Aust 2010 — Serrated Polyp Diagnostic Criteria (German consensus)

A German Society of Pathology consensus that turned serrated colorectal polyps into an operational decision rule — two of four architectural features in at least two crypts — and asserted its reproducibility without measuring it.

A German Society of Pathology consensus that turned serrated colorectal polyps into an operational decision rule — two of four architectural features in at least two crypts — and asserted its reproducibility without measuring it.

Read status: full text. Classified “REVIEW AND PERSPECTIVE” by the journal. It is a consensus proposal, not a study: there is no cohort, no experiment and no result, so AGENTS.md §8’s four model-paper fields do not apply and are not forced onto it below.

What it is

Output of a consensus meeting of the Working Group of Gastroenterological Pathology of the German Society for Pathology, held in Düsseldorf on 29 April 2008, working from 19 selected cases. Its purpose is nomenclature and diagnostic criteria for serrated colorectal polyps in the German-speaking countries, plus management recommendations aligned to the German S3 colorectal guideline.

The clinical motive is the serrated pathway: lesions once dismissed as hyperplastic are clonal proliferations carrying KRAS and BRAF mutations, progressing through CpG-island methylation (CIMP) with silencing of mismatch-repair genes, and so to MSI carcinoma. The paper reports that “some authors postulate” this route accounts for 30% of all sporadic colorectal carcinomas, and that progression may be faster than the classical adenoma–carcinoma sequence — the fastest cited being a case report of SSA to invasive cancer in 8 months.

The decision rules, which are the transferable content

Four entities. The paper’s own criteria tables, compressed:

	Site	Gross	Key microscopy
HP	Left colon, rectum	Slightly elevated, usually <5 mm	Elongated crypts; serration confined to upper half/third; small uniform basal nuclei; no atypia or architectural dysplasia
SSA	Right colon	Sessile, >5 mm	Serration into the lower third; T- and L-shaped crypts above muscularis mucosae; inverted crypts below it; columnar dilatation of the crypt base
TSA	Left colon (60%)	Pedunculated/villous	IEN by definition (90% LG, 10% HG); diffuse eosinophilic cytoplasm; ectopic crypt formation; prominent serration
Mixed	—	—	Conventional adenoma plus a serrated lesion, each component and its IEN grade named

Stated frequencies: HP 80–90%, SSA 15–20%, TSA 1–6%.

The operational rule, and the reason this paper is worth having filed. The consensus states plainly that “there are no quantitative criteria for the diagnosis of SSA”, then supplies one anyway, explicitly labelled pragmatic:

two of the four most important diagnostic features, present in at least two different crypts — and the crypts need not be adjacent.

The four are hyperserration into the lower third, T-/L-shaped crypts above the muscularis mucosae, inverted crypts below it, and columnar dilatation of the crypt base. That is a counting rule converting continuous architectural variation into a binary diagnosis, and the paper extends it to the appendix without separate justification.

Four further rules the group’s report tooling would have to encode:

1. SSP is not a synonym for SSA. Sessile serrated polyp is reserved for the case where basal mucosa is absent from the specimen so the decisive features cannot be assessed — and the report must then say why a definite diagnosis was impossible. The paper is explicit that it “should not be used as a wastebasket diagnosis.”
2. SSA with dysplasia/IEN is preferred over mixed polyp for a dysplastic SSA, because the international “mixed polyp” usage conflates adenoma-like (top–down) IEN with serrated dysplasia (bottom–up), which the group holds to be biologically different.
3. “Mixed polyp” is retained but must follow its components — e.g. sessile serrated and tubulovillous adenoma (mixed polyp) with low-grade IEN.
4. Three terms are banned outright: “serrated polyp with abnormal proliferation”, “sawtooth polyp”, and “SSA and HP (mixed polyp)”.

Two negatives worth recording because they close off shortcuts: subclassifying HPs (microvesicular / goblet cell / mucin poor) is not recommended in routine practice, and MUC6 is not advised as an HP-versus-SSA discriminator.

Management, from the S3 guideline: HP no follow-up; SSA, TSA and mixed polyp all 3 years; incomplete or piecemeal resection brings that to 2–6 months.

Methodological problems

Numbered so they can be referred to later. None of these makes the criteria wrong — they bound what can be claimed from them.

1. The evidence base is 19 selected cases at a one-day meeting. “Selected” is doing substantial work: cases chosen to illustrate criteria cannot also test them, because the ambiguous lesions that break a rule are exactly what a teaching set excludes.
2. The reproducibility claim carries no number. The criteria were used “with good interobserver concordance” — no kappa, no reader count, no design, no confidence interval. The paper concedes in the same sentence that this “has to be validated in future prospective studies”. This is the single most consequential gap: everything downstream inherits an agreement ceiling nobody has measured. See Interobserver Agreement.
3. The 2-of-4-in-2-crypts threshold is undermotivated. No derivation, no sensitivity or specificity, no comparison against 1-of-4 or 3-of-4, no ROC. It is a committee’s choice, and the paper says so. Treating it as validated because it is published would be the optimal-cutpoint error in Biomarker Cut Points arriving without even a dataset behind it.
4. Asymmetric reporting of the MUC6 evidence. Specificity is given as 82%; sensitivity is described only as “relatively low” and never stated. A recommendation against a marker is still a recommendation, and it should rest on two numbers.
5. Prevalence figures have no denominator. HP 80–90%, SSA 15–20%, TSA 1–6% — of what series, from which population, screened or symptomatic, is not given.
6. The progression claims rest on the weakest available designs. The 8-month figure is a case report; the 30%-of-sporadic-CRC figure is attributed to “some authors”. Both are cited accurately as such by the paper, and both are the kind of number that loses its qualifier when quoted onward.
7. No inter-country reconciliation. It is explicitly a nomenclature for German-speaking countries, and it argues against “sessile serrated lesion” while acknowledging the term is spreading. Whether SSA or SSL prevailed internationally is not something this paper can settle and is not traceable to any file in sources/ — so anyone using these criteria today must check them against the current WHO classification first. [unverified]

Why it earns space here

Not for the histology, which the reader knows. Three reasons, in order of what they change:

1. It is a worked example of where a diagnostic criterion actually comes from, and the answer is uncomfortable: a one-day meeting, a small teaching set, a threshold chosen for tractability, and a reproducibility claim asserted rather than measured. Interobserver Agreement argues that agreement is the ceiling on any reference standard; this is that ceiling being set, in public, without being measured.
2. It is directly encodable, and the group already has the tool. pathology-report-colon-QA in Pathology Report Checker Skills does narrative-to-synoptic gap analysis for colon resections. The four rules above — SSP only when basal mucosa is absent and with a stated reason, SSA-with-IEN over mixed polyp, components before the term, three banned strings — are mechanical string-and-structure checks, not judgement calls. The banned-terms list in particular is the cheapest possible report check.
3. It shows what a synoptic field cannot carry. Synoptic Reporting treats a structured field as unambiguous once filled. Here the field value SSA depends on a four-feature count across two crypts and on whether the specimen even contains basal mucosa. A checkbox records the conclusion and destroys the evidence for it — which is an argument for capturing the decisive features alongside the diagnosis, not only the diagnosis.

Related: Interobserver Agreement — the unmeasured concordance claim is the clearest example in this repository of a criteria set published with its agreement ceiling undefined. Related: Biomarker Cut Points — “two of four features in two crypts” is a categorisation threshold with no derivation, which is the same failure the optimal-cutpoint literature warns about. Related: Synoptic Reporting — the nomenclature rules here are exactly what a structured colorectal protocol has to encode, and the SSP caveat is a case where the structured field hides its own uncertainty. Related: Pathology Report Checker Skills — the group’s colon report QA tool is where the banned terms and the components-before-the-term rule would be enforced mechanically.