Background:lt has been suggested that 50 endosonographic examinations
are necessary to attain sufficient skill to become accurate in the sta
ging of esophageal carcinoma. Methods: Since July 1991 all esophageal
endosonographies have been performed by one investigator. The database
, containing all examinations since July 1991, was reviewed to investi
gate the presence of a learning curve in T-staging of esophageal carci
noma by comparing the endosonographic T-stage with the pathologic T-st
age. Results: From July 1991 until April 1993, 231 endosonographies we
re performed for esophageal malignancies. Eighty-nine patients had bee
n referred for preoperative staging and underwent surgical exploration
. Seventy-one tumors were resected. The patients were split in two gro
ups. Group I consisted of 36 patients examined between July 1991 and A
pril 1992. In this period 100 endosonographies were performed. Group I
I consisted of 35 patients examined between April 1992 and March 1993.
In this period 131 endosonographies were performed. Overall endosonog
raphic T-staging was correct in 70%. The accuracy in group I was 58%;
in group II 83% (p < 0.05). Overstaging was equally common in groups I
and II (14%). Understaging was common in group I (28%), but rare in g
roup II (3%). Conclusion: This study shows a definite learning curve f
or endosonographic T-staging of esophageal carcinoma. Acceptable accur
acy rates could only be achieved after 100 examinations.