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

Introduction

The use of ultrasound (US)-guided, fine-needle aspiration (FNA) cytology in the diagnosis of liver disease in dogs can have significant advantages over histologic diagnosis obtained either via percutaneous core biopsies or surgical biopsies in a clinical setting. FNA is less invasive, has a diminished risk of complications, yields faster results, and has lower costs than biopsy. Those advantages are balanced by the disadvantages of cytologic diagnosis, including absence of tissue architecture in assessing lesions and equivocal results due to low cellularity and artifacts.¹ Studies of the accuracy of US-guided FNA cytologic diagnosis versus histologic diagnosis used for evaluation of canine hepatic disease have yielded mixed results.^2–6 A previous study by Wang et al. (2004) reported an overall cytologic and histologic agreement of 30.3% in dogs with US-guided FNA of the liver.⁶ Another study by Roth (2001) that included dogs and cats reported complete agreement in 61% of cytology samples when compared with primarily core biopsies of the liver.⁷ In a study conducted by Weiss et al. (2001) there was overall sensitivity of 93% and specificity of 96% for the cytologic diagnosis of inflammatory liver disease in 30 canine patients; however, further review of the data indicates several cases of neoplasia were categorized as inflammatory disease.^1,3 Variable approaches to data collection and interpretation likely contribute to differing results. Wang et al. (2004) observed that selection bias was a likely contributor to the discrepancies based on the requirement for both cytology and histology results in retrospective data bases.⁶ Unfortunately, no attempts have been made to characterize the potential for bias in these types of correlation studies.

All three of the above mentioned studies were retrospective, and the criteria for inclusion in each of those studies did not specifically characterize the hepatic lesions sampled beyond review of records in which hepatic tissue was sampled by both cytology and histology. In the study by Weiss et al. (2001), the sampled specimens were obtained by a mix of blind and US-guided aspiration techniques, and there was no mention of imaging performed in the blindly obtained specimens.³ The study by Roth (2001) provided no specific information about specimen procurement (blind versus US-guided techniques), and specimen collection method was unknown in some cases.⁷ In the study by Wang et al. (2004), the specimens were all obtained by US-guided methods; however, there was no description of the US findings.⁶

It is generally accepted that either FNA or biopsy is required for diagnosis of focal liver lesions (FLL) on the basis that US features alone are of limited value in determining etiology.^8–10 The accuracy of US-guided FNA cytology compared with histologic diagnosis has not been investigated specifically for FLL. The purpose of this study was to evaluate the accuracy of US-guided FNA cytologic diagnosis compared to the histologic diagnosis obtained from US-detected FLL in 140 canine patients for which cytology and histology results were available. Furthermore, the authors evaluated the potential for selection bias in the patient population by statistically comparing the distribution of cytologic diagnoses for cases with histology and cases that were excluded from the comparison analysis because histology results were not available. Because differences in methodology and reporting may influence study conclusions, the Standards for Reporting of Diagnostic Accuracy were adopted to facilitate clear and accurate reporting of the data.¹¹

Materials and Methods

Results

In total, 140 cases met the inclusion criteria. Patient ages ranged from 2 yr to 15 yr (median, 11 yr; mean, 10.15 yr). Fifty-one different canine breeds were represented in this study, and the following breeds were the most abundant: beagle (n = 5), bichon frise (n = 7), cocker spaniel (n = 8), English springer spaniel (n = 4), German shepherd dog (n = 7), golden retriever (n = 14), Labrador retriever (n = 23), mixed-breed (n = 7), and Shetland sheepdog (n = 4). Two dogs were intact females, 76 were spayed females, 6 were intact males, and 56 were neutered males.

In the data set, cytology was most sensitive for the detection of vacuolar change (57.9%) and neoplasia (52.0%). Sensitivity was lower for inflammation (31.2%), necrosis (20.0%), and hyperplasia (13.8%). PPV was highest for neoplasia (86.7%) followed by vacuolar change (51.6%), inflammation (44.1%), necrosis (28.6%), and hyperplasia (25.0%). For the major pathologic categories of extramedullary hematopoeisis, cholestasis, and no morphologic abnormalities, too few cases were present to generate legitimate conclusions (Table 1).

TABLE 1 Comparison of the Diagnosis of Major Pathologic Processes Identified by Cytology and Histology in 140 Dogs with US-Detected FLL

FLL, focal liver lesions; PPV, positive predictive value; US, ultrasound.

View Fullscreen

Within the major category of neoplasia, cytology was most sensitive for round cell tumor (60%) and nonhepatocellular carcinoma (54.5%) followed by hepatocellular carcinoma (34.8%) (Table 2). Cytology was insensitive for sarcoma (16.7%), but few cases were available for analysis, which also negatively impacted the authors’ ability to evaluate the PPV. The PPV of cytology was excellent for hepatocellular carcinoma (100%) and good for both nonhepatocellular carcinoma (85.7%) and round cell tumors (75%). For inflammation, the sensitivity for all types was low; however, the sensitivity was slightly better for nonsuppurative inflammation (31.3%) compared with other types. The PPV for all types of inflammation was better than the sensitivity, but still low (Table 1). As with sensitivity, the PPV was best for nonsuppurative inflammation (45.5%) as shown in Table 2.

TABLE 2 Ability of Cytology to Detect Specific Pathologic Processes Diagnosed by Histology in US-Detected FLL in 140 dogs

FLL, focal liver lesions; PPV, positive predictive value; US, ultrasound.

View Fullscreen

The distribution of cytologic diagnoses of specific neoplastic and inflammatory processes diagnosed by histology have been summarized in Table 3. The χ² analysis comparing the frequency of occurrence of each pathologic process in dogs that had both liver cytology and histology compared with dogs that had only cytology but otherwise met all other inclusion criteria revealed a higher frequency of cytologic diagnosis of inflammation and necrosis than expected in the patients that ultimately had histologic evaluation performed than in the group that did not meet the inclusion criteria. The other categories of diagnoses (i.e., neoplasia, vacuolar change, hyperplasia, extramedullary hematopoeisis, cholestasis, no morphologic abnormalities) had similar frequencies of cytologic diagnoses between the two data sets (Table 4).

TABLE 3 Distribution of Cytologic Diagnoses of Specific Neoplastic and Inflammatory Processes Diagnosed by Histology in US-Detected FLL in 140 Dogs

FLL, focal liver lesions; US, ultrasound.

View Fullscreen

TABLE 4 Comparison of Frequency of Cytologic Diagnosis Among Cases Meeting the Inclusion Criteria (n = 140) and Those Excluded due to Lack of Histologic Sampling (n = 461)

χ² analysis indicated a significant difference in frequencies between the two groups. χ², 25.50; P < 0.0006 (based on 7 degrees of freedom and N = 727).

View Fullscreen

Discussion

The ability of cytology to accurately diagnose disease in canine FLL varies by pathologic process. In this study, cytology had the highest sensitivity for vacuolar change (57.9%) followed by neoplasia (52.0%). Within the category of neoplasia, cytology was most sensitive for the detection of round cell tumors and nonhepatocellular origin carcinomas. The cytologic diagnosis of neoplasia had the highest PPV (86.7%), with a PPV of 100% for hepatocellular carcinoma followed by good PPVs for both nonhepatocellular carcinoma and round cell tumors (85.7% and 75%, respectively). Therefore, 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 given that only approximately 50% of patients with a histologic diagnosis of hepatic neoplasia had neoplastic cells detected by cytology.

The ability to make a cytologic diagnosis of neoplasia depends in part on the exfoliative potential of the cell type and the presence of cytologic atypia. Exfoliative potential is generally highest for round cell tumors, but epithelial tumors are also reasonably exfoliative. Cytology is less sensitive for the diagnosis of sarcoma due to the poorly exfoliative nature of the cells. Well-differentiated hepatocellular carcinomas may exhibit little cytologic atypia, making diagnosis more challenging.¹³ PPV was highest for the diagnosis of hepatocellular carcinoma, possibly because cytologists at the authors’ institution reserve that diagnosis for cases that exhibit marked criteria of malignancy. The studies by Roth (2001) and Wang et al. (2004) described similar accuracy for cytologic diagnosis of hepatic neoplasia; however, those studies included fewer cases and imaging characteristics were not defined.^6,7 The larger sample size included in this study, as well as the inclusion of only cases of FLL with US-guided cytologic sampling, did not appear to impact the diagnostic performance of cytology for the detection of neoplasia.

In contrast to the results for neoplasia, cytology had low sensitivity and low PPV for inflammation, including all types evaluated in this study (i.e., suppurative, nonsuppurative, mixed). Those results are similar to the studies by Roth (2001) and Wang et al (2004).^6,7 The results in the study by Weiss et al. (2001) were different from both the current and previous studies.^3,6,7 The study by Weiss et al. (2001) described increased accuracy for cytologic diagnosis of suppurative inflammation and mixed inflammation, but similar results for cytologic diagnosis of nonsuppurative inflammation as seen in this study.³ Because this study included only dogs with FLL, the authors’ observations may disagree with results from studies that included diffuse liver disease; however, the findings reported herein are similar to those of Roth (2001) and Wang et al. (2004) that were not specific as to the US characteristics of the liver lesions.^6,7

In this data set, as well as the data in the study by Wang et al. (2004), there were a high number of false positive diagnoses of inflammation by cytology.⁶ Review of the data reveals that a majority of the cases with a false positive diagnosis of inflammation had major histologic findings of vacuolar change, hyperplasia, necrosis, or no morphologic abnormalities. Therefore, it is evident that inflammation is being over diagnosed, which could occur by over interpretation of cells originating from the blood or relatively small numbers of inflammatory cells that are not reflective of inflammatory lesions on histologic section.

The marginally good sensitivity for cytologic detection of vacuolar change in this study and the high number of false positives diagnosed by cytology in this patient population agreed with the results of the study by Wang et al. (2004), which was the only previous study to evaluate diagnostic accuracy of cytology for vacuolar change in dogs.⁶ The limited diagnostic performance of cytology for this diagnosis may be impacted by sample bias from selection of patients with FLL in this study because vacuolar change may be either patchy or highly localized. The questionable diagnostic value of vacuolar change when detected cytologically may also affect the agreement between cytology and histology because it is a relatively nonspecific finding, especially for FLL. Vacuolar change may be only selectively reported by pathologists who do not consider it a diagnostically relevant finding, especially if it appears to be mild. In contrast, the correlation of cytologic and histologic diagnoses of vacuolar hepatopathy in cats is better than in dogs because of the higher prevalence of vacuolar hepatopathy in cats and because vacuolar hepatopathy in cats is more likely a primary disease process (i.e., hepatic lipidosis) than in dogs.⁶

Cytology also had low sensitivity and PPV for diagnosis of hyperplasia and necrosis. The categories of vacuolar change, hyperplasia, and necrosis had high numbers of false positives, which may be either due to overinterpretation of subtle changes by cytologists or potential underreporting of these lesions on histology because they are perceived as nonspecific findings in FLL.

The limitations of this study include reliance on written records of pathology reports, potential workup bias in the study population, and lack of knowledge of specific biopsy techniques. Direct review of the cytologic and histologic slides was not performed; however, reports were reviewed by a single pathologist to verify diagnoses. Conformance to World Small Animal Veterinary Association guidelines for diagnosis of liver disease was not evaluated in this study.¹⁴ There was the potential for workup bias in these cases as the decision to biopsy may have been influenced by either the cytologic or US results. For example, cases in which cytologic results provided a diagnosis consistent with the clinical disease may not have undergone biopsy. In the cases with multiple FLL, it was not possible to confirm specifically which lesions were sampled for each technique. These limitations reflect realities in practice where clinicians must make decisions based on written reports and presume that samples obtained are representative of the disease process that is occurring, and were therefore incorporated into this study. Unfortunately, the method of biopsy was not documented in some of the patient records and, therefore, could not be analyzed. It would have been preferable to have that information because the method of collection of liver biopsy samples can influence the accuracy of the results.¹⁵ This study considered the histologic diagnosis to be the reference standard for diagnosis, which is typical in studies of the diagnostic accuracy of cytology. Critical evaluation of histology as a diagnostic technique reveals limitations (including interobserver variation and reliance on special staining for the confirmation of some tumor types), leading some studies to combine histology with clinical outcome as the reference standard for diagnostic outcome; however, that approach generally requires a prospective approach to generate reliable data.^1,16–20 It was not possible to evaluate the complications associated with liver biopsy of patients in this study because that information was often not documented in the medical records. Although previous studies have commented on the invasiveness of biopsy techniques, none have evaluated the complications associated with the sample populations.^3,6,7

The sample population in this study and the previous studies is representative of patients seen at referral centers, which may differ from the patient demographic seen in primary care practices. Furthermore, many primary care practices may not have access to US or have the proper training and experience to obtain US-guided biopsies, which could affect percentage of agreement between cytology and histology in this setting. The criteria for choosing which patients will undergo various types of sampling is determined by circumstances related to clinical decision making (including client preference), overall patient health status, and other factors, which may have introduced bias into the study. The potential for selection bias in the patient population was evaluated by determining if there were statistically significant differences in the relative distributions of cytologic diagnoses for cases with histology compared with cases excluded from the cytologic accuracy study because histology results were not available (Table 4). The χ² analysis revealed a higher frequency of cytologic diagnosis of inflammation and necrosis in the patients that ultimately had histologic evaluation performed. Those results suggest that given a diagnosis of inflammation, clinicians are more likely to proceed with biopsy possibly because it is generally accepted that cytology lacks a high degree of accuracy for inflammation in the liver and that evaluation of tissue architecture for the distribution of inflammation is important for diagnosis. When given a diagnosis of necrosis, clinicians are more likely to proceed with biopsy to look for underlying processes, such as neoplasia, in which necrosis may be a secondary finding. Ideally, a future prospective study conducted with a controlled sample population in which all patients with FLL receive both cytologic and histologic sampling would be beneficial to provide a more complete picture of the accuracy of US-guided cytologic sampling in the diagnosis of canine FLL.

Conclusion

Based on the study results, clinicians can have a high degree of confidence when a cytologic diagnosis of neoplasia is given for focal liver lesions. Cytology is less reliable for excluding the potential for neoplasia, detecting inflammatory disease, and confirming a diagnosis of vacuolar change, hyperplasia and necrosis.