Peptide Receptor Radionuclide Therapy Improves Survival in Patients Who Progress After Resection of Gastroenteropancreatic Neuroendocrine Tumors

Abstract

Background.

Peptide receptor radionuclide therapy (PRRT) is an effective treatment for advanced gastroenteropancreatic (GEP) neuroendocrine tumors (NETs). We investigated a 2-decade experience with PRRT to determine whether PRRT confers a survival advantage to patients who progress after surgery versus other therapies.

Methods.

We identified patients from our clinic who had resection/cytoreduction of GEP-NETs, then disease progression by Response Evaluation Criteria in Solid Tumors (RECIST) 1.1. The Kaplan–Meier method assessed progression-free survival (PFS) and overall survival (OS), calculated from progression after surgery (no-PRRT group) or the start of PRRT. Cox regression with time-dependent covariates controlled for immortal time bias and other confounders.

Results.

Overall, 237 patients progressed after surgery; 95 received PRRT and 142 did not. No differences existed in sex, T or N stage, tumor grade/differentiation, primary site, or time to progression; 94% of PRRT patients had metastases at diagnosis versus 77% in the no-PRRT group. Median PFS was longer in the PRRT group versus the no-PRRT group (32.4 vs. 11.0 months, p < 0.001), as was median OS (49.8 vs. 38.4 months; p = 0.009). In subgroup analysis, the PRRT group had improved PFS in small bowel NETs and pancreatic NETs. Time-dependent covariate analysis revealed a lower risk of death associated with PRRT (hazard ratio 0.61, p = 0.028) after adjusting for sex, age, M stage, tumor grade, and primary site.

Conclusion.

Surgical resection and cytoreduction is an effective treatment for patients with GEP-NETs, but most patients with metastatic disease develop recurrent disease. Surgery followed by PRRT after progression conferred superior PFS and OS over no PRRT/other therapies, and is an effective strategy for managing patients with GEP-NETs.

Neuroendocrine tumors (NETs) are the most common malignancies of the small bowel¹ and account for approximately 7% of all pancreatic neoplasms.² Survival of patients with NETs is often relatively long, even when there is distant disease (median overall survival [OS] of 70 months for metastatic small bowel NET [SBNET] patients and 24 months for pancreatic NETs [PNETs]).^3,4 This can often be improved by surgical resection of primary tumors and regional nodes, and cytoreduction of liver metastases.^5,6 However, even when metastases can be resected or cytoreduced, most patients will develop recurrence or progression, with rates as high as 94% in the 5-year follow-up.⁷

Fortunately, several medical therapies have proven their utility in randomized clinical trials for patients with advanced or metastatic NETs. Somatostatin analogs (SSAs) were first introduced in the 1980s⁸ and were found to be very effective for symptom control in patients with NETs.⁹ Long-acting SSAs were subsequently shown to improve progression-free survival (PFS) in patients with advanced midgut and gastroenteropancreatic (GEP) NETs.^10,11 Other medical options include everolimus for SBNETs, PNETs, and lung NETs,^12,13 sunitinib for PNETs,¹⁴ and capecitabine/temozolomide chemotherapy (CT) for PNETs.¹⁵ Peptide receptor radionuclide therapy (PRRT) has also been shown to be an effective treatment for NETs. This strategy takes advantage of the binding of radiolabeled SSAs to somatostatin receptors present on most NETs, which are then internalized, causing selective cytotoxicity to NET cells. The NETTER-1 randomized trial compared ¹⁷⁷Lu-DOTA-(Tyr³)-octreotate (DOTATATE) PRRT plus SSA with SSA alone in patients with advanced, progressive midgut NETs and demonstrated significantly improved PFS at 20 months (65.2% vs. 10.8%).¹⁶ As a result, the US FDA approved ¹⁷⁷Lu-DOTATATE PRRT as another option for patients with GEP-NETs in 2018.

One of the biggest challenges for clinicians taking care of patients with GEP-NETS is the sequencing of these therapeutic options as patients develop progressive disease.¹⁷ It has generally been our strategy to resect GEP-NETs and cytoreduce liver metastases where possible, and then treat patients with SSAs. When these patients inevitably progress, one must choose between these different systemic options, which have varying indications, response rates, and toxicities. In the United States (US), PRRT has become increasingly popular since its FDA approval, and it has been suggested that PRRT may be more effective in patients with lower liver tumor burden.¹⁸ However, the NETTER-1 trial only showed efficacy in midgut NETs, while longer-term follow-up studies failed to demonstrate a benefit in OS.¹⁹ Our institution was an early adopter of PRRT and we have treated or referred patients for this therapy for over 2 decades. The objective of this study was to determine whether there was a survival benefit of treating patients with PRRT versus other therapies after progression of their surgically resected GEP-NETs, and whether this was true for patients with PNETs as well as SBNETs.

METHODS

Patients

A single-institutional NET database spanning from 1999 to 2022 was reviewed for patients who had surgical resection and/or cytoreduction of GEP-NETs and later had disease progression. Informed consent was provided by all patients in the database in accordance with a protocol approved by the University of Iowa Institutional Review Board. Progression was determined using Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 criteria based on imaging (defined as a ≥20% increase in the diameter of target lesions with an increase of at least 5 mm, or the appearance of new lesions).²⁰ Patients had an histopathologically confirmed diagnosis of GEP-NET originating in the stomach, duodenum, small bowel, colon, or pancreas, or of unknown primary with predominantly gastrointestinal (GI)/abdominal tumor burden and molecular features consistent with GEP origin. Exclusion criteria included NETs of bronchopulmonary or hindgut origin, mixed neuroendocrine/non-neuroendocrine neoplasms (MiNENs), insufficient or inconclusive pathology and imaging data, or lack of follow-up data. Patients who received PRRT prior to surgical resection or who received PRRT after resection as planned ‘adjuvant’ therapy without documented disease progression were also excluded.

Patients were categorized into one of three groups for comparison: (1) patients who never received PRRT following progression after surgical resection (referred to as ‘no PRRT’); (2) patients who received PRRT as first-line therapy following disease progression after resection (referred to as ‘upfront PRRT’); and (3) patients who received PRRT as second-, third-, or fourth-line therapy after surgical resection (referred to as ‘delayed PRRT’). Clinicopathologic variables were recorded in a prospective database and compared between groups. Pathology samples were independently reviewed by an experienced NET pathologist and classified according to the WHO 2019 classification for GEP neuroendocrine neoplasms (NENs).²¹ Ki-67 proliferation index and somatostatin receptor type 2 (SSTR2) expression were assessed by immunohistochemistry (IHC). SSTR2 expression levels were evaluated using the antibody UMB-1 with methods previously described²² and classified, based on H-score, as either positive (>30), intermediate (1–30), or negative (0).²³ For imaging-based determination of SSTR2 expression, tumors were classified as ‘SSTR2-positive’ based on strong radiotracer avidity on DOTA-PET imaging and/or a Krenning score of >2 on Octreoscan.^24,25 Cancer staging followed the American Joint Committee on Cancer (AJCC) Cancer Staging Manual, Eighth Edition.²⁶

Statistical Analysis

Median OS was calculated as the time from surgical resection or initial progression (by RECIST 1.1) after surgery to the last date of follow-up or death from any cause. PFS was calculated in several ways. Initially, PFS from the date of operation (PFS1) until the time of initial progression by RECIST 1.1 was determined. The time to the next disease progression (PFS2) was calculated from the date of initial progression after surgical resection until the next progression by RECIST 1.1. To determine treatment-specific survival following PRRT, OS and PFS were also calculated separately from the time of initial PRRT dose administration for PRRT groups (PFS2postPRRT). The Kaplan–Meier method was used to estimate survival for all groups. Clinicopathologic factors were compared between groups using the Chi-square and Fisher’s exact tests for categorical variables and Wilcoxon or Kruskal–Wallis rank-sum tests for continuous variables. Univariate and multivariate Cox proportional hazards models were used to evaluate factors associated with survival. Variables that were significant by univariate analysis were included in the multivariate Cox model and were incorporated using a stepwise forward and backward selection process with significance set at p < 0.05.

In order to control for immortal time bias, rather than assigning patients to a fixed group (PRRT/no PRRT), all patients start in the no-PRRT group and switch into the PRRT group when they receive their first dose of PRRT. Cox regression models using time-dependent covariates were then performed to evaluate the effect of multiple variables on survival, assuming that the effect of PRRT on the hazard is proportional (i.e., constant across time).

RESULTS

Time Since Surgical Resection

An initial analysis comparing all patients who underwent surgical resection and subsequently received PRRT (n = 118) with those who did not (n = 468) indicated a significantly shorter PFS in the PRRT group (median 20.4 vs. 90.3 months, p < 0.001) (electronic supplementary material (ESM) Fig. 1). There was no significant difference in OS in the PRRT versus no-PRRT groups for all patients.

We then focused only on patients with documented progression after surgical resection by RECIST 1.1 criteria. Among these patients, the time to first progression after surgical resection (PFS1) in the PRRT (n = 95) and no-PRRT groups (n = 142) was the same (26.4 vs. 28.7 months, p = 0.77) (Fig. 1a). The median OS from the time of surgical resection was significantly longer in the PRRT group versus the no-PRRT group (median 156 vs. 106 months, p ≤ 0.001) (Fig. 1b). The main differences between the PRRT and no-PRRT groups were that there were more patients with liver metastases receiving PRRT (94% vs. 77%, p < 0.001), slightly younger median age (57 vs. 59 years, p = 0.014), and fewer SSTR2-negative tumors (1.3 vs. 5.8%, p = 0.017) (Table 1). The treatments received by the patients at initial progression, their year of progression, and type of PRRT given are reported in ESM Table 1.

FIG. 1.

a PFS and b OS in all patients having surgery with and without PRRT (calculated from the time of surgery). c PFS and d OS calculated from the time of initial progression after surgery with and without PRRT. PFS progression-free survival, OS overall survival, PRRT peptide receptor radionuclide therapy, NET neuroendocrine tumor, mPFS median progression-free survival, mOS median overall survival

TABLE 1.

Patient characteristics

Characteristica	No PRRT [n = 142]	PRRT [n = 95]	p-Valuea
Age, years [median (IQR)]	59 (52, 69)	57 (49, 64)
Sex			0.14
Female	61 (43)	50 (53)
Male	81 (57)	45 (47)
Differentiation			0.2
Well	105 (95)	83 (99)
Poor	5 (4.5)	1 (1.2)
WHO grade			0.2
G1	50 (36)	25 (27)
G2	69 (50)	58 (62)
G3	18 (13)	11 (12)
T stage			0.3
T1–T2	48 (34)	25 (26)
T3–T4	78 (55)	62 (65)
TX	16 (11)	8 (8.4)
N stage			0.12
N0	20 (14)	6 (6.3)
N1–N2	115 (81)	86 (91)
NX	7 (4.9)	3 (3.2)
M stage			< 0.001
M0	33 (23)	6 (6.3)
M1	109 (77)	89 (94)
Primary site			0.095
Small bowel	79 (56)	65 (68)
Pancreas	52 (37)	27 (28)
Other	11 (7.7)	3 (3.2)
SSTR2 status by IHC			0.017
Positive	61 (88)	77 (99)
Indeterminate	4 (5.8)	0 (0)
Negative	4 (5.8)	1 (1.3)
SSTR2 H-score [median (IQR)]	255 (179, 300)	270 (160, 300)

Bold values indicate statistical significance (p-value < 0.05)

Data are expressed as n (%) unless otherwise specified

^aWilcoxon rank-sum test, Pearson’s Chi-square test, Fisher’s exact test

PRRT peptide receptor radionuclide therapy, IQR interquartile range, WHO World Health Organization, IHC immunohistochemistry, SSTR2 somatostatin receptor type 2, H-score on a scale from 0 to 300 based on immunohistochemistry

Time Since Initial Progression

We next analyzed time to second progression and OS from the date of initial progression after surgical resection rather than from the date of surgical resection (PFS2). This revealed that PRRT dramatically extended both PFS (median 53 vs. 11 months, p < 0.001) and OS (median 74 vs. 38 months, p < 0.001) (Fig. 1c, d).

Sixty-one patients received PRRT within 12 months of progression at a median of 4.6 months (upfront PRRT; range 0.9–10.8 months). Another group of 34 patients received PRRT later (delayed PRRT; median 25.0 months after progression, range 12.1–94.1 months). For these delayed patients, PRRT followed one to three other treatment modalities. When the above PFS2 analysis is carried out separately for these groups, greater benefits were seen in the delayed-PRRT group for PFS (72 months vs. 37 months in the upfront group vs. 11 months for the no-PRRT group) and OS (167 months vs. 57 months in the upfront group vs. 38 months for the no-PRRT group).

The baseline for the PRRT patients was then shifted to coincide with the start of treatment (PFS2postPRRT; for the no-PRRT group, PFS2 is the time of second progression after surgical resection and other treatments) to reduce the effect of immortal time bias. The group receiving PRRT continued to demonstrate a benefit in both PFS (median 32 vs. 11 months, p < 0.001) (Fig. 2) and OS (median 50 vs. 38 months, p < 0.009). When this analysis is further separated into upfront and delayed groups, the benefit of PRRT was similar for both groups (ESM Fig. 2).

FIG. 2.

a PFS and b OS following initial progression after surgery (no-PRRT group) or start of PRRT (PRRT group). PFS progression-free survival, OS overall survival, PRRT peptide receptor radionuclide therapy, mPFS median progression-free survival, mOS median overall survival

Using these timepoints, univariate analysis for PFS revealed that WHO grade, primary site, and receiving PRRT were significantly correlated; on multivariate analysis, only PRRT remained significant (Table 2). Univariate analysis for OS revealed that age, M stage, and whether PRRT was received were all significantly correlated with survival and all remained significant on multivariate analysis (Table 2).

TABLE 2.

Univariate and multivariate testing for correlations with PFS and OS survival in 237 patients

Variable	Univariable			Multivariable
	HR	95% CI	p-Value	HR	95% CI	p-Value
Cox model for progression-free survival (from first progression or start of PRRT)
Age	1.01	1.00–1.02	0.2
Sex			0.4
Female	–	–
Male	0.88	0.64–1.22
WHO grade			0.041			0.075
G1–G2	–	–		–	–
G3	1.64	1.05–2.57		1.60	0.97–2.64
T stage			0.6
T1–T2	–	–
T3–T4	0.83	0.57–1.19
TX	0.83	0.47–1.45
N stage			0.7
N0	–	–
N1–N2	0.83	0.50–1.37
NX	0.69	0.27–1.76
M stage			0.6
M0	–	–
M1	0.90	0.59–1.36
Primary site			0.018			0.8
Pancreas	–	–		–	–
Small bowel	0.68	0.48–0.95		0.89	0.62–1.28
Other	1.45	0.76–2.77		1.03	0.51–2.07
Received PRRT			< 0.001			< 0.001
No PRRT	–	–		–	–
PRRT	0.35	0.24–0.49		0.30	0.21–0.44
SSTR2 by IHC			0.14
Positive	–	–
Indeterminate	2.56	0.79–8.31
Negative	2.43	0.87–6.77
Cox model for overall survival (from first progression or start of PRRT)
Age	1.02	1.01–1.04	0.003	1.02	1.01–1.04	0.002
Sex			0.4
Female	–	–
Male	1.16	0.81–1.68
WHO grade			0.3
G1	–	–
G2	1.02	0.68–1.53
G3	1.57	0.87–2.81
T stage			0.4
T1–T2	–	–
T3–T4	1.26	0.83–1.90
TX	0.93	0.47–1.84
N stage			0.9
N0	–	–
N1–N2	1.14	0.66–1.96
NX	0.99	0.33–2.98
M stage			0.002			< 0.001
M0	–	–		–	–
M1	2.33	1.27–4.25		3.05	1.65–5.63
Primary site			0.089
Pancreas	–	–
Small bowel	1.56	1.04–2.33
Other	1.43	0.60–3.42
PRRT			0.008			0.006
No PRRT	–	–		–	–
PRRT	0.59	0.40–0.88		0.57	0.38–0.86
SSTR2 by IHC			0.12
Positive	–	–
Indeterminate	3.41	1.03–11.3
Negative	2.02	0.72–5.68

Bold values indicate statistical significance (p-value < 0.05)

HR hazard ratio, CI confidence interval, PRRT peptide receptor radionuclide therapy, WHO World Health Organization, SSTR2 somatostatin receptor type 2, IHC immunohistochemistry

Time-Dependent Cox Modeling

Figure 3 shows the results of the Cox proportional hazards model using time-dependent covariates. This analysis indicates that PRRT leads to a 39% reduction in risk of death (p = 0.03) and a 25% reduction in risk of progression (p = 0.09), although the latter effect was not statistically significant. The analysis also indicates that younger patients and patients without metastatic disease have a lower risk of death. Furthermore, patients with SBNETs are less likely to experience progression compared with other primary tumor locations.

FIG. 3.

Time-dependent covariate analysis of a PFS and b OS in all 237 patients. PFS progression-free survival, OS overall survival, PRRT peptide receptor radionuclide therapy, AIC Akaike information criterion

Figure 4 presents the same analysis where we have excluded patients receiving PRRT >1 year after first progression from the Cox model. This analysis indicates a significantly stronger effect of PRRT on both OS (56% reduction in risk, p = 0.001) and PFS (65% reduction in risk, p < 0.001). The impact of grade upon both survival and progression can also be seen in this analysis, whereas it was not seen when the delayed-PRRT patients (who tended to have higher-grade tumors) were included.

FIG. 4.

Time-dependent covariate analysis of a PFS and b OS restricting to PRRT received within 1 year of progression (142 in the no-PRRT group, 61 in the upfront-PRRT group). PFS progression-free survival, OS overall survival, PRRT peptide receptor radionuclide therapy, AIC Akaike information criterion

Modified versions of Kaplan–Meier plots that were created to reflect time-dependent covariate analyses are shown in Fig. 5 (modified since patients shift from one group to the other as they begin PRRT) for a hypothetical patient who begins PRRT at 6 months (indicated by the dotted line). The red line indicates the survival probability for patients as if they never began PRRT, while the blue line indicates the survival probability after beginning PRRT. These lines are identical prior to 6 months as the treatment has not yet started. In all scenarios, a benefit for PRRT can be seen, although the benefit is considerably larger when the delayed-PRRT patients are excluded from the analysis (Fig. 5c, d).

FIG. 5.

Modified Kaplan–Meier curves from time-dependent covariate analyses of a PFS and b OS factoring in immortal time bias in all patients. c PFS and d OS factoring in immortal time bias for patients receiving PRRT within 1 year of progression. The dotted line shows the hypothetical assumption that PRRT is given at 6 months; prior to this, there is only one line (the no-PRRT group), and after this, the blue PRRT and the red no-PRRT lines begin to diverge. PFS progression-free survival, OS overall survival, PRRT peptide receptor radionuclide therapy

Subgroup Analysis by Primary Tumor Site

In subgroup analyses stratified by tumor site, patients with SBNETs and PNETs had PFS benefit from receiving PRRT. The median PFS in SBNET patients having PRRT (PFS2postPRRT; n = 65) was 32.0 months (31.2 months upfront and 32.4 months delayed) versus 12.4 months in the no-PRRT group (from initial progression, n = 79; p < 0.0001) (ESM Fig. 3a). The median OS from time of first progression was 40.5 (42.9 months upfront, n = 41; 40.5 months delayed, n = 24) versus 29.8 months in the no-PRRT group (p < 0.06) (ESM Fig. 3b). The median PFS in PNET patients having PRRT (PFS2postPRRT; n = 27) was 34.3 months (33.3 upfront, 34.3 delayed) versus 10.6 months in the no-PRRT group (from initial progression, n = 52; p < 0.003) (ESM Fig. 3c). Median OS from time of first progression was 123.4 months with PRRT (123.4 months upfront, n = 17; 91.5 months delayed, n = 10) versus 56.6 months in the no-PRRT group (n = 52, p = 0.2). Comparison of patients with SBNETs who had PRRT (n = 65) with those with PNETs (n = 27) revealed no significant difference in median PFS (32.0 vs. 34.3 months, p = 0.4) or OS (40.5 vs. 123 months, p = 0.09). When we performed time-dependent covariate analysis in SBNET patients, restricting to patients receiving PRRT within 1 year of progression, there was significant improvement in PFS and OS in patients receiving PRRT (ESM Fig. 4). M stage, age, and grade also influenced PFS and OS. In the comparable set of PNET patients, the benefit of PRRT on PFS remained significant, while that for OS did not (but there were only 17 patients in the PRRT group) (ESM Fig. 5).

DISCUSSION

The best chance for long-term survival in patients with GEP-NETs is surgical resection. This allows for removal of the primary tumor and regional nodal disease, and may prevent the development of distant metastases. Resection may also improve symptoms and prevent complications such as obstruction, bleeding, peritoneal seeding, and hormonal syndromes. However, as many as 30% of patients with SBNETs and 64% of those with PNETs present with metastatic disease,²⁷ and in these patients, cure is usually not possible. Surgical cytoreduction of metastases may improve survival in these patients^27–29 but few will be rendered tumor-free.⁷ Most patients will have some tumor left behind in addition to micrometastases that were not visible on preoperative imaging.³⁰ The optimal treatment upon progression has not been determined but might include medical treatment with SSAs, CT, targeted therapy (TT), liver-directed therapy, or PRRT.¹⁷

The value of systemic therapy for advanced, metastatic, or progressive GEP-NETs has been established in a variety of clinical trials. Randomized trials have demonstrated modest improvements in PFS using SSAs relative to placebo.^10,11 Treatment with everolimus has shown improved PFS versus placebo in patients with GI tumors, lung tumors, and PNETs,^12,13 as has sunitinib in PNET patients.¹³ None of these drugs have been shown to improve OS, possibly due to crossover from the placebo to active treatment arms. A recent phase II trial showed high response rates with capecitabine and temozolomide in PNET patients,¹⁵ suggesting another valuable option for those with advanced disease.

Somatostatin receptor-directed PRRT with octreotide analogs was first described in 1994 using ¹¹¹Indium-pentetreotide,³¹ and later, two phase I studies were performed using this isotope.^32,33 Limited tumor efficacy was seen with these γ-emitting agents, and it was believed that β particles might impart higher radiation doses to tumors.³² A phase I trial using the β-emitter ⁹⁰Yttrium-DOTATOC was reported from Basel in 1999,³⁴ followed by a phase II trial in 2001.³⁵ The larger phase II experience of this group was described for 1109 patients in 2011.³⁶ The group in Rotterdam (Erasmus Medical Center) switched from ¹¹¹In to the β- and γ-emitter ¹⁷⁷Lutetium-octreotate, which had similar uptake in the kidney and liver as ¹¹¹In-Pentetreotide, but three to four fold higher affinity for tumors.³⁷ They initially reported their experience in 131 patients in 2005,³⁸ followed by a larger group of over 500 patients in 2008.³⁹ Both ⁹⁰Y and ¹⁷⁷Lu radioisotopes conjugated to SSAs for PRRT in NET patients have shown reasonable response or disease stabilization rates with acceptable and limited toxicity. Whether treatment with ⁹⁰Y or ¹⁷⁷Lu is better is unclear, but renal toxicity is higher with ⁹⁰Y. Furthermore, the Erasmus group reported that response rates were higher at their institution using ¹⁷⁷Lu over ⁹⁰Y-DOTA octreotide analogs with a 2.1-fold higher tumor residence time.³⁹

Although these PRRT studies were groundbreaking, they were phase I and II studies and PRRT was not compared with other standard-of-care treatments. A randomized, phase III trial was finally performed in 41 centers (8 countries) between 2012 and 2016 (NETTER-1) that compared treating patients with advanced, progressive midgut NETs with ¹⁷⁷Lu-DOTATATE PRRT plus octreotide long-acting repeatable (LAR; 30 mg every 4 weeks) with high-dose octreotide LAR alone (60 mg every 4 weeks).¹⁶ The median PFS in these treatment arms was 25.0 and 8.5 months, respectively, showing a clear benefit for PRRT in this population. At a mean follow-up of 76 months, the median OS was 48 and 36 months, respectively (p = 0.30). The study also allowed for crossover, and 36% of patients in the high-dose octreotide LAR group later received PRRT, obscuring any OS benefit.¹⁹ These studies showed that ¹⁷⁷Lu-DOTATATE was well tolerated, with 6% developing grade 3–4 adverse events (2% myelodysplastic syndrome), and nephrotoxicity was similar in the two groups (5% vs. 4%). Based on the results of the NETTER-1 trial, and the efficacy seen for PNETs and other NETs from the Erasmus studies, the FDA approved ¹⁷⁷Lu-DOTATATE PRRT as a treatment for somatostatin receptor-positive GEP-NETs in 2018.⁴⁰

An ongoing question for patients with advanced NETs is the sequencing of therapies, which include surgery, liver-directed therapy, SSAs, CT, TTs, and PRRT. Most of these PRRT studies enrolled patients with advanced, metastatic, unresectable, or progressive tumors. Kwekkeboom et al. suggested that if the tumor burden is moderate, then a good strategy is to wait until tumor progression to treat with PRRT.³⁹ They also noted that remissions occurred more commonly in those with a limited number of liver metastases.³⁸ Campana et al. found that low tumor burden and lower grade correlated with improved PFS with PRRT.⁴¹ Ezziddin et al. reported that some of the independent predictors of decreased survival after PRRT were >25% hepatic tumor burden, Ki-67 >10%, and Karnofsky performance score of <70%.⁴² In the NETTER-1 trial, a PFS advantage was seen in patients receiving PRRT for those with low (<25%), moderate (25–50%), and high (>50%) liver tumor burdens. Patients whose target lesions were <3 cm in size and who were receiving PRRT had improved PFS over those with one or more lesions over 3 cm in size; tumor shrinkage measured at up to 72 weeks was 29% and 14% for lesions <3 cm and >3 cm, respectively.⁴³ Prior to many of these observations, our strategy at Iowa for the past 2 decades has been to remove primary tumors and involved regional lymph nodes, and to perform cytoreduction of liver metastases where possible. Part of this strategy was to potentially improve response rates to PRRT when patients progress after surgery, with the ultimate goal of improving the OS of our patients.

Before PRRT was accessible in the US, our NET group, led by Thomas and Sue O’Dorisio, recognized the promising role of PRRT from the reports coming out of Europe. We made this modality available to our patients beginning in 2001. Some patients traveled to Europe for PRRT, while others were treated on our multi-institutional trial of ⁹⁰Y-DOTATOC (⁹⁰Y-edotreotide). This trial enrolled 90 patients at 18 centers between July 2001 and August 2002, for patients with malignant carcinoid tumors and symptoms not controlled by SSAs, with metastatic disease, or showing disease progression.⁴⁴ Sharma et al. described our larger Iowa experience with 135 patients having PRRT between 2001 and 2011, where 38% of patients had SBNETs, 26% had PNETs, 13% had bronchial NETs, and 23% had unknown primaries or NETs of other sites.⁴⁵ Of these patients, 69% were treated in Basel, 26% were treated at the University of Iowa, and remainder were treated at other centers. Accordingly, 83% of these patients received ⁹⁰Y-DOTATOC and 17% received ¹⁷⁷Lu-DOTATATE. Later, patients began going to Erasmus for ¹⁷⁷Lu-DOTATATE, or received this isotope here as part of the NETTER-1 trial between September 2012 and January 2016. The current study describes results in a specific population of 95 GEP-NET patients who first had surgical resection of their primary tumors and/or metastases and who later received PRRT after progression of their disease. In this group, 72% received ¹⁷⁷Lu-DOTATATE, 15% received ⁹⁰Y-DOTATOC, and 13% of patients received both ¹⁷⁷Lu-DOTATATE and ⁹⁰Y-DOTATOC (ESM Table 1). These patients were compared with 142 patients who had surgical resection of their primary tumors and/or metastases and later progressed but were not treated with PRRT, making this a retrospective cohort study. It should be noted that 61% of patients receiving PRRT were treated prior to FDA approval in the US, and the reasons for treating patients with PRRT versus no PRRT varied. During some years, patients had to be willing to travel to Europe and potentially pay for this therapy out-of-pocket. During the NETTER-1 trial, patients needed to be willing to potentially be randomized to the octreotide-only group. Other patients may have been unwilling to be treated with radioisotopes, or did not meet the inclusion criteria for clinical trials (e.g. PNET patients were excluded from NETTER-1). Table 1 reveals that both groups were similar in terms of age, sex, differentiation, grade, and T and N stage. A higher percentage of the PRRT group had metastatic disease (94% vs. 77%, p < 0.001), were SSTR2-positive by IHC (99% vs. 89%, p = 0.017), and had SBNET primaries (68% vs. 56%, p = 0.095). Both groups were very similar in that their median time to progression from the date of surgical resection was essentially the same (28.7 months for PRRT, 26.4 months for non-PRRT, p = 0.7 (Fig. 1a), suggesting that time to progression was not an important factor in deciding upon therapy.

Accepting the shortcomings of comparing these retrospective cohorts, it is clear that OS was improved for the PRRT versus no-PRRT groups when determined from the date of surgery (156 vs. 106 months, p = 0.001) (Fig. 1b). Using the most conservative estimate of PFS, from the time of progression after surgery in the no-PRRT group and time of initiation of PRRT after progression in the PRRT group, there was also significantly improved PFS after PRRT (32.4 vs. 11.0 months, p < 0.001) (Fig. 2a). Further analysis revealed that two subgroups of patients were receiving PRRT—an early group treated at a median of 4.6 months after progression, and a delayed group who had PRRT after progression following other treatments (with PRRT at a median of 24.5 months after first progression after surgery). When calculated from the time of progression from initial surgery in both groups, PFS and OS were longest in those patients in whom PRRT was delayed until after other treatments. The longer PFS and OS in the delayed group using this starting point highlights the fact that these patients may have derived benefit from the other therapies received. PFS and OS in the upfront- and delayed-PRRT groups were essentially the same when calculated from the time of PRRT (ESM Fig. 2a, b), suggesting that the benefits of PRRT are similar regardless of whether it is given early or late.

These analyses using different corrections show that giving PRRT at progression after surgical resection significantly improves both PFS and OS (Figs. 1 and 2, and ESM Fig. 2), the latter of which was not seen in the NETTER-1 study. Furthermore, the PFS benefit of PRRT seemed to extend to both SBNETs and PNETs (ESM Fig. 3), whereas only the former group was included in NETTER-1. The median PFS observed in these patients was higher than that reported in NETTER-1 (31–32 months for SBNET patients vs. 25.0 months in NETTER-1). Although it is tempting to believe that this improvement is related to our strategy of trying to cytoreduce patients in order to increase the efficacy of PRRT, differences in patient groups between the two studies may also have been a significant contributor.

Studies focusing on the efficacy of PRRT in patients having surgery for NETs are few. Bertani at al. evaluated a population of patients with grade 1–2 PNETs with non-cytoreducible liver metastases who had either primary tumor resection before PRRT (31 patients) or PRRT alone (63 patients).⁴⁶ Patients having primary tumor resection followed by PRRT had better responses to PRRT, improved median OS (112 vs. 65 months, p = 0.011), and improved median PFS (70 vs. 30 months, p = 0.002). Univariate survival analyses revealed that primary tumor resection was the only variable significantly associated with improved survival. The observation that resection of the primary tumor followed by PRRT improved survival might have been due in part to lower tumor burden. A large series from Bad Berka reviewed outcomes in 889 patients with stage IV NENs who had received PRRT, comparing 486 patients who had their primary tumor resected prior to PRRT and 403 who did not have their primary tumors removed.⁴⁷ More patients with PNETs did not have their primary tumors resected (56%), while more SBNET primary tumors were resected (83%). The median OS in patients having their PTs resected followed by PRRT was 134 months versus 67 months in the PRRT-only group (hazard ratio [HR] 2.79, p < 0.001). Median PFS in these two groups was 18 and 14 months, respectively (HR 1.21, p = 0.012), again showing a potential benefit to resecting primary tumors in patients having PRRT. The authors concluded that resection of the primary tumor led to better outcomes in patients receiving PRRT, but acknowledged that these patients tended to be younger, have better Karnofsky scores, and were more likely to have SBNETs. Pusceddu et al. performed a retrospective study examining PFS in patients from 25 Italian centers with unresectable, locally advanced, or metastatic well-differentiated GEP-NETs treated with SSAs who later progressed between 2000 and 2020.⁴⁸ Upon progression, these patients were treated with either PRRT (329 patients) or CT/TT (179 patients). All patients progressing after CT/TT later received PRRT. Surgical resection of PTs was performed in 73% of the PRRT group and 54% of the CT/TT group. Propensity matching was performed to reduce bias (including resection of PTs and metastases) and the authors found that patients receiving upfront PRRT had significantly improved PFS over those who had upfront CT/TT (2.2 years vs. 0.6 years; HR 0.37, p < 0.0001). This was seen in patients with SBNETs and PNETs but not for patients with Ki-67 >10%. Moreover, there was no difference in OS between groups. The suggestion from this study was that the choice of PRRT first at progression over CT/TT led to improved PFS, but PRRT was still useful if given later, as OS was similar in the two groups. Our findings were similar, with comparable PFS in patients receiving upfront and delayed PRRT (Fig. 2 and ESM Fig. 3).

In retrospective studies, there is potential for selection bias. The PRRT and no-PRRT groups were fairly similar in this study except the former had a higher percentage of patients with metastatic disease and more SSTR2-positive tumors. There could also have been a bias towards treating patients with more concerning progression with PRRT, and conversely, patients who were able to access PRRT might have been more affluent and more engaged in their own care, yet the fact that initial progression after surgery was identical between the PRRT and no-PRRT groups suggests that the groups were well-balanced. Since we expected patients with smaller disease volumes to respond better to PRRT, it is also possible that the no-PRRT group had higher liver tumor burdens and might have received other therapies better suited to larger tumor volumes, such as embolization or reoperation. Although we did not quantitate the liver tumor burden in both groups, ESM Table 1 shows similar percentages of patients having embolization (9/142 [6.4%] in the no-PRRT group, and 4/45 [8.9%] in the PRRT group) and reoperation (10/142 [7.0%] in the no-PRRT group, and 5/95 [5.3%] in the PRRT group) following their first progression. One correction we performed to further help reduce bias was to not count PFS and OS in the PRRT group until PRRT was received, which reduced PFS from 53.4 months to 32.4 months (Figs. 1c, 2a), but compared with the no-PRRT group, the survival difference remained significant. The same was seen for OS (where the median PRRT group OS decreased from 73.9 to 49.8 months).

Unlike in a randomized trial, the treatment groups were not assigned at initial progression. Patients are presented at a tumor board, scheduled, and travel arrangements made, and they must have long-acting SSAs withheld for 1 month if they are to get PRRT. Therefore, there is usually a delay of at least several months before initiating therapy, and sometimes much longer. During this time, from the point of view of the above analysis, patients in the PRRT group cannot possibly experience death or recurrence. If they did, it would be recorded as a death or recurrence in the no-PRRT group. This is known in the survival analysis literature as immortal time bias: any delayed event or therapy is almost guaranteed to show a benefit if this is not accounted for. Time-dependent covariate analysis is a method to adjust for immortal time bias, which is widely accepted as one of the best methods for analyzing this type of observational study, in which the intervention of interest is not assigned at baseline.^49,50 Rather than assign patients to a fixed group (PRRT or no PRRT), all patients start in the no-PRRT group and switch into the PRRT group when they receive their first dose of PRRT. This can be accomplished in the Cox regression model by assuming that the effect of PRRT upon the hazard is proportional (i.e., constant across time). In these analyses, the reduction of hazard for all patients receiving PRRT lost significance for PFS (Fig. 3a) but was maintained for those receiving PRRT within 1 year of progression (Fig. 4a). The inclusion of patients who received PRRT long after their initial progression (>1 year) is problematic, as the reasons for waiting to give PRRT might be due to differences in tumor biology, extent of disease, or response to previous therapies. Identifying a comparable control group for these patients is also challenging. Both PRRT groups (all PRRT patients and those receiving PRRT within 1 year of progression) demonstrated significantly reduced HRs for OS as compared with the no-PRRT group (Figs. 3b, 4b). Modified Kaplan–Meier curves from time-dependent covariate analyses showed a greater magnitude of effect on PFS and OS for the PRRT within 1-year group (Fig. 5). Our finding of a significant OS benefit, even after eliminating the source of immortal time bias through a time-dependent covariate analysis, strongly supports the effectiveness of PRRT.

CONCLUSION

This retrospective study showed the benefit of PRRT after surgical resection of primary and/or metastatic tumors. These benefits include improved PFS and OS for SBNETs (ESM Fig. 4), PFS for PNETs (ESM Fig. 5), and for PFS if PRRT was given early or in a delayed fashion (ESM Fig. 3). Sequencing of therapies for patients with advanced NETs remains an understudied area but the results of our analysis suggest that the strategy of surgical resection followed by PRRT at some point after progression results in improved outcomes for patients over resection and other therapies alone. Although there are several ongoing randomized trials evaluating PRRT in various settings (COMPETE, NCT03049189; NETTER-2, NCT03972488; COMPOSE, NCT04919226; ComPareNET, NCT05247905), none are designed to look at the efficacy of PRRT after surgical cytoreduction and will therefore not help to determine whether reducing the volume of disease will improve the results of PRRT. This would ideally be addressed in a randomized clinical trial but this would be very difficult to carry out due to the need for extended follow-up, the possibility of crossover, and inconsistencies of surgical cytoreduction across sites. For the present, we can conclude that PRRT is a valuable adjuvant therapy following progression after surgical resection and cytoreduction.

Supplementary Material

SUPPLEMENTARY INFORMATION The online version contains supplementary material available at https://doi.org/10.1245/s10434-024-16463-7.