درمان تریکوتیلومانیا: فرا تجزیه و تحلیل اثرات درمان و ناظمان برای رفتاردرمانی و مهارکننده های بازجذب سروتونین

Few randomized controlled trials (RCTs) exist examining the efficacy of behavior therapy (BT) or serotonin reuptake inhibitors (SRIs) for the treatment of trichotillomania (TTM), with no examination of treatment moderators. The present meta-analysis synthesized the treatment effect sizes (ES) of BT and SRI relative to comparison conditions, and examined moderators of treatment. A comprehensive literature search identified 11 RCTs that met inclusion criteria. Clinical characteristics (e.g., age, comorbidity, therapeutic contact hours), outcome measures, treatment subtypes (e.g., SRI subtype, BT subtype), and ES data were extracted. The standardized mean difference of change in hair pulling severity was the outcome measure. A random effects meta-analysis found a large pooled ES for BT (ES = 1.41, p < 0.001). BT trials with greater therapeutic contact hours exhibited larger ES (p = 0.009). Additionally, BT trials that used mood enhanced therapeutic techniques exhibited greater ES relative to trials including only traditional BT components (p = 0.004). For SRI trials, a random effects meta-analysis identified a moderate pooled ES (ES = 0.41, p = 0.02). Although clomipramine exhibited larger ES relative to selective serotonin reuptake inhibitors, the difference was not statistically significant. Publication bias was not identified for either treatment. BT yields large treatment effects for TTM, with further examination needed to disentangle confounded treatment moderators. SRI trials exhibited a moderate pooled ES, with no treatment moderators identified. Sensitivity analyses highlighted the need for further RCTs of SRIs, especially among youth with TTM.

Hair pulling behaviors are common in the general population (Duke et al., 2009), with between 1% and 3% of adults reporting clinically significant hair pulling (Christenson et al., 1991b). Hair pulling disorder, commonly referred to as trichotillomania (TTM), is characterized by excessive hair pulling that can be automatic (e.g., outside of awareness) or focused (e.g., consciously pulled) in nature (American Psychiatric Association, 2013 and Stein et al., 2010). Individuals with TTM frequently experience co-occurring anxiety disorders, depressive disorders, and other body-focused repetitive behaviors (Duke et al., 2010 and Panza et al., 2013). Hair pulling behaviors can result in detrimental physical complications (Bouwer and Stein, 1998), psychosocial impairment (Diefenbach et al., 2005b, Stemberger et al., 2000, Wetterneck et al., 2006 and Woods et al., 2006a), and poor quality of life (Diefenbach et al., 2005b and Odlaug et al., 2010). In light of these deleterious physical and psychological consequences, effective and efficient treatments are needed. Several therapeutic approaches have been investigated to treat TTM symptoms (Franklin et al., 2011b), including behavior therapy (BT) and psychiatric medications (Franklin et al., 2008 and Woods et al., 2006a). Behavioral therapies such as habit reversal training (HRT) have demonstrated efficacy reducing hair pulling severity across several randomized controlled trials (RCTs) (Azrin et al., 1980, Diefenbach et al., 2006, Franklin et al., 2011a, Ninan et al., 2000 and van Minnen et al., 2003), with more recent BT trials including components of acceptance and commitment therapy (ACT) (Woods et al., 2006b), and dialectical behavior therapy (DBT) (Keuthen et al., 2012). Acute treatment gains obtained from BT have been generally maintained up to six months (Franklin et al., 2011a, Keuthen et al., 2011 and Woods et al., 2006b). Indeed, BT is recommended as a first-line treatment for youth and adults with TTM (Flessner et al., 2010). Despite noted efficacy, RCTs of BT have had relatively small sample sizes, with no examination of treatment moderators. Aside from BT, two types of psychiatric medications (antipsychotics and antidepressants) have demonstrated mixed efficacy in reducing hair pulling. While antipsychotic medications (e.g., olanzapine, aripiprazole) have demonstrated efficacy in open-label trials (Stewart and Nejtek, 2003 and White and Koran, 2011), only one RCT has evaluated the efficacy of olanzapine and identified its therapeutic benefit relative to placebo (Van Ameringen et al., 2010). Comparatively, antidepressant medications (e.g., clomipramine, fluoxetine, sertraline) remain the most frequently used treatment for individuals with TTM (Franklin et al., 2008 and Woods et al., 2006a), but have mixed evidence across RCTs (Christenson et al., 1991a, Dougherty et al., 2006, Ninan et al., 2000, Streichenwein and Thornby, 1995, Swedo et al., 1989 and van Minnen et al., 2003), with some evidence of long-term therapeutic benefit (Swedo et al., 1993). These medications share the commonality of inhibiting the reuptake of serotonin (referred to as serotonin reuptake inhibitors, SRIs) leading to the belief that deficiencies of serotonin may underlie hair pulling behaviors (Ferrão et al., 2009 and Mancini et al., 2009). Despite mixed efficacy, side effect profiles, and potential high relapse rates following discontinuation (Iancu et al., 1996 and Pollard et al., 1991); SRI medications are frequently used (Franklin et al., 2008 and Woods et al., 2006a) and recommended by experts as pharmacological treatment options (Chamberlain et al., 2007). More recently, N-acetylcysteine (i.e., an over-the-counter amino acid supplement that acts as a glutamate modulator; NAC) has been evaluated in the treatment of individuals with TTM. While demonstrating efficacy in an RCT of adults with TTM ( Grant et al., 2009), no significant benefit was found relative to placebo in a sample of youth with TTM ( Bloch et al., 2013). Although offering promise for some individuals ( Woods, 2013), the small number of RCTs limits inferences about NAC's efficacy for TTM. When making treatment recommendations, it is important to synthesize empirical evidence to guide clinical decisions (Murad and Montori, 2013). Relative to literature reviews and expert recommendations, meta-analyses provide a quantitative synthesis of treatment trials, and allow for the examination of moderators of treatment effects. Presently, only one meta-analysis has examined the efficacy of behavioral and pharmacological treatments for reducing hair pulling behaviors among individuals with TTM (Bloch et al., 2007). Bloch et al. (2007) found a large treatment effect for HRT [Standardized Mean Difference (SMD) = 1.14, 95% Confidence Interval (CI): −1.89, −0.38] compared to control conditions. Clomipramine (CMI) was found to be superior to comparison conditions (SMD = 0.68, 95% CI: −1.28, −0.07), with no significant effect found for selective serotonin reuptake inhibitor medications (SSRIs) relative to placebo (SMD = 0.02, 95% CI: −0.32, 0.35). While beneficial as the first quantitative synthesis of this area, this meta-analysis has several limitations in the present-day context (Bloch et al., 2007). First, since its publication, several additional RCTs of BT for TTM have been published (Diefenbach et al., 2006, Franklin et al., 2011a and Keuthen et al., 2012). Second, this meta-analysis relied upon treatment-blind ratings to limit reporting bias. Although a notable strength, this resulted in the use of some outcome measures that had limited psychometric evaluation (e.g., video-taped hair loss ratings), and interchanged measures that evaluated symptom severity and impairment rather than focusing solely on the construct of symptom severity (Bloch et al., 2007). Finally, this meta-analysis separated treatment effects between CMI and SSRI medications. Although CMI can impact other neurotransmitters (e.g., norepinephrine reuptake inhibitor), it serves as a strong post synaptic serotonin reuptake inhibitor. As CMI and SSRIs both strongly affect serotonin receptors, there is some benefit to examining their collective efficacy, as well as their individual efficacy. In an effort to address these limitations, this meta-analysis examined the efficacy of evidence-based treatments for individuals with TTM. Although antipsychotics and NAC have emerging empirical support, these treatments were considered too preliminary for inclusion due to the limited number of published RCTs. Thus, this meta-analysis examined the efficacy of BT and SRI treatment to reduce hair pulling severity among individuals with TTM. Additionally, clinically-relevant treatment moderators were examined that included: participant age; percentage of co-occurring anxiety and depressive disorders; outcome measure informant; average number of 1-h therapy sessions (for BT trials); study methodology; and intervention subtypes.

3.1. Included studies and study characteristics Initial search strategies produced 433 potential abstracts/citations, with 19 abstracts/citations being retrieved for a detailed review (see Fig. 1). Table 1 displays the 11 RCTs that met inclusion criteria that allowed for comparisons of seven BT trials (n = 182 participants) and six SRI trials (n = 118 participants). Full-size image (38 K) Fig. 1. Study selection and rationale for inclusion and exclusion. Figure options Table 1. Study characteristics for trials included in meta-analyses. Study Active Tx Control condition N Mean age Youth or adult sample Comorbid anxiety disorders Comorbid depressive disorders Outcome measure Self vs. clinician report Approx. contact hours Study quality Effect size Behavior therapy Azrin et al. (1980)a BT MNP 32 27 Mixed NR NR Number of daily hair pulling episodes Self-report 2 23 0.66 Ninan et al. (2000)b BT WL 10 33 Adult NR NR NIMH-TSS Clinician-report 9 19 1.94 van Minnen et al. (2003)b BT PLBO 29 31 Mixed 3% 17% MGH-HPS Self-report 6 36 1.28 Diefenbach et al. (2006) BT SP 24 40 Adult 25% 21% MGH-HPS Self-report 8 30 0.56 Woods et al. (2006b) BT WL 25 35 Adult 4% 32% MGH-HPS Self-report 10 28 2.14 Franklin et al. (2011a) BT MAC 24 13 Youth 33% 13% NIMH-TSS Clinician-report 8 35 1.38 Keuthen et al. (2012) BT MAC 38 31 Adult 11% 0% NIMH-TSS Clinician-report 11 29 2.35 Serotonin reuptake inhibitors Swedo et al. (1989) CMI DES 13 32 Mixed 8% 0% NIMH-TSS Clinician-report N/A 28 0.75 Christenson et al. (1991a) FLUX PLBO 16 32 Adult 13% 19% Weekly rating of hair pulling severity Self-report N/A 25 0.14 Streichenwein and Thornby (1995) FLUX PLBO 16 39 Adult 31% 6% Weekly severity of hair pulling episodes Self-report N/A 23 0.35 Ninan et al. (2000)b CMI PLBO 11 33 Adult NR NR NIMH-TSS Clinician-report N/A 20 0.61 van Minnen et al. (2003)b FLUX PLBO 26 32 Mixed 0% 19% MGH-HPS Self-report N/A 38 0.54 Dougherty et al. (2006) SERT PLBO 36 29 Adult 30% 16% PITS Clinician-report N/A 27 0.17 Note: CMI = clomipramine, DES = desipramine, FLUX = fluoxetine, SERT = sertraline, PLBO = placebo, BT = behavior therapy, MNP = Mass Negative Practice, WL = waitlist, SP = Supportive psychotherapy, MAC = Minimal Attention Control, NR = not reported, NIMH-TSS = National Institute of Mental Health Trichotillomania Severity Scale, MGH-HPS = Massachusetts General Hospital Hair pulling Scale, PITS = Psychiatric Institute Trichotillomania Scale, N/A = not applicable. a Used LOCF for participants without post-treatment assessment. Excluded two participants who did not use the same outcome metric. b PLBO group double counted for van Minnen et al. (2003) and Ninan et al. (2000) due to small sample size of PLBO group. Table options 3.2. Reliability of coding study characteristics There was excellent inter-rater agreement between the three raters on categorical and continuous study characteristics (100% agreement), as well as overall study methodological quality (ICC = 0.99, 95% CI: 0.98, 0.99). 3.3. Treatment effects of BT As seen in Fig. 2, a random effects meta-analysis identified a large treatment effect of BT compared to control conditions (SMD = 1.41, 95% CI: 0.87, 1.96, z = 5.07, p < 0.001). Visual inspection of the forest plot, Q statistic, and I2 statistic identified the presence of significant heterogeneity among ES across trials [Q(6) = 15.61, p = 0.02, I2 = 61.57%]. Given that Azrin et al. (1980) did not use a standardized rating scale, the summary effect was re-calculated with this trial excluded to examine the effects of BT for TTM with only standardized rating scales. Results identified a large treatment effect (SMD = 1.56, 95% CI: 0.99, 2.14, z = 5.37, p < 0.001), with less heterogeneity [Q(5) = 11.21, p = 0.05, I2 = 55.41%]. Given that Diefenbach et al. (2006) employed a group therapy format, the summary effect was re-calculated with this trial excluded to examine the effects of individual BT for TTM. Results identified a large treatment effect (SMD = 1.56, 95% CI: 1.00, 2.13, z = 5.43, p < 0.001), with less heterogeneity [Q(5) = 11.47, p = 0.04, I2 = 56.47%]. Full-size image (27 K) Fig. 2. Efficacy of behavior therapy relative to control conditions for the treatment of hair pulling severity. Figure options 3.4. Moderators of BT Table 2 presents results for moderator analyses. First, no significant association was observed between mean participant age and ES (p = 0.98). Furthermore, there were no significant difference among BT trials that included only adults (SMD = 1.71), only children (SMD = 1.39), or included both youth and adults (SMD = 0.94). Second, there was no significant association between ES and the percentage of study participants with either co-occurring anxiety disorders (p = 0.33) or depressive disorders (p = 0.61). Third, trials using treatment-blind clinician-raters (SMD = 1.90) had greater ES relative to unblind self-report raters (SMD = 1.11); however, this difference was not statistically significant (p = 0.09). Fourth, there was a significant association between number of therapeutic contact hours and ES (p < 0.01). Fifth, there was no significant association between study methodological quality and ES (p = 0.94). Finally, when examining BT treatment subtypes, a significant difference emerged between trials that used core BT (SMD = 1.02) compared to trials that used mood-enhanced BT (SMD = 2.26, p = 0.002). Table 2. Regression analyses and analog to ANOVA examining moderators of treatment effects in BT and SRI trials. Study characteristics BT trials (n = 7) SRI trials (n = 6) B SE z p B SE z p Average participant age −0.001 0.04 −0.02 0.98 0.02 0.06 0.26 0.79 Percentage of comorbid anxiety disorders −0.03 0.03 −0.97 0.33 −0.01 0.01 −0.87 0.39 Percentage of comorbid depressive disorders −0.02 0.03 −0.51 0.61 −0.02 0.02 −0.84 0.40 Average number of therapeutic contact hours 0.17 0.07 2.58 0.01 NA NA NA NA Study methodological quality rating −0.004 0.06 −0.08 0.94 0.01 0.03 0.09 0.92 Q (df) Q (df) Youth, adult, or mixed participants 2.03 2 0.36 1.13 1 0.29 Clinician vs. self-report ratings 2.93 1 0.09 0.03 1 0.87 Therapy subtype comparisons 10.01 1 0.002 1.09 1 0.30 Control condition comparison 2.09 2 0.35 0.90 1 0.34 Note: BT = behavior therapy, SRI = serotonin reuptake inhibitors, NA = not applicable. Table options 3.5. Publication bias and sensitivity analyses for BT trials Although visual inspection of the funnel plot suggested publication bias may exist, Eggers' test for bias indicated that publication bias was not significant (t = 1.31, p = 0.25). Duval and Tweedie's trim-and-fill method trimmed no studies, and a large treatment effect remained (SMD = 1.41). Rosenthal's and Orwin's Fail-safe N calculations identified that at least 115 and 25 unretrieved studies with an ES of zero were needed to reduce the summary ES of BT to a non-significant and/or trivial effect, respectively. Finally, no significant difference was identified across comparison conditions for WL trials (SMD = 2.14), PLBO trials (SMD = 1.43), and AC trials (SMD = 1.23, see Table 2). 3.6. Treatment effects of SRIs As seen in Fig. 3, a random effects meta-analysis identified a moderate summary effect of SRI medications relative to control conditions (SMD = 0.41, 95% CI: 0.06, 0.75, z = 2.29, p = 0.02). Visual inspection of the forest plot, Q statistic, and I2 indicated minimal heterogeneity among ES across trials [Q(5) = 1.74, p = 0.88, I2 = 0%]. Given that two of the six RCTs used unstandardized outcome measures ( Christenson et al., 1991a and Streichenwein and Thornby, 1995), the summary effect was re-calculated with these two trials excluded to examine the effects of SRIs for TTM using psychometrically validated rating scales. Results identified a moderate treatment effect (SMD = 0.46, 95% CI: 0.06, 0.86, z = 2.24, p = 0.03), with little heterogeneity [Q(3) = 1.38, p = 0.71, I2 = 0%]. Full-size image (28 K) Fig. 3. Efficacy of serotonin reuptake inhibitors compared to control conditions for the treatment of hair pulling severity. Figure options 3.7. Moderators of SRIs Table 2 presents results for moderator analyses. First, no significant association was observed between mean participant age and ES (p = 0.79). Furthermore, there were no significant differences between SRI trials that included only adults (SMD = 0.26) relative to SRI trials that included both older adolescents and adults (SMD = 0.65). Second, there was no significant association between ES and the percentage of study participants with co-occurring anxiety disorders (p = 0.39) or depressive disorders (p = 0.40). Third, no significant difference was identified between trials using treatment blind clinician-raters (SMD = 0.43) relative to unblind self-report informants (SMD = 0.37). Fourth, there was no significant association between study methodological quality and ES (p = 0.92). Finally, when specific SRI medication classes were examined, CMI trials (SMD = 0.71) exhibited greater ES relative to SSRI trials (SMD = 0.29); however, the difference between medication classes was not statistically significant (p = 0.30). 3.8. Publication bias and sensitivity analyses for SRI trials Visual inspection of the funnel plot and Egger's test for bias both indicated that publication bias was not significant (t = 0.46, p = 0.67). Duval and Tweedie's trim-and-fill method trimmed no studies, and a moderate significant effect remained (SMD = 0.41). Rosenthal's and Orwin's Fail-safe N calculations similarly identified that only three unretrieved studies with an ES of zero were needed to reduce the summary ES of SRI to a non-significant and/or trivial effect respectively. Finally, no significant difference was found across comparison conditions for PLBO trials (SMD = 0.32), and AC trials (SMD = 0.75, see Table 2).