Measurement of panic disorder by a modified panic diary

DEPRESSION AND ANXIETY 6:133–139 (1997) Research Articles MEASUREMENT OF PANIC DISORDER BY A MODIFIED PANIC DIARY Edwin de Beurs, Ph.D.,* Dianne L. Chambless, Ph.D., and Alan J. Goldstein, Ph.D. The psychometric characteristics of panic diary measures were investigated in a sample of 37 patients suffering from panic disorder with agoraphobia. Following recommendations made in the recent consensus development conference on the assessment of panic disorder, daily ratings included not only the occurrence of panic attacks but also fear of panic, expectancy of panic, and expected aversiveness of panic. These new measures were reliable and, on the whole, demonstrated good divergent and convergent validity. Further, adding such measures increased the incremental validity of panic disorder assessment. Depression and Anxiety 6:133–139, 1997. © 1997 Wiley-Liss, Inc. Key words: panic attack; self-monitoring; fear of panic; panic expectancy P INTRODUCTION anic attacks are the most striking feature of panic disorder. Consequently, assessment of panic frequency is considered essential for research on the psychopathology and the treatment of panic disorder (see Shear and Maser, 1994). For example, investigators often describe the percentage of patients in a treatment trial who reported no panic during a posttreatment monitoring period (e.g., Cross-national collaborative panic study, 1992). Although reporting panic frequency is essential, there are two reasons it is not sufficient. First, panic is but one feature of the disorder (see, e.g., American Psychiatric Association, 1994). Other defining features include persistent concerns about having additional attacks and worry about the possible consequences of an attack, such as insanity or death. Therefore, merely measuring frequency of panic is inadequate for proper assessment of this disorder. Second, measuring panic frequency is a complicated endeavor. Attacks typically occur unpredictably, and their frequency may fluctuate considerably over time—hardly desirable properties from a psychometric perspective. The instability of panic is especially marked for patients who suffer agoraphobic avoidance, the majority of panic sufferers who seek treatment (McNally, 1994). To some degree agoraphobic patients are successful in avoiding panic. Such patients may have a very low rate of panic before treatment and report increased panic only when their avoidance pattern is challenged by in vivo exposure (Cox et al., 1991; Craske and Barlow, 1988; de Beurs et al., 1993). For these patients, panic frequency per se is particularly deficient as the sole measure of their panic symptoms. © 1997 WILEY-LISS, INC. Panic disorder experts who attended the recent consensus development conference on the assessment of panic disorder (organized by the National Institute of Health) concluded that it is essential for investigators to measure an additional characteristic of panic disorder, so-called anticipatory anxiety (Shear and Maser, 1994); that is, apprehension about having another panic attack. Conferees further suggested that this construct be broken down in two components: (1) expectancy of having a panic attack; and (2) thinking that an attack is a terrible thing (and therefore something to be anxious about). Theoretically, different treatment modalities might have differential effects on these two aspects of panic. For example, pharmacological treatments might lower expectancy of panic while leaving patients frightened of what might happen to them if they were to panic. On the other hand, cognitive–behavioral therapy might teach patients to view panic as less catastrophic, but have less impact on their expectancy of panic. To capture these concepts, new measures need to be added to ones now in use. University of North Carolina at Chapel Hill, Chapel Hill, North Carolina Contract grant sponsor: NIMH; Contract grant number: R21MH49851. *Correspondence to: Edwin de Beurs, Dept. of Psychiatry, Vrije Universiteit, Valeriusplein 9, 1075 BG Amsterdam, The Netherlands. Received for publication 17 December 1997; Accepted 12 January 1998 134 de Beurs et al. Consensus conference experts (Shear and Maser, 1994) concluded that panic diaries were essential for measurement of the frequency of panic and related concerns. In such daily diaries, patients are asked to monitor the occurrence of each attack along with additional information. Rapee and Barlow (1991) provide a comprehensive description of a representative diary. Panic assessment research findings suggest that continuous monitoring of panic yields more accurate and conservative data than retrospective estimates of panic frequency gleaned from self-report questionnaires or clinical interviews (de Beurs et al., 1992; Margraf et al., 1987; Rapee et al., 1990). Nonetheless, little has been done to examine the psychometric aspects of panic diaries. Such is the focus of the current study. In the context of a clinical trial evaluating treatments for panic disorder with agoraphobia, patients kept a panic and anxiety diary for 2 weeks prior to starting treatment. We expanded the usual panic monitoring with additional ratings of panic-related concerns recommended by consensus conference experts (Shear and Maser, 1994). The diary comprised daily and weekly ratings of panic expectancy, ratings on how bad it would be to experience a panic attack (expected aversiveness of the attack), a daily rating of the highest fear of panic during the day, and whether panic had actually occurred. The purposes of the study were: 1) to examine the reliability and validity of these self-monitoring measures; 2) to determine whether assessment of panic is improved by adding daily ratings of fear of panic; and 3) to determine whether, as suggested by the consensus conference experts, breaking down fear of panic into two components—expectancy of panic and expected aversiveness—increases the incremental validity of panic assessment. METHODS SUBJECTS The sample consisted of 37 patients (30 female, 7 male) suffering from panic disorder with at least moderate agoraphobic avoidance (PDA) and accepted for participation in a clinical trial for their disorder.1 All patients were diagnosed by a clinical psychology doctoral student trained in using the SCID for DSM-IV (Spitzer et al., 1989). Patients’ mean age was 39.9 (SD = 9.2); their mean duration of complaints was 13.3 years (SD = 8.4). The inclusion criterion was a primary diagnosis of PDA of at least 1 year’s duration. No minimum number of panic attacks per week was required. Excluded were applicants with a SCID-I diagnosis of present substance dependence, current major depressive episode, or present or past psychosis; or a SCID-II diagnosis of paranoid, schizoid, schizotypal, antisocial, or 1 Other data on the same sample of patients are included in a paper on the Beck Anxiety Inventory (de Beurs et al., 1997). borderline personality disorder. Other comorbid diagnoses were acceptable unless they caused severe impairment in functioning according to SCID ratings. Also excluded were patients who used >2 mg alprazolam a day or equivalent dosages of other benzodiazepines, and those not yet stablized on medication. Those on medication (43% of the sample) were required to maintain a stable dosage throughout their participation. MEASURES Panic diary. As previously described, the panic diaries were used for continuous self-monitoring of fullblown panic attacks. In addition, each morning the patient rated the chance of having a panic attack during the ensuing day according to an 11-point scale (0 = “I will have no panic attack”; 10 = “I will definitely have a panic attack”), as well as how bad it would be to have an attack that day (0 = “Not bad at all”; 10 = “Extremely bad”). These are the daily panic expectancy and expected aversiveness ratings, respectively. At the end of the day, patients noted the number of panic attacks experienced that day (daily panic frequency) and retrospectively rated the highest level of fear they had experienced that day about having a panic attack (0 = “No fear at all”; 10 = “Extreme fear”). Finally, at the end of each week of monitoring, the subjects rated the chance of having a panic attack during the next week and how bad that would be (weekly panic expectancy and expected aversiveness ratings). Self-report questionnaires included the Mobility Inventory (MI) (Chambless et al., 1985), the Brief Symptom Checklist (BSI) (Derogatis, 1975), the Agoraphobic Cognitions Questionnaire and the Body Sensations Questionnaire (ACQ and BSQ) (Chambless et al., 1984), the Beck Anxiety Inventory (BAI) (Beck and Steer, 1990), the Beck Depression Inventory (BDI) (Beck and Steer, 1987), and the Panic Appraisal Inventory (PAI) (Telch et al., 1989). With the exception of the PAI, all are widely used reliable and valid measures of their constructs. The PAI is a relatively new measure designed to assess various aspects of panic disorder, especially cognitive features: anticipation of panic in circumscribed situations, panic-related concerns, and self-efficacy in coping with panic attacks in agoraphobic situations. Telch et al. (1989) have reported good psychometric properties. Test–retest reliability ranged from r = .81– .89, and internal consistency ratings ranged from α = .88 to α = .94). Data reported by Feske and de Beurs provided support for the PAI’s concurrent validity (Feske and de Beurs, 1997). Interviewer’s ratings. The assessor rated the severity of the patient’s psychopathology following the Panic Disorder Severity Scale (PDSS) (Shear et al., 1998). Each item on this 7-item scale is scored on a 5-point scale, ranging from 0 to 4. Three items address panic specifically: panic frequency, distress during panic (panic severity), and anticipatory anxiety (worry about future panic attacks). The other four Research Article: Modified Panic Diary items refer to agoraphobic fear and avoidance, interoceptive fear and avoidance, and level of impairment in vocational and social functioning due to the disorder. Items can be used individually or can be taken together to form one composite scale of severity of the patient’s condition. Shear and colleagues have reported positive findings on the psychometric properties of the PDSS. The PDSS items were demonstrated to form one factor with moderate internal consistency (α = .65) and excellent inter-rater and test–retest reliability (intraclass correlation coefficients of .87 and .88, respectively). Convergent and divergent validity of the individual items and the composite score were also generally supported. PROCEDURE Written informed consent was obtained from all patients before starting the assessment procedure. After the SCID, patients meeting entry criteria were carefully instructed on the definition of panic (Shear and Maser, 1994). The interviewer emphasized the sudden onset and limited duration of panic to distinguish it from more chronic forms of high anxiety. In addition, he described the distinction between full-blown panic attacks and limited symptom episodes and instructed patients to record fullblown panic attacks only. Patients were asked to record every panic attack immediately after its occurrence, to indicate which panic symptoms they experienced on a checklist, and to monitor their panic expectancy and fear of panic each day for 2 weeks prior to starting treatment. They received monitoring materials and written instruction on the definition of panic, including a list of the 13 panic attack symptoms of the DSM-IV. The instructions also contained a detailed description of each variable to be recorded. Patients monitored panic for 2 weeks prior to starting treatment. After 1 week, the patient returned to the clinic for an assessment session during which the results of the first week of monitoring were reviewed and discussed, and any instructions the patient found confusing were reiterated and clarified. The battery of self-report measures was administered and, finally, the interviewer (a doctoral level clinical psychologist with extensive experience with PDA) rated the severity of panic disorder according to the Panic Disorder Severity Scale. 135 RESULTS DESCRIPTIVE DATA The present analyses focus on pretreatment diary data, collected during baseline. For most patients (n = 27, 73%), panic frequency ranged from 1–12 attacks during the 2-week period. However, as is common for agoraphobic patients, 9 patients (24%) reported no panic during the same period. The data of one outlier, who reported 35 attacks, were excluded. Daily scores for panic expectancy, expected aversiveness, and fear of panic were aggregated into weekly averages to induce the amount of data and increase reliability. Descriptive analyses for the remaining 36 patients revealed that, as is typical, the panic frequency data had a highly peaked (kurtosis = 7.16) and positively skewed distribution (skewness = 2.10), due to the large number of subjects who reported no panic attacks (31%) or only one panic attack (14%) in the first week. All other diary variables had normal distributions. After square root transformation, both skewness (0.14) and kurtosis (–0.58) of the panic frequency variable were acceptable for parametric analysis. The mean scores of the first 2 weeks of monitoring are presented in Table 1. Comparing the means of weeks 1 and 2 gives an indication of the stability of the measures. The correlations between week 1 and 2 data provide information on short-term test–retest reliability. During baseline, panic frequency was quite reliable but decreased to a small, but statistically significant degree. Daily aggregated ratings were both reliable and stable, whereas weekly predictions were reliable but increased significantly from week 1 to week 2. CONVERGENT AND DIVERGENT VALIDITY Relationships among diary measures. To examine whether the three new diary ratings measured the same construct and to gauge their relationship to panic frequency, a correlation matrix was constructed. Presented in Table 2 are the correlations of the aggregated daily scores (in 1-week blocks) and the weekly prediction scores, as well as the average value for daily correlations. Aggregated scores should increase the reliability, and therefore the validity coefficients, of single-item data such as these ratings. Indeed, this was the case. TABLE 1. Comparison of panic measures in week 1 and week 2 Week 1 Panic frequency Fear of panic Panic expectancy Expected aversiveness (transformed) (aggregated) (aggregated) (next week) (aggregated) (next week) Week 2 Mean SD Mean SD t(34) P Test– retest r 1.24 4.25 4.00 6.50 5.12 6.17 1.01 2.22 1.92 2.34 1.88 2.05 1.00 4.45 4.20 5.60 5.28 5.63 0.88 2.72 2.26 2.29 2.20 2.16 2.26 –0.72 –0.66 3.44 –1.42 2.02 .03 .48 .51 .01 .16 .05 .78 .77 .84 .81 .91 .77 136 de Beurs et al. TABLE 2. Correlation coefficients among panic measures Fear of panic Week 1: Panic frequency Fear of panic: Panic expect.: Week 2: Panic frequency Fear of panic: Panic expect.: Panic expectancy Expected aversiveness daily (r– ) aggregated daily (r– ) aggregated next week daily (r– ) aggregated next week .40 .44** .26 .59 .49** .65** .11 .46 .14 .17 daily aggregated daily aggregated next week .67** .53** .51 .39* .53** .38 .48** daily aggregated daily aggregated next week .37 .72 .60** .54** .22 .65 .79** .29 –.06 .74** .55 .54** .45** *P < .05; **P < .01. Panic frequency was moderately correlated with fear of panic and panic expectancy, but not with expected aversiveness, suggesting this last variable has a somewhat different character from the other two new ratings. The pattern observed in week 1 was replicated in week 2. At both week 1 and week 2, the new variables (expectancy, aversiveness, and fear of panic) were all strongly intercorrelated, but the correlations between aversiveness and expectancy were lower than the correlations of fear of panic with each of the other two variables. These differences, tested with procedures suggested by Steiger (1980) for comparing dependent correlations, were statistically significant (all T2(33) > 2.19; all P < .05) for all contrasts in weeks 1 and 2, with one exception. The difference between the aversiveness–expectancy correlation and the fear of panic–aversiveness correlation for week 1 was in the same direction as the others (.39 vs. .53, respectively), but not statistically reliable, T2(33) = 1.16, P > .05. Overall, these data again suggest that the aversiveness and expectancy ratings are not simply two ways of asking the same question. Table 2 also includes correlations for the patient’s predictions for the upcoming week. Once again, the correlations of panic frequency with expected aversiveness were minimal. Correlations of expectancy and aversiveness are roughly comparable to those of the aggregated daily ratings. Relationships between diary measures and questionnaire and interviewer measures. Analyses of convergent and divergent validity with questionnaire and interviewer measures relied on the diary data from week 1 of monitoring, the week that corresponded to the time interval specified as the reference point for completing the other-method measures. The Beck Depression Inventory (BDI) was included as a measure of divergent validity. Because demoralization is common among panic patients, some correlations between the BDI and panic measures were expected, but these should not be as high as the convergent validity coefficients. In Table 3, key convergent validity coefficients are presented in italics. Convergent validity of the fear of panic rating was assessed via correlations with the PAI Coping subscale inquiring how confident patients are of their ability to cope in agoraphobic situations or when they have a panic attack. Following Telch et al. (1989), we reasoned that patients should be fearful of panic to the degree that they believe they are unable to cope with it. As expected, patients who rated themselves lower in self-efficacy reported significantly higher fear of panic. Further, as predicted, higher fear of panic ratings were correlated with higher scores on the PAI scale on which patients are asked to indicate the likelihood of panic in various agoraphobic situations, and with higher scores on the PDSS anticipatory anxiety item, designed to assess worry about further occurrence of panic. Contrary to expectation, the fear of panic rating was not significantly correlated with the Body Sensations Questionnaire, which assesses fear of physical symptoms of panic attacks, although a modest and positive correlation (P < .07) was obtained. Divergent validity is supported by the comparatively lower and nonsignificant correlation with the BDI. To examine convergent validity of the expectancy of panic rating, we correlated it with two scales we predicted to be related: the PAI anticipatory panic scale and the PDSS anticipatory anxiety item. Data supporting our predictions were statistically significant only for patients’ predictions for the next week, although correlations for the daily aggregated ratings tended to significance, both P < .10. Divergent validity vis-à-vis the BDI was clear for both weekly and daily aggregated ratings. For convergent validity of the expected aversiveness rating, the diary measure was correlated with the PAI Research Article: Modified Panic Diary 137 TABLE 3. Correlations of continuous monitoring measures with self-report measures and PDSS itemsa Panic expectancy Measure MI panic frequency PDSS panic frequency PAI coping PAI anticip. panic PDSS anticip. anxiety BSQ PAI concerns PDSS panic distress BDI MI alone BAI BSI PDSS composite Panic frequency .74** .75** –.14 .16 .09 .16 –.18 .41* .21 –.11 .34* .40* .40* Fear of panic Expected aversiveness Aggregated Next week Aggregated Next week .56** .53** –.26 .30 .28 .17 –.04 .28 .05 .31 .38* .37* .48** .57** .65** –.24 .43** .34* .36* .04 .41* .12 .21 .54** .46** .60** .25 .07 –.24 .34* .18 .36* .43* –.03 .18 .33* .42* .48** .28 .21 .17 –.02 .36* .35* .45** .25 .20 .21 .22 .50** .50** .46** .58** .56** –.40* .38* .45** .31 .28 .26 .22 .33* .52** .47** .55** a MI, Mobility Inventory; PDSS, Panic Disorder Severity Scale; PAI, Panic Appraisal Inventory; BSQ, Body Sensation Questionnaire; BAI, Beck Anxiety Inventory; BDI, Beck Depression Inventory. Italicized correlations are predicted covergent validity coefficients. *P < .05; **P < .01. Concerns scale, on which the patient rates concerns about negative consequences of attacks. Additionally, we predicted that the aversiveness rating would be associated with higher scores on the BSQ, indicating greater fear of the bodily sensations of panic. With one exception, both predictions were supported for daily aggregated as well as weekly data. Contrary to prediction, we did not find that patients who thought having a panic attack would be a very bad thing were those who were rated as having more severe attacks on the PDSS. CONSTRUCT VALIDITY To further examine the construct validity of the panic diary measures, we explored the incremental validity of adding the extra ratings to panic frequency for the measurement of important aspects of PDA psychopathology, including agoraphobic avoidance, generalized anxiety, overall severity of panic disorder, and global psychological distress. Two questions were addressed: 1) Will adding ratings of fear of panic improve the assessment of the given problem? 2) Does dividing fear of panic into panic expectancy and expected aversiveness further improve assessment? Four hierarchical multiple regression analyses were performed to address these questions. First, we investigated whether fear of panic, panic expectancy, and expected aversiveness ratings predicted additional variance, beyond that accounted for by panic frequency, in scores on the Mobility Inventory (Avoidance Alone), the Beck Anxiety Inventory, the General Severity Index of the Brief Symptom Inventory, and the clinician-rated PDSS composite rating. These criterion variables represent a range from those very specific to this PDA population (avoidance) to a measure of general distress (the BSI). In each equation followed by fear of panic, expectancy of panic, and expected aversiveness (week 1 aggregated ratings for all variables). At each step, we tested whether the next variable to enter would contribute to the predictive power of the equation to a statistically significant degree. The simple correlations of predictors with the criterion variables are given in Table 3 to facilitate interpretation. Results of the hierarchical regression equations are given in Table 4. The first column gives the multiple R2; the second column depicts the F value for testing whether any increase in explained variance is statistically significant. In every equation one or two of the new diary measures added to the prediction of PDA symptomatology. TABLE 4. Hierarchical regression analyses for panic measures as predictors of psychopathologya Dependent variable Avoidance (MI) Predictor Frequency of panic + Fear of panic + Expectancy of panic + Expected aversiveness Anxiety (BAI) Frequency of panic + Fear of panic + Expectancy of panic + Expected aversiveness General psychopathology Frequency of panic (BSI) + Fear of panic + Expectancy of panic + Expected aversiveness Severity rating (PDSS Frequency of panic composite) + Fear of panic + Expectancy of panic + Expected aversiveness a R2 .01 .19 .23 .25 .12 .29 .29 .32 .16 .27 .27 .36 .16 .34 .35 .35 Finc 7.42* 1.71 0.69 7.98** 0.00 1.46 4.92* 0.00 4.37* 8.83* 0.45 0.00 MI, Mobility Inventory; BAI, Beck Anxiety Inventory; BSI, Brief Symptom Inventory; PDSS, Panic Disorder Severity Scale. *P < .05; **P < .01. 138 de Beurs et al. Fear of panic, forced into the equation first, consistently added significantly to explanation of the variance. With one exception (the BSI), expected frequency and aversiveness ratings did not add further to the equations. As seen in Table 3, this was not necessarily because expected panic and aversiveness failed to be related to the criterion variables. The zero-order correlations were consistently significant for the BAI and almost uniformly so for the PDSS composite; expectancy was as strongly related to avoidance as was fear of panic. Further, examination of the regression diagnostics indicated that, although the three new diary measures are intercorrelated, multicollinearity was not problematic. Hence, it is likely that expectancy and aversiveness ratings fail to add significantly to the prediction of PDA psychopathology because the predictive portion of their variance (captured with the zero-order correlations) is the same portion that covaries with fear of panic ratings, which had the advantage of being forced into the equations immediately after panic frequency. Accordingly, we conducted additional analyses to further test the possible advantages of decomposing fear of panic into panic expectancy and expected aversiveness while avoiding the bias in favor of the fear of panic rating inherent in the hierarchical regression approach. We compared the predictive utility of fear of panic ratings to that of panic expectancy and expected aversiveness ratings, used jointly, in accounting for variance in the residualized criterion variables (controlling for panic frequency). The size of the effect for expectancy and aversiveness did not consistently exceed that for fear of panic alone. Even on those occasions when using these two variables did account for more of the variance than fear of panic alone, the difference never approached significance; all T2 < 0.68, all P > .20. These results confirm the conclusion drawn from the hierarchical regressions: Breaking down fear of panic into expectancy of panic and expected aversiveness does not seem to improve the assessment of panic-related concerns. DISCUSSION We tested the reliability and validity of three new ratings for a daily panic diary—fear of panic, expectancy of panic, and expected aversiveness of panic—along with the familiar measure of frequency of self-monitored panic attacks. All new measures demonstrated good short-term reliability whether we used daily ratings aggregated over a 1-week interval, or (in the case of expectancy and expected aversiveness) once-weekly predictions of the patient’s likely experience in the following week. Panic frequency was also reliable over the 2-week pretreatment monitoring interval. Divergent validity for all measures, assessed against the Beck Depression Inventory, was also consistently good. Self-monitored panic frequency was highly con- sistent with reports of panic on a questionnaire and to the clinical interviewer. However, these convergent validity coefficients are likely to be somewhat inflated. The very fact that patients had been self-monitoring is likely to have colored their report of panic to the clinical interviewer and on the questionnaire, probably by increasing the accuracy of their report. Convergent validity of the new measures of panicrelated concerns was more uneven. The majority of (although not all) tests were positive for fear of panic and expected aversiveness ratings; but, where expectancy of panic was concerned, convergent validity was only obtained for weekly predictions, not for daily ratings. Incremental validity was uniformly demonstrated for the fear of panic rating. When fear of panic was divided into expected occurrence of panic and expected aversiveness of panic, these ratings, used jointly, were equivalent to the fear of panic rating in incremental validity. However, they did not exceed the contribution made by the 1-item fear of panic rating alone. Because we found no evidence that the expectancy of panic rating and the expected aversiveness rating added to the prediction of PDA symptomatology already accomplished via frequency of panic and fear of panic ratings, our findings do not suggest that splitting fear of panic into two ratings of expected aversiveness and expectancy of panic is necessary. Just as much can be accomplished with one more economical, less burdensome fear of panic rating. In addition, convergent validity findings were stronger for the fear of panic rating than the expectancy of panic rating, to which it was highly correlated, making the former more psychometrically desirable than the latter. However, a cautionary note is required. All of our data concerned pretest symptoms only. It remains possible that the two separate ratings for components of fear of panic might prove useful in research on the process of change in treatment, or might demonstrate differential treatment outcome for different approaches to treatment. For the answer to this question, data from pharmacological and cognitive–behavioral treatment trials are needed. Second, expectancy of panic is interesting in its own right. There is a large body of literature suggesting that an important feature of anxiety disorders in general and panic disorder in particular is the sufferers’ tendency to overpredict fear or panic (Rachman, 1994). Because most of this research has been carried out in the laboratory, generalization of laboratory findings to naturally occurring panics in the patients’ environment requires investigation. Incorporating panic expectancy ratings in the continuous self-monitoring records would provide the necessary data. Another economy of measurement could be effected by using weekly ratings for panic expectancy and expected aversiveness. On the whole, comparing the results obtained with daily vs. weekly measures indicates that the weekly measures are equally reliable and valid. Panic researchers are plagued by the difficulty in obtaining consistent self-monitoring records (Shear and Research Article: Modified Panic Diary Maser, 1994). Reducing demands on patients with weekly rather than daily ratings might foster cooperation with this onerous assessment approach. This point is worth exploring in future research. For now, we caution researchers not to throw out daily ratings on the basis of these initial data. The quality of the weekly estimates may have been increased by the very fact that patients kept daily records of panic and anxiety and were, therefore, more apt to quantify their complaints accurately. We conclude that, as advocated by the researchers attending the consensus conference on panic assessment (Shear and Maser, 1994), adding a rating of fear of panic to panic diaries is highly desirable. We suggest that our study be replicated in a sample of patients without extensive agoraphobic avoidance. In studies conducted with questionnaire measures tapping the fear of panic construct, panic disorder patients with agoraphobia have demonstrated higher fear of panic than less avoidant patients (e.g., Chambless and Gracely, 1989; Cox et al., 1991). Thus, the incremental validity of fear of panic may be especially high in an agoraphobic population. Acknowledgments. This research was supported by NIMH grant R21-MH 49851. We thank Maria Belecanech and Marnita Floyd for serving as SCID interviewers. REFERENCES American Psychiatric Association (1994) DSM (IVrd ed.). Washington, DC. Beck AT, Steer RA (1987) Manual for the Revised Beck Depression Inventory. San Antonio, TX: Psychological Corporation. Beck AT, Steer RA (1990) Manual for the Beck Anxiety Inventory. San Antonio, TX: Psychological Corporation. Chambless DL, Caputo GC, Bright P, Gallager R (1984) Assessment of fear of fear in agoraphobics: The Body Sensations Questionnaire and the Agoraphobic Cognitions Questionnaire. J Consult Clin Psychol 52:1090–1097. Chambless DL, Caputo GC, Jasin SE, Gracely EJ, Williams C (1985) The Mobility Inventory for agoraphobia. Behav Res Ther 23:35–44. Chambless DL, Gracely EJ (1989) Fear of fear in the anxiety disorders. Cogn Ther Res 13:9–20. Cox BJ, Swinson RP, Norton GR, Kuch K (1991) Anticipatory anxiety and avoidance in panic disorder with agoraphobia. Behav Res Ther 29:363–365. 139 Craske MG, Barlow DH (1988) A review of the relationship between panic and avoidance. Clin Psych Rev 8:667–685. Cross-National Collaborative Panic Study (1992) Drug treatment of panic disorder: Comparative efficacy of alprazolam, imipramine and placebo. Brit J Psychiatry 160:191–202. de Beurs E, Lange A, Koele P, Van Dyck R (1993) Frequency of panic as an outcome measure in agoraphobia research: Latent effects of exposure on panic. J Anxiety Disorders 7:307–319. de Beurs E, Lange A, Van Dyck R (1992) Self-monitoring of panic attacks and retrospective estimates of panic: Discordant findings. Behav Res Ther 30:411–413. de Beurs E, Wilson KA, Chambless DL, Goldstein AJ, Feske U (1997) Convergent and divergent validity of the Beck anxiety inventory for patients with panic disorder and agoraphobia. Depression and Anxiety 6:147–153. Derogatis LR (1975) The Brief Symptom Inventory. Baltimore, MD: Clinical Psychometric Research. Feske R, de Beurs E (1997) Psychometric properties of the Panic Appraisal Inventory (PAI). Behav Res Ther 35:875–882. Margraf J, Taylor CB, Ehlers A, Roth WT, Agras WS (1987) Panic attacks in the natural environment. J Nerv Ment Dis 175:558–565. McNally RJ (1994) Panic Disorder: A Critical Analysis. New York: Guilford. Rachman S (1994) The overprediction of fear: A review. Behav Res Ther 32:683–690. Rapee RM, Barlow DH (1991) The cognitive–behavioral treatment of panic attacks and agoraphobic avoidance. In Walker JR, Norton GR, Ross CA (eds): Panic Disorder and Agoraphobia: A Comprehensive Guide for the Practitioner. Pacific Grove, CA: Brooks/Cole, pp 251–305. Rapee RM, Craske MG, Barlow DH (1990) Subject-described features of panic attacks using self-monitoring. J Anxiety Disorders 4:171–181. Shear M, Brown T, Barlow D, Money M, Sholomkas D, Woods S, Gorman J, Papp L (1997) Multicenter Collaborative Panic Disorder Severity Scale. Am J Psychiatry 154:1571–1575. Shear MK, Maser JD (1994) Standardized assessment for panic disorder research: A conference report. Archives of General Psychiatry 51:346–354. Spitzer RL, Williams JBW, Gibbon M (1989) Structured Clinical Interview for the DSM-III-R. Outpatient Version. New York: Biometrics Research Department, New York State Psychiatric Institute. Steiger JH (1980) Test for comparing elements of a correlation matrix. Psychological Bulletin 87:245–251. Telch MJ, Brouillard M, Telch CF, Agras WS, Taylor CB (1989) Role of cognitive appraisal in panic-related avoidance. Behav Res Ther 27:373–383.