During the past 2 decades, specific features of the 24-h blood pressure (BP) pattern determined by ambulatory blood pressure monitoring (ABPM) have been assessed as potential sources of injury to target tissues and triggers of cardiac and cerebrovascular events. For instance, the extent of the BP surge upon wakening has been associated with increased cardiovascular disease (CVD) morbidity and mortality in some but not all studies (1- 2). A growing number of studies (3- 6) have consistently shown an association between blunted asleep BP decline and increased incidence of fatal and nonfatal CVD events. Independent prospective studies have also concluded that the asleep BP mean is a better predictor of CVD risk than the awake or 24-h BP means (5- 10).

A limitation of all the previous studies on the prognostic value of ABPM is their reliance on a single baseline profile from each participant at the time of inclusion, without accounting for possible changes in the level and pattern of ambulatory BP during the years of follow-up, mainly associated with hypertension therapy, aging, and developing target organ damage and/or concomitant diseases. Thus, the potential reduction in CVD risk associated with modifying prognostic ABPM parameters (i.e., specifically reducing asleep BP) (9) is still a matter of debate. The MAPEC (Monitorización Ambulatoria para Predicción de Eventos Cardiovasculares [i.e., Ambulatory Blood Pressure Monitoring for Prediction of Cardiovascular Events]) study was specifically designed to investigate prospectively whether specific changes in the circadian BP profile results in reduced CVD risk (11). In particular, we investigated whether changes of ambulatory BP characteristics during follow-up might be related to increased survival in subjects who were systematically evaluated by periodic—at least annually—ABPM.

An extended version of Methods is available in the Online Appendix. In summary, the sample represents a population of Spanish subjects of both sexes ≥18 years of age. Inclusion criteria required subjects to be normotensive, untreated hypertensive, or resistant to treatment (uncontrolled BP according to the ABPM threshold values outlined in the following text while compliant to 3 optimally dosed hypertension medications of different classes, including a diuretic unless contraindicated or intolerant or any subject treated with >3 hypertension medications [12]). Exclusion criteria were pregnancy, history of drug/alcohol abuse, night/shift-work employment, diagnosis of acquired immunodeficiency syndrome, type 1 diabetes, secondary hypertension, CVD disorders (unstable angina pectoris, heart failure, life-threatening arrhythmia, kidney failure, and grade III to IV retinopathy), intolerance to ABPM, and inability to communicate and comply with all of the study requirements. Participants represent a consecutive series of persons fulfilling these exclusion/inclusion criteria, recruited among those referred to the hospital for ABPM evaluation, mainly to confirm/discard the diagnosis of hypertension suspected by daytime clinic cuff BP measurement in untreated subjects, or to evaluate BP control in already treated patients. Additional reasons to request ABPM included, among several others, impaired fasting glucose, metabolic syndrome, diabetes, chronic kidney disease, previous complications in pregnancy, suspicion of sleep disorders, and age over 60 years. This prospective single-center study (registered online with identifier code NCT00295542) was approved by the state Ethics Committee of Clinical Research. All participants gave written informed consent.

Between 2000 and 2007, we recruited 3,612 subjects fulfilling the inclusion/exclusion criteria. Among these, 3,344 (1,718 men/1,626 women, age 52.6 ± 14.5 years) provided all required information for the study. The remaining 268 subjects were eliminated due to inadequate ABPM sampling at baseline and their lack of consent for additional ABPM evaluations. We established a priori a minimum time of follow-up of 6 months for each subject and a minimum median follow-up of 5 years (11). Diagnosis of hypertension (in untreated subjects) was based on accepted ABPM criteria—an awake BP mean of ≥135/85 mm Hg for systolic blood pressure (SBP)/diastolic blood pressure (DBP) and/or an asleep BP mean ≥120/70 mm Hg (13). At the time of recruitment, 734 subjects were normotensive and 2,610 were hypertensive (776 with resistant hypertension) according to these ABPM criteria.

Patients with untreated hypertension were randomly assigned to 1 of 2 monotherapy treatment-time groups, either upon awakening in the morning or at bedtime at night, as extensively described elsewhere (14). Randomization of subjects to treatment-time (awakening or bedtime) was done separately for each allowed individual hypertension medication (valsartan, telmisartan, olmesartan, ramipril, spirapril, amlodipine, nifedipine GITS, nebivolol, torasemide). This ensured that the proportion of subjects treated with each medication was similar across the 2 treatment times. If subjects were uncontrolled based on ABPM criteria after 3 months of monotherapy, additional medications could be added in keeping with current clinical practice.

Participants with resistant hypertension were randomized to either: 1) modify the nature of their treatment, exchanging 1 of their medications with a new 1, thus without modifying the total number of medications but retaining the upon-waking ingestion time for all medications; or 2) shift the ingestion-time of 1 BP-lowering medication to bedtime (14). After this first randomization, investigators were allowed to exchange additional medications for others of different classes (keeping always the upon-waking ingestion time of all medications) in the first randomized group; progressively shift additional medications to bedtime in the second randomized group; or prescribe additional medications in either group. Changes in therapeutic scheme during follow-up in uncontrolled subjects (those with ambulatory BP above the thresholds provided in the preceding text) were always based on the results from periodic evaluation by ABPM (see the following text). The protocol did not allow dividing any prescribed medication in 2 or more doses. Thus, patients randomized to the bedtime-treatment group were never ingesting in the morning any of the medications ingested at bedtime.

Blood samples were obtained the same week each 48-h ABPM session was initiated. Participants reported to the hospital between 8:00 am and 9:00 am, after overnight fasting, for blood withdrawal from an antecubital vein. Samples were analyzed with routine automatic techniques in the hospital laboratory. Just before commencing ABPM, 6 clinic BP measurements were always obtained by the same investigator with a validated automatic oscillometric device (HEM-705IT, Omron Health Care, Inc., Vernon Hills, Illinois) after the subject had rested in a seated position for ≥10 min.

At inclusion as well as at each scheduled visit for ABPM during follow-up (see the following text), the SBP and DBP of each subject were automatically measured every 20 min between 7:00 am and 11:00 pm and every 30 min during the night for 48 consecutive hours with a calibrated SpaceLabs 90207 ABPM monitor (SpaceLabs Inc., Issaquah, Washington). Participants were instructed to do their usual activities with minimal restrictions but to adhere to a similar schedule during the 2 days of ABPM and avoid daytime napping. During monitoring, subjects maintained a diary listing the times of retiring to bed at night and awakening in the morning. The BP series were considered invalid for analysis if ≥30% of the measurements were missing, if data were lacking for an interval of >2 h, or if the nighttime sleep period was <6 or >12 h during ABPM.

All subjects wore an actigraph (Mini-Motion-Logger, Ambulatory Monitoring, Inc., Ardsley, New York) on the dominant wrist to monitor physical activity every min during 48-h ABPM. The actigraphy data, combined with patient diaries, were used to corroborate the absence of daytime napping and define the commencement and termination of the daytime awake and nocturnal asleep spans so the respective BP means for each subject could be accurately determined.

The same evaluation procedure described in the preceding text—including conventional clinic BP measurement, 48-h ABPM and wrist activity monitoring, and blood sampling—was scheduled annually in all participants or more frequently (after 3 months of any change in therapy in treated subjects) if the therapeutic scheme was modified to improve ambulatory BP control. Investigators blinded to the timed-treatment scheme of each participant (thus excluding those performing clinic evaluation at each visit to the hospital, clinic and ambulatory BP measurement, and/or statistical analyses) reviewed at least annually the complete clinical records of all enrolled subjects to assess CVD morbidity and mortality. Registered events included: death from all causes, myocardial infarction, angina pectoris, coronary revascularization, heart failure, acute arterial occlusion of lower extremities, thrombotic occlusion of the retinal artery, hemorrhagic stroke, ischemic stroke, and transient ischemic attack.

To correct for measurement errors and outliers, ABPM profiles were edited according to conventional criteria. Thus, SBP readings >250 or <70 mm Hg, DBP >150 or <40 mm Hg, and pulse pressure (difference between SBP and DBP) >150 or <20 mm Hg were automatically discarded. The sleep-time relative BP decline (an index of BP dipping), defined as the percentage decrease in mean BP during nocturnal sleep relative to the mean BP during daytime activity, was calculated as: ([awake BP mean − asleep BP mean]/awake BP mean) × 100, with all the data sampled by 48-h ABPM. For comparative purposes, a subject was defined as dipper if the sleep-time relative SBP decline was ≥10% and as nondipper otherwise. The morning BP surge was calculated as the difference between the average BP during the first 2 h after wake-up time and the hourly average centered on the lowest nocturnal sleep-time BP reading (1). The ambulatory arterial stiffness index was calculated as 1 minus the regression slope of DBP on SBP from ABPM (15).

The CVD risk was evaluated based on the: 1) baseline ABPM evaluation from every participant; and 2) changes in any tested parameter/participant during follow-up. The primary outcomes study endpoint was total CVD morbidity and mortality, which included all the events listed in the preceding text. We also used, as an additional primary endpoint, major CVD events (i.e., a composite of CVD deaths, myocardial infarction, and stroke). Demographic and clinical characteristics were compared among groups of subjects experiencing or not experiencing an event by t test (continuous variables) or nonparametric chi-square test (proportions). The Cox proportional hazards model, with adjustment for significant confounding variables, was used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) for events associated with each tested potential prognostic BP parameter at baseline; these HRs were standardized by calculating them for 1-SD increments for each BP parameter. By contrast, the prognostic value of BP reduction during follow-up was evaluated by entering the change in the tested ABPM parameter as a time-dependent covariate in the Cox regression analysis. For survival analysis, follow-up was established as either the time to the first documented event or the time to the last evaluation in event-free subjects. Statistical analyses were performed with SPSS software (version 13.0 for Macintosh, SPSS, Inc., Chicago, Illinois) and KaleidaGraph (version 3.6.4, Synergy Software, Reading, Pennsylvania).

During the median follow-up period of 5.6 years (range 0.5 to 8.6 years), we documented 331 first events (58 deaths, 45 myocardial infarctions, 51 angina pectoris, 35 coronary revascularizations, 44 cerebrovascular events, 46 heart failures, 21 cases of aortoiliac occlusive disease, and 31 thrombotic occlusions of the retinal artery). Event-subjects were: predominantly men; of older age; and who at baseline were likely to have type 2 diabetes, obstructive sleep apnea, metabolic syndrome, and/or abdominal obesity (Table 1). In addition, at baseline they had higher clinic BP (including pulse pressure, even after correcting for age), glucose, creatinine, uric acid, triglycerides, plasma fibrinogen, and erythrocyte sedimentation rate but lower estimated glomerular filtration rate (Table 1). All differences between groups reported in (Table 1) with a p value <0.002 stand as significant after correcting for multiple testing.

At baseline, the 48-h mean of SBP but not of DBP was significantly higher among event-subjects (Table 1). The largest difference between groups was in asleep SBP mean. The sleep-time relative BP decline was significantly lower (p < 0.001), and the prevalence of nondipping was significantly higher (73% vs. 46%) among event-subjects (p < 0.001) (Table 1).

(Table 2) shows the HR of CVD events estimated by the Cox proportional hazards model, calculated on the basis of the first ABPM profile from each subject. Adjustments were applied for sex, age, and diabetes (left column), because these influential factors were the only ones—among all the demographic and laboratory variables shown in (Table 1)—that were consistently significant in all tested models. Results documented increased CVD risk associated with older age (HR: 1.07, 95% CI: 1.06 to 1.08, p < 0.001, for each additional year of age), male sex (HR: 2.44, 95% CI: 1.93 to 3.08, p < 0.001, compared with women), and presence of diabetes (HR: 1.53, 95% CI: 1.21 to 1.94, p < 0.001, compared with no diabetes). Results remained mainly unchanged when the Cox proportional hazards models were further adjusted by baseline clinic BP (Table 2, right column); clinic BP was not a significant predictor of outcome in the models including asleep or 48-h BP mean. Moreover, a greater morning BP surge, calculated as defined (1), was significantly associated with lower CVD risk (Table 2). Finally, when asleep and awake SBP mean were used jointly in the same Cox regression model, only the former was a significant predictor of outcome. Most important, the asleep BP mean was also a most significant predictor of major CVD events (HR: 1.84, 95% CI: 1.60 to 2.11 for 1-SD elevation in asleep SBP mean, p < 0.001) (Table 2, right columns).

(Table 3) (left column) shows the results from time-dependent Cox regression analysis (adjusted by age, sex, diabetes, baseline BP, and number of hypertension medications used for treatment) for total CVD events. The progressive decrease in awake, asleep, and 48-h BP mean was associated with significantly increased survival. Changes in SD, morning BP surge, and ambulatory arterial stiffness index during follow-up were mainly not significantly associated with reduced/increased risk. The decrease in asleep BP mean during follow-up was also a significant predictor of survival from major events (HR: 0.81, 95% CI: 0.73 to 0.91 for SBP; HR: 0.68, 95% CI: 0.56 to 0.82 for DBP; p < 0.001) (Table 3, right column). When the changes in asleep and awake BP mean were used jointly in the same Cox regression model, only the decrease in asleep BP mean was significant associated with increased survival.

When subjects were categorized according to their baseline assessment with the ABPM threshold values previously outlined, the reduced HR associated with each 5-mm Hg decrease in asleep SBP mean during follow-up was significant for both subjects with either normal (HR: 0.81, 95% CI: 0.68 to 0.95, p = 0.005) or elevated BP (HR: 0.84, 95% CI: 0.79 to 0.89, p < 0.001). For this latter group, the increased survival associated with reducing asleep SBP mean was statistically significant independently of treatment time (HR: 0.88, 95% CI: 0.82 to 0.95, p < 0.001, for patients treated with all medications on awakening; HR: 0.76, 95% CI: 0.67 to 0.86, p < 0.001, for those treated with ≥1 medication at bedtime).

(Figure 1) shows, for the studied population divided in quintiles, the relationship between the decrease in asleep SBP mean from baseline and CVD risk. Adjusted HR was significantly higher (p < 0.001) in the last 3 quintiles, compared with the first 2 classes. There was a progressive reduction from 67% to 36% in the percentage of hypertensive subjects treated at bedtime across the quintiles, somehow indicating the association between bedtime-treatment, increased asleep BP reduction, and decreased CVD risk. (Figure 2) shows a J-shaped relationship between changes in clinic SBP during follow-up and CVD risk; the adjusted HR was significantly higher (p = 0.008) in the first than in the second quintile, and then it significantly increased again in the other 3 classes (p < 0.001, compared with the second quintile). The percentage of patients treated at bedtime was lower (45%) in the first as compared with the second quintile (56%).

The MAPEC study is the first to prospectively assess the prognostic value of changes in ABPM parameters during a follow-up time of sufficient duration by systematic periodic evaluation, with highly reproducible 48-h ABPM and wrist actigraphy, in a relatively large cohort of subjects with baseline BP ranging from normotension to hypertension. The results corroborate a higher prognostic value of ABPM over clinic BP measurements as documented in previous studies (3,5). Analyses based on a single ABPM profile from each subject obtained at baseline (as customary in all previous studies on this topic) indicate that the asleep SBP mean was a most significant predictor of CVD events (Table 2). Moreover, when asleep SBP mean was adjusted by awake SBP mean, only the former was a significant predictor of outcome.

Most important from the therapeutic point of view, the evaluation of changes in ABPM during follow-up documented that the progressive decrease in asleep BP mean was significantly associated with event-free survival. The relationship between decreasing asleep BP and reduced CVD risk was significant for subjects with either normal or elevated ambulatory BP analyzed separately, suggesting risk-reduction benefits even below the current diagnostic threshold of 120 mm Hg for asleep SBP mean. All together, these results not only corroborate that the asleep BP mean is the most significant prognostic marker of CVD morbidity and mortality, as previously suggested (5- 10), but also document for the first time that decreasing asleep BP mean significantly reduces CVD risk.

The findings summarized in (Table 3) from this prospective study suggest the asleep BP mean should be considered a novel therapeutic target for reduction of CVD risk. Along these lines, a number of previous prospective trials reviewed elsewhere (18) have detected meaningful morning/evening treatment-time differences in BP-lowering efficacy, duration of action, safety profile, and/or effects on the circadian BP pattern of different classes of hypertension medications. For instance, a once-daily evening, in comparison with morning, ingestion schedule of angiotensin receptor blockers and angiotensin-converting enzyme inhibitors results in greater therapeutic effect on asleep BP independently of the terminal half-life of each individual medication (18). Thus, the impact of hypertension treatment-time on sleep-time BP regulation might be clinically relevant. The previously reported results of the randomized part of the MAPEC study indicate that treatment with ≥1 medication at bedtime was significantly associated with greater asleep BP reduction and significantly lower CVD risk than treatment with all hypertension medications upon awakening (14).

Therapeutic intervention in hypertension consists of adequate control of BP, the goal being to reduce/avert CVD morbidity and mortality. Some studies found that too high a reduction in clinic BP might be associated with increased CVD risk, whereas moderate reduction in clinic BP would decrease it. Thus, it has been suggested that CVD outcomes have a J-shaped relationship to BP, decreasing as BP was lowered and rising again as BP decreased further (19- 21). We also found a J-shaped association in the relation between CVD risk and changes in clinic BP (Figure 2) but not with changes in asleep BP mean (Figure 1). The CVD risk decreased with progressive reduction in asleep BP mean. Moreover, the amount of the asleep BP reduction during follow-up was significantly correlated with increased number of patients treated at bedtime. It has been previously shown that increasing number of hypertension medications administered in the morning might lead to more intensive clinic BP reduction but also to an increased prevalence of nondipping as a consequence of the greater reduction in awake than asleep BP (22). We thus conclude that the actual controversy on the possible J-shaped relation with CVD risk, described so far only for clinic BP determined in patients presumably treated in the morning (19- 21), might not apply (when avoiding nocturnal hypotension) to asleep BP, a more significant predictor of CVD morbidity and mortality.

First, compared with other larger multicenter clinical trials on hypertensive patients entailing only clinic BP during follow-up, the sample size of the single-center MAPEC study might seem a limitation. However, the number of subjects participating in our study was considerably greater than that of most other published trials on the prognostic value of ABPM (3- 4). Second, like most other studies in this field, inclusion criteria restricted the study to subjects with suspected or previously diagnosed hypertension. Nonetheless, the design of the MAPEC study also incorporates several strengths. Although all previous trials on the prognostic value of ABPM had relied on a single baseline profile from each subject, the MAPEC study is the first to provide results that are based on systematic periodic multiple evaluations by ABPM throughout the median 5.6 years of follow-up. This so-far unique approach allowed determination of the influence on CVD risk of specific changes during follow-up in relevant ABPM parameters. Further strengths of the MAPEC study are the use of: 1) 48-h instead of the most common 24-h ABPM sampling, to increase the reproducibility of the BP findings (23); and 2) wrist actigraphy to precisely and individually determine the beginning and end of the activity and sleep spans for each subject to enable the accurate calculation of the awake and asleep BP means.

Current international guidelines have already recognized the prognostic value of ABPM and suggest limited clinical situations in which the technique is recommended. These situations include: suspected white-coat hypertension, resistant hypertension, hypotensive symptoms after treatment, episodic hypertension, and autonomic dysfunction (13). Our findings further support ABPM for proper CVD risk assessment, taking into account the prognostic value of asleep BP, a novel therapeutic target requiring patient evaluation by ABPM. Moreover, the decreased CVD risk associated with progressively reducing the asleep BP mean suggests the need to use ABPM for the routine evaluation of treatment efficacy and proper definition of BP control.

For the extended version of the Methods, please see the online version of this paper.