Fax: (011) 46-031415792.
Liquid-based cytology versus conventional Papanicolaou smear in an organized screening program†
A prospective randomized study
Article first published online: 27 AUG 2007
Copyright © 2007 American Cancer Society
Volume 111, Issue 5, pages 285–291, 25 October 2007
How to Cite
Strander, B., Andersson-Ellström, A., Milsom, I., Rådberg, T. and Ryd, W. (2007), Liquid-based cytology versus conventional Papanicolaou smear in an organized screening program. Cancer, 111: 285–291. doi: 10.1002/cncr.22953
Cytyc Corporation provided the ThinPrep 2000 machine free of charge.
- Issue published online: 11 OCT 2007
- Article first published online: 27 AUG 2007
- Manuscript Accepted: 3 MAY 2007
- Manuscript Revised: 2 MAY 2007
- Manuscript Received: 27 FEB 2007
- Region Vastra Gotaland, Sweden
- cervical intraepithelial neoplasia;
- liquid-based cytology;
- Papanicolaou smears;
- mass screening;
- randomized controlled trial
The objective of this study was to evaluate whether liquid-based cytology (LBC) can improve high-standard cervical cancer screening cytology further. The primary endpoint was histopathologic high-grade lesions in current and subsequent screening rounds. The secondary endpoints were cytologic diagnosis and inadequate samples.
Women were randomized to smear taking by conventional Papanicolaou (Pap) smear or LBC according to the time of appointment. Eight thousand eight hundred ten conventional Pap smears and 4674 LBC samples were included. Evaluations of atypical cytology and referral to colposcopy and treatment were performed as routine procedures. Histopathologic diagnoses were retrieved from a regional database 8 months after the study was closed. The mean follow-up was 2 years and 9 months.
Inadequate samples were observed in 0.3% of LBC samples versus 0.7% of Pap smears (P = .002). The total fraction of nonbenign diagnoses in cytology was 4.5% versus 3.5%, respectively (P < .001). Histopathologic evaluation was made on 570 patients constituting 4.6% of the LBC samples and 4% of the Pap smears. Forty percent more high-grade lesions were identified as a result of LBC sampling (1.20% vs 0.85%; P = .05). The influence of the sampling method was significant for all variables (odds ratio [OR], 1.60; 95% confidence interval [95% CI], 1.12–2.28) for high-grade lesions that were identified by histology when adjusting for age and screening unit in a logistic regression model. At the second follow-up 2 years and 1 month later, the OR was decreased only slightly (1.51; 95% CI, 1.13–2.01).
In the ongoing cervical screening program of western Sweden, liquid cytology produced a significantly higher yield of histologic high-grade lesions compared with conventional Pap smears. Cancer (Cancer Cytopathol) 2007. © 2007 American Cancer Society.
Incidence and mortality in cervical cancer have decreased substantially in all western countries as result of cytologic screening. In some settings, liquid-based cytology (LBC) has replaced the conventional Papanicolaou (Pap) test, offering practical advantages. The application of cells to the glass slide is standardized; cells are distributed evenly over the surface; mucus, blood, and inflammatory cells are reduced in the preparation; and fixation is effective and even. In several studies, these improvements have demonstrated higher detection rates of cervical intraepithelial neoplasia (CIN), lower rates of atypical squamous cells of undetermined significance (ASCUS), and/or a lower rate of smears that cannot be interpreted.1–13 Some studies also claim an improvement in detecting preinvasive and invasive glandular disease.14, 15 Furthermore, the residual liquid can be investigated for biomarkers, such as human papillomavirus (HPV) and p16, or for infectious agents. The method has spread rapidly in the U.S. and, in the U.K. and Canada, LBC is now the standard method of screening.
However, LBC has been questioned, and criticism has been raised regarding the design of the studies that asserted its superiority.16, 17 A recent systematic review18 came to the conclusion that, the better the study, according to the criteria set up by the authors, the less the difference between LBC and conventional Pap smear, with no difference at all in those that were classified as the best studies. These authors called for larger prospective, randomized trials.
Sweden has long experience of good-quality, widespread, low-cost cytology that has been the basis for a successful, organized screening program since the 1960s.19, 20 Rates of ASCUS and unsatisfactory samples are low. However, after 40 years of screening in Sweden, women still develop cervical cancer,21 and too many women feel disappointment and anger when they discover that, contrary to their expectation, they have not been protected against invasive cervical cancer by participation in the program. Although nonparticipation is the main reason for developing cervical cancer in regions with widespread screening,22 there is clearly a need for a more sensitive method of finding preinvasive lesions.
Compared with the conventional Pap smear, does LBC provide better protection against cervical cancer in an organized screening program that covers the entire female population with high-quality smear taking and high-standard conventional cytology? The objective of this study was to compare the effectiveness (primary endpoint: histopathologic high-grade lesions in current and subsequent screening round; secondary endpoint: cytologic diagnosis, inadequate samples) of cervical cancer screening using LBC and conventional Pap screening techniques in a prospective, randomized trial nested in the current screening program of western Sweden.
MATERIALS AND METHODS
In Sweden, all women are invited regularly to undergo a Pap smear. According to national guidelines, which are followed by nearly all counties in Sweden, a smear is taken every third year from age 23 years to age 50 years and then at ages 55 years and 60 years. The smears are taken at screening clinics by trained midwives.
Five screening clinics in the greater Goteborg area were engaged in the study and represented both inner-city and suburban populations. In these clinics and in most parts of Sweden, women are invited to a predefined appointment that can be changed at the initiative of the women. Women attending the clinics for prescription and counseling for contraceptives as well as pregnancy also are offered a smear if they are scheduled to be invited for a smear in the same calendar year. The midwives received special training in smear taking and also were trained in handling LBC samples. For the current study, women were randomized according to the time of their appointment for smear taking by conventional Pap smear or liquid cytology. After 1 week of LBC sampling, the midwives changed to Pap smears, and vice versa.
Conventional Pap smears were taken with a wooden spatula (Ayre) from the fornix and the portio and with a Cytobrush from the endocervix. LBC was taken similarly but with a plastic spatula and a cytobrush. Thus, material was taken from the fornix, portio, and endocervix from all women.
LBC specimens were placed in PreserveCyt solution and were processed in the ThinPrep 2000 machine (Cytyc Corporation, Boxborough, Mass). All slides for conventional cytology and LBC were stained according to the Pap method. All LBC specimens were screened by 1 of 3 cytotechnicians with special training in LBC. This group also screened 73% of the conventional Pap smears in the study. All positive cytology in the remaining 27% of smears was reviewed by 1 or 2 of these cytotechnicians. For conventional smears, the evaluation of whether slides were representative and for the presence of endocervical/metaplastic cells followed the Bethesda 2001 criteria. LBC was considered representative if the slide contained >5000 epithelial cells. Endocervical cells were considered present if the slide contained ≥2 groups of glandular/metaplastic cells with ≥5 cells each or if the slide contained ≥10 dissociated glandular/metaplastic cells. A smear that was taken within the organized mass screening program that lacked endocervical cells was not considered suboptimal or nondiagnostic and, thus, was not followed by a repeat smear.
Clinical management of atypical cytology and referral to colposcopy and treatment were performed as routine procedures within the screening program in the western region of Sweden. ASCUS and CIN type 1 (CIN1) led either to colposcopy after 4 months or to a repeat smear. Women with diagnoses of CIN2 to CIN3 were referred to colposcopy. Colposcopists were blinded to the cytology method that was used. Histopathology diagnoses were searched for in the Regional Database for Prevention of Cervical Cancer, which covers all 5 laboratories in the region, including, among other data, all histopathology related to cervical disease (biopsies, cones, and hysterectomy specimens). The highest grade of histopathologic diagnosis from the cervix uteri obtained after the index smear during the study period was used. Thus, histopathologic diagnoses were made as part of the clinical routine. The pathologists were blinded to the type of cytology that was used. This search was made on 2 occasions. The first search occurred 8 months after the study was closed. The mean follow-up was 1.5 years (range, from 9 months to 2 years and 5 months). The median time was equivalent to the mean. The maximum time from smear to follow-up histopathology was well within the period of 1 screening round. The second search in the regional database was made 2 years and 1 month later, before preparing this report. The mean follow-up at that time was 3 years and 7 months (range, from 2 years and 10 months to 4.5 years).
Because there was a difference in age (median, 3.5 years) and a discrepancy in the proportion of ThinPrep/Pap smears, which were not distributed evenly between the screening clinics, adjustment for these potential confounders were made by applying logistic regression analyses. The type of method was included as well as age and screening unit, reflected by several 0 to 1 variables, as independent variables in those analyses. Uncorrected differences in proportions between groups were calculated with the chi-square test using Stata software (version 9.2; StataCorp, College Station, TX). All P values were 2-sided.
The study was approved by the Ethics Committee of the University of Goteborg.
The study period was between May 2002 to December 2003, and 13,484 smears were entered into the database (8810 conventional Pap smears and 4674 LBC samples). There were 2 reasons for the uneven distribution of women to the 2 methods: 1) the clinics simply forgot to shift to the “new” method some weeks when they were supposed to; and 2) there were problems with the distribution of material for the LBC samples to the screening clinics, and midwives had to use conventional cytology. However, these aberrations occurred randomly, and no individual, midwife or patient, had any influence on which method was to be used for the individual woman. Thus, the distribution was random, although uneven. The mean age of the women at the time of smear taking was 40.27 years (40 years and 3 months). Data regarding baseline characteristics are provided in Table 1.
|Variable||No. of smears|
|Total no. of smears||8810||4674|
|Histopathology at first follow-up, %||4||4.6|
Tables 2 and 3 show the diagnosis of the smears and the histopathologic diagnostic outcome from follow-up. There was no difference noted with regard to the distribution of cytologic diagnosis or the proportion of Pap smears/LBC specimens among the 10,567 women who were invited to screening compared with the 2917 women who were offered a test because they were due for invitation to the screening program in the close future. There was no significant difference in the proportion of smears that were followed with histopathology (P = .71). The number of suboptimal smears was low in both groups but was significantly lower with LBC specimens compared with conventional Pap smears (Table 2).
|Variable||No.||Pap, %||LBC, %||OR||95% CI|
|Endocervical cells missing||713||4.2||7.5||1.8||1.54–2.10|
|No histology||Benign||Low grade||High grade|
Endocervical cells were lacking in more LBC specimens than in Pap smears. This difference diminished with increasing age of the participating women.
The ASCUS rate was nominally higher in the LBC group. This difference did not remain statistically significant after adjustment for age and screening unit in the logistic regression model.
The number of low-grade squamous intraepithelial lesions (LSIL) was significantly higher in the LBC group. There was also a significant difference in the diagnosis of high-grade cytologic dysplasias. Although the accuracy of high-grade squamous intraepithelial lesion (HSIL) cytology for predicting HSIL in histopathology and LSIL cytology versus LSIL in histopathology appeared to be greater for conventional Pap smears (Table 3); however, the numbers were rather small, and the difference was not statistically significant (P = .17 and P = .57, respectively).
The numbers of specimens with glandular cell atypia and adenocarcinoma were too low to be compared statistically. One LBC test identified atypical glandular cells of undetermined significance in which the histology was benign, and another LBC sample had a diagnosis of adenocarcinoma in situ (AIS) in which histology demonstrated benign glandular change and HSIL. Two conventional Pap smears correctly diagnosed histopathologically confirmed AIS.
Approximately 42% more high-grade lesions were identified by histopathology as a result of a screening test with LBC (P = .05) compared with conventional Pap smears (Table 4). This difference was even more significant after adjusting for age and screening unit in the logistic regression model. The crude difference was reduced to 30% when the follow-up was increased by 25 months, and this difference remained statistically significant in the regression model.
|Model||Conventional Pap||LBC||P||OR||95% CI|
|Follow-up at a mean of 1.5 y||75||0.85||56||1.20||.05|
|Adjusting for age||1.71||1.20–2.43|
|Adjusting for age and screening unit||1.60||1.12–2.28|
|Follow-up at a mean of 3 y, 7 mo||122||1.38||84||1.80||.06|
|Adjusting for age||1.62||1.22–2.16|
|Adjusting for age and screening unit||1.51||1.13–2.01|
In the period between the evaluations at 1.5 years and 3 years and 7 months, another 0.53% of the Pap cohort demonstrated high-grade lesions in histopathology, and the corresponding proportion for the LBC cohort was similar (0.60%). This period was considered to represent a new screening round for the majority of the women.
Six women in the study were diagnosed with invasive cancer, all of which were squamous cell cancers. Five of those women had a conventional Pap smear, and 1 woman had an LBC test. The LBC sample was inadequate, no diagnosis was provided, and the follow-up Pap smear demonstrated HSIL. Four of the index Pap tests showed HSIL, but the fifth test was diagnosed as within normal limits, and there was an interval of nearly 3 years from this smear to the cancer diagnosis. This cancer was identified in the second search for follow-up (mean follow-up, 3 years and 7 month), and the other 5 cancers were identified in the first search (mean follow-up, 1.5 years).
The results of the current prospective, randomized trial, which was conducted within an organized screening program, demonstrated that LBC yielded an increased rate of histopathologically proven, high-grade lesions in the routine follow-up of abnormal smears compared with conventional cytology. Extending the follow-up period to a mean/median of 3 years and 7 months, allowing us to find initially undetected, high-grade lesions in the next screening round, did not alter this conclusion.
This achievement has a cost in decreased specificity, because the number of women who received a notification that the smear was not normal increased to 30% from 3.5% or 4.5%. The cost of this is both psychological and economic, because resources for follow-up have to be increased concomitantly.
Conventional Pap smears demonstrated a tendency to be more accurate when histopathology was used as the gold standard. Approximately 93% of high-grade readings in Pap smears corresponded with high-grade lesions in histopathology compared with 83% for LBC. The reason that LBC yielded a better detection rate of high-grade lesions in histopathology in our study is that this method detected lesions to a greater extent that, with cytology, were classified as low grade but, with histopathology, were classified as high grade. This underlines the importance of the correct evaluation of cytologic low-grade lesions and calls into question the practice of accepting only a second smear as follow-up.
It is noteworthy that, despite a generally very low rate of inadequate smears in western Sweden with conventional cytology, this rate was lowered even further and was statistically highly significant with LBC. Many studies have been undertaken comparing LBC with conventional Pap cytology. Results both in primary studies and in meta-analysis have differed, and all studies have had their weaknesses. Common problems are lack of systematic or randomized allocation of cytology type, lack of blinding, lack of adequate reference standard, and questionable representativeness for population screening.16–18 The only other population-based randomized study we have identified to date23 did not demonstrate any significant superiority of either technique. However, that study was performed in an entirely different setting with a previously unscreened population in a South African township. Those authors pointed out that the high prevalence of cervicovaginal infections and possibly easily detected, large lesions in their study did not allow for close comparisons with a well-screened population in western countries. The African study, which had many assets, had half the number of participants compared with ours and only approximately two-thirds of the high-grade lesions in histopathology. An attempt was made to adjust for human immunodeficiency viral status, and women who recently had undergone cryotherapy were excluded, which decreased the power to detect any difference even further.
The strength of our study is the randomized design and large size, with sufficient power to answer the research questions. The randomization limped somewhat; however, despite its shortcomings, women still were selected randomly for either sampling method, and the only recognizable bias that occurred, age and (to a small extent) screening unit, could be adjusted for statistically. The main outcome was defined as the ability to detect high-grade lesions in histopathology, because cytologic findings only without a reference method is a quite inaccurate measure of true disease status.24 We believe that the essential objectives of cervical screening programs are detecting and removing histologically confirmed, high-grade lesions. Because not all women in our study were followed with histopathology, there is a possibility of verification bias, a weakness that was impossible to avoid for practical and ethical reasons. However, the proportions of follow-up in histopathology were quite equal between the 2 methods, and this potential bias appeared to be insignificant in this study. The study was conducted in a setting with experienced midwives taking smears, which resulted in a very low rate of inadequate conventional smears (0.7%) and a low rate of smears that lacked endocervical cells (4.2%).
It should be kept in mind that cervical dysplasia may regress and, for the most part, may not develop into invasive cancer. Thus, finding a greater proportion of high-grade lesions in a screening round will lead to a reduction in cancer incidence and mortality only if these lesions should progress and remain undetected until the event of cancer or death. It has been established that the sensitivity of the Pap smear is limited, reaching 80% in good hands.25 When it is progressing, CIN normally needs several years to develop into cancer, and it is possible to compensate for the limited sensitivity of the test by taking multiple tests during that period. Thus, a higher immediate detection rate of high-grade lesions, per se, does not imply a reduced incidence or mortality—the undetected lesion very well may be picked up in the next screening round. With the longer follow-up interval in our study, we demonstrated that this was not the true. The mean observation time of 3 years and 7 months permitted a majority of these women to go through a new screening round (with a conventional Pap smear), and the difference in the yield of high-grade lesions decreased only marginally. To our knowledge, there still is no indication that the number of high-grade lesions in a following screening round would decrease with the use of LBC. The reason for this is not obvious. We know little regarding the high-grade lesions that are not identified by Pap smears. Some, despite their potential to progress, very well may hide from detection with the now-proven, less sensitive, conventional Pap smear, even at the next screening round 3 years later. We believe that, unless proven otherwise, the increase in identifying high-grade lesions with LBC must be regarded as an increase in the detection of true precancerous lesions. However, if the number of high-grade lesions actually would have decreased in the next screening round, then an extended screening interval could have been considered; however, the results from this study do not support such a notion. A new search in the database at a later stage may spread more light on this important issue.
Several studies corresponding to our experience have demonstrated that even strict participation in regular screening does not provide full protection from cervical cancer, indicating a need for the improved detection of high-grade, precancerous lesions.21, 26 LBC in the form of ThinPrep slides demonstrates a >40% higher ability to detect high-grade cervical lesions in histopathology than conventional Pap smears within the framework of a Swedish cervical screening program. The corresponding cost in specificity, represented by a 30% increase in abnormal cytologic samples, is noteworthy, but we believe it is acceptable. It is probable that replacing conventional cytology with LBC in the Swedish screening program would further reduce the incidence of cervical cancer.
We thank Professor of Biostatistics Anders Oden for his work with the statistics, statistician Karl-Johan Dahlof for his work with the regional database, and cytotechnician Monica Dohse for her data filing. The dedicated work of the midwives at the 5 screening units was of course essential to this study as was the committed work of the cytotechnicians involved.
- 6Evaluation of the ThinPrep Papanicolaou test in clinical practice: 6-month study of 16,541 cases with histological correlation in 220 cases. Hong Kong Med J. 1999; 5: 233–239., , , .
- 19National Board of Health and Welfare. Cancer Incidence in Sweden 2003. Stockholm, Sweden: The National Board of Health and Welfare; 2005.
- 20How cancer in the uterine cervix became a rare disease? [hur cervixcancer blev en ovanlig sjukdom?]. In: LindbergB, ed. Swedish Gynecology During One Century [Svensk Gynekologi Under ett Sekel]. Uppsala, Sweden: SFOG; 2004: 181–185..