Trial design
The Nephropathy In Diabetes type 2 (NID-2) study is an open-label cluster randomized clinical trial in a population referred to 14 Italian diabetology clinics [10]. To maximize the contrast between the two approaches, we randomized clinics rather than patients. Indeed, in the latter modality of randomization similarities between the two interventions are expected to ensue over the long-term. Centres were randomly assigned to either MT therapy or SoC. A questionnaire ascertained that all participating physicians were well aware of the guidelines on T2DM management published at the time of the study [11,12,13,14].
All MACEs diagnoses were performed in each patient according to the diagnostic criteria defined by the international standards of care guidelines [15,16,17]. MACEs were evaluated by cardiologist blinded to the study arm (MT or SoC), either belonging to the same Centres or to hospitals where patients were referred for acute events.
Participants and procedures
We considered eligible T2DM patients with age ≥40 years, persistent albuminuria ≥30 mg/24 h in at least two of three 24 h-urine collections in the last 6 months), severe DR (according to the Wilkinson et al.) [18], diabetes onset at age > 30 years, absence of neoplastic/psychiatric diseases and follow-up at the centre ≥ 12 months. Exclusion criteria were previous MI or stroke, severe hepatic or cardiac failure.
Patients were enrolled between October 2005 and October 2008. The intervention phase was scheduled for a period of four years, and it was completed in December 2011. Then, patients were followed until May 2019 to achieve the number of events needed for the primary outcome.
The protocol was approved by the ethics committee of University of Campania “Luigi Vanvitelli” (clinicaltrials.gov: NCT00535925) and is in accordance with the 1976 Declaration of Helsinki and its later amendments. All participants signed their informed consent.
Randomization
All patients enrolled in each clinic were randomized, according to a cluster-randomization procedure, in two arms, MT and SoC. Randomization of centres was stratified based on their size, in order to reduce difference in the number of patients allocated to the two treatment arms.
The intensified therapy group was initiated to the therapeutic regimen summarized below and detailed and in Additional file 1: Appendix S1. Patients assigned to the conventional therapy group followed the therapy usually administered at their outpatient clinic; hence, they could receive any therapeutic change considered appropriate by their caregiver, under the respect of the good clinical practice rules.
Targets
In either arm participating physicians were required to adhere to guideline-based clinical targets recommended at the time of study initiation: [11,12,13,14] (a) systolic blood pressure (SBP) < 130 mmHg, (b) diastolic blood pressure (DBP) < 80 mmHg, (c) glycated haemoglobin (HbA1c) < 7%, (d) fasting serum LDL cholesterol < 100 mg/dL, (e) fasting serum HDL cholesterol > 40/50 mg/dL (for men/women, respectively), and (f) fasting total serum cholesterol < 175 mg/dL.
Study arms
In SoC group, the subjects received the therapy usually administered at their diabetic outpatient for the management of blood pressure, glycaemic and lipid control, and antiplatelet treatment. During the study, these patients received all therapeutic modifications considered appropriate by their physician, in the respect of the good clinical practice.
In MT group, the patients were treated with pre-specified algorithm for management of hypertension, glycol-metabolic control and dyslipidemia, including non-pharmacological and pharmacological treatment, as detailed in Additional file 1: Appendix 1. Briefly, specific recommendation for physical activity and low sodium diet were provided to patients in written form. In addition, renin–angiotensin system blockade was implemented by initial association of ACEi and ARBs with a strict monitoring of GFR and serum potassium, followed by stepwise addition of other anti-hypertensive drug classes. They received low-dose aspirin as primary prevention, unless contraindicated or not tolerated. Statin was added if non-pharmacological therapy was ineffective in reaching the target.
All patients, regardless of the study group, underwent control visits at their diabetes centre every six months to monitor laboratory and clinical parameters and compliance to therapies and lifestyle hints. During each visit, investigators carefully monitored the occurrence of adverse events. In MT group, additional visits could be planned if one or more risk factors resulted out of target. At each visit, adherence to pharmacological protocol as well as to lifestyle recommendations (see Additional file 1: Appendix S1) was strictly monitored and strengthened.
eGFR was estimated using the CKD-EPI equation and, since creatinine was not standardized, we reduced creatinine values by 5% [19].
Outcomes
Primary endpoint was a composite of fatal and non-fatal MACEs, including cardiovascular mortality, non-fatal MI (documented instrumentally and/or enzymatically), non-fatal stroke, coronary-artery by-pass, revascularization procedures (PTCA) and lower limbs major amputation, whichever occurred first. In both arms, all endpoints were captured and recorded by investigators in an electronic Case Report Form (CRF) at each visit.
Since the planned number of events was not reached during the initial 4-year time frame (interventional phase), incidence of the primary end point was assessed throughout the follow-up phase, that in the original study design was planned to assess the durability of effects of the intensified treatment.
During this extension phase, following the end of intervention, all patients enrolled in both arms were treated by their own physicians according to the good clinical practice.
As secondary endpoints, we considered each single component of primary endpoint, and all-cause death at the end of the follow-up phase, as well as MACEs and the achievement of BP, HbA1c and total, HDL and LDL cholesterol goals at the end of intervention phase.
Sample size
Study is powered to detect a Hazard Ratio (HR) of 0.67 in the comparison of the two groups, with an 80% power and a two-sided type I error of 5%, assuming an intraclass correlation coefficient of 0.01. For this purpose, with a sample size of about 420 patients, 14 overall clusters, an average of 30 subjects per cluster, and an expected surviving proportion of 30% at ten years in the SoC group, we determined a number of events needed of 258.
Statistical analysis
All statistical analyses were performed after the end of the follow-up and the achievement of number of events needed for the analysis of the primary outcome. A statistical analysis plan was prepared before the central database was locked for final data extraction and analysis. Categorical data were expressed as number and percentage, while continuous variables as either median and interquartile range or mean and standard deviation, based on their distribution assessed by the Shapiro–Wilk test. In order to check for imbalance in cluster randomization, we compared variables at baseline by using the method proposed by Leyrat et al. [20] Standardized differences (SDiff) were calculated for continuous and dichotomous variable. P-values to take into account clustering were computed by generalized estimating equations (GEE) model with cluster as group variable [21]. Distribution of dependent variable and link function was used as appropriate (gaussian and identity for continuous variable, binomial and logit for dichotomous variable). Comparison of groups at end of intervention was performed applying the same methodology, further adjusting for baseline values as covariate.
Criteria on SDiff cut-offs reported by Leyrat et al. [20] were used to establish covariates imbalanced at baseline. Moreover, to evaluate a global imbalance, c-statistic was calculated performing a logistic model with treatment arm as dependent variable and selected baseline variables as covariate.
Median follow-up time was calculated by the inverse Kaplan–Meier procedure. The primary endpoint was analysed according to the intention-to-treat principle, with event curves for the time-to-first event based on Kaplan–Meier analysis. Cox regression model was used to calculate HR and 95% Confidence Interval (CI). Due to the cluster randomized study design, a Cox shared-frailty model was fitted. Across centres, the frailties are assumed to be gamma-distributed latent random effects affecting the hazard multiplicatively. In the univariate analysis, only treatment group was included as covariate. In the multivariable analyses, depending on imbalance detection of each variable (Leyrat method), association with the outcome of interest and evidence from the literature, we adjusted the Cox regression models for age, sex, SBP, haemoglobin, eGFR, albuminuria, HbA1c, total cholesterol, triglycerides (log-scaled), statins and antiplatelets therapy at baseline to reduce risk of bias. Data were analysed using STATA 16.0 software (StataCorp. 2019. College Station, TX: StataCorp LLC).