Effects of ertugliflozin on kidney composite outcomes, renal function and albuminuria in patients with type 2 diabetes mellitus: an analysis from the randomised VERTIS CV trial

Aims/hypothesis In previous work, we reported the HR for the risk (95% CI) of the secondary kidney composite endpoint (time to first event of doubling of serum creatinine from baseline, renal dialysis/transplant or renal death) with ertugliflozin compared with placebo as 0.81 (0.63, 1.04). The effect of ertugliflozin on exploratory kidney-related outcomes was evaluated using data from the eValuation of ERTugliflozin effIcacy and Safety CardioVascular outcomes (VERTIS CV) trial (NCT01986881). Methods Individuals with type 2 diabetes mellitus and established atherosclerotic CVD were randomised to receive ertugliflozin 5 mg or 15 mg (observations from both doses were pooled), or matching placebo, added on to existing treatment. The kidney composite outcome in VERTIS CV (reported previously) was time to first event of doubling of serum creatinine from baseline, renal dialysis/transplant or renal death. The pre-specified exploratory composite outcome replaced doubling of serum creatinine with sustained 40% decrease from baseline in eGFR. In addition, the impact of ertugliflozin on urinary albumin/creatinine ratio (UACR) and eGFR over time was assessed. Results A total of 8246 individuals were randomised and followed for a mean of 3.5 years. The exploratory kidney composite outcome of sustained 40% reduction from baseline in eGFR, chronic kidney dialysis/transplant or renal death occurred at a lower event rate (events per 1000 person-years) in the ertugliflozin group than with the placebo group (6.0 vs 9.0); the HR (95% CI) was 0.66 (0.50, 0.88). At 60 months, in the ertugliflozin group, placebo-corrected changes from baseline (95% CIs) in UACR and eGFR were −16.2% (−23.9, −7.6) and 2.6 ml min−1 [1.73 m]−2 (1.5, 3.6), respectively. Ertugliflozin was associated with a consistent decrease in UACR and attenuation of eGFR decline across subgroups, with a suggested larger effect observed in the macroalbuminuria and Kidney Disease: Improving Global Outcomes in Chronic Kidney Disease (KDIGO CKD) high/very high-risk subgroups. Conclusions/interpretation Among individuals with type 2 diabetes and atherosclerotic CVD, ertugliflozin reduced the risk for the pre-specified exploratory composite renal endpoint and was associated with preservation of eGFR and reduced UACR. Trial registration ClinicalTrials.gov NCT01986881 Graphical abstract Supplementary Information The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-021-05407-5.


ESM Statistical Analysis Plan
Publication for Key Kidney Outcomes VERTIS-CV -Protocol MK8835-004-01/B1521021   Values, Section 6, no event has occurred in time-to-event endpoints, subjects will be censored at the time of "the last follow-up visit" to replace "discontinuation of study treatment". This was a typo in the previous version.

Introduction
This SAP summarizes the analysis plan for key kidney outcomes publication using data from patients treated with either ertugliflozin or placebo in the VERTIS-CV cardiovascular outcome trial, protocol MK-8835-004-01/B1521021. Note analyses that are already included in other SAP from this study will not be repeated.

Objectives
The objectives, which will be assessed by baseline renal function category, are: • To assess demographic and baseline characteristics of subjects randomized to ertugliflozin as compared with placebo

Analysis Sets
The full analysis set (FAS) defined in Section 5.1 of the Non-CV SAP will be used for the analysis of eGFR, UACR, and renal related endpoints (data collected during the treatment period).
For AE endpoints, the All Subjects as Treated (ASaT) analysis set defined in Section 5.3 of the Non-CV SAP will be used.
For analyses that use the constrained longitudinal data analysis (cLDA) model in Section 5, subjects require at least one baseline measurement or post-baseline measurement to be included in the model.
The CV intention-to-treat (CV ITT) analysis set will be used for renal composite endpoints, their individual components, and other time-to-event endpoints. This analysis set includes all randomized subjects. CV ITT will also be used to summarize baseline characteristics.
Data obtained after the initiation of glycemic rescue therapy or after bariatric surgery will be included. Data obtained more than 2 days after the last dose of study medication were excluded for the mean change from baseline of the eGFR and UACR analyses by timepoint. Sustained is defined as the occurrence of a value that meets the cut-off criteria (i.e., 2xSCr, or 40% decrease) which is followed, more than 30 days later, by a subsequent value that also meets the cut-off criteria. Values that meet the cut-off criteria do not need to be proximate (there can be an intervening value that does not meet the criterion).

Categorical Variables and Analysis Endpoints
• Change from Baseline in eGFR (mL/min/1.73m 2 ) with MDRD at Week 6, Week 18, Week 52 and yearly thereafter • Change from Baseline in eGFR (mL/min/1.73m 2 ) with CKD-EPI at Week 6, Week 18, Week 52 and yearly thereafter • Change from baseline in UACR (mg/g) 2 at Week 18, 52 and yearly thereafter 2 Serum creatinine was collected 6 times during the first year and three times per year thereafter, urine samples for the UACR were collected twice during the first year and annually thereafter. For the assessments of events related to changes in serum creatinine local laboratory results were also used.

Statistical Methods
The time-to-event endpoints will be analyzed using a stratified Cox proportional hazards (CPH) model including treatment group as a covariate. Cohort (defined in Section 6.2 of CV SAP) will be included as a stratification factor. For baseline subgroups (baseline eGFR category, KDIGO CKD risk categories, and UACR categories) analysis, each subgroup analysis model will include terms for treatment (categorical), subgroup (categorical), and treatment by subgroup interaction in the CPH model. The point estimates and two-sided 95% confidence intervals for the hazard ratio will be calculated from the CPH model. For subgroup analysis, the overall p-value of the interaction term will be presented.
Mean changes from baseline in eGFR (MDRD), eGFR (CKD-EPI), and UACR endpoints over time will be estimated using the constrained longitudinal data analysis (cLDA) model for the overall population. The model will contain fixed effects for treatment, time, treatment by time interaction, and baseline HbA1c, baseline systolic blood pressure. In the model, time will be treated as a categorical variable so that no restriction will be imposed on the trajectory of the means over time. The treatment difference in terms of mean change from baseline to a given time point will be estimated and presented. An unstructured covariance matrix will be used to model the correlation among repeated measurements. If covariance structure UN does not converge, TOEP will be used. If data do not converge due to the less samples in the later timepoints (e.g., greater than 2 years), then the later points will not be included in the model and be presented by descriptive statistics.
Due to the non-normal distribution of UACR, log transformation of UACR data will be performed.
The geometric means will be provided by timepoint in the data summary. Adjusted mean percentage change (derived from exponentiation of adjusted estimates) with 95% confidence intervals will be presented for each treatment. The difference between ertugliflozin treatment and placebo in mean percentage change in UACR from baseline will also be estimated.
For the subgroups: 3 eGFR categories, KDIGO CKD risk categories, and UACR categories, a repeated measures ANCOVA (RMANCOVA) method will be used. The RMANCOVA model will adjust for baseline value of response variable, baseline of HbA1c, treatment, subgroup, and treatment-by-subgroup interaction, and treatment-by-subgroup-by-time interaction. Time is treated as a categorical variable. An unstructured covariance matrix will be used to model the correlation among repeated measurements. For subgroup analyses based on factors that are already in the main model, the respective term will appear in the model only once.
Baseline demographics and disease characteristics will be summarized with the descriptive statistics by treatment (ertugliflozin 5 mg, 15 mg, pooled ertugliflozin, and placebo) for the overall population and by the following subgroups: 2 eGFR categories, 3 eGFR categories, KDIGO CKD risk categories, and UACR categories.
Section 8 contains a table which summarizes the overall analysis plan and compares to the SAP of study protocol.

Handling of Missing Values
No missing imputation of missing safety endpoints will be done. For analysis of UACR and eGFR, endpoints, the cLDA method uses a full likelihood model that does not require imputation of missing data and produces unbiased inference under the missingness mechanisms MCAR (missing completely at random) and MAR (missing at random). For time-to-event endpoints, if no event has occurred, subjects will be censored at the time of the last follow-up visit.

Graphical Summaries
Graphical summaries will include Kaplan-Meier curves of the time-to-event for each treatment group (pooled ertugliflozin doses vs. placebo) or separately by dose (ertugliflozin 5 mg, ertugliflozin 15 mg, and placebo) for the overall population and for subgroups.
The least squared mean (LSM) of change from baseline in eGFR (MDRD), eGFR (CKD-EPI) over time will be presented by treatment group for the overall population and for the following subgroups: 3 eGFR categories, UACR categories, and KDIGO CKD risk categories individually.
Additionally, adjusted mean percentage change (derived from exponentiation of adjusted estimates) in UACR over time will be presented by treatment group for the overall population and for the following subgroups: 3 eGFR categories, UACR categories, and KDIGO CKD risk categories individually.
10 | P a g e