The independent prognostic significance of different indices of blood pressure (BP) variability is not clear. We investigated the prognostic value of BP variability estimated as s.d. or average real variability (ARV) of daytime and night time BP, in hypertensive patients.
The occurrence of fatal and nonfatal cardiovascular events was evaluated in 1,280 sequential hypertensive patients (550 initially untreated and 730 initially treated) aged ≥40 years. Subjects with s.d. or ARV of daytime or night time systolic or diastolic BP below or above the median were classified as having low or high BP variability.
During the follow-up (4.75 ± 1.8 years), 104 cardiovascular events occurred. The event rate per 100 patient-years was 1.71 in the global population. After adjustment for other covariates, Cox regression analysis showed that cardiovascular risk was higher in subjects with high ARV of daytime systolic BP in initially untreated, initially treated, and all the subjects (high vs. low ARV, hazard ratio (HR) 2.29 (1.06–4.94), HR 1.90 (1.06–3.39), and HR 2.07 (1.31–3.28), respectively). ARV of daytime diastolic BP and night time BP, and s.d. of daytime and night time BP were not significantly associated with risk or were not independent predictors of outcome.
In this study, high ARV of daytime systolic BP resulted in an independent predictor of cardiovascular risk in hypertensive patients, while high s.d. did not. Our data suggest that, in comparison to s.d., ARV could be a more appropriate index of BP variability and a more useful predictor of outcomes.
American Journal of Hypertension 2009; 22:842–847 © 2009 American Journal of Hypertension, Ltd.
The impact of blood pressure (BP) variability, evaluated by noninvasive monitoring, on cardiovascular outcome is not yet completely clear. The prognostic significance of BP variability has been evaluated in different cohorts of initially untreated or treated hypertensive subjects1–8 and in general populations.9,10 In each situation, however, contrasting results have been reported.1–10 Some studies2,4,6–8,10 have reported no independent association between BP variability, expressed as s.d., and cardiovascular risk. Other studies,1,3,5–7,9 conversely, have reported an association between variability of daytime diastolic1 or systolic3,7,9 or night time systolic5–7,9 BP and cardiovascular outcome. Moreover, a divergent impact of daytime and night time systolic BP variability on cardiovascular morbidity and mortality,3,6,7,9 and on cardiac and cerebrovascular5,6 events has been observed in various studies. One study10 reporting no independent association between BP variability, evaluated as s.d., and cardiovascular mortality, showed an association between residual diastolic BP variability, evaluated by Fourier spectral analysis, and risk.
The appropriateness of s.d. as an index of BP variability has been recently questioned.11 Indeed, it only reflects the dispersion of values around the mean and does not account for the order in which BP measurements are obtained.11 As an alternative, it has been proposed a new index,11,12 named average real variability (ARV), inspired by the total variability concept12 of real analysis in mathematics and sensitive to the individual BP measurement order. In a preliminary study,11 by using s.d. and ARV as indices of BP variability, it has been shown that ARV, but not s.d., may predict cardiovascular risk. However, considering the relatively high event-rate observed in that study,11 its results cannot be completely extrapolated to a hypertensive population.
Thus, other studies are needed to better understand the prognostic significance of BP variability and to investigate the impact of various indices of BP variability on cardiovascular outcome.
The aim of this study was to evaluate the prognostic value of BP variability estimated with different indices, that is, s.d. and ARV of daytime and night time systolic and diastolic BP, in hypertensive patients.
Subjects. We studied 1,280 sequential hypertensive patients (550 initially untreated, that is, without drug therapy at baseline evaluation and 730 initially treated, that is, receiving drug therapy from a few years at baseline evaluation) aged ≥40 years recruited from January 2001 to June 2006 who were referred to our hospital outpatient clinic for evaluation of hypertension. The follow-up survey was between June and December 2008. Our databases of initially untreated or treated hypertensive subjects, including patients' characteristics and summary data of ambulatory BP monitoring, start from 1993. However, single readings are needed to calculate ARV. As these data were no more available on personal computer for patients evaluated up to 2000, we could include in this study only subjects examined after 2000. Subjects with secondary hypertension were excluded. All the patients underwent clinical evaluation, electrocardiogram, routine laboratory tests, echocardiographic examination and noninvasive ambulatory BP monitoring. Study population came from the same geographical area (Chieti and Pescara, Abruzzo, Italy). The study was in accordance with the Second Declaration of Helsinki and was approved by the institutional review committee. Subjects gave informed consent.
Office BP measurements. Clinical systolic and diastolic BP recordings were performed by a physician by using a mercury sphygmomanometer and appropriate-sized cuffs. Phase 5 was used to determine diastolic BP. Measurements were performed in triplicate, 2 min apart, and the average value was used as the BP for the visit. Clinic hypertension was defined as BP ≥140 and/or 90 mm Hg in at least two visits.
Ambulatory BP monitoring. Ambulatory BP monitoring was performed with a portable noninvasive recorder (SpaceLabs 90207, Redmond, WA) on a day of typical activity, within 1 week from clinic BP measurement. Each time a reading was taken, subjects were instructed to remain motionless and to record their activity on a diary sheet. Technical aspects have been previously reported.13 Ambulatory BP readings were obtained at 15-min intervals from 6 AM to midnight, and at 30-min intervals from midnight to 6 AM. The following ambulatory BP parameters were evaluated: daytime (awake period), night time (asleep period) and 24-h systolic and diastolic BP, and s.d. of daytime and night time systolic and diastolic BP. Then, readings were exported from the ambulatory BP monitoring software and ARV was calculated from these readings by a specifically designed software in Visual Basic that was based on the previously reported formula:11
where N is the number of valid BP measurements and K is the order of measurements from each subject monitoring.
Recordings were automatically edited if systolic BP was >260 or <70 mm Hg or if diastolic BP was >150 or <40 mm Hg and pulse pressure was >150 or <20 mm Hg (ref. 13). Subjects had recordings of good technical quality (at least 70% of valid readings). Patients with the s.d. or ARV of daytime and night time systolic/diastolic BP below or above the group median (10.9/8.4 or 8.7/6.6, and 9.1/7.8 or 7.9/6.7, respectively, for initially untreated subjects, and 11.7/8.4 or 8.9/6.5, and 10.0/8.1 or 8.4/6.8, respectively, for initially treated subjects) were classified as having low or high BP variability.
Echocardiography. End-systolic and end-diastolic measurements of interventricular septal thickness, left ventricular (LV) internal diameter, and posterior wall thickness were made according to the American Society of Echocardiography recommendations,14 within 1 month from clinic visit. LV mass was calculated using the formula introduced by Devereux et al.15 Individual values for LV mass were indexed by height2.7 and LV hypertrophy was defined as LV mass/height2.7 >50 g/m2.7 in men and >47 g/m2.7 in women.16
Follow-up. Subjects were followed-up in our hospital outpatient clinic or by their family doctors. The occurrence of cardiovascular events was recorded during follow-up visits or by telephone interview of the patient followed by a clinical visit. Hospital record forms were collegially reviewed by the authors of this study. Those reviewing the endpoints were blinded to BP variability data. Cardiovascular events included fatal and nonfatal myocardial infarction (at least two of three standard criteria: typical chest pain, ECG changes, transient elevation of conventional myocardial enzymes by more than twofold the upper normal limits), coronary or peripheral revascularization (bypass surgery or angioplasty), heart failure requiring hospitalization17 (acute pulmonary edema, paroxysmal nocturnal dyspnea, severe dyspnea on exertion), fatal and nonfatal stroke (rapid onset of localizing neurological deficit lasting ≥24 h with computer tomography evidence), and renal failure requiring dialysis.
Statistical analysis. Standard descriptive statistics were used. Unpaired t-test, χ2-test, and correlation were used, where appropriate. Event rates are expressed as the number of events per 100 patient-years based on the ratio of the observed number of events to the total number of patient-years of exposure up to the terminating event or censor. Cox regression analysis was used to evaluate univariate and multivariate association of factors with outcome in initially untreated, initially treated and all the subjects.18 First, univariate association between various variables or indices of BP variability and cardiovascular risk was evaluated. Then, multiple regression analysis was performed including in the final model variables and indices of BP variability that were significantly associated with outcome in univariate analysis (all the indices of BP variability that were associated with risk in at least one group were included). Cox models were also compared by using −2 log-likelihood statistics. Statistical significance was defined as P < 0.05. Analyses were made with the SPSS 12 software package (SPSS, Chicago, IL).
Characteristics and BP values of the study population according to daytime systolic ARV are summarized in Table 1. Subjects with high daytime systolic ARV were older, had higher clinic and ambulatory systolic BP and LV mass index, and higher prevalence of diabetes and LV hypertrophy.
In initially untreated subjects, at follow-up percentage using nondrug therapy, diuretics, β-blockers, calcium antagonists, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and α-blockers was 16.2, 42.2, 22.2, 26, 40.5, 15.3, and 7.5%, respectively. Single, double, and triple therapy were used in 27.8, 42.4, and 13.5% of patients, respectively. Those receiving antiplatelet and statin therapy were 12.2 and 5.8%, respectively.
In initially treated individuals, who were receiving drugs from a few years before entering the study, at follow-up percentage receiving diuretics, β-blockers, calcium antagonists, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and α-blockers was 57.8, 34.5, 30.5, 45.2, 28.2, and 15.6%, respectively. Antiplatelet and statin therapy was prescribed in 16.6 and 9.7% of patients, respectively.
Antihypertensive drug class distribution, as well as antiplatelet and statin therapy, was not significantly different among various BP variability groups in either initially untreated or treated patients.
Correlations between ARV and s.d., daytime and night time BP, and day-night BP ratio are reported in Table 2. As showed, partial or no correlations were found. The correlation coefficients between daytime and night time systolic/diastolic ARV were 0.20/0.20, 0.34/0.20, and 0.29/0.20 in initially untreated, initially treated, and all the patients, respectively.
During the follow-up (4.75 ± 1.8 years, range 0.2–7.5 years), 104 cardiovascular events occurred (43 in initially untreated and 61 in initially treated patients; the first one after 0.2 years and the last one after 7 years of follow-up). Specifically, there were 48 cardiac events (28 myocardial infarctions, 14 coronary revascularizations, and 6 heart failures), 48 cerebrovascular events, 7 peripheral events, and 1 renal failure requiring dialysis. The event rate of the population as a whole was 1.71 per 100 patient-years. Moreover, there were nine noncardiovascular deaths.
Only one subject included in the initially untreated group had an event before starting antihypertensive therapy. All the other subjects of both groups were on antihypertensive medication before the occurrence of events. Concerning the initially untreated group, the first event occurred after 0.75 years of follow-up and the last one after 7 years (duration of follow-up before events was 3.1 ± 1.7 years) and the duration of treatment before the occurrence of events (excluding the subject who experienced the event before starting therapy) was 2.8 ± 1.6 years (range 0.7–6 years).
The event rates and hazard ratios (HRs) according to BP variability subgroups are reported in Table 3. The following indices were significantly associated with cardiovascular risk in at least one subpopulation: (i) s.d. of daytime systolic BP, (ii) ARV of daytime systolic BP, (iii) ARV of daytime diastolic BP. The other indices of BP variability were not significantly associated with risk.
Multivariate analysis, after adjustment for other covariates, showed that (i) ARV of daytime systolic BP was an independent predictor of risk in all the groups, (ii) ARV of daytime diastolic BP tended to be a predictor of risk in initially untreated subjects, and (iii) s.d. of daytime systolic BP was not an independent predictor of risk (Table 4). In addition, when ARV was further adjusted for s.d. the HRs tended to increase, that is, 2.35 (1.08–5.11), 2.02 (1.11–3.69), and 2.19 (1.37–3.50) in initially untreated, initially treated and all the subjects, respectively.
When ARV was included in the model as a continuous variable it tended to be or remained an independent predictor of risk; HRs per 1 s.d. increase of the variable were 1.29 (0.98–1.69), 1.28 (1.02–1.60), and 1.27 (1.06–1.51) in initially untreated, initially treated, and all the patients, respectively. If ARV was further adjusted for s.d. the HRs tended to slightly decrease, that is, 1.26 (0.95–1.68), 1.28 (1.00–1.66), and 1.26 (1.04–1.53) in initially untreated, initially treated, and all the patients, respectively.
We also compared ARV and s.d. using −2 log-likelihood statistics. The addition of ARV to the model with daytime systolic BP resulted in a significant difference of the −2 log-likelihood value (P < 0.01), while the addition of s.d. did not (P > 0.1). Likewise, in the model with daytime systolic BP and ARV, the further addition of s.d. did not significantly change the value (P > 0.2).
Among subjects included in the initially untreated group, 110 (20%) had white coat hypertension (clinic BP ≥140 and/or 90 mm Hg and daytime BP <135 and/or 85 mm Hg) and one of them had a cardiovascular event. If these subjects were excluded from the analysis, the results did not change.
This study shows that ARV of daytime systolic BP is an independent predictor of cardiovascular events in initially untreated or treated hypertensive patients. ARV of daytime diastolic BP tends to be a predictor of risk in initially untreated subjects. ARV of night time systolic and diastolic BP, and s.d. of daytime and night time systolic and diastolic BP were not significantly associated with risk or were not independent predictors of outcome.
The impact of BP variability, evaluated as s.d., on cardiovascular risk has been explored in studies including initially untreated or treated1–8,11 hypertensive subjects or general populations.9,10 No independent association between BP variability and risk has been found in five studies2,4,8,10,11 and in subpopulations of two studies,6,7 while association has been observed in four studies1,3,5,9 and in subpopulations of two studies.6,7 In any case, in the studies reporting association between BP variability and risk, discrepant results have been obtained concerning the impact of daytime or night time systolic or diastolic BP variability on global cardiovascular morbidity,3,7 cardiovascular mortality,6,9 and incidence of cardiac or cerebrovascular events.5,6
Negative or discrepant results may be related to the relative weakness of BP variability in predicting cardiovascular outcome, inaccuracy of noninvasive monitoring in the evaluation of BP variability, or inappropriateness of s.d. as an index of BP variability.
Pitfalls of s.d. as an index of BP variability have been recently remarked.11 Indeed, (i) it only reflects the dispersion of values around the mean, (ii) does not account for the order in which BP measurements are obtained, and (iii) is sensitive to the relatively low sampling frequency of noninvasive BP monitoring. As an alternative, it has been proposed a new index11,12 named ARV, which is an average of the absolute differences of consecutive measurements. ARV is (i) sensitive to the individual BP measurement order and (ii) less sensitive to the relatively low sampling frequency of noninvasive monitoring. Indeed, it has been reported11,19 that subjects with different BP profiles may have the same s.d. but different ARV. As an example, we report 24-h BP profile of two patients with the aforementioned characteristic (Figure 1).
Twenty-four-hour blood pressure (BP) profile of two patients with similar s.d. but different average real variability (ARV).
Some years ago, Mena et al.11 have compared for the first time the prognostic impact of BP variability evaluated as s.d. or ARV. They11 studied 312 subjects, aged ≥55 years, of whom 71% had office hypertension and 14% were under antihypertensive medication. Patients were grouped according to tertiles of s.d. or ARV of daytime systolic BP and defined as having low, medium or high BP variability. During the follow-up (mean duration of ~2 years) there were 31 cardiovascular events representing an event rate of 5.38 per 100 patient-years. Cox regression analysis showed that risk was not significantly higher in subjects with high BP variability according to s.d. On the contrary, it was significantly higher in those with high BP variability according to ARV (high vs. low relative risk 4.55, 95% confidence interval 1.53–13.52, P < 0.02). The authors concluded that ARV is a more appropriate index of BP variability and is superior to s.d. in predicting cardiovascular outcome.
The aforementioned study11 is pioneering and opens new fields of investigation regarding BP variability and its prognostic relevance. However, it has a limitation. Indeed, the event rate reported in that study is exceedingly high considering the risk profile of the population.11 Thus, though the results are interesting they cannot be completely extrapolated to a hypertensive population and need to be confirmed in different contexts.
In our study, evaluating a hypertensive population with higher number of events with respect to the previous report11 and an event rate in line with the characteristics of the population, ARV was confirmed to be superior to s.d. in predicting cardiovascular risk in both initially untreated and treated hypertensive patients.
It can be hypothesized that ARV, taking into account the sequential order of BP changes between consecutive readings, is a more appropriate index of BP variability better describing the injury of additional intermittent stress on the cardiovascular system. It has been reported that intermittent BP load on cardiovascular structures may be as important as tonic BP load,20 and ARV seems to better describe this phenomenon. Thus, in addition to average BP, ARV may contribute to increased cardiovascular risk. Moreover, in contrast to s.d., ARV could be less obscured by other factors, such as age and BP level. In a previous study,10 a similar behavior has been reported for erratic BP variability that resulted superior to s.d. in predicting prognosis.
This study has some limitations. First, we studied only Caucasian subjects and our results cannot be applied to other ethnic groups. Second, concerning night time BP variability, it has been evaluated with a sampling interval of 30 min, which gives a less accurate estimation of BP variability;21 however, this interval was used also in other relevant studies.6,9 Third, it could be argued that daytime ARV, as well as average BP, could be influenced by usual daily activities; however, these parameters tend to represent the usual values of the subject, independently of an intrinsic or extrinsic origin. Fourth, at baseline the majority of subjects were treated from a few years and almost all the patients were treated during the study. Finally, it is unclear whether high ARV is a cause or a consequence of cardiovascular damage.
In conclusion, in this study high ARV of daytime systolic BP resulted in an independent predictor of cardiovascular risk in hypertensive patients, while high s.d. did not. Our data suggest that, in comparison to s.d., ARV could be a more appropriate index of BP variability and a more useful predictor of outcomes. Further studies are needed in other hypertensive populations to corroborate this finding.
The authors declared no conflict of interest.