1,012 articles were obtained from Medline (n=204) and Embase (n=808) and were screened
to remove duplicates. Titles and abstracts of the remaining 875 unique articles were
screened. 851 irrelevant studies were rejected, leaving 24 potentially relevant articles.
Two reviewers independently examined the full text of the 24 remaining articles and
assessed for inclusion. Eight articles were excluded due to participants being outside
the specified age range; two others because they described primary prevention interventions;
a further two because insufficient participant information was available and one because
the technology-based intervention was part of a multicomponent intervention. Therefore
a total of 11 studies were considered suitable for inclusion. No further articles
were retrieved from the bibliography search of these 11 included studies. Figure 2
demonstrates the study selection process. [See Additional file 6]
An overview of study characteristics extracted for this review is presented in Table
1 [See Additional file 1]. Year of publication ranged from 2006 (22) to 2017 (23).
Studies were all RCTs, three of which were cluster RCTs, where two randomised schools
(24,25) and one randomised a mixture of schools and YMCAs (15). The studies were carried
out in a range of countries, with the majority taking place in the USA (n=6) (8,15,22,23,26,27)
and others in New Zealand (n=2) (28,29), Canada (n=1) (30), Malaysia (n=1) (25) and
the Netherlands (n=1) (24). Study retention rates ranged from 70.2% (22) to 100% (25,29).
The total number of participants analysed varied from 26 (30) to 742 (24), with the
mean age ranging from 10 years (15) to 16 years (23). The majority of studies included
both genders, with a small number of studies conducted among females only (n=2) (22,23).
Two studies focused exclusively on participants of one particular ethnicity; one focused
on Chinese-American adolescents (26), while the other focused on African-American
girls (22). Two studies did not specify ethnicity (24,30), while the remaining seven
studies focused on children from a variety of ethnic origins (8,15,23,25,27–29).
All studies focused on technology-based interventions as secondary prevention, since
participants were overweight or obese prior to commencement of study. However, two
studies also included participants who were normal weight and presented this data
separately. In this case, the intervention also served as primary prevention (24,26).
One study included participants who engaged in binge eating or overeating behaviours
(8). Two studies had a family component, one of which provided three 15-minute internet
sessions for the parents so that they might acquire the skills to support their children
in leading a healthy lifestyle (26). The other study required the adolescents to participate
in the study with a parent (22).
There was some heterogeneity in how weight categories were defined. Seven studies
used the CDC definition (8,15,22,23,26,27,30) and two studies used the IOTF definition
to explain the terms normal, overweight and obese (24,29). Two studies did not employ
an established definition; one required participants to have a BMI > 25 kg/m2 (25), while the other classed its participants as overweight or obese with a BMI z-score
between 1.0 and 2.5 (28)
Details of the intervention and control groups, outcomes, follow-up time points and
key findings, including the effectiveness of interventions on BMI and other weight-related outcomes, for each included study are summarised in Tables 2, 3 and 4 [See Additional files
2, 3 and 4]. The mean difference of weight-related outcomes between intervention and
control groups was also calculated and presented, with the P-values in bold indicating
a statistically significant mean difference between groups.
Technology-based interventions were sub-divided into three main categories; active
video games or exergaming, internet or web-based interventions and mobile phone communications.
Intervention duration ranged from eight weeks (26) to two years (22,28). Intervention
intensity was reported in all but two studies (15,25) and was highly heterogenous.
Follow-up time points varied across studies from between two months (26,28) and 24
months (22,24,28). Four studies compared a technology-based intervention to a nil
intervention control group (8,23,24,27); four studies compared a technology-based
intervention to a non-technology-based intervention, and these included stationary
cycling to music (30), written pamphlets (25) and lifestyle programmes (15,28). Three
studies compared the effectiveness between different technology-based interventions,
however, the intervention of interest had an interactive or tailored component compared
to the passive or sedentary technology-based control intervention (22,26,29). The
study outcomes varied depending on the aims of the studies but they all included a
BMI outcome, which took the form of either BMI, BMI-score or BMI percentile. Other
common outcomes were other anthropometric, dietary and PA measurements.
Active Video Game Interventions
Five studies explored the effectiveness of active video gaming or exergaming [Table
2 – see Additional file 2] (15,23,27,29,30).
Adamo et al. (30) investigated the effectiveness of an interactive video game which
involved stationary cycling, known as ‘GameBike’, compared to stationary cycling to
music. The study focused on exercise adherence, energy expenditure, aerobic fitness,
body composition, and cardiovascular disease risk markers in overweight and obese
adolescents. The author found that the music group had significantly better adherence
and expended more energy than the ‘GameBike’ group. There were no other notable differences
between the two groups. Within both groups, statistically significant findings included
a reduction in peak heart rate (HR) at peak workload, an improvement in peak workload
and a reduction in time to exhaustion from pre to post intervention. There were no
statistically significant between or within group differences for BMI.
Furthermore, in the second study, Maddison et al. (29) explored the effect of active
video games on weight, body composition, PA and physical fitness compared to the effect
of sedentary video games. This study established that children in the active video
games group had significantly decreased their BMI, BMI z-score, % BF and body weight
compared to the children in the sedentary video games group.
The objective of the third study by Staiano et al. (23) was to examine the effect
of exergaming on adolescent girls’ body composition and their cardiovascular risk
factors compared to a control group who adhered to their normal PA level. No statistically
significant differences in body composition or cardiovascular risk factors between
the two groups were found at follow-up.
The aim of the fourth study by Trost et al. (15) was to evaluate the effects of active
video gaming on PA and weight loss in children participating in a community-based
weight management programme. Statistically significant increases in PA were confirmed
in the active video gaming group compared to the control group. The control group
participated only in the programme and had no access to active video gaming. Both
groups had statistically significant reductions in percentage of children overweight
and BMI z-scores, however, the active gaming group demonstrated statistically significantly
greater reductions.
Finally, Wagner et al. (27) investigated the impact of dance-based exergaming on obese
adolescents' perceived ability to exercise, their psychological adjustment and their
BMI compared to a wait-list control group. It was noted that there was a statistically
significant increase in self-reported perceived ability to exercise compared to the
control group. However, no statistically significant differences were found in BMI
z-score within or between the two groups.
In conclusion, two out of these five studies demonstrated a statistically significant
mean difference for the intervention compared to the control for all or some of their
weight-related outcomes; Maddison et al. (29) found beneficial results for four weight-related
outcomes. BMI, BMI z score, body weight and % BF, with a statistically significant
mean difference of -0.24, -0.06, -0.72 and -0.83 respectively when compared to control
group, with P-values of 0.02, 0.03, 0.02 and 0.02. Trost et al. (15) found a statistically
significant mean difference of -0.16 for BMI z-score when compared to the control
group with a P-value of <0.001. The remaining findings were not statistically significant
on between and within group analyses and displayed a degree of heterogeneity in estimates
in terms of trend of direction.
Internet-based Interventions
Five studies examined the effect of web-based or internet interventions [Table 3 –
see Additional file 3] (8,22,24–26).
Chen et al. (26) aimed to examine the efficacy of a theory-driven, family-based online
programme to promote healthy lifestyles and weights in Chinese-American adolescents.
This was compared to guidance given on a general health website. Outcomes included
BMI, Waist-to-hip ratio (WHtR), diet, PA and knowledge about PA and nutrition. Chen
et al. (26) found that adolescents in the intervention group, compared to the control
group, had statistically significantly decreased their WHR and their diastolic blood
pressure (DBP) while statistically significantly increasing their PA, improving their
diet and increasing their knowledge in regard to PA and nutrition. Statistically significant
within group changes for the intervention group included WHtR, DBP, PA, fruit and
vegetable intake and knowledge related to PA and nutrition. There were no statistically
significant within-group changes for any outcomes within the control group.
Similarly, Ezendam et al. (24) conducted a study to evaluate the short- and long-term
results of FATaintPHAT, a web-based computer-tailored intervention which aimed to
increase PA, decrease sedentary behaviour and promote healthy eating in adolescents.
The control group was a nil intervention group. Outcomes included self-reported behaviours
in terms of diet, PA and sedentary behaviour, step count, fitness measured by a shuttle
run test, as well as anthropometric measures. Analysis for this study was conducted
for the total study population and repeated for all at risk students, namely those
who did not meet behavioural recommendations at baseline. This study found that, for
at risk students, the intervention had no statistically significant effect on anthropometric
outcomes or on sedentary behaviour but did show a statistically significant increase
in fruit and vegetable consumption at four month follow-up and increased steps at
two year follow up. For the total study population, FATaintPHAT had no statistically
significant effect on BMI, WC or percentage of overweight or obese students. However,
at four month follow-up, the intervention group was less likely to report drinking
more than 400ml of sugar sweetened beverages per day compared to the control group.
At two year follow-up, this difference was not statistically significant.
Jones et al. (8) examined the effect of an internet-facilitated intervention for weight
maintenance and binge eating in adolescents. The control was a wait-list control group,
and outcomes included BMI, binge eating behaviours, fat and sugar intake, depression
and programme adherence. This study reported no statistically significant differences
in outcomes between groups. The only within group statistically significant finding
in the intervention group was a reduction in objective and subjective binge episodes
from baseline assessment to post-treatment and follow-up assessment.
A further web-based RCT by Nawi and Jamaludin (25) attempted to determine the effectiveness
of an internet-based intervention (obeseGO!) to address obesity among adolescents
in Kuala Lumpur compared to a control group who were provided with written health
education pamphlets. Outcomes measured were anthropometric measures; specifically,
BMI, WC and % BF. This study found no statistically significant reduction in any outcomes
between groups. However, on within-group analyses, mean BMI, WC and % BF in the obeseGO!
group were statistically significantly lower after the intervention.
Finally, Williamson et al. (22) conducted a study to test the efficacy of an internet-based
lifestyle behaviour modification programme for African-American girls over a two year
period. The control group was a passive internet health education programme. Outcomes
measured were anthropometric measures, weight loss behaviours and website use. At
the six month follow-up, participants in the intervention group in comparison to the
control group had statistically significantly reduced their % BF, however, this difference
was not sustained and at the end of the 24 month intervention period, there was no
difference in % BF between the two groups. Adolescents in both groups reported a statistically
significant improvement in exercise and a reduction in overeating, in comparison with
baseline.
To summarise, one study out of the five had a statistically mean difference for the
intervention compared to the control for one of its weight-related outcomes; Chen
et al. (26) found a small but statistically significant mean decrease in WHtR (0.01)
compared to the control group with a P-value of 0.02. The remaining findings were
not statistically significant when between group analyses were conducted and, as before,
displayed a degree of heterogeneity in estimates in terms of trend of direction.
Mobile Phone Communications
The third category of technology-based interventions, mobile phone communications,
was explored in a single study by Nguyen et al. (28), where the effect of supplementary
therapeutic contact as an additional support to a community-based weight-management
programme for overweight and obese adolescents was examined [Table 4 – see Additional
file 4]. Additional therapeutic contact took the form of telephone coaching, SMS or
email communications. The control group was the weight-management programme alone.
Outcomes measured were anthropometric measures, blood pressure, metabolic profile
and self-reported psychosocial and lifestyle changes. Both groups demonstrated statistically
significant reductions in BMI z-scores and WHtR as well as improvements in metabolic
and psychosocial profiles. The study found that this technology-based intervention
had no statistically significant impact on body weight, BMI, BMI z-score, WC and WHtR
compared to the control group.
In terms of statistically significant findings, active video gaming appeared to have
had the most positive impact. However, it is important to note that, within this review,
five out of eleven studies focused on the use of active video games, while only one
study explored mobile phone-based interventions, and this simple disparity in number
of studies conducted to date may partly account for this outcome.
Methodological quality was evaluated using the Cochrane Risk of Bias Tool for Randomised
Controlled Trials. The results of this tool were translated to the AHRQ standard to
facilitate comparison of study quality (21). All studies fell into the category of
poor quality. Figure 3 is a pictorial risk of bias summary for each of the studies.
[See Additional file 7]
The difficulty in making a definite judgment regarding the presence of selection,
performance, attrition and other bias in the included studies renders the evidence
less trustworthy and creates a challenge for making reliable recommendations for future
practice. The high risk of performance bias found in three studies is concerning particularly
in the lack of blinding of participants and personnel. Most studies used appropriate
random sequence generation, blinding of outcome assessment and selective reporting
so to reduce the likelihood of selection, detection and reporting bias.