Evidence summary

Population and studies description

This interventional procedures overview is based on about 1,965 patients from 9 studies reported in 8 publications. The evidence is comprised of 2 systematic review and meta-analyses (Frilling, 2019a; Ngo, 2021), a US-based multicentre retrospective comparative study of 248 people (Egger, 2020), a retrospective analysis of 170 people in an international prospective registry (RESiN Registry; Wong, 2022), an international retrospective case series of 244 people (Braat, 2019), a single-arm, single-centre trial of 30 people based in the Netherlands (Braat, 2020), a retrospective analysis of UK single-centre prospective data from 24 people (Frilling, 2019b), a retrospective case series of 93 people in the US (Tomozawa, 2018) and a prospective case series of 30 people from a single centre in the US (Cramer, 2017).

This is a rapid review of the literature, and a flow chart of the complete selection process is shown in Figure 1 in the appendix. This overview presents 9 studies reported in 8 publications as the key evidence in table 2 and table 3, and lists 74 other relevant studies in table 5.

Evidence evaluating SIRT against a comparator intervention was reported in a meta-analysis (Ngo, 2021) and in the retrospective comparative study of 248 people (Egger, 2020). The retrospective study of 248 people was included in the meta-analysis but includes additional outcome data. Of the total sample of 1,965 people in this overview, 1,316 had SIRT and 619 people had a comparative intra-arterial treatment (TACE). One study compared toxicity outcomes of SIRT when done as a unilobar treatment with bilobar treatment (Tomozawa, 2018). In this study, 48 people had unilobar and 45 had bilobar treatment.

The meta-analysis comparing SIRT with TACE included 6 retrospective case series published between 2013 and 2020 (Ngo, 2021), including the retrospective comparative study of 248 people that was also included in the main evidence (Egger, 2020). The second meta-analysis reported findings from 27 studies published between 2006 and 2018 (Frilling, 2019a). Both prospective and retrospective data from studies with between 6 and 148 people who had SIRT were included in this study.

Demographics

People's age was reported in 6 studies (Braat, 2020; Cramer, 2017; Egger, 2020; Frilling, 2019b; Tomozawa, 2018; Wong, 2022). Average age in these studies ranged from 58 to 66, and mostly had little variation around this within their samples, except for the prospective case series of 30 people in which the age range was 25 to 76 (Cramer, 2017). The average age of NET diagnosis in the UK is between 50 and 60 years. Between 27% and 57% of samples were women, among the studies that reported this information. The Surveillance, Epidemiology and End Results Programme (US) suggests slightly more women than men are diagnosed with NETs. UK hospital episode statistics (2021 to 2022) show that more women than men were diagnosed with secondary malignant neoplasm of liver and intrahepatic bile duct (C78.7). The meta-analyses (Frilling, 2019a; Ngo, 2021) and the retrospective case series of 244 people (Braat, 2019) did not report the age or gender distribution of the sample.

Follow up

When average follow up was reported, it ranged between 6 months (Braat, 2020) and 34 months (Egger, 2020). The 4 largest primary research studies reported data collected in registries or databases over several years; the RESiN registry analysis used data between 2015 and 2021 (Wong, 2022); the comparative retrospective case control study reported data between 2000 and 2018 (median=34 months; Egger, 2020); the retrospective study of 244 people used data collected between 2004 and 2016, with up to 12 years of follow up (Braat, 2019); and the retrospective case series of 93 people reported all patient data (until death or last known follow-up) between 2007 and 2015.

Previous therapies

Six of 9 studies reported whether patients had previous treatment. Most patients had at least 1 type of intervention before SIRT. Treatment history before SIRT was heterogeneous within and between studies.

The systematic review and meta-analysis comparing SIRT with TACE reported that history of resection or ablation was similar between groups but the SIRT group were more likely to have had previous systemic chemotherapy and octreocide therapy (both p=0.009; Ngo, 2021).

In the RESiN registry study (Wong, 2022), 14% of people had previous hepatic resection, 37% had previous liver-directed therapy, 85% had previous cytostatic or systemic therapy. This was non-exclusive meaning some people had multiple previous treatments.

In the retrospective case series of 244 people (Braat, 2019), 95% of the sample had multiple treatments before SIRT, 3% had surgical resection and 2% had no previous intervention.

In the single-arm trial of 30 people (Braat, 2020), everyone had 4 cycles of PRRT within 20 weeks before they had SIRT. Also, 8% of people had previous liver-directed therapy, 38% had surgical resection of the primary tumour, 71% had somatostatin analogues, 9% had sunitinib, 6% had everolimus and 6% had external-beam radiotherapy to the primary tumour.

In the retrospective analysis of prospectively collected UK data from 24 people (Frilling, 2019b), everyone had at least 1 previous treatment. Most commonly these were surgery, TACE, systemic chemotherapy, somatostatin analogues and PRRT.

In the retrospective case series of 93 people (Tomozawa, 2018), 62% of people had previous therapies. Most commonly, people had TACE or TAE (26%) and resection (23%). Other therapies included systemic chemotherapy (16%), and ablation (38%).

In the prospective case series of 30 people (Cramer, 2017) previous treatments included surgical research of the primary NET (n=13), surgical hepatic resection (n=8), chemotherapy (n=24) and locoregional therapy (n=6). These were not presented as percentages of the sample, so it is unclear what proportion of the sample had one, multiple or no previous treatment.

Details of previous therapies were not reported in the systematic review and meta-analysis of 27 studies (Frilling, 2019a), or in the retrospective comparative study of 248 people comparing SIRT with TACE (Egger, 2020).

Post-SIRT treatment

It was unclear whether people went on to have other therapies during their follow-up period in most studies. In the RESiN registry study (Wong, 2022), the authors reported that a median of 22 months (range 12 to 37) after SIRT, 5% (9/170) had PRRT. In the retrospective study of 244 people (Braat, 2019), 66% of people had additional therapy after SIRT. Time elapsed between SIRT and following treatments was not reported. In the retrospective analysis of prospectively collected UK data from 24 people (Frilling, 2019b), everyone had corticosteroids, ursodeoxycholic acid and omeprazole for 4 weeks after SIRT to minimize the incidence of post-radioembolisation syndrome. No other information was provided on other treatments during follow up in this study. In the retrospective case series of 93 people, 17% (9/52) of people who reached 1-year follow up had systemic chemotherapy and 1 person had TAE (Tomozawa, 2018).

Other studies did not comment on treatment following SIRT.

Disease profile

In the systematic review meta-analysis comparing SIRT with TACE, the location of primary tumour sites was similar; there was no statistically significant difference between groups in the proportion of people with pancreatic, lung and gastrointestinal primary tumours. The proportion of people with hepatic tumour burden greater than 50% was reported in 3 studies. This was not statistically different between groups (p=0.07) and there was no statistically significant difference in the proportion of people with grade 2 or 3 tumours (p=0.66, p=0.82, respectively). Proportion of people with functional status according to ECOG greater than or equal to 1 was not statically significant between groups (p=0.31).

In the retrospective study of 248 people comparing SIRT with TACE (Egger, 2020), the most common primary tumour sites were the small bowel (36%) and pancreas (25%). People who had TACE were more likely to have synchronous disease (p<0.05) and carcinoid syndrome (p<0.001). Tumour grade was higher in people with who had SIRT.

In the RESiN registry study, the most common primary tumour site was midgut (36%), followed by foregut (26%), pancreatic (24%) and hindgut (7%; Wong, 2022). Most tumours were graded as well differentiated (70%) and median tumour burden was 26%. 47% had bilobar tumours and 48% had extrahepatic metastases. According to ECOG assessment, baseline functioning was 0 in 48%, 1 in 37% and 2 or more in 15% of people.

In the retrospective case series of 244 people (Braat, 2019), 91% had progressive disease at baseline. The most common primary tumour sites were the pancreas (31%) and small bowel (35%). Most tumours were grade 1 (39%) or grade 2 (36%) and 66% had extrahepatic metastases. Tumour differentiation was not known for 71% of people. According to ECOG assessment, baseline functioning was 0 in 47%, 1 in 43%, and 2 or higher in 8%.

In the single-arm trial of 30 people (Braat, 2020), the most common primary tumour sites were the pancreas (32%) and small bowel (29%). Most tumours were grade 2 (65%; 35% were grade 1). Extrahepatic disease was present in 76% of people and 75% of people had less than 25% fractional liver involvement. According to ECOG assessment, baseline functioning was 0 in 59% of people, 1 in 38% of people, and 2 in 1 person (3%).

In the retrospective analysis of prospectively collected UK data from 24 people (Frilling, 2019b), the most common primary tumour sites were the small bowel (52%) and pancreas (25%). These were categorised as grade 1 in 46% and grade 2 in 42% of the sample. Everyone had progressive disease before treatment and 38% had extrahepatic metastases. At baseline, functioning score for everyone was less than 2 according to ECOG.

In the retrospective case series of 93 people (Tomozawa, 2017), 60% of tumours were classed as well differentiated and 16% were moderate to poorly differentiated. Most were carcinoid tumours (71%). Hepatic tumour burden was heterogeneous in this sample: 38% of people had less than 25% hepatic tumour burden, 33% of people had 25% to 50%, and 28% had more than 50%. Extrahepatic metastases were present in 34% of people. At baseline, functioning score according to ECOG was 0 in 51% of people, 1 in 44% and 2 in 5%.

In the prospective case series of 30 people (Cramer, 2017), the primary tumour location was most commonly unknown (37%) but 30% of primary tumours were in the pancreas. According to ECOG, everyone had a functioning score of 0 (47%) or 1 (50%).

Baseline data was not reported in the systematic review and meta-analysis of 27 studies (Frilling, 2019b).

Table 2 presents study details.

**Table 2 Study details**
Study no.	First author, date country	Patients (men: women; M:W)	Age	Study design	Inclusion criteria	Intervention	Follow up
1	Ngo, 2021 Included studies from Turkey, US, and one international study conducted in China, Germany and the US	n=643 (221 had SIRT and 422 had TACE; data gathered from 6 retrospective cohort studies) M:W pooled analysis showed OR=0.99 [95% CI 0.69 to 1.42] between SIRT and TACE	SMD between SIRT and TACE groups was -0.14 years [95%CI -0.32, 0.03]	Systematic review and meta-analysis of studies that compared SIRT with TACE between 2013 and 15 February 2020	Published, primary research studies reporting survival outcomes in people who had SIRT and TACE for mNETs. Studies had to have at least one distinct subgroup of people with mNETs treated with SIRT and one with TACE, and report mean or median overall survival. Systematic reviews, case reports and abstracts were excluded.	SIRT with ⁹⁰Y compared with TACE	Not reported
2	Frilling, 2019a	n=27 studies (n=842 patients/procedures) M:W not reported	Not reported	Systematic review and meta-analysis Search date: 30 April 2018	Publications in English, German, Dutch, French or Danish; at least 5 human subjects; response data; entire or subpopulation with mNETs who had SIRT.	SIRT with ⁹⁰Y microspheres (glass or resin)	Not reported
3	Egger, 2020 US	N=248 (n=51 had SIRT; n=197 had TACE) 45% men; 55% women	Median=60 years 45% men	Retrospective comparative study of people at 2 US academic medical centres	All adult patients with mNETs that had SIRT or TACE between 2000 and 2018. People received treatment if they had unresectable disease, minimal or no extrahepatic disease, adequate liver function.	SIRT with ⁹⁰Y glass microspheres (Therasphere, BTG)	Median=34 months
4	Wong, 2022 International registry data from 36 institutions.	n=170 56% men, 44% women	Median age=66 (IQR= 56 to 73)	Retrospective analysis of prospective international registry data (RESiN- Radiation Emitting SIR-Spheres in Non-resectable liver tumour registry) between 2015 and 2021	People aged 18 or over who could provide informed consent and were considered appropriate for arterial therapy.	SIRT with ⁹⁰Y resin microspheres (SIR-Spheres, Sirtex Medical).	Average follow-up was not reported. Registry data between 2015 and August 2021 were used.
5	Braat, 2019 8 centres across US, Germany, Belgium, UK, Netherlands	n=244 M:W not reported	Not reported	International, multicentre retrospective case series with data between July 2004 and May 2016	Histologically proven mNET of any origin treated with ⁹⁰Y resin microspheres, at least baseline and 3 (+ or -1.5) months follow-up cross-sectional imaging.	SIRT with ⁹⁰Y resin microspheres (SIR-Spheres, Sirtex Medical)	Range= 51 days to 12 years
6	Braat, Bruijnen, 2020 Netherlands	n=30 (n=34 in the original sample but 3 were excluded after consent because they did not meet inclusion criteria and 1 person died from overproducing insulinoma so did not complete follow up) 73% (22/30) men, 27% (8/30) women.	Mean=62 years (SD=8)	Single centre, single-arm trial (HEPAR PLuS) between October 2014 and September 2018	18 years or older, histologically confirmed grade 1 or 2 mNET, at least 3 unresectable measurable liver metastases, ECOG less than or equal to 2, life expectancy greater than 12 weeks, previous treatment with 4 cycles of PRRT. People with a previous cardiac event within 3 months or increased risk of liver toxicity were excluded.	¹⁶⁶Holmium SIRT within 20 weeks of finishing 4 cycles of PRRT	6 months
7	Frilling, 2019b UK	n=24 people 63% men, 37% women	Median=62 years (range 45 to 75)	Retrospective analysis of a prospective registry of all consecutive people with mNETs that had SIRT at an NHS Trust between January 2007 and December 2017.	People with non-resectable mNETs, grade 1/2 mNETs, disease progression following previous treatments, adequate haematological renal and hepatic function including clotting profile, and good performance status (ECOG less than 2). Decision to deliver SIRT was by MDT and low volume extrahepatic metastases and or primary tumour in situ were not exclusion criteria.	SIRT with ⁹⁰Y labelled resin microspheres (SIR-Spheres, Sirtex Medical, Sydney, Australia)	Median=18 months (range= 2 to 90)
8	Tomozawa, 2018 US	n=93 (n=48 people had unilobar treatment; n=45 had whole liver treatment) 50% men, 50% women	Mean=59 years (SD=14)	Retrospective case series from a single centre between 2007 and 2015	People with mNETs that: i) had CT or MR imaging available and were not surgical candidates; ii) had available details or treatment planning and delivery; iii) were aged 18 to 99. People who were considered unsafe to treat because of extra-arterial flow to the gastrointestinal tract, had hepatopulmonary shunt fraction greater than 20% or had previous external beam radiation therapy to the liver were excluded.	SIRT with ⁹⁰Y resin microspheres (SIR-Spheres; Sirtex Medical Ltd, North Sydney, Australia)	Average follow up was not reported. All patient data (until death or last known follow-up) between 2007 and 2015; 56% of people had greater than 1 year follow-up.
9	Cramer, 2017 US	n=30 43% men, 57% women	Median=60.5 years (range 25 to 76)	Prospective case series from a single centre between August 2009 to July 2013.	People with mNETs that were not surgical candidates, 18 or over, ECOG less than or equal to 2, life expectancy greater than 3 months. Lung shunt fraction greater than 20% were excluded.	SIRT with ⁹⁰Y labelled microspheres (SIR-Spheres, Sirtex, Australia)	24 months

**Table 3 Study outcomes**
First author, date	Efficacy outcomes	Safety outcomes
Ngo, 2021	Tumour response There was no statistically significant difference in tumour response within 3 months between the SIRT and TACE groups (OR=2.9, 95% CI= 0.8 to 10.2; p=0.1; n=2 studies). There was no statistically significant difference in tumour response at follow ups after 3 months between the SIRT and TACE groups (OR=1.0, 95%CI= 0.1 to 7.9; p=0.99; n=3 studies). Symptom response No statistically significant difference in symptom improvement after treatment (OR=0.3, 95% CI= 0.1 to 1.5; p=0.13, n=2 studies): 75% of people who had SIRT reported symptom improvement 52.6% of people who had TACE reported symptom improvement. Overall survival Median OS was statistically significantly better in the TACE group compared to SIRT (OR=1.92; 95%CI= 1.1 to 3.2, p=0.01; n=6 studies): median OS in the SIRT subgroup ranged from 14.5 to 66.8 months median OS in the TACE group ranged from 16.8 to 81.9 months. Hepatic progression free survival There was no statically significant difference in hepatic progression free survival between SIRT and TACE groups (OR= 1.0, 95%CI=0.8 to 1.4, p=0.96; n=5 studies).	Mortality Not reported. Major adverse events (clinical and biochemical) There was no statistically significant difference in major adverse event rates between groups (OR=1.16, 95% CI= 0.5 to 2.5, p=0.71, n=2 studies): 6.9% of people who had TACE reported major adverse events 8.5% of people who had SIRT. Minor adverse events (clinical and biochemical) There were no statistically significant differences in minor adverse event rates between groups (OR=1.08, 95%CI= 0.39–2.99; p= 0.88): 44.6% of people who had TACE reported minor adverse events 58.8% of people who had SIRT. Clinical toxicity events Common minor clinical adverse events included abdominal pain, nausea, vomiting, anorexia, fatigue, fever and flushing. Biochemical toxicity Common minor biochemical adverse events included elevations in aspartate aminotransferase or alanine aminotransferase.
Frilling, 2019a (systematic review & meta-analysis)	Tumour response The weighted mean objective response rate was 51% (95%CI= 47% to 54%; n=27 studies) The weighted mean disease control rate was 88% (95%CI= 85% to 90%; n=27 studies). Overall survival Median overall survival was 32 months (range 18 to 57). 1-year overall survival ranged from 71 to 95% (n= 9 studies). 2-year overall survival ranged from 18.5% to 87.4%. (n=8 studies; excluding the outlying study which reported 18.5%, the 2-year overall survival rate ranged from 45% to 87.4%). 3-year overall survival ranged from 45% to 77.6% (n=5 studies). 5-year overall survival rate ranged from 18.5% to 60% (n=3 studies).	Safety data was not reported in this meta-analysis.
Egger, 2020	Tumour response at 6 months (RECISTv1.1) Disease control rate was statistically significantly higher for people who had TACE compared with people who had SIRT (p=.004) 83% (34/42) of people who had SIRT showed disease control (complete response=4%, partial response=20%, stable disease=59%, progressive disease=17%) 96% (134/139) of people who had TACE showed disease control (complete response=4%, partial response=27%, stable disease=66%, progressive disease=4%). Median percent change in largest liver metastasis size Median percent decrease in size of largest liver metastasis was significantly greater in people who had TACE compared with people who had SIRT (p=0.05) Median decrease in largest liver metastasis was 9% in people who had SIRT (IQR=0, 27%) Median decrease in largest liver metastasis size was 19% (IQR=6 to 34%) Overall survival There was no statistically significant difference in overall survival between people who had SIRT and people who had TACE (p=0.3): Median overall survival was 35.9 months in people who had SIRT Median overall survival was 50.1 months in people who had TACE. This remained non-significant in a sensitivity analysis of only people with well-differentiated grade 1 or 2 tumours. There was no statistically significant difference in 5-year overall survival rates between people who had SIRT and people who had TACE (p=0.27). Progression free survival There was no significant difference in progression free survival (p=0.37) Median progression free survival was 15.9 months for people who had SIRT. Median progression free survival was 19.9 months for people who had TACE. This remained non-significant in a sensitivity analysis of only people with well-differentiated grade 1 or 2 tumours. Length of stay Length of stay was statistically significantly longer for people who had TACE than people who had SIRT (p<.001). median length of stay after SIRT was 0 days median length of stay after TACE was 1 day. 92% of people who had SIRT had treatment in outpatient appointments, compared with 99% of people who had TACE who spent at least 1 night in hospital.	Mortality There was no statistically significant difference in 30-day mortality between groups (p=1.0). The authors did not report whether these were treatment related deaths: 2% (1/51) of people who had SIRT died within 30 days 3.1% (6/198) of people who had TACE died within 30 days. There was no statistically significant difference in 90-day mortality between groups (p=0.2). The authors did not report whether these were treatment related deaths: 9.8% (5/51) of people who had SIRT died within 90 days 5.2% (10/197) of people who had TACE died within 90 days. Major adverse events This study reported 'major complication rates' as any grade 3 or 4 complication within 30 days of SIRT (Clavien-Dindo scale). The wording implies these are clinical events related to treatment, but it is not clear. There was no statistically significant difference in major complication rates between groups (p=0.58): 5.9% (3/51) of people who had SIRT had a grade 3 or 4 complication 9.2% (18/198) of people who had TACE had a grade 3 or 4 complication. Overall adverse events This study reported overall complication within 30 days of SIRT (Clavien-Dindo scale). The wording implies these are clinical events related to treatment, but it is not clear. There was no statistically significant difference in overall complication rates between people who had SIRT and people who had TACE (p=0.17): 13.7% (7/51) of people who had SIRT had a complication 22.6% (44/197) of people who had TACE had a complication. Clinical adverse events The wording of 'complication rates' (reported above) implies these were clinical events, but it is not clear. Biochemical toxicity People who had TACE had: significantly greater increase in post-procedure bilirubin (favours SIRT; +55% vs 0%, p < 0.001) significantly greater increase in post-procedure creatinine (favours SIRT; +12% vs -13%, p < 0.001)
Wong, 2022	Tumour response at 6 months (RECISTv1.1) Data was available for n=99/121 people with imaging data at 6 months. Percentages are as reported in the publication which used 121 as the denominator. 4% (n=5) had complete response 32% (n=39) had partial response 32% (n=39) had stable disease 13% (n=16) had progressive disease. Overall survival Median overall survival was 33 months (95%CI=25 to n/a): 1-year survival was 75% 2-year survival was 62% 3-year survival was 46%. There was no statistically significant difference in overall survival by primary tumour location (p=0.1) or by degree of tumour differentiation (well, moderate, poor; p=0.67). Hazard ratios indicated that overall survival was lower for people with ECOG greater than or equal to 2 (p=0.01) and if the person had ascites at baseline (p=0.049) Progression free survival Median progression free survival was 25 months (95%CI=22 to 35 months). 1-year progression free survival was 70% 2-year progression free survival was 54% 3-year progression free survival was 35%. There was no statistically significant difference in progression free survival by primary tumour location (p=0.3) or by degree of tumour differentiation (well, moderate, poor; p=0.96) Intrahepatic progression was seen in 42% (71/170) of people. This was in a previously treated region in 37% (26/71) of people with progressive disease. Extrahepatic progression was seen in 36% of (61/170) people. Hazard ratios indicated that ECOG greater than or equal to 2 (p=0.01) and baseline ascites (p=0.0001) was associated with shorter progression free survival. People who had unilobar treatment (p=0.03) and with 25% or greater tumour burden (p=0.049) were associated with longer progression free survival.	Mortality 39% (67/170) died within the follow-up period. Cause of death was available in 44 of these people. The most common cause was tumour progression (33/44 records). 1 person died of liver failure 11 months after treatment. This person had multiple prior chemotherapy treatments including cisplatin and irinotecan. Other grade 5 toxicities are listed in Table 4 of the publication. Major adverse events 139 adverse events that were rated grade 3, 4 or 5 (CTCAE v5) across 59 categories were listed in the publication. These were in 34% (58/170) of people 19% of toxicities (20/107 evaluated events) were attributed as definitely or probably related to treatment. Minor adverse events Minor adverse events (rated grade 1 or 2 according to CTCAE v5) were not reported in this study. Clinical adverse events Of the 20 grade 3 to 5 events that were classed as definitely or probably related to treatment, clinical events included: death, abdominal pain (grade 3 and 4), hepatic abscess (grade 3) and lymphopenia (grade 3). The most common grade 3 to 5 adverse events were abdominal pain (7/139) and anorexia (5/139) Of 15 patients with colonized bile ducts, one patient (7%) developed a hepatic abscess. Biochemical toxicity Of the 20 grade 3 to 5 events that were classed as definitely or probably related to treatment, liver function toxicities in the absence of progressive hepatic disease included: bilirubin (n= 5, 3%) new ascites (n= 8, 5%) alanine aminotransferase (n= 1, 0.6%) alkaline phosphatase (n= 1, 0.6%). of the 5 patients with Grade 3 hyperbilirubinemia, 2 normalized within 3 months, leaving 13 (7.6%) with durable hepatic function toxicities.
Braat, Kappadath, 2019	Disease control rate was 91% at all time points on all outcome measures not correlated with tumour grade. Tumour response at 3 months (RECIST v1.1) 244 people were assessed according to RECIST v1.1 criteria at a mean of 68 days (SD=34) after SIRT: 2% (4/244) had complete response 14% (34/244) had partial response 76% (185/244) had stable disease 9%(21/244) had progressive disease. Tumour response at 3 months (mRECIST) 126 people were assessed according to mRECIST at a mean of 89 days (SD=78) after SIRT: 8% (10/126) had complete response 35% (44/126) had partial response 48% (61/126) had stable disease 9% (11/126) had progressive disease Tumour response at 6 months (RECIST v1.1) 116 people were assessed according to RECIST v1.1 at a mean of 187 days (SD=48) after SIRT: 1% (1/116) had complete response 28% (32/116) had partial response 63% (73/116) had stable disease 9% (10/116) had progressive disease Tumour response at 6 months (mRECIST) 70 people were assessed according to mRECIST at a mean of 189 days (SD=38) after SIRT: 9% (6/70) had complete response 54% (38/70) had partial response 29% (38/70) had stable disease 9% (6/70) had progressive disease Symptom response Clinical symptom response was observed in 79% of people who reported symptoms at baseline (60% of 244 people): 44% reported improvement in symptoms 35% reported complete symptom resolution. Overall survival Median overall survival was 2.6 years (95%CI=2.2 to 3 years).	Mortality Treatment related mortality was not reported in this study. Major adverse events Only biochemical toxicity was described in terms of severity of event (CTCAE v4.03). The authors reported grade 3 to 4 biochemical and haematological toxicities were limited; most common was lymphocytopenia in 6.7% of people. Minor adverse events Only biochemical toxicity was described in terms of severity of event (CTCAE v4.03). These occurred in up to 52% of people. Details are described under biochemical toxicity. Clinical adverse events Note that the authors reported that clinical toxicities were not recorded for 12% and 39% of people at 3 and 6 months, respectively (missing data). At 3 months, 56% of people experienced known radioembolisation-related adverse events (fatigue, abdominal pain, nausea). At 6 months, 6% of people had persistent adverse events (mainly abdominal pain) There were no clinical toxicities in 32% of people at 3 months, and no clinical toxicities in 55% of people at 6 months. 3% (7/244) people had radiation-induced gastric ulcer 0.8% (2/244) of people had radiation-induced liver disease 1 person (of 244, 0.4%) had radiation pneumonitis 1 person (of 244, 0.4%) had liver abscess with bilioenteric anastomosis (without antibiotic prophylaxis) 1 person (of 244, 0.4%) had cholangitis with bilioenteric anastomosis (without antibiotic prophylaxis) Biochemical toxicity Grade 1 to 2 bilirubin elevation and/or decreased albumin levels occurred in 6%. Grade 1 to 2 lymphocytopenia in 52%, Thrombocytopenia occurred in 17%, Grade 1 to 2 anaemia or leukopenia occurred in 8%, Coagulation was unaffected as measured by the international normalized ratio. Other adverse events 0.8% (2/244) people had arterial dissection from the angiography procedure Treatment was stopped for 3 people because of complaints.
Braat, Bruijnen, 2020	Note that before SIRT in this study, all participants had PRRT. After the last of 4 cycles of PRRT, 17% (5/30) patients had progressive intrahepatic disease, 56% (17/30) had stable disease, and 27% (8/30) had partial response, according to RECIST v1.1. Tumour response at 3 months (RECIST v1.1) 0% (0/30) had complete response 43% (13/30) partial response 50% (15/30) stable disease 7% (2/30) progressive disease Tumour response at 3 months (mRECIST) 10% (3/30) had complete response 50% (15/30) partial response 27% (8/30) stable disease 0% (0/30) progressive disease 13% (4/30) could not be evaluated because they had hypovascular disease that precluded mRECIST evaluation. Tumour response at 6 months (RECIST v1.1) 0% (0/30) had complete response 47% (14/30) partial response 37% (11/30) stable disease 13% (4/30) progressive disease 3% (1/30) could not be evaluated because they died before follow-up. Tumour response at 6 months (mRECIST) 7% (2/30) complete response 50% (15/30) partial response 23% (7/30) had stable disease 3% (1/30) had progressive disease 17% (5/30) could not be evaluated because 1 person died before follow-up and the rest had hypovascular disease which precluded mRECIST measurement. Tumour-related hormonal symptom resolution 22% (2/9) people with symptoms after PRRT reported complete resolution of their symptoms 56% (5/9) reported an improvement 22% (2/9) reported stable symptoms.	Note that before SIRT in this study, all participants had PRRT. Mortality 1 person died from radioembolisation induced liver disease. Symptoms started 6 weeks after SIRT and the person died after 4 months. 1 person died because of hypoglycaemia caused by an overproducing insulinoma, considered unrelated to treatment. Major adverse events Treatment related clinical toxicity was assessed within 6 months and biochemical toxicity at 3 and 6 months (CTCAE v4.03) Major clinical toxicity considered related to SIRT was observed in up to 10% (3/31 had grade 3 abdominal pain; see clinical toxicity); 1 grade 5 related event was observed (see mortality). Major biochemical toxicity was observed in up to 47% at 3 months, and up to 50% at 6 months (grade 3 γ-glutamyltransferase increase); 3 grade 4 events happened at 3 months and 1 grade 4 event happened at 6 months. Minor adverse events Treatment related clinical toxicity was assessed within 6 months and biochemical toxicity at 3 and 6 months (CTCAE v4.03) Minor clinical toxicity considered related to SIRT was observed in up to 68% (21/31 had grade 1 to 2 abdominal pain; see clinical toxicity). Minor biochemical toxicity was observed in up to 80% at 3 months (grade 1 to 2 alkaline phosphatase increase), and up to 83% at 6 months (grade 1 to 2 aspartate aminotransferase increase). Clinical toxicity There were no grade 4 clinical toxicities related to treatment. Grade 1 to 2 treatment-related clinical toxicity events included: abdominal pain (68%, 21/31) fatigue (58%, 18/31) nausea (61%, 19/31) vomiting (42%, 13/31) malaise (25%, 8/31) subfebrile (13%, 4/31) shivering (10%, 3/31) oedema (6%, 2/31). Grade 3 treatment-related clinical toxicities included: abdominal pain (10%, 3/31) fatigue (3%, 1/31) nausea (3%, 1/31). Biochemical toxicity Biochemical toxicities included: Increases in: bilirubin, alkaline phosphatase, γ-glutamyltransferase, aspartate aminotransferase, alanine aminotransferase, lactate dehydrogenase, ammonia Decreases in albumin Anaemia, leucopoenia, neutropoenia, lymphocytopenia, thrombocytopaenia Other adverse events 1 person had a bleeding gastric ulcer and perforated cholecystitis. This was considered unrelated to treatment because of no evidence of microspheres on the histopathology.
Frilling, 2019b (retrospective case series of prospectively collected data)	Tumour response at 3 months (RECIST v1.1) 4% (1/24) had complete response 54% (13/24) had partial response 29% (7/24) had stable disease 8% (2/24) had mixed response 4% (1/24) had progressive disease. Tumour response at 6 months (mRECIST) 4% (1/24) had complete response 54% (13/24) had partial response 29% (8/24) had stable disease 8% (2/24) had mixed response 0% (0/24) had progressive disease. Symptom response at 6 months 90% (9/10) people who were symptomatic at baseline and completed a quality of life questionnaire reported being "not at all" symptomatic at 6 months. 10% (1/10) reported being "a little symptomatic" at 6 months. Overall survival Median overall survival was 57 months (95%CI=41 months to n/a). Overall survival rates were: 95% at 12 months 87% at 24 months 78% at 36 months. Progression free survival Median progression free survival was 41 months (95%CI=31 months to n/a). At a median of 18 months, 1 person had intrahepatic progression and 7 people had extrahepatic disease progression, 1 person had both intra- and extra-hepatic progression. Progression free survival rates were: 66% after 12 months 50% after 24 months 50% after 36 months.	Mortality There were no treatment related deaths. Major adverse events There were no grade 3 or 4 biochemical toxicities. Clinical adverse events were not graded. Minor adverse events 54% (13/24 people) had grade 1 to 2 biochemical toxicities. Clinical adverse events were not graded. Clinical adverse events 25% (6/24) of people had radioembolisation-related clinical side-effects (4/6 were abdominal pain; 2/6 were fatigue). No other radioembolisation-specific side-effects such as gastric ulcer, liver abscess, radioembolisation-induced liver disease or radiation pneumonitis were recorded during follow-up. One person was re-hospitalised for 24 hours for pain management. They had previous extensive abdominal surgery including right hemicolectomy and left hepatectomy. No peri-interventional carcinoid crisis or any intervention-related complications. Biochemical toxicity There were no grade 3 or 4 biochemical toxicities. 54% (13/24 people) had grade 1 to 2 haematological toxicity, including 12 cases of lymphocytopenia and 1 case of thrombocytopenia. No toxicity (affecting alkaline phosphatase, gamma-glutamyl transferase, aspartate/alanine transaminase or lactate dehydrogenase levels) were seen in 14 people. No changes in bilirubin levels or coagulation profile assessed by the international normalisation ratio occurred.
Tomozawa, 2018	Tumour response at 1 year (RECIST) At 1 year after SIRT: 25% (13/52) had partial response 67% (35/52) had stable disease 8% (4/52) had progressive disease During this time 17% (9/52) of people had systemic chemotherapy and 1/52 had TAE. Overall survival Median overall survival was 28 months (95%CI=16.2 to 40).	Mortality During the study period, 45% (42/93 people) died. The authors did not report whether deaths were considered treatment-related. Major adverse events At 6 months, 1 grade 3 event was seen (ascites). At 1 year, 10 grade 3 adverse events were seen (6/10 events were ascites). No grade 4 events were reported. Minor adverse events At 6 months and 1 year, the most common grade 2 event was elevated alkaline phosphatase (5 and 13 events reported at each follow-up, respectively). No grade 1 events were reported. Clinical toxicity At 1 year, 29% (15/52 people) developed imaging signs of cirrhosis-like morphology or portal hypertension after 12% (6/52) of people had ascites 13% (7/52) had cirrhosis-like morphology (defined as hepatic surface nodularity and enlargement of caudate lobe) 17% (9/52) had splenomegaly 4% (2/52) had varices. Of these 24 events, 19 were in people who had bilobar treatment. During this time 17% (9/52) of people had systemic chemotherapy and 1/52 had TAE. Biochemical toxicity People who had unilobar therapy only showed increased alkaline phosphatase levels. People with bilobar therapy had the following biochemical toxicities: Grade 2 biochemical toxicities included increases in: albumin, alkaline phosphatase, alanine aminotransferase, aspartate aminotransferase, bilirubin, haemoglobin, leukocytes, ascites. These were also seen at 1 year, as well as decrease in platelet count. Grade 3 events at 6 months included 1 case of ascites. At 1 year, changes to ascites, bilirubin, albumin and alkaline phosphatase were seen.
Cramer, 2017	Overall survival Mean overall survival was 32.9 months. At study completion, 37% (11/30) had died. Quality of Life (SF-36 v1) There was no statistically significant difference in any of the 8 health related quality of life domains at 1 and 3 months, compared to baseline. At 6 months, mean mental health domain score was significantly better than baseline (baseline score=74 compared with 6-month score=81, p=.007, scale=0 to 100). No other domains had a statistically significant difference at 6 months. At 12 months, the mean social functioning domain score was significantly better than baseline (baseline score=64 compared with 12-month score=89, p=.019, scale=0 to 100). There were no other statistically significant differences at 12 months. At 24 months, there were no statistically significant differences in any domain.	Adverse events were not reported in this study.

Procedure technique

Type of radioactivity and microsphere

Of 9 studies, 5 reported evidence from ⁹⁰Y labelled resin microspheres (Braat, 2019; Cramer, 2017; Frilling, 2019b; Tomozawa, 2018; Wong, 2022). The retrospective study of 248 people comparing SIRT with TACE (Egger, 2020) used ⁹⁰Y labelled glass microspheres. Both meta-analyses considered evidence from both glass and resin ⁹⁰Y microspheres and made no distinction between these in the results (Frilling, 2019a; Ngo, 2021). The HEPAR PLuS single-arm trial of 30 people was the only study that included people who had SIRT with ¹⁶⁶Ho labelled microspheres (Braat, 2020).

Treatment work-up and parameters
Methods of work-up, dosing and the administered activity were not uniformly reported across studies. When described, work-up was done mostly using ^99mTc‑labelled macroaggregated albumin to assess arterial anatomy and determine hepatopulmonary shunting. The body surface area method was most commonly used to calculate dose. Some studies reported using a dose reduction method when lung shunt fraction was high. When bilobar treatment was needed, there was a mixture of approaches within and between studies as to whether SIRT was delivered sequentially or in the same session. When the number of SIRT cycles was reported, this also varied within and between studies. The retrospective case series of 244 people (Braat, 2019) was the only study to explicitly extract outcomes after the first cycle of SIRT. The available details of work-up and procedure from each study are reported below.

The systematic review and meta-analysis comparing SIRT with TACE (Ngo, 2021) and the systematic review and meta-analysis of 27 studies (Frilling, 2019a) did not describe details of the procedure other than the type of microspheres used.

The retrospective study of 248 people comparing SIRT with TACE (Egger, 2020) reported that if lung shunt fraction was found to be less than 15% during the work-up, SIRT was done. In total, 58% of the SIRT group had bilobar treatment in which SIRT was delivered to the first lobe then to the second lobe, 4 weeks later. Methods used to calculate dose, treatment schedule or actual activity administered were not reported.

Work-up method was not described in the RESiN registry study (Wong, 2022). In this study, 48% (82/170) had bilobar treatment. Median activity delivered was 1.3 GBq (IQR= 0.9 to 1.5 GBq) and 1.9 GBq (IQR= 1.7 to 2.2 GBq) for unilobar and bilobar treatments, respectively. Dose was calculated according to body surface area method in 91% of the sample. Most people had a single cycle of SIRT; 98% had a single cycle and 4 people had 3 or 4 cycles.

In the retrospective case series of 244 people (Braat, 2019), body surface area method was used to calculate dose in 84% of the sample. Prophylactic intravenous octreotide infusion or prophylactic antibiotic treatment was given at the discretion of the treating physician and according to the institutes' guideline. Lung shunt fraction was calculated during work-up and dose was adjusted accordingly. Median lung shunt fraction was 5.6% (range 0.7% to 33%) and median administered activity was 1.8GBq. Activity reduction was applied in 7.4% of the sample and coil embolisation was reported in 50% of the sample. Whole liver treatment was administered in a single session in 56% of the sample, while 14% had whole liver sequentially; 26% had a single right lobar treatment and 3% had single left lobar. In this study, all results were extracted after the first cycle of SIRT.

In the single-arm trial of 30 people (Braat, 2020), they used a 'scout dose' of ¹⁶⁶Ho instead of Tc-labelled macroaggregated albumin to do the work-up. SIRT was either delivered in 1 session or split into 2 based on this. In total, 57% had whole liver single session and 23% had whole-liver sequential sessions. The remaining had single lobe treatment. Activity prescribed was calculated in relation to liver weight. Mean ¹⁶⁶Ho delivered was 6769MBq (SD=2419). Number of SIRT cycles was not reported.

In the retrospective analysis of prospectively collected UK data from 24 people (Frilling, 2019b), dose was modified according to pulmonary shunting detected in the work-up. People with shunt fraction of greater than 20% did not have treatment. Coil embolisation was used to seal non-target vessels in 11 people. Dose was calculated from body surface area, tumour volume and dose reduction was implemented based on lung shunt fraction. Peri-procedural octreotide infusion was given to minimise risk of carcinoid crisis. After SIRT, treatment with corticosteroids, ursodeoxycholic acid and omeprazole was done for 4 weeks to reduce risk of post-radioembolisation syndrome. Median net activity delivered was 1.4 GBq (range 0.5–2.4). Dose reduction of the activity prescribed was needed in 6 patients and number of sessions varies: whole liver single session (n=4), whole liver sequential (n=10), right lobe single session (n=10), and segmental (n=2). The number of cycles was not reported.

In the retrospective case series of 93 people (Tomozawa, 2018), body surface area was used to calculate dose. 52% of people had unilobar treatment and the rest had sequential bilobar treatment. Most people had 2 treatments (46%), followed by 40% who had 1 treatment and the remainder had 3 or 4. Total administered activity was 1.92GBq (SD=0.91).

In the prospective case series of 30 people (Cramer, 2017), they referenced general details of SIRT but these were not specific to the sample in their study. People with bilobar disease had sequential treatment with 1 month in between sessions. If, at 4 weeks, residual or recurrent tumour was observed, or if there were residual symptoms related to tumour burden. No other details were reported.

Efficacy

Tumour response

Eight of 9 studies reported tumour response outcomes. Most commonly, studies evaluated response according to RECIST v1.1 but 3 studies reported outcomes on both RECISTv1.1 and mRECIST. One study reported median change in largest liver metastasis size in addition to RECIST data.

Disease control rates (the sum of complete and partial response and stable disease rates) were reported in 7 of the 8 studies. This was consistently high across studies and response criteria. Disease control rates ranged from 83% at 6 months according to RECISTv1.1 in the retrospective comparative study (Egger, 2020) to 100% at 6 months according to mRECIST at 6 months in the UK-based study of 24 people (Frilling, 2019b; sum of all response categories except progressive disease). The retrospective case series of 93 people reported 92% disease control rate at 1 year (Tomozawa, 2018; sum of all response categories except progressive disease).

Complete response rates were low in all studies that reported this independently. The lowest rate was 0% at 3 months according to RECISTv1.1, reported in the HEPAR PLuS single-arm trial (Braat, 2020). The highest complete response rate was 10%, reported in the same study on the mRECIST criteria at 3 months (Braat, 2020).

Comparative evidence between TACE and SIRT was reported in 2 studies and showed mixed findings (Egger, 2020; Ngo, 2021). The disease control rate at 6 months was statistically significantly higher for people who had TACE than people who had SIRT in the retrospective comparative study of 248 people (Egger, 2020). This study also reported that decrease in largest liver metastasis was significantly greater in people who had TACE compared to SIRT, but this only just reached significance (p=0.05). But, when combined with the findings of two smaller studies with data beyond 3 months in the meta-analysis (Ngo, 2021), there was no statistically significant difference (n=3 studies). In this meta-analysis, there was also no statistically significant difference in disease control rate between SIRT and TACE within 3 months of treatment, but this trended toward favouring TACE (n=2 studies).

Symptom response

Symptom response was reported in 4 of 9 studies (Braat 2019; Braat 2020; Frilling, 2019b; Ngo 2021). Primary research studies reporting this data summarised rates in a subgroup of the sample that had symptoms at baseline. Symptom response was measured in different ways across studies but between 79% and 100% of people reported a response or stabilisation of symptoms; 22% (2/9) of people reported stable symptoms in the HEPAR PLuS single-arm trial (Braat, 2020) and 1/10 people in the UK-based study of 24 people reported being a little symptomatic (Frilling, 2019b).

Overall survival

Overall survival was reported in all 9 studies. The range of overall survival was broad within studies but average overall survival was quite consistent between studies. Median overall survival in people who had SIRT ranged from 14.5 months to 66.8 months in the meta-analysis comparing SIRT with TACE (Ngo, 2021). A similarly large range was reported in the meta-analysis of 27 studies, in which median overall survival ranged from 18 to 57 months (Frilling, 2019a). However, aggregated median overall survival among the 27 included studies was 32 months and this was similar to average estimates reported in most other studies included in the main evidence. Notably longer median overall survival of 57 months was reported in the single arm trial of 30 people (Braat, 2020).

Comparative evidence from the meta-analysis suggested that people who have TACE have statistically significant longer overall survival than people who have SIRT (Ngo, 2021). But the largest study in this meta-analysis was also included in the main evidence and this showed no statistically significant difference when SIRT was compared with TACE (Egger, 2020).

Progression free survival

Four studies reported progression-free survival (Egger, 2020; Frilling, 2019b; Ngo, 2021; Wong, 2022). The meta-analysis comparing SIRT with TACE found no significant difference in hepatic progression-free survival (Ngo, 2021) and this was consistent with the findings of the retrospective study of 248 people comparing SIRT with TACE (Egger, 2020). The RESiN registry study found that median progression free survival was 25 months and ranged from 22 to 35 months (Wong, 2022). This contrasted with the UK-based analysis of 24 people that found that median progression-free survival was 41 months (Frilling, 2019b). However, 1- and 2-year progression-free survival rates were similar between studies (66% and 70% at 1 year and 50% and 54% at 2 years in Frilling, 2019b and Wong, 2022, respectively).

Quality of life

The most complete report of quality of life data was reported in the prospective case series of 30 people (Cramer, 2017). Of the 8 domains measured at 1, 3, 6, 12 and 24 months, the only domains and timepoints to show a significant, positive difference to baseline were mental health at 6 months, social functioning domain at 12 months.

A NET-specific health related quality of life measure (EORTC QLQ-GINET 21 questionnaire) was used in the UK based analysis of 24 people (Frilling, 2019b) but this data was not presented in full. Only symptom response was reported and this has been reported in the summary above.

Quality of life was also measured in the single arm trial of 30 people (Braat, 2020) but this was only presented in a figure. Quantitative data was not reported in a way that could be extracted.

Length of stay

Length of stay was reported in the retrospective study of 248 people comparing SIRT with TACE (Egger, 2020). Length of stay was statistically significantly longer for people who had TACE than people who had SIRT; 92% of people who had SIRT had treatment in outpatient appointments, compared with 99% of people who had TACE, who spent at least 1 night in hospital.

Safety

Seven of 9 studies reported safety data. There were many types of adverse event, as exemplified by the list of 59 categories of CTCAE grade 3 or higher adverse events in the RESiN registry study (Wong, 2022). These are summarised here but refer to table 3 for more detail.

Mortality

Four studies reported mortality data with cause of death. The UK-based analysis of 24 people (Frilling, 2019b) reported no treatment-related deaths. The single-arm trial of 30 people reported 1 death caused by radiation-induced liver disease (Braat, 2020). The RESiN registry study reported that 1 person died of liver failure 11 months after treatment but said that the person had multiple previous chemotherapy treatments and did not state whether death was considered caused by SIRT (Wong, 2022). Other grade 5 events are reported in this publication. There was no statistically significant difference between SIRT and TACE on 30- and 90-day mortality in the study of Egger (2020), but the authors did not report whether these were treatment related deaths.

Major adverse event rates

Major adverse events are here classified as those which were reported as 'major' or which were classified as grade 3 or higher using a grading system. Event rates varied between studies and is likely a reflection of whether biochemical and clinical toxicities were aggregated or not. Major adverse events were most commonly grade 3 across all studies. The highest major adverse event rates were presented in the single arm trial of 30 people (Braat, 2020) in which major biochemical toxicities were seen in up to 50% of people at 6 months. Clinical toxicity rates were lower with up to 10% of people having grade 3 abdominal pain. Contrastingly, major biochemical adverse event rates were much lower in the retrospective case series of 244 people (Braat, 2019), with 77% of people having grade 3 to 4 events. Combined toxicity rates were also lower than the single arm trial in the RESiN registry study (Wong, 2022), 19% of all toxicities were attributed as definitely or probably associated with SIRT. Major adverse event rates were lower still across the rest of the studies that reported this information: in the retrospective comparative study of SIRT and TACE (Egger, 2020), major adverse events were reported in 66% of people. This study is not clear but implies these were clinical events only. In the UK study of 24 people, there were no grade 3 or 4 biochemical toxicities (Frilling, 2019b). Clinical events were not graded in this study.

Minor adverse event rates

Minor adverse event rates were higher across all studies. Between 52% (Braat, 2019) and 83% (Braat, 2020) of people experienced a grade 1 or 2 minor adverse events. When clinical and biochemical events were graded separately, clinical toxicity was lower.

Clinical adverse events

Most studies that reported rates of individual clinical events showed that abdominal pain was the most common. Other commonly reported clinical adverse events were fatigue, nausea, vomiting, anorexia, fever and flushing. The retrospective case series of 244 people (Braat, 2019) measured these at 3 and 6 months; by 6 months, only 6% of people had persistent adverse events (mainly abdominal pain) and 55% had no clinical toxicities. Less common adverse events included: hepatic abscess in a person with colonised bile duct, persistent abdominal pain at 6 months, radiation induced gastric ulcer, radiation-induced liver disease, radiation pneumonitis, cholangitis with bilioenteric anastomosis, malaise, subfebrile, shivering and oedema. Longer-term (1 year) adverse events are described below.

Biochemical toxicity

The following biochemical toxicities were reported: bilirubin, creatine, alanine aminotransferase, alkaline phosphatase, neutropoenia, lymphocytopenia, thrombocytopaenia, anaemia, leucopoenia, γ-glutamyltransferase, aspartate aminotransferase, alanine aminotransferase, lactate dehydrogenase, ammonia, albumin. Comparative evidence of unilobar compared with bilobar treatment biochemical adverse event rates showed that people who people who had unilobar therapy only showed increased alkaline phosphatase levels (Tomozawa, 2018). Multiple grade 2 and 3 biochemical toxicities were seen in people who had bilobar treatment.

Clinical toxicity at 1 year

One study explicitly measured clinical toxicity at 1 year (Tomozawa, 2018). Of 52 people, 29% (n=15) had cirrhosis like-morphology or portal hypertension at 1 year. This was more common in people who had bilobar treatment.

Comparison of other adverse event rates between SIRT and TACE

The meta-analysis comparing SIRT with TACE (Ngo, 2021) found that there were no statistically significant differences in the rates of major or minor adverse events between groups. In the retrospective comparative study of 248 people (Egger, 2020), the authors noted that whilst post-procedure biochemical toxicity favoured SIRT over TACE, there were no differences in 30-day overall or major complication rates between groups.

Anecdotal and theoretical adverse events

Expert advice was sought from consultants who have been nominated or ratified by their professional society or royal college. They were asked if they knew of any other adverse events for this procedure that they had heard about (anecdotal), which were not reported in the literature. They were also asked if they thought there were other adverse events that might possibly occur, even if they had never happened (theoretical).

They listed the following anecdotal adverse events:

carcinoid crisis caused by SIRT
gastritis
perforation of the bowl
ischaemic effects.

Nine professional expert questionnaires for this procedure were submitted. Find full details of what the professional experts said about the procedure in the specialist advice questionnaires for this procedure.

Validity and generalisability

Studies included in this overview reported consistently high rates of disease and symptom control. Complete response rates were low. Different RECIST criteria were used within and between studies to measure tumour response. These had different findings on granular outcome classifications such as complete and partial response and stable disease, but they tended to find the same proportions of people who were classed in the disease control and progressive disease groups.
Average overall survival was relatively consistent across studies, but overall survival varied within wide confidence intervals within studies.
There is some inconsistency in the evidence that compares tumour response outcomes between people who had SIRT and TACE.
There were many types of adverse event that ranged in severity from mild to death. Event rates varied between studies but biochemical toxicities were generally more common than clinical adverse event rates. There were reports of cirrhosis-like morphology and portal hypertension in the longer-term (evidence at 1 year; Tomozawa, 2018). Comparative evidence between people who had SIRT and TACE suggests that minor and major event rates are comparable between these 2 procedures.
The evidence includes some large, international studies with a range of study designs, including some comparative evidence. One study was done exclusively in the UK (Frilling, 2019b). A few studies had reasonable follow up but the extent to which long-term data contributed to safety data was unclear in some studies. Most studies reported tumour and symptom response at 3 and 6 months. One study (Tomozawa, 2018) explicitly examined 1-year safety data. The only randomised evidence found in this search was a pilot study. This was small (n=11) and had high risk of bias, so was not included in the main evidence.
Dose and planning methods varied between studies. Some studies used dose reduction methods if there was evidence of lung shunting over a threshold. Other studies split treatment sessions if bilobar treatment was needed. Professional expert questionnaires reflected this finding with advice that dose and planning methods are evolving. Professional expert questionnaires advise that because SIRT is embolic and a form of brachytherapy, precise dose planning and vascular radiology and nuclear medicine expertise is needed. They reported that toxicity may be related to experience of the team.
The type of microsphere (glass or resin) and radioisotope (⁹⁰Y or ¹⁶⁶Ho) varied between studies. Most frequently, studies used microspheres labelled with ⁹⁰Y. Only 1 study used ¹⁶⁶Ho.
Treatment histories and disease profiles were often complex and heterogeneous within and between studies. Studies generally had similar inclusion criteria and samples of a similar age. Professional expert questionnaires advised that patient selection is complex, and people with mNETs are a relatively rare and heterogeneous disease group. Treatment history is likely to be heterogeneous in practice. Some studies reported prognostic factors (not reported in this overview) that associate with better response to SIRT but there were many and authors commented that they are not yet well-recognised.
It was unclear whether people had further treatment after SIRT in most studies. Three studies explicitly reported that people who had SIRT went on to have other interventions after SIRT (Braat, 2019; Frilling, 2019b; Wong, 2022). No study explained whether or how this was accounted for in follow-up data and findings.
The systematic review and meta-analysis by Ngo (2021) was reasonable in quality and examined an important comparison between SIRT and another intra-arterial procedure (TACE). The meta-analysis only included 6 retrospective studies, and fewer studies contributed to the pooled analysis of tumour and symptom response data. The meta-analysis showed that other than overall survival, the procedures were comparable in tumour and symptom response and safety outcomes. The authors did not report whether differences in the length of follow up between studies included in pooled outcomes may have affected the findings. The more recent and larger studies included in the meta-analysis (which included the retrospective comparative analysis of 248 people by Egger, 2020 that was also included in the main evidence) showed similar efficacy between groups and a trend toward less minor adverse events in SIRT. The authors acknowledge that laboratory and clinical adverse events were not analysed separately. The authors concluded that comparative and better-quality evidence was needed to confirm the findings that TACE was better than SIRT for overall survival. There were differences in some baseline characteristics. The authors note that overall survival did not relate to disease control rate. They acknowledge that subsequent treatment could have affected survival outcomes in this analysis. They ask for research to clarify whether subgroups may benefit from most from SIRT and TACE. No conflicts of interest were reported for this study.
The meta-analysis of 27 studies (Frilling, 2019a) was of poor quality but included a lot of evidence and findings seem to be consistent with earlier meta-analyses that included fewer and less recent studies (for example, Devcic, 2014, in table 5). Outcomes were summarised together regardless of follow-up point. The authors did not report how this might have affected the findings. They report that the findings support the clinical effectiveness and safety of SIRT. The authors acknowledge that the optimal timepoint to measure response to SIRT is unclear. Overall, the authors reported that SIRT has good potential for people with carcinoid syndrome and acknowledge that this will probably be used in combination and sequence with other therapies. They state more research is needed to identify which patients are most likely to benefit from this treatment. Some conflicts of interest with one of the companies that manufacture SIRT device were declared. This was the same publication that reported the retrospective case series of 24 people in the UK.
In the retrospective study of 248 people that compared SIRT with TACE (Egger, 2020), there were some differences in baseline characteristics between groups. The authors state that both SIRT and TACE showed safe and effective control of unresectable mNETs. They highlight that the hospital stay is longer for TACE but that the short-term radiographic response was not as good. They report that the lower disease control rate in the SIRT group compared with the TACE and with outcomes reported in other studies of SIRT might be because more than 10% of the group had poorly differentiated tumours. The authors highlight the heterogeneity of the disease group and state that further, prospective and comparative research is needed to understand the patients most likely to benefit. This study was the largest study included in the Ngo (2021) meta-analysis. No conflicts of interest were declared.
In the RESiN registry study (Wong, 2022), the authors report that, given their findings reflect previous research, SIRT is effective for mNETs and has an acceptable toxicity profile. They acknowledge that dosimetry methods that became outdated in 2021 were mostly used. The authors acknowledge that there is no randomised evidence comparing SIRT with other intra-arterial therapies. Multiple authors reported conflicts of interest with one of the companies that manufacture a SIRT device.
In the retrospective study of 244 people (Braat, 2019), the authors reported that SIRT was shown to be safe and effective for the treatment of mNETs, although they also note high rates of missing data for clinical toxicity at 6 months (39%). The authors highlight the utility of SIRT in relieving symptoms of carcinoid syndrome and that positive outcomes for tumour response might be seen after 3 months. The authors state that randomised evidence is needed and more understanding about which people with mNETs might benefit most, and where SIRT sits in the sequencing of other treatment options, is needed. Several authors had received funding or consulted for one of the companies that manufacture a SIRT device. Travel and accommodation expenses were partially covered by this company.
In the single-arm trial of 30 people (Braat, 2020), the centre that did the trial owns royalties for the ¹⁶⁶Ho microspheres. This was the only study that reported data on this radionuclide. This study reported notably higher average overall survival than other studies. Everyone in the trial also previously had systemic therapy (PRRT). The authors report that the combination of SIRT with PRRT was safe and effective in their study, but that randomised evidence is needed. No quantitative quality of life data could be extracted from this study for the overview but the authors report that mostly small, temporary decreases in quality of life were observed following treatment.
In the retrospective case series of 93 people (Tomozawa, 2018), the authors acknowledge that therapies had before or after SIRT may have affected the long-term safety findings that they report. They also note that cause of death was not available for some people. They report that SIRT is a safe and promising treatment option for unresectable mNETs, although bilobar treatment did result in more frequent evidence of toxicity. The authors acknowledge that prospective evidence is challenging to collect but is needed. Some conflicts of interest including grants from companies that manufacture SIRT devices were reported.
In the prospective case series of 30 people (Cramer, 2017), no conflicts of interest were declared. In contrast with studies that presented patient reported symptom response, this study reported quality of life data using standardised measure of health-related quality of life. However, there was a high drop-out rate (50% of the sample did not complete a survey 1 month-post treatment and 17% completed the last follow-up at 24 months).

Ongoing trials:

ArTisaN trial protocol: a single Centre, open-label, phase II trial of the safety and efficacy of TheraSphere selective internal radiation therapy (SIRT) in the treatment of inoperable metastatic (liver) neuroendocrine neoplasia (NENs). DOI: 10.1186/s12885-022-09859-9. NCT04362436; Single arm trial; n=24. Due to complete September 2024.

How are you taking part in this consultation?

Interventional procedure overview of selective internal radiation therapy for neuroendocrine tumours that have metastasised to the liver