Agnieszka Onisko

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

OBJECTIVE Atypical glandular cell (AGC) Pap interpretations and screening for glandular neoplasias remain major challenges. We document the largest reported AGC histopathologic follow-up experience and include verification bias-adjusted data on laboratory screening sensitivity. METHODS AGC Pap tests of endocervical origin (AGC-EC), endometrial origin (AGC-EM), and not otherwise specified (AGC-NOS) were documented at a center serving an older low risk population. 98% of Pap tests were liquid-based cytology (LBC) specimens screened using computer-assisted screening. Follow-up diagnoses were correlated with cytology and stratified into age groups. Screening sensitivity was assessed by examining Pap results during 1 year preceding neoplastic diagnoses. Verification bias was adjusted with findings in over 2000 patients with hysterectomies. RESULTS Of 247,131 Pap tests, 1021 (0.41%) reported AGC results and 662 cases had tissue follow-up. Precancerous or malignant neoplastic histologic outcomes were documented in 101 patients (15.3%), including 8.3% cervical, 6.3% endometrial, and 0.6% ovarian. AGC results were most often associated with neoplastic cervical outcomes in women younger than 40 and with neoplastic endometrial outcomes in women 50 or older. AGC-NOS with a squamous cell abnormality and AGC-EC results suggested cervical neoplasia, while AGC-EM results suggested endometrial neoplasia. CONCLUSIONS AGC Pap results detected significant numbers of cervical and non-cervical neoplasias. Since 38 of 44 (86%) of AGC-detected carcinomas were endometrial or ovarian, HPV co-testing would not have aided screening in detecting the majority of malignancies diagnosed after AGC Pap results. Verification bias-adjusted Pap screening sensitivity in the laboratory for detection of significant neoplastic cervical disease was 93%.

Comparison of Rule-Based and Bayesian Network Approaches in Medical Diagnostic Systems

Almost two decades after the introduction of probabilistic expert systems, their theoretical status, practical use, and experiences are matching those of rule-based expert systems. Since both types of systems are in wide use, it is more than ever important to understand their advantages and drawbacks. We describe a study in which we compare rule-based systems to systems based on Bayesian networks. We present two expert systems for diagnosis of liver disorders that served as the inspiration and vehicle of our study and discuss problems related to knowledge engineering using the two approaches. We finally present the results of a simple experiment comparing the diagnostic performance of each of the systems on a subset of their domain.

Extension of the HEPAR II Model to Multiple-Disorder Diagnosis

The Hepar II system is based on a Bayesian network model of a subset of the domain of hepatology in which the structure of the network is elicited from an expert diagnostician and the parameters are learned from a database of medical cases. The model follows the assumption made in the database that each patient case is diagnosed with a single disorder, i.e., disorders are mutually exclusive.

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 Drug Administration. 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 models 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. RESULTS 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.

Follow-up outcomes for a large cohort of US women with negative imaged liquid-based cytology findings and positive high risk human papillomavirus test results ☆

OBJECTIVE This study aimed to follow a large group of US women with negative computer-imaged liquid-based cytology (LBC) and positive high risk (hr) HPV DNA results. METHODS Negative LBC and positive hrHPV cases were identified between July 1, 2005 and December 31, 2009. Cytologic and histopathologic follow-up results, repeat HPV results, and prior history were analyzed. RESULTS 1099 Patients with negative LBC and positive hrHPV results were identified. Eight hundred sixty-nine had repeat Pap or histopathologic follow-up results. Average age was 41.2 years. Average follow-up was 23.2 months. Two hundred ninety of 869 had colposcopic examination and biopsies, including 33 diagnostic excisional procedures and 10 hysterectomies. Cervical intraepithelial neoplasia (CIN) 1 and low-grade squamous intraepithelial lesions (CIN1/LSIL) and more severe lesions (CIN1/LSIL+) were detected in 211 of 689 (24.3%). CIN2+ was diagnosed in 21 (2.4%) (1 VAIN3, 2 adenocarcinoma in situ, 1 invasive cervical adenocarcinoma). Six hundred six had repeat HPV tests and 200 had multiple repeat HPV tests. More LSIL/CIN1+ was identified with repeat positive HPV results than with repeat negative HPV results (P<0.001). LSIL/CIN1+ was detected more often with a history of LSIL/CIN1+ than with a history of negative Paps (P<0.001). Eight of 105 (7.6%) cytology-negative HPV-positive patients tested positive for HPV 16 and/or HPV 18. CONCLUSION This is the largest study documenting follow-up on US cytology-negative hrHPV-positive patients screened with now widely utilized FDA-cleared methods of ciLBC and hrHPV testing. Of 869 patients followed for an average of almost 2 years, 20 cases of high grade intraepithelial neoplasia (2.3%) and one case of endocervical adenocarcinoma were detected. 90.5%(190/210) of intraepithelial neoplasias detected during follow-up were CIN1.

Immunohistochemistry and HPV in situ hybridization in pathologic distinction between endocervical and endometrial adenocarcinoma: a comparative tissue microarray study of 76 tumors.

Objectives It is clinically important to determine whether adenocarcinoma present in a biopsy or curettage is of endometrial or endocervical origin. When tumors are difficult to distinguish based on routine histologic sections, immunohistochemistry and human papillomavirus (HPV) in situ hybridization may be used. Materials and Methods We compare immunohistochemical profile and HPV expression in 76 tumors, including various types of endocervical adenocarcinoma and the most common endometrioid type of endometrial adenocarcinoma using tumor tissue microarray. Immunostaining for p16, mammaglobin, vimentin, monoclonal carcinoembryonic antigen, estrogen receptor, progesterone receptor, and PAX-8 as well as HPV in situ hybridization was performed in 37 endocervical adenocarcinomas and 39 endometrioid-type endometrial adenocarcinomas. The staining patterns were analyzed with Bayesian network model. Results The markers with the highest discriminatory values were p16, HPV, vimentin, estrogen receptor, and monoclonal carcinoembryonic antigen. The various histologic types of endocervical adenocarcinoma showed similar immunohistochemical profile, and most were positive for p16 (86%) and HPV (65%). Most (90%) of the endometrial adenocarcinomas were positive for vimentin and estrogen receptor, all were negative for HPV, and 97% were negative for carcinoembryonic antigen. Conclusions Immunohistochemical testing with multiple markers and HPV testing aids in diagnostic evaluation of adenocarcinomas of endocervix and endometrium and is recommended in tumors of uncertain origin.

An Experimental Comparison of Methods for Handling Incomplete Data in Learning Parameters of Bayesian Networks

Missing values of attributes in data sets, also referred to as incomplete data, pose difficulties in learning tasks, such as classification, data mining, or learning Bayesian network structure and its numerical parameters. Because of the predominance of incomplete data in practice, many methods have been proposed to deal with them while there are few studies that compare their performance. The Hepar II project presents an excellent opportunity to test experimentally how these methods perform on a real data set. We briefly review several popular methods for handling incomplete data and then compare them on the task of learning conditional probability distributions of a Bayesian network model, where the comparison criterion is the resulting diagnostic accuracy. While substitution of “normal” values of missing attributes seemed to perform best, we observed only a small difference in performance among the studied methods.

A validation study of quantum dot multispectral imaging to evaluate hormone receptor status in ductal carcinoma in situ of the breast

The assessment of hormone receptors, including estrogen receptor and progesterone receptor, has become a standard practice in breast cancer management. However, the need for multiple sections to evaluate each receptor individually by conventional immunohistochemistry may preclude the analysis on some core biopsies with a limited amount of tumors. The aim of the study was to validate the quantitative analysis of nuclear markers estrogen receptor and progesterone receptor by quantum dot-based immunohistochemistry using a multispectral imaging system in ductal carcinoma in situ of the breast. Consecutive sections from a total of 17 cases of ductal carcinoma in situ with excisional biopsies or mastectomies were stained with conventional immunohistochemistry and quantum dot-based, single- and double-labeled immunohistochemistry for estrogen receptor and progesterone receptor. The semiquantitative results from double-labeled, quantum dot-based immunohistochemistry were compared with those from single-labeled, quantum dot-based immunohistochemistry as well as from conventional immunohistochemistry. There was good concordance between double- and single-labeled quantum dot-based immunohistochemistry, and quantum dot-based immunohistochemistry correlated well with conventional immunohistochemistry (Spearman correlation coefficient range from 0.884 to 0.958, P < .001). The findings proved the validity and accuracy of quantum dot-based multiplex, multispectral technique in detecting 2 tumor markers in the same cellular compartment simultaneously on a single slide. This technique may enhance our ability to assess multiple breast tumor markers in specimens with limited available tissue. However, several technical and logistic issues await significant improvement before this novel technique can be justified for routine clinical application.

How to interpret the results of medical time series data analysis: Classical statistical approaches versus dynamic Bayesian network modeling

Background: Classical statistics is a well-established approach in the analysis of medical data. While the medical community seems to be familiar with the concept of a statistical analysis and its interpretation, the Bayesian approach, argued by many of its proponents to be superior to the classical frequentist approach, is still not well-recognized in the analysis of medical data. Aim: The goal of this study is to encourage data analysts to use the Bayesian approach, such as modeling with graphical probabilistic networks, as an insightful alternative to classical statistical analysis of medical data. Materials and Methods: This paper offers a comparison of two approaches to analysis of medical time series data: (1) classical statistical approach, such as the Kaplan-Meier estimator and the Cox proportional hazards regression model, and (2) dynamic Bayesian network modeling. Our comparison is based on time series cervical cancer screening data collected at Magee-Womens Hospital, University of Pittsburgh Medical Center over 10 years. Results: The main outcomes of our comparison are cervical cancer risk assessments produced by the three approaches. However, our analysis discusses also several aspects of the comparison, such as modeling assumptions, model building, dealing with incomplete data, individualized risk assessment, results interpretation, and model validation. Conclusion : Our study shows that the Bayesian approach is (1) much more flexible in terms of modeling effort, and (2) it offers an individualized risk assessment, which is more cumbersome for classical statistical approaches.

Dynamic Bayesian Network for Cervical Cancer Screening

In this chapter we will present the application of dynamic Bayesian networks to cervical cancer screening. The main goal of this project was to create a multivariate model that would incorporate several variables in one framework and predict the risk of developing cervical precancer and invasive cervical cancer. We were interested in identifying those women that are at higher risk of developing cervical cancer and that should be screened differently than indicated in the guidelines.