A randomised controlled trial of treatments of childhood anxiety disorder in the context of maternal anxiety disorder: clinical and cost‐effectiveness outcomes

Background This study evaluated whether clinical and economic outcomes from CBT for child anxiety disorders in the context of maternal anxiety disorders are improved by adding treatment focused on (a) maternal anxiety disorders or (b) mother–child interactions. Methods Two hundred and eleven children (7–12 years, 85% White British, 52% female) with a primary anxiety disorder, whose mothers also had a current anxiety disorder, were randomised to receive (a) child‐focused CBT with nonspecific control interventions (CCBT+Con), (b) CCBT with CBT for the maternal anxiety disorder (CCBT+MCBT), or (c) CCBT with an intervention targeting the mother–child interaction (CCBT+MCI). A cost‐utility analysis from a societal perspective was conducted using mother/child combined quality‐adjusted life years (QALYs). [Trial registration: https://doi.org/10.1186/isrctn19762288]. Results MCBT was associated with immediate reductions in maternal anxiety compared to the nonspecific control; however, after children had also received CCBT, maternal outcomes in the CCBT+MCI and CCBT+Con arms improved and CCBT+MCBT was no longer superior. Neither CCBT+MCBT nor CCBT+MCI conferred a benefit over CCBT+Con in terms of child anxiety disorder diagnoses post‐treatment [primary outcome] (adj RR: 1.22 (95% CI: 0.88, 1.67), p = .23; adj RR: 1.21 (95% CI: 0.88, 1.65), p = .24, respectively) or global improvement ratings (adj RR: 1.25 (95% CI: 0.99, 1.57), p = .06; adj RR: 1.18 (95% CI: 0.93, 1.50), p = .17) or six and 12 months later. No significant differences between the groups were found on the main economic outcome measures (child/mother combined QALY mean difference: CCBT+MCBT vs. CCBT+Con: −0.04 (95% CI: −0.12, 0.04), p = .29; CCBT+MCI vs. CCBT+Con: 0.02 (95% CI: −0.05, −0.09), p = .54). CCBT+MCI was associated with nonsignificantly higher costs than CCBT (mean difference: £154 (95% CI: −£1,239, £1,547), p = .83) but, when taking into account sampling uncertainty, it may be cost‐effective compared with CCBT alone. Conclusions Good outcomes were achieved for children and their mothers across treatment arms. There was no evidence of significant clinical benefit from supplementing CCBT with either CBT for the maternal anxiety disorder or treatment focussed on mother–child interactions, but the addition of MCI (and not MCBT) may be cost‐effective.


Introduction
Anxiety disorders are among the most common psychological disorders in childhood affecting 2.6%-5.2% of children under the age of 12 years (e.g. Costello, Egger, & Angold, 2004). They negatively affect children's functioning (Essau, Conradt, & Petermann, 2000), raise the risk for disorders in adolescence and adulthood (Woodward & Fergusson, 2001), and carry a substantial health and social cost (Bodden, Dirksen, & Bogels, 2008). There is considerable evidence for the efficacy of cognitive behaviour therapy (CBT) for these disorders, with 59% of anxious children no longer meeting criteria for an anxiety disorder following CBT (James, James, Cowdrey, Soler, & Choke, 2013). However, where a parent has an anxiety disorder, rates of child recovery have been found to be as little as half compared to when there is no parental anxiety disorder (Bodden, B€ ogels et al., 2008;Hudson et al., 2014). Two studies to date have examined whether specifically targeting elevated parental anxiety might benefit child treatment outcome (with treatments delivered in group formats). One of these studies reported a positive impact on anxious children of parents with high trait anxiety (Cobham, Dadds, & Spence, 1998), at least in the short term, while the other reported no benefit for anxious children of parents with a current anxiety disorder (Hudson et al., 2014). Notably, both of these studies administered relatively brief treatments for parental anxiety (4 and 3.75 hr, respectively) and neither brought about significant reductions in parental anxiety. The question therefore remains open whether successful treatment of parental anxiety might benefit child outcome.
It is possible that targeting parent-child interactions that are likely to interfere with the child's treatmentthat is reducing parent cognitions and behaviours that limit the child's autonomy and promote the child's focus on threat (Creswell, Murray, Stacey, & Cooper, 2011;Hudson & Rapee, 2004) may be of benefit for anxious children in the context of parental anxiety disorder (Settipani, O'Neil, Podell, Beidas, & Kendall, 2013). Indeed, parental cognitions characterised by expectations that the child will be frightened and feel out of control in the face of a challenge and behaviours characterised by intrusiveness and modelling of fear have been found to occur more commonly in anxious compared to nonanxious parents of children with anxiety disorders (Creswell, Apetroaia, Murray, & Cooper, 2013;Moore, Whaley, & Sigman, 2004).
In view of the roles of parental anxiety as well as parenting cognitions and behaviour in maintaining child anxiety, the current study tested the hypothesis that clinical outcomes for child anxiety disorders presenting in the context of parental anxiety disorders will be improved by supplementing individual child CBT with (a) treatment of parental anxiety disorders or (b) treatment focused on parent-child interactions.
This study also evaluated whether supplementing CBT with these adjunct treatments represents good value for money. The few published cost-effectiveness analyses of the treatment of child anxiety mainly concern the value of child CBT compared with a wider family approach (Bodden, B€ ogels et al., 2008;Bodden, Dirksen, & Bogels, 2008) or with a parent-directed treatment for child anxiety (Simon, Dirksen, B€ ogels, & Bodden, 2012). There is, however, a paucity of evidence on the short-term and long-term cost-effectiveness of alternative configurations of treatment types involving maternal and child treatment combinations, which are likely to be of particular relevance when maternal anxiety disorder is present. Furthermore, no economic evaluations of the treatment of child anxiety have taken into account both the direct effects to the child and 'spillover' effects, for example, to the parents and wider society. For outcomes, this is achieved by using quality-adjusted life years (QALY), a generic measure of outcome that can be aggregated across individuals (Weinstein, Torrance, & McGuire, 2009) and allows for examination of both direct treatment and spillover effects (Tilford et al., 2015). For costs, this is achieved by measuring broader impacts on education and employment. Aggregating direct and spillover effects may alter cost-effectiveness profiles, a particularly appropriate strategy when the economic evaluation adopts a societal perspective. Indeed, recommendations from a recent review of family-based therapies suggest that future research in this field should include such spillover effects to the rest of the family as well as consider cost and effectiveness over a longer time horizon (Tubeuf & Guthmuller, 2017). Further, few evaluations to date have formally aggregated QALYs as a means of capturing spillover effects. The current study therefore tested the value for money of the two supplementary interventions, considering spillover effects, societal impacts and including a one-year follow-up.

Participants
Participants comprised 211 children, aged 7-12 years (mean age = 10.22, SD = 1.58), with a current anxiety disorder, together with their mothers, who also had a current anxiety disorder. The study was powered to provide 90% power at the 5% (two-sided) significance level to detect a 30% difference in the primary outcomethat is the proportion of children who recovered from their primary anxiety disorder at the posttreatment assessment in the CCBT+MCI or CCBT+MCBT arms compared to the CCBT+Con arm. The estimated remission rate for the CCBT+Con group was 40% (based on Cobham et al., 1998). A difference of 30% in the proportion of anxiety-free children following completion of the treatment was considered to be the minimum that would be clinically worthwhile taking into account the increased resources required and change to service delivery that would be required if either of these interventions was found to be effective and implemented in practice. The sample size was defined on the basis of two independent trials as recommended by Machin, Campbell, Fayers, and Pinol (1997). The required sample size of 56 children per group was increased to allow for an estimated 20% loss to follow-up.
The study focussed on mothers as (a) intergenerational associations for anxiety disorders are most commonly found between mothers and their children (Cooper, Fearn, Willetts, Seabrook, & Parkinson, 2006), (b) mothers are most commonly the primary caregivers in the study region, and (c) maternal and paternal behaviours may differ in how they are associated with childhood anxiety (B€ ogels & Perotti, 2011).
Participants were recruited between June 2008 and May 2011, with the last follow-up assessment in February 2013. Participating children were all referred to a university research clinic for treatment by a health or educational professional. The inclusion criteria were (a) the child must (i) be aged 7-12 years; (ii) have a current primary diagnosis of a DSM-IV anxiety disorder (if specific phobia, this must be comorbid with another anxiety disorder); and (b) the child's mother must (i) be the primary carer; and (ii) meet criteria for a current DSM-IV anxiety disorder. The exclusion criteria were (i) significant physical or intellectual impairment (including autistic spectrum disorders) in the child or mother; (ii) current prescription of psychotropic medication for the child or mother that had not been at a stable dose for at least one month and without agreement to maintain that dose throughout the study; and (iii) the child had previously received six or more sessions of CCBT for an anxiety disorder. Participating children and mothers were given a complete description of the study orally and in writing before giving written informed assent/consent. A total of 676 children were assessed for eligibility; 435 families did not meet the inclusion criteria and 30 declined to participate (see Figure 1). Not assessed (n = 7) Reasons: • Mother in other therapy (n = 1) • Mother -did not attend (n = 5) • Withdrew-Child improved (n = 1) Not assessed (n = 5) Reasons: • Mother did not attend (n = 5) Not assessed (n = 1) Reasons: • Withdrew -Child improved (n = 1) Not assessed (n = 8) Reasons: • Did not attend (n = 5) • Treatment complete but declined assessment (n = 3) Not assessed (n = 5) Reasons: • Did not attend (n = 1) • Treatment complete but declined assessment (n = 4) Not assessed (n = 8)* Reasons: • Did not attend (n = 5) • Treatment complete but declined assessment (n = 2)* • Exclude -Mother serious mental health concerns (n = 1)
Randomisation was performed externally at the Centre for Statistics in Medicine (University of Oxford) on receipt of anonymised participant information by fax. Patients were randomised with a 1:1:1 ratio, with minimisation for child age and gender, type of child anxiety disorder, and baseline severity of both child and maternal primary anxiety disorder. These minimisation variables were selected as they have been most frequently (albeit not consistently) identified as predictors of response to treatment for child anxiety disorders (Ginsburg et al., 2011;Hudson et al., 2015). The minimisation algorithm included a 20% random element.
Assessments of maternal anxiety disorder and parenting were made before and immediately following the interventions. Assessments of child anxiety disorder status and severity were conducted before and following treatment, as well as at six and 12 months following treatment. Different research staff assessed the mother and the child. Assessments of maternal and child health-related quality of life were made on five occasions: before treatment (Assessment 1), mid-way through treatment (1B), immediately post-treatment (2), and at the six (3)-and 12-month (4) follow-up. Data on the use of other health and social care resources, as well as non-NHS cost-generating (e.g. educational) services and days off school or work, were obtained from parent report. All assessors were blind to treatment allocation. The trial was preregistered as follows https://doi. org/10.1186/isrctn19762288. The protocol is provided in Appendix S1.

Measures
Structured diagnostic interviews with children and parents. Children were assigned diagnoses on the basis of the Anxiety Disorders Interview Schedule for DSM-IV for Children -Child and Parent Versions (ADIS-C/P; Silverman & Albano, 1996). The presence or absence of a current maternal anxiety disorder was assigned on the basis of the Anxiety Disorders Interview Schedule for DSM-IV (ADIS-IV; DiNardo, Barlow, & Brown, 1994). For the ADIS-C/P overall diagnoses and clinical severity ratings (CSRs) were assigned if the child met diagnostic criteria on the basis of either child or parent report. Those with a CSR of 4 or more (at least moderate psychopathology) were considered to meet diagnostic criteria. Assessors were thoroughly trained to ensure high levels of reliability (presence/ absence of child diagnosis j = .98 (child report), .98 (mother report); CSR intraclass correlation = .99 (child report), .99 (mother report); presence/absence of maternal diagnosis j = .97; CSR intraclass correlation = .99). The primary outcome derived was absence of the child's primary (most impairing) anxiety disorder; secondary outcomes derived from the ADIS-c/ p were absence of all anxiety disorder diagnoses and CSR of the primary anxiety disorder, as prespecified in the trial protocol and statistical analysis plan (see Appendices S1 and S2).
Clinical Global Impressions: Improvement (CGI-I). Overall improvement in child anxiety was assessed using the Clinical Global Impression -Improvement scale (CGI-I), a seven-point scale from 1 = very much improved to 7 = very much worse; scores of 1 and 2 are accepted to represent treatment success (Guy, 1976). Overall mean inter-rater reliability for the assessment team was high (ICC = .96).
Symptoms of anxiety and comorbid difficulties were also assessed as secondary outcomes using the Spence Child Anxiety Scale (child, parent, teacher report), Child Anxiety Impact Scale (child, parent report), Short Mood and Feelings Questionnaire (child, parent report), conduct problems Total therapy sessions subscale of the Strengths and Difficulties Questionnaire (child, parent, teacher report) and a teacher report measure of Child Adjustment to School. Descriptions of measures and outcomes are provided in Table S1.
Assessment of parenting and parental expectations. In order to establish whether the intervention that targeted mother-child interactions did successfully alter maternal responses, we conducted laboratory observations under conditions of mild social, performance and physical challenge (Creswell et al., 2013). The social threat task involved the child preparing and delivering a speech, with their mother's support, to a research assistant with a hand-held video camera. The performance task involved the child attempting difficult tangram puzzles (following the procedure of Hudson & Rapee, 2001). The physical threat task required children to investigate the content of four chambers within a mysterious 'black box'. To account for prior experience, the assessment was modified at the posttreatment assessment point; in the social stress task, the child was required to present to a panel rather than a single research assistant, the tangram puzzles were more difficult, and the black box was accompanied by sound effects (i.e. rustling/scratching). Observers who were blind to treatment arm coded parental behaviours on scales developed by Murray et al. (2012) and adapted by Creswell et al. (2013) to be suitable for children aged 7-12 years. Ratings were given for each minute of the interaction on 5-point scales (1 = none, 5 = pervasive/strong). Since interactions varied in duration, total scores across tasks reflected the sum of the mean scores for each task. The following behaviours were considered: maternal expressed anxiety, overprotection, intrusiveness and positivity (warmth and encouragement). For each coder, in each task, a second coder independently scored a random sample of 25 videotapes. Intraclass correlations showed acceptable agreement across all indices (range .60-1.00; mean: .86).
Maternal expectations were assessed before initiating the challenge tasks (Creswell et al., 2013). Immediately after receiving the instructions for each task, mothers were taken to a separate room and asked to provide ratings regarding (a) how their child would feel about doing the task (0 = not scared at all, 10 = extremely scared); and (b) how much their child could do about how the task went (0 = nothing at all, 10 = a lot). Ratings were combined across the three tasks to represent their expectations across a range of challenge contexts.
Health-related quality of life. Mother and child healthrelated quality of life (HRQoL) was measured using responses to the EuroQol EQ-5D-3L (Dolan, Gudex, Kind, & Williams, 1995) and EQ-5D-Youth version (Wille et al., 2010) instruments, respectively. Responses for mother and child at each time point were converted into utility weights using the UK population tariff (Dolan et al., 1995). Total quality-adjusted life years (QALYs) were calculated for each mother and child, using the area under the curve approach after linear interpolation between the time points. Mother-child dyad QALYs were obtained by additively combining individual mother and child QALYs (Weinstein et al., 2009).
A broad, societal perspective (including parental productivity and school impacts) was adopted in assessing resource use and costs. In addition to the intervention costs (captured through therapists' logs completed at each therapy session), values were attached for measured school absence, time off work, and lost leisure time (for mothers), use of non-NHS services (e.g. educational services) and personal costs of medications, all of which was captured through parent diaries. An NHS perspective was also adopted as per NICE guidance (NICE, 2013).

Treatment
Child Cognitive Behaviour Therapy (CCBT). All children received eight-weekly one-hour sessions of individual CBT. Treatment was delivered by one of seven qualified clinical psychologists or cognitive behaviour therapists, following a manual based on the 'Cool Kids' (Lyneham, Abbott, Wignall, & Rapee, 2003) and 'Coping Cat' (Kendall & Hedtke, 2006) programmes (both of which have established efficacy; Hudson et al., 2009;Walkup et al., 2008). While these manuals include more than eight treatment sessions, there is evidence of good outcomes from similar CBT content for child anxiety disorders delivered in eight hours or less (e.g. Gallagher, Rabian, & McCloskey, 2004;Ginsburg & Drake, 2002). CCBT was supervised by a highly experienced, clinical psychologist and accredited cognitive behaviour therapist. Treatment focused on helping children to identify and challenge negative thinking styles, gradually increasing exposure to feared stimuli and developing problem-solving skills. Mothers were included at the beginning and end of each session to share progress and feedback. As described in Appendix S3, CCBT treatment adherence was equivalent across treatment arms.
CBT for maternal anxiety disorder (MCBT). MCBT consisted of eight one-hour weekly sessions delivered by one of five qualified clinical psychologists or cognitive behaviour therapists (all supervised by a highly experienced, clinical psychologist and accredited cognitive behaviour therapist) over 8 weeks. Given that mothers presented with a range of anxiety disorders and given evidence for the efficacy of transdiagnostic CBT treatments for anxiety disorders (Newby, McKinnon, Kuyken, Gilbody, & Dalgleish, 2015), we followed a manualised transdiagnostic treatment for adult disorders which aimed to reverse the putative cognitive behavioural maintaining mechanisms identified through individual formulation This manual has not previously been systematically evaluated; however, promising results were obtained in a case series evaluation (McManus, Clark, Muse, & Shafran, 2015).
In the treatment arms that did not involve MCBT, mothers received nondirective counselling (NDC), a supportive individual intervention that was not focussed specifically on reducing symptoms of anxiety. NDC was provided by one of four qualified counsellors (supervised by a highly experienced counsellor/psychotherapist) following the manual of Borkovec and Costello (1993). As shown in Appendix S3, the MCBT and NDC were clearly distinct. MCBT and NDC were delivered first, before the delivery of CCBT (see Table 1).
Mother-Child Interaction treatment (MCI). The MCI intervention consisted of 10 sessions delivered over 8 weeks by one of five qualified clinical psychologists or cognitive behaviour therapists (supervised by an experienced clinical psychologist): eight sessions were with the mother alone and two were with the mother and child together. This was a novel intervention designed to target potentially anxiogenic features of the mother-child relationship. Specifically, it aimed to enhance maternal autonomy promoting cognitions (such as confidence in the child's ability to face challenge) and behaviours, and reduce potentially anxiogenic behaviours. This was achieved through a combination of specific strategies from existing family interventions for childhood anxiety (Lyneham et al., 2003) with the addition of video-feedback techniques developed and piloted by the trial investigators. The two joint mother and child sessions involved the mother and child completing structured tasks that were video-recorded for reference in later sessions.
To balance therapist contact, sessions that focused on the promotion of a healthy lifestyle (Healthy Living Control, HLC) were delivered in the treatment arms that did not receive the MCI intervention. This intervention was principally concerned with family diet and exercise based on existing interventions applied within school settings (BDA, 2003) and was delivered by one of eight therapists (under supervision of an experienced clinical psychologist). MCI/HLC was delivered in parallel with CCBT. As shown in Appendix S3, MCI and HLC were clearly distinct.

Analysis
A comprehensive statistical analysis plan was prepared before embarking on the analysis (Appendix S4).
In line with our hypotheses and power analysis, we set out to make two sets of pairwise comparisons (MCBT+CCBT vs. CCBT+Con; and MCI+CCBT vs. CCBT+Con). Dichotomous outcome variables (e.g. diagnostic outcomes) were analysed using a modified Poisson regression approach with robust error variance adjusting for the minimisation factors [child age, child gender, type of child anxiety disorder (GAD, social phobia, SAD, other), baseline severity of the child's and the mother's primary anxiety disorder (ADIS Clinician Severity Rating)]. This approach allows for adjusted relative risks to be presented (Zou, 2004). All analyses are presented based on the intention-to-treat (ITT) population. Sensitivity analyses of the primary endpoints included: (a) no adjustment for minimisation criteria, (b) per-protocol population (i.e. those participants who had received at least half of the treatment sessions and had data for the post-treatment assessments) and (c) multiple imputation analysis (see Appendix S2). Primary outcome results did not change based on the prespecified sensitivity analyses, nor when best-and worst-case scenarios were used in the context of the multiple imputation analysis. Continuous outcomes were modelled using linear regression, adjusted for baseline scores and minimisation factors. No adjustments were made to the confidence intervals or p-values to account for multiplicity.
Current best-practice methods were adhered to for conducting and reporting economic evaluations alongside trials (Drummond, Sculpher, Claxton, Stoddart, & Torrance, 2015;Husereau et al., 2013;NICE, 2013;Petrou & Gray, 2011). A broad societal perspective cost-utility analysis framework was adopted for the base case to assess the cost-effectiveness of (a) CCBT+MCBT compared to CCBT; and (b) CCBT+MCI compared with CCBT. Costs of the nonspecific treatments (NDC/HLC) were excluded in the base-case analyses. This was in recognition of the fact that the main aim of any economic evaluation is to estimate the likely cost-effectiveness of a change in clinical practice in real-world settings, rather than in ideal controlled conditions (as in efficacy trials). Those costs were instead included in some of the sensitivity analyses. Costs were expressed in pounds sterling (£) at 2011/2012 prices. Given the short time-frame of the trial and follow-up, discounting was not applied to costs or effects. The base-case analyses were performed on an ITT basis. Mean imputation methods were used for missing resource use and health outcomes deemed highly deterministic (e.g. face-to-face therapists contact), and multiple imputation for other resources (e.g. use of medications), under the assumption of missing at random (Faria, Gomes, Epstein, & White, 2014). For each mother and child participant, all components of treatment costs, stratified by category of resource use and other broader societal costs (educational services, travel costs, time off school/work) were calculated by multiplying units of resource use by their unit costs (see Table S2). These values were then summed to obtain a total cost for each child/mother. Effects were identified and measured using child/mother QALYs, derived from the EQ-5D-Y and EQ-5D-3L child and mother report, respectively. Incremental mean costs and effects, and the associated 95% CIs, were estimated comparing the two intervention groups for each of the two comparisons (CCBT+MCBT vs. CCBT; and CCBT+MCI vs. CCBT). Incremental cost-effectiveness ratios (ICERs) were estimated and reported where relevant. Uncertainty in the cost-effectiveness results was analysed using cost-effectiveness acceptability curves (CEACs) over a range of potential threshold values that the health system and the wider society might be willing to pay for an additional QALY gained (Fenwick, Marshall, Levy, & Nichol, 2006). Extensive sensitivity analyses were performed for each of the two basecase comparisons (CCBT+MCBT vs. CCBT; and CCBT+MCI vs. CCBT) and are detailed in Table S3. All analyses were conducted using Stata Software: Release 12, StataCorp 2011.

Results
Baseline characteristics were well balanced across treatment arms (see Table 2). Numbers of available participants for each treatment arm are shown in Figure 1. No important harms due to any of the interventions or trial procedures were reported during the trial.

Manipulation checks
We first examined the extent to which delivery of the adjunct treatments was associated with change in the putative maintenance mechanisms that were targeted; maternal anxiety disorder status and change in maternal cognitions and behaviours. MCBT was successful in changing maternal anxiety disorder status. At Assessment 1B (i.e. after completing MCBT), 58.5% (n = 38) of mothers in the MCBT arm had recovered from their primary diagnosis compared to 36.5% (n = 23) in the CCBT+Con arm (who received NDC) (adj RR: 1.63 (95% CI: 1.13, 2.36), p = .009). The recovery rate in the MCI arm was not different from that in the CCBT+Con arm (who both received NDC) (adj RR: 1.22 (95% CI: 0.83, 1.81), p = .31).
Change in observed overprotection (across all three mother-child tasks) at the end of all treatment (Assessment 2) was greater in the CCBT+MCI arm compared to CCBT+Con (adjusted mean difference in change from baseline, (À0.03 (À0.06, À0.004), p = .03). There was no difference in change in observed overprotection between the CCBT+MCBT and CCBT+Con arms (À0.02 (À0.05, 0.009), p = .17). There were no significant differences between CCBT+Con and either CCBT+MCBT or CCBT+MCI in terms of change in observed intrusiveness, positive behaviours or maternal expressed anxiety (See Table S4 and S5).
Differences between the CCBT+Con and CCBT+MCI arms were also found on change in measures of maternal expectations of how scared the child would be (À0.67 (À1.26, À0.07); p = .03) and how in control the child would feel (0.53 (0.01, 1.05); p = .05), with mothers in the CCBT+MCI arm predicting that their child would be less scared and more in control. While a similar pattern was found, the corresponding differences between the CCBT+Con and CCBT+MCBT arms did not quite reach the conventional level of statistical significance (Scared: À0.53 (À1.12, 0.07), p = .08; Control: 0.50 (À0.03, 1.02), p = .06). A summary of the results of statistical analyses for maternal anxiety, behaviours and cognitions is given in Table S5.

Economic evaluation
Missing resource use data were prevalent (Table S6). Mother and child combined HRQoL results are presented in Table S7 (unadjusted differences) and Tables S8 and S9 (differences adjusted for child's and mother's baseline utilities). At 12 months, in the ADNOS, anxiety disorder not otherwise specified; CCBT+Con, child cognitive behaviour therapy + nonspecific control interventions; CSR, clinical severity rating; MCBT, CCBT + maternal cognitive behaviour therapy; MCI, CCBT + mother-child interaction treatment; MDD, major depressive disorder; OCD, obsessive-compulsive disorder; PTSD, post-traumatic stress disorder. a Children with specific phobia and selective mutism were only included if also met diagnostic criteria for another comorbid anxiety disorder. b These mothers had a current anxiety disorder but MDD had the highest severity.
Key cost drivers in both of these base-case comparisons were the differential treatment delivery costs in both the CCBT+MBCT and CCBT+MCI arms compared with CCBT only (Tables S11 and S12). While the CCBT+Con arm had higher absences from school and work costs than the other two arms, few differences were statistically significant although, notably, the lower costs of mother's lost days of work in the CCBT+MCI arm than in the CCBT+Con arm were significant (Table S12).
Taking sampling uncertainty into consideration, the cost-effectiveness acceptability curves (CEACs) for the two base-case analyses showed that, in view of the joint distribution of incremental mean costs and effects, the CCBT+MBCT arm was not likely to be a cost-effective alternative to CCBT (Figure 2, Panel 1, and Table S8, base-case line). The probability that the CCBT+MCI arm was cost-effective compared to CCBT (Figure 2, Panel 2, and Table S9 base-case line) was between 60% and 64% based on UK National Institute for Health and Care Excellence thresholds for accepted levels of willingness to pay for an additional QALY (usually between £20,000 and £30,000). Sensitivity analyses reinforced the findings of the base-case analyses, with the CCBT+MCBT arm not being cost-effective compared with the CCBT arm across the majority of sensitivity analyses (Table S8) [other than the scenario where the CCBT+MCBT arm was less costly but also less effective (SA 4 and SA 10)]. For the comparison between the CCBT+MCI arm and CCBT (Table S9), the CCBT+MCI arm may be a costeffective alternative to CCBT from a societal perspective (insignificantly higher costs or less costly, insignificantly higher QALYs), as reflected by the probability of cost-effectiveness being greater than 60% overall.

Discussion
Good outcomes were achieved for both children and their mothers across all treatment arms but neither adding treatment of maternal anxiety (MCBT) nor treatment of the mother-child relationship (MCI) to individual child-focused CBT conferred a significant benefit to children with anxiety disorders whose mothers also had a current anxiety disorder. Although adding both adjunct treatments achieved higher child recovery rates post-treatment than the CCBT plus control treatment arm (where neither maternal anxiety nor potentially anxiogenic parenting received specific therapeutic attention), the advantages were not statistically significant. However, from a broad societal perspective combining mother/child QALYs and costs, including productivity and school impacts, the economic evaluation indicated that CCBT+MCI arm (but not the CCBT+MCBT arm) may be cost-effective compared Table 3 Primary and secondary categorical (child) outcomes Initial assessment (1) Post-treatment assessment (2) 6-month post-treatment assessment (3) 12-month post-treatment assessment (4) to CCBT. Notably, the probability of the CCBT+MCI arm being cost-effective lessened when a more restricted NHS perspective was adopted and the gains from reduced school absenteeism and mothers' productivity/employment excluded. This indicates that costs associated with childhood anxiety disorders are likely to be underestimated if a purely healthcare provider perspective is adopted, and it affirms the importance of including wider family/ society born costs and outcomes in mental health economic evaluations (Creswell et al., 2017;Tilford et al., 2015;Tubeuf & Guthmuller, 2017). Despite these potential broader economic benefits, it is important to consider the lack of significant clinical effects of the two active adjunct interventions (MCI/MCBT). The study was powered to detect a 30% difference in the proportion of recovered children, based on a projected recovery rate from child CBT of 40% (Cobham et al., 1998). While a 30% difference in recovery between CBT plus the adjunct treatments and CBT plus the control interventions might be considered conservative (as smaller differences may still be clinically meaningful), an anticipated success rate of 40% from individual CBT could have been considered optimistic, given findings from a more recent trialthe first to examine treatment of child anxiety disorder in the context of diagnosed parental anxiety disorderwhere a recovery rate of only 33% was found (Hudson et al., 2014). However, the CCBT plus control treatment arm in the current study performed considerably better than expected, with 48% of children being free of their primary anxiety disorder and 64% 'much' or 'very much' improved immediately post-treatment, and over 70% free of their primary diagnosis and 'much/very much improved' a year after treatment. Indeed, these success rates are similar to those found from more intensive (14 session) CCBT for children with anxiety disorders, regardless of parental anxiety disorder status, where the proportions 'much'/'very much' improved were 60% at post-treatment (Walkup et al., 2008) and 72% at six-month follow-up (Piacentini et al., 2014), respectively. The unexpectedly high rate of recovery within the CCBT plus control arm in the current study is unlikely to be a function of particular features of our sample, as our study population was comparable to other clinic samples in the literature. There could, however, have been an added value of the nonspecific interventions, all of which provided some level of parental support. This raises the possibility that the modest outcomes typically found for anxious children in the context of parental anxiety disorder are in fact a result of other associated factorsfor example, other stressors experienced by the family (see, e.g. Schleider et al., 2015) which might have been addressed to some extent in this trial by the generic support received by all mothers.
Given the novelty of the adjunctive treatments, the degree to which they were successful in altering their respective targets also needs to be considered. In the one previous study that assessed the impact of adding CBT for parental anxiety disorders to CBT for child anxiety disorders (Hudson et al., 2014), the failure to find significant differences in child outcomes may have been attributable to the parental CBT not reducing parental anxiety. In the current study, however, compared to nondirective counselling, MCBT was associated with a significant Adjusted for child age, child gender, type of child anxiety disorder (GAD, social phobia, SAD, other), baseline severity (ADIS-C/P CSR) of the child's primary anxiety disorder and baseline severity (ADIS-IV mother self-report) of the mother's primary anxiety disorder. CCBT+Con, child cognitive behaviour therapy + nonspecific control interventions; CCBT+MCBT, CCBT + maternal cognitive behaviour therapy; CCBT+MCI, CCBT + mother-child interaction treatment; CGI-I, Clinical Global Impression -Improvement; Ref., reference category.   CAIS-c/p, Child Anxiety Impact Scale child/parent report; CCBT+Con, child cognitive behaviour therapy + nonspecific control interventions; CCBT+MCBT, CCBT + maternal cognitive behaviour therapy; CCBT+MCI, CCBT + mother-child interaction treatment; SCAS-c/p, Spence Children's Anxiety Scalechild/parent report; SDQ-p, Strengths and Difficulties Questionnaireparent report; SMFQ-c/p, Short Mood and Feelings Questionnairechild/parent report. reduction in the frequency of maternal anxiety disorder. However, after the children had received treatment, there was no longer a significant difference in maternal disorder status, with all treatment arms showing relatively high rates of recovery from maternal disorders. This is consistent with previous findings indicating that reductions in child anxiety are associated with reductions in parental anxiety (Settipani et al., 2013). We conducted observational assessments of parental responses before and after treatment for child anxiety disorders which provided evidence that, in terms of a relative change in overprotective behaviours, the mother-child interaction treatment was successful. The MCI intervention was also associated with change in maternal cognitions associated with confidence in child coping (i.e. reduced predictions regarding child fear and increased predictions regarding child control). Despite these positive benefits, no significant benefit to child outcome was conferred. Possible reasons for this may be that the changes were not of sufficient magnitude to be of benefit, or, as noted above, that change in these factors is not in fact critical to changing child anxiety. Notably, there was no specific benefit for the MCI intervention on measures of maternal expressed anxiety, intrusiveness or positive behaviours. While it is possible that the intervention was ineffective with respect to these dimensions, it may well be that these parental behaviours changed equally across groups in response to improvements in child anxiety (Silverman, Kurtines, Jaccard, & Pina, 2009). However, it may also reflect limitations in the extent to which these laboratory-based observational tasks assess parenting dimensions as they are expressed in the home.
The study had several notable strengths, including the use of reliable, blind raters to make assessments of child and maternal anxiety and maternal behaviours and cognitions before and after treatment, the inclusion of nonspecific interventions to balance therapist contact, a design which allowed for isolating the effects of specifically targeting maternal anxiety and parenting responses, and the inclusion of a broad societal perspective, prospectively designed economic evaluation measuring and valuing both mother and child costs and outcomes. These strengths need to be considered in the light of various limitations. Although we allowed for 20% loss to follow-up, by the one year post-treatment, assessment retention was down to 61% in the CCBT+Con arm. Although there were no clear baseline differences between completers and those who dropped out, it is of concern that the greatest dropout occurred during the eight-session maternal counselling phase which may not have been a sufficiently acceptable treatment approach for some families. Whether dropouts overrepresented those with good or bad treatment outcomes cannot be determined, although the sensitivity analyses that were conducted gave a consistent pattern of results which suggest that this was not the case. Other limitations include the relatively restricted demographic characteristics of the participants, who were predominantly of nonminority ethnicity and relatively high socioeconomic status. We elected to focus on middle childhood (7-12 years) as anxiety disorders can be reliably diagnosed at this age, particular parental behaviours have been observed in the context of parent anxiety disorder at this age (Creswell et al., 2013), and it is likely that the nature of parental influences on child anxiety varies with child age (Connell & Goodman, 2002;Maccoby, 1992). We also focused on intervening with mothers as parental influences have also been shown to differ according to parent gender (B€ ogels & Perotti, 2011). As a result, however, the findings cannot be generalised to younger children, to adolescents or to interventions with fathers or other caregivers. The study also included children and mothers with a broad range of anxiety disorders, and further work needs to be done which takes account of the precise form of parental and child anxiety.
For the health economics analyses, data on resource use and costs of additional health and personal social services beyond the NHS treatment costs were characterised by a high percentage of missing data. However, the economic results were confirmed by multiple sensitivity analyses, including a variation that best approximated a complete-case scenario. The base-case economic evaluations excluded the costs of the two nonspecific interventions (i.e. family healthy living and nondirective counselling for mothers), which were introduced to match for therapist time/contact across the trial arms, and are not commonly delivered in 'real-world' settings. While the sensitivity analyses that were conducted including those costs confirmed and reinforced the base-case results, it should be acknowledged that the nonspecific interventions may not have been completely neutral in their effect. However, in the absence of a valid counterfactual, we are unable to quantify that potential impact. Furthermore, this study should be regarded as providing an indication of the short-term likely cost-effectiveness of CCBT+MCBT or CCBT+MCI compared to CCBT, with further research required to determine cost-effectiveness in the longer term. Finally, generalising these economic findings to other populations and settings (e.g. different healthcare systems) is challenging and, with the added complexity associated with capturing the broader societal impact as well as the combined mother/child impacts, the cost-effectiveness results are best interpreted with caution.
In sum, the current study showed that good clinical outcomes can be achieved for children with anxiety disorders in the context of maternal anxiety disorder by providing high-quality individual child CBT together with some parental support. However, although offering CBT to mothers for their own anxiety disorders gave an early boost to the timing of maternal recovery, it did not lead to a significant clinical benefit in terms of child anxiety treatment outcome and (given the higher costs) was not costeffective. An adjunctive treatment to target parenting responses had a nonsignificant positive effect on child outcomes, which, despite the higher costs, may be cost-effective. The findings also suggest that reductions in child anxiety may have a positive effect on maternal mental health.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article: Appendix S1. Study protocol. Appendix S2. Statistical analysis plan. Appendix S3. Adherence and fidelity.

Key points
Treatment outcomes for children with anxiety disorders are impaired in the context of concurrent parental anxiety disorder but it is unclear how to improve outcomes.
In this RCT for children with anxiety disorders whose mothers also had a current anxiety disorder, individual child-focused CBT was supplemented with treatments to (a) improve maternal anxiety (MCBT) or (b) alter maternal responses to her child when faced with challenge (MCI).