Study population
This post hoc, exploratory analysis was conducted using a subset of stored plasma samples from the CANVAS trial (ClinicalTrials.gov identifier: NCT01032629), which was a randomized, multicenter, double-blind, parallel, PBO-controlled study that evaluated the efficacy and safety of CANA 100 and 300 mg (1:1:1 allocation). The inclusion and exclusion criteria have been described in previous publications [22]. In short, inclusion criteria were diagnosis of type 2 diabetes and age ≥ 30 years with a history of a CV event or age ≥ 50 years with a high risk of CV events, inadequate diabetes control (as defined by glycated hemoglobin [HbA1c] ≥ 7.0% to ≤ 10.5% at screening) and either (1) not currently on diabetes drug therapy or (2) on therapy with any approved class of diabetes drugs. Exclusion criteria were history of diabetic ketoacidosis, type 1 diabetes, pancreas or beta-cell transplantation, diabetes secondary to pancreatitis or pancreatectomy, history of ≥ 1 severe hypoglycemic episode within 6 months before screening. Median (range) follow-up time was 6.1 (6.1–7.1) years. A flow diagram for the used patient population is shown in Fig. 1.
Measurement of ETP by enzyme-linked immunosorbent assay (ELISA)
Baseline plasma samples from 3531 patients and urine from 3423 patients were available for this study. At year 3, plasma and urine samples were available for 2178 (61.7%) and 2070 (60.5%), respectively. ETP was measured in plasma (P-ETP) and urine (U-ETP) using a previously described competitive enzyme-linked immunosorbent assay, namely PRO-C6, developed by Nordic Bioscience, Denmark [21]. The unit for P-ETP is ng/mL, whereas the unit for U-ETP is ng/mmol creatinine.
Outcomes
The investigated outcomes have been described previously [22]. The outcomes investigated included hospitalization due to heart failure (HF), cardiovascular death (CVD), a composite of HF and CVD (HFCVD), and all-cause mortality. The first composite kidney endpoint (KCOM1) was defined as a sustained 40% decline of eGFR, end-stage kidney disease defined as an eGFR < 15 mL/min/1.73 m2, need for dialysis or kidney transplantation, or kidney death. The second kidney composite endpoint, KCOM2, included KCOM1 and CVD, and the third kidney composite endpoint, KCOM3, included KCOM1 and conversion to severely increased albuminuria.
Statistical analyses
Given the similar treatment effect from both canagliflozin treatment regimens (100 and 300 mg daily), data from patients receiving either of the 2 doses were combined in this analysis.
Associations between ETP and relevant variables were investigated using non-parametric Spearman-rank correlations.
To investigate which variables were independently associated with ETP, we used multivariable regression analysis with P-ETP as the dependent variable and age, sex, body mass index (BMI), diabetes duration, HbA1c, low density lipoprotein (LDL), high density lipoprotein (HDL), diastolic and systolic blood pressure (BP), eGFR, albumin:creatinine ratio (ACR), AST, ALT, and NTproBNP as the independent variables. The association of these variables with P-ETP were reported as rpartial.
A Mann–Whitney test was used to assess whether levels of P-ETP were elevated in patients with prior history of HF and prior history of CV disease. To assess whether P-ETP was higher in these patients independent of clinical covariates, a multivariate logistic regression analysis was performed with P-ETP, and the clinical covariates age, sex, BMI, diabetes duration, HbA1c, LDL, HDL, diastolic and systolic BP, eGFR, and ACR, AST, ALT, and NTproBNP.
Unadjusted Kaplan–Meier curves were used to visualize the association of tertiles of P-ETP and U-ETP with the investigated end-points.
Absolute change from baseline in plasma and urine levels of ETP (Nordic Bioscience, Herlev, Denmark) were analyzed in patients with data available at both baseline and Week 156 (n = 2178 and n = 2070, respectively). The absolute change for P-ETP (ΔP-ETP) and U-ETP (ΔU-ETP) was calculated as year 3 minus baseline ETP in plasma and urine, respectively. The unit for ΔP-ETP was ng/mL, whereas the unit for ΔU-ETP was ng/mmol.
We performed both univariable and multivariable Cox proportional hazard regression analysis. Treatment was included as an interaction term in all outcome analysis to assess whether there were differences in biomarker performance between placebo and canagliflozin treated patients. If no differences were observed, results from the combined treated arms were shown.
Backward selection using the Akaike information criterion (AIC) was employed to identify variables retained in a Cox proportional hazards model for each outcome. The variable input in the model build were age, sex, BMI, systolic BP, diastolic BP, HbA1c, diabetes duration, LDL, ACR, eGFR, AST, ALT, NTproBNP, prior history of HF, prior history of CV disease, treatment, smoking, and either P-ETP or U-ETP. To assess the robustness of the selected model, backwards selection was performed in 500 bootstrap iterations with random resampling. Hazard ratios (HRs) for ETP were presented adjusted for the clinical parameters included in the final model. The HRs for ETP at baseline were reported per doubling in biomarker levels (i.e., log2–transformation). As negative values were seen for the ΔP-ETP and ΔU-ETP, HRs for these analyses are presented per increments of 1 ng/mL and ng/mmol, respectively.
An ANOVA model with Tukey’s multiple comparisons test was used to determine overall differences in ΔP-ETP and ΔU-ETP levels between albuminuria stages.
Parametric analysis was performed for data having a Gaussian distribution (normally distributed) and non-parametric analysis was performed for data having a non-Gaussian distribution (non-normally distributed). Distribution of the data was assessed by the D’Agostino and Pearson normality test.
All 2-tailed p values of < 0.05 were considered significant. Statistical analyses were made by the MedCalc statistical software (MedCalc, Belgium), SAS software (version 9.4, SAS Institute, Cary, NC, USA) or R studio (Version 1.4.1106), and visualized using GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA).