003; Kulke 2017 (1); Pavel 2011; Phan 2015 (2); Raymond 2011 (1); Rinke 2009; Salazar 2018; Singh 2018 (1); Strosberg 2017; Xu 2020 (ep); Xu 2020 (p); Yao 2008 (1); Yao 2011; Yao 2016).

Excluded studies

During the first phase of record selection, we screened and excluded 991 records, which were not investigating therapeutic procedures in NET or did not fulfil the criteria of an RCT. Twenty‐three of the remaining 69 records were excluded after assessing the full‐text articles. They did not fulfil the criteria of an RCT, were duplicate reports or were not investigating therapeutic procedures in NET (see Characteristics of excluded studies for details).

Risk of bias in included studies

Summaries of the risk of bias for each domain and as percentages across all studies are presented in Figure 2 and Figure 3.

Allocation

Random sequence generation

Twenty‐nine studies described a random component in the sequence generation process and were at low risk of selection bias. The other 26 studies had a randomised controlled trial study design; but in 25 studies there was no further report on the sequence generation process and in one study the randomisation was performed by the study drug supplier (Jacobsen 1995). For these studies, we judged the risk of selection bias as unclear.

Allocation concealment

Twenty‐five studies reported on the method to conceal allocation and were at low risk of selection bias. Twenty‐eight studies provided no further information addressing allocation concealment and were considered to be at unclear risk of selection bias. Two studies without information on allocation concealment and identical numbers of people in all treatment groups were considered to be at unclear risk of selection bias (Kulke 2017 (2); Yao 2008 (1)).

Blinding

Blinding of participants and personnel (performance bias)

Twenty‐nine studies were double‐blinded and were at low risk of performance bias. Six studies (Kulke 2017 (1); Pavlakis 2020; Strosberg 2017; Strosberg 2020; Yao 2017; Zhang 2020) were designed as open‐label studies and in 14 studies participants and/or personnel were not blinded. They were considered to be at high risk of performance bias. Six studies provided no information and were at unclear risk of performance bias.

Blinding of outcome assessment (detection bias)

Eighteen studies reported blinding of outcome assessors and were at low risk of detection bias. Of the remaining studies, 29 studies were at unclear risk of detection bias due to missing information on the blinding of outcome assessment and eight studies were at high risk of detection bias due to a lack of evidence for a blinding of the outcome assessment.

Incomplete outcome data

Thirty‐five studies were at low risk and 16 studies were at unclear risk of attrition bias due to missing information. In three studies, a significant number of people were excluded after randomisation (Moertel 1992; O'Toole 2000; Zhang 2020) and in one study (Öberg 1989) a group cross‐over was performed without additional information, whether intention‐to‐treat or analysis per‐protocol was performed. These four studies were considered to be at high risk of attrition bias.

Selective reporting

Thirty‐two studies published a study protocol or reported all results of the endpoints stated in the methods section and were at low risk of reporting bias. Sixteen studies provided little information on primary or secondary endpoints and their definition and were judged to be at low or unclear risk for reporting bias, depending on a study‐level judgement. In seven studies, not all stated endpoints were reported (Meyer 2014; Moertel 1980; Singh 2018 (1); Strosberg 2017; Strosberg 2020; Yao 2016; Yao 2019). Hence, we judged the risk of reporting bias for these studies as high.

Other potential sources of bias

Two studies were at high risk for other potential sources of bias due to the use of investigator‐dependent measurement methods (Moertel 1980; Moertel 1992).

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4

Summary of findings 1. Estimates of effects, ranking, and certainty of evidence for different treatment options compared with placebo for disease control in pancreatic neuroendocrine tumours (pNET).

Total studies: 9 Total participants: 1757	Included trials	Median follow‐up (months)¹	Relative effect (95% CI)	Anticipated absolute effects²		Certainty of evidence³	P‐score⁴
Total studies: 9 Total participants: 1757	Included trials	Median follow‐up (months)¹	Relative effect (95% CI)	*Disease control with intervention*	*Disease control without intervention*	Certainty of evidence³	P‐score⁴
Everolimus (3 RCTs; 632 participants)	Kulke 2017 (1); Salazar 2018; Yao 2011	17	OR 3.29 (2.21 to 4.90)	80%	55%	Moderate*	0.83
Everolimus + SSA (2 RCTs; 589 participants)	Kulke 2017 (1); Pavel 2011	not reported	OR 2.89 (1.61 to 5.19)	84%	65%	Moderate‡	0.73
Interferon + SSA (2 RCTs; 171 participants)	Arnold 2005; Faiss 2003	not reported	OR 2.88 (1.16 to 7.13)	27%	11%	Very low*,‡,¶	0.71
Interferon (1 RCT; 66 participants)	Faiss 2003	not reported	OR 2.58 (0.75 to 8.81)	35%	17%	Very low**,‡,§§	0.63
SSA (4 RCTs, 804 participants)	Arnold 2005; Caplin 2014; Faiss 2003; Pavel 2011	not reported	OR 2.36 (1.43 to 3.88)	67%	47%	Moderate‡	0.56
Surufatinib (1 RCT; 172 participants)	Xu 2020 (p)	19	OR 1.99 (1.02 to 3.88)	74%	59%	High	0.48
Sunitinib (1 RCT; 171 participants)	Raymond 2011 (1)	60	OR 1.72 (0.91 to 3.27)	72%	60%	Low*,§	0.39
Placebo (4 RCTs; 957 participants)	Caplin 2014; Raymond 2011 (1); Xu 2020 (p); Yao 2011	27	Reference comparator	53%	‐	Reference	0.12
Dactolisib (1 RCT; 62 participants)	Salazar 2018	not reported	OR 0.56 (0.13 to 2.37)	61%	74%	Very low*,§§	0.06

Total studies: 10 Total participants: 2113	Included trials	Median follow‐up (months)¹	Relative effect (95% CI)	Anticipated absolute effect²		Certainty of evidence³	P‐score⁴
Total studies: 10 Total participants: 2113	Included trials	Median follow‐up (months)¹	Relative effect (95% CI)	*Median PFS with intervention (months)*	*Median PFS without intervention (months)*	Certainty of evidence³	P‐score⁴
Everolimus (3 RCT; 632 participants)	Kulke 2017 (1); Salazar 2018; Yao 2011	17	HR 0.36 (0.28 to 0.46)	12	4	Moderate*	0.75
Interferon + SSA (2 RCTs; 468 participants)	Faiss 2003; Yao 2017	not reported	HR 0.34 (0.14 to 0.80)	15	5	Very low**,‡	0.74
Everolimus + SSA (3 RCTs; 739 participants)	Kulke 2016; Kulke 2017 (1); Pavel 2011	not reported	HR 0.38 (0.26 to 0.57)	16	6	Low‡	0.68
Bevacizumab + SSA (1 RCT; 402 participants)	Yao 2017	not reported	HR 0.36 (0.15 to 0.89)	17	6	Very low**,‡,¶	0.65
Interferon (1 RCT; 66 participants)	Faiss 2003	not reported	HR 0.41 (0.18 to 0.94)	not reported	‐	Very low**,‡	0.58
Sunitinib (1 RCT; 171 participants)	Raymond 2011 (1)	60	HR 0.42 (0.26 to 0.67)	11	5	Moderate*	0.56
Everolimus + bevacizumab + SSA (1 RCT; 150 participants)	Kulke 2016	not reported	HR 0.48 (0.28 to 0.83)	17	8	Very low**,¶	0.42
Surufatinib (1 RCT; 172 participants)	Xu 2020 (p)	19	HR 0.49 (0.32 to 0.76)	11	5	High	0.41
Dactolisib (1 RCT; 62 participants)	Salazar 2018	not reported	HR 0.55 (0.25 to 1.21)	8	4	Low*,§	0.35
SSA (3 RCTs; 586 participants)	Faiss 2003; Pavel 2011; Phan 2015 (2)	not reported	HR 0.51 (0.34 to 0.77)	11	6	Moderate	0.33
Placebo (4 RCTs; 844 participants)	Phan 2015 (2); Raymond 2011 (1); Xu 2020 (p); Yao 2011	27	Reference comparator	6	‐	Reference	0.01

Total studies: 11 Total participants: 1338	Included trials	Median follow‐up (months)¹	Relative effect (95% CI)	Anticipated absolute effects²		Certainty of evidence³	P‐score⁴
Total studies: 11 Total participants: 1338	Included trials	Median follow‐up (months)¹	Relative effect (95% CI)	*Disease control with intervention*	*Disease control without intervention*	Certainty of evidence³	P‐score⁴
Bevacizumab + SSA (1 RCT; 44 participants)	Yao 2008 (1)	not reported	OR 45.0 (3.32 to 609)	95%	32%	Very low*,††,‡,¶¶,§	0.91
177‐Lu‐DOTATATE + SSA (1 RCT; 229 participants)	Strosberg 2017	14	OR 30.4 (8.19 to 113)	80%	12%	Very low**,¶,§	0.90
Everolimus + SSA (1 RCT; 39 participants)	Castellano 2013	not reported	OR 15.1 (2.55 to 88.9)	63%	10%	Very low‡,¶,§	0.78
Interferon + SSA (4 RCTs; 283 participants)	Arnold 2005; Faiss 2003; Kölby 2003; Yao 2008 (1)	76	OR 5.71 (1.90 to 17.2)	48%	14%	Very low*,††,‡,¶	0.60
Interferon (2 RCTs; 86 participants)	Faiss 2003; Öberg 1989	7	OR 4.03 (0.86 to 18.8)	55%	23%	Very low**,‡,¶,§§	0.48
Surufatinib (1 RCT; 198 participants)	Xu 2020 (ep)	14	OR 3.50 (1.21 to 10.1)	84%	61%	Moderate‡	0.45
SSA (7 RCTs; 796 participants)	Arnold 2005; Caplin 2014; Castellano 2013; Faiss 2003; Kölby 2003; Rinke 2009; Strosberg 2017	87	OR 2.93 (1.36 to 6.32)	43%	21%	Moderate‡	0.37
Everolimus (1 RCT; 302 participants)	Yao 2016	21	OR 2.53 (0.95 to 6.79)	82%	65%	Very low*,‡,§	0.35
Placebo (4 RCT; 789 participants)	Caplin 2014; Rinke 2009; Xu 2020 (ep); Yao 2016	35	Reference comparator	53%	‐	Reference	0.11
Streptozocin + 5‐FU (1 RCT; 20 participants)	Öberg 1989	12	OR 0.13 (0.00 to 4.58)	40%	83%	Very low**,‡,¶,§§	0.04

Dactolisib	0.17 (0.04 to 0.68)	0.19 (0.04 to 0.87)	0.22 (0.03 to 1.41)	0.19 (0.04 to 1.04)	0.56 (0.13 to 2.37)	0.24 (0.05 to 1.07)	0.32 (0.07 to 1.58)	0.28 (0.06 to 1.38)
5.89 (1.46 to 23.7)	Everolimus	1.14 (0.63 to 2.04)	1.27 (0.36 to 4.49)	1.14 (0.44 to 2.95)	3.29 (2.21 to 4.90)	1.40 (0.79 to 2.46)	1.91 (0.90 to 4.06)	1.65 (0.76 to 3.61)
5.18 (1.14 to 23.5)	0.88 (0.49 to 1.58)	Everolimus + SSA	1.12 (0.33 to 3.79)	1.00 (0.41 to 2.46)	2.89 (1.61 to 5.19)	1.23 (0.77 to 1.97)	1.68 (0.71 to 4.00)	1.46 (0.60 to 3.54)
4.62 (0.71 to 30.2)	0.78 (0.22 to 2.76)	0.89 (0.26 to 3.02)	Interferon	0.90 (0.29 to 2.79)	2.58 (0.75 to 8.81)	1.09 (0.36 to 3.37)	1.50 (0.37 to 5.98)	1.30 (0.32 to 5.26)
5.16 (0.96 to 27.8)	0.88 (0.34 to 2.26)	1.00 (0.41 to 2.43)	1.12 (0.36 to 3.47)	Interferon + SSA	2.88 (1.16 to 7.13)	1.22 (0.57 to 2.61)	1.67 (0.55 to 5.07)	1.45 (0.47 to 4.47)
1.79 (0.42 to 7.64)	0.30 (0.20 to 0.45)	0.35 (0.19 to 0.62)	0.39 (0.11 to 1.33)	0.35 (0.14 to 0.86)	Placebo	0.42 (0.26 to 0.70)	0.58 (0.31 to 1.10)	0.50 (0.26 to 0.98)
4.22 (0.94 to 19.0)	0.72 (0.41 to 1.26)	0.81 (0.51 to 1.31)	0.91 (0.30 to 2.81)	0.82 (0.38 to 1.75)	2.36 (1.43 to 3.88)	SSA	1.37 (0.61 to 3.08)	1.19 (0.51 to 2.73)
3.09 (0.63 to 15.1)	0.52 (0.25 to 1.11)	0.60 (0.25 to 1.42)	0.67 (0.17 to 2.67)	0.60 (0.20 to 1.82)	1.72 (0.91 to 3.27)	0.73 (0.32 to 1.65)	Sunitinib	0.87 (0.34 to 2.19)
3.56 (0.72 to 17.6)	0.60 (0.28 to 1.32)	0.69 (0.28 to 1.67)	0.77 (0.19 to 3.12)	0.69 (0.22 to 2.13)	1.99 (1.02 to 3.88)	0.84 (0.37 to 1.94)	1.15 (0.46 to 2.91)	Surufatinib

Bevacizumab + SSA	0.66 (0.21 to 2.13)	1.02 (0.42 to 2.47)	0.76 (0.31 to 1.90)	0.95 (0.41 to 2.19)	0.90 (0.41 to 1.96)	1.08 (0.85 to 1.37)	0.36 (0.15 to 0.89)	0.71 (0.32 to 1.58)	0.87 (0.32 to 2.38)	0.74 (0.28 to 2.01)
1.51 (0.47 to 4.83)	Dactolisib	1.53 (0.72 to 3.25)	1.15 (0.46 to 2.89)	1.43 (0.62 to 3.33)	1.35 (0.44 to 4.16)	1.62 (0.52 to 5.07)	0.55 (0.25 to 1.21)	1.08 (0.46 to 2.53)	1.31 (0.52 to 3.27)	1.12 (0.45 to 2.76)
0.98 (0.40 to 2.40)	0.65 (0.31 to 1.39)	Everolimus	0.75 (0.44 to 1.28)	0.94 (0.65 to 1.36)	0.89 (0.38 to 2.04)	1.06 (0.45 to 2.49)	0.36 (0.28 to 0.46)	0.70 (0.47 to 1.05)	0.85 (0.50 to 1.44)	0.73 (0.45 to 1.20)
1.31 (0.53 to 3.27)	0.87 (0.35 to 2.19)	1.33 (0.78 to 2.27)	Everolimus + bevacizumab + SSA	1.25 (0.86 to 1.82)	1.18 (0.50 to 2.78)	1.41 (0.58 to 3.41)	0.48 (0.28 to 0.83)	0.94 (0.60 to 1.47)	1.14 (0.55 to 2.34)	0.98 (0.48 to 1.96)
1.05 (0.46 to 2.41)	0.70 (0.30 to 1.62)	1.07 (0.73 to 1.55)	0.80 (0.55 to 1.17)	Everolimus + SSA	0.94 (0.44 to 2.04)	1.13 (0.51 to 2.50)	0.38 (0.26 to 0.57)	0.75 (0.58 to 0.96)	0.91 (0.49 to 1.68)	0.78 (0.43 to 1.41)
1.11 (0.51 to 2.43)	0.74 (0.24 to 2.27)	1.13 (0.49 to 2.60)	0.85 (0.36 to 2.00)	1.06 (0.49 to 2.29)	Interferon	1.20 (0.57 to 2.52)	0.41 (0.18 to 0.94)	0.80 (0.38 to 1.65)	0.96 (0.37 to 2.51)	0.83 (0.32 to 2.12)
0.93 (0.73 to 1.18)	0.62 (0.20 to 1.93)	0.94 (0.40 to 2.22)	0.71 (0.29 to 1.71)	0.89 (0.40 to 1.96)	0.84 (0.40 to 1.76)	Interferon + SSA	0.34 (0.14 to 0.80)	0.66 (0.31 to 1.42)	0.81 (0.30 to 2.15)	0.69 (0.26 to 1.81)
2.75 (1.12 to 6.71)	1.82 (0.83 to 4.02)	2.79 (2.19 to 3.55)	2.09 (1.21 to 3.63)	2.62 (1.75 to 3.91)	2.47 (1.07 to 5.70)	2.95 (1.25 to 6.98)	Placebo	1.96 (1.30 to 2.96)	2.38 (1.49 to 3.79)	2.04 (1.32 to 3.15)
1.40 (0.63 to 3.09)	0.93 (0.40 to 2.18)	1.42 (0.95 to 2.12)	1.07 (0.68 to 1.68)	1.33 (1.04 to 1.71)	1.26 (0.61 to 2.61)	1.50 (0.71 to 3.20)	0.51 (0.34 to 0.77)	SSA	1.21 (0.65 to 2.26)	1.04 (0.57 to 1.89)
1.15 (0.42 to 3.16)	0.77 (0.31 to 1.92)	1.17 (0.69 to 1.98)	0.88 (0.43 to 1.81)	1.10 (0.59 to 2.03)	1.04 (0.40 to 2.70)	1.24 (0.47 to 3.30)	0.42 (0.26 to 0.67)	0.82 (0.44 to 1.53)	Sunitinib	0.86 (0.45 to 1.62)
1.35 (0.50 to 3.63)	0.89 (0.36 to 2.20)	1.37 (0.83 to 2.24)	1.03 (0.51 to 2.06)	1.28 (0.71 to 2.31)	1.21 (0.47 to 3.10)	1.45 (0.55 to 3.79)	0.49 (0.32 to 0.76)	0.96 (0.53 to 1.75)	1.17 (0.62 to 2.20)	Surufatinib

	Direct evidence		Indirect evidence		Network meta‐analysis
Comparison	Odds ratio (95% CI)	Quality of evidence	Odds ratio (95% CI)	Quality of evidence	Odds ratio (95% CI)	Quality of evidence
Dactolisib vs everolimus	0.17 (0.04 to 0.68)	Low*,§			0.17 (0.04 to 0.68)	Low§
Dactolisib vs everolimus + SSA			0.19 (0.04 to 0.87)	Very low\|\|,§	0.19 (0.04 to 0.87)	Very low§
Dactolisib vs interferon			0.22 (0.03 to 1.41)	Very low\|\|\|,§§	0.22 (0.03 to 1.41)	Very low§§
Dactolisib vs interferon + SSA			0.19 (0.04 to 1.04)	Very low\|\|,¶,§§	0.19 (0.04 to 1.04)	Very low§§
Dactolisib vs placebo			0.56 (0.13 to 2.37)	Very low\|,§§	0.56 (0.13 to 2.37)	Very low§§
Dactolisib vs SSA			0.24 (0.05 to 1.07)	Very low\|,§§	0.24 (0.05 to 1.07)	Very low§§
Dactolisib vs sunitinib			0.32 (0.07 to 1.58)	Very low\|,§§	0.32 (0.07 to 1.58)	Very low§§
Dactolisib vs surufatinib			0.28 (0.06 to 1.38)	Very low\|,§§	0.28 (0.06 to 1.38)	Very low§§
Everolimus vs everolimus + SSA	1.41 (0.65 to 3.08)	Very low**,§	0.86 (0.35 to 2.08)	Very low\|,¶,§	1.14 (0.63 to 2.04)	Very low§
Everolimus vs interferon			1.27 (0.36 to 4.49)	Very low\|\|\|,§§	1.27 (0.36 to 4.49)	Very low§§
Everolimus vs interferon + SSA			1.14 (0.44 to 2.95)	Very low\|\|,¶,§	1.14 (0.44 to 2.95)	Very low§
Everolimus vs placebo	3.08 (2.01 to 4.72)	High	5.06 (1.68 to 15.2)	Very low\|\|,¶¶	3.29 (2.21 to 4.90)	High
Everolimus vs SSA			1.40 (0.79 to 2.46)	Low\|,§	1.40 (0.79 to 2.46)	Low§
Everolimus vs sunitinib			1.91 (0.90 to 4.06)	Moderate§	1.91 (0.90 to 4.06)	Moderate§
Everolimus vs surufatinib			1.65 (0.76 to 3.61)	Moderate§	1.65 (0.76 to 3.61)	Moderate§
Everolimus + SSA vs interferon			1.12 (0.33 to 3.79)	Very low\|\|\|,¶,§§	1.12 (0.33 to 3.79)	Very low§§
Everolimus + SSA vs interferon + SSA			1.00 (0.41 to 2.46)	Very low\|\|,¶,§	1.00 (0.41 to 2.46)	Very low§
Everolimus + SSA vs placebo			2.89 (1.61 to 5.19)	Moderate\|	2.89 (1.61 to 5.19)	Moderate
Everolimus + SSA vs SSA	1.36 (0.80 to 2.30)	Low‡,§	0.83 (0.29 to 2.37)	Very low\|\|,§§	1.23 (0.77 to 1.97)	Moderate
Everolimus + SSA vs sunitinib			1.68 (0.71 to 4.00)	Very low\|,§	1.68 (0.71 to 4.00)	Very low§
Everolimus + SSA vs surufatinib			1.46 (0.60 to 3.54)	Very low\|,§	1.46 (0.60 to 3.54)	Very low§
Interferon vs interferon + SSA	1.07 (0.31 to 3.72)	Very low**,‡,§§	0.39 (0.03 to 5.94)	Very low\|\|\|,¶,§§	0.90 (0.29 to 2.79)	Very low#,§§
Interferon vs placebo			2.58 (0.75 to 8.81)	Very low\|\|\|,§§	2.58 (0.75 to 8.81)	Very low§§
Interferon vs SSA	0.93 (0.28 to 3.16)	Very low**,‡,§§	2.64 (0.15 to 46.3)	Very low\|\|\|,¶,§§	1.09 (0.36 to 3.37)	Very low#,§§
Interferon vs sunitinib			1.50 (0.37 to 5.98)	Very low\|\|\|,§§	1.50 (0.37 to 5.98)	Very low§§
Interferon vs surufatinib			1.30 (0.32 to 5.26)	Very low\|\|\|,§§	1.30 (0.32 to 5.26)	Very low§§
Interferon + SSA vs placebo			2.88 (1.16 to 7.13)	Very low\|\|,¶	2.88 (1.16 to 7.13)	Very low
Interferon + SSA vs SSA	1.22 (0.57 to 2.61)	Very low*,‡,§			1.22 (0.57 to 2.61)	Very low§
Interferon + SSA vs sunitinib			1.67 (0.55 to 5.07)	Very low\|\|,¶,§§	1.67 (0.55 to 5.07)	Very low§§
Interferon + SSA vs surufatinib			1.45 (0.47 to 4.47)	Very low\|\|,¶,§§	1.45 (0.47 to 4.47)	Very low§§
Placebo vs SSA	0.38 (0.21 to 0.67)	Moderate‡	0.62 (0.22 to 1.75)	Very low\|\|,¶,§§	0.42 (0.26 to 0.70)	Moderate
Placebo vs sunitinib	0.58 (0.31 to 1.10)	Moderate§			0.58 (0.31 to 1.10)	Moderate§
Placebo vs surufatinib	0.50 (0.26 to 0.98)	High			0.50 (0.26 to 0.98)	High
SSA vs sunitinib			1.37 (0.61 to 3.08)	Low\|,§	1.37 (0.61 to 3.08)	Low§
SSA vs surufatinib			1.19 (0.51 to 2.73)	Low\|,§	1.19 (0.51 to 2.73)	Low§
Sunitinib vs surufatinib			0.87 (0.34 to 2.19)	Moderate§	0.87 (0.34 to 2.19)	Moderate§

177‐Lu‐DOTATATE + SSA	0.68 (0.05 to 10.1)	12.0 (2.33 to 62.1)	2.02 (0.30 to 13.8)	7.55 (1.37 to 41.6)	5.33 (1.42 to 20.0)	30.4 (8.19 to 113)	10.4 (3.59 to 30.1)	229 (6.16 to 8512)	8.69 (1.60 to 47.1)
1.48 (0.10 to 22.1)	Bevacizumab + SSA	17.8 (1.10 to 288)	2.99 (0.15 to 57.6)	11.2 (0.74 to 168)	7.87 (0.74 to 83.5)	45.0 (3.32 to 609)	15.4 (1.28 to 185)	338 (5.14 to 22282)	12.8 (0.77 to 214)
0.08 (0.02 to 0.43)	0.06 (0.00 to 0.91)	Everolimus	0.17 (0.02 to 1.28)	0.63 (0.10 to 3.91)	0.44 (0.10 to 1.94)	2.53 (0.95 to 6.79)	0.87 (0.25 to 3.02)	19.1 (0.48 to 752)	0.72 (0.17 to 3.08)
0.49 (0.07 to 3.38)	0.33 (0.02 to 6.45)	5.95 (0.78 to 45.3)	Everolimus + SSA	3.74 (0.47 to 30.0)	2.64 (0.44 to 15.7)	15.1 (2.55 to 88.9)	5.14 (1.04 to 25.5)	113 (2.51 to 5106)	4.30 (0.54 to 34.1)
0.13 (0.02 to 0.73)	0.09 (0.01 to 1.35)	1.59 (0.26 to 9.90)	0.27 (0.03 to 2.15)	Interferon	0.71 (0.18 to 2.70)	4.03 (0.86 to 18.8)	1.38 (0.36 to 5.22)	30.3 (1.25 to 735)	1.15 (0.18 to 7.47)
0.19 (0.05 to 0.71)	0.13 (0.01 to 1.35)	2.26 (0.51 to 9.89)	0.38 (0.06 to 2.26)	1.42 (0.37 to 5.41)	Interferon + SSA	5.71 (1.90 to 17.2)	1.95 (0.89 to 4.29)	43.0 (1.35 to 1365)	1.63 (0.35 to 7.54)
0.03 (0.01 to 0.12)	0.02 (0.00 to 0.30)	0.39 (0.15 to 1.06)	0.07 (0.01 to 0.39)	0.25 (0.05 to 1.16)	0.18 (0.06 to 0.53)	Placebo	0.34 (0.16 to 0.74)	7.52 (0.22 to 259)	0.29 (0.10 to 0.83)
0.10 (0.03 to 0.28)	0.07 (0.01 to 0.78)	1.16 (0.33 to 4.04)	0.19 (0.04 to 0.96)	0.73 (0.19 to 2.76)	0.51 (0.23 to 1.13)	2.93 (1.36 to 6.32)	SSA	22.0 (0.70 to 698)	0.84 (0.23 to 3.10)
0.00 (0.00 to 0.16)	0.00 (0.00 to 0.19)	0.05 (0.00 to 2.07)	0.01 (0.00 to 0.40)	0.03 (0.00 to 0.80)	0.02 (0.00 to 0.74)	0.13 (0.00 to 4.58)	0.05 (0.00 to 1.44)	Streptozocin + 5FU	0.04 (0.00 to 1.53)
0.12 (0.02 to 0.62)	0.08 (0.00 to 1.30)	1.38 (0.32 to 5.89)	0.23 (0.03 to 1.84)	0.87 (0.13 to 5.64)	0.61 (0.13 to 2.83)	3.50 (1.21 to 10.1)	1.20 (0.32 to 4.44)	26.4 (0.65 ‐ 1062)	Surufatinib

177‐Lu‐DOTATATE + SSA	0.40 (0.07 to 2.32)	0.13 (0.03 to 0.64)	0.62 (0.12 to 3.22)	0.26 (0.06 to 1.22)	0.32 (0.07 to 1.47)	0.07 (0.02 to 0.26)	0.21 (0.08 to 0.57)	0.22 (0.04 to 1.09)
2.51 (0.43 to 14.6)	Bevacizumab + SSA	0.32 (0.05 to 2.20)	1.55 (0.22 to 10.9)	0.66 (0.16 to 2.80)	0.79 (0.34 to 1.86)	0.18 (0.04 to 0.94)	0.53 (0.12 to 2.24)	0.55 (0.08 to 3.73)
7.75 (1.55 to 38.7)	3.09 (0.45 to 20.9)	Everolimus	4.78 (0.78 to 29.4)	2.05 (0.37 to 11.2)	2.45 (0.44 to 13.6)	0.56 (0.21 to 1.49)	1.63 (0.46 to 5.71)	1.70 (0.42 to 6.78)
1.62 (0.31 to 8.43)	0.64 (0.09 to 4.54)	0.21 (0.03 to 1.28)	Everolimus + SSA	0.43 (0.07 to 2.44)	0.51 (0.09 to 2.96)	0.12 (0.03 to 0.54)	0.34 (0.09 to 1.26)	0.35 (0.06 to 2.18)
3.79 (0.82 to 17.4)	1.51 (0.36 to 6.36)	0.49 (0.09 to 2.68)	2.34 (0.41 to 13.4)	Interferon	1.20 (0.38 to 3.81)	0.27 (0.07 to 1.10)	0.80 (0.25 to 2.51)	0.83 (0.15 to 4.55)
3.16 (0.68 to 14.8)	1.26 (0.54 to 2.96)	0.41 (0.07 to 2.27)	1.95 (0.34 to 11.3)	0.84 (0.26 to 2.66)	Interferon + SSA	0.23 (0.06 to 0.93)	0.66 (0.21 to 2.14)	0.69 (0.12 to 3.85)
13.8 (3.87 to 49.5)	5.51 (1.06 to 28.6)	1.79 (0.67 to 4.75)	8.54 (1.85 to 39.4)	3.65 (0.91 to 14.7)	4.37 (1.07 to 17.9)	Placebo	2.90 (1.32 to 6.38)	3.03 (1.14 to 8.07)
4.76 (1.75 to 13.0)	1.90 (0.45 to 8.05)	0.61 (0.18 to 2.16)	2.94 (0.79 to 10.9)	1.26 (0.40 to 3.97)	1.50 (0.47 to 4.83)	0.34 (0.16 to 0.76)	SSA	1.04 (0.30 to 3.66)
4.56 (0.91 to 22.8)	1.82 (0.27 to 12.3)	0.59 (0.15 to 2.35)	2.82 (0.46 to 17.3)	1.21 (0.22 to 6.62)	1.44 (0.26 to 8.01)	0.33 (0.12 ‐ 0.88)	0.96 (0.27 to 3.37)	Surufatinib

Treatment	Patients, no.	Grade 3 or 4 (total no.)¹	All grades (total no.)	Sources
177‐Lu‐DOTATATE + octreotide	111	10	95	Strosberg 2017
Capecitabine + streptozocin	43	33	195	Meyer 2014
Capecitabine + streptozocin + cisplatin	40	67	302	Meyer 2014
Chlorozotocin	51	29	198	Moertel 1992
Dactolisib	31	39	220	Salazar 2018
Doxorubicin + streptozocin	44	29	202	Moertel 1992
Etoposide + cisplatin	33	29	69	Zhang 2020
Everolimus	518	219	2095	Kulke 2017 (1), Salazar 2018, Yao 2011, Yao 2016
Everolimus + SSA	293	149	975	Kulke 2017 (1), Pavel 2011
Hepatic arterial chemoembolisation	12	3	15	Maire 2012
Hepatic arterial embolisation	14	2	12	Maire 2012
Interferon + SSA	33	1	7	Kölby 2003
Irinotecan + cisplatin	33	15	62	Zhang 2020
Placebo	670	107	1300	Caplin 2014, Raymond 2011 (1), Vinik 2016, Xu 2020 (ep), Xu 2020 (p), Yao 2011, Yao 2016
SSA	610	38	389	Caplin 2014, Kölby 2003, Pavel 2011, Strosberg 2017, Vinik 2016, Wolin 2015
Streptozocin + 5‐FU	42	86	271	Moertel 1992
Tyrosine kinase inhibitors	331	317	2590	Raymond 2011 (1), Xu 2020 (ep), Xu 2020 (p)

	("Neuroendocrine Tumors"[Mesh:NoExp] OR "Adenoma, Acidophil"[Mesh] OR "Adenoma, Basophil"[Mesh] OR "Adenoma, Chromophobe"[Mesh] OR "Apudoma"[Mesh] OR "Carcinoid Tumor"[Mesh] OR "Malignant Carcinoid Syndrome"[Mesh] OR "Carcinoma, Neuroendocrine"[Mesh] OR "Carcinoma, Medullary"[Mesh] OR "Carcinoma, Merkel Cell"[Mesh] OR "Somatostatinoma"[Mesh] OR "Vipoma"[Mesh] OR "Neurilemmoma"[Mesh] OR "Paraganglioma"[Mesh]) AND "Gastrointestinal Neoplasms"[Mesh]) OR ("Pancreatic Neoplasms"[Mesh:NoExp] AND neuroendocrine[tiab]) OR "Adenoma, Islet Cell"[Mesh] OR "Insulinoma"[Mesh] OR "Carcinoma, Islet Cell"[Mesh] OR "Gastrinoma"[Mesh] OR "Glucagonoma"[Mesh] OR ((gastroenteropancreatic OR gastro‐enteric pancreatic OR gastro‐entero‐pancreatic OR pancreas OR pancreatic) AND (neuroendocrine AND (tumor OR tumors OR tumour OR tumours OR neoplasm OR neoplasms OR carcinoma OR carcinomas)) OR GEPNET* OR GEP‐NET* OR GEPNEC* OR GEP‐NEC*
Therapy search filter	therapy[sh] OR "diet therapy"[sh] OR "drug therapy"[sh] OR radiotherapy[sh] OR surgery[sh] OR segmentectomy OR resection OR debulk* OR cryoablat* OR cryosurger* OR radioablat* OR radiofrequency ablat* OR radio‐frequency ablat* OR RFablat* OR thermoablat* OR "Cryosurgery"[Mesh] OR "Hepatectomy"[MeSH] OR Liver transplant OR local ablat* OR transarterial embolization OR transarterial embolisation OR transarterial chemoembolization OR transarterial chemoembolisation OR radioembolization OR radioembolisation OR somatostatin OR chemotherapy OR chemotherapies OR peptide receptor radiotherapy OR targeted molecular therapy OR radiopeptide OR DOTATOC OR DOTATATE OR PRRT
Study design filter	randomized controlled trial[pt] OR controlled clinical trial[pt] OR randomized[tiab] OR placebo[tiab] OR "drug therapy"[sh] OR randomly[tiab] OR trial[tiab] OR groups[tiab]) NOT ("animals"[mh] NOT ("humans"[mh] AND "animals"[mh])

	(('neuroendocrine tumor'/de OR 'gastroenteropancreatic neuroendocrine tumor'/de OR (adenoma NEAR/3 acidophil):ti,ab OR (adenoma NEAR/3 basophil):ti,ab OR 'chromophobe adenoma'/de OR 'apudoma'/de OR 'carcinoid'/de OR 'carcinoid syndrome'/de OR (carcinoma NEAR/3 neuroendocrine):ti,ab OR 'medullary carcinoma'/de OR 'merkel cell carcinoma'/de OR 'somatostatinoma'/de OR 'vipoma'/de OR 'neurilemoma'/de OR 'paraganglioma'/de) AND* ('gastrointestinal tumor'/de OR 'gastrointestinal stromal tumor'/de OR 'intestine tumor'/exp OR 'pancreas tumor'/exp OR 'stomach tumor'/exp)) or ('pancreas islet cell tumor'/de OR 'glucagonoma'/de OR 'insulinoma'/de OR 'pancreas islet cell carcinoma'/de OR 'gastrinoma'/de) OR (((gastroenteropancreatic OR 'gastro‐enteric pancreatic' OR 'gastro‐entero‐pancreatic' OR pancreas OR pancreatic) AND (neuroendocrine AND (tumor* OR tumour* OR neoplasm* OR carcinoma))) OR GEPNET OR 'GEP‐NET' OR GEPNEC* OR GEP‐NEC*)
Therapy search filter	('disease management':lnk OR 'drug therapy':lnk OR 'surgery':lnk OR 'therapy':lnk OR 'radiotherapy':lnk) OR segmentectomy OR resection OR debulk* OR cryoablat* OR cryosurger* OR radioablat* OR 'radiofrequency ablat' OR 'radio‐frequency ablat' OR RFablat* OR thermoablat* OR 'cryosurgery'/de OR 'liver resection'/exp OR 'liver transplant' OR 'local ablat*' OR 'transarterial embolization' OR 'transarterial embolisation' OR 'transarterial chemoembolization' OR 'transarterial chemoembolisation' OR radioembolization OR radioembolisation OR somatostatin OR chemotherapy OR chemotherapies OR 'peptide receptor radiotherapy' OR 'targeted molecular therapy' OR radiopeptide or DOTATOC or DOTATATE or PRRT
Study design filter	((random* OR factorial* OR crossover* OR cross NEXT/1 over* OR placebo* OR doubl* NEXT/1 blind* OR singl* NEXT/1 blind* OR assign* OR allocat* OR volunteer*):de,ab,ti OR 'crossover procedure'/exp OR 'double blind procedure'/exp OR 'randomized controlled trial'/exp OR 'single blind procedure'/exp) NOT ('animal'/exp NOT 'human'/exp)

	Placebo	SSA	Interferon + SSA	Telotristat	Capecitabine + Streptozocin	Capecitabine + Streptozocin + Cisplatin	Sunitinib	Surufatinib
Arnold 2005		11.4 ± 18.6	‐6.4 ± 18.6
Caplin 2014	‐4.87 ± 3.7	‐5.18 ± 3.73
Kulke 2017 (2)	8.5			21.6
Meyer 2014					2.2	‐3.8
Raymond 2011 (1)	‐2.7						‐4.6
Rinke 2009	‐2.1 ± 15.8	0.0 ± 18.5
Vinik 2016	1.2 ± 2.6	5.3 ± 2.1
Xu 2020 (ep)	‐6.43 ± 2.61							‐9.97 ± 1.87
Xu 2020 (p)	‐11.2 ± 2.6							‐8.8 ± 1.9

*Study characteristics*
Methods	Multicentre (16 countries), double‐blind, phase 3 study 1:1 randomisation by interactive voice response system Study group assignments were masked. Enrolment: January 2007‐April 2010 Subgroup analysis: effect of previous treatment with a long‐acting SSA on PFS in RADIANT‐2
Participants	Inclusion criteria Age ≥ 18 years Low‐grade or intermediate‐grade, unresectable locally advanced or distant metastatic neuroendocrine tumour Disease progression by radiological assessment within the past 12 months History of diarrhoea or flushing attributable to carcinoid syndrome Measurable disease according to RECIST version 1.0 WHO performance status ≤ 2 Adequate bone marrow, renal, and hepatic function and adequately controlled lipid concentrations Exclusion criteria Poorly differentiated or high‐grade neuroendocrine carcinomas RADIANT‐2 overall population Total patients: 429 Median age (study group 1 vs. study group 2): 60 vs. 60 Women, % (1 vs. 2): 55 vs. 42 WHO performance status 0/1/2, % (1 vs. 2): 55/39/6 vs. 66/29/5 Primary tumour site, %: Small intestine, (1 vs. 2): 51 vs. 53 Lung, (1 vs. 2): 15 vs. 5 Colon, (1 vs. 2): 6 vs. 7 Pancreas, (1 vs. 2): 5 vs. 7 Liver, (1 vs. 2): 3 vs. 5 Other, (1 vs. 2): 19 vs. 23 Missing, (1 vs. 2): 0 vs. 1 Grade (well differentiated/moderately differentiated/poorly differentiated), % (1 vs. 2): 77/18/1 vs. 82/14/1 Liver involvement, % (1 vs. 2): 92 vs. 92 Previous SSA treatment, % (1 vs. 2): 80 vs. 78 Previous systemic anti‐tumour drugs, % (1 vs. 2): 46 vs. 38 Chemotherapy, % (1 vs. 2): 35 vs. 26 Immunotherapy, % (1 vs. 2): 13 vs. 9 Targeted therapy, % (1 vs. 2): 7 vs. 8 Other, % (1 vs. 2): 10 vs. 13 Prior SSA treatment subgroup Total patients: 429 Previous SSA treatment, % (1 vs. 2): 80 vs. 78 Primary tumour site (overall in previous SSA treatment group): Foregut: 10% Midgut: 72% Hindgut: 11% Not classified/missing: 7% SSA naive, % (1 vs. 2): 20 vs. 22 Primary tumour site (overall in SSA naive group): Foregut: 32% Midgut: 51% Hindgut: 4% Not classified/missing: 13%
Interventions	Study group 1 (RADIANT‐2 overall: 216/429, prior SSA treatment subgroup: 173/339, SSA‐naive group: 43/90): 10 mg oral everolimus once daily plus intramuscular 30 mg octreotide LAR every 28 days Study group 2 (RADIANT‐2 overall: 213/429, prior SSA treatment subgroup: 166/339, SSA‐naive group: 47/90): matching placebo plus intramuscular 30 mg octreotide LAR every 28 days Treatment duration: until disease progression, withdrawal from treatment because of adverse events, or withdrawal of consent After disease progression in the placebo plus octreotide LAR group, cross over to open‐label everolimus plus octreotide LAR was permitted.
Outcomes	Primary endpoint: Progression‐free survival according to RECIST Secondary endpoints: Objective response rate according to RECIST Overall survival Changes from baseline in 5‐hydroxyindoleacetic acid and CgA concentrations Safety Supportive endpoint: Investigator‐assessed progression‐free survival Assessments: CT or MRI were done at baseline and repeated every 12 weeks. Serum CgA and 24‐h urine samples for 5‐hydroxyindoleacetic acid at baseline and on day 1 of each subsequent cycle (if raised at baseline) Monitoring of adverse events, vital signs and physical examinations every 4 weeks Chest radiographs every 12 weeks
Notes	Novartis funded the study and was involved in the study design, data collection and statistical analysis.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information about sequence generation
Allocation concealment (selection bias)	Low risk	Allocation was performed centrally.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blinded trial
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Central review for primary analysis of progression‐free survival by an independent, masked committee
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised patients accounted for efficacy analysis according to the intention‐to‐treat principle.
Selective reporting (reporting bias)	Low risk	Except for one secondary endpoint, every endpoint stated in the study protocol was reported in the publication.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Randomisation was performed by phone at the study centre and done by computer by using the method of random permuted blocks stratified by carcinoid syndrome versus other tumour entities, age ≤ 65 years versus > 65 years, luminal tumours (midgut tumours and duodenal tumours) versus non‐luminal (pancreatic) tumours, prior chemotherapy and prior octreotide treatment. Enrolment: January 1995‐March 1998 Follow‐up investigations were performed until April 2004.
Participants	Inclusion criteria Age ≥ 18 years Metastatic or locally advanced gastroenteropancreatic tumours without curative therapeutic option Primary within the pancreas, duodenum, and midgut; tumours of unknown origin believed to belong to the midgut as a result of the presence of a carcinoid syndrome or in nonfunctioning tumours as a result of histologic criteria Well differentiated histology by pathologic review Tumour progression documented on computed tomography (CT) or magnetic resonance imaging (MRI) according to World Health Organization (WHO) criteria Patients receiving ≤ 150 μg octreotide per day subcutaneously against flushing and/or diarrhoea caused by carcinoid syndrome Exclusion criteria Pretreatment with interferon‐alpha Pregnancy Karnofsky Index < 70 Previous hepatic artery embolisation Leukocytes < 2.0 g/L Thrombocytes < 75 g/L Autoimmune disorders History of major depression Decompensated organ insufficiency Drug or alcohol addiction Total randomised patients: 109 Total evaluable patients: 105 Age (study arm 1 vs. study arm 2): 58 vs. 57 Women, % (1 vs. 2): 47 vs. 44 Prior treatment, %: ≤ 150 μg octreotide per day (1 vs. 2): 14 vs. 11 Chemotherapy (1 vs. 2): 8 vs. 15 Primary tumour site, %: Pancreas (1 vs. 2): 31 vs. 41 Duodenum (1 vs. 2): 2 vs. 2 Midgut (1 vs. 2): 49 vs. 37 Unknown (1 vs. 2): 18 vs. 20 Nonfunctioning tumours, % (1 vs. 2): 53 vs. 56
Interventions	Study arm 1 (51/105): 200 μg octreotide, thrice daily, subcutaneous injection Study arm 2 (54/105): 200 μg octreotide, thrice daily, subcutaneous injection plus 4.5 × 10⁶ IU interferon‐alpha thrice weekly Treatment duration: until CT or MRI documented tumour progression Additional antiproliferative therapy was not allowed.
Outcomes	Primary endpoint: Time to treatment failure Secondary endpoints: Survival Adverse events Quality of life Symptomatic response (only in patients with carcinoid syndrome) Biochemical response (CgA in 40 patients, urine 5‐hydroxyindoleacetic acid levels in 26 patients) Assessments: Pretreatment evaluation: biochemical screening, chest radiography, octreoscan, and CT or MRI of pertinent index lesions Follow‐up investigations were performed at 3‐month intervals until tumour progression CT or MRI scans of pertinent indicator lesions were evaluated by one of the authors in a blinded fashion Biochemical response was evaluated only in patients treated in the hospital of the principal author
Notes	Novartis Pharma and Roche Pharma participated in the development of the study design and provided funding.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Stratified randomisation was done by computer by using the method of random permuted blocks.
Allocation concealment (selection bias)	Low risk	Allocation was performed centrally.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Study treatment was not blinded.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	CT or MRI scans were evaluated by one of the authors in a blinded fashion, but not by independent personnel.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Biochemical response was evaluated only in one centre. 109 patients were randomised but only 105 were evaluable.
Selective reporting (reporting bias)	Low risk	No study protocol available, but all endpoints were reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Randomised, double‐blind, placebo‐controlled, parallel‐group, multicentre, phase 3 study 48 secondary or tertiary care centres in 14 countries Duration: 96 weeks Computer‐generated randomisation, stratified by presence or absence of tumour progression at baseline and receipt or nonreceipt of previous therapies Conducted between June 2006 and April 2013
Participants	Inclusion criteria Adults (≥ 18 years of age) Sporadic well differentiated or moderately differentiated neuroendocrine tumours, located in the pancreas, midgut, hindgut or of unknown origin Unresectable locally advanced tumour, metastatic disease or declined surgery Measurable tumour according to RECIST (vers. 1.0) Ki‐67 index of less than 10% or a mitotic index of ≤ 2 mitoses per 10 high‐power fields Nonfunctioning tumours (except for gastrinomas that had been adequately controlled by means of proton‐pump inhibitors for 4 months or longer) Target lesion or lesions that were classified as grade 2 or higher on somatostatin‐receptor scintigraphy (0 (no uptake by tumour) to 4 (very intense uptake by tumour)) within the previous 6 months WHO performance score ≤ 2 A biopsy of the neuroendocrine tumour within 6 months before study entry was required for patients who had previous cancer and those with evidence of clinical progression. Exclusion criteria Previous treatment with interferon, chemoembolisation, or chemotherapy within 6 months before study entry, a radionuclide at any time, or a somatostatin analogue at any time (unless they had received it > 6 months previously and for < 15 days) Major surgery related to the neuroendocrine tumour within 3 months before study entry Multiple endocrine neoplasia Previous cancer (except: 1] treated or untreated in situ cervical or uterine carcinoma, or 2] basal‐cell skin carcinoma, or 3] other cancers that had been treated with curative intent and had been disease‐free for > 5 years) Baseline abnormalities or medical conditions that could jeopardise the patient's safety or interfere with the study Withdrawal Tumour progression (RECIST) Investigator's judgement Patient's request Adverse event that could jeopardise the patient's safety Total patients: 204 Age (lanreotide vs. placebo): 63 vs. 62 Women, % (lanreotide vs. placebo): 48 vs. 48 Prior treatment for neuroendocrine tumour, % (lanreotide vs. placebo): 16 vs. 16 Primary tumour resected, % (lanreotide vs. placebo): 40 vs. 38 Origin of tumour: Pancreas, % (lanreotide vs. placebo): 42 vs. 48 Midgut, % (lanreotide vs. placebo): 33 vs. 39 Hindgut, % (lanreotide vs. placebo): 11 vs. 3 Unknown, % (lanreotide vs. placebo): 15 vs. 11 Ki‐67 index, 0‐2%/3‐10%, % (lanreotide vs. placebo): 68/32 vs. 70/28 Hepatic tumour volume: 0%, % (lanreotide vs. placebo): 16 vs. 17 > 0‐10%, % (lanreotide vs. placebo): 33 vs. 39 > 10‐25%, % (lanreotide vs. placebo): 13 vs. 17 > 25‐50%, % (lanreotide vs. placebo): 23 vs. 12 > 50%,% (lanreotide vs. placebo): 16 vs. 16
Interventions	Intervention group (101/204): extended‐release aqueous‐gel formulation of lanreotide, 120 mg, without dose adjustment, deep subcutaneous injection, every 28 days to a maximum of 24 injections Control group (103/204): placebo (sodium chloride), deep subcutaneous injection, every 28 days to a maximum of 24 injections In case of disease progression while receiving placebo, patients crossed over to lanreotide.
Outcomes	Primary endpoint: Progression‐free survival or death within 96 weeks after the first injection of the study drug Secondary endpoints: Proportion of patients who were alive without disease progression at 48 and 96 weeks Time to tumour progression Overall survival Quality of life CgA levels Pharmacokinetic data Safety Exploratory endpoints: Data on other tumour biomarkers Assessments: Study visits: at weeks 1 (baseline), 12, 24, 36, 48, 72, and 96 CT or MRI of the chest, abdomen, and pelvis was performed twice during screening to determine the baseline disease‐progression status. Results of the second imaging test were considered to be the baseline findings and were used to determine target‐lesion sizes. Single scans were obtained at all post‐baseline visits. Disease progression was assessed centrally according to RECIST, version 1.0. Two quality of life questionnaires (QLQ‐C30 and QLQ‐GI.NET21) were completed at post‐screening visits. Serum chromogranin A levels: all visits and also at weeks 60 and 84 Serum lanreotide levels: prior to drug administration at all study visits and after the first and sixth administration Safety assessments: monitoring for adverse events, physical examination and monitoring of vital signs (at all visits), electrocardiography and ultrasonography of the gallbladder (at baseline and at weeks 48 and 96), and clinical laboratory tests (at screening, baseline, and at weeks 48 and 96)
Notes	The study was designed, funded, and conducted by Ipsen in collaboration with the European Neuroendocrine Tumor Society and the UK and Ireland Neuroendocrine Tumour Society.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation lists were created by a statistician employed by the sponsor who was independent of the study.
Allocation concealment (selection bias)	Low risk	The blinded database was held at a third‐party contract clinical research organisation.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blinded study design. Independent health professionals prepared and administered injections.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Disease progression was assessed centrally, but it remained unclear whether it was performed by independent personnel.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised patients accounted for in ITT analysis
Selective reporting (reporting bias)	Low risk	Study protocol available. One secondary endpoint mentioned in the protocol was not reported as a endpoint in the publication, but was reported in the supplementary appendix.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Multicentre, randomised, double‐blind, phase II study
Participants	Inclusion criteria Progressive low‐intermediate grade carcinoid tumours Radiologic progressive disease < 12 months Prior SSA mandated for midgut tumours Trial had 85% power to detect a difference in median PFS 14 v 9 mo (hazard ratio [HR] 0.64) at 1‐sided alpha = 0.1. Stratified log‐rank test based on intention‐to‐treat (ITT) principle used. Unblinding and cross‐over allowed if PD confirmed by central review Total patients: 171 Median age (overall): 63 Women (overall): 56% Small bowel primary (overall): 66% Concurrent SSA treatment (overall): 87%
Interventions	Intervention group (97/171): pazopanib, 800 mg/day, oral intake Control group (74/171): placebo Concurrent SSA allowed if previous progressive disease on SSA was documented. Cross‐over was allowed if progressive disease was confirmed by central review.
Outcomes	Primary endpoint: Progression‐free survival Secondary endpoints: Overall survival Objective response rate Safety
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information given
Allocation concealment (selection bias)	Low risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No information given
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Central review was mentioned, but it remained unclear, when and how it was performed.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No information given
Selective reporting (reporting bias)	Unclear risk	One secondary endpoint (objective response rate) was not reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	International (11 countries), multicentre, randomised, double‐blind, placebo‐controlled phase 3 companion study (TELECAST) 1:1:1 randomisation stratified by baseline u5‐HIAA levels Enrolment: April 2014 to April 2015 Subgroup analysis: assessment of efficacy and safety of telotristat in the TELECAST study population with 2 or fewer bowel movements per day
Participants	Inclusion criteria Age ≥ 18 years Histopathologically confirmed, well differentiated metastatic NETs Documented history of carcinoid syndrome No SSA treatment or stable‐dose SSA treatment (long‐acting release, depot or infusion pump) for at least 3 months prior to enrolment Average of < 4 bowel movements/day At least 1 of the following signs or symptoms: Daily stool consistency ≥ 5 on the Bristol Stool Form scale for ≥ 50% of the days during the screening period Average daily cutaneous flushing frequency of ≥ 2 Average daily rating of ≥ 3 for abdominal pain Nausea present ≥ 20% of days u5‐HIAA above the upper limit of normal For patients not receiving SSA therapy: at least 1 of the above symptoms or an average of ≥ 4 bowel movements/day Exclusion criteria Diarrhoea attributable to any condition other than carcinoid syndrome ≥ 4 BMs/day while on concomitant SSA therapy Enteric infection Karnofsky performance status ≤ 60% History of short bowel syndrome Chronic or idiopathic constipation Clinically important baseline elevation in liver function tests Tumour‐directed therapy within 4 weeks prior to screening Hepatic embolisation, radiotherapy, radiolabeled SSA therapy and/or tumour debulking within 12 weeks prior to screening TELECAST overall population Total patients: 76 Mean age (A vs. B vs. C): 62 vs. 64 vs. 63 Women, % (A vs. B vs. C): 50 vs. 44 vs. 40 SSA therapy at study entry, %: Octreotide (A vs. B vs. C): 46 vs. 68 vs. 64 Lanreotide (A vs. B vs. C): 54 vs. 20 vs. 12 Unknown (A vs. B vs. C): 0 vs. 0 vs. 4 Not on SSA (A vs. B vs. C): 0 vs. 12 vs. 20 Subgroup: ≤ 2 bowel movements per day population Total patients: 28
Interventions	Study group A (TELECAST overall: 26/76, subgroup: 9/28): placebo, oral doses, three times per day for 12 weeks Study group B (TELECAST overall: 25/76, subgroup: 10/28): telotristat ethyl 250 mg, oral doses, three times per day for 12 weeks Study group C (TELECAST overall: 25/76, subgroup: 9/28): telotristat ethyl 500 mg, oral doses, three times per day for 12 weeks Patients continued to receive their baseline stable‐dose SSA therapy. Rescue short‐acting SSA use was allowed. After the study, all patients were offered treatment with telotristat ethyl 500 mg, three times per day in a 36‐week open‐label extension.
Outcomes	Primary endpoints: Incidence of treatment‐emergent adverse events Percent change from baseline in 24‐h u5‐HIAA levels at week 12 Secondary endpoints: Change from baseline averaged over the 12‐weeks period for daily bowel movement frequency Stool consistency Cutaneous flushing episodes Abdominal pain Frequency of rescue short‐acting SSA treatment Additional endpoint: Durability of response to treatment Assessments: Screening period of at least 3 weeks Electronic patient diary (identical to the one used in the TELESTAR study) for patient‐reported measures
Notes	Trial supported by Lexicon Pharmaceuticals, Inc.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information given
Allocation concealment (selection bias)	Unclear risk	Nearly equal amount of participants per study group
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind study design
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	The majority of endpoints were self‐reported.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	In the main study, all randomised patients accounted for safety analysis but 10 of 76 (13%) randomised patients were excluded from the u5‐HIAA which was the second primary endpoint. It is not clear, if these patients would have been in this subgroup.
Selective reporting (reporting bias)	Low risk	No study protocol available, but every stated endpoint was reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Randomised phase II study Start: January 2014 Closed: September 2016
Participants	Inclusion criteria Multiple SI‐NET liver metastases Grade 1 or 2 Not accessible to curative resection or ablation Elevated serum chromogranin A (CgA) and/or 24‐h urinary 5‐HIAA excretion (du5‐HIAA). Exclusion criteria Remaining extrahepatic metastases Previous locoregional or systemic anti‐tumoural treatment (except SSA) Impaired liver function Tumour volume exceeding 50% of total liver volume Total patients: 11 Median age (RE vs. HAE): 66.5 vs. 67 Women, % (RE vs. HAE): 67 vs. 80 Primary tumour grade 1, % (RE vs. HAE): 83 vs. 40 Primary tumour grade 2, % (RE vs. HAE): 17 vs. 60 Functional tumours: not reported
Interventions	RE group (6/11): radioembolisation with bilobar infusion in a standard manner. Protective coil embolisation was used when necessary to prevent non‐target embolisation. The administered activity of ⁹⁰Y resin microspheres (SIR‐spheres™) was calculated using the partition model. HAE group (5/11): hepatic arterial embolisation was performed by infusion of PVA particles (45–150 μm) until stasis was achieved. The right liver lobe was treated first, embolising the remaining left lobe about 6 weeks later.
Outcomes	Primary endpoint: Treatment response of hepatic metastases at 3 months after therapy Secondary endpoints: Radiological response at 6 months Biochemical response Toxicity Evaluation of usefulness of early changes in diffusion‐weighted imaging parameters in predicting later treatment response Assessments: MRI or CT before treatment, 1 month after treatment followed by response evaluation with MRI or CT according to RECIST 1.1 at 3 and 6 months CgA in serum and du5‐HIAA were measured at 3 and 6 months after treatment. Toxicity was assessed weekly during the first month after treatment and 3 and 6 months after treatment by laboratory analysis.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information given
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	No evidence for blinding
Blinding of outcome assessment (detection bias) All outcomes	High risk	No evidence for independent assessment
Incomplete outcome data (attrition bias) All outcomes	Low risk	All patients accounted for final analysis
Selective reporting (reporting bias)	Low risk	No study protocol available. But all endpoints mentioned were reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Prospective, randomised, multicentre trial Stratified block‐wise randomisation, carried out centrally, stratified by primary tumour localisation (foregut, midgut, hindgut, unknown) and functional or non‐functional tumours Enrolment: July 1995‐October 1998
Participants	Inclusion criteria Documented tumour progression of neuroendocrine tumour disease Exclusion criteria ECOG performance score of 3 or 4 Previous therapy for more than 4 weeks with any of the study agents, any chemotherapy or chemoembolisation of liver metastases Leukocyte count less than 2.5 x 10⁹/L Platelet count less than 100 x 10⁹/L Any other concurrent or recent malignant disease Total patients: 80 Median age (lanreotide vs. interferon alfa vs. combination): 60 vs. 56 vs. 58 Women, % (lanreotide vs. interferon alfa vs. combination): 52 vs. 37 vs. 36 Functional tumour, % (lanreotide vs. interferon alfa vs. combination): 48 vs. 33 vs. 29 Liver metastases, % (lanreotide vs. interferon alfa vs. combination): 92 vs. 93 vs. 89 Localisation of the primary, %: Foregut (lanreotide vs. interferon alfa vs. combination): 56 vs. 37 vs. 43 Midgut (lanreotide vs. interferon alfa vs. combination): 32 vs. 41 vs. 39 Hindgut (lanreotide vs. interferon alfa vs. combination): 0 vs. 4 vs. 7 Unknown (lanreotide vs. interferon alfa vs. combination): 12 vs. 19 vs. 11 Previous surgical resection, % (lanreotide vs. interferon alfa vs. combination): 56 vs. 44 vs. 54
Interventions	Study arm 1 (27/80): lanreotide, 1 mg, three times a day, subcutaneous injection Study arm 2 (28/80): interferon alfa, 5 x 10⁶ U, three times a week, subcutaneous injection Study arm 3 (29/80): lanreotide, 1 mg three times a day, subcutaneous injection and interferon alfa, 5 x 10⁶ U, three times a week, subcutaneous injection Patients showing progressive disease while receiving the initially assigned treatment with lanreotide or interferon alfa received the combination of lanreotide and interferon alfa.
Outcomes	Primary endpoint: 1‐year tumour progression rate Secondary endpoints: Symptom control Biochemical response assessed by serum chromogranin A levels, serum serotonin levels, and urinary 5‐hydroxyindoleacetic acid (5‐HIAA) levels Assessments: Transabdominal ultrasound and CT scans every 3 months
Notes	Ipsen Pharma and Essex Pharma participated in the development of the study design, provided funding and participated also in the collection of the data.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A stratified block‐wise randomisation with block size 6 was carried out centrally by telephone using randomisation tables.
Allocation concealment (selection bias)	Low risk	Done centrally
Blinding of participants and personnel (performance bias) All outcomes	High risk	No masking
Blinding of outcome assessment (detection bias) All outcomes	High risk	Only critical cases were re‐reviewed by an independent radiologist.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	4 patients had to be excluded after randomisation.
Selective reporting (reporting bias)	Low risk	No study protocol available but all endpoints in 'methods' were reported in 'results'.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	3‐phase, multicentre study in 12 countries 16‐week randomised, double‐blind phase (reported here) 32‐week initial open‐label phase (not reported here) Long‐term open‐label extension (not reported here) 1:1 randomisation using 2 computer‐generated lists (one for the US and one for all other countries) stratified by previous treatment with any long‐ or short‐acting somatostatin analog or SSA‐naive patients Start: May 2009 End: May 2013 Subgroup analysis: Efficacy and safety of lanreotide in the ELECT study subgroup of patients with prior octreotide therapy
Participants	Inclusion criteria Age ≥ 18 years Histopathologically confirmed diagnosis of neuroendocrine tumour or a carcinoid tumour of unknown location with liver metastases (documented biopsy) History of carcinoid syndrome (flushing and/or diarrhoea) Positive somatostatin‐receptor scintigraphy SSA‐naive or responsive to conventional octreotide LAR doses (≤ 30 mg/4 weeks) or short‐acting octreotide (≤ 600 μg daily) Absence of tumour progression on 2 sequential computed tomography/magnetic resonance imaging scans ≥ 3 months apart Last scan ≤ 6 months from study entry Exclusion criteria History of treatment‐refractory carcinoid syndrome with conventional SSA doses Treatment with interferon, chemotherapy, and/or peptide receptor radionuclide therapy Tumour debulking < 3 months before study entry Hepatic artery embolisation/chemoembolisation and/or selective internal radiation therapy < 6 months before study entry Short‐bowel syndrome Uncontrolled diabetes Hypertension Severe renal and/or hepatic impairment Cardiac disease New York Heart Association classification > class 1 Any malignancy except NET, basocellular skin carcinoma, or in situ cervical carcinoma Life expectancy < 1 year ELECT overall population Total patients: 115 Mean age (lanreotide vs. placebo): 58 vs. 59 Women, % (lanreotide vs. placebo): 54 vs. 62 Prior SSA therapy, % (lanreotide vs. placebo): 56 vs. 55 Short‐acting octreotide during screening, % (lanreotide vs. placebo): 51 vs. 52 Subgroup: prior octreotide therapy Total patients: 64 Mean age (overall): 59 Women (overall): 55%
Interventions	Intervention group (ELECT overall population: 59/115, prior octreotide therapy subgroup: 33/64): lanreotide depot/autogel 120 mg, every 4 weeks by deep subcutaneous injection Control group (ELECT overall population: 56/115, prior octreotide therapy subgroup: 31/64): placebo (0.9% saline solution), every 4 weeks by deep subcutaneous injection Self‐injected subcutaneous short‐acting octreotide for symptom rescue at patients' discretion After ≥ 4 weeks in the double‐blind phase, patients could roll over into the open‐label phase if they used octreotide for ≥ 21 days of the 28‐day cycle and used a dose ≥ 300 μg/day for ≥ 14 of the 21 days.
Outcomes	Primary endpoint: Adjusted mean percentage of days short‐acting octreotide was used for symptom control Secondary endpoints: Average daily frequency of diarrhoea and flushing Percentage of days non‐octreotide rescue medications were used Proportion of patients who rolled over early into the initial open‐label phase Change from baseline to week 12 in: Health‐related quality of life Plasma chromogranin Urinary 24‐hour 5‐hydroxyindoleacetic acid levels Safety Assessments: Prior to randomisation, patients completed a 31‐day (± 3 days) screening period. Daily diary by Interactive Voice Response System (IVRS) or Interactive Web Response System (IWRS) (number and severity of diarrhoea and flushing events; and use and dose of short‐acting octreotide and any other rescue medications)
Notes	Trial funded by Ipsen
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated, stratified randomisation
Allocation concealment (selection bias)	Unclear risk	Insufficient information given
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blinded trial design with same injection schedules
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Patient‐reported results
Incomplete outcome data (attrition bias) All outcomes	Low risk	Efficacy analyses were performed with all randomised patients on an ITT principle.
Selective reporting (reporting bias)	Low risk	No study protocol available, but every stated endpoint was reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	International, multicentre, double‐blind, phase 3 study 82 centres in 18 countries worldwide Randomisation: Ratio 1:1 Stratified by whether or not patients have received prior cytotoxic chemotherapy and by WHO performance status (0 vs. 1‐2) at baseline Start: July 2007 Closed: May 2009
Participants	Inclusion criteria: 18 years of age or older Low‐grade or intermediate‐grade advanced (unresectable or metastatic) pancreatic neuroendocrine tumours Radiologic documentation of disease progression in the previous 12 months Measurable disease (RECIST, vers. 1.0) World Health Organization (WHO) performance status of 2 or less Adequate bone marrow, renal, and hepatic function Adequately controlled lipid and glucose concentrations Exclusion criteria: Hepatic‐artery embolisation within 6 months before enrolment or within 1 month if there were other sites of measurable disease or cryoablation or radiofrequency ablation of hepatic metastasis within 2 months before enrolment Severe or uncontrolled medical conditions Prior therapy with an mTOR inhibitor Long‐term treatment with glucocorticoids or other immunosuppressive agents RADIANT‐3 overall population: Total patients: 410 Median age (everolimus vs. placebo): 58 vs. 57 Women % (everolimus vs. placebo): 47 vs 42 WHO performance status 0 (everolimus vs. placebo): 67% vs. 66% Well differentiated % (everolimus vs. placebo): 82 vs. 84 Moderately differentiated % (everolimus vs. placebo): 17 vs. 15 Liver involvement, % (everolimus vs. placebo): 92% vs. 92% Functional tumours (overall): 24% Prior therapy for NET, %: Radiotherapy (everolimus vs. placebo): 23 vs. 20 Chemotherapy (everolimus vs. placebo): 50 vs. 50 Somatostatin analogue therapy (everolimus vs. placebo): 49 vs. 50 RADIANT‐3 Japanese subgroup: Total patients: 40 Median age (everolimus vs. placebo): 45 vs. 53 Women % (everolimus vs. placebo): 44 vs. 53 WHO performance status 0 (everolimus vs. placebo): 87% vs. 88% Well differentiated % (everolimus vs. placebo): 100 vs. 94 Moderately differentiated % (everolimus vs. placebo): 0 vs. 6 Prior therapy for NET, %: Radiotherapy (everolimus vs. placebo): 13 vs. 12 Chemotherapy (everolimus vs. placebo): 61 vs. 53 Somatostatin analogue therapy (everolimus vs. placebo): 22 vs. 35
Interventions	Intervention group (overall: 207/410; Japanese subgroup: 23): oral everolimus, at a dose of 10 mg once daily, in conjunction with best supportive care (e.g. somatostatin analogue therapy) Control group (overall: 203/410; Japanese subgroup: 17): oral matching placebo in conjunction with best supportive care (e.g. somatostatin analogue therapy) Length of treatment: until progression of the disease, development of an unacceptable toxic effect, drug interruption for 3 weeks or longer, or withdrawal of consent Patients who had been assigned to placebo initially could switch to open‐label everolimus after documented progression of disease (RECIST). Doses were delayed/reduced if patients had clinically significant adverse events that were considered to be related to the study treatment.
Outcomes	Primary endpoint: Progression‐free survival (RECIST) Secondary endpoints: Confirmed objective response rate (RECIST) Duration of response Overall survival Safety Assessments: Tumour measurements (computed tomography or magnetic resonance imaging): at baseline and every 12 weeks Safety assessments: monitoring and recording of all adverse events, haematologic and clinical biochemical levels and vital signs, and physical examinations every 4 weeks Data collection: sponsor's data management Data analysis: sponsor's statistical team
Notes	Funding/Sponsor: Novartis Oncology and Novartis Pharma K.K.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Centralised randomisation through interactive voice response system. Stratified by performance status and prior treatment (+/‐ chemotherapy)
Allocation concealment (selection bias)	Low risk	Centralised randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blinded trial design with same schemes for each study group
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Endpoints were documented by the local investigator according to RECIST, with independent adjudicated central assessment.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants accounted for in final analysis
Selective reporting (reporting bias)	Unclear risk	Not all secondary endpoints mentioned in the study protocol were published.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Randomised, double‐blind, cross‐over trial
Participants	Inclusion criteria Histologically proven neuroendocrine tumour with liver metastases One symptom related to the tumour: Symptoms had to interfere with daily activity. Exclusion criteria Previous therapy with octreotide Total patients: 11 Mean age (overall): 56.5 Women (overall): 55% Primary tumour site (overall): Pancreas: 18% Small intestine: 82%
Interventions	Intervention group: 100 μg octreotide, subcutaneous injection, twice daily for 4 weeks Control group: placebo, subcutaneous injection, twice daily for 4 weeks After the first 4 weeks, patients were shifted from placebo to octreotide and vice versa.
Outcomes	Endpoints: Quality of life Side effects Changes in urine 5‐HIAA concentration Change in diarrhoea and flushing episodes Assessments: Flushes and diarrhoea: daily 1 week prior to start of the study and during the duration of the entire study Biochemical marker: at the start and after 4 and 8 weeks Quality of life: at the start and after 4 and 8 weeks
Notes	The study drug was supplied by Sandoz AG.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was performed by the study drug supplier.
Allocation concealment (selection bias)	Unclear risk	At half time, the groups shifted from active treatment to placebo and vice versa. The inclusion criterion "the symptoms had to interfere with daily activity" was not precisely defined.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Double‐blind trial, but patients knew they would get both placebo and active treatments. Yet they did not know the order of administration.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information given
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Several patients left the study but it remained unclear whether they were accounted for in final analysis.
Selective reporting (reporting bias)	Unclear risk	No study protocol available and no clear endpoints stated
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Prospective randomised multicentre study 10 centres in Sweden Randomisation stratified by the presence or absence of carcinoid heart disease on ultrasonography and urinary 5‐HIAA level Start: April 1991 Enrolment closed: July 1998 Follow up: until April 2001
Participants	Inclusion criteria Verified midgut carcinoid tumour Primary tumour excised at surgery Presence of liver metastases on ultrasonography or CT Carcinoid symptoms (flush and/or diarrhoea) Urinary 5‐HIAA ≥ twice upper reference value Age ≤ 75 years Performance status WHO classification < IV Exclusion criteria Other concomitant malignancy Severe coronary heart disease Total patients: 68 Mean age (study arm 1 vs. 2): 62 vs. 63 Women (overall): 56% Ki‐67 index: not reported All patients underwent hepatic arterial embolisation before randomisation.
Interventions	Study arm 1 (35/68): Octreotide 100 µg twice daily. If there were persistent carcinoid symptoms, the dose was increased up to 200 µg three times daily. Study arm 2 (33/68): Octreotide 100 µg twice daily. If there were persistent carcinoid symptoms, the dose was increased up to 200 µg three times daily. With interferon‐α. Interferon treatment started with 3 × 10⁶ units on each of 3 days per week and was increased to a maximal dose of 5 × 10⁶ units on each of 5 days per week.
Outcomes	Endpoints: Death Progressive tumour growth Life‐threatening side effects Assessments: Clinical examination and laboratory investigations every 3 months Ultrasonography or CT of the liver and non‐invasive heart examination every 6 months
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stratified according to the presence or absence of carcinoid heart disease on ultrasonography (stenosis and/or regurgitation in the pulmonary and tricuspid valves) and urinary 5‐HIAA level more or less than 500 μmol per 24‐h; but it remained unclear how the randomisation process was performed.
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	Personnel and participants were not blinded.
Blinding of outcome assessment (detection bias) All outcomes	High risk	No evidence for independent assessment
Incomplete outcome data (attrition bias) All outcomes	Low risk	All patients accounted for in ITT analysis
Selective reporting (reporting bias)	Low risk	No protocol available. Every stated endpoint was reported in the results section.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Randomised phase II trial Randomisation: 1:1
Participants	Inclusion criteria: Advanced pNET Total patients: 150 Median age: 59 Women: 44% ECOG 0: 57%; ECOG 1: 43% Grade: not reported Functionality: not reported
Interventions	Study arm E: everolimus, 10 mg, p.o. qd. Study arm E + B: everolimus, 10 mg, p.o., qd; with bevacizumab, 10 mg/kg, i.v. q2 weeks. All patients received octreotide.
Outcomes	Primary endpoint: Progression‐free survival Secondary endpoints: Overall survival Response rate Safety
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information given
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	Two different application schemes for the study drugs
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information given
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information given
Selective reporting (reporting bias)	Unclear risk	Insufficient information given
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Randomised, global, multicentre, open‐label, phase 2 trial 1:1 randomisation, stratified by prior SSA treatment (yes or no) and the presence of elevated biomarkers at baseline Start: July 2011 Closed: December 2021
Participants	Inclusion criteria: Age ≥ 18 years Histologically confirmed, well differentiated, advanced pNET [WHO grade 1 or 2] Radiological documentation of disease progression within 12 months before randomisation Measurable disease (RECIST v1.0) WHO performance status ≤ 2 Adequate bone marrow, renal and hepatic function Exclusion criteria: Prior treatment with mTOR inhibitors Clinical requirement of SSA treatment Total patients: 160 Median age (everolimus + pasireotide LAR vs. everolimus): 57 vs. 59 Women % (everolimus + pasireotide LAR vs. everolimus): 51 vs. 42 WHO performance status 0‐1 (everolimus + pasireotide LAR vs. everolimus): 100% vs. 96% Grade 1 or 2, % (everolimus + pasireotide LAR vs. everolimus): 97.5 vs. 97.5 Functionality: not reported Prior antineoplastic treatment, % (everolimus + pasireotide LAR vs. everolimus): 65 vs. 62 Prior SSA treatment, % (everolimus + pasireotide LAR vs. everolimus): 33 vs. 33
Interventions	Study arm 1 (79/160): everolimus, 10 mg/day, per oral; with pasireotide LAR, 60 mg/28 days, intramuscular injection Study arm 2 (81/160): everolimus, 10 mg/day, per oral Length of treatment: until radiologically documented disease progression, start of a new anticancer therapy, intolerable toxicity or withdrawal of consent Dose modifications were permitted for any adverse event suspected to be drug related. Cross‐over: not allowed
Outcomes	Primary endpoint: Treatment effect on progression‐free survival (RECIST v 1.0) Secondary endpoints: Objective response rate Disease control rate Overall survival Pharmacokinetics Safety Biomarker response was evaluated as an exploratory analysis. Assessments: Tumour assessments: at screening and every 12 weeks from date of randomisation until radiologically documented disease progression Clinical suspicion of disease progression at any time required a physical examination and radiological confirmation. Patients who discontinued the study treatment prior to progression of disease continued to have tumour assessments performed every 12 weeks from randomisation until radiologically documented disease progression or start of a new antineoplastic therapy. Patients who discontinued the study treatment and were no longer followed for tumour evaluation were contacted every 12 weeks for survival. Blood samples: before and during treatment at prespecified time points for assessing pharmacodynamic markers
Notes	Funding: Novartis Pharmaceuticals Corporation
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation with stratification, but unclear how it was performed
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	Trial was open‐label.
Blinding of outcome assessment (detection bias) All outcomes	High risk	No evidence of independent assessment of radiological outcomes
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants accounted for; one patient in the combination arm did not receive study treatment.
Selective reporting (reporting bias)	Low risk	No protocol available. Not all endpoints mentioned were shown in the official publication, but can be found in the supplementary data.
Other bias	Low risk	No other potential sources of bias found

PERMALINK

Treatment for gastrointestinal and pancreatic neuroendocrine tumours: a network meta‐analysis

Martin A Walter

Cédric Nesti

Marko Spanjol

Attila Kollár

Lukas Bütikofer

Viktoria L Gloy

Rebecca A Dumont

Christian A Seiler

Emanuel R Christ

Piotr Radojewski

Matthias Briel

Reto M Kaderli

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Summary of findings

Background

Description of the condition

Description of the intervention

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Summary of findings and assessment of the certainty of the evidence

Results

Description of studies

Results of the search

1.

Included studies

Excluded studies

Risk of bias in included studies

2.

3.

Allocation

Random sequence generation

Allocation concealment

Blinding

Blinding of participants and personnel (performance bias)

Blinding of outcome assessment (detection bias)

Incomplete outcome data

Selective reporting

Other potential sources of bias

Effects of interventions

Summary of findings 1. Estimates of effects, ranking, and certainty of evidence for different treatment options compared with placebo for disease control in pancreatic neuroendocrine tumours (pNET).

Summary of findings 2. Estimates of effects, ranking, and certainty of evidence for different treatment options compared with placebo for progression‐free survival in pancreatic neuroendocrine tumours (pNET).

Summary of findings 3. Estimates of effects, ranking, and certainty of evidence for different treatment options compared with placebo for disease control in gastrointestinal neuroendocrine tumours (GI‐NET).

Summary of findings 4. Estimates of effects, ranking, and certainty of evidence for different treatment options compared with placebo for progression‐free survival in gastrointestinal neuroendocrine tumours (GI‐NET).

Treatment efficacy in pNETs

4.

1. Comparison of all treatment options from the network meta‐analysis of disease control in pancreatic neuroendocrine tumours (pNET).

5.

*Study characteristics*
Methods	International (12 countries), multicentre, randomised, double‐blind, placebo‐controlled phase III trial (TELESTAR) 1:1:1 randomisation
Participants	Inclusion criteria Age ≥ 18 years Histopathologically confirmed, well differentiated metastatic NETs Documented history of carcinoid syndrome Average of four or more bowel movements per day Stable‐dose SSA treatment (long‐acting release, depot or infusion pump) for ≥ 3 months before enrolment Exclusion criteria More than 12 watery bowel movements per day associated with volume contraction, dehydration or hypotension Enteric infection Karnofsky performance status ≤ 60% History of short bowel syndrome Clinically important baseline elevation in liver function tests Recently tumour‐directed therapy Total patients: 135 Mean age (A vs. B vs. C): 63 vs. 62 vs. 65 Women, % (A vs. B vs. C): 47 vs. 53 vs. 44 SSA therapy at study entry, %: Octreotide LAR (A vs. B vs. C): 67 vs. 89 vs. 73 Lanreotide depot (A vs. B vs. C): 33 vs. 11 vs. 27
Interventions	Study group A (45/135): placebo, oral doses, three times per day for 12 weeks Study group B (45/135): telotristat ethyl 250 mg, oral doses, three times per day for 12 weeks Study group C (45/135): telotristat ethyl 500 mg, oral doses, three times per day for 12 weeks Continued baseline SSA therapy for all 12 weeks Allowed rescue use of short‐acting octreotide and antidiarrhoeal agents After the study, all patients were offered treatment with telotristat ethyl 500 mg, three times per day in a 36‐week open‐label extension.
Outcomes	Primary endpoint: Mean reduction from baseline in daily bowel movements averaged over 12 weeks (self‐reported) Secondary endpoints: Change from baseline in u5‐HIAA at week 12 Number of daily flushing episodes (self‐reported) Abdominal pain severity (on a scale of 0 to 10) averaged over 12 weeks (self‐reported) Additional efficacy endpoints: Quality of life (self‐reported) Rescue short‐acting SSA use (self‐reported) Stool consistency (self‐reported) Proportion of days with urgency to defecate (self‐reported) Safety Assessments: Screening period of 3 or 4 weeks for baseline symptoms Self‐reporting by daily electronic diaries
Notes	Study was supported by Lexicon Pharmaceuticals.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information given
Allocation concealment (selection bias)	Unclear risk	No information regarding allocation concealment; all study groups with the same number of patients
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind trial design
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessment was done by self‐reporting in the majority of endpoints.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised patients accounted for in the efficacy analyses in ITT fashion
Selective reporting (reporting bias)	Low risk	No study protocol available, but all stated endpoints were reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Randomised, double‐blinded, placebo‐controlled trial Randomisation was stratified for diagnosis (gastrinoma vs. insulinoma) and for type of excision (enucleation vs. resection) Start: 1989 Closed: 1991
Participants	Inclusion criteria: Resection of pancreatic endocrine tumour at the National Institutes of Health Exclusion criteria: Diabetes No pancreatic incision Total patients: 21 Median age (octreotide vs. placebo): 47 vs. 46 Women, % (octreotide vs. placebo): 70 vs. 27 Functionality, % (octreotide vs. placebo): 100 vs. 100 Tumour grade: not reported Prior treatment for NET: not reported
Interventions	Experimental arm (10/21): octreotide, subcutaneous injection, beginning the day of surgery. Dosage: day 1, 50 μg every 8 hours; day 2, 100 μg every 8 hours; day 3 and for the duration of treatment, 150 μg every 8 hours Control arm (11/21): saline solution, subcutaneous injection, same schedule and in a volume to match that of the experimental arm Octreotide and saline solution injections were continued until 3 days after drain removal. Drain removal was regulated by a standardised algorithm.
Outcomes	Endpoints: Adverse reactions Development of gallstones Daily drain output Days to drain removal Total drainage Complications related to pancreatic drainage Assessments: Daily blood glucose tests Ultrasonography for assessment of gallstones before operation, each month during treatment and after drain removal Amylase content in drain fluid was measured on postoperative days 1, 3 and 7. It was not stated how the other endpoints were measured.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation with stratification, but unclear how it was performed
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Same protocols for each study arm
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information given
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants accounted for in final analysis
Selective reporting (reporting bias)	Low risk	No protocol available, but all endpoints stated in the paper as measured were reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	1:1:1:2 randomisation Enrolment: August 2017 to February 2019
Participants	Inclusion criteria High tracer uptake in tumour on ⁶⁸Ga‐DOTATATE PET/CT, evaluated within 1 week before inclusion Histological confirmed or inoperable/metastatic NET White blood cells ≥ 3 × 10⁹/L Platelets ≥ 60 × 10⁹/L Hemoglobin ≥ 10 g/dL Serum creatinine clearance > 40 mL/min No pregnancy or lactation Age > 18 Total patients: 33 Age (A vs. B vs. C vs. D): 43 vs. 55 vs. 55 vs. 50 Women, % (A vs. B vs. C vs. D): 67 vs. 29 vs. 33 vs. 50 Primary tumour site, %: Pancreas (A vs. B vs. C vs. D): 50 vs. 43 vs. 50 vs. 50 Duodenum (A vs. B vs. C vs. D): 0 vs. 29 vs. 17 vs. 21 Rectum (A vs. B vs. C vs. D): 0 vs. 14 vs. 0 vs. 14 Lung (A vs. B vs. C vs. D): 17 vs. 0 vs. 17 vs. 0 Ovary (A vs. B vs. C vs. D): 17 vs. 0 vs. 0 vs. 0 CUP (A vs. B vs. C vs. D): 17 vs. 0 vs. 0 vs. 7 MEN 1 (A vs. B vs. C vs. D): 0 vs. 0 vs. 17 vs. 0 Paraganglioma (A vs. B vs. C vs. D): 0 vs. 0 vs. 0 vs. 7 Pheochromocytoma (A vs. B vs. C vs. D): 0 vs. 14 vs. 0 vs. 0 Tumour grade, %: G1 (A vs. B vs. C vs. D): 50 vs. 29 vs. 17 vs. 21 G2 (A vs. B vs. C vs. D): 33 vs. 57 vs. 67 vs. 71 G3 (A vs. B vs. C vs. D): 17 vs. 14 vs. 17 vs. 7 Liver involvement, % (A vs. B vs. C vs. D): 100 vs. 100 vs. 83 vs. 100 Prior treatment, %: Surgery (A vs. B vs. C vs. D): 17 vs. 71 vs. 33 vs. 50 SSA (A vs. B vs. C vs. D): 83 vs. 29 vs. 83 vs. 36 Everolimus (A vs. B vs. C vs. D): 17 vs. 0 vs. 50 vs. 7 Tyrosine kinase inhibitor (A vs. B vs. C vs. D): 17 vs. 43 vs. 83 vs. 64 Chemotherapy (A vs. B vs. C vs. D): 50 vs. 43 vs. 50 vs. 43 Radiotherapy (A vs. B vs. C vs. D): 0 vs. 0 vs. 17 vs. 7 TACE (A vs. B vs. C vs. D): 17 vs. 14 vs. 17 vs. 21
Interventions	Group A (6/33): 3.7 GBq (100 mCi) ¹⁷⁷Lu‐DOTATATE, one dose Group B (7/33): 1.11 GBq (30 mCi) ¹⁷⁷Lu‐DOTA‐EB‐TATE, one dose Group C (6/33): 1.85 GBq (50 mCi) ¹⁷⁷Lu‐DOTA‐EB‐TATE, one dose Group D (14/33): 3.7 GBq (100 mCi) ¹⁷⁷Lu‐DOTA‐EB‐TATE, one dose
Outcomes	Endpoints: Tumour response referring to EORTC criteria Safety Assessments: ⁶⁸Ga‐DOTATATE PET/CT at baseline and 2–3 months post‐therapy Treatment‐related adverse events (AEs) recorded over a period of 2 months after the administration of PRRT Haematological parameters, liver and renal function at baseline, and 1‐week and 4‐week post‐therapy
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomised sequence was generated by computer.
Allocation concealment (selection bias)	Unclear risk	Was performed by different people
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No information given
Blinding of outcome assessment (detection bias) All outcomes	Low risk	All images were measured by the same physician who was masked to the clinical data.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All patients accounted for in analysis
Selective reporting (reporting bias)	Low risk	All stated endpoints were reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Prospective randomised trial 2 centres in France Central randomisation with an adaptive randomisation procedure stratified per centre and progression group Start: 2002 Closed: 2008 Follow‐up: 24 months after inclusion
Participants	Inclusion criteria Age ≥ 18 years Histologically confirmed endocrine liver metastases from midgut tumours Either progressive liver metastases within 12 months before the inclusion, i.e. progression of > 25% between two consecutive imaging procedures, or liver tumoural involvement of > 50% on CT scan Exclusion criteria Extrahepatic metastases, with the exception of lymph node involvement Liver dysfunction Renal dysfunction History of HAE or HACE Hepatic or portal vein thrombosis Cardiac insufficiency Unstable coronary disease Heart stroke within the previous 3 months Uncontrolled hyperthyroidism Karnofsky index < 70% Contrast allergy Pregnant, breastfeeding or fertile women without contraceptive Total patients: 26 Median age (HAE vs. HACE): 56 vs. 65 Women, % (HAE vs. HACE): 36 vs. 42 Median Karnofsky index (overall): 90 Carcinoid syndrome, % (HAE vs. HACE): 79 vs. 67 Liver involvement, < 25%/25‐50%/> 50%, % (HAE vs. HACE): 43/36/21 vs. 58/25/17 Ki‐67 index, ≤ 2%/3‐5%/6‐10%/unknown, % (overall): 62/19/4/15 Resection of primary tumour, % (HAE vs. HACE): 86 vs. 83 Previous resection of liver metastases, % (HAE vs. HACE): 14 vs. 17 Concomitant treatment with SSA, % (HAE vs. HACE): 79 vs. 67 Primary tumour location unknown (overall): 15%
Interventions	Study arm 1 (14/26): hepatic arterial embolisation (HAE): transfemoral, embolisation with gelatin sponge particles Study arm 2 (12/26): hepatic arterial chemoembolisation (HACE): transfemoral, doxorubicin (50 mg/m²) dissolved in normal saline and combined with 10–15 mL of iodised oil, injected into the branches of the hepatic artery distal to the gastroduodenal artery, followed by embolisation with gelatin sponge particles Treatment was administered after randomisation and was repeated 3 months thereafter. Carcinoid syndrome had to be controlled by somatostatin analogues. Patients with carcinoid syndrome were administered octreotide 200 µg subcutaneously before the procedure and every 8 h afterward during 48 h to prevent a carcinoid crisis.
Outcomes	Primary endpoint Progression‐free survival rate Secondary endpoints Side effects Morphological and biological response rates Overall survival rate Assessments Tumour assessment by the same imaging method throughout the follow‐up period after 3, 6, 12, 18 and 24 months or earlier if clinically indicated Physical examination, pain assessment using a visual analogue scale, and analgesic intake and toxicity were recorded after 3, 6, 12, 18 and 24 months.
Notes	Funding: Novartis Pharma
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was performed centrally with the use of an adaptive randomisation procedure stratified per centre and progression group, but it remained unclear how this adaptive randomisation procedure worked.
Allocation concealment (selection bias)	Low risk	Randomisation was performed centrally.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and personnel were not blinded.
Blinding of outcome assessment (detection bias) All outcomes	High risk	No evidence of independent assessment
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	All patients accounted for in primary endpoint. For morphological response and for biological response only 23 and 20 patients were evaluable.
Selective reporting (reporting bias)	Low risk	No study protocol available, but every endpoint mentioned in 'methods' was reported in 'results'.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Multicentre, randomised trial 13 United Kingdom centres 1:1 randomisation by stratified random block method Stratification factors: functional or non‐functional tumour, previous somatostatin analogues/interferon treatment versus none and known primary tumour site versus unknown Enrolment start: November 2006 Enrolment closed: October 2010
Participants	Inclusion criteria Chemo‐naive patients Histologically confirmed, unresectable, advanced and/or metastatic NETs of the pancreas, other gastrointestinal foregut, or unknown primary site suggestive of abdominal foregut origin Measurable disease by RECIST (version 1.0) ECOG performance status ≤ 2 Adequate bone marrow, hepatic and renal function with creatinine clearance > 60 mL/min Total patients: 86. Median age (CapStrep vs. CapStrepCis): 57 vs. 59 Women, % (CapStrep vs. CapStrepCis): 39 vs. 45 Site of origin, %: Pancreas (CapStrep vs. CapStrepCis): 45.5 vs. 50 Foregut (CapStrep vs. CapStrepCis): 20.5 vs. 19 Unknown (CapStrep vs. CapStrepCis): 33 vs. 31 Functional tumour, % (CapStrep vs. CapStrepCis): 30 vs. 43 Liver metastases, % (CapStrep vs. CapStrepCis): 93 vs. 81 Ki‐67 index (%) ≤ 9/10‐24/≥ 25, % (CapStrep vs. CapStrepCis): 46/33/21 vs. 50/27/24 Prior treatment received, %: SSA and interferon (CapStrep vs. CapStrepCis): 2 vs. 0 SSA (CapStrep vs. CapStrepCis): 27 vs. 29 Interferon (CapStrep vs. CapStrepCis): 2 vs. 0 None (CapStrep vs. CapStrepCis): 68 vs. 71
Interventions	CapStrep regimen group (44/86): capecitabine 625 mg/m² administered orally, twice daily on days 1–21, and streptozocin 1.0 g/m² (2‐h infusion intravenously in normal saline) on day 1 CapStepCis regimen group (42/86): capecitabine 625 mg/m² administered orally, twice daily on days 1–21, and streptozocin 1.0 g/m² (2‐h infusion intravenously in normal saline) on day 1 plus cisplatin 70 mg/m² (2‐h infusion intravenously in normal saline with hydration) on day 1, directly after the streptozocin infusion Treatment duration: six cycles (and beyond six cycles if there was evidence of benefit)
Outcomes	Primary endpoint: Objective response rate Secondary endpoints: Biochemical response Safety Progression‐free survival Overall survival Quality of life Assessments Adverse events: every cycle Disease progression: every 12 weeks Survival: every 12 weeks Tumour assessments with CT scans: baseline, every three cycles while on treatment and every 12 weeks until progression Retrospective central radiology review was undertaken for objective tumour response assessments in 10% of randomly selected patients who completed at least three treatment cycles. 24‐h urinary 5‐hyroxyindoleacetic acid (5‐HIAA) and serum chromogranin A (CgA): prior to treatment and if above the normal range were repeated every three cycles while on treatment and at 12 weeks from the end of treatment Patient quality of life: before randomisation, after three and six cycles of treatment or at the time of stopping treatment and at 12 weeks post‐treatment
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation by stratified random block method
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	No evidence for blinding
Blinding of outcome assessment (detection bias) All outcomes	High risk	Retrospective central radiology review in 10% of randomly selected patients who completed at least three treatment cycles
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No study protocol available, but every mentioned endpoint in 'methods' is reported in 'results'.
Selective reporting (reporting bias)	High risk	Four patients were not included in the primary analysis. There was a big loss of patient numbers in the 'quality of life' endpoint.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Multicentric, randomised trial 28 centres Randomisation was stratified according to: performance status, either functioning or nonfunctioning tumour, and use of either laboratory assay or measurable feature to assess objective response. Start: December 1972 Closed: December 1978
Participants	Inclusion criteria: Not resectable islet‐cell carcinoma Histologic proof of residual, recurrent or metastatic carcinoma Measurable malignant disease (laboratory assay (e.g. serum gastrin) or measurable area of known tumour). In case of liver involvement (bioptically confirmed), with liver edge extension at least 5 cm below the xiphoid or costal margin, malignant hepatomegaly was accepted as a measurable feature. Radioactive liver scans were also accepted, if a clearly demarcated perfusion defect of at least 5 cm was detectable. Recommendation to only include patients with symptoms or disability resulting from the malignant disease Exclusion criteria: Diagnosis without biopsy confirmation Prior therapy with either fluorouracil or streptozocin Active infectious process Severe malnutrition Frequent vomiting Leukocyte count below 4500 per cubic millimetre Platelet count below 150,000 per cubic millimetre Renal disease (creatinine above 1.5 mg/mL [133 μmol/L] or urea nitrogen above 30 mg per 100 mL [10.7 mmol/L] Exploratory surgery with biopsy within two weeks of treatment start and a resection or anastomosis within three weeks of treatment start Previous radiation therapy or treatment with cytotoxic drugs within one month after registration Present haematologic or renal toxic effect from therapy 103 patients were randomised; 19 were excluded. Mean age (study arm 1 vs. 2): 52 vs. 54 Women % (1 vs. 2): 57 vs. 45 ECOG performance status 0‐1 (1 vs. 2): 71% vs. 71% Functional tumour % (1 vs. 2): 52 vs. 44% Tumour grade: not reported. Prior chemotherapy (overall): 2%
Interventions	Study arm 1 (42/84): streptozocin, by rapid intravenous injection, 500 mg per square metre of body‐surface area, for five consecutive days, repeated every 6 weeks if disease improved or remained objectively stable Study arm 2 (42/84): streptozocin, by rapid intravenous injection, 500 mg per square metre of body‐surface area, for five consecutive days; and fluorouracil, by rapid intravenous injection, 400 mg per square metre of body‐surface area, for five consecutive days, concurrently with streptozocin, repeated every 6 weeks if disease improved or remained objectively stable Streptozocin dosage was reduced by 50%, if the patient had severe nausea and vomiting or any evidence of renal toxicity was present. Streptozocin was discontinued if these problems persisted after dose reduction. Flourouracil dosage was reduced by 25% if severe leukopenia or thrombocytopenia was present. Phenothiazine antiemetics were recommended for prophylaxis and therapy of nausea and vomiting.
Outcomes	No clear endpoints were set. Assessments: Therapeutic results: every six weeks before initiation of the next course of therapy White cell and platelet counts at weekly intervals after therapy Renal and liver function before of initiation of each course of therapy
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation with stratification, but unclear how it was performed
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information given
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information given
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	In two patients, the investigators failed to record serial tumour measurements.
Selective reporting (reporting bias)	High risk	19 patients were excluded after randomisation.
Other bias	High risk	Investigator‐dependent measurement methods were used.

*Study characteristics*
Methods	Multinational, randomised trial 25 centres in 3 countries Randomisation: stratified according to ECOG performance score and indicator of response (measurable tumour or laboratory assays) Start: November 1978 Closed: June 1985
Participants	Inclusion criteria: Histologic proof of unresectable or metastatic islet‐cell carcinoma Measurable indicator of response to therapy: 1) tumour on physical examination, or 2) chest films or well‐defined metastatic lesions in the liver on radioisotope or CT scanning > 5 cm, or 3) malignant hepatomegaly if the liver edge is at least 5 cm below the xiphoid process or the costal margins during quiet respiration, or 4) for patients without measurable tumour, laboratory assays demonstrating excessive hormone production Exclusion criteria: ECOG performance score of 4 Severe nutritional impairment Major surgery within three weeks Previous therapy with any of the study agents Chemotherapy or radiation therapy within the previous month Active infection Leukocyte count < 4×10⁹ per litre or a platelet count < 150 × 10⁹ per litre Active heart disease Serum creatinine level > 132.6 mmol per litre (1.5 mg per decilitre) or a blood urea nitrogen level > 10.7 mmol per litre (30 mg per decilitre) Any elevation of serum bilirubin Any other concurrent or recent malignant disease except cutaneous epitheliomas or cervical carcinoma in situ Total patients: 125; 18 patients were subsequently found to be ineligible and two withdrew from the study. Median age (1 vs. 2 vs. 3): 57 vs. 51 vs. 53 Women % (1 vs. 2 vs. 3): 61 vs. 41 vs. 53 ECOG performance status 0‐1 (1 vs. 2 vs. 3): 70% vs. 71% vs. 71% Nonfunctional tumours (overall): 52.4% Tumour grade: not reported Prior therapy for NET: not reported
Interventions	Study arm 1 (33/105): chlorozotocin, intravenous injection, 150 mg per square metre of body‐surface area, every seven weeks Study arm 2 (34/105): streptozocin, intravenous injection, 500 mg per square metre, for five consecutive days, every six weeks. And, fluorouracil intravenous injection, 400 mg per square metre, for five days, concurrently with streptozocin Study arm 3 (38/105): doxorubicin along with streptozocin, intravenous injection, 50 mg per square metre, days 1 and 22 of each six‐week treatment cycle, with a maximal total dose of 500 mg per square metre Dosages of streptozocin or chlorozotocin were reduced if 1) severe nausea or vomiting, stomatitis, diarrhoea, leukopenia, or thrombocytopenia occurred, or 2) creatinine level became elevated or persistent proteinuria developed. If these abnormalities persisted, treatment with these agents was discontinued. Length of therapy: until disease progression was noted
Outcomes	Endpoints: Disease progression Survival Rates of regression Assessments: Re‐evaluation every seven weeks for study arm 1 Re‐evaluation every six weeks for study arms 2 and 3 Leukocyte and platelet counts weekly, serum creatinine and urinalyses before each cycle of therapy
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation with stratification, but unclear how it was performed
Allocation concealment (selection bias)	Unclear risk	Insufficient information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	Different application schemes and control intervals for each study arm
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information given
Incomplete outcome data (attrition bias) All outcomes	High risk	20 patients were excluded after randomisation.
Selective reporting (reporting bias)	Low risk	No study protocol available, but the endpoints mentioned in 'methods' were reported in 'results'.
Other bias	High risk	Investigator‐dependent measurement methods were used.

*Study characteristics*
Methods	Prospective, open, comparative, cross‐over study 15 centres in France
Participants	Inclusion criteria Histologically confirmed carcinoid tumours At least one of the following symptoms: diarrhoea (at least 3 stools every 24 hours) or flushes (at least 1 flush every 24 hours) Exclusion criteria Previous treatment with SSA or discontinuation for a sufficient time to allow reappearance of clinical symptoms Symptoms of bowel obstruction Surgery for the tumour or its metastases scheduled in the 3 months after inclusion Withdrawal Development of bowel obstruction Requirement of another therapy (e.g. radiotherapy, chemotherapy, immunotherapy or chemoembolisation) Total patients: 33 Age (A vs. B): 63 vs. 64 Women, % (A vs. B): 50 vs. 53 Previous treatment with octreotide/lanreotide, % (A vs. B): 63/13 vs. 59/0 Primary tumour site, %: Intestine (A vs. B): 63 vs. 76 Pancreas (A vs. B): 0 vs. 6 Lung (A vs. B): 19 vs. 0 Unknown (A vs. B): 19 vs. 0 Stomach (A vs. B): 0 vs. 12 Ovary (A vs. B): 0 vs. 6 Metastases, % (A vs. B): 100 vs. 100
Interventions	Group A (16/33): octreotide, 200 mg, subcutaneously twice or thrice daily for 30 days followed by lanreotide, 30 mg, intramuscularly every 10 days on days 1, 10, and 20 for 30 days Group B (17/33): lanreotide, 30 mg, intramuscularly every 10 days on days 1, 10, and 20 for 30 days followed by octreotide, 200 mg, subcutaneously twice or thrice daily for 30 days A wash‐out period of at least 3 days was applied between the two treatments. Antidiarrhoea agents were prohibited during the study period.
Outcomes	Endpoints: Quality of life Clinical symptoms Tumour markers Adverse events Assessments: Day 1 and day 30 of each treatment period
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information given
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	Personnel and participants were not blinded.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Analysis was performed by an independent expert who was blinded to the treatment group.
Incomplete outcome data (attrition bias) All outcomes	High risk	5 of 33 (15%) randomised patients were excluded; therefore 28 patients accounted for in efficacy analysis (14 patients per group)
Selective reporting (reporting bias)	Low risk	No study protocol available, but all stated endpoints were reported.
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	1:1 randomised trial, stratified by urinary 5‐hydroxyindoleacetic acid level, sex and age
Participants	Inclusion criteria Malignant carcinoid tumour Total patients: 20 Mean age (overall): 61.5 Women, % (overall): 45% Liver metastases (overall): 100% Carcinoid symptoms (overall): 100% Primary tumour location (overall): 95% ileum, 5% bronchial Previous therapy for NET: not reported
Interventions	Study arm 1 (10/20): streptozotocin, 1 g, intravenous, for three consecutive days in combination with 5‐fluorouracil, 400 mg/m². Treatment was repeated every 6 weeks. Study arm 2 (10/20): interferon, 6 MU daily, subcutaneous injection; for the first three days, only half the dose was given. No other treatment for carcinoid syndrome was used. Cross‐over of 8 patients in the study arm 1 to study arm 2 after 6 months; and of 1 patient from study arm 2 to study arm 1
Outcomes	No clear primary or secondary endpoints stated Endpoints reported: Objective tumour response Adverse reactions Assessments: CT and ultrasound of the abdomen prior to treatment start and then every 3rd month Laboratory analysis prior to every new course of chemotherapy and every 3rd month in patients on interferon treatment Daily number of flush attacks and episodes of diarrhoea were monitored and evaluated in the subjective response.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No sufficient information given
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	Patients and personal not blinded. Different application and assessment schemes
Blinding of outcome assessment (detection bias) All outcomes	High risk	No evidence for independent evaluation
Incomplete outcome data (attrition bias) All outcomes	High risk	All patients accounted for in analysis, but group cross‐over was done. Not mentioned if ITT or per‐protocol analysis was performed
Selective reporting (reporting bias)	Unclear risk	No clear endpoints stated. No study protocol available
Other bias	Low risk	No other potential sources of bias found

*Study characteristics*
Methods	Non‐comparative randomised open‐label parallel‐group phase II trial 2:1 randomisation to PRRT/CAPTEM (experimental arm) vs. PRRT (mNETs control) and CAPTEM (pNETS control) Enrolment: December 2015–November 2018
Participants	Inclusion criteria Pancreatic and midgut neuroendocrine tumours Total patients: 75
Interventions	Experimental arm (33 mNETs and 19 pNETS/75): 7.8 GBq LuTate day 10, 8 weekly x 4, with twice a day oral CAP 750 mg/m² on days 1‐14 & TEM 75 mg/m² on days 10‐14, 8 wkly x 4 mNETs control (14/75): PRRT, 8 weekly x 4 pNETS control (9/75): CAPTEM, 8 weekly x 4
Outcomes	Primary endpoint: Progression‐free survival Secondary endpoints: Objective tumour response rate Clinical benefit rate Toxicity Quality of life
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information given
Allocation concealment (selection bias)	Unclear risk	No information given
Blinding of participants and personnel (performance bias) All outcomes	High risk	Open‐label trial
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information given
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No information given
Selective reporting (reporting bias)	Low risk	All stated endpoints were reported.
Other bias	Low risk	No other potential sources of bias found