Analysis of clinical and ultrasonographic data by use of logistic regression models for prediction of malignant versus benign causes of ultrasonographically detected focal liver lesions in dogs

Evaluation of a Probability-Based Predictive Tool on Pathologist Agreement Using Urinary Bladder as a Pilot Tissue

Simple Summary

There is a common joke in pathology—put three pathologists in a room and you will obtain three different answers. This saying comes from the fact that pathology can be subjective; pathologists’ diagnoses can be influenced by many different biases, and pathologists are also influenced by the presence or absence of animal information and medical history. Compared to pathology, statistics is a much more objective field. This study aimed to develop a probability-based tool using statistics obtained by analyzing 338 histopathology slides of canine and feline urinary bladders, then see if the tool affected agreement between the test pathologists. Four pathologists diagnosed 25 canine and feline bladder slides and they conducted this three times: without animal and clinical information, then with this information, and finally using the probability tool. Results showed large differences in the pathologists’ interpretation of bladder slides, with kappa agreement values (low value for digital slide images, high value for glass slides) of 7–37% without any animal or clinical information, 23–37% with animal signalment and history, and 31–42% when our probability tool was used. This study provides a starting point for the use of probability-based tools in standardizing pathologist agreement in veterinary pathology.

Abstract

Inter-pathologist variation is widely recognized across human and veterinary pathology and is often compounded by missing animal or clinical information on pathology submission forms. Variation in pathologist threshold levels of resident inflammatory cells in the tissue of interest can further decrease inter-pathologist agreement. This study applied a predictive modeling tool to bladder histology slides that were assessed by four pathologists: first without animal and clinical information, then with this information, and finally using the predictive tool. All three assessments were performed twice, using digital whole-slide images (WSI) and then glass slides. Results showed marked variation in pathologists’ interpretation of bladder slides, with kappa agreement values of 7–37% without any animal or clinical information, 23–37% with animal signalment and history, and 31–42% when our predictive tool was applied, for digital WSI and glass slides. The concurrence of test pathologists to the reference diagnosis was 60% overall. This study provides a starting point for the use of predictive modeling in standardizing pathologist agreement in veterinary pathology. It also highlights the importance of high-quality whole-slide imaging to limit the effect of digitization on inter-pathologist agreement and the benefit of continued standardization of tissue assessment in veterinary pathology.

Morphometric evaluation of canine hepatocellular carcinoma using computed tomography: a promising tool for predicting malignancy

The size of canine focal liver lesions (FLLs) is known to be one of the predicting criteria for malignancy. However, there are discrepancies for the measurement of maximum lesion size, resulting in contradicting results among studies and incidences of false positive outcomes. Thus far, the morphometric changes of FLLs for distinguishing malignancy from benignancy remains undocumented. This study aimed to investigate morphometric characteristics of FLLs using computed tomography (CT). CT images of 40 dogs with histopathological confirmation of 49 liver lesions, including 39 hepatocellular carcinomas and 10 nodular hyperplasias were retrospectively reviewed. The morphometric parameters including size (long and short axis diameters measured on transverse image), shape (measured by long to short axis (L/S) ratio), volume, and surface appearance of a liver lesion were evaluated using univariate and stepwise multivariate analyses, respectively. The results of univariate analysis showed that long and short axis diameters, L/S ratio, volume, and surface appearance of a lesion were significantly different between hepatocellular carcinomas and nodular hyperplasias. Multivariate analysis revealed that short axis diameter (>3.30 cm; odds ratio (OR): 36.1, 95% confidence interval (CI): 3.36–387.05, P=0.0031) and L/S ratio (>1.23; OR: 18.1, 95% CI: 1.61–205.12, P=0.0191) were independent predictors of malignancy, with the area under the curve of 0.9154. These results suggest that the combination of short axis diameter and L/S ratio is a promising tool for predicting liver malignancy with outstanding discriminating ability.

Predicting Diagnosis of Australian Canine and Feline Urinary Bladder Disease Based on Histologic Features

Anatomic pathology is a vital component of veterinary medicine but as a primarily subjective qualitative or semiquantitative discipline, it is at risk of cognitive biases. Logistic regression is a statistical technique used to explain relationships between data categories and outcomes and is increasingly being applied in medicine for predicting disease probability based on medical and patient variables. Our aims were to evaluate histologic features of canine and feline bladder diseases and explore the utility of logistic regression modeling in identifying associations in veterinary histopathology, then formulate a predictive disease model using urinary bladder as a pilot tissue. The histologic features of 267 canine and 71 feline bladder samples were evaluated, and a logistic regression model was developed to identify associations between the bladder disease diagnosed, and both patient and histologic variables. There were 102 cases of cystitis, 84 neoplasia, 42 urolithiasis and 63 normal bladders. Logistic regression modeling identified six variables that were significantly associated with disease outcome: species, urothelial ulceration, urothelial inflammation, submucosal lymphoid aggregates, neutrophilic submucosal inflammation, and moderate submucosal hemorrhage. This study demonstrated that logistic regression modeling could provide a more objective approach to veterinary histopathology and has opened the door toward predictive disease modeling based on histologic variables.

Computed tomographic features for differentiating benign from malignant liver lesions in dogs

Thus far, there are few computed tomography (CT) characteristics that can distinguish benign and malignant etiologies. The criteria are complex, subjective, and difficult to use in clinical applications due to the high level of experience needed. This study aimed to identify practical CT variables and their clinical relevance for broadly classifying histopathological diagnoses as benign or malignant. In this prospective study, all dogs with liver nodules or masses that underwent CT examination and subsequent histopathological diagnosis were included. Signalments, CT findings and histopathological diagnoses were recorded. Seventy liver nodules or masses in 57 dogs were diagnosed, comprising 18 benign and 52 malignant lesions. Twenty-three qualitative and quantitative CT variables were evaluated using univariate and stepwise multivariate analyses, respectively. Two variables, namely, the postcontrast enhancement pattern of the lesion in the delayed phase (heterogeneous; odds ratio (OR): 14.7, 95% confidence interval (CI): 0.82–262.03, P=0.0429) and the maximal transverse diameter of the lesion (>4.5 cm; OR: 33.3, 95% CI: 2.29–484.18, P=0.0006), were significantly related to the differentiation of benign from malignant liver lesions, with an area under the curve of 0.8910, representing an accuracy of 88.6%. These findings indicate that features from triple-phase CT can provide information for distinguishing pathological varieties of focal liver lesions and for clinical decision making. Evaluations of the maximal transverse diameter and postcontrast enhancement pattern of the lesion included simple CT features for predicting liver malignancy with high accuracy in clinical settings.

Predictive factors of malignancy in dogs with focal liver lesions using clinical data and ultrasonographic features

A definitive diagnosis of focal liver lesions (FLLs) requires invasive procedures for histopathologic examination. Thus, a simpler noninvasive diagnostic method, such as conventional ultrasonography combined with clinical data, is needed for the prediction of liver malignancy. The objective of this study was to examine the diagnostic accuracy of clinical data and ultrasonographic (US) features to differentiate benign and malignant liver lesions. Medical records and US images from dogs with FLLs that underwent abdominal US and histopathologic examinations following surgery or liver biopsy were retrospectively reviewed. Clinical data, including signalment, clinical signs and laboratory findings, and the US features of liver lesions that could act as predictive factors were assessed using univariate and multivariate analyses to evaluate the associations between predictive factors and liver malignancy. Based on the histopathologic results, 55 dogs with malignant lesions and 28 dogs with benign lesions were included in the study. The results of univariate analysis showed that several US features and platelet count were significantly associated with liver malignancy. Multivariate analysis revealed that the platelet count (thrombocytosis; odds ratio [OR]: 4.13, 95% confidence intervals [CI]: 1.81−9.41), lesion size (4.1 cm or greater; OR: 23.83, 95% CI: 3.74−151.95) and echotexture of FLLs (heterogenous; OR: 8.44; 95% CI: 1.37−51.91) were independent predictors for differentiating benign and malignant liver lesions, suggesting that a combination of clinical data and US findings of FLLs could predict liver malignancy in dogs.

Evaluation of canine hepatic masses by use of triphasic computed tomography and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography and correlation with histopathologic classification

OBJECTIVE To determine clinical relevance for quantitative and qualitative features of canine hepatic masses evaluated by use of triphasic CT and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography and to compare diagnostic accuracy of these modalities for predicting mass type on the basis of histopathologic classification. ANIMALS 44 client-owned dogs. PROCEDURES Dogs with histopathologic confirmation (needle core, punch, or excisional biopsy) of a hepatic mass were enrolled. Triphasic CT and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography of each hepatic mass were performed. Seventy quantitative and qualitative variables of each hepatic mass were recorded by 5 separate observers and statistically evaluated with discriminant and stepwise analyses. Significant variables were entered in equation-based predictions for the histopathologic diagnosis. RESULTS An equation that included the lowest delayed-phase absolute enhancement of the mass and the highest venous-phase mass conspicuity was used to correctly classify 43 of 46 (93.5%) hepatic masses as benign or malignant. An equation that included only the lowest delayed-phase absolute enhancement of the mass could be used to correctly classify 42 of 46 (91.3%) masses (with expectation of malignancy if this value was < 37 Hounsfield units). For ultrasonography, categorization of the masses with cavitations as malignant achieved a diagnostic accuracy of 80.4%. CONCLUSIONS AND CLINICAL RELEVANCE Triphasic CT had a higher accuracy than ultrasonography for use in predicting hepatic lesion classification. The lowest delayed-phase absolute enhancement of the mass was a simple calculation that required 2 measurements and aided in the differentiation of benign versus malignant hepatic masses.

Diagnostic Imaging of the Hepatobiliary System: An Update

Recent advances in diagnostic imaging of the hepatobiliary system include MRI, computed tomography (CT), contrast-enhanced ultrasound, and ultrasound elastography. With the advent of multislice CT scanners, sedated examinations in veterinary patients are feasible, increasing the utility of this imaging modality. CT and MRI provide additional information for dogs and cats with hepatobiliary diseases due to lack of superimposition of structures, operator dependence, and through intravenous contrast administration. Advanced ultrasound methods can offer complementary information to standard ultrasound imaging. These newer imaging modalities assist clinicians by aiding diagnosis, prognostication, and surgical planning.

Computed Tomography and MRI of the Hepatobiliary System and Pancreas

MRI and computed tomographic (CT) imaging are becoming more common in the diagnosis of hepatobiliary and pancreatic disorders in small animals. With the advent of multislice CT scanners, sedated examinations in veterinary patients are feasible increasing the use of this imaging modality. CT and MRI provide additional information for dogs and cats with hepatobiliary and pancreatic diseases because of lack of superimposition of structures, operator dependence, and through intravenous contrast administration. This added value provides more information for diagnosis, prognosis, and surgical planning.

A comparison of hepatic sonographic features and histopathologic diagnosis in canine liver disease: 138 cases

BACKGROUND

Ultrasound examination is commonly used in the diagnostic evaluation of liver disease in dogs.

HYPOTHESIS/OBJECTIVES

To determine if hepatic sonographic features were predictive of findings on liver histopathology. We hypothesized that there would be a relationship between sonographic features and the category of liver disease based on histologic assessment.

ANIMALS

One hundred and thirty-eight dogs in which the liver was evaluated by both abdominal ultrasound examination and histopathologic examination. Twenty-five dogs were included in each of the following categories based on histopathology: normal, degenerative, vascular, inflammatory, and neoplasia. Thirteen dogs had nodular regeneration.

METHODS

Retrospective study. Medical records of dogs from 2005 to 2010 were searched for cases in which the liver was evaluated by abdominal ultrasound examination as well as by histopathology. After independent evaluation of ultrasound images, the recorded sonographic features were analyzed to identify abnormalities associated with each histopathologic diagnosis or degree of fibrosis.

RESULTS

Sixty-four percent of sonographically unremarkable livers had histologic abnormalities. Both microhepatia and the identification of abnormal vasculature were significantly associated with a histopathologic diagnosis of vascular disease. Hepatic masses were significantly associated with a diagnosis of neoplasia. Dilated common bile duct and thickened gall bladder wall were significantly associated with hepatitis. There were no sonographic findings consistently present with hepatic fibrosis.

CONCLUSION AND CLINICAL IMPORTANCE

Although some ultrasonographic findings, including masses, microhepatia, anomalous veins, and biliary changes, are associated with specific histopathologic abnormalities, sonographic findings are inconsistently detected in many disorders. Overall, hepatic ultrasonographic abnormalities have substantial limitations in predicting the underlying disease.

Accuracy of US-Guided FNA of Focal Liver Lesions in Dogs: 140 Cases (2005–2008)

Medical records from dogs having abdominal ultrasound (US) performed between March 2005 and October 2008 were reviewed for detection of focal liver lesions (FLL) with both cytologic and histologic sampling. Samples were classified as to either the presence or absence of major categories of pathologic processes, including malignant neoplasia, inflammation, hyperplasia/benign neoplasia, vacuolar change, extramedullary hematopoeisis, cholestasis, necrosis, and no microscopic abnormalities. Evaluation of selection bias was performed by review of the relative distribution of cytologic diagnoses for cases with histology compared with cases excluded from the comparison analysis because histology results were not available. Cytology had the highest sensitivity for vacuolar change (57.9%), followed by neoplasia (52.0%). Cytology had the highest positive predictive value (PPV) for neoplasia (86.7%) followed by vacuolar change (51.6%). Cytology had lower sensitivity and PPVs for inflammation, necrosis, and hyperplasia. The ability of cytology to characterize disease in canine FLL varies by pathologic process. Clinicians can have a high degree of confidence when a cytologic diagnosis of neoplasia is given; however, cytology is less reliable for excluding the potential for neoplasia. Cytology has a low sensitivity and PPV for inflammation and a limited diagnostic performance for the diagnosis of vacuolar change.