Efficacy and toxicity

Systemic treatments, including chemotherapy and TKI, are options for dogs with non-resectable cMCTs when RT is not accessible, especially for those with tumors showing high metastatic-risk. For chemotherapy, VBL is commonly used as the first-line agent alone or combined with other chemotherapeutic agents, such as CTX [8]. For TKI, imatinib has not been thoroughly investigated as TOC or MAS, and its clinical application is only reported in some retrospective case studies with a small population [18-20, 43-45]. Our study retrospectively compared the efficacy and safety of VBL-based chemotherapy and imatinib in dogs with macroscopic high-risk cMCTs, which has not been reported at the time this study was ended.

The population of dogs in our study was similar to those of previous studies.

However, the retrospective nature of our study indeed resulted in some bias. Firstly, when comparing patient characteristics, body weight was significantly higher in the VBL group, and it was most likely a bias caused by the fact that the cost of imatinib was much higher than chemotherapeutic agents, making it unaffordable for some clients, especially for those with large-breed dogs. Despite that, in statistical analysis, the distribution of body weight seemed not to influence the clinical outcome. Secondly, there were more dogs classified as substage b in the VP subgroup compared to the VCP subgroup (80.0% versus 27.3%, P = 0.03), and it was caused by the different preferences between clinicians choosing the treatment regimen. The presence of systemic illness might cause the patients to be more susceptible to the treatment toxicity; however, the incidence of side effects was not significantly different between the two subgroups.

versus 35.0%, P = 0.606), but the CBR was significantly higher in VBL group (100.0%

versus 80.0%, P = 0.048). The difference in mechanism of action could explain it.

Imatinib inhibited the tumor growth by targeting mutated KIT proteins, so tumors that did not show clinical benefit was assumed to possess wild-type c-kit gene. For VBL, it acted by interfering with the mitosis of highly proliferative tumor cells, so at least SD could be expected initially in most cases.

There was no significant difference in PFI between two groups (83 days for the VBL group and 51 days for the imatinib group, P = 0.885). When comparing dogs that achieved an objective response, the PFI was longer for those in the imatinib group. Still, statistical significance was not reached (not reached versus 55 days for the VBL group, P = 0.080), and it might be due to the small case number of responders. However, the retrospective nature of our study could also result in bias that influences PFI. As there was no standard schedule for regular follow-up, disease progression would be recorded earlier for patients with shorter recheck interval and recorded latter for patients with longer recheck interval, resulted in longer PFI in the latter cases. A further prospective study with a larger population is required to confirm our findings.

One retrospective study reported an ORR of 63.6% for dogs with macroscopic, Patnaik grade II to III cMCTs treated with VCP protocol, which was higher than that of the VCP subgroup in our study (54.5%). The median PFI was 74 days in that study [8].

Two studies using VP protocol for dogs with similar disease status reported ORR of 43%

and 47%, which was also higher than that of the VP subgroup in our study (30.0%). The median PFI was 78 and 154 days [7, 24]. The distribution of histologic grade might be a reason for the difference in clinical outcome between previous studies and our study;

however, it could not be confirmed as the histologic grade was only reported in 7 cases in VBL group in our study.

In previous studies, the overall ORR for dogs with similar disease status treated with imatinib was 47.6% to 55.3%, with ORR up to 100.0% for tumors possessing mutated c-kit and 22.7% to 31.3% for tumors possessing wild-type c-c-kit; however, the median PFI

was not reported [18, 19]. In the present study, ORR seemed to be lower (35.0%).

However, the distribution of the mutation status of the c-kit gene in our population was not evaluated, so it remained unknown if it was the factor influencing our outcome or not.

Besides, imatinib was administered after the failure of chemotherapy and steroid treatment in about half of the cases. It is reported that steroids and some chemotherapeutic agents, such as VBL and chlorambucil, are P-glycoprotein inducers [46], and imatinib is a substrate to P-glycoprotein [47, 48]. Hence, the overproduction of P-glycoprotein caused by previous treatments might have resulted in resistance to imatinib, causing an inferior outcome. This phenomenon has been reported in MAS, which is also a substrate to P-glycoprotein [22].

Corticosteroids are widely used in cases of canine MCT due to their effect of decreasing the growth and reproductive rates of tumor cells [49], reducing the local inflammation caused by the tumor, and also improving the patients’ activity and appetite.

An early study reported an ORR of 20% for cutaneous and subcutaneous canine MCTs treated with oral prednisone [49]. However, the administration of corticosteroids is thought to be mildly beneficial and lack durable efficacy in other later studies, especially in the cases of high-grade MCTs [8, 50]. Most of the patients in our study were given prednisolone concurrently, but all of them had experienced PD when taking prednisolone before starting VBL or imatinib treatment, so the remission of the tumor was less likely to be caused by prednisolone. In contrast, imatinib and VBL are both substrates to P-glycoprotein [48], so the ORR and PFI in both groups might be affected negatively by the

prednisolone.

There was a significant difference in overall toxicity between the two groups. The proportion of dogs experienced side effects during treatment was twice higher in VBL group than in imatinib group (71.4% versus 35.0%, P = 0.019), and it could be a reason to advocate for the use of imatinib in patients expected to be more susceptible to adverse events and considering about quality of life. It was predictable that significantly more dogs were affected by neutropenia in the VBL group (23.8% versus 0.0%, P = 0.048) as bone marrow suppression was the dose-limiting toxicity of VBL, and was rarely reported for imatinib [18]. Lethargy was only reported in the VBL group (28.6% versus 0.0%, P = 0.021); however, as the interpretation of lethargy could be subjective, the inter-observer bias could exist. It should be noticed that although the effort was made to verify the side effects attributed to the treatment, the paraneoplastic syndromes caused by MCT, which also elicit lethargy, anorexia, and GI signs, still could not be excluded entirely.

Furthermore, the influence of concurrent prednisolone administration on the elevation of ALT in both groups could neither be excluded. Overall, both treatments were well tolerated as the toxicities were mild and self-limiting; in most instances, only two dogs required dose reduction due to side effects in the VBL group in our study.

The toxicity of VBL-based chemotherapy was reported in about 10% to 20% patients in two retrospective studies [7, 8]. Only the abnormality of complete blood count was evaluated, the result of serum biochemistry was not reported. Besides, the substage and tumor burden was not reported in both studies, which might also affect the susceptibility to adverse events. In a recent prospective clinical trial, the incidence of side effects was 89% [24], which was higher than that in our study. However, the starting dose of VBL was 2.5 mg/m² for all patients in that study and was 2 mg/m² for most patients in our study.

Side effects, including neutropenia, GI signs, and elevation of serum liver enzyme, are rarely reported in the clinical studies of imatinib with or without concurrent administration of steroids [18-20, 45]. In contrast, side effects were detected in 35% of patients in the present study. It could be explained by the fact that 45% of our patients were classified as substage b, causing them to be more susceptible to the toxicity elicited by the treatment.

VCP and VP protocols are commonly used in canine MCT, but there has been no research comparing the efficacy and toxicity of both protocols. In our study, the tendency to a better outcome and a higher incidence of side effects in the VCP subgroup were observed, but the differences were not significant. Although the median dose interval of VBL was longer in the VCP subgroup (3 weeks) than in the VP subgroup (2 weeks) when the administration of CTX was taken into account, the median dose interval of chemotherapeutic agents would be 1.5 weeks in VCP subgroup. The relative dose intensity of the combination would be 1.3 versus 1 for a single agent. As there were only 11 and 10 cases in each subgroup, a prospective study with a larger population is warranted to confirm this finding.

It is worth noting that TKI is classified as a cytostatic agent rather than a cytotoxic agent, which most often inhibit tumor growth and prevent metastasis, but not necessarily expected to shrink tumors. If WHO or RECIST criteria were used, such clinical benefit would be ignored. Consequently, the best way for response evaluation and the expected treatment endpoint for TKI may be different from those applied for chemotherapy, so other criteria should be incorporated, such as the saturation and modulation of the target or the alteration of the target-mediated pathway. Moreover, target therapies usually have maximal target inhibition at non-toxic doses, which is known as the optimal biological

imatinib has not been identified currently, and it is possibly lower than the dosage commonly used, which is 10 mg/kg/day. Durable CR was observed when imatinib was used at a dosage of 4.4 mg/kg/day for two dogs with measurable MCT and bone marrow involvement in a case study [20]. Hence, further research is required for the practical response evaluation of TKI, and also for finding out the OBD of imatinib.

5.2 Prognostic factors

The clinical appearance and biological behavior of canine cMCT are variable, and many factors are affecting the clinical outcome. Several negative prognostic factors have been reported in canine cMCTs, for example, histologic grade, clinical stage, local recurrence, and systemic signs [1]. However, in our study, the age and clinical stage at diagnosis were the only two factors that significantly influence the time to progression.

Dogs older than 11 years-old had significantly shorter median PFI than those younger than 11 years old (189 days versus 51 days; HR, 3.710; 95% CI, 1.110-12.408;

P = 0.048 and P = 0.033 for univariate and multivariate analysis, respectively). A similar

finding was only reported in a study of canine MCT treated with RT [52] and was thought to be associated with aging changes.

Dogs classified as stage 2 had significantly longer median PFI compared to stage 4 (not reached versus 35 days; HR, 0.223; 95% CI, 0.050-0.996; P = 0.005 and P = 0.049 for univariate analysis and multivariate analysis, respectively). This result was expectable as stage 4 might indicate a larger tumor burden or more aggressive tumor behavior compared to stage 2, and was also reported by other studies [1, 39]. A significant difference in PFI was not observed when comparing stage 1 to stage 4 in both univariate analysis and multivariate analysis, and it was likely due to the small number of patients

classified as stage 1. The median PFI between dogs classified as stage 3 and stage 4 was only significantly different in univariate analysis (189 days versus 35 days, P = 0.017), but not in multivariate analysis (P = 0.187). However, it should be noticed that abdominal ultrasound was not performed in 6 dogs classified as stage 3, so their clinical-stage might be underestimated.

In a retrospective study of canine MCT treated with MAS, the initial response to treatment was the only reliable prognostic factor [23]. Although the median PFI was longer for responders compared to non-responders in univariate analysis (160 days versus 40 days, P = 0.025) in the present study, the difference was not significant upon multivariate analysis (P = 0.231). The possible explanation was the small population of our study; hence, a prospective study with a larger population would be warranted to verify this finding.

Substage b and more significant target lesion, which were reported to be negative prognostic factors in previous studies, seemed to be associated with shorter median PFI in our study; however, a significant difference was not reached upon univariate analysis (P = 0.696 and P = 0.327, respectively). Local recurrence was also reported to be a negative prognostic factor; however, it was also not associated with the median PFI in our study. It might be due to the small population of our study or the selection bias that only patients with high metastatic-risk MCT were enrolled.

It was interesting that dogs received a higher median dosage of prednisolone during treatment had a trend toward shorter PFI, although the difference was not significant upon univariate analysis (P = 0.059). As mentioned, both VBL and imatinib were substrates to P-glycoprotein [48], so one possible explanation was that the administration of prednisolone might cause overproduction of P-glycoprotein, resulting in an inferior

sometimes concurrently prescribed for patients with more aggressive MCT that did not respond well to the chemotherapy or TKI treatment alone. Therefore, a prospective clinical trial would be required to figure out the influence of the administration of prednisolone during the treatment of canine cMCT.

Among the prognostic factors mentioned in previous studies, the histologic grade is the most predictive factor [1]. However, in our study, complete histological information was not available in more than half of the patients as most of their tumors were not amenable for surgery, thus, the influence of histologic grade was unable to investigate.