Early Detection and Diagnosis
Direct comparison of liquid-based and conventional cytology in a South African screening trial
Article first published online: 8 SEP 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 4, pages 957–962, 15 February 2006
How to Cite
Taylor, S., Kuhn, L., Dupree, W., Denny, L., De Souza, M. and Wright, T. C. (2006), Direct comparison of liquid-based and conventional cytology in a South African screening trial. Int. J. Cancer, 118: 957–962. doi: 10.1002/ijc.21434
- Issue published online: 13 DEC 2005
- Article first published online: 8 SEP 2005
- Manuscript Accepted: 7 JUN 2005
- Manuscript Received: 7 JAN 2005
- cervical cytology;
- papanicolaou (Pap) smear;
- liquid-based cytology;
- cervical intraepithelial neoplasia
Our study directly compares the performance of liquid-based (LBC) and conventional cytology for detecting high-grade cervical intraepithelial neoplasia and cancer (CIN 2+) in high-risk, previously unscreened women. As part of a larger randomized clinical trial assessing the efficacy and safety of a “screen and treat” program for cervical cancer prevention, 5,652 South African women, aged 35 to 65 years, were screened using either ThinPrep or conventional Papanicolaou cytology. The cytology method used (i.e., ThinPrep or conventional) was rotated on a 6-month basis for the duration of the study. Directly following collection of the cytology specimen, all women underwent colposcopy with endocervical curettage and biopsy of all colposcopic abnormalities. Assessment of cytology and histology results was blinded and results were compared using histology-confirmed CIN as the “gold standard.” The accuracy of LBC and conventional cytology was statistically equivalent, although the sensitivity of conventional cytology was at least 5 percentage points higher at all cutoff levels. For example, at a cytology cutoff level of low-grade squamous intraepithelial lesion, the sensitivity of LBC was 60.3% for CIN2+ vs. 69.1% for conventional cytology and specificity was 94.1% and 94.5%, respectively. LBC specimens were significantly less likely to be “satisfactory-but-limited-by” (6.5% vs. 27.9%) but significantly more likely to be unsatisfactory (2.2% vs. 0.8%). Thus, in this high-risk population, the sensitivity of LBC is no greater than the sensitivity of conventional cytology. Because of the higher unit cost of LBC, low resource settings should carefully consider the potential benefits and drawbacks of LBC before adopting this new technology. © 2005 Wiley-Liss, Inc.
Conventional cytology, i.e., the Papanicolaou smear, has been the mainstay of cervical cancer screening since the 1960s and is credited with successfully reducing the incidence and mortality of invasive cervical cancer in many developed countries.1, 2 Nonetheless, the accuracy of this important screening tool remains controversial, with several large meta-analyses suggesting that both the sensitivity and specificity of cervical cytology is relatively low (30% to 87% sensitivity, 86% to 100% specificity).3, 4, 5
As an alternative to conventional cytology, liquid based-cytology (LBC) was introduced in the mid-1990s. The purported advantages of LBC include a possible increase in the detection of high-grade cervical intraepithelial neoplasia (CIN2, 3), a reduction in the number of unsatisfactory and “satisfactory-but-limited-by” specimens, and providing residual cellular material for subsequent molecular testing (e.g., testing for “high-risk” types of HPV DNA). On the other hand, the unit cost of LBC methods is substantially more than that of conventional cytology. The substantial increase in unit cost is of particular concern in less-developed countries with limited resources to devote to cervical cancer screening programs and where the increased cost of LBC may not be justified unless substantial improvements in test performance can be achieved.
A number of studies have attempted to compare the performance of LBC and conventional cytology with respect to the abnormal rate, identification of CIN, time required for specimen evaluation and specimen adequacy. Although there is reasonable agreement that LBC improves specimen adequacy and reduces screening time compared to conventional cytology,6, 7 the relative sensitivity and specificity of the 2 methods remains controversial. In large measure this controversy is due to a lack of high-quality studies specifically designed to carefully compare test performance in an unbiased manner. Moreover, there are no comparative studies in high-risk populations from low-resource settings.
To address these needs, we conducted a direct comparison of LBC and conventional cytology in previously unscreened South African women enrolled in a cervical cancer prevention trial. In our study, all women were systematically assigned to receive either LBC or conventional cytology. In addition, to avoid verification bias, all women underwent colposcopy and biopsy of all cervical lesions and endocervical curettage. Both LBC and conventional cytology were read in a blinded fashion in a cytology laboratory with several years of experience in LBC processing.
Sample and study design
The sample derives from a population of previously unscreened, nonpregnant women, aged 35 to 65 years, with no history of hysterectomy or prior treatment for cervical intraepithelial neoplasia. All were participants in a randomized clinical trial evaluating the safety and efficacy of a “screen and treat” program for cervical cancer prevention8 and were recruited from 3 primary care clinical sites in Khayelitsha, which is a periurban, informal settlement outside Cape Town, South Africa. Participants were enrolled in the randomized trial between June 2000 and December 2002 and, upon enrolling into the trial, received a gynecologic exam, which included direct visual inspection (DVI) and testing for high-risk HPV DNA. Participants were then randomized to 1 of 3 arms: 1) treatment with cryotherapy if DVI positive, 2) treatment with cryotherapy if HPV positive and 3) delayed treatment for 6 months. Patients were asked to return in 6 months so that the impact of the “screen-and-treat” protocols could be evaluated.
Samples for the current study were collected at the clinic visit that took place 6 months after enrollment, by which time 14.5% had undergone cryotherapy by design in the overall trial. At this visit, all women regardless of trial arm or screening test results underwent a gynecologic examination. At the beginning of the examination a cytology specimen was collected using a plastic Ayre's type spatula and a cytobrush. Cells on the collection device were transferred either to a ThinPrep® collection vial or to a conventional glass slide. The cytology method used (i.e., ThinPrep® or conventional) was rotated on a 6 month basis for the duration of the study and performed by the same 6 gynecologists. Although the type of cytology method used was systematically assigned to ensure comparability between comparison groups, method type was not randomly assigned because it was not logistically feasible. Following collection of the cytology specimens, all women underwent colposcopy with endocervical curettage and biopsy of all colposcopic abnormalities by the same gynecologist who collected the cytology specimen. Therefore, at least 1 histology tissue specimen was collected from each study participant.
The study was approved by the Committee for the Study of Human Subjects at the University of Cape Town and by the Institutional Review Board of Columbia University.
Both LBC and conventional Papanicolaou cytology specimens were processed at the same laboratory in the United States (Health Networks Laboratory, Allentown, PA) within 4 weeks of being collected. This laboratory has been processing and reading both specimen types for more than 5 years. Cytology was reported using the 1991 modification of the Bethesda terminology,9 which utilizes 3 categories of specimen adequacy: satisfactory, satisfactory-but-limited-by and unsatisfactory. A single senior cytopathologist evaluated all abnormal cytology specimens (WD). All cervical biopsies and endocervical curettages were processed at Columbia University and read by a single experienced pathologist (TW). Histology was reported using the CIN terminology.10 Endocervical curettages containing high-grade cervical neoplasia (e.g., CIN2, CIN3 or cancer) were classified as positive for CIN 2+. Cytology and histology specimens were evaluated in a blinded fashion. At the end of the study, all biopsies originally classified as CIN (of any grade), as well as all biopsies from women who were HPV DNA or cytology positive (LSIL) or greater, were reviewed in a blinded fashion by the same pathologist. If the second blinded review did not agree with the first, another pathologist reviewed the slides and final diagnosis was obtained (concordance on 2 out of 3 reads).
HIV serostatus was determined using an enzyme linked immunosorbent assay (ELISA) for HIV antibody (Axsym®, HIV1/2, Abbot Laboratories, Chicago, IL). Positive results were confirmed with the Vironostick® HIV kit (Organon Teknika, Durham, NC).
Sensitivity, specificity, positive predictive value and negative predictive value of LBC and conventional cytology were estimated using histology-confirmed CIN as the “gold standard.” Two cut-off points were used for both the cytology results (ASCUS or higher and LSIL or higher) and the histology results (CIN1 or higher and CIN2 or higher). Measures of test performance were calculated twice: first excluding women with unsatisfactory cytology results and then treating women with unsatisfactory cytology results as normal. Exact 95% confidence intervals (CI) were calculated using the binomial option in SAS version 9.0 (SAS Institute, Cary, NC). Chi-squared tests11 were used to compare proportions and compute 2-tailed p-values.
A total of 5647 women with available cytology results were included in the final analysis. LBC specimens were obtained from 3,184 (56.4%) and conventional cytology specimens were obtained from 2,463 (43.6%) (Table I). Women who were screened using LBC and conventional cytology were equivalent in terms of their sociodemographic characteristics and prevalence of risk factors for cervical disease. Nearly 20% of both groups of women were aged 50 or older and nearly 40% had no high school level education as reported at enrollment. A high proportion of both tested positive for HIV (12.8% and 12.4%, respectively). Approximately 14% of women in both the LBC and conventional cytology group were treated with cryotherapy based on the protocol for the randomized trial.
|Type of cytology|
|35–39||1253 (39.4)||923 (37.5)|
|40–49||1317 (41.4)||1077 (43.7)|
|50–65||614 (19.3)||463 (18.8)|
|Treated with cryotherapy||449 (14.1)||362 (14.7)|
|Trichomonas vaginalis||340 (10.7)||261 (10.6)|
|HIV-seropositive||390 (12.8)||313 (12.4)|
|Moderate/severe vaginal discharge||693 (21.8)||608 (24.7)|
|Married||1621 (50.9)||1217 (49.4)|
|Age at first sex <16 years||1095 (34.4)||836 (33.9)|
|Five or more lifetime sexual partners||1069 (33.6)||873 (35.4)|
|Two or more sex partner previous month||47 (1.5)||32 (1.3)|
|Current smoker||226 (7.1)||207 (8.4)|
|Number of livebirths|
|None||111 (3.5)||92 (3.7)|
|1–4||2072 (65.1)||1631 (66.2)|
|5 or more||1001 (31.4)||740 (30.0)|
|No school||297 (9.3)||221 (9.0)|
|Some primary school||1212 (38.1)||902 (36.6)|
|Some high school||1408 (44.2)||1152 (46.8)|
|High school graduate||267 (8.4)||188 (7.6)|
|Currently employed||789 (24.8)||653 (26.5)|
|Current contraceptive use|
|Injectable||452 (14.2)||346 (14.1)|
|Oral||56 (1.8)||53 (2.2)|
The prevalence of biopsy-confirmed high-grade disease (CIN 2+) identified at the colposcopic examination performed immediately after the cytology was obtained was also similar between the 2 groups; 2.1% (68/3184) of women who received LBC at 6 months were diagnosed with CIN 2+ on histology vs. 2.3% (56/2463) of women who received conventional cytology. The estimated prevalences of CIN 2+ include 14 women with segments of high-grade cervical neoplasia on their endocervical currettage who were either not diagnosed with CIN 2+ on their biopsy (n = 6) or did not have a cervical biopsy (n = 8).
The performance of conventional cytology was statistically equivalent to the performance of LBC on all four measures of test performance and at all cutoff levels (Table II). However, there was some suggestion of improved performance in terms of sensitivity and positive predictive value of conventional cytology compared to LBC in this population. The sensitivity of conventional cytology was 5.6 to 13.0 percentage points higher than the sensitivity of LBC. The difference in sensitivity was greatest with a cytology cutoff of ASCUS and using a histology diagnosis of CIN2+. In this instance, conventional cytology correctly identified 83.6% (95% CI 71.2–92.2%) and LBC correctly identified 70.6% (95% CI 58.3–81.0%) of women with CIN2+ histology. This difference was not statistically significant (p = 0.09).
|Histology||≥ ASCUS||Histology||≥ LSIL|
|Liquid-based1 (n = 511/3114)||Conventional2 (n = 402/2444)||Liquid-based1 (n = 220/3114)||Conventional2 (n = 170/2444)|
|≥CIN1 (n = 152)||Sens||73.0 (65.2–79.9)||78.6 (70.1–85.7)||≥CIN1 (n = 117)||Sens||52.0 (43.7–60.1)||61.5 (52.1–70.4)|
|Spec||86.5 (85.2–87.7)||86.7 (85.2–88.0)||Spec||95.2 (94.4–96.0)||95.8 (94.9–96.6)|
|PPV||21.7 (18.2–25.6)||22.9 (18.9–27.0)||PPV||35.9 (29.6–42.6)||42.4 (34.8–50.2)|
|NPV||98.4 (97.9–98.9)||98.8 (98.2–99.2)||NPV||97.5 (96.8–98.0)||98.0 (97.4–98.6)|
|≥CIN23 (n = 68)||Sens||70.6 (58.3–81.0)||83.6 (71.2–92.2)||≥CIN24 (n = 55)||Sens||60.3 (47.7–71.9)||69.1 (55.2–80.9)|
|Spec||84.8 (83.5–86.1)||85.1 (83.6–86.5)||Spec||94.1 (93.2–94.9)||94.5 (93.5–95.4)|
|PPV||9.4 (7.0–12.3)||11.4 (8.5–15.0)||PPV||18.6 (13.7–24.4)||22.4 (16.3–29.4)|
|NPV||99.2 (98.8–99.5)||99.6 (99.2–99.8)||NPV||99.1 (98.7–99.4)||99.3 (98.8–99.6)|
Initial comparisons of test performance excluded 70 women who received LBC and 19 women who received conventional cytology with no cytology results due to specimen inadequacy. All had negative histology results except for 1 woman who received conventional cytology and was diagnosed with CIN 3. Treating unsatisfactory cytology results as negative slightly attenuated differences in sensitivity. When unsatisfactory cytology results were classified as negative, the sensitivity of conventional cytology decreased for CIN 1+ histology to 78.0% for an ASCUS cutoff and 61.0% for a LSIL cutoff. For CIN 2+ histology, sensitivity decreased to 82.1% for an ASCUS cutoff and 67.9% for a LSIL cutoff. Conversely, the sensitivity of LBC remained unchanged.
Reflecting to an extent the differences in sensitivity, conventional cytology also performed somewhat better than LBC in terms of positive predictive value. The difference was greatest for LSIL cytology cutoffs and CIN 1+ histology, where the positive predictive value of conventional cytology was 42.4% (95% CI 34.8–50.2%) vs. 35.9% (95% CI 29.6–42.6%). This difference was not statistically significant (p = 0.20). Treating unsatisfactory results as negative did not change the positive predictive value of either LBC or conventional cytology.
LBC and conventional cytology were equivalent in terms of specificity and negative predictive value. Both types of cytology exhibited strong specificity, approximately 85% for the ASCUS cytology cutoffs and 94% to 95% for the LSIL cytology cutoffs. Both had negative predictive values that neared 100%, regardless of cutoff level.
Stratification by HIV status and age group (<40 years vs. ≥40 years), separately, produced similar results. There were no substantial differences in specificity between LBC and conventional cytology. Conventional cytology had higher sensitivity for all strata and cutoff levels, except among HIV-positive women when the sensitivity of the 2 cytology methods was compared using CIN1+ histology. For this stratum, sensitivity of conventional cytology was lower for the ASCUS cutoff (85.5% for LBC vs. 75.0% for conventional, p = 0.19) and equivalent for the LSIL cutoff (60.0% for LBC vs. 59.1% for conventional). In all other strata sensitivity of conventional cytology was more than 10 percentage points higher. None of the differences were statistically significant.
In addition, since a portion of the women had previously been treated with cryotherapy [n = 445 (14.3%) women in the LBC group, n = 357 (14.6%) women in the conventional group], the analysis was rerun on the subset of women randomized to the no cryotherapy arm (n = 1,016 in the LBC group, n = 805 in the conventional group). Among this subgroup, sensitivity remained higher for conventional cytology by more than 10 percentage points for all comparisons and specificity was approximately comparable with increases of 3 percentage points or less for conventional cytology. Again, differences were not statistically significant.
Comparison of specific results
More detailed comparisons of specific cytology and histology results further indicated that LBC was not superior to conventional cytology at determining disease grade (Table III). For example, among women who received conventional cytology, 22.7% of women with CIN 2 and 12.1% of women with CIN 3 or cancer had negative cytology results compared to 34.3% and 21.4%, respectively, of women who received LBC. These differences were not statistically significant (p = 0.35 and 0.40, respectively). There was some suggestion that conventional cytology tended to diagnose women with biopsy-confirmed CIN 1 as HSIL or higher (19.4% ASCUS, 29.0% LSIL and 25.8% HSIL or higher), whereas LBC tended to diagnose women with CIN 1 as ASCUS (29.8% ASCUS, 34.5% LSIL and 11.1% HSIL or higher). Among women with negative histology results, there was essentially no difference in the distribution of LBC and conventional cytology results. The overall abnormal rates, including SIL and invasive carcinoma, for LBC and conventional cytology were the same (7.0%), although conventional cytology produced slightly more HSIL results (3.7% vs. 2.7%).
|Negative (n = 2962)||CIN1 (n = 84)||CIN2 (n = 35)||≥CIN3 (n = 28)||SONE1 (n = 5)||Total (n = 3114)|
|Negative||2,562 (86.5)||21 (25.0)||12 (34.3)||6 (21.4)||2 (40.0)||2,603 (83.6)|
|ASCUS||259 (8.7)||25 (29.8)||4 (11.4)||2 (7.1)||1 (20.0)||291 (9.3)|
|LSIL||95 (3.2)||29 (34.5)||7 (20.0)||4 (14.3)||0 (0.0)||135 (4.3)|
|≥HSIL||46 (1.6)||9 (10.7)||12 (34.3)||16 (57.1)||2 (40.0)||85 (2.73)|
|Conventional cytology||Negative (n = 2327)||CIN1 (n = 62)||CIN2 (n = 22)||≥CIN3 (n = 24)||SONE1 (n = 9)||Total (n = 2444)|
|Negative||2,017 (86.7)||16 (25.8)||5 (22.7)||4 (12.1)||0 (0.0)||2,042 (83.6)|
|ASCUS||212 (9.1)||12 (19.4)||3 (13.6)||4 (16.7)||1 (11.1)||232 (9.5)|
|LSIL||56 (2.4)||18 (29.0)||3 (13.6)||3 (12.5)||0 (0.0)||80 (3.3)|
|≥HSIL||42 (1.8)||16 (25.8)||11 (50.0)||13 (54.2)||8 (88.9)||90 (3.7)|
LBC and conventional cytology differed substantially in terms of specimen adequacy (Table IV). LBC was 4-fold less likely than conventional cytology to produce “satisfactory-but-limited-by” results (6.5% vs. 27.9%, p < 0.01). Conventional cytology results were mostly limited by obscuring blood (40.1%), obscuring inflammation (34.8%), poor preservation (19.4%) and thick smears (26.1%). In total, 93.6% of conventional cytology results that were satisfactory-but-limited were limited by at least 1 of these factors compared to 10.1% of LBC results. LBC results were mostly limited by scant cellularity. In addition, 16.3% of LBC results were limited because the endocervical component was either scant or absent versus 5.2% of conventional cytology results. This amounted to 10.9% and 1.5% of all satisfactory or satisfactory-but-limited LBC and conventional cytology results, respectively.
|Liquid-based (n = 3,184)||Conventional (n = 2,463)|
|Satisfactory||2,906 (91.3)||1,758 (71.4)|
|Blood present||309 (10.6)||119 (6.8)|
|Inflammation present||155 (5.3)||139 (7.9)|
|Poor preservation||0 (0.0)||5 (0.3)|
|Satisfactory but limited by||208 (6.5)||686 (27.9)|
|Obscuring blood||19 (9.1)||275 (40.1)|
|Obscuring inflammation||1 (0.5)||239 (34.8)|
|Poor preservation||1 (0.5)||133 (19.4)|
|Scant cellularity||163 (78.4)||27 (3.9)|
|Endocervical component absent or scant||34 (16.3)||36 (5.2)|
|Thick smear||0 (0.0)||179 (26.1)|
|Unsatisfactory||70 (2.2)||19 (0.8)|
|Obscuring blood||21 (30.0)||17 (89.5)|
|Obscuring inflammation||1 (1.4)||4 (21.0)|
|Poor preservation||0 (0.0)||3 (15.8)|
|Scant cellularity||70 (100.0)||6 (31.6)|
|Thick smear||0 (0.0)||4 (21.0)|
LBC results were nearly 3 times more likely than conventional cytology results to be unsatisfactory (2.2% vs. 0.8%, p < 0.01). All unsatisfactory LBC results suffered from scant cellularity compared to 31.6% of unsatisfactory conventional cytology results. Among conventional cytology results, obscuring blood (89.5%) was the major cause of specimen inadequacy.
Our data demonstrate that the sensitivity and specificity of LBC and conventional cytology are similar in a high-risk population in contrast to the opinion that LBC is better. In fact, point estimates of test sensitivity were higher, albeit nonsignificantly, for conventional cytology in comparison to LBC at all cytology cutoff levels for both the CIN1+ and CIN2+ histology endpoints. There was no difference in the specificity of the 2 methods. As found in previous studies,12, 13, 14, 15, 16, 17 LBC produced a substantially smaller proportion of “satisfactory-but-limited-by” results (27.9% vs. 6.5%), and a significantly higher proportion of unsatisfactory results (2.2% vs. 0.8%). All unsatisfactory LBC results were due to scant cellularity, perhaps reflecting high levels of cervical atrophy in our population of older women, aged 35–65. Some previous studies have also found that LBC has a higher unsatisfactory rate,13, 18, 19, 20 although others have not.12, 14, 16, 17, 21
A number of meta-analyses and systematic reviews have assessed the results of previous studies comparing LBC and conventional cytology with differing conclusions regarding test performance depending on the study inclusion criteria. Although several have concluded, in contrast to our findings, that the accumulated evidence suggests that LBC improves the detection of high-grade disease,7, 22, 23, 24 they have also unanimously noted the dearth of high-quality research needed to adequately answer this question. Indeed, in a 2001 review of new cervical cytology methods for the U.S. Preventive Services Task Force and the Agency for Healthcare Research and Quality, none of the 962 potentially relevant articles met the predefined inclusion criteria.25 The overwhelming reason for exclusion in this review and several other recent reviews22, 23, 24 was the lack of an adequate reference standard such as colposcopy/histology. For example, in lieu of colposcopy/histology, LBC comparison studies often use what is known as the “expert panel” review of selected cytology specimens and, instead of relating test findings to the true disease status of the cervix, compare abnormal rates (e.g., the proportion with LSIL or higher) as a proxy to comparisons of test sensitivity. Unfortunately, a higher rate of LSIL cytology diagnosis in no way ensures a higher “true”; disease detection rate, as a cytological diagnosis of LSIL is a relatively inaccurate measure of true disease status.3
Among the studies with an adequate reference standard such as colposcopy/biopsy, women are often selectively followed based on any previous detection of cytological abnormality or the presence of other risk factors for cervical disease. Besides the current study, only 2 studies to date (a Costa Rican study of 8,636 women from a population-based sample26 and a French study of 1,757 women attending public and private clinics for routine cytologic screening27) have avoided selective follow-up of high-risk or cytologically positive women and applied a definitive clinical reference standard to either a random sample of women with normal screening test results or the entire sample of women. This is critical to allowing the comparative sensitivity and specificity for LBC and conventional cytology to be calculated. Both were split-sample studies in which a single specimen was collected from each woman's cervix and both a conventional cytology and a ThinPrep liquid-based cytology preparation was made for the specimen. The Costa Rican study, based on cutoff levels of ASCUS or higher and CIN 2+, reported that LBC was significantly more sensitive (92.9% vs. 78.8%) but less specific (88.6% vs. 94.5%). In this study conventional cytology was evaluated in a service setting in Costa Rica, whereas LBC was evaluated by a single expert cytopathologist in the United States. In contrast, the more recent French study utilized comparable pathologists for both methods. Based on cutoff levels of ASCUS or higher and CIN 1+, the French study found that conventional cytology and LBC performed about the same in terms of both sensitivity (74% and 73%, respectively) and specificity (91% and 90%, respectively). Other methodological problems with LBC and conventional comparison studies include questionable comparability between comparison groups due to no systematic assignment of cytology type or use of historical controls, lack of blinding with respect to clinical status and comparisons initiated immediately following the training or retraining of laboratories in LBC techniques.
The current study attempted to address several of the above concerns by systematically assigning cytology type to ensure comparability between comparison groups, by blinding clinicians to the patients' clinical status, and by utilizing a laboratory with several years experience processing and reading both LBC and conventional cytology specimens. Most importantly, the study employed an adequate reference standard and avoided selective follow-up of patients as all participants underwent colposcopy and histological sampling immediately following collection of cytology specimens. Verification bias, which can occur when histological verification is obtained for only cytologically positive women or a selected subset of cytologically negative women,3 was completely avoided in our study since all women underwent colposcopy/biopsy.
The main limitation of the current study was the relatively small number of women with high-grade disease, which limited the precision of estimates of sensitivity and our ability to rule out random chance as plausible explanation for the observed differences. The study findings are also limited since the ThinPrep is only 1 of the several types of LBC currently on the market. In particular, the SurePath system, which is also widely used in developed countries and approved for use in the United States by the Food and Drug Administration, was not evaluated. These other LBC methods are currently being evaluated but no direct comparison studies have been completed as yet.
The generalizability of the findings may also be limited since the sample derives from a setting where cervicovaginal infections are extremely common, HIV is highly prevalent, and where most women have never been screened for cervical cancer. Indeed, one possible explanation for our failure to identify an increase in sensitivity for LBC compared to conventional cytology is that by screening women with no previous history of screening, we are identifying prevalent lesions that are likely larger and may shed more abnormal cells than lesions in populations with access to regular screening. A commonly cited advantage of LBC is that, unlike conventional smears, for which only a selective portion of cellular material is transferred to the slide, almost all sampled cells are rinsed into the liquid and dispersed throughout.28 As a result, LBC is thought to produce smears that are more representative of the entire sample, thereby, facilitating the detection of abnormal cells when only a few such cells have been adequately sampled.
We also detected a slightly lower than expected proportion of histology confirmed CIN2+ among women with HSIL cytology results (35.3% for LBC and 35.5% for conventional). The high positive rate for a HSIL cytology result was likely due to several factors. One is the fact that the study colposcopists were blinded to the cytology results, which was necessary to minimize the possibility of any differential measurement error. As shown in other studies, the performance of colposcopy is worse when the colposcopists are not expecting to find CIN 2,3.32 In addition, the high rates of cervical inflammation and atrophy in this older population of Black African women may also have increased the false positive rate or made it more difficult to colposcopically identify areas of CIN 2,3.
The results of our study have very different implications for developed and less-developed countries. In developed countries, even without significant improvements in sensitivity, it is probably still desirable to incorporate LBC into cervical cancer screening programs because of both the proven cost-effectiveness of using residual LBC specimens for “reflex” HPV DNA testing for triage of women with ASCUS results29, 30 and the substantial decrease in “satisfactory-but-limited-by” results. Both of these likely counterbalance costs by reducing the need for repeat smears and more aggressive follow-up.31 Conversely, in less-developed countries, where it is unlikely that reflex HPV DNA testing will be utilized in the foreseeable future and women are rarely screened more than once or twice in their lifetime, the increased cost of LBC may only be justifiable if there is definitive proof that LBC can increase the detection of high-grade disease in these populations.
In conclusion, the current study is a direct and rigorous comparison of the test performance of LBC and conventional cytology. Results confirm that ThinPrep substantially reduces “satisfactory-but-limited-by” results but do not show that ThinPrep increases the detection of histology-confirmed disease (CIN1+ or CIN2+). Because of the increased cost of LBC, less-developed countries, which do not use reflex HPV DNA testing, should carefully consider the potential benefits and drawbacks of LBC before adopting this new technology.
We thank Health Networks Laboratory for all of their work.
- 7Methods and techniques of cervical cancer screening. In: European guidelines for quality assurance in cervical cancer screening. Luxemburg: European Commision, 2004., , .
- 8Randomized clinical trial of the safety and efficacy of a “screen and treat” cervical cancer prevention program. 2006 (in preparation)., , , , .
- 9The Bethesda System for reporting cervical/vaginal cytologic diagnoses. New York: Springer-Verlag Telos, 1994., .
- 10Precancerous lesions of the cervix. In: Blaustein's pathology of the female genital tract. New York: Springer-Verlag, 1994., , .
- 11Statistical methods for rates and proportions, 2nd ed. New York: John Wiley & Sons, 1981..
- 26Utility of liquid-based cytology for cervical carcinoma screening: results of a population-based study conducted in a region of Costa Rica with a high incidence of cervical carcinoma. Cancer 1999; 87: 48–55., , , , , , , , , , .
- 31Evaluation of HPV/LBC cervical screening pilot studies: first report to the Department of Health on evaluation of LBC (December 2002, revised January 2003). www.cancerscreening.nhs.uk/cervical/lbc-pilot-evaluation.pdf, 2003., , , .