Role of postoperative endoscopy after peroral endoscopic myotomy: Evidence from a large cohort study

Abstract

BACKGROUND

Although the efficacy and safety of peroral endoscopic myotomy (POEM) are well established, evidence guiding optimal perioperative management remains limited. In particular, current international guideline recommendations regarding the role of programmed postoperative endoscopy offer limited procedural guidance and are supported by a restricted evidence base.

AIM

To identify risk factors for adverse events and complications following POEM and to evaluate the clinical value of postoperative endoscopy.

METHODS

This retrospective single-center study included 417 patients with achalasia or other primary esophageal motility disorders who underwent POEM at a tertiary care academic center (January 2013 to May 2025), using a prospectively maintained clinical database. Primary outcomes were intraoperative adverse events, pathological findings in postoperative endoscopy, and postoperative complications. Independent risk factors for intraoperative adverse events and pathological endoscopic findings were assessed using univariable and multivariable logistic regression. The number needed to screen (NNS) was calculated for programmed and risk-based postoperative endoscopy.

RESULTS

Intraoperative adverse events, predominantly mucosal injury (3.1%) and submucosal bleeding (3.8%), occurred in 10.1% of patients. Pathological findings requiring intervention were detected in 1.7% of programmed postoperative endoscopies. Overall, 3.1% of patients developed postoperative complications (Clavien-Dindo grade IIIa-IVb), of which 76.9% were (sub)mucosa-related. Obesity, myotomy length, and myotomy rate (minute/cm) were identified as independent risk factors for intraoperative adverse events. Adverse event rates declined in parallel with marked improvements in operative efficiency over time. Intraoperative adverse events predicted abnormal findings on programmed postoperative endoscopy. Risk-based endoscopic stratification reduced the NNS from 49.3 to 14 for findings requiring intervention.

CONCLUSION

Intraoperative adverse events during POEM predict pathological findings on programmed postoperative endoscopy. A risk-adapted postoperative endoscopy strategy maintains patient safety while substantially reducing unnecessary interventions.

Key Words: Achalasia; Peroral endoscopic myotomy; Adverse events; Postoperative endoscopy; Perioperative care

Core Tip: Although peroral endoscopic myotomy (POEM) is an established treatment for achalasia, evidence guiding optimal perioperative management remains limited. In this large single-center cohort of 417 patients, we identified independent risk factors for intraoperative adverse events and pathological findings in postoperative endoscopy following POEM. Risk stratification based on these factors substantially reduced the number needed to screen for clinically relevant endoscopic findings requiring intervention. These data provide evidence-based guidance for individualized perioperative management after POEM and support a selective, risk-adapted postoperative endoscopic strategy that preserves patient safety while reducing unnecessary procedures.

INTRODUCTION

Achalasia is a rare esophageal motility disorder characterized by impaired lower esophageal sphincter relaxation and absent or uncoordinated esophageal peristalsis[1-3]. Current therapeutic strategies aim to alleviate symptoms, decelerate disease progression, and prevent late complications[4,5]. Since its introduction in 2010, peroral endoscopic myotomy (POEM) has emerged as an effective treatment alongside pneumatic balloon dilatation and laparoscopic Heller myotomy (commonly referred to as LHM), being functionally non-inferior to LHM, with comparable or even superior functional outcomes[6-15].

Although technically demanding, POEM is considered safe when performed by experienced endoscopists[9,16-19]. Still, major and minor procedure-related adverse events (AEs) such as mucosal injury, bleeding, perforation, or insufflation-related events may occur and can ultimately cause severe complications[20,21].

Efforts to improve perioperative patient safety have focused on identifying risk factors for POEM-associated AEs and refining early detection and management strategies. The necessity of targeted postoperative re-evaluation and the appropriate diagnostic modalities remain subjects of ongoing debate[22-25].

Current evidence and guidelines generally discourage routine esophagram because of its limited diagnostic yield[24,26]. Endoscopy has been proposed as an alternative for early postoperative detection and, when necessary, immediate management of endoluminal pathologies[25] and is therefore implemented as part of routine postoperative care by some centers[25]. However, programmed postoperative endoscopy carries procedural risks, prolongs fasting, and requires substantial resources, while its overall clinical benefit remains uncertain and debated[25,27]. Consequently, current international guideline recommendations are cautiously worded and are not supported by robust evidence regarding the clinical relevance of postoperative endoscopy[5,19,27]. This large cohort study aims to address this evidence gap by identifying independent perioperative risk factors for AEs and by evaluating the clinical utility of programmed postoperative endoscopy relative to a risk-adapted approach.

MATERIALS AND METHODS

Study design and patient selection

This retrospective single-center cohort study was conducted at the Achalasia Center of University Hospital Würzburg, Germany, a tertiary care referral center for achalasia and other esophageal motility disorders. Data were extracted from a prospectively maintained database comprising the medical records of all consecutive patients who underwent POEM between January 2013 and May 2025 for achalasia or other esophageal motility disorders, classified according to the contemporaneous Chicago Classification. Diagnoses were confirmed by high-resolution manometry (commonly referred to as HRM) in accordance with European guidelines[5].

Diagnostic work-up

Symptom severity was quantified using the Eckardt score[28]. All patients underwent upper endoscopy with biopsies to exclude eosinophilic esophagitis. For diagnosis and classification, HRM was performed preoperatively; in more than 80% of cases, studies were conducted or repeated in-house. In-house diagnostics were performed in a standardized setting using the ManoScan^TM high-resolution esophageal manometry catheter (Medtronic GmbH, Meerbusch, Germany), with a median integrated relaxation pressure upper limit of 15 mmHg. Timed barium swallow images were available in all cases, while additional imaging was reserved for patients with risk factors for malignant pseudo-achalasia[5]. Treatment decisions followed a shared decision-making approach.

Pre-procedural care and POEM procedure

On the day before POEM, upper endoscopy was performed to clean the esophagus and to rule out thrush. Preoperatively, patients received only clear liquids and intravenous fluids as needed. All procedures were performed by two experienced endoscopic surgeons following a standardized protocol[10]. The first 10 cases of each surgeon were proctored by an experienced POEM specialist[29]. All procedures were performed under general anesthesia with preoperative antibiotic prophylaxis and carbon dioxide insufflation. POEM was initiated with submucosal injection of saline into the esophageal wall, followed by a 1.2-cm mucosal incision to access the submucosal space. A submucosal tunnel was created toward the lower esophageal sphincter and extended 3-4 cm beyond it, followed by myotomy of the circular and partial longitudinal muscle layers. The mucosal entry was closed with endoclips. Anterior myotomy was the standard approach; posterior or EndoFLIP^TM-guided (Endoflip^TM EF-322N; Medtronic GmbH) approaches were reserved for revisional or complex cases. Myotomy time, length, and rate (minute/cm) were recorded.

Post-procedural care

On the day of surgery, patients were monitored for 2 hours before transfer to the general ward. The lead endoscopic surgeon reviewed the patients starting on the evening of surgery and daily thereafter, together with the general ward staff, until discharge. Laboratory parameters were routinely assessed on the mornings of postoperative day (POD) 1 and POD 3, with additional testing performed when clinically indicated, such as in the presence of laboratory abnormalities (e.g., hemoglobin decline) or symptoms deviating from the expected postoperative course. Additionally, programmed endoscopy was performed on POD 1; the standardized inspection protocol and examination conditions are summarized in Supplementary Table 1. An abnormal finding was defined as any endoscopic observation deviating from the expected postoperative appearance after POEM. The decision to perform an endoscopic intervention was made based on the combination of endoscopic findings and the clinical context, and was ultimately determined through interdisciplinary team discussion (Supplementary Table 2). A soft diet was initiated after a normal endoscopic examination. A standard-dose proton pump inhibitor was administered twice a day for 6 weeks.

With the implementation of a selective POD 1 endoscopy strategy (from June 2024 onward), predefined criteria were used to prompt a decision for endoscopic reassessment. These criteria are outlined in Supplementary Table 3 and were applied consistently in clinical practice according to a decision-making algorithm.

Patients were typically discharged starting on POD 2 and onwards, once clinical recovery and laboratory parameters were deemed satisfactory. Discharge criteria included adequate oral intake, sufficient pain control with non-opioid analgesics, and stable laboratory findings, defined as hemoglobin levels within 2 g/dL of preoperative values, only moderate elevations of inflammatory markers (leukocyte count and C-reactive protein), and balanced electrolyte levels. At discharge, patients received standardized instructions to follow a pasty diet for 7 days and to immediately contact our center in case of fever (> 38.5 °C), pain, dysphagia, or other concerning symptoms. Patients were also encouraged to report any post-myotomy complications to our center, even if treated at other institutions, to ensure quality control and data integrity. Follow-up was performed in accordance with guidelines from the United European Gastroenterology and the European Society of Neurogastroenterology and Motility[5].

Outcomes, variables and variable definitions

The primary outcome was periprocedural patient safety, assessed by intraoperative AEs (iAEs) and postoperative complications. Independent risk factors for iAEs were identified. Special emphasis was placed on the findings of programmed postoperative endoscopy, and its clinical relevance was assessed by calculating a context-specific number needed to screen (NNS).

NNS: NNS was defined as the number of patients who need to undergo programmed postoperative endoscopy to identify 1 patient with postoperative abnormal findings after POEM. Abnormal findings were defined as any endoluminal deviation from the expected post-POEM appearance, including impaired entry-site closure, mucosal disruption, or (sub)mucosal bleeding. Lower NNS values indicate a higher diagnostic yield per screened patient.

AEs: An AE was defined as any undesirable event or finding associated with the procedure that deviates from the expected standard, including aspects related to operative technique or the procedural/patient setting. AEs directly related to the intraoperative course are referred to as iAEs and include submucosal bleeding, mucosal injury, insufficient entry-site closure, and insufflation-related AEs (e.g., subcutaneous emphysema, pneumothorax, pneumomediastinum, pneumoperitoneum, and pleural effusion). Insufflation-related AEs were categorized as such only if they had clinical relevance, such as causing pain or dysphagia, or requiring an intervention. Submucosal tunnel bleeding requiring the use of Coagrasper^TM hemostatic forceps (Olympus Medical, Tokyo, Japan) was classified as an iAE, given that even minor bleeding can reduce visibility and increase the risk of mucosal injury.

Postoperative complications: A postoperative complication was defined as an event following POEM that altered the normal postoperative course and required additional management. Postoperative complications were graded according to the Clavien-Dindo classification (CDC) based on their severity[30].

Length of stay: Length of stay (LOS) was defined as the period from the day of surgery until hospital discharge.

Time to surgery: Time to surgery was defined as the interval from the primary onset of symptoms to the time POEM procedure was performed at our institution.

Statistical analysis

Descriptive and inferential statistical analyses were performed using R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism version 10.1.1 (GraphPad Software, San Diego, CA, United States). Basic figures were generated using GraphPad Prism, whereas advanced statistical modeling and visualization were conducted in R using established packages including ggplot2, broom, dplyr, and logistf. Continuous variables are presented as mean ± SD for normally distributed data and as median with interquartile range for non-normally distributed data. Categorical variables are reported as absolute and relative frequencies. Distributional assumptions were assessed by visual inspection and appropriate statistical tests. Group comparisons were performed using Student’s t-test or the Mann-Whitney U test, as appropriate. Categorical variables were compared using the χ² test or Fisher’s exact test. To identify predictors of iAEs and pathological findings on postoperative endoscopy, univariable and multivariable logistic regression analyses were performed. Candidate predictors for the univariable analyses were selected a priori based on clinical relevance and previously reported associations. Variables entered into the multivariable models were selected based on a combination of clinical plausibility and signals of association observed in the univariable analyses in order to limit the number of model parameters relative to the available event counts. Supplementary Table 4 summarizes variables included in the different analyses. To reduce the risk of collinearity and overfitting, multivariable models were restricted to a predefined subset of clinically plausible variables, and closely related procedural surrogates were not included simultaneously where possible. In particular, procedural experience (modeled through case-number thresholds) and myotomy rate represent conceptually related measures; case-number thresholds were used to capture cumulative surgeon experience, whereas myotomy rate was included as a procedure-level marker of technical performance during an individual intervention. These variables were therefore interpreted with caution, acknowledging their potential interrelationship. For continuous variables, odds ratios (ORs) were scaled to the interquartile range to enhance clinical interpretability and comparability across variables. Achalasia type was modeled as a categorical three-level factor (types I-III), with type I serving as the reference category. In addition to level-specific ORs, the overall association of achalasia type with the outcome was evaluated using a global likelihood-ratio test. Logistic regression models were fitted using standard maximum likelihood estimation. In cases of complete or quasi-complete separation, Firth’s penalized logistic regression was applied using the logistf package in R to reduce small-sample bias. Results are reported as ORs with 95% confidence intervals (CIs). Multicollinearity within the multivariable models was assessed using variance inflation factor (VIF) and generalized VIF (GVIF) diagnostics, with GVIF ^{[1/(2 × degrees of freedom)]} values used for variables with multiple degrees of freedom. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. To further assess the association between iAEs and clinically relevant findings on POD 1 endoscopy, a contingency-table analysis was performed. Based on the resulting 2 × 2 table, effect estimates including ORs, relative risks, and absolute risk differences were calculated. In addition, diagnostic performance metrics (i.e., sensitivity, specificity, positive predictive value, and negative predictive value) were derived with corresponding 95%CIs. Patients who did not receive programmed POD 1 endoscopy (selective strategy) were analyzed separately. Missing data were minimal due to the use of a prospectively maintained institutional database; therefore, complete-case analyses were performed. Statistical expertise was available within the author team. The statistical methods and analyses applied in this study were reviewed by a biostatistician and were deemed appropriate for the study design and data structure. All analyses were performed in accordance with accepted biostatistical standards.

RESULTS

Baseline characteristics

Between January 2013 and May 2025, 417 patients underwent POEM performed by two leading endoscopic surgeons at our center. Tables 1 and 2 summarize the baseline characteristics and diagnostic criteria of the included patients.

Table 1 Baseline characteristics, n (%).

Characteristic	Total (n = 417)
Sex
Male	230 (55.2)
Female	187 (44.8)
ASA stage
I	36 (8.6)
II	351 (84.2)
III	30 (7.2)
Autoimmune disease
None	367 (88.0)
Hashimoto	23 (5.5)
Rheumatic disease	7 (1.7)
Other	17 (4.1)
Missing value	3 (-)
Pulmonary disease
COPD	5 (1.2)
Asthma	27 (6.5)
Type 2 diabetes
NIDDM	14 (3.4)
IDDM	5 (1.2)
Smoking	84 (20.1)
Anticoagulation
None	385 (92.3)
ASS	18 (4.3)
NMH/NOAK/Warfarin	14 (3.4)
Diagnosis
Achalasia	399 (95.7)
Type I	74 (17.7)
Type II	242 (58.0)
Type III	83 (19.9)
Other	18 (4.3)
EGJ-OO	2 (0.5)
Nutcracker	1 (0.2)
Jackhammer	11 (2.6)
Other	4 (1.0)
Previous treatment
Endoscopic	126 (30.2)
Dilatation	90 (21.6)
Botox	25 (6.0)
Botox and dilatation	11 (2.6)
Myotomy	7 (1.7)

ASA: American Society of Anesthesiologists; COPD: Chronic obstructive pulmonary disease; IDDM: Insulin-dependent diabetes mellitus; NIDDM: Non-insulin-dependent diabetes mellitus; ASS: Aspirin; NMH: Heparin; NOAK: Non–vitamin K oral anticoagulant; EGJ-OO: Esophagogastric junction outflow obstruction.

Table 2 Diagnostic criteria and intraoperative characteristics.

Characteristic	No.	Mean	SD
IRP in mmHg	392	30.7	11.4
DCI in mmHg/second/cm	262	3115.0	5045.0
Eckardt score	396	7.1	2.4
Dysphagia		2.7	0.7
Regurgitation		1.7	1.0
Chest pain		1.2	1.0
Weight loss		1.5	1.2
Time of surgery in minute	415	86.4	34.6
Myotomy length in cm	413	14.3	3.1
Myotomy rate in minute/cm	413	6.3	3.1

IRP: Integrated relaxation pressure; DCI: Distal contractile integral.

There was a near-equal sex distribution (44.8% female, 55.2% male). The mean age at intervention was 43.1 ± 17.1 years, with 5.0% of patients aged ≤ 16 years (Table 1). Only a minority of patients (7.7%) received any form of perioperative anticoagulation. Pre-existing autoimmune disease was present in 12.0% of patients.

Most patients were diagnosed with idiopathic achalasia (399/417, 95.7%), primarily type II (242/417, 58.0%) (Table 1). Other diagnoses (4.3%) included esophagogastric junction outflow obstruction, nutcracker esophagus, and jackhammer esophagus. HRM confirmed a disease-pathognomonic increase in integrated relaxation pressure (mean 30.7 ± 11.4 mmHg). The mean Eckardt score was 7.1 ± 2.4, with dysphagia as the leading symptom (score of 3 in 80.2% of patients) (Table 2 and Supplementary Figure 1).

Prior endoscopic or surgical treatment had been performed in 32.1% of patients (Table 1). Among the pretreated patients, 94.7% (126/133; 30.2%) had undergone endoscopic interventions (pneumatic balloon dilatation and/or botox injections). Only 2.1% of pretreated patients (7/133; 1.7% of all patients) had previously undergone myotomy (Table 1).

Peri-procedural aspects

An anterior myotomy was performed in the majority of cases. The mean operative time was 86.4 minutes, with a mean myotomy rate of 6.3 ± 3.1 minute/cm and a mean myotomy length of 14.3 ± 3.1 cm (Table 2). The relatively large standard deviation in myotomy rate reflects substantial variability between early and later procedures associated with increasing experience and procedural volume. After the first 100 procedures, the mean myotomy rate decreased to 4.6 ± 1.7 minute/cm.

iAEs occurred in 10.1% of patients (42/417), comprising a total of 48 events. More than half of all iAEs (29/48, 60.4%) were local events, including mucosal injury (13/417, 3.1%) and submucosal tunnel bleeding (16/417, 3.8%) (Table 3). Clinically relevant emphysema and pneumothorax occurred in 14 patients (3.4 %) (Table 3).

Table 3 Intraoperative adverse events, n (%).

Characteristic	n = 417
Patients with adverse events	42 (10.1)
Adverse events, total	48 (11.5)
Mucosal lesion, total	13 (3.1)
Esophagus	7 (1.7)
Cardia	6 (1.4)
Submucosal tunnel bleeding	16 (3.8)
Severe emphysema	9 (2.2)
Pneumothorax
Total	5 (1.2)
Requiring thoracic drain	3 (0.7)
Other	5 (1.2)
Conversion to LHM	4 (1.0)

LHM: Laparoscopic Heller myotomy.

A small proportion (4.8%) of patients required postoperative monitoring in an intermediate or intensive care unit. In only 2.6% of the overall cohort (11/417 patients), this was indicated owing to a procedure-related iAE or the development of a postoperative complication (Table 4). Even among patients with clinically relevant complications requiring intervention (CDC ≥ IIIa), 76.9% could be managed in a regular ward setting. Overall, the mean LOS was 4 ± 2.3 days. When stratified by complication severity, patients with an uneventful postoperative course had a mean LOS of 3.8 ± 1.7 days, whereas those with major complications (CDC ≥ IIIa) had a mean LOS of 8 ± 7.5 days. After implementation of the risk-stratified postoperative endoscopy protocol in June 2024, the mean LOS for the overall cohort decreased to 3.3 ± 1.4 days. During this period, LOS was 3.1 ± 1.0 days for patients without complications and 6.0 ± 1.9 days for those with CDC ≥ IIIa events. No readmissions due to postoperative complications occurred among the entire study cohort.

Table 4 Postoperative morbidity, n (%).

Characteristic	n = 417
Level of care
Ward	397 (95.2)
IMC/ICU, total	20 (4.8)
IMC/ICU, procedure-related	11 (2.6)
Clavien-Dindo classification
I-II	13 (3.1)
IIIa-IVb	13 (3.1)

ICU: Intensive care unit; IMC: Intermediate care unit.

Until June 2024, all patients underwent programmed postoperative endoscopy on POD 1; thereafter, postoperative endoscopy was performed selectively based on clinical considerations such as a technically difficult POEM procedure (Supplementary Figure 2 and Supplementary Table 3). Overall, 395 of 417 patients (94.7%) underwent POD 1 endoscopy. Abnormal findings were detected in 36/417 patients (8.6%; 9.1% of all patients with regular endoscopy), of whom 7/417 (1.7%; 1.8% of all patients with regular endoscopy) required endoscopic intervention (Table 5). One additional patient required endoscopy on POD 2 for clip placement owing to a small mucosal dehiscence, and another required stent placement for persistent bleeding primarily detected intraoperatively. Neither event was classified as a relevant pathological finding of the programmed POD 1 endoscopy (Table 5).

Table 5 Pathologic endoscopic findings, n (%).

Findings	Entire cohort (n = 417)	Endoscopy performed (n = 395)
No planned endoscopy	22 (5.3)
Normal findings	359 (86.1)	359 (90.9)
Pathological findings, all	36 (8.6)	36 (9.1)
Mucosal hematoma	3 (0.7)	3 (0.8)
Submucosal hematoma	8 (1.9)	8 (2.0)
Mucosal erosion/lesion	13 (3.1)	13 (3.3)
Ulcer	9 (2.2)	9 (2.3)
Insufficient closure of the entry	2 (0.5)	2 (0.5)
Active bleeding	1 (0.2)	1 (0.3)
Pathological findings, needing intervention	7 (1.7)	7 (1.8)
Endoscopic clip placement for entry closure	2 (0.5)	2 (0.5)
EVT for deep ulcer	1 (0.2)	1 (0.3)
Clip and adrenaline application	1 (0.2)	1 (0.3)
OTSC clip application for bleeding	2 (0.5)	2 (0.5)
Surgery	1 (0.2)	1 (0.3)

EVT: Endoscopic vacuum therapy; OTSC: Over-the-scope clip.

Severe complications (CDC IIIa-IVb) were documented in 3.1% of all cases, with 76.9% (10/13) being related to the esophageal wall (Table 4 and Supplementary Table 5)[30]. Overall, two procedure-related complications required surgical intervention. In 1 case (0.2%), esophageal resection was necessary, and in another patient, laparoscopic transhiatal evacuation of a hematoma was required. All other complications involving injuries of the esophageal or gastric wall were successfully managed endoscopically using clip placement or endoscopic vacuum therapy without long-term sequelae (Table 5).

Independent predictors of AEs

Univariable and multivariable logistic regression analyses were performed to identify (additional) predictors of periprocedural AEs and complications. In univariable analyses, obesity [body mass index (BMI) > 30 kg/m²; OR = 2.34, 95%CI: 1.04-5.24], limited surgical experience (< 25 cases; OR = 2.17, 95%CI: 1.13-4.14), and surgical-technical factors specifically, myotomy rate (OR = 1.44, 95%CI: 1.15-1.81) were independent risk factors for iAEs (Figure 1A and Supplementary Table 6). In multivariable analysis, obesity (OR = 3.78, 95%CI: 1.36-10.52), myotomy length (OR = 2.46, 95%CI: 1.48-4.08), and myotomy rate (minute/cm; OR = 1.78, 95%CI: 1.30-2.44) were ultimately identified as independent risk factors for the occurrence of iAEs (Figure 1B and Supplementary Table 7).

Figure 1 Forest plots of univariable and multivariable logistic regression analyses for intraoperative adverse events and pathological findings on postoperative endoscopy. A and B: Odds ratios with 95% confidence intervals are shown for univariable models; C and D: Odds ratios with 95% confidence intervals are shown for multivariable models. Achalasia type was modeled as a categorical variable with three levels (types I-III), with type I serving as the reference category. Continuous variables are expressed per unit increase or per interquartile range, as indicated. Procedural experience thresholds, with the number of peroral endoscopic myotomy procedures performed at our center by a surgeon as markers of cumulative surgeon’s experience. iAE: Intraoperative adverse events; OR: Odds ratio; IQR: Interquartile range; GLM: Generalized linear model; BMI: Body mass index; myotomy rate: Time required to perform the myotomy, expressed as minutes per centimeter (minute/cm); prev. myotomy: Previous myotomy; OP Nr.: Cumulative number of procedures performed; prev. dilatation: Previous dilatation; prev. endoscopy: Any kind of previous endoscopic intervention (botox injection and/or dilatation) before the first peroral endoscopic myotomy at our center; TTS: Time to surgery (period from initial disease onset/initial diagnosis to first peroral endoscopic myotomy at our center); prev. botox: Previous botox injection; ASA: American Society of Anesthesiologists; LRT: Likelihood ratio test; ES: Eckardt score; NS: Not significant.

Univariable analysis identified achalasia type II (OR = 0.35, 95%CI: 0.16-0.80, reference achalasia type I), myotomy rate (OR = 1.55, 95%CI: 1.12-1.98), low BMI (< 18.5 kg/m²), and the presence of iAEs (OR = 4.81, 95%CI: 2.16-10.72), especially mucosal lesions and submucosal bleeding (OR = 9.40, 95%CI: 3.95-22.41), as factors associated with abnormal findings on POD 1 endoscopy (Figure 1C and Supplementary Table 8).

In multivariable logistic regression analysis, previous iAEs (OR = 4.81, 95%CI: 2.16-10.72), especially mucosal lesions and submucosal bleeding (OR = 9.40, 95%CI: 3.95-22.41), emerged as the strongest independent predictors of abnormal findings on programmed POD 1 endoscopy (Figure 1D and Supplementary Table 9). Low BMI (< 18.5 kg/m²) also remained statistically significant. Achalasia type II was inversely associated with pathological findings on postoperative endoscopy compared with the reference category of achalasia type I.

VIF/GVIF diagnostics demonstrated no relevant multicollinearity among the primary clinical and procedural variables included in the multivariable models, including myotomy length, myotomy rate, and procedural experience threshold (< 25 procedures), with all adjusted GVIF values remaining low (< 1.5 for iAEs and < 1.1 for pathological findings). As expected, relevant collinearity was observed between iAEs and bleeding/mucosal lesion as a subtype of iAEs, reflecting their close clinical and conceptual overlap.

Role and clinical relevance of programmed postoperative endoscopy

To assess the clinical relevance and efficiency of programmed postoperative endoscopy, the NNS was calculated. To detect any abnormality after POEM, 10.6 endoscopies were required. To identify findings necessitating endoscopic intervention, endoscopy had to be performed in 49.3 patients (Figure 2A).

Figure 2 Number needed to screen for the detection of pathological findings on postoperative endoscopy. Bar charts depict the number needed to screen (NNS) required to detect pathological findings in patient subgroups. A-D: Subgroup-specific NNS estimates for unselected study population (A), patients with intraoperative adverse events (iAEs) (B), patients without iAEs (C), and patients who did not undergo programmed endoscopic evaluation (D); E: Summary of NNS shown in panels A-D. Within all panels, individual bars are labeled groups A-J and correspond to the categories described below each subfigure. All estimates are presented descriptively, and no formal statistical comparisons between categories were performed. NNS: Number needed to screen; iAE: Intraoperative adverse events; EGD: Esophagogastroduodenosocopy.

Restricting postoperative endoscopy to patients with known iAEs yielded an NNS of 3.8 for detecting any pathological finding. Further focusing on patients with evidence of intraoperative mucosal lesions or (sub)mucosal bleeding resulted in a marked reduction of the NNS from 10.6 to 2.6. Limiting endoscopy to patients with iAEs reduced the NNS for detecting findings necessitating intervention from 49.3 to 14.0 (Figure 2B). In contrast, among patients without documented iAEs, the NNS increased markedly to 14.1 for detecting any abnormal finding and to 88.3 for detecting lesions requiring therapeutic intervention (Figure 2C).

Given the high NNS, our postoperative protocol was revised in June 2024, after which re-endoscopy on POD 1 was performed in an individualized and selective manner based on patient- and procedure-specific factors. In 22 patients with an uneventful intraoperative course, uncomplicated esophageal anatomy, and unremarkable postoperative recovery, programmed endoscopy was intentionally omitted. None of these patients required unplanned endoscopic evaluation, and all were discharged without complications on POD 2 or POD 3. Among patients in whom endoscopy was deliberately performed during this period, pathological findings were detected with an NNS of 2.3 (vs 10.6), and findings requiring intervention with an NNS of 4.6 (vs 49.3) (Figure 2D). Figure 2E summarizes all calculated stratified and unstratified NNS values in our cohort. Interim 1-year follow-up data were available for 91% of patients in whom programmed endoscopy had been omitted, without evidence of previously unrecognized clinically relevant delayed procedure-related complications.

Conversely, to further explore the association between iAEs and intervention-requiring findings on POD 1 endoscopy and thereby to better inform the safety of omitting programmed POD 1 endoscopy in selected patients we performed a 2 × 2 contingency analysis with calculation of diagnostic performance metrics.

Among the 395 patients included in this analysis (exclusion of 22 patients with no POD 1 due to the selective approach), iAEs occurred in 42 (10.6%) and intervention-requiring endoscopic findings were identified in 7 (1.8%) overall. Such findings were more frequent in patients with an iAE than in those without an iAE [4/42 (9.5%) vs 3/353 (0.85%); Fisher’s exact test, P = 0.027].

When considering iAEs as a potential indicator of intervention-requiring findings, diagnostic performance analysis demonstrated a sensitivity of 57.1% (95%CI: 25.0%-84.2%) and a specificity of 90.2% (95%CI: 86.6%-92.9%). While the positive predictive value was 9.5% (95%CI: 3.8%-22.1%), the negative predictive value was 99.2% (95%CI: 97.6%-99.8%). The corresponding results are summarized in Supplementary Tables 10-12.

DISCUSSION

POEM has emerged as a safe and effective treatment for achalasia and other hypercontractile esophageal motility disorders. Nevertheless, perioperative AEs and complications can still occur[20,22,31]. iAEs primarily include insufflation-related complications and esophageal wall-associated incidents. Esophageal wall-associated AEs should be proactively sought during POEM procedures[32], as early recognition and management are crucial. However, intraoperative detection may be challenging owing to carbon dioxide-related tissue emphysema, prompting the adoption of a routine postoperative re-evaluation in some centers.

Currently, there is no consensus regarding routine postoperative assessment, and existing guideline recommendations are constrained by insufficient evidence[4,5,19,27,32-34]. The American Gastroenterological Association explicitly acknowledges this evidence gap, noting the lack of evidence and standardization on specific postprocedural care strategies[27].

In line with the current literature, our institutional protocol does not incorporate programmed postoperative esophagram owing to its low specificity and poor correlation with clinically significant outcomes[24,26,27,35]. Accordingly, its routine use is not recommended[24].

Instead, early postoperative endoscopy is considered more effective for identifying and managing (sub)mucosal lesions. Accordingly, since the implementation of POEM at our center, a programmed re-endoscopy has been adopted as a primary safety strategy. However, considering the already low overall rate of postoperatively detected lesions (8.6%) and the even lower rate of complications associated with inadequately addressed wall-associated lesions (1.7%), the necessity of programmed POD 1 endoscopy is questionable.

The proportion of pathological findings requiring intervention in our cohort aligns with prior studies[25]. Fujiyoshi et al[25] reported abnormal postoperative endoscopic findings in 14.3% of patients, with 2.4% experiencing newly detected AEs that altered the normal course of treatment by necessitating prolonged fasting, extended hospitalization, or endoscopic intervention. While these authors advocated for programmed re-endoscopy before oral intake, others have questioned its overall benefit[36].

Our findings support the latter view and challenge the value of unstratified programmed re-endoscopy. Given a NNS of 10.6 for any pathologic finding and 49.3 for lesions requiring intervention, programmed re-endoscopy in all patients undergoing POEM, irrespective of postprocedural risk stratification, is not justified. To improve the effectiveness and clinical value of postoperative endoscopy, we evaluated potential selection criteria within our patient cohort. iAEs, especially mucosal lesions and (sub)mucosal bleeding, were identified as independent risk factors for abnormal postoperative endoscopic findings. Restricting postoperative endoscopy to these at-risk groups reduced the NNS from 10.6 (for any pathological finding) and 49.3 (for lesions requiring interventions) to 3.8 and 2.6, respectively.

In contrast, the NNS was markedly higher among patients with an uneventful intraoperative course, particularly when considering clinically relevant findings requiring intervention (49.3 for all patients vs 14.0 for those with an iAE vs 88.3 for those without an iAE). These findings underscore the value of a risk-stratified endoscopic approach based on the presence of iAEs. In addition, it is consistently confirmed that clinical warning signs should reliably prompt postoperative endoscopy[36-38]. The substantially higher diagnostic yield in the subgroup of patients with an iAE may help optimize endoscopic resource utilization while maintaining patient safety.

However, an important concern when omitting programmed POD 1 endoscopy in the absence of iAEs might be the potential risk of missed lesions requiring intervention. The contingency analysis assessing the association between iAEs and intervention requiring findings on POD 1 endoscopy in our cohort mitigates this concern. The very high negative predictive value (99.2%) indicates that the absence of an iAE was strongly associated with an uncomplicated postoperative (endoscopic) course although, this finding must be interpreted in the context of the overall low event rate. In our cohort, clinically relevant findings requiring intervention were identified in only 3 of 353 patients without an iAE (0.85%), indicating a very low residual risk in this subgroup.

From a clinical perspective, the value of iAEs as a marker therefore lies less in its positive predictive performance than in its ability to identify a subgroup with a very low likelihood of clinically relevant postoperative endoscopic pathology when iAEs are absent. This observation is particularly relevant when considering postoperative management strategies following POEM, as the extremely low event rate among patients without an iAE suggests that programmed POD 1 endoscopy may not be necessary for all patients and could potentially be omitted within a risk-adapted, selective strategy. However, given the limited sensitivity of iAEs for predicting intervention-requiring findings in our cohort, iAEs could be regarded as the primary risk stratification marker within a selective postoperative management strategy, while postoperative clinical vigilance and attention to warning signs and symptoms remain essential; this approach nevertheless warrants further prospective validation.

Our multivariable analysis identified BMI and procedural parameters (myotomy length and rate) as independent predictors of iAEs, thereby providing a basis for risk-adapted patient stratification. Ultimately, postoperative wall-associated lesions closely correlate with the presence of lesion-related iAEs.

In contrast to the findings of Wang et al[39], prior myotomy did not emerge as a risk factor for mucosal injury in our cohort, likely owing to the small number of such cases (1.7%). Notably, prior endoscopic treatment, which may induce esophageal wall scarring, also did not emerge as an independent risk factor. These findings have important implications for postoperative management, as same-day discharge appears feasible and safe[36,37,40]. In our cohort, the longer total hospital stay reflects several contextual factors, and LOS decreased substantially with the progressive establishment of the procedure and associated increasing procedural experience. Moreover, enhanced recovery and fast-track concepts have only recently been more widely implemented in Germany. Taken together, these findings support an individualized postoperative strategy. Stratifying patients according to intraoperative findings and clinical course may reduce unnecessary procedures without compromising safety.

Limitations

This study has several limitations. First, its retrospective single-center design may limit the generalizability of the findings, and residual confounding cannot be entirely excluded. In addition, interpretation bias may arise from the fact that all endoscopic procedures and postoperative assessments were performed within the same institution. However, all programmed re-endoscopies were conducted by a dedicated and experienced team following a standardized post-POEM protocol (Supplementary Table 1), allowing consistent identification and interpretation of predefined postoperative findings and reliable differentiation between expected post-procedural changes and atypical findings requiring intervention on POD 1. Furthermore, the present study was conducted at a high-volume tertiary referral center with substantial institutional experience in POEM and standardized peri-interventional management pathways. Although direct transferability of the findings to lower-volume centers may be limited, the proposed approach primarily emphasizes key principles that should be applicable across different practice settings, including awareness and structured assessment of iAEs as well as consistent clinical postoperative evaluation. In this context, a risk-stratified use of post-POEM endoscopy may even help optimize resource utilization and reduce unnecessary procedures, being particularly advantageous in smaller centers.

Second, the study period encompasses both the introduction and maturation of the POEM procedure at our institution. Learning-curve effects cannot be completely excluded and may partly overlap with procedural variables such as myotomy rate or procedure time. However, VIF/GVIF diagnostics demonstrated no relevant multicollinearity between procedural experience and procedure-related parameters within the final multivariable models.

Third, structured 30-day follow-up was not uniformly available for patients in whom programmed endoscopy was omitted following the transition to a selective strategy. However, these patients continued to be prospectively followed within the institutional registry, and interim 1-year follow-up data were available for 91% of patients, without evidence of clinically relevant delayed procedure-related complications. Nevertheless, the limited size of this subgroup restricts the interpretability and generalizability of these observations, particularly with regard to uncommon delayed AEs. Prospective validation of the selective post-POEM management approach in larger and systematically followed patient cohorts therefore remains warranted.

Finally, the comparatively long myotomy length observed in this cohort (mean 14.3 cm) reflects earlier procedural standards during the study period. Recent comparative studies suggest that shorter myotomy strategies may reduce procedural complexity, iAEs, and postoperative reflux without compromising clinical efficacy; although, available data primarily focus on procedural and reflux-related outcomes rather than postoperative endoscopic findings. Therefore, the present findings should be interpreted within the context of these historical procedural characteristics and evolving contemporary POEM techniques favoring shorter, more tailored myotomy strategies[41-43]. In addition, the relatively small number of patients with prior myotomy limits the ability to draw firm conclusions regarding its role as a potential risk factor.

CONCLUSION

Implementing risk stratification for iAEs potentially associated with procedure-related complications appears rational. iAEs, particularly (sub)mucosal defects and bleeding, were strong predictors of abnormal findings on postoperative endoscopy. While programmed postoperative endoscopy offers limited clinical benefit owing to a high NNS, a risk-adapted strategy focused on iAEs offers a more efficient and patient-centered approach. Our findings address a crucial evidence gap and may inform future guideline recommendations and clinical practice in the perioperative management of POEM.