Ultrasound is a poor predictor of early or overt liver or spleen metastasis in dogs with high‐risk mast cell tumours

Conflicting evidence exists regarding the importance of routine abdominal ultrasound (US) with hepatic and splenic fine needle aspiration (FNA) cytology during staging of canine mast cell tumours (MCT). The objective of this study was to correlate ultrasonographic and cytologic findings in dogs with strictly defined high-risk MCTs and to determine the influence on outcome. Our hypothesis was that US poorly predicts visceral metastasis in high-risk MCTs and that early metastasis is associated with improved outcome when compared to overt metastasis. US of liver and spleen correlated to cytologic results, categorised as no metastasis, early metastasis or overt metastasis. Of 82 dogs prospectively enrolled, 18% had early visceral metastasis and seven % had overt metastasis on cytology; 67% with visceral metastasis had regional LN metastasis. US was a poor predictor of metastasis with sensitivity, specificity, positive predictive value and negative predictive value for the spleen of 67%, 68%, 21%, and 94%, respectively and for the liver of 29%, 93%, 56%, and 82%, respectively. Median time to progression (TTP) for dogs with no metastasis, early metastasis, and overt metastasis was not reached, 305 days and 69 days, respectively (p<0.001). Median survival time (MST) for the three groups were not reached, 322 days and 81 days respectively (p<0.001). High Patnaik or Kiupel grade, early metastasis, overt metastasis and adequate local control were significantly associated with outcome. Early visceral metastasis was associated with poorer outcome compared to dogs without metastasis, however a subset of dogs experienced long-term control. This article is protected by copyright. All rights reserved.

While current literature provides information on dogs with distant metastasis, precise staging data is often lacking clear documentation of whether visceral metastasis was classified on the basis of cytologic findings or US appearance alone. 11,18,30,[34][35][36][37][38] When normal on US, liver and spleen may not be aspirated and several studies include patients without full staging. 12,19,[39][40][41][42][43] The utility of US to predict cytologic findings is controversial. One study demonstrated that sensitivity of US to detect liver and spleen metastasis was 0% and 43%, respectively. 44 Another study suggested a sensitivity of 71% for both liver and spleen. 45 A recent study reported that all dogs with cytologic confirmation of metastasis had abnormal ultrasound findings. 27 Routine aspiration of the liver and spleen is also debatable. The presence of mast cells is not definitively diagnostic for metastasis as they can be present in healthy dogs and dogs with nonneoplastic conditions. 28,[46][47][48][49][50][51][52] Cytologic guidelines for the designation of visceral MCT metastasis have been described. 27 Using these criteria, dogs with distant metastasis were shown to have a poor outcome with a median survival time (MST) of 34 and 100 days compared to >733 and 291 days, respectively, for dogs without metastasis. 27,44 In the authors' experience, some dogs have cytologic findings that do not fit the criteria for overtly positive metastasis. These dogs have FNA results characterized by, a mild subjective increase in numbers of mast cells of normal morphology, not associated with connective tissue elements, where early metastasis cannot be ruled out. This cytologic finding therefore creates a challenge when determining treatment decisions.
The overall value of routine US and cytology of the liver and spleen in dogs with MCTs is unclear, because dogs with low-tointermediate grade tumours without negative prognostic factors have a favourable prognosis after surgery, even following incomplete or narrow margins. 20,50,53,54 However, routine hepatic and splenic cytologic assessment is probably useful for dogs at high-risk for visceral metastasis. 27,44 Cytologic evidence of early metastasis may be prognostic and alter treatment decisions. The objective of this study was to prospectively correlate US and cytologic findings in dogs with highrisk MCTs, and to determine the outcome of dogs with early metastasis compared to dogs without metastasis and those with overt visceral metastasis. The hypothesis was that US is poorly predictive of hepatic or splenic metastasis when selecting for dogs at high-risk for MCT metastasis, and that dogs with cytologic evidence of early metastasis have improved outcomes compared to dogs with overt metastasis.

| Dogs
Dogs that presented to a referral oncology service between March 2013 and May 2017 for staging of high-risk MCTs diagnosed on cytology and/or histopathology were prospectively enrolled. The study was conducted following approval by the University veterinary ethical review committee. Written owner consent was obtained for all dogs. A subset of the cases had previously been published in a study on escalating vinblastine protocol in dogs with MCTs. 55 Dogs were included if they had negative prognostic indicators consistent with biologically aggressive behaviour and if US and FNA cytology of liver and spleen were scheduled to be performed as part of full staging.
Strict criteria for designation of "high risk" included the following negative prognostic indicators: grade III Patnaik and/or high grade Kiupel), mitotic count >5, LN metastasis, recurrent disease, recent rapid growth and ulceration, and high-risk location (mucocutaneous, preputial, scrotal or visceral).  Rapid growth was defined as volume increase of >200% in the 30 days prior to clinical exam.
LN evaluation on cytology was performed using previously published guidelines and histopathological evaluation from 2015 onwards was performed according to the classification published in 2014. 15,32 Both HN2 and HN3 were documented as metastatic for the current study as these had been reported to be associated with outcome.

| Staging
Staging included physical examination with MCT measurement where present, haematology, serum biochemistry, regional LN evaluation and FNA cytology, three view thoracic radiographs, and abdominal US with FNA cytology of liver and spleen. All palpable regional LNs were sampled; non-palpable LNs were sampled where feasible with USguidance. If surgical removal was elected, the regional LN was excised and submitted for histopathology. US with FNA of the liver and spleen was performed by board-certified specialists in diagnostic imaging.
Sedation was used to facilitate imaging and FNA. Liver and spleen appearance on US was characterized as nodules, masses, altered echogenicity, enlargement or normal appearance. FNA of the liver/ spleen was attempted regardless of the US findings and when abnormalities were present, both abnormal and normal regions were aspirated. FNA was performed using a 23-gauge needle attached to a 5 mL syringe without any active aspiration at the time of sampling to avoid blood contamination, based on radiologist preference. Slides were stained with May-Grünwald Giemsa and cytologic evaluation was performed by one of two board-certified pathologists. Prior to sampling the liver and spleen, manual platelet counts were assessed and congenital coagulopathies were ruled out based on clinical history. If there was a history of liver disease, partial thromboplastin time and activated partial thromboplastin time were assessed in house.

| Cytologic interpretation of liver and spleen
Dogs were assigned to 3 groups based on a modification of previous recommendations and included: no metastasis, early metastasis and overt metastasis. 27 Criteria used can be seen in Table 1 Kiupel) or incomplete or narrow surgical margins followed by definitive-intent radiation therapy (RT) of the primary tumour site and regional LN bed where feasible. 54,[56][57][58][59] Chemotherapy with escalating vinblastine and prednisolone was recommended for all dogs diagnosed with MCT at high-risk for metastasis following local control. 55 TTP was defined as time from date of diagnosis to the date of progression of the primary tumour, local recurrence or development of metastasis. The development of de novo MCTs was not documented as progression unless greater than five MCTs appeared rapidly (within 14 days) in the face of therapy, in which case skin metastasis was presumed. Survival time was documented as time from date of diagnosis to date of death. Recommended follow-up included recheck examination once every 3 months for 18 months, followed by once every 6 months for 18 months, followed by yearly revisits. At the time of recheck, physical examination, regional LN measurement +/− FNA cytology, and US and with liver/spleen FNA were recommended. Cox-proportional hazards regression model was used to determine if high-risk location, MCT recurrence, rapid growth and ulceration, grade III Patnaik, high-grade Kiupel), high MC, LN metastasis, early hepatic or splenic metastasis, overt hepatic or splenic metastasis, surgery, adequate loco-regional control or chemotherapy had a significant impact on outcome. The proportional hazard assumption was assessed by graphical examination of the Schoenfeld residuals for each variable against time and a null hypothesis test that there is not a time dependent trend in these residuals (using the cox.zph function in the R survival package). Multivariable analysis was performed to evaluate for factors related to outcome within groups and between groups. For this analysis, variables were grouped into primary tumour characteristics (rapid growth and ulceration, high risk location, recurrence, grade III Patnaik or Kiupel high grade, high MC), metastasis status (no metastasis, LN metastasis, early liver or spleen metastasis, overt liver or spleen metastasis) and treatment variables (surgery of the primary tumour +/− regional LN, adequate local control, chemotherapy). A P-value of <.05 was considered statistically significant.

| RESULTS
Eighty-two dogs met the inclusion criteria and were enrolled. One dog previously received targeted therapy with masitinib for a MCT, however, this medication was discontinued 1 month prior to referral  (Table 4). Haematology and serum biochemistry were performed in all dogs. Thoracic radiography was performed in 74 cases (90%) and none revealed lymphadenopathy or pulmonary metastasis.
All regional LNs were assessed (for MCTs on the caudal half of the T A B L E 1 Criteria used to assign metastasis status on cytology of the liver and spleen were assessed on palpation, however, for axillary, inguinal and medial iliac lymph nodes ultrasound was used. All lymph nodes assessed were sampled where possible, however, axillary, inguinal and medial iliac lymph nodes were difficult to sample when normal in size. Regional lymphadenomegaly was documented in 33 dogs (40%). Sixty-one dogs (74%) had at least one regional LN assessed cytologically (n = 55) and/ or histologically (n = 48). Overall 41 dogs (50%) were diagnosed with metastasis ( Table 5).  robust multivariable statistical analysis and in order to evaluate a number of prognostic factors, we would need 10 times as many events as prognostic variables to detect statistical significance. 66 Even with comprehensive multivariable analysis on large case numbers, it is challenging to interpret the results as there is a complex relationship between primary tumour characteristics, metastatic status and treatment variables. For example, dogs with overt visceral metastasis achieved adequate local control less commonly (33% vs 67% of the dogs without metastasis). This is possibly due to more aggressive biologic behaviour, however, this cannot be concluded with certainty as definitive treatment options are less likely to be pursued in the face of visceral metastasis. However, the study was designed to clearly define "increased numbers of mast cells" and to follow dogs with high-risk F I G U R E 2 Kaplan-Meier curves showing survival times for dogs with no metastasis (n = 61), early metastasis (n = 15) and overt metastasis (n = 6). For dogs without metastasis, the median survival time (MST) was not reached, while for dogs with early and overt visceral metastasis, the MST was 322 and 81 days, respectively (P < .001). The full, striped and dotted curves represent dogs without metastasis, with early metastasis and overt visceral metastasis, respectively T A B L E 8 Univariable analysis for multiple variables evaluating association with time to progression (TTP) and median survival time (MST); Likelihood ratio test with P < .05 considered significant (indicated by *) tumours to evaluate if definitive treatment is rational in dogs with early visceral metastasis.
As aspiration of the liver and spleen were previously demonstrated to be safe with no obvious complications 27,44 ; repeat imaging was not routinely performed the same day following sampling to rule out intra-abdominal haemorrhage. No complications were seen in this study.
Fifty percent of the dogs with high risk MCTs had LN metastasis, 18% had early liver and/or spleen metastasis and 7% had overt metastasis. The LN metastatic rate in this study was high, which reflects the population of high-risk MCTs recruited for this study as was seen in previous literature. 12,30,39,43 The rate could have been biased as LN metastasis was one of the inclusion criteria. Early metastasis to the liver and/or spleen has not been previously associated to outcome to the authors' knowledge, and the rate of early metastasis using our criteria was higher than expected at 18%. for metastasis rather than positive when determining the therapeutic plan.
Adequate local control was defined at the start of the study, recognizing that there is no clear standard of what defines loco-regional radiation therapy boundaries or exactly what margin is required to control grade 3 MCTs. 56 However we were consistent with current literature in the standards set forth at the start of the study. 53,56,72 Labradors were overrepresented at 45% of the canine population in this study. This is markedly higher than multiple previous reports that included 11%-27% Labradors. 11,12,19,28,30,31,34,35,37,39,41,74 In two studies that used similar inclusion criteria to ours, Labradors represented 41% and 37% of their study populations, supporting the possibility that Labradors are not only more prone to developing MCTs but are also more likely to develop high-risk MCTs. 18,44 A recent study did not identify the Labrador Retriever as a breed with an increased risk of developing higher grade MCTs, however, the study did not take other prognostic variables into consideration. 75 Epidemiologic studies would be required to evaluate the proportion of Labradors in the overall population in this region and in our referral hospital prior to drawing any conclusions.
Only dogs with high-risk MCTs that fit to our specific criteria were included in the study. It is possible that some dogs were not included in the study even though they had negative prognostic factors that we were unaware of, particularly as we did not have tumour size for all dogs if they were referred following treatment at the referring practice. 61 Size has previously been reported as prognostic for dogs with MCTs, however, was not evaluated as a prognostic indicator in this study as recent rapid growth was more easily assessed, even in cases that sought initial surgery with the referring practice. 61 Size was not consistently recorded for cases which did not initially seek treatment with oncology, and as MCTs can shift size dramatically, we felt that recent rapid growth was more reliable. Additionally, others have suggested that rapid growth rather than size might be a more accurate predictor of prognosis. 8 Additional studies could also include size, to better determine if it is an independent negative factor. 19,21,28,56 It is currently difficult to conclude a true visceral metastasis rate in low-risk MCTs (defined as MCTs lacking negative prognostic indicators as used in the current study) as previous literature did not consistently clarify all negative factors nor report consistent hepatic or splenic aspiration cytology. 12