Behavioural and Cognitive Psychotherapy, 1998, 26, Cambridge University Press. Printed in the United Kingdom

Transcription

1 Behavioural and Cognitive Psychotherapy, 1998, 26, Cambridge University Press. Printed in the United Kingdom PSYCHOLOGIST, PATIENT, AND GENERAL PRACTITIONER RATINGS OF OUTCOME OF PHARMACOLOGICAL AND PSYCHOLOGICAL TREATMENTS FOR PANIC DISORDER AND AGORAPHOBIA IN PRIMARY CARE Donald M. Sharp and Kevin G. Power University of Stirling and Tayside Area Clinical Psychology Department, U.K. Abstract. Recent discussion has emphasized the importance of assessments of clinical significance of outcome in treatment outcome studies. Several methods exist to assess clinical significance, one of which is the use of global ratings of outcome. This study reports the use of such measures in a treatment outcome study investigating pharmacological and psychological treatments for panic disorder and agoraphobia in primary care. Outcome is reported in terms of global ratings of symptom severity and of change in symptoms following treatment, completed by the psychologist, referring General Practitioners (GPs) and patients. Treatments involving cognitive behaviour therapy showed greater efficacy than those employing pharmacological treatments alone. Considerable agreement was demonstrated between psychologist and patient ratings. Such agreement was much less obvious for ratings completed by referring GPs. Possible reasons for this are advanced. Keywords: Panic disorder, outcome ratings, primary care. Introduction Panic disorder with or without agoraphobia is a prevalent condition whose sufferers make heavy demands on primary care treatment services (Simpson, Kazmierczak, Power, & Sharp, 1994). Effective treatment is therefore of considerable importance, particularly those treatments employed in the primary care setting, given that this is where the largest proportion of patients with panic disorder seek treatment (Katerndahl & Realini, 1995). Pharmacological (Ballenger, 1994), psychological (Chambless & Gillis, 1994), and combined treatments (Telch, 1988) have all been shown to produce statistically significant improvements in patients with panic disorder and agoraphobia. It is recognized, however, that whilst changes on target measures following treatment may achieve statistical significance, this does not mean that they are clinically significant (Jacobson & Reprint requests to D. M. Sharp, Anxiety and Stress Research Centre, Department of Psychology, University of Stirling, Stirling FK9 4LA, Scotland, U.K British Association for Behavioural and Cognitive Psychotherapies

2 14 D. M. Sharp and K. G. Power Ravenstorf, 1988; Jacobson & Truax, 1991). It is possible for pre-treatment to posttreatment changes on target measures to be robust enough to achieve conventional statistical significance without being large enough to represent a real clinical improvement for the patient. The true value of clinical treatments may therefore be better indicated by measures that assess their impact on the clinical condition of patients. Measures of clinical significance may also prove useful in allowing clinicians to differentiate between treatments in terms of the clinical significance of their impact. Attempts have been made to assess clinical significance of outcome in treatments with panic disorder and agoraphobia patients. The most common of these involves the definition of high end-state functioning. This method has been widely used by a number of research groups investigating both psychological treatments (Mavissakalian, 1986; Barlow, Craske, Cerny, & Klosko, 1989; Michelson et al., 1990; Cote, Gauthier, Laberge, Cormier, & Plamondon, 1994), and more unusually psychological versus pharmacological treatments (Clark et al., 1994). To be classified as achieving high endstate functioning, patients must show a post-treatment response of sufficient magnitude to satisfy established a priori criteria. These criteria will often include panic free status, and low ratings on specified rating scales. Unfortunately, criteria for high end-state functioning vary between research groups, making cross-study comparison difficult. Furthermore, whilst high end-state functioning criteria are undoubtedly an improvement over simple assessments of statistical significance, the methods chosen are rarely justified in terms of the ecological validity of the criteria chosen. Further investigation is required to establish that patients achieving high end-state functioning criteria do show genuine clinically significant improvement. There is a need therefore for a theoretically coherent, widely accepted method of assessing clinical significance that can be adopted as a standard, thus facilitating comparison between studies. An example of an attempt to construct theoretically based criteria of clinically significant change is found in the work of Neil Jacobson and colleagues (Jacobson, Follette, & Ravenstorf 1984; Jacobson & Ravenstorf, 1988; Jacobson & Truax, 1991). These methods rely on the reasonable assumption that patients enter an effective therapy as members of a dysfunctional population and leave it having moved towards the distribution of a functional population. Jacobson and Truax (1991) recommend three methods by which this move from dysfunctional to functional population can be operationalized. These methods all rely on the creation of cut-off scores, on chosen target measures, which represent a defined level above which clinically significant change can be assumed to have occurred. Patients completing treatment can therefore be regarded as clinically significantly improved if post-treatment scores on target measures: (a) fall outside the range of the dysfunctional population by at least two standard deviations from the pre-treatment mean of that population in the direction of functionality; or (b) fall within the range of the functional or normal population, defined as falling within two standard deviations of the mean of the normal population; and finally (c) place the post-treatment patient closer to the mean of the functional than dysfunctional population. This latter criterion is the least arbitrary of the three, relying on a calculation of probability of a given score falling within the functional as opposed to the dysfunctional populations. These criteria are increasingly becoming accepted as a standardized method of outcome. As the latter two methods (b and c) require data on normal populations that are

3 Panic outcome ratings in primary care 15 rarely available for the assessment measures used in clinical practice, method (a) is the most widely employed technique. Jacobson and Truax (1991) acknowledge, however, that their methods are not the only means by which clinical significance might be assessed, and that other methodologies might repay investigation. Their methods have also been criticized for using absolute cut-off points and thus failing to take account of magnitude of change (Hollon & Flick, 1988), and for failing to relate criteria of clinically significant change to the social validity or relevance of that change (Baer, 1988). A further problem with Jacobson s methods lies in their validity being entirely dependant upon the construct validity of the target measures chosen. The process of establishing a cut-off score on a given measure and assuming that scores above that cut-off score reflect a clinically significant change will be valid only to the extent that the chosen target measure accurately reflects the clinical condition being assessed. That is to say, the target measure must have sufficient construct validity. For studies on panic disorder and agoraphobia, the most widely used measuring scale common to most studies (Jacobson, Wilson, & Tupper, 1988), is the Fear Questionnaire (Marks & Mathews, 1979), particularly the agoraphobia sub-scale of this measure. Indeed, this measure has often been used as the only psychometrically standardized outcome measure. Assessments of clinically significant change using the Fear Questionnaire Agoraphobia Sub-scale as the target measure all therefore assume that avoidance behaviour is the most salient feature of panic disorder and agoraphobia, and therefore the most representative index of change following treatment. It has been suggested (Kellner & Uhlenhuth, 1991), however, that there is no consensus on the boundaries of the construct(s) and thus this focus on specific factors such as avoidance behaviours may be premature. Early discussions on the issue of clinical significance (Kazdin, 1977; Strupp & Hadley, 1977) suggested that clinical significance might also be assessed by obtaining global ratings of change and levels of distress disruption from patients, clinicians or other significant observers. This method has the appealing directness of asking those undergoing a treatment to rate its effect upon them and as such may have more face validity for patients than completing complex questionnaires enquiring about symptoms that may or may not be relevant to an individual patient s own condition. In the case of panic disorder and agoraphobia, a patient may find more face validity in general questions regarding their levels of distress or general improvement following treatment, than in rating whether or not they continue to avoid a range of specific situations or experience a range of highly specific physical sensations as would be required by more traditional anxiety or avoidance rating scales. Global ratings may also avoid the problems of construct validity of chosen target measures discussed earlier, given that the constructs that global ratings reflect are less specific; for example, global severity distress or global change or improvement. Global ratings have also been noted to have roughly equal sensitivity to clinical change as do questionnaire measures (Kellner & Uhlenhuth, 1991). Indeed, a recent consensus conference on standardizing assessments for panic disorder research (Shear & Maser, 1994) recommended that all treatment studies should employ global ratings of outcome completed by the patient, the treating clinician, and if possible, by an additional (independent) clinician. The Clinical Global scales (Guy, 1976) provide a suitable tool for this purpose. These 7-point Likert type

4 16 D. M. Sharp and K. G. Power scales rate global symptom severity, and improvement, and can be easily and quickly completed by patient, clinician or other rater. Whilst these scales have been used in some studies, this has been for patient or therapist ratings only (Shear et al., 1991; Black, Wesner, Bowers, & Gabel, 1993). One recent study (Power, Simpson, Swanson, & Wallace, 1990) of the efficacies of diazepam, placebo, and cognitive behaviour therapy used alone and in combination in the treatment of generalized anxiety disorder in primary care employed global symptom severity and clinical global improvements ratings from patients, psychologist therapist, and referring general practitioner (GP). The latter ratings by GPs are of interest firstly, as they are unique in this field, and secondly, as referring GPs have the constant duty of care for the patients they refer to secondary treatment services, and are responsible for their care on completion of treatment, their views are of considerable relevance. The present study reports the use of global ratings of outcome in a study investigating the relative and combined efficacies of fluvoxamine, placebo, and cognitive behaviour therapy in the treatment of panic disorder and agoraphobia in primary care (Sharp et al., 1996). Ratings on the Global Symptom Severity and Clinical Global Improvement scales were made by the clinical psychologist, patients and GPs. These global ratings were compared with more commonly employed questionnaire assessments, firstly the Hamilton Anxiety Scale (HAM-A) (Hamilton, 1959), a rating scale assessing a broad range of anxiety related symptoms, completed by the treating psychologist; and secondly, the Kellner and Sheffield Symptom Rating Test (SRT) (Kellner & Sheffield, 1973), which again assesses a broad range of anxiety related symptoms, this time completed by patients themselves. The third questionnaire measure used was the Agoraphobia sub-scale of the Fear Questionnaire (FQ-AG) (Marks & Mathews, 1979). This patient rated scale assesses agoraphobic avoidance behaviour only and is the assessment common to most previous treatment outcome studies on panic disorder and agoraphobia. The study therefore investigates the use of global ratings as an alternative to both Jacobson s clinically significant change, and high end-state functioning criteria. The design also permits the comparison of global ratings of symptom severity and clinical improvement with scores on questionnaires rating broader constructs of distress and anxiety (HAM-A, SRT), and the more specific construct of avoidance behaviour alone (FQ-AG). This is the first time global ratings have been completed by the psychologist, patient and referring GP in a study of treatments for panic disorder and agoraphobia, in primary care. Method The global measures investigated in the current study were employed as part of a wider assessment package in a larger study investigating the relative and combined efficacies of the selective serotonin re-uptake inhibitor fluvoxamine, placebo, and cognitive behaviour therapy in the treatment of panic disorder with or without agoraphobia (Sharp et al., 1996). Patients were randomized to one of five treatment groups, namely, fluvoxamine (FL), placebo (PL), fluvoxamine plus cognitive behaviour therapy (FLCCBT), placebo plus cognitive behaviour therapy (PLCCBT), and cognitive behaviour therapy alone (CBT). All patients were required to be free of concurrent psychotropic medication throughout the active treatment period and all patients were

5 Panic outcome ratings in primary care 17 seen in the primary care setting. Procedural detail relevant to the current investigation is given below, a more detailed description and discussion of the full study design is available elsewhere (Sharp et al., 1996). Subjects Patients were referred by general practitioners and were seen for all treatment sessions in their local GP health centre or surgery. Patients were seen following referral for entry assessment by a clinical psychologist. Patients entered the study if they provided written consent and conformed to the following entry requirements; panic disorder with or without agoraphobia to DSM-III-R (APA, 1987) criteria, score a minimum of 15 on the Hamilton Anxiety Scale (Hamilton, 1959) and a maximum of 20 on the Montgomery Asberg Depression rating scale (Montgomery & Asberg, 1970); symptoms lasting at least 3 months; no psychotropic medication in the 28 days prior to study entry and no non-study psychotropic medication throughout the study treatment period, age between 18 and 70 years inclusive. A total of NG193 patients were entered into the study. Three patients dropped out during the placebo wash-in week (Day 7 today 0), and ng41 patients dropped out during treatment. Thus a total of ng149 patients completed treatment. The number of drop-outs for each treatment group were FL, ng 7 (19.4%), PL, ng9 (24.3%), FLCCBT, ng9 (23.7%), PLCCBT, ng3 (8.3%), and CBT, ng13 (30.2%). Treatments All patients were seen to an identical schedule of contact for appointments of equivalent duration in an attempt to balance for therapist contact. Patients were seen for 9 appointments of between 30 and 60 minutes over 13 weeks with sessions on Day 7, Days 0, 7, 14, 28, 42, 56, 70 and 84. Patients receiving medication received one week of single blind placebo then 12 weeks of either fluvoxamine or placebo. Fluvoxamine was given in 50 mg tablets at a starting dose of 50 mg day, rising to 100 mg day after 7 days to 150 mg day after another 7 days, thereafter remaining at 150 mg day throughout the study period. Patients receiving placebo were given an identical number of tablets thus maintaining the double blind status. Patients receiving cognitive behaviour therapy followed an identical schedule of contact and were supplied with a written treatment manual. The cognitive behaviour therapy employed emphasized both gross exposure and behavioural and cognitive panic management techniques. The approach to treatment was similar to that of Barlow and co-workers (Barlow, 1988) in that it emphasized the alteration of the action tendencies associated with the patient s panic attacks and also the hypervigilant and avoidant information processing and behaviours typical of panic disorder and agoraphobia. Treatment emphasized the importance of patients confronting their panic attacks, within the context of exposure outings if necessary, and attempting to replace behavioural and cognitive avoidances with more approach centred actions. Patients receiving either fluvoxamine or placebo plus cognitive behaviour therapy received medication to the identical protocol and cognitive therapy to the identical protocol to those given above.

6 18 D. M. Sharp and K. G. Power Measures This report will concentrate on patient, psychologist and GP global ratings of outcome and will compare these with both the psychologist s and patient s self ratings of anxiety and avoidance behaviour, pre- and post-treatment. Further measures employed in the study are reported elsewhere (Sharp et al., 1996). The main global measures of outcome were the Global Symptom Severity, and Clinical Improvements scales (Guy, 1976). These are two simple 7-point rating scales. The Global Symptom Severity scale assesses severity of symptoms in response to the question Considering your total clinical experience with this condition, how emotionally distressed is the patient now?. Responses were rated on a 7-point scale ranging from 1 normal, through 2 borderline, 3 mild, 4 moderate, 5 marked, 6 severe, to 7 extreme. This scale was completed by the psychologist and the referring GP at entry (Day 7) and end point (Day 84). The Clinical Global Improvement scale rates change in symptoms following treatment in response to the question Compared to his her condition on entry to the study, how much has the patient changed?. Responses were rated on a 7-point scale from 1 very much improved, through 2 much improved, 3 minimally improved, 4 no change, 5 minimally worse, 6 much worse, to 7 very much worse. This scale was completed by the psychologist, referring GP, and patient (with an appropriately altered wording of the question), and at end point (Day 84). Measures of anxiety used were the psychologist rated Hamilton Anxiety Scale, HAM-A (Hamilton, 1959) and the patient rated Symptom Rating Test, SRT (Kellner & Sheffield, 1973). Agoraphobia avoidance was measured using the Agoraphobia subscale of the Fear Questionnaire, FQ-AG (Marks & Mathews, 1979). Results are given for entry and end point only on these measures. Statistical analysis Results Repeated measures analysis of variance with a between subjects factor, treatment group and a within subjects factor assessment point were conducted, with simple effects oneway analyses of variance to investigate between group differences. Post hoc Scheffe tests were employed to detect significance of between group differences. Within group comparisons of pre and post scores were conducted using paired two-tailed t-tests. There were no statistically significant differences between treatment groups on any of the demographic or clinical measures at entry, meaning that groups were equivalent prior to treatment. Table 1 gives the means and standard deviations for the HAM-A, SRT and FQ-AG pre and post-treatment. Analysis of variance for HAM-A scores revealed significant group (F(4, 144)G6.45, pf.0001), time (F(1, 144)G625.73, pf.0001), and interaction (groupbtime), (F(1, 4)G8.44, pf.0001) effects, indicating differential improvements between groups over treatment. Within group comparisons of HAM-A scores revealed a significant reduction on scores for all groups FL (pf.0001), PL (pf.001), FLCCBT ( pf.0001), PLCCBT ( pf.0002) and CBT ( pf.0001). Between groups comparison of

7 Panic outcome ratings in primary care 19 Table 1. Means (SD) for HAM-A, SRT, and FQ-AG for each group pre- and post-treatment FL PL FLCCBT PLCCBT CBT ng29 ng28 ng29 ng33 ng30 F HAM-A Pre n.s. (3.7) (4.5) (3.6) (4.2) (3.6) Post ** (7.0) (8.9) (5.7) (6.7) (4.4) tg 10.19*** 4.09** 15.28*** 14.5*** 13.88*** SRT Pre n.s. (17.2) (18.2) (15.5) (14.3) (16.6) Post ** (17.1) (23.1) (16.3) (15.2) (12.2) tg 4.96** 2.61* 8.41*** 8.17*** 6.21*** FG-AQ Pre n.s. (10.0) (11.4) (11.5) (14.9) (11.2) Post *** (7.9) (13.2) (7.8) (10.6) (5.4) tg 3.44** n.s. 7.24*** 5.9*** 4.82*** *GpF.01, **GpF.001, ***GpF end point scores was significant ( pf.0001) with post hoc Scheffe tests indicating significant difference from the PL group for FL ( pf.05), PLCCBT ( pf.01) and FLCCBT and CBT (both pf.001). Analysis of variance for SRT scores revealed significant group (F(4, 144)G3.27, pf.01), time (F(1, 144)G172.77, pf.0001), and interaction (groupbtime) (F(1, 4)G 3.56, pf.01) effects, indicating differential improvements between groups over treatment. Within groups comparison of SRT scores revealed a significant reduction in scores for all groups, FL ( pf.001), PL ( pf.01) FLCCBT ( pf.0001), PLCCBT ( pf.0001), and CBT ( pf.0001). Between groups comparison of end point scores was significant ( pf.001), with post hoc Scheffe tests indicating a significant difference from PL for, PLCCBT (pf.05), and FLCCBT and CBT, (both pf.01), but not for FL. Analysis of variance for FQ-AG scores revealed significant time (F(1, 144)G95.57, pf.0001), and interaction (groupbtime) (F(1, 4)G8.12, pf.0001) effects, but a nonsignificant group effect. Within groups comparison of FQ-AG scores revealed a significant reduction in scores for all groups, FL ( pf.01), FLCCBT ( pf.0001), PLCCBT (pf.0001), and CBT (pf.0001), with the exception of PL (n.s.). Between groups comparisons of end point scores was significant ( pf.0001) with post hoc Scheffe tests indicating a significant difference from PL for, FL, FLCCBT, and CBT, (all pf.05), but not for PLCCBT. On these measures treatment can therefore be seen to be effective for all treatment groups to a greater or lesser extent. The smallest pre- to post-treatment gains were shown by the PL group, with those on the FQ-AG failing to reach significance. The largest and most consistent gains in treatment were shown by the CBT and FLCCBT

8 20 D. M. Sharp and K. G. Power Table 2. Means (SD) for GP and psychologist ratings of Global Symptom Severity for each group pre- and post-treatment FL PL FLCCBT PLCCBT CBT ng29 ng28 ng29 ng33 ng30 F GP Global Symptom Severity Pre n.s. (0.72) (0.84) (0.78) (0.84) (0.90) Post n.s. (3.47) (2.99) (3.61) (3.17) (3.37) tg n.s. n.s. n.s. n.s. n.s. Psychologist Global Symptom Severity Pre n.s. (0.77) (0.76) (0.65) (0.79) (0.87) Post *** (1.22) (1.48) (1.06) (1.33) (0.95) tg 10.13*** 2.64* 11.07*** 11.03*** 10.34*** * pf.01, ** pf.001, *** pf groups, with patients receiving these treatments showing the largest gains pre to post on all three measures and with these gains being consistently greater than those of the placebo (PL) group. It is with this pattern of efficacy that the global ratings can be compared. Global Symptom Severity Means and standard deviations for psychologist ratings of Global Symptom Severity pre- and post-treatment are given in Table 2. For psychologist therapist ratings of Global Symptom Severity, analysis of variance revealed significant group (F(4, 144)G 5.16, pf.001), time (F(1, 144)G389.91, pf.0001) and interaction (F(1, 4)G10.98, pf.0001) effects indicating differential changes between groups. There were no significant differences between groups for psychologist Global Symptom Severity ratings pretreatment, with all groups showing a mean of 4 or higher, indicating that the mean severity of symptoms pre-treatment was around the moderate level according to the psychologist s rating. Within group comparisons of pre- vs. post-treatment psychologist rated Global Symptom Severity indicated a significant reduction for all groups, FL ( pf.0001); PL ( pf.01); FLCCBT ( pf.0001); PLCCBT ( pf.0001) and CBT ( ph.0001); thus all five groups had shown improvements during treatment on this measure. Significant differences existed between groups post-treatment, however, ( pf.0001), with significant differences from PL shown by FL ( pf.05), and by FLCCBT, PLCCBT and CBT (all pf.001). An inspection of means indicates that the psychologist s mean ratings for each group on symptom severity post-treatment were, PL mild, FL and PLCCBT borderline, with only the FLCCBT and CBT groups achieving the normal rating level.

9 Panic outcome ratings in primary care 21 Table 3. Means (SD) for GP, psychologist, and patient ratings of Clinical Global Improvement for each group post-treatment FL PL FLCCBT PLCCBT CBT ng29 ng28 ng29 ng33 ng30 F GP Clinical Global n.s. Improvement (3.52) (3.35) (3.71) (3.39) (3.61) Psychologist Clinical ** Global Improvement (1.71) (1.30) (0.68) (1.07) (0.72) Patient Clinical ** Global Improvement (2.08) (1.80) (0.85) (1.57) (0.68) * pf.01, ** pf.001. In contrast to the above, with the GP Global Symptom Severity ratings, analysis of variance indicated no significant differences between groups on GP ratings both before and after treatment. Also within group comparisons of pre- vs. post-treatment GP ratings showed no significant difference between entry and end point scores for any group. Mean ratings for all groups remained at the moderate level pre- and posttreatment with the exception of the PL group where mean rating post-treatment moved to the marked level. Clinical Global Improvement Means and standard deviations for psychologist, GP and patient rating of change in treatment are given in Table 3. Between groups analysis of psychologist Clinical Global Improvement ratings showed significant differences between groups ( pf.001), with differences from PL from PLCCBT and CBT (both pf.01), and FLCCBT ( pf.001), but not for FL. Patients Clinical Global Improvement ratings showed a similar pattern of significance (pf.001), with differences from PL for PLCCBT and CBT (both pf.01), and for FLCCBT ( pf.001), but not for FL. Mean Clinical Global Improvement ratings for psychologist were also similar with the mean ratings for the FL and PL groups placing them in the much improved range, whilst those for FLCCBT, PLCCBT and CBT fell in the very much improved range. Thus treatments involving cognitive behaviour therapy showed a greater improvement, the difference being greatest for the FLCCBT group. Ratings for GP Clinical Global Improvement again failed to reveal significant differences between groups, with all groups falling in either the no change or minimally improved range. It is clear from the above results that the GP s ratings differed from those of both the psychologist and patient, which in turn appeared to be in broad agreement. The strength of these associations was investigated using correlations. GP s ratings of Global Symptom Severity showed a weak but significant relationship with the psychologist s rating of Global Symptom Severity using both pre-treatment (Pearson rg.27, pf.01), and post-treatment (Pearson rg.26, pf.01) scores. There was

10 22 D. M. Sharp and K. G. Power Table 4. Correlations between post-treatment scores on HAM-A, SRT, FQ-AG, GP and psychologist Global Symptom Severity, and GP, psychologist and patient Clinical Global Improvement HAM-A SRT FQ-AG GP Global Symptom Severity 0.33** 0.35** 0.18* GP Clinical Global Improvement 0.31** 0.32** 0.14 n.s. Psychologist Global Symptom Severity 0.95** 0.85** 0.70** Psychologist Clinical Global Improvement 0.80** 0.72** 0.56** Patient Clinical Global Improvement 0.76** 0.70** 0.51** * pf.05, ** pf.01. therefore little agreement between referring GPs and the psychologist as to global severity of symptoms either before or after treatment. When ratings for Clinical Global Improvement were compared GP ratings showed a weak but significant correlation with the psychologist s ratings (Pearson rg.21, ph.01), and with the patient s ratings (Pearson rg.18, pf.05). Ratings of psychologist and patient were, however, strongly correlated (Pearson rg.89, pf.01), indicating a good agreement between psychologist and patient as to the degree of improvement following treatment. This is in contrast to the lack of agreement with GP ratings. Results presented thus far do not indicate, however, whether the GP s or the psychologist s and patient s ratings more accurately reflect the patients actual clinical presentation. This was investigated again using correlational analysis to assess the strength of relationship between the global ratings (Global Symptom Severity and Clinical Global Improvement) and the standardized measures (HAM-A, SRT, and FQ-AG). Results are given in Table 4. GP Global Symptom Severity ratings correlated weakly with HAM-A, SRT and FQ-AG scores whereas the psychologist s Global Symptom Severity ratings correlated strongly with HAM-A, SRT and FQ-AG. Thus the GP s ratings symptom severity were not strongly related to either the psychologist s ratings or to the standardized measures of anxiety and avoidance. By contrast, the psychologist s ratings of symptom severity were in close agreement with anxiety as measured by the HAM-A and SRT, and avoidance as rated on the FQ-AG. A similar pattern emerged for GP, psychologist and patient Clinical Global Improvement, with GP ratings correlating weakly with HAM-A and SRT scores while the correlation with FQ-AG scores failed to reach significance. By contrast, psychologist and patient ratings of Clinical Global Improvement showed strong correlations with HAM-A and SRT scores, whilst the correlations with FQ-AG scores tended to be slightly lower but remained statistically significant. Thus for improvement following treatment a greater degree of agreement between global ratings and standardized measures was found for psychologist and patient ratings than for GP ratings. It would appear therefore that GP global ratings symptom severity and improvement following treatment did not accurately reflect the clinical picture as presented by the psychologist and patient ratings and by standardized measures of anxiety and avoidance behaviours.

11 Panic outcome ratings in primary care 23 The lower correlations found with the FQ-AG suggest that this measure of agoraphobic avoidance behaviour was less strongly related to the global measures of symptom severity, and in particular clinical global improvement than were the measures of anxiety (HAM-A, SRT). Discussion Overall the global ratings employed in this study proved to be a useful and valid indication of treatment effectiveness and reinforce the recent calls for global measures to be used in treatment outcome studies in panic disorder (Shear & Maser, 1994). This position is further strengthened by the strong correlations found between these global ratings and the more traditional measures of anxiety and avoidance (HAM-A, SRT, FQ-AG) particularly for psychologist and patient ratings. This relationship between the global ratings of symptom severity and improvement and the measures of anxiety and avoidance merits further comment. The strongest relationship with the global ratings was found for the measures of anxiety (HAM-A, SRT), rather than the measure of behavioural avoidance (FQ-AG). Both the patients and the psychologist s perceptions of symptom severity and global improvement were therefore more closely in accord with their ratings of general anxiety than those for behavioural avoidance. This is potentially of relevance given that the Fear Questionnaire, and the Agoraphobia subscale in particular, is one of the most widely used outcome measures in treatment outcome research on panic disorder and agoraphobia. The current finding suggests that, for the raters in this study, the construct of general anxiety was of more relevance to the assessment of clinical outcome than the construct of behavioural avoidance. This finding may be attributable in part to the way in which the questions relating to the global ratings were phrased. For the Global Symptom Severity rating assessors were asked how emotionally distressed the patient was; this may have led to a reduced emphasis on the behavioural, avoidance based, aspects of the patient s presentation. For the Clinical Global Improvement ratings, the question referred simply to improvement compared to status at entry, thus the rater had considerable freedom as to which facets of the problem should be considered. Whilst these procedural issues are of relevance, the finding here of a weaker relationship between the global ratings and the FQ-AG nonetheless raises a question over the validity of assessments of outcome and clinical significance based solely on measures of behavioural avoidance such as the FQ-AG. It also argues for the inclusion of measures of general anxiety as standard in treatment outcome studies on panic disorder and agoraphobia, something that has not occurred widely to date. This question warrants further investigation, specifically of the interrelationships between these two methods of assessing clinical outcomes, namely those based on the use of global ratings and those based on the use of standardized measures of both avoidance behaviour and general anxiety. This again reinforces the need for global ratings to be incorporated as standard into treatment outcome methodology. Global ratings do, however, rely on the accuracy of the raters used, as demonstrated by the GP s ratings of symptom severity and global improvement. The GP s ratings differed significantly from those of the psychologist and the patients themselves. This difference was further reinforced by the noticeably weaker correlations between the GP

12 24 D. M. Sharp and K. G. Power ratings and the measures of anxiety and avoidance. A more forthright statement of the finding would be that GPs were unable to detect the large changes following treatment that were obvious to both the psychologist and the patients. One possible explanation may be simply the length of time spent with the patients, and thus familiarity with the clinical picture. The psychologist had 9 appointments with each patient over a 13 week period, representing a maximum of 9 hours contact. The patients familiarity with their own condition needs little elaboration. GPs, however, are unlikely to have had anything like the same amount of concentrated contact with the patients, and thus the accuracy of their assessments may have been reduced as a result. It may also be the case that the GPs in the present study were insufficiently experienced in rating outcomes of panic disorder and agoraphobia. In the only previous study to employ GP ratings, (Power et al., 1990) the referring GPs were all members of a research group and had received training in research assessment methods. The Power et al. (1990) study found that GP ratings were highly correlated with those of the psychologist. This would suggest that training of referring GPs might be required prior to their ratings being used in future studies. If replicated in future studies, the lack of acuity demonstrated by GP raters in this study would have relevance beyond the measurement of experimental treatment outcome. Researchers investigating, for example, service usage or number of drug prescriptions following an experimental treatment may have to be aware that such post study treatment would be likely to be provided by the patient s GP, and its provision would be informed by the GP s perception of treatment outcome, whether accurate or not. The overall picture to emerge from the study reported here is that global ratings of outcome can serve as useful adjuncts to more traditional measures in treatment studies of panic disorder and agoraphobia. They may also help to clarify which of these traditional measures best reflects the overall clinical condition of the patient. Psychologist and patient ratings also differentiated between treatments, with an advantage being shown for those treatments including cognitive behaviour therapy, (FLCCBT, PLCCBT, CBT) with the CBT and FLCCBT groups showing the most consistent advantage. Some caution should be exercised, however, in interpreting the results obtained for the CBT alone group in this study. The CBT treatment showed a higher drop-out rate (32.2%, ng13) than the other treatment groups in the present study. This drop-out rate from CBT is also higher than that reported in some other studies of CBT for panic disorder and agoraphobia. Similar drop-out rates (circa 20 25%) from psychological treatments were, however, found for panic disorder patients in a series of studies by Michelson and colleagues (Michelson, Mavissakalian, & Marchione, 1985; Michelson et al., 1990) using graded therapist assisted exposure and programmed practice, and by Black and colleagues (Black et al., 1993) using cognitive therapy. Interestingly, these are the only other reports of drop-out rates from psychological treatments given in the controlled absence of concurrent psychotropic medication. It is also possible that the primary care treatment setting of the present study may have had an influence on drop-out rates. In a review of drop-out rates from psychotherapy delivered in a variety of settings Hunt and Andrews (1992) reported that drop-out rates were around 8% in controlled explanatory studies delivered to restricted populations in research centres, and rose to circa 17 20% for treatments delivered in specialist

13 Panic outcome ratings in primary care 25 centres. They noted, however, that drop-out rates rose to between 30 60% for psychotherapies delivered in community facilities. Thus, accepting firstly the large assumption that treatments were of equivalent quality in each facility, the setting in which a treatment is delivered may have a considerable influence on drop-out rate. The drug group (FL) was superior to placebo (PL) on the psychologist s rating of Global Symptom Severity post-treatment, but did not show significant differences from placebo (PL) on either psychologist or patient ratings of Clinical Global Improvement following treatment. In terms of treatment efficacy, therefore, the drug fluvoxamine occupies an intermediate position being more effective than placebo but less effective than treatments including cognitive behaviour therapy. The validity of such between group differences obviously rests in the strengths and weaknesses of the overall study design. These are discussed in detail elsewhere (Sharp et al., 1996). Acknowledgements This research was supported in part by Solvay Duphar Ltd (Grant No. H114928). References AMERICAN PSYCHIATRIC ASSOCIATION (1987). Diagnostic and statistical manual of disorders (3rd edn. revised, DSM-III-R). Washington, DC: Author. BAER, D. M. (1988). If you know why you re changing a behaviour you ll know when you ve changed it enough. Behavioural Assessment, 10, BALLENGER, J. C. (1994). Overview of the pharmacotherapy of panic disorder. In B. E. Wolf & J. D. Maser (Eds.), Treatment of panic disorder: A consensus development conference. Washington, DC: American Psychiatric Press. BARLOW, D. H. (1988). Anxiety and its disorders: The nature and treatment of anxiety and panic. New York: Guilford Press. BARLOW, D. H., CRASKE, M. G., CERNY, J. A.,& KLOSKO, J. S. (1989). Behavioural treatment of panic disorder. Behaviour Therapy, 20, BLACK, D. W., WESNER, R., BOWERS, W., & GABEL, J. (1993). A comparison of fluvoxamine, cognitive therapy, and placebo in the treatment of panic disorder. Archives of General Psychiatry, 50, CHAMBLESS, D. L., & GILLIS, M. M. (1994). A review of psychosocial treatments for panic disorder. In B. E. Wolf & J. D. Maser (Eds.), Treatment of panic disorder: A consensus development conference. Washington, DC: American Psychiatric Press. CLARK, D. M., SALKOVSKIS, P. M., HACKMAN, A., MIDDLETON, H., ANASTASIADES, P., & GELDER, M. (1994). A comparison of cognitive therapy applied relaxation and imipramine in the treatment of panic disorder. British Journal of Psychiatry, 164, COTE, G., GAUTHIER, J. G., LABERGE, B., CORMIER, H. J.,& PLAMONDON, J. (1994). Reduced therapist contact in the cognitive behavioural treatment of panic disorder. Behaviour Therapy, 25, GUY, W. (1976). Assessment manual for psychopharmacology. Washington, DC: U.S. Government Printing Office. HAMILTON, M. (1959). The assessment of anxiety states by rating. British Journal of Medical Psychology, 32, HOLLON, S. D.,& FLICK, S. N. (1988). On the meaning and methods of clinical significance. Behavioural Assessment, 10,

14 26 D. M. Sharp and K. G. Power HUNT, H. C.,& ANDREWS, G. (1992). Drop-out rate as a performance indicator in psychotherapy. Acta Psychiatrica Scandinavica, 85, JACOBSON, N. S., FOLLETTE, W. C., & RAVENSTORF, D. (1984). Psychotherapy outcome research: Methods for reporting variability and evaluating clinical significance. Behaviour Therapy, 15, JACOBSON, N. S.,& RAVENSTORF, D. (1988). Statistics for assessing clinical significance of psychotherapy techniques: Issues problems and new developments. Behavioural Assessment, 10, JACOBSON, N. S.,& TRAUX, P. (1991). Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59, JACOBSON, N. S., WILSON, L., & TUPPER, C. C. (1988). The clinical significance of treatment gains resulting from exposure-based interventions for agoraphobia: A reanalysis of outcome data. Behaviour Therapy, 19, KATERNDAHL, D. A.,& REALINI, J. P. (1995). Where do panic attack sufferers seek care? Journal of Family Practice, 40, KAZDIN, A. E. (1977). Assessing the clinical or applied importance of behaviour change through social validation. Behaviour Modification, 1, KELLNER, R., & SHEFFIELD, B. F. (1973). A self-rating scale of distress. Psychological Medicine, 3, KELLNER, R., & UHLENHUTH, E. H. (1991). The rating and self-rating of anxiety. British Journal of Psychiatry, 159 (suppl. 12), MARKS, I. M.,& MATHEWS, A. M. (1979). A brief standard self-rating for phobic patients. Behaviour Research and Therapy, 17, MAVISSAKALIAN, M. (1986). Clinically significant improvement in agoraphobia research. Behaviour Research and Therapy, 24, MICHELSON, L., MARCHIONE, K., GREENWALD, M., GLANZ, L., TESTA, S., & MARCHIONE, N. (1990). Panic disorder: Cognitive-behavioural treatment. Behaviour Research and Therapy, 28, MICHELSON, L., MAVISSAKALIAN, M.,& MARCHIONE, K. (1985). Cognitive and behavioural treatments of agoraphobia: Clinical behavioural and psychophysiological outcomes. Journal of Consulting and Clinical Psychology, 53, MONTGOMERY, S. A.,& ASBERG, M. (1979). A new depression rating scale designed to be sensitive to change. British Journal of Psychiatry, 134, POWER, K. G., SIMPSON, R. J., SWANSON, V., & WALLACE, L. A. (1990). Controlled comparisons of pharmacological and psychological treatment of generalised anxiety disorder in primary care. British Journal of General Practice, 40, SHARP, D. M., POWER, K. G., SIMPSON, R. J., SWANSON, V., MOODIE, E., ANSTEE, J. A., & ASHFORD, J. (1996). Fluvoxamine placebo and cognitive behaviour therapy used alone and in combination in the treatment of panic disorder and agoraphobia. Journal of Anxiety Disorders, 10, SHEAR, M. K., BALL, G., FITZPATRICK, M., JOSEPHSON, S., KLOSKO, J., & FRANCES, A. (1991). Cognitive behaviour therapy for panic disorder: An open study. Journal of Nervous and Mental Disease, 179, SHEAR, M. K.,& MASER, J. D. (1994). Standardized assessment for panic disorder research. Archives of General Psychiatry, 51, SIMPSON, R. J., KAZMIERCZAK, T., POWER, K.G.,&SHARP, D. M. (1994). Controlled comparison of the characteristics of panic patients in primary care. British Journal of General Practice, 44,

15 Panic outcome ratings in primary care 27 STRUPP, H. H.,& HADLEY, S. W. (1977). A tripartate model of mental health and therapeutic outcomes. American Psychologist, 32, TELCH, M. J. (1988). Combined pharmacological and psychological treatments for panic sufferers. In S. Rachman & J. D. Maser (Eds.), Panic: Psychological perspectives. New Jersey: Lawrence Erlbaum.