+ Models

BIOPSY-5897; No of Pages 8

Biological Psychology xxx (2007) xxx–xxx www.elsevier.com/locate/biopsycho

Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia Xavier Bornas a,*, Jordi Llabre´s a, Miquel Tortella-Feliu a, Miquel A. Fullana b, Pedro Montoya a, Ana Lo´pez c, Miquel Noguera d, Joan M. Gelabert e a

University Research Institute on Health Sciences (IUNICS), Department of Psychology, University of the Balearic Islands, Spain b Department of Psychiatry, Autonomous University of Barcelona, Spain c Departament of Experimental Psychology, University of Sevilla, Spain d Department of Applied Mathematics 2, Technical University of Catalonia, Spain e Department of Psychology, University of the Balearic Islands, Spain Received 18 July 2006; accepted 25 July 2007

Abstract In the present study a computer-assisted exposure-based treatment was applied to 54 flight phobics and the predictive role of vagally mediated heart rate (HR) variability (high frequency, 0.15–0.4 Hz band power) and heart rate entropy (HR time series sample entropy) on treatment outcome was investigated. Both physiological measures were taken under controlled breathing at 0.2 Hz and during exposure to a fearful sequence of audiovisual stimuli. Hierarchical regression analyses were conducted to assess the predictive power of these variables in these conditions on treatment self-report measures at the end of treatment and at 6 months follow-up, as well as on the behavioral treatment outcome (i.e. flying at the end of treatment). Regression models predicting significant amounts of outcome variance could be built only when HR entropy was added to the HR variability measure in a second step of the regression analyses. HR variability alone was not found to be a good predictor of neither selfreported nor behavioral treatment outcomes. # 2007 Elsevier B.V. All rights reserved. Keywords: Heart rate variability; Sample entropy; Complexity; Specific phobia; Fear of flying

Specific phobia is one of the most prevalent mental disorders and the most common anxiety disorder (Kessler et al., 2005). The most successful treatment for specific phobias is exposure in vivo (Marks, 1987). Effective virtual reality or computerassisted exposure procedures are also currently available for several phobias (Bornas et al., 2006a; Botella et al., 2005; Maltby et al., 2002; Mu¨hlberger et al., 2003). However, a number of patients with specific phobia do not respond to exposure-based treatments. This raises the question of whether some pre-treatment variables are associated with a different prognosis. As Steketee and Chambless (1992) point out, ‘‘If we can identify characteristics of clients that are associated with poor response to treatment, we may (. . .) be able to match clients to treatments that work best for those with their particular characteristics.’’ (p. 387). Research on treatment

* Corresponding author. E-mail address: [email protected] (X. Bornas).

outcome predictors is an appropriate way to identify such characteristics but this kind of research in anxiety disorders, including specific phobia, has been disappointing so far. The use of demographic or clinical variables as outcome predictors ¨ st, has yielded very inconsistent results (see Hellstrom and O 1996; Fullana, 2000, for a review). Psychophysiological variables have been less studied, but with more promising results. According to Foa and Kozak’s model of emotional processing of fear (Foa and Kozak, 1986) heightened initial arousal, as indexed by increased heart rate (HR), is predictive of superior outcome (Foa and McNally, 1996; Lang et al., 1970; Watson and Marks, 1971). In a study with spider, injection or blood phobics who received exposure-based treatments, significant outcome predictors were found only for blood phobia, where diastolic blood pressure during a behavioral test ¨ st, predicted 16% of the outcome variance (Hellstrom and O 1996). In the study by Beckham et al. (1990) on flight phobia, participants who attended the 2-month follow-up session were divided into two groups: those who had flown and those who

0301-0511/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.biopsycho.2007.07.007

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007

+ Models

BIOPSY-5897; No of Pages 8

2

X. Bornas et al. / Biological Psychology xxx (2007) xxx–xxx

had not after the last treatment session, which consisted of an exposure flight. Then the authors compared the mean HR of both groups during that flight (HR was measured just before taking-off) and found that the mean HR of the group of participants who had flown was significantly higher than the mean of the group that had not flown. That is, increased physiological activation when confronted to the feared stimuli was related to positive treatment outcome. In recent years, a new psychophysiological measure has gained interest from researchers: heart rate variability (HRV). One of the most studied sources of HRV is the variability due to the inhibitory action of the parasympathetic or vagal system, i.e. the so-called vagally mediated HRV. This variability can be successfully measured in the frequency domain since it is widely accepted that the spectral power in the 0.15–0.4 Hz band (usually called the high frequency band) reflects exclusively or overwhelmingly the vagal influence on HR (Camm et al., 1996). Research on HRV has suggested a link between low vagally mediated HRV and anxiety disorders (see Friedman, 2007, for a review). Friedman and Thayer (1998) compared panic disorder patients, blood phobics, and normal controls using several measures of HRV, and found lower high frequency (0.18–0.35 Hz) power and shorter mean successive differences (MSD) in panic disorder patients than in blood phobics, and lower high frequency power and shorter MSD in blood phobics than in controls. Johnsen et al. (2003) found decreased root mean square successive differences (RMSSD) in a sample of dental phobics during exposure to feared stimuli (videoclips) as well as during a Stroop task performed after exposure and during a 5 min recovery period following the task. Regarding fear of flying, Wilhelm and Roth (1998) found decreased respiratory sinus arrhythmia – high frequency power measured during the first 128 s after taking off – in flight phobics. Bornas et al. (2005) found that low vagally mediated HRV fearful fliers reported higher levels of anxiety than high HRV fearful fliers when confronted with flight-related stimuli. These data suggest that having low vagally mediated HRV is less advantageous for health than having high HRV. Thayer and Lane (2000) have presented an interesting theoretical justification of this statement within a dynamic systems framework. Briefly stated, highly variable systems are more flexible and more able to adapt themselves (and their own behavior) to the demands of an ever-changing environment. Therefore, a relationship between vagally mediated HRV and therapy outcome could be expected. Patients with more variable heart rate would show greater adaptability to therapy and would have better treatment outcomes. However, to date, the relationship between vagally mediated HRV and treatment response has not been explored. Other features of heart rate deserve attention. From a dynamic systems perspective, the cardiovascular system can be characterized by a number of properties (e.g. complexity, nonlinearity, predictability, regularity, and so on) which can be measured on the HR time series derived from the ECG. Yeragani and co-workers have conducted most of the studies in this field since the early 1990s (Rao and Yeragani, 2001; Yeragani et al., 2000, 2002a,b), and several interesting links

between nonlinear properties of HR and anxiety disorders have begun to emerge. For example, it was found that panic disorder patients have increased complexity (assessed by the minimum embedding dimension, MED) and predictability (assessed by the largest Lyapunov exponent, LLE) than controls (Rao and Yeragani, 2001). Lower LLE has also been reported in patients with major depression (Yeragani et al., 2002a,b). However, several nonlinear methodologists recommend caution when using measures derived from the study of low dimensional chaos to investigate physiological signals (Heath, 2000; Kantz and Schreiber, 1997; Sprott, 2003) because biological systems are probably much more complex (i.e. they have higher dimensionality) than the mathematical ones. Partly as a solution to this problem, approximate entropy (ApEn) was introduced as a measure of regularity and complexity (Pincus, 1991, 1995) in relatively short and noisy time series typical from living systems. While variability refers to the degree of dispersion that successive values in a HR time series show around a central value (e.g. the mean), the concept of regularity refers to the time order of the values. Regularity can be as important as variability since the output from healthy systems is characterized by a greater irregularity (Goldberger et al., 2002; Guastello, 2004). Richman and Moorman (2000) introduced the sample entropy (SampEn) as a new and related complexity measure which ‘‘is largely independent of record length and displays relative consistency under circumstances where ApEn does not’’ (p. H2039). The SampEn index, like the ApEn index, is a nonnegative number assigned to a time series, with larger values corresponding to greater irregularity in the process, and smaller values corresponding to more recognizable patterns in the data. Hence, periodic data (e.g. a sine wave) should have an index closed to zero. Following the idea that complexity (like variability) is a sign of health in biological systems, and adding to a growing literature confirming the association between low entropy (complexity) and disease (Pincus, 2000; Vigo et al., 2004; Wagner and Persson, 1998; Wessel et al., 2000), Bornas et al. (2006b) found lower SampEn values in ECG (mV) time series of fearful fliers than in non-fearful controls in several threatening conditions but also during baseline and relaxing situations, thus revealing reduced complexity in the ECG output from fearful fliers. Further, a multiscale entropy analysis (MSE; Costa et al., 2002) – which uses the SampEn measure – of the ECG mV time series of a large group of flight phobics revealed a significant fear induced complexity loss (Bornas et al., 2006c). To sum up, a growing body of research provided support for the idea that enhanced complexity and predictability in the cardiac system could be associated to panic disorder and depression, though some of those studies measured properties that characterize low dimensional systems more than the high dimensional ones of living organisms. Entropy measures (e.g. ApEn and SampEn) should better estimate the complexity of such systems. Unfortunately, HR entropy has been much less investigated than HRV, but studies using entropy measures show that the output from healthy systems is more complex than the output from impaired or impoverished systems.

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007

+ Models

BIOPSY-5897; No of Pages 8

X. Bornas et al. / Biological Psychology xxx (2007) xxx–xxx

The aim of the present study was to evaluate the predictive power of vagally mediated HRV and HR SampEn on both the behavioral and the self-reported treatment outcomes of an exposure-based treatment which has shown to be effective in the treatment of flight phobia (i.e. the computer assisted fear of flying treatment, CAFFT; Bornas et al., 2001a,b, 2002). Our hypothesis was that high vagally mediated HRV as well as high HR SampEn would be related with better behavioral and selfreported treatment outcomes. We also compared the predictive power of these psychophysiological variables to the more widely used self-reported measures. The higher cost of the former ones can be justified only if they predict treatment outcome better than a low cost measure like a 30-item questionnaire. 1. Method 1.1. Participants Fifty-four flight phobics (35 women) with a mean age of 37.91 years (S.D. = 10.34) started treatment and 48 were completers. All of them were older than 18 and fulfilled DSM-IV (American Psychiatric Association, 1994) criteria for specific phobia (situational). The avoidance mean score was 6.31 (S.D. = 3.52) and the interference mean was 5.96 (S.D. = 1.18) on the anxiety disorders interview schedule for DSM-IV (ADIS-IV; Brown et al., 1994). They had avoided flying for a mean of 35.85 months (S.D. = 73.75). At 6-month follow-up 33 patients attended the assessment session. They did not differ significantly from the non-attenders in any of the pre-treatment measures.

1.2. Procedure Participants were recruited through advertisements in local newspapers. Seventy-two subjects asked for treatment and attended the first session at the Clinical Laboratory of the University. The ADIS-IV and the fear of flying questionnaire (FFQ; Bornas et al. (1999)) were used to assess all subjects. The FFQ is a 30-item questionnaire that includes a subscale about anxiety experienced before getting on the plane, i.e. previous situations subscale. It also includes a general discomfort 1–9 single item scale. Clinical exclusion criteria were: not fulfilling DSM-IV criteria for specific phobia, being in psychological treatment, taking psychotropic medication, suffering from any other mental disorder requiring immediate treatment, having a history of psychotic symptoms or current psychotic disorder, suffering from cardiovascular or respiratory illness, or being pregnant. Upon arrival to the laboratory, they were given general information about the study and signed a written consent form. Each subject was then seated approximately 1 m from a 17-in. monitor and sensors were attached for psychophysiological recording. A 3-min adaptation phase was followed by a 5-min resting baseline period, a 5-min paced breathing (0.2 Hz) task (PB), and a 5-min period of exposure to a threatening flying sequence (E). Physiological measures were taken while controlling for respiration (see Grossman et al., 1991; Tripathi, 2004) and under threatening situations, because both respiration and fear may influence the HR variability and complexity. The whole process was controlled by a computer (Pentium IV 1.6 GHz). Short text messages appeared on the screen before the beginning of each task. During baseline the subject was asked to look at a red cross in the computer’s screen and be relaxed. Five minutes later, the paced breathing task began. The breathing phase was paced by means of a picture of a human face with arrows pointing to the mouth when inspiring and going out from the mouth when expiring; a scrolling bar on the bottom of the screen also indicated the inspiration and expiration phases, which had the same duration. The message ‘‘now we need to evaluate your responses to some stimuli related to flying’’ appeared on the screen before the 5 min exposure phase. Subjects did not know that feared stimuli would be presented to them until this moment. A modified version of the take-off sequence of the CAFFT was used as the threatening stimulus. We chose this

3

sequence because take-off is usually the most feared moment among flight phobics (see Fullana, 2000, for a review). Based on the results of the clinical interview and self-report measures, 13 potential participants were excluded. Five additional subjects with anomalous ECG signals, mostly due to body movements, were also excluded. The remaining 54 patients were called a few days later to start treatment.

1.3. Treatment Treatment sessions were conducted individually in the same place where subjects had been assessed. During each session, patients were seated in front of a 17-in. screen and were instructed by the therapist to carefully look at the images and listen to the related sounds through the headphones. Each exposure session lasted about 1 h, although the therapist avoided ending the session if the patient was still experiencing too much fear in a certain exposure sequence. CAFFT exposure treatment lasted until the patient’s anxiety decreased to match the software requirements (see below) when exposed to any of the six exposure sequences. Patients were randomly assigned to one of two young therapists (one male and one female) with equivalent clinical experience, and who were proficient in the use of the CAFFT. Six patients did not finish treatment because they did not attend the scheduled therapy sessions. After the last treatment session, 48 patients completed the FFQ and were advised by the therapist to take a flight as soon as possible. All patients were called 6 months after they finished treatment and were assessed again with the initial protocol (except for the ADIS-IV). Thirty-three patients attended this follow-up session.

1.4. Stimuli and apparatus Sessions were conducted in a dimly lit and sound-attenuated room. ECG was recorded in a lead II configuration (a positive electrode on the left ankle, a negative electrode on the right wrist, and the ground electrode on the right ankle) using 10 mm Ag/AgCl electrodes. Participants were instructed to avoid arm movements during the experiment. The signal was recorded on a BIOPAC MP150 monitoring system and the sample rate was set to 200 Hz. The rationale for CAFFT is as follows: air travel can be conceptualized as a series of chronological events with critical moments. CAFFT, now in Version 2.2, divides air travel into five sequential stages: (1) preparation for travel, (2) preflight activities the day of the flight, (3) boarding the plane and take off, (4) in-flight condition, and (5) landing. Though most people with flight phobia usually experience anxiety during all stages of air travel, most patients experience idiosyncratic patterns of anxiety intensity throughout the flight experience. That is, they fear certain critical moments more than others. The CAFFT automatically configures the patient’s fear hierarchy based on his or her answers to the FFQ integrated into the program. Each item is associated with one of the exposure sequences. The CAFFT calculates the mean score for each stage of flight and then ranks the presentation of sequences from the one with the lowest score to the one with the highest one. Each stage of flight consists of a chronological series of photographs (at home, at the airport, walking onto the plane, etc.) shown on the screen of a personal computer with paired sounds taken in real settings. In addition to these five sequences, the CAFFT also includes a sixth sequence of pictures and matching audio stimuli related to aircraft accidents. This sixth sequence was included in the CAFFT since anxious apprehension about the airplane crashing is hypothesized to be a key component of many flight phobics’ fear (Howard et al., 1983; van Gerwen et al., 1997; Wilhelm and Roth, 1997). Exposure to this sequence does not seek to eliminate the instinctual and adaptive fear response to an actual plane crash, but to reduce the extreme anxiety that some flight phobic patients experience anticipating an accident or when they for example see a plane crash on TV. The sequences include about 15 pictures and last between 3 and 4 min each. After being presented with all the audiovisual stimuli in a sequence, the patient rates his or her anxiety on a 1–9-point Likert scale. The program repeats the sequence until the patient rates his/her anxiety as a 1 or 2. Once the patient’s anxiety has habituated, the program advances to the next sequence in the patient’s fear hierarchy. The patient completes therapy after he/she has habituated to all stages of flight. At this point the patient can choose to receive additional exposure to any of the six sequences (overexposure phase). At pretreatment assessment, after the FFQ has been filled out, the patient is asked if

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007

+ Models

BIOPSY-5897; No of Pages 8

4

X. Bornas et al. / Biological Psychology xxx (2007) xxx–xxx

flying at night and/or with a bad weather condition is associated with more fear. When this is the case, he/she is exposed to pictures and sounds reproducing these special conditions. CAFFT also includes a detailed on-line help utility in order to allow full self-application if required.

Table 1 Means and standard deviations of self-reported measures at pre-treatment (N = 54), post-treatment (N = 48) and follow-up (N = 33) and physiological measures at pre-treatment and follow-up

1.5. Data reduction and analysis After visual inspection of the ECG recordings to detect anomalous signals, five subjects were excluded from further analysis. An automatic R-wave detector was used to identify the interbeat intervals (IBI) in milliseconds. Instantaneous HR was obtained from the ECG recordings using the algorithm developed by Berger et al. (1986) with a sampling rate of 4 Hz. Only the first 1100 points were used in order to standardize the length of HR time series across subjects. The high frequency band (0.15–0.4 Hz) power (HF) was calculated with the fast Fourier transform method on these HR time series (N = 1100). The HRV analysis software (Version 1.1; Niskanen et al., 2004) developed by the Biosignal Analysis and Medical Imaging Group at the University of Kuopio was used to detect the high frequency spectral peak during baseline and paced breathing conditions to be sure that the baseline free breathing rate was not significantly different from the scheduled rate (0.2 Hz), and therefore this condition could be easily accomplished by patients. None of the patients were excluded because of extreme breathing rate. SampEn was calculated using the software available in Physionet (Goldberger et al., 2000). The length of each HR time series was 1100 points, r was set to 20% of the standard deviation and m was set to 2.

1.6. Statistics Following Steketee and Chambless (1992), residual gain scores were used instead of raw FFQ scores as outcome. Residual gain (RG) control for the initial between-patients differences and the measure error inherent in the use of repeated measures with the same assessment instrument. They are calculated with the formula RG = Z2 Z1r12, where Z2 is the final score, Z1 the initial score and r12 is their correlation (Steketee and Chambless, 1992, p. 394). To test the effectiveness of the treatment, self-reported scores were compared across the three assessments by overall one-way repeated-measures multivariate analysis of variance (MANOVA). Then univariate analyses of variance (ANOVA) were performed on each self-report measure. Alpha level was set to .05 and adjusted (Bonferroni) for multiple comparisons. Physiological measures with highly skewed distributions were ln transformed before any analysis. Hierarchical regression analyses were performed to study the predictive power of pre-treatment physiological measures on post-treatment and follow-up self-reported scores. The predictive role of these measures on the behavioral treatment outcome (flying or not flying at the end of treatment) was analyzed using logistic regression. The regression models included ln HF under PB condition and ln HF during condition E in the first step and HR SampEn in the same conditions were added in the second step. All the analyses were computed using SPSS 14.0 for Windows (SPSS Inc., 1989–2005).

2. Results 2.1. Treatment outcome Means and standard deviations of self-reported measures at pre-treatment, post-treatment and follow-up and HR, heart rate sample entropy, and high frequency band power measures at pre-treatment and follow-up are depicted in Table 1. Fear of flying self-reported scores were compared across three assessments by overall one-way repeated-measures MANOVA, revealing a significant main effect for time (Wilks’ l = .29, F(6, 126) = 17.67, p = .000, h2 = .46). Univariate F tests showed differences in each variable: FFQ total score F(2, 64) = 55.62, p = .000, h2 = .63; FFQ previous subscale

FFQ FFQ-prev General discomfort HR BL PB E HR SampEn BL PB E ln HF BL PB E

Pre-treatment

Post-treatment

6 m follow-up

M

S.D.

M

S.D.

M

S.D.

181.29 58.60 8.33

34.23 13.55 .97

98.85 28.60 4.83

35.42 13.30 2.01

102.58 29.09 4.48

56.57 19.45 2.30

78.79 81.98 79.27

11.53 11.47 10.84

76.22 79.44 78.34

9.04 7.44 8.50

.654 .615 .550 5.58 6.35 5.58

.140 .117 .125 .89 1.03 .82

– – –

– – –

– – –

– – –

– – –

– – –

.632 .614 .539 5.21 6.21 5.25

.131 .084 .121 1.12 1.05 1.06

BL = baseline; E = exposure; FFQ = fear of flying questionnaire; FFQprev = FFQ previous subscale; HF = high frequency band power; HR = heart rate; HR SampEn = HR sample entropy; PB = paced breathing.

F(2, 64) = 61.58, p = .000, h2 = .66; and general discomfort, F(2, 64) = 62.70, p = . 000, h2 = .66. Further, pair comparisons revealed that in all the cases the pre-treatment mean was significantly higher than any other, and there were no differences between post-treatment and follow-up means in any measure. On the other hand, 35 patients out of 48 (63%) flew after treatment. 2.2. Regression analysis To examine the predictive power of vagally mediated HRV and HR SampEn on self-reported fear of flying decrease, two hierarchical regression analyses were conducted using FFQ total and previous subscale RG scores at the end of treatment and at 6 months follow-up as dependent variables. One more hierarchical regression analysis was performed with the behavioral measure (take a flight or not) as the dependent variable. Pre-treatment ln HF power and HR SampEn, both in PB and E conditions, served as independent variables. ln HF at PB and E, as a set of variables, was entered first into the model, and HR SampEn at PB and E were then entered, also as a set of variables, in a second step as we were interested in knowing the additional amount of the variance that could be explained by the entropy variable. For self-reported dependent variables, HR SampEn explained the greatest percentage of variance, both at posttreatment (see Table 2) and at follow-up (see Table 3). In fact, ln HF did not show any significant predictive power over selfreported measures. It is noteworthy that the percentage of variance explained by psychophysiological measures is greater on the FFQ previous subscale than in FFQ total scores. Regarding behavioral outcome, logistic regression revealed a significant regression equation (x2(4) = 11.22, p = .024), in

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007

+ Models

BIOPSY-5897; No of Pages 8

X. Bornas et al. / Biological Psychology xxx (2007) xxx–xxx

5

Table 2 Summary of hierarchical regression analysis for variables predicting residual gain scores on FFQ and FFQ previous subscale at post-treatment (N = 48) FFQ B

FFQ previous subscale

a

Step 1 ln HF PB ln HF E Step 2 ln HF PB ln HF E HR SampEn PB HR SampEn E

S.E. B

a

b

a

Ba

S.E. Ba

ba

.30 .18

.21 .27

.33 .15

.16 .08

.22 .27

1.18 .07

.32 .31 1.00 2.29

.21 .26 1.46 1.19

.36 .27 .12 .31

.21 .23 .48 3.14

.21 .26 1.43 1.17

.24 .20 .06 .42

2

2

2

p = .010

2

Note: FFQ R = .055 for Step 1, n.s.; DR = .132 for Step 2 ( p = .043). FFQ previous subscale R = .018 for Step 1, n.s.; DR = .194 for Step 2 ( p = .010). HF PB = high frequency band power during paced breathing; HF E = high frequency band power during exposure; HR SampEn PB = heart rate sample entropy during paced breathing; HR SampEn E = heart rate sample entropy during exposure. a Variable.

which only one variable entered, HR SampEn measured during pre-treatment exposure (B = 8,15, S.E. = 3.47, Wald = 5.51, p = .019). The Hosmer–Lemshow goodness-of-fit test found no differences between observed and model-predicted values (x2(7) = 6.69 p = .462), showing and acceptable fit to the data for the model’s estimates. Therefore, increases in HR SampEn augment the probability of taking the post-treatment flight. To determine whether physiological measures would improve the treatment outcome (FFQ scores) prediction based on self-reported measures taken before treatment, the FFQ pretreatment scores were entered in a first step of a hierarchical regression analysis, ln HF measures entered in a second step, and HR SampEn measures entered in the third step. Only when HR SampEn measures entered in the analysis, a significant predictive regression model (see Table 4) explaining up to 18.9% of the FFQ scores at post-treatment could be built. Neither the FFQ pre-treatment scores nor the ln HF measures accounted for significant amounts of outcome variance. 3. Discussion In this study the computer assisted fear of flying treatment (Bornas et al., 2006a, 2002, 2001b) was applied to 54 flight phobics. The expected behavioral outcome was that patients would take a flight at the end of treatment. A reduction in

self-report measures of fear at the end of treatment and at 6 months follow-up was also expected. The aim of the study was to elucidate the predictive power of vagally mediated HRV and HR SampEn on both the behavioral and self-report treatment outcomes. Further, we were interested in knowing if these physiological measures improve the outcome prediction based on pre-treatment low cost self-reported measures. We had hypothesized a significant negative relation between the physiological measures and self-reported fear at the end of treatment. Hierarchical regression analyses on the RG scores lend partial support to this hypothesis. The ln HF power alone did not show significant predictive power. The predictive model included the entropy measures to explain up to 18.7% of the variance of the residual gain FFQ total scores and 21.2% of the variance of the residual gain FFQ previous subscale scores. Attending to the beta coefficients in this regression model, the highest ones correspond to the HR SampEn measured during exposure conditions (E), thus indicating that this variable plays the main role in the model. On the other hand, the negative signs of these coefficients mean that the higher the HR SampEn the less fear post-treatment. In other words, the model predicts that those phobics showing high HR SampEn during exposure to fearful stimuli at pre-treatment will report less fear at the end of treatment. At follow-up, we found a very similar pattern of predictive relations between the physiological measures and the FFQ RG

Table 3 Summary of hierarchical regression analysis for variables predicting residual gain scores on FFQ and FFQ previous subscale at follow-up (N = 33) FFQ B Step 1 ln HF PB ln HF E Step 2 ln HF PB ln HF E HR SampEn PB HR SampEn E 2

FFQ previous subscale

a

S.E. B

a

b

a

Ba

S.E. Ba

ba

.35 .53

.29 .39

.44 .49

.31 .32

.28 .38

.40 .30

.22 .54 2.05 3.12

.27 .38 1.60 1.15

.28 .49 .27 .50

.20 .37 2.40 2.96

.27 .37 1.59 1.14

.25 .35 .33 .49

2

p = .012 2

p = .015

2

Note: R = .061 for Step 1, n.s.; DR = .200 for Step 2 ( p = .039); FFQ previous subscale R = .043 for Step 1, n.s.; DR = .195 for Step 2 ( p = .046). HF PB = high frequency band power during paced breathing; HF E = high frequency band power during exposure; HR SampEn PB = heart rate sample entropy during paced breathing; HR SampEn E = heart rate sample entropy during exposure.

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007

+ Models

BIOPSY-5897; No of Pages 8

6

X. Bornas et al. / Biological Psychology xxx (2007) xxx–xxx

Table 4 Summary of hierarchical regression analysis for variables predicting FFQ total scores at post-treatment (N = 48) Variable Step 1 FFQ

B

S.E. B

b

.276

.147

.269

Step 2 FFQ ln HF PB ln HF E

.242 10.89 5.72

.150 7.62 9.68

.236 .325 .135

Step 3 FFQ ln HF PB ln HF E HR SampEn PB HR SampEn E

.267 11.63 10.73 33.73 81.33

.144 7.53 9.46 52.43 42.56

.260 .347 .254 .114 .296

Note: R2 = .072 for Step 1, n.s.; DR2 = .054 for Step 2, n.s.; DR2 = .120 for Step 3 ( p = .048). FFQ = fear of flying questionnaire pre-treatment score; HF PB = high frequency band power during paced breathing; HF E = high frequency band power during exposure; HR SampEn PB = heart rate sample entropy during paced breathing; HR SampEn E = heart rate sample entropy during exposure.

scores. A regression model including both the variability and the entropy measures is required to explain up to 26.1% of the variance of the residual gain FFQ total scores and 23.8% of the variance of the residual gain FFQ previous subscale scores. The variability measures alone (i.e. HF power in PB and E conditions) did not explain a significant percentage of the dependent variables variance. The highest values of the Beta coefficients correspond again to the HR SampEn measured during exposure conditions (E), and the sign of these coefficients is negative. In our opinion this is a remarkable result because the percentage of explained variance is rather high but also because we are assessing the predictive role of measures that were taken 6 months before. A difference between this and previous treatment trials on fear of flying is that the post-treatment flight was not part of the treatment, but only a recommendation by the therapist. Thus, patients had to fly without a therapist, pay for the flight tickets, and could fly wherever they wished. As expected, most of them flew after treatment, although 13 patients (a relatively high number of them) did not fly. Logistic regression analysis confirmed our hypothesis that patients with high HR SampEn would be more likely to fly after treatment. The last aim of this study was to see whether adding psychophysiological measures to the self-reported ones would result in a significantly better predictive model. Ideally, a low cost measure like a short questionnaire could be used to answer the question ‘‘Will this patient improve with this treatment?’’. If we could answer yes or no based on the patient’s responses to the questionnaire then recording his or her ECG and calculating the HR SampEn would be unnecessary. The results of our last hierarchical regression analysis, however, did not lend support to this idea. On the contrary, a predictive model could not be built with the pre-treatment FFQ scores alone. It should include the psychophysiological variables as predictors. Further, HR

SampEn seemed to be better predictor than vagally mediated HRV. Beyond clear conceptual differences between variability and regularity (complexity), these findings suggest that differences exist and are meaningful also at an applied, clinical level. As stated before, to date previous research on outcome predictors has been very inconsistent. The most powerful outcome predictor in specific phobias so far has been diastolic blood pressure for blood phobia. This is likely to be a predictor only for this very particular phobia, where the phobic response, unlike other phobias, is characterized by syncope or presyncope and where an underlying autonomic dysregulation has been identified (Accurso et al., 2001). The use of ‘‘new’’ psychophysiological variables, such as vagally mediated HRV and HR SampEn, and emerging theoretical frameworks such as dynamic systems theory, could provide a new impetus for research in the area of outcome predictors for emotional disorders. The next step would be to replicate the value of those variables as outcome predictors in other phobias and other anxiety disorders to find out whether they are as useful as they look in flight phobia. It would be also interesting to know whether the same variables are significant predictors for other treatments (e.g. pharmacological). Several limitations of the present study should be noted. First, although our sample size was bigger than some previous studies on outcome predictors in anxiety disorders, it was not big enough to conduct a cross-validation study, which some authors regard as the best procedure to study treatment ¨ st, 1996; Scho¨ling and Emmelkpredictors (Hellstrom and O amp, 1999). Second, ECG sampling rate was set to 200 Hz. This is a common setting in ECG studies, although we acknowledge that a greater effective resolution would be better for R-wave detection and HRV measures. Third, although significant, the amount of variability explained by HR SampEn was low. Other variables may have also a significant role in the prediction of treatment outcome. The exclusion of HR reactivity from our analysis could be seen as a limitation of the study because it made not possible to replicate previous research where increased HR in front of fear stimuli has been identified as a predictor of treatment outcome. However, it must be stressed that there was a phase of cued-breathing between the baseline and exposure, and this phase was critical in our design. Since our measure of vagally mediated HRV had to be taken during cued-breathing, we did not have a true measure of the anxiety elicited by the fear stimuli, i.e. the heart rate increase between the baseline and the exposure, since these were not consecutive. It would be interesting to test whether HRV and HR SampEn measures exceed simple HR reactivity in the prediction of treatment outcome but it would make no sense to introduce HF power during ‘‘free’’ breathing. Finally, the clinical utility of our findings is uncertain given the paradox that in clinical settings anxiety is evaluated exclusively by self-report, even when many anxiety symptoms have physiological origins and most logically should be measured physiologically (Wilhelm and Roth, 2001). To summarize, in this study we compared two heart rate measures, variability and entropy, and tried to evaluate their

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007

+ Models

BIOPSY-5897; No of Pages 8

X. Bornas et al. / Biological Psychology xxx (2007) xxx–xxx

significance as predictors of treatment outcome in flight phobia patients. We found that HR SampEn has a greater predictive power on treatment outcomes, behavioral or self-reported. Low HR SampEn flight phobics had a poorer response to treatment than high HR SampEn phobics. Despite some limitations, our results suggest that heart rate entropy may have a role in the prediction of outcome and further research in this area seems warranted. Acknowledgement This study was funded by Spanish Ministry of Science and Technology (Grants SEJ2005-00899/PSIC and SEJ200614301/PSIC). References Accurso, V., Winnicki, M., Shamsuzzaman, A.S., Wenzel, A., Johnson, A.K., Somers, V.K., 2001. Predisposition to vasovagal syncope in subjects with blood/injury phobia. Circulation 104, 903–907. American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders, 4th ed. American Psychiatric Association, Washington, DC. Beckham, J.C., Vrana, S.R., May, J.G., Gustafson, D.J., Smith, G.R., 1990. Emotional processing and fear measurement synchrony as indicators of treatment outcome in fear of flying. Journal of Behavior Therapy and Experimental Psychiatry 21, 153–162. Berger, R.D., Akselrod, S., Gordon, D., Cohen, R.J., 1986. An efficient algorithm for spectral analysis of heart rate variability. IEEE Transactions on Biomedical Engineering 33, 900–904. Bornas, X., Tortella-Feliu, M., Garcı´a de la Banda, G., Fullana, M.A., Llabre´s, J., 1999. Validacio´n factorial del cuestionario de miedo a volar (QPV) [The factor validity of the fear of flying questionnaire]. Ana´lisis y Modificacio´n de Conducta 25, 885–907. Bornas, X., Fullana, M.A., Tortella-Feliu, M., Llabre´s, J., Garcı´a de la Banda, G., 2001a. Computer-assisted therapy in the treatment of flight phobia: a case report. Cognitive and Behavioral Practice 8, 234–240. Bornas, X., Tortella-Feliu, M., Llabre´s, J., Fullana, M.A., 2001b. Computerassisted exposure treatment for flight phobia: a controlled study. Psychotherapy Research 11, 259–273. Bornas, X., Tortella-Feliu, M., Llabre´s, J., Barcelo´, F., Pauli, P., Mu¨lhberger, A., 2002. Clinical usefulness of a simulated exposure treatment for fear of flying. International Journal of Clinical and Health Psychology 2, 247–262. Bornas, X., Llabre´s, J., Noguera, M., Lo´pez, A.M., Barcelo´, F., Tortella-Feliu, M., et al., 2005. Looking at the heart of low and high heart rate variability fearful flyers: self-reported anxiety when confronting feared stimuli. Biological Psychology 70, 182–187. Bornas, X., Tortella-Feliu, M., Llabre´s, J., 2006a. Do all treatments work for flight phobia? Computer-assisted exposure versus a brief multicomponent nonexposure treatment. Psychotherapy Research 16, 41–50. Bornas, X., Llabre´s, J., Noguera, M., Lo´pez, A., 2006b. Sample Entropy of ECG time series of fearful flyers: preliminary results. Nonlinear Dynamics, Psychology, and Life Sciences 10, 301–338. Bornas, X., Llabre´s, J., Noguera, M., Lo´pez, A., Gelabert, J.M., Vila, I., 2006c. Fear induced complexity loss in the electrocardiogram of flight phobics: a multiscale entropy analysis. Biological Psychology 73, 272–279. Botella, C.M., Juan, M.C., Banos, R.M., Alcaniz, M., Guillen, V., Rey, B., 2005. Mixing realities? An application of augmented reality for the treatment of cockroach phobia. Cyberpsychology & Behavior 8, 162–171. Brown, T.A., DiNardo, P.A., Barlow, D.H., 1994. Anxiety Disorders Interview Schedule for DSM-IV (ADIS-IV) Adult Version. Graywind Publications Inc., Albany, NY. Camm, A.J., Bigger, J.T., Breithardt, G., Cerutti, S., Cohen, R.J., Coumel, P., et al., 1996. Heart rate variability. Standards of measurement, physiolo-

7

gical interpretation, and clinical use. European Heart Journal 17, 354–381. Costa, M., Goldberger, A.L., Peng, C.K., 2002. Multiscale entropy analysis of complex physiologic time series. Physical Review Letters 98 (6) pp. 068102-1–068102-4. Foa, E.B., Kozak, M.J., 1986. Emotional processing of fear: exposure to corrective information. Psychological Bulletin 99, 20–35. Foa, E.B., McNally, R.J., 1996. Mechanisms of change in exposure therapy. In: Rapee, R.M. (Ed.), Current Controversies in the Anxiety Disorders. Guilford Press, New York, pp. 229–343. Friedman, B.H., 2007. An autonomic flexibility–neurovisceral integration model of anxiety and cardiac vagal tone. Biological Psychology 74, 185–199. Friedman, B.H., Thayer, J.F., 1998. Autonomic balance revisited: panic anxiety and heart rate variability. Journal of Psychosomatic Research 44, 133–151. Fullana, M.A., 2000. Variables predictores del resultat terape`utic en el tractament de la por de volar amb el programa CAFFT [Outcome predictors of fear of flying treatment with CAFFT]. Unpublished PhD dissertation, Universitat Auto`noma de Barcelona. Goldberger, A.L., Amaral, L.A., Glass, L., Hausdorff, J.M., Ivanov, P.C., Mark, R.G., et al., 2000. PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation 101, E215–E220. Goldberger, A.L., Amaral, L.A., Hausdorff, J.M., Ivanov, P., Peng, C.K., Stanley, H.E., 2002. Fractal dynamics in physiology: alterations with disease and aging. Proceedings of the National Academy of Sciences USA 99 (Suppl. 1), 2466–2472. Grossman, P., Karemaker, J., Wieling, W., 1991. Prediction of tonic parasympathetic cardiac control using respiratory sinus arrhythmia: the need for respiratory control. Psychophysiology 28, 201–216. Guastello, S.J., 2004. Progress in applied nonlinear dynamics: welcome to NDPLS Volume 8. Nonlinear Dynamics, Psychology, and Life Sciences 8, 1–15. Heath, R.A., 2000. Nonlinear Dynamics. Techniques and Applications in Psychology. Lawrence Erlbaum Associates, Mahwah, NJ. ¨ st, L.-G., 1996. Prediction outcome in the treatment of specific Hellstrom, K., O phobia. A cross-validation study. Behaviour Research and Therapy 34, 403– 411. Howard, W.A., Murphy, S.M., Clarke, J.C., 1983. The nature and treatment of fear of flying: a controlled investigation. Behavior Therapy 14, 567–587. Johnsen, B.H., Thayer, J.F., Laberg, J.C., Wormnes, B., Raadal, M., Skaret, E., et al., 2003. Attentional and physiological characteristics of patients with dental anxiety. Journal of Anxiety Disorders 17, 75–87. Kantz, H., Schreiber, T., 1997. Nonlinear Time Series Analysis. Cambridge University Press, Cambridge, UK. Kessler, R.C., Chiu, W.T., Demler, O., Merikangas, K.R., Walters, E.E., 2005. Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry 62, 617–627. Lang, P.J., Melamed, B.G., Hart, J., 1970. A psychophysiological analysis of fear modification using an automated desensitization procedure. Journal of Abnormal Psychology 76, 220–234. Maltby, N., Kirsch, I., Mayers, M., Allen, G.J., 2002. Virtual reality exposure therapy for the treatment of fear of flying: a controlled investigation. Journal of Consulting and Clinical Psychology 70, 1112–1118. Marks, I.M., 1987. Fears, Phobias, and Rituals. Oxford University Press, Oxford. Mu¨hlberger, A., Wiedemann, G., Pauli, P., 2003. Efficacy of a one-session virtual reality exposure treatment for fear of flying. Psychotherapy Research 13, 323–336. Niskanen, J.P., Tarvainen, M.P., Ranta-Aho, P.O., Karjalainen, P.A., 2004. Software for advanced HRV analysis. Computer Methods and Programs in Biomedicine 76, 73–81. Pincus, S., 1991. Approximate entropy as a measure of system complexity. Proceedings of the National Academy of Sciences USA 88, 2297–2301. Pincus, S., 1995. Approximate entropy (ApEn) as a complexity measure. Chaos 5, 110–117.

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007

+ Models

BIOPSY-5897; No of Pages 8

8

X. Bornas et al. / Biological Psychology xxx (2007) xxx–xxx

Pincus, S., 2000. Approximate entropy in cardiology. Herzschrittmachertherapie und Elektrophysiologie 11, 139–150. Rao, R.K.A., Yeragani, V.K., 2001. Decreased chaos and increased nonlinearity of heart rate time series in patients with panic disorder. Autonomic Neuroscience: Basic & Clinical 88, 99–108. Richman, J.S., Moorman, J.R., 2000. Physiological time-series analysis using approximate entropy and sample entropy. American Journal of Phisiology. Heart and Circulatory Physiology 278, H2039–H2049. Scho¨ling, A., Emmelkamp, P.M.G., 1999. Prediction of treatment outcome in social phobia: a cross-validation study. Behaviour Research and Therapy 37, 659–670. Sprott, J.C., 2003. Chaos and Time-series Analysis. Oxford University Press, Oxford. Steketee, G.S., Chambless, D.L., 1992. Methodological issues in the prediction of treatment outcome. Clinical Psychology Review 12, 387–400. Thayer, J.F., Lane, R.D., 2000. A model of neurovisceral integration in emotion regulation and dysregulation. Journal of Affective Disorders 61, 201–216. Tripathi, K.K., 2004. Respiration and heart rate variability: a review with special reference to its application in aerospace medicine. Indian Journal of Aerospace Medicine 48, 64–75. van Gerwen, L.J., Spinhoven, P., Diekstra, R.F., van Dyck, R., 1997. People who seek help for fear of flying: typology of flying phobics. Behavior Therapy 28, 237–251. Vigo, D.E., Siri, L.N., Ladro´n de Guevara, M.S., Martı´nez-Martı´nez, J.A., Fahrer, R.D., Cardinali, D.P., et al., 2004. Relation of depression to heart rate nonlinear dynamics in patients >60 years of age with recent unstable

angina pectoris or acute myocardial infarction. The American Journal of Cardiology 93, 756–760. Wagner, C.D., Persson, P.B., 1998. Chaos in the cardiovascular system: an update. Cardiovascular Research 40, 257–264. Watson, J.P., Marks, I.M., 1971. Relevant and irrelevant fear in flooding: a crossover study of phobic patients. Behavior Therapy 2, 275–293. Wessel, N., Voss, A., Malberg, H., Ziehmann, C., Voss, H.U., Schirdewan, A., et al., 2000. Nonlinear analysis of complex phenomena in cardiological data. Herzschrittmachertherapie und Elektrophysiologie 11, 159–173. Wilhelm, F.H., Roth, W.T., 1997. Clinical characteristics of flight phobia. Journal of Anxiety Disorders 11, 241–261. Wilhelm, F.H., Roth, W.T., 1998. Taking the laboratory to the skies: ambulatory assessment of self-report, autonomic, and respiratory response in phobic anxiety. Psychophysiology 35, 596–606. Wilhelm, F.H., Roth, W.T., 2001. The somatic symptom paradox in DSM-IV anxiety disorders: suggestions for a clinical focus in psychophysiology. Biological Psychology 57, 105–140. Yeragani, V.K., Pohl, R.B., Jampala, V.C., Balon, R., Ramesh, C., Srinivasan, K., 2000. Increased QT variability in patients with panic disorder and depression. Psychiatry Research 93, 225–235. Yeragani, V.K., Pesce, V., Jayaraman, A., Roose, S., 2002a. Major depression with ischemic heart desease: effects of paroxetine and nortriptyline on longterm heart rate variability measures. Biological Psychiatry 52, 418–429. Yeragani, V.K., Rao, K.A.R.K., Smitha, M.R., Pohl, R.B., Balon, R., Srinivasan, K., 2002b. Diminished chaos of heart rate time series in patients with major depression. Biological Psychiatry 51, 733–744.

Please cite this article in press as: Bornas, X., et al., Vagally mediated heart rate variability and heart rate entropy as predictors of treatment outcome in flight phobia, Biol. Psychol. (2007), doi:10.1016/j.biopsycho.2007.07.007