Intent-to-treat analyses were conducted including all 488 randomized participants, regardless of study completion and/or compliance. Among them, 439 (90%) had complete Week 12 assessment data. Multiple imputation was used to impute data for the 49 participants who did not provide Week 12 data.
Remission Versus Response Rates
presents remission and response rates for each treatment group. McNemar chi-square tests, estimated within a multiple imputation framework, examined whether response and remission rates differed significantly from each other. Results indicated that the sample remission rate was significantly lower than the sample response rate for all definitions of remission, the loss of AD diagnoses (f = 29.11, df = 1, ddf = 100, p < .0001), the CGI-S score of 1 or 2 (f = 19.35, df = 1, ddf = 77, p < .0001), and the CGI-I score of 1 (f = 63.43, df = 1, ddf = 86, p < .0001).
Response and Remission Rates (With Confidence Intervals in Brackets) of Child/Adolescent Anxiety Multimodal Study Subjects at Week 12
Remission Rates Across Treatment Conditions
To investigate whether remission status varied by treatment condition, we conducted a series of three logistic regression models. Each model also included several covariates: child's age, gender, minority status, socioeconomic status, and treatment site. Six contrasts for each model (comparing each of the four treatment conditions to each other) were conducted. Results for the loss of AD diagnoses indicated that participants in all three active treatments were significantly more likely to remit than participants in the PBO condition (see ). Furthermore, participants in the COMB condition were more likely to remit than participants in the SRT and CBT conditions. SRT and CBT remission rates did not significantly differ from each other.
Between-Group Comparisons of Remission Rates Based on Loss of Anxiety Disorder Diagnoses
Results from the models for CGI-S and CGI-I remission revealed only one significant difference (see and ). Using these definitions for remission, participants in the COMB treatment condition were more likely to remit than participants in the CBT condition. Despite this, neither COMB, CBT, nor SRT participants exhibited significantly different CGI-I or CGI-S remission rates from participants in the PBO condition.
Between-Group Comparisons of Remission Rates Based on Clinical Global Impression Severity Scale Score of 1 or 2
Between-Group Comparisons of Remission Rates Based on Clinical Global Impression Improvement Scale Score of 1
Predictors of Remission
In a second set of analyses, we conducted a series of single-predictor logistic regressions, within a multiple imputation framework, in which each of 10 potential variables was examined as predictors of remission. Four demographic variables (i.e., child's age, gender, minority status, and socioeconomic status), four baseline child clinical variables (i.e., anxiety severity as measured by the CGI-S, primary AD diagnosis, presence of a comorbid internalizing disorder [i.e., anxiety or depressive disorder] other than the primary three [i.e., GAD, SAD, and SOP] treated as part of this study, presence of a comorbid externalizing disorder), and two baseline measures of parental psychopathology (BSI Global scale and STAI total score) were examined. In a last step, we combined significant predictors together in one model to examine which uniquely contributed to predicting remission rates.
Age (measured in year increments) significantly predicted remission status on the ADIS-C/P, b = −0.007, t(464) = −2.65, p = .01, odds ratio (OR) = 0.92, 95% confidence interval (CI) [0.86, 0.98], and CGI-S, b = −0.006, t(443) = −2.11, p = .04, OR = 0.54, 95% CI [0.42, 0.71], such that older children were less likely to enter remission (see ). Minority status also predicted remission on the ADIS-C/P, b = −0.73, t(331) = −2.97, p = .003, OR = 2.07, 95% CI [1.94, 2.21], such that minority children were less likely to enter remission (see ). Neither gender nor socioeconomic status significantly predicted remission status for any definition of remission. None of the demographic variables significantly predicted CGI-I remission status.
Remission Rates by Treatment Condition and Minority Status
Baseline Child Clinical Variables
Higher anxiety on the CGI-S significantly predicted reduced likelihood of remitting for all definitions of remission, loss of AD diagnoses on the ADIS-C/P: b = −0.61, t(437) = −4.51, p = .0001, OR = 0.54, 95% CI [0.42, 0.71]; CGI-S: b = −0.64, t(387) = −4.62, p = .0001, OR = 0.64, 95% CI [0.48, 0.84]; CGI-I: b = −0.45, t(461) = −3.16, p = .002, OR = 0.52, 95% CI [0.40, 0.69].
Binary variables coded to denote the presence or absence of a diagnosis at baseline for GAD, SAD, and SOP were entered together in models for each definition of remission. There were no significant interactions among diagnoses for any definition of remission; thus, interaction terms were not included in the models. Results revealed that participants with baseline SOP were significantly less likely to achieve remission than those participants without SOP on the ADIS-C/P, b = −0.91, t(473) = −3.65, p = .003, OR = 0.40, 95% CI [0.24, 0.66], and the CGI-S, b = −0.65, t(415) = −2.64, p = .01, OR = 0.52, 95% CI [0.32, 0.85], at Week 12 but not on the CGI-I score of 1, b = −0.36, t(461) = −1.43, p = .15 (see ). No significant differences were found for participants with or without GAD and SAD.
Remission Rates by Treatment Condition for Participants With and Without Social Phobia
At baseline, 44% of the sample (n
= 215) met criteria for one or more of the following internalizing disorders, in addition to the child's target diagnosis/diagnoses of GAD, SAD, and/or SOP: selective mutism, obsessive-compulsive disorder, panic disorder, agoraphobia, specific phobia, major depressive disorder, major depressive disorder not otherwise specified, and dysthymic disorder. Eighteen percent (n
= 90) met criteria for one or more of the following externalizing disorders: attention-deficit/hyperactivity disorder, attention-deficit/hyperactivity disorder not otherwise specified, conduct disorder, and oppositional defiant disorder (see Kendall et al., 2010
, for a full description of comorbidities in the CAMS sample). The presence of a comorbid internalizing disorder was associated with a reduced likelihood of achieving remission in Week 12 as assessed by the ADIS-C/P, b
= −0.38, t
(459) = −2.04, p
= .04, but was not statistically associated with the CGI-S, b
= −0.36, t
(306) = −1.84, p
= .07, or CGI-I, t
(437) = −0.89,p
= .37, remission status (see ). Comorbid externalizing disorder did not significantly predict remission status for any definition of remission, ADIS-C/P: t
(396) = −0.81, p
= .42; CGI-I: t
(322) = 0.04, p
= .89; and CGI-S: t
(261) = −0.81, p
Remission Rates by Treatment Condition for Participants With and Without Additional Internalizing Comorbidity
Parent Global BSI scores did not significantly predict Week 12 remission status on the ADIS-C/P, t(381) = −0.07, p = .94; the CGI-I, t(418) = −0.12, p = .90; or the CGI-S, t(254) = 0.29, p = .77. Similarly, parental trait anxiety did not predict Week 12 remission status, ADIS-C/P: t(439) = 0.69, p = .49; CGI-I: t(411) = −0.13, p = .90; and CGI-S: t(392) = 1.01, p = .31.
Combined Predictors of Remission
In this last step in these analyses, we tested models in which baseline CGI-S score, the presence of SOP, and internalizing comorbidity, together with age and minority status, predicted remission status for each different definition of remission. Results indicated that baseline CGI-S score, b = −0.53, t(432) = −3.80, p = .0002; SOP, b = −0.77, t(467) = −2.93, p = .004; and minority status, b = −0.68, t(321) = −2.69, p = .001, all significantly predicted ADIS-C/P remission status when included in the same model. Internalizing comorbidity, b = −0.37, t(452) = −1.88, p = .06, and age, b = −0.05, t(456) = −1.77, p = .08, were not predictors of ADIS-C/P remission in this model. In contrast, only baseline CGI-S score, b = −0.64, t(384) = −4.20, p < .0001, and SOP, b = −0.51, t(403) = −1.98, p = .05, significantly predicted CGI-S remission status when included, together with comorbid internalizing, age, and minority status, in the same model. Last, only baseline CGI-S significantly predicted CGI-I remission status, b = −0.42, t(460) = −2.87, p = .004, when all significant predictors were included in the same model.