‘Since the information which the pulse affords is of so great importance, and so often consulted, surely it must be to our advantage to appreciate fully all it tells us, and to draw from it every detail that it is capable of imparting’ F.A. Mahomed 1872 [1]

It is now possible to generate the ascending aortic pressure wave from the arterial pressure pulse, recorded noninvasively by applanation tonometry in the radial or carotid artery. This represents a blend of nineteenth century sphygmography with cuff sphygmomanometry, and is made possible by introduction of high fidelity tonometers, by characterization of arterial hydraulic properties in the upper limb and neck, and through application of mathematical engineering techniques in modern computer systems. This review will consider historical development, theoretic background, present status and future potential, as well as comparing this technique with radial and carotid tonometry alone, and with analysis of flow pulse and volume pulse waveforms as determined by Doppler or photoplethysmographic techniques.

The arterial pulse is the most fundamental sign in clinical medicine, and has since antiquity been identified with the physician and the art of medicine. Palpation of the pulse forms the crest of the Royal College of Physicians of London, which was established to improve the scientific basis and practice of medicine and in an era when William Harvey, as anatomist to that college, wrote his classic text ‘de Motu Cordis…’ [2]. William Bright [3] based his diagnosis of high blood pressure on ‘hardness’ of the pulse, and on the pressure required to extinguish the pulse. A scientific basis only arose after Marey [4], and then Mahomed [1] developed graphic methods to record the arterial pulse. By the beginning of the twentieth century, sphygmography was well established in medical journals and in medical textbooks and had been used to describe heart block and effects of antianginal medication as well as hypertension and other conditions [5–9]. In life insurance examinations sphygmocardiography was widely used for detecting persons with ‘arterial senility’ and increased risk of premature death [10]. Unfortunately sphygmography lapsed with introduction of the cuff sphygmomanometer, which provided numbers for the extremes of the pulse, and a veneer of scientific accuracy.

Frederick Akbar Mohamed established the foundation of pulse wave analysis in a short medical lifetime from 1872 to 1884. He described the normal radial pressure waveform and showed the difference between this and the carotid wave [1]. He showed the effect of high blood pressure on the radial waveform, and used the waveform to describe the natural history of essential hypertension, and the difference between this and chronic nephritis [7, 8]. He also described the effects of arterial degeneration with ageing on the arterial pulse [7]. These features were identified and utilized in the life insurance studies of the late nineteenth century [10].

Mahomed's sphygmogram, and the popular Dudgeon sphygmogram which followed, and which was used by Sir James MacKenzie [9] were mechanical devices, awkward to use and prone to artifact. Modern tonometer systems are piezo-electric and are far more accurate, reliable, and easy to use. While originally introduced clinically to measure intraocular pressure, they have been adapted for vascular use by Drzwiecki [11], Millar and others [5, 6].

While Mahomed was the first to recognize the difference between pressure waves in central and peripheral arteries, McDonald [12] was responsible for explaining this phenomenon on the basis of wave reflection, and for introducing transfer functions to characterize properties of vascular beds in the frequency domain, and (with his colleague J.R. Womersley) for establishing the validity of assuming linearity in the arterial tree [13]. The work of McDonald, Womersley, Taylor and others, originally from Harvey's own hospital (St Bartholomew's, London) has led on to the techniques described here for pulse wave analysis.

The technique of noninvasive aortic pulse wave analysis, as described here, depends on accurate recording of the radial pressure wave, its calibration against brachial pressure, then generation of the ascending aortic pressure waveform through use of a generalized transfer function in a computerized process. Ascending aortic waveforms are ensemble averaged into a single calibrated wave whose different features can be identified automatically with clinically important measures of pressure and time intervals measured and printed out in an interpretive report (Figure 1). Steps in the process are described below.

Figure 1 The Sphygmocardiograph: computerized report on analysis of radial artery and synthesized aortic pressure waves. A series of radial artery pressure waves, recorded over an 8 s period (upper continuous tracing) are used to synthesize a series of ascending (more ...)

Validation of our own specific methods have been described by groups in Japan [26], Sweden [27], and the USA [28]. Validation studies confirm that the radial and brachial artery pressure values are very different to those in the ascending aorta especially with tachycardia, during vasodilator infusion and during interventions such as the Valsalva manoeuvre, while the calculated values for systolic, diastolic and pulse pressure fall within the range specified for the AAMI and B.H.S. requirements [29], at least for directly recorded (and calibrated) radial waveforms.

Claims for accuracy of this technique for noninvasively recorded waveforms need to be guarded. Calibration of systolic and diastolic values for the tonometric wave with a cuff sphygmomanometer relies on the innate accuracy of this method for determining intra-arterial pressure. If one accepts the innate inaccuracy of the cuff and Korotkov method for determining intra-arterial pressure [30], then use of this technique is a step forward towards greater accuracy in determining central pressure from what currently exists.

Formal studies show good repeatability and low interobserver variation for tonometry and derived indices [31–33, Table 1]. These results accord with our own observations. Persons interested in using these techniques are advised to undertake their own studies of reproducibility, repeatability and interobserver error, as well as comparing results from both radial arteries and from the carotids. Persons would reasonably be advised not to proceed with use of such instrumentation unless they and their technical staff could achieve results similar to those in Table 1.

Table 1 Reproducibility

The curious pressure pulse waveform in infants, and paradoxical similarity to the elderly is readily explained on the basis of early return of wave reflection from peripheral to central vessels. In infants this is due to short body length, despite the aorta being very distensible and having relatively low pulse wave velocity. In the elderly, the aorta is very stiff, and its high pulse wave velocity accounts for earlier return of wave reflection despite normal adult body dimensions. Infants have relatively long left ventricular ejection periods, despite their small bodies and relatively fast heart rate [5, 35]. Experimental animals of similar size and heart rate have much shorter ejection duration, so that secondary reflected waves are apparent in diastole rather than systole [5].

The generalized transfer function used by ourselves and others to generate ascending aortic pressure waves [17–19, 22, 23] have been established in fully grown adults, and are not applicable to children whose shorter body lengths affect wave travel and reflection in the upper limb as well as in the trunk and lower limbs (whence arises most wave reflection).

Figure 2 Change in contour and amplitude of pressure waves recorded in the radial artery (left) and carotid (right) in normal subjects between the first and eighth decade of life. Data are ensemble averaged into decades from 1005 different subjects. From Kelly (more ...)

Augmentation is the boost to late systolic pressure after the initial systolic shoulder, and can be expressed in mmHg or in relative terms as pulse pressure divided by the pressure to the first systolic shoulder, or as augmentation divided by pulse pressure. For the aortic wave in Figure 1, augmentation is measured as 20 mmHg or as 159% (or as 37% of pulse pressure).

In late adolescence, amplitude of the ascending aortic systolic pressure wave approximates the first, with augmentation indexes around 100%. This appears to increase progressively throughout life with values of around 160% found in elderly persons. In very elderly persons, radial pressure wave augmentation is close to 100%, with a secondary pressure wave of like amplitude to the first, having risen from around 40% in late adolescence.

Progressive increase in augmentation during adult life is attributable to early return of wave reflection from the trunk and lower body. Wave reflection occurs at the myriad of arterial terminations throughout the body where low resistance conduit arteries join high resistance arterioles. Early return of reflection is caused by arterial (and particularly aortic) stiffening, which increases pulse wave velocity [5]. Such aortic stiffening is apparent as increase in pulse wave velocity, which has also been demonstrated by our group in normal western subjects and in normal Chinese, who have very low prevalence of atherosclerotic disease [5, 20]. Aortic wave velocity more than doubles between age 20 and 100 years [5, 20]. This is a normal ageing phenomenon, and clearly is responsible for the increase in pulse pressure and in systolic pressure which is seen with advancing years [38]. The phenomenon is accentuated by hypertension and by arterial degeneration, and is modified by cardiac as well as by arterial disease, as discussed below.

While widespread improvement of endothelial function may explain decreased augmentation in physically fit individuals, there is no known relationship between endothelial dysfunction and pulse wave contour. The demonstrated impairment of upper limb endothelial function with age in different diseases is not associated with definite alterations in pulse wave velocity or other indices of arterial stiffness in the upper limb [5, 21, 24].

It appears that the upper limb generalized transfer function remains valid for generation of central from peripheral pressure in the presence of hypertension, since distensibility of the brachial artery and its radicles, and transfer function changes little with elevation of arterial pressure [5, 21, 24].

Figure 3 Above) Systolic dysfunction and heart failure. Differing effects of wave reflection (depicted on third tracing from top) on brachial and aortic pressure waves (top two tracings) and on ascending aortic flow (bottom trace) with development of elevated (more ...)

In systolic left ventricular dysfunction, ejection duration is decreased on account of early wave reflection having a greater effect on flow than on pressure [56, 57]. This phenomenon accounts for the appearances of a dicrotic pulse in systolic heart failure, with ejection duration shortened, with central pressure augmentation minor or absent, and with prominence of wave reflection in diastole [5, 6, 58, 59]. We have used this phenomenon to separate systolic and diastolic dysfunction, and to gauge the severity of systolic dysfunction [5, 60] as well as to separate systolic from diastolic left ventricular dysfunction as the predominant cause of cardiac failure. Value is enhanced through use of glyceryl trinitrate which, through reduction in wave reflection, decreases left ventricular load and increases cardiac output and left ventricular ejection duration; capacity to increase ejection duration is related to the severity of left ventricular dysfunction [60].

Pulse wave analysis uncovers the favourable effects of vasodilator drugs which are either not apparent or not fully apparent from conventional cuff recordings of arterial pressure. Decrease in amplitude of the secondary systolic wave in the radial artery, even with maintenance of the initial peak indicates decrease in ascending aortic and left ventricular systolic pressure (Figure 4). Such a change in Murrell's radial artery tracing after nitroglycerin [61] (Figure 5), explains the antianginal effect of this drug is due at least in part to decrease in left ventricular load [5, 6, 62, 63] (Figure 4). Anggard (personal communication) uses this change to quantify effects of nitroglycerin and other vasodilators. We prefer to determine these effects from the synthesized aortic pressure pulse, partly for convenience and to provide other indices of ventricular vascular interaction, and partly because it is often not possible to identify the secondary systolic shoulder on the radial pressure wave after nitroglycerin. Anggard's approach is to measure change in relative height of the dicrotic notch in the finger photoplethysmograph.

Figure 4 Effects of nitroglycerin on contour of the ascending aortic and brachial pressure waveforms. Pressure waves recorded in the ascending aorta (above), and brachial artery (below), of an adult man at cardiac catheterization under control conditions (left) (more ...)

Figure 5 Radial artery sphygmographs showing the effects of nitroglycerin as causing progressive reduction in the late systolic shoulder of the radial artery pressure pulse. Reproduced from the first description (by Murrell in 1879) of the therapeutic value of (more ...)

Our initial observation of relative maintenance of peripheral systolic pressure despite considerable fall in central aortic and LV systolic pressure after vasodilators such as nitroglycerin has been confirmed [64, 65], and indicates a more potent effect of such drugs on the heart than measured by conventional methods. The findings are explained on the basis of decrease in wave reflection from the lower body. The fact that brachial transfer function is little altered by nitroglycerin [17] and that the ascending aortic pressure wave during nitroglycerin infusion can accurately be determined from the radial [27, 28] indicates that vasodilation has a lesser effect on wave reflection in the upper limb than in the lower body.

We regularly utilize the synthesized aortic pulse to monitor the effects of drug therapy in hypertension, angina pectoris, and cardiac failure. Our principal aim with drug therapy is to reduce wave reflection, and thereby to reduce aortic augmentation and so, aortic and left ventricular systolic pressure, and to optimize indices of ventricular/vascular interaction.

In systolic failure (left ventricular failure due to impaired myocardial contractility) ejection duration of the aortic pulse waveform is shortened, as described previously; through reducing wave reflection, vasodilator agents prolong ejection and thereby increase stroke volume. Such prolongation of ejection duration can be seen on the radial, carotid, and derived aortic pulse [5, 60]. The degree of prolongation is related to severity of cardiac failure, and has been suggested as a measure of this [60]. Japanese workers have used aortic flow waveforms to describe the ill effects of early wave reflection on the ascending aortic flow and to describe the beneficial effects of reduction in wave reflection with vasodilator agents [66, 67].

Pulse wave analysis is particularly useful in studying beneficial effects of nitroglycerin, since this drug in conventional doses has no detectible effects on arterioles, or on peripheral resistance. Beneficial effects in reducing wave reflections are readily explained on the basis of dilating small conduit arteries [5, 62]. At present, interaction between sildenafil (Viagra) and nitrates is of great clinical importance, and the rather draconian restrictions recommended by the joint ACC/AHA Committee [68], have been made without consideration of pulse wave analysis, and on the basis of relatively large fall of blood pressure in volunteers when full doses of each were used concurrently. Our own preliminary studies of sildenafil/nitrate interaction, based on pulse wave analysis, carries no surprise, and suggests that sildenafil potentiates the effect of nitroglycerin approximately two fold, and prolongs this for up to 8 h [69]. Pulse wave analysis, with attention to late systolic augmentation, should provide a method for monitoring this interaction, and should permit titration of nitroglycerin (if required after sildenafil), at least by the intravenous and transdermal route.

Potential ill effects of drugs can be apparent from the arterial pressure pulse. β-adrenoceptor blocking drugs tend to increase wave reflection [70] as well as ejection duration and in some individuals cause substantial increase in central augmentation and in aortic and left ventricular systolic pressure [5]. This is rare; the beneficial effects of β-adrenoceptor blocking drugs in angina pectoris and diastolic left ventricular dysfunction are attributable to disproportionate increase in diastolic (compared with systolic) duration with bradycardia [5, 71].

Calcium channel antagonists reduce aortic systolic pressure augmentation through their vasodilator effect, and this is desirable. However, unlike nitrates, they may prolong ejection duration and decrease diastolic period, even in persons without cardiac failure [5]. This effect can predispose to angina pectoris by reducing coronary perfusion time, especially if tachycardia occurs concurrently. This effect could well explain provocation of angina and of myocardial ischaemia with short-acting dihydropyridines.

At present, we employ pulse wave analysis routinely in clinical practice. The cuff sphygmomanometer provides information only on the limits between which pressure fluctuates through the cardiac cycle in the upper limb arteries; this is distorted, inaccurate and incomplete information on left ventricular load and function. At best (with generation of the ascending aortic waveform), pulse wave analysis provides more accurate left ventricular systolic and end-systolic pressure, and ejection duration, with the ability to determine other indices of ventricular/vascular interaction. Generation of a sphygmocardiograph report, using a hand held tonometer takes no longer than an ECG.

We find this useful in hypertension – in diagnosis and therapy. We can confirm the diagnosis of hypertension by noting the presence of substantial late systolic augmentation in the aortic pressure wave, and question the diagnosis when brachial systolic pressure is high, but this feature is not present [1, 5, 7]. We can make a diagnosis of spurious systolic hypertension of youth, sparing other investigations and lifting career restrictions that would otherwise be placed on a perfectly well young man [5, 72]. In patients with systolic hypertension we would be more concerned about a patient with substantial augmentation than one without. In patients with cardiac failure, diabetes mellitus, nephropathy, in the grey area of systolic 130–140 mmHg, a decision of when and how aggressively to treat, is aided by interpretation of the arterial pulse. In treatment itself we can gauge peripheral vasodilation from fall in mean pressure, wave reflection from fall in augmented pressure (in patients without heart failure), and beneficial effect of a β-adrenoceptor blocking drug from increase in diastolic period as well as decrease in heart rate (Figure 6).

Figure 6 Sphygmocardiograph reports in a patient with hypertension above (before) and below (after) treatment with perindopril. Reduction in aortic systolic pressure of 27 mmHg is attributable to reduction of augmentation (17 mmHg) and of mean pressure (10 mmHg). (more ...)

For patients in cardiac failure, we can separate predominantly systolic from predominantly diastolic causation from wave contour and measurement of ejection duration [5]. Severity of systolic dysfunction can be gauged from contour and ejection duration, and additionally from the change induced by nitroglycerin.

In patients with angina pectoris, findings of inappropriately long duration of ejection (and short diastole) at rest or with exercise suggests that coronary narrowing may be sufficiently minor not to warrant angioplasty or bypass surgery [5, 71]. Such a finding may argue for deferring coronary angiography, and for intensification of medical therapy. Response to such therapy can be gauged from pulse wave contour.

Other uses for pulse wave analysis are certain to be encountered in the future. We are puzzled by findings of inappropriately high augmentation in some persons, suggestions of premature arterial ageing, but have not been able completely to explain this from conventional risk factors. We have not yet seen anything remarkable in the pulse waves of individual patients with coronary atherosclerosis or diabetes mellitus that might assist in screening of asymptomatic persons, but recognize that such data may become available [51, Brooks & Yue, personal communication].

Possibilities are substantial, but have not been explored systematically to date. We have seen (as did Mahomed [7] and his contemporaries [10] innate cardiovascular events develop in persons with greater than expected augmentation for age (or heart rate, arterial pressure), and have noted that such persons often have adverse family histories. We believe that these persons have premature arterial degeneration that has not been uncovered by other methods. The Cornell Group [73] has shown a strong association between carotid augmentation and left ventricular hypertrophy, and are exploring the value of radial, carotid and aortic augmentation in the ongoing STRONG study. The Framingham Group reported association of pulse wave contour with stroke in 1981 [74], but with dubious technology; review with new technology is presently being undertaken at Framingham.

It is ironic that Mahomed, who first made a quantitative sphygmocardiogram, and first explored clinical utility, described the ominous prognostic implication of late systolic augmentation of the radial pulse in an asymptomatic, apparently normal individual ‘by his pulse you will know him’ [8]. Mahomed was in the process of arranging a multicentre cumulative clinical record in association with the British Medical Association at the time of his premature death at age 35 years. This is the forerunner of modern epidemiological studies and of the American College of Cardiology National Cardiovascular Data Registry. There is little doubt that had it proceeded, pulse waveform analysis would have been included, and as much might now be known about the arterial pulse as a prognostic tool as is known about sphygmomanometric blood pressure – a method which was introduced more than 10 years after Mahomed's death. Likewise, use of pulse wave analysis by life insurance companies lapsed when Fisher's work in 1917 established the value of the Riva Rocci technique [10].

While particular emphasis is given here to the most sophisticated form of pulse waveform analysis (generation of the ascending aortic pressure wave), comment needs to be made on the utility of analysing other pulse waveforms.

The radial pressure pulse contains all the basic information from which the ascending aortic pulse is generated. All that is determined quantatively from the aortic pulse can be inferred from the radial. Duration of systole and diastole are measured directly from the radial pulse, and its change with ageing and with drugs can usually be followed in the same way as the aortic, though it is more difficult to identify inflections in late systole on the radial pulse. Aortic and left ventricular systolic pressure cannot be determined, nor end systolic pressure, nor the indices from which Buckberg's subendocardial viability ratio is calculated [5, 74, 75]. Analysis of the radial pulse alone may be adequate for many, and certainly improves on sphygmomanometric measurement of brachial systolic and diastolic pressure alone.

Cohn and colleagues of Minneapolis have developed a method for analysis of the diastolic part of the peripheral pressure waveform [76]. This does not consider pulse pressure nor the period of systole, and has practical and theoretic problems. The diastolic part of the pulse is the most variable when recorded noninvasively. End systolic pressure (the beginning of diastole) is some 12 mm lower in the radial artery than in the aorta [27, 28]. The method requires noninvasive calculation of cardiac output from arterial pressure and a model of the arterial tree, and seeks to unscramble the incomplete dampened fluctuation due to wave reflection from the exponential pressure decline during diastole. A commercial product has been developed and is currently under evaluation.

We use the carotid pulse waveform to check against the radial when this appears unusual. Many use the carotid pulse exclusively as a surrogate of the ascending aortic. Certainly its contour is much closer to the aortic than the radial. However it cannot be calibrated with the same degree of confidence as the radial, intra and interobserver variability is greater than the radial, and many persons find the procedure quite uncomfortable. Baro-receptors are apt to be stimulated with reflex cardiac slowing. We are also concerned about plaque disruption; though never reported in the literature, this is at least a theoretic possibility.

Doppler flow pulses are recorded from the carotid and peripheral arteries to give variable information, unavailable with tonometry, on arterial narrowing [5]. Stenosis is inferred from increased flow velocity and/or spectral broadening. The Doppler technique is particularly valuable for measuring ascending aortic flow and for exposing the characteristic patterns of late systolic reduction in flow that indicate the presence of impaired cardiac contractility and systolic heart failure [66, 67]. Unusual flow patterns may also be useful in the diagnosis of ‘hypertensive’ or ‘idiopathic’ hypertrophic cardiomyopathy [5, 77, 78].

The photoplethysmogram is widely used for measuring finger volume pulsations in intensive care wards and during operative procedures. Little attention is usually paid to the waveform. Takazawa et al. [79] have shown that this resembles the carotid pressure pulse, and bears a resemblance to the aortic pulse. Japanese investigators have paid particular attention to the second derivative of the finger photoplethysmograph [79], and use this as an index of vascular ageing and drug effect. This field, highly developed in Japan, warrants more attention in the western world.

The potential value of this technique is apparent from previous argument. Weakness is in validation. Special emphasis is being given to this at present by our group so that others will know what degree of confidence is warranted. Articles still appear in reputable journals claiming there is insignificant difference between aortic and brachial pressure [80, 81]. This may be the case under control conditions in older persons such as those studied after diagnostic coronary angiography, but it does not apply to the young, nor to older persons during interventions, with tachycardia, with heart failure, or with vasodilator drug therapy. The technique described here appears to retain accuracy under a variety of interventions [17, 18, 27, 28]. With respect to differences between central and peripheral systolic pressures, the most frequently quoted article [82] claimed a difference of mean 0.8, s.d. 3.5 mmHg. One has to doubt the way that data were selected when the most recent report (which is typical of ‘blinded’ studies) gave a mean offset using the Korotkov method of 7 mmHg, with 95% confidence intervals of almost 50 mmHg [83]. Results of this study are similar to our own [30], and those of Cohn et al. [77].

In evaluating the system we have used, Cameron et al. [84] suggested that a linearized correction could be applied to determine aortic systolic pressure from brachial. Our data from over 15 000 reports fits in well with those of Cameron et al. (Figure 7) but we are dissatisfied with the large scatter which has 95% confidence limits embracing 30 mmHg. Our technique, which incorporates pulse waveform, results in very close approximation of estimated to measured aortic systolic pressure, and with small scatter (Figure 7).

Figure 7 a) Data from 15 533 separate Sphygmocor reports in 1604 different subjects/patients, showing the relationship between estimated aortic systolic pressure (ordinate mmHg), and brachial artery systolic pressure, determined by conventional cuff sphygmomanometry (more ...)

The greatest problem in accurate noninvasive estimation of aortic pressure relates not to the tonometric pressure wave method or convolutional process but to calibration from the sphygmomanometer cuff. Data presented here argue convincingly that the radial waveform, calibrated from intra-arterial, invasively recorded arterial pressure, can be used to generate an accurate ascending aortic pressure waveform under a variety of conditions. But the same cannot be said for calibration against the sphygmomanometer cuff. Many studies have shown relatively huge differences between intra-arterial and cuff blood pressure measurements [30, 77, 83]. Any attempt to determine ascending aortic pressure noninvasively will be at the mercy of cuff inaccuracies. However it may be argued that epidemiological studies have been based on the cuff method, rather than on invasive pressures, so that the former rather than the latter constitutes the ‘gold standard’. If this argument were accepted, we could claim that the ascending aortic pressure waveform that we generate and indices therefrom may overcome some of the shortcomings of the cuff sphygmomanometer, and so improve on the information that it has provided over the past century. Such a proposition is worthy of consideration and test.

This review is based on Mahomed's proposition: ‘Surely it must be to our advantage to appreciate fully all the pulse tells us, and to draw from the pulse all that it is capable of imparting’. But it is worth recalling William Harvey's second essay to Jean Riolan [85], and his strong views on the danger of extreme distrust on the one hand or of over-enthusiasm on the other, in evaluation of a new concept.

Michael O'Rourke is a director of PWV Medical, a company developing products for pulse wave analysis.