Journal of the American College of Cardiology Vol. 57, No. 8, by the American College of Cardiology Foundation ISSN /$36.

Transcription

1 Journal of the American College of Cardiology Vol. 7, No. 8, by the American College of Cardiology Foundation ISSN 73-97/$36. Published by Elsevier Inc. doi:.16/j.jacc..9.4 Vascular Disease Development and Validation of a Novel Method to Derive Central Aortic Systolic Pressure From the Radial Pressure Waveform Using an N-Point Moving Average Method Bryan Williams, MD,* Peter S. Lacy, PHD,* Peter Yan, MBBS, Chua-Ngak Hwee, Chen Liang, PHD, Choon-Meng Ting, MBBS Leicester, United Kingdom; and Singapore Objectives Background Methods Results Conclusions The purpose of this study was to develop and validate the novel application of a simple n-point moving average (NPMA) a mathematical low pass filter to noninvasively derive central aortic systolic pressure (CASP) from the radial artery pressure waveform (RAPWF) in humans. CASP may be an important independent determinant of clinical outcomes. Development of simple, well-validated methods to noninvasively derive CASP is necessary to facilitate the routine clinical measurement of CASP. Three studies in different population cohorts were used to develop and validate the NPMA method to derive CASP in humans: 1) a development study (n 217), describing the optimal application of the NPMA to derive CASP; 2) a validation study comparing NPMA-CASP with CASP derived using a generalized transfer function (GTF-CASP [SphygmoCor system, AtCor, Sydney, Australia]) using,349 RAPWFs from the CAFE (Conduit Artery Function Evaluation) study; and 3) an invasive validation study (n ) comparing NPMA-CASP with direct aortic root pressure measurements during cardiac catheterization. In the development study, when using the NPMA, a denominator of n/4 (where n tonometer sampling frequency) most accurately defined CASP relative to GTF-CASP. Validation of NPMA-CASP using RAPWFs from the CAFE study revealed excellent correlation and agreement (r 2.993, mean difference.3 1. mm Hg). The agreement remained robust after stratification by sex, age, treatment, and diabetes status. There was also excellent correlation and agreement (r 2.98, p.1) between directly measured aortic root systolic pressures (Millar s catheter) at cardiac catheterization versus NPMA-CASP, derived simultaneously from noninvasive RAPWFs. We show that an NPMA with a denominator of one-quarter of the tonometer sampling frequency accurately defines CASP when applied to noninvasively acquired RAPWFs in man. These novel findings have important implications for the simplification of noninvasive CASP measurement and its wider application in clinical trials and clinical practice. (J Am Coll Cardiol 11;7:91 61) 11 by the American College of Cardiology Foundation Blood pressure (BP) has been conveniently measured over the brachial artery using a sphygmomanometer for more than a century. However, brachial BP may not always accurately represent the corresponding pressures in the aorta (1). It is now clear that many factors including age, heart rate, body height, sex, and drug therapies can all influence the relationship between brachial and central aortic pressure (2,3). Furthermore, evidence is beginning to emerge that central aortic systolic and/or pulse pressures may more accurately predict cardiovascular structural damage and cardiovascular outcomes when compared to brachial pressures (1,4 9). From the *Department of Cardiovascular Sciences, University of Leicester, and the Leicester NIHR Biomedical Research Unit in Cardiovascular Diseases, Leicester, United Kingdom; Gleneagles Medical Centre, Singapore; and Healthstats International, Singapore. Dr. Williams worked in a scientific collaboration with Healthstats, Singapore, to develop the concepts contained in the report, and is a nonsalaried advisor to Healthstats, Singapore; and an NIHR Senior Investigator. Drs. Williams and Lacy and Mr. Ting are faculty members of the Leicester National Institute for Health Research (NIHR) Biomedical Research Unit in Cardiovascular Diseases, which supported this study. Dr. Lacy and Mr. Yan have no conflicts of interest in relation to this work. Chua-Ngak Hwee is a board member of Healthstats. Dr. Liang is an employee of Healthstats. Mr. Ting is founder and Chairman of Healthstats. Manuscript received September 7, ; revised manuscript received September 13,, accepted September 17,.

2 92 Williams et al. JACC Vol. 7, No. 8, 11 Moving Average and Central Systolic Pressure February 11:91 61 Abbreviations and Acronyms BP blood pressure CASP central aortic systolic pressure GTF generalized transfer function NPMA n-point moving average PP pulse pressure RAPWF radial artery pressure waveform SBP systolic blood pressure With regard to the variation between central aortic and brachial pressures, it is assumed that the mean arterial and diastolic pressure remains largely unchanged from aortic root to brachial artery, and that it is variation in amplification of the pulsatile pressure component, namely, systolic pressure, that accounts for the central-to-brachial pressure differences (). Thus, focus has been on the accurate derivation of central aortic systolic pressure (CASP). Mindful of the potential for central aortic pressures to provide incremental value in predicting cardiovascular morbidity and mortality, the challenge is to develop simple and convenient ways to accurately estimate central aortic pressures in routine clinical practice. The gold standard is direct measurement of aortic root pressures using a pressure transducer introduced into the aortic root at the time of cardiac catheterization, but this is clearly invasive and unsuitable for routine clinical practice. A popular alternative approach has been analysis of the radial artery waveform obtained by noninvasive tonometry. In this method, the radial waveform is usually calibrated to brachial blood pressure, measured using standard sphygmomanometry, thereby generating a calibrated radial artery pressure waveform (RAPWF). Mathematical generalized transfer functions (GTFs) in the frequency or time domain have then been used to derive central aortic pressures and related aortic hemodynamic indices from the RAPWF (11 13). This method has, however, been criticized because of concerns that it may not be appropriate to apply a GTF generated in 1 cohort of patients to all patients with different disease states, at different ages, and receiving different treatments, and so forth (14,1). Nevertheless, applying a GTF to the RAPWF remains the most commonly used method for the noninvasive assessment of central aortic pressure indices (16). More recently, an alternative approach to estimating CASP from the RAPWF has been proposed. This requires the accurate identification of an inflection point on the RAPWF that is said to correspond to the superimposition of the peak of the reflected wave onto the outgoing pressure wave (17 19). Numerous recent studies have suggested that this inflection point, so-called SBP2, corresponds to the peak CASP and is a reasonably accurate way of noninvasively assessing CASP, without the need to apply a GTF (,21). In this paper, we report for the first time a simple approach for the accurate estimation of CASP in humans, using an n-point moving average (NPMA). A moving average is a mathematical model, used extensively in many applications beyond medicine, that acts as a low pass filter to smooth signal data, to better determine underlying trends. We hypothesized that this methodology could be applied to the RAPWF to generate a representative smoothed waveform, the peak of which would correspond to CASP. This reasoning is based on the fact that amplification of radial systolic pressure from the aortic root is due to changes in the relative timings of the incident and reflected pressure waves with increasing distance from the heart, together with the nonuniform elasticity and viscous damping within the arterial system (22,23). These effects result in a spatial amplification of the arterial pressure wave without additional energy input. This has been described as a distortion rather than true amplification of the pressure signal, which translates into the altered morphology of the radial waveform (24). Accordingly, we hypothesized that application of a NPMA might eliminate this effect, revealing the amplitude of the original signal and, thus, CASP. Such an approach, if valid, would have the advantage of utilizing patient-specific data contained within the entire RAPWF, removing the need for a more complex GTF, or the need to identify any specific points on the waveform, namely, an inflection point. It would provide a novel and simple system that could potentially be applied routinely to estimate CASP, wherever a radial waveform is captured by tonometry. Methods Study design. This work involved 3 related studies, each using different population cohorts: 1) an initial development study to define the optimal denominator for the NPMA (i.e., extent of filtering) required to estimate CASP relative to the reference noninvasive GTF method (see the following text); 2) a noninvasive validation study versus the GTF-CASP; and 3) an invasive validation study at cardiac catheterization. Reference GTF method for noninvasive measurement of CASP. To develop and validate the NPMA method for deriving CASP from the RAPWF, a reference method was required. We used the GTF embedded within the Sphygmo- Cor system (AtCor, Sydney, Australia), which is, to date, the most widely used method for noninvasively deriving central pressures and hemodynamics. The SphygmoCor GTF is reported to be based on a Fourier transformation, namely, a frequency domain method (11,2). Estimation of CASP from individual RAPWFs using a moving-average method. A moving average generates an array of incrementally averaged data points, based on a specified and constant denominator. We hypothesized that this denominator would be related to the sampling frequency of the tonometer used to acquire the RAPWF. The optimal NPMA denominator to derive CASP from the RAPWF could have been one of a range of fractions of the sampling frequency of the tonometer (n), namely, n/2, n/3, n/4, and so on. Thus, if the optimal fraction for the NPMA was ultimately found to be n/4, then using a sampling frequency of 128 Hz, incremental blocks of 128/4, in other words, 32 data points, would be generated. The resulting array of

3 JACC Vol. 7, No. 8, 11 February 11:91 61 Williams et al. Moving Average and Central Systolic Pressure 93 A Pressure (mmhg) B Pressure (mmhg) RAPWF data points 1 Moving average curve (n/4) Average of RAPWF data points Average of RAPWF data points Average of RAPWF data points Average of RAPWF data points Average of RAPWF data points Average of RAPWF data points Average of RAPWF data points Average of RAPWF data points SBP2 CASP Time (Seconds) GTF CASP Moving Average CASP etc... Radial Waveform Aortic Waveform Moving Average (Radial) Time (seconds) Figure 1 Methods for Estimating CASP From RAPWFs (A) Illustration of the application of an n-point moving average (NPMA) with a denominator of n/4 (n sampling frequency) to a radial artery pressure waveform (RAPWF) acquired using a tonometer with a sampling frequency of 128 Hz. Successive averages are calculated for each incremental block of 32 RAPWF sampling data points (i.e., 128/4 32) and plotted as the moving average curve. The blocks move in increments of sampling points, namely, 1-32, 2-33, 3-34, and so forth. Central aortic systolic pressure (CASP) equates to the maximum value of the moving average array. (B) Derivation of CASP from arterial pressure waveforms as 1) the inflection point on the down-stroke of systole for the radial pressure waveform (SBP2-CASP) (black line); 2) derived from the central aortic waveform (red line) after generalized transfer function (GTF) transformation of the radial pressure waveform (GTF-CASP); and 3) derived from applying a moving average model (blue line) to the radial pressure waveform (moving average-casp). averages yields a maximum value that we hypothesized would equate to CASP (Fig. 1). Study to develop and optimize the NPMA method to noninvasively derive CASP in humans. Using a study cohort comprising 217 volunteers, RAPWFs were sampled using the SphygmoCor tonometer (sampling ƒ 128 Hz), calibrated to seated brachial BP, measured using the MC3 oscillometric device (AAMI standard-sp, 1992, Food and Drug Administration approved, Healthstats, Singapore). To define the optimal denominator for the NPMA method to predict CASP, various denominators, namely, proportions of the tonometer sampling frequency ranging from n/3 through to n/6 were tested. The resulting peak value from the moving average arrays was then compared to the GTF-CASP generated by the SphygmoCor device to determine which denominator generated a curve, the peak of which best equated to GTF-CASP (see preceding text and Fig. 1 for details). The noninvasive NPMA validation study. The optimal NPMA method for deriving CASP as determined in the development study was then tested versus the alternative, 2 currently used approaches for the noninvasive derivation of CASP from the RAPWF, namely, GTF-CASP and SBP2- CASP. The population used for this study was different

4 94 Williams et al. JACC Vol. 7, No. 8, 11 Moving Average and Central Systolic Pressure February 11:91 61 from that used for the development study. The RAPWFs used for these analyses had been collected for the CAFE (Conduit Artery Function Evaluation) study, which has been previously described in detail (1). We used the RAPWFs collected from patients who attended the host CAFE study center (Leicester, United Kingdom [n 383]) and had repeated samplings of RAPWFs, at scheduled study follow-up visits, over the course of years, yielding,349 individual RAPWFs for these analyses. The RAPWFs had been sampled using radial applanation tonometry (Sphygmo- Cor system) calibrated to brachial blood pressure (Omron 7CP), as previously described (1). Ensemble-averaged waveforms were manually extracted as text files to facilitate comparisons of the signal processing models to estimate CASP. The reference method for these noninvasive validation studies was the GTF embedded within the Sphygmo- Cor system (GTF-CASP) (24), which was compared to CASP generated by our NPMA method and the SBP2 method. The SBP2 method for deriving CASP (SBP2- CASP) relies on the identification of an inflection point on the systolic down slope of the untransformed ensembleaveraged RAPWFs. The SBP2 was defined by the Sphygmo- Cor software using a derivative plot and was taken to represent SBP2-CASP, as previously described (,21). The invasive validation study. This study was conducted at the Gleneagles Medical Center in Singapore. Twenty adult patients undergoing routine diagnostic cardiac catheterization were invited to participate. Before cardiac catheterization, a high-fidelity micromanometer-tipped catheter (Millar SPC-44D, Millar Instruments, Houston, Texas) was set to zero and introduced through the femoral artery, and with fluoroscopic guidance, the tip of the catheter was positioned in the proximal aortic root, 1 cm above the aortic valve, until stable aortic pressure waveforms were obtained. At the same time as the Millar catheter was introduced, a tonometer attached to a wrist strap (A-pulse, Healthstats, Singapore) was applied to the wrist to allow continuous sampling of RAPWFs. The A-pulse tonometer had a sampling frequency of 6 Hz; thus, the denominator for the moving average was 1 (i.e., 6/4). With both devices in place, the brachial BP was measured over the same arm as the tonometer using an oscillometric device (MC3, Healthstats). These measurements were used to calibrate the noninvasive RAPWFs. An average of 3 brachial BP readings, not 3 s apart, were used for this calibration. The output from the Millar catheter, by a multichannel monitor, was synchronized in time with the output from the A-pulse tonometer, allowing simultaneous, real-time comparison of the moving average-casp (NPMA-CASP) with direct aortic root measurement of CASP. Both the invasive and noninvasive devices averaged waveform data in synchronous, time stamped, -s blocks; and the first stable blocks obtained after calibration was completed were used for data analysis, over a sampling period of up to 3 min. This allowed comparison of the simultaneous beat-to-beat averaged generation of CASP from the RAPWF through the NPMA versus the actual invasive aortic root pressures. All studies were approved by local research ethics review boards, and all participants provided written informed consent for participation in these studies. Statistical analysis. Comparisons of the reference GTF method with the NPMA or SBP2 for the development and noninvasive validation studies, namely, when data from the same waveforms were processed using different algorithms, used paired Student t tests. The impact of the CAFE study BP-lowering treatments, comparing the reference GTF- CASP with the moving average and SBP2 methods, was analyzed using an unpaired Student t test because 2 different treatment populations were being compared. For the invasive validation study, where waveforms were collected using different sampling devices for direct comparisons between noninvasive NPMA-CASP and invasive aortic root pressures, a paired Student t test was used. Additional comparisons used linear regression and Bland-Altman plots. Results Study to develop and optimize the NPMA method to derive CASP in humans. The demographics of the population used for the development study (n 217) to define the optimal denominator for the NPMA to derive CASP from the RAPWF is shown in Table 1. An NPMA with a denominator of n/4 consistently produced a moving average curve with a peak value (i.e., NPMA-CASP) that most closely approximated to the reference GTF-CASP (Fig. 2). The other denominators either underestimated (n/3) or overestimated (n/, n/6) relative to GTF-CASP (Fig. 2). Thus, NPMA-CASP was best defined relative to GTF- CASP using a denominator of one-quarter of the sampling frequency of the tonometer. Demographics Recruited Into Demographics for theparticipants NPMA Development for Participants Study Table 1 Recruited Into the NPMA Development Study Variable Mean SD or n (%) Range Age, yrs Sex Male 112 (1.6%) Female (48.4%) Blood pressure Normal 16 (79%) High* 2 (24%) SBP, mm Hg DBP, mm Hg Ethnic group Chinese 212 (97.7%) Malay 4 (1.8%) Caucasian 1 (.%) Diabetes mellitus 3 (1.4%) Hypertension 17 (7.8%) Data are mean SD or n (%). *High blood pressure defined as 14/9 mm Hg. DBP diastolic blood pressure; SBP systolic blood pressure.

5 JACC Vol. 7, No. 8, 11 February 11:91 61 Williams et al. Moving Average and Central Systolic Pressure 9 NPMA-CASP minus GTF-CASP (mmhg) NPMA-CASP minus GTF-CASP (mmhg) n/ Average CASP (mmhg) n/ Average CASP (mmhg) NPMA-CASP minus GTF-CASP (mmhg) NPMA-CASP minus GTF-CASP (mmhg) n/ Average CASP (mmhg) n/ Average CASP (mmhg) Figure 2 Determination of Optimal NPMA Denominator for Estimating CASP Bland-Altman comparisons of generalized transfer function (GTF) central aortic systolic pressure (CASP) with n-point moving average (NPMA)-CASP derived using differing denominators (n/3 through n/6) in the generation cohort where n tonometer sampling frequency. The dashed line shows the mean difference between measurements; the dotted lines show the extent of 2 SDs from the mean. Noninvasive validation study and comparison of NPMA- CASP with CASP derived using the GTF and SBP2 methods. The characteristics of the patients included in the noninvasive validation study are shown in Table 2. The mean age was 63 years; 8% of patients were male; and 9% were white Caucasian. All had treated hypertension, and 1% had diabetes mellitus at baseline. The CASP could be derived from all,349 RAPWFs analyzed in this validation study, using either the GTF or the NPMA method. By contrast, the systolic inflection point, upon which SBP2-CASP is dependent, could not be identified in 164 (3.1%) RAPWFs. Figure 1B shows an example of the 3 methods for deriving CASP from a typical RAPWF and the remarkable correspondence for CASP according to each of these methods. Pair-wise comparisons of the NPMA or SBP2 methods to derive CASP, versus GTF-CASP, are shown in Table 3. Using the GTF-CASP as the reference, the NPMA-CASP differed by an average of only.3 mm Hg. In contrast, SBP2-CASP differed from GTF-CASP by an average of 1.6 mm Hg. Regression analyses. Regression analyses comparing the reference GTF-CASP with NPMA-CASP or SBP2- CASP are shown in Figure 3. Good correlations were seen with both methods; however, the correlation was improved for NPMA-CASP relative to SBP2-CASP (NPMA-CASP versus GTF-CASP: r 2.99, p.1; SBP2-CASP versus GTF-CASP: r 2.9, p.1). Bland-Altman analyses. Comparison of GTF-CASP versus either NPMA-CASP or SBP2-CASP showed good agreement in the Bland-Altman plots. However, the agreement was stronger and the data less scattered for the GTF versus NPMA-CASP (average difference.3 mm Hg, 2 SD 2.1 mm Hg) (Fig. 4), when compared to the GTF versus SBP2-CASP (average difference 1.6 mm Hg, 2 SD 6.1 mm Hg). The data also suggested a bias with the SBP2 method, namely, increasing difference from the GTF-CASP with increasing pressure (Fig. 4). There was no such bias in the comparison between the GTF and NPMA-CASP. Impact of sex, age, and diabetes. There was no impact of sex, age ( years vs. years), or presence versus absence of diabetes on performance of the NPMA in defining CASP relative to the reference GTF-CASP in the noninvasive validation study (see the Online Appendix for supplementary results).

6 96 Williams et al. JACC Vol. 7, No. 8, 11 Moving Average and Central Systolic Pressure February 11:91 61 Baseline Patients RecruitedUsed Characteristics From Baseline for the the Leicester Characteristics Waveform CAFE Analysis, Study for Center Table 2 Patients Used for the Waveform Analysis, Recruited From the Leicester CAFE Study Center Treatment Allocation Amlodipine Atenolol Variable (n 194) (n 189) Demographics and clinical characteristics Age, yrs Sex, women 32 (16.%) 28 (14.8%) Ethnicity, white 177 (91.2%) 17 (92.6%) Current smoker (2.8%) 4 (21.2%) Systolic blood pressure, mm Hg Diastolic blood pressure, mm Hg Heart rate, beats/min BMI, kg/m Total cholesterol, mg/dl LDL cholesterol, mg/dl HDL cholesterol, mg/dl Glucose, mg/dl Creatinine, mg/dl Medical history Previous stroke/tia 14 (7.2%) (.3%) Diabetes mellitus 28 (14.4%) 31 (16.4%) LVH, according to ECG or 38 (19.6%) 36 (19.%) echocardiography* Peripheral vascular disease 6 (3.1%) (2.6%) Drug therapy Lipid-lowering therapy 8 (4.1%) 12 (6.3%) Aspirin use 36 (18.6%) 32 (16.9%) Data are mean SD or n (%). *Left ventricular hypertrophy (LVH) by echocardiography was assessed as 116 g/m 2 in men and 4 g/m 2 in women. The electrocardiography (ECG) LVH was defined using either Cornell voltage duration product ( 2,44) or Sokolow-Lyon criteria ( 38 mm). Assessed using a validated questionnaire or from evidence of a recent history of surgical intervention for peripheral vascular disease. BMI body mass index; HDL high-density lipoprotein; LDL low-density lipoprotein; TIA transient ischemic attack. Impact of BP-lowering treatments on CASP derived by different processing methods. The patients in the CAFE study from which the RAPWFs were derived had been randomized to treatment with 2 different blood pressurelowering strategies (amlodipine perindopril, as required vs. atenolol bendroflumethiazide, as required) (1). The brachial blood pressures throughout the study on both treatment regimens were similar. Stratification by treatment arm allowed examination of whether the differential effects of treatment (e.g., on heart rate and/or vasodilation) introduced any variation in the relationship between the reference GTF-derived CASP and the NPMA or SBP2 methods. Table 4 shows the average value for brachial BP and the corresponding CASP values derived using the 3 different signal processing methods. With each method, amlodipine-based therapy was associated with lower CASP values versus atenolol-based therapy. There was excellent agreement between GTF- CASP and the NPMA-CASP irrespective of treatment. There was also good agreement between GTF-CASP and SBP2-CASP but with wider scatter. Invasive validation study comparing direct aortic root systolic pressure with NPMA-CASP. The demographics of the patients recruited for the invasive validation study are shown in Table. The data showing a comparison of invasive aortic BP versus the corresponding noninvasive brachial BP at the time of calibration and derived NPMA- CASP are shown in Table 6. It is notable that noninvasive brachial systolic and diastolic pressures are overestimated by conventional oscillometric BP measurement versus direct aortic root pressures (aortic systolic BP mm Hg versus brachial systolic BP mm Hg, p.1) (Table 6); this was most noticeable for diastolic BP (aortic diastolic BP mm Hg vs. brachial diastolic BP mm Hg, p.1). In contrast, there was remarkable correspondence between invasive CASP measured at the aortic root versus CASP derived noninvasively from RAPWFs using the NPMA (invasive CASP mm Hg vs. NPMA-CASP mm Hg) (Table 6). The data in Figure shows the average CASP for each -s block (i.e., data points per patient) for both invasive and noninvasive measurements. Noninvasive NPMA-CASP from the RAPWFs did not differ from invasively measured aortic root pressure (invasive CASP mm Hg vs. NPMA-CASP mm Hg, p NS), and a high degree of correlation (r 2.98, p.1) (Fig. ) was seen between these 2 parameters. Bland-Altman analysis also showed good agreement between values with little scatter (mean difference: NPMA-CASP minus invasive CASP mm Hg) (Fig. ). Pair-Wise Table 3 Comparisons Pair-Wise Comparisons Between GTF- Between and NPMA-CASP GTF- and NPMA-CASP or GTF- and SBP2-CASP or GTF- and SBP2-CASP CASP Values for Paired Comparison Mean (mm Hg) n Difference (mm Hg) 9% CI for Difference (mm Hg) p Value GTF-CASP* ,349 NPMA-CASP , to.3.1 GTF-CASP ,18 SBP2-CASP , to Data represent average SD of all analyzable waveforms collected during CAFE follow-up. Comparisons were made using a paired Student s t test. *Central aortic systolic pressure derived by applying a generalized transfer function (GTF) to the radial pressure waveform. Central aortic systolic pressure derived by applying an n-point moving average (NPMA) to the radial pressure waveform. Central aortic systolic pressure defined using the radial artery pressure waveform inflection point on the down-stroke of systole (2SBP). CASP central aortic systolic pressure; CI confidence interval.

7 JACC Vol. 7, No. 8, 11 February 11:91 61 Williams et al. Moving Average and Central Systolic Pressure 97 CASP has been validated against the most commonly used method to derive CASP from the RAPWF, and against direct aortic root pressures. Methods for deriving central systolic pressure from radial artery pressure waveforms. There has been much debate about the use of a GTF to derive central aortic pressures and hemodynamics from the RAPWF (1) and, in particular, whether a generalized function is appropriate for people at different ages and with different disease states (14,28,29). Most concur, however, that the low frequency harmonics of the pressure wave do allow CASP to be accurately determined by the use of a GTF (3 33). Furthermore, the GTF-derived CASP used in the present study has been validated against direct invasive measurement of central Figure 3 Discussion Linear Regression of GTF-CASP Versus NPMA- or SBP2-CASP in the CAFE Study Cohort (A) Linear regression between central aortic systolic pressure (CASP) derived from individual radial artery pressure waveforms (RAPWF) using a generalized transfer function (GTF)-CASP and an n-point moving average (NPMA)-CASP. (B) Linear regression between CASP derived from individual RAPWF as the inflection point on the down-stroke of systole for the radial artery pressure waveforms (SBP2-CASP) and a moving average (NPMA- CASP). Data show individual values calculated from all waveforms collected in the CAFE (Conduit Artery Function Evaluation) study cohort. We have demonstrated that the application of a simple moving average method to a noninvasively acquired radial artery waveform, calibrated to brachial blood pressure, is sufficient to accurately derive central aortic systolic pressure in humans. The correspondence between NPMA-CASP and direct measurement of aortic root pressure was remarkable. In the invasive study, the agreement between direct invasive measurement and noninvasive NPMA-CASP was well within agreed standards for validation of blood pressure measuring devices (26,27). Applying these standards, the average of all sampling blocks for each patient revealed that the vast majority (96%) of data points were comfortably within mm Hg variance from the mean (Fig. ). Moreover, when NPMA-CASP was compared to GTF-CASP, all but 1 of, data points was within these standards. Thus, in the present study, the NPMA- A NPMA-CASP minus GTF-CASP (mmhg) B SBP2-CASP minus GTF-CASP (mmhg) Figure 4 1 Average CASP (mmhg) 1 Average CASP (mmhg) Bland-Altman Comparison of GTF-CASP Versus NPMA- and SBP2-CASP in the CAFE Study Cohort (A) Bland-Altman analysis comparing central aortic systolic pressure (CASP) derived from individual radial artery pressure waveforms (RAPWF) using a generalized transfer function (GTF)-CASP and an n-point moving average (NPMA)- CASP. (B) Bland-Altman analysis comparing CASP derived from individual radial artery pressure waveforms using a GTF-CASP and SBP2-CASP defined as the inflection point on the down-stroke of systole for the radial artery pressure waveforms. Data show individual values calculated from all waveforms collected in the CAFE (Conduit Artery Function Evaluation) study cohort. The dashed line shows the mean difference between measurements; the dotted lines show the extent of 2 standard deviations from the mean.

8 98 Williams et al. JACC Vol. 7, No. 8, 11 Moving Average and Central Systolic Pressure February 11:91 61 Comparison Central Systolic Comparison Between Pressure CAFE Between Estimated Study CAFE Treatment as GTF-P, StudyRegimens Treatment NPMA, and forregimens SBP2-CASP Brachial and for Brachial and Table 4 Central Systolic Pressure Estimated as GTF-P, NPMA, and SBP2-CASP Parameter Treatment Mean (mm Hg) SD n Brachial SBP Atenolol ,331 Difference (mm Hg) p Value Amlodipine ,37. NS GTF-CASP* Atenolol ,331 Amlodipine , NPMA-CASP Atenolol ,331 Amlodipine , SBP2-CASP Atenolol ,317 Amlodipine , Data represent average values for all measurements collected during CAFE follow-up where each measurement comprises the average of at least duplicate samplings. Comparisons were made using unpaired Student t tests. *Central aortic systolic pressure (CASP) derived by applying a generalized transfer function (GTF) to the radial pressure waveform. Central aortic systolic pressure derived by applying an n-point moving average (NPMA) to the radial pressure waveform. Central aortic systolic pressure defined using the radial artery pressure waveform inflection point on the down-stroke of systole (SBP2). NS not significant; SBP systolic blood pressure. aortic pressures (2,3). The main concern about the GTF has related to its application to higher frequency harmonics upon which other derived central hemodynamic parameters depend, for example, central augmentation index (12,33) these are much less well validated. The alternative to a GTF for the derivation of CASP from the RAPWF has been the use of SBP2, first described by Takazawa (17). This method is critically dependent on the accurate determination of the inflection point on the down-stroke of the radial systolic pressure wave. Identifying this point is dependent on the morphology of the radial artery waveform and is not always possible. In our study of, waveforms, identification of SBP2 was not possible for 3.1% of waveforms. Furthermore, the correspondence between SBP2 and GTF-derived CASP was less robust than that seen with our NPMA-CASP versus GTF-CASP. We also detected a bias with the SBP2 method, which was accentuated at higher pressures. The same bias has also been reported by others (,21). Nevertheless, despite these limitations, the SBP2 method does allow CASP to be estimated from the RAPWF with reasonable accuracy (19). A key finding of the present study is the demonstration that there is no need for the added complexity of a GTF or the limitations of SBP2 to derive CASP from the RAPWF. We show that the information required to accurately derive CASP is contained within the RAPWF and can be revealed by a Demographics the Invasive Demographics Blood for Patients Pressurefor Recruited Measurement Patients Into Recruited Study Into Table the Invasive Blood Pressure Measurement Study Variable Mean SD or n (%) Range Age, yrs Sex, male 1 (7%) Systolic blood pressure, mm Hg Diastolic blood pressure, mm Hg Heart rate, beats/min Hypertension 6 (3%) Coronary heart disease 17 (8%) Diabetes mellitus 3 (1%) simple low pass filter, in other words, the moving average. We tested a range of denominators as integers for the moving average and determined that one-quarter of the tonometer s sampling frequency is the most accurate denominator for deriving CASP from the RAPWF in humans. Importantly, this applies irrespective of the sampling frequency of the tonometer, as illustrated by the fact that we used 2 different tonometers for the noninvasive and invasive validation studies, sampling at 128 Hz and 6 Hz, respectively. Use of a low pass filter in assessing CASP. We show that a simple low pass filter, performing a smoothing function, can be used to accurately define the amplitude of the original pressure wave signal, and thus CASP from the RAPWF. It could be argued that a Fourier transformation in the time or frequency domains, as used in a GTF, is also functioning as a low pass filter. However, the noted controversy about the use of a GTF to derive other central hemodynamic parameters has obscured recognition of the simplicity of applying a low pass filter to accurately derive CASP as described here. To our knowledge, this is the first report describing the use of a simple moving average to noninvasively derive CASP from the RAPWF. This application of the moving average for deriving CASP is not dependent on any generalized adjustment or assumptions; it simply filters data that are actually present in the waveform, specific to the patient. Accordingly, we observed no obvious impact of sex or age on the correlations between GTF-CASP and the NPMA-CASP, and no evidence of any blood pressure-related bias. Why one-quarter sampling frequency of the tonometer should be the specific denominator required to reveal CASP from the RAPWF in humans requires further evaluation. Simplicity in the moving average method. A frustrating feature of existing technologies to derive central pressure and hemodynamics noninvasively is that they are expensive and often exist as a black box from which is it is difficult to define the specific algorithms used to derive the final data. The NPMA method reported here is simple and will allow CASP to be derived from any RAPWF imported into a spreadsheet, facilitating further evaluation of the method

9 JACC Vol. 7, No. 8, 11 February 11:91 61 Williams et al. Moving Average and Central Systolic Pressure 99 Comparison Noninvasive Comparison of Brachial Invasive Pressure Aortic of Invasive Root Measurements Pressures Aortic Root and With Pressures NPMA-Derived Oscillometric, With Oscillometric, Noninvasive CASP Table 6 Noninvasive Brachial Pressure Measurements and NPMA-Derived Noninvasive CASP SBP (mm Hg) DBP (mm Hg) Mean Difference vs. Invasive SBP (mm Hg) Mean Difference vs. Invasive DBP (mm Hg) Brachial * * Invasive NPMA Data represent mean SEM or mean standard error of difference per patient (n ). *p.1, paired Student t test. Abbreviations as in Tables 1 and 3. in a variety of patient groups from past and future studies. The main potential error in the method, as with all of the noninvasive methods, comes from the inherent error in the conventional measurement of brachial blood pressure that is required for calibration of the radial artery waveform. This error is generally less for brachial systolic pressure than for diastolic pressure (Table 6) (34). Thus, any error in CASP estimation from the RAPWF should be no greater than the error in conventional brachial BP measurement. The moving average method is designed to accurately derive CASP and does not generate an aortic waveform. Thus, it cannot be used to derive other indices such as aortic augmentation index, the validation and utility of which remain controversial (12,33,3). An alternative method for deriving CASP is the use of carotid artery waveforms calibrated to brachial mean and diastolic pressures. This method does not require a GTF or moving average as the carotid waveform is considered to be a central waveform. However, carotid tonometry is more difficult than radial tonometry and is much less suited for routine measurements of CASP when compared to the aforementioned radial waveform approach. Study limitations. There are a number of limitations to our study that we acknowledge. In the development study and noninvasive validation studies, the reference method for estimating CASP was the GTF embedded within the SphygmoCor device. This noninvasive method has itself, as discussed in the preceding text, been criticized with regard to the extent of its validation. It is, however, the most widely used method, and our study is strengthened by our also reporting an invasive validation of the NPMA method, showing excellent correlation with direct measurement of aortic root pressure. With regard to the comparison versus GTF-CASP, our noninvasive validation study population was mainly male but did include 8 RAPWFs from female patients, and no sex bias was noted when formally tested (Online Fig. 1A). Another limitation is that our noninvasive validation study population was predominantly white Caucasian, consistent with a general paucity of data on pulse wave analysis in other ethnic groups, which needs addressing. However, our population for the NPMA development and invasive validation studies were Chinese Asian, confirming applicability of the method to these populations as well. As with almost all studies of this kind, our populations for both validation studies were mostly older than years, with manifest cardiovascular disease or at high cardiovascular risk. We saw no evidence of a bias related to extremes of age in the performance of the NPMA method (Online Fig. 2A). It is noteworthy that the demographics and ethnic origins of the 3 cohorts used in these studies of NPMA-CASP were different from each other but still yielded remarkably consistent data supporting the robustness of A B Figure Comparison of Invasive Aortic Root Systolic Pressure With Noninvasively Derived CASP (A) Linear regression of central aortic systolic pressure (CASP) measured invasively (invasive central systolic pressure) and noninvasively estimated CASP. GTF-CASP versus NPMA- Bland-Altman plot comparing central systolic blood pressure (SBP) measured invasively (invasive central systolic pressure) and noninvasively estimated CASP. Invasive CASP was measured using a Millartipped catheter. Noninvasive CASP was estimated using moving average analysis of radial pressure waveforms acquired using a tonometer embedded within a wrist strap (A-pulse, Healthstats, Singapore), calibrated to oscillometric brachial blood pressure. Data points represent each individual, -s samplings for each patient ( blocks per patient, patients, n ).

10 96 Williams et al. JACC Vol. 7, No. 8, 11 Moving Average and Central Systolic Pressure February 11:91 61 the method. However, further studies would be desirable in younger people and children. Finally, regarding the potential impacts of drug therapy on the performance of the NPMA, our noninvasive validation study cohort, from the CAFE study, were treated with 2 different BP-lowering regimens (atenolol bendroflumethiazide or amlodipine perindopril) (1). We have previously reported that these treatments had markedly different effects on heart rate and CASP as determined using the SphygmoCor GTF (1,36). It is notable that the differential impact of these treatments on CASP were perfectly replicated by application of the NPMA method to the same RAPWFs (Table 3), indicating no inherent bias or degeneration of the performance of the NPMA-CASP due to treatment and/or heart rate effects. Conclusions We report a simple and accurate method for estimating CASP from noninvasively acquired RAPWFs in humans. This method has been compared to existing reference methods and validated against direct invasive measurements of aortic pressures, using different study populations. There is currently tremendous interest in the noninvasive measurement of CASP in clinical studies and clinical practice (31). This has been driven by emerging evidence that CASP may provide a better assessment of the impact of medications on aortic pressures and incremental value, over and above conventional brachial blood pressure measurement, in the assessment of cardiovascular risk (1,4 9). Wider application of CASP measurement in clinical practice has been hindered by a lack of data regarding its clinical relevance, the unnecessary complexity and cost of acquiring the data, and longstanding controversy regarding the use of a GTF. The simplicity of the moving average method, the transparency of the technique, its widespread acceptance as a universal tool for signal processing, and the detailed development and validation studies reported here, could lead to greater acceptance and use of noninvasive CASP measurements in clinical studies and ultimately, clinical practice. Acknowledgments The authors thank Tom Williams, BSc (Hons), University of Warwick United Kingdom, for help in manually extracting wave forms for the noninvasive validation study. Part of this study used radial artery waveforms collected in Leicester, United Kingdom, for the CAFE study, a substudy of the ASCOT study. Therefore, the authors thank the CAFE and ASCOT study investigators for their contributions. Reprint requests and correspondence: Dr. Bryan Williams, Department of Cardiovascular Sciences, University of Leicester, Clinical Sciences Wing, Glenfield Hospital, Leicester LE3 9QP, United Kingdom. bw17@le.ac.uk. REFERENCES 1. The CAFE Investigators for the ASCOT Investigators. Differential impact of blood pressure lowering drugs on central aortic pressure and clinical outcomes principal results of the Conduit Artery Function Evaluation study: the CAFE study. Circulation 6;113: McEniery CM, Yasmin, Hall IR, Qasem A, Wilkinson IB, Cockcroft JR, on behalf of the ACCT Investigators. Normal vascular aging: differential effects on wave reflections and aortic pulse wave velocity. The Anglo Cardiff Collaborative Trial (ACCT). J Am Coll Cardiol ;46: Cecelja M, Chowienczyk P. Dissociation of aortic pulse wave velocity with risk factors for cardiovascular disease other than hypertension: a systematic review. Hypertension 9;4: Williams B, Lacy PS. Central aortic pressure and clinical outcomes. J Hypertens 9;27: Roman MJ, Devereux RB, Kizer JR, et al. Central pressure more strongly relates to vascular disease and outcome than does brachial pressure: the Strong Heart Study. Hypertension 7;: Pini R, Cavallini MC, Palmieri V, et al. Central but not brachial blood pressure predicts cardiovascular events in an unselected geriatric population. The ICARe Dicomano study. J Am Coll Cardiol 8;1: Jankowski P, Kawecka-Jaszcz K, Czarnecka D, et al. Pulsatile but not steady component of blood pressure predicts cardiovascular events in coronary patients. Hypertension 8;1: Wang K-L, Cheng H-M, Chuang SY, et al. Central or peripheral systolic or pulse pressure: which one best relates to target organs and future mortality. J Hypertens 9;27: Roman MJ, Okin PM, Kizer JR, Lee ET, Howard BV, Devereux RB. Relations of central and brachial blood pressure to left ventricular hypertrophy and geometry: the Strong Heart Study. J Hypertens ;28: Nichols WW, O Rourke M. McDonald s Blood Flow in Arteries: Theoretical, Experimental and Clinical Principles. London, UK: Arnold,. 11. Karamanoglu M, O Rourke MF, Avolio AP, Kelly RP. An analysis of the relationship between central aortic and peripheral upper limb pressure waves in man. Eur Heart J 1993;14: Chen CH, Nevo E, Fetics B, et al. Estimation of central aortic pressure waveform by mathematical transformation of radial tonometry pressure. Validation of generalized transfer function. Circulation 1997;9: Fetics B, Nevo E, Chen CH, Kass DA. Parametric model derivation of transfer function for noninvasive estimation of aortic pressure by radial tonometry. IEEE Trans Biomed Eng 1999;46: Hope SA, Meredith IT, Cameron J. Generalizability of radial-aortic transfer function for the derivation of central aortic waveform parameters. J Hypertens 7;2: Hope SA, Meredith IT, Cameron JD. Arterial transfer functions and the reconstruction of central aortic waveforms: myths, controversies and misconceptions. J Hypertens 8;26: Gallagher D, Adji A, O Rourke MF. Validation of the transfer function technique for generating central from peripheral upper limb pressure waveform. Am J Hypertens 4;17: Takazawa K, Tanaka N, Takeda K, Kurosu F, Ibukiyama C. Underestimation of vasodilator effects of nitroglycerin by upper limb blood pressure. Hypertension 199;26: Pauca AL, Kon ND, O Rourke MF. The second peak of the radial artery pressure wave represents aortic systolic pressure in hypertensive and elderly patients. Br J Anaesth 4;92: Millasseau SC, Patel SJ, Redwood SR, Ritter JM, Chowienczyk PJ. Pulse wave reflection assessed from the peripheral pulse: is a transfer function necessary? Hypertension 3;41:16.. Adji A, Hirata K, Hoegler S, O Rourke MF. Non-invasive pulse waveform analysis in clinical trials: similarity of two methods for calculating aortic systolic pressure. Am J Hypertens 7;: Hickson SS, Butlin M, Mir FA, et al., on behalf of the ACCT Investigators. The accuracy of central SBP determined from the second systolic peak of the peripheral pressure waveform. J Hypertens 9;27: O Rourke MF. Mechanical principles. Arterial stiffness and wave reflection. Pathol Biol (Paris) 1999;47: O Rourke M. Arterial stiffness, systolic blood pressure and logical treatment of arterial hypertension. Hypertension 199;1:

11 JACC Vol. 7, No. 8, 11 February 11:91 61 Williams et al. Moving Average and Central Systolic Pressure Avolio AP, Van Bortel LM, Boutouyrie P, et al. Role of pulse pressure amplification in arterial hypertension: experts opinion and review of the data. Hypertension 9;4: Soderstrom S, Nyberg G, O Rourke MF, Sellgren J, Ponten J. Can a clinically useful aortic pressure wave be derived from a radial pressure wave? Br J Anaesth 2;88: White WB, Berson AS, Robbins C, et al. National standard for measurement of resting and ambulatory blood pressures with automated sphygmomanometers. Hypertension 1993;21: O Brien E, Pickering T, Asmar R, et al. Working Group on Blood Pressure Monitoring of the European Society of Hypertension international protocol for validation of blood pressure measuring devices in adults. Blood Press Monit 2;7: Hope SA, Tay DB, Meredith IT, Cameron JD. Comparison of generalized and gender-specific transfer functions for the derivation of aortic waveforms. Am J Physiol Heart Circ Physiol 2;283:H Hope SA, Tay DB, Meredith IT, Cameron JD. Use of arterial transfer functions for the derivation of central aortic waveform characteristics in subjects with type 2 diabetes and cardiovascular disease. Diabetes Care 4;27: Pauca A L, O Rourke MF, Kon ND. Prospective evaluation of a method for estimating ascending aortic pressure from the radial artery pressure waveform. Hypertension 1;38: Agabiti-Rosei E, Mancia G, O Rourke M, et al. Central blood pressure measurements and antihypertensive therapy: a consensus document. Hypertension 7;: Laurent S, Cockcroft J, Van Bortel L, et al., on behalf of the European Network for Non-invasive Investigation of Large Arteries. Expert consensus document on arterial stiffness: methodological issues and clinical applications. Eur Heart J 6;27: Kass DA, Chen C-H, Nevo E, et al. Estimation of central aortic pressure waveform by mathematical transformation of radial tonometry pressure data. Circulation 1998;98: Umana E, Ahmed W, Fraley MA, Alpert MA. Comparison of oscillometric and intraarterial systolic and diastolic blood pressures in lean, overweight and obese patients. Angiology 6;7: Williams B, Lacy PS. Central haemodynamics and clinical outcomes: going beyond brachial blood pressure? Eur Heart J ;31: Williams B, Lacy PS. Impact of heart rate on central aortic pressures and hemodynamics: analysis from the CAFE (Conduit Artery Function Evaluation) study. J Am Coll Cardiol 9;4:7 13. Key Words: central aortic systolic pressure y generalized transfer function y noninvasive y n-point moving average y radial artery pressure waveforms. APPENDIX For detailed Methods, Results, and supplementary figures, please see the online version of this article.