Multiple behavior interventions to prevent substance abuse and increase energy balance behaviors in middle school students

ABSTRACT

This study examined the effectiveness of two transtheoretical model-tailored, computer-delivered interventions designed to impact multiple substance use or energy balance behaviors in a middle school population recruited in schools. Twenty middle schools in Rhode Island including sixth grade students (N = 4,158) were stratified and randomly assigned by school to either a substance use prevention (decreasing smoking and alcohol) or an energy balance (increasing physical activity, fruit and vegetable consumption, and limiting TV time) intervention group in 2007. Each intervention involved five in-class contacts over a 3-year period with assessments at 12, 24, and 36 months. Main outcomes were analyzed using random effects modeling. In the full energy balance group and in subsamples at risk and not at risk at baseline, strong effects were found for physical activity, healthy diet, and reducing TV time, for both categorical and continuous outcomes. Despite no direct treatment, the energy balance group also showed significantly lower smoking and alcohol use over time than the substance use prevention group. The energy balance intervention demonstrated strong effects across all behaviors over 3 years among middle school students. The substance use prevention intervention was less effective than the energy balance intervention in preventing both smoking and alcohol use over 3 years in middle school students. The lack of a true control group and unrepresented secular trends suggest the need for further study.

The major causes of preventable morbidity and mortality in the USA are the result of modifiable behaviors including tobacco smoking, alcohol abuse, physical inactivity, and poor diet [1]. These problematic health behaviors often begin during adolescence and follow a worsening trajectory over time. Past 30-day tobacco use doubles between sixth grade (3.3 %) and eighth grade (7.7 %) and then increases from 3 to 5 percentage points per year through the 12th grade [2]. Between ages 13 and 21 years, the percentage of young people who report past month binge drinking increases from about 1 to 50 % [3]. Rapid declines in healthy patterns of physical activity/exercise and healthful diet are also observed during adolescence [4–6], while media usage (television, video games, etc.) increases [7]. Nearly 80 % of 11- to 15-year-old adolescents had multiple risk behaviors, including low physical activity (PA), 2 h or more of TV time per day (TV), high-fat diet, and inadequate fruit and vegetable intake (FV) [8]. Relative to younger cohorts, behavioral risks peak during the high school years while readiness to change reaches an all time low [9]. This transition towards unhealthy lifestyle behaviors during adolescence underscores the importance of preventive interventions targeting the middle school years.

Results of smoking and alcohol use prevention programs with adolescents have produced mixed results. One large, intensive, and well-controlled smoking prevention program among students in 40 school districts (grades 3–12) found no significant separation between treatment and control schools over 9 years [10]. However, more recent well-controlled studies have successfully reduced and prevented smoking and alcohol use in adolescents. One review of 66 adolescent smoking cessation and prevention programs found that collectively these programs doubled quit rates (12 % in treatments groups versus 7 % in controls) and that the most effective programs were theory-based, motivationally enhanced, or contingency-based, and classroom- or computer-based [11]. One successful study targeting multiple addictive behaviors was the Communities That Care program which applied evidence-based interventions at school, home, and community levels [12]. In this large randomized community trial, treatment group adolescents reported about 50 % lower rates of starting to smoke and drink alcohol between sixth and eighth grades.

Computer and internet-based interventions are increasingly used as platforms for smoking and alcohol interventions. Among youth, computer-based interventions are well established and their effectiveness is comparable to other classroom- and parent-involved programs [11, 13]. One ten-session prevention program targeting middle school youth reduced past-month alcohol and cigarette use and instances of heavy drinking, as well as negative alcohol-related consequences among intervention group youth at 6-year follow-up [13]. Another 12-session internet-based program reduced cannabis and alcohol use among eighth grade students [14]. Whether briefer computer-based interventions can produce comparable results remains to be seen.

Transtheoretical model (TTM)-tailored interventions have demonstrated and replicated efficacy for smoking cessation with adults [15, 16], even when smoking was included as one of multiple behaviors targeted [17–20]. However, results with adolescents have been more limited and mixed. One study randomized a large population of adolescents recruited in primary care settings to receive a computerized smoking cessation or prevention intervention. Results supported the efficacy of the smoking cessation intervention but showed no effect for smoking prevention [21]. Another study with diverse adolescent females attending family planning clinics found comparable results [22].

Despite the need for the promotion of PA and energy balance behaviors [4] and enthusiasm for interactive technologies as a platform for population-based prevention [23, 24], there are few rigorously tested computer-based interventions for energy balance behaviors among youth. Literature reviews concluded that technology-based interventions are promising but in need of more and better research [25–28]. The effectiveness of computer-based interventions was unclear in part, since they have rarely been examined alone, typically being combined with other school- or home-based components [29–32]. Follow-up periods were often too short [31, 33], and several interventions showed little to no impact on energy balance behaviors. A few randomized controlled trials, however, have included lengthy follow-ups (1 to 2 years), intervention tailoring, and recruitment of large representative samples of students [30, 32, 34, 35]. Mauriello and colleagues [35] demonstrated the efficacy of a computer-delivered TTM-tailored program delivered in high school classrooms targeting physical activity, fruits and vegetables, and TV time outcomes. Although these results are promising, their generalization to middle school students remains untested.

Populations with multiple behavioral risks suffer greater morbidity, disability, and premature mortality [36, 37], and most of the US population has multiple behavioral risks [38, 39] that either develop or worsen during adolescence. To address these growing concerns, a 2009 NIH report [40] recommended that multiple behavior change research should be a top priority for all NIH preventive interventions. With accumulating evidence that interventions can impact multiple behaviors on a population basis, researchers have concluded that such strategies are promising for addictive behaviors [41] and necessary for obesity prevention [23, 42, 43].

CURRENT STUDY

This study addresses major gaps in prevention research targeting multiple risk behaviors (smoking, alcohol use, exercise, diet, and TV viewing) for entire populations of adolescents in middle schools. This study reports the 36-month results of a randomized two-arm comparison trial of TTM-tailored, multiple behavior change interventions delivered to entire populations of middle school students. Both interventions were based on the transtheoretical model of behavior change (TTM) [44]. The smoking and alcohol prevention program targeted smoking and alcohol acquisition using cluster risk-profile tailoring [45, 46]. The energy balance intervention targeted physical activity (PA, at least 60 min for at least 5 days per week), fruit and vegetable consumption (FV, at least five servings of fruits and vegetables each day), and limited TV viewing (TV, 2 h or less of TV time each day). Each intervention group acted as the comparison condition for the other, and untreated outcomes were examined for transfer effects [47]. Both interventions were relevant for all adolescents, regardless of weight or behavior risk status. The TTM-tailored energy balance intervention [48, 49] was found to be effective with high school populations [35]. The TTM-tailored adolescent smoking cessation intervention [50] was also found effective in an older adolescent primary care population [21].

METHODS

Procedure

This school-based randomized control trial was conducted between 2007 and 2011. Both groups received five tailored intervention sessions. Sessions were delivered in the following manner: the baseline assessment session and the first intervention session were administered at the beginning of sixth grade (2007–2008), students received three intervention sessions approximately 2 months apart during seventh grade (2008–2009), and the final intervention session occurring at the beginning of eighth grade (2009–2010). Computer-administered health risk assessments were administered once at the beginning of each school year during middle school with the final assessment (month 36) occurring during ninth grade after students transitioned into high school (2010–2011). All intervention and assessment sessions were delivered in school computer laboratories using laptops provided by the study and assisted by research assistants who were not blind to group assignment.

School recruitment and randomization

Middle school students were recruited from 20 middle schools in Rhode Island. There were 11,312 potential participants in Rhode Island public schools during the 2007–2008 academic year, but 6,037 were excluded because the school required active consent (n = 1,780), the school declined to participate (n = 1,900), or the middle school included only students in seventh and eighth grade (n = 2,357), leaving a potential sample of 5,275 students who met the criteria. The selected 20 schools included 4,158 students who were randomized by school. Eligibility requirements included that students were English speaking and in the sixth grade at the beginning of the trial. Twenty schools were randomized to group utilizing the multiattribute utility measurement approach [51] to balance school-level characteristics. Participating schools were matched on available school-level data (sixth grade class size, percent free lunch eligible, percent English as second language, percent attending college, racial/ethnic composition, smoking rate, alcohol use rate) to form matched pairs of schools which were then randomized to each group. This study had two treatment arms with each group serving as the comparison group for the other. Both groups received comparable computerized assessment and TTM-tailored intervention feedback with multimedia components; however, each intervention targeted different sets of behaviors. Ten middle schools received the energy balance (EB) intervention, and ten schools received a substance use prevention (SP) intervention.

Schools were recruited by project staff who visited school administrators and presented information regarding overarching goals of the grant. All students in sixth grade were included in the invitation to participate. Parents received a letter describing the research and opt-out forms approximately 1 month prior to the baseline session. Few parents (n = 104) refused permission, and only five students refused to participate. Human subjects’ aspects of this research were approved by the University of Rhode Island Institutional Review Board.

Interventions

Each participant interacted with five 30-min computerized TTM-tailored intervention sessions that were group specific: one in sixth grade, three times in seventh grade, and one in eighth grade.

SP intervention

This intervention aimed to reduce tobacco and alcohol use in all adolescents. The effective TTM-tailored cessation intervention was included for adolescent smokers [21, 22, 50]. Since the majority of adolescents were nonsmokers and nondrinkers, a novel approach using cluster membership [45, 46] to provide TTM-tailored prevention feedback was implemented. Tailoring prevention feedback based on these cluster profiles was based on their predictive validity [45, 46, 52]. Details of this cessation intervention have been described [50].

EB intervention

To fit the multiple health behavior intervention into one class period, a combination of full, moderate, and minimal tailoring was used [35, 53]. A fully tailored intervention for PA was used at each intervention time point and included tailored feedback based on assessments of all TTM constructs. Interventions for FV and TV alternated between moderately and minimally tailored. Moderate tailoring included limited assessments and a combination of tailored and stage-matched feedback, and minimally tailored interventions included stage-matched feedback only. Moderately-tailored interventions offered first for FV included feedback on stage of change, decisional balance (pros), and stage-matched feedback on processes of change. Details on the intervention, formative research, pilot study, and outcome results with high school students have been described [35, 48, 49].

Outcome measures

Students in both groups completed a total of four computerized health risk assessments early in each year of the project (sixth, seventh, eighth, and ninth grades).

Stages of change for smoking acquisition and cessation

Since this assessment was administered to all adolescents, the first items determined current smoking status [45, 54–56] so that current and former smokers could receive the cessation intervention, while nonsmokers could receive the prevention intervention. Never smokers and experimental smokers were classified as nonsmokers and asked subsequent questions to determine their stage of smoking acquisition, whereas current and former smokers were then asked questions relevant to determine their stage of smoking cessation.

With smoking status measured, smokers and nonsmokers answered separate questions to assess stages of cessation or acquisition. For nonsmokers, two questions measured acquisition stage [54–56]. Nonsmokers were asked if they were thinking about or planning to try smoking within the next 30 days (acquisition-preparation stage) or 6 months (acquisition-contemplation stage). Participants who reported that they were not thinking of trying smoking in the next 6 months were classified into the acquisition-precontemplation (aPC) stage. Adolescents in the aPC stage were asked Decisional Balance and Temptation measures [45, 46, 55, 56] and placed into clusters based on their answers.

Stages of change for alcohol acquisition and cessation

Since this assessment was administered to all adolescents, the first items determined current drinking status [45, 46, 57, 58] so that current drinkers could receive a cessation intervention and nondrinkers could receive a prevention intervention. Never drinkers and experimental drinkers were asked subsequent questions to determine their stage of drinking acquisition, whereas current and former drinkers were then asked questions to determine their stage of cessation.

For the majority of sixth graders who had not tried alcohol, two questions measured acquisition stage. These adolescents were asked if they were thinking about or planning to try drinking within the next 30 days (acquisition-preparation stage) or 6 months (acquisition-contemplation stage). Participants who reported they were not thinking of trying alcohol in the next six months were classified into the aPC stage. All aPC stage adolescents were asked Decisional Balance and Temptation measures [45, 46, 57, 58] and placed into clusters based on their answers.

Stages of change for energy balance behaviors

Stage of change was assessed using the same algorithm for all energy balance behaviors: (1) precontemplation (not meeting criteria and not planning to meet criteria in the next 6 months), (2) contemplation (not meeting criteria but planning to meet criteria in the next 6 months), (3) preparation (not meeting criteria but planning to meet criteria in the next 30 days), (4) action (meeting criteria for less than 6 months), and (5) maintenance (meeting criteria for more than 6 months). The criteria for each behavior were (a) physical activity (at least 60 min of physical activity for at least 5 days per week), (b) fruit and vegetable consumption (at least five servings of fruits and vegetables each day), and (c) limited TV viewing (2 h or less of TV time each day). Energy balance measures have been validated against established measures [35].

Continuous energy balance measures

Outcomes reporting for the addictive behaviors typically include only point prevalence, whereas outcomes for energy balance behaviors include both point prevalence (percent at criteria) and continuous outcome measures. The continuous outcome measures reported in this study include: (a) physical activity, students reported the number of days per week that they got at least 60 min of PA; (b) fruit and vegetable consumption, students reported the number of servings of FV they ate each day, and (c) limited TV viewing, students reported the number of hours of TV/DVDs watched per day.

Statistical analyses

A 2 × 4 factorial design of repeated measures compared the EB group and the SP group at 0, 12, 24, and 36 months. Efficacy was assessed separately for each target behavior. Outcomes by group differences were evaluated in three ways: (1) energy balance risk reduction goals: stage of change measures identified students who were at risk (not at public health criteria) for each health behavior at baseline (pre-action stages) and progressed to being at criteria (action/maintenance) at follow-up, (2) energy balance relapse prevention goals: stage of change measures identified students who were at criteria (action/maintenance) at baseline and remained at criteria at follow-up, and (3) substance use prevention goals: measures identified students who were not smoking or drinking at baseline and continued to abstain at follow-up. Outcomes reported for addictive behaviors typically include only point prevalence, but outcomes for energy balance behaviors include both point prevalence (percent at criteria) and continuous measures.

Following protocols of other school-based trials, the generalized estimating equation method of analysis controlled for intra-class correlations within each school [59, 60]. Complete datasets were analyzed using complete data methodology—in this case, generalized estimating equations (GEEs) examining repeated measures’ effects and random effects of school as the unit of assignment—and the results were pooled by using SAS v9.1 PROC MIXED and PROC MI for the continuous outcomes and PROC GLIMMIX and PROC MI for the categorical outcomes. Statistical significance of the pooled results was evaluated using t test and degrees of freedom that take into account the uncertainty in the data and the uncertainty due to missing values. GEE allows analysis of longitudinal data for continuous and categorical outcomes [61]. All GEEs, which were run with an unstructured variance matrix, used the comparison group as reference. All time points were included in GEE analyses.

RESULTS

Participants

Table 1 presents the demographics and baseline stage distributions for each group and for the full sample and demonstrates that baseline randomization successfully achieved balanced groups. The sample (N = 4,158) was 48 % female, 65 % White, and 15.6 % Hispanic. About 44 % of students were in action/maintenance (A/M) for PA, about 30 % of students were in A/M for FV, and about 56 % were in A/M for TV at baseline. Most students did not smoke or drink alcohol and did not intend to try smoking or drinking (98–99 %) at baseline (see Table 1).

Table 1.

Baseline demographics by group

Variable	Energy balance (N = 2,184)		Substance use (N = 1,974)		Total (N = 4,158)
	%	N	%	N	%	N
Gender
Male	52.2	1,134	52.3	1,028	52.2	2,162
Female	47.8	1,038	47.7	938	47.8	1,976
Ethnicity
American Indian/Alaskan Native	2.3	51	2.1	42	2.2	93
Asian	2.3	50	2.6	51	2.4	101
Black, not Hispanic	3.1	67	4.6	90	3.8	157
Pacific Islander	0.4	9	0.7	13	0.5	22
White, not Hispanic	64.0	1,393	66.2	1,303	65.0	2,696
Combination	16.1	351	17.2	338	16.6	689
Unknown/not reported	11.8	257	6.6	130	9.3	387
Hispanic	18.7	397	12.2	238	15.6	635
Physical activity stage
PC	27.4	596	32.2	631	29.7	1,227
C	14.4	313	14.6	286	14.5	599
PR	11.5	251	10.9	214	11.2	465
A	19.9	434	18.0	353	19.0	787
M	26.7	582	24.3	476	25.6	1,058
Fruit and vegetable stage
PC	12.2	267	15.3	301	13.7	568
C	22.5	491	20.3	401	21.5	892
PR	34.4	751	35.0	690	34.7	1,441
A	3.4	75	4.3	85	3.9	160
M	27.4	597	25.1	494	26.3	1,091
TV time stage
PC	7.9	172	7.5	147	7.7	319
C	13.1	285	12.9	254	13.0	539
PR	22.1	481	23.6	466	22.8	947
A	8.6	188	6.7	133	7.7	321
M	48.4	1,055	49.3	971	48.8	2,026
Smoking
aPrecontemplation	99.1	2,142	98.9	1,918	99.0	4,060
aContemplation	0.2	5	0.6	12	0.4	17
aPreparation	0.7	15	0.5	9	0.6	24
Smoker	1.0	22	1.7	29	1.2	51
Alcohol
aPrecontemplation	96.1	2,074	95.2	1,853	95.6	3,927
aContemplation	1.1	24	1.5	30	1.3	54
aPreparation	0.8	17	1.1	21	0.9	38
Drinker	2.0	44	2.2	43	2.1	87
Age	Mean (SD)	N	Mean (SD)	N	Mean (SD)	N
	11.38 (.68)	2,129	11.41 (.70)	1,918	11.40 (.69)	4,047

Retention

Figure 1 presents the school- and student-level retention rates at months 12, 24, and 36. School-level retention was 100 % across time points. Student-level retention did not differ significantly between the EB and SP groups at 12 months (82.6 versus 82.8 %), 24 months (76.7 versus 76.4 %), or 36 months (72.3 versus 71.1 %). Overall, 82.7 % were retained at 12 months, 76.6 % at 24 months, and 71.7 % at 36 months.

Fig 1.

Recruitment and retention of subjects across 36 months

Substance use prevention outcomes

Analyses assessed group differences in smoking and alcohol acquisition among those who were nonsmokers and nondrinkers at baseline.

Smoking acquisition

The EB group had less smoking acquisition than the SP group at 12 months (1.2 versus 2.6 %; t(9,460) = 2.40, p < .05, h = .11), 24 months (3.7 versus 6.4 %; t(9,460) = 2.09, p < .05, h = .15), and 36 months (5.7 versus 9.2 %; t(9,460) = 1.99, p < .05, h = .11).

Alcohol acquisition

The EB group had less alcohol acquisition than the SP group at 12 months (2.2 versus 4.5 %; t(9,465) = 2.87, p < .01, h = .13), 24 months (5.3 versus 8.6 %; t(9,465) = 2.48, p < .05, h = .13), and 36 months (10.1 versus 14.4 %; t(9,465) = 2.02, p < .05, h = .12).

Substance abuse cessation

There were too few students (1–2 %) who were smoking or using alcohol at baseline to perform a meaningful analysis (see Table 1).

Energy balance behavior outcomes

Analyses assessed group differences in movement to A/M stages among those in a pre-action stage at baseline for each behavior (see Table 2 and Fig. 2).

Table 2.

Means (SD) by group across time points for three energy balance behaviors for acquisition, prevention, and total sample

	Time point	Energy balance	Substance use	Effect Size
		Mean (SD)	Mean (SD)
Entire sample
Physical activity days	BL	4.71 (1.9)	4.78 (2.0)
	12 months	4.82 (1.8)	4.82 (2.0)	0.00
	24 months*	4.90 (1.8)	4.74 (1.9)	0.09
	36 months**	4.90 (1.9)	4.57 (2.0)	0.17
FV servings	BL	3.79 (2.1)	3.67 (2.1)
	12 months**	4.05 (2.2)	3.48 (2.1)	0.27
	24 months**	4.11 (2.1)	3.33 (2.1)	0.37
	36 months**	3.85 (2.1)	3.33 (2.1)	0.25
TV hours	BL	2.98 (2.3)	3.20 (2.4)
	12 months	2.61 (2.0)	3.09 (2.2)	0.22
	24 months	2.54 (2.1)	2.92 (2.2)	0.18
	36 months*	2.46 (2.0)	3.05 (2.3)	0.27
Pre-action at baseline
Physical activity days	BL	2.78 (1.2)	2.75 (1.2)
	12 months	4.04 (1.8)	3.92 (2.0)	0.06
	24 months**	4.32 (1.8)	3.87 (1.9)	0.24
	36 months**	4.31 (1.9)	3.86 (1.9)	0.24
FV servings	BL	2.61 (1.1)	2.59 (1.1)
	12 months**	3.57 (1.9)	3.05 (1.9)	0.27
	24 months**	3.65 (1.9)	3.00 (1.9)	0.34
	36 months**	3.48 (1.9)	3.01 (2.0)	0.24
TV hours	BL	4.65 (2.1)	4.58 (2.0)
	12 months	3.47 (2.2)	3.77 (2.4)	0.13
	24 months	3.18 (2.3)	3.51 (2.3)	0.15
	36 months*	3.02 (2.2)	3.53 (2.4)	0.22
Action/maintenance at baseline
Physical activity days	BL	6.13 (0.9)	6.25 (0.9)
	12 months	5.39 (1.6)	5.46 (1.7)	0.05
	24 months	5.33 (1.7)	5.37 (1.7)	0.02
	36 months***	5.33 (1.7)	5.09 (1.8)	0.13
FV servings	BL	6.41 (1.6)	6.33 (1.5)
	12 months***	5.11 (2.2)	4.53 (2.1)	0.27
	24 months**	5.13 (2.1)	4.15 (2.2)	0.37
	36 months**	4.67 (2.2)	4.12 (2.1)	0.25
TV hours	BL	1.40 (0.7)	1.49 (0.6)
	12 months*	1.79 (1.4)	2.17 (1.7)	0.24
	24 months	1.94 (1.7)	2.10 (1.8)	0.10
	36 months***	1.93 (1.6)	2.39 (1.8)	0.26

A/M action/maintenance, FV fruit and vegetable, TV television.

*p < .05; **p < .01; ***p < .001.

Fig 2.

Point prevalence acquisition for energy balance behaviors

Physical activity

The EB group had greater percentages than the SP group progressing to A/M at 24 months (48.6 versus 36.1 %; t(4,122) = 3.59, p < .001, h = .24) and 36 months (49.3 versus 37.3 %; t(4,122) = 3.39, p < .001, h = .25).

Fruit and vegetable consumption

The EB group had greater percentages than the SP group progressing to A/M at 12 months (27.7 versus 19.2 %; t(6,687) = 3.46, p < .001, h = .20), 24 months (28.3 versus 18.0 %; t(6,687) = 4.15, p < .001, h = .25), and 36 months (24.9 versus 17.2 %; t(6,687) = 3.23, p < .01, h = 19).

Limited TV viewing

The EB group had greater percentages than the SP group progressing to A/M at 12 months (38.3 versus 30.6 %; t(5,137) = 2.56, p < .05, h = .15), 24 months (47.6 versus 37.2 %; t(5,137) = 3.20, p < .01, h = .20), and 36 months (47.7 versus 39.1 %; t(5,137) = 2.60, p < .01, h = .17).

Energy balance behavior relapse prevention outcomes

Analyses assessed group differences in stability within the A/M stages among those who were in A/M at baseline (see Table 2 and Fig. 3).

Fig 3.

Prevention of relapse for energy balance behaviors

Physical activity

The EB group had greater percentages than the SP group remaining in A/M at 36 months (73.4 versus 65.8 %; t(5,434) = 3.13, p < .01, h = .16).

Fruit and vegetable consumption

The EB group had greater percentages than the SP group remaining in A/M at 12 months (57.9 versus 43.5 %; t(2,870) = 3.85, p < .001, h = .29), 24 months (58.3 versus 34.6 %; t(2,870) = 6.18, p < .001, h = .48), and 36 months (48.6 versus 34.8 %; t(2,870) = 3.55, p < .001, h = .28).

Limited TV viewing

The EB group had greater percentages than the SP group remaining in A/M at 12 months (79.2 versus 70.0 %; t(4,383) = 2.70, p < .01, h = .23) and 36 months (77.6 versus 65.8 %; t(4,383) = 3.22, p < .01, h = .27).

Continuous energy balance outcomes

Group differences for changes in continuous outcome variables for each energy balance behavior were assessed within the entire sample and then within subgroups of students who were either at risk (pre-action stages) or who met criteria at baseline (in A/M stages, see Table 2).

Physical activity

• I.Entire sample. The EB group reported greater numbers of days doing at least 60 min of physical activity than the SP group at 24 months (4.90 versus 4.74; t(14,000) = 2.06, p < .05, h = .09) and 36 months (4.90 versus 4.57; t(14,000) = 3.39, p < .001, h = .17).
• II.Pre-action at baseline. The EB group reported greater numbers of days doing at least 60 min of physical activity than the SP group at 24 months (4.32 versus 3.87; t(5,925) = 4.26, p < .001, h = .24) and 36 months (4.31 versus 3.86; t(5,925) = 4.18, p < .001, h = .24).
• III.Action/maintenance at baseline. The EB group reported greater numbers of days doing at least 60 min of physical activity (5.33) at 36 months than the SP group (5.09) (t(7,779) = 32.66, p < .001, h = .13).

Fruit and vegetable consumption

• I.Entire sample. The EB group reported eating significantly more servings of fruits and vegetables than the SP group at 12 months (4.05 versus 3.48; t(14,000) = 4.75, p < .001, h = .27), 24 months (4.11 versus 3.33; t(14,000) = 6.21, p < .001, h = .37), and 36 months (3.85 versus 3.33; t(14,000) = 4.11, p < .001, h = .25).
• II.Pre-action at baseline. The EB group reported eating significantly more servings of fruits and vegetables than the SP group at 12 months (3.57 versus 3.05; t(9,585) = 5.08, p < .001, h = .27) and 24 months (3.65 versus 3.00; t(9,585) = 5.93, p < .001, h = .34), and 36 months (3.48 versus 3.01; t(9,585) = 4.19, p < .001, h = .24).
• III.Action/maintenance at baseline. The EB group reported eating significantly more servings of fruits and vegetables than the SP group at 12 months (5.11 versus 4.53; t(4,119) = 3.96, p < .001, h = .27), 24 months (5.13 versus 4.15; t(4,119) = 6.44, p < .001, h = .37), and 36 months (4.67 versus 4.12; t(4,119) = 3.76, p < .01, h = .25).

Limited TV viewing

• I.Entire sample. The EB group reported a lower mean number of hours of TV (2.46) at 36 months than the SP group (3.05) (t(14,000) = −2.07, p < .05, h = .27.
• II.Pre-action at baseline. The EB group reported a lower mean number of hours of TV (3.02) at 36 months than the SP group (3.53) (t(7,425) = −2.38, p < .05, h = .22.
• III.Action/maintenance at baseline. The EB group reported watching fewer hours of TV than the SP group at 12 months (1.79 versus 2.17; t(6,226) = −2.55, p < .05, h = .24) and 36 months (1.93 versus 2.39; t(6,226) = −3.21, p < .05, h = .26).

DISCUSSION

This two-arm cluster randomized comparison trial examined the efficacy of two five-session computer-delivered TTM-tailored interventions: an intervention to prevent or reduce smoking and alcohol use (SP) and an intervention to increase or maintain energy balance behaviors (EB) in an entire population of middle school students. Results indicated that students in the energy balance intervention condition who did not meet criteria at baseline were more likely to initiate all three energy balance behaviors at follow-up time points and students at criteria (i.e., already in A/M at baseline) were less likely to relapse to pre-action stages at 12 and 36 months for PA, at all time points for FV, and at 12 and 36 months for TV. Outcome effect sizes were predominantly in the medium to large range. Effects of this magnitude, particularly over a 36-month follow-up period, are uncommon in population-based research and indicate that the energy balance intervention had a substantial impact on energy balance behaviors in these middle school students.

These results replicated and extended a previous randomized controlled trial conducted among high school students [35] to middle school students. Among high school students, some of the observed effects diminished over time; however, in the current study, the magnitude of effects typically increased over time with the largest effects observed at 24 and 36 months. This energy balance intervention, while effective among high school students, had an even greater impact on energy balance behaviors in middle school students. These findings add to a growing body of evidence indicating that computer-tailored interventions (CTIs) are an efficacious way to improve energy balance behaviors among adolescents at the population level [29–35].

This energy balance intervention was effective not only in initiating and maintaining energy balance behaviors but also in reducing smoking and alcohol acquisition in these early adolescents. Among adolescents not intending to smoke, a significantly smaller percentage in the energy balance intervention group were smoking at 36 months (5.7 versus 9.2 %). Similarly, with alcohol use in adolescents not intending to drink, a significantly smaller percentage in the energy balance intervention group reported drinking at 36 months (10.1 versus 14.4 %). These effects compare favorably to the Communities That Care outcomes, in spite of the fact that that study was much larger [12]. Contrary to our hypotheses, rates of smoking and alcohol use in schools receiving the smoking and alcohol prevention intervention did not decrease relative to schools receiving the energy balance intervention.

Several explanations for the disappointing smoking and alcohol use prevention results should be considered. First, despite the fact that the energy balance intervention did not directly target smoking and alcohol acquisition, the focus on increasing healthy lifestyle behaviors may have increased healthy behaviors overall. It should be noted that despite a comparable design, the PACE study showed efficacy on both energy balance behaviors [32] in one group and on sun protective behaviors in the comparison group [62]. Nevertheless, a positive change in some untreated behaviors following health-behavior change interventions has been reported in the literature [47, 63, 64]. Second, the overt focus on addictive behaviors in the smoking and alcohol prevention intervention may have inadvertently increased reactivity and defensiveness in middle school students despite efforts to present positive motivational messages that were nonconfrontational and nonjudgmental. However, this explanation is not supported by independent data collected during the time this intervention was taking place. We compared rates of smoking and alcohol use in independent data collected in schools statewide, comparing those that received either the SP or EB intervention in this study with all nonstudy schools [65]. Overall, 7.5 % of students in the study group reported past 30-day smoking as compared to 8.6 % in nonstudy schools. Alcohol use in the past 6 weeks was reported by 13.1 % of students in the study group compared to 15.9 % of students in nonstudy schools [65]. These data suggest that the smoking and alcohol prevention intervention may have had some effect that simply did not exceed that of the energy balance comparison condition. However, without a true no-treatment control condition, this explanation remains speculative. Follow-up analyses may be able to identify mediators and moderators of intervention efficacy in these groups.

Strengths and limitations

This study represents the first randomized controlled trial of a stand-alone population-based CTI for multiple substance-related and energy balance behaviors delivered to middle school students. Results of the energy balance intervention (Health in Motion) replicate past success among high school students [35], and the content is consistent with the national standards for primary prevention efforts targeting health risk behaviors related to overweight and obesity among youth [66]. Potential effects on untreated behaviors are intriguing and require further empirical investigation. The methodology in this trial was strong and included a randomized design, theoretically-based tailoring, and the application of five iterative intervention sessions. The interventions targeted a population-based sample and achieved high retention over a 3-year follow-up period. These programs are easily integrated into school curricula and are self-directed, and each 30-min session can be administered in a single class period. Past research has had difficulty engaging students in computer-tailored interventions [29] and encountered logistical difficulties when implementing the technology in schools [33]; therefore, it is notable that this study successfully implemented the computer-based interventions in 20 schools and was able to retain and engage schools and students over a 3-year follow-up period.

An important limitation of this study is the lack of a true no-treatment control group. Without a true control condition, it cannot be determined whether the smoking and alcohol prevention program resulted in significant reductions in smoking and alcohol use. This limitation is mitigated in part by the consistent multiple behavior changes in the energy balance group and other TTM research targeting exercise as the primary behavior that also produced transfer effects on untreated behaviors [64]. Despite this limitation, conducting a two-treatment comparison trial was more cost-effective than conducting two stand-alone intervention trials. This choice also maximized school participation. Due to curriculum demands, administrators prefer to take part in comparison trials that guarantee some intervention participation.

Future directions

Further investigation of computer-tailored interventions is needed to determine the best practices for smoking and alcohol use prevention among early adolescents. This energy balance intervention is a stand-alone intervention that is efficaciously administered in school settings; however, it has the potential for even more widespread dissemination via other channels including primary care clinics, community-based organizations, and fitness/wellness centers. The program is easily implemented, self-directed, and applicable to adolescents ranging from 11 to 18 years old making it appropriate for a range of school- or community-based initiatives. Future research should determine the effectiveness of this energy balance intervention nationally among diverse students and through various dissemination channels.