The colposcopic impression: Is it influenced by the colposcopist's knowledge of the findings on the referral Papanicolaou smear?

OBJECTIVE: To determine whether the colposcopic impression is influenced by the colposcopist's knowledge of the referral Papanicolaou smear. STUDY DESIGN: Using a community hospital database, the accuracy of the colposcopic impression (accuracy=proportion of women with histology greater than cervical intraepithelial neoplasia [CIN] 2 that have colposcopic impressions of greater than CIN 2) when referral smears were atypical squamous cells of uncertain significance (ASCUS), atypical glandular cells of uncertain significance (AGUS) or low grade squamous intraepithelial lesion (LSIL) was compared to that when smears showed high grade squamous intraepithelial lesion (HSIL) or cancer. The analysis was repeated with a screening study database in which colposcopic impression was assigned without knowledge of the Papanicolaou smear. Univariate and logistic regression analysis of the second database determined the relative importance of size and grade of lesion and Papanicolaou result to the accuracy of the colposcopic impression. RESULTS: In the community database, colposcopic accuracy was 60/510 (12%) when smears were ASCUS, AGUS or LSIL and 77/132 (58%) when smears were HSIL or cancer (P<.001); in the second database, it was 2/19 (11%) when smears subsequently were reported as negative, ASCUS, AGUS or LSIL and 33/65 (54%) when smears were HSIL or cancer (P<.005). An accurate colposcopic impression was seen in 5/39 (13%) women with one-quadrant lesions, 8/18 (44%) with two-quadrant lesions and 23/27 (85%) with three- or four-quadrant lesions (P<.O05). None of 19 women with smears reported as negative, ASCUS, AGUS or LSIL had lesions involving three or foud quadrants of the cervix, while 27/65 (42%) women with smears reported as HSIL or cancer had such lesions (P<.005). With logistic regression, the more quadrants of the cervix involved, the more accurate the colposcopic impression. Once controlled for lesion size, there was no improvement when worst histologic grade or Papanicolaou smear result was considered. CONCLUSION: Through lesions greater than CIN 2 were more often overlooked when referral smears were negative, ASCUS, AGUS or LSIL than when they were HSIL or cancer, the real reason that the lesions were not detected by colposcopy was that they were small.