A randomised comparison of SurePath liquid-based cytology and conventional smear cytology in a colposcopy clinic setting

Authors


Dr PH Sykes, Department of Obstetrics and Gynaecology, University of Otago, Christchurch, Private Bag 4711, Christchurch, New Zealand. Email peter.sykes@otago.ac.nz

Abstract

Objective  The objective of this study was to compare the sensitivity of cervical cytology using conventional smears and SurePath liquid-based cytology (LBC).

Design  Prospective randomised evaluation of diagnostic test.

Setting  A single institution colposcopy clinic.

Population  Women attending first visit colposcopy appointments were offered entry into the study.

Methods  Cervical cytology samples from 913 women of age 16–75 years were randomly processed as SurePath LBC or conventional smears. Conventional smears were taken for 453 women and a SurePath LBC taken for 451 women. Cytology results were correlated with colposcopic findings and histology from colposcopic biopsies, treatment and follow up.

Main outcome measures  To compare the sensitivity of SurePath LBC and conventional smears for histologically proven abnormality. Other outcome measures include a comparison of their sensitivity for high-grade abnormalities and their satisfactory rate.

Results  Accounting for all randomised samples, there was a trend towards improved sensitivity for SurePath LBC (79.1 versus 73.7%, P = 0.1). However, excluding unsatisfactory cytology (and samples not taken) eliminated this trend; the sensitivity for both LBC and conventional smears for any epithelial abnormality was 81%. With a threshold of atypical squamous cells of uncertain significance (ASC-US), both SurePath LBC and conventional smears had a sensitivity of 92% for high-grade lesions. SurePath LBC was less likely to be reported as unsatisfactory (2.7 versus 9.1%, P < 0.0001).

Conclusions  In this context, with a threshold of ASC-US, both SurePath LBC and conventional smears offer high sensitivity for the detection of CIN2/3, but SurePath LBC is less likely to be reported as unsatisfactory.

Introduction

ThinPrep (Cytyc Corporation, Boxborough, MA, USA) liquid-based cytology (LBC) system was approved by the US Food and Drug Administration for cervical screening in 1996. Subsequently, other liquid-based systems have been approved including the SurePath (formerly Autocyte PREP) LBC system (BD, formerly TriPath Imaging Inc., Burlington, NC, USA) in 1999. Since this time, there has been a considerable debate on the merits or otherwise of LBC compared with conventional cytology. However, LBC has been adopted by several major screening programmes, including the NHS Cervical Screening Programme in the UK.

The differences between LBC and conventional smears include ease of transportation, processing costs, the preservation, presentation and numbers of cells on the slide. In addition, there is the possibility of ancillary testing and automated primary screening. The key issues, however, relate to its performance in cervical screening in comparison to conventional smears. Two important claims have been made in this regard: first, that LBC offers greater sensitivity for the detection of cervical intraepithelial neoplasia (CIN), in particular high-grade squamous intraepithelial lesions (HSIL) (CIN2/3), and second, that LBC is associated with a lower unsatisfactory rate than conventional smears.

There have been many studies and several systematic reviews on the subject most comparing the ThinPrep system with conventional smears. Unfortunately, the superiority or otherwise of LBC still remains controversial. There have been a limited number of large randomised studies, and many studies have had methodological weaknesses.1 Despite contrary conclusions in previous meta-analyses,2,3 a recent systematic review concluded that the evidence ‘does not support a conclusion that LBC is better than conventional cytology’ with regard to sensitivity or satisfactory rates.1

Fewer studies are available comparing the SurePath system to conventional cytology. There are several large studies on screening populations using split samples or comparison between nonrandomised populations that report a higher sensitivity and lower unsatisfactory rates for SurePath compared with conventional slides.4–11 However, we have been unable to identify a published, randomised study including colposcopic evaluation of cytologically negative women.

At this time, on the basis of systematic reviews,12,13 New Zealand (NZ) and Australia have not opted to fully fund LBC smears, but they are considered of diagnostic equivalence14 and they are available to women at varying additional cost. At times, LBC smears have been marketed as a superior test. As a result, approximately 25% of smears in NZ are performed with LBC methods. The laboratory serving our colposcopy clinic, Canterbury Health Laboratories, offers LBC screening using the SurePath system. In 2005, Canterbury Health Laboratories reported 20 000 LBC and 2400 conventional smears.

Aim

The aim of this study was to compare the performance of SurePath LBC with conventional cytology using randomised smears taken at the time of first assessment colposcopy. The cytology results were compared with the worst histological diagnosis at either first colposcopy, treatment or follow-up colposcopy.

The main outcome measure was to compare the sensitivity of LBC and conventional smears for histologically proven abnormality, human papillomavirus (HPV) changes or worse. The secondary outcome measures were to compare the sensitivity of LBC and conventional cytology for high-grade histological abnormalities (CIN2/3) and for low-grade histological abnormalities (CIN1/HPV). A comparison of unsatisfactory rate and a measure of specificity were also considered important.

Materials and methods

It was estimated that approximately 90% of new patients at our colposcopy service had abnormal histology. It was therefore calculated that a trial with 450 cytology specimens per arm (about 400 cases) offered a power of 80% to detect an increase in sensitivity of LBC from 75 to 83%, a difference of 8%.

Therefore, from 22 December 2004 to 24 July 2006, all new patients referred to the colposcopy clinic at Christchurch Women’s Hospital for the assessment of cervical abnormalities were identified. These women received written information regarding the study.

As per our routine practice, a cervical cytology specimen was taken at the commencement of colposcopic assessment. Subject to the patient’s agreement, a randomisation was performed to decide whether the cervical cytology specimen taken would be processed by conventional means or as a SurePath LBC sample. The randomisation was performed with the use of sequential sealed envelopes. The randomisation sequence was provided by an independent biostatistician who used SAS version 8 with random block lengths with a maximum block length of 24. All randomised samples were accounted for and included in the analysis. All colposcopies were performed by consultant colposcopists or registrars under their supervision in the colposcopy clinic. Colposcopic assessment was recorded on the departmental database and clinical notes.

Cytology and histology specimens were handled and reported in a routine fashion by appropriately trained and qualified staff of Canterbury Health Laboratories. Cytology was reported by the Bethesda 2001 (NZ modified) system.15 Previous histology and cytology reports were available to colposcopists and cytopathology staff through the National Cervical Screening Programme (NCSP) register. Routine internal quality assurance practice was applied as required by the NCSP.

The women were managed in a routine fashion by the colposcopy clinic. Colposcopy, cytology and histology results were collated from the departmental database and notes retrieved to ensure completion of data. An audit of the notes confirmed the accuracy of the database.

The cytology result was correlated with colposcopy findings and histology specimens taken from their colposcopic biopsy, treatment biopsy or colposcopic biopsy at a follow-up visit up to 12 months following the first visit. This information was included in order that the highest grade histology was recorded.

Results were tabulated, and statistical analysis was performed using Statcalc within Epi-Info 3.3.2 (http://www.cdc.gov/epiinfo/) for uncorrected chi-square test for contingency tables and 95% CI for differences in percentages in CIA.16 This study received ethical approval from the Canterbury B Ethics Committee.

Results

Samples from 913 women were randomly allocated to conventional Pap smear or SurePath LBC cytology (Figure 1). The allocated test was performed for 904 samples, but due to oversight by the colposcopist, nine samples were not taken. Four hundred and fifty-three women had a conventional smear and 451 had LBC. Colposcopic assessment of the cervix was recorded for all women. All women whose samples were randomised were accounted for in the analysis. The women’s ages ranged from 16 to 75 years, and the median age was 30 years.

Figure 1.

Flowchart.

The cytological findings taken at initial colposcopy are presented in Table 1. Of the 904 women who had cytology taken, it was reported as unsatisfactory in 6%. Normal cytology was reported in 31%. An epithelial abnormality was reported in 63% of the samples. A HSIL was reported in 21% of the samples and 6.3% were reported as ASC-H (atypical squamous cells present—a HSIL cannot be excluded). 22.7% were reported as a low-grade squamous intraepithelial lesion (LSIL) and 12.9% were reported as ASC-US (atypical squamous cells of undetermined significance). There were no significant differences between the conventional arm and the LBC arm in reporting rates for cytological abnormalities, but SurePath LBC was less likely to be reported as unsatisfactory (2.7 versus 9.1%, P < 0.0001).

Table 1.  Cytology results presented as per randomisation and comparison of cytology result by sample type
RandomisationSurePath LBC (S)Conventional (C)Total S+CComparison S and C
Number randomised457456913
Study smear result*Number S% SNumber C% CNumber S+C% S+CDifference % (S−C)Difference 95% CIP value
  • AGC, abnormal glandular cells any subclassification; ASC-US, atypical squamous cells of undetermined significance; SCC, squamous cell carcinoma.

  • Cytology results for all study samples are tabulated and results are compared for SurePath LBC and conventional smears. The differences are compared and 95% CI applied.

  • *

    Smear results as for modified Bethesda classification 2001 (www.nsu.govt.nz/Files/Bethesda__2001.pdf).

  • **

    Two women had both HSIL and AGC reported and are included in both rows.

SCC00.010.210.1 
HSIL (CIN2/3)9420.89521.018920.9−0.1−5.4 to 5.20.96
ASCH265.8316.8576.3−1.1−4.3 to 2.10.50
LSIL (CIN1/HPV)11024.49521.020522.73.4−2.0 to 8.90.22
ASC-US6113.55612.411712.91.2−3.2 to 5.50.60
AGC**10.220.430.3 
Normal14832.813329.428131.13.7−2.4 to 9.70.23
Unsatisfactory122.7419.1535.9−6.6−9.7 to −3.6<0.0001
Total smears taken451 453 904 
Not taken6 3 9 

The colposcopic and histological assessments of all women were as follows. Nine percent had a single colposcopic assessment and no biopsy, 20% had a single colposcopic assessment and a biopsy, and the remainder had at least two colposcopic assessments with or without a treatment biopsy. The worst histology was identified subsequent to the first colposcopic assessment in 19% (173) of women. There was no difference in assessment between the two groups.

A histologically proven abnormality was identified in 72% (659) of women. The worst histological findings are described in Table 2. Eleven (1.2%) women had microinvasive squamous carcinoma of the cervix, 37.2% had CIN2/3 and 32.7% had CIN1/HPV. In addition, nine (1%) women had adenocarcinoma in situ (ACIS) and one woman was subsequently discovered to have an endometrial adenocarcinoma. There was no significant difference in the proportion of histological abnormalities in the SurePath or conventional arm.

Table 2.  Worst histology by mode of cytology
Worst histologySurePath LBCSurePath LBC% (S)ConventionalConventional % (C)Total%Difference S−C95% CI (S−C)
  1. SCC, squamous cell carcinoma.

  2. The worst or highest grade cervical histology identified is tabulated by randomisation to conventional smears or SurePath LBC. Differences are calculated and 95% CI applied. There are no significant differences between the two groups.

Microinvasive SCC81.830.7111.21.1−0.3 to 2.5
CIN2/316536.117538.434037.2−2.3−8.5 to 4.0
HPV/CIN114231.115734.429932.7−3.4−9.4 to 2.7
Adenocarcinoma in situ51.140.991.00.2−1.1 to 1.5
Normal6814.96113.412914.11.7−2.8 to 6.2
No specimen6915.15612.312513.72.6−1.9 to 7.1
Total457 456 913 
Overall χ2= 5.07, df = 5, P = 0.41 

The relationship of cytology to worst histological diagnosis is identified in Table 3. When all the study patients (including those with unsatisfactory smears and those in whom smears were not performed) are taken into consideration, the overall sensitivity of cytology for any histologically proven abnormality was 76%. There was a trend towards greater sensitivity in the SurePath arm (79.1 versus 73.7%, P = 0.1). This nonsignificant trend was repeated when considering sensitivity for CIN2/3 with a threshold of ASC-US (89.7% for LBC versus 84.6% for conventional, P = 0.1) and HPV/CIN1 (66.2 versus 61.1%, respectively, P = 0.1). It is noted that in the conventional cytology arm, 13 women who had histologically proven high-grade abnormalities had unsatisfactory smears. A further 13 women with low-grade abnormalities had unsatisfactory smears. In comparison, in the SurePath arm, only two women with high-grade and one woman with low-grade histological abnormalities had unsatisfactory smears.

Table 3.  Correlation of sensitivity of cytology in conventional and SurePath arms of study. Unsatisfactory and not taken samples included
Worst histologyCytologySurePath (S)Conventional (C)Difference S–C95%CI difference
No. detectedTotal UnsatisfactoryTotal Not TakenTotal No. of casesSensitivity %No. detectedTotal UnsatisfactoryTotal Not TakenTotal No. of casesSensitivity %
  • ACIS, adenocarcinoma in situ; AGC, abnormal glandular cells any subclassification; SCC, squamous cell carcinoma.

  • Table correlating the worst histological result with the study cytology taken at the time of colposcopic assessment including samples that were not taken or reported as unsatisfactory. The sensitivities are calculated and compared for LBC and conventional smears. The differences are calculated and 95% CI applied.

  • *

    Two women with AIS (one each SurePath and conventional cytology) had both HSIL and AGC noted on cytology these are included in AIS and HSIL row.

Any abnormalityAny abnormal (ASC-US/worse)2533532079250263339735.3−1.2 to 11.8
SCCSCC0 800 30 
SCCHSIL/ASC-H7018883003100 
HSILHSIL (ASC-H/worse)1022216562109131175625.1−2.0 to 12.2
HSILAny abnormal (ASC-US/worse)148221659014813117585 
ACISAGC*1 5201 425 
ACISHSIL/ASC-H400580301475 
LSILAny abnormal (ASC-US/worse)94121426696131157615.1−5.8 to 15.9

If the analysis is performed excluding samples that were not taken or reported as unsatisfactory (Table 4), this trend of improved sensitivity in the LBC arm is no longer seen. Overall, the sensitivity for both LBC and conventional smears was 81%. With a threshold of ASC-H, SurePath had a sensitivity of 63% and conventional smears 68% (P = 0.41) for CIN2/3. With a threshold of ASC-US or worse, both SurePath and conventional smears had a sensitivity of 92%. The sensitivity of cytology for LSIL (CIN1/HPV) was 68% for SurePath and 67% for conventional smears.

Table 4.  Correlation of cytology result and histology excluding samples recorded as unsatisfactory or not taken
Worst histologyCytologySurePath LBC (S)Conventional (C)Difference S–C95%CI (S–C)
No. detectedNo. of cases% (S)No. detectedNo. cases% (C)
  1. This correlates the worst histological result with the cytology result and excludes samples that were not taken or reported as unsatisfactory. The sensitivity and specificity are calculated and compared for LBC and conventional smears. The differences are calculated and 95% CI applied.

 Sensitivity Sensitivity 
Any abnormality HPV or worseAny abnormal (ASC-US/worse)25331281250310811−5.7 to 6.6
CIN2/3ASC-H or worse)1021616310916168−4.4−14.7 to 6.0
CIN2/3(ASC-US or worse)14816192148161920.0−6.0 to 6.0
CIN1/HPVASC-US or worse941396896143670.5−10.5 to 11.4
 Specificity Specificity 
Normal biopsyNormal4164 3551 −4.6−21.9 to 12.8
No biopsy/normal colposcopyNormal4256 3348 6.9−10.6 to 24.4
Combined no biopsy/normal colposcopy and normal biopsyNormal83120696899690.8−11.6 to 13.2

This study was not designed to identify a difference in specificity between the two modalities. However, specificity, as defined by the number of women for whom the worst histology was normal or who had normal colposcopy and no biopsy divided by the number with normal cytology reports, was 69% for both SurePath and conventional smears.

There were 11 women with a histological diagnosis of microinvasive squamous carcinoma. None of these women was identified cytologically by a smear suggestive of invasion, but all ten satisfactory smears reported were at least ASC-H.

There were nine cases of histologically confirmed ACIS, all occurred with CIN2/3. In eight women, HSIL was reported on the smear; in addition, two women (one SurePath and one conventional) also had abnormal glandular cells reported on cytology, and in the last patient, LSIL only was reported.

Discussion

This randomised single institution study compares the sensitivity of SurePath LBC and conventional smears in a colposcopy clinic setting. The overall sensitivity of a single cytology sample for a histologically identified abnormality was 76%. When all the intended cytology samples were taken into account, there was a trend towards improved sensitivity for LBC samples. However, the rate of unsatisfactory cytology samples was significantly higher for the conventional smears, and if unsatisfactory samples (and samples not taken) are excluded from the analysis, the trend towards improved sensitivity in LBC samples was eliminated. With a threshold of ASC-US or worse, both SurePath LBC and conventional smears had a sensitivity of 92% for CIN2/3. A similar specificity was demonstrated in LBC and conventional smears. Neither LBC nor conventional smears specifically identified cases of microinvasive squamous cell carcinoma or the majority of cases of ACIS.

As this is a single institution colposcopy-based study, results cannot be directly translated to the performance of these tests in the population screening situation. However, many studies of the sensitivity of conventional smears are associated with some degree of compromise.1,13,17 In population screening, the incidence of high-grade abnormalities is low, and consequently, very large studies are required to identify a difference in sensitivity between two cytology modalities. As well, it is impractical to colposcope all women with negative smears, which compromises the ability of studies based on population screening to identify false negatives. As a result, such randomised trials are difficult to perform and are rare.

By contrast, in this study, the incidence of high-grade abnormalities is high, the study is randomised and false-negative cytology samples can be identified against a gold standard of colposcopy and biopsy. As well, follow up and treatment biopsies were also included, so that abnormalities not identified at first colposcopy were accounted for. However, it can be argued that in a high incidence setting that the cytology tests sensitivity may differ from those in population screening just because screeners will be expecting to see more abnormalities in the former.

Despite a plethora of studies suggesting improved sensitivity in LBC,2,18 it is notable that a recent large randomised screening study,19 a smaller randomised study with colposcopic verification20 and two recent meta-analyses1,17 did not identify increased sensitivity for high-grade abnormalities from LBC. In this study, when all randomised cases are included, there was a nonsignificant trend towards improved sensitivity in the LBC arm. The study was powered to detect an 8% difference in sensitivity, and therefore, an improvement in sensitivity for a single LBC smear smaller than 8% is possible.

If the results are analysed having excluded unsatisfactory cytology, the sensitivity of LBC and conventional smears was almost identical. The trend towards improved sensitivity in the SurePath arm appears to be explained by the reduced number of women in this group, with histological abnormalities, having unsatisfactory smears. A similar phenomenon was noted by Bergeron et al.21 Currently, in NZ, about 70% of women with unsatisfactory smears have a repeat smear within 12 months.

It is therefore considered that in this clinical context, the two smear modalities offer equivalent sensitivity. The loss of sensitivity due to unsatisfactory smears, however, has the potential to be clinically relevant.

The reduced rate of unsatisfactory results with LBC in our study is similar to that seen in many studies.3,6,7,19,21,22,23 However, in this study, the unsatisfactory rate (5.9%), particularly for conventional smears (9.1%) is higher than the average unsatisfactory rates for Canterbury Health Laboratories (1.5%) and indeed for NZ laboratories (4.4%) This may relate to differences of technique specific to the colposcopy clinic. Given the wide range of unsatisfactory rates reported by NZ laboratories 1.5–6.8%,24 we do not think that our data provides information with respect to whether LBC would be cost-effective in population screening, with relation to reduction of the unsatisfactory rate.

This study was not designed to assess other measures of accuracy of cervical cytology, namely specificity, identification of invasive carcinoma and ACIS. Bearing this in mind, no differences were noted.

In conclusion, in this clinical setting, the conventional smear and SurePath LBC should be considered to have similar sensitivity for the detection of CIN. The advantage of LBC relates to the reduced number of unsatisfactory samples if this technique is used.

Disclosure of interests

The authors have no material interest in the results of this research. Two authors M.W. and H.N. have attended SurePath training courses at BD in Burlington, USA, at the expense of BD.

Contribution to authorship

P.H.S., D.Y.H., H.N. and D.P. contributed to study design, data collection, analysis and manuscript. A.M., M.W. and J.E.W. contributed to study design, analysis and manuscript.

Details of ethics approval

Ethics approval was received from Canterbury B Ethics Committee ref (ctb/04/10/183).

Funding

This work was funded in part with a grant from Canterbury District Health Board.

Acknowledgements

The authors wish to acknowledge the support of the staff and patients of the Christchurch Women’s hospital colposcopy clinic.

Ancillary