The diagnostic accuracy of cervical smears lacking a transformation zone component has been discussed for over 20 years.1–3 Some studies have reported that smears without a transformation zone component are more likely to miss abnormal cells.4–8 However, others claim that the adequacy of cytologic sampling is not guaranteed by the presence of endocervical or squamous metaplastic cells9,10 because smears with a transformation zone component may miss a lesion that is present, and those with an inadequate component may nevertheless be positive for neoplastic cells. This view is supported by a number of longitudinal studies, which have reported no difference in frequency of abnormal smear diagnoses according to the presence or absence of a transformation zone component.9,11
In 1996, the US Food and Drug Administration cleared the ThinPrep Papanicolaou test (Cytyc Corp., Boxborough, MA) as a replacement for conventional Papanicolaou smears. Although the sensitivity of liquid-based smears is improved compared with the conventional method of Papanicolaou smear preparation, there has been some suggestion that ThinPrep smears may be more likely than conventional smears to miss an endocervical or metaplastic cell component.12,13 Whether this characteristic affects the ability to detect cervical abnormalities has not yet been examined.
A screening study was undertaken to evaluate whether the absence of a transformation zone component in ThinPrep smears limits the ability to detect high-grade cervical lesions. Results from screening ThinPrep smears were compared with histology for confirmation of diagnoses.
MATERIALS AND METHODS
The population consisted of women attending three Planned Parenthood of Western Washington clinics between 1997 and 2001. Informed consent was obtained from all study participants. The study protocol was approved by the Institutional Review Board of the University of Washington. Women who were pregnant, chronically immune impaired, without a cervix, or had prior treatment for cervical neoplasia (including electrocautery, cryocautery, conization, laser surgery, or loop electrosurgical excision procedure) were not eligible to participate.
Papanicolaou smears from the cervix and samples for human papillomavirus (HPV) testing were obtained at the screening visit according to the following protocol: an Ayres spatula was used for transformation zone cell collection, and a cytobrush was used to sample the endocervix. Both the brush and the spatula were then directly rinsed in vials containing 20 mL of cytology fluid (PreservCyt, Cytyc Corp., Marlborough, MA) for processing the final ThinPrep cytology slide. An HPV DNA cervical specimen was obtained after the cytology specimen. A Dacron-tipped swab was inserted into the cervical os, rotated, withdrawn, and then swabbed on the ectocervical epithelium. This swab was then placed in Specimen Transport Medium (Digene diagnostics, Silver Spring, MD) for transportation to the laboratory for HPV DNA testing.
Women with 1) atypical squamous cells of undetermined significance (ASCUS), low-grade or high-grade squamous intraepithelial lesions (LSIL or HSIL) on Papanicolaou smear, or 2) an HPV DNA test positive by polymerase chain reaction for high-risk types (16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 55, 56, 58, 59, 68, 73, 82, and 83) were referred for colposcopy and biopsy. In addition, a random sample of women with normal Papanicolaou smears and negative HPV DNA tests were also referred for colposcopy and biopsy. Eighteen women with HSIL (n = 128) did not return to the study clinic for colposcopy and biopsy. All were phoned repeatedly and sent certified letters. The median interval between the screening visit and the colposcopy visit was 60 days (range 13–991 days).
At the colposcopy visit, detailed questionnaires were completed, and cervical samples for ThinPrep cytology slides and HPV testing were repeated using the same collection technique as carried out at the screening visit. Colposcopies were performed by a nurse practitioner with extensive training in colposcopy, or by a board-certified gynecologist, both of whom had at least 7 years of clinical practice in colposcopy. The DenVu computerized colposcopy system (DenVu, Tucson, AZ) was used to record information and pictures of the cervix. Biopsies of the worst visible lesions were obtained using a Wittner or Tessler biopsy forceps; women with negative cytology and colposcopy results underwent colposcopically directed biopsy at a 12 o'clock site. Endocervical curettage was performed if the colposcopy was unsatisfactory or if the lesion(s) seen did not correlate with the Papanicolaou smear results (n = 662). Women with histologically confirmed cervical intraepithelial neoplasia grade 2 to 3 (CIN 2–3) or invasive cervical cancer were referred for treatment. Patients with CIN 1 or cervical atypia on histology returned to Planned Parenthood for a repeat Papanicolaou smear in 4–6 months with follow-up care provided by Planned Parenthood.
Specimens were processed at Harborview Medical Center, University of Washington, by cytotechnologists who received training at Cytyc (Boston, MA). The HPV DNA was detected from the cervical swab sample in Specimen Transport Medium after the sample was digested with protease and the DNA was ethanol precipitated. The HPV L1 consensus primer (MY09/MY11 and HMB01) polymerase chain reaction amplification assay and Roche reverse line blot hybridization were used for amplification and typing of HPV DNA,14 and the primers PC04 and GH20 were used for β-globin detection. One-fiftieth of the original 1-mL sample was amplified by polymerase chain reaction. The PreservCyt vial was loaded into the Cytyc ThinPrep Processor (ThinPrep 2000) along with a glass microscope slide, filter cylinder, and vial containing a fixative solution and processed according to the manufacturer's specifications. The processed slide preparation was then stained with the Papanicolaou stain, screened by a cytotechnologist, and reviewed by pathologists at Harborview Medical Center, without knowledge of other laboratory or clinical data.
The presence of glandular epithelium and metaplastic epithelial cells was recorded to serve as an indicator of the adequacy of the cervical samples. Smears with no glandular cells present, too few cells present (less than 100), obscuring inflammation, or air-drying artifact were regarded as unsatisfactory. Smears were considered limited if endocervical cells or at least ten metaplastic cells were absent. Based on these criteria, 78.8% of the total screening smears were satisfactory, 19.3% were satisfactory but limited, and 1.9% were unsatisfactory. Unsatisfactory smears were not significantly more likely to have been limited by a transformation zone component (21.4%) compared with satisfactory but limited smears (15.6%). Women with an unsatisfactory smear were asked to return for a repeat smear; the interval to the repeat smear ranged from 17 to 212 days (median 54 days). The proportion of repeat smears that were limited because of a transformation zone component was similar to that at the initial screening. Therefore, data from the repeat screening visit were used for women whose initial smear was unsatisfactory, excluding 60 women with an unsatisfactory smear who did not return for a repeat smear or whose specimen adequacy was not noted. Of the remaining 4389 women who were screened, 1086 (24.7%) women attended the colposcopy visit and had a cervical biopsy.
Cytologic findings were recorded in terms of the Bethesda System classification,15 which defined ASCUS as mild nuclear enlargement or dyskeratosis including parakeratosis and hyperkeratosis, whereas LSIL included koilocytotic atypia and mild dysplasia, and HSIL included moderate and severe dysplasia and carcinoma in situ. A random 10% sample of all slides read out as normal were rescreened manually as mandated by Federal law (Clinical Laboratory Improvement Amendments).
Hemotoxylin and eosin-stained slides of biopsy specimens were prepared and reviewed by the pathologist without knowledge of other clinical or laboratory data. Specimens were classified as negative, atypia, CIN 1, 2, or 3, and microinvasive or invasive cancer. The highest histologic diagnosis from either the cervical biopsy or the endocervical curettage was used for analysis.
Patient characteristics, including age, current use of oral contraceptives (OCs), parity, Papanicolaou smear history, date of last menstrual period, and use of douching or vaginal medications were available from a questionnaire administered at the screening visit.
As χ2 analysis was used to test for significant differences between proportions. Mean and median values and standard deviations were calculated for continuous variables. Univariate and multivariate logistic regression analyses were performed to estimate odds ratios (ORs) and 95% confidence intervals (CIs) (SAS 6.12, SAS Institute Inc., Cary, NC).
During the first 4 years of the study, 4389 women aged 18–50 were screened (mean age 25.2 years). Overall, 700 smears (15.9%) lacked a transformation zone component. The possibility that patient's age, current OC use, abnormal Papanicolaou history, time since last menstrual period, gravidity, lifetime number of male sex partners, and douching or vaginal medication use might influence the collection of cells from the transformation zone was considered (Table 1). The percentage of transformation zone-negative smears increased with age (χ2 test for trend, P < .001). Compared with women aged 20 years and younger, the likelihood of having a transformation zone-negative smear was greatest among women older than 30 years (OR 1.5, CI 1.1, 1.9). In addition, women with transformation zone-negative smears were more likely to be current OC users (45.9%) compared with women with transformation zone-positive smears (34.0%) (OR 1.7, CI 1.4, 2.0), and also more likely to be past the 14th day of their last menstrual period (64.0% versus 53.9%, respectively) (OR 1.6, CI 1.3, 1.9). Other demographic factors were not significantly associated with transformation zone status. Women with a transformation zone-positive smear were more likely to have a high-risk HPV type (19.7%) at the screening visit compared with those with a transformation zone-negative smear (13.7%) (OR 1.6, CI 1.3, 2.0).
Table 2 relates the presence or absence of a transformation zone component to the cytologic diagnosis of cervical abnormalities at the screening visit. The frequency of an abnormal cytology diagnosis was significantly higher in transformation zone-positive smears (18.3%) compared with transformation zone-negative smears (10.7%) (OR 1.9, CI 1.4, 2.4). Adjusting for the presence of atypical metaplastic cells, smears lacking a transformation zone component were 1.4 times more likely to be diagnosed as ASCUS (CI 1.0, 1.9). There was a trend of increasing disparity in diagnoses with increasing grades of dysplasia (χ2 test for trend = 10.0, P = .001). Transformation zone-positive smears were more likely to have a cytologic diagnosis of LSIL (OR 2.5, CI 1.5, 4.2) and HSIL or higher (OR 5.1, CI 2.1, 16.1) compared with transformation zone-negative smears. The relationship between presence of a transformation zone component and abnormal cytology was confounded only by timing in the menstrual cycle and presence of high-risk HPV DNA (see Table 2 for adjusted estimates).
Cytology diagnoses at the colposcopy visit were reviewed according to the presence or absence of a transformation zone component and cytology diagnoses at the screening visit. Among women with normal cytology at screening, a diagnosis of ASCUS or higher (OR 1.3, CI 0.7, 2.7), LSIL or higher (OR 1.3, CI 0.5, 3.2), or HSIL or higher (OR 1.5, CI 0.3, 6.5) at the colposcopy visit was not significantly higher among transformation zone-positive compared with zone-negative smears. Adjusted for OC use, the association between transformation zone status and a diagnosis of HSIL or higher was elevated (OR 1.8, CI 0.4, 8.0).
Within each category of cytology diagnoses at screening (ie, normal, ASCUS, LSIL, HSIL or higher), follow-up attendance for biopsy did not differ significantly by transformation zone status. Overall, the colposcopically directed cervical biopsies and/or endocervical curettage revealed atypia in 150 (13.9%) women, CIN 1 in 191 (17.7%) women, CIN 2 in 53 (4.9%) women, and CIN 3 or higher in 98 (9.1%) women. A biopsy diagnosis of CIN 2 or higher was observed in 14.8% (140 of 946) of transformation zone-positive smears and 8.3% (11 of 133) of transformation zone-negative smears at the screening visit (OR 1.9, CI 1.0, 3.6) (Figure 1). This estimate was not confounded by other factors.
Among the subset of women with normal cytology at screening, 17.7% (14 of 79) of transformation zone-negative smears at screening compared with 9.8% (45 of 460) of transformation zone-positive smears had CIN 1 detected by biopsy at the colposcopy visit (OR 2.0, CI 1.0, 3.8) (Figure 1). A biopsy diagnosis of CIN 2 or higher was not significantly higher among women whose screening smears were transformation zone positive (7.8%) versus negative (6.3%) (OR 1.3, CI 0.5, 3.3). Among women with a smear diagnosed as ASCUS or LSIL at screening, a diagnosis of CIN 2 or higher was observed among 11.7% of transformation zone-positive screening smears compared with 10.0% of transformation zone-negative smears (OR 1.2, CI 0.5, 3.2). The likelihood of a screening smear missing a biopsy-detected lesion grade 2 or higher did not differ significantly between screening smears without (7.5%) versus with (8.5%) a transformation zone component (OR 0.9, CI 0.4, 1.7).
Several studies have reported that, in comparison to a conventional Papanicolaou smear, the ThinPrep Papanicolaou test is more likely to be classified as “satisfactory but limited” because of the absence of a transformation zone component, with estimates of this finding ranging from 9.8%12 to 20.0%.13 In our study, 15.9% of screening ThinPrep smears lacked a transformation zone component. However, absence of a transformation zone component did not compromise cytologic-based detection of biopsy-confirmed CIN.
Two main approaches have been used to determine whether conventional Papanicolaou smears limited by a transformation zone component are less diagnostic of cervical abnormalities compared with smears with adequate endocervical and/or metaplastic cells. One approach has been to compare the frequency of cytologic abnormalities in smears with versus without these cells. According to this strategy, we found that smears limited by a transformation zone component were more likely than their counterparts to be categorized as cytologically normal. Similarly designed studies have reported comparable results.4–7 We also observed an increasing disparity in frequency of abnormalities between the two groups with higher grades of dysplasia. This finding has been reported previously,4,5 but the origins of the disparity are not well understood. Accounting for potential differential misclassification of atypical metaplastic cells as ASCUS rather than higher-grade dysplasia did not explain the difference between groups in our study.
Another approach has been to compare subsequent cytology and/or histology findings according to the presence or absence of a transformation zone component at screening. In our analysis, we restricted evaluation of follow-up cytology and histology to women whose first smear was diagnosed as normal. The approach has merit over the first cross-sectional strategy because if specimens are obtained shortly after the initial smear, it is possible to assess whether abnormalities present on later smears represent missed lesions at initial screening, rather than incident lesions or progression of dysplasia. Furthermore, this is an important subgroup to analyze because women who have a falsely normal test result may not be reexamined for another year or more, by which time their lesion could progress such that it becomes more difficult to treat.
We found that among women with a normal screening smear, abnormal cytology at follow-up and missed high-grade lesions (as defined by the presence of biopsy-confirmed CIN 2–3 at the colposcopy visit) were not significantly more likely among those without versus with a transformation zone component at screening. Similarly, among women who had ASCUS or LSIL at screening, the presence of a biopsy-confirmed lesion of CIN 2 or higher did not differ significantly according to transformation zone status at screening. These findings are consistent with several studies of conventional Papanicolaou smears. Kivlahan and Ingram16 found that among women who had a normal initial smear, those without (n = 2532) versus with (n = 3443) endocervical cells were not significantly more likely to have cytologically detected atypia on the second screening (OR 1.1, CI 0.8, 1.5). Likewise, in a study of 20,222 women with negative cytology reports during 1987, the incidence of definite or equivocal cytologically detected CIN at a second smear did not differ according to the presence or absence of an endocervical component on the first smear.17 Vooijs et al7 reported that women with an initial smear lacking endocervical cells were more likely to have cytologically detected dysplasia on a later smear. However, the interpretation of the data is difficult because 3 years elapsed between the first and second smears, the follow-up analysis did not exclude initial abnormal smears, and the authors did not report the number of patients treated after finding the initial abnormalities. Therefore, it is not possible to distinguish whether the excess risk was attributed to missed lesions, incident lesions, or progression of dysplasia.
To our knowledge, there has been only one well-designed, large follow-up study that has compared detection of histologically proven CIN 2–3 and cancer according to the presence or absence of a transformation zone component on an earlier smear.18 A retrospective analysis of data from women in the Netherlands who obtained a cervical smear for preventive reasons between 1990 and 1991 and had a normal cytology result (n = 515,146 smears) was performed. Those who had abnormal smears in the previous 4 years were excluded. Women were followed-up for 6.25–8.25 years, during which time the incidence of CIN 2 or higher did not differ according to the presence or absence of endocervical cells on the initial smear (OR 1.03, 95% CI 0.92, 1.15). Similarly, the risk of an invasive cervical carcinoma within 8 years after a cytologically normal smear without endocervical cells was equal to the risk after a normal smear with endocervical cells (OR 1.01, 95% CI 0.68, 1.49). These associations were not confounded by age or length of the screening interval.
To date, most other studies with biopsy follow-up have been conducted among select populations (eg, specialized referral clinics), have not examined CIN 2–3 specifically, or did not restrict the analysis to the subgroup of women with normal screening smears. The available evidence from these studies suggests that the absence of a transformation zone component does not compromise the adequacy of conventional Papanicolaou smears. Sidawy et al11 examined 84 patients (aged 15–50 years, 136 smears total) who were referred to a colposcopy clinic on the basis of condyloma, dysplasia, or persistent inflammatory atypia, and were found to have histologic evidence of dysplasia or carcinoma. Discrepancies between cytologic diagnoses from the referral smears and histologic diagnoses at colposcopy did not differ significantly according to whether an endocervical component was present (18.4%) or absent (20%) on the referral smear. Furthermore, of the 13 undercalled cases of HSIL and microinvasive carcinoma, only one involved a smear without an endocervical component; the four negative referral smears with biopsy-detected LSIL and two negative smears with biopsy-detected HSIL were divided equally according to the presence or absence of an endocervical component.
In the present study, we found no association between the presence of a transformation zone component at screening and gravidity, history of an abnormal Papanicolaou smear, or use of vaginal medications or douches 24 hours before the examination. However, we did find that absence of a transformation zone component was associated with current OC use and timing of the menstrual cycle. These findings are consistent with those of Vooijs et al,19 where better quality smears were obtained during the first half of the menstrual cycle compared with the second half, and smears without endocervical columnar cells were more common among OC users compared with nonusers (OR 2.6, CI 2.5, 2.8). We also found that age was inversely associated with the presence of a transformation zone component. Mauney et al5 similarly reported a trend of decreasing endocervical cells with increasing age, although the decline in the percentage of smears lacking these cells was gradual and was not apparent until after age 55 (the age range included was 14–85 years). Bos et al18 also reported a decline with increasing age (17.5% of women 65 years or older had smears lacking endocervical cells), although those younger than 25 years had a similarly high proportion of smears lacking endocervical cells (18.4%). Elias et al4 found no decline in the presence of endocervical cells through age 55, although the study did not examine women younger than age 35. In our study, we found a higher frequency of smears with a transformation zone component among women who were positive for high-risk HPV types. The association remained after adjusting for potential confounders.
This study lacked adequate power to detect small differences in risk of high-grade lesions according to the presence or absence of a transformation zone component; therefore, caution needs to be taken when drawing conclusions about nonsignificant differences for these outcomes. However, our study is the largest to date to examine biopsy-confirmed CIN 2–3 among a group of women with normal ThinPrep Papanicolaou smear findings. The fact that our results are consistent with similarly designed studies of conventional Papanicolaou smears lends credibility to our findings.
Another limitation of the study relates to the generalizability of our results. Our study population consists of women aged 18–50 years, and therefore our findings do not extend to women of older age groups, who may be more likely to have smears lacking a transformation zone component. However, given the low incidence of dysplasia in postmenopausal women, this limitation is of minimal significance. Also, unlike the referral criteria used in most clinical settings, women in this study were referred to colposcopy if they had an abnormal smear diagnosis, tested positive for high-risk HPV types, or were among a random sample who were cytologically normal and negative for high-risk HPV. These sampling criteria enabled us to identify the women with CIN 2–3 who may not otherwise have undergone colposcopy and biopsy. The experience of the practitioners who performed the colposcopies, along with that of the cytopathologists and cytotechnologists, lends further validity to our results.
Finally, there are many different designs of cervical smear collection devices, and numerous studies have examined the relationship between these designs and rates of inadequate smears. In this study, we collected endocervical samples using a combination of spatula and cytobrush. A systematic review and meta-analysis suggest that this combination is superior to other devices for collecting endocervical cells.20 Therefore, although the possibility of nondifferential misclassification of transformation zone status exists, the tools we used for endocervical sampling are the best currently available.
In sum, although a number of studies have suggested that liquid-based smears are more likely to lack a transformation zone component compared with conventional Papanicolaou smears, this quality does not appear to compromise cytologic detection of biopsy-confirmed cervical abnormalities. If smears without a transformation zone component were more likely to miss important abnormalities, then the frequency of CIN 2–3 should be highest in follow-up smears or biopsies among women whose first smear lacked these cells. In our study, the frequency of CIN 2–3 did not differ significantly according to the presence or absence of a transformation zone component at screening. Based upon our data, we do not recommend repeat screening of reproductive-aged women with negative liquid-based tests and no cytologic evidence of a transformation zone component. If our findings are confirmed by future studies, a definition of smear adequacy specific to the ThinPrep method should be developed. Such a revision could limit the effort and expense caused by the need to repeat smears, thereby improving the cost-effectiveness of screening as well as minimizing the psychologic cost of repeating smears with an uncertain diagnosis.