Management of canine insulinomas with toceranib phosphate: 30 cases (2009-2019).

BACKGROUND: Insulinomas are the most common tumour of the endocrine pancreas in dogs. These are malignant tumours with a high metastatic rate and limited ecacious chemotherapeutic options available. Recent literature supports the use of the multi-receptor tyrosine kinase inhibitor sunitinib malate for treatment of metastatic insulinoma in people. Toceranib phosphate is a veterinary targeted therapy that may provide benet in treatment of canine insulinomas. The primary objectives of this study were to describe the duration of clinical benet, dened as absence of clinical signs associated with hypoglycemia and/or measurable response according to the response evaluation criteria for solid tumours in dogs (cRECIST), and measures of outcome, namely progression free interval (PFI) and overall survival time (OST) in dogs diagnosed with insulinoma treated with toceranib. A secondary objective was to describe the adverse effects of toceranib in dogs with insulinoma. RESULTS: A medical record search identied 30 dogs diagnosed with insulinoma and treated with toceranib at ve university hospitals and eight veterinary specialty referral hospitals. A majority (66.7%) of dogs with measurable disease experienced either complete response (CR), partial response (PR), or stable disease (SD) with toceranib therapy. The overall median progression free interval (PFI) was 561 days (95% condence interval: [246, 727 days]). The overall median survival time (OST) was 656 days [310, 1045 days]. Larger dogs were at increased risk for disease progression (P = 0.0310) and death (P = 0.0064), with every 1 kg increase in body weight resulting in hazard ratios (HRs) of 1.045 [1.003, 1.084] and 1.05 [1.012, 1.090], respectively. In addition, time to disease progression was associated with use of therapies prior to toceranib (P = 0.0050) and type of veterinary practice (P = 0.0025). The most common adverse events with toceranib therapy were grade 1 or 2 gastrointestinal toxicities. Clinical benet was reported in the majority of dogs diagnosed with insulinoma treated with toceranib, but randomized, prospective studies are needed to assess and quantify the effect of this therapy. The most commonly observed adverse events (AEs) were gastrointestinal AEs.


Background
Insulinomas are insulin-secreting tumours which arise from pancreatic β islet cells. Insulinomas are the most common tumour of the endocrine pancreas in dogs, yet are relatively rare [1][2][3]. The majority of canine insulinomas are highly malignant tumours. Macroscopic metastatic lesions are present in 45 -50% of dogs at the time of diagnosis, and it is clinically anticipated that the majority of dogs will develop disease recurrence or metastasis despite attempts at local disease control with surgery [1][2][3][4].
The most debilitating clinical signs observed with insulinoma are the result of neuroglycopenia, and include weakness, ataxia, disorientation, behavior changes, and seizures [1,5]. Initial therapy is aimed at managing these clinical signs, while intermediate to longer-term therapy aims to treat the primary tumour and/or metastasis [1][2][3].
Consequently, there is a need to investigate alternative, e cacious, and well-tolerated medical interventions to manage insulinomas in dogs.
Sunitinib malate (Sutent ®; P zer, Inc., New York, NY, USA) is an oral small molecule inhibitor initially approved by the Food and Drug Administration (FDA) for the treatment of renal cell carcinoma and imatinib-resistant gastrointestinal stromal tumour in people, but additional applications have been subsequently identi ed [13,14]. Based on the results of a multinational, randomized, double-blind, placebo-controlled phase 3 clinical trial, sunitinib was recently approved by the FDA for treatment of locally advanced or metastatic pancreatic neuroendocrine tumours (pNETs) in people, a classi cation encompassing all tumours arising from the multipotent stem cells of pancreatic ductal epithelium, including insulinoma [15]. Toceranib phosphate (Palladia®; Zoetis Animal Heath, Madison, NJ, USA) is a small molecule inhibitor with similar molecular targets to sunitinib, including the cell surface receptors for vascular endothelial growth factor (VEGF)-1 and -2, platelet-derived growth factor (PDGF)-β, and stem cell factor (SCF) [16,17].
The primary objectives of this study were to describe the duration of clinical bene t, de ned as absence of clinical signs associated with hypoglycemia and/or measurable response according to the response evaluation criteria for solid tumours in dogs (cRECIST), and measures of outcome, namely progression free interval (PFI) and overall survival time (OST) in dogs diagnosed with insulinoma treated with toceranib [29]. A secondary objective was to describe the adverse effects of toceranib in dogs with insulinoma.

Page 4/22
The medical record search, and application of inclusion and exclusion criteria, identi ed 30 dogs diagnosed with insulinoma and treated with toceranib at ve university hospitals and eight veterinary specialty referral hospitals between June 2009 and March 2019. The study population included eight mixed breed dogs, two Boston terriers, two Chihuahuas, two Labrador retrievers, and one of each of the following breeds: Afghan hound, Australian shepherd, chowchow, cocker spaniel, coonhound, dachshund, Doberman pinscher, Irish setter, Jack Russel terrier, papillion, Pekingese, Pomeranian, Scottish terrier, Shar Pei, West Highland white terrier, and Yorkshire terrier. There were 14 male neutered dogs and 16 female spayed dogs. The median age at diagnosis was 9 years (range, 5 -15 years). The median weight at diagnosis was 14.9 kg (range, 2.9 -44.4 kg).

Presentation, Diagnosis, and Staging
The majority of the dogs (n = 25/30) initially presented with clinical signs associated with neuroglycopenia, including seizures (n = 13), collapse (n = 7), ataxia (n = 5) muscle tremors (n = 3), and twitching (n = 3). Two dogs presented with lethargy and two dogs presented with vomiting at diagnosis. Four dogs were reported to be asymptomatic for insulinoma at the time of diagnosis. One of these dogs was presented for evaluation of a soft tissue sarcoma, and the other three were presented for routine annual examination. Each of these dogs were further evaluated for possible insulinoma when hypoglycemia was noted on blood work.
Twenty-one dogs underwent partial pancreatectomy with histopathology of the pancreas, con rming the diagnosis of insulinoma. Four dogs had an ultrasound-guided ne needle aspirate biopsy (FNA) with cytology of a pancreatic mass to con rm diagnosis. All of the dogs with either a histologic or cytologic diagnosis also had hypoglycemia reported at the time of diagnosis, additionally, 23 of these dogs had a documented abnormal paired fasting insulin and glucose ratio. Two additional dogs were diagnosed with an abnormal paired fasting insulin and glucose ratio and a pancreatic nodule on CT. Another two dogs were diagnosed with an abnormal paired fasting insulin and glucose ratio and a pancreatic nodule on abdominal ultrasound. The nal dog was diagnosed with clinical signs consistent with neuroglycopenia including seizures, weakness, and collapse, and an abnormal paired fasting insulin and glucose ratio, despite no signi cant abnormalities noted on abdominal ultrasound. In all cases, other causes of hypoglycemia, including sepsis, hepatic failure, adrenocortical insu ciency, and toxin ingestion, were reasonably investigated and eliminated prior to treatment with toceranib.
All 30 dogs had a reported low fasting blood glucose concentration (de ned as <3.33 mmol/L) at the time of diagnosis, and 28 dogs were reported to have hypoglycemia in the presence of a concurrent normal or increased fasting insulin measurement, as was reported by each individual reference range.
The remaining two dogs did not have baseline insulin levels measured. The median fasting blood glucose concentration was 2.16 mmol/L (range, 1.89 -3.27 mmol/L), and the median fasting insulin concentration was 101.8 µIU/mL(range, 14.6 -647.3 µIU/mL).
Records of baseline complete blood counts were available for 25 dogs, and serum chemistry pro les were available for 29 dogs at the time of toceranib initiation. The most commonly reported chemistry abnormality was hypoglycemia (n = 25). Concurrent urinalysis was recorded for 14 dogs. Two dogs had hyposthenuria and one dog had isosthenuria. Six dogs had a baseline blood pressure performed, of which three were hypertensive (systolic blood pressure (BP) >140 mmHg).
Staging tests at the time of toceranib initiation included thoracic radiographs (n = 21), abdominal ultrasound (n = 22), and/or computed tomography (CT) scan of the thorax and abdomen (n = 12). The ndings of these imaging tests are summarized in Table 1.  [30]. There were eight dogs with at least stage I , 13 with at least stage II, and nine with at least stage III disease. Locations of metastasis con rmed by either FNA or tissue biopsy included pancreatic lymph nodes (n = 13), liver (n = 9), mesenteric lymph nodes (n = 4), hepatic lymph nodes (n = 1), and spleen (n = 1).

Treatment
Twenty-one dogs underwent a partial pancreatectomy for treatment of insulinoma prior to initiating toceranib. One dog had a second surgery performed for progressive disease two years after an initial partial pancreatectomy, and then commenced toceranib. One dog also underwent regional lymph node extirpation at the time of partial pancreatectomy. Reasons for toceranib initiation for the dogs that had a prior partial pancreatectomy included regional lymph node metastasis at diagnosis (n = 8), hepatic metastasis +/-lymph node metastasis at diagnosis (n = 7), inability to excise pancreatic mass (n = 2), recurrence of clinical signs associated with hypoglycemia (n = 3), recurrent pancreatic nodule 1458 days following partial pancreatectomy (n = 1). The median duration of time between initial surgery and toceranib treatment was 93 days (range, 4 -1458 days).
Four dogs received cytotoxic chemotherapy prior to toceranib initiation. One of these dogs was initiated on toceranib 21 days after it received four doses of doxorubicin (30 mg/m2 every 3 weeks) in the adjuvant setting following partial pancreatectomy, one received adjuvant streptozotocin with prednisone following partial pancreatectomy, one received adjuvant vinorelbine with marbo oxacin, and one received cytarabine for meningoencephalitis approximately one year prior to starting toceranib.
Six dogs received prednisone as a sole therapy for insulinoma prior to toceranib initiation. One dog received glucagon with prednisone prior to toceranib initiation, yet clinical signs persisted and glucagon was discontinued only 4 days following diagnosis. All of the dogs that received prednisone as a sole therapy were initiated on toceranib when recurrence of clinical signs associated with hypoglycemia was noted. The median duration between diagnosis and toceranib treatment for the dogs treated with prednisone alone was 50 days (range, 13 -453 days).

Page 7/22
Three dogs did not receive any therapies for insulinoma prior to commencing toceranib. For all three of these dogs, toceranib was offered as a rst-line therapy as an alternative to surgical excision. The median duration between initial diagnosis of insulinoma to treatment with toceranib for dogs treated with toceranib as a rst-line therapy was 13 days (range, 0 -22 days).
The median initial toceranib dose administered was 2.67 mg/kg by mouth (range, 2.1 -3.27 mg/kg). Twenty-four dogs initially received toceranib on a Monday, Wednesday, Friday (MWF) schedule and six dogs on an every other day (EOD) schedule. Fourteen dogs had a dose reduction instituted during treatment with toceranib. All of these dogs had gastrointestinal AEs. All AEs observed during toceranib treatment are listed by grade of toxicity in Table 2.
*Suspected but not con rmed Five dogs were concurrently administered prednisone at a median dose of 0.5 mg/kg q 24 hr. (range, 0.5 -0.8 mg/kg q 24 hr). One dog received both prednisone and streptozotocin (dose and frequency not speci ed) during treatment with toceranib. Two dogs received both prednisone and diazoxide (dose and frequency not speci ed) during treatment with toceranib.

Response to Toceranib
As a retrospective study, methods of response assessment and intervals between assessments varied between institutions. Response to therapy was de ned as an absence of clinical signs associated with hypoglycemia, or in the case of measurable disease, designated according to the criteria described by cRECIST [29].
cRECIST response was reported for 15 dogs [29]. Response to therapy was determined using the absence of clinical signs associated with hypoglycemia (n = 15), repeated imaging with abdominal ultrasound (n = 13) +/-thoracic radiographs (n = 6), or CT-scan (n = 2). Intervals between repeated imaging were reported to be either monthly (n = 4), bi-monthly (n = 1), or tri-monthly (n = 10). Of the dogs with cRECIST response reported, 66.7% of dogs experienced a measurable response; six (40%) dogs experienced CR, one (6.7%) dog experienced a PR, three (20%) dogs experienced SD, and ve (33.3%) dogs experienced PD. All of the dogs that had either CR, PR, or SD were also reported to be normoglycemic at each response assessment.
Twelve dogs had no new reported clinical signs and repeated blood glucose measurements reported as methods utilized to monitor toceranib response. The median duration of reported normoglycemia was 275 days (range, 12 -727 days). The remaining three dogs did not have repeated imaging or regular blood glucose measurements reported, but were monitored for recurrence of clinical signs associated with hypoglycemia.
Nineteen dogs ultimately discontinued toceranib. Seventeen of these dogs discontinued toceranib when PD was observed. One dog discontinued toceranib after 883 days due to increased liver enzyme values on monitoring bloodwork. One dog discontinued toceranib after 1261 days due to gastrointestinal adverse events. One dog had splenic metastasis con rmed by FNA after 288 days of toceranib therapy, yet continued to receive toceranib despite PD. The median duration of toceranib therapy for all dogs was 281 days (range, 10 -727).
One dog that developed progressive disease on toceranib was treated with masitinib following discontinuation of toceranib, but clinical signs worsened acutely, and masitinib was discontinued. Following discontinuation of masitnib, metronomic cyclophosphamide (15 m/m2 every 24 hours) was initiated for 6 days before the dog was euthanized for progressive clinical signs.
Time-to-event outcomes: Kaplan-Meier survival curves Disease progression during toceranib therapy was ultimately observed in 20 dogs (66.6%). In absence of documentation of PD, disease progression was considered not observed for 10 dogs, up until the last veterinary visit, and PFI was right-censored for these cases. The Kaplan-Meier median overall PFI from time of toceranib initiation was 561 days (95% con dence interval ( Time-to event outcomes: Proportional Hazards modeling Proportional hazards modeling was implemented on each of the two right-censored variables (i.e. PFI and OST). Potential explanatory non-time-dependent covariates (e.g. sex, age, weight, treatment facility, tumour stage, therapies prior to toceranib (e.g. partial pancreatectomy, chemotherapy, and/or prednisone), dosing frequency, and toceranib dose) and all 2-way interactions were considered for inclusion in the models.
For overall PFI, there was evidence of a signi cant association between time to progression and therapies prior to toceranib (P = 0.0050), type of veterinary practice (P = 0.0025), and weight (P = 0.0301). In evaluation of OST, the only relevant explanatory variable was weight at diagnosis, whereby for every 1 kg increase in body weight at diagnosis, there was an increased hazard of death by an estimated multiplier of 1.050 [1.012, 1.090]. There was no evidence of any association between OST and any of the other proposed explanatory covariates at a 5% level of signi cance (P > 0.18).

Discussion
The objectives of this retrospective study were to describe the duration of clinical bene t, de ned as absence of clinical signs associated with hypoglycemia and/or measurable response according to cRECIST, and measures of outcome in dogs diagnosed with insulinoma treated with toceranib [29]. The initial clinical signs attributed to hypoglycemia noted at diagnosis were reported to be improved with the addition of toceranib therapy in all cases. The majority (66.7%) of dogs with measurable disease at the time of toceranib initiation were not reported to have recurrent clinical signs, and experienced either a CR, PR, or SD. For all 30 dogs, there was an overall median PFI and OST of 561 days and 656 days, respectively. Overall, toceranib was reported to be well-tolerated; the majority of adverse events reported were grade 1 or 2 gastrointestinal toxicities.
Prior studies that have evaluated OST for dogs treated for insulinoma have found that young age, stage II or III disease, and medical therapy were poor prognostic factors for survival, none of which were found to be signi cantly associated with either PFI or OST in our study [4,7,31]. These inconsistences are most likely a result of variation in the study population and small numbers, precluding robust statistical analyses. In the present study, the retrospective application of the WHO TNM stage based upon the available case information may have underestimated some case stages. To better elucidate prognostic factors and the in uence that disease stage has on outcome, a standardized approach to diagnosis and case management would be ideal.
In the present study, dogs that had therapy for insulinoma prior to toceranib had a greater hazard for disease progression than dogs that received toceranib as a rst-line therapy. This could be a result of case selection bias. The majority of the dogs that had prior therapies had residual disease or metastatic disease at the time of toceranib initiation, and dogs that had prior therapies may have also had disease that was inherently or had eventually become resistant to therapy.
Additionally, dogs in our study that were treated at academic institutions had an increased hazard for disease progression, relative to those treated at a private practice. Previous retrospective studies describing therapy for canine insulinoma were performed primarily at academic hospitals [4,7,10,31]. Our study's nding could be explained by biases not speci cally evaluated in this study, such as potentially more comprehensive follow-up documentation at academic institutions, or perhaps the enrollment of dogs with more advanced disease. The clinical relevance of this nding is unknown, and comparisons of temporally and demographically disparate and distinct study populations could be misleading. Prospective, randomized or controlled cohort studies would be required to compare these treatment settings and evaluate the clinical implications.
Finally, in the present study, there was also an increased hazard of both disease progression and death for every 1 kg increase in body weight at diagnosis. Canine insulinomas have commonly been reported in medium and large-breed dogs, yet body weight has not been previously described to be associated with  [32,34].
Additional AEs reported in this study included a combination of grades 1 -3 constitutional, hematologic, and biochemical AEs. These AEs did not result in overt clinical signs, nor necessitate dosing modi cations. Therefore, when comparing medical treatment options for insulinoma, the AE pro le of toceranib is seemingly more tolerable than that described for alloxan or streptozotocin [10,12]. Meleo et al. described the use of alloxan for treatment of ve canine insulinomas, and reported complications including renal tubular necrosis, acute renal failure, and persistent hyperglycemia [12]. Development of diabetes mellitus was reported in 42.1% of the dogs treated with streptozotocin in a prospective study by Northrup et al., resulting in euthanasia in the majority of dogs (6/8; 75%) experiencing this AE [10].
In humans, there are reports of diabetic and non-diabetic patients experiencing signi cant decreases in blood glucose measurements following treatment of various neoplasms with sunitinib [35]. This phenomenon has also been reported humans treated with other tyrosine kinase inhibitors, including dasatinib, imatinib, and sorafenib [35]. The mechanism by which these tyrosine kinase inhibitors (TKIs) affect blood glucose levels is not well-understood, but one theory hypothesizes that it may be secondary to inhibition of the downstream pathway normally initiated by the binding of stem cell factor (SCF) to its receptor [35]. The gene responsible for encoding the receptor for SCF, c-kit, has been demonstrated to be expressed by insulin-producing pancreatic β-cells in rats [36]. Considering that blood glucose measurements are an integral part of monitoring PFI in patients with insulinoma, use of TKIs in treatment of insulinoma may cause iatrogenic hypoglycemia and impair the ability to monitor treatment response.
The authors are unaware of documented hypoglycemia associated with toceranib administration in dogs, but it is important to be aware of this potential issue when selecting toceranib for the management of insulinoma.
Further support for the utilization of toceranib in the management of canine insulinomas could be provided by molecular studies. Human malignant pNET tissues have widespread expression of the receptor tyrosine kinases (RTKs) for platelet-derived growth factor (PDGF)-α/β, and vascular endothelial growth factor (VEGF)-2 and -3, and these tumours are clinically responsive to receptor tyrosine kinase inhibition [15,37]. Although there have been no veterinary studies verifying the expression of RTKs on endocrine pancreatic tumours such as insulinomas, other neoplastic neuroendocrine histotypes, including thyroid carcinomas and apocrine gland anal sac adenocarcinomas, have been shown to express targets of toceranib [16,17].
One challenge of this study was accurately de ning the extent of disease and objectively measuring response to therapy. As a retrospective study, the diagnostic and staging tests were preexisting and the decisions in uenced by owner and clinician preferences. In addition to paired fasting insulin and glucose ratio, the work-up of a dog with suspected insulinoma ideally incorporates both thoracic and abdominal imaging [1,2,38]. In the present study, all 30 dogs had either thoracic radiographs and abdominal ultrasound, and/or CT scan of the thorax and abdomen performed for complete staging purposes. In the 22 dogs for which thoracic radiographs were performed, no overt evidence of pulmonary metastatic disease was noted at diagnosis. This nding is consistent with previous reports of dogs with insulinoma in which pulmonary metastasis were not observed at diagnosis [4,[6][7][8][9][10]. Twenty-three dogs had abdominal ultrasound reports available. For these dogs, a pancreatic mass was identi ed in just 10 dogs (43.5%), which is comparable to previous reports that describe abdominal ultrasound sensitivities ranging from 28% to 75% for detection of insulinoma in dogs [6, 7, 48]. CT scan identi ed a pancreatic mass in 10/12 (83%) dogs in which it was performed, including three dogs where the mass was not evident on ultrasound. Although CT scan has been shown to identify 71.4% of primary insulinomas in dogs, contrast-enhanced CT has also been reported to inaccurately locate insulinomas in dogs, and intraoperative evaluation of the pancreas and potential metastatic lesions is proposed to be superior for staging purposes [38,39].
There is also the potential for misdiagnosing metastatic lesions utilizing diagnostic imaging, with one study indicating that 37.5% of suspected intra-abdominal metastases were negative for insulinoma [7]. The original TNM classi cation for dogs diagnosed with insulinoma recommends that lymph nodes be examined by laparotomy or laparoscopy [30]. Additionally, multiple veterinary studies have observed the advantage of surgical treatment of insulinomas even when intra-abdominal metastatic disease was present [6,8,9]. Considering that exploratory laparotomy has both diagnostic and therapeutic value, it should ideally be performed in all cases where insulinoma is suspected.
There were several limitations to this study that are inherent to its retrospective and multi-institutional design. As previously discussed, there was a lack of standardization in the methods used to con rm the diagnosis of insulinoma. Histopathology reports were not available for nine dogs, relying instead on a variable combination of clinical signs attributed to neuroglycopenia, paired blood glucose and insulin measurements, and cytology results, and/or supportive imaging ndings. While not ideal, this clinical approach is consistent with previous publications describing the diagnosis of canine insulinoma [1,2,38].
Additionally, our study included one dog that had a previous diagnosis of a hepatocellular carcinoma.
Paraneoplastic hypoglycemia and secretion of insulin-like growth factor type-II (IGF-2) has been reported in cases of canine hepatocellular carcinoma [40,41]. Although this dog met the diagnostic inclusion criteria for our study, the historical diagnosis of hepatocellular carcinoma may have contributed to the hypoglycemia diagnosis and later complicated follow-up blood glucose monitoring. Another dog in our study was previously diagnosed with a metastatic mast cell tumour. Once again, although this dog met the diagnostic inclusion criteria for the present study, monitoring response to therapy in this dog could have been confounded by this prior diagnosis as canine mast cell tumours have been reported to respond to toceranib phosphate [42,43]. The methods used to monitor treatment response were also not standardized, and therefore the retrospectively applied stage at diagnosis and progression timepoints may have been underestimated. To more accurately evaluate treatment response and measures of outcome, a prospective study with standardized diagnostic and staging criteria are ideal. Finally, the lack of appropriate control populations precluded comparative analyses of endpoints.
The clinical endpoints reported in this study were also potentially confounded by the fact that the clinical setting in which toceranib was initiated was variable. Twenty-one dogs in our study were treated with partial pancreatectomy prior to initiating toceranib therapy. Prior studies report long OSTs ranging from

Methods
The primary objectives of this study were to describe the duration of clinical bene t, de ned as absence of clinical signs associated with hypoglycemia and/or measurable response according to cRECIST, and measures of outcome in dogs diagnosed with insulinoma treated with toceranib [29]. A secondary objective was to describe the adverse effects of toceranib in dogs with insulinoma.

Case Selection
The medical record databases of ve university teaching hospitals and eight veterinary specialty referral centers were searched for cases in which toceranib phosphate was used to treat dogs diagnosed with neuroendocrine pancreatic neoplasia. Inclusion criteria were: 1) documentation of fasting hyperinsulinemia with paired hypoglycemia, and/or cytologic or histologic diagnosis of a primary pancreatic tumour with neuroendocrine morphology, 2), other causes of hypoglycemia, including sepsis, hepatic failure, adrenocortical insu ciency, and toxin ingestion were reasonably investigated and eliminated 3) details of any prior, concurrent or subsequent treatments, 4) toceranib dosage and schedule, 5) at least one documented follow-up assessment during the toceranib treatment period.
Exclusion criteria were: 1) absence of an acceptable clinical diagnosis of insulinoma, 2) insu cient details regarding toceranib treatment, as described in the inclusion criteria, and 3) lack of a documented follow-up assessment. Endpoints were response to therapy, progression-free interval (PFI), and overall survival time (OST).
Response to therapy was de ned as an absence of clinical signs associated with hypoglycemia and/or measurable response according to cRECIST [29] (Table 3).

Statistical Analysis
The Kaplan-Meier estimator was used to estimate the survival distribution of each of two right-censored response variables (i.e. PFI and OST). This estimator is a non-parametric statistic that is commonly used in the medical literature to estimate the fraction of patients that have not shown an undesirable event, such as death or disease progression, for certain amount of time. Key to Kaplan Meier curves is that they can take into account right-censored data, which occurs in cases that are lost-to-follow-up or for which the event of interest has not been observed at last follow-up. Censoring criteria were lack of progression diagnosis and alive status for PFI and OST, respectively, at end of the follow-up period. Computations were conducted using the LIFETEST procedure of SAS (Version 9.4, Cary, NC). Con dence intervals at given time point were calculated based on log-log transformations.
A Cox proportional hazards (PH) model was tted to each response PFI and OST using the censoring criteria described in the previous paragraph. The Cox PH model is a parametric survival method that further re nes the Kaplan Meier approach by enabling covariate adjustments, thus enabling assessment of potential risk factors. The linear predictor considered the following explanatory non-time-dependent covariates, namely sex, age at diagnosis, weight at diagnosis, type of veterinary practice, tumour stage, therapies prior to toceranib, dosing frequency, toceranib dose, veterinary clinic, and all 2-way interactions.  Figure 1 Estimated Kaplan-Meier survival curve for median overall progression free interval for 30 dogs with insulinoma treated with toceranib. In absence of documentation of PD, disease progression was considered not observed for 10 dogs, up until the last veterinary visit, and PFI was right-censored for these cases.