Motivational Interviewing to Treat Overweight Children: 24-Month Follow-Up of a Randomized Controlled Trial
BACKGROUND: Pediatrician-led motivational interviewing can be an effective way of controlling BMI in overweight children in the short term. Its long-term efficacy is unknown. The primary aim was to determine whether the short-term (12-month) impact of family pediatrician-led motivational interviews on the BMI of overweight children could be sustained in the long term (24 months), in the absence of any other intervention.
METHODS: Children were recruited in 2011 by family pediatricians working in the province of Reggio Emilia, Italy, and randomly allocated to receive either 5 interviews delivered over a 12-month period or usual care. Eligible participants were all 4- to 7-year-old overweight children resident in the province of Reggio Emilia who had been receiving care from the pediatrician for ≥12 months. The primary outcome of this study was individual variation in BMI between the baseline visit and the 24-month follow-up, assessed by pediatricians not blinded to treatment group allocation.
RESULTS: Of 419 eligible families, 372 (89%) participated; 187 children were randomized to receive intervention and 185 to usual care. Ninety-five percent of the children attended the 12-month follow-up, and 91% attended the 24-month follow-up. After the 12-month intervention period, BMI in the intervention group increased less than in the control group (0.46 and 0.78, respectively; difference −0.32; P = .005). At the 24-month follow-up, the difference had disappeared (1.52 and 1.56, respectively; difference −0.04; P = .986).
CONCLUSIONS: The intervention lost its effectiveness within 1 year of cessation. Sustainable boosters are required for weight control and obesity prevention.
- CI —
- confidence interval
- MI —
- motivational interviewing
- PA —
- physical activity
What’s Known on This Subject:
Childhood obesity can seriously affect health outcomes. Motivational interviewing in primary care has been shown to be effective in BMI control, but previous studies measured its efficacy only just after the intervention ended. There are no available long-term follow-up data.
What This Study Adds:
Despite very encouraging initial results, 12 months after intervention ended, children who received motivational interviewing lost all the advantage in terms of BMI, compared with the control group.
The number of overweight or obese children (0–5 years old) increased from 31 million globally in 1990 to 44 million in 2012.1 Childhood obesity is associated with a higher chance of obesity, premature death, and disability in adulthood. More than 60% of children who are overweight before puberty will be overweight in early adulthood, reducing the average age at which noncommunicable diseases become apparent and greatly increasing the burden on health services, which have to provide treatment during much of their adult lives.2 At present, there is lack of consensus worldwide on which approaches and which combinations of interventions are likely to be most effective at preventing childhood obesity in different contexts and societies.3 Motivational interviewing (MIs) has been applied to pediatric populations over the last 2 decades4,5 and is recommended6 and widely used7 for weight control and obesity prevention. Nevertheless, published studies8–13 evaluating the influence of MI or behavioral counseling on the BMI of overweight children showed inconsistent results. Although individual studies reported mixed success, a recent meta-analysis demonstrated that behavioral strategies to improve diet and physical activity (PA) were efficacious in reducing children’s BMI.14 In 2011, we conducted a randomized trial in the province of Reggio Emilia, Italy, to evaluate the efficacy of pediatrician-led MI in controlling BMI in overweight children aged 4 to 7 years.15 Our study concluded that MI was effective at controlling the BMI of overweight children (BMI percentile ≥85th and <95th) in the short term. Despite the encouraging short-term effect, evidence of long-term efficacy is required.16,17
In this article, we report the results of the follow-up conducted 12 months after the end of intervention.
Study Design and Setting
The study is an individually randomized controlled trial and was conducted in the province of Reggio Emilia, Italy, from 2011 to 2013. Reggio Emilia had a resident population of 530 543 on January 1, 2011; 15.2% were children aged 0 to 14 years18 under the care of 82 public health service family pediatricians. In 2010, the estimated prevalence of overweight children was 22%.19 The methods used to conduct this trial are published elsewhere15 and are described briefly below.
Children were recruited through their family pediatrician. A maximum of 12 overweight children (based on a previous survey) per participating pediatrician were randomly selected for eligibility assessment. During a baseline visit for eligibility assessment, the family pediatrician proposed the study and gave consent forms to parents of eligible children. Eligible participants were all overweight children (BMI percentile ≥85th and <95th)20 aged between 4 and 7 years, resident in the province of Reggio Emilia, and under the care of said pediatrician for ≥12 months. Exclusion criteria were metabolic pathologic conditions and all pathologic conditions related to obesity and being overweight. Moreover, those families who did not consider their children being overweight to be an issue and were not interested in the negative consequences or advice on how to lose weight (families in the “precontemplation stage”) were also excluded. Recruitment took place from June to August 2011.
The epidemiology unit used the RALLOC package of Stata software to randomly allocate eligible children whose parents signed the informed consent form to an intervention or control group.21 The allocation rule depended on the number of eligible children. Pediatricians were not blinded to the group allocation.
The intervention consisted of 5 MIs delivered at 1, 4, 7, and 12 months after the baseline visit. Before enrollment began, all pediatricians attended a 4-hour training course on how to accurately measure weight and height and how to calculate the BMI percentile, as well as a 20-hour training course on MI conducted by specialist psychologists from “Luoghi di Prevenzione,” the regional reference center for training in health promotion.22 As previously reported, 94% of participants allocated to the intervention group completed all 5 MIs.15 The intervention was based on the transtheoretical model of addiction and behavioral change.23 The child and parents always had to leave the meeting having agreed on 2 clearly defined and achievable objectives (1 concerning food and 1 concerning PA improvements). During each subsequent interview, the extent to which the objectives set at the previous meeting had been achieved was assessed. The objectives were then reinforced or redefined and recorded accordingly. Participants who were randomly assigned to the control group attended the baseline and 12-month visits. They received a booklet with the main information on obesity prevention (eg, opportunistic healthy diet recommendations if the pediatrician was seeing the child for other reasons). All children were invited to attend a follow-up visit 24 months after the baseline, eg, 12 months after the intervention had finished.
Data Collection and Outcomes
Baseline and follow-up data were collected by means of a corporate Intranet Web form customized for the trial, compiled by the pediatrician. Dietary and PA habits were assessed using the questionnaire from the Italian survey on childhood obesity “OKKIO alla Salute,”24 conducted periodically by Italy’s National Institute of Health and part of the World Health Organization European region Childhood Obesity Surveillance Initiative25; some items were supplemented with regional food names. Sociodemographic data on the parents and child was collected at the baseline visit. Weight, height, BMI, BMI percentile, and dietary and PA habits were assessed at baseline and at 12- and 24-month follow-up visits. The primary outcome of the study was the individual BMI score variation (Δ0–24BMI) between the baseline and 24-month follow-up visits (long-term effect).26 The secondary outcomes were the percentage of positive changes in parent-reported dietary behavior and PA between the baseline and 24-month follow-up visits.
The study was sized to detect a between-group difference in ΔBMI of ≥0.5 with a 5% significance level and a power of 90%, assuming an SD of ΔBMI of 1. The smallest possible sample size was 85 children per group. Considering a 30% dropout rate, recruitment of at least 110 children per arm was planned.
Data analysis was performed using Stata statistical software, version 13.0. Primary objective analysis was conducted on the basis of an intention-to-treat approach, replacing missing Δ0–24BMI values with the mean BMI variation observed in the control group. All inferential analyses were performed using weights to balance allocation within strata.15 Mean Δ0–24BMI, Δ0–24BMI z-score, and relative 95% confidence interval (CI) were presented by study group. Δ0–24BMI, Δ0–12BMI (short-term effect), and Δ12–24BMI (rebound effect) in the intervention and control groups were compared using nonparametric Wilcoxon–Mann–Whitney test because the outcome was not normally distributed. Multilevel linear models were developed to measure the influence of pediatricians on long-term effectiveness. Two multilevel models were specified: random intercept and random intercept and slopes. The intrapediatrician correlation coefficient was reported with the likelihood ratio test to compare models. Post hoc subgroup analyses were performed by gender, age at baseline (<6 or ≥6 years), and mother’s level of education (<13, 13, or >13 years of schooling). Changes in PA and dietary habits in intervention and control groups were compared using Wilcoxon rank sum test.
We performed a formal mediation analysis to distinguish the direct effect of the intervention on BMI and its mediated effect through diet and physical activity. The analysis comprised 3 steps: (1) checking the association between the intervention and the putative mediator; (2) checking the association between the putative mediator and BMI; and (3) calculating the mediation percentage, eg, the proportion of the intervention effect on BMI mediated by the mediator. Twelve-month variations were considered, eg, 0–12 and 12–24, and the intervention was categorized in 3 ways: untreated, treatment period, and posttreatment. This analysis included all the dietary and PA variables associated with the BMI z-score (P < .05). Separate age-adjusted models were developed for each putative mediator.
Four hundred nineteen parents were asked to consent to their child’s participation in the study, and 372 signed the informed consent form; 187 were randomized to intervention and 185 to usual care. The intervention and usual care groups had similar baseline characteristics (gender, breastfeeding, overweight before the age of 5 years, gestational age at birth, small for gestational age, parent level of education, parent weight, baseline age, and baseline BMI).15
Participation in 12-Month Follow-Up
Of the 187 participants in the intervention group, 167 (89.3%) attended the 24-month follow-up visit, as did 170 (91.9%) of the 185 participants in the usual care group. One treated child was excluded from the analysis because, during a follow-up assessment, a pediatrician noted that the weight and height measurements had been recorded incorrectly at the baseline visit and, on the basis of the up-to-date information, the child was not eligible to take part27 (Fig 1). There were no differences in baseline characteristics between participants who provided data at the 12-month follow-up and those who did not. At 24 months, a greater proportion of children with Italian-born parents attended the follow-up visit compared with children with ≥1 immigrant parent (94% vs 62%; P < .001).
Effect on BMI
At the 24-month follow-up, there were no significant differences between groups in terms of the Δ0–24BMI score: −0.04 (95% CI −0.36 to 0.28). During the postintervention period (between the 12- and 24-month follow-ups), mean BMI increased more among children in the intervention group (Δ12–24BMI 1.06% [95% CI 0.90 to 1.22]) than in the usual care group (Δ12–24BMI 0.78% [95% CI 0.59 to 0.97]) (rebound effect). (Table 1) The proportion of pediatrician-level variance on the overall variation in Δ0–24BMI (intraclass correlation coefficient) was 7.2%. Mean Δ0–24BMI in the control group differed significantly across pediatricians (likelihood ratio test, random-intercept versus linear-regression model: χ2 = 7.72, P = .0211). The lack of long-term effectiveness of MI did not vary significantly across pediatricians (likelihood ratio test, random-intercept and slope versus random-intercept model: χ2 = 0.30, P = .8602). The results were similar if only the rebound effect was taken into consideration.
The long-term effect was similar by gender (interaction test P = .332), although breaking down the results into short-term and rebound effects showed that the short-term effect was stronger in girls (Δ0–12BMI −0.54 and 0.0 for females and males, respectively, interaction test P = .053), as was the rebound effect (Δ12–24BMI 0.36 and 0.16 for females and males, interaction test P = .421) (Table 2; Fig 2). There was no interaction between age and intervention (interaction test P = .333) (Table 2). MI had a positive long-term effect on Δ0–24BMI in children whose mother had a high (Δ0–24BMI −0.73% [95%CI −1.65 to 0.18]) or medium (Δ0–24BMI −0.31% [95% CI −0.74 to 0.13]) level of education, whereas it had a negative long-term effect in children whose mother had a low level of education (Δ0–24BMI 0.66% [95% CI 0.08 to 1.23) (interaction test P = .008). The same results were observed in the short term (Table 2; Fig 3).
Effect of Diet and PA on BMI Changes
Several improvements in behaviors were evident in the MI group in the short term,15 but almost no improvements were observed during the follow-up period. As a result, the long-term effect showed only modest improvements in nonorganized PA (P = .072), fruit intake (P = .070), and the consumption of sweet snacks/candies (P = .066), desserts (P = .047), and sweetened drinks (P = .004) (Table 3). The only 2 exceptions were having breakfast and eating fruit, for which improvements were observed during both periods (Table 3), but they were not associated with BMI. Only parent-reported nonorganized PA and vegetable and dessert consumption were associated with BMI variations (Table 4). The mediation analysis showed that the increase in nonorganized PA and the decrease in dessert consumption accounted for 7.7% (95% CI 4.5 to 22.7) and 8.6% (95% CI 5.0 to 27.7), respectively, of the total effect of the intervention on the BMI z-score.
In 2013, we published the positive results of MI intervention conducted by pediatricians in an Italian province.15 We concluded that intervention was effective in the short term, although the results were less convincing for boys and for children whose mothers had low levels of education. Our positive conclusions have now been dampened by the follow-up results for the children enrolled in the trial, as reported in this article: 12 months after the end of the intervention, the advantages observed in the intervention group had almost disappeared and, in fact, the rebound effect in the intervention group produced a greater increase in BMI than in the control group.
Gender and mother’s level of education were found to play an important role in determining the outcome. Whereas benefits disappeared after the 12-month follow-up visit for children whose mothers had spent >13 years at school, the effects of intervention seem counterproductive in the long term for children whose mothers had received <13 years of education. The child’s gender also modified the effect of our proposed MI, but unlike the low level of maternal education, the lack of an effect in boys was not followed by a rebound effect, meaning that the intervention does not help boys, but also that it does not harm them.
Our results are consistent with previous results in adults,28–30 where the success of initial motivational interviewing in BMI control and changing diet and levels of PA is often followed by a lack of long-term efficacy. A relapse once MI has been discontinued is considered an intrinsic characteristic of these interventions.31,32
Nevertheless, 2 more recent studies reported the effect of MI over the course of 24 months.12,13 In Resnicow et al, the study population and the intervention were quite similar to those of our own study, although only 4 interviews were distributed throughout the entire 24-month period and 1 of the intervention groups included a dietitian interview.12 The study observed a larger δ in BMI variation between the intervention group and the usual care group than we observed during our 12-month intervention period. This larger effect could be because the study analyzed only children who were not lost during follow-up, the longer period of intervention, the involvement of paid volunteer pediatricians, or the inclusion of dietitians in the intervention. On the other hand, the authors only report data at the end of intervention, which is not comparable with our 24-month follow-up data and, to date, it is not known whether a rebound would also occur after the 24-month period covered by their study.
Taylor et al reported positive long-term results of family-based intervention addressing overweight and obese children aged 4 to 8 years.13 This intervention is tailored to the needs of each family and consists of frequent, low-dose contact over a 2-year period (total contact time was 6 to 7 hours per family). The Δ0–24BMI z-score difference between the intervention and control group was −0.12 (95% CI −0.20 to −0.04), which is consistent with our findings after 12 months of intervention. The long-term efficacy after the intervention has not been investigated.
It is encouraging that 3 trials12,13,15 conducted in very different contexts and health systems showed improvements after MI interventions delivered in a primary care setting, at least for as long as the intervention was still in progress, in line with results of a recent systematic review.14
The family, particularly the mother, also proved to be an important effect modifier for the long-term maintenance of the effects in other studies.33,34 In our setting, some pediatricians reported that the motivational diagnosis was more difficult and less accurate when the mother had a lower level of education. The potentially harmful effect of MI when administered in precontemplative subjects has been postulated in theory and also observed in experimental settings.35 The observed results stress the need for an accurate motivational diagnosis in all families.
In our population, the changes in BMI in the intervention and control group were consistent with the changes in diet and physical activity at the end of the intervention.15 During the follow-up period, the changes in parent-reported diet and PA no longer showed any benefit in the intervention group compared with the control group. We conducted an analysis trying to understand which dietary and PA changes played the biggest part in the mechanism leading to BMI reduction after intervention and BMI rebound after intervention cessation. Only 2 behavioral factors were found to be weak mediators of the intervention effect on BMI: nonorganized PA and dessert consumption.
Strengths and Limitations
The follow-up visit was well attended, and there was good compliance with treatment, with positive feedback from families and children. The study maintained a good level of statistical power, at least for the primary endpoint.
One of the main limitations of the follow-up study was the risk of randomized group contamination after the end of the intervention. In fact, the study protocol recommended offering usual care only once the intervention was over, to both treated children and those from the control group. However, it cannot be ruled out that the pediatricians may have adopted a different approach to usual care after training and experience with MI techniques developed during the trial. Moreover, we did not ask them to record contact with participants during the follow-up period. Interestingly, we observed a rebound effect during the follow-up period rather than an improvement in the control group, as would be expected in the case of contamination. What is more, the rebound effect did not differ between pediatricians, suggesting an effect independent from the pediatrician’s approach.
Another limitation of the study is that parent-reported dietary habits may be strongly affected by desirability bias, and the bias may be stronger for those who received the intervention with 4 MIs in which they had to agree on dietary habit and PA targets than for families in the control group. This bias can lead to an overestimation relating to the effect of intervention on dietary and PA habits and could also explain why at 24 months we still observed an effect on some lifestyle habits but not on BMI. Nevertheless, desirability bias does not have an excessive influence on the results of the mediation analysis, where we compare the relative influence of different behavioral variables in both groups during the intervention and follow-up periods.
Considerations on Designing Effective Long-Term Intervention
According to the review by Janicke et al14 and the experiences reported by Resnicow et al12 and Taylor et al,13 it seems that longer intervention with fewer interviews per year or with more shorter interviews tailored to the needs of each family could at least maintain the effectiveness of MI for 24 months, with similar consumption of resources.
Some studies in the child obesity literature have reported the positive long-term effects of different individual interventions: long-term interventions, including boosters, based on family-based behavioral intervention in primary care36; parent-centered dietary-modification programs combined with child-centered physical-activity skill-development programs37; or structured outpatient training programs consisting of physical exercise, nutritional education, and behavioral therapy.38,39
Wilfley et al40 demonstrated that maintenance treatments based on behavioral skills and social facilitation improve the long-term efficacy of weight loss treatments. Researchers have underlined the need for strong maintenance treatments to sustain effects after weight loss treatment.34 We are therefore planning MI reinforcement for the treated children, while continuing to monitor the control group.
The intervention proposed in this trial has been designed to be sustainable and feasible when scaled up to the entire population of overweight 5-year-olds in the province of Reggio Emilia.15 A longer intervention risks disrupting the sustainability of the intervention. The cost-effectiveness and cost opportunity of longer individual interventions should be compared with the costs of community-based interventions.
Finally, it must be considered that even interventions proven to be effective in some groups can be completely ineffective or damaging in others, highlighting the importance of considering pretreatment variables (BMI at baseline, family support) as modifiers of the effectiveness of different strategies.34 This was the case for our intervention, which had no effect on boys or children whose mothers had low levels of education.
The long-term effectiveness of MI in BMI control cannot be inferred on the basis of immediate effectiveness. There is a need for sustainable and effective boosters and maintenance strategies for MI interventions.
We are grateful to all pediatricians working in the Province of Reggio Emilia for their participation in this project. Their work and attention to details contributed to the success of the study. We wish to thank Paola Albertini (Local Health Authority, Reggio Emilia) for her support and assistance in providing the customized corporate Intranet Web form for the trial.
- Accepted October 15, 2015.
- Address correspondence to Serena Broccoli, Epidemiology Unit, Local Health Authority of Reggio Emilia, via Amendola 2, Reggio Emilia, Italy. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- World Health Organization (WHO)
- Branca F,
- Nikogosian H,
- Lobstein T
- Barlow SE; Expert Committee
- Walpole B,
- Dettmer E,
- Morrongiello BA,
- McCrindle BW,
- Hamilton J
- Taylor RW,
- Cox A,
- Knight L, et al
- Janicke DM,
- Steele RG,
- Gayes LA, et al
- Davoli AM,
- Broccoli S,
- Bonvicini L, et al
- ↵Istat. Demography in Figures [in Italian]. Available at: http://demo.istat.it/. Accessed June 1, 2013
- Fabbri A,
- Palomba A
- Luoghi di Prevenzione
- Prochaska JO,
- Velicer WF
- OKkio alla SALUTE.
- Wijnhoven T,
- Branca F
- Fergusson D,
- Aaron SD,
- Guyatt G,
- Hébert P
- West DS,
- DiLillo V,
- Bursac Z,
- Gore SA,
- Greene PG
- Hardcastle SJ,
- Taylor AH,
- Bailey MP,
- Harley RA,
- Hagger MS
- Collins RL
- Fröhlich G,
- Pott W,
- Albayrak Ö,
- Hebebrand J,
- Pauli-Pott U
- Prochaska JO,
- DiClemente CC
- Collins CE,
- Okely AD,
- Morgan PJ, et al
- Copyright © 2016 by the American Academy of Pediatrics