Association between somatostatin analogues and diabetes mellitus in gastroenteropancreatic neuroendocrine tumor patients: A Surveillance, Epidemiology, and End Results‐Medicare analysis of 5235 patients

Abstract Background Gastroenteropancreatic neuroendocrine tumors (GEP‐NETs) are increasingly common malignancies and tend to have favorable long‐term prognoses. Somatostatin analogues (SSA) are a first‐line treatment for many NETs. Short‐term experiments suggest an association between SSAs and hyperglycemia. However, it is unknown whether there is a relationship between SSAs and clinically significant hyperglycemia causing development of diabetes mellitus (DM), a chronic condition with significant morbidity and mortality. Aim In this study, we aimed to compare risk of developing DM in patients treated with SSA vs no SSA treatment. Methods and Results Using the Surveillance, Epidemiology, and End Results (SEER) database and linked Medicare claims (1991‐2016), we identified patients age 65+ with no prior DM diagnosis and a GEP‐NET in the stomach, small intestine, appendix, colon, rectum, or pancreas. We used χ 2 tests to compare SSA‐treated and SSA‐untreated patients and multivariable Cox regression to assess risk factors for developing DM. Among 8464 GEP‐NET patients, 5235 patients had no prior DM and were included for analysis. Of these, 784 (15%) patients received SSAs. In multivariable analysis, the hazard ratio of developing DM with SSA treatment was 1.19, which was not statistically significant (95% CI 0.95‐1.49). Significant risk factors for DM included black race, Hispanic ethnicity, prior pancreatic surgery, prior chemotherapy, tumor size >2 cm, pancreas tumors, and higher Charlson scores. Conclusion DM was very common in GEP‐NET patients, affecting 53% of our cohort. Despite prior studies suggesting an association between SSAs and hyperglycemia, our analysis found similar risk of DM in SSA‐treated and SSA‐untreated GEP‐NET patients. Further studies are needed to better understand this relationship. As NET patients have increasingly prolonged survival, it is crucial to identify chronic conditions such as DM that these patients may be at elevated risk for.


| INTRODUCTION
Gastroenteropancreatic neuroendocrine tumors (GEP-NETs) are an increasingly common group of malignancies, with diagnosis rates increasing by more than sixfold in the last four decades and continuing to rise. 1,2 Even in patients who present with advanced stage tumors, the long-term prognosis is relatively favorable, with median survivals of 5 years or more. 3

,4 A common first-line treatment for
NETs is a somatostatin analogue (SSA), such as octreotide or lanreotide. Given the often-prolonged duration of treatment with SSAs and long survival of patients with GEP-NETs, it is important to understand the potential long-term effects of these treatments.

Diabetes mellitus (DM) is a serious condition that has been
reported in approximately 21.4% of all individuals over age 65. 5 DM is associated with significant complications and morbidity, decreased quality of life, and increased mortality. [6][7][8][9][10] SSAs have been shown to alter glucose homeostasis. In clinical trials of the octreotide acetate long-acting injection in NET patients, hyperglycemia was observed in 27%. 11 Additionally, in small studies in non-NET patients, intravenous or subcutaneous octreotide was found to be effective in treating sulfonylurea-induced hypoglycemia in the acute setting, 12,13 thought to be via octreotide's inhibition of insulin release from pancreatic beta cells. However, the effects of SSAs on glucose regulation are complex. While SSAs inhibit insulin and glucagon-like peptide 1 (GLP-1) secretion, which increases blood glucose; they also inhibit secretion of the counterregulatory hormones, growth hormone and glucagon, thereby reducing insulin resistance and blood glucose levels. [14][15][16] Though less common than instances of hyperglycemia, hypoglycemia was observed in 4% of NET patients in the octreotide clinical trials. 11 In NET patients, small studies have suggested an association between GEP-NETs and metabolic syndrome 17 and higher fasting glucose, 18 but these initial studies showed no statistically significant difference in these measures in patients treated with SSAs. Overall, although hyperglycemia is more often suggested than hypoglycemia, the net effect of SSAs on glucose homeostasis in GEP-NET patients is not well-understood. Additionally, whether SSAs predict clinically significant and long-term hyperglycemia causing subsequent diabetes mellitus has not been studied.
Therefore, our primary aim in this retrospective study is to determine whether SSA treatment independently increases risk of developing DM in GEP-NET patients.

| Study population
The study was approved by the Icahn School of Medicine Institutional stage. Participants were defined as having the clinical symptoms of carcinoid syndrome if they had two or more diagnosis codes for "diarrhea," "flushing," or "carcinoid syndrome" using International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes.
Prior NET treatments with SSAs (octreotide or lanreotide), pancreatic surgery, radiation, or chemotherapy were identified by ICD-9 diagnosis codes and Healthcare Common Procedure Coding System (HCPCS) codes, billing codes which identify services, procedures, and supplies. All treatments were considered binary variables, such that patients who received a treatment for any period of time were treated equally as having received the treatment. Charlson comorbidity index was calculated using ICD-9 code claims preceding the date of NET diagnosis for each patient. Patients were defined as having developed DM by determining the date of the first occurrence of an ICD-9 code for DM across all of an individual's claims after the date of GEP-NET diagnosis.
The primary outcome was the risk of developing DM in SSA-treated vs SSA-untreated GEP-NET patients. Secondary outcomes were identification of other significant predictors of developing DM in adjusted analysis.

| Statistical analysis
We compared patients who received SSA with those who did not receive SSA using chi-squared tests. We used a multivariable Cox proportional hazards model to identify independent risk factors for developing DM. In this time-to-event analysis, the target event was the time of DM diagnosis, and the time origin was the GEP-NET diagnosis date.
Receipt of SSA treatment was analyzed as a time-dependent variable, such that the value of the binary "received SSA" variable is reassigned at each time point that there is an event (ie, a diagnosis of diabetes).
This accounts for the possibility that SSA may have been started at varying time points before or after DM diagnosis. Individuals who had no DM diagnosis by the date of last follow-up were treated as censored observations. Kaplan-Meier survival analysis was performed to compare survival in DM patients vs non-DM patients, for the entire cohort of NET patients, and for subgroups of pancreatic primary tumors compared to tumors of other primary sites. A P-value of <.05 was considered statistically significant. All statistical analyses were performed using SAS (Cary, North Carolina).

| RESULTS
We identified 8464 patients with a pathologically confirmed GEP-NET from the SEER database and linked Medicare claims from 1991 to 2016. Of these, 3229 (38% of all GEP-NET patients) had a diagnosis of DM predating their NET diagnosis and were excluded from the main analysis. The remaining 5235 patients were included for analysis.  This is the first study to describe a markedly higher prevalence of DM in individuals with GEP-NETs compared to the general population over age 65. One group has previously observed an association between well-differentiated GEP-NETs and metabolic syndrome and visceral obesity. 17,18 Higher fasting glucose and glucose intolerance were described previously in two small samples of patients with NETs; however, the prevalence of DM diagnoses in the GEP-NET population has not been studied. 17,25 It has been proposed that DM and neuroendocrine tumors share components in their pathogenesis, though the exact mechanism is not known. 26 This is supported by studies showing that some medications for diabetes such as metformin have been associated with slower tumor progression in NET patients. 27,28 Further studies are needed to better explain these relationships.
We found significantly increased risk of diabetes in non-Hispanic black and Hispanic populations, which is consistent with prior data in both the general population and the elderly. [29][30][31] A variety of socioeconomic (eg, neighborhood, income) and biologic (eg, BMI, waist circumference) risk factors have been linked to the higher rates of DM in these groups and are likely also at play in our patient population. [32][33][34] Our finding that pancreatic NETs in particular predicted higher risk of DM has also been described previously, though the direction of causality has been unclear. 35 Strengths