Efficacy and safety of canagliflozin in insulin dysregulated horses: a 4-week multi-arm, double-blind, randomized, clinical trial

Abstract

Background

Sodium-glucose cotransporter-2 (SGLT2) inhibitors are promising drugs for treatment of hyperinsulinemia in insulin-dysregulated (ID) horses.

Hypothesis/Objectives

Compare short-term effects of the SGLT2 inhibitor canagliflozin versus placebo on glucose and insulin responses during an oral sugar test (OST) and a forage-based feed challenge test (FCT), and to assess adverse effects.

Animals

Forty-two privately owned severely ID horses.

Methods

Multi-center, randomized, double-blind, placebo-controlled, parallel-group study. Horses were allocated (1:1:1) to receive canagliflozin (0.6 or 1.2 mg/kg PO q24h) or placebo. The study included a 5-day baseline evaluation, a 4-week at-home double-blind treatment period, and a 5-day follow-up evaluation. During each clinical visit, an OST (day 3) and FCT (day 4) were performed to characterize glucose and insulin responses. Adverse effects were recorded. Data are presented as geometric least squares mean (95% CI).

Results

Peak insulin concentrations (μIU/mL) during the OST and the FCT were lower after treatment with canagliflozin 0.6 and 1.2 mg/kg compared with placebo (P ≤ .01); OST: 122.0 (96.6-154.0) and 108.6 (85.6-137.8) vs 288.7 (227.1-366.9) and FCT: 78.3 (37.6-162.9) and 54.2 (25.7-114.5) vs 140.3 (68.5-287.3). Canagliflozin caused a dose-dependent increase in triglyceride concentrations (P ≤ .01). Post-treatment triglyceride concentrations were 0.5 (0.4-0.8), 1.3 (0.9-2.0), and 2.9 (1.9-4.4) mmol/L for placebo, 0.6 mg/kg, and 1.2 mg/kg groups, respectively.

Conclusions and clinical importance

Canagliflozin effectively decreases hyperinsulinemia in ID horses but induces a dose-dependent increase in triglyceride concentrations.

Introduction

Hyperinsulinemia in insulin-dysregulated (ID) horses is a major risk factor for the most common form of laminitis, hyperinsulinemia-associated laminitis (HAL).^1–3 Consequently, decreasing hyperinsulinemia is central to preventing laminitis. Most ID horses respond favorably to management interventions, including diets low in non-structured carbohydrates (NSC), structured weight loss programs, and increased exercise.⁴ However, when these strategies fail to adequately control postprandial hyperinsulinemia, adjunctive pharmacological approaches may be required. Sodium-glucose cotransporter 2 (SGLT2) inhibitors are a relatively new class of drugs used to treat type 2 diabetes mellitus (T2DM) in humans. By inhibiting renal glucose reabsorption in the proximal tubule, these agents increase urinary glucose excretion (UGE), leading to weight loss and decreased hyperglycemia.⁵ Although no veterinary drugs currently are licensed for the treatment of ID in horses, SGLT2 inhibitors recently have been proposed as a complementary treatment to aid in controlling excessive insulin responses.^6–10 Glucose is the primary stimulus for insulin secretion from pancreatic β-cells and the hypothesis for using SGLT2 inhibitors in ID horses is that decreased postprandial plasma glucose concentrations will lead to attenuated insulin responses and thereby decrease the risk of HAL.¹¹

To date, the SGLT2 inhibitors, velagliflozin,^8,9,12 canagliflozin,^6,13,14 ertugliflozin,^10,15 and dapagliflozin,⁷ have been reported for use in horses. Several published reports have used single samples or basal insulin concentrations as the measure of treatment efficacy.^7,9,15,16 In contrast, assessment of postprandial insulin trajectories after forage-based feed challenges may provide more detailed insight into how SGLT2 inhibitors influence the amplitude, duration, and overall pattern of insulin responses in ID horses. The commonly suggested effective daily PO dosage of canagliflozin ranges from 0.3 to 0.6 mg/kg,^6,16 but it remains unknown whether higher doses produce larger decreases in postprandial hyperinsulinemia or increase the risk of adverse effects, such as hypertriglyceridemia. Weight loss is observed in SGLT2-treated horses^6,7 and is presumed to result from negative energy balance secondary to glucosuria, although the possible contributions of natriuresis or decreased plasma volume have not been investigated in horses.^17–19

Our first objective was therefore to compare the short-term effects of 2 canagliflozin dosages with placebo on glucose and insulin responses during an oral sugar test (OST) and a forage-based feed challenge test (FCT) in ID horses. The second objective was to determine the effects of canagliflozin on body weight, plasma electrolyte concentrations, plasma volume, liver enzyme activities, and triglyceride concentrations.

Materials and methods

Detailed materials and methods are provided in the Appendix.

Study design and horses

This 4-week, multicenter, randomized, parallel-group, double-blind, placebo-controlled trial evaluated the efficacy and adverse effects of canagliflozin treatment in ID horses. Client-owned horses and ponies previously diagnosed with severe ID were eligible for enrollment. A flow diagram of enrollment is provided in Figure 1. Eligible animals were randomized (1:1:1) to receive once-daily PO placebo or canagliflozin at 0.6 mg/kg or 1.2 mg/kg (Invokana, 100 or 300 mg tablets, Mundipharma AB, Gothenburg, Sweden). Investigators and owners were blinded to treatment allocation throughout the trial.

Figure 1.

Flow diagram for enrollment, randomization, allocation, follow-up, and data analysis for the 4-week study. Abbreviation: ID = insulin dysregulation.

Experimental protocol

Horses were admitted to the University Animal Hospitals in Sweden or Norway for an initial 5-day baseline evaluation and returned after 24 days of at-home treatment for a 5-day follow-up evaluation. During the second visit, double-blind treatments were administered. At each visit, horses were allowed a 2-day acclimatization period. Blood samples were collected from IV jugular catheters.

On the morning of day 3 at both visits, feed was withheld and a 12-sample, 300-min OST (0.5 mL/kg body weight of Dan Sukker Glykossirap; Nordic Sugar A/S, Copenhagen, Denmark) was started at 7 a.m. On Day 4 at both visits, an 11-sample, 300-min FCT using each horse’s regular forage (hay or haylage; 0.4 kg dry matter/100 kg body weight) was conducted at 7 a.m.

The experimental protocol included an optional fifth day at each visit for a secondary study assessing β-cell responsiveness to IV glucose using a graded glucose infusion test. This study will be reported separately.

Primary endpoints

Primary endpoints were the differences between treatment groups in the area under the plasma insulin vs time curve (insulin AUC_0-300) and the maximal plasma insulin concentration (insulin C_max) during the OST and FCT.

Statistical methods

Endpoints were analyzed using mixed effects models with treatment group (placebo and canagliflozin 0.6 or 1.2 mg/kg) as fixed effect and the corresponding baseline as a continuous covariate. Study center (Sweden or Norway) was included as a random effect.

Adjusted means are presented as least squares (LS) means ± standard error of the mean (SEM). Variables with non-normally distributed residuals are reported as geometric least squares (LS) means with 95% CI after back transformation. Pairwise comparisons between treatments were conducted using Tukey–Kramer post hoc tests. Statistical significance was defined as P < .05.

Results

Study subjects

Of 42 horses randomized to receive placebo (n = 14), canagliflozin 0.6 mg/kg (n = 14), or canagliflozin 1.2 mg/kg (n = 14), all but one completed the study. The single withdrawal occurred in the placebo group when the horse sustained an accident during transport home after the first visit (Figure 1). The study population comprised British native pony breeds (n = 14), crossbreed ponies (n = 4), Scandinavian native pony breeds (n = 6), Icelandic horses (n = 12), warmbloods (n = 4), and cold blood breeds (n = 2). Demographic variables and baseline clinical characteristics were comparable across treatment groups (Table 1).

Table 1.

Baseline clinical characteristic of participating insulin dysregulated horses.

Variable	Placebo    (n = 14)	Canagliflozin 0.6    (n = 14)	Canagliflozin 1.2 (n = 14)
Age (years)	14.9 ± 3.1	12.3 ± 4.5	15.1 ± 3.4
Sex (mare/gelding)	11/3	10/4	10/4
Weight (kg)	338 ± 128	381 ± 138	362 ± 93
BCS	6.5 (1.75)	7.25 (0.75)	7 (2.625)
CNS	3.25 (1.0)	3.75 (1.125)	3.75 (1.0)
Previous laminitis (yes/no)	13/1	12/2	10/4
Fasting plasma glucose (mmol/L)	5.4 ± 0.8	5.1 ± 0.4	5.1 ± 0.4
90 min OST plasma glucose (mmol/L)	9.2 ± 1.9	8.9 ± 1.0	9.0 ± 1.6
Fasting plasma insulin (μIU/mL)	24.4 (76.4)	31.9 (26.0)	43.3 (42.3)
90 min OST plasma insulin (μIU/mL)	243.6 (319.1)	205.0 (172.7)	277.8 (426.2)

Values are mean ± SD, median (IQR), or n/n (count number).

Abbreviations: BCS = body condition score; CNS = cresty neck score; OST = oral sugar test.

Fasting insulin and postprandial insulin responses

Least square mean differences or ratios for primary and secondary endpoints are presented in Table S1. Treatment with canagliflozin (0.6 and 1.2 mg/kg) resulted in lower adjusted geometric mean fasting plasma insulin concentrations on the mornings of both the OST and FCT compared with placebo (P < .01), with no significant difference between the two canagliflozin dosages (P ≥ .56; Table 2).

Table 2.

Comparison of primary and secondary endpoints between placebo (PBO), canagliflozin 0.6 mg/kg (CAN 0.6), and canagliflozin 1.2 mg/kg (CAN 1.2) treated insulin dysregulated horses.

Parameter	Placebo (n = 13)	Canagliflozin 0.6 (n = 14)	Canagliflozin 1.2 (n = 14)	Pairwise comparison (P-value)
				CAN 0.6 vs PBO	CAN 1.2 vs PBO	CAN 1.2 vs CAN 0.6
Oral sugar test
Fasting plasma insulin (μIU/mL)	35.8 (28.9-44.2)	16.9 (13.8-20.2)	14.1 (11.6-17.3)	<.001	<.001	.56
Insulin Cmax (μIU/mL)	288.7 (227.1-366.9)	122.0 (96.6-154.0)	108.6 (85.6-137.8)	<.001	<.001	.84
Insulin AUC0-300 (μIU/mL min)	53942 (43154-67427)	21162 (17520-26683)	16577 (13330-20616)	<.001	<.001	.34
Fasting plasma glucose (mmol/L)	5.1 ± 0.2	4.8 ± 0.2	4.4 ± 0.2	.35	.01	.20
Glucose Cmax (mmol/L)  Covariate at 7.8 (mmol/L)   Covariate at 9.9 (mmol/L)   Covariate at 12.0 (mmol/L)	7.3 ± 0.6 9.3 ± 0.4 11.3 ± 0.5	7.5 ± 0.6 7.5 ± 0.5 7.5 ± 0.9	6.5 ± 0.5 6.9 ± 0.4 7.3 ± 0.6	1.00 .01 .01	.94 <.001 <.001	.88 .91 1.00
Glucose AUC0-300 (mmol/L min)   Covariate 1873 (mmol/L min)   Covariate 2399 (mmol/L min)   Covariate 2925 (mmol/L min)	1759 ± 127 2284 ± 101 2808 ± 104	1840 ± 141 1932 ± 99 2024 ± 171	1435 ± 112 1623 ± 98 1812 ± 124	1.00 .03 .01	.30 <.001 <.001	.21 .06 .96
Feed challenge test
Fasting plasma insulin (μIU/mL)	34.8 (17.5-69.2)	16.9 (7.3-38.9)	13.5 (5.9-30.9)	.01	<.001	.57
Insulin Cmax (μIU/mL)	140.3 (68.5-287.3)	78.3 (37.6-162.9)	54.2 (25.7-114.5)	.01	<.001	.06
Insulin AUC0-300 (μIU/mL min)	26851 (17540-41105)	12085 (7769-18798)	8229 (5002-13537)	<.001	<.001	.02
Fasting plasma glucose (mmol/L)	5.3 ± 0.15	4.9 ± 0.14	4.5 ± 0.14	.28	.01	.16
Glucose Cmax (mmol/L)	6.6 ± 0.17	6.5 ± 0.15	5.5 ± 0.16	.85	.01	.01
Glucose AUC0-300 (mmol/L min)	1772 ± 43	1701 ± 42	1460 ± 40	.60	<.001	.01
Clinical biochemistry
Triglycerides (mmol/L)	0.5 (0.4-0.8)	1.3 (0.9-2.0)	2.9 (1.9-4.4)	.01	<.001	.01
GLDH (nkat/L)	113 (68-186)	189 (117-303)	298 (185-481)	.25	.01	.31
GGT (μkat/L)	0.4 (0.2-0.5)	0.5 (0.3-0.7)	0.7 (0.5-1.1)	.63	.06	.33
Na+ (mmol/L)	138.8 ± 0.5	138.4 ± 0.5	137.8 ± 0.5	.84	.32	.62
K+ (mmol/L)	3.9 ± 0.1	3.8 ± 0.1	3.7 ± 0.1	.36	.01	.28
Cl− (mmol/L)	103.0 ± 0.7	101.2 ± 0.7	100.6 ± 0.7	.12	.03	.78
Protein (g/L)	63 ± 1	64 ± 1	65 ± 1	.90	.39	.64
Osmolality (mOsm/kg)	279.1 ± 1.2	279.2 ± 1.2	279.5 ± 1.2	1.00	.96	.97
PCV (%)	31.3 ± 0.6	31.1 ± 0.6	30.5 ± 0.6	.99	.69	.78
Body weight
Body weight (kg)	355.9 ± 1.9	345.0 ± 1.8	341.9 ± 1.8	<.001	<.001	.45
Change in body weight (%)	0.6 ± 0.5	−2.9 ± 0.4	−3.9 ± 0.4	<.001	<.001	.40

Adjusted means for baseline values are presented as least square means ± SEM or geometric least square means with 95% CI. Abbreviations: insulin C_max = maximal insulin concentration; insulin AUC_0-300 = area under the insulin vs time curve at 0-300 min; GGT = gamma glutamyl transferase; glucose C_max = maximal glucose concentration; glucose AUC_0-300 = area under the glucose vs time curve at 0-300 min; GLDH = glutamate dehydrogenase.

After PO administration of glucose syrup, the postprandial insulin responses were decreased with canagliflozin treatment compared with placebo (main effect of treatment, P < .001; Figure 2A). Horses receiving placebo or canagliflozin 0.6 mg/kg did not result in return of postprandial insulin concentrations to fasting concentrations within the 300-min sampling period whereas horses treated with canagliflozin 1.2 mg/kg resulted in a return to fasting insulin concentrations by 270 min (Figure 2A). Both maximal insulin concentration (insulin C_max) and area under the plasma insulin vs time curve (insulin AUC_0-300) were decreased after 4 weeks of canagliflozin treatment (P < .001), with no difference between the 2 doses (P > .34; Table 2). Relative to placebo, the geometric LS mean insulin AUC_0-300 was decreased by 60% and 69% with canagliflozin dosages of 0.6 and 1.2 mg/kg, respectively. Three of 14 horses in each canagliflozin group had insulin C_max results <65 μIU/mL, the diagnostic cut-off for ID, whereas all placebo horses had insulin C_max > 65 μIU/mL.

Figure 2.

Adjusted geometric mean and 95% CI (A and C) or adjusted mean ± SEM (B and D) concentration-time profiles. Plasma insulin (A) and plasma glucose (B) after a 0.5 mL/kg dose of oral glucose syrup (Dan Sukker). Plasma insulin (C) and plasma glucose (D) after a feed challenge test with the horses’ own hay or haylage (0.4 kg on dry matter basis per 100 kg of body weight). PBO, placebo; CAN 0.6, canagliflozin 0.6 mg/kg; CAN 1.2, canagliflozin 1.2 mg/kg. Range of time points (postprandial phase) within a treatment that differed (P < .05) from time 0 (fasting samples) for: placebo, blue dashed arrow; canagliflozin 0.6 mg/kg, dotted red arrow; canagliflozin 1.2 mg/kg, solid gray arrow.

No differences (P ≥ .32) were observed in water-soluble carbohydrates (WSC), water-soluble carbohydrates minus fructanes (WSC-f), free glucose, free fructose, or sucrose content of the forage fed during the FCT among the treatment groups (placebo, canagliflozin 0.6, and 1.2 mg/kg; Table S2). After consumption of the standardized forage meal, postprandial insulin concentrations were lower in horses treated with canagliflozin (0.6 and 1.2 mg/kg) compared with placebo (main effect of treatment, P < .001; Figure 2C). Horses receiving canagliflozin (0.6 and 1.2 mg/kg) experienced a return to fasting insulin concentrations earlier than horses receiving placebo (180 vs 270 min). Insulin C_max decreased after 4 weeks of canagliflozin treatment (0.6 and 1.2 mg/kg; P < .001), with no difference between dosages (P = .06; Table 2). Compared with placebo, geometric LS mean insulin AUC_0-300 was decreased by 55% and 69% with canagliflozin 0.6 and 1.2 mg/kg, respectively (P < .001). The geometric LS mean insulin AUC_0-300 was lower in the canagliflozin 1.2 mg/kg group compared with the 0.6 mg/kg group (P = .02; Table 2).

Dose–response relationships between the percentage decrease in insulin AUC_0-300 and the administered PO dose of canagliflozin are shown in Figure 3. Despite differences in test type, the treatment effects, expressed as percentage decrease, were similar for the OST and FCT.

Figure 3.

Mean ± SD treatment effects, expressed as percentage change in area under the insulin concentration vs time curve from 0 to 300 min (insulin AUC_0-300) from baseline to post treatment, for the oral sugar test (OST) and forage-based feed challenge test (FCT). Positive percentage values indicate a reduction in insulin AUC_0-300 whereas negative values indicate an increase in insulin AUC_0-300. Fitted curves represent sigmoidal dose–response relationships for administered canagliflozin doses (OST: r² = 0.58; FCT: r² = 0.62).

Fasting glucose and postprandial glucose responses

Least square mean differences or ratios for secondary endpoints are presented in Table S1. Fasting plasma glucose concentrations on the mornings of the OST and the FCT were lower after 4 weeks of treatment with canagliflozin 1.2 mg/kg compared with placebo (P = .01), whereas results of treatment with canagliflozin 0.6 mg/kg did not differ from placebo (P > .28; Table 2).

For maximal glucose concentration (glucose C_max) and area under the plasma glucose vs time curve (glucose AUC_0-300) during the OST, the linear mixed effects model indicated a significant covariate-by-treatment interaction (P < .001). Accordingly, comparisons of adjusted treatment means were made along the regression lines at three prespecified levels of the covariate (Table 2). As pre-treatment glucose C_max and glucose AUC_0-300 increased, the magnitude of the decrease in glucose C_max and AUC_0-300 with canagliflozin relative to placebo increased. However, no significant difference was observed between the two dosages of canagliflozin (0.6 vs 1.2 mg/kg; P ≥ .06).

After consumption of the standardized forage meal, postprandial plasma glucose concentrations were significantly decreased with canagliflozin 1.2 mg/kg compared with placebo, whereas no significant decrease was observed with canagliflozin 0.6 mg/kg (Figure 2D; Table 2). This outcome was reflected by lower glucose C_max and lower glucose AUC_0-300 in the canagliflozin 1.2 mg/kg treated horses compared with both the 0.6 mg/kg and placebo groups (P ≤ .01; Table 2). The duration of the postprandial glucose responses was 30 min shorter in both canagliflozin-treated groups (0.6 and 1.2 mg/kg) compared with placebo (Figure 2D).

Clinical biochemistry

Least square mean differences or ratios for secondary endpoints are presented in Table S1. Serum triglyceride concentrations increased in both canagliflozin groups compared to placebo (P ≤ .01), and this effect was dose dependent. Triglyceride concentrations were higher after 4 weeks of treatment with canagliflozin 1.2 mg/kg compared with 0.6 mg/kg (P = .01; Table 2). Treatment with canagliflozin 1.2 mg/kg, but not 0.6 mg/kg, increased serum glutamate dehydrogenase enzyme activity compared with placebo (P = .01 and P = .25, respectively). Serum gamma glutamyl transferase (GGT) enzyme activity did not differ among treatment groups at 4 weeks (P ≥ .06). Canagliflozin treatment did not affect serum sodium or total protein concentrations, osmolality, or PCV (P ≥ .32). Mild decreases in serum potassium and chloride concentrations were observed in the canagliflozin 1.2 mg/kg group (P ≤ .03), but all results remained within reference limits.

Body weight, body condition score, and cresty neck score

Least square mean differences or ratios for secondary endpoints are presented in Table S1. Canagliflozin (0.6 and 1.2 mg/kg) resulted in larger decrease in body weight, both in absolute kg and as percentage loss, compared with placebo (P ≤ .001; Table 2). Body condition score (BCS)²⁰ and cresty neck score (CNS)²¹ remained unchanged for the placebo treated horses after 4 weeks of treatment (Figure S1). Compared with placebo, the decrease in BCS was larger in horses treated with canagliflozin at 1.2 mg/kg (P = .02), but not in horses treated with 0.6 mg/kg (P = .30). No significant differences in CNS decrease were observed between canagliflozin treated horses and placebo (P = .30).

Physical examinations, horse owners’ scoring, and reported adverse effects

Physical examination findings, with the exception of BCS and CNS, remained within normal limits in all horses both before and after 4 weeks of treatment. Owners’ assessments of attitude, appetite, and urine volume did not differ between treatment groups (P ≥ .31; Figure S2). Increased water consumption was reported by six owners (canagliflozin 0.6 mg/kg, n = 4; canagliflozin 1.2 mg/kg, n = 2). Increased sweating was reported by 2 owners of horses in the canagliflozin 0.6 mg/kg group, and 1 owner of a horse in the canagliflozin 0.6 mg/kg group reported very mild, transient diarrhea during the first few days of treatment.

Discussion

Our randomized clinical trial demonstrates that short-term treatment with canagliflozin at both investigated dosages effectively attenuated postprandial insulin responses. This effect was characterized by a decrease in peak insulin concentrations and a shortening of the postprandial insulin response. Treatment efficacy increased with canagliflozin dosage, consistent with a dose–response relationship. However, increasing the dosage from 0.6 to 1.2 mg/kg resulted in only a modest additional decrease in insulin concentrations, while being accompanied by a more than twofold increase in serum triglyceride concentrations. Notably, canagliflozin treatment was associated with weight loss without indications of decreased plasma volume, because PCV, serum protein concentrations, and plasma osmolality remained unchanged.

Hyperinsulinemia is strongly associated with the development of laminitis in horses.^1–3,12 Consequently, therapeutic strategies that effectively attenuate hyperinsulinemia are critically important for minimizing the risk of clinical laminitis in ID horses. Among currently available pharmacologic options, SGLT2 inhibitors have been shown to be particularly effective in decreasing postprandial insulin concentrations in ID horses.^6–9,15 In agreement with previous reports, canagliflozin administration in our study resulted in a marked decrease in postprandial insulin responses. However, the magnitude of insulin decrease required to confer protection against laminitis has not yet been defined. Horses with ID often exhibit exaggerated postprandial hyperinsulinemia, largely driven by increased pancreatic β-cell responsiveness to PO carbohydrates.²² This raises the question of whether increasing the dose of SGLT2 inhibitors could provide additional therapeutic benefit in individuals with more severe ID. In our study, increasing the canagliflozin dosage from 0.6 to 1.2 mg/kg did not result in further decreases in fasting glucose or insulin concentrations, nor did it significantly alter postprandial glucose or insulin responses during the OST. In contrast, the higher dosage significantly decreased postprandial insulin concentrations during the FCT. Assessment of dose–response relationships across all treatment groups (placebo, 0.6, and 1.2 mg/kg) provided visual evidence of a diminishing incremental decrease in insulin concentrations between the 0.6 and 1.2 mg/kg dosages during both the OST and FCT. Collectively, these findings indicate that a higher canagliflozin dosage (1.2 mg/kg) may confer some additional benefits in decreasing postprandial insulin concentrations under more physiologically relevant feeding conditions, such as consumption of hay or haylage. However, escalation of the dosage may be limited by the potential for adverse effects.²³

Although most clinical trials evaluating the efficacy of SGLT2 inhibitors have relied on carbohydrate sources rich in starch and sugars,^6,8,12 our study additionally evaluated horses maintained on their regular forage-based diet consisting of hay and haylage. Compared with the OST, the carbohydrate source used during the FCT is more variable and contains a lower proportion of NSC. Nevertheless, when treatment effects were expressed as percentage decrease from baseline insulin AUC values, the FCT yielded results comparable to those obtained with the OST. Furthermore, evaluating postprandial glucose and insulin responses after normal feeding in ID horses provides valuable insight into glucose and insulin dynamics under typical management conditions. Canagliflozin treatment at both dosages (0.6 and 1.2 mg/kg) significantly altered the postprandial glucose and insulin response curves compared with placebo. The duration of the postprandial insulin period during the FCT was 90 min shorter in canagliflozin-treated horses. Assuming a feeding frequency of three meals per day, this decrease would translate into an additional 4.5 h of fasting per day compared with placebo, increasing total daily fasting time from approximately 12 to 16.5 h. Prolonged fasting is not a natural physiological state for horses, because they are grazing animals. Extended daily fasting periods may contribute to the hypertriglyceridemia observed in horses treated with SGLT2 inhibitors, potentially reflecting increased reliance on lipid mobilization during extended periods of relative insulin suppression. Further investigation is warranted to clarify the metabolic consequences of altered postprandial insulin dynamics associated with SGLT2 inhibition in horses.

In agreement with previous studies evaluating canagliflozin, velagliflozin, ertugliflozin, or dapagliflozin in horses,^7,15,24 serum triglyceride concentrations increased with canagliflozin treatment in our study, without concurrent clinical signs of hypertriglyceridemia. A previously published case series reported a dose-dependent effect of canagliflozin on serum triglyceride concentrations,²³ and this relationship was confirmed in our study. Increasing the canagliflozin dosage from 0.6 to 1.2 mg/kg resulted in a more than two-fold increase in the geometric mean serum triglyceride concentration. Furthermore, 5 of 14 horses treated with canagliflozin at 1.2 mg/kg developed serum triglyceride concentrations > 5 mmol/L, whereas all horses receiving 0.6 mg/kg maintained triglyceride concentrations ≤ 3.2 mmol/L. Overall, geometric mean serum triglyceride concentrations increased in a positive linear relationship with canagliflozin dose. In contrast, treatment efficacy, expressed as the percentage decrease in insulin AUC, increased with dosage in a characteristic dose–response pattern. Together, these findings suggest that increasing the canagliflozin dosage above 0.6 mg/kg is associated with only a modest additional improvement in treatment efficacy, but at the cost of a disproportionate and clinically relevant increase in serum triglyceride concentrations.

In humans, SGLT2 inhibitors shift substrate utilization from carbohydrates to lipids.²⁵ This metabolic adaptation is characterized by increased endogenous glucose production (EGP), accelerated lipolysis with enhanced release of free fatty acids into the circulation, increased fatty acid oxidation, and augmented ketogenesis.²⁶ These effects are largely mediated by decreased insulin secretion and a concomitant increase in glucagon release.^26,27 In contrast, administration of SGLT2 inhibitors in horses is commonly associated with the development of hypertriglyceridemia.^6,7,9 Beyond this observation, little is currently known about alterations in substrate utilization in horses receiving SGLT2 inhibitors. Importantly, the hypertriglyceridemia observed in equine patients appears to be mild in the majority of reported cases, is not associated with triglyceride-induced renal or hepatic dysfunction, and has not been linked to treatment-associated weight loss.^6,7,9

Several clinical studies in humans have demonstrated that SGLT2 inhibitors induce weight loss by creating a negative energy balance secondary to caloric loss through UGE.^18,28,29 Consistent with these studies, as well as previous studies in horses using SGLT2 inhibitors,^6,7 treatment with canagliflozin in our study resulted in significant weight loss after 4 weeks of treatment. No dose-dependent effect was observed, because neither relative nor absolute weight loss differed between horses receiving 0.6 or 1.2 mg/kg of canagliflozin.

In humans, counterregulatory mechanisms in response to weight loss with SGLT2 inhibitors have been suggested such as compensatory increases in appetite and caloric intake.^27,29 In our study, horse owners were instructed to maintain feeding and exercise at a constant level one month before and throughout the study period to have the weight loss determined as accurately as possible. Furthermore, no increase in appetite was reported in horses receiving active treatment. Nevertheless, such compensatory changes may occur with longer treatment durations. In humans with T2DM, weight loss plateaus after 26 weeks of treatment despite sustained UGE, potentially because of an adaptive increase in energy intake.¹⁹

In humans treated with SGLT2 inhibitors, osmotic diuresis leads to fluid losses, and decreased plasma volume contributes to the decrease in body weight.^17,30 In our study, there were no indications of decreased plasma volume, because PCV, serum protein concentrations, and plasma osmolality remained unchanged after canagliflozin treatment. However, changes in total body water or plasma volume were not directly measured. In humans treated with SGLT2 inhibitors, decreases in plasma volume are accompanied by increases in PCV and hemoglobin concentration,¹⁷ which were not observed in our study. Consequently, the weight loss observed after 4 weeks of canagliflozin treatment appears to be primarily attributable to a decrease in adipose tissue, although loss of lean tissue cannot be excluded. Nonetheless, definitive evidence for a primary decrease in adipose tissue is lacking, because direct measurements of fat mass were not performed in our study.³¹

Sodium-glucose cotransporter-2 inhibitors increase basal EGP in normoglycemic humans, and this amount approximates the amount glucose excreted in the urine.³² A similar increase in basal EGP may occur in horses, requiring glucogenic substrates such as amino acids, glycerol, and propionic acid for gluconeogenesis. Increased basal EGP therefore could contribute to loss of muscle tissue, and some proportion of the observed weight loss may reflect decreases in lean mass. Although treatment with SGLT2 inhibitors has not been associated with decreased muscle mass in humans,³³ the relevance of this finding in horses remains unknown.

Serum sodium concentrations were unchanged after 4 weeks of canagliflozin treatment. Mild decreases in serum potassium and chloride concentrations were observed in horses receiving canagliflozin at 1.2 mg/kg, but these changes were not considered clinically relevant. In contrast to previous reports,^7,15 horse owners did not report increased urine output during canagliflozin treatment. Because the study was double-blinded and horse owners were unaware of the specific drug used, expectation bias was minimized. It remains likely that urine volume increased over 24 h in treated horses, but not to an extent readily detectable to owners.

Our study had some limitations. Sample size was limited, and treatment duration was relatively short. Consequently, metabolic adaptations, including changes in triglyceride concentrations and body weight, may differ with long-term treatment. Horses receiving prolonged treatment may adapt by shifting substrate utilization from carbohydrates toward lipid metabolism, resulting in improved regulation of lipid homeostasis over time. Additionally, assessment of UGE and hepatic lipid content would have provided additional insight into treatment effects. Finally, the study population included multiple breeds rather than a limited number of predefined breeds, which may have introduced variability in metabolic responses.

Taken together, our findings demonstrate that canagliflozin effectively decreases the postprandial insulin response after both an OST and an FCT in ID horses. Treatment efficacy increased with higher canagliflozin dose, but the magnitude of the treatment effect appeared to plateau at the higher dosage. Hypertriglyceridemia represents a potential safety concern during SGLT2 inhibitor treatment, and our study provides evidence of a dose-dependent increase in plasma triglyceride concentrations associated with canagliflozin administration.

Abbreviations

BCS

body condition score

CI

confidence interval

CNS

cresty neck score

EGP

endogenous glucose production

FCT

forage-based feed challenge test

GLDH

glutamate dehydrogenase

GGT

gamma glutamyl transferase

Glucose AUC_0-300

area under the plasma glucose versus time curve during the OST or FCT

Glucose C_max

maximal glucose concentration during OST or FCT

HAL

hyperinsulinemia-associated laminitis

ID

insulin dysregulation

Insulin AUC_0-300

area under the plasma insulin versus time curve during OST or FCT

Insulin C_max

maximal insulin concentration during OST or FCT

LS means

least squares means

NSC

non-structural carbohydrates

OST

oral sugar test

PCV

packed cell volume

SD

standard deviation

SEM

standard error of the mean

SGLT2 inhibitor

sodium-glucose cotransporter 2 inhibitor

T2DM

type 2 diabetes mellitus

UGE

urinary glucose excretion

WSC

water-soluble carbohydrates

WSC-f

water-soluble carbohydrates minus fructans

Appendix

Materials and methods

Study design

This 4-week, multicenter, randomized, parallel-group, double-blind, placebo-controlled study evaluated the efficacy and safety of PO canagliflozin administered at dosages of 0.6 or 1.2 mg/kg in horses with insulin dysregulation (ID). The study was conducted at the University Animal Hospitals of the Swedish University of Agricultural Sciences and the Norwegian University of Life Sciences between January 2021 and July 2023.

Client-owned horses and ponies previously diagnosed with severe ID and referred to the University Animal Hospitals for further evaluation were eligible for enrollment. Written informed consent was obtained from all owners before inclusion. The study protocol was approved by the Ethics Committees in Sweden (Dnr: 5.8.18-09082-2020) and Norway (FOTS ID 123285).

Study population

The diagnosis of ID was based on a single-sample oral sugar test (OST) performed in the field by referring veterinarians. Horses received PO glucose syrup (Dan Sukker® Glykossirap, Nordic Sugar A/S, Copenhagen, Denmark; 0.2 mL/kg body weight), and blood samples were collected 60–90 minutes after administration. Insulin concentrations >45 μIU/mL were considered diagnostic for ID.⁶ Horses and ponies aged ≥4 years with insulin concentrations >100 μIU/mL on the OST were eligible for inclusion. Plasma insulin concentrations were analyzed using an equine insulin ELISA (Equine Insulin ELISA; Mercodia AB, Uppsala, Sweden). Horses were excluded from the study if they had an ongoing acute episode of laminitis based on clinical examination or had clinical evidence of pituitary pars intermedia dysfunction including assessment of EDTA plasma adrenocorticotropic hormone (ACTH) concentrations adjusted for season. Pregnant or lactating mares were excluded. Additional exclusion criteria included treatment with drugs and access to grass pasture for 1 month or less before enrollment and systemic disease other than ID. A flow diagram detailing enrollment and exclusions is presented in Figure 1.

Randomization and blinding

Eligible client-owned horses with ID were randomized in a 1:1:1 ratio using a web-based randomization service (www.randomizer.com) to receive once-daily PO administration of placebo or canagliflozin at a dosage of 0.6 mg/kg or 1.2 mg/kg (100 or 300 mg tablets; Invokana, Mundipharma AB, Gothenburg, Sweden). Treatments were administered between 8 and 9 p.m. Investigators, laboratory personnel, and horse owners were blinded to treatment allocation throughout the study period. Details regarding the formulation of canagliflozin tablets and placebo have been described previously.⁶

Experimental protocol

The horses arrived at the University Animal Hospitals in Sweden or Norway, for a 5-day visit to obtain baseline data and a second time after 24 days of treatment at home for a 5-day follow-up period similar to the initial baseline period but with double-blind treatment. On both clinical visits, the horses were acclimated to the environment at the University Animal Hospitals for 2 days. The horses were confined to individual stalls with daily access to a sand paddock and maintained on their regular diet divided into 4 rations a day. During the second visit, double-blind treatment was administered PO at 9 p.m. every day, one hour before the last feeding of the day.

On second day, for both clinical visits, BCS²⁰ and CNS²¹ were obtained and clinical examination was performed. An IV catheter (Intranule, 2.0 x 105 mm, Vygon France or Milacath®−extended use 1.7 x 75 mm, Mila, USA, depending on the size of the horse) was inserted aseptically under local anesthesia (EMLA, AstraZeneca AB, Södertälje, Sweden) into a jugular vein. All blood samples were collected from the jugular catheter after withdrawal of 5 mL of waste blood.

On both clinical visits, feed was withheld in the morning of the third day, and a 300-minute OST started at 7 a.m. Baseline blood samples for the OST were obtained within 10 minutes before PO administration of glucose syrup (0.5 mL/kg body weight; Dan Sukker Glykossirap®, Nordic Sugar A/S, Copenhagen, Denmark). Subsequent blood samples were collected at 15, 30, 60, 90, 120, 150, 180, 210, 240, 270 and 300 minutes after PO glucose administration.

On Day 4 of both visits, an FCT using the horses’ regular forage of hay or haylage was started at 7 a.m. Horses were fed 0.4 kg forage (dry matter basis) per 100 kg body weight. Forage samples were collected from part of the feed that was fed to the horses. Blood samples were collected immediately before and at 30, 60, 90, 120, 150, 180, 210, 240, 270 and 300 minutes after the ration was fed.

The experimental protocol was extended with an optional fifth day on both clinical visits for a second study evaluating the horses’ β-cell responsiveness to IV glucose during placebo or active treatment with canagliflozin (0.6 mg/kg or 1.2 mg/kg) using a 240-minute IV graded glucose infusion test.¹³ This study will be reported separately.

All blood samples collected on Days 3 and 4 were transferred into lithium heparin vacutainer tubes and centrifuged within 5 minutes of collection. Blood samples from the fourth day, before conducting the FCT, were analyzed for PCV using a microhematocrit analyzer or transferred to vacutainer tubes without additive and allowed to clot before centrifugation. Plasma or serum was harvested. Plasma was immediately frozen at -80°C until further analyses in Sweden. Serum aliquots were frozen at -80°C until further analyses in Sweden or analyzed for sodium, potassium, chloride, total protein, and triglyceride concentrationsd, and GLDH and GGT activity on sampling day using automated clinical chemistry analyzers (Architect c4000, Abbott Diagnostics, Abbott Park, USA and Attelica Solution, Cl Analyzer, Siemens Healthineers, Germany for samples obtained in Sweden and Norway, respectively). Plasma osmolality was analyzed using an osmometer (Micro-Osmometer, Löser Messtechnik, Germany). Plasma insulin concentrations were analyzed in duplicate using an equine-optimized ELISA (Equine Insulin ELISA; Mercodia AB, Uppsala, Sweden). Plasma glucose concentrations were analyzed using an automated clinical chemistry analyzer (YSI 2500 glucose/lactate analyzer; Yellow Springs Instruments, Yellow Springs, USA). All forage samples were frozen and stored at −20 °C until chemical analysis. Concentrations of free glucose, free fructose, sucrose, and fructans were determined using an enzymatic spectrophotometric method.³⁴ Total water-soluble carbohydrate (WSC) content was calculated as the sum of free glucose, free fructose, sucrose, and fructans. Dry matter content was estimated as previously described.³⁵

Outcomes

Primary endpoints were differences in area under the plasma insulin vs time curve (insulin AUC_0-300) and the maximal insulin concentration (insulin C_max) between treatments during an OST and an FCT. The secondary endpoints were the corresponding values for area under the plasma glucose vs time curve (glucose AUC_0-300) and the maximal glucose concentration (glucose C_max). Additional secondary endpoints were differences in body weight, BCS, CNS, fasting plasma glucose and insulin concentrations, serum triglyceride and electrolyte concentrations, liver enzyme activities, osmolality, PCV, total serum protein concentrations and horse owners’ averaged rating scores between treatments.

Horse owners’ rating of the horses’ attitude, appetite and urine volume

Horse owners were asked to assess their horses’ overall attitude, appetite, and urine volume on a weekly basis during each treatment week (weeks 1–3), using the week preceding study inclusion as the reference period. Assessments were collected using a short questionnaire. Each variable was rated using a predefined five-point ordinal scale: very increased, mildly increased, unchanged, mildly decreased, or very decreased. For data analysis, these categories were converted to a numerical scale by the investigators (very increased = 2, mildly increased = 1, unchanged = 0, mildly decreased = −1, very decreased = −2). An overall score for each variable was calculated as the mean of the scores for the three weeks.

Statistical methods

A sample size of 12 horses per treatment group was estimated to provide > 80% power to detect differences in primary outcomes using repeated measures analysis of variance (ANOVA) with three groups and a moderate effect size of 0.28 and an alpha of 0.05. A priori calculations were performed using G*Power 3.1 software (Heinrich Heine Universität, Düsseldorf, Germany). A maximum of 2 horses per group that did not complete the study was expected. Thus, 42 horses were enrolled and randomly assigned to the three treatment groups.

The insulin AUC_0-300 and glucose AUC_0-300 were calculated using the trapezoidal rule. Data for baseline clinical characteristics were summarized and displayed as mean ± SD, median (IQR) or count (n).

Plasma insulin or glucose responses over time after treatment were analyzed using a mixed effects model with treatment (placebo, canagliflozin 0.6 or 1.2 mg/kg), sampling time (0 – 300 min) and treatment by sampling time interaction as categorical fixed effects. Center (Sweden or Norway) and horse nested within center were included as random effects. Corresponding pre-treatment baseline results for insulin AUC_0-300 or glucose AUC_0-300 were included as continuous covariates. The initial model included several interaction terms with the covariate. If not significant, they were excluded from the model in a stepwise manner. Insulin data were log-transformed before analysis to meet assumptions of normality and homogeneity of variance among residuals.

Primary and secondary endpoints were analyzed using analysis of linear mixed effects model with treatment (placebo, canagliflozin 0.6 or 1.2 mg/kg) included as a fixed effect and the corresponding baseline value as a continuous covariate to adjust for baseline values. Center (Sweden or Norway) was included as a random effect. The initial model included an interaction between treatment and the covariate. If not significant, the interaction term was removed from the model and data were analyzed using a linear mixed effects model with equal slopes. If the interaction was significant, treatment effects were compared at three covariate levels (mean - 1 SD, mean and mean + 1 SD). Residual vs. predicted plots and Q-Q plots, were performed to evaluate assumptions of homoscedasticity and normality. When necessary, log transformation was applied to address violations of assumptions. Furthermore, each horse’s WSC or WSC-f contents were included as covariates in the models to assess whether individual horse-level variation in WSC or WSC-f influenced glucose AUC_0-300 or insulin AUC_0-300 during the FCT. Neither covariate was significant (P ≥ 0.10) and they therefore were excluded from the final models.

Adjusted means are reported as least square (LS) means ± SEM and data with non-normally distributed residuals are reported as geometric LS means and 95% CI after back-transformation to their original scale. Comparison between time points within treatments or between treatments at different levels of the covariate were performed using Tukey-Kramer post hoc test. Least square mean differences are reported with 95% CI whereas LS mean differences for logarithmic data are reported as ratios with 95% CI.⁶ Consistent with the CONSORT 2010 extension for reporting multi-arm parallel-group randomized trials with different doses, reporting a dose-response relationship for intervention outcomes may be more informative than conducting pairwise comparisons between treatment groups.³⁶ Accordingly, dose-response relationships were evaluated for the percentage change in insulin AUC_0-300 from baseline to post-treatment during the OST and FCT across all groups (placebo, canagliflozin 0.6 mg/kg and 1.2 mg/kg). Sigmoidal dose-response curves were fitted to the administered canagliflozin doses using commercial software (GraphPad Prism version 10.6.1 for Windows, GraphPad Software, Boston, MA).

Differences between groups in BCS and CNS pre-test and post-test scores, as well as treatment differences in owner-reported ratings of overall attitude, appetite, and urine production, were analyzed using the Kruskal–Wallis test. Values where P < .05 were considered statistically significant. All statistical analyses were performed using JMP Pro version 19.0.0 (SAS Institute Inc, Cary, NC).

Author contributions

Johan Bröjer (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Visualization, Writing—original draft, Writing—review & editing), Siv Hanche-Olsen (Conceptualization, Funding acquisition, Investigation, Project administration, Writing—review & editing), Constanze Fintl (Conceptualization, Investigation, Writing—review & editing), Elin Svonni (Conceptualization, Investigation, Writing—review & editing), Ingunn Risnes Hellings (Conceptualization, Investigation, Writing—review & editing), Cecilia Elisabeth Müller (Conceptualization, Investigation, Writing—review & editing), Sanna Lindåse (Conceptualization, Funding acquisition, Investigation, Project administration, Writing—review & editing)

Conflicts of interest

Authors declare no conflicts of interest.

Funding

The study was funded by the Swedish-Norwegian Foundation for Equine Research, H-19-47-479.

Off-label antimicrobial declaration

Authors declare no off-label use of antimicrobials.

Institutional animal care and use committee or other approval declaration

The study was approved by the Ethical Committee for Animal Experiments, Uppsala, Sweden and the Norwegian Food and Safety Authority (Dnr: 5.8.18-09082-2020 and FOTS ID 23285, respectively). Written informed consent was obtained from all owners prior to inclusion.

Human ethics approval declaration

Authors declare human ethics approval was not needed.