The Pittsburgh cervical cancer screening model: a risk assessment tool.
* Context.--Evaluation of cervical cancer screening has grown
increasingly complex with the introduction of human papillomavirus (HPV)
vaccination and newer screening technologies approved by the US Food and
Objective.--To create a unique Pittsburgh Cervical Cancer Screening Model (PCCSM) that quantifies risk for histopathologic cervical precancer (cervical intraepithelial neoplasia [CIN] 2, CIN3, and adenocarcinoma in situ) and cervical cancer in an environment predominantly using newer screening technologies.
Design.--The PCCSM is a dynamic Bayesian network consisting of 19 variables available in the laboratory information system, including patient history data (most recent HPV vaccination data), Papanicolaou test results, high-risk HPV results, procedure data, and histopathologic results. The model's graphic structure was based on the published literature. Results from 375 441 patient records from 2005 through 2008 were used to build and train the model. Additional data from 45 930 patients were used to test the model.
Conclusions.--The PCCSM compares risk quantitatively over time for histopathologically verifiable CIN2, CIN3, adenocarcinoma in situ, and cervical cancer in screened patients for each current cytology result category and for each HPV result. For each current cytology result, HPV test results affect risk; however, the degree of cytologic abnormality remains the largest positive predictor of risk. Prior history also alters the CIN2, CIN3, adenocarcinoma in situ, and cervical cancer risk for patients with common current cytology and HPV test results. The PCCSM can also generate negative risk projections, estimating the likelihood of the absence of histopathologic CIN2, CIN3, adenocarcinoma in situ, and cervical cancer in screened patients.
Conclusions.--The PCCSM is a dynamic Bayesian network that computes quantitative cervical disease risk estimates for patients undergoing cervical screening. Continuously updatable with current system data, the PCCSM provides a new tool to monitor cervical disease risk in the evolving postvaccination era.
(Arch Pathol Lab Med. 2010;134:744-750)
Cervical cancer (Diagnosis)
Cervical cancer (Care and treatment)
Health risk assessment (Methods)
Health risk assessment (Usage)
Austin, R. Marshall
Druzdzel, Marek J.
|Publication:||Name: Archives of Pathology & Laboratory Medicine Publisher: College of American Pathologists Audience: Academic; Professional Format: Magazine/Journal Subject: Health Copyright: COPYRIGHT 2010 College of American Pathologists ISSN: 1543-2165|
|Issue:||Date: May, 2010 Source Volume: 134 Source Issue: 5|
|Geographic:||Geographic Scope: United States Geographic Code: 1USA United States|
Evaluation of the factors involved in cervical cancer (CxCa)
screening have grown increasingly complex with the introduction of human
papillomavirus (HPV) vaccination1 and the newer screening technologies
approved by the US Food and Drug Administration (FDA). Use of HPV
vaccines enter a CxCa screening environment that has shifted
dramatically in the United States during the past 6 to 13 years with the
introduction of FDA-approved liquid-based cytology (LBC), (2,3)
computer-assisted screening, (4,5) widespread high-risk HPV (hrHPV) DNA
reflex testing after findings of atypical squamous cells of undetermined
significance (ASCUS) from Papanicolaou (Pap) tests, (6) and the
introduction of Pap and HPV DNA cotesting in women 30 years and older.
(7) The many testing and prevention options raise several new challenges
when assessing how current test results and prior clinical history
affect the risk stratification of individual patients.
We employed decision science tools of a dynamic Bayesian-network model analysis to form a unique Pittsburgh Cervical Cancer Screening Model (PCCSM) (8) that addresses these CxCa screening risk assessment challenges in an environment predominantly using the newer screening technologies now prevalent in the United States. The model is a dynamic Bayesian network that combines several sources of knowledge, including published medical literature, expert opinion, and extensive objective hospital data. The model is a convenient tool for assessing a patient's prospective risk for histopathologic diagnosis of cervical precancer (cervical intraepithelial neoplasia [CIN] 2, CIN3, and adenocarcinoma in situ [AIS]) or invasive CxCa. The model's quantitative assessments were initially constructed to assist in routine laboratory identification of patients at high risk for the quality-control rescreening mandated in federal regulations. (9) Furthermore, the model quantifies how multiple current testing and historic data variables together influence prospective risk for a histopathologic diagnosis of cervical precancer or invasive CxCa.
MATERIALS AND METHODS
This study was approved by the Magee-Womens Hospital (MWH) Institutional Review Board (Pittsburgh, Pennsylvania). The data available for analysis consisted of 421 371 Pap test results collected during a 4-year period (2005-2008) at the MWH of the University of Pittsburgh Medical Center (Table). The MWH serves a significantly older-than-average, (10) relatively lower-risk, US population and uses LBC, location-guided computer-assisted screening, and HPV testing as its primary cervical screening tools. Most (97%-98%) of the Pap tests performed during the study period were liquid-based ThinPrep (Hologic Inc/Cytyc Corporation, Marlborough, Massachusetts) Pap tests, which were screened using the ThinPrep Imaging System (Hologic). (5) The reporting profile of the laboratory is documented in many recent publications. (11-16) Approximately 11% of cytology-screened cases were followed by surgical and histopathologic procedures (45 115 data entries), whereas about 19% of all cytologic data entries were associated with hrHPV DNA test results (81 848 hrHPV DNA test results). Some of the cases were accompanied by additional patient history data from screening test-requisition forms, which, most recently, included the history of HPV vaccination. The HPV vaccination history was first recorded by offices on Pap requisition forms in January 2008 and is the only new variable to be introduced during the 4-year study period. As of June 30, 2009, 841 patients have had a history of HPV vaccination recorded on Pap requisition forms. Depending on the historic variable, history data were available for 2% to 30% of all cytology entries. Each patient was further characterized by the demographic variables of age and race.
Dynamic Bayesian Networks
Bayesian networks, (17) also called belief networks or causal networks, are acyclic-directed graphs that model probabilistic influences among variables. The graphic part of a Bayesian network forms the structure of the modeled problem, whereas local interactions among neighboring probability distributions are quantified using actual system data. Bayesian networks have proven to be powerful tools for modeling complex problems involving uncertain knowledge. They have been practically employed in a wide variety of fields, including engineering, the physical sciences, and medicine, with some models reaching the size of hundreds or thousands of variables. Dynamic Bayesian networks are a temporal extension of Bayesian networks that allow for modeling of dynamic processes. The hidden Markov model is considered to be the simplest dynamic Bayesian network. (18) Considerable work on dynamic models in medicine has been carried out by Leong and collaborators, (19,20) who, in addition to developing Bayesian networks and dynamic Bayesian networks, have successfully used a combination of graphic models with Markov chains to address problems in different medical domains, including colorectal cancer management, neurosurgical intensive care unit monitoring, and cleft lip and palate management. (21) Other applications of dynamic Bayesian networks in medicine include NasoNet, a system for diagnosis and prognostication of nasopharyngeal cancer, (22) and a dynamic Bayesian network developed by investigators in the Netherlands for management of patients with carcinoid tumor. (23) Other reported biologic applications of dynamic Bayesian networks include predicting the secondary structure of a protein, (24) modeling peptide fragmentations (25) and cellular systems, (26) and identifying gene regulatory networks from time-course, microarray data. (27)
Pittsburgh Cervical Cancer Screening Model
The PCCSM is a dynamic Bayesian network consisting of 19 variables, including the Bethesda System (28) cytologic variables (source of Pap test, Pap test type, Pap test result, and Pap test adequacy category), histopathologic and surgical data (surgical procedures and histopathologic results), and hrHPV DNA test results (as positive, negative, or not available). The model also includes patient history data available in the laboratory information system, such as history of infections, history of cancer, history of contraception, history of abnormal cytology, menstrual history, HPV vaccination history since January 2008, and the demographics age and race. We based the structure of the model, that is, the interactions among the modeled variables, on existing published medical evidence enhanced with expert opinion and independence tests performed on patient records. Most of the variables were discrete. The only continuous variable (age) was discretized into 5 intervals: younger than 20, 20 to 29, 30 to 44, 45 to 59, and 60 years and older. Figure 1 represents the graphic structure of a static version of the PCCSM, which preceded the building of a dynamic version of the model capable of computing future risk estimates. The model was quantitatively parameterized by means of training data collected for almost 4 full years (2005-2008) and consisting of 375 441 patient records. The model was created and tested using SMILE (Structural Modeling, Inference, and Learning Engine), an inference engine, and GeNIe, a development environment for reasoning in graphic probabilistic models, both developed in the Decision Systems Laboratory of the University of Pittsburgh and available in the public domain at no cost to users (http://genie.sis.pitt.edu/; accessed October 20, 2009). After building and training the model, 45 930 patient cases from a 5-month period (April-August 2008) not included in the training data set were analyzed. Patient ages ranged from 12 to 95 years (mean, 42.17 years; SD, 15.71). Current cytology result data from these patients were entered into the model along with patient history findings and current available hrHPV DNA test results. These data are referred to as testing data. The dynamic PCCSM can generate CIN2, CIN3, AIS, and CxCa risk projections over variable future periods. Results presented in the following section represent the projected risk for histopathologically verifiable CIN2, CIN3, AIS, and CxCa over specified periods, ranging from the time of screening to 3 years. Quantitative outputs of the PCCSM can be further analyzed with statistical tests of significance (eg, Z tests, analyses of variance). (29) Risk projections for different categories of current cytologic and HPV results can be compared to investigate whether or not risk variations between specific groups of patients are statistically significant.
Given current available cytology and HPV test results and patient history data, the PCCSM was first used to estimate the relative risk of histopathologically verifiable CIN2, CIN3, AIS, and CxCa for groups of individuals, with each discrete Bethesda System category of current cytologic result and each possible current hrHPV DNA test result category (positive, negative, or not available). Figure 2 shows quantitative relative risk projections at 2 years. Quantitative risk projections varied substantially according to the degree of current cytologic abnormality. Risk also varied according to hrHPV DNA test status. The degree of cytologic abnormality was the largest positive predictor of histopathologic precancer or CxCa risk, whether or not hrHPV DNA test results were available. The highest risk was projected for the group of patients with current cytologic findings of high-grade squamous intraepithelial lesions or worse (including cytologic findings of cancer or cytologic findings suspicious for cancer) and those with a positive hrHPV DNA test result. The lowest risk was projected for the group of patients with current negative cytology results and those who had negative findings for hrHPV DNA cotest results. Other combinations of cytology results and hrHPV DNA test results, including no HPV test result available, showed intermediate risk projections, as presented in Figure 2, arranged with lower risk levels on the right and higher risk levels on the left.
[FIGURE 1 OMITTED]
Effect of Prior History on CIN2, CIN3, AIS, and CxCa Risk
The PCCSM was next used to assess the effect of different prior history records on quantitative risk for histopathologic CIN2, CIN3, AIS, and CxCa of patients with common current Pap and HPV results. The PCCSM allows the entry for each patient to include all available history from the 2004 to 2008 study period. Figure 3 shows the 2-year relative-risk projections for histopathologic CIN2, CIN3, AIS, and CxCa for patients presenting with current ASCUS Pap and positive hrHPV DNA test results. Model results showed that risk varied substantially with different patient history. For example, patients with all previously negative findings on Pap tests constituted the lowest-risk group, whereas those patients who, in the past, had any histopathologically verified CIN2, CIN3, AIS, or CxCa or any cytologic high-grade squamous intraepithelial lesion results constituted the highest-risk group. Figure 4 similarly shows that the 2-year relative risk for CIN2, CIN3, AIS, and CxCa also varied with different prior history records for patients with current ASCUS Pap test and negative hrHPV DNA test results; however, the levels of risk were generally lower than those shown in Figure 3 for patients with current ASCUS Pap test results and a positive hrHPV DNA test result.
The PCCSM was also used to estimate the negative predictive value of current test results. Figure 5 shows model projections for the likelihood of not having a histopathologic diagnosis of CIN2, CIN3, AIS, and CxCa over periods ranging from 1 to 3 years for 5 groups with various combinations of current low-risk screening test results. Figure 5 includes projections for patients with current negative Pap and current negative hrHPV DNA cotest results, for those with current negative Pap results and no available current hrHPV DNA test results, for those with any current Pap result and current negative hrHPV DNA test results, for those with current ASCUS Pap results and negative reflex hrHPV DNA test results, and finally for those with current negative Pap and positive hrHPV DNA combined test results. The greatest likelihood for absence of a CIN2, CIN3, AIS, and CxCa diagnosis was in the group with current negative Pap and HPV cotest results. Even though results appear similar in Figure 5 for the group with negative Pap results and no HPV test results available (NA), statistical analysis confirmed that each of the slightly higher negative predictive values associated with double-negative Pap and HPV cotest results was statistically significant (Z test, P , .01).
[FIGURE 2 OMITTED]
The PCCSM constitutes a large, dynamic Bayesian network-modeled database that reflects prevalent, current use in the United States of newer, FDA-approved screening technologies. (2,5-7) The model allows for computation of quantitative estimates of the effect of multiple variables of risk for a studied diagnostic outcome, that is, detection of histopathologic cervical precancer (CIN2, CIN3, or AIS) and invasive CxCa. The model, therefore, is able to identify groups of patients who are at progressively lower or higher risk for having a subsequent histopathologic diagnosis of CIN2, CIN3, AIS, and CxCa. Furthermore, the PCCSM prospective-risk projections reflect not only variable combinations of current screening test results but also prior history data.
A number of modeling approaches to CxCa have been reported to address questions about the role and cost-effectiveness of new screening or prevention techniques, optimal screening frequency, the potential of risk-adapted screening policies, and adherence to screening policy. (30-36) These models have primarily been transition models that simulate the natural history of disease. Such models have been widely used and regarded as valid approaches for epidemiologic projections and as guidance for CxCa screening, diagnosis, and treatment decisions. One significant limitation of many modeling approaches has been reliance on published, older, historic, and international data sets, which may precede or differ substantially from newer, FDA-approved, cervical screening methods of the modern era that are in widespread use and are prevalent in the United States, including liquid-based cytology, computer-assisted imaging, and adjunctive HPV reflex or routine cotesting. (2-7) According to manufacturer estimates reported periodically to the Securities and Exchange Commission, LBC now comprises around 95% of the annual US Pap-test market, and computer-assisted imaging is being used on most of those samples. More than 85% of findings of ASCUS from Pap tests in the United States are now being followed by reflex HPV testing, and more than 25% to 30% of women 30 years and older are being routinely screened with cytology and HPV cotesting. Older data sets may also not reflect the significantly increasing glandular histology of CxCa noted in the United States and elsewhere during the past several decades. (37,38) The PCCSM is the first reported model, to our knowledge, based on extensive, recent, US data using newer, FDA-approved, cervical screening technologies. (11-15) Reliance of models on international data from health systems where screening methods and policies on screening frequency, follow-up, and treatment are substantially different from the United States may have less application in the United States. (39,40)
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
Both the current and intermediate (1-3 years) future risk of histopathologic CIN2, CIN3, AIS, or CxCa diagnoses in the PCCSM were most strongly correlated with the degree of cytologic abnormality. These observations are consistent with numerous published observations on the positive predictive value of cervical cytology. (39) The PCCSM projections also reflect the utility of adjunctive HPV cotesting with equivocal abnormal cytology results because the combination of equivocally abnormal cytology and positive HPV test results show increased positive predictive values, which support diagnostic testing referral, whereas the negative predictive value of negative HPV tests allows efficient triage of patients to routine, periodic retest screening. (6,13) The quantitative effects of prior history on risk projections for patients with common current cytology and HPV test results (Figures 3 and 4) has not, to our knowledge, been previously reported but is generally consistent with the broad literature indicating the effect of prior history on future risk. (41,42) The PCCSM projections of high, negative predictive value for double-negative (cytology negative and hrHPV negative cotest results) are also consistent with observations that patients with both negative Pap test and hrHPV test results are at very lowrisk for current or prospective diagnosis of cervical precancer or CxCa. (43,44) The high, negative predictive values projected by the PCCSM for imaged LBC are consistent with recently published results reflecting the use of imaging and LBC (11,12) and with recent international clinical trial data using LBC. (45) The PCCSM projections differ from some other results reported in screening trials using manually screened conventional Pap smears. (46) These observations emphasize the importance of examining data sets reflecting the use of newer, FDA-approved screening technologies.
[FIGURE 5 OMITTED]
Beginning in January 2008, the University of Pittsburgh Medical Center Pap test requisition forms began listing HPV vaccination history among clinical history variables to be potentially checked off by system cervical screening providers and office staff. Since then, 841 patients have had this history recorded on Pap test requisition slips. This small number of data entries has not yet measurably affected PCCSM risk projections. Although this method of documenting HPV vaccination history clearly has major limitations, nevertheless the continuously updated risk projections of the PCCSM should prove to be of great interest over time as a new modeling method to assess postvaccination screening strategies. (31,47) Furthermore, the vaccination component of the database could be updated in the future if a local or national vaccine registry were to be developed. The PCCSM quantitative risk assessments may also prove to be of use in the future as a personalized aid in clinical management and follow-up decision-making. Early efforts exploring these possibilities are underway in our health system. The PCCSM identifies numerous promising paths for research and investigation.
We thank Karen Lassige, MS, for her invaluable help in retrieving data from MWH's CoPathPlus database (Cerner DHT, Inc, Waltham, Massachusetts). We also acknowledge MWH cytology manager, Nancy Mauser, MDM, for assistance in reviewing individual cytology reports, and MWH lead cytotechnologist, Jonee Matsko, BS, for assistance in identifying cytologyhistology correlates.
(1.) Garland SM, Hernandez-Avila M, Wheeler CM, et al. Quadrivalent vaccine against human papillomavirus to prevent anogenital diseases. N Engl J Med. 2007;356(19):1928-1943.
(2.) Lee KR, Ashfaq R, Birdsong GG, Corkill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and a fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol. 1997;90(2):278-284.
(3.) Bishop JW, Bigner SH, Colgan TJ, et al. Multicenter masked evaluation of AutoCyte PREP thin layers with matched conventional smears including initial biopsy results. Acta Cytol. 1998;42(1):189-197.
(4.) Wilbur DC, Prey MU, Miller WM, Pawlick GF, Colgan TJ. The AutoPap System for primary screening in cervical cytology: comparing the results of a prospective, intended-use study with routine manual practice. Acta Cytol. 1998;42(1):214-220.
(5.) Davey E, d'Assuncao J, Irwig L, et al. Accuracy of reading liquid-based cytology slides using the ThinPrep Imager compared with conventional cytology: prospective study. BMJ. 2007;335(7609):31.
(6.) Solomon D, Schiffman M, Tarone R. Comparison of three management strategies for patients with atypical squamous cells of undetermined significance: baseline results from a randomized trial. J Natl Cancer Inst. 2001;93:293-299.
(7.) Zahn CM. Human papillomavirus DNA testing as an adjunct in primary screening: is it prime time? Obstet Gynecol. 2004;103(4):617-618.
(8.) Austin RM, Onisko A, Druzdzel M. The Pittsburgh Cervical Cancer Screening Model. Cancer Cytopathol. 2008;14S:345A.
(9.) Austin RM, Onisko A, Druzdzel M. Bayesian network model analysis as a quality control and risk assessment tool in cervical cancer screening: biennial meeting of ASCCP. J Lower Genital Tract Dis. 2008;12:160-161A.
(10.) ACS. 2006 US Census Bureau American Community Survey (ACS). http:// www.census.gov/acs/www/. Accessed June 23, 2009.
(11.) Zhao C, Florea A, Onisko A, Austin RM. Histologic follow-up results in 662 patients with Pap test findings of atypical glandular cells: results from a large academic womens hospital laboratory employing sensitive screening methods. Gynecol Oncol. 2009;114(3):383-389.
(12.) Bansal M, Austin RM, Zhao C. High-risk HPV DNA detected in less than 2% of more than 25,000 cytology negative imaged liquid-based Pap test samples from women 30 and older. Gynecol Oncol. 2009;115(2):257-261.
(13.) Bandyopadhyay S, Austin RM, Dabbs D, Zhao C. Adjunctive human papillomavirus DNA testing is a useful option in some clinical settings for disease risk assessment and triage of females with ASC-H Papanicolaou test results. Arch Pathol Lab Med. 2008;132(12):1874-1881.
(14.) Zhao C, Austin RM. Human papillomavirus DNA detection in ThinPrep Pap test vials is independent of cytologic sampling of the transformation zone. Gynecol Oncol. 2007;107(2):231-235.
(15.) Zhao C, Austin RM. Adjunctive high-risk human papillomavirus DNA testing is a useful option for disease risk assessment in patients with negative Papanicolaou tests without an endocervical/transformation zone sample. Cancer Cytopathol. 2008;114(4):242-248.
(16.) Armah H, Austin RM, Dabbs D, Zhao C. Follow-up findings for women with human papillomavirus-positive and atypical squamous cells of undetermined significance screening test results in a large women's hospital practice. Arch Pathol Lab Med. 2009;133(9):1426-1430.
(17.) Pearl J. Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. San Mateo, California: Morgan Kaufmann Publishers Inc; 1988.
(18.) Friedman, N., Murphy, K., Russell, S. Learning the structure of dynamic probabilistic networks. In: Proceedings of Fourteenth Conference on Uncertainty in Artificial Intelligence; July 24-26, 1998; Madison, WI. San Mateo, CA: Morgan Kaufmann; 1998:139-147.
(19.) Harmanec D, Leong TY, Sundaresh S, et al. Decision analytic approach to severe head injury management. In: Proceedings of the 1999 AMIA Annual Fall Symposium; Washington, DC: American Medical Informatics Association; 1999: 271-275.
(20.) Leong TY. Multiple perspective dynamic decision making. Artif Intell. 1998;105(1-2):209-261.
(21.) Xiang Y, Poh KL. Time-critical dynamic decision modeling in medicine. Comput Biol Med. 2002;32(2):85-97.
(22.) Galan SF, Aguado F, Diez F, Mira J, NasoNet, modeling the spread of nasopharyngeal cancer with networks of probabilistic events in discrete time. Artif Intell Med. 2002;25(3):247-264.
(23.) van Gerven MAJ, Taal BG, Lucas PJF. Dynamic Bayesian networks as prognostic models for clinical patient management. J Biomed Inform. 2008;41(4): 515-529.
(24.) Yao XQ, Zhu H, She ZS. A dynamic Bayesian network approach to protein secondary structure prediction. BMC Bioinformatics. 2008;9:49.
(25.) Klammer AA, Reynolds SM, Bilmes JA, MacCoss MJ, Noble WS. Modeling peptide fragmentation with dynamic Bayesian networks for peptide identification. Bioinformatics. 2008;24(13):i348-i356.
(26.) Ferrazzi F, Sebastiani P, Isaac S, et al. Dynamic Bayesian networks in modelling cellular systems: a critical appraisal on simulated data. In: CBMS 2006: The 19th IEEE International Symposium on Computer-Based Medical Systems; June 22-23, 2006; Salt Lake City, UT. New York, NY: IEEE Computer Society; 2006:544-549.
(27.) Zou M, Conzen SZ. A new dynamic Bayesian network (DBN) approach for identifying gene regulatory networks from time course microarray data. Bioinformatics. 2005;21(1):71-79.
(28.) Solomon D, Davey D, Kurman R, et al. The 2001 Bethesda System: terminology for reporting results of cervical cytology. JAMA. 2002;287(16):2114- 2119.
(29.) Kirkwood B, Sterne J. Essentials of Medical Statistics. Malden, MA: Blackwell Science; 2003.
(30.) Muhlberger N, Sroczynski G, Esteban E, et al. Cost-effectiveness of primarily human papillomavirus-based cervical cancer screening in settings with currently established Pap screening: a systematic review commissioned by the German federal Ministry of Health. Int J Technol Assess Health Care. 2008;24(2): 184-192.
(31.) Goldhaber-Fiebert JD, Stout NK, Salomon JA, Kuntz KM, Goldie SJ. Cost-effectiveness of cervical cancer screening with human papillomavirus DNA testing and HPV-16,18 vaccination. J Natl Cancer Inst. 2008;100(5):308-320.
(32.) Goldhaber-Fiebert JD, Stout NK, Ortendahl J, Kuntz KM, Goldie SJ, Salomon JA. Modeling human papillomavirus and cervical cancer in the United States for analyses of screening and vaccination. Popul Health Metr. 2007;5:11.
(33.) Siebert U, Sroczynski G, Hillemanns P, et al. German cervical cancer screening model: development and validation of a decision-analytic model for cervical cancer screening in Germany. Eur J Public Health. 2006;16(2):185-192.
(34.) Bidus MA, Maxwell GL, Kulasingam S, et al. Cost-effectiveness analysis of liquid-based cytology and human papillomavirus testing in cervical cancer screening. Obstet Gynecol. 2006;107(5):997-1005.
(35.) Holmes J, Hemmett L, Garfield S. The cost-effectiveness of human papillomavirus screening for cervical cancer. Eur J Health Econ. 2005;5(1):30-37.
(36.) Myers ER, McCrory DC, Nanda K, Bastian L, Matchar DB. Mathematical model for the natural history of human papillomavirus infection and cervical carcinogenesis. Am J Epidemiol. 2000;151(12):1158-1171.
(37.) Smith HO, Tiffany MF, Qualls CR, Key CR. The rising incidence of adenocarcinoma relative to squamous cell carcinoma of the uterine cervix in the United States--a 24-year population-based study. Gynecol Oncol. 2000;78(2):97-105.
(38.) Kinney W, Sawaya GF, Sung HY, Kearney KA, Miller M, Hiatt RA. Stage at diagnosis and mortality in patients with adenocarcinoma and adenosquamous carcinoma of the cervix diagnosed as a consequence of cytologic screening. Acta Cytol. 2003;47(2):167-171.
(39.) Miller AB. The (in)efficiency of cervical screening in Europe. Eur J Cancer. 2002;38(3):321-326.
(40.) Ronco G, van Ballegooijen M, Becker N, et al., Process performance of cervical screening programmes in Europe, Eur J Cancer. 2009;45(15):2659-2670.
(41.) Strander B, Andersson-Ellstrom A, Milsom I, Sparen P. Long term risk of invasive cancer after treatment for cervical intraepithelial neoplasia grade 3: population based cohort study. BMJ. 2007;335(7629):1077.
(42.) Miller MG, Sung HY, Sawaya GF, Kearney KA, Kinney W, Hiatt RA. Screening interval and risk of invasive squamous cell cervical cancer. Obstet Gynecol. 2003;101(1):29-37.
(43.) Belinson J, Quao YL, Pretorius R, et al. Shanxi Province cervical cancer screening study: a cross sectional comparative trial of multiple techniques to detect cervical neoplasia. Gynecol Oncol. 2004;83(2):439-444.
(44.) Khan MJ, Castle PE, Lorincz AT, et al. The elevated 10-year risk of cervical precancer and cancer in women with human papillomavirus (HPV) type 16 or 18 and the possible utility of type-specific HPV testing in clinical practice. J Natl Cancer Inst. 2005;97(14):1072-1079.
(45.) Kitchener HC, Almonte M, Thomson C, et al. HPV testing in combination with liquid-based cytology in primary cervical screening (ARTISTIC): a randomised controlled trial. Lancet Oncol. 2009;10(7):672-682.
(46.) Cuzick J, Clavel C, Petry KU, et al. Overview of European and North American studies on HPV testing in primary cervical cancer screening. Int J Cancer. 2006;119(5):1095-1101.
(47.) Massad LS, Einstein M, Myers E, Wheeler CM, Wentzensen N, Solomon D. The impact of human papillomavirus vaccination on cervical cancer prevention efforts. Gynecol Oncol. 2009;114(2):360-364.
R. Marshall Austin, MD, PhD; Agnieszka Onisko, PhD; Marek J. Druzdzel, PhD
Accepted for publication November 5, 2009.
From the Department of Pathology, Magee-Womens Hospital, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania (Drs Austin and Onisko); and the Decision Systems Laboratory, School of Information Sciences, University of Pittsburgh (Dr Druzdzel).
The authors have no relevant financial interest in the products or companies described in this article.
Reprints: R. Marshall Austin, MD, PhD, Department of Pathology, Magee-Womens Hospital, University of Pittsburgh Medical Center, 300 Halket St, Pittsburgh, PA 15213-3180 (e-mail: firstname.lastname@example.org).
The Magee-Womens Hospital Data From Test Results Collected During a 4-Year Period (2005-2008) 2005 2006 2007 2008 Total Pap test results 97 144 111 019 113 197 100 011 421 371 hrHPV DNA test results 9120 18 652 30 150 23 926 81 848 Histopathologic data 11 009 10 590 11 798 11 718 45 115 Abbreviations: hrHPV, high-risk human papillomavirus; Pap, Papanicolaou.
|Gale Copyright:||Copyright 2010 Gale, Cengage Learning. All rights reserved.|