Abstract
Objectives: In clinical trials of pediatric trichotillomania (TTM), three instruments are typically employed to rate TTM severity: (1) the Massachusetts General Hospital Hair Pulling Scale (MGH-HPS), (2) the National Institute of Mental Health Trichotillomania Severity Scale (NIMH-TSS), and (3) the Trichotillomania Scale for Children (TSC). These instruments lack standardized definitions of treatment response, which lead researchers to determine their own definitions of response post hoc and potentially inflate results. We performed a meta-analysis to provide empirically determined accuracy measures for percentage reduction cut points in these three instruments.
Methods: MEDLINE was searched for TTM clinical trials. A total of 67 studies were initially identified, but only 5 were clinical trials focused on TTM in pediatric populations and therefore were included in this meta-analysis (n = 180). A Clinical Global Impressions Improvement score ≤2 was used to define clinical response. Receiver operating characteristic principles were employed to determine accuracy measures for percentage reduction cut points on each one of the instruments. Meta-DiSc software was employed to provide pooled accuracy measures for each cut point for each instrument. The Youden Index and the distance to corner methods were used to determine the optimal cut point.
Results: The optimal cut points to determine treatment response were a 45% reduction on the MGH-HPS (Youden Index 0.40, distance to corner 0.20), a 35% reduction on the NIMH-TSS (Youden Index 0.42, distance to corner 0.17), a 25% reduction on the TSC child version (TSC-C; Youden Index 0.40, distance to corner 0.18), and a 45% (distance to corner 0.30) or 50% reduction (Youden Index 0.33) on the TSC parent version (TSC-P). The TSC-C had less discriminative ability at determining response in younger children in comparison to older children; no age-related differences were observed on the TSC-P.
Conclusions: This study provides empirically determined cut points of treatment response on three instruments that rate TTM severity. These data-driven cut points will benefit future research on pediatric TTM.
Keywords: trichotillomania, children and adolescents, clinical trials, meta-analysis, Massachusetts General Hospital Hair Pulling Scale (MGH-HPS), National Institute of Mental Health Trichotillomania Severity Scale (NIMH-TSS), Trichotillomania Scale for Children (TSC)
Introduction
Trichotillomania (TTM) (or hair pulling disorder) is a psychiatric condition characterized by pulling out of one's own hair, which results in noticeable hair loss (American Psychiatric Association 2013). TTM typically develops during early adolescence and then progresses into adulthood with a chronic course of waxing-and-waning symptom severity throughout the lifetime (Bloch 2009), so early diagnosis is critical. Studies of college students suggest that between 1% and 3.5% of the general population engage in hair pulling that leads to significant distress (Christenson et al. 1991b; Odlaug and Grant 2010). Children with TTM report that their hair pulling interferes moderately with their social life and academic performance (Franklin et al. 2008), as peer teasing is common and leads to avoidance of daily activities. Children with TTM frequently present with comorbid anxiety and depressive disorders (Lewin et al. 2009) and there is some evidence that these comorbidities develop after the onset of TTM, highlighting the importance of early intervention for TTM in childhood (Woods et al. 2006; Franklin et al. 2008).
Currently, there is no sufficient evidence to support any intervention as a first-line treatment of pediatric TTM as there are only four randomized, controlled clinical trials (RCTs). Of these, two evaluated behavioral therapy with habit reversal training (HRT) procedures (Franklin et al. 2011; Rahman et al. 2017), one evaluated behavioral therapy augmented with response inhibition training (Lee et al. 2018), and one evaluated a pharmacological agent—N-acetylcysteine (NAC), a glutamatergic modulator (Bloch et al. 2013). Findings indicated that HRT was more efficacious than the control condition, whereas NAC did not separate from placebo. These results are in line with those of RCTs for adult TTM, as meta-analyses suggest behavioral therapy is the first-line intervention for the management of TTM, while there is no clear first-line pharmacological treatment (Bloch et al. 2007; McGuire et al. 2014).
Additional clinical trials are warranted to identify efficacious therapeutic options for pediatric TTM. RCTs are considered the optimal study design to evaluate the efficacy and safety of novel interventions in medicine, including psychiatry (Emsley and Hawkridge 2009; Fleischhacker and Goodwin 2009). Yet clinical trials for pediatric TTM face several important issues, including that rating scales of treatment efficacy often have not been specifically validated in children and lack standard definitions of treatment response. The lack of standard definitions of response to treatment is a significant issue for at least three reasons. First, it unintentionally encourages researchers to develop their own definitions of treatment response through post hoc analysis, which can inflate the results. Second, it creates challenges for comparing results across different trials. Third, it precludes assessment of clinical benefit as statistically significant differences between active and control groups do not necessarily correspond to individuals actually experiencing clinical improvement with the proposed intervention (Kraemer and Kupfer 2006).
We have recently addressed the lack of empirical definitions of response to treatment on two instruments typically employed to measure TTM symptom severity in RCTs of adults—the Massachusetts General Hospital Hair Pulling Scale (MGH-HPS) and the National Institute of Mental Health Trichotillomania Severity Scale (NIMH-TSS) (Houghton et al. 2015; Farhat et al. 2019). The MGH-HPS is a 7-item self-report, while the NIMH-TSS is a 6-item clinician-rated instrument. Our study showed that a 35% or 7-point reduction on the MGH-HPS or a 50% or 6-point reduction on the NIMH-TSS were considered optimal cut points to classify an individual as a responder to the treatment in RCTs for TTM. Importantly, our results also showed that both the MGH-HPS and the NIMH-TSS had greater discriminative ability to determine response when pediatric samples were excluded from the analyses. Our study was limited considerably by the paucity of pediatric trials and separate analyses of exclusively pediatric populations were not carried out due to the small sample size available at the time. Yet, the results suggest that both the MGH-HPS and the NIMH-TSS might not be appropriate to measure response to treatment in RCTs for pediatric TTM, and separate analyses, including exclusively pediatric populations, are still warranted.
Although the MGH-HPS has been widely employed as a self-report instrument to evaluate TTM severity in clinical trials for pediatric TTM, significant issues should be considered. First, the MGH-HPS has never been validated in pediatric TTM populations, and some of its items feature language that is not appropriate for children and adolescents. In addition, there is no parent-report version of the MGH-HPS, so parents are not able to provide their perspective on their children's hair pulling behavior on this instrument. The self- and parent-report Trichotillomania Scale for Children (TSC) was developed to address these limitations of the MGH-HPS to rate symptom severity of pediatric TTM (Tolin et al. 2008). The TSC comprised 12 items divided into 2 subscales—severity and distress/impairment; the TSC has shown strong psychometric properties and, importantly, it can be completed by children and/or their parents, using the child version (TSC-C) and/or parent version (TSC-P). Indeed, the inclusion of parent- and child-reported TTM severity is consistent with guidelines for an evidence-based assessment in youth with TTM (McGuire et al. 2012).
In this meta-analysis, we present data that address these gaps in the literature of pediatric TTM. The objective of this meta-analysis is to determine the discriminative ability of the MGH-HPS, NIMH-TSS, TSC-C, and TSC-P to determine response to treatment in clinical trials for pediatric TTM. A Clinical Global Impressions Improvement (CGI-I) (Busner and Targum 2007) score of either 2 (“much improved”) or 1 (“very much improved”) was determined as the criterion measure to consider individuals as responders to treatments. In addition, as younger children may struggle to recognize hair pulling behavior, we performed separate analyses for younger and older children to examine sensitivity of these benchmarks of treatment response.
Methods
Search strategy for identification of studies and study selection
On February 2, 2019, MEDLINE was searched with the search strategy Trichotillomania OR “hair pulling disorder.” The search was limited to clinical trials. Language restrictions were not applied. Meta-analyses, systematic reviews, and review articles were carefully read to look for other potentially eligible articles or unpublished research. No further efforts were made to look for unpublished research. Titles and abstracts of references obtained by this search strategy were analyzed by two independent reviewers (L.C.F. and M.H.B.) to determine if they met the inclusion criteria required for our meta-analysis. To be included, studies had to (1) include a patient population with a primary diagnosis of TTM, (2) be a clinical trial, either an RCT or an open-label (OL) trial, and (3) include exclusively a pediatric sample (age range 4–17 years of age). A few TTM trials with a predominantly adult population also enrolled some adolescents—usually 15–17 years of age. In a separate recent analysis of adult TTM trials, we attempted to contact these authors, but we were unable to attain these data (described in Farhat et al. 2019), and therefore it was considered unavailable. Any discrepancies between the two independent reviewers were resolved through discussing and reading the full text of the article.
An e-mail was sent to either the corresponding or the senior author of articles fulfilling the inclusion criteria for our meta-analysis. In the e-mail, de-identified individual participant data were requested; specifically, data requested included, for each participant, (1) age, (2) gender, (3) CGI-I score posttreatment, and (4) TTM severity scores pretreatment and posttreatment. We did not adopt a hierarchy of preferred TTM severity outcomes; authors could send pretreatment and posttreatment scores from any of the following instruments: MGH-HPS, NIMH-TSS, TSC-C, and/or TSC-P. To be definitively included in the meta-analysis, authors had to provide at least the CGI-I posttreatment score as well as pretreatment and posttreatment scores on at least one instrument to measure TTM severity for each individual included in their studies.
Meta-analytic and receiver operating characteristic procedures
A similar analytic approach was employed as previously described by our group (Farhat et al. 2019). Analyses were initially carried out separately for each trial included in the meta-analysis. Individuals were classified as responders to treatment if they presented a CGI-I score of either 2 (“much improved”) or 1 (“very much improved”). Afterward, the percentage reduction in the score of TTM severity from pretreatment to posttreatment was calculated for each individual for each available instrument. Cut points for percentage reduction were then established from 5% to 70% at intervals of 5%. For each cut point, true positives, false positives, true negatives, and false negatives were calculated; these values were then employed to calculate sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) for each cut point, for each instrument.
Meta-DiSc software (Zamora et al. 2006) was employed to meta-analyze sensitivity, specificity, PLR, and NLR across the studies for each cut point. Meta-DiSc computes accuracy estimates from each study and provides pooled measures with confidence intervals. Heterogeneity was evaluated through a chi-square comparison. If heterogeneity was significant (p < 0.05), a random-effects model was employed; otherwise, a fixed-effects model was employed.
Both the Youden Index and distance to the corner methods were employed to determine the optimal cut point to dichotomize individuals into responders and nonresponders for each TTM severity instrument. The Youden Index is an approach commonly employed to maximize both sensitivity and specificity, and is calculated by summing the sensitivity and the specificity, and then subtracting one from the result. The distance to the corner is defined by the Euclidean distance between the receiver operating characteristic (ROC) curve and the (0,1) point (Unal 2017). Optimal cut points should maximize the Youden index value and minimize the distance to the corner.
For sensitivity analyses, we performed the same analytical procedures described above separately for younger and older children. Given previous research indicates younger children are more likely to experience automatic pulling in comparison to older children, which are more likely to experience focused pulling (Keuthen et al. 2008), we divided our entire sample in two age groups: 8–13 and 14–17 years of age. Separate age-related analyses for the MGH-HPS were not carried out as there were only two studies that provided MGH-HPS data, and the resultant sample size of each age group would be too small.
Results
Characteristics of trials
Supplementary Figure S1 illustrates the selection of studies for inclusion in the meta-analysis. The search strategy initially yielded 67 articles as potentially eligible. We were able to include data from all five clinical trials of pediatric TTM published to date (Tolin et al. 2007; Franklin et al. 2011; Bloch et al. 2013; Rahman et al. 2017; Lee et al. 2018). Of these five studies, four were RCTs and one was OL. The interventions examined in these studies were HRT (n = 3), computerized response inhibition training (n = 1), and NAC (n = 1). Although we did not specifically look for additional unpublished studies, we were also able to include unpublished data from an RCT that compared HRT against supportive therapy for the treatment of pediatric TTM (NCT00917098). Table 1 illustrates the characteristics of all six trials included in this meta-analysis. In total, data from 210 participants were included in the analyses and the majority of participants were female (80.99%).
Table 1.
Characteristics of Included Trials
| ID | Active | Max dose (mg/day) | Control | Duration (weeks) | Sample size | Age, mean (standard deviation) | Females (%) | TTM outcomes |
|---|---|---|---|---|---|---|---|---|
| Bloch et al. (2013) | NAC | 2400 | PCB | 12 | 35 | PCB 13.1 (3.1) | 85.71 | MGH-HPS |
| TSC-C/TSC-P | ||||||||
| NIMH-TSS | ||||||||
| NAC 14 (2.4) | ||||||||
| Franklin et al. (2011) | HRT | — | MAC | 8 | 32 | 12.5 (2.7) | 83.60 | TSC-C/TSC-P |
| Lee et al. (2018) | RIT | — | WL | 4 | 20 | RIT 12.91 (2.39) | 75 | TSC-C/TSC-P |
| NIMH-TSS | ||||||||
| WL 13.11 (2.52) | ||||||||
| Rahman et al. (2017) | HRT | — | TAU | 8 | 40 | 11.85 (3.14) | 83.33 | MGH-HPS |
| Tolin et al. (2007) | HRT | — | — | 8 | 22 | 12.6 (3.0) | 63.63 | TSC-C/TSC-P |
| NIMH-TSS | ||||||||
| NCT00917098 | HRT | — | ST | 61 | 13.49 (2.45) | 83.6 | TSC-C/TSC-P | |
| NIMH-TSS |
HRT, Behavioral Therapy with Habit Reversal Training Procedures; ID, identification; MAC, minimal attention control; MGH-HPS, Massachusetts General Hospital Hair Pulling Scale; NAC, N-acetylcysteine; NIMH-TSS, National Institute of Mental Health Trichotillomania Severity Scale; PCB, placebo; RIT, Response Inhibition Training; ST, supportive therapy; TAU, treatment as usual; TTM, trichotillomania; TSC-C, Trichotillomania Scale for Children Child Version; TSC-P, Trichotillomania Scale for Children Parent Version; WL, wait list.
MGH-HPS and NIMH-TSS
Figure 1 depicts the ROC curves for the analyses of the MGH-HPS (Fig. 1A) and NIMH-TSS (Fig. 1B) percentage reduction cut points. The first panel of Figure 1 features the ROC curve depicting the results from the overall analysis, including data from all children. For the MGH-HPS, when data from all children (n = 75) were included in the analysis, the area under the curve (AUC) for the ROC curve was 0.68, indicating that the MGH-HPS had a fair ability to determine treatment response as measured by a CGI-I ≤ 2. The second panel of Figure 1 features three ROC curves; one of the curves depicts the results from the overall analysis, including data from all children, while the remaining two depict results from the separate analyses for children 8–13 and 14–17 years of age. For the NIMH-TSS, when data from all children (n = 126) were included in the analysis, the AUC was 0.75, indicating that the NIMH-TSS had a fair ability to determine treatment response as measured by a CGI-I ≤ 2. When data from children 8–13 (n = 65) and 14–17 (n = 61) years of age were analyzed separately, the AUCs for the ROC curves were 0.72 and 0.76, respectively, indicating that the NIMH-TSS retained its fair discriminative ability to determine treatment response among both age groups.
FIG. 1.
Receiver operating curves for the MGH-HPS and NIMH-TSS. (A, B) Depict the ROC curves of sensitivity versus one-specificity for the MGH-HPS and NIMH-TSS percentage reduction cut points, respectively. For the MGH-HPS when all children were included, the AUC for the ROC curve was 0.68, indicating that the MGH-HPS has a fair ability to determine response to treatment as determined by a CGI-I ≤ 2. For the NIMH-TSS when all children were included, the AUC for the ROC curve was 0.75, also indicating that the NIMH-TSS has a fair ability to determine response to treatment as determined by a CGI-I ≤ 2. AUC, area under the curve; CGI-I, Clinical Global Impressions Improvement; MGH-HPS, Massachusetts General Hospital Hair Pulling Scale; NIMH-TSS, National Institute of Mental Health Trichotillomania Severity Scale; ROC, receiver operating characteristic.
Tables 2 and 3 show pooled accuracy measures and parameters to determine the optimal percentage reduction cut point for each cut point on the MGH-HPS (Table 2) and NIMH-TSS (Table 3) when data from all children were included in the analyses. For the MGH-HPS, a 45% reduction on the scores from pretreatment to posttreatment was considered the optimal cut point to determine treatment response, as measured by a CGI-I ≤ 2, by both the Youden Index (0.37) and the distance to corner (0.20) methods. This 45% reduction cut point on the MGH-HPS had a sensitivity of 0.67 (95% CI [0.35–0.90]), a specificity of 0.70 (95% CI [0.57–0.81]), a PLR of 2.31 (95% CI [0.39–13.70]), and an NLR of 0.57 (95% CI [0.11–3.09]). For the NIMH-TSS, a 35% reduction on the scores from pretreatment to posttreatment was considered the optimal cut point to determine treatment response, as measured by a CGI-I ≤ 2, by both the Youden Index (0.42) and the distance to corner (0.17) methods. This 35% reduction cut point on the NIMH-TSS had a sensitivity of 0.65 (95% CI [050–0.78]), a specificity of 0.77 (95% CI [0.66–0.86]), a PLR of 2.18 (95% CI [1.24–3.83]), and an NLR of 0.55 (95% CI [0.25–1.20]).
Table 2.
Accuracy Measures, Youden Index, and Distance to Corner for Cut Points for the Massachusetts General Hospital Hair Pulling Scale
| Cut point, % | Sensitivity (95% CI) | Specificity (95% CI) | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Youden | Distance |
|---|---|---|---|---|---|---|
| 5 | 0.83 (0.52–0.98) | 0.41 (0.29–0.54) | 1.54 (0.62–3.80) | 0.49 (0.06–3.83) | 0.24 | 0.38 |
| 10 | 0.75 (0.43–0.95) | 0.43 (0.31–0.56) | 1.24 (0.19–8.19) | 0.69 (0.02–25.32) | 0.18 | 0.39 |
| 15 | 0.75 (0.43–0.95) | 0.49 (0.36–0.62) | 1.40 (0.22–8.77) | 0.59 (0.02–16.88) | 0.24 | 0.32 |
| 20 | 0.75 (0.43–0.95) | 0.52 (0.39–0.65) | 1.48 (0.27–8.25) | 0.56 (0.02–12.90) | 0.27 | 0.29 |
| 25 | 0.75 (0.43–0.95) | 0.56 (0.43–0.68) | 1.61 (0.27–9.57) | 0.52 (0.02–12.51) | 0.31 | 0.26 |
| 30 | 0.67 (0.35–0.90) | 0.62 (0.49–0.74) | 1.78 (0.36–8.67) | 0.64 (0.11–3.69) | 0.29 | 0.25 |
| 35 | 0.67 (0.35–0.90) | 0.62 (0.49–0.74) | 1.78 (0.36–8.67) | 0.64 (0.11–3.69) | 0.29 | 0.25 |
| 40 | 0.67 (0.35–0.90) | 0.65 (0.52–0.77) | 1.97 (0.36–10.79) | 0.61 (0.11–3.51) | 0.32 | 0.23 |
| 45 | 0.67 (0.35–0.90) | 0.70 (0.57–0.81) | 2.31 (0.39–13.70) | 0.57 (0.11–3.09) | 0.37 | 0.20 |
| 50 | 0.58 (0.28–0.85) | 0.71 (0.59–0.82) | 1.75 (0.06–47.46) | 0.63 (0.04–10.20) | 0.29 | 0.26 |
| 55 | 0.50 (0.21–0.79) | 0.79 (0.67–0.89) | 2.84 (0.07–123.7) | 0.69 (0.11–4.5) | 0.29 | 0.29 |
| 60 | 0.50 (0.21–0.79) | 0.81 (0.69–0.90) | 2.97 (0.08–118.2) | 0.68 (0.11–4.18) | 0.31 | 0.29 |
| 65 | 0.33 (0.1–0.65) | 0.83 (0.71–0.91) | 2.55 (0.10–62.6) | 0.85 (0.37–1.99) | 0.16 | 0.48 |
| 70 | 0.33 (0.1–0.65) | 0.84 (0.73–0.92) | 2.70 (0.12–59.36) | 0.84 (0.38–1.89) | 0.17 | 0.47 |
Bold indicates optimal cut point.
Table 3.
Accuracy Measures, Youden Index, and Distance to Corner for Cut Points for the National Institute of Mental Health Trichotillomania Severity Scale
| Cut point, % | Sensitivity (95% CI) | Specificity (95% CI) | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Youden | Distance |
|---|---|---|---|---|---|---|
| 5 | 0.81 (0.67–0.91) | 0.41 (0.30–0.53) | 1.35 (1.02–1.78) | 0.46 (0.1–2.13) | 0.22 | 0.38 |
| 10 | 0.81 (0.67–0.91) | 0.47 (0.36–0.59) | 1.46 (1.13–1.90) | 0.4 (0.09–1.78) | 0.28 | 0.32 |
| 15 | 0.77 (0.63–0.88) | 0.54 (0.42–0.65) | 1.51 (1.05–2.16) | 0.48 (0.15–1.49) | 0.31 | 0.27 |
| 20 | 0.77 (0.63–0.88) | 0.58 (0.46–0.69) | 1.6 (1.08–2.36) | 0.45 (0.14–1.42) | 0.35 | 0.23 |
| 25 | 0.77 (0.63–0.88) | 0.6 (0.49–0.71) | 1.64 (1.14–2.37) | 0.44 (0.15–1.29) | 0.37 | 0.21 |
| 30 | 0.71 (0.56–0.83) | 0.65 (0.54–0.76) | 1.7 (1.08–2.68) | 0.56 (0.27–1.16) | 0.36 | 0.21 |
| 35 | 0.65 (0.50–0.78) | 0.77 (0.66–0.86) | 2.18 (1.24–3.83) | 0.55 (0.25–1.20) | 0.42 | 0.17 |
| 40 | 0.6 (0.45–0.74) | 0.8 (0.69–0.88) | 2.3 (1.05–5.06) | 0.57 (0.30–1.1) | 0.4 | 0.20 |
| 45 | 0.56 (0.41–0.71) | 0.85 (0.75–0.92) | 2.63 (1.38–5.02) | 0.58 (0.33–1.01) | 0.41 | 0.22 |
| 50 | 0.52 (0.37–0.67) | 0.87 (0.78–0.94) | 2.91 (1.40–6.08) | 0.651 (0.39–0.95) | 0.39 | 0.25 |
| 55 | 0.46 (0.31–0.61) | 0.94 (0.86–0.98) | 4.44 (1.45–13.59) | 0.64 (0.43–0.95) | 0.4 | 0.30 |
| 60 | 0.42 (0.28–0.57) | 0.94 (0.86–0.98) | 4.12 (1.26–13.43) | 0.7 (0.52–0.94) | 0.36 | 0.34 |
| 65 | 0.33 (0.20–0.48) | 0.94 (0.86–0.98) | 3.09 (1.19–8.02) | 0.78 (0.65–0.94) | 0.27 | 0.45 |
| 70 | 0.31 (0.19–0.46) | 0.96 (0.89–0.99) | 4.17 (1.56–11.20) | 0.77 (0.65–0.92) | 0.27 | 0.48 |
Bold indicates optimal cut point.
TSC child and parent versions
Figure 2 depicts the ROC curves for the analyses of the TSC-C (Fig. 2A) and TSC-P (Fig. 2B) percentage reduction cut points. Each panel of Figure 2 features three ROC curves; one of the curves depicts the results from the overall analysis, including data from all children, while the remaining two depict results from the separate analyses for children 8–13 and 14–17 years of age. For the TSC-C, when data from all children (n = 98) were included in the analysis, the AUC for the ROC curve was 0.75, indicating that the TSC-C had a fair ability to determine treatment response as measured by a CGI-I ≤ 2. When data from children 8–13 (n = 52) and 14–17 (n = 46) years of age were analyzed separately, the AUCs for the ROC curves were 0.71 and 0.80, respectively, indicating that the TSC-C had a fair and good-to-great ability to determine response among children 8–13 and 14–17 years of age, respectively. For the TSC-P, when data from all children (n = 109) were included in the analysis, the AUC was 0.63, indicating that the TSC-P had a fair ability to determine treatment response as measured by a CGI-I ≤ 2. When data from children 8–13 (n = 65) and 14–17 (n = 44) years of age were analyzed separately, the AUCs for the ROC curves were 0.61 and 0.64, respectively, indicating that the TSC-P retained its fair ability to determine response to treatment among both age groups.
FIG. 2.
Receiver operating curves for the TSC-C and TSC-P Versions. (A, B) Depict the ROC curves of sensitivity versus one-specificity for the TSC-C and TSC-P percentage reduction cut points, respectively. For the TSC-C, when data from all children were included in the analysis, the AUC for the ROC curve was 0.75, indicating that the TSC-C has a fair ability to determine response to treatment as determined by a CGI-I ≤ 2. For the TSC-P, when all children and adolescents were included in the analysis, the AUC for the ROC curve was 0.63, also indicating that the TSC-P has a fair ability to determine response to treatment as determined by a CGI-I ≤ 2. AUC, area under the curve; CGI-I, Clinical Global Impressions Improvement; ROC, receiver operating characteristic; TSC-C, Trichotillomania Scale for Children Child Version; TSC-P, Trichotillomania Scale for Children Parent Version.
Tables 4 and 5 show pooled accuracy measures and parameters to determine the optimal percentage reduction cut point for each cut point on the TSC-C (Table 4) and TSC-P (Table 5) when data from all children were included in the analyses. For the TSC-C, a 25% reduction on the scores from pretreatment to posttreatment was considered the optimal cut point to determine treatment response, as measured by a CGI-I ≤ 2, by both the Youden Index (0.40) and the distance to corner (0.18) methods. This 25% reduction cut point on the TSC-C had a sensitivity of 0.74 (95% CI [0.56–0.87]), a specificity of 0.66 (95% CI [0.53–0.77]), a PLR of 2.27 (95% CI [1.46–3.54]), and an NLR of 0.52 (95% CI [0.18–1.48]). For treatment response as measured by CGI-I ≤ 2, the optimal cut point for reduction in TSC-P scores from pretreatment to posttreatment was 45% by the distance to corner (0.30) method and 50% by the Youden Index (0.33). The 45% reduction cut point on the TSC-P had a sensitivity of 0.49 (95% CI [0.31–0.66]), a specificity of 0.81 (95% CI [0.70–0.89]), a PLR of 3.58 (95% CI [1.35–9.45]), and an NLR of 0.66 (95% CI [0.44–1.0]); the 50% reduction cut point on the TSC-P had a sensitivity of 0.46 (95% CI [0.29–0.63]), a specificity of 0.87 (95% CI [0.77–0.93]), a PLR of 4.39 (95% CI [1.19–16.27]), and an NLR of 0.64 (95% CI [0.46–0.88]).
Table 4.
Accuracy Measures, Youden Index, and Distance to Corner for Cut Points for the Trichotillomania Scale for Children-Child Version
| Cut point, % | Sensitivity (95% CI) | Specificity (95% CI) | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Youden | Distance |
|---|---|---|---|---|---|---|
| 5 | 0.9 (0.73–0.98) | 0.38 (0.26–0.52) | 1.44 (1.12–1.86) | 0.34 (0.13–0.92) | 0.28 | 0.39 |
| 10 | 0.86 (0.68–0.96) | 0.47 (0.33–0.60) | 1.57 (1.16–2.13) | 0.35 (0.15–0.82) | 0.33 | 0.30 |
| 15 | 0.83 (0.64–0.94) | 0.55 (0.42–0.68) | 1.91 (1.31–2.79) | 0.43 (0.18–1.04) | 0.38 | 0.23 |
| 20 | 0.77 (0.59–0.89) | 0.63 (0.50–0.74) | 1.9 (1.03–3.50) | 0.54 (0.18–1.69) | 0.40 | 0.19 |
| 25 | 0.74 (0.56–0.87) | 0.66 (0.53–0.77) | 2.27 (1.46–3.54) | 0.52 (0.18–1.48) | 0.40 | 0.18 |
| 30 | 0.68 (0.50–0.83) | 0.69 (0.56–0.80) | 2.33 (1.43–3.82) | 0.6 (0.25–1.40) | 0.37 | 0.20 |
| 35 | 0.56 (0.38–0.73) | 0.75 (0.63–0.85) | 2.41 (1.32–4.40) | 0.68 (0.40–1.15) | 0.31 | 0.26 |
| 40 | 0.5 (0.32–0.68) | 0.78 (0.66–0.88) | 2.6 (1.34–5.01) | 0.73 (0.46–1.15) | 0.28 | 0.30 |
| 45 | 0.44 (0.27–0.62) | 0.88 (0.77–0.94) | 3.36 (1.47–7.68) | 0.73 (0.54–0.98) | 0.32 | 0.33 |
| 50 | 0.44 (0.27–0.62) | 0.91 (0.81–0.97) | 4.28 (1.67–10.97) | 0.71 (0.53–0.96) | 0.35 | 0.32 |
| 55 | 0.35 (0.20–0.54) | 0.92 (0.83–0.97) | 3.55 (1.30–9.74) | 0.76 (0.60–0.96) | 0.27 | 0.43 |
| 60 | 0.29 (0.15–0.48) | 0.94 (0.85–0.98) | 3.32 (1.07–10.33) | 0.8 (0.65–0.99) | 0.23 | 0.51 |
| 65 | 0.27 (0.13–0.44) | 0.94 (0.85–0.98) | 3.09 (0.99–9.68) | 0.82 (0.67–1.00) | 0.21 | 0.54 |
| 70 | 0.24 (0.11–0.41) | 0.94 (0.85–0.98) | 2.81 (0.90–8.90) | 0.85 (0.71–1.02) | 0.18 | 0.58 |
Bold indicates optimal cut point.
Table 5.
Accuracy Measures, Youden Index, and Distance to Corner for Cut Points for the Trichotillomania Scale for Children-Parent Version
| Cut point, % | Sensitivity (95% CI) | Specificity (95% CI) | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Youden | Distance |
|---|---|---|---|---|---|---|
| 5 | 0.74 (0.57–0.88) | 0.32 (0.22–0.44) | 1.15 (0.79–1.68) | 1.17 (0.33–4.11) | 0.06 | 0.53 |
| 10 | 0.71 (0.54–0.85) | 0.35 (0.24–0.47) | 1.15 (0.81–1.62) | 1.14 (0.38–3.37) | 0.06 | 0.51 |
| 15 | 0.69 (0.51–0.83) | 0.41 (0.29–0.53) | 1.23 (0.78–1.92) | 1.06 (0.35–3.24) | 0.1 | 0.44 |
| 20 | 0.66 (0.48–0.81) | 0.49 (0.37–0.61) | 1.38 (0.74–2.56) | 0.93 (0.31–2.81) | 0.15 | 0.38 |
| 25 | 0.63 (0.45–0.79) | 0.57 (0.45–0.68) | 1.58 (0.69–3.63) | 0.9 (0.29–2.75) | 0.2 | 0.32 |
| 30 | 0.54 (0.37–0.71) | 0.68 (0.56–0.78) | 2.23 (1.27–3.93) | 0.74 (0.43–1.29) | 0.22 | 0.31 |
| 35 | 0.49 (0.31–0.66) | 0.73 (0.61–0.83) | 2.36 (1.13–4.94) | 0.73 (0.47–1.15) | 0.22 | 0.33 |
| 40 | 0.49 (0.31–0.66) | 0.77 (0.66–0.86) | 3.11 (1.47–6.58) | 0.68 (0.44–1.05) | 0.26 | 0.31 |
| 45 | 0.49 (0.31–0.66) | 0.81 (0.70–0.89) | 3.58 (1.35–9.45) | 0.66 (0.44–1.00) | 0.3 | 0.30 |
| 50 | 0.46 (0.29–0.63) | 0.87 (0.77–0.93) | 4.39 (1.19–16.27) | 0.64 (0.46–0.88) | 0.33 | 0.31 |
| 55 | 0.37 (0.22–0.55) | 0.87 (0.77–0.93) | 3.66 (1.06–12.61) | 0.7 (0.54–0.90) | 0.24 | 0.41 |
| 60 | 0.34 (0.19–0.52) | 0.88 (0.78–0.94) | 3.52 (1.13–10.97) | 0.7 (0.56–0.91) | 0.22 | 0.45 |
| 65 | 0.29 (0.15–0.46) | 0.89 (0.80–0.95) | 3.15 (0.96–10.29) | 0.76 (0.61–0.95) | 0.18 | 0.52 |
| 70 | 0.2 (0.08–0.37) | 0.93 (0.85–0.98) | 2.45 (0.71–8.41) | 0.87 (0.74–1.04) | 0.13 | 0.65 |
Bold indicates optimal cut point.
Discussion
This meta-analysis compares three commonly used instruments to assess TTM severity in clinical trials of pediatric populations: the MGH-HPS, NIMH-TSS, and TSC-C/P. Specifically, this study provides accuracy measures of percentage reduction from pretreatment to posttreatment on these three instruments that correspond to treatment response defined by a CGI-I score of 2 (“much improved”) or 1 (“very much improved”). Furthermore, this study also derives optimal percentage reduction cut points for each instrument to classify responders and nonresponders.
Our results indicate that a 45% reduction on the MGH-HPS could be considered the optimal cut point to determine response to treatment among children and adolescents with TTM according to both the Youden Index (0.37) and distance to corner (0.20) methods. Our results also indicated that the MGH-HPS had a poorer discriminative ability to determine treatment response in comparison to the TSC-C (AUC = 0.68 vs. 0.75), the other self-report instrument employed to measure TTM severity. This could be explained by the fact that the MGH-HPS features language that is likely not appropriate for children; for instance, instead of inquiring about hair pulling behavior itself, severity items of the MGH-HPS query about urges (e.g., “On an average day, how often did you feel the urge to pull your hair?”), which might be somewhat difficult for children to understand. Consistent with this hypothesis, the MGH-HPS showed a good discriminative ability to determine response among adults (AUC = 0.82) in our previous study (Farhat et al. 2019). Yet, it should be noted that for this study, only two trials had data available on the MGH-HPS and we were unable to perform separate analyses for different age groups as sensitivity analyses, given the small sample sizes. In addition, although the AUC for the MGH-HPS is smaller than the AUC for the TSC-C, we were not able to compare both AUCs to evaluate if these numerical differences reach significance threshold. Future research should look further into possible age-related differences in the ability of the MGH-HPS to discriminate response among youth of different ages. Given that younger children may have more difficulty with the abstract language, it would be interesting to see if the MGH-HPS has a higher discriminative ability among adolescents when compared with young children. Nevertheless, considering that the MGH-HPS has never been validated among children and adolescents, data from this meta-analysis reinforce the notion that caution is required about employing the MGH-HPS as a primary outcome measure in trials for pediatric TTM.
This meta-analysis also showed that a 35% reduction on the NIMH-TSS could be considered the optimal cut point to determine response to treatment among children and adolescents with TTM according to both the Youden Index (0.42) and distance to corner (0.17) methods. Our results also indicated that the NIMH-TSS had a fair ability to determine response to treatment in clinical trials for pediatric TTM (AUC = 0.75). Although both the NIMH-TSS and the CGI-I are clinician-rated instruments, the NIMH-TSS is a semistructured interview that queries about specific subjects regarding hair pulling, whereas the CGI-I is a unitary score that assesses the patient's global well-being and functioning. It is possible that some of the questions on the NIMH-TSS may be inappropriate for young children because they require respondents to estimate intervals of time (e.g., “How much time did you spend pulling hairs yesterday?”); this may explain why the NIMH-TSS had less discriminative ability among pediatric populations in comparison to adults (AUC = 0.92) (Farhat et al. 2019). At the present moment, considering the NIMH-TSS has also never been validated among pediatric populations, researchers should also be cautious about employing the NIMH-TSS in their clinical trials as a primary outcome to rate TTM severity.
Finally, our results suggest that for the TSC-C, a 25% reduction on the TSC-C (Youden Index = 0.40, distance to corner = 0.18) could be considered the optimal cut point to determine response to treatment among children and adolescents with TTM. For the TSC-P, a 45% reduction could be considered optimal according to the distance to corner method (0.30); according to the Youden index (0.33), a 50% reduction could be considered the optimal cut point. Although both the 45% and 50% reduction cut points could be considered optimal, it should be noted that the 45% cut point, in comparison to the 50% cut point, had a slightly larger sensitivity (0.49 vs. 0.46) and slightly smaller specificity (0.81 vs. 0.87). When deciding on which one of these evidence-based cut points to adopt, researchers and clinicians should consider these differences in the accuracy measures of both cut points. Both the TSC-C (AUC = 0.75) and the TSC-P (AUC = 0.63) had a fair ability to determine response to treatment, but the self-report measure had a higher discriminative ability to determine treatment response than the parent measure. Children often feel ashamed of their hair pulling behavior and try to conceal or minimize their pulling in their everyday life; it is possible that this could interfere with parental perceived change of hair pulling severity during the course of a treatment trial.
Previous research on TTM (Keuthen et al. 2008) has shown that parent-youth concordance for awareness of hair pulling and anxiety scores among younger children (10–12 years old) is lower than among older adolescents (15–17 years old). Phenotypical research on TTM could help explain this discrepancy as two distinct hair pulling styles have been described: automatic and focused (Christenson et al. 1991a; Flessner et al. 2007). While the automatic pattern describes pulling that occurs out of one's own awareness, the focused pattern describes pulling with awareness. Cross-sectional studies report findings that suggest that children may experience a transition from automatic to focused hair pulling, as younger children are less aware of urges and less focused on their pulling in comparison to older children (Diefenbach et al. 2002; Panza et al. 2013). It is likely that the lower concordance of pulling awareness observed in parent-child dyads could be explained by the fact that young children may be more likely to pull their hair without being consciously aware of doing so. In that sense, it is possible that the TSC-C would be a less appropriate measure of treatment response in RCTs with young children. This meta-analysis provides some evidence to support this hypothesis, as the TSC-C had a higher discriminative ability among older in comparison to younger children (AUC = 0.80 vs. 0.71). Yet, we were unable to investigate whether these age-related differences are mediated by differences in the hair pulling styles, given the small sample sizes. Future research on this topic is warranted when more data are available from trials for pediatric TTM.
At the moment, TTM remains a difficult-to-treat condition, especially among children and adolescents. Pediatric TTM is considerably understudied as currently there are only 5 published clinical trials evaluating treatments for pediatric TTM. Importantly, these trials exhibit significant methodological limitations, such as small sample sizes and primary outcome measures that were not validated among pediatric samples and lack standardized definitions of treatment response. These methodological limitations hamper further advancements in the development of novel treatments for pediatric TTM. In this meta-analysis, we demonstrated the discriminative ability of different TTM severity instruments to determine response to treatment in clinical trials for pediatric TTM; we also provided accuracy measures for percentage reduction cut points, as well as parameters to determine the optimal cut point to dichotomize children as responders and nonresponder for each instrument. This meta-analysis has several significant limitations. For instance, although all pediatric TTM clinical trials were included in this meta-analysis, the sample size (n = 210) was considerably small and limited considerably the findings we reported for the sensitivity analyses. Furthermore, we only employed as criterion measure the CGI-I posttreatment score. Although the CGI-I is typically employed to determine response to treatment in clinical trials for pediatric TTM, it does not correspond to a clinically meaningful response to treatment as defined by Jacobson and Truax (1991). These authors developed a reliable change (RC) index that compares the individual's change in scores from pretreatment to posttreatment with the distribution of the possible changes in scores from pretreatment to posttreatment if no change at all had occurred; an RC index higher than 1.96 is highly suggestive (p < 0.05) that the posttreatment score indeed represents a clinically significant change. Nevertheless, although the use of CGI-I as a criterion measure does not consider this complex literature on clinically meaningful response to treatment, we decided to employ it as the gold standard of response, given this is currently the standard in clinical trials for TTM, including pediatric TTM. Therefore, using the CGI-I as a criterion measure has important practical considerations for previous, as well as future, clinical trials for pediatric TTM. Future research should investigate how TTM severity instruments correspond to clinically meaningful treatment response.
Conclusions
To conclude, our findings provide evidence-based optimal cut points for the three most commonly used TTM severity instruments to help guide clinicians and researchers for assessing treatment response in children and adolescents. We recommend that clinical trials for pediatric TTM determine response to treatment a priori as post hoc definitions of response are likely to inflate the results of the clinical trials, limiting the accuracy of their findings. Given our findings that these three instruments show similar, modest discriminative abilities to determine treatment response in pediatric populations, an expert consensus is warranted to determine an agreed-upon definition of response to treatment for pediatric TTM clinical trials. Data from this meta-analysis, as well as from the previous one performed by our group (Farhat et al. 2019) and others (Houghton et al. 2015), should be considered in forming expert consensus recommendations. Further continued collaborative efforts to validate rating scales for pediatric TTM particularly in relation to symptom improvement are needed. Importantly, two of the most commonly used hair pulling severity instruments, the MGH-HPS and NIMH-TSS, have not yet been validated or tailored for pediatric populations, highlighting a need for further research on these measures.
Clinical Significance
Pediatric TTM is an often underrecognized and difficult-to-treat condition. Currently, there are no clear first-line pharmacological agents for hair pulling and efficacious behavioral interventions are often difficult to access in community settings. Additional research is warranted to identify novel, efficacious treatments for pediatric TTM. Future clinical trials evaluating interventions for TTM should define treatment response a priori to facilitate comparisons across studies as well as to avoid potential problems with post hoc definitions of response. This study provides evidence-based definitions of treatment response of three most commonly used TTM severity instruments in pediatric populations. In that way, this study might be helpful in guiding researchers and clinicians in assessing treatment response.
Disclosures
Mr. Farhat reports no disclosures. Dr. Olfson receives research support from the American Academy of Child & Adolescent Psychiatry and the Alan B. Slifka Foundation through the Riva Ariella Ritvo endowment. Ms. Levine reports no disclosures. Mr. Li reports no disclosures. Dr. Franklin reports no conflicts of interest. Dr. Lee has received research grant from the Trichotillomania Learning Center Research Foundation. Dr. Lewin has received research grants from CDC, All Children's Hospital, and Tourette Association of America. He also received an honorarium from Springer and has been awarded travel grants from the IOCDF Scientific & Clinical Advisory Board. He is board of the board of directors from the American Board of Clinical Child & Adolescent Psychology. He is a consultant for Bracket LLC. Dr. McGuire has received research support from the Tourette Association of America, American Academy of Neurology, the Brain Research Foundation, American Psychological Foundation, and the Hilda and Preston Davis Foundation. He receives royalties from Elsevier, and serves as a consultant for Bracket, Syneos, Health, and Luminopia. Dr. Rahman reports receiving research support from Neurocrine Biosciences, Roche, Otsuka, and Nuvelation, and consulting fees from Bracket. Dr. Storch is a consultant for Levo therapeutics. He receives research support from the National Institute of Health, the Red Cross, ReBuild Texas, the Texas Higher Education Coordinating Board, and the Greater Houston Community Foundation. He also reported receiving royalties from Elsevier Publications, John Wiley & Sons, Inc., American Psychological Association, Springer, and Lawrence Erlbaum. Dr. Tolin reports no disclosures. Dr. Zickgraf is funded by the T32MH082761-10 NIH/NIMH, Midwestern Regional Eating Disorders Training Grant. Dr. Bloch is on the Scientific Advisory Board of Therapix Biosciences and receives research support from Biohaven Pharmaceuticals, Janssen Pharmaceuticals, Neurocrine Biosciences, and Therapix Biosciences. Dr. Bloch also receives research support from the National Institute of Health, Tourette Association of America, the Brain & Behavior Research Foundation (formerly NARSAD), and the Patterson Foundation.
Supplementary Material
Supplementary Material
References
- American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders, 5th ed. Varlington, VA: American Psychiatric Publishing; 2013 [Google Scholar]
- Bloch MH: Trichotillomania across the life span. J Am Acad Child Adolesc Psychiatry 48:879–883, 2009 [DOI] [PubMed] [Google Scholar]
- Bloch MH, Landeros-Weisenberger A, Dombrowski P, Kelmendi B, Wegner R, Nudel J, Pittenger C, Leckman JF, Coric V: Systematic review: Pharmacological and behavioral treatment for trichotillomania. Biol Psychiatry 62:839–846, 2007 [DOI] [PubMed] [Google Scholar]
- Bloch MH, Panza KE, Grant JE, Pittenger C, Leckman JF: N-Acetylcysteine in the treatment of pediatric trichotillomania: A randomized, double-blind, placebo-controlled add-on trial. J Am Acad Child Adolesc Psychiatry 52:231–240, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Busner J, Targum SD: The clinical global impressions scale: Applying a research tool in clinical practice. Psychiatry (Edgmont) 4:28–37, 2007 [PMC free article] [PubMed] [Google Scholar]
- Christenson GA, Mackenzie TB, Mitchell JE: Characteristics of 60 adult chronic hair pullers. Am J Psychiatry 148:365–370, 1991a [DOI] [PubMed] [Google Scholar]
- Christenson GA, Pyle RL, Mitchell JE: Estimated lifetime prevalence of trichotillomania in college students. J Clin Psychiatry 52:415–417, 1991b [PubMed] [Google Scholar]
- Diefenbach GJ, Mouton-Odum S, Stanley MA: Affective correlates of trichotillomania. Behav Res Ther 40:1305–1315, 2002 [DOI] [PubMed] [Google Scholar]
- Emsley R, Hawkridge S: The quest for a meaningful evidence base in psychiatry. World Psychiatry 8:33–34, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Farhat LC, Olfson E, Li F, Telang S, Bloch MH: Identifying standardized definitions of treatment response in trichotillomania: A meta-analysis. Progr Neuropsychopharmacol Biol Psychiatry 89:446–455, 2019 [DOI] [PubMed] [Google Scholar]
- Fleischhacker WW, Goodwin GM: Effectiveness as an outcome measure for treatment trials in psychiatry. World Psychiatry 8:23–27, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Flessner CA, Woods DW, Franklin ME, Keuthen NJ, Piacentini J, Cashin SE, Moore PS: The Milwaukee Inventory for Styles of Trichotillomania-Child Version (MIST-C): Initial development and psychometric properties. Behav Modif 31:896–918, 2007 [DOI] [PubMed] [Google Scholar]
- Franklin ME, Edson AL, Ledley DA, Cahill SP: Behavior therapy for pediatric trichotillomania: A randomized controlled trial. J Am Acad Child Adolesc Psychiatry 50:763–771, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Franklin ME, Flessner CA, Woods DW, Keuthen NJ, Piacentini JC, Moore P, Stein DJ, Cohen SB, Wilson MA: The child and adolescent trichotillomania impact project: Descriptive psychopathology, comorbidity, functional impairment, and treatment utilization. J Dev Behav Pediatr 29:493–500, 2008 [DOI] [PubMed] [Google Scholar]
- Houghton DC, Capriotti MR, De Nadai AS, Compton SN, Twohig MP, Neal-Barnett AM, Saunders SM, Franklin ME, Woods DW: Defining treatment response in trichotillomania: A signal detection analysis. J Anxiety Disord 36:44–51, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jacobson NS, Truax P: Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. J Consult Clin Psychol 59:12–19, 1991 [DOI] [PubMed] [Google Scholar]
- Keuthen NJ, Flessner CA, Woods DW, Franklin ME, Piacentini J, Khanna M, Moore P, Cashin S, The TLC-SAB: Parent-Youth Rating Concordance for Hair Pulling Variables, Functional Impairment and Anxiety Scale Scores in Trichotillomania. Child Fam Behav Ther 30:337–353, 2008 [Google Scholar]
- Kraemer HC, Kupfer DJ: Size of treatment effects and their importance to clinical research and practice. Biol Psychiatry 59:990–996, 2006 [DOI] [PubMed] [Google Scholar]
- Lee HJ, Espil FM, Bauer CC, Siwiec SG, Woods DW: Computerized response inhibition training for children with trichotillomania. Psychiatry Res 262:20–27, 2018 [DOI] [PubMed] [Google Scholar]
- Lewin AB, Piacentini J, Flessner CA, Woods DW, Franklin ME, Keuthen NJ, Moore P, Khanna M, March JS, Stein DJ: Depression, anxiety, and functional impairment in children with trichotillomania. Depress Anxiety 26:521–527, 2009 [DOI] [PubMed] [Google Scholar]
- McGuire JF, Kugler BB, Park JM, Horng B, Lewin AB, Murphy TK, Storch EA: Evidence-based assessment of compulsive skin picking, chronic tic disorders and trichotillomania in children. Child Psychiatry Hum Dev 43:855–883, 2012 [DOI] [PubMed] [Google Scholar]
- McGuire JF, Ung D, Selles RR, Rahman O, Lewin AB, Murphy TK, Storch EA: Treating trichotillomania: A meta-analysis of treatment effects and moderators for behavior therapy and serotonin reuptake inhibitors. J Psychiatr Res 58:76–83, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Odlaug BL, Grant JE: Impulse-control disorders in a college sample: Results from the self-administered Minnesota Impulse Disorders Interview (MIDI). Prim Care Companion J Clin Psychiatry 12, 2010. DOI: 10.4088/PCC.09m00842whi [DOI] [PMC free article] [PubMed]
- Panza KE, Pittenger C, Bloch MH: Age and gender correlates of pulling in pediatric trichotillomania. J Am Acad Child Adolesc Psychiatry 52:241–249, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rahman O, McGuire J, Storch EA, Lewin AB: Preliminary randomized controlled trial of habit reversal training for treatment of hair pulling in youth. J Child Adolesc Psychopharmacol 27:132–139, 2017 [DOI] [PubMed] [Google Scholar]
- Tolin DF, Diefenbach GJ, Flessner CA, Franklin ME, Keuthen NJ, Moore P, Piacentini J, Stein DJ, Woods DW: The trichotillomania scale for children: Development and validation. Child Psychiatry Hum Dev 39:331–349, 2008 [DOI] [PubMed] [Google Scholar]
- Tolin DF, Franklin ME, Diefenbach GJ, Anderson E, Meunier SA: Pediatric trichotillomania: Descriptive psychopathology and an open trial of cognitive behavioral therapy. Cogn Behav Ther 36:129–144, 2007 [DOI] [PubMed] [Google Scholar]
- Unal I: Defining an optimal cut-point value in ROC analysis: An alternative approach. Comput Math Methods Med 2017:3762651, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Woods DW, Flessner C, Franklin ME, Wetterneck CT, Walther MR, Anderson ER, Cardona D: Understanding and treating trichotillomania: What we know and what we don't know. Psychiatr Clin North Am 29:487–501, ix, 2006 [DOI] [PubMed] [Google Scholar]
- Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A: Meta-DiSc: A software for meta-analysis of test accuracy data. BMC Med Res Methodol 6:31, 2006 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.