• Funding source: None.


Magnetic resonance imaging (MRI) has been recommended for staging and surgical planning in cats with injection site sarcomas (ISS). The purpose of this retrospective study was to describe low-field MRI characteristics of confirmed injection site sarcomas in a group of cats. Low-field MR images, thoracic radiographs, histopathology findings, and medical records of cats that fulfilled histological criteria of injection site sarcoma were retrieved and reviewed retrospectively. Presence or absence of tumor mineralization and pulmonary metastases were recorded from thoracic radiographs. Characteristics recorded from low-field MRI studies included tumor number, volume (ellipsoid method), intensity relative to surrounding musculature, homogeneity, regions of signal void (mineralization) or cavitation, degree and pattern of contrast enhancement, tumor margination, presence of a peripheral T2W hyperintense zone, and bone contact. A total of 19 cats met inclusion criteria. Cats with multiple tumors were more likely to have had previous excisional biopsy, and were less likely to undergo definitive surgery. All tumors were hyperintense relative to surrounding musculature on T1W and T2W images. Larger tumors were more likely to exhibit mineralization (P < 0.05). Tumor volume could not predict tumor-free margins at definitive surgery. The majority of tumors showed moderate to marked heterogeneous contrast enhancement. Infiltrative margins and the presence of a peripheral T2W hyperintense zone were more prevalent following excisional biopsy, while cavitation was more prevalent following incisional biopsy. Findings indicated that low-field MRI characteristics of injection site sarcoma may vary widely and may be affected by prior incisional or excisional biopsy.


Feline injection site sarcomas (ISS, vaccine-associated sarcomas, VAS) are a heterogeneous group of sarcomas that occur at injection sites in cats months to years postinjection, and include fibrosarcoma, malignant fibrous histiocytoma, and chondrosarcoma.[1] Injection site sarcomas can be defined as those tumors that develop in an area used for injections (traditionally the scruff or interscapular area in the case of vaccinations), have characteristic histological features, display aggressive growth patterns, and tend to recur.[2] Criteria for histologic diagnosis include the presence of blue-gray aggregates within macrophages (identified as adjuvant and oxygen), marked peritumoral inflammation, high mitotic activity, and a necrotic center.[3-5] The link between vaccination and sarcoma formation was first suggested in 1991, following the introduction of Pennsylvanian State laws making vaccination of cats for rabies compulsory.[3, 6] This coincided with the introduction of both a killed aluminium adjuvanted rabies vaccine and a feline leukaemia virus vaccine in 1985. The causal and temporal relationship between vaccination and sarcoma formation has been supported by a shift in the distribution of sarcomas subsequent to the Vaccine Associated Feline Sarcoma Task Force (VAFSTF) recommendation for differential vaccination sites in the hindlimbs and right shoulder area.[7] Radical first surgery is likely to be the most important prognostic factor, however both pre- and postoperative radiotherapy and the administration of chemotherapeutics have been reported to be useful in the management of injection site sarcoma.[2, 8-11] The use of cross-sectional imaging to assess the size and extent of human soft tissue sarcomas is critical for surgical success, with magnetic resonance imaging (MRI) considered the ideal imaging modality.[12, 13] The appearance of feline fibrosarcomas using MRI, and injection site sarcoma using CT has been reported, with both modalities considered useful for surgical planning.[14, 15] The purpose of this retrospective study was to describe low-field MR imaging characteristics of confirmed injection site sarcoma in a group of cats.

Materials and Methods

The imaging database of the Queen's Veterinary School Hospital, University of Cambridge, was searched from 2002 to 2010 for cats that had undergone MRI and that had a histological diagnosis of injection site sarcoma. Only animals with a history of prior vaccination, a tumor in the dorsal neck or inter-scapular area (consistent with current UK practice), and a documented mortality status (i.e., alive or dead at the study endpoint), as obtained by contacting the referring veterinarian or owner, were selected. The endpoint was the date of death, or the date of final data capture in those animals still alive at the end of the study. All data for the study were recorded after consensus was reached by a third-year resident and board-certified radiologist (M.E.H.). Imaging, surgical, and histopathological reports, signalment, and history provided by the referring veterinarian were reviewed. Time points for medical record data collection included the dates of first presentation and surgery at the referring veterinary surgeon (including both incisional and excisional biopsy), MR imaging, definitive surgery, and death. Definitive surgery was defined as excisional surgery by a board-certified surgeon with the intent to obtain tumor-free margins, performed after MRI. Tumor-free margins were defined based on reports by the original examining histopathologist. Mounted hematoxylin and eosin sections were re-reviewed by a pathologist when they were available. Two lateral thoracic radiographs that were taken as part of clinical staging to identify metastatic disease were also examined and the presence of pulmonary metastases or mineralization of the tumor were recorded. Tumor dimensions in three planes were recorded using standard image viewing software and electronic callipers (Visbion, Surrey, UK). T1W images were used to define the tumor margins, and the largest length (L), width (W), and height (H) were determined from all planes. Tumor volume was calculated by the ellipsoid method (L × W × H × π/6).[16] Tumor number was noted (single or multiple), and the volumes of multiple tumors were measured separately and summated. Tumor signal intensity was recorded relative to muscle, and classified as homogeneous or heterogeneous on T1W and T2W images. The MR images of cats with evidence of radiographic mineralization were also evaluated for regions of signal void in a consistent position on T1W and T2W images. Cavitated lesions were defined as T2W homogenously hyperintense areas (as compared to the rest of the tumor) with clearly defined borders within the tumor, and identified as either present or absent.

In cases that received contrast agent, the degree of contrast enhancement was divided into one of two categories: none to mild, or moderate to marked. The pattern of contrast enhancement was also described (homogenous or heterogeneous). Tumor margination was classified as either clearly defined with an expansile growth pattern, or infiltrative (where the margin of the tumor was irregular or infiltrative cords extended into neighboring tissues). The presence of a peripheral T2W hyperintense zone exterior to the T1W gross tumor margin was recorded as either present or absent. Signal intensity for this zone was described relative to normal muscle. Tumor margination and causes of the peripheral T2W hyperintense zone were not characterized histologically because sample labeling was not standardized. Bone involvement was scored as direct contact between the tumor and bone cortex without any interposed fascial planes, or the presence of osteolysis.

Statistical tests were selected and performed by a veterinary statistician (M.A.H.). All analyses were performed using statistical software (SPSS v.20 2011, IBM, New York, NY). Statistical significance was set at P < 0.05. Right censored data were included in the survival calculation (i.e., data from cats that were still alive at the endpoint of the study, providing a minimum survival figure but not their actual survival periods). Statistical analysis for the hypothesis that previous biopsy type would influence survival outcome was performed with a Cox's proportional hazard model. Comparison of volumes for mineralized and nonmineralized tumors was tested using a Student's t-test. The ability of tumor volume to predict tumor-free margins at definitive surgery was tested using binary logistic regression.


A total of 19 cats were included in the study, of which 14 were male (74%), and 5 were female (26%). All animals were neutered. Seventeen cats were domestic shorthaired breeds, 1 cat was domestic longhaired, and 1 cat was Persian. Seven of the nineteen cats (37%) were alive at the endpoint of the study. All cats that died prior to the endpoint of the study were euthanized due to tumor recurrence. The mean age of presentation at referral was 9.5 years (median 9 years, range 4–15 years). Prior to referral, 6 cases had undergone fine-needle aspiration (of which none were cytologically diagnostic), 10 cases had undergone incisional biopsy, and 9 cases had undergone excisional biopsy. All of the excisional biopsy cases had local recurrence prior to referral even though 2/9 (22%) had had apparently tumor-free histological margins following the first surgery. No thoracic metastases were detected in any cases that had thoracic radiographs taken (17/19).

MR imaging was performed prior to October 2008 with a 0.2T scanner (Esaote VetMR s.p.a, Via Siffredi, Genoa, Italy) and subsequently with a 0.25T scanner (Esaote MR Grande s.p.a, Via Siffredi, Genoa, Italy). All cats had been anesthetized and placed in sternal recumbency. Protocols varied between cases. T1W (T1W, TR = 520–1020 ms, TE = 18–26 ms) and T2W (T2W, TR = 2430–3000 ms, TE = 80–90 ms) sequences were available for all cases. Postcontrast T1W sequences were available for 12 cats, after administration of an intravenous bolus of 0.1 mmol/kg of gadobutrol (Gadovist, Bayer Schering Pharma, Berkshire, UK). Other nonroutine sequences obtained included Gradient Echo STIR (GE STIR, TR = 1100–1680 ms, TE = 25 ms) in three cats, T2W STIR (T2W STIR, TR = 2060 ms, TE = 80 ms) in two cats, high-resolution gradient echo (Hi Res GE, TR = 325 ms, TE = 16 ms) in one cat, and proton density (PD, TR = 2430–2600 ms, TE = 28 ms) sequence in three cats.

Five cases underwent radiation therapy after imaging. Three cats received four preoperative fractions of 350–400 cGy at 2- to 3-day intervals, one cat received two preoperative fractions of 900 cGy (days 0 and 7) from a 6 MV linear accelerator, and one cat received four postoperative weekly fractions of 800 cGy from a 4 MV linear accelerator, all in parallel-opposed beam configuration. Preoperative fractions were with the intent of downsizing of the mass, while the postoperative fractions were for treatment of dirty margins. Thirteen of the nineteen cats (68%) underwent definitive surgery and of these histologically tumor-free margins were obtained in 11/13 cats (84%). The minimum distance between tumor margin and surgical margin was specified in 5/11 histopathological reports and ranged from 2 mm to 15 mm. All reports had findings consistent with injection site sarcoma, most frequently describing an intermediate- to high-grade fibrosarcoma, with an inflammatory component consisting of lymphocytes, plasma cells, macrophages, reactive fibroblasts, and multinucleate giant cells. No reports specifically stated the presence of blue-gray aggregates. Immunohistochemistry and electron probe X-ray microanalysis were not performed.

The reasons cited for not treating the remaining six cases were that the tumor was inoperable (4/6) or that the procedure was declined by the owners (2/6). The median survival time postimaging of the cases that did not undergo definitive surgery at referral was 127 days (range 6–162 days). In cases undergoing definitive surgery, median survival time postsurgery was 431 days, (range 91–2125 days). Of these, there was no difference in survival outcomes, measured as time from definitive surgery until death, between cases that had previous excisional biopsy (median 488 days, range 150–1245 days) and those that had previous incisional biopsy (median 424.5 days, range 99–2125 days; P = 1.0). Microchips were present in seven of our cases, and all manifested as magnetic susceptibility artifacts, but in none of these were the images considered nondiagnostic for the study. Size measurement was minimally affected, as we could obtain the three maximum dimensions required for the ellipsoid volume calculation by using all three planes. The mean of the single largest tumor dimension recorded for each tumor was 4.92 cm (median 4.7 cm, range 1.9–12 cm).

Mean tumor volume was 36.4 cm3 (median 23 cm3, range 2–138 cm3). Tumor volume could not predict the ability to obtain tumor-free histological margins at subsequent definitive surgery (P = 0.537). At the time of imaging, 15/19 (79%) cats had a single tumor and 4/19 (21%) cats had multiple tumors. Multiple tumors on MR imaging were more prevalent in animals that had excisional biopsy (33%, 3/9) compared to those that had incisional biopsy (10%, 1/10). Two of four (50%) cases with multiple tumors proceeded to definitive surgery at referral, whereas 11/15 (73%) of cases with single tumors did the same. Definitive surgery yielded tumor-free margins in 10/11 (91%) of single tumors, and 1/2 (50%) of multiple tumors. All tumors (19/19, 100%) were hyperintense relative to muscle on T1W and T2W images. On T2W images, all tumors (100%) were heterogeneous, while on T1W images, 14/19 (74%) tumors were heterogeneous. Of the 5/19 (26%) tumors that were homogeneous on T1W images, 3 had previous excisional biopsy, and 1 had previous incisional biopsy.

Radiographic mineralization of the tumor was noted in 3/17 cases (18%; Fig. 1). The three tumors that had radiographic evidence of mineralization, had signal void areas within the tumor in a consistent position on T1W and T2W images. Two of these had undergone previous excisional biopsy and the other had undergone previous incisional biopsy. In the only case where both radiographic mineralization was observed and H&E slides were available for review, multiple small mineralized areas were confirmed on light microscopy. There was a statistically significant difference in volume between nonmineralized tumors (mean 27 cm3, median 20 cm3, range 2–86 cm3) and mineralized tumors (mean 91 cm3, median 113 cm3, range 22–138 cm3; P = 0.008).

Figure 1.

Right lateral thoracic radiograph (A), transverse plane T1W (B), and T1W postcontrast (C) MR images taken at the level of T3 in a cat with an injection site sarcoma. Radiographic tumor mineralization is present in locations that match signal void areas in the MR image B (white arrows). Marked and heterogeneous contrast enhancement is seen (C, black arrows). A magnetic susceptibility artifact on the MR images is consistent with the location of a microchip seen on the radiograph (white stars).

Tumor cavitation was present in 58% of cases, comprising 70% (7/10) of cases that had had incisional biopsy, and 44% (4/9) of cases that had had excisional biopsy.

Only 1 of the 12 that received contrast agent (8.3%) had no to mild contrast enhancement, and only one case (8.3%) had a homogenous pattern of contrast enhancement. The remainder (10/12, or 83%) showed both heterogeneous and moderate to marked enhancement (Figs. 1 and 2).

Figure 2.

Three transverse plane MR images taken at the level of T5 in a cat with an injection site sarcoma: T1W (A), T1W postcontrast (B), T2W (C). The tumor directly contacts the dorsal margins of both scapulae (A, white arrows). Infiltrative margins of the tumor are evident (B, white arrows). The tumor is moderately and heterogeneously contrast-enhancing. Cavitations are seen on the T2W image (C) as regions that are homogenously hyperintense relative to the rest of the tumor (white arrow).

Infiltrative margins were present in 15/19 cases (79%; Figs. 1-3). All cases that had excisional biopsy (9/9, 100%), and 4/10 (40%) of cases that had incisional biopsy, had infiltrative margins.

Figure 3.

Transverse plane MR images taken at the level of T6 in a cat with an injection site sarcoma: T1W (A), T2W (B). The tumor margin was classified as clearly defined or expansile (white arrow, A), with the absence of peripheral T2W hyperintense zone (white arrows, B).

The presence of a T2W hyperintense zone peripheral to the T1W gross tumor margin was seen in 3/10 (30%) and 8/9 (89%) of cases that had incisional and excisional biopsies, respectively (Fig. 4).

Figure 4.

Dorsal plane T2W MR image at the level of the thoracic inlet of a cat with an injection site sarcoma that extends to the left shoulder region. The trachea (T), left (L), and right (R) humerus are labeled. The ventral portion of the tumor is labeled (star), and its edges are defined by the small open arrows. The tumor is surrounded by a zone of peripheral T2W hyperintensity relative to adjacent musculature (white arrows) that may reflect peritumoral inflammation, hemorrhage, edema, or extension of the tumor.

In the cats with bone contact (13/19), none had evidence of osteolysis. The presence of bone contact prompted ostectomy of affected regions in all cats undergoing definitive surgery. Seven out of ten (70%) cases that had previous incisional biopsy and 6/9 (67%) cases that had previous excisional biopsy had bone contact.


Low-field MR imaging characteristics of feline injection site sarcoma recorded in this study were similar to those routinely used for staging and surgical planning of human soft tissue sarcomas.[17-19]

The median survival of cases undergoing definitive surgery after MRI was 431 days. Median survival time of 576 days have previously been reported.[9, 11] Recently, a median survival time of 901 days was reported in cats undergoing radical (5 cm margin) surgery.[10] The survival time in currently reported cases that underwent definitive surgery was not related to the method of initial biopsy. This finding was consistent with two previous reports that also found no association between number of previous surgeries or presentation with locally recurrent disease and prognosis,[1, 10] but was inconsistent with another report and the VAFSTF recommendations that initial biopsy should be incisional or needle-core.[9, 20]

No pulmonary metastases were detected on thoracic radiographs. The use of a two view technique to screen dogs for pulmonary metastases has previously been validated.[21] Inclusion of an orthogonal view (DV or VD) to a two lateral-view series has been reported to increase diagnostic sensitivity by 12%, although at increased cost and personnel exposure.[22] As such we cannot exclude that a low percentage of pulmonary metastatic disease may have remained undetected. In human soft tissue sarcomas, tumor volume on MR images is predictive of operative size and prognostically important.[18, 23, 24] Tumor volume of feline injection site sarcomas measured by callipers has previously been reported not to be prognostically important.[1] In this group of cats, tumor volume could not predict tumor-free margins at definitive surgery. Direct comparison between tumor-free margins obtained at surgery and prognosis was not possible in this study due to our small sample size. The large variation in tumor volume in our cases (range 2–138 cm3), was similar to that seen in a CT-based volumetric analysis of feline vaccine-associated sarcomas (range 1.6–384.2 cm3).[15] The median MRI tumor volume of our cases (23 cm3) was lower than that seen in the CT study (median 44.7 cm3 on precontrast images). Although this may partly be due to minor difference in the volumetric formulas used, it could also be due to variation in the patient population, or differences inherent in the imaging modalities used.

In the present study, multiple tumors were more common in animals that had undergone excisional biopsy prior to imaging. A previous report found both multiple recurrences after surgery, and multiple tumors at independent vaccination sites.[25] We speculate that, in our cases, these were due to islands of neoplastic cells remaining after primary surgery. In all four cases with multiple tumors, the lesions were in close proximity and each cat had initially presented to the referring veterinarian with a single tumor. Multiple tumors may influence the decision to proceed with complete surgical excision, and the likelihood of obtaining tumor-free margins. Both cases that had dirty margins following complete surgical excision had undergone excisional biopsy prior to referral. Patients presenting with recurrence following excisional biopsy may thus be disadvantaged relative to those with prior incisional biopsy, as previously reported.[9]

Radiographic tumor mineralization was consistently seen on MRI, and was more likely to be observed in larger tumors (P < 0.05). The presence of tumor mineralization was not reported in a CT study of feline vaccine associated sarcomas.[15] We could only obtain histological confirmation of radiographically evident mineralization in one case, and speculate that mineralization represents zones of dystrophic mineralization in necrotic portions of the tumor.

Assessment of the periphery of sarcomas in people is prognostically important. Tumors with a benign appearance of the periphery (i.e., expansile or “pushing” type of growth pattern, or only focal invasion) carry a better prognosis compared to those with a diffusely infiltrative appearance of the periphery.[17] In the present study, infiltrative margins were more prevalent in cases that had excisional biopsy compared to incisional biopsy (100% vs. 40%). The histology of the peripheral T2W hyperintense zone has been studied in people. It is commonly referred to as the “reactive zone” and this term is misleading because, in addition to edema, it may also reflect neovascularity or satellite tumor cells in the periphery of the gross tumor. In 15 people with soft tissue sarcoma, two thirds of samples of the tumor periphery displayed tumor cells up to 4 cm from the gross tumor margin, with no correlation between the presence or extent of the MRI peripheral T2W hyperintense zone and the presence of tumor cells beyond the tumor margin.[26] In our study the presence of a peripheral T2W hyperintense zone was more prevalent in cases that had excisional biopsy compared to incisional biopsy (89% vs. 30%). Thus this may also represent residual scarring or inflammation. The nonstandardization of sample labeling precluded a comparison of tumor margin histology to MR appearance in this study.

The presence of osteolysis was not seen in any cat in this study, in any of the sequences examined. This is similar to findings in a CT study of feline vaccine-associated sarcoma.[15] The use of MRI to detect osteolysis has been described in high-field MRI in people, where proton density and T1W sequences are preferred due to their reduced susceptibility to metallic artifact.[27] Bone involvement was therefore defined in this study as direct contact of tumor with cortical bone, and was similarly distributed in those cases that had previous incisional biopsy (7/10, 70%), and previous excisional biopsy (6/9, 67%).

Limitations of this study primarily were related to the retrospective nature of the study, that is, small sample size, nonstandardization of MRI protocols, lack of follow-up diagnostic imaging, and lack of samples for histological review. The wide variation in tumor-free margin status for our cats could have been the cause for the widely varying outcomes. Although we could not definitively prove that all our cases were injection site sarcomas, we used generally accepted criteria, including position of the tumor, history of vaccination, and histological appearance. The use of mortality date rather than time to first recurrence may have biased outcome results, given the variation in owner and veterinarian perception of quality of life in those cases with recurrence. Mortality date was selected as a final criterion due to the limited availability of patient history after discharge.

The presence of microchips in seven of our cases limited complete evaluation of the tumor due to magnetic susceptibility artifact. Volume measurements could still be acquired from the three planes, and character of the tumor and margins could still be assessed from the images. Surgical removal of the microchip to improve image quality was not performed due to reluctance to disrupt the soft tissues prior to definitive surgery. Five of 11 cases that were scored as having a peripheral T2W hyperintense zone had no postcontrast T1W images. This limited confident characterization of the zone, and would have been useful to differentiate peritumoral edema from tumor infiltration, or inflammation subsequent to previous biopsy.

In conclusion, low-field MRI characteristics of feline injection site sarcomas are highly variable. The majority presented as a single tumor, with a broad range of volumes. All of the tumors were hyperintense relative to surrounding musculature on T1W and T2W images, and most (T1W) or all (T2W) were heterogeneous. Tumor mineralization may be more commonly seen in larger masses, and is expected to be evident both on radiographs and MRI. Just over half the tumors appeared cavitated. Of the cases in which postcontrast images were available, a large proportion showed moderate to marked, heterogeneous contrast enhancement. The majority of tumors displayed infiltrative margins and a peripheral T2W hyperintense zone, however no standardized histologic comparisons were available and some of these findings could have been due to previous biopsy procedures. Cases that had previous excisional biopsy were more likely to present with multiple tumors, exhibit infiltrative margins and a peripheral T2W hyperintense zone, while those that had incisional biopsy were more likely to be cavitated.