Population Prevalence of Tourette Syndrome: A Systematic Review and Meta-Analysis Jeremiah M. Scharf, MD, PhD,1,2* Laura L. Miller, MSc,3 Caitlin A. Gauvin, BS,1 Janelle Alabiso, MA,1 Carol A. Mathews, MD,4 and Yoav Ben-Shlomo, MBBS, PhD3 1

Psychiatric and Neurodevelopmental Genetics Unit, Center for Human Genetics Research, Departments of Neurology and Psychiatry, Massachusetts General Hospital, Boston, Massachusetts, USA 2 Division of Cognitive and Behavioral Neurology, Department of Neurology, Brigham and Women’s Hospital, Boston, Massachusetts, USA 3 School of Social and Community Medicine, University of Bristol, United Kingdom 4 Department of Psychiatry, University of California, San Francisco, California, USA


The aim of this study was to refine the population prevalence estimate of Tourette Syndrome (TS) in children and to investigate potential sources of heterogeneity in previously published studies. A systematic review was conducted and all qualifying published studies of TS prevalence were examined. Extracted data were subjected to a random-effects meta-analysis weighted by sample size; metaregressions were performed to examine covariates that have previously been proposed as potential sources of heterogeneity. Twenty-six articles met study inclusion criteria. Studies derived from clinically referred cases had prevalence estimates that were significantly lower than those derived from population-based samples (P 5 0.004). Among the 21 population-based prevalence studies, the pooled TS population prevalence estimate was 0.52% (95% confidence interval CI: 0.32-0.85). In univariable meta-regression analysis, study sample size (P 5 0.002) and study date (P 5 0.03) were significant predictors of TS prevalence. In the final multivariable model including sample size, study date, age, and diag-


*Correspondence to: Dr. Jeremiah M. Scharf, Psychiatric and Neurodevelopmental Genetics Unit, Center for Human Genetics Research, Massachusetts General Hospital, 185 Cambridge Street, 6th floor, Boston, MA 02114, USA; [email protected] Dr. Scharf and Miss Miller contributed equally to the article. Funding agencies: This work was funded by the Tourette Syndrome Association and the National Institutes of Mental Health (MH080857).

Relevant conflicts of interest/financial disclosures: Nothing to report. Full financial disclosures and author roles may be found in the online version of this article. Received: 28 June 2013; Revised: 17 October 2014; Accepted: 23 October 2014 Published online 8 December 2014 in Wiley Online Library ( DOI: 10.1002/mds.26089

nostic criteria, only sample size (P < 0.001) and diagnostic criteria (omnibus P 5 0.003; Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision [DSM-IV-TR]: P 5 0.005) were independently associated with variation in TS population prevalence across studies. This study refines the population prevalence estimate of TS in children to be 0.3% to 0.9%. Study sample size, which is likely a proxy for case assessment method, and the use of DSM-IV-TR diagnostic criteria are the major sources of heterogeneity across studies. The true TS population prevalence rate is likely at the higher end of these estimates, given the methodological limitations of most studies. Further studies in large, well-characterized samples will be helpful to determine the burden of disease in the genC eral population. V 2014 International Parkinson and Movement Disorder Society

K e y W o r d s : Tourette syndrome; tics; prevalence studies; meta-analysis; developmental disorders

Tourette syndrome (TS) is a neurodevelopmental disorder defined by the presence of multiple motor and phonic tics that begin in childhood and persist for more than 1 year.1 Early prevalence reports based on clinically ascertained cases suggested that TS is a rare disorder with rates of approximately 5 per 10,000 for school-age children.2 However, clinical case series are known to underestimate the actual population prevalence of TS at least 10-fold.3 Over the past 20 years, a wide range of population prevalence estimates (0.03%-5.26%) have been reported, and a 2007 review concluded that existing data were insufficient to provide an accurate estimate.4-6

Movement Disorders, Vol. 30, No. 2, 2015


S C H A R F,




Many factors have been proposed to explain the discrepancy in TS prevalence estimates between published studies, including differences in diagnostic criteria, case assessment procedures, subject age range, geographical region, and ascertainment bias.6-8 However, these various hypotheses have never been formally examined using an empirical approach. Determination of an accurate TS prevalence is critical for understanding the overall disease burden. A recent systematic review and metaanalysis of tic disorders reported a TS prevalence of 0.77% (95% confidence interval [CI]: 0.39-1.51) in children and adolescents and a prevalence of 0.05% (95% CI: 0.03-0.08) in adults.9 However, though the investigators identified significant heterogeneity between studies in school-aged children, they did not examine potential sources of variation. Here, we expand on and refine the previously published literature with a systematic review and meta-analysis of published TS prevalence studies in children and a meta-regression to examine sources of study heterogeneity, including case assessment method, TS diagnostic criteria, study date, sample size, geographical location, and subject age.

in the present study. Reference lists of remaining articles were checked for additional studies, and results of this search were also checked against recent reviews7,9,43 to ensure completeness and revise the search criteria if needed. One additional article was identified.8 A secondary search was also conducted to identify TS prevalence studies reported in the context of population-based screens of all Axis I psychiatric disorders in children and adolescents and identified five additional candidate articles.44-48 Four articles were excluded for examination of tics rather than TS44,46,47 or for lacking tic screening in the first-stage sample.48 The remaining article was included.45 For studies in which screening and assessment procedures were not clear, investigators were contacted for clarification of methodology.11,20,45

Materials and Methods

Prevalence Rates

Literature Search MEDLINE (PubMed) and EMBASE were systematically searched in October 2012 by bachelor’s-level research assistants (J.A. and C.A.G., supervised by J.M.S.). The search included the period from 1966 to 2012 and was performed in accord with the Metaanalysis of Observational Studies in Epidemiology (MOOSE) criteria.10 Detailed search strategies are provided in Appendix 1 (see the Supporting Information). These searches yielded 291 articles in PubMed and 261 in EMBASE. Abstracts were reviewed and articles selected based on prespecified inclusion/exclusion criteria (Supporting Table 1; Supporting Fig. 1). Studies were required to contain raw data on TS prevalence (either population or clinic based), including total sample size, sample population, and detailed screening methods. Articles were excluded if they studied only adults (age >18) or specific clinical subsamples (autistic or special education populations). Incidence studies, follow-up studies, treatment studies, diagnostic instrument studies, studies of medication use, prevalence in first-degree relatives, prevalence within another disorder, and prevalence of specific symptoms, traits, or characteristics in a TS sample were also excluded. The remaining 34 articles were searched by hand (by J.A. and C.A.G., supervised by J.M.S.).2,3,11-42 Ten articles were subsequently excluded because (1) the study examined tics rather than TS,19,26,27 (2) raw data were not provided,28,29 (3) data were presented in a previous study,30-33 or (4) the study focused only on special education populations or children with behavioral problems.34 Twenty-four articles remained and were included


Movement Disorders, Vol. 30, No. 2, 2015

Data Extraction The following data were extracted from each article for meta-analysis and meta-regression: study date; sample size; age of sample population; number of TS cases; case assessment method; diagnostic criteria used; and geographical region.

TS prevalence rates were calculated directly from the raw data. In one study, the calculated rate (0.25%) was lower than that reported in the article (0.5%)21; in this case, the calculated rate was used, consistent with the approach taken in a previous meta-analysis.9 Of note, throughout this article, we use the term “prevalence” to indicate TS population prevalence, given that studies of clinically ascertained cases still aimed to estimate population prevalence, albeit with a known ascertainment bias.

Study Date Study date was calculated by taking the median date within the study recruitment period. In studies where the date performed could not be ascertained, 3 years were subtracted from the date of publication and used as a proxy for the median study date (Table 1).

Age Subject ages ranged from 4 to 18 years. Minimum and maximum ages were extracted from each study; for the heterogeneity analyses and meta-regression, mean age was used. Where studies reported separate prevalence rates for nonoverlapping age subgroups, agespecific rates were examined separately to assess agerelated heterogeneity.17,22,37,38 However, for analyses of non-age-related variables, age subgroups were collapsed so that there was only one data point per study.

Case Assessment Method Each study was assigned to one of the following five case assessment methods: (1) clinical series: previously

M E T A - A N A L Y S I S




TABLE 1. Published TS prevalence studies and extracted study data included in the meta-analysis and meta-regression Study No.

1 1 1 1 2 3 4 4 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 20 20 20 20 20 21 22 23 23 23 23 24 25 26 27

First Author

Publication Year

Study Datea


Diagnostic Criteria Used


Mean Age

Sample Size

Total No, of TS Cases

Rate (%)

Khalifa (age 7-15) Khalifa (7-9) Khalifa (10-12) Khalifa (13-15) Kurlan Scahill CDC (age 6-17) CDC (6-11) CDC (12-17) Stefanoff Wang Mason Comings Nomoto Wong Kadesjo (a) Kadesjo (b) Hornsey Costello Caine Burd Apter Atladottir Peterson Jin et al. (age 7-16) Jin et al. (7-9) Jin et al. (9-11) Jin et al. (11-13) Jin et al. (13-15) Jin et al. (15-16) Heiervang Zohar Cubo (age 5-17) Cubo (6-9) Cubo (10-14) Cubo (15-17) Kraft Schlander Amiri Scharf

2003 2003 2003 2003 2001 2006 2009 2009 2009 2007 2003 1998 1990 1990 1992 2000 2000 2001 1996 1988 1986 1992 2007 2001 2005 2005 2005 2005 2005 2005 2007 1992 2011 2011 2011 2011 2012 2011 2012 2012

2000 2000 2000 2000 1994-1998 1991 2007-2008 2007-2008 2007-2008 2004 2003 1995 1987 1988 1989 1996 1998 1998 1993 1981-1982 1983 1989 2004 1983 2002 2002 2002 2002 2002 2002 2002-2003 1989 2008 2008 2008 2008 2005 2003 2009 2009

Europe Europe Europe Europe North America North America North America North America North America Europe Far East Europe North America Far East Far East Europe Europe Europe North America North America North America Middle East Europe North America Far East Far East Far East Far East Far East Far East Europe Middle East Europe Europe Europe Europe Europe Europe Middle East Europe


Two-stage Two-stage Two-stage Two-stage Directly assessed Two-stage Clinical series Clinical series Clinical series Two-stage Two-stage 2-stage plus Two-stage Questionnaire only Two-stage Two-stage Clinical series Two-stage Two-stage Clinical series Clinical series 2-stage plus Clinical series Directly assessed 2-stage plus 2-stage plus 2-stage plus 2-stage plus 2-stage plus 2-stage plus Two-stage Directly assessed 2-stage plus 2-stage plus 2-stage plus 2-stage plus Two-stage Clinical series 2-stage plus Questionnaire only

11 8 11 14 13 9 11.5 8.5 14.5 13.5 9 13.5 9.5 8 9.7 11.1 10 13.5 11 11.5 12 16.5 11.5 13.7 11.5 8 10 12 14 15.5 8 16.5 11 7.5 12 16 14 12.5 9.5 13

4,479 1,850 1,277 1,352 1,255 910 64,034 27,776 36,258 1,579 2,000 166 2,972 1,218 718 435 38,786 918 4,067 142,636 140,580 28,037 402,315 776 9,742 2,192 1,742 1,945 2,940 923 9,430 562 741 264 373 104 5,974 292,568 1,658 6,768

25 11 10 4 10 3 225 61 164 9 11 5 17 3 3 5 58 7 4 41 73 12 259 2 42 10 11 10 9 2 15(4)c 1 39 16 18 6 33 114 22 50

0.56 0.59 0.78 0.30 0.80 0.33 0.30 0.19 0.40 0.60 0.56 3.00 0.57 0.25 0.40 1.10 0.15 0.76 0.10 0.03 0.05 0.04 0.06 0.26 0.43 0.46 0.63 0.51 0.31 0.22 0.16 0.18 5.26 6.06 4.83 5.77 0.55 0.04 1.33 0.74


Date performed. CCMD-3 (Chinese Classification of Mental Disorders, 3rd edition) criteria are equivalent to DSM-IV criteria and thus treated as “DSM-IV” in diagnostic criteria analyses; c Per contact with the corresponding author, 4 TS cases were identified in the interviewed subsample. This represents an estimated 15 cases in the target population. b

diagnosed clinical cases within a predefined geographical region; (2) questionnaire only: study of a school or population-based sample employing screening questionnaires by nonprofessionals without a subsequent formal diagnostic evaluation; (3) two-stage assessment: screening questionnaires followed by a full diagnostic evaluation of screen-positive subjects; (4) two-stage assessment plus observational screen: screening questionnaires and a brief observation period followed by a full diagnostic evaluation of questionnaire-positive or observational screen-positive individuals using a formal diagnostic instrument; and (5) direct assessment of all subjects using a formal diagnostic instrument.

Diagnostic Criteria Formal diagnostic criteria used to define TS were recorded: Diagnostic and Statistical Manual of Mental Disorders (DSM)-III, DSM-III-R, DSM-IV, and DSM-IVTR. Studies using the Chinese Classification of Mental Disorders (CCMD), 3rd Edition,49 were included in the DSM-IV category based on the presence of an impairment criterion (I.C. Chou, personal communication).

Geographical Region Geographical region was divided into categories based on continent or subcontinental region: North America, Europe, Middle East, and the Far East.

Movement Disorders, Vol. 30, No. 2, 2015


S C H A R F,




Statistical Analyses Statistical analyses were performed in Stata software (v11; StataCorp LP, College Stattion, TX). Study heterogeneity was assessed with der Simonian and Laird random-effects modeling, which incorporates an estimate of between-study variation (s2) as well as the inverse sampling variance. We used a random-effects model, rather than a fixed-effects model, because we postulated a priori that there would be between-study variability as a result of methodological differences. Degree of heterogeneity was evaluated using the Higgins I2 statistic and Cochran’s Q test.50 In addition, each covariate was evaluated with univariable metaregression followed by multivariable regression using log prevalence as the dependent variable. Study date was centered on median year (2001) and examined as a continuous variable. We also attempted to examine study date in tertiles to assess for nonlinear effects, but were unable to do so because of collinearity. Age was grouped into categories (6-9, 10-12, and 13-18 years) for stratified meta-analysis and metaregression. For categorical predictors, the reference level was selected as the level that contained the most studies to maximize power (two-stage for method, DSM-III-R for diagnostic criteria, age >12 for age, and Europe for geographical region). Categorical variables were assessed first using a multilevel omnibus F-test and then a category-level (v2) test. Candidate predictor variables were selected for multivariable analysis either based on nominal significance (P < 0.05) in the univariable omnibus test or based on a priori clinical grounds as potential mediators or confounders (age and diagnostic criteria).51

Results Systematic Review of Previous TS Prevalence Studies Twenty-six articles met all inclusion/exclusion criteria (Table 1 and Supporting Table 1; Supporting Fig. 1). One article contained two distinct studies, one in a clinically referred sample and the other in a population-based school cohort.3 Therefore, 27 independent TS prevalence studies were included in the meta-analysis. Four articles included data from multiple, nonoverlapping age groups for age-specific analyses.17,22,37,38 TS prevalence rates varied widely across studies (0.03%-5.26%; Table 1).

Meta-Analysis of Previous TS Prevalence Studies Meta-analysis with random-effects modeling confirmed the presence of significant heterogeneity across studies (I2 assuming a fixed-effects model 5 98.4%; P < 0.001; Fig. 1). The pooled prevalence estimate from studies of clinically ascertained cases (0.08%; 95% CI:


Movement Disorders, Vol. 30, No. 2, 2015

0.03-0.17) was significantly lower than those derived from population-based samples (relative risk [RR] 5 0.17; 95% CI: 0.06-0.54; P 5 0.004; Supporting Table 2). Thus, the six studies based on clinical series were excluded from subsequent analyses. Among the 21 population-based prevalence studies, the pooled TS prevalence estimate was 0.52% (95% CI: 0.32-0.85; Fig. 1). Repeating analyses after removing each study individually revealed no undue influence of a single study (data not shown). Stratified meta-analysis demonstrated significant residual heterogeneity even after excluding clinical series studies (I2 5 93.7%; P < 0.001; Fig. 1 and Supporting Fig. 2).

Metaregression Given that heterogeneity tests between strata are invalid under a random-effects model, meta-regression analyses were conducted to evaluate the potential sources of heterogeneity across the 21 populationbased prevalence studies. In univariable metaregression analyses, only study sample size and study date were significant predictors of TS prevalence (Table 2); there were no significant differences in prevalence based on case assessment method, diagnostic criteria, subject age, or geographic region (omnibus Ftests; P > 0.05). Sample size accounted for 47% of the variance between studies (adjusted r2 5 47.4%; P 5 0.002; Supporting Fig. 3). For study date, TS prevalence estimate increased by 7% per year (P 5 0.03). Univariable meta-regression for diagnostic criteria suggested a 2-fold increase in prevalence for studies that used DSM-IV-TR criteria, compared to the DSM-III-R reference level (RR 5 3.53; P 5 0.048). Though the omnibus tests for age and diagnostic criteria did not reach formal significance, both variables were included in the multivariable analysis because both are expected clinically to influence TS prevalence. A multivariable model incorporating sample size, study date, subject age, and diagnostic criteria identified both sample size (P < 0.001) and diagnostic criteria (omnibus, P 5 0.003; DSM-IV-TR: RR 5 2.81, P 5 0.005) as independent predictors of variance in TS prevalence estimates (Table 2). Post-hoc analyses of DSM-IV-TR versus DSM-III and DSM-IV were consistent with the results using DSM-III-R as a reference (DSM-IV-TR vs. DSM-III: RR 5 6.39, P 5 0.085; DSM-IV-TR vs. DSM-IV: RR 5 2.67, P < 0.001). Age was not significantly associated with prevalence. Of note, study date was no longer a predictor of TS prevalence after controlling for other covariates (RR 5 1.02; P 5 0.45).

Discussion Previously reported TS prevalence estimates have varied widely, an inconsistency that has significantly

M E T A - A N A L Y S I S




FIG. 1. Forest plot of TS prevalence meta-analysis using random-effects modeling. Random-effects meta-analysis of all 27 TS prevalence studies, including clinical series, stratified by clinical assessment method. Black dashed line indicates the overall point estimate including clinical series; gray dashed line indicates the overall point estimate after removing clinical series studies. ES, effect size, that is, prevalence per 100 children; I-squared, Higgins I2 statistic; p, Cochran’s Q test for heterogeneity.

limited the understanding of disease burden.6 Although many factors have been proposed to account for the 200-fold difference in reported prevalence rates, no formal heterogeneity analyses have previously been conducted. In the present study, we sought both to refine the TS prevalence estimate and investigate these potential sources of heterogeneity. We initially found that case-assessment method was a primary source of cross-study heterogeneity; studies based on clinically ascertained cases had 6-fold lower TS prevalence rates than population-based methods (Supporting Table 2). This is consistent with a previous report in which two parallel studies, one clinical series and one using population-based assessments, were conducted simultaneously in the same Swedish county and

found that the clinical series had a 10-fold lower TS rate than the comparable population-based study (0.15% vs. 1.1%), presumably because the vast majority of the population-based sample never sought treatment.3 After restricting analyses to population-based studies, only sample size and diagnostic criteria were significant predictors of cross-study heterogeneity. Sample size accounted for 47% of the variance in TS prevalence estimates between population-based studies. This strong inverse correlation between sample size and prevalence may be capturing differences between case-assessment procedures in a more sensitive manner than classifying assessment methods using categorical variables. For example, prevalence studies using the most reliable assessment method (direct

Movement Disorders, Vol. 30, No. 2, 2015


S C H A R F,




TABLE 2. Uni- and multivariable meta-regression of TS population-based prevalence studies Univariable Model RR (95% CI)



Diagnostic criteria


Study date Sample size

Omnibus test Two-stage (reference) Questionnaire only Two-stage plus Directly assessed Omnibus test Europe (reference) North America Far East Middle East Omnibus test DSM-III-R (reference) DSM-III DSM-IV DSM-IV-TR Omnibus test Age >12 (reference) Age

Population prevalence of Tourette syndrome: a systematic review and meta-analysis.

The aim of this study was to refine the population prevalence estimate of Tourette Syndrome (TS) in children and to investigate potential sources of h...
360KB Sizes 48 Downloads 4 Views