Critical Evaluation of Transbronchial Biopsy Practice in the Diagnosis of Lung Cancer at the Hershey Medical Center
Open Access
- Author:
- Sinha, Anju
- Graduate Program:
- Molecular Medicine
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 26, 2008
- Committee Members:
- Rebecca Bascom, Committee Chair/Co-Chair
John Peter Richie Jr., Committee Member
Shantu G Amin, Committee Member
Eugene Joseph Lengerich, Committee Member
Dani S Zander, Committee Member - Keywords:
- Hershey Medical Center
diagnostic yield
lung cancer
transbronchial biopsy
molecular markers - Abstract:
- The goal of this project is to determine the pre-test probability and diagnostic yield of transbronchial biopsy (TBBx) in detecting lung cancer and to identify gaps in the current process of care associated with use of TBBx for lung cancer diagnosis at the Hershey Medical Center. The hypotheses of this project are that the current TBBx practice at the Hershey Medical Center has a low diagnostic yield, high false negative results, especially for small peripheral lesions, provides insufficient clinical data to the interpreting pathologist, and the current process of care for lung cancer patients has variable time gaps between CT scan, TBBx and lung resection. The specific aims are: 1. To define the current clinical process of care and to identify clinical datasets used for key elements of lung cancer diagnosis. 2. To critically evaluate a retrospective cohort of bronchoscopically acquired TBBxs and estimate the pre-test probability of diagnosing lung cancer. 3. To identify and integrate matched pairs of TBBx and lung resections in order to validate the performance of TBBx against lung resection as the gold standard. 4. To benchmark TBBx pathology reports as a function of digitally acquired computerized tomography (CT) lesions using a novel classification scheme and identify characteristics of lesions commonly biopsied and lesions frequently reported to be positive for malignancy. 5. Assessment of pulmonary lesion characteristics accurately and inaccurately diagnosed to be malignant by TBBx. 6. To estimate the time intervals between CT scan, TBBx and lung resection in order to identify time gaps in the current process of care for lung cancer patients at the Hershey Medical Center. Methods: Specific Aim 1: Process of care and quality measures are assessed by defining the audience of this project, the clinical area being evaluated in this study, the assessment team, the component of the process being measured, preliminary tests, scoring and analytic specifications. To understand and define the current process of care at the Hershey Medical Center for diagnosing, staging and managing patients with lung cancer, various physician resources utilized are conferences with pulmonologists, pathologists, oncologists, radiologists, and cardiothoracic surgeons. Electronic resources utilized to assess patient information are billing records, Power Chart, digital CT scan repository, “Natural Language IIa search”, a software program available in the pathology department, and AJCC guided lung cancer classification protocol adopted by the Department of Pathology. Specific Aim 2: Patients who underwent TBBx are identified through billing records, their pathology reports are acquired from Power Chart. An epidemiological codebook is developed to document data. Outcome measures evaluated are: Pre-procedure suspicion of lung cancer, pathology report positive for malignancy, smoking history, specimen site biopsied, average aggregate volume of tissue, statistical analysis of average aggregate tissue volume differences between samples positive for malignancy and those negative for malignancy, histopathological type of malignancy, gender distribution by histopathological type of malignancy, estimate of the pre-test probability of a positive pathology report for malignancy among samples obtained from patients clinically suspected to have lung cancer, cellular atypia, non-malignant findings, and staining. Specific Aim 3: “Natural Language IIa search”, software is used to identify and acquire pathology reports of lung resection surgeries. Patient name and MRN are used to identify patients common in the TBBx and lung resection datasets. An epidemiological codebook is developed to document lung resection data. TBBx reports are categorized as test positive and negative based on presence or absence of malignancy. Lung resection reports are used as gold standard to assess TBBx performance in diagnosing lung cancer. Data are integrated to identify four groups of cases, true positive, true negative, false positive, and false negative. Specific Aim 4: The digital CT scans IDX/GE electronic image repository is used to retrieve digital images of pulmonary lesions biopsied for patients included in the TBBx dataset. A single transverse slice of each CT scan image centered on the region of interest (ROI) is extracted using Snagit®. A novel image classification scheme is used to classify ROIs into focal and diffuse lesions. Each type of lesion if further divided into subcategories based on size. Outcome measures evaluated are: availability of digital CT scans with time, most commonly biopsied lesion characteristic, and characteristic of lesion most commonly reported to be malignant by TBBx. Specific Aim 5: TBBx pathology report, lung resection pathology report and digital CT scan image data are integrated in order to identify lesion characteristics in the true and false positive and negative categories. The outcome measures established are: location of lesion biopsied and resected, congruency of site biopsied and site resected, histopathological type of malignancy reported by TBBx and lung resection, and digital CT scan lesion characteristics. Specific Aim 6: Availability of lung resection and digital CT scan data with time are established. In order to assess time to diagnosis (TDx) and time to treatment (TRx) only pre-TBBx CT scans and post TBBx resections are included. TDx and TRx are established for the four groups of cases to identify gaps in process of care. Results: Specific Aim 1: The audience of this project includes patients, physicians and researchers involved and interested in lung cancer. The clinical area being evaluated is TBBx and its efficiency in diagnosing lung cancer. The assessment team is primarily composed of the author and her thesis advisor. The component of the TBBx process being measured is primarily the pre-procedure differential diagnosis of cancer. Datasets utilized in this study include billing department records, Power Chart (Cerner), digital CT scan image repository, “Natural Language IIa search” software available in the Department of Pathology. Specific Aim 2: Pathology reports from 801 TBBx samples obtained from 653 procedures done over the 5 years 10 months period of this study are obtained and data organized and analyzed. Patients’ age averaged 58yrs 2mths +/- 15yrs 2mths (mean +/- SD) and 360/653 (55.13%) are male. Lung Cancer was suspected pre-procedure in 122/653 (18.7%) cases. Pathology report is positive for malignancy in 111/801 (13.86%) TBBx samples. Pre-test probability of diagnosing lung cancer is highest (22.45%) among patients suspected to have lung cancer, intermediate (13.89%) in the group where differential diagnosis is not specified and minimum (0.7%) in the group where lung cancer is not suspected pre-procedure. Average aggregate volume of samples positive for malignancy is significantly higher (0.062cm3) than the average aggregate volume of samples negative for malignancy (0.047cm3). Adenocarcinoma is the most common histopathological type of cancer seen in our study. A lack of transfer of clinical information to the diagnosing pathologist is noticed. Specific Aim 3: Fourteen of 72 cases common to TBBx and lung resection datasets are true positive, 28 are false negative, 29 are true negative and 1 is false positive. Sensitivity of TBBx in diagnosing lung cancer is estimated to be 0.33, specificity is 0.97, false negative rate (FNR) =0.67, false positive rate (FPR) =0.03, predictive value positive (PVP) =0.93, predictive value negative (PVN) =0.51. Specific Aim 4: Digital CT scan data are available for 298/653 TBBx procedures. Focal ROIs are more commonly biopsied (180/298, 54.05%) compared to diffuse lesions (153/298, 45.95%). Focal ROIs 1-3cm in size are the most commonly biopsied lesions. Among diffuse lesions, ROIs involving <25% of the lobe are most commonly biopsied. However, malignancy is most commonly reported in focal ROIs >3cm in size. Among diffuse lesions, malignancy is most commonly reported in ROIs involving >50% of the lobe. Specific Aim 5: Of the 29 true negative cases diffuse lesions involving <25% of the right lower lobe is most commonly seen. Among the 14 true positive cases adenocarcinomas, focal lesions 1-3 cm in size, located in the right upper lobe are most commonly seen. One false positive sample is reported to be carcinoma in situ, squamous cell carcinoma in the left upper lobe. CT scan for this case is not available. Of the 28 false negative cases, adenocarcinomas, focal lesions 1-3cm in size and lesions in the right upper lobe are most commonly seen. Specific Aim 6: Digital CT scan images are available for 298/653 (45.64%) and lung resection data are available for 74/653 (11.33%) of TBBx procedures. The median time to diagnosis (TDx) is 7 days pre-TBBx (range 0-57 days) and the median time to treatment (TRx) is 63.5 days post-TBBx (range 4-1493). No significant difference is seen in TDx and TRx between the true and false positive and negative groups. However the number of cases in these groups is small (4-14) therefore these results need to be confirmed in a study including larger number of patients. Conclusions: Disparate electronic datasets can be linked to critically evaluate TBBx yield in diagnosing lung cancer. Essential clinical information is not provided to the diagnosing pathologist frequently. Pre-test probability of diagnosing lung cancer is highest among patients suspected to have lung cancer. Average aggregate volume is significantly higher in samples reported positive for malignancy compared to non-malignant samples. This is an important factor when assessing feasibility of using TBBx tissue for molecular markers. Validation of TBBx performance in diagnosing lung cancer against lung resection as gold standard provided a sensitivity= 0.33, and specificity= 0.97 with a high false negative rate of 0.67. Malignant focal lesions >3cm in size are more frequently detected accurately than adenocarcinomas, focal lesions 1-3cm in size that have a high false negative rate. The median TDx is 7 days pre-TBBx and the median TRx is 63.5 days post-TBBx. Variability is seen in TDx and TRx but no statistically significant difference is seen in TDx and TRx between the true and false positive and negative groups.