Study design
The Steno-2 study was initiated in 1993, enrolling 160 Danish patients of European descent with type 2 diabetes mellitus and microalbuminuria [11–13]. The design and methods for patient inclusion, randomisation, treatment and initial follow-up have been reported in detail previously [11–13].
Briefly, patients were randomised 1:1, stratified in blocks by sex, age, known diabetes duration and urinary albumin excretion rate (<100 mg/day vs >100 mg/day) using sealed envelopes, to receive either conventional multifactorial treatment with treatment goals at all times according to existing national guidelines or to receive intensified, multifactorial treatment targeting co-existing risk factors for late diabetic complications. Eighty patients were assigned to each group. Conventional-therapy patients were followed by their general practitioner, but at all times had the opportunity of being referred to specialist treatment. The treatment in the intensive-therapy group at a specialised diabetes clinic was target-driven with stepwise implementation of both behavioural and pharmacological treatment following a structured approach. Treatment goals for the groups are shown in Section 1 of the electronic supplementary material (ESM) in ESM Table 1.
Patients completed up to six study visits at Steno Diabetes Center, namely at baseline and after an average of 1.9, 3.8, 7.8 and 13.3 years, respectively, and at the termination visit after 21.2 years.
Study population
Of the 160 patients originally enrolled in the Steno-2 study, 130 completed the interventional part of the study. Patient flow is depicted in Fig. 1. All patients completing the interventional part of the study provided written informed consent to continue participation in the observational post-trial follow-up. In the entire follow-up period, one patient in the conventional group and none in the intensive group were lost to follow-up. The loss was due to emigration and data collection on that patient terminated in 2007. All other patients were followed to death or study termination.
Due to ethical concerns, following the marked risk reductions found with intensive therapy after 7.8 years, all patients were subsequently offered intensified multifactorial treatment according to the original protocol using targets resembling current EASD/ESC and ADA guidelines [14, 15].
The protocol for the follow-up trial was in accordance with the declaration of Helsinki and approved by the local ethics committee (Ethics committee, Capital Region of Denmark; protocol ID number: H-KA-99035-GS, add. 41104) and by the Danish Data Protection Agency (J.Nr. 2015-41-4042).
Procedures, measurements and endpoints
The follow-up investigations were initiated to investigate the difference in median time to 50% mortality in each of the two original treatment groups. This follow-up study was done as an amendment to the original study protocol.
The primary endpoint for the present study was the difference between the two treatment groups in the survival time after randomisation without and with cardiovascular disease (CVD). The secondary endpoint was defined as a composite of cardiovascular events (time to incident CVD and number of cardiovascular events) as well as mortality and CVD rates. The tertiary endpoints were rate of incident diabetic nephropathy, rate of end-stage renal disease and the development or progression of diabetic retinopathy and neuropathy.
For the follow-up examinations, all living patients were invited to an outpatient clinic visit assessing anthropometric, biochemical and physiological variables as well as detailed status on micro- and macrovascular complications. The examinations were conducted by a single laboratory technician who was unaware of the original randomisation assignments.
Supplementary clinical evaluations were collected from nationwide registries when not obtainable at the study visit. All measurement methods were in accordance with previous follow-up protocols [13].
For endpoint assessment, all patients were tracked by their unique, static Central Personal Registration number using the Danish Civil Registration System. All hospital contacts diagnosed to possibly fulfil the pre-specified criteria (see ESM Section 2) or relating to the criteria (including ICD-10 [www.who.int/classifications/icd/en/] observation-diagnoses [DZ0.3x] and symptom-diagnoses [e.g. DR0.74, chest pain]) were extracted from the Danish National Patient Registry and used as cases for the endpoint committee. Complete health records for cases were obtained from nationwide electronic health records and adjudicated for endpoints by an external committee masked for patients’ original treatment allocation. The registries have proven highly valid and complete [16, 17]. Hospitals report mandatorily to the registries and the reported data on diagnoses and procedures translates directly to the funding of the hospitals. The data provide information from 1978 onwards. Hospitals mandatorily keep individual patient records related to hospitalisations and outpatient treatment for at least 10 years after the last record.
Cardiovascular events were defined as a composite of death from cardiovascular causes, myocardial infarction, stroke, amputation due to ischaemia and cardiac or peripheral revascularisation. If an event occurred as a direct consequence of, or as an adverse event related to another, immediately prior event (e.g. percutaneous transluminary coronary angioplasty immediately after acute myocardial infarction; see ESM section 2.1), this latter event was not considered in the analyses of separate endpoints. Self-reported cardiovascular events prior to randomisation were not considered. Microvascular complications were defined, measured and evaluated as described in ESM Section 3.
Statistical analyses
All statistical analyses were conducted using the intention-to-treat principle. Adherence to the treatment was assessed at study visits in both groups based on interviews with patients.
Overall cumulative survival, as well as CVD-free survival, by time since randomisation was calculated for the two treatment groups by the Kaplan–Meier estimator. Median survival time from randomisation was compared between the treatment groups. CIs for the medians were calculated using bootstrapping. We specifically chose difference in median survival time between groups to make a clear statement in easily understandable terms in addition to the more extensively used, but more difficult to interpret, HR and RR reduction.
Proportional hazards Poisson models were fitted for mortality (both from cardiovascular causes and from other causes) and cardiovascular events, taking treatment group, age, sex and current CVD status into account, using smooth underlying hazards (time since baseline). Current CVD status was included in models for all-cause mortality as 0, 1, 2 or 3 or more cardiovascular events after randomisation. Thus, we modelled both death and (extra) CVD events as outcome, and the models were used to compute the average years of life lived and the years of life free of CVD during the study period and thereby the average number of years gained by intensive treatment. Models were checked for proportional hazards between randomisation groups by likelihood-ratio tests. Diabetes duration at baseline, as well as the interaction with treatment allocation, was included in a model to assess the effect of diabetes duration on mortality.
Except for end-stage renal disease (where exact date of transplantation or first dialysis treatment was known), the status of microvascular complications was only assessed at study visits, and the exact event date is therefore unknown. For analysis purposes a random date between the last day without and first day with the specified complication was imputed as the event date. When the progression between states jumped more than one category (e.g. from EURODIAB-score 2 to 5 between two observation points), random dates between the two observations for the steps were generated and used in analyses of transition rates. Sensitivity analyses were performed by repeating the random allocation of dates.
Transition rates between states and group comparison were analysed as for cardiovascular events. For each of the four types of outcome (retinopathy, autonomic neuropathy, peripheral neuropathy and albuminuria) we estimated the rates of transition between states of increasing severity and used these to construct plots showing the fraction of patients in each state at different times after randomisation.
A fuller account of statistical approaches and detailing of analyses can be found in ESM Sections 4–7. For the anthropometric, biochemical and physiological variables, the t test, Mann–Whitney U test and χ2 statistics were applied for the comparison of means, medians and proportions, respectively. In the results section all estimates are followed by 95% CIs in parentheses. Analyses were conducted using R, version 3.3.0, using the Epi package version 2.5 or Stata/IC version 14.1 for Windows (StataCorp LP, College Station, TX, USA).