Clinical relevance of splenic nodules or heterogeneous splenic parenchyma assessed by cytologic evaluation of fine‐needle samples in 125 dogs (2011‐2015)

Abstract Background Splenic nodules and heterogeneous parenchyma are seen frequently in abdominal ultrasound examinations of dogs, but the clinical importance of these lesions remains unclear. Objectives To determine whether specific ultrasonographic findings correlate with clinically relevant cytologic diagnoses and determine what sonographic features are correlated with these diagnoses. Another objective was to develop a scoring rubric to help clinicians make decisions on whether or not certain ultasonographic findings of the spleen warrant evaluation by fine‐needle cytology. Animals One‐hundred twenty‐five adult client‐owned dogs with ultrasonographically identified splenic nodules, heterogeneous parenchyma, or both. Methods Medical records were retrospectively searched for ultrasound‐guided splenic fine‐needle cytology reports. Ultrasonographic images were assessed for nodule number, size, echogenicity and distal enhancement, degree of splenic heterogeneity, and peritoneal fluid. Dogs were divided into 2 groups: those with clinically important or clinically irrelevant cytologic findings. Potentially useful and discriminatory ultrasonographic findings were identified by statistical analysis, and the most useful findings were used to generate the scoring rubric. Results The clinically important group included 25 of 125 dogs (22 malignancies, 3 suppurative inflammation). Splenic nodules 1‐2 cm in diameter, peritoneal fluid, and >1 targetoid nodule were associated with clinically important cytologic findings. Receiver operator characteristic analysis showed that the scoring rubric was useful for identifying dogs in the clinically important group. Conclusions and Clinical Importance Splenic fine‐needle cytologic findings identified a clinically relevant diagnosis in 20% of dogs, and larger nodule size, number of targetoid lesions, and presence of peritoneal fluid increase the likelihood of detection of clinically important disease.


K E Y W O R D S
canine, FNA, spleen, splenic nodules

| INTRODUCTION
Sonographically detected splenic parenchymal changes can be found during abdominal ultrasound examination in asymptomatic patients or patients with unrelated disorders. Splenic nodules without associated splenomegaly are a relatively common finding in older dogs, 1 are likely to be benign, and may require no further action in this age group. 1 Nodular hyperplasia, hematoma, extramedullary hematopoiesis, congestion, and lymphoid hyperplasia are the most common non-neoplastic lesions found in the spleen of dogs at necropsy or on biopsy, [2][3][4] whereas the most common malignant neoplastic lesions of the spleen in dogs are hemangiosarcoma, histiocytic sarcoma, non-angiomatous or nonlymphoid sarcomas, and lymphoma. 5,6 There is overlap in ultrasonographic appearance of benign and malignant splenic lesions, and they cannot be differentiated based on gray-scale ultrasonography alone. 7 Some studies found that neoplastic and non-neoplastic splenic lesions had similar ultrasonographic appearance, [8][9][10] whereas other studies found that specific appearances of nodules, such as targetoid nodules or multiple similar-appearing nodules, more often were associated with malignancy and single lesions more often were benign. 11,12 Furthermore, diffuse parenchymal changes such as a "moth-eaten" pattern resulting in moderately to severely heterogeneous splenic parenchyma can be seen in lymphoma and some other splenic disorders (e.g., systemic mast cell disease, leukemias, and splenitis) in dogs. 13 Fine-needle cytology is commonly performed in patients with splenic nodules, and fine-needle sample collection is considered safe even in the presence of coagulopathies or thrombocytopenia. 11,14 In a study of human patients, splenic aspirates were diagnostically accurate in 85% (253/298) of the cases. 15 In veterinary patients, splenic fineneedle sample cytology has been reported to be as beneficial as needle core biopsy for diagnosing splenic neoplasia. 16 In 1 study comparing the results of cytology and histopathology, splenic cytology correctly identified the underlying problem in 61% of cases, whereas histopathology was required to distinguish between neoplastic and reactive conditions in 23% of the cases. 11 Similar results were reported in a study comparing cytological and histopathological diagnoses of splenic disorders in dogs, with a 59% agreement rate. 17 The clinical relevance of small, nodular lesions in the spleen and the optimal clinical approach to patients with these lesions remains unclear. This is in contrast to patients with larger splenic masses (>2 cm diameter) or known underlying disease with a high rate of splenic involvement (e.g., mast cell tumor, lymphoma), in which it is generally accepted that fine-needle aspirates of the splenic tissue are indicated. 18 Our objective was to determine whether specific ultrasonographic findings correlate with a clinically relevant cytologic diagnosis and determine what sonographic features were correlated with these diagnoses. A further objective was to develop a scoring rubric to help clinicians make decisions on whether or not certain ultasonographic findings of the spleen warrant fine-needle cytology.

| Case Selection
The medical records database of the Oregon State University Veterinary Teaching Hospital was retrospectively searched to identify dogs in which ultrasound-guided splenic sample cytology was performed between July 2011 and June 2015. Criteria for inclusion were ultrasonographic examination of the spleen in which splenic nodules or heterogeneous splenic parenchyma were identified followed by splenic fine-needle sample cytology on the same day. Nodules were defined as being <2 cm, whereas masses were defined as being >2 cm. In patients with splenic nodules identified on multiple visits, the ultrasound examination and cytology results from the first visit only were used in the study. The following criteria were used for exclusion: presence of a splenic mass (>2 cm), dogs known to have mast cell neoplasia or lymphoma before ultrasound examination, absence of a cytology report or ultrasonographic images, and dogs with homogenous splenic parenchyma without any nodules. Dogs with a diagnosis of lymphoma or mast cell neoplasia made during the first visit to the veterinary teaching hospital were not excluded from the retrospective analysis.
Information collected from medical records included signalment, date of imaging and sampling, presenting complaint, and cytologic diagnosis.

| Procedures
Cytologic specimens were obtained by ultrasound-guided fine-needle sample collection performed by board-certified veterinary radiologists.
All ultrasonographic examinations and ultrasound-guided sample collection was performed using an 8 MHz curvilinear transducer. Fine-needle sample collection was performed using 22-gauge, 1.5-inch-long needles using a nonaspiration technique. 19,20 A needle was guided into the spleen and then gently moved up and down along the needle tract in an attempt to harvest cells; no negative pressure was applied to the syringe. 20 Splenic nodules that had characteristic ultrasonographic appearance of myelolipomas were not routinely sampled. 21,22 All ultra- Targetoid lesions were defined as nodules with a hypoechoic rim and a hyperechoic or isoechoic center. 12 In cases with multiple nodules, the size determination was based on the largest nodule. A "clinical importance score" rubric ( Table 2) was created based on the ultrasonographic findings and their statistical relevance (Table 1).
For each of these findings, a specific score was given with a maximum possible total score of 13 points.

| Statistical analysis
For analysis, dogs were classified into group 1 or group 2 based on the cytologic diagnosis and its impact on the management of the case.
Potential marker findings for clinically relevant changes initially were analyzed using contingency tables. Binary variables (e.g., nodules present, peritoneal fluid present) were analyzed using Fisher's Exact Test, whereas variables that were stratified (e.g., parenchymal heterogeneity, total nodule number) were analyzed using the chi-squared test for trend. For the initial univariate analyses, statistical significance was set at P < .05. After univariate analysis, a multiple factor "clinical importance score" rubric was created. Briefly, variables from the univariate analysis that showed a calculated P value < .1 were identified, and numeric scores assigned for each finding. Scores were weighted by relative risk identified in the contingency analysis, with findings that had a high relative risk for the presence of clinically relevant disease given a higher score. For variables that were stratifiable, increasingly more severe findings were given higher weightings in the scoring system. The scoring system was applied to the ultrasonographic data, and the diagnostic performance of the scoring system assessed using receiver operator characteristic (ROC) analysis ( Figure 1). The scoring

| RESULTS
The medical records search identified 245 dogs meeting the initial inclusion criteria. One-hundred sixteen dogs were excluded from the study: 64 dogs were diagnosed with mast cell tumor or lymphoma before splenic fine-needle sample cytology, 45 dogs had a mass ≥ 2 cm in diameter, 6 dogs did not have cytology reports or ultrasound images available, 1 dog had homogenous splenic parenchyma without any nodules, and 1 dog had a nondiagnostic sample.
These excluded cases resulted in a sample size of 125 dogs that met the final criteria for assessment.  Table 1.
The scoring rubric differentiated irrelevant from clinically important splenic lesions, as shown in Table 2. A cutoff value of >5 resulted in a sensitivity of 56% and a specificity of 92% in this patient population (Table 2). A cutoff value >4 resulted in an increased sensitivity of 84% and decreased specificity of 58%. The ROC analysis showed an AUC of 0.804, with a 95% CI of 0.697 to 0.911 (P < .0001; Figure 1).

F I G U R E 1
The receiver operator characteristic (ROC) analysis. The ROC analysis showed an area under the curve (AUC) of 0.804, with a 95% confidence interval (CI) for AUC of 0.697 to 0.911, P < .0001. This indicates that this scoring system is significantly better than random chance (AUC of 0.5) at detecting clinically important disease in the spleen

| DISCUSSION
Our results showed that malignancies, particularly lymphoma, were the most common cytologic findings in dogs with clinically relevant diagnoses. Significant differences in the size of nodules and the presence of peritoneal fluid were found between the groups. The presence of multiple nodules had a P value < .1 between the groups and was incorporated into the overall clinical importance scoring system. No significant differences in age, breed, and sex of dogs as well as splenic parenchymal heterogeneity, margination of nodules, distal enhancement of nodules, and nodule echogenicity were found between the 2 groups.
In our study, 100 of 125 dogs had benign splenic pathology or normal splenic tissue, as interpreted by cytologic evaluation. The ratio of non-neoplastic to neoplastic splenic disease in dogs varies among previous studies. Studies that included all cases of splenomegaly or masses showed >50% prevalence of non-neoplastic diseases. 1,2,11 We found that the presence of splenic nodules 1-2 cm in diameter was predictive of a clinically important cytologic diagnosis. A change in diameter from 1.0 to 1.25 cm is associated with a doubling in volume for a spherical mass, 21 and based on these findings, we recommend performing fine-needle cytology of the spleen in dogs with splenic nodules 1-2 cm in diameter.
In our study, the presence of peritoneal fluid was more common 16 multiple lesions were associated with a malignant neoplasm. 12 In our study, the presence of >1 targetoid lesion was associated with increased relative risk for a clinically important diagnosis on splenic sample cytology, but the number of animals having >1 targetoid lesion (n = 3 dogs, all in group 1) was low, and these results should be interpreted with caution.
The scoring system described here is based on the combination of ultrasound findings that showed potential utility for the presence of a clinically important disease. A score > 5 is suggestive of clinically important ultrasonographic findings with an excellent specificity of 92%, but poor sensitivity of 56%. Decreasing the cutoff score to >4 improves the sensitivity of the scoring system to 84% but decreases its specificity to 58%. The AUC of the ROC analysis of 0.804 (95% CI, 0.697-0.911) indicates that this scoring system is significantly better than random chance (AUC of 0.5) at detecting clinically important disease in the spleen ( Figure 1). Our results suggest that splenic fine-needle sample cytology should be considered in dogs with a score >4-5.
As with most retrospective studies, our study had some limitations.
The first limitation was the absence of histopathologic confirmation of the cytologic diagnosis. Although cytologic diagnoses often reflect histologic results, if incorrect or inadequate sampling occurs or cytology is unable to distinguish between reactive and neoplastic conditions, accurate diagnosis with fine-needle sample cytology may not be possible. 11,15 To minimize the possibility of error, a blinded retrospective review of the cytologic slides was performed by an independent clinical pathologist. Excellent agreement was found between the second slide review and the initial evaluation that did not change the results of the statistical analysis. Our study, however, was not intended to compare or evaluate diagnostic accuracy of cytology as compared to histopathology but to determine whether specific ultrasonographic findings correlated with a clinically important cytologic diagnosis.
A second limitation also is related to the retrospective nature of the study. It was not definitively documented that the splenic fine-T A B L E 2 Scoring system for the likelihood of achieving a cytologic diagnosis that will affect case management based on ultrasonographic findings needle cytology sample was obtained from the nodules or surrounding parenchyma. It is our general procedure to sample nodules as well as some non-nodular parenchyma, but we cannot be sure that this was always the case.
A third limitation relates to the retrospective evaluation of the static ultrasonographic images of the spleen and radiology reports because only saved images could be evaluated and images may not have been acquired of all nodules. All ultrasound examinations were performed by 1 of 2 board-certified radiologists, and multiple images of the spleen were available for all animals. The combination of this consistency combined with a radiology report written within 24 hours of the ultrasound examination and the presence of static images documenting the appearance of the spleen decreases the possibility that the ultrasonographic data incorrectly represented the disease of the patient.
In conclusion, splenic fine-needle sample cytology is a minimally invasive diagnostic modality that identified a clinically important diagno-