Sage Journals: Discover world-class research

Abstract

Background and aims

Despite recent gains in the amount and quality of early autism intervention research, identifying what works for whom remains an ongoing challenge. Exploiting data from the Preschool Autism Communication Trial (PACT), we undertook secondary analysis to explore prognostic indicators and predictors of response to one year of PACT therapy versus treatment as usual within this large and rigorously characterised cohort recruited across three UK trial sites.

Methods

In this secondary analysis of variability in child gains on the primary trial outcome measure – social-communication symptom severity – we used a pragmatic and data-driven approach to identify a subgroup of children who showed reliable improvement and a subgroup showing clear lack thereof. We then tested which among several baseline child and family factors – including measures routinely collected in research trials and clinical practice – varied as a function of child outcome status and treatment group.

Results

Greater baseline child non-verbal ability was a significant prognostic indicator of symptom reduction over time (i.e. irrespective of treatment group). By contrast, parent synchrony presented as marginal predictor, and trial recruitment site as a significant predictor, of differential outcome by treatment group. Specifically, lower parent synchrony showed some association with poorer outcomes for children from families assigned to treatment as usual (but with no such effect for those assigned to PACT). Similarly, children at one recruitment site were more likely to have poorer outcomes if assigned to treatment as usual, compared to children at the same site assigned to PACT.

Conclusions

The current data contribute to an evidence base indicting that early non-verbal ability is a robust indicator of generally better prognosis for young children with autism. Lower parent synchrony and a broadly more deprived socio-geographical context may inform the appropriate targeting of PACT. That is, given that the former factors predicted poorer outcome in children from families assigned to treatment as usual, the receipt of a relatively low-dose, parent-mediated and communication-focused therapy might be developmentally protective for young children with autism. Nevertheless, results from this study also highlight the paucity of meaningful predictors of outcome among routine clinical characterisation measures such as those investigated here.

Implications

Understanding the factors associated with differential treatment outcomes is critical if we are to individualise treatment decisions for children with autism. Inherently tied to this objective is a need to delineate those factors which specifically predict positive response (or lack of response) to one or other treatment option, versus those that indicate generally better (or poorer) prognosis, irrespective of treatment.

Keywords

Autism early intervention prognostic indicators predictors of response symptom severity

Autism is highly heterogeneous, with core symptoms, cognitive and adaptive skills, and associated behaviour difficulties varying widely (Jeste & Geschwind, 2014). While systematic reviews and meta-analyses demonstrate the promise of various early childhood intervention approaches (e.g. Kasari, Gulsrud, Paparella, Hellemann, & Berry, 2015; Oono, Honey, & McConachie, 2013; Smith & Iadarola, 2015), heterogeneity remains a major barrier to determining best-practice (Romanczyk, Callahan, Turner, & Cavalari, 2014; Spence & Thurm, 2010). Even where trials demonstrate overall efficacy, some children show little improvement (Magiati, Tay, & Howlin, 2012) and, conversely, some children make considerable gains in the absence of intervention (Howlin, Magiati, & Charman, 2009). To further the field, it is important that we identify those children with autism most likely to benefit from a given intervention approach and any for whom the approach may be contraindicated (Kraemer, Wilson, & Fairburn, 2002). Three key barriers to this end include (1) inconsistent use of terminology and methods, in studies reported to date, for delineating predictors of positive response (or lack of response) to a particular treatment option versus prognostic indicators of generally better (or poorer) outcome irrespective of treatment; (2) lack of guidelines for what constitutes meaningful improvement (or lack thereof) for young children with autism; and (3) limited numbers of adequately designed and well-powered studies to address this broad aim.

Prognostic indicators versus predictors of response

The extant literature highlights a variety of characteristics associated with more positive outcomes for young children with autism. For example, milder autism symptom severity and better cognitive, linguistic and social skills have been shown to be associated with more positive outcomes in intervention studies (see Vivanti, Prior, Williams, & Dissanayake, 2014) as have parent/family factors such as lower stress, greater motivation for and fidelity to treatment, higher socio-economic status (SES) and stronger support networks (e.g. Karst & Van Hecke, 2012; Sherer & Schreibman, 2005). Further – and particularly relevant to the context of parent-mediated interventions – child outcomes have been associated with features of the parent–child relationship (Perryman et al., 2013). What remains unclear, however, is whether and which of these represent true predictors of positive response to intervention, as opposed to prognostic indicators of greater ‘natural’ improvement over time.

Many studies inadequately separate predictors from prognostic indicators. Evidence of a prognostic indicator presents as a main effect (e.g. where children with characteristic X are observed to have better outcomes than those with characteristic Y). By contrast, evidence of a predictive effect presents only in a significant interaction term (e.g. where, in treatment condition A, children with characteristic X are observed to have better outcomes than those with characteristic Y, but this is not the case in treatment condition B). Hence, evaluation of predictors versus prognostic indicators must necessarily occur within the context of some comparative or controlled investigation of one treatment condition versus another. However, the lack of sufficient rigorously conducted, well-powered autism intervention studies means that we are not yet able to draw definitive and appropriate conclusions about what works for whom.

Further complicating efforts toward this end, the existing evidence concerning associations between baseline characteristics and later outcomes for children with autism is also inconsistent. While milder symptoms and better cognitive/linguistic abilities tend to correlate with positive outcomes (e.g. Virués-Ortega, 2010), this is not true across all trials (e.g. Howlin et al., 2009) and these same variables often present as prognostic indicators of outcome in the absence of intervention (Magiati, Tay, & Howlin, 2014). Findings are similarly mixed regarding characteristics of parents and features of parent-child interaction behaviour (e.g. Perryman et al., 2013; Siller & Sigman, 2008; Vivanti et al., 2014). Further, other potentially relevant sources of prognostic or predictive influence – such as family ethnicity and cultural values (Kasari & Patterson, 2012) and genetic and endophenotypic factors (Parr, Gray, Wigham, McConachie, & LeCouteur, 2015) – remain virtually unexplored in the field of autism intervention.

What constitutes meaningful improvement?

An associated challenge for the field concerns the operationalisation of positive child outcomes. Rather than following best-practice guidelines to specify primary and secondary outcomes a priori (e.g. National Institute for Health and Care Excellence, 2013), autism intervention researchers often examine multiple outcomes, including cognitive/linguistic skills and adaptive behaviours (Magiati et al., 2012). In the context of intervention for pre-schoolers with autism, a key objective is often to achieve reduction in core symptom presentation (Bieleninik et al., 2017). Achieving this aim has proved challenging (e.g. Waddington, van der Meer, & Sigafoos, 2016), and there is also little consensus on what might constitute meaningful reduction in core symptoms (e.g. see McConachie et al., 2015) nor on how to determine the appropriate target difference when planning a randomised controlled trial (Cook et al., 2014). Gold-standard measures to quantify autism symptom presentation (Lord et al., 2000; Lord, Rutter, & Le Couteur, 1994) were developed to inform diagnostic decision-making, rather than to measure sensitively change over time or in response to treatment. Nevertheless, as recently synthesised by Bieleninik et al. (2017), several prospective cohort studies as well as a small number of randomised controlled trials (e.g. Dawson et al., 2010; Green et al., 2010; Solomon, Van Egeren, Mahoney, Quon Huber, & Zimmerman, 2014) have adopted gold-standard diagnostic measures as ambitious targets by which to evaluate intervention outcome.

Background to this study: The Preschool Autism Communication Trial

As already described, a third key barrier to understanding how to individualise early autism intervention concerns the lack of studies adequately designed to address predictors of response. Given the proliferation of available treatment options, it is critical to determine whether children are more or less likely to benefit from particular approaches, and this requires thorough assessment of factors associated with differential outcome (e.g. such as for students with learning disabilities (Fuchs, Fuchs, & Compton, 2012) but is rare in studies of autism). Beyond formal intention-to-treat (ITT), moderation and mediation analyses, secondary exploratory analysis of clinical trial datasets may elucidate factors informative for designing subsequent, hypothesis-driven trials (Kraemer et al., 2002).

The Preschool Autism Communication Trial (PACT; Green et al., 2010) was an efficacy randomised controlled trial (RCT) of a parent-mediated, communication-focussed therapy conducted across three UK sites. Among 152 parent–child dyads, around half were assigned to receive PACT therapy (n = 77) versus community treatment as usual (TAU; n = 75). While Green et al. reported no significant overall ITT effect on the a priori nominated primary outcome measure – severity of child social-communication symptoms – at 13-month treatment endpoint (d = −.24; 95% CI: −.59, −.11), substantial variability in outcomes was apparent. Nevertheless, pre-specified moderator analysis revealed no significant predictive effects (i.e. statistical interaction terms) among the following baseline measures: child age, symptom severity, non-verbal ability, receptive language, family SES or trial recruitment site.

Current aims and hypotheses

We undertook secondary analysis to explore the substantial variability in child outcomes apparent within the PACT dataset. Using a pragmatic and data-driven approach, we examined which among various measures of child and parent/family factors collected at baseline might differentiate those children who went on to make improvement and clear lack of improvement in their core social-communication symptoms across the year-long trial period. To mirror the pragmatic context of a clinician seeking to tailor treatment to characteristics of the presenting client, we focused our examination on baseline measures commonly collected both within research trials and as part of clinical service provision – initial child symptom severity and cognitive, linguistic and adaptive skills, and family socio-demographic characteristics – noting that these have also previously been examined as potential associates of outcome within past research (e.g. Vivanti et al., 2014). As this was a multi-site trial and PACT was a parent-mediated intervention, we also included recruitment site in our investigation (see Carr et al., 2015; Nunes et al., 2010) and a measure of parent interaction with the child (i.e. parent synchrony; Aldred, Green, & Adams, 2004; Hudry et al., 2013).

Lack of consensus in the literature precluded us from drawing strong conclusions around likely specific predictors of response to PACT. However, following Magiati et al. (2014) and Siller and Sigman (2008), we hypothesised that child cognitive/linguistic abilities and parent synchrony might be prognostic indicators of symptom reduction (i.e. presenting as main effects of child outcome). Our particular interest, however, was to understand whether these or other factors might predict specific response to PACT versus TAU.

Methods

Participants and design

Among the 152 dyads enrolled into PACT, 146 (96%) were retained to 13-month outcome (Green et al., 2010). At trial entry, children were aged 24 to 59 months, had ≥12 months non-verbal age-equivalence (NVAE; Mullen, 1995) and met algorithm cut-offs for autism on at least two domains of the Autism Diagnostic Interview – Revised (ADI-R; Lord et al., 1994) as well as algorithm cut-off for autism on the Autism Diagnostic Observation Schedule – Generic (ADOS-G; Lord et al., 2000).

Outcome measure

Our primary interest was to ascertain meaningful change on the primary trial outcome measure – ADOS-G social-communication algorithm score (Lord et al., 2000) – with higher scores indicating greater symptom severity. As described above, the ADOS-G was administered at baseline as part of our procedure to confirm child diagnosis. This was re-administered at 13-month trial outcome, with slight adaptation to standard protocol, for the measure of primary outcome.

Typically, the appropriate ADOS-G module is selected according to participant age and language level and then comprises a set of relevant semi-structured tasks focusing on social interaction between the examiner and the participant. Module 1 is used with children who are minimally verbal and Module 2 with those using short phrase speech. However, as reported by Green et al. (2010), children in this trial were administered the same module at outcome assessment as had been administered at baseline, to avoid discontinuity of scores due to the change in module that would typically occur for a child who developed phrase speech. Further, slight adaptation to the conversion of raw item codes to the algorithm was applied, such that raw codes of 3 were retained as such (rather than rescaled to 2) and raw codes of 8 were rescaled to 3 (rather than to 0), thereby ensuring more sensitive measurement of improvement over time.

Administration was by individuals trained to high standard and participating in regular coding meetings to ensure maintenance of research-level reliability. Coding of outcome ADOS-G assessments was conducted blind to treatment group. Intra-class correlation (ICC) computed from 66 ratings made across 15 tapes evidenced strong agreement among researchers for the standard social-communication algorithm score (ICC = 0.79) and when coding modifications were applied (ICC = 0.83).

Characterisation measures

Additional standardised examiner-administered and parent-report assessments were conducted at baseline assessment, alongside the collection of parent–child interaction footage. As described above, for further confirmation of child diagnosis, the Autism Diagnostic Interview-Revised (ADI-R; Lord et al, 1994) was administered, providing a second metric of autism symptom severity. The ADI-R is a structured interview conducted with parents, spanning the child’s full developmental history. It comprises three principal diagnostic domains – social interaction skills and difficulties, communication skills and difficulties, and restricted and repetitive behaviours – and includes cutoff scores suggestive of autism on each of these as well as for a symptom onset domain.

Developmental and language abilities were assessed in several ways. The visual reception and fine motor scales of the examiner-administered Mullen Scales of Early Learning (MSEL; Mullen, 1995) were used to quantify non-verbal developmental abilities in the children, with a NVAE score computed as the average of these domains. The parent-report MacArthur-Bates Communicative Development Inventories, Words and Gestures form (MCDI, Fenson, Dale, & Reznick, 1993) was administered to obtain raw counts of child receptive vocabulary knowledge. An expressive vocabulary count was also available from the MCDI, and Auditory Comprehension and Expressive Language standard scores were obtained from administration of the Preschool Language Scales – 3rd Edition, UK Adaptation (Zimmerman, Steiner, & Pond, 1997). However, these were omitted from the main analysis due to observed multi-collinearity with MSEL NVAE and MCDI receptive vocabulary.

Parent–child interaction

Child pragmatic communication toward the parent, and also parent interaction behaviour toward the child, was quantified from filmed 8-minute free-play interaction samples via the Dyadic Communication Measure for Autism coding scheme (see Aldred et al., 2004; Green et al., 2010; Hudry et al., 2013). Specifically, coding of child communication focused on classifying initiations (i.e. spontaneous communication acts directed toward the parent) and responses (i.e. acts following on from a previous parental contribution). Coding of parent behaviour focused on classifying communication acts as synchronous (i.e. supportive of the child’s attentional focus and commenting on the child’s play/activity) or asynchronous (i.e. directive/redirective or placing some demand on the child’s attention or behaviour). For analysis, proportionate Child Initiation was computed as the proportion of all child communication acts that were initiations, and proportionate Parent Synchrony was computed as the proportion of all parent communication acts that were synchronous. Moderate to good inter-rater agreement was evident across these scales (computed from 66 ratings made across 22 tapes): Child Initiation, ICC = .59; Parent Synchrony, ICC = .80.

Family and treatment-related factors

Demographic data were collected via semi-structured parent interview and questionnaire responses, and the following were dichotomous variables were coded for analysis: family composition (dual vs. single parent status), ethnicity (both parents white vs. one/both parents non-white), household Income (less than vs. equal to or greater than £40,000 per annum) and parental educational attainment (less than vs. equal to or greater than one parent with some qualification post 16 years of age). As already described, families were recruited from three trial sites, situated in South London, Manchester and the North East of England. Among families assigned to the PACT arm, the number of treatment sessions attended by families was also recorded for a measure of PACT dose.

Analysis plan

We sought to delineate subgroups of children who improved/did not improve on their ADOS-G scores. While a 4-point reduction has been suggested to represent clinically meaningful improvement (Aldred et al., 2004), this was based on data from a relatively small sample. Hence, we adopted Jacobson and Truax’s (1991) Reliable Change Index (RCI), enabling identification of children whose improvement/deterioration could be considered psychometrically reliable (vs. attributable to variability of scores/measurement error; see also Eldevik, Jahr, Eikeseth, Hastings, & Hughes, 2010).

Following Jacobson and Truax (1991; formula below), we computed RCI z scores for the change in ADOS-G from baseline to outcome. Our index of reliability (r) was the correlation between baseline and outcome ADOS-G scores among children randomised to TAU (n = 72; test–retest r = .735), and our index of variance was the standard deviation (SD) of ADOS-G scores at baseline (n = 152; SD = 4.135):
$\begin{matrix} RCI z score \\ = \frac{(ADOS baseline score - ADOS outcome score)}{\sqrt{2 (SD \sqrt{(1 - r)})^{2}}} \end{matrix}$

We then sought to delineate two child subgroups – reliable Improvers and Non-Improvers – where reliable improvement was defined by RCI z scores <−1.96 and reliable deterioration by RCI z scores >1.96 (i.e. both signifying change unlikely to have occurred by chance; p < .05).

We then planned to conduct two-way between-subjects analyses of variance (ANOVAs) – two outcome status (Improvers and Non-Improvers) × two treatment groups (PACT and TAU) – on continuous baseline characteristics of interest or χ² contingency analyses in the case of categorical measures. Again, preliminary analysis indicated multi-collinearity among many measures of development and language ability/knowledge taken with child participants. Hence, key baseline characteristics retained for these analyses were:
parent-reported child autism symptoms: sum of ADI-R algorithm items,

assessment of child developmental level: NVAE from MSEL, with log-10 transformation was applied to address significant positive skew in the raw data,

parent-reported receptive vocabulary count from the MCDI,

proportionate Child Initiation and Parent Synchrony coded from parent–child interaction, with square-root transformation applied to address significant positive skew in raw data for the former,

parent-reported demographic characteristics: family composition, ethnicity, household income and highest level of education, and

trial recruitment site and, where relevant, dose of PACT therapy received.

From these analyses, prognostic indicators would be evidenced by significant main effects of outcome status, and predictors of treatment response would be evidenced by significant two-way (treatment group by outcome status) interaction terms. Subsequently, we planned to include significant baseline characteristics within a logistic regression to identify which singly/together contributed predictive value for child Improver versus Non-Improver outcome status.

Results

Reliable symptom improvement (i.e. RCI z < −1.96) was identified for 43 children, with this threshold corresponding to a ≥6-point decrease in ADOS-G scores. No child showed reliable deterioration in autism symptoms (i.e. RCI z > 1.96), so we made the post hoc decision to classify as Non-Improvers those children who made no change or showed any increase in ADOS-G scores over time. Hence, Non-Improvers comprised those 41 children whose RCI z ≥0, reflecting 0–5-point increase in ADOS-G scores.

Table 1 shows baseline data for the whole cohort and for Improver and Non Improver subgroups. We also show data for the remaining 62 children who showed intermediate, non-reliable change (i.e. RCI z = −1.96 to 0) but were then excluded from further analysis.
Table 1.
Baseline characteristics of total cohort and subgroups of reliable Improvers, Non-Improvers and children making intermediate gains.

Cohort (N = 146) Improvers (N = 43) Intermediate (N = 62) Non-Improvers (N = 41)

Child characteristics

Age (months) 44.7 (7.9) 45.0 (7.7) 44.7 (8.0) 44.7 (8.2)

ADI-R algorithm score 34.3 (6.7) 34.0 (6.9) 34.8 (6.6) 33.8 (6.8)

MSEL non-verbal age-equivalence^a 26.3 (10.0) 29.5 (10.6) 24.6 (8.5) 25.5 (10.8)

MCDI receptive vocabulary^b 162 (120) 178 (110) 144 (125) 172 (122)

Parent–child interaction measures

DCMA Child Initiation 13.7 (9.8) 14.9 (8.7) 12.2 (9.8) 14.8 (10.6)

DCMA Parent Synchrony 28.4 (11.9) 29.2 (10.9) 28.6 (11.7) 27.3 (13.4)

Family measures

Single (vs. dual) parent family 23% 16% 21% 32%

Both parents white (vs. other) 58% 67% 55% 54%

Income <£40K (vs. ≥£40K) 65% 65% 63% 68%

Parental education (≥1 post 16 years) 73% 74% 71% 76%

Site^a: South London N = 50 N = 9 N = 21 N = 20

Manchester N = 50 N = 18 N = 21 N = 11

North East England N = 46 N = 16 N = 20 N = 10

Note. Data are mean (standard deviation) unless otherwise indicated. Descriptive statistics based on scores prior to log-10/square-root transformation.

ADI-R: Autism Diagnostic Interview – Revised (Lord et al., 1994); MSEL: Mullen Scales of Early Learning (Mullen, 1995); MCDI: MacArthur Bates Communicative Development Inventories (Fenson et al., 1993); DCMA: Dyadic Communication Measure for Autism (Aldred et al., 2004; Hudry et al., 2013).

^aItalicised values indicate significant between-group differences.

^bData missing on five cases.

Full data supporting the principal analyses reported below are shown in the online Appendix A. Table 2 shows ANOVA results for the continuous baseline characterisation measures. A significant main effect of outcome status presented for baseline NVAE, such that this was greater among children who were Improvers (M = 29.5, SD = 10.6) versus Non-Improvers (M = 25.5, SD = 10.8), with moderate effect size (d = .46). The two-way interaction term approached significance for parent synchrony, such that this was somewhat lower among parents from families assigned to TAU where children were Non-Improvers (M = 23.9, SD = 9.2), compared to both those families assigned to TAU where children were Improvers (M = 30.5, SD = 10.2) and to those assigned to PACT, irrespective of child outcome status (Improvers M = 28.3, SD = 11.4; Non-Improvers M = 31.3, SD = 16.4).
Table 2.
Results of ANOVAs comparing continuous baseline characteristics across Improver and Non-Improver subgroups randomised to PACT and TAU.

Main effects
Two-way interaction

Outcome status Treatment group

Chronological age F(1,80) = .08, p = .783, d = .04 F(1,80) = .13, p = .723, d = .07 F(1,80) = .48, p = .492, η²= .01

ADI-R autism symptoms F(1,80) = .17, p = .685, d = .04 F(1,80) = 1.54, p = .218, d = .27 F(1,80) = 1.05, p = .310, η²= .01

MSEL non-verbal ability F(1,80) = 4.08, p = .047, d = .46 F(1,80) = .06, p = .806, d = .11 F(1,80) = .10, p = .757, η²= .00

MCDI receptive vocabulary F(1,78) = .08, p = .774, d = .05 F(1,78) = .55, p = .462, d = .16 F(1,78) = .19, p = .667, η²= .00

DCMA Child Initiation F(1,79) = .29, p = .591, d = .13 F(1,79) = .87, p = .353, d = .22 F(1,79) = 2.07, p = .154, η²= .03

DCMA Parent Synchrony F(1,79) = .45, p = .506, d = .15 F(1,79) = .93, p = .337, d = .23 F(1,79) = 3.28, p = .074, η²= .04

Note. ADI-R: Autism Diagnostic Interview – Revised (Lord et al., 1994); MCDI: MacArthur Bates Communicative Development Inventories (Fenson et al., 1993); DCMA: Dyadic Communication Measure for Autism (Aldred et al., 2004; Hudry et al., 2013); MSEL: Mullen Scales of Early Learning (Mullen, 1995); PACT: Preschool Autism Communication Trial; TAU: treatment as usual; ANOVA: analysis of variance.

Table 3 shows χ² contingency test results for the categorical baseline characterisation measures. A significant association presented for trial recruitment site such that, among children from South London, there were more Non-Improvers than Improvers. Children recruited from Manchester and North-East England, however, were more often Improvers than Non-Improvers. Follow-up analysis indicated this association to apply only within the TAU trial arm such that, among families assigned to TAU, there were more Non-Improvers than Improvers at the South London site (vs. similar proportion of Improvers/Non-Improvers in TAU at the other two sites). Finally, among families assigned to PACT, therapy dose did not differentiate Improvers from Non-Improvers, z = .45 p = .650.
Table 3.
Results of chi-square contingency analyses comparing categorical baseline characteristics across improver and non-improver subgroups randomised to PACT and TAU.

Overall contingency Within PACT group Within TAU group

Family composition: single versus dual parent χ²(1) = 2.75, p = .097 χ²(1) = .82, p = .365 χ²(1) = 1.66, p = .198

Ethnicity: both parents white versus other χ²(1) = 1.67, p = .196 χ²(1) = 1.46, p = .227 χ²(1) = .55, p = .458

Household income: <£40K (vs. ≥£40K) χ²(1) = .10, p = .758 χ²(1) = .89, p = .345 χ²(1) = 1.46, p = .226

Parental education: <1 ≥ parent post 16 years χ²(1) = .02, p = .900 χ²(1) = .63, p = .426 χ²(1) = .01, p = .945

Recruitment site χ²(1) = 7.20, p = .027 χ²(1) = 1.84, p = .398 χ²(1) = 8.57, p = .014

Note. PACT: Preschool Autism Communication Trial; TAU: treatment as usual.

We included child NVAE and interaction terms of treatment group by each of recruitment site and parent synchrony within a logistic regression on Improver versus Non-Improver outcome status. The overall model was statistically significant – χ²(3) = 13.63, p = .009; Cox & Snell R²= .151; Nagelkerke R²= .202 – increasing correct classification from 50.6% (no predictors entered) to 66.3%. However, only the treatment group × recruitment site interaction (i.e. South London vs. others) carried unique predictive value (Wald = 5.28, p = .022), with non-significant unique effects for NVAE (Wald = 2.64, p = .104) and treatment group × parent synchrony (Wald = 0.19, p = .663).

Given the unique association of recruitment from the South London site with poorer child outcomes – particularly for families not assigned to the PACT trial arm – we conducted further post hoc exploration of cross-site differences in attempt to identify possible explanations for this effect at this particular site. As shown in Table 4 and consistent with the ANOVA results described above, families recruited from this site had parents with lower synchrony and children with lower verbal and non-verbal abilities and greater autism symptom severity. Families from this site were also more often single-parent and in the lower household income band than those at both other sites and had somewhat lower educational attainment than families from North-East England.
Table 4.
Baseline characterisation differences between families recruited into TAU trial arm at three sites.

South London (N = 26) Manchester (N = 24) North East England (N = 22) Difference between South London and other two sites

Child characteristics

Age (months) 42.8 (8.2) 45.4 (8.5) 47.4 (7.4) t(70) = 1.83, p = .072; d = .40

ADI-R algorithm score 35.9 (7.0) 35.0 (7.6) 33.4 (5.4) t(70) = .97, p = .336; d = .21

MSEL non-verbal age-equivalence (months) 22.2 (7.6) 25.6 (8.3) 29.1 (12.1) t(70) = 2.20, p = .022; d = .59

MCDI receptive vocabulary 136 (122) 162 (110) 199 (141) t(67) = 1.41, p = .163; d = .35

Parent–child interaction measures

DCMA Child Initiation 11.3 (8.5) 15.4 (10.0) 15.2 (11.3) t(70) = 1.37, p = .176; d = .30

DCMA Parent Synchrony 23.9 (11.3) 28.3 (11.7) 30.7 (10.1) t(70) = 2.06, p = .044; d = .49

Family measures

Single (vs. dual) parent family 39% 21% 14% χ²(1) = 3.93, p = .047

Both parents white (vs. other) 15% 75% 82% χ²(1) = 26.60, p < .001

Household income <£40K (vs. ≥£40K) 85% 75% 64% χ²(1) = 2.01, p = .157

Parental education (≥1 post 16 years) 54% 54% 77% χ²(1) = .90, p = .342

Note. Data are mean (standard deviation) unless otherwise indicated. Italicised values indicate significant between-group differences.

ADI-R: Autism Diagnostic Interview – Revised (Lord et al., 1994); MSEL: Mullen Scales of Early Learning (Mullen, 1995); MCDI: MacArthur Bates Communicative Development Inventories (Fenson et al., 1993); DCMA: Dyadic Communication Measure for Autism (Aldred et al., 2004; Hudry et al., 2013); PACT: Preschool Autism Communication Trial; TAU: treatment as usual.

Furthermore, examination of data on families’ access to TAU services revealed that families recruited from South London versus from Manchester and North East England had additional contacts with health, social and other therapeutic supports across the trial period. These included, on average, two more contacts with health service providers (including outpatient hospital visits or inpatient/emergency admissions, appointments with general practitioners, nurses, health visitors, paediatricians, dentists, ophthalmologists, audiologists and dieticians/nutritionists), five more contacts with providers of therapy services (including portage workers, educational/clinical psychologists, occupational therapists, physiotherapists and speech language therapists), and three more contacts with social care services (including social workers, family support workers, community autism specialists, home care workers, play workers or visits to walk-in centres). As shown in Table 5, however, none of these between-group differences was statistically significant and so differential access to TAU services is unlikely to explain the observed site-specific effects on child outcome.
Table 5.
Numbers of contacts with services by families assigned to the TAU arm at the South London versus two other trial recruitment sites.

South London Manchester/ North East England Coefficient 95% CI p

N 26 43

Health contacts 7.31 (5.56) 9.26 (6.65) −1.95 (−5.05 to 1.16) 0.215

Therapy contacts 13.78 (13.43) 19.39 (18.26) −5.61 (−14.31 to 3.09) 0.202

Social care contacts 6.69 (15.60) 3.21 (8.39) 3.48 (−2.28 to 9.24) 0.232

Note. PACT: Preschool Autism Communication Trial; TAU: treatment as usual; CI: confidence interval.

Discussion

Recent reviews highlight the need for treatment to be better adapted to individual children with autism, particularly in light of limited resources (e.g. Romanczyk et al., 2014). However, identifying factors associated with differential outcomes presents a significant challenge, and existing studies have often failed to differentiate true predictors of treatment response from indicators of generally better prognosis. Exploiting the large PACT dataset (Green et al., 2010) and retaining focus on the a priori nominated primary outcome measure of change in core symptom severity – for which substantial variability was observed – we adopted a pragmatic and data-driven approach to try to identify for whom this particular parent-mediated, communication-focussed treatment might have been most appropriate. We identified subgroups of children who were reliable improvers and non-improvers in autism symptom severity and then examined whether commonly available baseline characteristics varied between these subgroups, seeking to mirror the process that might be adopted by community professionals to guide individualised treatment decision-making for pre-schoolers with autism.

Non-verbal developmental level as a prognostic indicator

Among the large cohort, followed over around one year, ∼30% of children showed decreases in their social-communication symptoms of autism to an extent that was highly unlikely to have occurred by chance (and corresponding to a ≥6-point reduction on the ADOS-G modified algorithm total score). Similarly, a subgroup of ∼28% of children showed a clear lack of improvement or some worsening in their symptoms across the same period. Nevertheless, among the various baseline factors available for our analysis – child age, initial autism symptom severity, receptive language knowledge, social initiations and parent/family socio-demographic factors – we found a striking lack of differences between our Improver and Non-Improver subgroups.

One clear exception to this pattern of results was for child non-verbal ability, which presented as a significant prognostic indicator of outcome for children in this trial. This replicates findings from other studies (see Magiati et al., 2014) such that children who made reliable improvement in their autism symptoms had shown greater initial non-verbal ability than those whose symptoms did not improve. That this effect was prognostic – applicable to children who received the PACT intervention or community TAU – rather than predictive of response within one particular condition or other, was identifiable within this dataset given sound experimental design with random assignment of children to intervention arms. Conversely, many studies of early autism intervention are prospective cohort studies, rather than intervention trials (see Bieleninik et al., 2017) within which it is not possible to disentangle prognostic indicators – such as non-verbal ability – from true predictors of treatment response.

The significance of parental communicative synchrony

Alongside our clear prognostic effect of NVAE, we also observed a trend toward significant predictive effect for parent communicative synchrony. That is, a marginal interaction of outcome status by treatment group presented, such that baseline parent synchrony was somewhat lower specifically among TAU families with children who were non-improvers, compared to both TAU families with children who were improvers and also to families assigned to PACT therapy irrespective of child outcome. This effect should be interpreted with caution, pending future replication work. However, we have recently demonstrated the mechanistic importance of parent synchrony, through mediation analysis on the PACT dataset, for (1) proximal gains in child initiation skills and (2) more distal reduction in child symptoms. That is, to the extent that participation in PACT had an overall (small, non-significant) effect on child symptoms (Green et al., 2010), this was achieved through a large proximal treatment effect on parent synchrony which brought moderate downstream improvements in child communicative initiations, and (attenuated) reductions in core symptom presentation (Pickles et al., 2015). The current result – a marginally significant predictive effect of reduced parent synchrony for poorer child outcomes, in the context of one year of receipt of community TAU – provides further support for the importance of parent interaction behaviour in the context of early childhood autism. In the absence of participation in a therapy that focuses on supporting parent sensitive interaction with their child, greater ‘natural’ parent synchrony may be a protective factor against (and lower ‘natural’ synchrony a risk factor for) poorer child outcome during the preschool period. That is, for parents with more limited communicative synchrony, participation in PACT therapy may mitigate the odds of the child maintaining or increasing their core social-communication symptoms.

The significance of trial recruitment site

Finally, we observed a significant predictive effect of trial recruitment site, such that children recruited from South London showed poor outcomes if assigned to TAU, but more balanced odds of core symptom improvement if assigned to receive PACT therapy. We had included trial site within our analysis, because of its potential impact on factors such as treatment session attendance and adherence to therapy processes (e.g. see Carr et al., 2015).

When entered into the logistic regression on child outcome status, recruitment site carried unique predictive value in this dataset, with no further variance explained by child non-verbal ability or parent synchrony. Exploratory post hoc analysis of community services accessed by families across the three trial sites provided no evidence that this effect might be underpinned by the amount or type of TAU available to families in South London versus Manchester and North East England. However, this former cohort differed from those at the latter two sites in having higher representation of (a) children with greater baseline autism symptoms, (b) poorer verbal and non-verbal skills, (c) parents with lower communicative synchrony and (d) more single-parent and culturally/ethnically diverse households. As noted by Kasari and Patterson (2012) research with under-represented populations is very limited in the field of autism, with few trials involving ethnically or culturally diverse samples. The current findings suggest that the provision of routine TAU may be inadequate, particularly for more disadvantaged families, even within relatively high resource countries.

Considerations and future directions

Within the PACT cohort, we were able to identify a subgroup of children who showed clear improvement in their autism symptoms, using Jacobson and Truax’s (1991) RCI. We did not, however, identify any children who experienced reliable symptom deterioration. While this is clinically encouraging, evidence from adult mental health studies indicates that the predictors of positive outcomes may differ from those associated with negative treatment response (Starcevic & Brakoulias, 2008). Thus, further work is required to understand not only for whom particular interventions may be helpful but also whether there are any children for whom an approach may be contraindicated.

This field also lacks good, functional indicators of outcome for young children with autism. Nevertheless, evaluation of the natural developmental growth suggests the possibility of divergent trajectories for core symptom presentation versus adaptive behaviour, thereby justifying ongoing focus on differential types of intervention outcome (Szatmari et al., 2015). Here, retaining the a priori aim of the PACT trial (Green et al., 2010; Pickles et al., 2015), our analysis focus was on changes in child core autism symptoms. However, other constructs may be equally important indicators of outcome and other measures may be more sensitive to change over time (e.g. social-communication skills, adaptive behaviour, quality of life; McConachie et al., 2015). Future consideration is therefore warranted regarding the predictors of change in secondary outcome measures within this RCT dataset, including parent synchrony as the direct proximal target of PACT therapy sessions and child initiation as the proximal target outcome for children.

Statistical power for the current analysis may have been limited. However, we adopted this pragmatic alternative to formal moderation analysis to explore the predictors of clearly (i.e. reliably) positive child outcome or clear lack thereof. Moreover, we verified that our observed associations of symptom change with NVAE, parent synchrony and site also held when more conventional analysis was undertaken retaining both the full participant sample and symptom change as a continuous metric. Furthermore, our failure to identify additional prognostic or predictive associations is unlikely to be due to limited statistical power as the effect sizes observed here for non-significant between-group comparisons were small. Rather, it seems likely that broad characterisation measures – such as age, core symptom severity and developmental/cognitive ability – are insufficiently sensitive to serve as predictors of treatment response in young children with autism. Indeed, other groups have begun to explore more specific indicators of early skills – such as the exploration and functional use of objects, joint attention, imitation, and social motivation and response to social reward – as potential predictors of treatment response (see Vivanti et al., 2014).

Summary and conclusion

In summary, this study contributes to a growing empirical evidence base for the prognostic value of child non-verbal ability – contextualising this clearly as a prognostic indicator of child outcome irrespective of therapy group assignment. Further, these data provide some further indication of the potential importance of parent synchrony as a protective factor in autism (see also Pickles et al., 2015). Finally, these results highlight the importance of considering trial recruitment site and what comprises TAU when interpreting outcomes from intervention trials. These findings require replication in the context of other evaluations of autism treatment efficacy – including other parent-mediated approaches and therapist-delivered interventions alike. Our approach should equally be applied within community-based evaluations of treatment effectiveness.

Tasked with identifying and providing treatment approaches best suited to particular children and families, clinicians face the potential misuse of costly or limited resources, and possible exacerbation of child behaviour problems and parent stress from wasted time and effort (March, 2009; Winburn et al., 2014). Hence, it is critical to expand our understanding of what works for whom to develop an evidence base around individualising treatment plans for young children and their parents.

Supplemental Material

Appendix -Supplemental material for Predictors of reliable symptom change: Secondary analysis of the Preschool Autism Communication Trial

Supplemental material, Appendix for Predictors of reliable symptom change: Secondary analysis of the Preschool Autism Communication Trial by Kristelle Hudry, Helen McConachie, Ann Le Couteur, Patricia Howlin, Barbara Barrett, Vicky Slonims and The PACT Consortium in Autism & Developmental Language Impairments

	Cohort (N = 146)	Improvers (N = 43)	Intermediate (N = 62)	Non-Improvers (N = 41)
Child characteristics
Age (months)	44.7 (7.9)	45.0 (7.7)	44.7 (8.0)	44.7 (8.2)
ADI-R algorithm score	34.3 (6.7)	34.0 (6.9)	34.8 (6.6)	33.8 (6.8)
MSEL non-verbal age-equivalence^a	26.3 (10.0)	29.5 (10.6)	24.6 (8.5)	25.5 (10.8)
MCDI receptive vocabulary^b	162 (120)	178 (110)	144 (125)	172 (122)
Parent–child interaction measures
DCMA Child Initiation	13.7 (9.8)	14.9 (8.7)	12.2 (9.8)	14.8 (10.6)
DCMA Parent Synchrony	28.4 (11.9)	29.2 (10.9)	28.6 (11.7)	27.3 (13.4)
Family measures
Single (vs. dual) parent family	23%	16%	21%	32%
Both parents white (vs. other)	58%	67%	55%	54%
Income <£40K (vs. ≥£40K)	65%	65%	63%	68%
Parental education (≥1 post 16 years)	73%	74%	71%	76%
Site^a: South London	N = 50	N = 9	N = 21	N = 20
Manchester	N = 50	N = 18	N = 21	N = 11
North East England	N = 46	N = 16	N = 20	N = 10

	Main effects	Two-way interaction
Chronological age	F(1,80) = .08, p = .783, d = .04	F(1,80) = .13, p = .723, d = .07	F(1,80) = .48, p = .492, η²= .01
ADI-R autism symptoms	F(1,80) = .17, p = .685, d = .04	F(1,80) = 1.54, p = .218, d = .27	F(1,80) = 1.05, p = .310, η²= .01
MSEL non-verbal ability	F(1,80) = 4.08, p = .047, d = .46	F(1,80) = .06, p = .806, d = .11	F(1,80) = .10, p = .757, η²= .00
MCDI receptive vocabulary	F(1,78) = .08, p = .774, d = .05	F(1,78) = .55, p = .462, d = .16	F(1,78) = .19, p = .667, η²= .00
DCMA Child Initiation	F(1,79) = .29, p = .591, d = .13	F(1,79) = .87, p = .353, d = .22	F(1,79) = 2.07, p = .154, η²= .03
DCMA Parent Synchrony	F(1,79) = .45, p = .506, d = .15	F(1,79) = .93, p = .337, d = .23	F(1,79) = 3.28, p = .074, η²= .04

	Overall contingency	Within PACT group	Within TAU group
Family composition: single versus dual parent	χ²(1) = 2.75, p = .097	χ²(1) = .82, p = .365	χ²(1) = 1.66, p = .198
Ethnicity: both parents white versus other	χ²(1) = 1.67, p = .196	χ²(1) = 1.46, p = .227	χ²(1) = .55, p = .458
Household income: <£40K (vs. ≥£40K)	χ²(1) = .10, p = .758	χ²(1) = .89, p = .345	χ²(1) = 1.46, p = .226
Parental education: <1 ≥ parent post 16 years	χ²(1) = .02, p = .900	χ²(1) = .63, p = .426	χ²(1) = .01, p = .945
Recruitment site	χ²(1) = 7.20, p = .027	χ²(1) = 1.84, p = .398	χ²(1) = 8.57, p = .014

	South London (N = 26)	Manchester (N = 24)	North East England (N = 22)	Difference between South London and other two sites
Child characteristics
Age (months)	42.8 (8.2)	45.4 (8.5)	47.4 (7.4)	t(70) = 1.83, p = .072; d = .40
ADI-R algorithm score	35.9 (7.0)	35.0 (7.6)	33.4 (5.4)	t(70) = .97, p = .336; d = .21
MSEL non-verbal age-equivalence (months)	22.2 (7.6)	25.6 (8.3)	29.1 (12.1)	t(70) = 2.20, p = .022; d = .59
MCDI receptive vocabulary	136 (122)	162 (110)	199 (141)	t(67) = 1.41, p = .163; d = .35
Parent–child interaction measures
DCMA Child Initiation	11.3 (8.5)	15.4 (10.0)	15.2 (11.3)	t(70) = 1.37, p = .176; d = .30
DCMA Parent Synchrony	23.9 (11.3)	28.3 (11.7)	30.7 (10.1)	t(70) = 2.06, p = .044; d = .49
Family measures
Single (vs. dual) parent family	39%	21%	14%	χ²(1) = 3.93, p = .047
Both parents white (vs. other)	15%	75%	82%	χ²(1) = 26.60, p < .001
Household income <£40K (vs. ≥£40K)	85%	75%	64%	χ²(1) = 2.01, p = .157
Parental education (≥1 post 16 years)	54%	54%	77%	χ²(1) = .90, p = .342

	South London	Manchester/ North East England	Coefficient	95% CI	p
N	26	43
Health contacts	7.31 (5.56)	9.26 (6.65)	−1.95	(−5.05 to 1.16)	0.215
Therapy contacts	13.78 (13.43)	19.39 (18.26)	−5.61	(−14.31 to 3.09)	0.202
Social care contacts	6.69 (15.60)	3.21 (8.39)	3.48	(−2.28 to 9.24)	0.232

Footnotes

Supplementary materials

Supplementary material for this paper can be found at

Acknowledgements

The authors thank all the families and the referring professionals who contributed to the trial. The authors are grateful for comments on this paper from members of the PACT Consortium, in particular, Tony Charman, Sarah Byford, Jonathan Green and Andrew Pickles.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research represents secondary analysis of data from the Preschool Autism Communication Trial (PACT) which was sponsored by the University of Manchester and funded by the Medical Research Council (G0401546), the UK Department for Children, Schools and Families and the UK Department of Health. PACT was also supported by trial steering, data monitoring and ethics committees, the UK Mental Health Research Network and the UK National Autistic Society.

Members of the PACT Consortium

Catherine Aldred, Laura Blazey, Katy Bourne, Jonathan Green, Clare Harrop, Dharmi Kapadia, Kathy Leadbitter, Wendy Macdonald, Carol Taylor, Lydia White (University of Manchester); Tori Houghton (Stockport Primary Care Trust); Sarah Byford, Tony Charman, Andrew Pickles (Institute of Psychiatry, Psychology & Neuroscience, King’s College London); Karen Beggs (Southwark Primary Care Trust); Julia Collino (Lewisham Primary Care Trust); Anna Cutress, Sue Leach, Kathryn Temple (Newcastle University); Sam Barron, Ruth Colmer, Sarah Randles (North Tyneside Primary Care Trust).

References

Aldred

Green

Adams

(2004) A new social communication intervention for children with autism: Pilot randomised controlled treatment study suggesting effectiveness. Journal of Child Psychology and Psychiatry 45: 1420–1430. doi:10.1111/j.1469-7610.2004.00848.x.

Bieleninik

Posserud

M. B.

Geretsegger

Thompson

Elefant

Gold

(2017) Tracing the temporal stability of autism spectrum diagnosis and severity as measured by the Autism Diagnostic Observation Schedule: A systematic review and meta-analysis. PLoS One 12: e0183160. doi:10.1371/journal.pone.0183160.

Carr

Shih

Lawton

Lord

King

Kasari

(2015) The relationship between treatment attendance, adherence, and outcome in a caregiver-mediated intervention for low-resourced families of young children with autism spectrum disorder. Autism 20: 643–652. doi:10.1177/1362361315598634.

Cook

J. A.

Hislop

Adewuyi

T. E.

Harrild

Altman

D. G.

Ramsay

C. R.

Vale

L. D.

(2014) Assessing methods to specify the target difference for a randomised controlled trial: DELTA (Difference ELicitation in TriAls) review. Health Technology Assessment 18(28): v–vi, 1–175. doi: 10.3310/hta18280.

Dawson, G., Rogers, S., Munson, J., Smith, M., Winter, J., Greenson, J., Donaldson, A., Varley, J. (2010). Randomized, controlled trial of an intervention for toddlers with autism: The Early Start Denver Model. Pediatrics, 125, e17–e23.

Eldevik

Jahr

Eikeseth

Hastings

R. P.

Hughes

C. J.

(2010) Cognitive and adaptive behavior outcomes of behavioral intervention for young children with intellectual disability. Behavior Modification 34: 16–34. doi:10.1177/0145445509351961.

Fenson

Dale

Reznick

(1993) MacArthur communicative development inventories: User’s guide and technical manual, San Diego, CA: Singular Publishing Group.

Fuchs

Compton

(2012) Intervention effects for students with comorbid forms of learning disability: Understanding the needs of nonresponders. Journal of Learning Disability 46: 534–548. doi:10.1177/0022219412468889.

Green, J., Charman, T., McConachie, H., Aldred, C., Slonims, V., Howlin, P., … PACT Consortium. (2010). Parent-mediated communication-focused treatment in children with autism (PACT): A randomised controlled trial. Lancet, 375, 2152–2160. doi:10.1016/S0140-6736(10)60587-9.

10.

Howlin

Magiati

Charman

(2009) Systematic review of early intensive behavioral interventions (EIBI) for children with autism. American Journal on Intellectual and Developmental Disabilities 114: 23–41. doi:10.1352/2009.114:23-41.

11.

Hudry, K., Aldred, C., Wigham, S., Green, J., Leadbitter, K., Temple, K., … PACT Consortium. (2013). Predictors of parent-child interaction style in dyads with autism. Research in Developmental Disabilities, 34, 3400–3410. doi:10.1016/j.ridd.2013.07.015.

12.

Jacobson

Truax

(1991) Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology 59: 12–19. doi:10.1037/0022-006X.59.1.12.

13.

Jeste

Geschwind

(2014) Disentangling the heterogeneity of autism spectrum disorder through genetic findings. Nature Review of Neurology 10: 74–81. doi:10.1038/nrneurol.2013.278.

14.

Karst

Van Hecke

(2012) Parent and family impact of autism spectrum disorders: A review and proposed model for intervention evaluation. Clinical Child and Family Psychology Review 15: 247–277. doi:10.1007/s10567-012-0119-6.

15.

Kasari

Gulsrud

Paparella

Hellemann

Berry

(2015) Randomized comparative efficacy study of parent-mediated interventions for toddlers with autism. Journal of Consulting and Clinical Psychology 83: 554–563. doi:10.1037/a0039080.

16.

Kasari

Patterson

(2012) Interventions addressing social impairment in autism. Current Psychiatry Reports 14: 713–725. doi:10.1007/s11920-012-0317-4.

17.

Kraemer

Wilson

Fairburn

(2002) Mediators and moderators of treatment effects in randomized clinical trials. Archives of General Psychiatry 59: 877–883.

18.

Lord

Risi

Lambrecht

Cook

E. H.

Jr Leventhal

B. L.

DiLavore

P. C.

Rutter

(2000) The autism diagnostic observation schedule-generic: A standard measure of social and communication deficits associated with the spectrum of autism. Journal of Autism and Developmental Disorders 30: 205–223. doi:10.1023/A:1005592401947.

19.

Lord

Rutter

Le Couteur

(1994) Autism diagnostic interview-revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive development disorders. Journal of Autism and Developmental Disorders 24: 659–685.

20.

Magiati

Tay

Howlin

(2012) Early comprehensive interventions for children with autism spectrum disorders: A critical synthesis of recent review findings. Neuropsychiatry 2: 543–570. doi:10.2217/NPY.12.59.

21.

Magiati

Tay

Howlin

(2014) Cognitive, language, social and behavioural outcomes in adults with autism spectrum disorders: A systematic review of longitudinal follow-up studies in adulthood. Clinical Psychology Review 34: 73–86. doi:10.1016/j.cpr.2013.11.002.

22.

March

(2009) The future of psychotherapy for mentally ill children and adolescents. Journal of Child Psychology and Psychiatry 50: 170–179. doi:10.1111/j.1469-7610.2008.02034.x.

23.

McConachie

Parr

Glod

Hanratty

Livingstone

Oono

I. P.

Williams

(2015) Systematic review of tools to measure outcomes for young children with autism spectrum disorder. Health Technology Assessment 19: 1–506.

24.

Mullen

(1995) Mullen scales of early learning, Minneapolis, MN: Pearson.

25.

National Institute for Health and Care Excellence (2013) Autism: The management and support of children and young people on the autism spectrum (NICE guidelines), London, England: Author.

26.

Nunes

Ball

Booth

Brigham

Calsyn

D. A.

Carroll

Woody

(2010) Multisite effectiveness trials of treatments for substance abuse and co-occurring problems: Have we chosen the best designs? Journal of Substance Abuse Treatment 38: S97–S112. doi:10.1016/j.jsat.2010.01.012.

27.

Oono, I., Honey, E., & McConachie, H. (2013). Parent-mediated early intervention for young children with autism spectrum disorders (ASD). Cochrane Database of Systematic Reviews. doi:10.1002/14651858.CD009774.pub2.

28.

Parr

Gray

Wigham

McConachie

LeCouteur

(2015) Measuring the relationship between the parental Broader Autism Phenotype, parent-child interaction and children’s progress following parent-mediated intervention. Research in Autism Spectrum Disorder 20: 24–30. doi:10.1016/j.rasd.2015.07.006.

29.

Perryman

Carter

Messinger

Stone

W. L.

Ivanescu

A. E.

Yoder

P. J.

(2013) Brief report: Parental child-directed speech as a predictor of receptive language in children with autism symptomatology. Journal of Autism and Developmental Disorders 43: 1983–1987. doi:10.1007/s10803-012-1725-3.

30.

Pickles, A., Harris, V., Green, J., Aldred, C., McConachie, H., Slonims, V., … PACT Consortium. (2015). Treatment mechanism in the MRC preschool autism communication trial: Implications for study design and parent-focused therapy for children. Journal of Child Psychology and Psychiatry, 56, 162–170. doi:10.1111/jcpp.12291.

31.

Romanczyk

Callahan

Turner

Cavalari

(2014) Efficacy of behavioral interventions for young children with autism spectrum disorders: Public policy, the evidence base, and implementation parameters. Review Journal of Autism and Developmental Disorders 1: 276–326. doi:10.1007/s40489-014-0025-6.

32.

Sherer

Schreibman

(2005) Individual behavioral profiles and predictors of treatment effectiveness for children with autism. Journal of Consulting and Clinical Psychology 73: 525–538. doi:10.1037/0022-006X.73.3.525.

33.

Siller

Sigman

(2008) Modeling longitudinal change in the language abilities of children with autism: Parent behaviors and child characteristics as predictors of change. Developmental Psychology 44: 1691–1704. doi:10.1037/a0013771.

34.

Smith

Iadarola

(2015) Evidence base update for Autism Spectrum Disorder. Journal of Clinical Child and Adolescent Psychology 44: 897–922. doi:10.1080/15374416.2015.1077448.

35.

Solomon

Van Egeren

L. A.

Mahoney

Quon Huber

M. S.

Zimmerman

(2014) PLAY project home consultation intervention program for young children with autism spectrum disorders: A randomized controlled trial. Journal of Developmental and Behavioural Paediatrics 35: 475–485. doi:10.1097/DBP.0000000000000096.

36.

Spence, S J., Thurm, A. (2010). Testing autism interventions: Trials and tribulations. The Lancet, 375, 2124–2125.

37.

Starcevic

Brakoulias

(2008) Symptom subtypes of obsessive-compulsive disorder: Are they relevant for treatment? Australian and New Zealand Journal of Psychiatry 42: 651–661. doi:10.1080/00048670802203442.

38.

Szatmari, P., Georgiades, S., Duku, E., Bennett, T. A., Bryson, S., Fombonne, E., … Pathways in ASD Study Team. (2015). Developmental trajectories of symptom severity and adaptive functioning in an inception cohort of preschool children with autism spectrum disorder. JAMA Psychiatry, 72, 276–283. doi:10.1001/jamapsychiatry.2014.2463.

39.

Virués-Ortega

(2010) Applied behavior analytic intervention for autism in early childhood: Meta-analysis, meta-regression and dose-response meta-analysis of multiple outcomes. Clinical Psychology Review 30: 387–399. doi:10.1016/j.cpr.2010.01.008.

40.

Vivanti

Prior

Williams

Dissanayake

(2014) Predictors of outcomes in autism early intervention: Why don’t we know more? Frontiers in Pediatrics 2: 58. doi:10.3389/fped.2014.00058.

41.

Waddington

van der Meer

Sigafoos

(2016) Effectiveness of the Early Start Denver Model: A systematic review. Review Journal of Autism and Developmental Disorders 3: 93–106. doi:10.1007/s40489-015-0068-3.

42.

Winburn

Charlton

McConachie

McColl

Parr

O’Hare

Le Couteur

(2014) Parents’ and child health professionals’ attitudes towards dietary interventions for children with autism spectrum disorders. Journal of Autism and Developmental Disorders 44: 747–757. doi:10.1007/s10803-013-1922-8.

43.

Zimmerman

I. L.

Steiner

Pond

R. E.

(1997) Preschool language scales 3rd UK edition, Oxford, England: Psychological Corporation. (Adapted by J. Boucher & V. Lewis).

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.08 MB

	Main effects		Two-way interaction
	Outcome status	Treatment group	Two-way interaction
Chronological age	F(1,80) = .08, p = .783, d = .04	F(1,80) = .13, p = .723, d = .07	F(1,80) = .48, p = .492, η²= .01
ADI-R autism symptoms	F(1,80) = .17, p = .685, d = .04	F(1,80) = 1.54, p = .218, d = .27	F(1,80) = 1.05, p = .310, η²= .01
MSEL non-verbal ability	F(1,80) = 4.08, p = .047, d = .46	F(1,80) = .06, p = .806, d = .11	F(1,80) = .10, p = .757, η²= .00
MCDI receptive vocabulary	F(1,78) = .08, p = .774, d = .05	F(1,78) = .55, p = .462, d = .16	F(1,78) = .19, p = .667, η²= .00
DCMA Child Initiation	F(1,79) = .29, p = .591, d = .13	F(1,79) = .87, p = .353, d = .22	F(1,79) = 2.07, p = .154, η²= .03
DCMA Parent Synchrony	F(1,79) = .45, p = .506, d = .15	F(1,79) = .93, p = .337, d = .23	F(1,79) = 3.28, p = .074, η²= .04