Evidence for increasing usage of low-grade squamous intraepithelial lesion, cannot exclude high-grade squamous intraepithelial lesion (LSIL-H) Pap test interpretations

Authors


Abstract

BACKGROUND

Pap test (PT) interpretations of low-grade squamous intraepithelial lesion (LSIL), cannot exclude high-grade squamous intraepithelial lesion (HSIL), or LSIL-H, are used in many laboratories; however monitoring its usage for quality assurance purposes is understudied.

METHODS

PTs from 2005 to 2010 were collected, and yearly frequencies of LSIL, HSIL, LSIL-H, and atypical squamous cells, cannot exclude HSIL (ASC-H) as a function of total PTs and total squamous intraepithelial lesions (SILs) were calculated. Two-year risk of cervical intraepithelial neoplasia 2 (CIN2) or worse (CIN2+) and CIN 3 or worse (CIN3+) was calculated.

RESULTS

A total of 352,220 PTs were identified including 17,301 abnormal PTs. LSIL-H usage increased from 2005 to 2010 (from 0.28% of total PTs in 2005 to 0.61% in 2010, P < .01; from 5.8% of total SILs in 2005 to 12% in 2010, P < .001). HSIL usage decreased significantly from 2005 to 2010 (from 0.7% of total PTs in 2005 to 0.48% in 2010, P = .048; from 14.5% of total SILs in 2005 to 9.5% in 2010, P < .01). Usage of LSIL and ASC-H did not change. Two-year risk of CIN2+ and CIN3+ for HSIL increased significantly from 2005 to 2010 (P < .01). Two-year risk of CIN2+ and CIN3+ for LSIL-H did not change significantly from 2005 to 2010.

CONCLUSIONS

The frequency of LSIL-H interpretations is significantly increasing at our institution, with a significant decrease in HSIL interpretations over the same period. Two-year risk of CIN2+ and CIN3+ for HSIL increased significantly as usage of LSIL-H increased and that of HSIL decreased. Laboratories using LSIL-H may benefit from monitoring its frequency to ensure its appropriate use. Cancer (Cancer Cytopathol) 2014;122:123–7. © 2013 American Cancer Society.

Although not recognized by the 2001 Bethesda System (TBS) for Pap test (PT) classification,[1] the cytologic designation of “low-grade squamous intraepithelial lesion (LSIL), cannot exclude high-grade squamous intraepithelial lesion (HSIL),” or LSIL-H, is now being used regularly in many pathology laboratories. LSIL-H refers to smears fulfilling criteria for both LSIL and atypical squamous cells, cannot exclude HSIL (ASC-H) or smears with squamous intraepithelial lesions (SIL) of indeterminate grade. Various studies investigating LSIL-H relative to the other TBS diagnostic categories have consistently found an intermediate risk of histologic follow-up of cervical intraepithelial neoplasia 2 or greater (CIN2+) relative to LSIL and HSIL and significantly greater prevalence of high-risk HPV (hrHPV) relative to ASC-H.[2-12] For these reasons several investigators have concluded that LSIL-H warrants a distinct diagnostic category in cervical cytology reporting.

Indeterminate categories in any reporting system raise the valid concern of overuse and impact on clinical management. The ratio of atypical squamous cell interpretations to squamous intraepithelial lesion interpretations (ASC/SIL) has been used as a quality assurance measure in Pap testing and is a component of the College of American Pathology (CAP) self-reported gynecologic cytology quality metrics. The goal of monitoring ASC/SIL is to identify overuse of ASC interpretations and give feedback to individual pathologists so criteria can be adjusted appropriately. Similar to the interpretation of ASC, with LSIL-H, there is potential for overuse of in Pap testing, something opponents of using LSIL-H have noted. The potential and impact of overutilization of LSIL-H are understudied in the peer-reviewed literature. The aim of this study was to evaluate the temporal relationship of the LSIL-H diagnosis within our institution to the other TBS diagnostic categories over a period of 6 years.

MATERIALS AND METHODS

The study was approved by the institution review board of the University of Massachusetts Medical School (Project #H-10396). Computer search was used to identify all patients with a cervical PT at the University of Massachusetts Medical School, a tertiary-care academic medical center, between 2005 and 2010. From these cases, all cases of LSIL, ASC-H, LSIL-H, and HSIL were selected for this study. Our laboratory does not have defined criteria or any restrictions in place for LSIL-H usage. Consult PTs were excluded. The yearly frequencies for each of these diagnoses were calculated as a function of total PTs and for LSIL, HSIL, and LSIL-H as a function of total SILs. Follow-up histologic studies including cervical biopsies, endocervical currettings, cervical excisional biopsies (loop electrosurgical excision procedure and conization), and hysterectomy specimens for 2 years were retrieved for LSIL-H and HSIL PTs, and the highest histologic lesion was recorded.

Cytologic Preparation

All PTs were prepared into ThinPrep smears by transferring exfoliated cells into a vial containing PreservCyt fixative (Hologic Corporation, Boxborough, MA) by vigorous agitation of the sampling device against the vial wall. The vials were then transported to the processing laboratory. The ThinPrep 2000 or 3000 automated processor (Hologic) was used to homogenize, collect and transfer a cellular monolayer to a ThinPrep slide. An aliquot of solution was used for hrHPV testing in cotested women or reflexed for testing on the basis of findings of ASCs. Slides were Pap-stained with Tissue-Tek DRS 2000 (Sakura Finetek, Torrance, CA) automated slide stainer using ThinPrep stain protocol and recommended reagents including ThinPrep nuclear stain, ThinPrep rinse solution, ThinPrep bluing solution, ThinPrep Orange G solution, and ThinPrep EA solution. Automated review with the ThinPrep Imaging System (Hologic) identified fields of interest for a cytotechnologist to review. Abnormal smears were referred to a cytopathologist.

Statistical Analysis

Linear regression was used to assess the change in yearly prevalence of HSIL, LSIL-H, LSIL, and ASC-H as a function of both total PTs and SIL (for HSIL, LSIL-H, and LSIL) over time. Log binomial regression was used to assess the linear change in the relative risk of histologic outcomes of CIN2+ and cervical intraepithelial neoplasia 3 or greater (CIN3+) tests over time separately for PTs interpreted as HSIL and LSIL-H. Analysis was performed using STATA version 12.

RESULTS

A total of 352,220 PTs were identified and included in the study. From these cases 12,908 cases of LSIL, 1912 cases of HSIL, 1002 cases of ASC-H, and 1479 cases of LSIL-H were identified. The yearly frequencies of diagnoses for PTs as a function of total PTs and SILs are summarized in Table 1. Interpretations of LSIL-H increased significantly, from 0.28% of total PTs and 5.8% of total SILs in 2005 to 0.61% of total PTs and 12.0% of total SILs in 2010 (P < .01 and P < .001, respectively). Interpretations of HSIL decreased significantly from 0.70% of total PTs and 14.5% of total SILs in 2005 to 0.48% of total PTs and 9.5% of total SILs in 2010 (P = .048 and P < .01, respectively). Interpretations of LSIL and ASC-H and the annual percentage of SILs did not significantly change during the study period. The LSIL-H/HSIL ratio increased steadily, from 0.4 in 2005 to 1.27 by 2010 (Table 2).

Table 1. Yearly Frequencies of HSIL, LSIL, LSIL-H, and ASC-H
 Pap Tests, Year (n) 
 2005 (54,256)2006 (54,680)2007 (53,110)2008 (52,059)2009 (61,288)2010 (76,827)P
HSIL, n (%)383 (0.70)321 (0.58)239 (0.45)294 (0.56)309 (0.50)366 (0.48).048
LSIL, n (%)2106 (3.9)2005 (3.66)1772 (3.33)1762 (3.38)2235 (3.65)3028 (3.94).91
LSIL-H, n (%)153 (0.28)192 (0.35)147 (0.28)219 (0.42)303 (0.49)465 (0.61)< .01
ASC-H, n (%)152 (0.28)125 (0.22)101 (0.19)138 (0.27)201 (0.33)285 (0.37).14
 Squamous Intraepithelial Lesions, Year (n) 
 2005 (2642)2006 (2518)2007 (2158)2008 (2275)2009 (2847)2010 (3859)P
  1. Abbreviations: HSIL, high-grade squamous intraepithelial lesions; LSIL, low-grade squamous intraepithelial lesions; LSIL-H, low-grade squamous intraepithelial lesion, cannot exclude high-grade squamous intraepithelial lesion; ASC-H, atypical squamous cells, cannot exclude HSIL.

% Total Paps4.94.64.14.44.65.0.73
HSIL, n (%)383 (14.5)321 (12.7)239 (11.1)294 (12.9)309 (10.9)366 (9.5)<.01
LSIL, n (%)2106 (79.7)2005 (79.6)1772 (82.1)1762 (77.4)2235 (78.5)3028 (78.4).06
LSIL-H, n (%)153 (5.8)192 (7.6)147 (6.8)219 (9.6)303 (10.6)465 (12.0)< .001
Table 2. LSIL-H/HSIL Ratio by Year
YearLSIL-H/HSIL Ratio
20050.40
20060.60
20070.62
20080.75
20090.98
20101.27
Total0.77

In patients with a referring diagnosis of HSIL, the percentage of patient with CIN2+ follow-up increased significantly, from 66.9% in 2005 to 79.1% in 2010 (P < .001), and the percentage of patients with CIN3+ follow-up increased significantly, from 36.1% in 2005 to 48.1% in 2010 (P < .001). In patients with a referring diagnosis of LSIL-H, the percentage of patients with CIN2+ follow-up and CIN3+ follow-up did not significantly change over the course of the study (Table 3).

Table 3. Two-Year Risk of CIN2+ and CIN3+ for LSIL-H and HSIL by Year
 200520062007200820092010P
LSIL-H
CIN 2+, %37.443.736.944.243.537.1.77
CIN 3+ %12.113.412.613.614.513.1.76
HSIL
CIN 2+66.969.575.876.677.379.1< .001
CIN 3+36.134.943.245.047.248.1< .001

DISCUSSION

LSIL-H is a cytologic interpretation that is equivocal for HSIL and is now recognized to harbor an intermediate risk of CIN2+ compared with LSIL and HSIL.[2-12] Our study of more than 350,000 PTs including more than 17,000 abnormal tests demonstrates that the frequency of LSIL-H interpretations increased at our institution over a 6-year period with a simultaneous decrease in HSIL interpretations. During the same period, interpretations of LSIL and ASC-H did not significantly change. The predictive value of LSIL-H for CIN2+ and CIN3+ did not change during the study period. However, the predictive value of HSIL for CIN2+ and CIN3+ did increase during the study period.

Our study of LSIL-H adds to the literature in several important ways. The majority of studies to date focusing on LSIL-H have aimed to determine the risk of precancerous cervical lesions and the prevalence and types of hrHPV in PTs interpreted as LSIL-H. Investigating the pattern of usage of LSIL-H as a quality control measure has not been a major focus of the majority of prior studies. Nishino and colleagues noted in a recent study variation in the predictive value of LSIL-H for CIN2+ among the pathologists in their group.[13] The authors concluded that monitoring the frequency of LSIL-H usage and monitoring histologic outcomes associated with LSIL-H would be useful feedback for individual pathologists to adjust criteria as appropriate. The increasing year-over-year usage of LSIL-H observed at our institution supports monitoring frequency of LSIL-H interpretations.

Our data definitively demonstrate an increase in the frequency of usage of LSIL-H in Pap testing at our institution. Alsharif and colleagues briefly addressed the frequency of usage of LSIL-H in their study of histologic outcomes from 2004 to 2007.8 They reported that usage of LSIL-H increased from 0.2% of all PTs in 2004 to 0.41% of PTs in 2007. The authors noted that at the beginning of their study LSIL-H had just been introduced to the group and that the increase in interpretations of LSIL-H seemed related to fewer cases being interpreted as ASC-H. The increase in usage of LSIL-H in the current study corresponds to and may be related to the observed decrease in HSIL interpretations and thus differs from Alsharif and colleagues. The differences between the current study and Alsharif's probably relate to the years included in the study. From 2004 to 2007 there were fewer publications and by extension less awareness of LSIL-H as a diagnostic entity in Pap testing, and some increase in usage during this earlier time could simply reflect uptake of the diagnostic entity by groups and individual pathologists. The current study period (2005-2010) is more recent and encompasses a period when more was known about LSIL-H. Our findings suggest that LSIL-H usage is increasing because of changing criteria, specifically an increased threshold for diagnosing HSIL. Scrutiny of the literature suggests that LSIL-H usage could be increasing at other institutions for similar reasons. Table 4 summarizes several studies of LSIL-H by year(s) of study. The frequencies of LSIL-H as a function of total SIL were consistently lower in earlier studies, and LSIL-H/HSIL ratios were typically higher in later studies. In 2010 our group diagnosed LSIL-H more frequently than HSIL (eg, LSIL-H/HSIL ratio >1) for the first time, and another more recent study of LSIL-H has similarly shown an LSIL-H/HSIL ratio >1.13

Table 4. Summary of Studies Investigating LSIL-H.
 Year(s) of Studyn% SILLSIL-H/HSIL% Paps% CIN2+
Elsheikh et al 200642001-2003129,9116.00.340.1540.7
Owens et al 20075200321,2209.20.740.3840
Shidham et al 200710200477,9797.80.760.1933
Al-Nourhji et al 20089200550,9230.6124
Alsharif et al 200982004-2007235,64511.30.780.3033.1
Ince et al 201172004-2009163,3959.40.840.2440
Current study2005-2010352,2209.10.770.4240.4
Nishino et al 2012132009-201011.51.3129.4

Considering the evidence from our study that usage of LSIL-H may be increasing, it follows to ask the questions: what are the effects on the predictive value of LSIL-H and other TBS categories on precancerous cervical lesions, and is this change acceptable? Our results indicate that the risk of CIN2+ and CIN3+ for LSIL-H PT interpretations is not changing despite the increased usage of LSIL-H and despite most additional LSIL-H cases appearing to be cases that would have been previously classified as HSIL. However, the risk of both CIN2+ and CIN3+ for HSIL significantly increased over the period of the study. We hypothesize that as LSIL-H usage has increased in our group; borderline cases of HSIL have gradually been removed from the HSIL category and are now being classified as LSIL-H. The smaller fraction of cases remaining in the HSIL category at the end of the study period are probably smears that were more definitive, and thus the increase in the predictive value of CIN2+ and CIN3+ for HSIL as the category contracted is not surprising. The end result is increased specificity of HSIL for CIN2+ and CIN3+. This may be desirable considering that HSIL cytology can trigger cervical ablative procedures even if cervical biopsies fail to identify CIN2+ and that cytologic HSIL is sometimes followed by a “see and treat” cervical ablative procedure.

The impact of increasing usage of LSIL-H on the management of this interpretation is another question raised by our study. The management of LSIL-H is not entirely straightforward. The 2006 American Society for Colposcopy and Cervical Pathology guidelines for management of abnormal PTs do not offer specific recommendations for LSIL-H,[14] and management is largely based on the clinical judgment of the treating clinician.[15] Because of the intermediate risk of LSIL-H for CIN2+, most agree that initial management should be similar to that of ASC-H with colposcopic referral.[16] A management dilemma could arise if colposcopic evaluation following LSIL-H does not identify a CIN2+ lesion. Even as usage of LSIL-H increased in our cohort, the risk of CIN2+ did not substantially change and was similar to that in prior studies. Thus, we would advocate that LSIL-H be followed by prompt colposcopic examination with biopsy of any identified lesions. At our institution cervical biopsies with referring cytology of LSIL-H are examined with deeper step sections and often p16 immunostaining if initial levels do not reveal CIN2+. Anecdotally, we see occasional cases of CIN2+ that were not diagnosable on initial sections.

In this study we have shown that the frequency of LSIL-H interpretations at our institution is significantly increasing with a significant decrease in HSIL interpretations over the same period. The diminished HSIL category is more predictive of CIN2+ and CIN3+ follow-up, which is advantageous for certain management strategies. Our findings suggest that laboratories using LSIL-H might benefit from following the frequency of LSIL-H usage and associated histologic outcomes to ensure consistent and appropriate use of the diagnosis.

FUNDING SOURCES

No specific funding was disclosed.

CONFLICT OF INTEREST DISCLOSURES

The authors make no disclosures.

Ancillary