Behavioural improvements of children with attention-deficit hyperactivity disorder (ADHD) and oppositional defiant disorder (ODD) following a restricted elimination diet (RED), may be due to concurrent changes in family environment.
Behavioural improvements of children with attention-deficit hyperactivity disorder (ADHD) and oppositional defiant disorder (ODD) following a restricted elimination diet (RED), may be due to concurrent changes in family environment.
Twenty-four children with ADHD, were randomized to either a 5-week RED intervention, or a control intervention consisting of healthy food advices in a pilot study.
No differences in family environment were found, neither at baseline nor when comparing the start and end measurements of both groups.
In this pilot study, the effects of an RED on ADHD and ODD are not mediated by improvement of family environment in families motivated to follow an RED. Replication of this preliminary study in larger groups of children is advised.
Attention-deficit hyperactivity disorder (ADHD), characterized by problems in attention, impulse control and activity regulation (American Psychiatric Association, 2000), is one of the most common psychiatric disorders, with a strong genetic disposition (Faraone & Mick, 2010). Both biological as well as psychosocial environmental factors are related to ADHD, including prenatal maternal smoking, prematurity, low birth weight, foetal distress, foster placing and disturbed parent–child relationships (Banerjee, Middleton & Faraone, 2007; Biederman & Faraone, 2005; Deault, 2010). An important, albeit controversial, environmental factor that may trigger ADHD is food (Arnold, 1999). Research investigating the effects of additives like colourings and preservatives on ADHD, has shown that although additives may have some effects on the behaviour of all children [effect size (ES) 0.3], additives do not cause ADHD (McCann et al., 2007). Conversely, recent research investigating the effects of a restricted elimination diet (RED), that is, eliminating many kinds of foods from the child's diet, has shown statistically significant and clinically relevant results, with ES on the ADHD DSM-IV Rating Scale varying from 1.7 according to the open teacher measurements to 2.0 according to the blinded paediatrician measurements (Pelsser et al., 2011). These results confirm the outcomes of seven previous randomized controlled trials, investigating the effects of an RED on ADHD, with an overall ES of 1.6 (Pelsser et al., 2011).
One could argue that the children's behavioural changes might be due to concomitant improvement of parental behavioural strategies, caused by the strict parental supervision necessary to comply with the RED. Research has shown that ADHD is associated with disruptive parent–child relationships and poor parenting structure (Foley, 2011; Harpin, 2005; Pressman et al., 2006; Verhoeven, Junger, Aken van, Devokić & Aken van, 2010), even more when children are suffering from comorbid oppositional defiant disorder (ODD; Deault, 2010). Conversely, consistent parenting and positive parent–child interactions are associated with improvements of child behaviour (Kaminski, Valle, Filene & Boyle, 2008). This suggests the possibility that behavioural improvements assessed in the RED trials might be mediated by the strict parenting structure necessary to follow the RED, rather than be a direct result of the diet.
The present pilot study uses a subsample of the Impact of Nutrition on Children with ADHD (INCA) study and investigated whether the effects of an RED on ADHD and ODD symptoms as previously reported (Pelsser et al., 2011), can be explained by changes in family structure during the intervention. A group of no-ADHD control children was included as a comparison group, to investigate changes in family structure over time. Our study is the first study investigating the effects of following an RED on parenting abilities.
Children with ADHD were recruited as part of the INCA study, investigating the effects of an RED on ADHD and ODD in children, the results of which have been reported elsewhere (Pelsser et al., 2011). All parents of children entering the INCA study after the summer holiday 2009, that is, the last 24 children (ADHD RED group n = 11, ADHD control group n = 13), took part in this pilot study. INCA inclusion criteria were (1) the children were diagnosed ADHD any subtype (American Psychiatric Association, 2000), (2) children were of age 4–8, (3) their parents had sufficient command of the Dutch language and (4) parents were motivated to follow an RED during a 5-week period. Exclusion criteria were children taking medication for ADHD or receiving behavioural therapy, children already following a diet, or family circumstances that were likely to impede completion of the study. The INCA study was approved by the Medical Ethics Committee of Wageningen University and by the executive board and ethics committee of Catharina-Hospital Eindhoven.
A control group (n = 23) consisting of children without ADHD (no-ADHD control group), aged 4–8 years, was recruited through the teachers of the participating INCA children. All teachers were contacted by phone and were asked to distribute an information leaflet, concerning a request to participate in this study, to parents of children without any behavioural problems at school. Interested parents filled in the ADHD DSM-IV Rating Scale (ARS) and subsequently were contacted by phone. Parents of all children participating in this trial gave written informed consent before the start of the study.
Three questionnaires were used to assess outcome: (1) the Dutch version of the Family Environment Scale (FES), [Jansma and Coole (1996)], to assess family relationships and parenting structure; (2) the 18-item ADHD DSM-IV Rating Scale (ARS; Dupaul, 1991) to assess ADHD; and (3) a semi-structured, DSM-IV-based, psychiatric interview ([SPI] Pelsser et al., 2009), to assess ODD. The FES consists of 77 yes/no questions related to seven subscales: (1) cohesion (family commitment and support), (2) expressiveness (expression of feelings), (3) conflict (expression of anger and aggression), (4) organization (structure and planning of family life), (5) control (rules used in family life), (6) family values (opinion about norms and behaviour) and 7) social orientation (involvement in the social environment). Each subscale consists of 11 questions and scores range from 0 to 11, higher scores indicating a more positive environment, with the exception of the conflict scale. In this study two index scores were used: the family relationships index (FRI), based on three subscales (i.e. cohesion, expressiveness and conflict) and the family structure index (FSI), based on two subscales (i.e. organization and control). Higher scores indicate better family relationship and parental structure (Gazendam-Donofrio et al., 2007). Both FES subscales and FES indices have shown good reliability and adequate validity (Jansma & Coole, 1996). The two subscales which are not linked to the two indices, that is, family values and social orientation, are not included in this article.
The ARS, based on the DSM-IV criteria for ADHD, was used to assess ADHD, and consists of 18 criteria, nine inattention and nine hyperactivity/impulsivity criteria, using a 4-point scale (0 = never [less than once a week], 1 = sometimes [several times a week], 2 = often [once a day], and 3 = very often [several times a day]), with a maximum of 54 points.
Comorbid ODD was assessed by the SPI, based on the eight DSM-IV-ODD-criteria, with a maximum of eight points. A detailed description of the behavioural questionnaires has been published elsewhere (Pelsser et al., 2011).
The study design is shown in Figure 1. The assessment points of all questionnaires, in both ADHD groups and in the no-ADHD control group, were at baseline and at the end of the trial, and data were collected in all children participating in the study (n = 47). After the baseline assessments the ADHD group (n = 24) was randomized to the ADHD RED group (n = 11) or the ADHD control group (n = 13). The ADHD control group received healthy food advices according to the guidelines of the Dutch Nutrition Centre, the ADHD RED group followed a 5-week individually composed RED. The RED, of which the details are described elsewhere (Pelsser, Frankena, Buitelaar & Rommelse, 2010), was based on the few foods diet, consisting of rice, meat, vegetables, pears and water. This diet was initially supplemented with specific foods such as potatoes, fruits and wheat, in order to create a less restrictive elimination diet, thus resulting in a diet easier for children and their parents to adhere to. If the parents reported no behavioural changes by the end of the second diet week, the diet was gradually restricted to the few foods diet only. After the baseline assessments, the control group without ADHD (n = 23) did not receive any intervention.
The FES was filled in by the parents in both ADHD groups and the no-ADHD control group. The ADHD and ODD behavioural ratings in both ADHD groups were executed by a paediatrician blinded for treatment assignment (Pelsser et al., 2011), whereas the behavioural ratings in the no-ADHD control group were unmasked parent assessments.
Statistical analyses were done with STATA version 10 and SPSS version 15. Statistical significance was based on α = 0.05, two sided, and clinical relevance was expressed by means of ES, with ES ≥ 0.5 indicating a clinically significant effect (Nakagawa & Cuthill, 2007).
Family Environment Scale-outcomes at baseline before randomization were compared between ADHD group, no-ADHD control group and norm score provided by the FES manual, based on 941 mother reports (Jansma & Coole, 1996). p-values for differences between groups were obtained using general linear model (GLM) and additionally Cohen's d was calculated as effect size estimate. p-values for comparisons with the norm were obtained by the Welch-Satterthwaite equation.
To assess any mediating effects of FES on the behaviour of the children, the FES outcomes at the end of the study were analysed by GLM, using group (ADHD RED, ADHD control, no-ADHD control) as independent variable and the scores at start of the trial as covariate to adjust for differences that were potentially present at start already.
Attention-deficit hyperactivity disorder and ODD analyses were by intention-to-treat, last observation carried forward, and based on the blinded measurements in both ADHD groups. A detailed description has been published elsewhere (Pelsser et al., 2011). Repeated measurement models were used to separately analyse the moderating role of the two FES indices (Relationship and Structure) on the effect of the RED on ADHD and ODD scores. Independent variables were (1) group (ADHD RED or ADHD control), (2) measurement point (start or end), and FES indices (FRI of FSI). Child was added as repeated effect to adjust for potential intra-child correlation (generalized estimated equations, Gaussian distribution, exchangeable covariance structure). Dependent variables were ADHD and ODD scores. All two-way interactions (FRI and FSI × group, FRI and FSI × measurement point, group × measurement point), and three-way interactions (FRI and FSI × group × measurement point) were evaluated.
In both the ADHD control group and the no-ADHD control group two families left the study prematurely and one family did not complete the second FES-questionnaire. None of the families in the ADHD RED group left the trial. Most FES questionnaires were completed by the mother (43/47), and most children were boys (38/47). The FES was completed twice for each child, at start and at and of the trial, with a total number of 94 observations. No significant differences regarding age, family size, number of siblings and single parent families were found between the ADHD groups and No-ADHD control group (see Table 1). In the ADHD group 11/24 children (46%) were also diagnosed with comorbid ODD (6/11 children in de ADHD RED group, 5/13 in the ADHD control group). In the no-ADHD control group one child met the criteria for ODD, none of them meeting the criteria for ADHD.
|ADHD RED n = 11||ADHD control n = 13||No-ADHD control n = 23||p-value Fisher exact|
|Boys||10 (91%)||10 (77%)||18 (78%)||0.708|
|Age in years (mean (SD))||7.7 (0.9)||7.2 (1.2)||6.8 (1.4)||0.115a,b|
|Step/single parent family||0.1 (0%)||1 (8%)||2 (9%)||1.000|
|FES completed by mother||11 (100%)||11 (85%)||21 (91%)||0.668|
|Family size, number of siblings||0.073b|
|Only child||0 (0%)||3 (23%)||0 (0%)||0.028c|
|1 sibling||5 (45%)||8 (63%)||16 (70%)||0.431c|
|2 siblings||4 (36%)||2 (15%)||6 (26%)||0.562c|
|3 siblings||2 (18%)||0 (0%)||1 (4%)||0.203c|
|Meeting criteria ADHD/ODD|
|ADHD||11d (100%)||13d (100%)||0d (0%)|
|ODD||6 (55%)||5 (38%)||1 (4%)||0.001|
At baseline, the FRI and the FSI scores of the ADHD group preceding randomization (n = 24) were similar to the scores of the no-ADHD control group (n = 23; see Table 2). Also no differences were found between the ADHD RED group (n = 11) and the ADHD control group (n = 13). Both ADHD groups showed significantly more conflicts on the conflict subscale than the no-ADHD control group, whereas on the other subscales no significant differences were found. When compared to the norm scores provided by the FES-manual, significantly higher scores of FRI and FSI were found in both ADHD groups and in the no-ADHD control group at baseline.
|1 n = 24||2 n = 23||3 n = 941||Difference 1–2a||Difference 1–3b||Difference 2–3b|
|ADHD group||No-ADHD Control group Mean (SD)||Norm||Mean difference||p-value||Cohen's d||Mean difference||p-value||Cohen's d||Mean difference||p-value||Cohen's d|
|Cohesion||8.9 (1.7)||9.1 (0.9)||8.5 (2.0)||−0.2 (−1.0–0.4)||0.59||−0.2||0.4||0.134||0.2||0.6||0.003||0.4|
|Expressiveness||10.1 (1.3)||9.7 (1.2)||8.3 (2.2)||0.3 (−0.4–1.1)||0.352||0.3||1.8||<0.0001||1.0||1.4||<0.0001||0.8|
|Conflict||4.7 (2.2)||3.0 (2.1)||5.0 (2.5)||1.7 (0.5–2.9)||0.007||0.8||−0.3||0.256||−0.1||−2.0||<0.0001||−0.9|
|Organization||8.3 (1.5)||8.7 (1.0)||8.0 (2.3)||−0.4 (−1.2–0.3)||0.230||−0.3||0.3||0.175||0.2||0.7||0.002||0.4|
|Control||9.5 (1.3)||9.5 (1.2)||8.5 (2.1)||0.0 (−0.7−0.7)||0.951||0.0||1.0||0.001||0.6||1.0||<0.0001||0.6|
|FRI||25.3 (3.8)||26.8 (2.9)||22.9 (3.4)||−1.5 (−3.5–0.4)||0.123||−0.4||2.4||0.003||0.7||3.9||<0.0001||1.2|
|FSI||17.8 (2.1)||18.3 (1.7)||16.5 (3.8)||−0.5 (−1.6–0.7)||0.429||−0.3||1.3||0.003||0.4||1.8||<0.0001||0.6|
The FRI and FSI baseline and endpoint scores of the ADHD RED group and ADHD control group are presented in Table 3. No intervention effect was found on the FRI and FSI, neither in the ADHD RED group nor in the ADHD control group. Furthermore, no intervention effect was found on the two FES subscales not linked to the two indices, that is, family values and social orientation.
|ADHD RED N = 11||ADHD Control N = 12||End rating ADHD Control versus ADHD RED group, adjusted for score at start|
|Mean (SD)||Mean difference (95% CI) start-end||p-value||Cohen's d||Mean (SD)||Mean difference (95% CI)||p-value||Cohen's d||Mean difference (95% CI)||p-value|
|ADHDa||44.6 (4.1)||18.1 (12.9)||26.5 (18.2–34.9)||<0.0001||2.8||46.0 (4.2)||43.7 (7.1)||2.3 (−0.–5.6)||0.156||0.4||25.1 (16.4–33.8)||<.0001|
|ODDa||3.3 (2.0)||2.0 (2.0)||1.3 (−0.2–2.8)||0.101||0.7||3.2 (2.3)||2.9 (2.6)||0.2 (−0.3–0.8)||0.411||0.1||1.0 (−0.5–2.5)||.201|
|Cohesion||8.7 (1.4)||8.9 (1.1)||−0.2 (−0.9–0.5)||0.606||−0.2||8.9 (2.0)||9.0 (1.6)||−0.2 (–0.7.–0.4)||0.575||−0.1||−0.1 (−0.8–0.7)||.883|
|Expressiveness||10.1 (1.2)||10.0 (1.6)||0.1 (−0.2–0.4)||0.576||0.1||10.2 (1.5)||10.0 (1.1)||0.2 (−0.4–0.7)||0.538||0.2||0.1 (−0.6–0.7)||.840|
|Conflict||5.6 (2.2)||4.7 (2.2)||0.8 (0.0–1.6)||0.041||0.4||4.3 (1.6)||3.8 (1.5)||0.5 (−0.1–1.1)||0.083||0.3||−0.1 (−1.0–0.9)||.904|
|Organization||8.4 (1.4)||8.6 (0.7)||−0.2 (−0.8–0.5)||0.606||−0.2||8.1 (1.6)||8.0 (1.6)||0.1 (−0.6–0.7)||0.804||0.1||0.4 (−0.4–1.2)||.332|
|Control||9.4 (1.0)||9.8 (1.0)||−0.5 (−0.8–0.1)||0.004||−0.4||9.5 (1.4)||9.6 (1.5)||−0.1 (−0.6–0.5)||0.772||−0.1||0.3 (−0.3–1.0)||.295|
|FRI||24.3 (3.2)||25.2 (3.2)||−0.9 (−2.1–0.3)||0.136||−0.3||25.8 (4.0)||26.3 (2.8)||−0.5 (−1.8–0.8)||0.455||−0.1||−0.1 (−1.6–1.4)||.882|
|FSI||17.7 (1.7)||18.4 (1.0)||−0.6 (−1.4–0.2)||0.121||−0.5||17.6 (2.7)||17.6 (2.8)||0.0 (−1.1–1.1)||1.000||0.0||0.7 (−0.6–1.9)||.282|
The effects of the diet intervention on ADHD and ODD are shown in Table 3. The analysis of the ADHD score using a repeated measurement design showed that FRI was significantly associated with the ADHD score, independent of ADHD group and measurement moment (estimate −0.8, 95% CI = −1.5–−0.1, p = .024, see Figure 2). Consequently, there seems to be an inverse relationship between the number of ADHD sympoms and the family relationship: the higher the ADHD symptom score, the less the FES Relationships Index score. The two-way and three-way interactions with group and measurement point were nonsignificant. No effect of FSI on ADHD score was found (estimate 0.8, 95% CI = −0.2−1.8, p = .134), and the two-way and three-way interactions were non-significant, suggesting that changes in both FRI and FSI did neither mediate nor moderate the results of the RED on ADHD symptoms.
A similar analysis for ODD score equally showed a significant association for FRI, independent of ADHD group and measurement moment (estimate −0.2, 95% CI = −0.4–−0.03, p = .022), indicating that a higher FRI was related to less ODD symptoms (see Figure 2). No effect of FSI on ODD score was found (estimate 0.2, 95% CI = −0.1–0.5, p = .238). Again two-way and three-way interactions were non-significant, suggesting that changes in both FRI and FSI did neither mediate nor moderate the results of the RED on ODD symptoms.
The main aim of this pilot study was to investigate whether the beneficial behavioural effects in children with ADHD following an RED (Pelsser et al., 2011), may be explained by an improvement of family environment, that is, family relationships and family structure. Our findings indicate that (1) family environment in families of children with ADHD, motivated to follow a 5-week RED, is similar to the family environment of a no-ADHD control group and better than the FES-manual norm; (2) family relationships and structure are not affected by following a 5-week RED; (3) the effects of an RED on ADHD and ODD symptoms are not mediated by changes in family relationships and structure (in fact, there were no changes in FRI and FSI); (4) family relationships but not family structure are inversely associated with ADHD and ODD.
At baseline we did not find any significant differences in family environment between the ADHD group (n = 24) and the no-ADHD control group (n = 23), nor between the ADHD RED group (n = 11) and the ADHD control group (n = 13). All groups, however, showed a better family environment compared to the norms. Consequently, relationships and structure of ADHD families taking part in this study were (more than) adequate. These findings are consistent with the findings of Pimentel, Vieira-Santos, Santos and Vale (2011), who found that parental practices of mothers of children with ADHD were similar to those of a validation sample. Conversely, our results are contrary to the findings of Kepley and Ostrander (2007) who found that families with ADHD were less cohesive and expressive than families without ADHD. This discrepancy may be caused by the fact that, in advance, parents were adequately informed about the stringency of the diet. It is further conceivable that parents who dreaded this challenge would decide not to participate, and that only parents who were confident of their parenting capacities, would decide to participate. Consequently, our sample may have consisted of families with an above-average family environment only.
We found no differences in family relationships and structure when comparing the start and end measurements of both ADHD groups. Consequently, in our sample the RED did not affect family environment; the restricted diet instigated neither any favourable nor any deleterious effects on the parent–child relationship. Unfortunately, considering that all families in the ADHD RED group completed the study, we could not investigate the parenting capacities of parents not complying with the diet.
The results of the RED on the behavioural outcomes of the ADHD group, that is, statistically significant and clinically relevant improvements of ADHD and ODD in 64% of the children, have been discussed elsewhere (Pelsser et al., 2011). Considering the analyses of two-way and three-way interactions of the family relationships and structure indices and behaviour, with group and measurement point as independent variables, we did not find evidence that family relationships and structure mediate the effect of the RED on ADHD and ODD symptoms. However, we found an inverse association between family relationship and both ADHD and ODD symptoms, that is, a higher FRI score co-occurred with less ADHD and ODD symptoms. These findings are consistent with the findings of Deault (2010), showing that ADHD is associated with conflicted parent–child relationships. We found no significant association between family structure and both ADHD and ODD symptoms.
Limitations of this study should be noted. First, the ADHD group only consisted of families with a good family environment. It is conceivable that families with less adequate family environments were deterred from following an RED, which may have led to a sample bias. We would like to argue that all clinical trials asking specific commitment and efforts may inevitably lead to a highly motivated sample of subjects entering the trial, thus resulting in a study population not reflecting the real-life heterogeneity of subjects. Consequently, our study results are applicable only to those families with both a good family environment and sufficient motivation to undertake a 5-week diet. Concurrently, our study may indicate that in families meeting the conditions of both a good family environment and a positive attitude towards a 5-week diet an RED is a feasible intervention in children with ADHD. Second, it is conceivable that parenting problems might impede the compliance to and the completion of the RED, but this aspect could not be investigated because all families assigned to the ADHD RED group completed the diet. Third, although promising, the sample size in this study is small, and replication in a larger group of children is advised. Still, we would like to argue that our study results confirm the findings of two double-blind placebo-controlled diet studies using a restricted verum diet (diet group) and a restricted sham diet (control group) [Kaplan, McNicol, Conte & Moghadam,1989; Schmidt et al., 1997]. Both studies resulted in behavioural improvements in the diet group only (abbreviated Conners' Scale effect sizes 0.55 and 0.59 respectively), thus indirectly showing that the children's behavioural changes were not due to changes in parenting capacities.
Concluding, the results of our pilot study suggest that, in families with good family environment and motivated to follow an RED, the behavioural improvements of children with ADHD and ODD following an RED are not mediated by improvements of family relationships and structure. Replication of this study, in larger groups of children, is advised.
This study was supported by grants from the Foundation of Child and Behaviour.
LMP is franchiser of the ADHD Research Centre. In the past 3 years, JKB has been a consultant, member of advisory board, speaker, or a combination thereof for Janssen Cilag, Eli Lilly, Bristol-Myers Squibb, Schering Plough, UCB, Shire, Medice and Servier.
All other authors declare that they have no conflicts of interest.
We thank all parents, children and teachers who participated in this study.