Objective
We sought to determine the validity of colposcopically directed cervical biopsies as a diagnostic test to define the degree of cervical intraepithelial neoplasia (CIN).
Study Design
In a prospective multicenter trial, patients undergoing excisional procedures of the transformation zone additionally had colposcopy and up to 3 guided cervical biopsies in a single procedure. Cervical biopsies were regarded as a diagnostic test to detect high-grade lesions (CIN 2,3), with the cone specimen as reference standard.
Results
In all, 488 biopsies were performed in 244 cases, with 2 biopsies done in 192 cases. Cervical biopsies underestimated the severity of lesions in 46.7% of cases. Sensitivity, specificity, and positive and negative predictive values were 66.2% (95% confidence interval [CI], 59.4–72.3), 95.0% (95% CI, 83.5–98.6), 98.5% (95% CI, 94.8–99.6), and 35.5% (95% CI, 27.1–44.9), respectively.
Conclusion
Our data suggest that cytologically suspected high-grade lesions (CIN 2,3) can be confirmed by biopsy in many cases, but they cannot be excluded.
Diagnosing the grade of cervical intraepithelial neoplasia (CIN) by colposcopically directed biopsies is a widely used method in Western countries. While confidence in the validity of cervical biopsies is generally high, contradictory results are continually discussed in literature. These articles compare cervical biopsies with outcomes of the reference standard, histopathology of cone specimens or other surgical material obtained subsequently.
In 1980, Kirkup and Hill reported an excellent correlation between histological diagnosis from colposcopically directed punch biopsies and definitive diagnosis after conization or hysterectomy. However, many subsequent studies showed only moderate correlation. Significant discrepancies generated concern. Only a few authors concluded that directed cervical biopsies provide a consistent estimate of the final grade of CIN.
Overestimation may cause overtreatment, and excisional methods to remove the transformation zone are associated with an increased risk of pregnancy-related morbidity such as preterm delivery, low birthweight, cesarean section, and premature rupture of the membranes. Cold knife conization (CKC) may be a risk factor for perinatal mortality.
Underestimation, on the other hand, may have even more serious implications. Some undetected high-grade lesions may ultimately progress to invasive cervical carcinoma during the observation period. Therefore, defining extent and severity of intraepithelial lesions remains an important clinical problem. One of 4 patients with mild dysplasia on directed biopsy may have high-grade lesions found in excision specimens. These issues highlight the risk of a false-negative biopsy, and challenge the established role of guided cervical biopsies, while at the same time numerous studies use diagnosis of cervical biopsies as an endpoint. Therefore, deficiencies of the guided biopsy need to be addressed. Influencing factors need to be analyzed.
In the present study we correlate the degree of CIN found on cone specimen with the degree of CIN on colposcopically guided cervical biopsies. Since the severity of biopsy results leads to the appropriate treatment, the aim of this study was to show if there are diagnostic limitations of cervical biopsies that should influence our clinical decision-making. If women with cytology indicating high-grade squamous intraepithelial lesions (HSIL) undergo colposcopy, a significant number of CIN 2,3 lesions are missed. These are cases of special interest and need further evaluation.
As the time span between 2 diagnostic tests allows for both regression and progression of lesions, we performed cervical biopsies as an adjunct at the time of conization.
Materials and Methods
Participants
In a prospective multicenter trial, 2 tertiary care centers and 1 county hospital in Austria recruited consecutive nonpregnant patients scheduled for excisional procedures of the transformation zone. Patients were grouped by whether the indicating disease (either cytologic or histologic) was high grade (indications with preoperative evidence of high-grade disease) and those in whom neither the indicating cytologic nor histologic studies were high grade and at least 1 of them was low grade (indications with preoperative evidence of low-grade disease). Patients with cervical Papanicolaou smears classified as atypical squamous cells of undetermined significance were not included in the study, unless a preoperative biopsy revealed CIN. Patients with suspected cervical cancer were not included in the study either.
Women were included in the study if the entire squamocolumnar junction (SCJ) of the cervix was visible (satisfactory colposcopy) or with part of the SCJ visible. If the SCJ could not be visualized, women were not included in the study.
Procedures were also grouped by whether women recently underwent cervical biopsies prior to the study, and those who did not. Since most of the indications for conization are biopsy proven, it was obvious that patients with prior biopsies had to be included in the study. But for the purpose of this study we focused on recruiting patients without recent biopsies. Many of these patients were women with HSIL cytology following see-and-treat approach, or with cytology indicating low-grade disease following repeat cervical cytologic testing, showing persistence over ≥12 months.
The study was conducted after local ethics committee approval and written informed consent of all participants.
Test methods
Patients who underwent conization had colposcopy and guided cervical biopsies in a single procedure, performed immediately before conization. Biopsies were taken in the area subsequently excised with the cone specimen.
Investigators were gynecologic oncologists and certified obstetrician-gynecologists who had practiced colposcopy for at least 3 years. Residents were also accepted as investigators if they had successfully completed colposcopy training, and were under direct supervision of an experienced obstetrician-gynecologist.
After assessing the visibility of the SCJ, colposcopy was evaluated by describing the size of the lesion; sharp or diffuse margins and differences in surface levels; regular, absent, or atypical blood vessels; distinct or milky acetowhite lesions; as well as iodine staining with distinct or faintly yellow areas, and graded as described by Strander et al. In an effort to find influencing factors that contribute to inadequate accuracy of guided biopsies, the lesion characteristics of both concordant and discrepant cases were compared.
Uterine biopsy forceps with 5- to 6-mm jaws were used, yielding 3- to 4-mm biopsies. The first biopsy had to be taken from the site considered most suspicious. Then the examiners had to decide whether to take a second and third biopsy. Findings were noted on a colposcopy chart.
Conizations were performed using either CKC or loop electrosurgical excision procedure. Cone specimens were weighed prior to formalin fixation.
When correlating results found on guided biopsies and cone specimens, a number of studies evaluated the correlation of CIN “within 1 degree.” From a practical approach CIN 2, as opposed to CIN 1, is considered a high-grade lesion and will often initiate an excisional procedure. Therefore, we analyzed sensitivity, specificity, positive predictive value, and negative predictive value for guided cervical biopsies as a diagnostic test to detect high-grade lesions (ie, CIN 2,3), with the cone specimen as reference standard. For this analysis biopsies and cone specimens showing CIN 1 were treated as results without dysplasia.
Histology report
Biopsies were formalin fixed, paraffin embedded, and four 4 μm–thick serial sections were cut. Cone specimens were formalin fixed, paraffin embedded, and processed sagittally in 2 mm–thick sections. Slides of 4-μm thickness were cut at intervals of 250 μm and hematoxylin-eosin stained.
As reproducibility of cervical pathology specimens is reported to be only moderate, slides from the different centers were collected and 1 expert pathologist (M.H.) reviewed all histologic samples. These results were used for the present analysis.
Diagnosis of dysplasia was made according to the World Health Organization Classification of Tumors. Inflammatory and reactive changes were considered negative for the present analysis.
Statistical methods
Apart from standard methods of descriptive statistics (means, SDs, frequencies), data were analyzed by cross-tabulation, the χ 2 test, McNemar test, the t test, and Cohen Kappa statistic, using SPSS software (version 16; SPSS, Inc., Chicago, IL). Confidence intervals (CI) were calculated using the CIA software (BMJ Books, London, England) provided by Altman et al. In particular, the Wilson method was used to calculate CI for proportions, since standard methods are not applicable for proportions near 1. P values < .05 were considered statistically significant. Due to rounding, percentages in Table 1 may not add up to 100%.
| Cone specimen | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| CIN 0 | CIN 1 | CIN 2 | CIN 3 | Total | ||||||
| Guided cervical biopsies | n | % | n | % | n | % | n | % | n | % |
| CIN 0 | 12 | 19.4 | 15 | 24.2 | 12 | 19.4 | 23 | 37.1 | 62 | 100.0 |
| CIN 1 | 2 | 4.4 | 9 | 20.0 | 19 | 42.2 | 15 | 33.3 | 45 | 100.0 |
| CIN 2 | 1 | 1.9 | 1 | 1.9 | 21 | 39.6 | 30 | 56.6 | 53 | 100.0 |
| CIN 3 | 84 | 100.0 | 84 | 100.0 | ||||||
| Total | 15 | 6.1 | 25 | 10.2 | 52 | 21.3 | 152 | 62.3 | 244 | 100.0 |
Results
Between December 2005 and November 2007 we recruited 287 women ( Figure ). At the time of colposcopy 18 cases were not included because the SCJ was not visible. Among the 269 included cases we observed 11 cases where histopathology slides or colposcopy charts were not retrievable, 10 protocol violations (resident without supervision), 2 cases of cervical glandular intraepithelial neoplasia detected in the cone specimen, 1 case with fragmented biopsy specimen showing CIN that could not be graded, and 1 case without any biopsy that had to be excluded. Therefore, 244 cases met all criteria for analysis according to the study protocol.
