Long-term lumen-apposing metal stents in endoscopic ultrasound-guided gallbladder drainage: Late adverse events and stent removal

Abstract

Endoscopic ultrasound-guided gallbladder drainage with a lumen-apposing metal stent (LAMS) is an accepted alternative to percutaneous drainage for acute cholecystitis in patients unfit for surgery. Short-term technical and clinical success exceeds 95% in expert series, although rates in nationwide data are lower. The optimal subsequent management of the stent remains unresolved: It may be removed, exchanged for a plastic stent, or left in place indefinitely. This minireview summarizes the PubMed-indexed evidence on long-term indwelling LAMSs, focusing on late adverse events and removal. We group late events—food impaction, occlusion, buried LAMS syndrome, tissue overgrowth, bleeding, migration, and recurrent cholecystitis—by mechanism rather than name, and treat the three management strategies as a competing risks problem in which death is a competing event, not simply a censoring event. Removal eliminates stent-specific late events but introduces the risk of recurrence, whereas the apparently benign course of permanent stents partly reflects the reliance on crude proportions from cohorts whose follow-up is truncated by death. We propose deciding the stent’s fate at the index procedure, according to whether the patient is temporarily unfit for surgery or permanently inoperable. Prospective registries reporting cumulative incidence and a randomized comparison of removal vs retention are needed before this decision can become evidence-based.

Key Words: Acute cholecystitis; Endoscopic ultrasound; Gallbladder drainage; Lumen-apposing metal stent; Stent removal; Late adverse events; Competing risks

Core Tip: Lumen-apposing metal stents were designed for short-term drainage, yet after endoscopic ultrasound-guided gallbladder drainage, many remain in place for years. We reframe the choice to remove, exchange, or retain as a competing risks decision. Removal prevents late stent-specific events but risks recurrent cholecystitis, whereas the apparent safety of permanent stents partly reflects the truncated follow-up of a high-mortality population. We argue for deciding the fate of the stent at the index procedure based on the clinical trajectory of the patient, reserving permanent placement for those who will never become surgical candidates, and we outline the prospective studies needed.

INTRODUCTION

Acute cholecystitis (AC) is one of the most common reasons for emergency abdominal surgery, and early laparoscopic cholecystectomy remains the standard of care for patients fit for surgery[1,2]. A substantial minority of patients, however, are elderly, frail, or so burdened by comorbidities that the perioperative risk is prohibitive. The CHOCOLATE trial showed that laparoscopic cholecystectomy outperformed percutaneous catheter drainage in high-risk patients, defined by an Acute Physiology and Chronic Health Evaluation II score of 7 or higher, and its authors rightly caution against routine drainage in anyone who can still tolerate surgery. The trial, however, enrolled patients who remained operative candidates; therefore, it says nothing about the smaller, more extreme group for whom surgery is simply not feasible[3]. For this specific group—excluded from the CHOCOLATE trial—drainage remains the only realistic option, and the Tokyo Guidelines 2018 incorporated it into the treatment algorithm alongside the Charlson Comorbidity Index and American Society of Anesthesiologists physical status[1,4].

Within this drainage-dependent population, endoscopic ultrasound-guided gallbladder drainage (EUS-GBD) has moved in little more than a decade from an experimental maneuver to a procedure that several societies now recommend over percutaneous transhepatic gallbladder drainage (PTGBD). The technique became practical with the lumen-apposing metal stent (LAMS), first described by Itoi, Binmoeller, and colleagues for pancreatic pseudocyst and gallbladder drainage[5]. Its fistula-forming design was subsequently applied to transenteric gallbladder drainage[6], and the procedure was simplified by the electrocautery-enhanced delivery system, which allows puncture and deployment in a single step[7]. Both the European Society of Gastrointestinal Endoscopy guidelines and the American Gastroenterological Association (AGA) Clinical Practice Update present EUS-GBD as a practical alternative to PTGBD where such expertise is available, citing lower adverse event (AE) and reintervention rates in short-term comparisons[8,9]. Three caveats deserve emphasis. These recommendations rest mainly on short-term endpoints; the AGA document is a commentary-style practice update rather than a formal guideline; and the guidance covering the indwelling phase is explicitly conditional and of low certainty[8,9].

The short-term data are robust. However, an evidence-based consensus on the subsequent management of the stent is lacking, which is the focus of this review. LAMSs were designed for temporary lumen apposition and tract formation. After EUS-GBD, they are increasingly left in place for years, in patients who—by the very logic of their selection—may never become operative candidates. Whether an indwelling LAMS should be removed, exchanged, or left in situ is therefore not a procedural footnote but a decision node that shapes late morbidity, surveillance burden, and cost. Recent comprehensive reviews and a multidisciplinary consensus have appraised EUS-GBD broadly, covering indications, technique, equipment, and postprocedural care[10,11]. The i-EUS consensus issued recommendations on stent choice and postprocedural strategy, but classified almost all of its statements as conditional and acknowledged that indwelling stent management rests on low-quality evidence[11].

Three uncertainties dominate the management of the indwelling stent, and this review is organized around them. First, no randomized trial has compared elective removal with permanent retention, so the choice is currently governed by local practice. Second, the cohorts that report reassuring late-event rates are drawn from patients whose median age often exceeds 80 years, so death frequently ends follow-up before a late event can occur. Death here is a competing event rather than a simple censoring event, and the direction in which it distorts a published estimate depends on how that estimate was constructed; for the crude proportions that dominate this literature, the distortion skews toward optimism. Third, no consensus exists on whether the stent should be removed, exchanged, or retained, nor is there an agreed-upon definition of “long-term”, because the follow-up windows of the published series range from a few months to several years. We address these three problems in turn. We reframe the choice to remove, exchange, or retain as a competing-risks decision in which death is modeled explicitly rather than treated as a censoring event. Furthermore, we scrutinize the degree to which the reassuring late-event data are confounded by competing mortality. Finally, we propose a trajectory-based framework that fixes the fate of the stent at the index procedure rather than by default.

HOW THE EVIDENCE WAS IDENTIFIED

This is a narrative minireview, not a systematic review. We searched PubMed from inception to June 2026 using combinations of the terms “gallbladder drainage”, “EUS-guided”, “lumen-apposing metal stent”, “acute cholecystitis”, “stent removal”, “long-term”, “buried stent”, and “cholecystoscopy”, and we hand-searched the reference lists of the retrieved guidelines, consensus documents, and reviews. We gave priority to randomized trials, prospective series, multicenter cohorts, and systematic reviews reporting outcomes beyond 30 days; single case reports were used only to illustrate mechanisms or rescue techniques. Because most included studies are retrospective and use different definitions of an AE, we report numerators and denominators wherever the source provides them and label pooled estimates as such. Where a figure derives from a single study, we state this explicitly, so that readers do not mistake it for a pooled rate. The grading of the evidence supporting each management strategy in Table 1 was assigned by the authors using an approach adapted from the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system. We based this assessment on study design, the number and size of the contributing studies, the directness of the population and intervention, the presence of a comparator, and the risk-of-bias arising from competing mortality. Notably, no formal GRADE assessment was performed.

Table 1 Competing strategies for managing the indwelling stent.

Dimension	Strategy A: Elective removal after tract maturation	Strategy B: Scheduled exchange for long-term plastic stent	Strategy C: Permanent indwelling LAMS
Target patient profile	Temporarily unfit for surgery; possible future cholecystectomy	Intermediate; needs ongoing drainage but a metal stent is undesirable long term	Definitively inoperable (never a surgical candidate)
Principal rationale	Preempts late stent-specific events; often combined with cholecystoscopic stone clearance	Maintains a low-profile conduit through an established tract	Avoids repeat instrumentation in frail patients
Supporting evidence	Kamata et al[58] (self-expandable metal stent, not LAMS): Removal at about 4 weeks with no migration or impaction; Choi et al[59] and Walter et al[60]: Long-term outcomes after removal or exchange	Commentary recommendation for patients with extended life expectancy[44]; DRAC-1 protocol[27]; no dedicated trial	Yuste et al[43]; recurrence not increased by retention[36]; most patients die with the stent in situ[44]
Principal risk traded	Recurrent cholecystitis if the cystic duct is obstructed and no conduit is left	Requires repeat endoscopy; uncertain durability	An indefinite, albeit low, risk of burial, erosion, and impaction; it may also complicate subsequent surgery[62,63]
Evidence quality1	Low	Very low	Low to moderate
Basis for the grading	Small single-center series, one using a self-expandable metal stent rather than a LAMS; indirectness of the population; no comparator[58,59]	No study reports the strategy as an intervention; entirely inferential from protocol descriptions and expert opinion[27,44]	Two or more retrospective single-center and multicenter cohorts with follow-up beyond 1 year but serious risk-of-bias from competing mortality and short median follow-up[36,43,44]
Outstanding question	Optimal timing of removal is undefined	Whether exchange outperforms retention	Whether the apparent safety of permanence is real or an artifact of short survival

¹Grading was assigned by the authors using an approach adapted from the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) and applied to the body of evidence supporting each strategy, as described in the section “How the evidence was identified”. Low to moderate = two or more cohorts with follow-up beyond 1 year but serious risk-of-bias; Low = single small series or indirect population; Very low = no study reports the strategy as an intervention. No formal GRADE assessment was performed, and no strategy is supported by randomized data. A higher grade reflects the quantity and design of the available evidence and not a stronger recommendation. DRAC-1: Randomized trial of endoscopic ultrasound-guided vs percutaneous gallbladder drainage; LAMS: Lumen-apposing metal stent.

FROM RESCUE PROCEDURE TO DEFINITIVE THERAPY

The evolution of EUS-GBD into its present role is highly relevant to the debate on stent removal. Endoscopic gallbladder drainage began as a transpapillary technique, in which a stent is advanced through the cystic duct; this route still has a place, with efficacy comparable to that of the percutaneous approach in selected patients[12-14]. Transmural access through the gastrointestinal wall followed as endoscopic ultrasound (EUS) matured, achieving technical success comparable to the percutaneous route with fewer AEs[12,15]. Comparative studies showed that EUS-guided transmural drainage and PTGBD were equally effective in resolving cholecystitis, but that the internal route avoided the need for an external drain, along with the associated discomfort and risk of dislodgment[12,16]. Pooled analyses of the expanding case series then suggested that transmural stenting was not merely equivalent but better, reducing reintervention and unplanned readmission[17,18]. A technique adopted as a rescue measure has, in other words, become a definitive treatment—and the stent has come with it.

DEVICE DESIGN AND THE GALLBLADDER ENVIRONMENT

The design of a LAMS largely determines how it behaves once left in place. The dumbbell geometry, with two retention flanges and a short saddle, generates the apposing force that holds the gastrointestinal lumen against the gallbladder wall. An ex vivo study quantified how this force varies between EUS-specific stents and suggested that it may relate both to secure tract formation and to the potential for wall trauma[19]. The original lumen-apposing concept and its first fistula-forming versions were designed to overcome the shortcomings of tubular plastic and metal stents in nonadherent transluminal drainage[5,6,20].

The same wide channel that secures apposition and allows the passage of gallstones or an endoscope predisposes to occlusion by food and debris—a failure mode largely absent when sterile collections are drained. The gallbladder is in any case a hostile environment for a permanent prosthesis. Unlike a walled-off pancreatic collection, which resolves once drained, the gallbladder remains a contractile organ that produces mucin and lithogenic bile and often still contains stones. These features sustain a continuing substrate for sludge, stone formation, and inflammation around the stent. Tract maturation is central to the removal debate. It is conventionally assumed to be largely complete by approximately 4 weeks, although this interval rests on customary practice rather than on any formal validation study. The fibrosis that makes the stent safely removable is also the process by which it becomes progressively epithelialized and buried.

What comparative clinical data support the intuition that a larger or stiffer device carries a higher risk of stent-related complications? No trial has randomized patients to different LAMS diameters, and the available evidence is indirect. Three strands are relevant. First, in the Italian nationwide cohort of 116 patients, univariable analysis found technical success less likely when a LAMS narrower than 10 mm was used, although this did not survive multivariable analysis; the authors argue that a gallbladder distended by AC is at least 40 mm across and therefore accommodates a wider stent, so diameters of 10 mm or more are conventional in EUS-GBD[21]. Second, an indirect meta-analysis of 18 observational studies compared the two most widely used devices. For gallbladder drainage, pooled technical and clinical success rates were 96.2% and 92.7% for the Axios stent (Boston Scientific, Marlborough, MA, United States; 11 studies, 433 patients) and 95.9% and 94.2% for the Spaxus stent (Taewoong Medical, Gimpo, South Korea; 7 studies, 242 patients), but total AE rates were 23.6% vs 9.5% and bleeding rates 4.8% vs 1.8%[22]. The comparison is indirect and the confidence intervals overlap, so it cannot establish superiority; it does show that AE rates differ by device even when success rates do not. Third, and most directly relevant to the issue of stent removal, a prospective international trial of a self-approximating LAMS with a deliberately lower apposing force reported that buried stent syndrome and bleeding were rare over long-term follow-up in 27 EUS-GBD and 26 choledochoduodenostomy patients[23]. A separate comparison of LAMSs against antimigration tubular self-expandable metal stents found no difference in effectiveness or safety in high-risk patients[24]. Taken together, these data support the mechanistic argument—apposing force and channel caliber plausibly drive mural and luminal late events—while falling well short of a diameter-specific recommendation. This gap in the literature warrants highlighting.

PATIENT SELECTION

Every subsequent management decision rests on appropriate patient selection. The Tokyo Guidelines 2018 diagnostic criteria and severity grading provide a framework for identifying candidates for drainage, and the accompanying flowchart positions drainage as the option for those unfit for early cholecystectomy[4,25]. Series examining predictors of mortality after EUS-GBD found that outcomes are driven more by the underlying condition than by the procedure, serving as a reminder that the technique is applied to a population with very limited physiological reserve[26].

SHORT-TERM EFFICACY IS SETTLED; DURABILITY IS NOT

The short-term efficacy of EUS-GBD is now well established (Table 2). The DRAC-1 trial, the first international randomized comparison against PTGBD in very-high-risk patients, showed that EUS-GBD reduced 1-year AEs (25.6% vs 77.5%), as well as 30-day AEs, reinterventions, unplanned readmissions, and recurrent cholecystitis, with comparable technical and clinical success[27]. A matched cohort had already shown fewer AEs than the percutaneous route[28], and a comparison against transpapillary cholecystostomy favored the transmural approach[29]. Successive systematic reviews and meta-analyses, including network analyses across all three modalities, largely agree that LAMS-based drainage offers a favorable safety and readmission profile[30-33]. A meta-analysis of proportions with meta-regression of 27 studies and more than 1000 patients reported pooled technical success of approximately 98% and clinical success above 95%[34].

Table 2 Representative studies and pooled analyses of endoscopic ultrasound-guided gallbladder drainage with lumen-apposing metal stents.

Ref.	Design and patients	Technical success	Clinical success	Follow-up	Key long-term findings
Teoh et al[27], 2020	Randomized trial vs PTGBD; 80 patients recruited (39 EUS-GBD, 41 PTGBD); 39 vs 40 analyzed	97.4%	92.3%	1 year	1-year adverse event rates were 10/39 (25.6%) vs 31/40 (77.5%)b for PTGBD; recurrent cholecystitis 1/39 (2.6%) vs 8/40 (20%); reintervention beyond 30 days 1/39 (2.6%) vs 12/40 (30%)
Binda et al[21], 2024	Multicenter cohort; 116 patients; mean age 82.7 years	94%	87.1%	Mean 309 days	Adverse events in 12/116 (10.3%); 8 intraprocedural, 1 early, 3 late; no fatal adverse events; no recurrence among 40 patients followed beyond 1 year
Dollhopf et al[35], 2017	Multicenter cohort; 75 patients; electrocautery-enhanced LAMS	98.7%	95.9%	Not reported as a single value	Overall adverse events 8/75 (10.7%); late events in 6 patients: Recurrent cholecystitis (3), migration (2), Bouveret syndrome (1); all managed without surgery
David et al[36], 2025	Multicenter cohort; 109 patients; 18 United States centers	99.1% (108/109)	Initial 97.2% (106/109); long-term 89.9% (98/109)	Median 140 days (IQR 76-330; range 30-1188)	Events beyond 30 days in 10/109 (9.2%), of which 7/109 (6.4%) were LAMS-specific; LAMSs removed in 24/109 (22%); recurrent cholecystitis 1/24 (4.2%) after removal vs 7/85 (8.2%) after retention, with no statistically significant difference; the decline from initial to long-term clinical success and the short median follow-up limit long-term inferences
Yuste et al[43], 2019	Single-center series; 22 of 47 patients with dwell time > 1 year	NR1	NR1	Median 24.4 months (IQR 18.2-42.4)	No stent-related events beyond 12 months; 1 patient (4.5%) readmitted for gallstone-related disease; one completely buried stent at 61.5 months with patent fistula
Martinez-Moreno et al[44], 2023	Single-center registry; 50 patients analyzed; 3-year design	91.2% (62/68)2	98% (49/50)	Median 25.0 months	Adverse events in 18%, 20%, and 26% of patients during years 1-3; 14 LAMS-related events in 11 patients (22%) at a median of 674 days; only 5/14 (35.7%) symptomatic, and late events increasingly asymptomatic; all 7 migrations (14%) asymptomatic; symptomatic events accounted for 66.7% of events associated with a transgastric stent vs 12.5% of those from a transduodenal stenta; recurrent cholecystitis 2/50 (4%); no stent-related bleeding or stent-related death
Fabbri et al[34], 2022	Proportion meta-analysis; 27 studies; 1004 patients	98.0% (95%CI: 96.3-99.3)	95.4% (95%CI: 92.8-97.5)	Varied across studies	Pooled procedure-related adverse event rate was 14.8%; meta-regression identified institutional volume and stent type as outcome modifiers
Canakis et al[45], 2026	Meta-analysis restricted to follow-up ≥ 1 year; 18 studies; 701 patients	95.8% (95%CI: 93.9-97.2)	94.3% (95%CI: 92-96)	≥ 1 year by design	Recurrent cholecystitis rate at or beyond 1 year was 4.2%; 1-year readmission rate was 19%; reintervention rate 6%; repeat endoscopy due to occlusion 2.9%; mean patency 418.8 days (I2 = 98%); approximately half the patients did not receive a LAMS

^aP < 0.05, transgastric vs transduodenal stent position.

^bP < 0.01, endoscopic ultrasound-guided gallbladder drainage vs percutaneous transhepatic gallbladder drainage.

¹The series was restricted to the 22 of 47 patients whose stent remained in situ for more than 12 months; procedural success was therefore not an endpoint and is not reported.

²Denominator is the 68 patients assessed for gallbladder drainage; of the 62 technically successful procedures, 12 were performed for indications other than acute cholecystitis and were excluded, leaving 50 patients for analysis.

EUS-GBD: Endoscopic ultrasound-guided gallbladder drainage; IQR: Interquartile range; LAMS: Lumen-apposing metal stent; NR: Not reported; PTGBD: Percutaneous transhepatic gallbladder drainage.

Real-world data temper these figures. The Italian nationwide study of 116 patients (mean age 82.7 years) reported technical and clinical success of 94% and 87.1%, below the rates of selected high-volume series, which suggests that multi-operator practice is associated with lower success rates[21]. Multicenter series using electrocautery-enhanced LAMSs illustrate the upper bound achievable in expert hands: 98.7% and 95.9% in 75 patients[35], and 99.1% and 97.2% in a United States registry of 109 patients[36]. Pooled and three-way comparisons against transpapillary and percutaneous drainage, together with further meta-analyses and early international registries, reach concordant conclusions[37-41]. Outcomes achieved by early-career endoscopists suggest the technique is becoming more broadly reproducible, which will steadily increase the number of indwelling stents requiring a long-term decision[42].

The methodological trap common to almost all of these studies is that “success” conflates the success of the index procedure with the long-term outcome. The largest United States multicenter cohort makes the distance between the two explicit: Initial clinical success was 97.2% (106/109), but long-term clinical success was 89.9% (98/109), and even that lower figure was obtained over a median follow-up of only 140 days[36]. A 7-percentage-point drop in clinical success within a few months of observation is not a negligible decline, and it is the part of the outcome that the overall rates conceal. Durable stent-dependent drainage, and the hazards that accompany it, remain comparatively uncharacterized.

WHAT COUNTS AS LONG TERM?

One obstacle to synthesizing the late-event literature is that “long term” has no agreed-upon meaning. Follow-up in the studies cited above ranges from a median of 140 days in the largest United States multicenter cohort[36] to a mean of 309 days in the Italian registry[21], a median of 24.4 months in a series restricted to dwell times beyond 1 year[43], and 3 years by design in a Spanish registry, in which the observed median follow-up was 25.0 months[44]. Some studies report a median, others a mean, and few report dwell time as a continuous variable at all. AE windows differ in the same way: “late” means beyond 15 days in one cohort[21], beyond 30 days in another[36], and beyond 12 months in a third[43]. Because late events accrue with time, these differences alone can generate a several-fold variation in reported rates without any true difference in device performance.

A recent systematic review and meta-analysis restricted to studies with at least 1 year of follow-up illustrates both the value and the limits of pooling. Across 18 studies and 701 patients, it found technical and clinical success of 95.8% and 94.3%, a 4.2% rate of recurrent cholecystitis at or beyond 1 year, a 1-year readmission rate of 19%, a reintervention rate of 6%, a 2.9% rate of repeat endoscopy due to stent obstruction, and mean stent patency of approximately 419 days[45]. The I² heterogeneity statistic for stent patency was 98%, and approximately half of the pooled patients did not receive a LAMS at all, so the patency figure in particular should not be read as a property of the device. For the remainder of this review, we therefore use explicit definitions: An event is early if it occurs within 30 days, late if it occurs thereafter, and we reserve “long-term dwell” for stents left in situ for 12 months or more. We recommend that future studies adopt the same convention, report dwell time continuously, and present late events as cumulative incidence at prespecified landmarks with death modeled as a competing event, rather than as crude proportions. We deliberately do not recommend rates per person-year, which are often proposed for this purpose but which assume a constant hazard and cannot be reconciled with the early clustering of events described below.

SPECTRUM OF LATE ADVERSE EVENTS

The range of late complications—food impaction, occlusion, buried LAMS syndrome, tissue overgrowth, delayed bleeding, migration, and recurrent cholecystitis—is easier to conceptualize when organized by mechanism than by name. Luminal events reflect the patency of a wide channel exposed to bile and enteric contents; mural events reflect the chronic interaction between flange and wall; and mechanical events reflect loss of apposition. A dedicated systematic review and meta-analysis of LAMS-specific AEs in gallbladder drainage helps quantify the overall burden, and broader analyses of AEs in EUS-guided transluminal drainage provide context for mural and bleeding complications[46,47]. Consistent figures across studies will require standardized reporting, whether through the American Society for Gastrointestinal Endoscopy (ASGE) lexicon[48] or, more recently, the Adverse Events in Gastrointestinal Endoscopy (AGREE) classification[49].

The evidence regarding late AEs are reassuring in some aspects but ambiguous in others. Several cohorts show that the hazard of symptomatic stent-specific AEs is highest shortly after placement. In a single-center series of 22 patients with LAMS dwell times beyond 1 year, no late stent-related AEs occurred after the first 12 months, and only one patient (4.5%) was readmitted for gallstone-related disease[43]. A 3-year follow-up registry of 50 patients recorded 14 LAMS-related events in 11 patients (22%) at a median of 674 days after placement, with AEs affecting 18%, 20%, and 26% of patients during years 1, 2, and 3, respectively. Only 5 of the 14 events (35.7%) were symptomatic, and the authors concluded that late events became progressively more likely to be asymptomatic[44]. A United States multicenter study similarly found that most long-term AEs were LAMS-specific but affected only 6.4% of patients beyond 30 days, with no significant difference in recurrent cholecystitis according to whether the stent was removed[36].

This latter finding warrants cautious interpretation. Median follow-up in that cohort was 140 days, so the absence of an observed difference between removal and retention reflects a short observation window rather than demonstrated long-term equivalence[36]. In the 75-patient electrocautery-enhanced series, late events comprised recurrent cholecystitis (three patients), migration (two patients), and Bouveret syndrome (one patient), all managed without surgery[35]. A more recent single-center series using a different electrocautery-enhanced device reported late AEs in three of 55 patients (5.4%) and recurrent AC in two (3.6%), again in a cohort too small for rare events to surface reliably[50].

Two caveats accompany this relatively benign picture. The first concerns the access route. In the 3-year registry, symptomatic events accounted for 66.7% of the LAMS-related events associated with a transgastric route but for only 12.5% of those arising from a transduodenal route, a difference that the authors reported as statistically significant. Furthermore, all seven migrations (14% of patients) were asymptomatic, and no stent-related bleeding or stent-related death occurred. The excess symptomatic burden of the transgastric route therefore reflected occlusion, burial, and gastric outlet obstruction rather than migration[44]. AEs also tended to recur in patients who had already had one[36,44]. The second caution is that the true incidence is almost certainly underestimated. These cohorts consist of frail older adults with a mean age typically exceeding 80 years[21], and competing mortality often truncates the observation period before a late event can manifest. Because these events are almost always reported as crude proportions, this competing mortality may overstate the apparent long-term safety of the indwelling strategy—a limitation that the more optimistic conclusions of single-center studies, and the short-term multicenter data, seldom acknowledge. The distortion is not, however, uniform in direction, and the next section explains why it matters which estimator a study has used.

Buried lumen-apposing metal stent syndrome

Given its direct implications for stent removability, buried LAMS syndrome deserves separate treatment. The enteric-side (proximal) flange becomes progressively covered by hyperplastic tissue and epithelium until it is embedded in, or has passed through, the enteric wall. No cohort has reported its incidence as a primary endpoint, and the published estimates are fragmentary. In the series restricted to dwell times beyond 1 year, imaging identified a completely buried cholecystogastric stent 61.5 months after deployment, with the fistula still patent[43]. The prospective international trial of a lower-force self-approximating LAMS was designed partly to address this problem and reported a low rate of buried stent syndrome over long-term follow-up[23]. Registry data show that tissue overgrowth continues to accumulate after the first year even when it causes no symptoms[44].

The plausible risk factors are derived from the mechanism rather than a formal multivariable analysis: Long dwell time, high radial and apposing force, a transgastric route (where gastric wall thickness and peristalsis favor stent burial), and the tissue hyperplasia that follows electrocautery-assisted deployment. Prevention rests on the same logic. Several strategies reduce the opportunity for burial, although none has been tested against burial as an endpoint: Deciding at the index procedure whether the stent is temporary, planning removal or conversion once the tract has matured in patients for whom that is appropriate, choosing a device with a lower apposing force when prolonged dwell is anticipated, and placing a coaxial double-pigtail stent. When burial has already occurred, removal is no longer routine: Extraction can precipitate bleeding or perforation, and the reported rescue strategies are endoscopic. Deploying a second LAMS through the buried one (the LAMS-in-LAMS technique) re-establishes drainage and allows the buried device to be retrieved[51], and placement of a double-pigtail stent has been used to control bleeding from a buried cholecystogastric LAMS[52]. These are single-case techniques, and their availability is an argument for anticipating burial rather than for tolerating it.

DEATH AS A COMPETING RISK, NOT AS INFORMATIVE CENSORING

The assertion that competing mortality biases the results in favor of the indwelling strategy is made loosely in most of this literature, including in the sections above. It is worth stating this precisely, because the direction of the distortion depends on the estimator a study has used.

Three quantities need to be distinguished: (1) The cause-specific hazard is the instantaneous rate of a stent-related event among patients still alive and still carrying a stent; it is the quantity that characterizes the device; (2) The cumulative incidence function—the probability that a patient experiences a stent-related event before dying of some other cause—is the quantity that characterizes the patient’s prognosis, and it is what a clinician implicitly weighs when deciding whether to leave a stent in situ; and (3) The crude proportion reported by almost every published series is neither of these, but an estimate of the cumulative incidence truncated at the end of an administratively short period of observation.

The consequences are not symmetrical. Where a series reports a crude proportion, patients who die early contribute little observation time, few events are counted, and the figure underestimates the risk that the same device would generate in a longer-lived patient; in that sense, the literature presents an overly optimistic view of the device, and this is the bias to which most commentaries, including ours, refer. When a Kaplan-Meier estimator is applied with death treated as a censoring event, the complement of the Kaplan-Meier (KM) survival estimate (1 − KM) overestimates the cumulative incidence of stent-related events—a bias that arises from the competing-risk structure itself, and is further aggravated when the censoring is informative. It is therefore not correct to say that competing mortality always biases late-event rates downward. What is true is that the published estimates are of uncertain interpretation because the estimand is almost never stated, and that the short observed follow-up of even the largest cohorts—a median of 140 days in the largest multicenter series[36]—leaves the tail of the distribution unobserved.

Recognizing this sharpens the clinical argument, but it also has implications that are easily overlooked. If death is modeled explicitly as a competing event, then in a patient whose expected survival is short, the cumulative incidence of burial, erosion, and impaction is genuinely low, and permanent retention is not merely defensible but rational. The same calculation reverses as expected survival lengthens: The longer a patient is likely to live, the greater the proportion of the cause-specific hazard that is converted into observed events, and the more the balance shifts toward removal or conversion. The competing-risks framework is thus not an argument against permanent retention as such. It is the statistical expression of the trajectory-based framework proposed below, and it identifies realistic life expectancy—rather than any fixed threshold of dwell time—as the variable on which the decision should turn.

Two practical consequences follow for study design. First, late events should be reported as cumulative incidence at prespecified landmarks (e.g., at 6 months, 12 months, and 24 months), estimated with the Aalen-Johansen or Fine-Gray approach and with death from any cause modeled as a competing event, alongside the cause-specific hazard wherever device performance rather than patient prognosis is the question. Second, and for the same reason, the crude proportions collected in Table 3 and elsewhere in this review should be read as descriptions of the cohorts that generated them rather than as risks transferable to an individual patient.

Table 3 Late adverse events of indwelling lumen-apposing metal stents organized by mechanism.

Mechanism	Affected domain	Specific events	Typical timing	Evidence and source
Luminal	Channel patency and drainage	Food impaction; sludge- or stone-related occlusion; recurrent cholecystitis	Accumulates with longer dwell time	Drives recurrent cholecystitis and provides the rationale for coaxial pigtail placement and proactive cholecystoscopy[9,55]; recurrent cholecystitis in 3 of 75 patients[35], in 2 of 55 patients[50] and in 2 of 50 patients[44]; no significant difference by removal status[36]; pooled recurrence beyond 1 year 4.2%[45]
Mural	Stent-wall interaction	Tissue overgrowth; buried LAMS syndrome; delayed bleeding from flange erosion	More apparent with longer dwell time	Mechanistically expected and informed by a dedicated adverse-event meta-analysis[46]; no cohort reports incidence as a primary endpoint; a completely buried stent was documented at 61.5 months[43]; a device with a lower apposing force showed a low rate of buried stent syndrome[23]
Mechanical	Loss of apposition	Migration into the gallbladder lumen or distal migration	Early or late	Migration in 2 of 75 patients[35] and in 7 of 50 patients (14%) in a 3-year registry, in which every migration was asymptomatic[44]; frequently asymptomatic when the stent migrates internally

Rates given are single-study figures unless labeled as pooled. Symptomatic late events cluster in the first year, whereas asymptomatic tissue overgrowth and migration continue to accumulate thereafter[43,44]. Every value in this table is a crude proportion derived from a cohort with high competing mortality and limited observation, and therefore describes that cohort rather than a risk transferable to an individual patient. LAMS: Lumen-apposing metal stent.

COAXIAL DOUBLE-PIGTAIL STENTS: MITIGATION OR CONFOUNDER?

Placing a double-pigtail plastic stent coaxially through the LAMS is a widely adopted refinement, intended to anchor the device, keep the channel patent, and prevent food impaction. In our own practice, one or two 7 Fr pigtail stents, 4 to 7 cm long, are advanced over a guidewire through the LAMS at the index procedure or at a subsequent session, with the internal end coiled inside the gallbladder and the external end in the duodenum or stomach; published series rarely specify caliber, length, or number, so these details reflect our practice rather than an established standard. The rationale is mechanistically sound and the practice is intuitively attractive. It was the protocol followed in DRAC-1, in which the metal stent was later exchanged for double-pigtail stents to maintain the fistula[27]. It is also the maneuver most often used to salvage a stent that has occluded or that has begun to bleed[50].

The evidence base is nevertheless thin and largely indirect. No randomized trial in EUS-GBD has compared LAMS alone with LAMS plus coaxial pigtail, and the retrospective series that report the practice do not apply it uniformly or describe it consistently. Its adoption tends to track operator experience and procedural caution, so retrospective data cannot separate the effect of the plastic stent from the effect of the endoscopists who favor it. The strategy also changes the distribution of failure modes rather than abolishing them: It plausibly reduces impaction and migration while doing little to prevent tissue overgrowth or burial, which complicates any pooled comparison of outcomes of LAMS alone. In practice, the maneuver is cheap, quick, and low-risk, and we continue to use it; but it should be recorded prospectively as a covariate rather than treated as an established preventive measure.

PERORAL CHOLECYSTOSCOPY AND THE STONE-FREE GALLBLADDER

A matured fistula allows peroral cholecystoscopy, through which residual stones can be cleared by basket extraction, mechanical lithotripsy, or laser and electrohydraulic lithotripsy—a distinctive advantage of large-bore LAMSs[53-55]. The principal late AEs associated with indwelling LAMSs and their underlying mechanisms are summarized in Table 3.

Proponents argue that a stone-free gallbladder removes the substrate for recurrent cholecystitis and therefore justifies subsequent stent removal, with the matured tract or a double-pigtail stent preserving drainage. The figures frequently cited to support this argument derive from a single retrospective series of 25 patients in whom 29 cholecystoscopies were attempted and 27 were successful (93.1%): 14 patients (56%) had cleared their stones spontaneously, residual stones were found in 11 and cleared in 8, and overall stone clearance reached 88% after a mean of 1.25 sessions[55]. Those data, rather than any independent dataset, underlie the protocol of routine cholecystoscopy at 4 weeks to 6 weeks set out in the AGA Clinical Practice Update and an accompanying technical review. According to this protocol, the LAMS could be removed once clearance and cystic duct patency were confirmed[9,56]. The same approach has been used to maintain chemotherapy in oncological patients, in whom recurrent biliary events would otherwise interrupt treatment[57]. A protocol resting on 25 patients from one center is a reasonable basis for practice but a weak one for a general recommendation, and whether proactive clearance genuinely reduces late events, as opposed to simply shifting when the intervention occurs, has never been tested against a retention strategy in a controlled fashion.

CENTRAL CONTROVERSY: REMOVE, EXCHANGE, OR LEAVE IN SITU?

Here, the literature openly diverges, and three competing strategies can be discerned, each implicitly assuming a different patient (Table 1). The first favors elective removal once the tract has matured. Its clearest support comes from a series in which a self-expandable metal stent—not a LAMS—was removed at about 4 weeks after clinical improvement. In that study, there was no migration or food impaction, and recurrence was confined to one patient left without a residual conduit[58]. Earlier work reporting long-term outcomes after transmural metal stent removal or exchange also supports this approach[59,60]. The second favors scheduled exchange for a long-term double-pigtail plastic stent within the fistula, converting an impaction- and erosion-prone device into a lower-profile conduit; commentators recommend it particularly for patients with a longer life expectancy[44]. The third accepts a permanent indwelling LAMS, pointing to series in which no stent-related events emerged beyond the first year and to multicenter data showing no difference in recurrent cholecystitis between removal and retention[36,43]. A frequently cited observation is that most patients die of their underlying disease with the stent in place, which proponents interpret as evidence that the device can serve as definitive therapy[44].

The core of the controversy is that these strategies involve a trade-off between different categories of risk. Removal eliminates late LAMS-specific events but reintroduces the risk of recurrent cholecystitis if the cystic duct is obstructed and no alternative drainage is secured. Leaving the stent in situ avoids repeat instrumentation but exposes the patient to an indefinite, if low, hazard of burial, erosion, and impaction. This is the competing-risks problem set out above, and the current evidence base—retrospective cohorts with heterogeneous removal timing and confounding by indication—cannot adjudicate it. Current consensus recommendations emphasize that the choice between LAMS removal, exchange, and permanent retention should be individualized because evidence on long-term stent management remains limited[11].

INDWELLING STENT AND SUBSEQUENT SURGERY

One consideration separates patients who are temporarily unfit for surgery from those who are permanently inoperable more sharply than any other: The effect of an indwelling transmural stent on a later cholecystectomy. EUS-GBD has been studied as a bridge to surgery, and cholecystectomy is achievable after drainage in appropriately selected patients[11,61]. A transmural fistula, however, complicates the subsequent operation, particularly the cholecystogastric fistula created by a transgastric stent, and surgeons have drawn attention to the long-term problems that the surgical approach inherits from an indwelling LAMS[62]. Whether cholecystectomy after EUS-GBD is beneficial remains unsettled[63]. The observation that surgical risk is often reversible, with a meaningful minority of patients improving enough to undergo cholecystectomy after drainage[62], argues directly against the reflexive use of a permanent stent in anyone whose inoperability might prove temporary. Where surgery is contemplated, removal of the LAMS before the operation with endoscopic closure of the fistula has been recommended, especially for the transgastric route[62]. Even the timing of cholecystectomy after percutaneous drainage affects morbidity, which underlines that drainage is rarely a neutral act with respect to later surgery[64].

TOWARD A PATIENT-STRATIFIED FRAMEWORK

Progress likely depends more on stratifying patients by clinical trajectory than on a universal rule or a static high-risk label (Figure 1). Such a framework separates patients who are temporarily unfit for surgery, in whom the stent should be treated as temporary and removal or conversion is planned, from permanently inoperable patients, in whom retention with surveillance is defensible. The decision variables that should outweigh any fixed dwell-time threshold are cystic duct patency, residual stone burden, stent location, and realistic life expectancy. This is the clinical expression of the competing-risks argument developed above: Life expectancy determines how much of the cause-specific hazard of stent-related events a patient will live long enough to experience, and it is therefore the variable that should drive the decision. Permanent retention is reasonable for patients who are definitively inoperable due to malignancy, advanced cirrhosis, or irreversible cardiopulmonary disease, whereas a permanent transgastric stent may preclude a future minimally invasive option in patients whose comorbidities could still improve[28,62].

Figure 1 Trajectory-based framework for managing the indwelling lumen-apposing metal stent after endoscopic ultrasound-guided gallbladder drainage. After clinical success, the clinical trajectory is reassessed at tract maturation. Patients temporarily unfit for surgery are directed toward stone clearance and planned removal or conversion, whereas permanently inoperable patients are candidates for retention with surveillance. Three strategies are shown alongside their respective risk trade-offs: A: Elective removal; B: Scheduled exchange for a long-term plastic stent; C: Permanent indwelling stent. The lower panel organizes late adverse events by mechanism (luminal, mural, or mechanical), framed against death from the underlying condition as the competing risk, which determines how much of the cause-specific hazard a patient survives long enough to experience. EUS-GBD: Endoscopic ultrasound-guided gallbladder drainage; LAMS: Lumen-apposing metal stent.

The boundary between temporary and definitive therapy may also be shifting. It is sometimes asserted that noninferiority data now support endoscopic drainage even in surgical candidates. That claim is not supported by current evidence. The most directly relevant study is a propensity-score analysis with 1-year follow-up in which 30 patients treated by EUS-GBD were matched against 30 undergoing laparoscopic cholecystectomy; technical and clinical success, 30-day AEs, mortality, reinterventions, and recurrent biliary events did not differ between groups[65]. The comparison is small and nonrandomized, and the EUS-GBD patients were poor surgical candidates while the surgical patients were fit, so confounding by indication runs in a direction that favors drainage. A recent systematic review and meta-analysis found that interval cholecystectomy after EUS-GBD is technically feasible but remains uncommon and technically challenging[66]. Until more robust comparative evidence becomes available, the extension of EUS-GBD to operable patients should be regarded as a hypothesis rather than as a basis for leaving stents permanently in patients who are surgical candidates.

SURVEILLANCE, REMOVAL, AND THE RISKS OF INTERVENTION

Any active management strategy carries its own iatrogenic cost. No validated surveillance protocol exists, and there is no consensus on how often indwelling stents should be monitored; furthermore, indefinite surveillance imposes a real burden on a frail population[44,62]. Delayed removal is not always straightforward: Stent burial complicates retrieval, and where tissue overgrowth or buried LAMS syndrome has developed, extraction can precipitate bleeding or perforation[10,52]. Ultimately, every option, including watchful waiting, carries inherent risks, and leaving the stent in situ should be understood as a legitimate active choice rather than as a passive default.

LIMITATIONS OF THE EVIDENCE BASE

The deficiencies of the current literature should be stated plainly. This review itself is limited by a single-database, nonsystematic search strategy without formal risk-of-bias assessment, so selective retrieval cannot be excluded. Despite the availability of the ASGE lexicon and the newer AGREE classification[48,49], almost all long-term data come from retrospective single-center or registry cohorts subject to selection bias, heterogeneous technique (cautery vs noncautery LAMS, variable coaxial stenting, inconsistent stone clearance), and inconsistent AE definitions. The high competing mortality of the population truncates observation; because almost every series reports crude proportions, this introduces an optimistic bias into their safety estimates. Conversely, the few analyses that apply survival methods with death treated as a censoring event err in the opposite direction; in neither case is the estimand stated. The short median follow-up of even the largest multicenter cohort compounds the problem[36]. Dwell time is not reported uniformly as a continuous variable, no severity grading specific to late LAMS events exists, and—most consequentially—no randomized comparison of removal strategies has been performed. On present evidence, the conclusion that a permanent LAMS is safe reflects short observed survival as much as any demonstrated property of the device.

Notably, recent authoritative appraisals converge on this same gap rather than closing it. Broad narrative reviews focus their unanswered questions on patient selection and on extending the indication to surgical candidates, leaving the timing and necessity of removal in established nonoperative patients largely unexamined[10], while the i-EUS consensus could offer only conditional, low-certainty guidance on postprocedural stent management[11]. That the gap persists across independent expert efforts is an argument that the issue of stent removal deserves a dedicated study rather than incidental coverage within general reviews.

FUTURE PERSPECTIVES

Resolving these questions will require more than the customary call for further studies. Three concrete directions are warranted. First, a multicenter prospective registry must be created with mandated long-term follow-up and dwell time recorded as a continuous covariate, serving as the foundation for a randomized trial of elective removal or conversion vs permanent stent retention in definitively inoperable patients. This trial should be powered to detect differences in the cumulative incidence of late stent-specific events, with death from any cause modeled explicitly as a competing event rather than as a censoring event. Second, reporting must be standardized through a consensus lexicon and severity grading for late LAMS-specific AEs, built on the AGREE framework, together with an agreed operational definition of long-term follow-up. Without these, cross-study synthesis will remain unreliable[49]. Third, further device and technique innovations must be pursued—such as antimigration and anti-tissue ingrowth coatings, lower-force and more readily removable designs informed by ex vivo mechanical characterization, and protocolized cholecystoscopic stone clearance—any of which could reframe the issue of stent removal by removing the substrate for late events[19,23,55]. Until such data exist, addressing the issue of stent removal prospectively at the index procedure, individualized to the trajectory of the patient, is the most defensible posture.

CONCLUSION

Long-term retention of a LAMS after EUS-GBD is feasible and, in appropriately selected nonoperative patients, appears to provide durable drainage with an acceptable late-event profile. Feasibility has nevertheless outpaced evidence. The reassurance offered by current cohorts is limited by their retrospective design and by a competing mortality that conceals as much as it reveals, and the absence of any randomized comparison leaves the question of whether to remove, exchange, or retain genuinely unresolved. The practical implication for clinicians is to avoid resorting to permanent retention as a default strategy, reserving it exclusively for patients who will never become surgical candidates, and to decide the fate of the stent when it is placed rather than by omission. Because the technology has been adopted more rapidly than its long-term evidence base has developed, management of the indwelling stent—rather than the drainage procedure itself—is likely to be the defining clinical and research challenge of the coming years.

ACKNOWLEDGEMENTS

The authors thank the staff of the Department of Gastroenterology, Eastern Chiba Medical Center, for their support during the preparation of this manuscript.