Somatic Symptoms and Physiologic Responses in Generalized Anxiety Disorderand Panic Disorder:  An Ambulatory Monitor Study FREE

Anxiety is a biological warning system that prepares us for action.Considering subjective and objective body reactions observed in nonanxiousindividuals during acute stress,^1- 4 itis reasonable to assume that patients with chronic anxiety disorders exhibitphysiologic hyperarousal at rest or heightened physiologic responses to stressors.This, however, is not uniformly the case. The most consistent finding in patientswith anxiety is increased muscle tension.¹ Autonomicchanges are found less consistently. Our group⁵ foundthat patients with generalized anxiety disorder (GAD) showed normal heartrate, skin conductance, and respiration values while at rest. During laboratorystress, patients with GAD actually had a lower skin conductance response thancontrols.⁵ In some other studies,^6- 8 patientswith GAD exhibited an increased heart rate and decreased cardiac vagal toneduring rest and during mental stress.

Similarly, in some studies, patients with panic disorder (PD) exhibitednormal heart rate, skin conductance, and respiration values while at rest,whereas other studies found increases in those physiologic functions.^1,9- 10 The most consistentlaboratory finding in patients with chronic anxiety was diminished physiologicflexibility (DPF), namely, a diminished range or variability of physiologicresponses to stressors found in most,^1,10- 13 butnot all,¹⁴ studies.

Moreover, most studies of patients with anxiety rely on self-reportsof somatic symptoms, despite the fact that somatic manifestations often correlatepoorly with physiologic states and reactions, and cannot be taken at facevalue.¹⁵

Most physiologic studies have been conducted in the laboratory, whichprovides a controlled environment and permits multiple simultaneous recordingsbut rarely reflects stressors experienced in everyday situations. Therefore,physiologic states and responses in laboratory studies may differ from thoseexperienced by patients in real life.

The aim of this study is to measure self-reports and physiologic statesin patients with GAD, patients with PD, and controls in their natural environmentusing an ambulatory monitor. We are interested in how these 3 groups differwhile they are not anxious, in their responses to stress or panic attacks,and in the relationship between experienced states and actual physiologicresponses. We are also interested in determining whether diminished autonomicflexibility occurs in natural settings and represents an altered physiologicresponse pattern in patients with chronic anxiety or whether it representsa laboratory artifact due to diminished attention to tasks that seem irrelevantto the anxious patient.

A total of 113 individuals, consisting of physically healthy patientswith GAD or PD and volunteers without a psychiatric disorder (controls), wererecruited by advertising and gave informed consent to participate in the study. Structured Clinical Interview for DSM-IV criteria¹⁶ were used to confirm the diagnoses, which, in caseof doubt, were reviewed by a second investigator. All other psychiatric disorders,including substance abuse, were excluded, except mild specific phobias thatdid not interfere with the participant's functioning. A physical examination,including a urine toxicology screen, was performed during the initial screening.All participants had to abstain from medications that affect the central andautonomic nervous systems for at least 2 weeks before entering and duringthe study. To be included in the study, patients with GAD and PD were requiredto score 38 or greater on the trait scale of the State-Trait Anxiety Inventory¹⁷ and 18 or more on the Hamilton Anxiety Rating Scale(HAM-A).¹⁸ Patients with PD had to have atleast 1 panic attack per week during the 4 weeks before testing. On entryinto the study, participants also completed the Beck Depression Inventory,¹⁹ the Sheehan Disability Scale,²⁰ theBarsky Amplification Scale,²¹ and the BodySensations Questionnaire.²² Participants reportedthat they engaged in moderate daily physical activities.

The study sample was predominantly female (76%), with a mean age of36 years (range, 19-55 years). Of the 90 participants who provided usabledata, 24 were controls, 40 were diagnosed as having GAD, and 26 were diagnosedas having PD. More patients with GAD were recruited because we were also interestedin patients with high vs low levels of cardiac symptoms.²³ Inthis article we compare the entire GAD group with the other 2 groups. Thecontrol group consisted of 17 whites, 3 African Americans, and 4 Asian Americans.The GAD group consisted of 36 whites, 2 African Americans, 1 Hispanic, and1 Asian American. The PD group consisted of 23 whites and 3 African Americans.As compensation, participants with an anxiety disorder were given free treatmentafter completion of the study, and controls were paid.

All participants attended a training session before actual assessment.They wore the ambulatory monitoring device while sitting in a reclining chair.For a half hour, participants were acclimated to the physiologic monitoringequipment and were trained to complete the rating forms. Subsequently, for4 weekdays, they wore the ambulatory monitor and rated themselves on variousscales for at least 6 hours between 9 AM and 5 PM whilepursuing their usual daily activities.

On study days the participants were required to provide informationon self-report assessment forms at 30-minute intervals throughout the day.The Daily Rating Form allowed participants to rate the severity of symptomscommonly experienced during heightened anxiety on a scale from 0 ("absolutelynone") to 100 ("as bad as it could possibly be"). Symptoms consisted of severityof "rapid heart beat or palpitations," "sweating," "difficulty breathing normally,""feeling tense," and "worrying." A 5-point scale, asking, "How nervous orupset are you now?" with the response choices of "not at all" to "extremely,"allowed participants to rate their level of tension or anxiety while theyrated the symptoms listed on the Daily Rating Form. The Physical ActivityLevel Form allowed participants to rate their level of activity at a particulartime as "low," "mild," "moderate," or "high."

Physiologic measures were monitored continuously throughout the testdays using an ambulatory monitoring device described by Thakor et al.²⁴ The monitor was placed in a carrying case that hadshoulder and waist straps. The key physiologic measures of heart interbeatinterval (IBI), skin conductance level, and respiration rate were obtained,along with the contextual variables of ambient temperature and activity. Ambienttemperature was measured so as not to attribute changes in temperature thataffect physiologic states, including skin conductance, to changes in an emotionalstate. Measures of physical activity levels provide an indicator of when autonomicchanges should be attributed to physical activity. Heart IBI and breathingwere measured using standard electrocardiographic electrodes attached to thesides of the chest. Skin conductance was measured using silver and silverchloride electrodes attached to 2 fingers of the nondominant hand. For skinconductance, a Unibase (Parke-Davis, New York, NY) and isotonic sodium chloridesolution preparation was used as the electrolyte,²⁵ andthe fingers were individually wrapped in self-adhesive gauze. Adhesive diskswere used to allow skin exposure to the electrodes of exactly 1 cm in diameter.Activity level and ambient temperature were measured by sensors located withinthe ambulatory monitoring device. Participants were expected to wear the monitorfor 8 hours each day, but technical problems caused data loss for many participantsin the latter part of the day, so a 6-hour period was established as the standardfor the study. The device automatically stored data in 6-minute epochs every30 minutes. At each recording epoch, the monitor emitted a beep to alert theperson who was wearing it that subjective ratings should be completed. Inaddition to this routine monitoring every 30 minutes, a button was availableon the device to allow participants to signal the occurrence of other importantevents. In this study, patients with GAD and controls were asked to pressthe button to indicate a "stressful occurrence." To avoid possible confusionbetween panic attacks and nonpanic anxiety, patients with PD were asked topress the button only to indicate a panic attack but not a stressful occurrence.Physiologic data for the 3 minutes before and the 3 minutes after the buttonpress were automatically stored. Pressing the button also produced an audiblebeep that served to cue the participants to complete the subjective data forms.

Heart IBI was measured 800 times per second; skin conductance, 4 timesper second; and respiration, 10 times per second. Activity level and ambienttemperature were stored once per second and once per minute, respectively.At the end of the recording time, the data were transferred to a personalcomputer for storage and statistical analysis. Data available for analysisincluded up to 16 epochs per participant session. Twelve 6-minute epochs wereavailable for the routine sampling at 30-minute intervals during the session.Additional 6-minute epochs were allocated for measuring the physiologic statussurrounding each button-press event reported by the participant.

Analyses of overall daily effects focused on data obtained on the firstday of the study. Independent analyses for the physiologic and subjectivevariables were undertaken using the statistical package BMDP-5V (StatisticalSoftware Inc, Los Angeles, Calif), as required for an unbalanced repeated-measuresmodel with structured covariance matrices.^26- 28 Eachanalysis predicted the dependent variable (D) as a function of group membership,measurement occasion (recording epoch), and the interaction between thesefactors. The general form of the model was as follows: D = Status + Time +[Status × Time]. The analysis assumed a first-order autoregressive within-subjectcovariance matrix. A maximum likelihood method was used to estimate parameters.Missing data were computed based on the estimated conditional mean of themissing response, given the values of the responses that were present. "Status"reflected the diagnostic category of the participant, whereas "time" reflectedthe average of the dependent variable during successive measurement occasionsthroughout the day. The first 12 measurement occasions of the daily sessionwere used in these analyses.

Participants in the 3 diagnostic groups who reported stress (or panicfor the PD group) were examined in more detail to determine whether differencesin response to stress were evident. Any stress period during the experimentalsessions was included in the analysis, so some participants were representedby more than 1 data point. The unbalanced repeated-measures approach usedfor the analysis, implemented through BMDP-5V, took this lack of independenceinto account. Diagnostic group was a between-group factor, whereas the repeated-measurefactor was time in reference to the report of stress (before, during, andafter the button press).

The relation between objective and subjective responses for the 3 diagnosticgroups was examined for measures of basal level (daily average, not includingbutton-press periods) and change due to stress (as indicated by the buttonpress). A hierarchical set regression approach²⁹ wasused. This approach investigated the relationship between the objective levelof physiologic activity and the subject's report of the subjective state usuallyassociated with that response. In these analyses, the subjective measure wasconsidered the dependent measure, and the physiologic and diagnostic variableswere regarded as predictors. Between-group differences were found in age,and this variable was used as a covariate in all analyses. The statisticalpackage BMDP-2R (Statistical Software Inc) was used to perform the analyses.

Variables were forced into the regression in sets in the following order:age, group membership/physiologic response, linear interactions with group,and quadratic interactions with group. Interaction terms were carried by productvariables as a function of their order of entry into the equation. The primarypurpose of these analyses was to evaluate whether the diagnostic variablemodified the nature of the relationship between physiologic level or physiologicchange (for stress response) and the individual's report of the associatedsubjective variable.

Demographic characteristics of the PD, GAD, and control groups are givenin Table 1. Race and sex distributionsdid not differ as a function of diagnostic category, but control subjectswere younger on average than those diagnosed as having GAD or PD. No interactionsbetween diagnostic groups and sex were found. Both anxiety groups rated higherthan controls on the total HAM-A, the cardiovascular symptoms item of theHAM-A, the state and trait scales of the State-Trait Anxiety Inventory, allitems of the Sheehan Disability Scale, the Barsky Amplification Scale, andthe Body Sensations Questionnaire. The PD group differed from the GAD groupon the total HAM-A, the cardiovascular symptoms item of the HAM-A, and theBody Sensations Questionnaire.

Table 2 summarizes the significanteffects during the first daily monitoring session. Responses on all of thesubjective measures showed significant differences between groups, as didthe physiologic measures of mean IBI, IBI variance, and skin conductance variance.In general, the analyses distinguished the participants with anxiety disordersfrom the controls. According to Mann-Whitney post hoc comparisons, both anxietygroups rated themselves higher than the controls on rapid heart beat, sweating,difficulty breathing, feeling tense, and worry; however, the anxiety groupsdid not differ from each other.

These findings were consistent with orthogonal contrasts, performedas part of the overall analysis comparing the physiologic measures of controlsubjects with those of patients with GAD and PD. The following results wereobtained: mean heart rate was lower (IBI was greater) in controls (z = 1.89; P = .06), but IBI variance (z = 2.74; P = .006) and skin conductancevariance (z = 2.87; P =.004) were greater in controls.

Participants who reported stress or panic during the study did not differin terms of age. However, the PD subgroup reporting panic attacks was predominantlyfemale (88% women), whereas the control subgroup reporting stress was predominantlymale (71% men). The GAD subgroup reporting stress was equally divided (50%women).

Table 3 gives means andstandard deviations for subjective responses that showed a significant differencebetween the last rating before the button press and during the button pressindicating stress or anxiety or, in patients with PD, a panic attack. Patientswith PD showed increased response at the time of button press compared withthe preceding recording for difficulty breathing, rapid heart rate, sweating,feeling tense, and worry. The GAD group showed an increase only for rapidheart rate and feeling tense. The control group showed no significant changes.None of the groups showed differences between ratings obtained before andafter the button press. When the groups were compared for difference in scoresbetween baseline and button press, the PD group differed from the controlgroup in rapid heart beat, sweating, difficulty breathing, and feeling tense,whereas the GAD group differed from the control group only in rapid heartbeat. The PD group differed from the GAD group in sweating, feeling tense,and worry, along with a tendency toward difficulty breathing. There were nostatistically significant differences in measures of skin conductance, heartrate, or respiration. In the control group, 75% of button presses were accompaniedby an increase in heart rate; in the GAD group, 71% of button presses wereaccompanied by an increase in heart rate; and in the PD group, 74% of buttonpresses were accompanied by an increase in heart rate.

In addition, an examination of button-press data from all 4 days ofthe experiment revealed between-group differences in the frequency of buttonpressing (Table 4). During theexperiment, 29% of controls (n = 7) indicated at least 1 stressful periodcompared with 42% of patients with GAD (n = 17) and 65% of patients with PD(n = 17) who experienced a full-blown panic attack at least 1 time duringthe study (χ² = 6.87; P = .03).

Relations between subjective and physiologic data were examined usinghierarchical regression analysis and data from the basal period for mean IBIand mean skin conductance level. Figure 1 and Figure 2 illustratehow the relation between basal objective and subjective measures differedas a function of diagnostic category. An overall negative linear relationshipwas identified between mean IBI (inversely related to heart rate) and perceptionof rapid heart rate (F_2,78 = 5.62; R² inc = 0.121; P = .006), but this overalleffect was more pronounced for controls than for individuals with PD or GAD(F_1,79 = 10.3; R² inc = 0.111; P = .001) (Figure 1).Thus, although there was a general trend in all participants to associatelower IBI values with a subjective report of a more rapid heart beat, thetrend was stronger in nonanxious control subjects. This finding suggests that,overall, the participants in the anxiety disorder groups were less "responsive,"as expressed by the slope of subjective vs physiologic relationships, to differencesin the IBI than were nonanxious controls. Again using data from the basalperiod, the relation between skin conductance level and perceived sweatingwas statistically significantly different among the diagnostic groups (Figure 2). The controls associated more accuratelythe 2 conditions (F_2,77 = 5.15; R² inc = 0.112; P = .008). No statisticallysignificant main effects or interactions were found in the analysis of therelationship between basal respiration and reported difficulty of breathing.Patients in each anxiety disorder group differed significantly more amongthemselves in the accuracy of estimation of bodily functions than did controls.

This study described and compared the subjective and physiologic responsesof patients with GAD, patients with PD, and controls who wore a speciallyconstructed ambulatory monitor during daily activities. For calculating subjectiveand physiologic states when not feeling tense or anxious, we used the datafrom the first day's recording because they were most complete. For calculatingchanges when stressed, we included button presses and the recording of thepreceding automatic half-hour recording from all 4 recording days. Becausea button press can occur at any time after the last 6-minute recording period,the time between the 2 recordings, without overlapping, could have been 27minutes or less.

Patients with GAD and PD rated themselves higher on psychic and somaticanxiety, on disability scales, and on sensitivity to body sensations thancontrols. However, patients with PD and GAD differed little from each otheron self-ratings, except on the Body Sensations Questionnaire, where patientswith PD rated higher than patients with GAD, indicating heightened concernwith bodily functions. This finding is consistent with the idea that patientswith PD interpret physical sensations as dangerous and patients with GAD interpretthem as anxiety but is at variance with findings from previous studies^30- 32 indicating that patientswith PD reported significantly more autonomic symptoms than patients withGAD. There are several possible explanations for these differences. Patientswith GAD and PD are not homogenous groups and may vary considerably in typeand severity of physical symptoms. Moreover, preoccupation with somatic symptomsmay not relate to the degree of physiologic change.¹⁵ Patientsalso may have different anchoring points of severity for symptoms,³³ and self-ratings may depend on the instructions givento the participant.³⁴

Patients with PD, patients with GAD, and controls showed little differencein their physiologic responses when not registering anxiety, except for atrend in patients with PD to have a faster heart rate throughout the day.This finding corresponds with results obtained in some, but not all, laboratorystudies. Differences in the severity of PD may contribute to differences inheart rate. For example, Charney et al³⁵ foundthat physiologic responses to yohimbine challenge in patients with PD correlatedwith the average number of panic attacks. Another possibility is that patientswith PD did less physical exercise than nonanxious subjects. Physical exerciseaffects the physiologic state of an individual.³⁶ However,our patients, by interview and by self-ratings while wearing the monitor,pursued normal daily activities that were comparable to those of the othergroups. If the groups had differed significantly in physical condition, onewould expect differences in their physiologic data, which was not the case.Our data suggest that patients with PD experienced not only heightened sensitivityto bodily sensations but slightly higher autonomic arousal levels than patientswith GAD and controls. Respiration rate did not differentiate the groups,but the monitor recorded only frequency, not volume. Several studies^37- 39 suggest respiratoryirregularity and higher tidal volume in patients with PD that our recordingdevice missed.

The most prominent physiologic finding of this study was the decreasedvariance in heart IBI and skin conductance throughout the day in both anxietygroups compared with controls. Our group found DPF in the laboratory in patientswith GAD,⁵ patients with PD,⁴⁰ andpatients with obsessive-compulsive disorder⁴¹;other laboratories confirmed these findings in patients with GAD⁸ andPD.^10,12 Other studies found DPFin patients with phobic anxiety,⁴² posttraumaticstress disorder,⁴³ depression,⁴⁴ premenstrualsyndrome,⁴⁵ and alcoholism⁴⁶ andin individuals with high neuroticism or social maladjustment.^3,47- 48 Furthermore,DPF manifests itself in anxious individuals as decreased catecholamine^3,47 and cortisol^49- 51 excretionand in electroencephalographic responses to challenges.^52- 53 Thus,diminished responsiveness to stressors is a nonspecific central and peripheralmanifestation that accompanies prolonged anxiety or stress. There are severalpossible explanations for this phenomenon. First, DPF does not represent a"ceiling effect," as proposed by some investigators,^53- 54 becausebaseline values of patients with chronic anxiety and their response to stressorsdo not often differ from those of nonanxious subjects. Constitutional factors,as seen in shy children,⁵⁵ may predispose individualsto DPF and anxiety disorders, but the developmental course and clinical implicationsof such possible effects are not well understood. Diminished physiologic flexibilitymay represent a partial but inadequate attempt by the body to adapt to thephysiologic changes induced by chronic anxiety.^56- 57 Apsychological explanation is also plausible. Anxiety, particularly worry,preoccupies anxious individuals with internal events and diminishes theirattention to stimuli that are unrelated to their pathologic condition.^48,58 Thayer and Lane⁵⁹ presenteda model in which diminished cardiac vagal tone, manifesting itself in diminishedheart beat variability, represents the peripheral manifestation of inadequatecentral inhibition of the autonomic system in anxious subjects; a high vagaltone is associated with greater behavioral flexibility. According to thismodel, the Central Autonomic Network, a functional unit that appears to supportgoal-directed behavior and adaptability, includes the anterior cingulate,the insular and ventromedial prefrontal cortices, the periaqueductal gray,and nuclei of the hypothalamus, the striatum, and the pontine regions. Itsprimary output system is mediated through the preganglionic sympathetic andparasympathetic neurons. The system interprets visceral, humoral, and environmentalinformation and coordinates autonomic, endocrine, and behavioral responsesto environmental challenges. Anxiety leads to inhibition of the parasympatheticsystem and to dominance of the sympathetic system, which manifests itselfin decreased responsivity of the cardiovascular system to rapid changes inenvironmental demands. Although their model may explain many physiologic responsepatterns, a change in cardiac vagal tone is not invariably associated withDPF. We did not find diminished vagal tone in patients with GAD (D.R.M. andR.H.-S., unpublished data, 2000), and other researchers have not found itin patients with PD⁶⁰ or in depression despitean increased heart rate.⁶¹ Using functionalmagnetic resonance imaging, our group⁶² foundthat patients with GAD exhibited strong BOLD responses in the prefrontal andlimbic regions to statements that described a personal worry and to neutralstatements. Reduction of anxiety with citalopram therapy led to weaker BOLDresponses to both but particularly to neutral statements. These findings^62- 63 suggest that during high anxiety,cerebral responses to stimuli become indiscriminate to the nature of the stimulus,leading to dysfunctional central processing of information. The indiscriminantresponses to stimuli may lead to limited modulation of physiologic reactivitywithout necessarily involving the vagal system. Further clarifications ofthe biological function of DPF and its long-term effects on health are needed.^59,64

At the time of button press, all 3 groups registered higher reportsof rapid heart beat, sweating, difficulties breathing, feeling of tension,and worry. These increases were strongest in patients with PD and weakestin controls. The greatest increase in self-ratings was in difficulty breathing,which increased 6-fold. In addition, patients with PD pressed the button significantlymore frequently (indicating a panic attack) than did patients with GAD andcontrols, although the latter groups were instructed to press the button wheneverthey felt tense or anxious. Thus, patients with PD experienced not only moresevere but also more frequent anxiety.

Reports of physiologic changes during panic attacks indicate that marked,mild, or no physiologic changes may accompany attacks. Heart rate changes,recorded spontaneously in the laboratory or by ambulatory monitors duringpanic attacks, range from no change to 38 beats per minute.^65- 71 In1 study,⁶⁹ heart rate did not change disproportionatelyduring 42% of recorded panic attacks. This demonstrates that panic attackscan occur without substantial cardiac changes. Similar to previous ambulatorymonitor studies,^68- 69,71 74%of our patients with PD had an increase in heart rate during the time of thebutton press. Thus, 26% of the panic attacks occurred without cardiac changes.However, the GAD and control groups had similar heart rate increases, whichranged from no change to 12 beats per minute. Thus, heart rate increase seemsto be associated with increased tension or anxiety irrespective of the presenceof panic attacks or anxiety disorder and despite prominent cardiac symptomsaccompanying panic attacks. However, without obtaining constant analog recordingswe may have missed brief changes in heart rhythm.

As reported by other researchers,^22,72 patientswith PD were more sensitive to body sensations on self-ratings. Despite greatersensitivity, patients with PD and GAD were less accurate than controls intheir perception of bodily states and showed substantially more variability.As demonstrated elsewhere,^15,73 patientsare accurate in the estimation of the direction of bodily changes but notin their estimation of the degree of change and differ in levels of attentionand expectations, which modify their perceptions. The discrepancy betweenself-reports of physical changes during panic attacks and the lack of suchchanges may be explained by a sensitization of the patients to body changesduring early panic attacks. Recurrent somatic experiences may than create"engrams" in the brain that can induce bodily sensations with minimally alteringperipheral physiologic function. Such centrally induced panic attacks havebeen demonstrated by Lenz et al⁷⁴ who found,during surgery for intractable pain in a patient with panic attacks, thatthe stimulation of an area in the thalamus evoked panic attacks without causingthe physical changes that the patient described.

Ambulatory monitors measure responses in real-life situations, and theirdata have greater external validity than those obtained in the laboratory.However, the number of parameters that can be recorded and the length of therecordings are limited by memory, battery power, and the potential for technicalfailures. Moreover, the type and severity of stressors is unpredictable. Verificationof panic attacks is easier in the laboratory than during ambulatory recording.Despite these shortcomings, we obtained adequate information on physiologicstates throughout a day and responses before and after indication of stress,anxiety, or panic. Further studies are needed to clarify the relationshipbetween reported somatic manifestations and physiologic responses in patientswho differ in diagnosis, severity of illness, and family history and to integratethe data into the framework of imaging studies.

Correspondence: Rudolf Hoehn-Saric, MD, Department of Psychiatry,The Johns Hopkins Medical Institutions, 600 N Wolfe St, Meyer Bldg, Room 113,Baltimore, MD 21287-7113 (rhoehn@mail.jhmi.edu).

Submitted for publication January 13, 2003; final revision receivedMarch 10, 2004; accepted March 16, 2004.

This study was supported by grant MH42579 from the National Instituteof Mental Health, Bethesda, Md.