Alexithymia as a predictor of treatment response in post-traumatic stress disorder

Journal of Traumatic Stress, Vol. 5, No. 4, 1992

Alexithymia as a Predictor of Treatment Response in Post-Traumatic Stress Disorder Thomas R. Kosten, 1 John H. Krystal, 1 Earl L. Giller, Jr., 1 Julia Frank,. and Elisheva Dan 1

Fifty-seven veterans with post-traumatic stress disorder (PTSD) completed the Alexithymia Provoked Response Questionnaire (APRQ) upon entering an 8-week randomized trial comparing phenelzine, imipramine, and placebo. Low alexithymia on the A P R Q significantly predicted improvement on the avoidance items of the Impact of Events Scale (IES) particularly among patients treated with placebo, but was not associated with changes in the intrusion items of the scale. KEY WORDS: alexithymia; post-traumatic stress disorder; imipramine; phenelzine.

INTRODUCTION Alexithymia was introduced as a concept by Sifneos (Sifneos et al., 1977) in 1972 to describe a cognitive-affective style exhibited by patients with psychosomatic illness. Patients with alexithymia substitute action and descriptive words for affective language (Taylor, 1984). When these patients are asked "How would you feel if you saw a truck coming at you at 90 mph?" instead of responding "I would feel scared," they might reply "I don't know. No feeling; I'd get out of the way." (This was a response by a Vietnam veteran with post-traumatic stress disorder~PTSD.) The prevalence of alexithymia among early adult men, who might develop PTSD after combat exposure in Vietnam, has been estimated as 8% (Blanchard et al., 1981). Because of its relatively high prevalence in young males, alexithymia is an important risk factor for developing PTSD after exposure to combat 1Department of Psychiatry and the West Haven V. A. Medical Center, Yale University SchooI of Medicine and University of Connecticut Health Center. 563 0894-9867/92/1000-0563506.50/0© 1992PlenumPublishingCorporation

564

Kosten et al.

stress. The inability to verbalize feelings (alexithymia) may predispose a young man to develop delayed onset PTSD, rather than "working through" the combat or other trauma soon after the event, and this potential predisposition for PTSD gives alexithymia its importance as a clinical problem. Patients with PTSD have been described by Krystal as having alexithymic characteristics, and alexithymia may impact their treatment response (Krystal, 1971, 1977). Alexithymia has been considered "possibly the most important single factor diminishing the success of psychoanalysis and psychodynamic psychotherapy" (Krystal, 1982). Alexithymic patients with psychosomatic disorders have been considered instead as candidates for educational therapy approaches (Apfel-Savitz et al., 1977; Sifneos, 1975), and they may be particularly in need of antidepressant pharmacotherapy before psychotherapy can be effective (Blumer and Heilbronn, 1982; Taylor, 1984). This potential interaction of alexithymia with antidepressants and psychotherapy has not been studied in PTSD patients. The current study is an attempt to confirm this association among PTSD patients in a randomized clinical trial of pharmocotherapy with concurrent psychotherapy. Pharmacotherapy for PTSD has been evaluated using various scales to assess the DSM-IIIR criteria included in this diagnosis. Two classes of symptoms in these criteria--intrusive and avoidance symptoms----can be assessed using a standardized instrument--the Impact of Events Scale (IES) (Horowitz et al., 1972). While this scale does not cover all DSM-IIIR criteria for PTSD, it is a reliable and valid instrument that covers the salient features of PTSD. Among patients meeting DSM-IIIR criteria for PTSD at admission to treatment, the IES can be an excellent measure of treatment response just as various depression scales can be excellent outcome measures in patients with major depressive disorders. In controlled trials, antidepressant medications including amitriptyline, phenelzine, and imipramine have improved the symptoms of PTSD including nightmares, flashbacks and startle responses, as well as avoidance symptoms (Davidson et al., 1990; Frank et aL, 1988). These avoidance symptoms include efforts to avoid talking or thinking about the traumatic event, emotional numbing, and general avoidance of feelings. These avoidance symptoms are strikingly similar to some characteristics of alexithymia, but in alexithymia the avoidance of emotional responses and of expressing feelings in words is not limited to the traumatic event, but is generalized to everyday life. Thus, patients with substantial alexithymia may be particularly predisposed to have severe avoidance symptoms, when they develop PTSD, and these symptoms may respond better to antidepressants than to the usual psychological therapies for PTSD.

Alexlthymia as a Predictor of Treatment Response

565

In order to examine the relationship of alexithymia to treatment response, we assessed alexithymia before starting treatment, then randomly assigned PTSD patients to imipramine, phenelzine, or placebo, all in combination with psychotherapy (Frank et al., 1988; Kosten et al., 1992). In a previous study with PTSD veterans, we had developed an alexithymia scale called the Alexithymia Provoked Response Questionnaire (APRQ) (Krystal et al., 1986). On the APRQ, PTSD inpatients, who had failed outpatient treatment, had more severe alexithymia than outpatients, who were being adequately managed without hospitalization. Thus, more severe alexithymia appeared to predict a poorer outpatient treatment response in PTSD. The present study tested this prediction by examining whether alexithymia was associated with a poor treatment response on our two IES scales. A second issue was suggested by previous findings that alexithymic patients responded better to antidepressants than to psychotherapy alone (Taylor, 1984). If this assertion held for PTSD patients, then we expected that PTSD patients with high alexithymia would have a poorer treatment response to psychotherapy alone (placebo group) than to psychotherapy combined with pharmacotherapy (the two medicated groups). This relationship would be reflected in a significant correlation between alexithymia and outcome in the group without pharmacotherapy (the placebo group), and minimal correlation between alexithymia and outcome in the two antidepressant treated groups.

METHODS Subjects Fifty-seven outpatient male Vietnam combat veterans who were recruited through the New Haven and Hartford, Connecticut Veterans Outreach Programs, and who met DSM-III criteria for PTSD, completed the APRQ structured interview and entered into a randomized clinical trial of placebo (n = 16), imipramine (n = 23), or phenelzine (n = 18). Subjects with schizophrenia, bipolar disorder, or current substance abuse were excluded. Although subjects with substance abuse within the previous month were rejected, substantial rates (57%) of past substance abuse, including alcoholism, were found, Minor depression by RDC was found in 27 of the veterans. The sample of 57 included 7 (12%) nonwhite subjects; mean age (_+SD) was 39 _+2.3 years. All subjects gave written informed consent to participate in this randomized clinical trial.

566

Kosten et al.

Study Design Subjects in this study were interviewed using the Alexithymia Provoked Response Questionnaire (APRQ) before being randomly assigned to receive active medication or placebo under double-blind conditions in an 8-week study. The procedures for randomization involved a table of random numbers where digits 1 to 3 represented placebo, 4 to 6 represented imipramine, and 7 to 9 represented phenelzine assignment (zeros were not used). With this digit representation a random list of medication assignment was generated, and as each patient entered the study, he was given the next sequential medication assignment on this randomized list. Once weekly individual psychotherapy, medication monitoring, and symptom assessments were conducted. The psychotherapy was provided by a Masters level psychologist using a short-term psychodynamic approach (Marmor, 1979). While several subjects had been involved in "rap groups" for a few sessions in the past, this was the first exposure to individual psychotherapy for these patients. Details of the medication dosing and blood monitoring for imipramine levels and MAO inhibition can be found elsewhere (Frank et al., 1988). The mean maximal imipramine dose was 225 +_55, mg and the mean maximal dose of phenelzine was 68 +20 mg. Treatment response was assessed using the Impact of Events Scale (IES) both as a baseline measure of PTSD symptoms and as a weekly outcome measure throughout the trial (Horowitz et al., 1972). The IES includes two subscales with seven items assessing intrusion symptoms such as nightmares, flashbacks, and intrusive recollections and eight items assessing avoidance symptoms such as social withdrawal and emotional numbing. Combat-related atrocity exposure was assessed using a 6-item, 18-point scale with scores of three or higher indicating significant combat atrocity exposure. While this self-report has been used in other studies of Vietnam veterans, no published reliability or validity data are available (Egendorf et al., 1981). Scoring on the APRQ was structured so that a lower score indicated more alexithymia. The 17 items in the APRQ scored 0 for an alexithymic and 1 for a nonalexithymic response. Responses were scored as alexithymic if they: (1) described an action but not affect, e.g., "I'd leave," (2) contained detailed descriptions of the situation but no affective response, e.g., "It depends on how large the knife was. If it was small I might fight," (3) describe physical sensations rather than affect. In a preliminary study, inpatients with PTSD had a score of 5.9 +3.6, while outpatients had a score of 8.7 +_5.5. The score for inpatients with affective disorder was 11 +4.8. The APRQ also showed good correlation with the Beth Israel Psychosomatic Questionnaire, which is one of the most widely used clinical measures of alexithymia (Apfel and Sifneos, 1979; Krystal et aL, 1986; Lolas et al., 1980; Sifneos et

Alexithymia as a Predictor of Treatment Response

567

al., 1977), and good interrater reliability on the total A P R Q score with a kappa of 0.84. Since our preliminary study with the A P R Q , we have replicated its correlation with clinical assessments of alexithymia in other populations (Pierce et aL, 1989, 1990). In present study an additional test/retest reliability assessment of the A P R Q was possible, since 41 of the patients completed the alexithymia scale a second time about 2 months after the original testing. The 16 patients who did not complete the A P R Q a second time were equally distributed across the three medication groups.

Data Analysis Prediction of treatment outcome using the A P R Q was done with Pearson correlation coefficients, correlating A P R Q score with the final rating on the IES. Both the avoidance and intrusion subscales of the IES were examined. In order to adjust for baseline levels of symptomatology, two approaches were taken. First, difference scores between the initial IES scores and the final IES scores were computed, and these change scores were correlated with the A P R Q . Second, covariance analyses were done to adjust for baseline correlations between IES and the A P R Q . For analyses that involved comparison across medicated and placebo groups analysis of covariance was used with covariants including baseline scores on the IES and atrocity scores from the combat scale.

RESULTS

Alexithymia and Overall Treatment Response The mean A P R Q score at intake was 9.2+4.2 (standard deviation) and the mean score on the IES at termination was 24.6_+15 with subscale scores of 11.2+10.5 for the intrusion subscale and 12.8+7.2 for the avoidance subscale. Significant improvement was shown on both subscales of the IES. On the intrusion subscale there was an improvement of 4.9 points (t = 4.3; df = 55; p < 0.0001). On the avoidance subscale there was a mean change of 4.5 points (t = 4.4; df = 55; p < 0.0001). The percentage improvement from baseline was 30% for the intrusion subscale and 26% for the avoidance subscale in the total patient sample. The A P R Q significantly predicted treatment outcome on the IES. The terminal avoidance scale score of the IES was significantly correlated with the A P R Q (r = 0.30; df = 55; p < 0.05). The change from baseline on the avoidance scale was also significantly correlated with the A P R Q (r

568

Kosten a aL

= +0.26; p < 0.05). Using analysis of covariance with baseline avoidance scores, the relationship between alexithymia and outcome on the avoidance scale remained significant (F = 15.0; df = 2, 56; p < 0.0001). The A P R Q was not significantly correlated with either the terminal intrusion scale score (r = +0.06) or the change in the intrusion scale score (r = -0.01). Thus, alexithymia differentially predicted improvement in avoidance symptoms, but not improvement in intrusion symptoms.

Medication Impact on Treatment Response and Alexithymia Treatment improvement was significantly different across the three medication groups, and the range of response was relatively wide within each group, as shown in Table I. For the purposes of these analyses about the relationship of alexithymia to treatment response within medication groups, a wide range of outcomes within each group was more important than was the difference in mean scores across the three groups. The key question was whether the patients with good treatment outcomes could be predicted by the baseline alexithymia scores, which did not differ at baseline across the three groups as shown in Table I. The prediction of change in the avoidance subscale by the A P R Q was quite strong among the sixteen patients in the placebo group (r = +0.62; df = 14; p < 0.01), but no relationship was found within the 41 patients in the two medicated groups (r = +0.1). These two correlation coefficients were significantly different using the Z test (p < 0.05), and using an analysis of variance to adjust for medication vs. placebo group yielded a significant relationship between the A P R Q and the change in avoidance subscale (F = 4.7; df = 1, 56; p < 0.05). Thus, greater degrees of alexithymia predicted a poorer treatment response in the unmedicated, psychotherapy only group, but had no predictive power in the antidepressant treated groups. Because alexithymia may be developed secondarily to severe trauma (Krystal, 1971, 1977), we examined the relationship of the severity of these veterans' war related trauma to both the A P R Q score and to our outcome measures on the IES. The severity of war atrocity exposure was substantial with a mean score of 8.4+4.0 on the atrocity scale. While we found that the atrocity scale was significantly correlated with treatment response on both the terminal avoidance (r = 0.38;p < 0.05) and the intrusion subscales (r = 0.30; p < 0.05), the war atrocity scale was not correlated with A P R Q scores. This lack of correlation indicated that the severity of war atrocity exposure did not influence levels of alexithymia significantly. However, to adjust for the association between atrocity exposure and treatment response on the avoidance scale, a partial correlation was done between the A P R Q

Alexithymia as a Predictor of Treatment Response

569

Table I. Baseline Alexithymia and Pre/Post-treatment Impact of Events Scale Scores by Medication Group: Means (Standard Deviations) (n = 57) a Medication Group Scale Alexithymia Impact of events: Intrusion Avoidance

Pre/Post Pre Pre Post Pre Post Pre Post

Phenelzine (n = 18) 9.4 30.0 16.3 14.1 6.0 16.4 9.8

(3.6) (15.4) (11.2) (10.3) (7.0) (8.0) (6.3)

Imipramine (n = 23) 8.9 36.5 27.4 18.1 13.3 18.5 13.8

(4.1) (16.7) (16.3) (10.4) (12.4) (9.6) (7.4)

Placebo (n = 16) 9.6 32.0 30.0 16.1 14.7 16.8 15.3

(5.1) (13.9) (15.3) (8.5) (9.2) (6.6) (8.1)

p ns ** * ns

ap value for pre-post ANOVA, time by medication group interaction: ,,p < 0.05; p < 0.01.

and the avoidance scale scores. The partial correlation remained statistically significant (r = -0.27; p < 0.05). Thus, alexithymia predicted poor responses in the treatment of the avoidance symptoms of PTSD independent of severity of war trauma. Other potentially confounding variables were examined including age, sex, race, the duration of PTSD symptoms, and substance abuse. In comparing the 21 patients without a history of substance abuse to the 36 with a substance abuse history, we found no statistically significant difference (t = 1.3; df = 55; p < 0.2) in A P R Q scores (9.8+4.4) vs (8.3 _+3.2). Both age of onset of PTSD symptoms (r = +0.3; p < 0.03) and duration of PTSD symptoms (r = -0.29; p < 0.03) were significantly correlated with A P R Q scores. Men with more alexithymia developed PTSD when they were younger and had PTSD for a longer time. Since age was not correlated with alexithymia and most men went to Vietnam in their late teens/early twenties, alexithymia appeared to be associated with a shorter latency between end of combat and onset of PTSD symptoms. Although age of onset was not associated with treatment outcome, we used analysis of covariance to adjust for age of onset in the association between the A P R Q and the terminal avoidance score. The A P R Q association with outcome remained highly significant (F = 7.1; df = 2, 52; p < 0.002).

Reliability of the APRQ Test/retest reliability of the A P R Q was then examined among the 41 patients who had repeat testing, at a median of 6 weeks after the initial

Kosten et aL

570

testing. As an indication of the stability of the trait measured by the APRQ, the mean APRQ score was not significantly different at the follow-up testing with the initial score being 8.9+4.5 and the follow-up score being 9.2+__4.6 (t = 0.7; ns), Scores on the two administrations of the test also were highly correlated with a Pearson correlation coefficient of 0.79 for the total score. This test/retest correlation did not significantly differ across the three treatment groups with the correlation being 0.9 for the placebo group, 0.8 for the imipramine group, and 0.75 for the phenelzine group. Similarly, the change in alexithymia scores did not correlate with the change from baseline either in avoidance score (r = +0.1) or overall IES (r = -0.2). Thus, the APRQ showed good test/retest reliability across all three types of treatment.

DISCUSSION In this clinical trial testing pharmacotherapy to treat post-traumatic stress disorder (PTSD) we found that the Alexithymia Provoked Response Questionnaire (APRQ) was a significant predictor of treatment outcome (Frank et al., 1988; Kosten et al., 1992). The APRQ is a structured interview designed to "provoke" the patient to produce affective material by asking him to imagine himself in a variety of potentially stressful settings. Those patients who produced very little affective material during this interview tended to have a poor treatment response during this eight week clinical trial. Clinical response was assessed using the Impact of Events Scale (IES) which measures post-traumatic stress symptoms. This scale is divided into two subscales for intrusive type symptoms and for avoidance type symptoms. Most patients with PTSD have both intrusive and avoidance type symptoms, although some patients may have a preponderance of one type or the other. The avoidance symptoms include emotional numbing, distance from others, and efforts to avoid thinking about the traumatic event. High degrees of alexithymia predicted a poor response in these avoidance symptoms, but not in the intrusive symptoms. This poor treatment response of avoidance type symptoms was found particularly in the group treated with placebo rather than with active medication. Since all three treatment groups received psychotherapy in addition to the medications, the implication of this finding in the placebo group suggests that psychotherapy alone may not be effective for reducing avoidance symptoms in FTSD patients with alexithymia. In these alexithymic patients medications may therefore have a particularly critical role, in agreement with previous studies of alexithymic psychosomatic patients, who do not respond to psychodynamic

Alexithymia

as a

Predictor of Treatment Response

571

therapy and benefit from antidepressants (Krystal, 1982; Sifneas, 1975; Taylor, 1984). We also examined whether the severity of traumatic exposure or history of substance abuse would influence scores on the APRQ, since previous work had suggested that both severe trauma and substance abuse could be associated with a "secondary" alexithymia (Krystal, 1971, 1977). We found no significant associations, although atrocity exposure was associated with treatment outcome. Veterans participating in more severe atrocities had a poorer treatment response, but this poorer response was independent of the alexithymic trait. Since patients with current substance abuse were excluded, this may have accounted for the lack of association between those with a history of substance abuse and more alexithymia. In "secondary" alexithymia associated with substance abuse, the alexithymic characteristics are reduced when sustained drug abstinence is attained, as in the current sample of patients. For the purposes of the current study, however, the important conclusions were that neither severity of trauma nor history of substance abuse accounted for the association between alexithymia and treatment outcome, The finding that age of onset for PTSD symptoms was associated with alexithymia suggests that veterans with alexithymia may be at greater risk for the relatively early onset of PTSD after traumatic exposure. Since alexithymic men have an inability to express directly their emotional responses to stressful events, even at the time of the events, they may delay their responses and rapidly go on to develop a protracted PTSD syndrome. Further work will examine this course of illness issue more fully, but this study provides an interesting lead. In this study we also assessed the reliability of the APRQ in relationship to the initial study from which it was developed using a test-retest design. The APRQ was initially given before patients were randomized into treatment, and the APRQ was repeated in 41 patients at the end of treatment. Since most of these patients' PTSD symptoms improved, this provided an excellent setting for a retest study that examined whether the APRQ measured a trait characteristic in these patients. This study showed excellent test-retest reliability (r = 0.8) with no significant difference between initial and follow-up scores at six to eight weeks after initial administration. This excellent test-retest reliability held across all three treatment groups and improvement in symptoms was not correlated with any change in alexithymia score. In comparison to our previous use of this instrument, our APRQ scores were quite consistent with the outpatient PTSD patients reported in the previous study (8.7 vs 9.2 initially and 9.2 at follow-up). Thus, since our previous study had shown excellent interrater reliability, the APRQ appears to be a reliable measure for assessing alexithymia.

Kosten et aL

572

This study has several limitations in its generalizability and utility. While the associations were statistically significant, the variance in outcome explained by our alexithymia measure was modest and under 10% overall. For the placebo group the explained variance was substantially greater (almost 40%), but this was a relatively small group of patients. Replication in a larger sample treated with psychotherapy alone or perhaps with medications alone would be informative. A similar issue of generalizability arises due to the use of male Vietnam combat veterans, and replication in a nonveteran and a female sample would be important. The original studies with psychosomatic illnesses included both men and women suggesting that our findings might be replicated in women. The trauma associated with natural disasters or rape in women would provide interesting groups for replications. In future replications the APRQ alexithymia interview might also be supplemented by some of the other alexithymia assessments that have been developed (Krystal et aL, 1986; Sifneos et aL, 1977). Overall, our findings seem promising for both theoretical and practical purposes, since the APRQ is relatively easy and quick to administer and it may provide a clear indication for pharmacotherapy rather than psychodynamic therapy alone. Replication of this study in larger samples getting either pharmaco- or psychotherapy seems well indicated.

ACKNOWLEDGMENTS We thank the veterans and staff at the Connecticut readjustment counseling centers for participation in this study. Supported in part by the Veteran's Administration Research Funds and Research Scientist Award DA00112 (TRK). Presented at the American Psychiatric Association, New Research Session, New York, New York, May 1990.

REFERENCES Apfel-Savitz, R. J., Silverman, D. and Bennett, M. (1977). Group psychotherapy of patients with somatic illness and alexithymia. Psychother. Psychosom. 28: 323-329. Apfel, R. J., and Sifneos, T. E. (1979). Alexithymia: Concept in measurement. Psychother. Psychosom. 32: 180-190. Blanchard, E. B., Arena, J. G, and Pallmeyer, T. P. (1981). Psychosomatic properties of a scale to measure alexithymia. Psychother. Psychosom. 35: 64-71. Blumer, D. and Heilhronn, M. (1982). Chronic pain as a variant of depressive disease. 3. Nerv. Merit. Dis. 170: 381-406. Davidson, J., Cudler, H., Smith, R., Mahorney, S. L., et aL (1990). Treatment of posttraumatic stress disorder with amitriptylene and placebo. Arch. Gen. Psychiatry 47: 259-269. Gendorf, A., Kadushin, D., Laufer, R. S., et aL (1981). Legacies of Vietnam: Comparative Adjustments of Veterans and Their Peers, Center for Policy Research, New York.

Alexithymla as a Predictor of Treatment Response

573

Frank, J. B., Kosten, T. R., Giller, E. L, and Dan, E. (1980). A preliminary study of phenelzine and imipramine for post traumatic stress disorder. Am. J. Psychiatry 145: 1289-1291. Horowitz, M., Wilner, N., and Alvarez, W. (1972). Impact of events scale: A measure of subjective distress. Psychosom. Med. 41: 209-218. Kosten, T. R., Frank, J. B., Dan, E., McDougle, C. J., and Giller, E. L. (1992). Pharmacotherapy for post-traumatic stress disorder using phenelzine or imipramine. J. Nervous Mental Dis. 179: 366-371. Krystal, H. (1971). Trauma: Consideration of severity and chronicity. In Krystal, H., and Niederland, W. (eds.), Psychic Traumatization, Little, Brown, Boston, pp. 11-28. Krystal, H. (1977). Self-representation in the capacity for self care. Ann. PsyehoanaL 6: 209-246. Krystal, H. (1982). Alexithymia and the effectiveness of psychoanalytic treatment. Int. Z PsyehoanaL Psychother. 9: 353-378. Krystal, J. H., Giller, E. L., and Cicchetti, D. V. (1986). Assessment of alexithymia in post-traumatic stress disorder and somatic illness: Introduction of a reliable measure. Psyehosom. Med. 48: 84-94. Lolas, F., DelaParra, G., Aronsobn, S., et al., (1980). On the measurement of alexithymic behavior. Psychother. Psychosom. 33: 139-146. Marmor, J. (1979). Short term dynamic psychotherapy. Am. J. Psychiatry 136: 149-155. Pierce, M., Faryna, A., Davidson, A., Markert, R., and Krystal, J. H. (1989). A comparison of interview and self-report alexithymia measures. Psychosom. Med. 51: 254-255. Pierce, M., Krystal, J. H., Faryna, A., Markert, R., and Davidson, A. (1990). A comparison of alexithymia measures. Proe. American Psychiatric Association, New Research, New York. Sifneos, P. E. (1975). Problems of psychotherapy of patients with alexlthymic characteristics and physical disease. Psychother. Psyehosom. 26: 65-70. Sifneos, P. E., Apfel-Savitz, R., and Frankel, F. (1977). The phenomena "alexithymia": Observations in neurotic and psychosomatic patients. Psychother. Psyehosom. 28: 47-57. Taylor, G. J. (1984). Alexithymia: Concept, measurement, and implications for treatment. Am. Z Psychiatry 141: 725-732.