Accuracy of blood pressure measurement: sphygmomanometer calibration and beyond

The importance of accurate blood pressure measurement in the diagnosis of arterial hypertension, and in the precise titration of antihypertensive treatment, is largely acknowledged in clinical practice. The relevance of this issue has recently been further emphasized by hypertension management guidelines [1–3]. Health care givers as well as patients are now aware of the need to follow carefully standardized procedures when measuring blood pressure, aimed at achieving the highest possible accuracy of blood pressure estimates. Conversely, the possible negative impact that poor maintenance of blood pressure-measuring devices may have on the accuracy of patients' blood pressure assessment, is still largely underestimated or disregarded, even by physicians and nurses. This phenomenon should be considered with concern because a technical problem such as sphygmomanometer calibration might importantly affect the effectiveness of hypertension management and, surprisingly enough, its influence may be greater now than some years ago. This is due to the more and more frequent tendency among hospital staff and general practitioners to shift from use of the classical mercury sphygmomanometer to the adoption of other blood pressure-measuring techniques, such as aneroid sphygmomanometers and electronic devices. In this issue of the journal, Turner et al. [4] provide further evidence to support the clinical relevance of sphygmomanometer calibration. By means of mathematical simulations, the authors have extended the results of previous studies [5,6], and have estimated the combined effects exerted on the detection of arterial hypertension by random blood pressure variability (including intra-individual blood pressure variability as well as random measurement errors) and by systematic errors due to inadequate sphygmomanometer calibration. The main findings of their mathematical simulation are summarized in Fig. 2 of their paper [4]. The percentage of subjects correctly labelled as hypertensives (a measure of the sensitivity of the diagnostic procedure) increased significantly with the increase in the number of doctor's visits, when correctly calibrated blood pressure-measuring devices were employed. This increase reflects a reduction in random error obtained by averaging the blood pressure values measured over successive visits. Conversely, when non-calibrated devices were used not only the sensitivity of the diagnostic procedure was lower at the first visit, but also the increment in sensitivity with repeated visits, observed over subsequent measurements when using calibrated manometers, was less pronounced. This is because the non-calibrated sphygmomanometer introduces a systematic error in the blood pressure estimate that cannot be reduced by averaging several measurements. Because of this, the sensitivity curves typical of calibrated and of non-calibrated devices displayed a tendency to diverge with the increase in the number of visits. A specific feature of the study by Turner et al. [4], which deserves some discussion, is represented by the mathematical model employed to assess the clinical relevance of sphygmomanometer calibration, a model which may not necessarily be familiar to some researchers. The authors used a simple Monte Carlo simulation to estimate the combined effects of systematic sphygmomanometer error and day-to-day random variability of blood pressure measurements on the over- and under-detection of hypertension in adults aged 18 years and older. Monte Carlo methods are a class of computational algorithms for simulating the behaviour of various physical and mathematical systems. These techniques are used in situations where the analytic solution of a problem is either hardly achievable or time consuming. In other words, a Monte Carlo algorithm is a method used to find solutions to mathematical problems (which may be characterized by many variables) that cannot easily be solved, for example, by integral calculus, or by other numerical methods. Its efficiency when compared with other numerical methods increases with the increase in the dimension of the problem to be solved. A Monte Carlo technique can be differentiated from other simulation methods because of its stochastic nature (i.e. because it is characterized by non-deterministic algorithms, through use of random numbers). Monte Carlo techniques have a long history in mathematics, with their earliest application probably being represented by Buffon's Needle in 1733. The name ‘Monte Carlo’ was given by Nicholas Metropolis (in honour of his uncle Ulam, who was a gambler) during the Manhattan Project of World War II, and specifically refers to the famous casino in Monaco. The use of randomness and the repetitive nature of the analytic process characteristic of a Monte Carlo model are both analogous to the gambling activities conducted in a casino. Perhaps the most famous early use of this model was made by Enrico Fermi in 1930, when he employed a random method to calculate the properties of the newly-discovered neutron. Monte Carlo methods were central to the simulations required for the Manhattan Project, although they were strongly limited by the computational tools available at that time. It was only after electronic computers became available (i.e. from 1945 onwards) that Monte Carlo methods began to be studied in depth. In the 1950s, they were used at Los Alamos for some early work aimed at the development of the hydrogen bomb, and became popular in the fields of physics and operations research. The Rand Corporation and the US Air Force were two of the major organizations responsible for funding and disseminating information on Monte Carlo methods during this period, which found wide applications in many different fields. The use of Monte Carlo methods requires large amounts of random numbers and, indeed, it was their use that spurred the development of pseudorandom number generators, which were far quicker to use than the tables of random numbers previously used for statistical sampling. Monte Carlo methods have diverse applications, ranging from the simulation of complex physical phenomena, such as radiation transport in the earth's atmosphere or simulation of the esoteric subnuclear processes in high-energy physics experiments, to daily entertainments, such as the simulation of a Bingo game. The analogy of Monte Carlo methods to games of chance is intriguing, but the ‘game’ when addressing a scientific problem is a physical system, and the outcome of the game is not a pot of money or a stack of chips (unless simulated) but rather a satisfactory solution for the problem under assessment. Factors involved in the accuracy of blood pressure measurements: device calibration and beyond The classic approach to blood pressure measurement is based on use of the mercury sphygmomanometer, which has long been regarded as the gold standard for blood pressure readings in a clinical setting since its first description by Scipione Riva-Rocci, published in 1896 on the Gazzetta Medica di Torino [7,8]. Its use is based on the theory of the occlusion technique: the pressure of air within the arm cuff is balanced by the weight of a mercury column contained in a graded glass tube. The height of the column therefore measures the air pressure in the arm cuff in mmHg. In spite of its worldwide use, the conventional sphygmomanometric measurement of blood pressure through the Riva-Rocci manometer coupled with use of Korotkoff sounds is known to be affected by a number of limitations. A first important problem is the high between-measurement variation in office blood pressure readings, which reflects both the physiological variations in blood pressure level and frequently also errors in the technique of measurement, due to conditions of measurement, arm position, observer's attitude. In 1964, Geoffrey Rose [9] classified the observer error into three categories: (i) systematic errors, which lead to both intraobserver and interobserver error; (ii) terminal digit preference, which results in the observer rounding off the pressure reading to a preferred digit, most often to zero [10]; and (iii) observer prejudice or bias, leading the observer simply to adjust the blood pressure readings to meet his or her preconceived notion of what the blood pressure should be. All these problems contribute to the between-observer variability in the accuracy of blood pressure readings, mostly due to differences in skill and training [3]. Other factors, not of a technological nature, can introduce systematic errors in sphygmomanometric blood pressure measurements. These include errors in arm cuff positioning, as well as the choice of an inappropriate cuff-size, in relation to patients' arm circumference. The latter problem has recently received more attention due to the progressive increase in the prevalence of obesity and overweight in developed countries, without a parallel increase in the frequency of large arm-cuff use [11]. A second major problem affecting the accuracy of sphygmomanometric blood pressure readings is the discrepancy between blood pressure values obtained in the office and those measured outside of a clinic setting (i.e. the so-called ‘white-coat effect’). This phenomenon shows high inter-individual variations, limited reproducibility and its prognostic importance is still a matter of controversy, but it certainly interferes with the assessment of subjects' ‘true’ blood pressure levels [12–16]. Beyond all these factors possibly involved in determining the accuracy of blood pressure measurements, another critical issue in this context is represented by performance of a regular calibration of the device employed. This is the case for mercury, oscillometric and electronic devices, even with important technical differences. Given the simple structure of the mercury device, its maintenance is relatively simple [3], and a loss of calibration is observed rarely. The latter is mainly due to mercury oxidation, a process that occurs in a relatively slow manner. When such a phenomenon is observed, the instrument can be easily recalibrated by simply replacing the mercury in the reservoir. This procedure, even if some attention is needed due to mercury toxicity, can be easily carried out by the sphygmomanometer users, and does not require any intervention by the manufacturer. However, increasing concerns about the toxicity of mercury for individuals using or servicing mercury sphygmomanometers, and, more importantly, the great concern about the toxic effects of mercury on the environment, have determined a progressive shift towards other kind of blood pressure-measuring devices [17,18]. The most commonly employed alternative to mercury manometers is represented by aneroid sphygmomanometers. As indicated by the name itself (aneroid means without fluid, in the Greek language), these devices do not contain mercury, and are generally of smaller size and easier to handle than the classic mercury manometers. Typically, they consist of a thin brass corrugated bellows that is expanded by the air pressure in the cuff. The expansion moves a pin resting on the bellows. The movement is mechanically amplified and transmitted to a dial. This bellows and lever system is mechanically more complex than the mercury reservoir and column [19]. Moreover, besides being prone to all the problems of the auscultatory technique (namely observer bias and terminal digit preference), aneroid manometers are easily damaged and can go out of calibration without detection [3], due to mechanical shocks and changes in environmental temperature related to everyday use. They usually lead to falsely low readings with the consequent underestimation of blood pressure. When calibrated against a mercury sphygmomanometer a mean difference of 3 mmHg is considered to be acceptable; however, 58% of aneroid sphygmomanometers have been shown to have errors greater than 4 mmHg, with approximately one-third of these having errors higher than 7 mmHg [3,20]. Aneroid sphygmomanometers must therefore be checked every 6 months against an accurate mercury sphygmomanometer over the entire pressure range, which can be carried out by connecting the aneroid sphygmomanometer via a Y-piece to the tubing of the mercury sphygmomanometer and inflating the cuff around a metal or wood cylinder. If inaccuracies or other faults are found, the instrument must be returned to the manufacturer or supplier for repair, given the difficulties which users often have to face when trying to re-calibrate them [3]. In practice, in spite of their easier use, this implies that these devices are often employed under suboptimal conditions, or even with evident calibration problems and loss of measurement linearity. The observer-related errors as well as the errors related to the inaccuracy of aneroid manometers, could in theory be eliminated by using automated electronic devices, such as those proposed for self blood pressure measurement at home [3]. Accurate automated sphygmomanometers, capable of providing print-outs of systolic, diastolic and mean blood pressure together with heart rate and the time and date of measurement, would eliminate errors of interpretation and abolish observer bias and terminal digit preference. Moreover, the need for observer training would no longer be necessary, although a period of instruction and assessment of proficiency in using the automated device will always be necessary. Another advantage carried by automated blood pressure-measuring devices is their ability to store data for later analysis and/or to allow for remote teletransmission. These favourable properties have promoted the diffusion of these devices, and a number of large research studies are already employing automated technology to measure blood pressure instead of the traditional mercury ‘gold standard’. However, most of these devices are based on the oscillometric method for blood pressure measurement. This implies that only mean blood pressure values are measured, whereas systolic and diastolic blood pressure are determined through proprietary and undisclosed algorithms, built in each device, and different between different manometers [3]. Moreover, the cuff deflation rate may be different in different devices, thus introducing another possible source of error in automated blood pressure measurements. All these problems may lead to significant differences in the blood pressure values provided by automated oscillometric devices produced by different manufacturers [3,21,22]. Thus, although automated devices may eliminate observer-related systematic errors, they may unfortunately introduce other ‘manufacturer-related’ errors. Moreover, most of the available automated devices were designed for self-measurement of blood pressure at home, and may not necessarily be suitable for clinical application. Evidence is still lacking concerning whether they will remain accurate with long-term use in a clinic setting, although some of them have been used successfully in hospital practice and/or in large hypertension studies. Furthermore, oscillometric techniques cannot measure blood pressure in all situations, particularly in patients with dysrhythmias such as rapid atrial fibrillation. Finally, many doctors do not trust automated measurements, and prefer manual blood pressure readings. To improve this situation, and to minimize problems of accuracy related to their use, only automated devices validated according to internationally accepted protocols should be considered [23–27]. Complying with this recommendation may help to reduce systematic errors, while the known stability of electronic transducers means that calibration needs only be performed at relatively long time intervals. The latter advantage applies also when electronic devices are used ‘manually’. Notwithstanding the blood pressure-measuring technique selected, it should be noted that the parameter under assessment (i.e. blood pressure) is by no means constant over time, and is characterized by a pronounced variability not only between different behavioural conditions, but also on a beat-by-beat basis at rest [28–30]. This means that, even when making use of a perfect blood pressure-measuring tool in ideal environmental conditions, the same observer will unavoidably measure different blood pressure values over successive visits and even at different times within the same visit. This is exemplified by calculations made in our laboratory over 10-min segments of continuous finger blood pressure recordings performed in healthy volunteers by a Portapres device (Finapres Medical Systems, Amsterdam, The Netherlands) [31]. Spontaneous systolic and diastolic blood pressure variability in the sitting position, expressed as the standard deviation of the average blood pressure level determined over this time window, amounted to 7 and 4 mmHg, respectively. In conclusion, given the crucial clinical importance of obtaining accurate blood pressure measurements, a great deal of methodological care is needed whenever assessing blood pressure values, also on the background of the spontaneous variability that characterizes blood pressure. This includes not only compliance with the procedures recommended in blood pressure measurement guidelines [3], but also regular performance of sphygmomanometer calibration, regardless of the type of blood pressure-measuring device employed.