Drug treatment of panic disorder: early response to treatment as a predictor of final outcome Albus M, Lecrubier Y, Maier W, Buller R, Rosenberg R, Hippius H. Drug treatment of panic disorder: early response to treatment as a predictor of final outcome. Acta Psychiatr Scand 1990: 82: 359-365.
1
One of the core problems in clinical research is the detection of early changes in target symptoms that predict future therapeutic outcome. To analyze potential predictors of outcome, data of a multicenter study on patients with panic disorder were used. A total of 1010 patients were randomly allocated either to alprazolam, imipramine or placebo treatment. Early improvement in the number of spontaneous panic attacks within the first week of treatment predicted outcome exclusively in the alprazolam group. In contrast, placebo responders and nonresponders were differentiated by early changes in anticipatory anxiety intensity. For tricyclic antidepressants such as imipramine an evaluation period of more than one week is required to allow conclusions about outcome.
One of the core problems in clinical research is the prediction of response to different treatment procedures. Two problems are involved: defining response or outcome and searching for variables that improve at an early stage and predict final outcome. Response or improvement may be specific to a few target symptoms or may form a broad spectrum across many symptoms (1). The definition of target symptoms in panic disorder depends at least partially on the underlying concepts of the disease. So far, the dominant view in United States psychiatry, which is reflected in DSM-III-R (2),is that panic attacks are primary and anticipatory anxiety and phobic avoidance appear secondarily (3). In contrast, some leading behaviorists argue for a more important role for phobic avoidance (4). Outcome measures considered in therapeutic studies published to date include: presence or absence of spontaneous panic attacks (5), phobic anxiety and avoidance, panic and overall severity score (6), the Hamilton Rating Scale for Anxiety (HRSA) (7, 8) the Hopkins Symptom Checklist (9), the Symptom Check List-90 (lo), frequency of panic attacks, anticipatory anxiety episodes, and overall phobia (1 1). However, the outcome measures that should be considered as most appropriate in patients with panic disorder is still controversial. A least one target symptom, the number of spontaneous panic attacks, has previously been found to be a good outcome measure (5).
M. Albus’, Y. Lecrubier*, W. Maier3, R. Buller3, R. Rosenberg4, H. Hippius’
’
Psychiatric Hospital, University of Munich, Federal Republic of Germany, Charge de Recherche, INSERM Unite 302, Paris, France, Psychiatric Hospital, University of Mainz, Federal Republic of Germany, Department of Psychiatry, Rigshospitalet, Copenhagen, Denmark
Key words: alprazolam; imipramine; panic disorder; placebo; prediction Margot Albus, Psychiatric Hospital, University of Munich, Nussbaumstr. 7, D-8000 Munich
2, FRG Accepted for publication June 24, 1990
The assumption that early response to treatment predicts long-term outcome is derived from early observations on schizophrenic patients (12). Several treatment studies showed that reduction of productive symptoms within 7-10 d predicts a more favorable long-term outcome in schizophrenic patients (13-15). Although reduction of productive symptoms by neuroleptics can be considered as an empirically supported predictor for final outcome, no consistent predictors for clinical response to anxiolytic drugs have been reported until now. To date, only qualitatively different responses to the different anxiolytic drugs seem proven. The drugs most frequently used in the treatment of panic disorder patients, tricyclic antidepressants (TCD) and benzodiazepines, are effective on a variety of symptoms. Thus, one can assume that early changes in symptoms other than in the target symptoms may predict the outcome. There are some traces of evidence that anxiolytic drugs show different early and late effects on different parameters. Downing & Rickels (9) showed that the most consistent predictors in their studies, anxiety, somatization and depression, entail contrasting relationships between response to chlordiazepoxide and placebo. The authors conclude that these predictors are basically not predictors of response to the treatment situation as a whole but are factors that differentially mediate response to chlordiazepoxide and placebo. Other investigators (16) found compelling evidence in favor of a specific 359
Albus et al.
pharmacological effect of imipramine on phobia and depression measures, with no striking differences on panic measures as compared with placebo. McNair & Kahn (17) postulated an association between the antidepressant and antiphobic effects but not the antipanic effects of imipramine. Evidence from another study (18) suggests that imipramine may possess a direct effect on the reduction of phobic responses. Optimal response may require doses of 150 mg imipramine/day or more, but side effects are likely to interfere with the attainment of optimal dose (16). These side effects are most likely responsible for the high drop-out rates in patients treated with imipramine (19) and therefore mar the prediction of potential response on the target symptoms. Placebo seems to work in the treatment of panic patients as well. After only 1 week of placebo, the percentage of patients experiencing spontaneous panic attacks dropped from 72% to 58% with an additional decrease to 25 % by the end of week 8 (5). This finding contrasts with the results of another study (8), in which only 5.4% of anxiety patients showed a positive placebo response. In clinical practice, the search for early changes in certain symptoms that predict final outcome is at least as relevant as research on target symptoms and overall therapeutic response. However, the evidence that early changes in variables are stable and thus predictors of outcome is sparse and inconclusive (5). This is partially because of unsolved methodological issues : the higher drop-out rates of patients treated with TCDs, which have more initial side effects and a later onset of therapeutic action, and the diverging effects of different anxiolytics or placebo on different target symptoms make a separate analysis of different treatment groups necessary. Thus, the low numbers of subjects investigated in the above-cited studies do not provide a sufficient base for appropriate statistical analysis. In a multicenter trial of 526 patients with panic disorder, agoraphobia, and panic attacks, half of the improvement in panic attacks, phobia and disability was already seen by the end of the first week in patients treated with alprazolam and was significantly greater than that resulting from placebo (1 1). Based on these findings, in a cross-national collaborative panic study, the following question was approached. Which early changes in measures relevant to panic disorder predict final outcome in patients treated with alprazolarn, imipramine or placebo?
pramine and placebo in patients with panic disorder or panic disorder with agoraphobia (20). Since the patients, based on a structured interview (21), met DSM-I11 criteria for panic disorder (22), panic attacks were considered as primary target symptoms, whereas anticipatory anxiety and avoidance were seen as secondary target symptoms. Thus, the number of situational, spontaneous and total panic attacks as well as phobic avoidance and anticipatory anxiety, as measured by overall phobia and intensity of anticipatory anxiety, were used to define categorical and dimensional responses to treatment. As secondary measures for outcome, the total anxiety score of the HRSA and physician’s global improvement as well as work and social disability were considered. In defining predictors, the following considerations had to be taken into account. Since only patients with a complete data set for all ratings could be included in the analysis, considering all potential outcome measures as predictors would have reduced the sample size too much. This was especially important for the primary target symptoms total, situational and spontaneous panic attacks. Since all these data were log-transformed, each data set where the log transformation was zero had to be defined as missing data. Therefore, only the number of spontaneous panic attacks was chosen as predictor, since spontaneous panic attacks are considered to be a core symptom of panic disorder (3). Of the secondary measures of outcome the total anxiety score of the HRSA was choosen, since this is the most widely used outcome measure in anxiolytic drug trials. The physician’s global improvement reflects a change score already, with no data available at baseline. Therefore, this measure could not be used. Thus, balancing out the requirement of a representative sample and the consideration of the most relevant outcome measures according to the literature to date resulted in a limitation to only four predictors. Consequently, the change in spontaneous panic attacks, overall phobia, anticipatory anxiety intensity and HRSA total score between week 1 and baseline was defined as early response to treatment. Change in these measures was then used in analyzing the predictive value for good or poor response in the outcome measures in each of the 3 treatment groups.
Material and methods Study design
A logistic regression model for each medication group was carried out. In that model the change variable was defined by the value at week 1 divided by the baseline value. Maximum likelihood analysis of variance (CATMOD procedure) was calculated for the following parameters: change in HRSA total
The cross-national collaborative panic study was a multicenter study comparing in a double-blind randomized design the efficacy of alprazolam, imi-
360
Statistical analysis
Prediction of therapeutic outcome score, spontaneous panic attacks, anticipatory anxiety intensity and overall phobia. In the imipramine and placebo groups, the rate of drop-out caused by lack of improvement or side effects was considerably higher than in the alprazolam group, and the week-8 comparison of study completers therefore has a bias favoring placebo and imipramine. In contrast, endpoint analyses, which have become a standard method in such situations, correspond to a pragmatic result in the largest data sample available by using the week-8 data when present and the last available value carried forward when they are not. This type of analysis may favor the alprazolam group, since carrying forward the last values for drop-outs caused by lack of improvement may underestimate how well these patients would have done had they remained in the study. In both types of analyses, only the data of patients that had remained at least for 3 weeks in the active treatment phase of the study were included. Thus, both week-8 and endpoint values for the outcome measures situational, spontaneous and total panic attacks, HRSA total score, overall phobia, anticipatory anxiety intensity and physician’s global improvement were then regressed against and above-defined predictor variables in each medication group-
number of situational panic attacks at endpoint. In contrast, in the imipramine group, none of the measures of early improvement predicted the number of situational panic attacks at the end of the drug trial. In the alprazolam group, early improvement in the number of spontaneous panic attacks predicted the number of spontaneous panic attacks at week 8 and endpoint (Table 2). In the placebo group, early response in the number of spontaneous panic attacks predicted the improvement in the number of spontaneous panic attacks at endpoint, and the early response in anticipatory anxiety predicted the number of spontaneous panic attacks at week 8 as well as at the endpoint. Again, in the imipramine group, none of the early improvement measures predicted spontaneous panic attacks, neither at week 8 nor at endpoint. In the alprazolam group, the change in the number of spontaneous panic attacks predicted the number of total panic attacks at week 8 and endpoint; the change in anticipatory anxiety intensity and in HRSA total score predicted the number of total panic attacks at endpoint (Table 3). In the placebo group, the change in anticipatory anxiety intensity predicted the number of total panic attacks at week 8 and endpoint; the change in HRSA total score predicted the number of total panic attacks at endpoint. As for spontaneous and situational panic attacks, none of the 4 early improvement measures predicted total panic attacks at the end of the study in the imipramine group. In the alprazolam group, the change in anticipatory anxiety intensity predicted anticipatory anxiety intensity at endpoint (Table 4). In the imipramine group, change in the number of spontaneous panic attacks as well as early response in anticipatory anxiety intensity predicted anticipatory anxiety intensity at week 8 and endpoint. In the placebo group,
Results
In the alprazolam group, the change in HRSA total score between week 1 and baseline predicted the number of situational panic attacks at week 8 and endpoint and the change in anticipatory anxiety intensity predicted the number of situational panic attacks at endpoint (Table 1). In the placebo group, the change in anticipatory anxiety intensity predicted the number of situational panic attacks at week 8, and the change in HRSA total score predicted the
Table 1. Prediction of the number of situational panic attacks (log-transformedl by the change in the 4 predictor variables between week 1 and baseline subvided by treatment group for week 8 (completer) as well as endpoint; type Ill analysis (only significant results are listed) Alprazolam n = 190 (week 8) 252 (endpoint)
F
P (2-tailed)
Log spontaneous panic attacks change week l/baseline
week 8 endpoint
Overall phobia change week 1 /baseline
week 8 endpoint
Anticipatory anxiety intensity change week 1/baseline
week 8 endpoint
6.36
1
One of the core problems in clinical research is the detection of early changes in target symptoms that predict future therapeutic outcome. To analyze potential predictors of outcome, data of a multicenter study on patients with panic disorder were used. A total of 1010 patients were randomly allocated either to alprazolam, imipramine or placebo treatment. Early improvement in the number of spontaneous panic attacks within the first week of treatment predicted outcome exclusively in the alprazolam group. In contrast, placebo responders and nonresponders were differentiated by early changes in anticipatory anxiety intensity. For tricyclic antidepressants such as imipramine an evaluation period of more than one week is required to allow conclusions about outcome.
One of the core problems in clinical research is the prediction of response to different treatment procedures. Two problems are involved: defining response or outcome and searching for variables that improve at an early stage and predict final outcome. Response or improvement may be specific to a few target symptoms or may form a broad spectrum across many symptoms (1). The definition of target symptoms in panic disorder depends at least partially on the underlying concepts of the disease. So far, the dominant view in United States psychiatry, which is reflected in DSM-III-R (2),is that panic attacks are primary and anticipatory anxiety and phobic avoidance appear secondarily (3). In contrast, some leading behaviorists argue for a more important role for phobic avoidance (4). Outcome measures considered in therapeutic studies published to date include: presence or absence of spontaneous panic attacks (5), phobic anxiety and avoidance, panic and overall severity score (6), the Hamilton Rating Scale for Anxiety (HRSA) (7, 8) the Hopkins Symptom Checklist (9), the Symptom Check List-90 (lo), frequency of panic attacks, anticipatory anxiety episodes, and overall phobia (1 1). However, the outcome measures that should be considered as most appropriate in patients with panic disorder is still controversial. A least one target symptom, the number of spontaneous panic attacks, has previously been found to be a good outcome measure (5).
M. Albus’, Y. Lecrubier*, W. Maier3, R. Buller3, R. Rosenberg4, H. Hippius’
’
Psychiatric Hospital, University of Munich, Federal Republic of Germany, Charge de Recherche, INSERM Unite 302, Paris, France, Psychiatric Hospital, University of Mainz, Federal Republic of Germany, Department of Psychiatry, Rigshospitalet, Copenhagen, Denmark
Key words: alprazolam; imipramine; panic disorder; placebo; prediction Margot Albus, Psychiatric Hospital, University of Munich, Nussbaumstr. 7, D-8000 Munich
2, FRG Accepted for publication June 24, 1990
The assumption that early response to treatment predicts long-term outcome is derived from early observations on schizophrenic patients (12). Several treatment studies showed that reduction of productive symptoms within 7-10 d predicts a more favorable long-term outcome in schizophrenic patients (13-15). Although reduction of productive symptoms by neuroleptics can be considered as an empirically supported predictor for final outcome, no consistent predictors for clinical response to anxiolytic drugs have been reported until now. To date, only qualitatively different responses to the different anxiolytic drugs seem proven. The drugs most frequently used in the treatment of panic disorder patients, tricyclic antidepressants (TCD) and benzodiazepines, are effective on a variety of symptoms. Thus, one can assume that early changes in symptoms other than in the target symptoms may predict the outcome. There are some traces of evidence that anxiolytic drugs show different early and late effects on different parameters. Downing & Rickels (9) showed that the most consistent predictors in their studies, anxiety, somatization and depression, entail contrasting relationships between response to chlordiazepoxide and placebo. The authors conclude that these predictors are basically not predictors of response to the treatment situation as a whole but are factors that differentially mediate response to chlordiazepoxide and placebo. Other investigators (16) found compelling evidence in favor of a specific 359
Albus et al.
pharmacological effect of imipramine on phobia and depression measures, with no striking differences on panic measures as compared with placebo. McNair & Kahn (17) postulated an association between the antidepressant and antiphobic effects but not the antipanic effects of imipramine. Evidence from another study (18) suggests that imipramine may possess a direct effect on the reduction of phobic responses. Optimal response may require doses of 150 mg imipramine/day or more, but side effects are likely to interfere with the attainment of optimal dose (16). These side effects are most likely responsible for the high drop-out rates in patients treated with imipramine (19) and therefore mar the prediction of potential response on the target symptoms. Placebo seems to work in the treatment of panic patients as well. After only 1 week of placebo, the percentage of patients experiencing spontaneous panic attacks dropped from 72% to 58% with an additional decrease to 25 % by the end of week 8 (5). This finding contrasts with the results of another study (8), in which only 5.4% of anxiety patients showed a positive placebo response. In clinical practice, the search for early changes in certain symptoms that predict final outcome is at least as relevant as research on target symptoms and overall therapeutic response. However, the evidence that early changes in variables are stable and thus predictors of outcome is sparse and inconclusive (5). This is partially because of unsolved methodological issues : the higher drop-out rates of patients treated with TCDs, which have more initial side effects and a later onset of therapeutic action, and the diverging effects of different anxiolytics or placebo on different target symptoms make a separate analysis of different treatment groups necessary. Thus, the low numbers of subjects investigated in the above-cited studies do not provide a sufficient base for appropriate statistical analysis. In a multicenter trial of 526 patients with panic disorder, agoraphobia, and panic attacks, half of the improvement in panic attacks, phobia and disability was already seen by the end of the first week in patients treated with alprazolam and was significantly greater than that resulting from placebo (1 1). Based on these findings, in a cross-national collaborative panic study, the following question was approached. Which early changes in measures relevant to panic disorder predict final outcome in patients treated with alprazolarn, imipramine or placebo?
pramine and placebo in patients with panic disorder or panic disorder with agoraphobia (20). Since the patients, based on a structured interview (21), met DSM-I11 criteria for panic disorder (22), panic attacks were considered as primary target symptoms, whereas anticipatory anxiety and avoidance were seen as secondary target symptoms. Thus, the number of situational, spontaneous and total panic attacks as well as phobic avoidance and anticipatory anxiety, as measured by overall phobia and intensity of anticipatory anxiety, were used to define categorical and dimensional responses to treatment. As secondary measures for outcome, the total anxiety score of the HRSA and physician’s global improvement as well as work and social disability were considered. In defining predictors, the following considerations had to be taken into account. Since only patients with a complete data set for all ratings could be included in the analysis, considering all potential outcome measures as predictors would have reduced the sample size too much. This was especially important for the primary target symptoms total, situational and spontaneous panic attacks. Since all these data were log-transformed, each data set where the log transformation was zero had to be defined as missing data. Therefore, only the number of spontaneous panic attacks was chosen as predictor, since spontaneous panic attacks are considered to be a core symptom of panic disorder (3). Of the secondary measures of outcome the total anxiety score of the HRSA was choosen, since this is the most widely used outcome measure in anxiolytic drug trials. The physician’s global improvement reflects a change score already, with no data available at baseline. Therefore, this measure could not be used. Thus, balancing out the requirement of a representative sample and the consideration of the most relevant outcome measures according to the literature to date resulted in a limitation to only four predictors. Consequently, the change in spontaneous panic attacks, overall phobia, anticipatory anxiety intensity and HRSA total score between week 1 and baseline was defined as early response to treatment. Change in these measures was then used in analyzing the predictive value for good or poor response in the outcome measures in each of the 3 treatment groups.
Material and methods Study design
A logistic regression model for each medication group was carried out. In that model the change variable was defined by the value at week 1 divided by the baseline value. Maximum likelihood analysis of variance (CATMOD procedure) was calculated for the following parameters: change in HRSA total
The cross-national collaborative panic study was a multicenter study comparing in a double-blind randomized design the efficacy of alprazolam, imi-
360
Statistical analysis
Prediction of therapeutic outcome score, spontaneous panic attacks, anticipatory anxiety intensity and overall phobia. In the imipramine and placebo groups, the rate of drop-out caused by lack of improvement or side effects was considerably higher than in the alprazolam group, and the week-8 comparison of study completers therefore has a bias favoring placebo and imipramine. In contrast, endpoint analyses, which have become a standard method in such situations, correspond to a pragmatic result in the largest data sample available by using the week-8 data when present and the last available value carried forward when they are not. This type of analysis may favor the alprazolam group, since carrying forward the last values for drop-outs caused by lack of improvement may underestimate how well these patients would have done had they remained in the study. In both types of analyses, only the data of patients that had remained at least for 3 weeks in the active treatment phase of the study were included. Thus, both week-8 and endpoint values for the outcome measures situational, spontaneous and total panic attacks, HRSA total score, overall phobia, anticipatory anxiety intensity and physician’s global improvement were then regressed against and above-defined predictor variables in each medication group-
number of situational panic attacks at endpoint. In contrast, in the imipramine group, none of the measures of early improvement predicted the number of situational panic attacks at the end of the drug trial. In the alprazolam group, early improvement in the number of spontaneous panic attacks predicted the number of spontaneous panic attacks at week 8 and endpoint (Table 2). In the placebo group, early response in the number of spontaneous panic attacks predicted the improvement in the number of spontaneous panic attacks at endpoint, and the early response in anticipatory anxiety predicted the number of spontaneous panic attacks at week 8 as well as at the endpoint. Again, in the imipramine group, none of the early improvement measures predicted spontaneous panic attacks, neither at week 8 nor at endpoint. In the alprazolam group, the change in the number of spontaneous panic attacks predicted the number of total panic attacks at week 8 and endpoint; the change in anticipatory anxiety intensity and in HRSA total score predicted the number of total panic attacks at endpoint (Table 3). In the placebo group, the change in anticipatory anxiety intensity predicted the number of total panic attacks at week 8 and endpoint; the change in HRSA total score predicted the number of total panic attacks at endpoint. As for spontaneous and situational panic attacks, none of the 4 early improvement measures predicted total panic attacks at the end of the study in the imipramine group. In the alprazolam group, the change in anticipatory anxiety intensity predicted anticipatory anxiety intensity at endpoint (Table 4). In the imipramine group, change in the number of spontaneous panic attacks as well as early response in anticipatory anxiety intensity predicted anticipatory anxiety intensity at week 8 and endpoint. In the placebo group,
Results
In the alprazolam group, the change in HRSA total score between week 1 and baseline predicted the number of situational panic attacks at week 8 and endpoint and the change in anticipatory anxiety intensity predicted the number of situational panic attacks at endpoint (Table 1). In the placebo group, the change in anticipatory anxiety intensity predicted the number of situational panic attacks at week 8, and the change in HRSA total score predicted the
Table 1. Prediction of the number of situational panic attacks (log-transformedl by the change in the 4 predictor variables between week 1 and baseline subvided by treatment group for week 8 (completer) as well as endpoint; type Ill analysis (only significant results are listed) Alprazolam n = 190 (week 8) 252 (endpoint)
F
P (2-tailed)
Log spontaneous panic attacks change week l/baseline
week 8 endpoint
Overall phobia change week 1 /baseline
week 8 endpoint
Anticipatory anxiety intensity change week 1/baseline
week 8 endpoint
6.36
