Outcome and Prognostic Indicators for Hemangiopericytomas in Dogs: 167 Cases (2009–2016)
ABSTRACT
This retrospective study evaluated the postoperative outcome and clinical parameters associated with prognosis in 167 client-owned dogs with 167 hemangiopericytomas. Parameters that were reviewed for an association with long-term outcome included signalment, clinical history, results of staging tests, tumor and surgical variables, and administration of adjunctive therapy. History of previous surgery, type of surgery performed, status of surgical margins, tumor location, and whether adjunctive therapy was performed were associated with tumor recurrence. The distal forelimb was the most common location reported overall (46/167 [27.5%]). Dogs with tumors located at the tail/perineum had the fastest recurrence rate, with a median disease-free interval of ∼16 mo (505 days). Tumor grade alone was not associated with recurrence (P = .069), but when analyzing tumor grade and margin, low-grade tumors with dirty margins had a significantly shorter time to recurrence than low-grade tumors with either clean or narrow margins. Tumor location should be considered when assessing the treatment plan and follow-up recommendations for any hemangiopericytomas. Aggressive initial surgical treatment is recommended when possible to reduce the chance of local tumor recurrence.
Introduction
Perivascular wall tumor (PWT) is a subcategory of soft tissue sarcomas (STS) arising from nonendothelial vascular mural cells. PWTs include hemangiopericytoma (HPC), myopericytomas, angioleiomyomas, angiofibroma, and angiomyofibroblastoma. Diagnosis is characterized histopathologically by specific perivascular patterns and variable expression of specific antigens on immunohistochemistry.1–6 Previous studies classified HPC as a type of STS, but more recent data classified it as specific subtype of PWT.1,2 Limited clinical studies have focused on the prognosis and treatment of HPC. Recommendations for treatment of this tumor type have been extrapolated from research regarding management of STS as a whole.7,8 Despite numerous studies evaluating treatment options and prognostic factors for disease-free interval (DFI) and median survival time (MST) of STS, there is no single tumor variable that has been determined to predict STS behavior.9 A recent study suggested that HPC may be less aggressive than other STS.10 The goals of this study were to evaluate the prognosis of dogs diagnosed with HPC that were treated surgically and to identify clinical and tumor variables associated with outcome in a group of 167 dogs with this disease. We hypothesized that tumor variables including location, grade, and surgical margins for patients with HPC would affect DFI and MST.
Materials and Methods
Patient Selection and Evaluation
Medical records at Red Bank Veterinary Hospital were evaluated retrospectively for canine HPC between 2009 and 2016. Dogs with histologically confirmed, surgically excised HPC were eligible for inclusion in the study. Data was collected through review of the medical records, recheck physical examinations, and telephone conversations with the owners. If full medical records were not available, then cases were excluded. In total, 167 cases of HPC in 167 client-owned dogs fit the inclusion criteria.
Clinical parameters reviewed included signalment, history, results of staging tests, tumor and surgical variables, administration of adjunctive therapy, and response to treatment. Tumor characteristics included size, location, grade, and surgical margins. Tumors were grouped as <3, 3–6, and >6 cm. Tumor locations included the thorax, proximal forelimb, distal forelimb, hindlimb, pelvis/hip, abdomen, and tail/perineum. Surgical margins were recorded as clean, narrow, or dirty. Margins were considered clean if surgical borders were separated from tumor cells by ≥3 mm on histopathology, narrow if surgical borders were separated from tumor cells by <3 mm, and dirty if tumor cells were seen at the surgical margins.
Surgical variables included the type of surgery performed and surgical history. For cases that had a history of previous surgery, the reason for additional surgery was determined to be because of incomplete excision, if no gross tumor was present at the time of surgery but previous surgical margins were narrow or dirty on histopathology, or tumor recurrence if gross disease was present irrespective of previous histopathology results. The type of surgery performed was determined by the extent of gross margins at the time of surgery. Surgical wound margins included radical, wide, and marginal. Radial surgical margins were defined as removal of an entire structure, such as a limb amputation. Surgical margins were wide when the tumor and capsules were not penetrated during excision and marginal when tumor was shelled out at the level of the capsule.
A complete blood count and serum biochemical analysis were performed on all dogs. Additional staging tests for each patient were variable but included fine needle aspiration and thoracic radiographs. Histological diagnosis was available for all cases. All tumors in this paper were determined to be HPCs by board-certified pathologists, but immunohistochemistry was not performed consistently. For the purposes of this paper, the term HPC will be used to describe PWTs.
Adjuvant Therapy and Outcome
Follow-up data regarding use of adjunctive treatment, DFI, patient status, and survival time were obtained through a combination of in-house and referral veterinarian medical records and communication with the owners via telephone. Adjunctive therapies were categorized as chemotherapy, radiation therapy, or nonsteroidal anti-inflammatory drugs (NSAIDs) alone, none, or a combination of two or all of the aforementioned therapies. For dogs receiving adjuvant therapy, the specific protocol was recorded along with response, duration of the response, and rescue protocols. DFI, patient status, and survival time were recorded. DFI was defined as the time interval between initial tumor occurrence and tumor recurrence as noted by the owner. For dogs who were still alive with no tumor recurrence at the completion of the study, DFI was defined as the time from surgery to the end of the study. For dogs who died without tumor recurrence, DFI was defined as the time from surgery to the date of death. Median DFI was defined as the middle DFI for the cases, with 50% of cases having tumor recurrence before this time and 50% of cases having tumor recurrence after this time or no tumor recurrence by the completion of the study. Median DFI was not reached if 50% of cases did not experience recurrence by the completion of the study or before patient death. Survival time was defined as the time interval between initial tumor occurrence and either date of death or the end date of the study. For deceased dogs, the cause of death and whether it was related to the patient’s HPC or another cause were recorded.
Statistical Analysis
Quantitative descriptive data for metric variables are presented as medians, range, and mean ± standard deviation. Associations between categorical variables were tested using Pearson’s χ2 test or Fisher exact test. Curves for survival times, DFI in this case, were generated using the Kaplan-Meier method. Dogs were censored in the analysis if they were alive at the time of statistical analysis, lost to follow-up, or died of causes unrelated to their HPC. Variables examined for predictors of recurrence in the univariate analysis included sex, age, tumor size, tumor location, tumor grade and surgical margins of the excised tumor, adjunctive therapy, and patient status.
The log-rank test was used to determine if there was a statistical difference between survival functions for different groups within a variable, such as male and female within sex. A value of P < .05 was considered significant for differences in two such groups. For variables with more than two groups, once a significant difference was found between the groups, pairwise comparisons were performed to determine which of the groups were different, using a Bonferroni correction by lowering the P value for significance to .05/n, where n was the number of groups. Variables considered clinically relevant to the study were further evaluated by multivariate analysis. Multivariate analysis was performed with Cox proportional hazards regression using the guidelines in Hosmer, Lemeshow, and Maya for assessing confounding and interactions using elimination procedures, and for assessing the proportional hazards assumption. Beginning with all variables with univariate P values of .3 or less in the regression model, the P values of the Wald statistic of each variable, as well as the P value of the partial likelihood ratio test for the overall model, were used to eliminate variables from the model one at a time and then to determine whether the removed variable was a confounder. If the variable was a confounder, it would be added back into the model. For multivariate analysis, a value of P = .1 was used to remove a variable from the model. If the Cox regression model would not converge using interaction terms, interactions were assessed using stratified Kaplan-Meier analysis. All analyses were performed using SPSS statistical softwarea.
Results
There were 167 cases of surgically treated HPC in 167 client-owned dogs. Of the 167 dogs, 77 (46.1%) were female and 90 (53.9%) were male, with a median age of 10 yr (range: 3–17 yr). The median weight was 29.7 kg (range: 2.4–81.7 kg), with the majority of dogs between 10 and 40 kg (n = 112; 64.7%). Median tumor size was 5 cm in diameter (range, 0.5–25 cm).
DFI (1–4 yr) for significant univariate variables are reported in Table 1. The distal forelimb was the most common tumor location (n = 46 [27.5%]), followed by the pelvis/hip (n = 32 [18.5%]), thorax (n = 32 [19.2%]), hindlimb (n = 18 [10.8%]), proximal forelimb (n = 16 [9.6%]), abdomen (n = 12 [7.2%]), head/neck (n = 7 [4.2%]), and tail/perineum (n = 7 [4.2%]). Most tumors were low grade (n = 99 [59.3%]) followed by intermediate (n = 52 [31.1%]) and high (n = 12 [7.2%]). Tumor margins were clean in 42.5% (71/167), narrow in 43.7% (73/167), and dirty in 13.8% (23/167).
Adjunctive therapy was provided for 26 tumors. Sole (adjunctive) treatment included radiation therapy (n = 14), chemotherapy (n = 7), and NSAIDs (n = 1). Dual (adjunctive) therapy of chemotherapy and NSAIDs was administered for three tumors and radiation therapy and NSAIDs for one. At the end of the study, which was 350 days after the last HPC surgery reported here, 108 dogs were deceased, with 9 having been euthanized or having died from tumor-related causes. None of these dogs developed metastasis from their HPC.
Overall tumor recurrence was 13.8% (23/167). Nineteen of the 167 tumors underwent more than one surgery as a result of incomplete excision (n = 7) or recurrence (n = 12). Four of the tumors that recurred did not undergo additional surgery. There was significant difference in survival functions between dogs who had more than one surgery (median DFI = 541 days) and those who did not (median DFI > 2500 days; P < .0001; Figure 1).
Citation: Journal of the American Animal Hospital Association 55, 4; 10.5326/JAAHA-MS-6807
In univariate analysis, there was a significant difference in DFI between type of surgery performed (P = .014). Tumors that were marginally removed (3/9 [33.3%]) had a shorter time to recurrence (median DFI = 615 days) compared with those that had a wide resection (20/158 [12.7%]; DFI > 2500 days). All marginally resected tumors had dirty margins (9/9 [100%]). Of the 158 tumors removed with wide resection, 71 [44.9%] had clean margins, 73 [46.2%] had narrow margins, and 14 [8.9%] had dirty margins. Recurrence occurred in 2/71 [2.8%], 13/73 [17.8%], and 5/14 [35.7%], respectively.
Univariate analysis revealed that tumor location was associated with recurrence (P = .013), with the tail/perineum being the most common location for recurrence (4/7 [57.1%]) and the hindlimb being the location with the highest number of recurrences (5/18 [27.8%]; Table 2). There was a significant difference in median DFI across locations (P = .001). Tail/perineum tumors had the shortest median DFI (505 days), and this was significantly shorter than median DFI at either the thorax (P = .001), where no median DFI was reached, or the proximal forelimb, where median DFI was 2009 days (P < .001; Figure 2).
Citation: Journal of the American Animal Hospital Association 55, 4; 10.5326/JAAHA-MS-6807
Tumor grade alone was not associated with recurrence (P = .069) in univariate analysis, and there was no significant difference in DFI across the three tumor grade categories (P = .082). Margins were associated with recurrence (P < .001), and univariate analysis showed that there was a significant difference in DFI across the three categories of tumor margin (P = .082). Tumors with dirty margins recurred most commonly, with a 34.8% (8/23) recurrence rate and a median DFI of 792 days, compared with tumors with clean margins, which had a 4.1% recurrence rate (3/74) and a median DFI of 2009 days (Figure 3). Tumors removed with narrow margins had an 18.4% recurrence rate (13/73) and a median DFI of >2500 days.
Citation: Journal of the American Animal Hospital Association 55, 4; 10.5326/JAAHA-MS-6807
Multivariate analysis using Cox regression showed that grade was not a confounding variable for margin. The possible interaction between margin and grade was not able to be tested using this method as a result of nonconvergence of the model with the interaction term present. However, multivariate analysis using Kaplan-Meier stratified pairwise analysis (with Bonferroni corrected alpha) showed that low-grade tumors with clean or narrow margins (44/99 [44.4 %], median DFI = 2009 days; 46/99 [46.4%], median DFI > 2500 days, respectively) had a significantly longer time to recurrence than low-grade tumors with dirty margins (median DFI = 792 days, 9/99 [9.1%]; P < .001 [clean versus dirty], P = .002 [narrow versus dirty], respectively). Intermediate-grade tumors with clean margins (21/52 [40.3%], median DFI > 2500 days) had a significantly longer time to recurrence than those with both narrow margins (21/52 [40.3%], median DFI = 1020 days, P = .001) and dirty margins (10/52 [19.2%], median DFI > 2500, P = .004). High-grade tumors with clean margins (3/12 [25%], median DFI > 2500 days) had a longer time to recurrence than those with both narrow (5/12 [41.7%], median DFI not reached) and dirty margins (4/12 [33.3%], median DFI = 505 days). However, these were not found to be significant (P = .246 [clean versus narrow], P = .157 [clean versus dirty], respectively). There was also no significant difference in DFI between high-grade tumors with narrow and dirty margins (P = .894).
Adjunctive therapy was associated with recurrence (P = .045) and with type of margins (P = .001) but not with tumor grade (P = .237). Survival curves were significantly different between dogs who had any form of therapy (median DFI > 1500 days) and those who did not (DFI > 2500 days; P = .030). Multivariate analysis using Cox regression showed no confounding between tumor grade, margin type, and adjunctive therapy. Interactions could not be assessed with Cox regression as a result of nonconvergence of the regression model when interaction terms were included. However, stratified Kaplan-Meier pairwise analysis (with Bonferroni corrected alpha) resulted in the following. When stratified by margin type, multivariate analysis showed that there were no significant differences in DFI between dogs receiving adjunctive therapy and those who did not in any of the margin types (P = .804 [clean], P = .522 [narrow], P = .760 [dirty]). When stratified by adjunctive therapy (yes/no), significant differences in DFI were found between some of the different margin types for dogs not undergoing adjunctive therapy (clean margin median DFI = 2009 versus narrow margin median DFI = 2500, P = .005, and clean margin median DFI versus dirty margin median DFI = 792, P < .001). For dogs undergoing therapy, there were no significant differences in DFI in any pairwise comparisons between margin types.
In univariate analysis, there were not any statistically significant differences in survival functions within the following variable groups: sex (P = .712), dog size (P = .120), tumor size (P = .490), and whether gross tumor was present if a second or third surgery was performed (P = .537).
Discussion
In this study population, dogs with HPC were most commonly medium- to large-breed middle-aged to older dogs. This is consistent with signalment information reported in other studies.4,10–14 The distal forelimb was the most common location in this population (46/167 [27.5%]). This is a lower reported frequency than seen in previous studies (63–81%).4,11–16 This discrepancy is likely a result of study sample population.
Although the distal forelimb was the most common location for tumor development, it was not the location that experienced the fastest recurrence rate. Dogs with tumors located at the tail/perineum experienced recurrence at a much faster rate than at the proximal forelimb or at the thorax, with a median DFI of ∼16 mo (505 days). This is likely because of the difficulty in obtaining clean excision in this area. However, results may have been attributable to the small sample size for tail/perineum (7/167) and the fact that all but one case had either dirty or narrow surgical margins. It is also possible that tumors in this location behave more aggressively than those located on other parts of the body, especially because two of seven and four of seven tumors were intermediate and high grade, respectively. Additional studies are needed to further investigate the biological behavior of these tumors based on location of growth.
Tumor depth has been shown to be prognostic for DFI, with tumors invading muscle having a much shorter DFI compared with tumors confined to the cutis and subcutis.2 Although these specific tumor details were not present in the medical records for this study, the minimal subcutaneous tissue in the tail/perineum allows for more potential for invasion into deeper tissue. In addition, tumor size was previously described as being prognostic; this was not shown to be significant in the overall sample population.2,17
In univariate analysis, tumors with clean margins had a significantly longer time to recurrence than tumors with narrow or dirty margins. This finding is consistent with previously reported results.17 However, when analyzing tumor grade and margins together, low-grade tumors with clean or narrow margins had a significantly longer time to recurrence than tumors with dirty margins. This result may be attributable to the fact that low-grade tumors behave less aggressively than intermediate- or high-grade tumors because of their histologically lower criteria of malignancy. Thus, regrowth may be absent or slower for low-grade tumors with clean or narrow margins. However, this was not found for intermediate-grade tumors, of which only those with clean margins had a significantly longer time to recurrence when compared with those with either narrow or dirty margins. This observed discrepancy when performing multivariate analysis is likely multifactorial. Intermediate tumors tend to behave more aggressively than low-grade tumors, thus increasing the likelihood of tumor recurrence when margins were not clean. Tumors with narrow margins can have normal-looking cells at the tissue periphery, but it is possible that because the excised tissue submitted for histopathology is large, not every three-dimensional section can be accurately examined. This is unlike tumors with dirty margins, in which malignant cells are left behind to multiply, leading to bulky disease reoccurrence.
Results in this study were largely consistent with previous results. Prior studies evaluating STS showed that clean surgical margins were an important determinant for local recurrence, regardless of tumor grade.18,19 When clean surgical margins were not met, both tumor grade and mitotic index were prognostic for tumor recurrence.18 When specifically evaluating HPC, grade was determined to affect the probability for metastasis and was considered to be a poor variable in predicting local tumor recurrence.2 In addition, the results here are consistent with previous findings that HPC has a low chance of metastasis.3,11,15,16 Of the 9/167 dogs who died from tumor-related causes, none died as a result of tumor spread. Although the study variables were not assessed as potential predictors of metastasis, this again highlights that HPC may be a less aggressive type of STS.10
However, contrary to the findings in previous studies, tumor size was not an important prognostic variable.2,17 Surgical margins alone were associated with recurrence, whereas this was not the case for tumor grade. When tumor grade and margins were analyzed together, significant results were found. A possible explanation for this discrepancy may be that previous reports focus on univariate analysis, and not multivariant analysis, to help determine prognostic factors for tumor DFI and MST. This underscores how no single factor is predictive of tumor recurrence.
This was a retrospective study and thus had inherent flaws including nonuniform data and data collection based on phone interviews with owners. Adjunctive therapy was not consistent and was dependent on the treating clinician and owners’ willingness to pursue treatment. The majority of cases did not have adjunctive therapy, resulting in a small sample size. Data was also limited on intermediate-grade tumors because of a small number of cases with tumor recurrence. Thus, these results should be interpreted judiciously. Further studies examining the use of adjunctive therapies with a large sample size are indicated to determine whether additional treatment is beneficial to prolong tumor recurrence.
Conclusion
Several prognostic factors were identified for dogs undergoing surgery for HPC. History of previous surgery, type of surgery performed, status of surgical margins, tumor location, and whether adjunctive therapy was performed were associated with tumor recurrence. Local recurrence was higher in locations where clean surgical margins were more challenging to achieve, such as the tail/perineum. This should be considered when assessing the treatment plan and follow-up recommendations for any tumor located in these areas. Aggressive initial surgical treatment is also recommended when possible to reduce the chance of local tumor recurrence. Further study will be needed to investigate the utility of adjunctive therapy postoperatively.
Kaplan-Meier curve depicting disease-free interval for dogs with and without previous surgery for hemangiopericytoma. Dogs who did not have a previous surgery experienced a significantly longer disease-free interval.
Kaplan-Meier curve depicting disease-free interval for hemangiopericytoma based on location.
Kaplan-Meier curve depicting disease-free interval for tumor margins irrespective of tumor grade.
Contributor Notes
A. Kravitz’s present affiliation is Surgery Departments, Gold Coast Center for Veterinary Care, Huntington, New York.
