Optical Diagnosis of Diminutive Rectosigmoid Hyperplastic Polyps

Leave-in-situ is only for diminutive rectosigmoid lesions called hyperplastic with high confidence, and only if audited NPV for adenoma is at least 90%. It is not the same strategy as resect-and-discard. DISCARD worked in experts. DISCARD2 failed in community hospitals. Do not leave a proximal NICE 1 lesion. NICE 1 is the serrated class, not HP only.

Experienced teaching points

Clinical Pearls

Leave-in-situ is only for diminutive rectosigmoid lesions called hyperplastic with high confidence on a validated system, and only if audited NPV for adenoma is at least 90%.
NICE type 1 and JNET type 1 lump hyperplastic polyps with sessile serrated lesions. A pale right-colon mucus-capped lesion is the lookalike you must not leave or discard.
Resect-and-discard is a separate strategy. ESGE 2024 still wants retrieval and histology for resected polyps and reserves discard for experts.
DISCARD worked in experts. DISCARD2 failed in community hospitals. DISCARD3 recovered interval concordance in BCSP colonoscopists with a photodocumentation standard.
Gastric hyperplastic polyps are a reactive gastric disease tied to H. pylori or autoimmune gastritis. Do not mix them with rectosigmoid optical diagnosis.

Clinical Bottom Line

Lesion	Optical call	Action
Diminutive (<=5 mm) rectosigmoid, high-confidence NICE 1, no WASP SSL features [11]	Hyperplastic	Leave in situ only if the operator (or audited AI-assist program) meets PIVI-1 [1] / SODA leave-in-situ thresholds. Photodocument.
Same lesion, low confidence or mixed NICE signs	Uncertain	Resect and send to pathology.
Diminutive, high-confidence NICE 2	Adenoma	Resect (cold snare). Discard without pathology only inside an expert or audited resect-and-discard program that meets PIVI-2. Otherwise retrieve.
Any size proximal to sigmoid, NICE 1, mucus cap, clouded surface, indistinct border	Treat as SSL until histology	Resect and retrieve. Do not leave. Do not discard. SSL management belongs on the SSL post.
NICE 3 or JNET 3	Deep submucosal invasion	Do not snare. Targeted biopsy of the worst area. Surgical evaluation.
Gastric "hyperplastic polyp"	Different disease	Test and treat H. pylori. Biopsy background mucosa. Not a colon optical-diagnosis problem (AGA 2026 CPU).

What do I do tonight?

If it is diminutive (<=5 mm), rectosigmoid, high-confidence NICE 1, no WASP SSL features, and you (or an audited AI program) meet PIVI-1 / SODA leave-in-situ thresholds: leave it and photodocument.
If confidence is low or NICE signs are mixed: resect and send to pathology.
If it is diminutive high-confidence NICE 2: resect with a cold snare. Discard without pathology only inside an expert or audited resect-and-discard program that meets PIVI-2. Otherwise retrieve.
If it is proximal to sigmoid, NICE 1, mucus-capped, or has WASP SSL features: resect and retrieve. Never leave it.
If this is a gastric hyperplastic polyp, stop. That is a different disease (AGA 2026 CPU).

What do the societies say?

Do not leave proximal NICE 1 polyps in situ. Do not treat leave-in-situ as the same decision as resect-and-discard. Do not mix gastric hyperplastic polyps with rectosigmoid optical diagnosis.

Three-card summary: leave-in-situ for diminutive rectosigmoid NICE 1, resect-and-discard is a separate expert strategy, proximal NICE 1 is never left. — Figure 2. Leave-in-situ is not resect-and-discard. Proximal pale mucus-capped lesions are SSL until proven otherwise.

Leave-in-situ or resect-and-discard?

Leave-in-situ (diagnose-and-leave) and resect-and-discard are related and not identical. Leave-in-situ avoids polypectomy for a high-confidence diminutive rectosigmoid hyperplastic polyp. PIVI-1: NPV >=90% for adenomatous histology. ESGE SODA: sensitivity >=90% and specificity >=80% for neoplasia in the rectosigmoid. Resect-and-discard still resects a diminutive polyp, skips pathology, and assigns surveillance from the optical call plus histology of polyps >5 mm. PIVI-2: >=90% agreement with pathology-based intervals. SODA for discard is easier to audit (>=80% sensitivity and specificity for 1-5 mm neoplasia anywhere).

ASGE's 2011 PIVI is a performance threshold, not a mandate that every US endoscopist discard today. AGA 2024 uses NICE and WASP to choose how to resect, not to stop the pathology jar. ESGE 2024 is the current polypectomy text: diminutive rectosigmoid hyperplastic polyps have negligible progression risk, so diagnose-and-leave is appropriate. Retrieve resected polyps. Reserve discard for experts.

Does optical diagnosis work outside expert centers?

DISCARD (St Mark's, four colonoscopists, polyps <10 mm): adenoma sensitivity 0.94, HP specificity 0.89, surveillance match 98% BSG and 95% USMSTF in the 82 patients who got an optical interval. Expert, polyp-enriched, not a community RCT, not leave-in-situ outcomes.

DISCARD2 [8] (six UK general hospitals, 28 colonoscopists trained on NICE): patient-level adenoma sensitivity 83.4%, powered for 95%. Surveillance-need sensitivity 73%. Polyp-level adenoma sensitivity 76.1%. If two or more NICE adenoma features were present, sensitivity 99.4%. Neither expertise, stated confidence, nor HD independently rescued accuracy in the adjusted model. The paper's conclusion: NBI optical diagnosis cannot currently be recommended for routine clinical practice. That is the honest counterweight.

DISCARD3 [9] (eight English BCSP colonoscopists, structured training): <=5 mm high-confidence accuracy 81.5%, adenoma sensitivity 93.0%, PPV 90.8%. Surveillance concordance for <=5 mm high-confidence optical diagnosis: USMSTF 91.3%, ESGE 98.3%, BSG 98.7%. That meets PIVI-2 in this selected screening group. No cancers <10 mm. England later used this work for national BCSP optical diagnosis. That is a screening-program implementation, not proof that US community practice already qualifies.

Abu Dayyeh 2015: NBI pooled NPV 91%, but community 88.3% and novices 87%. Experts plus high-confidence reads are the subgroup that actually passes. i-SCAN and FICE pooled NPVs were lower.

Can I leave a proximal NICE 1 polyp?

Hewett's NICE type 1 vs type 2 was built for use with or without magnification. USMSTF 2020 is explicit: type 1 = hyperplastic polyps and sessile serrated lesions. JNET type 1 lumps them the same way. WASP adds four SSL features onto NICE 1: cloud-like surface, indistinct borders, irregular shape, dark spots inside crypts (Hazewinkel). Two or more of those features: call SSL, resect, retrieve.

A pale right-colon mucus-capped lesion is not a diminutive rectal HP you can leave. Proximal NICE 1 is SSL until histology. Do not expand this page into the serrated-pathway textbook. That is the SSL post.

Cancer in diminutive polyps is vanishingly rare. Ponugoti: zero cancers in 36,107 polyps <=5 mm. Gupta N: advanced histology 0.5% in diminutive polyps. The failure mode is not missing a 3 mm cancer. It is misclassifying SSL and assigning the wrong interval. Unmarked diminutive rectosigmoid polyps you cannot photograph with confidence get resected.

Can AI replace histology for diminutive polyps?

Alaoui 2021 (live SIMPLE/iSCAN): surveillance-interval agreement for <=5 mm was 93.5% (meets PIVI-2), but rectosigmoid NPV was 86.7% (misses PIVI-1). Meeting discard math is not the same as meeting leave-in-situ math.

ABC (CAD-EYE, diminutive rectosigmoid polyps): AI-assisted NPV 91.0% (meets PIVI-1). CADx alone 86.7% (does not). Expert AI-assisted accuracy 91.9% vs nonexperts 82.3%. Do not cite ABC as a pass for proximal leave-in-situ or discard. Do not trust CADx alone.

DISCARD3 error analysis: 50 of 184 high-confidence discordance cases were pathology or lab error, not optical error. Histology is an imperfect gold standard. That cuts both ways. Proximal serrated lesions still go in the jar.

Hassan 2010 modelled $33 million in annual US savings. That is a Markov model at 23% screening adherence, not an observed budget number. Cost is real. It is secondary to not leaving an SSL.

Pitfalls

Pitfall	Why it matters
Treating leave-in-situ and resect-and-discard as one suction-trap maneuver	Leave-in-situ avoids polypectomy. Discard still resects. Different thresholds, different QA stories.
NICE 1 in the right colon = "tiny HP, discard"	NICE 1 is the whole serrated class. Proximal NICE 1 is SSL until proven otherwise.
Community NBI after a slide deck, no audit	DISCARD2 failed its 95% bar.
Using SIMPLE NPV as if it met PIVI-1	Alaoui met PIVI-2 and missed PIVI-1.
Trusting CADx alone, or using it proximal to the sigmoid	ABC was powered for diminutive rectosigmoid polyps.
Leaving a distal lesion with WASP SSL features	Call SSL, resect, retrieve.
Importing colon optical-diagnosis rules to gastric HP	Gastric hyperplastic polyps are a reactive gastric disease (AGA 2026 CPU).

Selected references

Questions doctors ask

Can I leave a 6 mm rectosigmoid NICE 1 polyp?

Only if it is diminutive (<=5 mm), high-confidence, no SSL features, and your audited NPV for adenoma is at least 90%. Otherwise resect and retrieve.

Is leave-in-situ the same as resect-and-discard?

No. Leave-in-situ is PIVI-1 / SODA. Resect-and-discard is PIVI-2. ESGE 2024 still wants histology for resected polyps except in expert discard programs.

Does optical diagnosis work in community practice?

DISCARD worked in experts. DISCARD2 failed in community hospitals. DISCARD3 recovered interval concordance in BCSP colonoscopists with photodocumentation.

Can I leave a proximal NICE 1 lesion?

No. NICE 1 lumps hyperplastic polyps with sessile serrated lesions. A pale right-colon mucus-capped lesion is the lookalike you must not leave.

Topics

Polyps & Polypectomy

For your teaching file

Save this article as a PDF

Drop your email and we'll open a print-ready version you can save as a PDF — and you'll start getting our weekly GI endoscopy newsletter.

← All articles