Context.—Recent randomized controlled trials have shown a significant decrease in unsatisfactory rates for liquid-based cytology (LBC) compared with conventional Papanicolaou test (CP). The underlying causes and relevance of unsatisfactory results for LBC and CP have never been compared within the setting of a randomized controlled trial.
Objective.—To examine differences in causes and relevance of unsatisfactory and satisfactory but limited by (SBLB) results for LBC and CP.
Design.—Data from the Netherlands ThinPrep Versus Conventional Cytology (NETHCON) trial were used, involving 89 784 women. Causes and relevance of unsatisfactory and SBLB results were analyzed.
Results.—The primary cause for unsatisfactory results for CP and LBC was scant cellularity. Other causes for unsatisfactory CPs were virtually eliminated with LBC. The same was true for SBLB subcategories, with the exception of SBLB absence of transformation zone component and SBLB scant cellularity. The SBLB absence of transformation zone component showed a statistically significant 22% and SBLB scant cellularity a 12% nonsignificant increase with LBC. The detection rates of abnormalities found during 18 months of follow-up of unsatisfactory test results did not differ significantly between the 2 study arms, nor did they differ from the initial test positivity rates from the NETHCON trial.
Conclusions.—Liquid-based cytology shows an almost complete elimination of most causes for unsatisfactory CP, with scant cellularity remaining as the sole cause for unsatisfactory LBC. On the other hand, with LBC a significant increase of smears without a transformation zone component was noted. Women with an unsatisfactory test result are not at increased risk for cervical abnormalities either with LBC or with CP.
Trial Registration.—Nederlands Trial Register, NTR1032, www.trialregister.nl.
Liquid-based cytology (LBC) has been largely adopted as an alternative to cervical cancer screening with conventional Papanicolaou tests (CP) in many Western countries. However, recent evidence from 2 randomized controlled trials (RCTs)1,2 and a meta-analysis comparing the performance of LBC and CP3 supports the conclusion of equal accuracy for both methods. Yet the RCT has shown a significant decrease in unsatisfactory rates for LBC as compared with CP.4
Unsatisfactory tests account for a number of screening failures, and are a source of distress to women and a waste of resources.5 It is assumed that LBC decreases the unsatisfactory rate by reducing the number of cases that are inadequate because of obscuring inflammatory cells and blood or because of inhomogeneous distribution or bad fixation of cells. However, to our knowledge differences in the underlying causes and relevance of unsatisfactory smear results for LBC as compared with CP have never been investigated within the setting of an RCT.
The aim of the study is to examine the underlying causes for inadequacy in LBC and CP, using the data of the Netherlands ThinPrep Versus Conventional Cytology (NETHCON) trial. This comparison includes the Bethesda 1991 category “satisfactory but limited by” (SBLB),6 with special emphasis on slides without endocervical cells or transformation zone component (ec−). This was considered of relevance in view of previous studies that reported increased numbers of tests without endocervical cells while using LBC.7–11 In addition, the clinical relevance of unsatisfactory results was assessed with follow-up cytology and histology.
The data for this study were extracted from the NETHCON trial database. This trial has been described in detail previously.2,4 In short, it concerns a cluster-randomized controlled clinical trial, involving 89 784 women recruited from the Dutch cervical cancer screening program.
The family practices associated with the 2 clinical laboratories were cluster randomized to prevent selection bias by patients' or physicians' preference and for practical reasons. All practices (n = 246) were eligible and were randomly assigned to the experimental arm (LBC with ThinPrep; Hologic Corp, Marlborough, Massachusetts; n = 122) or the control arm (CP; n = 124) and all consented to the outcome of the randomization. The Dutch Ministry of Health, Welfare, and Sport granted ethical approval for the trial. Informed refusal was offered to the participants by means of an information folder. None of the women recruited from the randomized family practices refused and all 89 784 individual participants were included in the study (experimental arm 49 222, and control arm 40 562 individuals). Only participants who received the correct test (ie, the study arm their family practice had been assigned to by randomization; n = 85 076) were included in a per-protocol analysis (experimental arm 46 066 and control arm 39 010 individuals).
The family physician or his/her assistant was responsible for sample taking. Before the start of the trial, family practices allocated to the LBC arm received material for processing LBC samples accompanied by written instructions and additional training by a regional course or by in-office instruction by the manufacturer. Practices assigned to the CP arm were provided with the regular material for preparing CPs. Both LBC and CP samples were collected with the Rovers Cervex-Brush (Rovers Medical Devices BV, Oss, the Netherlands). Cell samples for conventional cytology were spread on a glass slide and fixed within seconds, whereas cell samples for LBC were transferred from the brush to a vial containing PreservCyt fixative and transport medium by rotating and pushing the brush vigorously against the vial wall 10 times.12 The LBC samples were processed at the laboratory with the ThinPrep 3000 processor (Hologic).
Cytologic reporting was done using the Dutch CISOE-A classification system;13 these results are easily translatable to the Bethesda 1991 subcategories atypical squamous cells of undetermined significance (ASCUS)/atypical glandular cells of undetermined significance (AGUS), low-grade squamous intraepithelial lesion (LSIL), and high-grade squamous intraepithelial lesion (HSIL).14 Specimen adequacy was additionally reported, including the subcategories unsatisfactory and SBLB, considering (1) obscuring blood, (2) obscuring inflammation, (3) scant cellularity, (4) poor fixation, (5) mechanical damage, (6) cytolysis, (7) thick area, (8) too few squamous cells with predominantly endocervical cells, and (9) no endocervical or transformation zone component (SBLB ec−). A result of SBLB does not provoke early repeating and serves only as a statement on the limited quality of the smear. The registration system does not facilitate registration of multiple causes, and when applicable the most striking reason was recorded.
The criteria for an unsatisfactory result for CPs were semiquantitative and based on Bethesda 2001,15 as Bethesda 1991 did not define criteria for an unsatisfactory LBC or CP smear. The CP was considered unsatisfactory when more than 75% of the epithelial cells were obscured or could not be clearly visualized. Cellularity was assessed semiquantitatively by counting the number of squamous cells in 25 random fields of view at a 10-fold magnification. A minimum of 25 well-preserved and clearly visualized squamous cells in each field of view is needed for an adequate CP. For LBC, a minimum of 10 fields of view (including the center of the slide) with a ×40 objective should contain a minimum of 4 clearly visualized and preserved squamous cells to achieve a minimum of 5000 cells per slide. However, when atrophic cellular changes were found, these criteria were applied more liberally, both for CP and for LBC. For SBLB, no specific criteria were applied, and this was subjective by nature except for SBLB ec−. Both LBC and CP were considered positive for the presence of transformation zone component when a minimum of 10 endocervical or metaplastic cells or 2 groups of 5 endocervical cells were seen. Criteria for SBLB scant cellularity were limited cellularity with more than 5000 cells per slide for LBC and limited cellularity but with more than 8000 to 12 000 cells per slide for CP.
Follow-up results of unsatisfactory smears included both repeat cytology and histology assessed within 18 months if biopsies were taken. Follow-up data were retrieved from the local pathology databases as well as the national pathology database, which contains the results of all pathology specimens in the Netherlands.16
Proportions are compared with χ2 tests in per-protocol analysis. Crude and adjusted odds ratios (controlling for age, urbanization level, study period, and laboratory site) are computed with univariate and multivariate logistic regression. Period was defined as the first or second half of the study, using the median preparation date as separator. Relative reduction in unsatisfactory and SBLB results was defined as [(Proportionlbc − Proportioncp)/Proportioncp], and the reduction in percentage of total as (Proportionlbc − Proportioncp). Analyses were performed with SPSS 16.0.01 (SPSS Inc, Chicago, Illinois) software.
Between April 2003 and July 2006 a total of 89 784 participants were recruited, of whom 85 076 received the correct test corresponding with the study arm to which their family practice had been assigned (46 066 LBC from 122 practices and 39 010 CP from 124 practices).
In Table 1 the relation between the different determinants and causes for unsatisfactory or SBLB smears are explored with univariate analysis. There was a small but statistically significant difference in unsatisfactory results between low and high urbanization, with more unsatisfactory results found in highly urbanized areas. This was caused mainly by unsatisfactory results based on obscuring blood. Unsatisfactory smears were also related to higher age, which was attributable to a higher proportion of smears that were unsatisfactory because of obscuring inflammation. Unsatisfactory smears were significantly more frequently diagnosed at the Eindhoven study site, with scant cellularity being the only relevant underlying causal factor. Unsatisfactory results were more often diagnosed in the second study period as a result of small variations in smears that were unsatisfactory because of obscuring inflammation, scant cellularity, and too few squamous cells with predominantly endocervical cells.
SBLB results were significantly more frequently found in high-urbanization areas. This was exclusively because of a significantly higher proportion of SBLB ec− results in highly urbanized regions. The overall proportion of SBLB was also related to higher age. The main factor was SBLB ec−, but also SBLB obscuring inflammation and SBLB scant cellularity were significantly more often diagnosed in older women. In contrast, SBLB cytolysis was significantly more frequently found in younger women. Between the study sites, there were significant differences for SBLB for most subcategories. However, the most striking differences were found for SBLB ec−, which was diagnosed more often at the Eindhoven site (15.6%) as compared with Nijmegen (11.1%). Between the first and the second study period, small differences were found for SBLB obscuring inflammation, SBLB scant cellularity, SBLB cytolysis, and SBLB thick area. It is noteworthy that the proportion of SBLB ec− results was not related to study period, in contrast to urbanization, age, and study site.
As shown in Table 2, the primary cause for unsatisfactory results with LBC was scant cellularity (0.30%). This was also the main cause for unsatisfactory CP results, followed by obscuring blood and inflammation. The crude and adjusted odds ratios comparing LBC with CP for the reasons for unsatisfactory and SBLB smears are also shown in Table 2. Adjustment involved logistic regression including the covariates described in Table 1. All subcategories of unsatisfactory and SBLB results showed a significant decrease when using LBC, with 2 exceptions. The proportion of SBLB ec− was significantly increased in the LBC arm, whereas SBLB scant cellularity did not differ significantly between LBC and CP.
In Table 3 the reductions in unsatisfactory and SBLB results for the various subcategories are shown. In the LBC arm, 100% reductions were found for unsatisfactory because of obscuring blood, poor fixation, cytolysis, and insufficient spreading of cells (thick area) and also for SBLB poor fixation. Marked reductions were seen also for unsatisfactory because of obscuring inflammation, mechanical damage, and too few squamous cells with predominantly endocervical cells (87.9%, 92.3%, and 84.6%, respectively). Unsatisfactory because of scant cellularity showed the slightest reduction (37.1%). Most SBLB subcategories were markedly reduced (64.0%–100%). Exceptions were SBLB ec− and SBLB scant cellularity. The proportion of SBLB ec− increased significantly (21.9%); SBLB scant cellularity also increased 11.6%, but this was not statistically significant.
In Table 4 the results of short-term follow-up (cytology or histology) are presented. Only 6.5% (LBC) and 10.1% (CP) of the cases were lost to follow-up. Within 18 months following an unsatisfactory result, 3 cases of atypical squamous cells of undetermined significance or more severe (2.1%) were detected in the LBC arm. In the same time span, 11 cases of atypical squamous cells of undetermined significance or more severe (2.8%) were found in the conventional arm. One case of LSIL+/CIN1+ (0.7%) was found in the LBC arm and 5 cases (1.3%) in the control arm. For HSIL+/CIN 2+, the figures were 1 case (0.7%) and 3 cases (0.8%), respectively. These results all concerned histologically confirmed CIN 2+ lesions. Differences in the detection rates between the 2 study arms were not statistically significant. The test positivity rates of cervical abnormalities found during follow-up of unsatisfactory tests were in line with the cytologic test positivity rates for LBC and CP in the NETHCON trial. The differences were statistically not significant (χ2 test).
Liquid-based cytology significantly reduced the proportion of unsatisfactory specimens from 1.1% to 0.3% and eliminated obscuring blood, poor fixation, cytolysis, and insufficient spreading of cells as causes of unsatisfactory results. Liquid-based cytology also reduced the proportion of SBLB but the relative impact was smaller. Specimen quality and the reasons for unsatisfactory or suboptimal quality (SBLB) varied by study characteristics (urbanization, age, study site, and study period). However, adjustment for these potentially confounding factors did not modify the main findings.
Various subcategories of unsatisfactory and SBLB smears were shown to be impacted by urbanization level, age, study site, and study period. Women living in high-urbanization areas were more likely to have an unsatisfactory result because of obscuring blood but also to have a SBLB ec− result. One can only speculate on the underlying causes. This could be population based, but it is also possible that sampling in general is of lower quality in high urbanization areas. Age also impacted the unsatisfactory rate. Older women were more likely to have a sample called unsatisfactory. As published earlier, this was specifically found in the LBC arm and mainly caused by smears from women in the oldest age group (55–60 years).17 As shown in the Figure, A and B, the same effect of the oldest age group on unsatisfactory results is also found when analyzing the 2 relevant subcategories of unsatisfactory due to obscuring inflammation and scant cellularity separately. Obscuring inflammation was the most important factor for unsatisfactory smears in relation to age, in which atrophic colpitis may be an important determinant. The criteria for cellularity may play a role, too, especially in cases with epithelial atrophy. Age had a variable impact on the SBLB subcategories. The most striking finding was that SBLB ec− was significantly more often diagnosed in women aged 45 years or older. This may also be an effect of less accessibility of the transformation zone in older women12 and more frequent presence of cellular atrophy in older women, in which endocervical cells are more difficult to diagnose. The impact of the study site on the unsatisfactory rate was exclusively due to unsatisfactory result based on scant cellularity. In contrast, most SBLB subcategories showed significant but variable differences between the 2 study sites, illustrating the subjectivity of criteria applied for SBLB in this trial. Apart from SBLB ec−, criteria for SBLB were not consistently defined. Small differences were found for certain subcategories of unsatisfactory and SBLB between the first and second study period. However, SBLB ec−, which is impacted by urbanization, age, and study site, showed consistency during the study, as no significant difference was found between the first and second study period.
The major findings of this study were that the specimen adequacy was significantly improved with LBC with a near elimination of unsatisfactory tests due to obscuring blood, obscuring inflammation, poor fixation, mechanical damage, cytolysis, thick area, and too few squamous cells with predominantly endocervical cells. With LBC, only a limited reduction of 37% was found in unsatisfactory results due to scant cellularity, which remained as the sole cause for unsatisfactory LBC tests. A comparable reduction in the LBC arm was found for most SBLB subcategories, with the exception of SBLB ec−, which showed a statistically significant increase of 21.9%. In contrast to the reduced rate of unsatisfactory because of scant cellularity with LBC, SBLB scant cellularity showed an 11.6% increase compared with CP. However, this increase was not statistically significant.
Only a few other studies have compared the causes for unsatisfactory and SBLB test results for LBC and CP. Lee et al7 found a 29% overall reduction in SBLB but a 19% increase in unsatisfactory results for LBC, mainly caused by scant cellularity. They also reported a 68% increase in SBLB ec−. The authors ascribed this to the split-sample design of the study, in which a CP is prepared first and the less cellular residue is left for LBC. A companion “direct-to-vial” study by Corkill et al18 showed, in contrast to the results presented here, an equivalent rate of SBLB ec−, and they concluded that “it is expected that the rate of representing endocervical component will be maintained when the ThinPrep preparation method is used routinely in place of the conventional cytologic smear method.” 18,(p39) A more recent 2-cohort study by Duggan et al8 also reported causes for unsatisfactory and SBLB test results. However, they found that the frequency of unsatisfactory and SBLB results was similar for ThinPrep and CP. In concordance with our results, they reported a 47% statistically significant increase of SBLB ec− with LBC and also found scant cellularity as the main cause for unsatisfactory results for both LBC and CP. However, this was ascribed to inexperience of the smear takers with the new sampling device that was used for LBC. Moriarty et al19 performed a survey among 674 US laboratories and also found that the most common reason for an unsatisfactory test result was scant cellularity, regardless of whether LBC or CP was applied. They also showed that a large majority of the US laboratories used the Bethesda 2001 criteria for specimen cellularity. More recently, Alsharif et al20 investigated causes and significance of unsatisfactory results with SurePath and also found scant cellularity as primary cause of an unsatisfactory test result.
Scant cellularity is thought to be induced by problems with smear taking and transfer of cells from the sampling device to the slide. In the case of LBC, deficiencies during processing of the slide may also occur. Besides that, subjective assessment or imprecise use of criteria for cellularity by the cytotechnologist also impacts unsatisfactory by scant cellularity rates. This RCT shows that scant cellularity is the major cause for unsatisfactory CP, and de facto almost the sole cause for unsatisfactory LBC results. In this trial the sampling device for both LBC and CP was the Rovers Cervex-Brush, and therefore a difference in sampling device does not explain the difference we found in unsatisfactory rate due to scant cellularity between LBC and CP. Possibly this difference is caused by a less adequate transfer of the sampled cells from the collection device to the slide with CP. But also, the thresholds that are used impact the rates of unsatisfactory by scant cellularity. A low threshold for cellularity may introduce false negative results. Inappropriate reporting of unsatisfactory samples as negative has been shown to be a factor in false-negative cytology preceding cancer and provides clinicians and patients with false reassurance.21,22 A high threshold results in unacceptable unsatisfactory rates with coinciding anxiety and unnecessary costs. In this RCT criteria for cellularity were based on Bethesda 2001. Though the relation between cellularity and the detection potential of a cytologic smear is not a subject of discussion, the minimum degree of cellularity certainly is. The Bethesda 2001 criteria with a minimum of 8000 to 12 000 squamous cells for CP and 5000 cells for LBC were set arbitrarily and are controversial.21,22 Some authors favor setting the threshold for LBC at 10 000 or even more squamous cells as a minimum.
In contrast with the often conflicting results from other studies,7–11,18,23 we found with a high level of evidence that LBC leads to significantly higher proportions of SBLB ec− results. The reason for this observation is unclear. As with scant cellularity, problems with cell sampling with or transfer of cells from the sampling device or during processing of the slide, as well as differences in criteria used for evidence of transformation zone sampling, can induce differences in SBLB ec− rates between CP and LBC. As with unsatisfactory smears due to scant cellularity, a difference in cell sampling between CP and LBC can be excluded, because in both arms the Rovers Cervex-Brush was used. However, the transfer of cells differs between CP and LBC. Whereas for CP the sampled cells are smeared from the Cervex-Brush on the slide, for LBC the Cervex-Brush is rinsed in the vial. It is believed that with LBC most harvested cells are transferred to the vial. However, one could speculate that endocervical cells are less adequately transferred to the vial because they are more likely to be trapped in endocervical mucus that is held by the collection hairs of the Cervex-Brush and thus are less easily dispersed in the PreservCyt fixative. Also, trapping of cylindrical cells in endocervical mucus during the processing of the LBC sample could potentially affect the recovery of these specific cells. The criteria we used for the presence of endocervical cells were based on Bethesda 2001. For both CP and LBC a minimum of 10 endocervical cells or 2 groups of 5 endocervical cells was required to sign the test out as adequately sampled transformation zone (endocervical component present). It is unlikely that these criteria could explain the difference in SBLB ec− rates we found between the 2 study arms. Nevertheless, the actual reasons for the observed difference have to be investigated in more detail in future studies. In most countries, the absence of an endocervical component has no serious consequences. In the 1990s, SBLB ec− was an indication for a repeat smear in the Netherlands. This repeat advice was abolished in view of 2 Dutch studies that found that SBLB ec− did not indicate increased risk for high-grade lesions.24,25 However, a SBLB ec− result still provokes a repeat smear within 6 weeks in the Netherlands when the transformation zone of the cervix is not seen or is considered abnormal by the smear taker. Thus, in the setting of the Dutch cervical screening program, a higher SBLB ec− rate has negative consequences.
In this trial, 18 months of follow-up of unsatisfactory smears showed no statistically significant differences in detection of abnormalities between the 2 study arms, indicating that there is no difference in the relevance of an unsatisfactory result with LBC as compared with CP. The detection rates of cervical abnormalities during 18 months of follow-up of unsatisfactory smears were in line with the cytologic test positivity rates for LBC and CP in the NETHCON trial. This indicates that women with unsatisfactory results by either LBC or CP were not at higher risk for cervical abnormalities during the 18 months of follow-up in this trial. In contrast to our results, Alsharif et al20 found that unsatisfactory SurePath tests had a higher risk of significant abnormalities on follow-up. However, a significant verification bias might have influenced their results.
The unsatisfactory rates in the Netherlands are relatively low. Therefore, in spite of a relative reduction of 70% in unsatisfactory results with LBC, the absolute reduction was only 0.78%. Thus, the added value of LBC in the Netherlands with respect to the reduction in unsatisfactory smears is very limited.
In conclusion, with this RCT we showed with a high level of evidence that LBC results in almost complete elimination of most underlying causes for unsatisfactory CP smears. Scant cellularity remains as the sole cause for an unsatisfactory LBC result. On the other hand, we observed a statistically significant 22% increase in SBLB ec− rates for LBC as compared with CP. Follow-up during 18 months showed that women with unsatisfactory test results by either LBC or CP were not at increased risk for cervical abnormalities.
This study was supported by the European Commission (DG Sanco, Luxembourg, Grand Duchy of Luxembourg) through the European Network for Cervical Cancer Screening (contract SPC.2002475), the European Cancer Network, the European Cooperation on Development and Implementation of Cancer Screening & Prevention Guidelines (ECCG network) and the Dutch Ministry of Health, through the National Institute for Public Health and Environment (contract 25141652). Dr Arbyn was supported by (1) the Belgian Foundation against Cancer (Brussels) and (2) the European Commission (Directorate of SANCO, Luxembourg, Grand-Duchy of Luxembourg), through the ECCG project (IARC, Lyon, France) and the 7th Framework Programme of DG Research of the European Commission through the PREHDICT project (grant 242061, coordinated by the Vrije Universiteit Amsterdam, the Netherlands).
From the Department of Pathology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands (Drs Siebers and Bulten); the Laboratory of Pathology, PAMM Laboratories, Eindhoven, The Netherlands (Dr Klinkhamer); the Department of Pathology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands (Ms Vedder); and the Unit of Cancer Epidemiology, Scientific Institute of Public Health, Brussels, Belgium (Dr Arbyn).
The authors have no relevant financial interest in the products or companies described in this article.