Journal of Applied Physiology

Carotid artery pulse wave time characteristics to quantify ventriculoarterial responses to orthostatic challenge

Koen D. Reesink, Evelien Hermeling, M. Christianne Hoeberigs, Robert S. Reneman, Arnold P. G. Hoeks


Central blood pressure waveforms contain specific features related to cardiac and arterial function. We investigated posture-related changes in ventriculoarterial hemodynamics by means of carotid artery (CA) pulse wave analysis. ECG, brachial cuff pressure, and common CA diameter waveforms (by M-mode ultrasound) were obtained in 21 healthy volunteers (19–30 yr of age, 10 men and 11 women) in supine and sitting positions. Pulse wave analysis was based on a timing extraction algorithm that automatically detects acceleration maxima in the second derivative of the CA pulse waveform. The algorithm enabled determination of isovolumic contraction period (ICP) and ejection period (EP): ICP = 43 ± 8 (SD) ms (4-ms precision), and EP = 302 ± 16 (SD) ms (5-ms precision). Compared with the supine position, in the sitting position diastolic blood pressure (DBP) increased by 7 ± 4 mmHg (P < 0.001) and R-R interval decreased by 49 ± 82 ms (P = 0.013), reflecting normal baroreflex response, whereas EP decreased to 267 ± 19 ms (P < 0.001). Shortening of EP was significantly correlated to earlier arrival of the lower body peripheral reflection wave (r2 = 0.46, P < 0.001). ICP increased by 7 ± 7 ms (P < 0.001), the ICP-to-EP ratio increased from 14 ± 3% (supine) to 19 ± 3% (P < 0.001) and the DBP-to-ICP ratio decreased by 7% (P = 0.023). These results suggest that orthostasis decreases left ventricular output as a result of arterial wave reflections and, presumably, reduced cardiac preload. We conclude that CA ultrasound and pulse wave analysis enable noninvasive quantification of ventriculoarterial responses to changes in posture.

  • hemodynamics
  • orthostasis
  • systolic time intervals
  • vascular ultrasound

carotid artery ultrasonography is an established source of clinical and experimental information. Measurement of the structural and functional properties of the carotid arterial segment yields a host of indexes for the assessment of cardiovascular risk, vascular adaptation, and therapeutic efficacy (3, 17, 22, 25, 37). Because it is noninvasive, carotid artery ultrasound is particularly suited to the study of large populations, but it could be equally valuable in monitoring cardiovascular remodeling and therapeutic progress in individual patients. Although the close relation between arterial and cardiac function has been well recognized and has been investigated extensively in experimental and clinical studies (3, 9, 15), to the best of our knowledge, few attempts have been made to quantify left ventricular function by means of carotid artery ultrasound.

The majority of approaches to noninvasive evaluation of left ventricular function utilize temporal characteristics of the central or peripheral arterial pulse wave (12, 26, 35) or cardiac Doppler signal analysis (32, 33, 38). In general, reduction of ejection duration and concomitant lengthening of the isovolumic contraction and relaxation periods (ICP and IRP, respectively) are considered indicative of ventricular dysfunction. The preejection period (PEP)-to-ejection period (EP) ratio (PEP/EP) is widely used to quantify left ventricular function, because it can easily be obtained from simultaneous continuous ECG and blood pressure recordings (12, 26, 35) or accelerometry in pacemaker recipients (24). However, PEP as obtained by these and related methods comprises electromechanical delay, ICP, and arterial transit time, depending on the site of measurement. Hence, PEP/EP is a complex function of electrophysiological and contractile cardiac properties and arterial stiffness. To exclude the spurious components in PEP and related indexes, cardiac Doppler and tissue Doppler techniques have been employed to discriminate systolic time intervals. Tei et al. (32) discriminated the ICP, IRP, and EP by timing analysis of mitral and aortic flow velocity patterns. In a comparative study in normal subjects and patients with dilated cardiomyopathy, the myocardial performance index [MPI = (ICP + IRP)/EP] and the ICP-to-EP ratio (ICP/EP) had greater discriminatory power to identify left ventricular dysfunction than mitral deceleration time, early-to-atrial peak velocity ratio (E/A), PEP/EP, stroke index, and cardiac index (32). Using echo/phonocardiography, Rhodes et al. (26) introduced a noninvasive estimate of the rate of left ventricular isovolumic pressure development (dP/dt) based on approximation of isovolumic pressure development and measurement of ICP. This method, however, may be particularly sensitive to error because of the approximation of ventricular end-diastolic pressure.

Because the temporal characteristics of the carotid artery diameter match those of the local transmural pressure waveform (2, 12), pulse wave analysis applied to the pulsatile carotid artery diameter (distension waveform) obtained by ultrasound may provide an interesting alternative. Recently, our group showed that the onset and duration (ICP) of isovolumic contraction can be determined automatically from the late diastolic phase of the carotid artery distension waveform (34). This allows us to monitor central arterial and left ventricular hemodynamics by a single noninvasive measurement technique. To test the feasibility of extracting systolic time intervals (specifically ICP and EP) from the distension waveform, we obtained M-mode recordings of the common carotid artery diameter in a group of young healthy volunteers. In addition to extraction of systolic time intervals, analysis included arterial transit time by identification of the lower body peripheral reflection (PR) wave in early diastole (19). Normal orthostatic stress, i.e., a change in position from the supine to the sitting position, was induced in these subjects to alter arterial and cardiac loading conditions to establish the precision and the discriminatory properties of carotid artery pulse wave analysis. In normal subjects, orthostatic stress induces reduced cardiac preload and increased peripheral resistance, arterial blood pressure, and heart rate but has little effect on intrinsic ventricular contractility (6, 8, 10, 11, 16).



Carotid artery distension waveforms were obtained from 21 presumed healthy volunteers [19–30 (mean 21) yr of age, 10 men and 11 women]. None of the subjects had a history of cardiovascular disease, diabetes, hypercholesterolemia, or established hypertension, and none was receiving medication that would affect cardiovascular function. The study was approved by the Medical Ethical Committee of Maastricht University and the Academic Hospital Maastricht, The Netherlands. All volunteers gave written informed consent before enrollment in the study.


Measurement sessions were held in the afternoon in a quiet and temperature-controlled (20–22°C) room. Before initiation of any measurement, the subjects were allowed to acclimatize for 10 min in the supine position (27). The subjects were studied successively in the supine (baseline) and the sitting position. After position change, 15 min were allowed for hemodynamic stabilization. All measurements were obtained by one investigator.

After the right common carotid artery was localized with an ultrasound scanner in B mode (7.5-MHz linear array; model HDI-9, Advanced Technology Laboratories, Bothell, WA), multiple M-mode recordings were obtained. The ultrasound probe was positioned such that the M-line intersected the common carotid artery 2–3 cm proximal to the carotid bifurcation. In each position, at least three (range 3–5) repeated measurements were obtained per subject. Recording length was 5 s, and thus each measurement consisted of four to seven consecutive heartbeats. Single-lead ECG (II) was acquired simultaneously for triggering purposes. During each recording, the subjects performed an end-expiratory breath-hold maneuver to minimize hemodynamic fluctuation due to respiration. In each position, arterial blood pressure was measured with a brachial cuff positioned at heart level (model 705CP, Omron, Matsusaka, Japan). Six repeated blood pressure measurements were taken per position: three before M-mode measurement and one after each of the first three M-mode recordings.

Signal processing.

The M-mode and R-wave trigger sample rate was 1 kHz (pulse repetition frequency). The ultrasound radio-frequency (RF) data were acquired at a sampling rate of 20 MHz and stored on hard disk for offline analysis. The RF data were preprocessed in the depth direction by application of a fourth-order band-pass filter matching the frequency range of the ultrasound probe (5- to 6-MHz effective center frequency, 4-MHz bandwidth). Arterial wall velocity was extracted by means of in-house echo-tracking software (5). Markers placed on the first acquired RF signals of the anterior and posterior artery wall facilitated wall tracking by the complex cross-correlation method (4). A 10-ms temporal correlation window with an overlap of 50% resulted in an effective sample rate of 200 Hz for the wall velocity. The distension waveform, i.e., the change in diameter over time, was obtained by integrating wall velocity over time (yielding wall displacement) and calculating the difference between the displacement signals of both walls (5). The distension waveforms were filtered using a three-point (15 ms) rectangular moving average filter without phase delay. The filtered distension signal was then interpolated by a cubic spline method (Matlab, Mathworks, Natick, MA) to reestablish a sample interval of 1 ms. The second derivative of the distension waveform, i.e., diameter acceleration, was calculated by passing the interpolated signal through a cascade of two first-order recursive high-pass filters with a cutoff frequency of ∼40 Hz. The second filter in the cascade was applied in reverse order to cancel out the phase shift of the first filter, yielding zero phase delay (34). For suppression of residual respiratory interference and spurious intrarecording baseline excursions, the acceleration signal was postprocessed by an additional second-order zero-phase-delay recursive high-pass filter with a cutoff frequency of 20 Hz.

Timing analysis.

With the use of the acceleration waveform, a proprietary algorithm implemented in Matlab determined the intrabeat positions (referenced to the R wave) of the following time points (Fig. 1) (34): start of left ventricular isovolumic contraction [SIC, local acceleration maximum preceding aortic valve opening (AVO)] (7, 34), AVO (maximum of the 2nd derivative within 150 ms after the R-wave peak), and aortic valve closure (AVC, maximum acceleration following peak distension). The intrabeat position of the lower body PR wave was determined by identification of peak distension in early diastole (after AVC) (19).

Fig. 1.

Characteristics of the distension waveform and the derived acceleration signal. R, ECG R wave peak; SIC, start of isovolumic contraction; AVO, aortic valve opening; AVC, aortic valve closure; PR, lower body peripheral reflection wave; dD/dtmax, maximum distension velocity; au, arbitrary units.


From the primary intrabeat positions, the ICP (AVO − SIC) and the EP (AVC − AVO) were derived (Fig. 1).

End-diastolic carotid artery diameter directly followed from the distension waveform in combination with the initial identification of the anterior and posterior wall positions. Carotid artery distension, i.e., peak-to-peak change in diameter, was defined as the difference between the maximum and the preceding minimum distension. Maximum distension velocity (dD/dtmax) was determined by calculation of the incremental slope at the peak distension velocity position (1st zero crossing of the 2nd derivative of the distension following AVO).

We calculated ICP/EP (in %) as a measure of left ventricular function (1, 32). Similar to the approach of Rhodes et al. (26), the ratio of diastolic blood pressure (DBP) to ICP (DBP/ICP) was calculated. To explore the applicability of the diameter (D) acceleration as an alternative measure of systolic left ventricular function, we calculated (dD/dt)/ICP (in mm/s2).

Arterial pulse pressure (PP) was calculated from systolic blood pressure (SBP) and DBP as follows: PP = SBP − DBP. Mean arterial pressure (MAP) was estimated as follows: MAP = DBP + PP/3.

Statistical analysis.

For each primary hemodynamic variable, group mean and standard deviation, average difference between supine and sitting positions, and intrasubject standard deviation (precision) were calculated (34). Statistical difference between supine and sitting positions was tested by Student's two-tailed paired t-test. Sex differences in response to posture were analyzed by Student's two-tailed two-sample t-test with the assumption of equal variance (Excel, Microsoft, Redmond, WA). Associations between pairs of variables were judged by scatterplots (Excel, Microsoft, Redmond, WA), and linearity was evaluated by Pearson's correlation coefficient. P < 0.05 was considered statistically significant. Values are means ± SD, unless stated otherwise.


An example of a distension waveform and the acceleration waveform derived from it is given in Fig. 1. Table 1 shows the values of all variables as obtained in supine (baseline) and sitting positions. Extraction of the various characteristic time points and systolic time intervals showed good reproducibility. For each time point, the intrasubject standard deviation (precision) was smaller than the observed difference and intersubject standard deviation with postural intervention (Table 1).

View this table:
Table 1.

Effect of posture on central hemodynamics


In the supine position, average heart rate was 60 beats/min, MAP was 81 ± 9 mmHg, and PP was 46 ± 7 mmHg (Table 1). On average, intrasubject R-R variability was 5%. Relative distension was 11.8%, and dD/dtmax was 10.8 ± 2.0 mm/s. At the level of the carotid arteries, SIC and AVO were 68 ± 8 ms and 111 ± 9 ms after the R wave, respectively, yielding an ICP of 43 ± 8 ms. AVC was detected after 412 ± 21 ms, yielding a mean EP of 302 ± 16 ms. Arrival of the lower body PR wave was timed at 522 ± 31 ms. Baseline ICP/EP was 14 ± 3%. Average isovolumic pressure development (DBP/ICP) was 1,572 ± 369 mmHg/s, and average systolic diameter acceleration [(dD/dt)/ICP] was 263 ± 67 mm/s2.

Postural intervention.

Changing posture from the supine to the sitting position induced a decrease in R-R interval of 49 ± 82 ms (P = 0.013), whereas mean and diastolic central arterial pressures increased (P < 0.001, Table 1). Systolic pressure remained unaffected (P = 0.09), and, consequently, PP significantly decreased in the sitting position (P < 0.001). A decrease in pulsatility was also seen at the carotid artery level: distension decreased from 0.74 ± 0.15 to 0.69 ± 0.14 mm (P = 0.012). However, posture-induced changes in carotid artery distension and PP were not correlated (r2 = 0.003). Carotid artery diameter did not change (P = 0.449); thus relative distension decreased from 11.8% to 11.1%. In the sitting position, dD/dtmax was reduced to 9.4 ± 1.7 mm/s (P < 0.001).

Figure 2 illustrates the time-related changes in the carotid artery pulse wave with postural intervention (beats obtained from a single subject). The systolic upstroke of the pulse wave (SIC and AVO) reached the carotid artery later in the sitting than in the supine position (P < 0.001; Table 1), reflecting decreased pulse wave velocity in the cranial direction. Changes in SIC and AVO were significantly correlated (r2 = 0.42, P < 0.01; Table 2). ICP significantly increased by 7 ± 7 ms (+16%, P < 0.001). Changes in ICP were not correlated to changes in AVO, DBP, R-R interval, or EP (Table 2). In the upright position, timing of AVC was earlier and, consequently, EP decreased to 267 ± 19 ms (302 ± 16 ms in the supine position, P < 0.001; Table 1). Advanced arrival of the lower body PR wave and earlier AVC were significantly correlated (r2 = 0.40, P < 0.01). Shortening of EP was significantly correlated to arrival of the lower body PR wave (Fig. 3, Table 2). Changes in EP were correlated to changes in R-R interval (P < 0.05) but not to changes in DBP or ICP (Table 2).

Fig. 2.

Changes in distension waveform timing with postural intervention. ICP, isovolumic contraction period; EP, ejection period. Time 0 corresponds to the peak of the R wave. Bars at the bottom pertain to systolic time intervals.

Fig. 3.

Correlation between change in arrival of the lower body PR wave and change in EP.

View this table:
Table 2.

Associations between posture-induced changes in hemodynamics

In response to the postural intervention, ICP/EP increased by 5 ± 3% (P < 0.001) and (dD/dt)/ICP and DBP/ICP decreased (P < 0.05). Changes in ICP/EP, DBP/ICP, and (dD/dt)/ICP were not correlated to changes in R-R interval or changes in blood pressure (Table 2).

Except for EP and ICP/EP, no significant differences related to postural change were found between men (n = 10) and women (n = 11). The average increase in ICP/EP was 6.2 ± 2.5% in men, but only 3.2 ± 2.5% in women (P = 0.011). Similarly, EP decreased more in men than in women (42 ± 12 vs. 29 ± 15 ms, P = 0.039). The advance in arrival of the PR wave was similar in men and women (P = 0.12).


In the present study, we were able to determine ICP and EP from carotid artery distension waveforms with good precision. Our results demonstrate the feasibility of quantifying systolic left ventricular function and ventriculoarterial interaction by means of carotid artery ultrasonography and pulse wave analysis.

Ventricular function indexes.

The ICP and EP as measured at baseline were similar to those measured by cardiac Doppler techniques and phonographic methods (Table 3) (20). ICP/EP corresponds to previously reported values in normal subjects as well (32, 33). Compared with the dP/dt estimates reported by Rhodes et al. (26), our DBP/ICP index value is closest to the dP/dtmax value derived from invasive left ventricular pressure. Interestingly, when left ventricular pressure development is offset by 10 mmHg, accounting for end-diastolic pressure, we arrive at values similar to those reported by Rhodes et al. (Table 3). These investigators studied a heterogeneous group of patients that were younger or older than our healthy individuals, which is reflected by the respective standard deviations. Average systolic diameter acceleration, reflected by (dD/dt)/ICP, is smaller by a factor of 10–20 than left ventricular myocardial isovolumic acceleration as obtained by tissue Doppler imaging. Both indexes are related to systolic ventricular function, in that they are used to characterize local tissue velocity as a function of time. However, the measured (dD/dt)/ICP may be influenced by central arterial stiffness and blood pressure, whereas isovolumic acceleration has been shown to be load independent (36).

View this table:
Table 3.

Noninvasive systolic time intervals and related indexes in the literature

Changes in ventricular function indexes with orthostatic challenge.

By decreasing carotid artery transmural pressure, orthostatic stress elicits a baroreflex response that temporarily increases heart rate and arterial blood pressure (6, 11, 16, 28, 30). These basic effects were reproduced in our study. Additionally, orthostatic challenge is associated with reductions in venous return and, hence, cardiac preload and left ventricular output (6, 10, 11, 28).

In our subjects, ICP, EP, and all derived function indexes consistently suggested a significant reduction in pump function. Comparable observations were made by Ovadia et al. (24), who measured accelerometer-derived systolic time intervals in young patients with upright syncope. In their study, PEP increased, EP decreased, and PEP/EP increased in the upright position. Using head-up tilt, Vijayalakshmi et al. (35) reported a similar response of these systolic time intervals in healthy subjects. Although ICP/EP has been shown to be a valid measure of systolic left ventricular function in patients (1, 32), the implied reduction in output due to orthostatic stress in healthy subjects is likely to result from changes in loading conditions. Inasmuch as orthostatic challenge affects preload by passive (blood pooling) effects (11, 28) and afterload by active (reflex) mechanisms (6, 14), it is difficult to discriminate the various factors that modulate ICP, EP, and the derived function indexes. Tolerance of orthostatic stress in normal subjects is primarily related to intact autonomic chronotropic response and reflex constriction of peripheral resistance vessels (6, 11, 14, 31), and not to inotropic effects (6, 8, 11, 16). ICP is correlated to dP/dtmax (38) and is, as such, preload and afterload dependent (36). The preload reduction and afterload increase, as induced by orthostatic challenge, increase the duration of isovolumic contraction. Similarly, the reduction in EP may be indicative of reduced ventricular filling and, hence, diminished stroke volume (6, 10, 11). Moreover, an increase in peripheral resistance, as expected with orthostatic challenge (6, 16), would allow a decrease in left ventricular output while maintaining or even enhancing DBP (Table 1). ICP and EP are rate dependent (Table 2) (32), but in our subjects heart rate increase was only modest. The changes in ICP/EP were independent of R-R interval and DBP (Table 2). The rate independence of ICP/EP has previously been demonstrated (32).

Theoretically, a DBP increase (Table 1) with constant ICP would suggest a contractility increase reflected by DBP/ICP. Apparently, the rise in ICP with postural intervention outweighs the DBP increase in our subjects, because DBP/ICP decreased significantly, independently of R-R interval and DBP. Compared with the other indexes, the change in DBP/ICP was borderline significant (P < 0.023). The accuracy of DBP/ICP may be limited because of the errors made in approximating left ventricular isovolumic pressure development (26). Furthermore, the precision of the DBP/ICP index will be dependent on reproducible timing extraction (34) and precise manometer leveling. In the present study, depending on chest size, the upper arm cuff might have been below heart level in the supine position, leading to overestimation of mean blood pressure and DBP and subsequent underestimation of the change in blood pressures upon sitting. Because the observed changes are moderate and similar to published values in normal subjects (6, 11, 28), it appears unlikely that the outcome of our study is influenced by manometer level errors.

The (dD/dt)/ICP index changed significantly and was independent of changes in R-R interval and DBP as well. The variability in response among subjects, however, was considerable. Moreover, the reproducibility of dD/dtmax and (dD/dt)/ICP was acceptable but not convincing (Table 1). The validity of diameter acceleration in characterizing ventricular contraction or ejection remains to be determined, whereas the index is potentially confounded by changes in blood pressure and arterial stiffness.

The significant difference in the ICP/EP and EP responses between men and women might reflect greater muscle exercise capacity and related vascular capacity in our male subjects. If this assumption is valid, then reflex peripheral vasoconstriction may have been more potent in our male subjects, because the change in DBP was not significantly different between men and women.

The borderline significance of the decrease in DBP/ICP might suggest that intrinsic contractility barely changed in our study, which is consistent with related studies (6, 8, 10, 11, 16). In our subjects, increased ICP and ICP/EP and decreased DBP/ICP and (dD/dt)/ICP should be considered indicative of changes in ventricular loading due to passive and active peripheral responses to orthostasis.

Ventriculoarterial interaction.

Orthostasis induces increased hydrostatic pressure in the arteries of the lower body. Because the stiffness of conductance arteries increases with increasing distending pressure (2, 23, 25), orthostasis is likely to affect pulse wave propagation velocity, such that waves reflected in the lower body periphery tend to return earlier at heart level. Our results confirm the earlier arrival of the PR wave in the sitting position. Moreover, the PR wave was significantly correlated to AVC and EP shortening (Fig. 3). This intrabeat ventriculoarterial interaction appears to outweigh heart rate- and blood pressure-related effects on ejection time (Table 2). In our subjects, we identified the PR wave by detecting the maximum diameter in diastole following the dicrotic notch (Fig. 1). With age, the PR waves may gradually shift into the systolic phase of the cardiac cycle (21, 25). Therefore, in older subjects or in cardiovascular patients, this detection method may not be feasible or valid.

Accuracy and precision of timing extraction.

The prerequisite for successful application of timing analysis to the carotid distension waveform is maintenance of the pulse wave contour during its propagation from the aortic root to the site of measurement in the common carotid artery. Studinger et al. (29) observed similar distensibility coefficients in proximal aortic and carotid arterial segments in normal subjects, which implies continuous pulse wave propagation velocity. Impedance mismatch between the aorta and the carotid artery due to the difference in diameter (20) is unlikely to affect the timing characteristics of the transmitted wave from the aorta to the common carotid artery: reflected waves travel in the opposite direction, away from the site of measurement. Distortion of the pulse wave timing patterns due to the viscoelastic properties of normal large arteries appears negligible when measured by the appropriate technique (2, 13). The similarity of our ICP, EP, and ICP/EP values to those obtained in related studies in healthy subjects suggests that time-related features of the pulse wave are preserved in transit (12, 26, 32, 34, 35).

Intrasubject variability due to low-frequency interference (unstable probe position and unstable breath hold) was effectively suppressed by the acceleration postprocessing filter. Therefore, the precision of ICP is slightly better in the present study (4 ms in the supine position) than in our previous study (5 ms) (34). More importantly, the reproducibility is on the same order of magnitude for both postures (Table 1). For all timing parameters and indexes obtained by pulse wave analysis, precision was such that individual differences could be discriminated. Using tissue Doppler echocardiography, Tekten et al. (33) reported a reproducibility of <3 ms for ICP and ∼20 ms for EP. Because their analysis was performed manually, it remains difficult to discriminate measurement- and analysis-based variability. Phonographic determination of systolic time intervals is similarly dependent on manual or dedicated signal processing (12, 26, 38).

Study limitations.

The postural intervention used in our healthy young subjects is probably limited in terms of hemodynamic challenge. In hypertensive and diabetic patients or in the elderly, normal orthostatic challenge may have more pronounced and, likely, more differentiating effects (9, 14, 15, 18, 21). Evaluation of the ventriculoarterial interaction as a function of posture in these populations could enable further discrimination of the various (patho-)physiological mechanisms in regard to baroreflex function and orthostatic intolerance (6, 16, 18, 24, 28).

In the present study, no gold-standard measurement of systolic time intervals was obtained. Because the noninvasive alternatives (Doppler echocardiography and phonocardiography) are subject to similar reproducibility errors (33), they were considered to be unsuitable to serve as the gold standard. Furthermore, invasive pressure measurements are unethical in the population studied. The validity of extracting ICP from the arterial pulse waveform was previously demonstrated on a data set containing simultaneously recorded intraventricular and aortic pressures (34).

The present study was not intended to establish that any of the indexes conclusively quantify true ventricular or arterial function but, rather, to determine the feasibility (precision) of the measurements. The use of carotid artery ultrasound to quantify ventricular systolic time intervals and its validity in characterizing ventriculoarterial interaction in patients requires further investigation.

In conclusion, ultrasound-based carotid artery pulse wave analysis enables noninvasive quantification of the hemodynamic ventriculoarterial responses to postural changes.


This study was supported by Ministry of Economic Affairs (The Netherlands) Grant TSGE3162.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract