Serum insulin concentration in dogs with insulinoma as a clinical marker for presence of metastasis at the time of diagnosis

Abstract Background Information regarding serum insulin concentration in dogs newly diagnosed with insulinoma and its association with clinical stage and survival time is lacking. Objective Examine association between serum insulin concentration and survival and clinical disease stage in dogs with insulinoma. Animals. Fifty‐nine client‐owned dogs with a diagnosis of insulinoma from 2 referral hospitals. Method Retrospective observational study. The χ 2 test was used to compare the proportion of dogs with increased insulin concentration in groups with or without metastasis at the time of diagnosis. Linear mixed‐effect models were built to compare differences in insulin concentration between dogs with and without evidence of metastasis at the time of original diagnosis. Cox's proportional hazards regression and Kaplan‐Meier graphs were used to evaluate the association between insulin concentration and insulin groups and survival. Results Median serum insulin concentration was 33 mIU/L (range, 8‐200 mIU/L) in dogs with World Health Organization (WHO) stage I disease and 45 mIU/L (range, 12‐213 mIU/L) in dogs with WHO stage II and III disease. No difference was found in the proportion of dogs with increased insulin concentration with or without metastasis (P = .09). No association was identified between insulin concentration and survival (P = .63), and between dogs grouped by insulin concentration and survival (P = .51). Conclusions and Clinical Importance Serum insulin concentrations were not different between dogs with or without metastasis at diagnosis. The degree of insulinemia does not provide further information regarding the stage of the disease and is not associated with survival time in dogs with insulinoma.

K E Y W O R D S insulinemia, prognosis, stage, survival

| INTRODUCTION
Insulinoma is a functional tumor of pancreatic beta cells, characterized by insulin secretion independent of blood glucose concentration. 1 Insulinoma has been reported in different species including humans, 2,3 dogs, 4 and cats. [5][6][7] Many different clinical signs are observed, including seizures, collapse, generalized weakness, muscle twitching or trembling, ataxia, exercise intolerance, pelvic limb weakness, and disorientation. [8][9][10] These signs are thought to be the result of either neuroglycopenia or hypoglycemia-induced catecholamine secretion. [11][12][13] In human medicine, as many as 90% of insulinomas are reported to be benign whereas, in veterinary medicine, most dogs have microscopic or macroscopic metastasis at the time of diagnosis. 1,6 The most common sites of metastasis are locoregional lymph nodes, liver, and lungs; metastasis to the spleen, mesentery, gastrointestinal tract, kidney, spinal cord, and bone has been described but are much less common. 1,4,8,9,13,14 Clinical staging of insulinomas according to the WHO defines stage I as T 1 N 0 M 0 (primary tumor with no evidence of metastasis), stage II as T 1 N 1 M 0 (primary tumor with locoregional lymph node metastasis) and stage III as T 1 N 1 M 1 or T 1 N 0 M 1 (primary tumor with distant metastasis with or without locoregional lymph node metastasis). 15 Treatment of choice is surgical resection of the primary mass and lymph nodes with metastases if present. 4,8,14 For nonsurgical candidates, medical treatment can be considered. This treatment includes attempts to decrease insulin secretion (e.g., diazoxide, octreotide), to increase insulin resistance and promote gluconeogenesis (e.g., glucocorticoids) or cytotoxic treatment (e.g., streptozocin, tyrosine kinase inhibitors). 1,4,8,[16][17][18][19][20] Detection of metastases at the time of diagnosis is fundamental to select the most appropriate treatment approach. Their presence may influence cost of treatment, treatment success rate, and prognosis. Unfortunately, despite advanced diagnostic techniques such as contrast-enhanced computed tomography (CT), abdominal ultrasonography, and fine needle aspiration of suspicious intra-abdominal lesions, metastases can be missed. In addition, economic constraints of clients and the inability for some veterinary facilities to perform such procedures should be considered. Therefore, development of a simple inexpensive method to predict the presence of metastasis at time of diagnosis would be useful in the management of these cases. Two recent studies have described clinically relevant differences between benign and malignant insulinomas in people. 21,22 Compared to benign tumors, malignant insulinomas tended to be larger, be associated with higher plasma insulin concentrations and to be associated with metastases. 21  with concurrent inappropriate serum insulin concentration (either within or above the reference interval). Blood glucose concentration was measured on whole blood using a portable glucose meter (AlphaTrak) or on serum obtained by centrifugation within 15 minutes after collection of a blood sample in serum gel tube using a biochemical analyzer (Beckman Coulter AU480), both methods validated in dogs. 23,24 Serum insulin concentration was measured either by radioimmunoassay (RIA) or immunoradiometric assay (IRMA). The RIA has been validated in dogs, 25 whereas the IRMA was internally validated against the RIA by the laboratory (NationWide Laboratories, UK) using approved procedures which included parallelism, intra-assay and inter-assay variation as well as limit of detection and recovery studies using samples from dogs. The same IRMA assay was used in a recent study measuring insulinemia in dogs undergoing surgical management of insulinoma. 14 The same serum sample in which hypoglycemia was identified was used to quantify insulinemia, once other causes of low blood glucose concentration were excluded. If the sample volume was not sufficient or had not been maintained at or below 4 C, a new blood sample was obtained and, if hypoglycemia was confirmed, the sample was placed in a refrigerator pack and sent to the respective external laboratory for insulin measurement. In all cases, serum insulin concentration was assessed using the same sample in which hypoglycemia was documented.

| Case identification and eligibility criteria
Evidence of a pancreatic nodule on abdominal imaging (CT or ultrasonography) also was required for inclusion. In addition, thoracic CT or 3-view thoracic radiography also was required to stage each dog using the TNM system. Finally, data needed to be available from history, physical examination, hematology, serum biochemistry and, if available, either the cytological or histopathological report of the pancreatic nodule and of the suspected metastatic lesion. The diagnosis of insulinoma still was considered likely even in dogs without cytological or histopathological examination of the pancreatic mass, based on compatible clinical signs, normal or increased serum insulin concentration with concurrent hypoglycemia, and presence of a pancreatic mass on CT or ultrasonography. Dogs with relapsed disease and those in which staging had either not been performed or had been performed

| Data collection and grouping
The electronic client data management systems of the 2 referral hospitals were searched for dogs with a diagnosis of insulinoma. Data

| Statistical analysis
Data were entered into an electronic spreadsheet (Microsoft Excel for Mac version 16.19) and checked for errors, and an online open-access statistical language and environment (R, version 4.2.0) 26 was used for all statistical manipulations, and the level of statistical significance was set at P < .05 for 2-sided analyses. A χ 2 test was used to compare the proportion of dogs with increased serum insulin concentration in groups with or without metastasis at the time of diagnosis, and the sensitivity and specificity of having an increased serum insulin concentration above the reference interval also was determined.
Given that different insulin assays were used at the different veterinary hospitals, linear mixed-effects models were built (using the

| Animals
Sixty dogs with a diagnosis of insulinoma were identified from the 2 referral hospitals. However, 1 dog was excluded because of a pre-existing diagnosis of hyperadrenocorticism. Therefore, 59 dogs ultimately were included in the study (Table 1)

| DISCUSSION
Our study is the first to assess serum insulin concentration in dogs with insulinoma in relation to clinical staging. Previous studies demonstrated an association between the presence of metastasis at time of diagnosis and survival time, highlighting the importance of staging dogs with insulinoma. 14,28,29 This consideration is particularly important because insulinoma in dogs is more often malignant, compared to insulinoma in human patients. Two studies in human medicine investigated the differences between benign and malignant insulinomas.
One study reported that, when patients were presented within 24 hours of symptoms appearing, the fasting time for the occurrence of hypoglycemia was <8 hours, the serum insulin concentration was ≥28 mU/mL and C-peptide concentration was ≥4 ng/mL at the glycemic nadir, malignant insulinoma was more likely. 21 A tumor >2.5 cm in diameter also was more frequently malignant in people. 21 Another study confirmed these results and also determined that median insulin study also confirmed that malignant tumors were larger than benign tumors (4.2 vs 1.8 cm). 22 The criteria used in human medicine to differentiate a benign vs malignant insulinoma rely on the presence of metastasis, but histological and immunohistochemical analysis is sometimes necessary in those patients without extrapancreatic disease. 30 In contrast, for dogs included in our study, no difference in serum insulin concentration was found between those with or without metastasis at diagnosis, nor could serum insulin concentrations distinguish benign from malignant primary tumors. Therefore, in dogs, serum insulin concentration at the time of diagnosis has limited prognostic value. The discrepancy between studies in human medicine and our study could be explained by the fact that, in studies of humans, insulin concentration was compared between cases of benign and malignant insulinomas whereas, in our study, all dogs had malignant neoplasia, albeit of different stages. A previous study hypothesized that increased insulin and C-peptide concentrations might be associated with a larger-sized primary mass rather than the presence of metastasis, 22 implying that the amount of insulin produced by metastases might be negligible and thus explaining the lack of difference in insulin concentration in dogs with different disease stages seen in our study. That said, an association between hypoglycemia and clinical signs has been reported in dogs with stage II and III disease, making control more difficult and leading to worse prognosis. 14 Furthermore, to our knowledge, no previous studies have examined associations between the size of the primary pancreatic mass and serum insulin concentration, regardless of the stage of disease. The size of the primary pancreatic mass was not recorded in many dogs in our study and, therefore, it was not possible to evaluate this aspect. However, no association was found between tumor size and prognosis in a previous study. 14 Finally, the average duration of clinical signs in the population of our study was 4 weeks, which is consistent with recent studies reporting 4 to 6 weeks since the onset of clinical signs. 10,14 Our study showed an association between duration of clinical signs and survival, indicating that the longer the duration of the clinical signs, the shorter the survival time. This observation is in contrast with previous studies, 10,14 in which no significant differences were found between these 2 variables. However, dogs having clinical signs for a longer period of time either because of delayed diagnosis or delayed intervention or both, potentially have more time to develop metastasis, which is associated with worse prognosis. 10,14 In addition, although the association between the size of the pancreatic mass and a poor prognosis has been demonstrated only in human patients, 21 F I G U R E 1 Kaplan-Meier curve for survival of dogs classified according to insulin level as high or normal.
with time the insulinoma can become larger with consequently higher insulin production. Prolonged hyperinsulinemic hypoglycemia could become more difficult to control medically and also could induce more severe pancreatic β-cell dysfunction, with higher risk of postoperative diabetes mellitus.
Our small observational study had some limitations. First, timing of sampling for insulin concentrations was not recorded. In both healthy humans and dogs, insulin secretion follows a circadian rhythm, 31 which might be disrupted by the presence of insulinoma.
Such effects were not considered in our study. Second, given the small population and lack of sample size calculation, it is possible that the study was underpowered, with the potential for type I statistical error. Third, the retrospective and multicenter nature of the study might have introduced variability as a result of differences in diagnostic and therapeutic approach. Most notable was the use of 2 different insulin assays, although an attempt was made to correct for this concern using mixed-effects and stratification in regression.
Fourth, some cases were staged only based on thoracic and abdominal radiographs, wherein sensitivity for metastasis is relatively poor. 32 Abdominal ultrasonography and conventional CT have sensitivity of approximately 28% to 75% and 71%, respectively, for the detection of a primary mass. Sensitivity can be increased up to 96% using contrast-enhanced CT. 33 However, primary masses still can be missed, and surgical exploration might be necessary to confirm them and localize their presence by direct palpation of the pancreas, as noted in a previous study. 14 Furthermore, contrast-enhanced CT has a reported sensitivity for detecting lymph node and hepatic metastasis of approximately 67% and 75%, respectively. 33 This aspect of our study could have caused incorrect classification of some cases. In addition, not all dogs underwent a complete diagnostic evaluation for hypoglycemia. However, the inappropriate insulinemia in relation to the hypoglycemia and presence of a pancreatic nodule or mass made most other causes of hypoglycemia unlikely.
Finally, some dogs were included in our study despite no cytological or histological confirmation of insulinoma being performed. One case report described a dog affected by nesidioblastosis, 34 a condition caused by the hyperplasia of the β-cells and no detectable pancreatic nodule or mass, and diagnosis was reached only after histological evaluation of pancreatic surgical biopsy samples. In our cases, all dogs without cytological or histological assessment had evidence of a pancreatic nodule or mass.
In conclusion, serum insulin concentration was not associated with either the stage of disease or survival in dogs with insulinoma.
These findings confirm and reinforce the importance of clinical staging in dogs with insulinoma in order to provide a more accurate prognosis.

ACKNOWLEDGMENT
No funding was received for this study.