ARTICLE

Cardiac baroreflex gain is frequency dependent: insights fromrepeated sit-to-stand maneuvers and the modified Oxford methodHelen M. Horsman, Karen C. Peebles, Duncan C. Galletly, and Yu-Chieh Tzeng

Abstract: Cardiac baroreflex gain is usually quantified as the reflex alteration in heart rate during changes in blood pressurewithout considering the effect of the rate of change in blood pressure on the estimated gain. This study sought to (i) characterizebaroreflex gain as a function of blood pressure oscillation frequencies using a repeat sit-to-stand method and (ii) comparebaroreflex gain values obtained using the sit-to-stand method against the modified Oxford method. Fifteen healthy individualsunderwent the repeated sit-to-stand method in which blood pressure oscillations were driven at 0.03, 0.05, 0.07, and 0.1 Hz.Sixteen healthy participants underwent the sit-to-stand and modified Oxford methods to examine their agreement. Sit-to-standbaroreflex gain was highest at 0.05 Hz (8.8 ± 3.2 ms·mm Hg−1) and lowest at 0.1 Hz (5.8 ± 3.0 ms·mm Hg−1). Baroreflex gains at0.03 Hz (7.7 ± 3.0 ms·mm Hg−1) and 0.07 Hz (7.5 ± 3.3 ms·mm Hg−1) were not different from the baroreflex gain at 0.05 Hz. Therewas moderate correlation between phenylephrine gain and sit-to-stand gain (r values ranged from 0.52 to 0.75; all frequencies,p < 0.05), but no correlation between sodium nitroprusside gain and sit-to-stand gain (r values ranged from –0.07 to 0.22; allp < 0.05). Bland–Altman analysis of phenylephrine gain and sit-to-stand gain showed poor agreement and a positive proportionalbias. These results show that baroreflex gains derived from these 2 methods cannot be used interchangeably. Furthermore,cardiac baroreflex gain is frequency dependent between 0.03 Hz and 0.1 Hz, which challenges the conventional practice ofsummarizing baroreflex gain as a single number.

Key words: blood pressure, heart rate, baroreceptor.

Résumé :On quantifie habituellement le gain du baroréflexe cardiaque par l'altération réflexe du rythme cardiaque au cours desvariations de la pression sanguine sans prendre en compte le taux de variation de la pression sanguine. Cette étude se propose(i) de caractériser par la répétition de la méthode assis–debout le gain du baroréflexe en fonction de la fréquence d'oscillation dela pression sanguine et (ii) de comparer les valeurs du gain du baroréflexe dans la méthode assis–debout aux valeurs de laméthode modifiée d'Oxford. Quinze sujets en bonne santé participent a la répétition de la méthode assis–debout au cours delaquelle on règle les oscillations de la pression sanguine a 0,03, 0,05, 0,07 et 0,1 Hz. Seize sujets en bonne santé participent auxdeuxméthodes, assis–debout et Oxfordmodifiée, pour en vérifier la concordance. Le gain du baroréflexe au cours de la méthodeassis–debout est le plus élevé a 0,05 Hz (8,8 ± 3,2 ms·mm Hg−1) et le plus faible a 0,1 Hz (5,8 ± 3,0 ms·mm Hg−1). Le gain dubaroréflexe a 0,03 Hz (7,7 ± 3,0 ms·mmHg−1) et a 0,07 Hz (7,5 ± 3,3 ms·mmHg−1) ne diffère pas de celui a 0,05 Hz. On observe unecorrélation modérée entre le gain suscité par la phényléphrine et le gain enregistré dans la méthode assis–debout (r entre0,52 et 0,75; p < 0,05 a toutes les fréquences), mais on n'observe pas de corrélation entre le gain suscité par le nitroprussiate desodium et le gain enregistré dans la méthode assis–debout (r entre –0,07 et 0,22; p < 0,05 a toutes les fréquences). L'analyse deBland–Altman des gains suscités par la phényléphrine et la méthode assis–debout révèle une faible concordance et un biaispositif proportionnel. D'après ces observations, le gain du baroréflexe suscité par ces deux méthodes ne peut pas être utilisé defaçon interchangeable. De plus, le gain du baroréflexe cardiaque est lié a la fréquence entre 0,03 Hz et 0,1 Hz, ce quimet en doutela pratique courante de décrire le gain du baroréflexe au moyen d'un seul nombre. [Traduit par la Rédaction]

Mots-clés : pression sanguine, rythme cardiaque, barorécepteur.

IntroductionThe arterial baroreflex plays an important role in modulating

heart rate (HR) and peripheral resistance in response to changes inblood pressure (BP). The assessment of baroreflex gain has beenused extensively to study baroreflex physiology in healthy andpatient populations. Numerous studies have shown that barore-flex impairment is associated with the early development of car-diovascular disease, increased mortality after myocardial infarct,hypertension, diabetes, stroke, and other disease states (Grassiet al. 1995; La Rovere et al. 1998). This knowledge has led to con-siderable interest in the use of baroreflex gain as a clinical risk-stratification index.

Although a variety of techniques have been developed to assessbaroreflex gain, studies indicate that there is little comparabilityacrossmethods (Goldstein et al. 1982; Colombo et al. 1999; Lipmanet al. 2003; La Rovere et al. 2008). One commonly applied tech-nique is the modified Oxford method (Smyth 1969), which usessodium nitroprusside (SNP) and phenylephrine (PE) to pharmaco-logically drive increases and decreases in BP. This method is oftencriticized for its reliance on vasoactive drugs that may affect baro-receptor transduction. However, another potentially importantdrawback is that the technique affords little control over the timecourse of the BP perturbation either between or within individu-

Received 15 November 2012. Accepted 21 February 2013.

H.M. Horsman and Y.-C. Tzeng. Cardiovascular Systems Laboratory, Centre for Translational Physiology, University of Otago, 23A Mein Street, Wellington South, New Zealand; Departmentof Surgery and Anaesthesia, University of Otago, Wellington South, New Zealand.K.C. Peebles. Cardiovascular Systems Laboratory, Centre for Translational Physiology, University of Otago, 23A Mein Street, Wellington South, New Zealand; Department of Surgery andAnaesthesia, University of Otago, Wellington South, New Zealand; Department of Physiology, University of Otago, Wellington South, New Zealand.D.C. Galletly. Cardiovascular Systems Laboratory, Centre for Translational Physiology, University of Otago, 23A Mein Street, Wellington South, New Zealand.

Corresponding author: Yu-Chieh Tzeng (e-mail: shieak.tzeng@otago.ac.nz).

753

Appl. Physiol. Nutr. Metab. 38: 753–759 (2013) dx.doi.org/10.1139/apnm-2012-0444 Published at www.nrcresearchpress.com/apnm on 26 February 2013.

App

l. Ph

ysio

l. N

utr.

Met

ab. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

NIV

ER

SIT

Y O

F N

OR

TH

TE

XA

S L

IBR

AR

Y o

n 11

/23/

14Fo

r pe

rson

al u

se o

nly.

als across repeated tests. Thus, interpretation of cardiac barore-flex gain is usually carried outwithout considering the effects thatdifferences in the rate of change in BP may have on the estimatedgain. In a recent study, Zhang et al. (2009) examined the practical-ities of using repeated squat-to-stand maneuvers performed at0.05 Hz and 0.1 Hz to assess baroreflex function. The study dem-onstrated that baroreflex gain differed between these frequen-cies. However, to our knowledge, no studies have comparedbaroreflex sensitivity measured using the repeated sit-to-standmethod performed over a range of frequencies against the modi-fied Oxford method.

Because the modified Oxford method affords little control overthe time course of BP alterations, the aim of this study was tocompare baroreflex gain assessed using the sit-to-stand and mod-ified Oxford methods, and thereby determine the extent to whichthe methods may be used interchangeably. The repeated sit-to-stand method was performed over an expanded physiologicallyrelevant range to ensure comprehensive evaluation of cardiacbaroreflex gain. We hypothesized that cardiac baroreflex gainwould depend on the time course of BP changes and that the2 methods would show weak correlation and poor agreement.

Materials and methods

ParticipantsParticipants were recruited to perform the sit-to-stand and

modified Oxford methods. Participants (n = 15; 7 males and8 females; ages 31 ± 10 years (mean ± SD); body mass index, 24 ±2.9 kg·m−2) who performed only the sit-to-stand method wereused in protocol 1, which determined the frequency-dependentcharacteristics of the cardiac baroreflex (aim 1). Individuals (n = 16;7 males and 9 females; ages 33 ± 11 years (mean ± SD); body massindex, 26 ± 3.7 kg·m−2) who performed bothmethods were used inprotocol 2, which compared the sit-to-stand procedure with themodified Oxford method (aim 2). Participants were free from re-spiratory, cardiovascular, neurologic, and endocrine disease, andwere not taking any prescription or nonprescriptionmedications,except contraceptive pills. Females were not pregnant and werestudied within the early follicular phase of their menstrual cycleor during their pill-free days. This study was approved by theCentral Regional Ethics Committee and conformed to the stan-dards set by the Declaration of Helsinki. All participants gavewritten informed consent.

MeasurementsContinuous systolic and diastolic arterial BP (SAP and DAP, respec-

tively) was measured using finger photoplethysmography (Finometer,Finapres Medical Systems, Amsterdam, the Netherlands). Meanarterial pressure (MAP) was calculated directly from SAP and DAP.The finometer cuff was fitted around the participant's index ormiddle finger, which was positioned at heart level. Photoplethys-mographic BP measurements were checked against oscillometricBP measurements (Seinex Electronics Ltd., UK) throughout theexperimental procedure. HR was recorded using a 3-lead ECG(ADInstruments, Colorado Springs, Colo., USA). Partial pressure ofend-tidal CO2 (PETCO2) levels were sampled from a nasal cannulaand measured using a CO2 gas analyzer (model ML206, ADInstru-ments). All signals were digitalized using an analog-to-digital con-verter (Powerlab/16SP ML795, ADInstruments). Waveforms weresampled at 1 kHz and stored for offline analysis.

Experimental designParticipants arrived at the laboratory (ambient temperature �22 °C)

at 0900 h, having refrained from caffeine, alcohol, and strenuousexercise for at least 12h.All participantswere thenpositioned supinefor instrumentation. In addition, a venous cannula (18GABD,Nexiva,Utah, USA) was inserted into the antecubital vein of participantsundertaking the modified Oxford method to enable drug infusions.After �10 min of rest to acclimatize to the equipment, the experi-

mental protocol began. All participants underwent 5 min of restingbaseline data collection. Thereafter, protocol 1 participants per-formed the sit-to-stand procedure, whereas protocol 2 participantsunderwent the modified Oxford method before the sit-to-stand pro-cedure. The details of these procedures are given subsequently.

Sit-to-stand methodBefore commencing the sit-to-stand procedure, participants

were repositioned in a straight back chair with their feet flat onthe floor. The arm bearing the finometer was placed in a sling toensure it was relaxed. Participants then performed the sit-to-standmaneuver at 4 frequencies in random order: (i) 0.03 Hz (16.6-s sitfollowed by 16.6-s stand), (ii) 0.05 Hz (10-s sit followed by 10-sstand), (iii) 0.07 Hz (7.1-s sit followed by 7.1-s stand), and (iv) 0.1 Hz(5-s sit followed by 5-s stand). Each sit-to-stand frequency was per-formed for 5 min and separated by at least 5 min of rest forrecovery. The target frequency was controlled using an electronicmetronome. During sit-to-stand maneuvers, participants were in-structed to breathe normally (i.e., avoid the Valsalva maneuver)and to minimize excessive forward flexion when standing up.PETCO2 levels were monitored closely to ensure that each partici-pant was breathing regularly and that no exercise effect was in-curred during the maneuver.

Modified Oxford methodThe modified Oxford method has been described elsewhere

(Tzeng et al. 2009). Briefly, a bolus injection of SNP (100–200 �g)was followed, 60 s later, by a bolus injection of PE (100–300 �g). Aregistered medical practitioner delivered these drugs, and thedoses were titrated to achieve 15-mm Hg decreases (SNP) and in-creases (PE) in BP. The modified Oxford method was performedtwice (trial 1 and trial 2), with at least 15 min of separation. Partic-ipants, who were in a supine position, were monitored closely forECG irregularities and untoward cardiovascular symptoms be-fore, during, and after the method.

Data analysisAll data analysis was conducted using custom-written software

in LabView 8.2 (National Instruments).

Sit-to-stand methodWe assessed the cardiac baroreflex during the sit-to-stand pro-

cedure as described by others (Iwasaki et al. 2000; Zhang et al.2009). The recorded ECG and BP measures for each frequencywere checked for the presence of artifacts, and erroneously de-tected or missed R waves were corrected by linear interpolation.All time series were filtered and resampled at 4 Hz to obtainequidistant data points and were then linearly detrended. Theseseries were then passed through a Hanning window before under-going a fast Fourier transformation. Power spectral analysis wasused to derive the R–R interval spectral density and BP spectraldensity, which are measures of R–R interval and BP variability,respectively. The product of the SAP signal (input) and the R–Rinterval (output) provided the cross-spectrum from which trans-fer function gain, phase, and coherence, at each sit-to-stand fre-quency, were derived. Transfer function gain, which representsthe change in output over the change in input, reflects cardiacbaroreflex function. Transfer function phase, which representsthe lag between the input and output signals, is indicative of thelatency of the baroreflex response. Transfer function coherence,which reflects the fraction of the output power that can be linear-ity related to the input power, is a measure of the correlationbetween the 2 time series. Sit-to-stand gain (GainSS) was taken asan estimate of cardiac baroreflex function, provided the esti-mated squared coherence was >0.6, indicating that the input andoutput signals were related.

754 Appl. Physiol. Nutr. Metab. Vol. 38, 2013

Published by NRC Research Press

App

l. Ph

ysio

l. N

utr.

Met

ab. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

NIV

ER

SIT

Y O

F N

OR

TH

TE

XA

S L

IBR

AR

Y o

n 11

/23/

14Fo

r pe

rson

al u

se o

nly.

Modified Oxford methodWe calculated cardiac baroreflex gain during the modified Ox-

ford method as described previously (Taylor et al. 2011). Each rawdata set was checked for the presence of artifacts, and erroneouslydetected or missed R waves were corrected by linear interpola-tion. The R–R interval was plotted against SAP to visually inspectthe relationship and to identify the saturation and threshold re-gions (Lipman et al. 2003). The saturation points and thresholdwere identified automatically using a piecewise linear regressionapproach as described previously (Studinger et al. 2007). The slopeof the linear regression was taken as an estimate of baroreflexgain only when the correlation coefficient was >0.8 (Willie et al.2011; Taylor et al. 2011). This criterion was achieved in all partici-pants. The changes in SAP were matched to the concurrent heartperiod, or, to allow for baroreflex delays at shorter heart periods(between 500 and 800 ms), a 1-beat delay was used. To account forrespiratory oscillations in R–R interval and SAP readings, bothtime series were averaged across 3mmHg bins (Joseph et al. 2005).The final baroreflex gain estimates were taken as the average of the2 trials. We derived 3 cardiac baroreflex gain measures from themodified Oxford method: (i) SNP gain (GainSNP), (ii) PE gain (GainPE),and (iii) combined SNP and PE gain (GainCOM).

Statistical analysesData were tested for normality using the Shapiro–Wilk test.

Differences in GainSS and BP spectral density at each frequencywere assessed using linear mixed models accounting for fixedeffects, and p values were adjusted using the Sidak correction tocontrol for type I errors. SAP changes, rate of BP change, andcardiac baroreflex gain at each sit-to-stand frequency and duringthe modified Oxford method were assessed using 1-way repeated-measures analysis of variance (ANOVA) (Greenhouse–Geisser cor-rected). The Bonferroni–Dunn test was used for post hoc analysiswhen a significant effect was found. All analyses were conductedusing SPSS 20 (SPSS Inc.). Between-method relationships werequantified using Pearson's correlation coefficient. At each sit-to-stand frequency, we analyzed correlations between GainSS and(i) GainSNP, (ii) GainPE, and (iii) GainCOM. In the case of a significantcorrelation, the Bland–Altman analysis (Bland and Altman 1986)was used to assess agreement betweenmethods and possible bias.Provided the differences between the methods followed a normaldistribution, 95% prediction bands were calculated as describedpreviously (Hamilton and Stamey 2009). Correlation and Bland–Altman analyses were conducted with GraphPad Prism (version5.0c). Significance was defined at an � level of p < 0.05 for allcomparisons. Data are presented as means ± SD.

ResultsAll participants completed the study, and the resultant datawas

normally distributed. Resting cardiovascular parameters were notsignificantly different between protocols. For protocol 1 and pro-tocol 2, the HR was 64 ± 7 beats·min−1 and 65 ± 8 beats·min−1, SAPwas 121 ± 13 mm Hg and 119 ± 15 mm Hg, DAP was 64 ± 9 mm Hgand 66 ± 11 mm Hg, and MAP was 81 ± 12 mm Hg and 83 ± 13 mmHg, at rest, respectively.

Assessment of the cardiac baroreflex during the sit-to-standmethod

Table 1 summarizes the mean cardiovascular parameters ob-served during the sit-to-stand maneuver at each frequency. HR(and R–R interval), SAP, DAP, and MAP were not significantly dif-ferent between frequencies. Table 2 summarizes the transferfunction metrics for the 15 participants undertaking protocol 1.Linear mixed-model analysis showed a significant main effect forfrequency (p < 0.01), indicating that the cardiac baroreflex gaindiffered with frequency. Post hoc analysis showed that GainSS at0.03, 0.05, and 0.07 Hz was significantly higher than GainSS ob-served at 0.1 Hz. The highest GainSS was observed at 0.05 Hz (8.8 ±

3.2 ms·mm Hg−1), and the lowest at 0.1 Hz (5.8 ± 3.0 ms·mm Hg−1).GainSS at 0.03 Hz and 0.07 Hz were lower than at 0.05 Hz, but notsignificantly different. Linear mixed-model analysis of BP spectraldensity data showed a significant main effect (p < 0.05) for fre-quency, indicating that BP spectral density also differed with fre-quency. Post hoc analysis showed that BP spectral density wassignificantly higher at 0.07 Hz (107 ± 83 mm Hg2 × 10−2) than at0.03Hz, 0.05Hz, and0.1Hz (25± 13, 58±49), and53±33mmHg2× 10−2,respectively). There was no significant difference in BP spectral densityat 0.03, 0.05, or 0.1 Hz.

Comparison of the modified Oxford method and the sit-to-stand method

The peak SAP and the SAP range, the rate of BP change, and thecardiac baroreflex gain during the modified Oxford method andtheir comparison with the sit-to-standmethod are summarized inTable 3. The peak SAP was the same during both methods; how-ever, the rate of BP change and the baroreflex gain were different.In general, the rate of BP change increased with sit-to-stand fre-quency. Irrespective of frequency, the rate of BP change duringthe sit-to-stand maneuver was significantly faster than the rate ofBP change during the modified Oxford method. However, GainPEand GainCOMwere greater than GainSS and reached significance at0.03 Hz and 0.1 Hz.

Table 4 summarizes the correlation between the baroreflex gainevaluated using the modified Oxford method and the sit-to-standmethod. No correlation was found between GainSNP and GainSS atany frequency (r values ranged from –0.07 to 0.22; NS). However,GainPE and GainSS showedmoderate correlation at all frequencies (rvalues ranged from 0.52 to 0.75; p < 0.05), whereas GainCOM andGainSS showedmoderate correlation at 0.03, 0.05, and 0.07Hz only (rvalues were 0.70, 0.52, and 0.55, respectively; p < 0.05) (Table 4).Probing this relationship further, Bland–Altman analysis showedthat across frequencies, the mean differences in baroreflex gain(GainPE minus GainSS, and GainCOM minus GainSS) were clusteredwithin the 95% prediction bands. The upper and lower 95% predic-tion bands were �7ms·mmHg−1 higher and lower than the averageof the 2 measures, respectively, and varied proportionately acrossthe range of the averages (Figs. 1 and 2). Bland–Altman analysis alsorevealed that asGainPE increased, so toodid thediscrepancybetweenthemethods. These findings are indicative of a positive proportionalbias whereby GainPE consistently overestimated GainSS at all fre-quencies (Fig. 1).

Discussion

Main findingsThe main findings of our study are twofold. First, in healthy

adults, we demonstrated that cardiac baroreflex gain assessed us-ing the repetitive sit-to stand method differed with the frequencyof BP oscillation and was highest at 0.05 Hz. GainSS at 0.1 Hz wassignificantly lower than GainSS at 0.03, 0.05, and 0.07 Hz. Thesefindings support the hypothesis that the cardiac baroreflex gainexhibits frequency dependence between 0.03 Hz and 0.1 Hz. Sec-ond, we showed a weak-to-moderate correlation between GainPEand GainSS, and GainCOM and GainSS, but found no correlation

Table 1. Cardiovascular parameters at each sit-to-stand frequency.

Sit-to-stand frequency

0.03 Hz 0.05 Hz 0.07 Hz 0.1 Hz

HR (beats·min–1) 79±8.8 81±9.2 81±9.8 86±9.5R–R Interval (s) 0.78±0.08 0.76±0.08 0.76±0.09 0.71±0.08SAP (mm Hg) 129±20.0 129±14 125±16 128±19.0DAP (mm Hg) 70±10.0 67±7.8 62±7.2 66±10.7MAP (mm Hg) 88±14 85±9.4 81±9.0 85±13

Note: Data are presented as means ± SD. There were no statistical differencesbetween groups. HR, heart rate; SAP, systolic arterial blood pressure; DAP, dia-stolic arterial blood pressure; MAP, mean arterial blood pressure.

Horsman et al. 755

Published by NRC Research Press

App

l. Ph

ysio

l. N

utr.

Met

ab. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

NIV

ER

SIT

Y O

F N

OR

TH

TE

XA

S L

IBR

AR

Y o

n 11

/23/

14Fo

r pe

rson

al u

se o

nly.

betweenGainSNP andGainSS. Bland–Altman analysis of GainPE andGainSS, and GainCOM and GainSS, revealed poor levels of agree-ment; additionally, GainPE and GainSS showed a positive propor-tional bias. These findings indicate that there is poor concordancebetween the cardiac baroreflex gain estimated using themodifiedOxford method and the repeated sit-to-stand method. Beforethese findings are examined and the implications considered,methodologic considerations are discussed.

Methodologic considerationsThe use of the sit-to-stand maneuver to drive BP oscillations

means that the observed changes in HR are likely to reflect thefeedback component of the baroreflex loop. This is an importantconsideration because non-baroreflex-mediated HR changes alsooccur. These changes may be mediated by feed-forward mecha-nisms, such as the central processes associated with respirationand cognitive processing (Lipman et al. 2003). Therefore, it is nec-essary to reduce the contribution from the feed-forward compo-nent by driving BP oscillations to open the reflex loop (Kamiyaet al. 2011; Akimoto et al. 2011; Tan and Taylor 2011). Although it isnot practical to completely open the baroreflex loop in humanstudies, partial opening of the loop can be achieved by activelyengaging the baroreflex (Diaz and Taylor 2006). Even so, we rec-ognize that it is not possible to quantify the contributions madeby either the feedback or the feed-forward components. Addition-ally, whether the feedback and feed-forward contributions varywith the frequency of the sit-to-stand maneuver and thus whateffect there is on baroreflex gain estimation are not known.

We did not examine the effect of posture in this study andcannot explicitly define any effects it may have on thesemethods.

However, it is relevant to acknowledge that the modified Oxfordmethod was carried out in the supine position, whereas the sit-to-stand method was carried out in the upright position. Changes inposture alter the relative contributions from the sympathetic andparasympathetic nervous systems, and standing, as opposed tosupine, may cause a reduction in cardiac baroreflex gain and anincrease in sympathetic gain (O'Leary et al. 2003). Moreover,head-up tilt has been shown to directly reduce cardiovagal baro-reflex gain and may also account for some of the differences be-tween methods (Steinback et al. 2005). Additionally, it should benoted that the BP spectral density varied across all frequenciesand was significantly higher at 0.07 Hz. The extent to whichchanges in the magnitude of BP oscillations contributed to ourresults is not known.

We also acknowledge that the modified Oxford method, byconvention, separates the baroreflex response into decreases andincreases in BP, whereas the sit-to-stand method does not. This isdue to the application of transfer function analysis for data han-dling. An inherent limitation of transfer function analysis is thatit treats pressure and R–R interval changes as sinusoidal oscilla-tions without distinction between the “up” and “down” phases ofBP waves. Therefore, the method cannot be used to characterizebaroreflex hysteresis (Rudas et al. 1999). This difference in datahandling betweenmethodsmay have influenced our results whencomparing the 2 methods; therefore, we calculated the average ofGainSNP and GainPE (GainCOM) to minimize this effect.

Frequency-dependent characteristics of the baroreflexOur results show that when healthy individuals perform re-

peated sit-to-stand maneuvers across a range of frequencies, therelationship between the R–R interval and SAP differs. We foundthat the greatest GainSS occurred at 0.05 Hz and the lowest at0.1 Hz, which corroborates with and expands on the findings inhumans (Zhang et al. 2009) and rats (Kawada et al. 2011). Althoughthe main effect for frequency was significant, indicating thatthere is frequency dependence, post hoc analysis showed that theonly significant comparison was between GainSS at 0.1 Hz andGainSS at 0.03, 0.05, and 0.07 Hz. Our inability to detect significantdifferences in other comparisons may be due to insufficient sta-tistical power after correcting for multiple comparisons.

Our findings also showed that the peak SAP, the SAP range, andthe BP spectral density were similar between the 0.05-Hz and 0.1-Hzfrequencies. By implication, these results suggest that the difference

Table 2. Coherence, gain, and BP spectral density at each sit-to-stand frequency.

Sit-to-stand frequency

0.03 Hz 0.05 Hz 0.07 Hz 0.1 Hz p

Coherence (AU) 0.81±0.10 0.92±0.08 0.97±0.02 0.96±0.04 NAGain (ms·mm Hg–1) 7.7±3.0 8.8±3.2 7.5±3.3 5.8±3.0�,†,‡ <0.01BP PSD (mm Hg2·Hz–1 × 10–2) 25±13 58±49 107±83�,†,§ 53±33 <0.01

Note:Data are presented asmeans ± SD. The p values denote themain effects for frequency in a linearmixed model. BP, blood pressure; BP PSD, blood pressure power spectral density; NA, not applicable. �,p < 0.05 compared with 0.03 Hz; †, p < 0.05 compared with 0.05 Hz; ‡, p < 0.05 compared with 0.07 Hz;§, p < 0.05 compared with 0.1 Hz.

Table 3. BP changes and cardiac baroreflex gain at each sit-to-stand frequency and during the modified Oxford method.

Sit-to-stand frequency Modified Oxford method

0.03 Hz 0.05 Hz 0.07 Hz 0.1 Hz PE SNP Combined p

Peak SAP (mm Hg) 149±23 148±17 144±20 150±22 146±18 (130±13) NA NSSAP range (mm Hg) 120–149 118–148 110–144 120–150 110–146 102–130 NA NARate of BP change (mm Hg·s–1) 1.8±0.8 3.1±1.1�,†,‡ 4.7±1.6�,§ 6.0±2.7�,§ 1.0±0.4�,†,‡,§ 0.9±0.4�,†,‡,§ NA <0.01Gain (ms·mm Hg–1) 7.4±2.9 8.8±3.1 7.8±3.0 5.7±3.2 11.8±7.3�,‡ 8.5±3.9 10.2±4.2�,‡ <0.01

Note: Data are presented as means ± SD unless indicated otherwise. Peak SAP after SNP infusion was not included in the statistical analysis because SNP evokes adecrease in BP. The p values denote the main effects for frequency in 1-way repeated-measures analysis of variance (ANOVA). BP, blood pressure; PE, phenylephrine;SNP, sodium nitroprusside; Combined, averaged SNP and PE gains; SAP, systolic arterial blood pressure; NA, not applicable; NS, not significant. �, p < 0.05 comparedwith 0.03 Hz; †, p < 0.05 compared with 0.07 Hz; ‡, p < 0.05 compared with 0.1 Hz; §, p < 0.05 compared with 0.05 Hz.

Table 4. Relationships between cardiac baroreflex gain measured us-ing the sit-to-stand and the modified Oxford methods.

Correlationwith sit-to-standmethod

0.03 Hz 0.05 Hz 0.07 Hz 0.1 Hz

Modified Oxford methodGainSNP 0.01 0.22 –0.07 0.01GainPE 0.75� 0.52� 0.69� 0.58�

GainCOM 0.70� 0.52� 0.55� 0.48

Note: GainSNP, sodium nitroprusside gain; GainPE, phenylephrine gain; GainCOM,combined sodiumnitroprusside and phenylephrine gain. Values are Pearson's correla-tion coefficient. �, p < 0.05.

756 Appl. Physiol. Nutr. Metab. Vol. 38, 2013

Published by NRC Research Press

App

l. Ph

ysio

l. N

utr.

Met

ab. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

NIV

ER

SIT

Y O

F N

OR

TH

TE

XA

S L

IBR

AR

Y o

n 11

/23/

14Fo

r pe

rson

al u

se o

nly.

in GainSS between 0.05 Hz and 0.1 Hz is frequency rather than mag-nitudedependent. Itwasnot the intentionof this study to determinethemechanisms underlying frequency dependence. However, thesefindings have led us to consider whether different frequencies of BPoscillationmayalter thefiringpatternofbaroreceptor afferentnervefibres, therebyaccounting for alterations inbaroreflexgain. Support-ing this suggestion are 2 subtypes of baroreceptor afferents (A and Cfibres). Both fibre types operate at normal BPs but differ in theirfiring rates, sensitivities to BP, operational thresholds, and satura-tion points. For example, it has been shown in an animalmodel thatbaroreceptor afferents display differentfiringpatterns during rampsof increasing and decreasing BP (Bolter et al. 2011). Furthermore, thebaroreceptors are known to respond to both the frequency aswell asthemagnitude of BP change (Hammer and Saul 2005). It is thereforepossible that the different gain values found across sit-to-stand fre-

quencies may be due to different recruitment patterns of afferentfibres.

Comparison of the modified Oxford method and the sit-to-stand method

Our results showed no correlation between GainSNP and GainSS

(r < 0.1) at any frequency, mild-to-moderate correlation betweenGainPE and GainSS at all frequencies (r range, 0.52–0.75), andmild-to-moderate correlation between GainCOM and GainSS at 0.03,0.05, and 0.07 Hz (r range, 0.52–0.70). Interestingly the highestcorrelation was observed between GainPE and sit-to-stand at 0.03 Hz,and the lowest correlation was found at 0.05 Hz. Moderatecorrelation between methods does not necessarily imply goodagreement between methods, and further analysis using theBland–Altman method of differences showed a positive propor-

Fig. 1. Bland–Altman plots for comparison of cardiac baroreflex gain obtained from the modified Oxford method (phenylephrine infusion)and the sit-to-stand method (0.03, 0.05, 0.07, and 0.1 Hz). GainPE, phenylephrine gain; GainSS, sit-to-stand gain; r2 denotes coefficient ofdetermination.

-10

0

10

20

GainPE vs. GainSS (0.03 Hz) r2 = 0.73 (p < 0.01)

Gai

n PE

- Gai

n SS

(ms ·m

m H

g–1)

GainPE vs. GainSS (0.07 Hz) r2 = 058 (p < 0.01)

0 5 10 15 20

-10

0

10

20

(GainPE + GainSS)/2 (ms·mm Hg–1)

Gai

n PE

- Gai

n SS

(ms ·m

m H

g–1)

GainPE vs. GainSS (0.05 Hz) r2 = 0.51 (p < 0.01)

GainPE vs. GainSS (0.1 Hz) r2 = 054 (p < 0.01)

0 5 10 15 20

(GainPE + GainSS)/2 (ms·mm Hg–1)

Fig. 2. Bland–Altman plots for comparison of cardiac baroreflex gain obtained from the modified Oxford method (combined sodiumnitroprusside and phenylephrine gain) and the sit-to-stand method (0.03, 0.05, and 0.07 Hz). Bland–Altman plots were not constructed betweenGainCOM and GainSS at 0.1 Hz because the Pearson's correlation coefficient was not significant. GainCOM, combined sodium nitroprusside and phen-ylephrine gain; GainSS, sit-to-stand gain; r2 denotes coefficient of determination.

GainCOM vs. GainSS (0.03 Hz) r2 = 0.19 (NS)

0 5 10 15 20

-10

0

10

20

(GainCOM + GainSS)/2 (ms·mm Hg–1)

Gai

n CO

M -

Gai

n SS

(ms ·m

m H

g–1)

GainCOM vs. GainSS (0.05 Hz) r2 = 0.11 (NS)

0 5 10 15 20

(GainCOM + GainSS)/2 (ms·mm Hg–1)

GainCOM vs. GainSS (0.07 Hz) r2 = 0.08 (NS)

0 5 10 15 20

(GainCOM + GainSS)/2 (ms·mm Hg–1)

Horsman et al. 757

Published by NRC Research Press

App

l. Ph

ysio

l. N

utr.

Met

ab. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

NIV

ER

SIT

Y O

F N

OR

TH

TE

XA

S L

IBR

AR

Y o

n 11

/23/

14Fo

r pe

rson

al u

se o

nly.

tional bias combined with wide limits of agreement betweenGainPE and GainSS. GainPE was consistently higher than GainSS; insome cases, the difference between methods was greater thandouble the gain itself, as indicated by the wide prediction bands.Moreover, the disparity between GainPE and GainSS widened asGainPE levels increased. Because higher values of gain are gener-ally accepted to be an indicator of good baroreflex function, acomparison between GainPE and GainSS in a healthy population isless likely to agree across methods than a comparison in individualswith reduced baroreflex function. Interestingly, Bland–Altman anal-ysis between GainCOM and GainSS did not show a positive propor-tional bias. However, the prediction bands are of a similar widthto those observed when comparing GainPE and GainSS, indicatingthat combining GainSNP and GainPE does not improve the agree-ment betweenmethods. Our results show that the potential discrep-ancy between the modified Oxford and sit-to-stand methods is�7 ms·mm Hg−1. It is plausible that an equivalent drop in barore-flex gain may occur after an intervention such as a 40-min bout ofmoderate exercise (Willie et al. 2011); therefore, we suggest thatthe baroreflex gains assessed by these methods cannot be usedinterchangeably. Furthermore, reference ranges derived from themodified Oxford method should not be applied to the sit-to-standmethod.

It is unclear why these 2 methods gave different estimates ofbaroreflex gain, although several possibilities warrant consider-ation. First, given the suggestion that a vasoconstricting stimulusincreases baroreceptor sensitivity (O'Leary et al. 2005), we consid-ered whether PE infusion and sit-to-stand evoked different levelsof vasoconstriction, thereby explaining the gain differences. How-ever, in the context of our current findings, this seems unlikelybecause the peak SAP (and SAP range) were not significantly dif-ferent between methods. Second, we considered whether the dif-ferences in gain may have been related to different rates of BPchange, the implication being that different rates of mechanicaldistortion may cause different patterns of baroreceptor afferentnerve recruitment (Bolter et al. 2011), thus giving rise to differentgain values. Interestingly, there were differences in the rate of BPchange between methods. Specifically, the rate of BP change wasslower after PE (and SNP) than during sitting-to-standing. GainPEwas generally higher than GainSS, which suggests that the cardiacbaroreflex response is more effective when the rate of BP changeis slowest. Given these considerations, it would appear that thehigher gain after PE infusion than sit-to-stand relates to differ-ences in the rate of BP change rather than to the magnitude of BPincreases between methods.

Implications of cardiac baroreflex frequency dependenceThese findings support the notion that cardiac baroreflex gain

is frequency dependent (Kawada et al. 2011; Zhang et al. 2009) andthat full characterization of the baroreflex may require assess-ment of baroreflex sensitivity (BRS) across a range of frequencies(Parati et al. 1995). Additionally, when the sit-to-stand method iscompared with the modified Oxford method, both the correlationand the agreement are poor, indicating that these 2 methods yielddifferent information about the baroreflex. This calls into questionthe validity of assigning, even colloquially, the term “gold standard”to any onemethod. These results also highlight the conceptual inad-equacies of summarizing and reducing baroreflex function to a sin-gle number. The prognostic usefulness of assessing baroreflex gainhas been obtained mainly from studies using variations of the Ox-ford method (La Rovere et al. 2008). The Autonomic Tone and Re-flexes After Myocardial Infarction (ATRAMI) study (La Rovere et al.1998) suggests that values of <3 ms·mm Hg−1 are indicative of a sig-nificant cardiac risk, whereas a BRS >6 ms·mmHg−1 indicates a bet-ter prognosis. BRS values of �15 ms·mm Hg−1 indicate normalbaroreflex function (La Rovere and Raczak 2006). Our data showedthat GainPE was consistently higher than GainSS, and if the abovereference ranges were applied to sit-to-stand data, the results would

be considered low andwould offer a poor prognosis to an apparentlyhealthy individual. Thus, robust clinical application of BRS esti-mated using different methods requires the development ofmethod-specific reference ranges.

ConclusionIn conclusion, this study showed that there is poor agreement

between baroreflex gain measured using the modified Oxford andthe sit-to-stand methods. Indeed, because the baroreflex gain exhib-its frequency dependence, we suggest that it cannot be summarizedcomprehensivelybya singlenumber.Accordingly, quantificationsofbaroreflex gain assessed using these methods cannot be used inter-changeably.

AcknowledgementsWethankthevolunteers for theirparticipationandDr. JamesStanley

for statistical advice. HMH, KCP, DCG, and YCT declare they have noconflictsof interest. Thisworkwas supportedbyHealthResearchCoun-cil of New Zealand Grant 11/125 (YCT).

ReferencesAkimoto, T., Sugawara, J., Ichikawa, D., Terada, N., Fadel, P.J., and Ogoh, S. 2011.

Enhanced open-loop but not closed-loop cardiac baroreflex sensitivity duringorthostatic stress in humans. Am. J. Physiol. Regul. Integr. Comp. Physiol.301(5): R1591–R1598. doi:10.1152/ajpregu.00347.2011. PMID:21900646.

Bland, J., and Altman, D.G. 1986. Statistical methods for assessing agreementbetween two methods of clinical measurement. Lancet, 327(8476): 307–310.doi:10.1016/S0140-6736(86)90837-8.

Bolter, C.P., Turner, M.J., and Barrett, C.J. 2011. The relationship between aorticbaroreceptor activity and arterial pressure is not monotonic. J. Physiol. Sci.61(2): 151–160. doi:10.1007/s12576-011-0132-4. PMID:21240644.

Colombo, R., Mazzuero, G., Spinatonda, G., Lanfranchi, P., Giannuzzi, P.,Ponikowski, P., Coats, A.J.S., andMinuco, G. 1999. Comparison between spec-tral analysis and the phenylephrine method for the assessment of baroreflexsensitivity in chronic heart failure. Clin. Sci. 97: 503–513. doi:10.1042/CS19990035. PMID:10491351.

Diaz, T., and Taylor, J.A. 2006. Probing the arterial baroreflex: is there a ‘spon-taneous’ baroreflex? Clin. Auton. Res. 16(4): 256–261. doi:10.1007/s10286-006-0352-5. PMID:16732466.

Goldstein, D., Horwitz, D., and Keiser, H. 1982. Comparison of techniques formeasuring baroreflex sensitivity in man. Circulation, 66(2): 432–439. doi:10.1161/01.CIR.66.2.432. PMID:7094250.

Grassi, G., Seravalle, G., Cattaneo, B.M., Lanfranchi, A., Vailati, S.,Giannattasio, C., Del Bo, A., Sala, C., Bolla, G.B., and Pozzi, M. 1995. Sympa-thetic activation and loss of reflex sympathetic control in mild congestiveheart failure. Circulation, 92(11): 3206–3211. doi:10.1161/01.CIR.92.11.3206.PMID:7586305.

Hamilton, C., and Stamey, J.D. 2009. Using a prediction approach to assessagreement between two continuous measurements. J. Clin. Monit. Comput.23(5): 311–314. doi:10.1007/s10877-009-9198-4. PMID:19701687.

Hammer, P.E., and Saul, J.P. 2005. Resonance in a mathematical model ofbaroreflex control: arterial blood pressure waves accompanying posturalstress. Am. J. Physiol. Regul. Integr. Comp. Physiol. 288(6): R1637. doi:10.1152/ajpregu.00050.2004. PMID:15718393.

Iwasaki, K.I., Zhang, R., Zuckerman, J.H., Pawelczyk, J.A., and Levine, B.D. 2000.Effect of head-down-tilt bed rest and hypovolemia on dynamic regulation ofheart rate and blood pressure. Am. J. Physiol. Regul. Integr. Comp. Physiol.279(6): R2189–R2199. PMID:11080085.

Joseph, C.N., Porta, C., Casucci, G., Casiraghi, N., Maffeis, M., Rossi, M., andBernardi, L. 2005. Slow breathing improves arterial baroreflex sensitivity anddecreases blood pressure in essential hypertension. Hypertension, 46(4):714–718. doi:10.1161/01.HYP.0000179581.68566.7d. PMID:16129818.

Kamiya, A., Kawada, T., Shimizu, S., and Sugimachi, M. 2011. Closed-loop spon-taneous baroreflex transfer function is inappropriate for system identifica-tion of neural arc but partly accurate for peripheral arc: predictabilityanalysis. J. Physiol. 589(7): 1769–1790. doi:10.1113/jphysiol.2010.203455. PMID:21486839.

Kawada, T., Shimizu, S., Kamiya, A., Sata, Y., Uemura, K., and Sugimachi, M. 2011.Dynamic characteristics of baroreflex neural and peripheral arcs are pre-served in spontaneously hypertensive rats. Am. J. Physiol. Regul. Integr.Comp. Physiol. 300(1): R155–R165. doi:10.1152/ajpregu.00540.2010. PMID:21048073.

La Rovere, M., and Raczak, G. 2006. How to measure baroreflex sensitivity. FoliaCardiol. 13(7): 630–637.

La Rovere, M.T., Bigger, J.T., Jr., Marcus, F.I., Mortara, A., and Schwartz, P.J. 1998.Baroreflex sensitivity and heart-rate variability in prediction of total cardiacmortality after myocardial infarction. ATRAMI (Autonomic Tone and Re-

758 Appl. Physiol. Nutr. Metab. Vol. 38, 2013

Published by NRC Research Press

App

l. Ph

ysio

l. N

utr.

Met

ab. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

NIV

ER

SIT

Y O

F N

OR

TH

TE

XA

S L

IBR

AR

Y o

n 11

/23/

14Fo

r pe

rson

al u

se o

nly.

flexes After Myocardial Infarction) Investigators. Lancet, 351(9101): 478. doi:10.1016/S0140-6736(97)11144-8. PMID:9482439.

La Rovere, M.T., Pinna, G.D., and Raczak, G. 2008. Baroreflex sensitivity: mea-surement and clinical implications. Ann. Noninvasive Electrocardiol. 13(2):191–207. doi:10.1111/j.1542-474X.2008.00219.x. PMID:18426445.

Lipman, R.D., Salisbury, J.K., and Taylor, J.A. 2003. Spontaneous indices areinconsistent with arterial baroreflex gain. Hypertension, 42(4): 481–487. doi:10.1161/01.HYP.0000091370.83602.E6. PMID:12975383.

O'Leary, D., Kimmerly, D., Cechetto, A., and Shoemaker, J. 2003. Differentialeffect of head-up tilt on cardiovagal and sympathetic baroreflex sensitivity inhumans. Exp. Physiol. 88(06): 769–774. doi:10.1113/eph8802632. PMID:14603376.

O'Leary, D., Steinback, C.D., Cechetto, A.D., Foell, B.T., Topolovec, J.C.,Gelb, A.W., Cechetto, D.F., and Shoemaker, J.K. 2005. Relating drug-inducedchanges in carotid artery mechanics to cardiovagal and sympathetic barore-flex control. Can. J. Physiol. Pharmacol. 83(5): 439–446. doi:10.1139/y05-030.PMID:15897926.

Parati, G., Saul, J.P., Di Rienzo, M., andMancia, G. 1995. Spectral analysis of bloodpressure and heart rate variability in evaluating cardiovascular regulation: acritical appraisal. Hypertension, 25(6): 1276–1286. doi:10.1161/01.HYP.25.6.1276. PMID:7768574.

Rudas, L., Crossman, A.A., Morillo, C.A., Halliwill, J.R., Tahvanainen, K.U.,Kuusela, T.A., and Eckberg, D.L. 1999. Human sympathetic and vagal barore-flex responses to sequential nitroprusside and phenylephrine. Am. J. Physiol.276(5): H1691–H1698. PMID:10330255.

Smyth, H.S. 1969. Reflex regulation of arterial pressure during sleep in man.Circ. Res. 24: 109–121. doi:10.1161/01.RES.24.1.109. PMID:4303309.

Steinback, C.D., O'Leary, D.D., Bakker, J., Cechetto, A.D., Ladak, H.M., andShoemaker, J.K. 2005. Carotid distensibility, baroreflex sensitivity, and or-thostatic stress. J. Appl. Physiol. 99(1): 64–70. doi:10.1152/japplphysiol.01248.2004. PMID:15731395.

Studinger, P., Goldstein, R., and Taylor, J.A. 2007. Mechanical and neural contri-butions to hysteresis in the cardiac vagal limb of the arterial baroreflex.J. Physiol. 583(3): 1041. doi:10.1113/jphysiol.2007.139204. PMID:17640929.

Tan, C.O., and Taylor, J.A. 2011. Feedback and feedforward sympathetic haemo-dynamic control: chicken or egg? J. Physiol. 589(8): 1867–1867. doi:10.1113/jphysiol.2011.208199. PMID:21498380.

Taylor, C.E., Atkinson, G., Willie, C.K., Jones, H., Ainslie, P.N., and Tzeng, Y.C.2011. Diurnal variation in the mechanical and neural components of thebaroreflex. hypertension, 58(1): 51–56. doi:10.1161/HYPERTENSIONAHA.111.171512. PMID:21632473.

Tzeng, Y., Sin, P.Y.W., Lucas, S.J.E., and Ainslie, P.N. 2009. Respiratory modula-tion of cardiovagal baroreflex sensitivity. J. Appl. Physiol. 107(3): 718. doi:10.1152/japplphysiol.00548.2009. PMID:19608928.

Willie, C.K., Ainslie, P.N., Taylor, C.E., Jones, H., Sin, P.Y.W., and Tzeng, Y.C. 2011.Neuromechanical features of the cardiac baroreflex after exercise. Hyperten-sion, 57(5): 927. doi:10.1161/HYPERTENSIONAHA.110.164616. PMID:21422379.

Zhang, R., Claassen, J.A.H.R., Shibata, S., Kilic, S., Martin-Cook, K.,Diaz-Arrastia, R., and Levine, B.D. 2009. Arterial-cardiac baroreflex function:insights from repeated squat–stand maneuvers. Am. J. Physiol. Regul. Integr.Comp. Physiol. 297(1): R116. doi:10.1152/ajpregu.90977.2008. PMID:19420293.

Horsman et al. 759

Published by NRC Research Press

App

l. Ph

ysio

l. N

utr.

Met

ab. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

NIV

ER

SIT

Y O

F N

OR

TH

TE

XA

S L

IBR

AR

Y o

n 11

/23/

14Fo

r pe

rson

al u

se o

nly.