Design: Systematic review and meta-analysis of 35 immunohistochemistry studies
Population: 1,146 histological and 35 cytological salivary gland tumour samples
Primary outcome: Pooled SOX10 positivity by tumour type
Effect: Pleomorphic adenoma 95.8%, adenoid cystic 96.6%; Warthin 2.3%, salivary duct 3.7%

This meta-analysis pooled 35 studies of SOX10 immunohistochemistry in 1,146 histological and 35 cytological samples of primary salivary gland tumours.

Pooled positivity was above 90% for pleomorphic adenoma (96%), canalicular adenoma (99%), basal cell adenoma (91%), adenoid cystic carcinoma (97%), acinic cell carcinoma (97%), polymorphous adenocarcinoma (97%), microsecretory carcinoma (96%) and intercalated duct-type intraductal carcinoma (91%). Myoepithelioma was less consistent (75%). Warthin tumour (2%), salivary duct carcinoma (4%), oncocytoma (8%) and apocrine intraductal carcinoma (10%) were essentially negative. Polyclonal antibodies and clones EP268, N-20 and A-2 performed alike; clone BC34 stained fewer positive tumours.

In practice SOX10 is most useful as a split between broadly acinar or intercalated-duct and myoepithelial lineages, which stain, and oncocytic or apocrine-type lesions, which do not. It does not distinguish between the positive entities, so it belongs in a panel, not on its own. Note the clone in your report; a negative result with BC34 means less than one with another clone.

Use SOX10 to separate SOX10-positive tumours (pleomorphic adenoma, adenoid cystic, acinic cell) from oncocytic or apocrine-type lesions.
A SOX10-positive oncocytic-looking tumour should prompt a second look before calling it oncocytoma or Warthin.
A SOX10-negative high-grade carcinoma with apocrine features favours salivary duct carcinoma; confirm with AR.
Record the antibody clone in the report — BC34 stains fewer tumours than other clones.
Do not rely on SOX10 to tell positive entities apart; combine it with morphology and other markers.

Why it matters

A pooled list replaces scattered series, and the clone difference explains some conflicting results.

The statistics, in plain English

Some confidence intervals run above 100% or below 0% — an artefact of the meta-analysis method with small studies and near-100% or near-0% results. Read them as 'nearly always' or 'nearly never'. Myoepithelioma's wide interval (52% to 98%) means its SOX10 status is genuinely variable.