Twenty-four-hour ambulatory blood pressure profiles of high-risk patients in general practice: data from an ambulatory blood pressure monitoring registry

‘Essential hypertension represents a quantitative deviation from the norm. To select a dividing line between normal and abnormal blood pressure is evidently arbitrary’ ‘The conclusions drawn are often chiefly based not on facts, but by an entirely arbitrary procedure which is absolutely without justification, either from theory or from fact’ (Sir George Pickering, 1961) Assessing the actual blood pressure (BP)-related risk of cardiovascular complications in a given patient is often a difficult task, with such a quantification depending both on the accuracy in estimating the BP behaviour over 24 h and on the level of total cardiovascular risk. In this context, important information can be offered by ambulatory BP monitoring (ABPM), due to the ability of this approach to provide us not only with data on average BP values during daytime and night-time, but also on the amplitude and frequency of BP variations during different activities [1–5]. Because of these appealing features, clinical use of 24-h BP recordings has been rapidly increasing all over the world. Its diffusion is further supported by the data, obtained either from general population studies or from cohort studies in hypertensive patients on the ability of ABPM to predict cardiovascular morbidity and mortality over and above the information provided by clinic BP measurement [6–17]. In this issue of the journal, Gorostidi et al. [18] offer an additional demonstration of the clinical value of ABPM, by reporting data on the features of ABP profiles in a large number of hypertensive patients at high cardiovascular risk. This is carried out through a cross-sectional analysis of a 20 000-patient database built from data obtained in general practice throughout Spain. Collecting data on daily life BP from a large number of subjects in a general practice setting is not an easy task, and the efforts made by the Spanish Society of Hypertension to develop a project aimed at reaching this goal, with the endorsement of the European Society of Hypertension, should be acknowledged and appreciated. The availability of such a database is likely to provide a very important support for longitudinal follow-up based on ABPM parameters in clinical practice. Although only cross-sectional data are reported in the study, and even if the results are partly confirmatory of previous work, the information provided by this paper is important because such a large sample size was never obtained in previous studies. Among the observations reported by Gorostidi et al. [18], two issues may be of particular interest. The first is the difference between office and 24-h average BP in patients categorized as being at low or at high added risk of cardiovascular events. Several previous papers have shown that, on average, office BP is higher than 24-h ABP mean levels. This was the case in a general population as well as in selected groups of patients [19–26]. A difference between clinic and 24-h ABP can also be observed when assessing the response to antihypertensive treatment. In a meta-analysis of studies in which the BP response to treatment was simultaneously assessed in the same patients by both clinic readings and 24-h ABPM, the drug-induced changes in clinic BP were always greater than those in 24-h average ABP [27]. In the study by Gorostidi et al. [18], a mean difference between clinic and 24-h ABP of approximately 16/9 mmHg was seen in patients categorized as being at low to moderate added risk (clinic BP 144/87 versus 24-h ABP 128/78 mmHg). Such a difference was greater in patients categorized as being at high risk, based on the number of their cardiovascular risk factors, amounting to 23/13 mmHg (clinic BP 159/90 versus 24-h ABP 136/77 mmHg). As already shown in the PAMELA study [21], this might mainly depend on the higher clinic BP values found in these patients, with higher clinic BPs being usually characterized by a greater difference with corresponding average ABP values. The practical implication of these data, which extend previous observations, is that clinic BP readings and ABPM may provide us with definitions of patients' BP levels that are variably discrepant in different categories of patients. This is relevant to the identification of target BP levels to be achieved by treatment. Although it has been suggested that clinic BP should be reduced to values < 140/90 mmHg in patients at lower risk, lower values would be desirable in patients at higher risk with evidence of cardiac and/or renal damage. Due to the varying and non-linear discrepancy between clinic and ambulatory BP levels, the definition of the corresponding ABPM targets in the latter category of patients are more difficult to be identified. While these targets can hardly be derived from a linear regression between clinic and ambulatory BP values they should better be defined based on outcome data. The variable difference between clinic and ambulatory BP found in different patients may have several explanations. One factor is certainly the interference from a white-coat effect raising clinic BP at the time of its measurement [19,28,29]. Another factor is inclusion of the lower nocturnal BP levels in the calculation of 24-h BP averages. A final important component is the variability of isolated clinic readings, which may be more pronounced in patients with higher clinic BP and at higher risk, and which may lead to variably wide differences between clinic and ambulatory BP. This may particularly be the case when the former, as in this study, is obtained from a few readings only. The amplitude of such a difference could be reduced after repeated clinic BP measurements due to a possible reduction in the magnitude of the white-coat effect over time, and because of a progressive regression towards the mean phenomenon, which affects isolated clinic BP measurements more than average ambulatory data. Indeed, the average of 24-h ABP values already includes the effects of a ‘regression to the mean’. Thus, when only a few clinic readings are compared to 24-h average ABP, as in the study by Gorostidi et al. [18], the occurrence of a wide difference between the two sets of values may more importantly depend on the above phenomena, which appear to be more pronounced in more severe hypertensive patients. Data are available indicating that 24-h ABPM is a useful method for minimizing regression towards the mean phenomenon, as well as any regression dilution bias when BP is related to subsequent cardiovascular risk. The second interesting issue addressed by Gorostidi et al. [18] is the relationship between the degree of day–night BP difference, leading to classification of patients into ‘dipper’ or ‘non-dipper’ groups, and the level of patients' cardiovascular risk. Indeed, as already mentioned, a key advantage of ABPM compared to isolated office readings is represented by its ability to provide information not only on average BP levels, but also on BP variations over time. Twenty-four-hour BP fluctuations include several components [25], the three most important patterns, in terms of prognostic implications, probably being the nocturnal BP fall [11], the morning BP surge [6,30] and the degree of overall BP variability (quantified as the standard deviation of average BP of the recording period) [4,31]. Although the clinical importance of different BP variability patterns appears to be supported by an increasing amount of evidence [5], the precise mechanisms responsible for the prognostic consequences of these phenomena are still a matter of debate. This in particular is the case of nocturnal BP dipping. Most likely, non-dipping is related to the degree of concomitant target organ damage, to the severity of cardiovascular disease and also to alterations in sleep quality and duration [25]. These different mechanisms may have a different prognostic value. A ‘true’ and reproducible non-dipping pattern, due to an impairment in cardiovascular regulation, has been reported to carry an adverse prognosis [32,33]. This does not appear to be the case for a reduced nocturnal BP fall simply due to lack of nocturnal sleep during the recording time (in patients who thus should be better defined as ‘non-sleepers’ rather than ‘non-dippers’). In true non-dipper subjects, it is still unclear whether the amplitude of changes in BP between day and night does itself contribute to cardiovascular damage, or whether the latter actually depends on the mean absolute levels of nocturnal BP achieved as a result of an altered circadian BP profile. In the study by Gorostidi et al. [18], a non-dipping status was highly prevalent in the group of high-risk hypertensive individuals, reaching approximately 60%. This figure is much higher than that observed in studies dealing with a general population, where a non-dipping pattern may be found in approximately 25–35% of the subjects screened. Non-dipping status increased with increasing levels of 24-h systolic ABP; the higher the BP, the higher the prevalence of non-dipping state. Overall, these results [18] suggest the occurrence of an inverse association between level of total cardiovascular risk in hypertensive patients and the degree of BP reduction at night [32]. In spite of the interest raised by the above findings, a few critical aspects of the study by Gorostidi et al. [18] should be considered. Indeed, the great potential of a large database built from general practice observations needs to be carefully weighed against the background of the possible limitations affecting data collection in such a difficult condition. First, the results obtained by Gorostidi et al. [18] might have been affected by methodological and logistic problems. Out of 20 000 24-h ABP recordings initially considered 17 200 tracings were found to be valid for analysis, which is an impressive figure, although the editing criteria adopted to identify high-quality recordings are not reported in detail. However, data collected in a general practice setting may be less reliable than those obtained in the strictly controlled environment of a research laboratory because of the inability to properly account for confounding factors. Indeed, when considering correlations between BP patterns and clinical outcome within a set of cross-sectional data, and even more so when considering causal links during a longitudinal follow-up, meticulous adjustments for a number of other risk factors and for possible confounders is of ultimate importance for defining the precise clinical value of 24-h ABPM parameters. Second, most patients were treated [only 5284 (30.7%) of the total sample of 17.219 hypertensive individuals were untreated at baseline] and no information is available on the type and dose of the prescribed drugs, nor on their administration time. These factors might have deeply affected changes in 24-h BP profiles and, in particular, the rate of dipping. Indeed, in the study by Gorostidi et al. [18] it is difficult to discriminate between true non-dipping because of alterations in cardiovascular control mechanisms, and lack of nocturnal BP fall due to insufficient coverage by treatment. Third, patients were classified as dippers or non-dippers based on a single 24-h ABPM. Data are available indicating that the definition of the features characterizing a 24-h BP profile, including the dipping pattern, is affected by a poor reproducibility when derived from analysis of a 24-h ABPM performed once only [28,32]. In this context, repeated performance of a 24-h ABPM or 48-h ABPM would provide a more accurate classification of circadian patterns, but the application of these procedures in large-scale studies and/or in daily practice is clearly unfeasible. A similar limitation applies to clinic BP readings. In the study by Gorostidi et al. [18], data stem from more than 900 physicians reporting to the database, but clinic BP was defined on the basis of two BP readings from one visit only. This may have contributed to a lesser accuracy of the results obtained. Repeated readings obtained on different visits would have led to a more reliable definition of clinic BP levels, as well as of the ‘true’ difference between clinic and ambulatory BP. Fourth, some methodological aspects related to patients classification and recruitment should also be considered. The categorical classification of patients as dippers/non-dippers and their assignment to a given level of cardiovascular risk, as performed in the the study by Gorostidi et al. [18], is not free from drawbacks. This methodology could be improved by managing the considered parameters as continuous variables (i.e. by using the night-to-day ratio to define the dipping pattern and by adopting a quantitative approach to risk stratification and to assessment of target organ damage at the cardiac or renal site). Some problems may also have affected patient recruitment, because, unfortunately, data were not collected by considering hypertensive patients consecutively seen in physicians' offices. Patients were included in the database only if they fulfilled ESH guidelines for indications to ABPM performance, such as suspected white-coat hypertension, resistant hypertension, assessment of dipper status, assessment of drug treatment efficacy and high-risk hypertension. This may limit the applicability of the information obtained. We also cannot exclude a selection bias for patients with suspected white-coat hypertension being classified in the low-to-moderate risk group and patients with uncontrolled hypertension being classified in the high-risk group. This potential bias could have led to a wider than expected dichotomization of risk between low-to-moderate and high-risk groups. Fifth, given that a low rate of BP control is one of the most important issues in treating hypertension, it is an interesting finding that the control rate of ABP was somewhat higher compared to the control rate defined from clinic BP measurement. For those defined as being at low/moderately high risk, BP control rate based on few clinic readings was 12.8% compared to 27.0% based on 24-h ABPM. The control rate of patients classified as being at high added risk was 23% with clinic BP versus 36% with ABP, respectively. Again, a regression towards the mean phenomenon might have come into play in explaining such a difference between clinic and ambulatory data. However, it should be emphasized that the finding of a better control of ambulatory than of clinic BP by treatment is not in line with previous observations, obtained in more controlled conditions. In the above-mentioned meta-analysis of drug studies including both clinic and ABP measurements, the opposite was found [27] (i.e. the control rate of ABP was always less than the control rate of clinic BP). An additional problem affecting the assessment of BP control in the study by Gorostidi et al. [18], focusing on patients at high cardiovascular risk, is related to the cut-off value selected to identify normality targets for clinic and for ABP. A cut-off value < 130–125/80 mmHg for defining normal 24-h average ABP was proposed in relation to a cut-off value for office BP < 140/90 mmHg in a general population [34]. Both targets may be inappropriate for hypertensive patients identified as being at high cardiovascular risk, in particular for those with previous cardiovascular events in their history, with diabetes mellitus or with an impaired renal function. In these patients, a lower BP goal would be desirable, which suggests that a corresponding lower target should be considered also for 24-h ABP. However, no precise indications are available in this regard, and this important issue is presently investigated by outcome studies. Finally, it would have been of interest to quantify the rate of patients with a so-called white-coat hypertension (or isolated clinic hypertension) in this large general practice database. The issue of ‘isolated clinic hypertension’ is not quite settled, even though the evidence indicates that it is not a completely innocent condition but carries a somewhat increased risk [35]. Future analysis from the very large sample of patients included in the database built by Gorostidi et al. [18] might be useful to further clarify this issue. This could also be the case for another burning issue; namely, the prognostic implication of the so-called ‘masked hypertension’ (i.e. the condition characterized by a normal clinic BP and by an elevated out-of-clinic BP), which has been suggested to carry an increased risk of cardiovascular complications by a number of recent studies [35]. In conclusion, most of the evidence currently available on the diagnostic and prognostic value of 24-h ABPM parameters comes from studies including a variable number of individuals selected either from a general population or from well defined cohorts of hypertensive patients, whose sample size, however, was never comparable to the one characterizing this nationwide registry of general practice data. The possible importance of observations from the very large database built by Gorostidi et al. [18] is thus easily evident, and may depend not only on the possibility to improve the statistical power for calculations, but also on the ability to define the clinical relevance of different 24-h ABP profile patterns in a real-life setting from a wide population sample. If the methodological problems listed above are adequately managed, data from such a solid national BP database will offer major advantages for future cross-sectional as well as longitudinal research. At any rate, an important contribution provided by Gorostidi et al. [18] is the further demonstration of the clinical importance of the information provided by 24-h ABPM. In particular, the results reported by Gorostidi et al. [18] strongly support the recommendation of a wider use of ABPM in high-risk hypertensive patients.