Inclusion and exclusion criteria
Complete details of the rationale and design of the MAPEC study are described in previous publications [5, 6, 11, 12, 20–22, 30]. In summary, the sample consisted of Spanish individuals ≥18 years of age who adhered to a routine of daytime activity and night-time sleep. Inclusion criteria required participants to be either normotensive, untreated hypertensive or resistant to treatment  when ingesting all their prescribed BP-lowering medications upon awakening. Exclusion criteria were pregnancy, history of drug/alcohol abuse, night/shift-work employment, acquired immunodeficiency syndrome (AIDS), type 1 diabetes, secondary hypertension, CVD disorders (unstable angina pectoris, heart failure, life-threatening arrhythmia, nephropathy and grade III–IV retinopathy), intolerance to ABPM, and inability to communicate and comply with all study requirements. This prospective single-centre study (ClinicalTrial.gov registration no. NCT00295542) was approved by the state Ethics Committee of Clinical Research. All participants gave written informed consent.
Participants and diagnostic criteria
Between 2000 and 2007, we recruited 3,612 persons fulfilling the inclusion/exclusion criteria, with 3,344 (1,718 men/1,626 women, 52.6 ± 14.5 [mean ± SD] years of age) providing all required information for study. The remaining 268 individuals were excluded due to inadequate ABPM sampling at baseline and non-consent for additional ABPM evaluations. At the time of recruitment, 688 patients already had a diagnosis of type 2 diabetes and, therefore, were also excluded from analyses. The final evaluated population for the hypotheses tested herein thus consisted of 2,656 individuals (644 normotensive and 2,012 hypertensive according to the ABPM criteria provided above) without diabetes, 1,292 men and 1,364 women, 50.6 ± 14.3 years of age. A major goal of the study was to assess the effect of treatment-time regimen of prescribed BP-lowering medications on CVD, new-onset diabetes and renal outcomes. Thus, hypertensive patients were randomised either to ingest all BP-lowering medications upon awakening or the complete daily dose of ≥1 of them at bedtime and the remaining ones (if any) upon awakening (see Electronic Supplementary Material [ESM] Methods) [20–22, 30, 32].
New-onset diabetes was defined as fasting glucose ≥7.0 mmol/l on at least two clinical assessments ≥3 months apart in participants without prior history of diabetes or glucose-lowering treatment . Chronic kidney disease (CKD) was defined as either an estimated GFR <60 ml min−1 1.73 m−2, albuminuria (urinary albumin excretion ≥30 mg/24 h urine or albumin/creatinine ratio ≥3 mg/mmol), or both, on at least two occasions ≥3 months apart . GFR (ml min−1 1.73 m−2) was estimated by the CKD–Epidemiology Collaboration (CKD-EPI) equation . Diagnosis of the metabolic syndrome was established by the National Cholesterol Education Program Adult Treatment Panel III revised definition . Diagnosis of obstructive sleep apnoea (apnoea/hypopnoea index ≥10) was corroborated by overnight in-hospital polysomnography when the participant or bed-mate reported sleepiness during daytime and loud snoring, choking, interrupted breathing and/or awakenings during night-time sleep.
ABP, wrist activity and other assessments
At inclusion and at every scheduled visit for ABPM during follow-up, the SBP and DBP of each participant were automatically measured every 20 min between 07:00 and 23:00 hours and every 30 min during the night for 48 consecutive hours with a calibrated SpaceLabs 90207 ABPM monitor (SpaceLabs Issaquah, WA, USA). A 48 h, instead of the most common 24 h, monitoring span was chosen to improve reproducibility of results, as accurate calculation of ABP characteristics (including mean BP values) and dipping classification depends markedly on the duration of ABPM . Participants were instructed to adhere to their usual activities with minimal restrictions but to keep a similar activity–rest schedule and avoid daytime napping during the two consecutive days of ABPM. In keeping with current recommendations , BP series were considered invalid for analysis, and thus ABPM repeated within the same week, if ≥30% of the measurements were missing, if data were lacking for an interval of >2 h, if data were obtained when the rest–activity schedule was irregular during the 2 days of monitoring, or if the night-time sleep period was <6 h or >12 h.
All participants also wore an actigraph (Mini-Motion-Logger, Ambulatory Monitoring, Ardsley, NY, USA) on the dominant wrist to record the level of physical activity at 1-min intervals during each 48 h ABPM. Actigraphy data were used to verify absence of daytime napping and to precisely define the commencement and termination of the daytime awake and night-time asleep spans of each participant using dedicated software  to accurately derive the respective ABP means.
Blood withdrawal from an antecubital vein was done in the clinic between 08:00 and 09:00 hours, after overnight fasting, the same week when each 48 h ABPM session was initiated. The patients collected their urine during the first 24 h of ABPM. Blood and urine samples were analysed using routine automatic techniques in the hospital laboratory. Just before commencing each 48 h ABPM session, the same investigator obtained six consecutive clinic BP measurements using a validated automatic oscillometric device (HEM-705IT, Omron Health Care, Vernon Hills, IL, USA) after the participant had rested in a seated position for ≥10 min. Proper cuff size for clinic and ABP assessment was determined by measurement of upper arm circumference at each study visit.
Participants underwent the same evaluation procedure described above, including conventional daytime clinic BP measurement, 48 h ABPM/wrist activity monitoring and blood/urine analysis, plus other complementary tests as ordered by physicians (e.g., electrocardiogram, funduscopic evaluation, echocardiogram, etc.), annually or more frequently (3 months after any doctor-ordered change in therapy to improve ABP control in treated patients). Investigators blinded to the treatment scheme of hypertensive patients and not involved in clinic evaluations, ABP measurement and statistical analyses assessed the development of new-onset diabetes, as defined above, among other outcome variables of interest. Complete clinical records of all enrolled participants, including all periodic laboratory tests performed during follow-up, were reviewed at least annually plus the year following their last ABPM.
ABPM profiles were edited according to conventional criteria to correct for measurement errors and outliers; SBP readings >250 or <70 mmHg, DBP >150 or <40 mmHg and pulse pressure (PP, SBP minus DBP) >150 or <20 mmHg were automatically discarded. The ‘48 h ABP mean’ was calculated as the average of all valid readings obtained throughout the 48 h assessment span. Awake and asleep ABP means were calculated from the 48 h monitoring as the average of all valid readings obtained during the hours of daytime activity or night-time sleep, respectively, as differentiated by wrist actigraphy. Sleep-time relative BP decline (index of BP dipping), defined as % decrease in mean BP during night-time sleep relative to mean BP during daytime activity, was calculated as: ([awake ABP mean − asleep ABP mean] / awake ABP mean) × 100, incorporating all the data sampled by 48 h ABPM. For comparative purposes, a participant was defined as dipper if the sleep-time relative SBP decline was ≥10%, and as non-dipper otherwise [28, 29].
The risk of new-onset diabetes was evaluated based on the: (1) baseline clinic and ABPM evaluation from every participant; and (2) changes in any tested clinic cuff and ABP variable per participant during follow-up. Demographic and clinical characteristics were compared among groups of individuals who did or did not develop new-onset diabetes by t test (quantitative variables) or nonparametric χ
2 test (proportions). The Cox proportional-hazard model, adjusted for significant confounding variables, was used to estimate HRs with 95% CI for events associated with each tested potential prognostic BP variable; we standardised these HRs by calculating them for 1-SD increments for each BP variable. All demographic, anthropometric and clinical laboratory variables listed in Table 1 were tested as potential confounding variables. Adjustments were applied for fasting glucose, waist circumference, age, hypertension treatment-time and CKD diagnosis, as these influential factors were the only ones consistently significant in all tested Cox regression models. On the other hand, the prognostic value of BP changes during follow-up was evaluated by entering the change in the BP variable of interest as a time-dependent covariate in the Cox regression analysis. For survival analysis, follow-up was established as either the time to the confirmed diagnosis of new-onset diabetes or the time to the last clinical evaluation in individuals who did not develop new-onset diabetes.