Is salvage treatment necessary after incomplete resection of rectal neuroendocrine tumors: A systematic review and meta-analysis

Abstract

BACKGROUND

Incomplete resection of rectal neuroendocrine tumors (NETs), defined by positive or indeterminate margins or lymphovascular invasion, raises concerns regarding residual disease and recurrence. However, the benefits of salvage treatment in these cases remain unclear.

AIM

To evaluate the oncologic impact of salvage treatment compared with observation after incomplete endoscopic resection of rectal NETs through a systematic review and meta-analysis.

METHODS

We conducted a systematic review and meta-analysis in accordance with the preferred reporting items for systematic reviews and meta-analyses guidelines. MEDLINE, EMBASE, and the Cochrane Library were searched from their inception until May 2025. Eligible studies included patients with incompletely resected rectal NETs managed with salvage treatment or observation, reporting outcomes of residual tumors or recurrence. Pooled estimates were calculated using random-effects models with Hartung-Knapp adjustments.

RESULTS

Thirty-four studies with 2279 cases of incomplete endoscopic resection met the inclusion criteria. The incomplete resection rates differed markedly according to the initial resection method (17 studies): 73.1% for cold snare polypectomy, 29.8% for conventional endoscopic mucosal resection (EMR), 28.4% for modified EMR, and 14.7% for endoscopic submucosal dissection. Among the 19 studies that evaluated salvage treatment, the pooled residual tumor rate was 25.0% [95% confidence interval (CI): 12.0%-40.0%]. The crude recurrence rates from 31 studies favored salvage treatment over observation (0.96% vs 2.96%, P = 0.003). However, a meta-analysis of nine comparative studies found no statistically significant difference in recurrence risk (odds ratio = 0.89; 95%CI: 0.40-2.02).

CONCLUSION

Given the relatively high residual tumor rate and low incidence of recurrence, salvage treatment may be justified as both a diagnostic and therapeutic approach after incomplete resection of rectal NETs. Although its benefits in preventing recurrence remain unclear, clinical decisions should be individualized, as these findings are based on low-certainty evidence.

Key Words: Rectum; Neuroendocrine tumors; Residual neoplasm; Salvage therapy; Observation

Core Tip: Incomplete resection of rectal neuroendocrine tumors raises concern for residual disease and recurrence, yet the optimal management remains debated. In this systematic review and meta-analysis of 34 studies, residual tumors were identified on salvage pathology in about 25% of cases, whereas recurrence was rare in both salvage and observation groups. Although crude data suggested fewer recurrences with salvage treatment, comparative meta-analysis showed no significant difference. These findings highlight salvage resection as a diagnostic and therapeutic option, but its routine use for recurrence prevention is uncertain. Clinical decisions should be individualized.

INTRODUCTION

Rectal neuroendocrine tumors (NETs) are well-differentiated neoplasms that most commonly arise from enteroendocrine L cells of the distal gut, which produce peptide YY and glucagon-like peptides. In contrast, small-intestinal NETs typically derive from enterochromaffin cells that secrete serotonin[1-3]. Although rectal NETs account for a small proportion of colorectal neoplasms, their incidence has been steadily increasing, largely because of the widespread use of screening colonoscopy and advances in endoscopic imaging techniques[4,5]. A population-based study using surveillance, epidemiology, and end result data demonstrated that rectal NETs, along with small intestinal NETs, have the highest incidence among gastrointestinal NETs over the past four decades. This trend is particularly pronounced in Asian countries[6].

Small rectal NETs measuring less than 10 mm in diameter are generally considered suitable candidates for curative endoscopic resection. Techniques, such as conventional endoscopic mucosal resection (cEMR), modified EMR (mEMR) (including cap-assisted or band ligation-assisted EMR techniques), and endoscopic submucosal dissection (ESD) are widely utilized[7-10]. Rectal NETs measuring up to 20 mm in diameter may also be amenable to endoscopic resection, provided that high-risk features, such as lymphovascular invasion, positive margins, or involvement of the muscularis propria are not present[11,12]. However, incomplete resection, defined as a positive or indeterminate resection margin or the presence of lymphovascular invasion, raises concerns regarding residual tumor or recurrence[13,14].

Currently, there is no consensus for the optimal management of incompletely resected rectal NETs. While some guidelines support close surveillance for carefully selected low-risk lesions, others endorse salvage treatments, including repeat endoscopic resection, transanal endoscopic microsurgery, or radical surgery[1,3,15,16]. Given the variability in clinical practice and the lack of robust data, the benefit of salvage therapy remains uncertain.

A clearer understanding of the recurrence risk, residual lesion rate, and treatment outcomes after incomplete resection is essential to inform evidence-based recommendations. This systematic review aimed to synthesize existing data on the natural course of incompletely resected rectal NETs and evaluate the oncological impact of salvage treatments.

MATERIALS AND METHODS

Protocol registration

This systematic review and meta-analysis followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) 2020 guidelines[17]. This study was prospectively registered in PROSPERO (registration No. CRD420251054723).

Search strategy and study selection

A comprehensive literature search was performed across three major databases: MEDLINE, EMBASE, and the Cochrane Library. The final search was completed on May 21, 2025. Detailed search strategies for each database are provided in Supplementary material.

Eligible studies investigated outcomes in patients with incompletely resected rectal NETs, defined as tumors with positive or indeterminate resection margins or the presence of lymphovascular invasion in the resected specimen. An indeterminate margin was typically described as one that could not be reliably evaluated due to cautery artifact, tissue fragmentation, or tangential sectioning, making it difficult to determine tumor involvement. Studies were included if they reported residual tumors or recurrence and evaluated salvage treatment, observation, or both. A residual tumor was defined as histologically or macroscopically confirmed tumor tissue identified during additional treatment (e.g., salvage resection) following an initial incomplete resection. Recurrence was defined as the reappearance of the tumor after a period of no detectable disease during follow-up, as confirmed by colonoscopy or cross-sectional imaging, such as computed tomography. The exclusion criteria were as follows: Case reports or small case series (n < 5); Review articles, editorials, or letters; Studies involving tumors > 2 cm in size; Studies evaluating primary surgical treatment as initial management, and studies lacking outcome data on residual tumors or recurrent disease.

Two reviewers (Kim JH and Lee JW) independently screened the titles and abstracts of the identified records, followed by full-text reviews of potentially eligible articles. Discrepancies were resolved by discussion and consensus.

Data extraction and outcomes

Data were extracted from each study in a standardized format. For each included study, we collected information regarding the first author, publication year, country of origin, and study design. Patient-related data including the number of participants, sex distribution, and age were also recorded. Tumor-related characteristics, such as the size and completeness of resection, were documented, along with the duration of follow-up and the type of management strategy applied, whether salvage treatment or observation.

The primary outcomes were the residual tumor rate following salvage treatment and the recurrence rate in patients who underwent salvage treatment compared with those managed with observation.

Statistical analysis

We conducted a meta-analysis to estimate the pooled prevalence of residual tumors after salvage treatment in patients with incomplete resection. This single-arm proportion meta-analysis was performed using the metaprop function of the meta-package in R (version 4.2.2). The Freeman-Tukey double-arcsine transformation was applied to stabilize variance and minimize bias in pooled estimates, particularly when proportions were close to 0% or 100%. Study-level 95% confidence intervals (CIs) were calculated using the Clopper-Pearson exact method and pooled estimates were obtained using a random-effects model with Hartung-Knapp adjustment. Forest plots were used to display the individual study proportions and event counts.

To evaluate the recurrence outcomes, we first conducted a pooled crude analysis comparing the recurrence rates between patients who received salvage treatment and those managed with observation. The total number of recurrence events and total number of patients were aggregated across the studies for each group. Group differences were assessed using either the χ² test or Fisher’s exact test, as appropriate. A bar plot was generated to visualize the group-level proportions and absolute event counts.

For comparative analysis, we performed a meta-analysis limited to studies that reported recurrence outcomes in both treatment arms. Two-by-two contingency tables were constructed for each study and odds ratios (ORs) with 95%CIs were calculated. Pooled ORs were estimated using a random-effects model based on the DerSimonian-Laird method, with Hartung-Knapp adjustment to account for uncertainty. Studies with zero events in both arms were excluded, and a continuity correction of 0.5 was applied when one arm reported zero events. Between-study heterogeneity was evaluated using Cochran’s Q statistic, I², and τ². A forest plot was generated to illustrate both the individual and pooled effect estimates.

All statistical analyses were performed using the R software (version 4.2.2). Meta-analyses were conducted using the meta package and all visualizations were produced using ggplot2.

Risk of bias assessment

The risk of bias in nonrandomized studies was assessed using the risk of bias in nonrandomized studies of interventions (ROBINS-I) tool. This tool evaluates the bias across seven domains: Confounding factors, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of reported results. Each domain was rated as having low, moderate, or serious risk of bias. Two reviewers independently assessed each study and any discrepancies were resolved through discussion and consensus.

Certainty of evidence assessment

The certainty of evidence for the main outcomes, residual tumor rate, and recurrence risk was assessed using the grading of recommendations assessment, development, and evaluation (GRADE) approach[18]. This approach evaluates five domains: Risk of bias, inconsistency, indirectness, imprecision, and publication bias. Based on these assessments, the certainty of the evidence was rated as high, moderate, low, or very low.

RESULTS

Study selection and baseline characteristics

A total of 1353 records were identified through database searches (MEDLINE: 381; EMBASE: 967; Cochrane Library: 5), from which 409 duplicates were removed. After screening 944 titles and abstracts, 105 full-text articles were assessed for eligibility, and 34 studies met the inclusion criteria. The study selection process is illustrated in the PRISMA flow diagram (Figure 1).

Figure 1 Preferred reporting items for systematic reviews and meta-analyses 2020 study selection flow diagram. The flow diagram illustrates the study selection process according to the preferred reporting items for systematic reviews and meta-analyses 2020 guidelines. The number of records identified (n = 1353), screened (n = 944), and excluded (with reasons) are shown. The systematic review and meta-analysis included 34 studies.

All included studies were observational in design, and the majority were retrospective, single-center studies. Most studies were conducted in Asia, primarily in Korea (n = 19), China (n = 6), and Japan (n = 6) with one study each from the United States, France, and Italy. A total of 6761 patients with rectal NETs were included in 34 studies. Among these, 2279 cases of incomplete endoscopic resection, defined as positive or indeterminate resection margins or the presence of lymphovascular invasion, have been reported. However, because some studies enrolled only patients with incomplete resection, this figure does not represent the true prevalence of incomplete resection in all rectal NET cases.

Patient characteristics such as age, sex distribution, tumor size, and follow-up duration varied across studies. Most cohorts consisted of patients who were considered appropriate for endoscopic treatment. The follow-up duration varied substantially, reflecting heterogeneity in clinical settings and institutional protocols. Salvage treatments include a range of endoscopic techniques such as cEMR and mEMR, ESD, and in select cases, surgical interventions. The detailed study characteristics are summarized in Tables 1, 2 and 3[13,14,19-50].

Table 1 Baseline characteristics.

Ref.	Design	Country	Period	Rectal NET cases (n)	Primary endoscopic resection method
					cEMR	mEMR	ESD	Hybrid ESD	CSP	CFP	EC	Missing data
Cha et al[19]	Retrospective	Korea	Between 2008 and 2018	260	126	113	21
Cha et al[20]	Retrospective	Korea	Between January 2005 and December 2016	322	97	122	103
Chatila et al[21]	Retrospective	United States	Between January 2023 and October 2023	10	2					4		4
Cheminel et al[22]	Retrospective	France	Between January 2000 and December 2021	86	14				23	48		1
Cho et al[23]	Retrospective	Korea	Between January 2003 and December 2012	411	300		111
Choi et al[24]	Retrospective	Korea	Between December 2008 and November 2015	134		81	53
Chung et al[25]	Retrospective	Korea	Between January 2013 and April 2018	337	81	143	113
Dell’Unto et al[26]	Retrospective	Italy	Between 2010 and 2021	31		2			14			15
Duan et al[27]	Retrospective	China	Between December 2012 and June 2023	527	213	199	113	2
Heo et al[28]	Retrospective	Korea	Between September 2007 and October 2012	82	34	48
Hwang et al[29]	Retrospective	Korea	Between 2003 and 2014	407	19				68			320
Inoue et al[30]	Retrospective	Japan	Between January 2003 and December 2012	115	43	25	47
Jeon et al[31]	Retrospective	Korea	Between January 2007 and December 2009	31	6				25
Ju et al[32]	Retrospective	Korea	Between January 2018 and December 2022	121	8		2		9	95		7
Kaneko et al[33]	Retrospective	Japan	Between April 2001 and August 2013	46		22	24
Kim et al[34]	Retrospective	Korea	Between October 1999 and November 2010	107	105				2
Kim et al[35]	Retrospective	Korea	Between October 2004 and December 2019	401
Kim et al[36]	Retrospective	Korea	Between January 2012 and October 2021	68	10				11	47
Kim et al[37]	Retrospective	Korea	Between January 1995 and March 2007	30	22	6	2
Kusuhara et al[38]	Retrospective	Japan	Between September 2007 and May 2023	50
Lee et al[39]	Retrospective	Japan	Between 2009 and 2018	346	119	72	155
Li et al[40]	Retrospective	China	Between January 2010 and September 2019	428	211		179			12	26
Moon et al[14]	Retrospective	Korea	Between March 2007 and December 2009	35			35
Moon et al[41]	Retrospective	Korea	Between January 2000 and November 2012	407	211	92	38		66
Nakamura et al[42]	Retrospective	Japan	Between April 2001 and March 2012	170	11	108	51
Oh et al[43]	Retrospective	Korea	Between January 2010 and November 2023	212	103	22	8		35	44
Park et al[44]	Retrospective	Korea	Between January 2020 and August 2023	378	35	5			25	30
Seo et al[45]	Retrospective	Korea	Between 2008 and 2023	202
Sun et al[13]	Retrospective	China	Between January 2010 and December 2016	95			436
Sung et al[46]	Retrospective	Korea	Between January 2000 and June 2010	77	14	58	5
Takita et al[47]	Retrospective	Japan	Between January 2005 and March 2014	139	10	117	12
Wang et al[48]	Retrospective	China	Between February 2011 and September 2018	272			161	111
Zheng et al[49]	Retrospective	China	Between August 2010 and October 2019	98		36	62
Zhuang et al[50]	Retrospective	China	Between January 2013 and December 2021	326	39	6	247		12	32

NET: Neuroendocrine tumor; CSP: Cold snare polypectomy; cEMR: Conventional endoscopic mucosal resection; mEMR: Modified endoscopic mucosal resection; ESD: Endoscopic submucosal dissection; CFP: Cold forceps polypectomy; EC: Electrocoagulation.

Table 2 Clinicopathological characteristics of patients with incomplete resection, mean ± SD.

Ref.	Incomplete resection	Resection margin				Salvage treatment	Observation	Sex (M/F)	Age
		Positive	Indeterminate	LVI	Missing data
Cha et al[19]	77	25	5	4	43	21	10	19/12	54 (range: 27-84)
Cha et al[20]	87	57	19	11		44	43	52/35	48.6 ± 11.3
Chatila et al[21]	10	10				10	0	4/6	53.4
Cheminel et al[22]	86					86	0	51/35	54 (range: 18-87)
Cho et al[23]	67					5	62		49.6 ± 11.3
Choi et al[24]	20	6	13	1		1	19	13/6	50.8 ± 11.3
Chung et al[25]	19					9	8		47.5 (range: 39.6-57.1)
Dell’Unto et al[26]	31	31				16	15	18/31	60.5 (range: 22-85)
Duan et al[27]	181					76	105	109/72	59 (range: 50-66)
Heo et al[28]	7	6	0	1		0	7	4/3	49.8 ± 11.6
Hwang et al[29]	87					0	87	55/32	51.7 ± 11.7
Inoue et al[30]	23	7	16			7	16	14/9	62 (range: 27-85)
Jeon et al[31]	31	6	24		1	31	0	21/10	52.0 ± 11.8
Ju et al[32]	121	20	2		99	121	0	73/48	49 (range: 39-58)
Kaneko et al[33]	11					2	6	7/4	58 ± 12
Kim et al[34]	54	17	37	0		0	4	5/4	49.7
Kim et al[35]	118		34	84			54	31/23	55 (range: 27-82)
Kim et al[36]	68	59				9			68
Kim et al[37]	9	9				0			5
Kusuhara et al[38]	50		50	0			0	49/19	49.53 ± 11.20
Lee et al[39]	64	41				12	10	1	7
Li et al[40]	54	54	0			0	57	38/26	50.4 ± 10.8
Moon et al[14]	9		1	8	6/3	49.0 (range: 32-74)	54	33/21	49.9 ± 10.9
Moon et al[41]	148	20				14	7	107	14
Nakamura et al[42]	24	24	0			7	134	88/60	49.7 ± 10.9
Oh et al[43]	212		212	0			17	16/8	56.84 ± 13.27
Park et al[44]	146					95	48	79/44	43 (range: 21-76)
Seo et al[45]	202					76	126
Sun et al[13]	116	76				0	3	37	43
Sung et al[46]	19		6	13	12/7	52.3 ± 12.0	73	63/53	55.8 ± 11.3
Takita et al[47]	10					7	3
Wang et al[48]	22	22				0			3
Zheng et al[49]	13	13	0			0	19	13/9	49.3 ± 10.7
Zhuang et al[50]	83	70	13			18	13	8/5	48.3 ± 12.1

LVI: Lymphovascular invasion; M: Male; F: Female.

Table 3 Additional treatment modalities and follow-up duration, mean ± SD.

Ref.	Tumor size (mm)	Additional treatment method							Follow-up duration (month)
		cEMR	mEMR	ESD	Hybrid ESD	EFTR	Surgery	APC
Cha et al[19]	5.2 ± 1.7		16				5		39.8 (range: 24.2-119.7)
Cha et al[20]	4.72		33				11		40.5 (range: 4.27-102.7)
Chatila et al[21]	8.6			10
Cheminel et al[22]	5 (range: 2-10)	19	15	44	1	7			24 (range: 2-216)
Cho et al[23]	5.8 ± 3.2
Choi et al[24]	5.5 ± 2.4								27.4 ± 16.5
Chung et al[25]	4.0 (range: 3.0-6.0)		3	2			4		35 (range: 24.4-48.8)
Dell’Unto et al[26]	6 (range: 1.0-16.0)								36
Duan et al[27]	8.4 (range: 7.5-10.8)								72
Heo et al[28]	6.1 ± 2.7								13 (range: 3-57)
Hwang et al[29]	8.6 ± 3.6								29.1 ± 31.3
Inoue et al[30]	5 (range: 3-13)								24.5 (range: 6-108)
Jeon et al[31]	8.9 ± 3.2	31							12
Ju et al[32]	4 (range: 2-6)	27		94
Kaneko et al[33]	6.4								61.6 (range: 12-150)
Kim et al[34]	5.0 (range: 1.0-10.0)								31 (range: 13-121)
Kim et al[35]									39.3
Kim et al[36]	4.46 ± 2.11								51.85 ± 22.92
Kim et al[37]	6.3 ± 3.1								19.3 (range: 0-122)
Kusuhara et al[38]	5 (range: 2-10)	2							27
Lee et al[39]	5.4 ± 3.1			21					39
Li et al[40]			38 (range: 10-110)	3			4		48.8 (range: 12-131)
Moon et al[14]	4.7 (range: 1-9)								25 (range: 12-43)
Moon et al[41]	5.7 ± 3.1	2							1
Nakamura et al[42]	5.6 ± 2.4		48.3	2			7	2	56.8 (range: 24-227.7)
Oh et al[43]
Park et al[44]	4.7 ± 2.7	1							71
Seo et al[45]			35.0 ± 37.6	16			7		19.0 (range: 1.2-57.5)
Sun et al[13]	9.0 ± 4.0
Sung et al[46]	7.0 ± 2.8			43					61.4 (range: 33.4-125.3)
Takita et al[47]	5.7						4	2	18.5 (range: 5-107)
Wang et al[48]	6.9 ± 3.3								31 (range: 6-97)
Zheng et al[49]	6.3 ± 2.9								28.3 (range: 6-98)
Zhuang et al[50]	6.5

cEMR: Conventional endoscopic mucosal resection; mEMR: Modified endoscopic mucosal resection; ESD: Endoscopic submucosal dissection; EFTR: Endoscopic full-thickness resection; APC: Argon plasma coagulation.

Residual tumors following salvage treatment for incompletely resected rectal NETs

First, the risk of incomplete resection was assessed using the initial resection method. Among the 17 studies included in this analysis, the pooled incomplete resection rates varied substantially according to technique (Figure 2). Cold snare polypectomy (CSP) exhibited the highest rate at 73.1% (95%CI: 38.3%-95.6%), followed by cEMR at 29.8% (95%CI: 24.3%-34.5%) and mEMR at 28.4% (95%CI: 25.9%-35.8%). In contrast, ESD showed the lowest incomplete resection rate at 14.7% (95%CI: 12.9%-16.4%).

Figure 2 Incomplete resection rate by initial resection method. The forest plot shows pooled incomplete resection rates across different primary endoscopic techniques based on 17 studies[19,20,24,27,28,30,33,34,37,40-42,46-50]. Cold snare polypectomy exhibited the highest rate (73.1%), whereas endoscopic submucosal dissection had the lowest rate (14.7%). Conventional endoscopic mucosal resection and modified endoscopic mucosal resection had intermediate incomplete resection rates. This figure emphasizes the impact of the resection technique on histological completeness. CSP: Cold snare polypectomy; cEMR: Conventional endoscopic mucosal resection; mEMR: Modified endoscopic mucosal resection; ESD: Endoscopic submucosal dissection.

Next, we evaluated the frequency of residual lesions identified on salvage pathology in patients with an incomplete initial resection. In 19 studies, residual tumor status was reported based on pathological findings from salvage procedures. A single-arm meta-analysis using a random-effects model estimated the pooled residual lesion rate to be 25.0% (95%CI: 12.0%-40.0%). The inter-study heterogeneity was substantial (I² = 95.1%). These findings suggest that a considerable proportion of patients harbor residual disease as confirmed on salvage pathology (Figure 3).

Figure 3 Residual tumor rate following salvage treatment. The forest plot shows the pooled residual lesion rate after salvage treatment in patients with incompletely resected rectal neuroendocrine tumors based on a single-arm meta-analysis of 19 studies. The pooled rate was 25.0% (95% confidence interval: 12.0%-40.0%). This indicates that a substantial proportion of patients may harbor remnant tumors despite additional intervention. CI: Confidence interval.

Crude recurrence rates: Salvage treatment vs observation

Crude recurrence rates were calculated based on the data from 31 studies that reported recurrence outcomes. Among the 732 patients who received salvage treatment, 7 (0.96 %) experienced recurrence. In contrast, 34 recurrences occurred in 1148 patients (2.96%) who were managed with observation. This difference was statistically significant (P = 0.003), favoring the salvage group. These results are presented in Figure 4, which illustrates group-level recurrence proportions and absolute event counts.

Figure 4 Crude recurrence rates: Salvage treatment vs observation. The bar plot summarizes the crude recurrence rates from 31 studies by treatment group. Recurrence occurred in 0.96% of the patients receiving salvage treatment (7/776) and 2.96% of those managed with observation (34/1246), showing a statistically significant difference (P = 0.003) in favor of salvage treatment. O: Observation; S: Salvage treatment.

Comparative meta-analysis of recurrence risk: Salvage treatment vs observation

To further evaluate the effect of salvage treatment on recurrence, a meta-analysis was conducted using nine studies that reported recurrence outcomes in both the salvage and observation groups. The pooled OR for recurrence in the salvage group compared to the observation group was 0.89 (95%CI: 0.40-2.02). Although the point estimate numerically favored salvage treatment, the difference was not statistically significant. Between-study heterogeneity was negligible (I² = 0%, τ² = 0,P = 0.83). These findings (Figure 5) suggest that while salvage treatment may be associated with a lower recurrence rate, current comparative evidence is insufficient to confirm its superiority over observation.

Figure 5 Meta-analysis of recurrence risk: Salvage treatment vs observation. The forest plot compares the recurrence risk between salvage treatment and observation based on nine studies reporting events in both groups. The pooled odds ratio was 0.89 (95% confidence interval: 0.40-2.02), with no statistical significance and negligible heterogeneity (I² = 0%), suggesting inconclusive comparative effectiveness. OR: Odds ratio; CI: Confidence interval.

Risk of bias across studies

All the included studies were observational and exhibited varying levels of methodological rigor. According to the ROBINS-I assessment, most studies were judged to have a moderate to serious overall risk of bias, primarily because of insufficient adjustment for confounders and incomplete outcome data. A detailed breakdown of the risk assessments across domains is shown in Supplementary Figure 1.

Certainty of evidence based on GRADE

The certainty of the evidence for recurrence outcomes was rated low, primarily because of the serious risks of bias and imprecision. Although the point estimate (OR = 0.84) numerically favored salvage treatment over observation, the wide CI (95%CI: 0.42-1.68) and limited number of comparative studies contributed to serious imprecision. The certainty of evidence was rated as very low for the residual lesion rate after salvage treatment, reflecting both a serious risk of bias and high heterogeneity across studies. These assessments are summarized in the GRADE summary of findings table (Table 4).

Table 4 Grading of recommendations assessment, development, and evaluation summary of findings.

Outcome	Number of participants (studies)	Effect (95%CI)	Risk of Bias	Inconsistency	Indirectness	Imprecision	Publication Bias	Certainty of evidence (GRADE)
Residual tumor after salvage treatment	1223 (19 studies)	Approximately 25% (12%-40%)	Serious	Serious	Not serious	Not serious	Not detected	Very low
Recurrence (salvage vs observation)	2022 (9 studies)	OR = 0.84 (0.42-1.68)	Serious	Not serious	Not serious	Serious	Not detected	Low

GRADE: Grading of recommendations assessment, development, and evaluation; OR: Odds ratio; CI: Confidence interval.

DISCUSSION

This systematic review and meta-analysis evaluated whether salvage treatment offers a clinical advantage over observation after incomplete endoscopic resection of rectal NETs. Two principal outcomes were analyzed: (1) The prevalence of residual tumors after salvage intervention; and (2) The risk of recurrence based on the management strategy.

A pooled analysis of 19 studies showed that residual tumors were detected in approximately 25% of the patients who underwent salvage treatment[13,14,19,20,27,29,33,35-46]. This notable prevalence underscores the limitations of initial endoscopic resection, particularly in the presence of positive or indeterminate margins, or lymphovascular invasion. Accordingly, salvage interventions may serve not only as therapeutic strategies but also as diagnostic procedures for identifying and managing residual diseases.

Importantly, the likelihood of incomplete resection appeared to depend significantly on the initial resection technique. A pooled analysis of 17 studies showed that CSP had the highest incomplete resection rate (73.1%), followed by cEMR (29.8%), mEMR (28.4%), and ESD showing the lowest rate (14.7%)[19,20,24,27,28,30,33,34,37,40-42,46-50]. These findings highlight the importance of selecting the appropriate technique for treating suspected rectal NETs. Given the high incomplete resection rate, CSP may be suboptimal in this setting and should be avoided when margin assessment is critical. Where sufficient expertise is available, ESD should be considered the primary resection technique, as it offers the highest likelihood of achieving complete (R0) resection and reducing the need for subsequent salvage procedures.

For recurrence outcomes, crude pooled data from 31 studies showed a statistically significant benefit associated with salvage treatment; recurrence occurred in 0.96% of patients in the salvage group vs 2.96% in the observation group (P = 0.003)[13,14,19-35,37-39,41-43,45-50]. However, a meta-analysis limited to nine studies that included both treatment groups found no statistically significant difference in recurrence risk (OR = 0.89; 95%CI: 0.40-2.02). This discrepancy likely reflects confounding by indication and selection bias, as higher-risk patients those with positive margins, larger tumors, or lymphovascular invasion were more likely to undergo salvage treatment in the included observational studies. Therefore, the crude analysis should be interpreted with caution, as it may overestimate the effect of salvage treatment. In contrast, the comparative meta-analysis, despite its wider CIs, provides a more valid estimate of the treatment effect because it accounts for between-study variance and excludes double-zero studies.

These findings suggest the potential benefit of salvage treatment in reducing recurrence; however, the evidence remains inconclusive. Notably, the recurrence rates were low in both groups, reaffirming the favorable prognosis of small rectal NETs when appropriately managed. Routine salvage treatment may therefore offer limited additional benefits in terms of recurrence prevention and should be considered in the context of individual patient risk and procedural burden.

This study has several limitations. First, all the included studies were observational, mostly retrospective, single-center case series, introducing the potential for selection bias, reporting bias, and unmeasured confounding factors. Second, the quality of evidence was generally limited, with most studies judged to have a moderate to serious risk of bias based on the ROBINS-I assessment. Third, substantial heterogeneity was observed in the residual lesion meta-analysis (I² = 95.1%), likely due to the variability in tumor characteristics, endoscopic techniques, definitions of incomplete resection, and follow-up protocols. Although potential sources of heterogeneity were carefully reviewed, quantitative subgroup analyses (e.g., by tumor size, lymphovascular invasion, or salvage type) were not feasible due to incomplete reporting across studies. Fourth, the exclusion of studies with zero events in both treatment arms may have reduced the precision of recurrence estimates. Finally, the low frequency of recurrence events limited the statistical power and precluded meaningful subgroup analyses to explore prognostic factors such as tumor grade, margin distance, and lymphovascular invasion.

CONCLUSION

Given the relatively high rate of residual tumors and low overall recurrence rate, salvage treatment after incomplete resection of rectal NETs may be justified as both a diagnostic and therapeutic intervention. However, based on current comparative evidence, its efficacy in reducing recurrence remains uncertain. Therefore, clinical decisions should be individualized, taking into consideration the tumor characteristics, resection margin status, and patient preferences. As the certainty of evidence is low, these conclusions should be interpreted with caution. To better inform treatment strategies and clinical guidelines, high-quality prospective multicenter studies with standardized definitions and consistent outcome reporting are urgently required.