Disproportional Decrease in Office Blood Pressure Compared With 24-Hour Ambulatory Blood Pressure With Antihypertensive TreatmentNovelty and Significance
Dependency on Pretreatment Blood Pressure Levels
The long-term relationship between 24-hour ambulatory blood pressure (ABP) and office BP in patients on therapy is not well documented. From a registry we included all patients in whom antihypertensive therapy needed to be uptitrated. Drug treatment included the direct renin inhibitor aliskiren or an angiotensin-converting enzyme inhibitor/angiotensin receptor blocker or drugs not blocking the renin–angiotensin system, alone or on top of an existing drug regimen. In all patients, office BP and 24-hour ABP were obtained at baseline and after 1 year with validated devices. In the study population of 2722 patients, there was a good correlation between the change in office BP and 24-hour ABP (systolic: r=0.39; P<0.001; diastolic: r=0.34; P<0.001). However, the numeric decrease in office BP did not correspond to the decrease in ABP in a 1:1 fashion, for example, a decrease of 10, 20, and 30 mm Hg corresponded to a decrease of ≈7.2, 10.5, and 13.9 mm Hg in systolic ABP, respectively. The disproportionally greater decrease in systolic office BP compared with ABP was dependent on the level of the pretreatment BP, which was consistently higher for office BP than ABP. The white coat effect (difference between office BP and ABP) was on average 10/5 mm Hg lower 1 year after intensifying treatment and the magnitude of that was also dependent on pretreatment BP. There was a disproportionally greater decrease in systolic office BP than in ABP, which for both office BP and ABP seemed to depend on the pretreatment BP level.
In their most recent 2013 guidelines, the European Society of Hypertension and the European Society of Cardiology1 have identified lower threshold values for diagnosing arterial hypertension with 24-hour ambulatory blood pressure (ABP; ≥130/80 mm Hg) than with office BP readings (≥140/90 mm Hg). Since 2011, the National Institute for Health and Clinical Excellence Guidance has recommended the use of ABP monitoring to confirm the diagnosis of arterial hypertension if office BP was ≥140/90 mm Hg.2 In contrast, data on target BP based on 24-hour ABP to guide antihypertensive treatment are not provided in the guidelines.1 The reason seems to be that no large-scale randomized clinical trial in patients on antihypertensive treatment has been conducted to analyze the effect of antihypertensive therapy on cardiovascular prognosis with 24-hour ABP as target BP.
To develop a new consensus statement for ABP monitoring, most recently, several critical analyses of the best available evidence from clinical and observational studies were performed.3,4 The results showed that the agreement between ABP and office BP is not simply linear and that changes in ABP do not necessarily correspond to office BP in a 1:1 fashion. However, the long-term relationship between 24-hour ABP and office BP in hypertensive patients on treatment and the change of BP because of therapeutic intervention remains ill-documented.
In the 3A Registry, patients were prospectively followed for at least 1 year and had both office BP readings and 24-hour ABP monitoring before and 1 year after intensifying antihypertensive medication to achieve target systolic office BP <140 mm Hg.5,6 This database represents, therefore, a tool allowing us to compare the changes of office BP measurements to the changes of ABP obtained under real-life conditions.
The present analysis is based on the data of 3A Registry.6 The 3A Registry is a prospective, observational, noninterventional, multicenter registry listed under clinicaltrials.gov (NCT01454583) and the VfA database, a resource for noninterventional studies (http://www.vfa.de/de/arzneimittel-forschung/datenbanken-zu-arzneimitteln/nisdb/nis-details/_616). Details of the study design and baseline data have been published elsewhere.5
In brief, consecutive patients with known or newly diagnosed arterial hypertension in whom the physician had decided independently and per best clinical judgment to initiate or intensify antihypertensive therapy were eligible for inclusion. Exclusion criteria were participation in a randomized controlled clinical trial and foreseeable problems to perform follow-up visits. Depending on the initiated medication, patients were part of 1 of the 3 study groups: treatment with (1) direct renin inhibitor aliskiren, or (2) an angiotensin-converting enzyme inhibitor or angiotensin receptor blocker, or (3) drugs not blocking the renin–angiotensin system.
Reflecting the utilized medication of the 3 study groups, the registry was called 3A—aliskiren, angiotensin receptor blocker/angiotensin-converting enzyme inhibitor, and others (Andere in German, ie, others). Medication was given alone or on top of an existing drug regimen.
The data were collected in web-based format with a standardized questionnaire (electronic case report form). Measures of quality control included automated plausibility checks during data entry, queries after data entry, and in 10% of patients, on-site monitoring with source data verification. All data, if available, were collected during the clinical examination or from the review of patient chart. Data were recorded at inclusion (baseline) and during follow-up visits.
In 6139 patients, 24-hour ABP monitoring was performed at baseline visit. At the 1-year examination, in 2722 hypertensive patients 24-hour ABP monitoring was repeated in parallel to office BP measurements. Office BP was assessed with the standard devices (all were oscillometric devices) available at the physicians’ office (manual sphygmomanometers or semiautomated devices), which according to German legislation must have a calibration validation. Furthermore, the German guidance for measuring office BP (sitting position, after 5 minutes of rest, at least 2 repeated measurements) had to be followed. Twenty-four–hour ABP monitoring was also only performed with validated devices (for German guidelines, see http://www.hochdruckliga.de/blutdruckmessgeraete-mit-pruefsiegel.html) routinely used in physicians’ office. Average of office BP readings and means of 24-hour ABP (minimum requirement of ≥50 measurements during at least a ≥22-hour period), daytime ABP, and night-time ABP were entered into the database.
Because at the time of inclusion patients had uncontrolled hypertension and the physician had decided to initiate or intensify antihypertensive therapy, we have specified the 1-year follow-up examination for our analyses when a stable situation was achieved.
Continuous variables were summarized with descriptive statistics (absolute numbers, means, SD, or medians with 25 and 75 percentiles as appropriate). Categorical data were described by the number and percentage of subjects in each category. As univariate test of location, we applied the signed-rank test. Statistical comparisons between groups were performed by Pearson χ2 for categorical variables or Kruskal–Wallis test for continuous measures. Percentages were calculated on the basis of patients with data for each respective parameter. All variables showed moderate deviations from a normal distribution as evidenced by the Kolmogorow–Smirnow test.
Like other biological measures, changes in both ABP and office BP may depend on baseline BP levels (Wilder law).7 Previous analyses comparing office BP versus ABP applied a linear regression statistical model.8 It remains to be determined whether the relationship between office BP and ABP can be described by simple, proportional, or linear formulas, because BP (which should not fluctuate according to the assumptions of the statistical model) is in fact a highly variable biological parameter. Thus, the statistical premises to run simple regression models are not entirely fulfilled, although widely used.9,10 Evaluating ABP and office BP by means of univariate or bivariate models hence runs the risk of inadequately simplifying a complex reality. We, therefore, further developed and applied a multivariate (4-variate) model of BP change (see online-only Data Supplement).
We conducted all analyses with SAS 9.3 (SAS Institute, Inc, Cary, NC). P values ≤0.05 (2-sided) were considered significant.
In a total of 2722 patients, office BP and 24-hour ABP were obtained both at baseline and after 1 year. Clinical characteristics of the study population were: mean age 64 years, mean body mass index 28.4 kg/m2, 45% women, mean hypertension duration 6.9 years, and average regimen 3 antihypertensive medications. Of the 2722 patients, 85% had hyperlipidemia, 14% were current smokers, 30% had diabetes mellitus, 31% had cardiovascular diseases, and 9% had chronic kidney diseases (Table 1).
Decrease in Office BP Versus ABP
In the whole study population, both office BP and 24-hour ABP decreased after 1 year of treatment. Office BP decreased by 18.7±20/9.6±12 mm Hg, that is, from 156.2±18/90.6±11 mm Hg at baseline to 137.5±14/81.0±8 mm Hg at 1 year. In parallel, 24-hour ABP decreased by 10.1±15/6.1±10 mm Hg, from 146.2±15/85.5±11 mm Hg at baseline to 136.2±13/79.4±8 mm Hg after 1 year. Daytime and night-time ABP decreased accordingly (daytime from 151±16/88±11 to 140±13/82±9 mm Hg; night-time from 136±17/79±12 to 126±15/73±9 mm Hg). The change in office BP and 24-hour ABP correlated highly significantly with each other (systolic: r=0.39; P<0.001; diastolic: r=0.34; P<0.001; Figure 1). Similar correlations were observed for daytime (Figure S3A and S3B) and night-time ABP (Figure S3C and S3D). However, the numeric decrease in office BP did not correspond to the decrease in ABP in a 1:1 fashion, neither for 24-hour ABP nor for daytime and night-time ABP. For any given fall of office BP, the fall in ABP was clearly less than the one in office BP.
By applying a linear regression model, for any given pretreatment office BP (including 156/91 mm Hg reflecting the average value of our population), a decrease of 10, 20, and 30 mm Hg in systolic office BP corresponded to a decrease of 7.1, 10.5, and 13.9 mm Hg in systolic ABP, respectively (Figure 2A). By applying the nonlinear multivariate additive mixed model (given for the average BP 156/91 mm Hg), decreases of 10, 20, and 30 mm Hg in systolic office BP corresponded to significantly smaller decreases in systolic ABP of 7.2, 10.6, and 14.0 mm Hg, respectively (Figure 2A). Similar findings were obtained for changes in diastolic BP (Figure 2B). When the analysis was repeated for daytime and night-time ABP, a similar striking difference between the change in office versus daytime and night-time ABP was observed (Figure S4A–S4D). Direct comparison of the 2 models showed that the nonlinear multivariate additive mixed model yielded even greater differences between office BP and ABP values than the linear regression model, in particular in the higher range of pretreatment office BP (see Figure S2).
BP Decrease and Pretreatment BP Level
When the actually measured pretreatment office BP values were stratified into 10 mm Hg groups (excluding those with systolic <140 mm Hg and diastolic <90 mm Hg), the measured decrease in both office BP and 24-hour ABP (in absolute terms and in percentage) was highly dependent on pretreatment BP (Table 2). In patients with a complete data set (N=2157), for example, with a pretreatment systolic office BP in the range of 140 to 149 mm Hg, the decrease in office BP and ABP was −11.3 mm Hg (or −7.6%) and −8.9 mm Hg (or −5.7%), respectively (Figure 3). In contrast, with pretreatment systolic office BP ≥180 mm Hg, the decrease in office BP and ABP was −51.4 mm Hg (or −26%) as opposed to a decrease in ABP by −18.5 mm Hg (or −11%). Hence, the disproportional decreases in systolic office BP compared with ABP were dependent on pretreatment systolic BP, irrespective whether changes are given in absolute or percent terms. A similar result was found for the relation of changes in diastolic office BP versus ABP (Table 2).
The analysis for daytime and night-time ABP displayed similar results (Table S1A–S1D). For example, with a pretreatment systolic office BP in the range of 160 to 169 mm Hg, the decrease in office BP and day-time ABP were −23.6 mm Hg (or −14.6%) and −13.3 mm Hg (or –8.1%), respectively.
White Coat Effect and Pretreatment BP Level
The white coat effect defined as difference between office BP and 24-hour ABP decreased after 1 year by 8.6±20/3.5±12 mm Hg (baseline 10±18/5.1±11; follow-up 1.3±13/1.6±8.7 mm Hg; P<0.001). Likewise, when defined as difference between office BP and daytime ABP, the white coat effect declined by 7.5±20.3/3.2±13 mm Hg after 1 year. Interestingly, the decrease of white coat effect after 1 year was dependent on pretreatment BP (Figure 4). Thus, the observed decrease of white coat effect contributed to the disparate changes in office BP and ABP after 1 year.
Several important findings evolve from the present study. First, we showed that changes in BP after initiating or uptitrating antihypertensive medication are dependent on pretreatment BP values. This was found to be true for both office BP and ABP. This phenomenon that the pretreatment level determines to a large extent the change per se is not restricted to change in BP only but has also been observed for changes in heart rate11 or low-density lipoprotein cholesterol.12 The law of initial value (German: Ausgangswertgesetz) was first described by Josef Wilder in 1927 (published in 1932), who proposed that the direction of response of body function to any agent depends to a large degree on the initial value of that function.7 However, even today it remains uncertain whether the Wilder law of initial value represents a real biological phenomenon or simply a statistical artifact.
From a clinical perspective, the Wilder law of initial value is an important concept, because it predicts that in the most severe hypertensive patients, the fall in BP will be greater with the same medication than in those with less severe hypertension. If the effect was similar (ie, the fall in BP was independent of pretreatment level), we would encounter many more clinical complications related to hypotension with antihypertensive treatment. Wilder law of initial value thus indicates that we have to take the pretreatment level into account when comparing the efficacy of various antihypertensive medications in clinical trials.
Our second finding is that changes in office BP and ABP are not related to each other in a 1:1 fashion. When the decreases of systolic 24-hour daytime/night-time ABP were plotted against the decrease of systolic office BP, the regression lines were not the line of identity. This observation again has important implications. Based on the average value of pretreatment systolic BP of our population (156 mm Hg), we calculated that decreases of 10, 20, and 30 mm Hg in office systolic BP corresponded to decreases of 7.1, 10.5, and 13.9 mm Hg in systolic ABP (Figure 2A). The ratio between these 2 changes is obviously not constant; it depends on the fall in office BP and ABP, respectively. Similar observations were also made comparing changes in daytime and night-time ABP with office BP changes (Figures S4A–D). Likewise, analyzing diastolic instead of systolic BP, consistent findings were observed through with numerically lower magnitude. By using a similar approach, Mancia et al4 documented in a meta-analysis that a fall of 10 mm Hg in office systolic BP corresponded to a fall in 24-hour systolic BP to nearly the same extent, whereas a fall of 30 mm Hg in office systolic BP corresponded to a fall of ≈20 mm Hg in ABP only. However, a direct comparison of the 2 studies is not possible, because the basis of analysis and pretreatment BP differed substantially between (based on mean BP of study cohorts) Mancia et al’s4 meta-analysis and our large patient-based analysis.
Our third finding is that the white coat effect in patients on antihypertensive therapy decreases over time by ≈10/5 mm Hg on average, but was still present after 1 year. Such a decrease of white coat effect during 1 year appeared to be dependent on pretreatment BP as well (Figure 4), being neglible if pretreatment BP is close to target BP <140/90 mm Hg, but substantial with severely elevated BP values. This result is reflected by the volatile BP component (Figure S1A and S1B, dark grey column) that decreases after 1 year (Figure S1A and S1B, dark orange column) to a large extent if pretreatment BP is high. Thus, the greater fall in office BP in patients with severely elevated pretreatment BP is caused to a large extent by the reduction of the white coat effect. Because 24-hour ABP is void of white coat and placebo effect, the changes in ABP are smaller and the pretreatment BP level lower,13–15 thereby explaining at least in part why changes in office BP and ABP are not related to each other in a 1:1 fashion. The statistical phenomenon of regression to mean might be also one contributing factor, but it is impossible to quantify the effect size from our data set.
When analyzing clinical studies with the primary objective of assessing BP changes, our findings have significant implication for their interpretation. The average systolic pretreatment BP in the 3A Registry population was 156 mm Hg and a decrease of office systolic BP of ≈20 mm Hg was observed. If we assume that pretreatment ABP was also 156 mm Hg (in fact it was 146 mm Hg), we would, according to the Wilder law of initial value, expect a greater fall in ABP than the actually observed fall in ABP of 10 mm Hg only.
The Symplicity-HTN 2 study illustrates the importance of our findings. Systolic BP dropped by 32/12 mm Hg by office BP with pretreatment BP of 178/96 mm Hg, whereas the changes in ambulatory systolic BP were 11/7 mm Hg observed from a pretreatment level of 146/86 mm Hg.16 Similar observations were made in our 3A Registry. BP in 274 patients whose pretreatment systolic office BP was between 170 and 179 mm Hg decreased by 31 mm Hg in office values but only 13 mm Hg in ABP values (Table 2).
Our study was not a randomized controlled clinical trial but one of a noninterventional observation study, which may be considered as a limitation. Both designs have their strengths and weaknesses discussed otherwise in detail.5 The use of different methods and devices for BP measurements may have created a greater variability of the results, but the same device was used at baseline and after 1 year, and all devices were validated according to German legislation following the recommendations and validation according to the German Hypertension Society. Furthermore, time of office BP measurements and the peak levels of multiple drugs taken by the patients were not assessed because of the obvious inherent difficulties of such an effort in a large-scale trial. The white coat effect was determined by the difference of office BP minus 24-hour or daytime ABP, which represents an indirect approach of the white coat effect because the ABP measurement depends also on other factors (eg, level of activities), which are inherent to ABP measurements.
Both changes in office BP and ABP are dependent on pretreatment BP level. Because the pretreatment BP of ABP is usually lower, decreases in ABP are, therefore, smaller than those observed with office BP. A simple recipe to overcome these limitations in interpretation of changes in ABP is unfortunately not available because of the complexity of the phenomenon. Thus, changes in ABP in clinical studies as well as in individual patients need careful judgment and analysis bearing in mind Wilder law of the initial value.
Sources of Funding
This 3A registry has been sponsored by Novartis Pharma GmbH, Nuremberg, Germany.
The online-only Data Supplement is available with this article at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.113.03140/-/DC1.
- Received March 26, 2014.
- Revision received April 8, 2014.
- Accepted July 14, 2014.
- © 2014 American Heart Association, Inc.
- Mancia G,
- Fagard R,
- Narkiewicz K,
- et al
- Krause T,
- Lovibond K,
- Caulfield M,
- McCormack T,
- Williams B
- Zeymer U,
- Dechend R,
- Deeg E,
- Kaiser E,
- Senges J,
- Pittrow D,
- Schmieder R
- Head GA,
- Mihailidou AS,
- Duggan KA,
- et al
- Chesher A
- Camm J
- Palmer MK,
- Nicholls SJ,
- Lundman P,
- Barter PJ,
- Karlson BW
Novelty and Significance
What Is New!
We found a strong dependency of treatment-induced changes in blood pressure (BP) from pretreatment levels.
What Is Relevant?
At any given pretreatment BP, we observed a disproportionally greater decrease in systolic office BP compared with ambulatory BP.
The white coat effect was on average 10/5 mm Hg lower 1 year after intensifying treatment and also dependent on pretreatment BP.
In this study cohort of 2722 patients, changes in office BP were not related to changes in ambulatory BP in a 1:1 fashion. Our results should be taken into account when judging decrease in BP in individual patients and clinical studies.