Endoscopic submucosal dissection of early GI neoplasia — marking, submucosal lift, en-bloc dissection and R0 margin verification
Every endoscopic submucosal dissection (ESD) begins not with a knife but with careful optical characterization. Chromoendoscopy (indigo carmine or acetic acid) and narrow-band imaging (NBI) with magnification reveal the surface (pit) pattern and microvascular architecture of the lesion, allowing the endoscopist to predict histology and depth of invasion before a single biopsy is taken. Electrocautery marking dots are then placed circumferentially, 5–10mm outside the visible lesion edge, to guide a incision line that guarantees a clear lateral margin.
Before any tissue is cut, the endoscopist must answer three questions: is this lesion neoplastic, what is its likely depth of invasion, and can it be removed endoscopically with curative intent?
Chromoendoscopy sprays a contrast dye (0.4% indigo carmine in the colon, acetic acid in Barrett's esophagus) that pools in mucosal grooves, revealing the surface pit pattern under white light or magnification. Narrow-band imaging (NBI) uses filtered blue/green light (415/540 nm) absorbed by hemoglobin to highlight the microvascular network without dye.
In the colon, the Kudo pit-pattern classification (types I–V) and the simplified JNET classification (Types 1, 2A, 2B, 3) correlate surface and vessel appearance with histology: Type IIIL/IV pits and JNET 2A predict adenoma or superficial cancer suitable for ESD, while JNET Type 3 (loose vessel areas, avascular zones) suggests deep submucosal invasion (SM2/SM3) that should instead go to surgery.
In the stomach, the VS (vessel-plus-surface) classification system serves an analogous role, and in the esophagus the Japan Esophageal Society (JES) classification of intrapapillary capillary loops (IPCL types B1–B3) predicts invasion depth with high accuracy.
Optical diagnosis with NBI magnification predicts deep submucosal invasion with roughly 85–95% accuracy in expert hands — comparable to endoscopic ultrasound — allowing real-time triage between ESD and referral for surgery without waiting for biopsy results.
Unlike the colon, where mucosal folds provide some visual reference, the stomach in particular has notoriously indistinct lesion borders under white light — differentiated-type early gastric cancer can blend almost seamlessly into surrounding atrophic or intestinalized mucosa. For this reason, Japanese guidelines mandate circumferential marking dots for all gastric ESD, typically placed with argon plasma coagulation (APC) or the tip of the ESD knife itself, spaced roughly 3–5mm apart, 5mm outside the NBI-defined tumor edge.
In the esophagus and colon, where borders are often more distinct on chromoendoscopy, marking is still frequently performed — particularly for flat or laterally spreading lesions, lesions with a scarred or previously biopsied segment (which can obscure the true margin), and any lesion where the endoscopist has any doubt about the true extent.
Marking dots serve a second purpose beyond guiding the mucosal incision: they act as a permanent, gas-independent reference frame throughout the sometimes hour-long procedure, letting the endoscopist reorient after suctioning, irrigation, or scope withdrawal for accessory exchange.
A single confirmatory biopsy is usually taken before ESD is scheduled, but extensive multi-quadrant biopsying is deliberately avoided: biopsy-induced submucosal fibrosis is one of the strongest predictors of a difficult, non-lifting dissection and a higher risk of piecemeal resection. Endoscopic ultrasound (EUS) is reserved for lesions with equivocal optical features or suspected deeper invasion, since EUS depth staging has accuracy of only about 65–85% for differentiating T1a (mucosal) from T1b (submucosal) disease — meaningfully lower than expert optical diagnosis for superficial lesions.
Cross-sectional imaging (CT, and for esophageal/gastric lesions sometimes PET-CT) is added only when submucosal invasion is suspected, to exclude nodal or distant disease that would change the treatment plan from endoscopic to surgical.
Submucosal injection is the single technical step that separates safe ESD from a high-risk free-hand dissection. By expanding the submucosal space with a viscous fluid, the endoscopist creates a physical buffer between the mucosa/lesion and the muscularis propria, both facilitating precise dissection and reducing the chance of a full-thickness burn or perforation.
The ideal submucosal injection fluid creates a durable, easily visualized cushion with minimal tissue damage. Options used worldwide include:
• Normal saline: cheapest and always available, but low viscosity means it diffuses and is absorbed within 5–10 minutes, often requiring repeated injections during a long dissection. • Glycerol solutions (10% glycerol/5% fructose): intermediate viscosity and duration, mild osmotic effect prolongs the cushion versus saline. • Sodium hyaluronate (0.4%): the standard in Japan and increasingly worldwide — high viscosity produces a durable cushion lasting 30–60 minutes or longer, reducing the number of re-injections needed for large lesions. • Succinylated gelatin, hydroxyethyl starch, and other colloid solutions offer similar prolonged-lift properties at lower cost than hyaluronate in some markets.
Indigo carmine or methylene blue dye (a few drops per syringe) is routinely mixed into the injectate to stain the submucosa blue, sharply contrasting it against the white/glistening muscularis propria fibers and making the correct dissection plane visually unmistakable. Dilute epinephrine (1:100,000–1:200,000) is often added for prophylactic vasoconstriction of submucosal vessels.
Injection begins at the anal/proximal edge of the planned incision (so that fluid gravity and scope trajectory work together) and proceeds around and beneath the lesion in stages, re-injecting as the cushion is consumed during cutting. The needle is inserted at a shallow angle just beneath the marking dots, and correct submucosal placement is confirmed by a visible, symmetric bleb rising beneath the mucosa — deep intramuscular injection produces a firm, poorly mobile bulge and should be avoided.
A lesion that fails to lift adequately after adequate volume injection — the "non-lifting sign" — is a critical red flag. It suggests either deep submucosal fibrosis (from prior biopsy, prior resection attempts, or chronic inflammation) or, more concerningly, invasive cancer that has anchored the mucosa to deeper layers. A positive non-lifting sign should prompt reassessment of the resection plan and, in lesions where cancer invasion is suspected, referral for surgical resection rather than continued endoscopic attempts.
A well-formed submucosal cushion increases the working distance between the knife tip and the muscularis propria from roughly 1–2mm to 5–10mm — the single largest modifiable factor reducing perforation risk during ESD.
For large or technically difficult lesions, supplemental traction devices are increasingly used to keep the submucosal plane open and under tension as dissection proceeds, meaningfully shortening procedure time in randomized trials. Common techniques include clip-with-line (an endoclip anchored to a lesion edge, tethered by dental floss or a specialized line to pull the flap open), the S-O clip (a clip-and-spring device providing continuous countertraction), and double-scope/double-channel techniques where a second instrument grasps and retracts tissue. These adjuncts are particularly valuable for colonic ESD, where the thin muscularis propria and difficult scope angulation in the right colon make maintaining visualization of the dissection plane especially challenging.
With the submucosal cushion in place, the electrosurgical knife makes a continuous circumferential incision through the mucosa and superficial submucosa, connecting the marking dots into a closed ring that fully separates the lesion — including its 5–10mm safety margin of normal-appearing tissue — from the surrounding gastrointestinal wall.
ESD knives fall into two broad families. Insulated-tip knives (IT knife, IT knife-2, IT knife-nano) have a ceramic ball at the tip that prevents deep tissue contact, allowing safer cutting with a lateral, shaving-like motion — favored for gastric ESD where the wall is relatively thick. Needle-type knives (Dual knife, Hook knife, Flush knife, HybridKnife) have an exposed, retractable blade offering more precise, perpendicular cuts and easier maneuverability around folds and in tight spaces such as the colon and esophagus, at the cost of requiring more operator experience to avoid inadvertent deep cuts.
Modern electrosurgical generators (e.g., ERBE VIO 300D) alternate between cutting current (Endocut Q or I modes, which pulse cut and coagulate automatically) for the incision itself and forced or soft coagulation modes for hemostasis and controlled submucosal dissection — the operator toggles between these settings dozens of times during a single procedure via a foot pedal.
The incision typically begins on the near (proximal) side of the lesion, working circumferentially, because starting distally can cause the scope to lose its working position as the lesion becomes mobile. A short initial mucosal incision is made at one marking dot, the knife is then walked along the marking line, alternating between short cutting strokes and re-injection as the cushion is consumed.
A critical technical error is finishing the circumferential incision completely before starting submucosal dissection: once fully circumscribed, a large flat lesion loses the fixed traction provided by attached surrounding mucosa, becomes mobile and difficult to grasp, and gravity can cause it to flop over the dissection plane, obscuring the operator's view. For this reason, many endoscopists complete only part of the circumference initially, then interleave incision with early submucosal dissection at the same site — an approach sometimes called the "pocket-creation method."
Small vessels are frequently encountered and coagulated at this stage using the knife tip in soft-coagulation mode or with dedicated hemostatic forceps, since incisional bleeding — while rarely dangerous on its own — can obscure the surgical field for the more demanding dissection phase to follow.
Developed to address the "mobile lesion" problem, the pocket-creation method makes a small (2–3cm) mucosal incision at one point, then immediately tunnels into the submucosal space to create a "pocket" beneath the lesion before the circumferential incision is completed. This preserves normal mucosal attachments around most of the lesion perimeter, keeping it anchored and under tension, so gravity and lesion mobility work in the endoscopist's favor rather than against them. The remaining circumferential incision is then completed from within this stable pocket, and full-thickness submucosal dissection proceeds outward from the tunnel — a technique credited with meaningfully shortening procedure times for large and technically difficult lesions, particularly in the colon.
Submucosal dissection is the technical heart of ESD and the step most responsible for the procedure's steep learning curve. Working millimeter by millimeter, the endoscopist divides the connective-tissue fibers of the submucosa while maintaining a consistent plane just above the muscularis propria — close enough to remove all submucosa attached to the specimen (needed for accurate depth staging), yet far enough to avoid burning through the muscle layer.
The blue-stained submucosal fibers, contrasted against the pale, glistening, circular/longitudinal muscle fibers of the muscularis propria, provide the visual roadmap for dissection. Cutting proceeds in a controlled, layer-by-layer fashion: the knife divides small bundles of connective tissue and traversing blood vessels, working in short strokes parallel to the muscle layer rather than perpendicular to it, which reduces the chance of inadvertently deepening into the muscle if the tissue moves unexpectedly.
Experienced operators aim to leave a thin, uniform layer of submucosa attached to the muscularis propria across the entire dissection bed, and an equally thin, even layer attached to the underside of the specimen — an even margin here both minimizes perforation risk and yields the best possible specimen for pathological depth staging. Novice operators, by contrast, tend to produce an uneven plane: excessively deep in some areas (approaching or exposing the muscle, raising perforation risk) and excessively shallow in others (leaving submucosal tissue behind, risking incomplete/non-curative resection).
Visible circular muscle fibers of the muscularis propria are the single most important landmark in ESD — the entire procedure is, at its core, an exercise in staying reliably 1–2mm above them across an area that can exceed 20–30 cm² for the largest lesions.
Microperforation appears as a sudden, well-demarcated dark opening exposing extraluminal fat or peritoneum, sometimes accompanied by abrupt luminal collapse (in colonic ESD, due to peritoneal insufflation of CO2). Most intraprocedural perforations recognized immediately are managed endoscopically with clips (through-the-scope or over-the-scope) without need for emergency surgery, particularly since CO2 insufflation (rather than room air) is standard for ESD — CO2 is absorbed roughly 150 times faster than nitrogen, dramatically reducing the risk and severity of tension pneumoperitoneum if a perforation does occur.
Risk factors for perforation include larger lesion size, submucosal fibrosis (from prior biopsy or resection attempts), difficult anatomic location (e.g., gastric fundus, colonic flexures, ileocecal valve), and — consistently across published series — lower operator case volume. Delayed perforation, recognized hours after the procedure by new abdominal pain and free air on imaging, is less common but carries higher morbidity and more often requires surgical intervention.
ESD is widely regarded as one of the most technically demanding procedures in interventional endoscopy, requiring bimanual-equivalent scope control, tissue-plane recognition, and electrosurgical judgment developed over dozens to well over a hundred supervised cases. Published learning-curve analyses suggest gastric ESD proficiency (consistent en-bloc resection with acceptable complication rates) is typically reached after 30–40 cases for endoscopists already experienced in basic endoscopic resection, while colorectal ESD — technically harder due to a thinner muscularis propria, more difficult scope angulation (particularly right colon and flexures), and greater looping tendency — often requires 80–100 or more cases.
This learning curve is precisely why ESD training worldwide has moved toward structured pathways: ex-vivo animal model practice, live-animal (porcine) courses, proctored early clinical cases, and staged progression from straightforward rectal lesions to increasingly difficult locations, rather than unsupervised independent practice from the outset.
The endoscopic procedure ends when the specimen is finally freed and retrieved — but the clinical question ESD exists to answer is only settled once the pathologist reports on it. En-bloc resection (removal in one intact piece) is a prerequisite for meaningful margin assessment; R0 resection (histologically clear lateral and vertical margins) combined with favorable depth, differentiation, and absence of lymphovascular invasion together define a curative resection.
ESD was developed specifically to overcome the central limitation of standard EMR: lesions larger than about 20mm generally cannot be captured in a single snare and must be removed piecemeal, in several overlapping fragments. Piecemeal resection makes it impossible to reliably assess true lateral margins on the reassembled specimen (fragment edges cannot be definitively matched to the true excision edge), and it leaves microscopic islands of residual neoplastic tissue between snare bites far more often than en-bloc technique — the mechanistic reason piecemeal EMR carries a 10–20% local recurrence rate at follow-up colonoscopy versus under 1–2% after curative ESD.
By dissecting the entire lesion free in one continuous submucosal plane, ESD achieves en-bloc resection in roughly 90–95% of cases across gastric, esophageal, and colorectal series in expert centers — a figure that falls with increasing lesion size, submucosal fibrosis, and difficult location, and rises with operator experience.
Immediately after retrieval, the specimen is pinned flat, mucosal side up, onto a cork or foam board — stretching it to approximate its in-vivo dimensions without overstretching, which would distort measured margins — and fixed in formalin. The pathologist sections the specimen in thin (2–3mm) parallel slices perpendicular to the closest margin, examining every slice for tumor extent, depth of invasion, differentiation grade, and lymphovascular invasion.
Lateral (horizontal) margin: measured as the distance from the tumor edge to the cut edge of normal mucosa; R0 requires tumor-free lateral margins on all slices. Vertical (deep) margin: the submucosal cut surface at the base of the specimen; R0 requires no tumor at the deep margin, confirming the dissection plane stayed within the wall. An R1 resection (microscopically positive margin) or Rx (margin status indeterminate, e.g. due to cautery artifact or fragmentation) both fall short of a curative resection, even if the lesion was removed en-bloc.
A clear (R0) margin is necessary but not sufficient for a curative resection. Guidelines (Japanese Gastric Cancer Association for stomach; ESGE/European and multiple Asian society guidelines for colorectal and esophageal lesions) define curability using a composite of: R0 margins, depth of invasion below a defined submucosal threshold, absence of lymphovascular invasion, favorable histologic differentiation, and (for gastric cancer) absence of ulceration for larger intramucosal lesions.
For colorectal lesions, submucosal invasion depth of less than 1000 micrometers (1mm) below the muscularis mucosae (so-called SM1) carries an estimated lymph node metastasis risk of roughly 1–3% and is generally considered low enough to accept endoscopic resection as definitive treatment when combined with the other favorable features. Deeper invasion (SM2/SM3, ≥1000µm) carries a substantially higher nodal metastasis risk (roughly 10–15% in some series) and, even with clear margins, usually prompts discussion of additional surgical resection with lymphadenectomy.
When a resection is en-bloc and R0 but fails a depth, differentiation, or lymphovascular criterion, it is termed a "non-curative" (but still complete, R0) resection — a distinct category from a positive-margin (R1) resection, and one that requires individualized discussion of surveillance versus additional surgery.
A resection can be simultaneously en-bloc, R0, and still "non-curative" — margin clearance and cancer cure are related but distinct questions, and both must be answered before a patient can be told ESD alone was sufficient treatment.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Lateral / deep margin | R0 — no tumor at cut edge | Assessed on serial 2–3mm perpendicular sections of the pinned specimen | Necessary for any curative claim |
| Depth of invasion | Intramucosal or SM <1000 µm | Measured from muscularis mucosae to deepest invasive front | Predicts <1–3% nodal metastasis risk |
| Differentiation grade | Well / moderately differentiated | Poor differentiation associated with higher nodal spread | Favorable grade supports endoscopic cure |
| Lymphovascular invasion | Absent | Histologic identification of tumor in lymphatic/vascular channels | Presence mandates consideration of surgery |