Pooled analysis of the performance of liquid-based cytology in population-based cervical cancer screening studies in China

Authors

  • Qin-Jing Pan MD,

    Corresponding author
    1. Department of Pathology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    • Corresponding author: You-Lin Qiao, MD, Department of Epidemiology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, 17 South Panjiayuan Lane, PO Box 2258, Beijing 100021, People's Republic of China; Fax: (011) 86-10-6771-3648; qiaoy@cicams.ac.cn

      and Qin-Jing Pan, MD, Department of Pathology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, 17 South Panjiayuan Lane, PO Box 2258, Beijing 100021, People's Republic of China; Fax: (011) 86-10-6771-3648; pqjing@hotmail.com

    Search for more papers by this author
  • Shang-ying Hu MD, PhD,

    1. Department of Epidemiology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    Search for more papers by this author
  • Xun Zhang MD,

    1. Department of Pathology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    Search for more papers by this author
  • Pu-wa Ci MPH,

    1. Department of Epidemiology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    Search for more papers by this author
  • Wen-hua Zhang MD,

    1. Department of Gynecological Oncology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    Search for more papers by this author
  • Hui-qin Guo MD,

    1. Department of Pathology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    Search for more papers by this author
  • Jian Cao MD,

    1. Department of Pathology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    Search for more papers by this author
  • Fang-hui Zhao MD, PhD,

    1. Department of Epidemiology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    Search for more papers by this author
  • Alice Lytwyn, MD,

    1. Department of Pathology and Molecular Medicine, Juravinski Hospital, Hamilton, Ontario, Canada
    Search for more papers by this author
  • You-lin Qiao MD, PhD

    Corresponding author
    1. Department of Epidemiology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People's Republic of China
    • Corresponding author: You-Lin Qiao, MD, Department of Epidemiology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, 17 South Panjiayuan Lane, PO Box 2258, Beijing 100021, People's Republic of China; Fax: (011) 86-10-6771-3648; qiaoy@cicams.ac.cn

      and Qin-Jing Pan, MD, Department of Pathology, Cancer Institute and Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, 17 South Panjiayuan Lane, PO Box 2258, Beijing 100021, People's Republic of China; Fax: (011) 86-10-6771-3648; pqjing@hotmail.com

    Search for more papers by this author

  • We thank the local physicians and the women who participated in the current study from Beijing, Gansu, Jiangsu, Jiangxi, Henan, and Shanxi.

Abstract

BACKGROUND

Liquid-based cytology (LBC) has been widely used for cervical cancer screening. Despite numerous studies and systematic reviews, to the authors' knowledge few large studies to date have focused on biopsy-confirmed cervical lesions and controversy remains concerning its diagnostic accuracy. The objective of the current study was to assess LBC for detecting biopsy-confirmed cervical intraepithelial neoplasia (CIN) and cancer.

METHODS

A pooled analysis of LBC using data from 13 population-based, cross-sectional, cervical cancer screening studies performed in China from 1999 to 2008 was performed. Participants (n = 26,782) received LBC and human papillomavirus testing. Women found to be positive on screening were referred for colposcopy and biopsy. The accuracy of LBC for detecting biopsy-confirmed CIN of type 2 or worse (CIN2+) as well as CIN type 3 or worse (CIN3+) lesions was analyzed.

RESULTS

Of 25,830 women included in the analysis, CIN2+ was found in 107 of 2612 with atypical squamous cells (4.1%), 142 of 923 with low-grade squamous intraepithelial neoplasia (15.4%), 512 of 784 with high-grade squamous intraepithelial neoplasia (65.3%), 29 of 30 with squamous cell carcinoma (96.7%), 4 of 27 with atypical glandular cells (14.8%), and 85 of 21,454 with normal cytology results (0.4%). No invasive cancers were found to have atypical squamous cells, atypical glandular cells, or cytologically normal slides. The overall sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of LBC for detecting CIN2+ were 81.0%, 95.4%, 38.3%, 99.3 %, and 94.9%, respectively. Although Hybrid Capture 2 was more sensitive than LBC, the specificity, positive predictive value, and overall accuracy of LBC were higher than those of Hybrid Capture2 at 85.2%, 18.6%, and 85.5%, respectively.

CONCLUSIONS

The results of the current study indicate that the performance of LBC can effectively predict the risk of existing CIN2+ and may be a good screening tool for cervical cancer prevention in a developing country. Cancer (Cancer Cytopathol) 2013;121:473-82. © 2013 American Cancer Society.

INTRODUCTION

Because of its large population and the imbalance in economic development, the disease burden of cervical cancer varies greatly in China. Whereas women in urban areas may have the opportunity to access cervical screening, in rural areas, due to a lack of medical resources, the incidence of cervical cancer is relatively high and has reached 81 per 100,000 population in some areas.[1] To explore a strategy for cervical cancer screening suited to conditions in China, the Cancer Institute and Hospital of the Chinese Academy of Medical Science (CICAMS) in Beijing, China, in collaboration with institutions at home and abroad, conducted a series of population-based, cross-sectional cervical cancer screening studies from 1999 through 2008.[2-15]

Liquid-based cytology (LBC) was first introduced into China in 1999, and was evaluated in all the studies above-mentioned. LBC is now widely used for cervical cancer screening in China and is preferred to conventional cytology for its lower rate of unsatisfactory slides, shorter time needed for interpretation, and the ability to use the same sample for human papillomavirus (HPV) and other molecular testing. Although some studies have reported that LBC increased the detection of cytological abnormalities and demonstrated an improved sensitivity, to our knowledge few large studies to date have focused on biopsy-confirmed cervical lesions,[16, 17] and none have been based on an indigent population at high risk for the development of cervical cancer. To better delineate cervical cancer screening diagnoses and outcomes of abnormal LBC, we performed a pooled analysis of LBC screening at CICAMS using the data from 13 population-based, cross-sectional, cervical cancer screening studies. We assessed the overall accuracy of LBC for the detection of biopsy-confirmed cervical intraepithelial neoplasia (CIN) and cancer in high-incidence areas in China.

MATERIALS AND METHODS

Participants

CICAMS, in collaboration with the Cleveland Clinic (Cleveland, Ohio), Program for Appropriate Technology in Health (PATH) (Seattle, Wash), and the International Agency for Research on Cancer (IARC) (Lyon, France), screened women from 1999 to 2008. Our pooled analysis used individual patient data from 13 cross-sectional, population-based studies in which LBC was performed at CICAMS. The 13 studies were: 4 projects from the Shanxi Province Cervical Cancer Screening Study (SPOCCS 1, SPOCCS 2, SPOCCS 3-Henan, and SPOCCS 3-Xiangyuan), which were performed with the Cleveland Clinic from 1999 to 2006[2, 3, 5-10, 15]; 5 projects of the Screening Technologies to Advance Rapid Testing (START 2003, 2004, 2005, 2006, and 2007), which were performed with PATH[11, 15]; 1 project performed in 2004 with IARC in Yangcheng, Shanxi province[4, 15]; the FastHPV trial performed in 2007[14, 15]; the HPV and CIN Prevalence Survey performed in Jiangsu in 2008[12, 15]; and the Hybrid Capture 2 (HC2) clinical trial performed in 2008.[13, 15] Eligible women were aged 15 years to 59 years, sexually active, and not pregnant; had an intact uterus; and had no history of CIN, cervical cancer, or pelvic radiation. None had been screened for cervical cancer within the past 5 years, and all provided written informed consent. Recruitment for all studies was based on community lists to minimize selection bias. The Human Subjects Review Boards of CICAMS, the Cleveland Clinic, PATH, and IARC approved these studies.

Procedures

Women included in the pooled analysis all concurrently received LBC (SurePath [BD Diagnostics, Franklin Lakes, NJ] or ThinPrep [Hologic, Bedford, Mass]), HPV DNA testing with the HC2 assay (Digene Corporation, Gaithersburg, Md), and direct visual inspection of the cervix after the application of 5% acetic acid solution (VIA). The study methods for each individual study have been outlined in detail elsewhere,[2-15] and are described briefly below.

All participants underwent collection of cervical specimens and VIA in the local health centers. All specimens were collected by physicians. Specimens for HPV tests were obtained from the endocervix with a conical-shaped brush. The conical brushes were placed in Universal Collection Media (Digene Corporation). Specimens for LBC were obtained with a broom-type sampling device (Cervex-Brush; Rovers Medical Devices B.V., Oss, the Netherlands) and then transferred into a vial containing a fixative liquid. For processing the liquid-based cervical samples, 9 studies used the SurePath slide preparation technique in which cervical samples were placed into SurePath preservative fluid. Three studies (SPOCCS 1, START 2004, and the HPV and CIN prevalence study) used the ThinPrep slide preparation technique in which cervical samples were placed in PreservCyt solution (Hologic Inc, Marlborough, Mass) and in 1 study (START 2003), samples from the first half of screening participants were placed in SurePath preservative fluid and samples from the second half of the screening participants were placed in PreservCyt solution.[18] All specimens were stored at room temperature and sent weekly to CICAMS (Beijing, China) to be processed in a centralized laboratory. After the collection of cervical specimens, all participants underwent VIA. The participants also used visual inspection with Lugol iodine (VILI) if VIA was negative in 6 of the studies (START 2004, START 2005, START 2006, IARC-Yangcheng, FastHPV, and HC2 trial). The definition of positive VIA and VILI results have been described in detail elsewhere.[11] All visual inspection was performed by trained Chinese gynecologists.

In the centralized laboratory of CICAMS, HPV tests and LBC were performed. For LBC, ThinPrep 2000 (Hologic Inc) and SurePath were used. Cytology results were reported according to the Bethesda System (TBS). The cytological classifications were normal, atypical squamous cells (ASC), low-grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion (HSIL), squamous cell carcinoma (SCC), atypical glandular cells (AGC), adenocarcinoma in situ (AIS), or adenocarcinoma. To match the data, ASC represents atypical squamous cells of undetermined significance (ASC-US) for the projects conducted before 2003 (ie, SPOCCS 1 and SPOCCS 2), according to TBS 1991; for the projects conducted from 2003 onward, ASC represents the combined categories of ASC-US and ASC-H (atypical squamous cells, cannot exclude high-grade squamous intraepithelial lesion), in accordance with TBS 2001.[19]

The cytological diagnoses were made by cytopathologists at CICAMS. In addition, in the 5 START studies, all abnormal cytology slides and a percentage of negative slides that were randomly selected were reviewed by an external international expert; any disagreements were reviewed together with the external consultant and the cytopathologists of CICAMS, and a consensus was reached for the final diagnoses. For HPV tests, carcinogenic HPV DNA testing was performed with the high-risk probe set of HC2, which detects a pool of 13 carcinogenic HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68). HPV DNA positivity was defined according to the manufacturer's recommendation of a positive cutpoint of 1.0 relative light units per cutoff (approximately equal to 1.0 pg DNA/mL).[2] In addition to using the specimen collected directly by physicians, 7 studies (14,329 samples) used the liquid remaining after the LBC slide preparation for HPV DNA testing (the SPOCCS 1 study, all 5 START studies, and the IARC-Yangcheng study).

Regardless of screening status, all women in SPOCCS 1 underwent colposcopy and biopsy after VIA. In SPOCCS 2, SPOCCS 3, START 2003, and START 2004, women who were found to have positive results on any screening test were referred for colposcopy and biopsy, and in the other studies, women underwent colposcopy and biopsy if they had ASC-US and were positive for HPV, or if their cytology was ASC-H (ie, ASC-H, LSIL, HSIL, SCC, AGC, AIS, or adenocarcinoma). Directed biopsy was taken from all visible cervical lesions. When the 4-quadrant punch biopsy method was indicated, biopsies were taken at positions of 2 o'clock, 4 o'clock, 8 o'clock, and 10 o'clock depending on the quadrant and an endocervical curettage was performed. Histological diagnoses were categorized as negative, CIN type 1 (CIN1), CIN type 2 (CIN2), CIN type 3 (CIN3), SCC, AIS, or adenocarcinoma. All the biopsies were performed within 2 months of cytology sampling in the local health centers. Biopsy diagnoses were made by pathologists of CICAMS, and were reviewed by international experts in 11 studies.

Verification of Disease Endpoints

The current study combined individual raw data from 13 studies and estimated pooled sensitivity and specificity for the detection of histological CIN2 or worse (CIN2+) and CIN3 or worse (CIN3+). In all the studies, the clinical results and other test results were masked from the laboratory personnel who performed the HPV analyses and the cytopathologists and histopathologists who did the reviews and made diagnoses did not know the results of other tests. Results of the specific screening tests were masked from colposcopists, but they were aware that one of the tests was positive.

To minimize verification bias, we unified the criteria to verify disease status for final diagnoses and applied it to all participants. The gold-standard diagnosis was a histologically confirmed biopsy result. However, women with no biopsy results but either negative or ASC-US findings on cytology and who were negative for HPV DNA were deemed to have true-negative results (very low risk for high-grade lesions) on the basis of findings from SPOCCS 1 that there was only 1 case of CIN2 and no case of CIN3+ (1 of 1511 cases [0.07%]). Women without a biopsy result were also categorized as being negative for disease if they were positive for HPV or had an inadequate HPV test but had negative cytology and a negative colposcopic examination. Women who had no biopsy results were judged as having incomplete data if they had ASC-US findings and were positive for or had inadequate HPV, had ASC-H+, were positive for HPV with negative cytology but had missing data or a positive colposcopy findings, had unsatisfactory cytology results, or had inadequate HPV testing results (ie, the specimen was not sufficient for HC2 testing) in circumstances other than when both cytology and colposcopy was negative.

Statistical Analysis

We assessed the accuracy of LBC testing for the detection of CIN2+ and CIN3+. Women with incomplete data were excluded from the analyses. Sensitivities were pooled with a fixed-effect model, and specificities were pooled with a random effect model because interstudy heterogeneity was found to be statistically significant. We used forest plots to display the variations of sensitivity and specificity in the individual studies and pooled measures. The Q test and I2 test were used to assess the heterogeneity between the studies. Sensitivities, specificities, positive predictive values (PPVs), and negative predictive values (NPVs) were calculated. Pooled analyses were performed with Meta-DiSc (version 1.4; available at http://www.hrc.es/investigacion/metadisc_en.htm). The corrected accuracy was computed with the estimated number of cases in which the lesion was CIN2+ and CIN3+. Trends in pathology with cytological results were calculated using the Cochran-Armitage test for trend. Continuous variables were examined by calculating the means, medians, and standard deviations. A P value ≤ .05 was considered to be statistically significant. R 2.11.1 statistical software (www.r-project.org) was used for all other analyses.

RESULTS

A total of 26,782 women from 13 population-based studies were screened. Of these, 560 (2.1%) women were excluded because they did not have satisfactory cytology results (Fig. 1). Altogether, 26,222 women were found to have satisfactory cytological results and were included in the current study. The mean age of the women who were included in the cytologic analysis was 40.07 years ± 6.32 years (range, 18 years-59 years); 1063 women were aged < 30 years and 1497 women were aged > 50 years. The majority of women were from rural areas (26,116 women; 99.6%).

Figure 1.

Flow chart of study participants is shown. LSIL indicates low-grade squamous intraepithelial lesion; ASC-US, atypical squamous cells of undetermined significance; HPV, human papillomavirus.

The LBC diagnoses for each project are shown in Table 1. The rate of abnormal cytology, including ASC, LSIL, HSIL, AGC, and SCC, was 17.4% among 26,222 women, and ranged from 8.9% in the START 2005 trial to 25.7% in the SPOCCS 1 trial. The rate of ASC diagnoses was 10.3%, ranging from 6.3% in START 2005 to 14.9% in SPOCCS 1. The rate for LSIL, HSIL, and SCC combined was 7.1%, ranging from 2.6% in START 2005 to 10.9% in SPOCCS 1. For AGC, the overall rate was 0.1%. The ratio for ASC compared with LSIL, HSIL, and SCC combined was 1.5 (range 1.1-3.2).

Table 1. Reporting Rates of LBC for Various Bethesda Categories by Project
Project Name (Year; Location)LBC, No. (%)
NormalASCaLSILHSILAGCSCCTotal
  1. Abbreviations: AGC, atypical glandular cells; ASC, atypical squamous cells; HC2, Hybrid Capture 2; HPV, human papillomavirus; HSIL, high-grade squamous intraepithelial lesion; IARC, International Agency for Research on Cancer; LBC, liquid-based cytology; LSIL, low-grade squamous intraepithelial lesion; SCC, squamous cell carcinoma; SPOCCS, Shanxi Province Cervical Cancer Screening Study; START, Screening Technologies to Advance Rapid Testing.

  2. a

    To match the data, ASC represents atypical squamous cells of undetermined significance (ASC-US) for the projects conducted before 2003, according to the Bethesda System 1991; for the projects conducted from 2003 onward, ASC represents the combined categories of ASC-US and atypical squamous cells cannot exclude high-grade squamous intraepithelial lesion, in accordance with the Bethesda System 2001.

SPOCCS 1 (1999; Xiangyuan County, Shanxi Province)1480 (74.3)296 (14.9)91 (4.6)96 (4.8)18 (0.9)12 (0.6)1993 (100.0)
SPOCCS 2 (2001; Xiangyuan and Yangcheng County, Shanxi Province)6643 (78.2)990 (11.7)535 (6.3)307 (3.6)15 (0.2)7 (0.1)8497 (100.0)
SPOCCS 3-Henan (2006; Ximi, Henan Province)787 (89.5)67 (7.6)13 (1.5)10 (1.1)0 (0.0)2 (0.2)879 (100.0)
SPOCCS 3-Xiangyuan (2006; Xiangyuan County, Shanxi Province)734 (83.0)114 (12.9)16 (1.8)19 (2.2)0 (0.0)1 (0.1)884 (100.0)
START 2003 (2003; Xiangyuan County, Shanxi Province)1666 (83.8)202 (10.2)52 (2.6)62 (3.1)0 (0.0)6 (0.3)1988 (100.0)
START 2004 (2004; Xiushui County, Jiangxi Province)2017 (85.7)184 (7.8)49 (2.1)103 (4.4)0 (0.0)1 (0.0)2354 (100.0)
START 2005 (2005; Wudu County, Gansu Province)1863 (91.1)129 (6.3)30 (1.5)22 (1.1)1 (0.1)1 (0.1)2046 (100.0)
START 2006 (2006; Qinxian County, Shanxi Province)2000 (85.1)240 (10.2)51 (2.2)58 (2.5)0 (0.0)0 (0.0)2349 (100.0)
START 2007 (2007; Xiangyuan and Wuxiang, County, Shanxi)2018 (86.2)214 (9.2)46 (2.0)62 (2.7)0 (0.0)0 (0.0)2340 (100.0)
IARC-Yangcheng (2004; Yangcheng County, Shanxi Province)603 (81.9)81 (11.0)20 (2.7)30 (4.1)2 (0.3)0 (0.0)736 (100.0)
FastHPV trial (2007; Qinxian County, Shanxi Province)674 (82.4)96 (11.7)16 (2.0)30 (3.7)1 (0.1)1 (0.1)818 (100.0)
Prevalence survey (2008; Binhai and Jintan County, Xuzhou City, Jiangsu Province)257 (85.4)28 (9.3)10 (3.3)5 (1.7)1 (0.3)0 (0.0)301 (100.0)
HC2 trial (2008; Xiangyuan County, Shanxi Province)909 (87.7)69 (6.7)36 (3.5)22 (2.1)1 (0.1)0 (0.0)1037 (100.0)
Total21,651 (82.6)2710 (10.3)965 (3.7)826 (3.2)39 (0.2)31 (0.1)26,222 (100.0)

Among 26,222 participants with satisfactory cytology results, 392 without biopsy results were excluded (Fig. 1), leaving a total of 25,830 women included in the final analysis. The presence of CIN and SCC according to cytology diagnosis in studies that used TBS 1991and TBS 2001 terminology is shown in Tables 2 and 3, respectively. The overall rates of CIN2+; LSIL, HSIL, and SCC combined; and LSIL/HSIL+ (including HSIL and SCC) before 2003 were 4.4%, 10.0%, and 1.5%, respectively, and were higher than those of the studies conducted from 2003 to 2008 (2.7%, 4.5%, and 0.8%, respectively; P < .0001). Using data from all 13 studies, 3.4% of women had CIN2+. CIN 2+ was found in 107 of 2612 women with ASC (4.1%), 142 of 923 women with LSIL (15.4%), 512 of 784 women with HSIL (65.3%), 29 of 30 women with SCC (96.7%), and 4 of 27 women with AGC (14.8%). No histologic adenocarcinoma was identified. Of 21,454 women with normal cytology results, 0.4% (85 women) had a diagnosis of CIN2 or CIN3. No women with invasive cancers had ASC, AGC, or cytologically normal slides.

Table 2. Correspondence Between LBC Diagnoses Using TBS 1991 and Biopsy Outcomes for the 2 Projects Before 2003
 Biopsy, No. (%)
LBCNormalCIN1CIN2CIN3SCCTotal
  1. Abbreviations: AGC, atypical glandular cells; ASC-US, atypical squamous cells of undetermined significance; CIN, cervical intraepithelial neoplasia; HSIL, high-grade squamous intraepithelial lesion; LBC, liquid-based cytology; LSIL, low-grade squamous intraepithelial lesion; SCC, squamous cell carcinoma; TBS, the Bethesda System.

Normal7914 (97.7)136 (1.7)37 (0.5)12 (0.2)0 (0.0)8099 (100.0)
ASC-US1158 (90.4)83 (6.5)26 (2.0)14 (1.1)0 (0.0)1281 (100.0)
LSIL351 (56.5)171 (27.5)57 (9.2)41 (6.6)1 (0.2)621 (100.0)
HSIL75 (18.6)76 (18.9)91 (22.6)138 (34.2)23 (5.7)403 (100.0)
AGC18 (81.8)1 (4.6)2 (9.1)1 (4.6)0 (0.0)22 (100.0)
SCC0 (0.0)1 (5.3)3 (15.8)6 (31.6)9 (47.4)19 (100.0)
Total9516 (91.1)468 (4.5)216 (2.1)212 (2.0)33 (0.3)10,445 (100.0)
Table 3. Correspondence Between LBC Diagnoses Using TBS 2001 and Biopsy Outcomes for the 11 Projects Conducted Between 2003 and 2008
 Biopsy, No. (%)
LBCNormalCIN1CIN2CIN3SCCTotal
  1. Abbreviations: AGC, atypical glandular cells; ASC-US, atypical squamous cells of undetermined significance; ASC-H, atypical squamous cells cannot exclude high-grade squamous intraepithelial lesion; CIN, cervical intraepithelial neoplasia; HSIL, high-grade squamous intraepithelial lesion; LBC, liquid-based cytology; LSIL, low-grade squamous intraepithelial lesion; SCC, squamous cell carcinoma; TBS, the Bethesda System.

Normal13,169 (98.6)150 (1.1)25 (0.2)11 (0.1)0 (0.0)13,355 (100.0)
ASC-US1084 (87.2)119 (9.6)24 (1.9)16 (1.3)0 (0.0)1243 (100.0)
ASC-H51 (58.0)10 (11.4)17 (19.3)10 (11.4)0 (0.0)88 (100.0)
LSIL138 (45.7)121 (40.1)33 (10.9)10 (3.3)0 (0.0)302 (100.0)
HSIL68 (17.9)53 (13.9)99 (26.0)153 (40.2)8 (2.1)381 (100.0)
AGC4 (80.0)0 (0.0)0 (0.0)1 (20.0)0 (0.0)5 (100.0)
SCC0 (0.0)0 (0.0)1 (9.1)3 (27.3)7 (63.6)11 (100.0)
Total14,514 (94.3)453 (2.9)199 (1.3)204 (1.3)15 (0.1)15,385 (100.0)

Excluding cases with AGC, the rate of CIN2+ generally increased with the cytology grade (chi-square test, 7313.4; P < .0001). In the 11 studies that used the TBS 2001 classification, CIN2+ was present in 3.2% of women with ASC-US and in 30.7% of women with ASC-H.

The sensitivities and specificities of LBC testing for detecting CIN2+ are shown in Figure 2 and those for the detection of CIN3+ are shown in Figure 3. For the 2 studies using TBS 1991 terminology, the threshold for a positive LBC test was LSIL (ie, either LSIL, HSIL, SCC, AGC, AIS, or adenocarcinoma) and for the other 11 studies using TBS 2001, the threshold was ASC-H (ie, ASC-H, LSIL, HSIL, SCC, AGC, AIS, or adenocarcinoma). The pooled sensitivities for CIN2+ and CIN3+ were 81% and 89%, respectively. The sensitivity varied little by project. The pooled specificity of LBC for CIN2+ was 95%, with 11 of 13 projects having specificities of ≥ 96%. The pooled specificity of LBC for CIN3+ was 94%, with 10 of 13 studies having specificities of ≥ 95%. Specificities varied significantly between studies performed before 2003 and those performed from 2003 onward for both CIN2+ and CIN3+ (both P < .0001).The pooled specificities of LBC for both CIN2+ and CIN3+ had narrow 95% confidence intervals, which varied from < 2% (CIN2: 95% confidence interval, 95%-96%; CIN3, 95% confidence interval, 94%-95%).

Figure 2.

Forest plots of pooled and individual study sensitivities and specificities of liquid-based cytology for cervical intraepithelial neoplasia of type 2 or worse are shown. SPOCCS indicates Shanxi Province Cervical Cancer Screening Study; START, Screening Technologies to Advance Rapid Testing; IARC, International Agency for Research on Cancer; HPV, human papillomavirus; HC2, Hybrid Capture 2; df, degrees of freedom; 95% CI, 95% confidence interval.

Figure 3.

Forest plots of pooled and individual study sensitivities and specificities of liquid-based cytology for cervical intraepithelial neoplasia of type 3 or worse are shown. SPOCCS indicates Shanxi Province Cervical Cancer Screening Study; START, Screening Technologies to Advance Rapid Testing; IARC, International Agency for Research on Cancer; HPV, human papillomavirus; HC2, Hybrid Capture 2; df, degrees of freedom; 95% CI, 95% confidence interval.

To compare the accuracy of LBC and HC2 testing, the results from the 25,404 (of 25,830) women with adequate HPV testing and satisfactory LBC were examined. Although the sensitivities of HC2 for detecting CIN2+ and CIN3+ were significantly higher than the sensitivities of LBC (P < .0001 for all), the specificities, PPV, and accuracy of LBC for detecting CIN2+ and CIN3+ were greater than those of HC2 (P < .0001) (Table 4).

Table 4. Accuracy of LBC Compared With HC2 Testing for Detecting CIN2+ and CIN3+
DetectingSensitivity, % (95% CI)Specificity, % (95% CI)PPV, % (95% CI)NPV, % (95% CI)Youden's index % (95% CI)Accuracy, % (95% CI)
  1. Abbreviations: 95% CI, 95% confidence interval; CIN2+, cervical intraepithelial neoplasia of type 2 or worse; CIN3+, cervical intraepithelial neoplasia of type 3 or worse; HC2, Hybrid Capture 2; LBC, liquid-based cytology; NPV, negative predictive value; PPV, positive predictive value.

LBC (N=25,404)      
CIN2+81.0 (78.2-83.5)95.4 (95.1-95.7)38.3 (36.0-40.5)99.3 (99.2-99.4)76.4 (73.3-79.2)94.9 (94.6-95.2)
CIN3+88.5 (85.2-91.3)94.3 (94.0-94.6)22.0 (20.1-24.0)99.8 (99.7-99.8)82.8 (79.2-85.9)94.2 (93.9-94.5)
HC2 (N=25,404)      
CIN2+96.3 (94.8-97.4)85.2 (84.7-85.6)18.6 (17.4-19.7)99.9 (99.8-99.9)81.5 (79.5-83.1)85.5 (85.1-86.0)
CIN3+97.4 (95.4-98.6)83.9 (83.4-84.3)9.9 (9.0-10.8)99.9 (99.9-100.0)81.2 (78.8-82.9)84.1 (83.6-84.5)

In 10,656 cases, HPV testing was performed on the liquid that remained after a sample of the LBC was removed for cytology, and in these cases, the unsatisfactory rate was 0.1% (9 of 10,656 cases). The accuracy of HPV testing performed on the remaining liquid sample was similar to that performed on the physician-obtained cervical swabs (sensitivity, 94.9%; specificity, 88.3%; Youden's index, 83.9%; accuracy, 88.5%; PPV, 15.9%; and NPV, 99.9% for detecting CIN2+).

DISCUSSION

To the best of our knowledge, the current study, with pooled analysis from multiple studies performed in China and including large numbers of women with abnormal cytology results, is the first report of a large-scale, population-based study to examine the diagnostic accuracy of LBC as a primary screening tool for histologically confirmed CIN2+ and CIN3+ detection in China.

Greater than 99% of the women in the current study lived in rural areas. None of the study participants, regardless of residence, had undergone cervical screening within the 5 years before study enrollment. The prevalence of CIN2+ ranged from 1.2% to 4.4%. These rates are least 8 times higher than those reported in developed countries.[20]

The unsatisfactory rate for LBC specimens in the current study was 2.1%, which is consistent with other studies and lower than that observed in the studies using conventional Papanicolaou (Pap) tests.[16, 21]

Our overall rates for ASC (10.3%) and for LSIL, HSIL, and SCC combined (7.1%) were higher than those reported by other studies.[21] Nevertheless, the ratio of ASC to LSIL, HSIL, and SCC combined (1.52 in the current study) was consistent with most high-quality studies,[21] and indicates that the higher rates of abnormal cytology reports in our center were unlikely due to over-interpretations.

As is not unexpected with screening tests, the correlation between LBC and the biopsy diagnosis was not perfect. However, the percentage of women with negative outcomes fell as the severity of the LBC diagnosis increased, from 87.8% among women with ASC to 53.1% after LSIL, 18.2% after HSIL, and 0.0% after a SCC smear. Conversely, the probability of a biopsy demonstrating CIN2+ rose with the increasing grade of cytological abnormality, from 4.1% among women with ASC-US to 15.6%, 65.3%, and 96.7%, respectively, among those with LSIL, HSIL, and SCC. Similarly, the risk of invasive cancer rose from 0.11% among women with LSIL to 3.9% among women with HSIL and 53.3% among those with SCC smears. The vast majority (> 98%) of cytologically normal cases also had normal histology. CIN3 and CIN2 were present in few cases (0.1% and 0.3%, respectively) of normal cytology, and no invasive cancers were found in women with smears that were determined to be ASC or normal.

Similar to other studies, ASC constituted the majority (59.3%) of abnormal cytology interpretations in the current study, and among these, few had positive biopsy findings. Using TBS 2001 classification in our studies conducted from 2003 onward helped to refine the risk of CIN2+ among women with ASC (while ASC-H constituted only 6.6% of cases termed ASC, CIN2+ was present in 30.7% of women with ASC-H and in only 3.2% in those with ASC-US). Although not part of the current study, the literature has shown that women with ASC-US can be triaged further using high-risk HPV DNA[22, 23]; LBC has the advantage over conventional Pap tests in that HPV testing can be performed on residual LBC material. In the current study, the majority of LBC had sufficient residual material for testing; only 0.1% of cases could not be tested because of insufficient material.

In the current study, AGC was uncommon and was reported in only 0.1% of cases. None of these women was found to have a glandular lesion on further investigation; 18.5% (5 of 27 women) had CIN and 81.5% had negative evaluations. The low detection rates for cervical granular lesions in this study were also indicated by the literature.[21, 24] Although we cannot exclude the possibility that AGC was underdiagnosed at our institution, the finding that squamous lesions predominate among women with AGC smears has been reported generally.[24, 25] Because the reliable diagnoses for women with AGC may require cervical conization, endometrial biopsy, and other potentially invasive and costly procedures,[23] the AGC smears in these studies deserve further validation.

By pooling our results from several studies, we presented an overall picture of the sensitivity, specificity, PPV, and NPV of LBC. Over the last 10 years, numerous studies have reported the accuracy of LBC,[16, 17, 26-32] but to the best of our knowledge only a limited number have used biopsy-confirmed cervical lesions as the reference standard[16, 17, 26, 31] and few studies have performed population-based studies with such a large number of women with cervical abnormalities.[16, 30, 31] Our overall sensitivity, specificity, PPV, and NPV for detecting CIN2+ were 81.0%, 95.4%, 38.3%, and 99.3 %, respectively; and those for detecting CIN3+ were 88.5%, 94.3%, 22.0%, and 99.8%, respectively. The studies appeared homogeneous in their estimates of sensitivities. Statistically significant heterogeneity was detected among the estimates for specificity for the detection of both CIN2+ and CIN3+. This is attributed to the difference in specificities in projects performed before 2003 compared with those from 2003 onward. Nevertheless, specificities in all cases were over 90%, and we do not think that the differences before and from 2003 onward were significant from a practical or “clinical” perspective.

The overall accuracy of LBC in the current study appears to be higher than those in most reports on conventional Pap testing and LBC.[16, 17, 26, 30, 31] However, there were several important differences between the current study and other studies. First, the current study population consisted of women who had rarely or never been screened and who had a high prevalence of high-grade CIN2+. Most other studies involved women who were well screened, having been enrolled from areas in which cytologic screening had been in place for many years, and therefore comparably fewer CIN2+ lesions were present.[16, 17, 31] Second, our interval from screening to colposcopy examination and biopsy was generally < than 2 months. Other studies have included lesions detected within 1 year or more from screening to colposcopy.[16, 31] It is possible that immediate referral led to the detection of a higher percentage of lesions that otherwise would have regressed, and thus increased the accuracy of the LBC screen. Finally, published reports have been a summary of test results from multiple laboratories,[16, 30, 31] whereas the data in our study were all from 1 laboratory. The potential of variation of test accuracy from different laboratories was avoided in the current study.

When we compared LBC with HC2 testing, we found that LBC testing had higher accuracies for detecting CIN2+ and CIN3+ than HC2 (94.9% and 94.2% vs 85.5% and 84.1%). This was primarily due to the higher specificity of LBC. However, HC2 was more sensitive than LBC for detecting CIN2+ and CIN3+.

The results of the current study indicate that the performance of LBC is effective for the detection of CIN2+ and cervical cancer in a high-incidence area. All the studies summarized in this article were performed in a high-quality laboratory in Beijing, and may not be generalizable to all the cytology laboratories in China. However, the results from this study can serve as a relevant baseline for the risk of histological abnormalities after an abnormal LBC diagnosis and indicate that if centralization of cytology services to a high-quality laboratory is possible, then LBC may be a good screening tool for cervical cancer prevention in a developing country.

FUNDING SUPPORT

Supported by the Fogarty International Clinical Research Scholars Program (Fogarty International Center, National Institutes of Health) (R24 TW007988) and the Academic Capacity Development Program of the Beijing Municipal Commission of Education (grant XK100230447).

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary