胆管細胞診の正診率は？

下部胆管狭窄の患者さんのERCPにて、胆汁および胆管擦過細胞診にてClass Vが出た。臨床経過その他より、原発性胆汁性胆管炎もしくはIgG4関連硬化性胆管炎だと思っていたのでびっくり。今後は外科コンサルトになるが、胆管細胞診がどれくらい当てになるのか調べてみた。日本語の文献で「ERCPにおける胆汁細胞診の正診率は約60％」という記述があった。結構低い。Hepatologyが参考文献に挙げられていたので、サマリーだけ読んでみた。

Kurzawinski TR et.al, A prospective study of biliary cytology in 100 patients with bile duct strictures., Hepatology. 1993 Dec;18(6):1399-403

In patients with obstructive jaundice due to biliary tract stricture a tissue diagnosis is essential because of the varied treatment options available. Radiological imaging of a biliary stricture may suggest that it is malignant, but only a tissue diagnosis can be conclusive. The difficulty of obtaining biopsy tissue has encouraged the use of cytology in this field. This study prospectively analyzed the diagnostic value of exfoliative bile and brush cytology methods. One hundred consecutive patients with biliary strictures diagnosed at endoscopic retrograde cholangiopancreatography or percutaneous transhepatic cholangiography (60 men and 40 women; median age = 71 yr, range = 31 to 91 yr) underwent biliary cytology and were divided into two groups. Group 1 comprised the first 47 patients, who were studied by means of bile cytology alone; and group 2 comprised the subsequent 46 patients, who were studied by means of bile and brush cytology techniques. Seven patients were excluded from analysis because of inadequate follow-up information. A single experienced cytologist examined all samples to determine whether they were neoplastic. Eighty-one patients had malignant strictures and 12 had benign strictures. Combined bile and brush cytology (group 2) was more sensitive than bile cytology alone (group 1) (69% [27 of 39] vs. 33% [16 of 42], p < 0.01). In the patients studied by means of bile and brush cytology methods (group 2), cytologic study of brushings was more sensitive (69% vs. 26%, p < 0.01). No false-positive results were reported in either group (specificity = 100%).

胆汁細胞診のみ（47人） v.s. 胆汁細胞診＋胆管擦過細胞診（46人）の比較。タイトルは100人となっているのは、7人が不適切なフォローアップ情報のため除外されたから。あたりまえの話だが、胆管擦過細胞診を組み合わせたほうがより正しく診断できる。胆汁細胞診のみ47人のうち、42人が悪性、5人が良性だった。悪性の42人のうち、診断が正しかったのは16人（16/42=33％）。残りは見落とされた。胆汁細胞診＋胆管擦過細胞診46人のうち、39人が悪性、7人が良性だった。その39人の悪性のうち診断が正しかったのは27人（27/39=69％）。悪性と診断された人で実は良性だった（false positive）というのはゼロ。

正診率は、胆汁細胞診のみで(16+5)/47=54%、胆汁細胞診＋胆管擦過細胞診で(27+7)/46=74%になるような気がする。いずれにせよ、偽陽性ゼロであれば、今回のClass Vは信頼してよいということになろう。ただ、検査技師の腕や考え方にも左右されるであろうが。