Interventions to Prevent Illicit and Nonmedical Drug Use in Children, Adolescents, and Young Adults: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force

Abstract

Importance.

Illicit and nonmedical (use in ways other than instructed) drug use is common in adolescents and young adults and increases the risk of harmful outcomes such as injuries, violence, and poorer academic performance.

Objective:

To review the benefits and harms of interventions to prevent illicit and nonmedical drug use in children, adolescents, and young adults to inform the US Preventive Services Task Force.

Data Sources:

MEDLINE, PubMED, PsycINFO, and the Cochrane Central Register of Controlled Trials; surveillance through September 20, 2019.

Study Selection:

Clinical trials of behavioral counseling interventions to prevent initiation of illicit and nonmedical drug use among young people.

Data Extraction and Synthesis.

Critical appraisal was completed independently by 2 investigators. Data were extracted by 1 reviewer and checked by a second. Random effects meta-analysis was used to estimate the effects sizes associated with the interventions.

Main Outcomes and Measures:

Number of times illicit drugs were used, any illicit drug or any cannabis use.

Results:

Twenty-eight trials (N=17,482) met inclusion criteria. Health, social, or legal outcomes such as mental health symptom, family functioning, consequences of drug use, and arrests were reported in 17 trials and most showed no group differences. The effects on illicit drug use in 25 trials among non-pregnant youth were highly variable; the pooled effect size did not show a clinically important or statistically significant association with illicit drug use (standardized mean difference =−0.08 [95% CI, −0.16 to 0.01], 23 effects [from 22 studies], n=11,932, I²=58.2%). The percent of participants using illicit drugs ranged from 2.3% to 38.6% in the control groups and 2.4% to 33.7% in the intervention groups, at 3 to 32 months follow-up. The median absolute risk difference between groups was −2.3%, favoring the intervention group (range, −11.5% to +14.8%). The remaining 3 trials provided a perinatal home-visiting intervention to pregnant Native American youth. One found a reduction in illicit drug use at 38 months, but not at earlier follow-up assessments. Across all 28 trials, only one trial reported on harms and found no statistically significant group differences.

Conclusions and Relevance:

The evidence for behavioral counseling interventions to prevent initiation of illicit and non-medical drug use among adolescents and young adults was inconsistent and imprecise, with some interventions associated with reduction in use, and others associated with no benefit or increased use. Health, social, and legal outcomes were sparsely reported and few showed improvements.

In 2017, 5,455 adolescents and young adults aged 15 to 24 years died from an illicit or prescription drug overdose; a 50 percent increase from 2007 to 2017.¹ Illicit and nonmedical drug use (use of medications in ways other than instructed; hereafter illicit and nonmedical drug use are collectively referred to as “illicit drug use”) is associated with additional negative consequences in young people, including an increased risk of motor vehicle accidents,^2,3 violence,⁴ suicidal behavior,⁵ and lower educational achievement.^6,7 Some long-term negative effects have been associated specifically with heavy cannabis use in adolescence, including anxiety in midlife;⁷ impaired development of emotional resiliency;⁸ impairments in abstract thinking, attention, learning, and psychomotor functioning;^9,10 and increased risk of psychosis.^11,12

Illicit drug use is common in adolescents and young adults. The 2018 Monitoring the Future report on adolescent drug use estimated that that 47.8% of 12^th graders in the United States had ever used an illicit drug, including cannabis, which was the most frequently reported drug, as well as inhalants, hallucinogens, amphetamines, prescription drugs, and heroin.¹³ Based on the 2016 National Survey of Drug Use and Health results, it was estimated that young adults ages 18 to 25 years had the highest rate of illicit and nonmedical drug use, with 23.2% reporting use in the past month.¹⁴

In 2014, the US Preventive Services Task Force (USPSTF) concluded that evidence was insufficient to assess the balance of benefits and harms of primary care-based behavioral interventions to prevent or reduce illicit drug or nonmedical pharmaceutical use in children and adolescents (I statement).¹⁵ The current review was undertaken to help the USPSTF update their recommendation on this topic.

METHODS

Scope of Review

This review addressed 3 key questions (KQs), shown in Figure 1. A draft of the analytic framework, review questions, and inclusion and exclusion criteria was posted on the USPSTF website from May 10 to June 7, 2018 for the purpose of gathering public input. Detailed methods (e.g., more detailed information about quality rating criteria, data elements that were abstracted, study and intervention characteristics examined in meta-regression or subgroup analyses, methods for grading the strength of evidence for key questions, expert review and public comment process) are available in the full evidence report (URL not yet available).

Figure 1.

Analytic Framework

^aRefer to USPSTF Procedure Manual for interpretation of the analytic framework¹⁹

Evidence reviews for the US Preventive Services Task Force (USPSTF) use an analytic framework to visually display the key questions that the review will address to allow the USPSTF to evaluate the effectiveness and safety of a preventive service. The questions are depicted by linkages that relate interventions and outcomes. A dashed line indicates a health outcome that immediately follows an intermediate outcome

Data Sources and Searches

In addition to evaluating all trials included in the previous review¹⁶ and references excluded from the previous review that would be eligible because of expanded inclusion criteria, MEDLINE, PubMed (for publisher-supplied records only), PsycINFO, and the Cochrane Central Register of Controlled Trials were searched for relevant English-language literature published beginning January 1, 2013 for children and adolescents or January 1, 1992 for young adults up to age 25 and ending June 14, 2018 (eMethods). The database searches were supplemented with bibliographies of other relevant reviews, suggestions from experts, and monitoring news and table-of-contents alerts. Ongoing surveillance was conducted after June 2018 through September 20, 2019 to identify newly published studies that may affect the findings of the review. This was accomplished through review of publications in high-impact-factor journals and article alerts. One relevant RCT was identified during the surveillance window, however was it judged not to substantially change the review results or affect our conclusions.¹⁷

Study Selection

Two reviewers independently reviewed abstracts and full-text articles against prespecified inclusion criteria (Figure 2). Discrepancies were resolved via discussion and consultation with another reviewer as needed. Randomized clinical trials (RCTs), including cluster randomized trials, and nonrandomized controlled intervention studies were included if they were published in the English language and assessed behavioral counseling interventions designed to prevent or reduce illicit and nonmedical drug use in children, adolescents, and young adults (age ≤25 years), including pregnant females, who did not regularly use illicit drugs or medications for nonmedical psychoactive effects were included. Interventions could target other risk behaviors in addition to illicit drug use (e.g., alcohol use, tobacco use, risky sexual behavior) but were included only if there was some intervention content that directly addressed illicit drug use and the study reported a drug use outcome. A minimum of 3 months’ follow-up was required. Interventions that utilized components that could not be replicated in a health care setting, such as broad public health, media, or policy interventions were excluded.

Figure 2.

Literature Flow Diagram

^aReason for Exclusion

Geography: Not a country with a very high HDI ranking

Setting: Excluded on the basis of setting alone (e.g., substance abuse treatment centers, school classrooms, worksites, inpatient/residential, other institutions (e.g., juvenile detention facilities)

Comparative Effectiveness: Control group received active intervention

Outcomes: No relevant outcomes

Population: Does not target youth or young adults, or average age of study sample >22 years old; targets youth with health conditions that limit generalizability (schizophrenia/psychosis, HIV), individuals in juvenile justice system, court-mandated

Condition: Non-psychoactive drugs, or >50% with regular drug use (weekly use, injection use, positive screener), harmful/hazardous use, or diagnosable disorder

Intervention: Not a primary care-relevant behavioral counseling intervention, or drug misuse is not a primary target of the intervention

Study Design: Not an RCT or CCT

Follow-up: <3 months (12 weeks) follow-up post baseline (does not apply to harms)

Study Relevance: Study aim not relevant

Poor Quality Rating: Study was poor quality

Non-English: Publication not in English

Conference abstract: Study abstract from conference only

^bStudies may appear in more than one Key Question

Trials in countries rated as “Very High” human development according to the United Nations¹⁸ were included if they were conducted in health care settings or other settings judged to be generalizable to primary care, including research, community, virtual, and school health clinics. Studies conducted in most other school settings were excluded because of concerns that generalizability to primary care may be limited. Studies were included if they used schools only for recruitment purposes, however, as long as they recruited from multiple schools and most participants were not attending sessions at their own schools, or if they studied interventions that were conducted entirely online and did not involve interactions among students at the same school or between students and teachers.

RCTs and nonrandomized controlled intervention studies that did not have a control groupwere excluded, and allowable control groups included no intervention (e.g., usual care, wait list), a minimal intervention (e.g., pamphlets, links to pre-existing internet resources, or no more than a single brief contact per year), and attention controls (similar format and intensity but a different content area).

Data Extraction and Quality Assessment

Two reviewers applied USPSTF design-specific criteria (eTable1 in the supplement)¹⁹ to assess the methodological quality of all eligible studies and assigned each study a quality rating of “good,” “fair,” or “poor.” (eTable 1) Discordant quality ratings were resolved through discussion or consultation with another reviewer. Studies were rated as poor quality and excluded from the review if they had a major flaw, such as very high attrition (generally >40%); differential attrition between intervention groups (generally >20%); substantial lack of baseline comparability between groups without adjustment; or major concerns about the trial conduct, analysis, or reporting of results. One reviewer extracted key elements of included studies into standardized forms in DistillerSR (Evidence Partners, Ottawa, Canada). A second reviewer checked the data for accuracy. Study design details were abstracted, along with population characteristics, intervention characteristics, and results.

Data Synthesis and Analysis

Summary tables were created for all KQs showing study, population, intervention characteristics, and outcomes. Three trials provided very intensive pre- and post-natal home visits to pregnant American Indian youth. Because these trials (the Family Spirit trials) were substantially different from the other included trials, both due to the population of interest and the nature of the intervention, their results will be discussed separately from the remaining trials, which are referred to as the “general prevention” trials.

Illicit drug use was selected as the primary outcome for meta-analysis. Nineteen trials reported a continuous measure, most commonly the number of times illicit drugs were used over a specified period, and thirteen trials reported the dichotomous outcomes of any illicit drug use or any cannabis use. Continuous measures were converted to Hedges g, which is a standardized mean difference (SMD), based on either change from baseline or mean post-test scores, after converting all “times used” variables to the same time window of the previous 3 months. For dichotomous outcomes, log-ORs were converted to Cohen’s d and then to Hedges g using standard formulae.²⁰ Hedge’s g can be interpreted as a Cohen’s d, where a small effect is typically considered to be 0.20 to 0.50.²¹ Odds ratios were either extracted from the studies directly or calculated based on the study-reported numbers of persons with and without the event for each group.

Pooled analyses of the general prevention trials were conducted using random effects meta-analyses on SMDs for three categories of substances: illicit drugs, alcohol, and tobacco. Where multiple intervention groups or follow-up timepoints were provided, the intervention group with the most intensive or comprehensive drug prevention component was selected, reported at 6 to 12 months or the closest to that time frame. Dichotomous measures were selected over continuous measures, 1-month timeframes were selected over longer observation windows, and outcomes assessing the use of any illicit drug over the use of a single drug (e.g., cannabis) were selected when multiple outcomes were reported for the same study. In addition, ORs and between-group mean differences were pooled separately to better understand effects in the native units. Additionally, analyses of cannabis-specific results in native units are provided.

The DerSimonian and Laird (DL) model was used to calculate the pooled effect estimate across studies. In addition, because the DL method is prone to insufficient coverage of the full 95% confidence intervals when the number of studies is small and statistical heterogeneity is high (and I² were typically near or above 50% in this review), restricted maximum likelihood (REML) models with the Knapp-Hartung correction for small samples were also run when pooling fewer than 10 trials. Funnel plots were generated and Egger’s test was run to explore small-study effects, which can be an related to publication bias.²² Additionally, for the primary drug use outcome, meta-regression and subgroup analyses were conducted to explore study and intervention characteristics that were associated with effect size. Stata version 15.1 (StataCorp LP, College Station, TX) was used for all analyses. All significance testing was 2-sided, and results were considered statistically significant if the p-value was 0.05 or less.

RESULTS

Twenty-eight studies (N=17,482) were identified that met inclusion criteria, including 27 RCTs and 1 nonrandomized controlled trial,²³ reported in 37 publications.^23–58 Across all trials most participants were between the ages of 10 and 18, although two trials focused on young adults^37,43 and others covered a wide range, including young people up to age 20 to 24 years.^39,48,49,56 See eTable2 for a list of all included studies, and eTable3 and eTable4 for summaries of the study and population characteristics.

Nineteen (68%) of the included studies addressed broad populations for universal prevention of drug use, but some focused on selected groups at increased risk of substance use or harms from substance use, such as pregnant American Indian youth,^26,28,56 girls in foster care,⁴¹ sexual minority teens (self-identifying as lesbian, gay, bisexual, transgender, or questioning),⁵⁵ youth with asthma,⁴⁸ youth who were truant³¹ or had other school-related behavior problems,³³ or who had some early signs of risky substance use.⁴⁶ Planned intervention dose was variable, with a median of 3 sessions (IQR 1–10 session), but a range of 1 to 46 sessions. Duration ranged from 1 day to over 3 years, with a median of 7 weeks (IQR, 1 day–26 weeks). Interventions for 12 of the trials were delivered exclusively through a computer.^{25,33,35,43,45,50,51,53–55,57,58} Six trials (with 8 intervention groups) took place in primary care settings.^{23,39,49,57–59}

Only 4 trials appeared to focus on illicit drugs without explicit discussion of other substances or behaviors.^43,53–55 Eight trials focused broadly on substance use in their intervention messages, including alcohol or alcohol and tobacco in addition to illicit drug use.^{23,37,45,46,48,57–59} The remaining used even more broadly-targeted interventions that addressed additional behaviors such as family functioning,^{24,33,35,36,38,41,50,51} risky sexual behavior,^{25,33,39,41,49} mental health and emotional well-being (including social skills training),^{25,35,41,49–51} truancy and delinquent behaviors,³¹ and breastfeeding and infant care.^26,28,56 Three trials examined a home-visiting intervention for pregnant American Indian adolescents and young adults recruited through the Indian Health Service covering a range of preventive health topics.^26,28,56

KQ1. Do primary care-feasible or referable interventions [EO: We prefer to keep the original wording, consistent with the wording in the posted Research Plan. We have updated the figure with this language so the text and figure are consistent] to prevent drug use in children, adolescents, and young adults improve health outcomes or other related outcomes?

Health, social, or legal outcomes were reported in 16 of the general prevention trials^{24,25,31,33,35,36,38,41,43,49–51,54,57–59} and all 3 Family Spirit trials.^26,28,56 No single outcome was widely reported. Mental health outcomes such as depression, anxiety, and externalizing symptoms, and family functioning were the most commonly-reported health outcomes, reported by 9 of the general prevention trials^{24,25,35,36,38,41,50,51,54} and all 3 Family Spirit trials.^26,28,56 Most general prevention trials found no group differences on mental health symptom scales after 3 to 24 months (Figure 3), and results were mixed in the Family Spirit trials. Measures of family functioning were reported in 5 of the general prevention trials.^{35,50,51,33, 36} Improvement in several family functioning outcomes (family communication, parental monitoring, and maternal closeness) were found in 3 trials of computer-based interventions targeting middle school-aged females and their mothers^35,50,51 (Figure 4). Across all timepoints (up to 24 months), differences in change between groups in these trials most commonly fell between 0.3 and 0.6 on a 5-point scale. The other two trials did not find group differences on measures of communication,^33,36 parental monitoring,³³ or positive parenting, although data needed to include these results on the forest plot were not provided.³³

Figure 3.

Mental health outcomes summary (KQ1) among the Family Spirit trials, mean difference between intervention and control groups, by outcome

CES-D = Center for Epidemiologic Studies Depression Scale; CG = control group; CI = confidence interval; IG = intervention group; SD = standard deviation

*Mean value at follow-up, rather than change from baseline

Note: Effect sizes are study-reported mean differences at follow-up (for Barlow, 2006, and Barlow, 2013) and beta-weight (Walkup, 2009)

Figure 4.

Mental health and family functioning outcomes summary (KQ1) among the general prevention trials, standardized mean difference between intervention and control groups, by outcome, for main timepoint only

CG = control group; CI = confidence interval; IG = intervention group; MH = mental health; SD = standard deviation

*Mean value at follow-up, rather than change from baseline

Note: Effect sizes include a variety of measures reported by studies, if available, or a calculated between-group difference if study-reported values were not reported; effects include mean difference in change between groups, mean difference between groups at follow-up, regression parameter estimates (e.g., beta-weights, b-weights), Cohen’s d

KQ2. Do primary care-feasible or referable interventions to prevent drug use in children, adolescents, and young adults improve drug use outcomes?

The effects of the general prevention interventions on illicit and nonmedical drug use were wide ranging, and the pooled analysis indicated a very small effect size that was not statistically significant and was unlikely to be clinically important (pooled SMD= −0.08 [95% CI, −0.16 to 0.01], 23 effects [from 22 studies], n=11,932, I²=58.2% (Figure 5, Table 1).²¹ The pooled OR for any illicit drug use or any cannabis use was 0.85 (95% CI, 0.67 to 1.07, 11 effects [10 studies], n=8162, I²=43.1% (Table 1). The percent using illicit drugs at follow-up ranged from 2.3 to 38.6% in the control groups and 2.4 to 33.7% in the intervention groups. The median absolute risk difference between groups was −2.3%, favoring the intervention group (range, −11.5% to +14.8%). When examining times used in the previous 3 months, the pooled mean difference (MD) between groups was −0.21 times (95% CI, −0.44 to 0.02, 11 studies, n=3651, I²=51.0%, Table 1).

Figure 5.

Primary drug use outcome (KQ2) for general prevention trials, standardized mean difference between intervention and control group, sorted by specific outcome

CG = Control group; CI = Confidence interval; CZ = Czech Republic; IG = Intervention group; NR = Not reported; NS = Not significant; SD = Standard deviation; US = United States; Vertical dashed line indicates the overall measure of effect

*Mean value at follow-up, rather than change from baseline

Table 1.

Summary of meta-analysis results for substance use outcomes for general prevention trials (KQ2)

Outcome	No. studies	Type of effect	Pooled result (95% CI)	No. studies (effects) in MA	I2, %	Tau2	N	Range of effectsa	Median (IQR) effectsa
Primary drug outcome	25	SMD	−0.08 (−0.16, 0.01)	22 (23)	58.2	0.022	11,932	−0.58 to 0.69	−0.10 (−0.22 to 0.05)
% Any illicit drug use	11	OR	0.85 (0.67, 1.07)	10 (11)	43.1	0.058	8162	0.42 to 3.52	0.81 (0.64 to 0.98)
% Any illicit drug use	11	ARD	--	--	--	--	--	−11.5 to 14.8	−2.3 (−3.8 to 0.5)
% Any cannabis use	6	OR	0.79 (0.59, 1.06)c	5 (6)	16.1	0.019	5651	0.51 to 1.34	0.78 (0.67 to 0.90)
% Any cannabis use	6	ARD	--	--	--	--	--	−11.5 to 2.85	−2.5 (−3.8 to −0.6)
Times used in previous 3m	12	MD	−0.21 (−0.44, 0.02)	11 (11)	51.0	0.037	3651	ΔΔ: −7.5 to 1.0 Δ: −1.1 to 1.5	ΔΔ: −0.3 (−1.6 to 0.0) Δ: 0.7 (−0.4 to 1.3)
Times used cannabis in previous 3m	10	MD	−0.23 (−0.48, 0.01)	10 (10)	58.1	0.045	3616	ΔΔ: −2.7 to 1.0 Δ: −1.1 to 1.5	ΔΔ: −0.3 (−0.9 to 0.0) Δ: 0.7 (−0.4 to 1.3)
Primary alcohol outcome	23b	SMD	−0.11 (−0.16, −0.06)	21 (22)	4.9	0.001	11,438	−0.36 to 0.40	−0.04 (−0.12 to 0.10)
% Any alcohol use	5	OR	0.79 (0.59, 1.06)c	5 (5)	0	0.009	4985	0.62 to 1.40	1.07 (0.82 to 1.25)
% Any alcohol use	5	ARD	--	--	--	--	--	−8.2 to 10.2	2.5 (0.6 to 5.8)
% Risky alcohol use	5	OR	0.92 (0.72, 1.17)c	5 (5)	0	0.0	5078	0.77 to 1.45	0.94 (0.88 to 1.20)
% Risky alcohol use	5	ARD	--	--	--	--	--	−4.7 to 8.9	0.8 (−2.4 to 4.6)
Times used alcohol in previous 3m	8	MD	−0.29 (−0.53 to −0.05)c	8 (8)	20.7	0.014	3192	ΔΔ: −1.2 to 0.8 Δ: −1.9 to −0.5	ΔΔ: −0.2 (−0.4 to 0.2) Δ: −0.6 (−1.3 to −0.5)
Total drinks in previous 3m	3	MD	--	--	--	--	--	ΔΔ: −3.8 to 2.8 Δ: NA (0 trials)	ΔΔ: 1.4 (−2.2 to 2.5) Δ: NA (0 trials)
Primary tobacco outcome	16b	SMD	−0.09 (−0.15 to −0.03)	15 (15)	0	0.0	8366	−0.41 to 0.29	−0.06 (−0.14 to 0.04)
% Any tobacco use	7	OR	0.91 (0.73 to 1.14)c	6 (6)	0	0.0	5373	0.63 to 1.69	1.08 (0.88 to 1.32)
% Any tobacco use	7	ARD	--	--	--	--	--	−8.6 to 8.5	0.8 (−2.1 to 5.8)
Times used tobacco in previous 3m	8	MD	−0.27 (−0.55 to 0.01)c	8 (8)	0	0.0	2893	ΔΔ: −5.5 to −0.2 Δ: 0.54	ΔΔ: −1.0 (−2.2 to −0.3) Δ: NA (1 trial)

Abbreviations: Δ = difference between group at follow-up; ΔΔ = difference between groups in change from baseline; 3m = 3 months; ARD = absolute risk difference; IQR = interquartile range; k = number of effects in the meta-analysis; MD = mean difference between groups; NA = not applicable; OR = odds ration; obs. = observations; SMD = standardized mean difference (Hedges g)

^aRange of effects for all study groups and timepoints, i.e., not limited to records in the meta-analysis

^bNumber of trials reporting the specific substance use outcomes (any use, risky use, times used, total drinks) does not add up to the total number of trials reporting any outcome because some trials reported only a continuous scale score and are not shown in this table

^cEffect based on restricted maximum likelihood model. Remaining effects based on DerSimonian & Laird model

Only 4 trials reported the effect of their interventions on misuse of prescription medications specifically. All 4 were computer-based interventions and all were broadly targeted at substance use and other non-substance-related outcomes.^33,35,50,51 All 4 reported greater reductions in misuse of prescription medications with the intervention, ranging from 0.1 (95% CI NR)³³ to 11.3 (95% CI −22.6 to −0.08)³⁵ fewer times used over the previous 3 months, at up to 24 months’ follow-up.

Despite the small pooled effect, some interventions did show statistically significant reductions in illicit drug use at one or more follow-up timepoint.^{23,33,35,39,49–51,53,57,58} Among the general prevention interventions, these included the computer-based interventions targeting young adolescent females,^35,50,51,53 the computer-based version of Familias Unidas intervention (which has been widely studied in school settings) targeting eighth-graders with behavior problems,³³ a primary care clinician training intervention,⁴⁹ and the 46-session intervention for eighth-grade girls in foster care and their foster parents.⁴¹ For example, one of the largest effects was found with the Families Unidas intervention. At 12 months’ follow-up, it found that participants had used cannabis an average of 2.7 fewer times (95% CI −3.7 to 0.5, p<.01) over the previous 3 months, had misused prescription medications 0.2 fewer times (95% CI −1.8 to 1.6, p<0.01), and had used inhalants 1.4 fewer times (95% CI −3.5 to 0.77, p<0.001). Effects were generally maintained through 12 months or beyond in these trials. Most of these interventions involved 9 or more intervention sessions, all but 1⁵³ included components for parents or caregivers as well as the youth themselves, and all addressed a broad range of skills and topics.

However, many interventions showed no clear evidence of benefit and 2 reported increased illicit drug use in youth participating in the interventions for at least 1 drug-related outcome.^38,40 Ten trials had less than 12 months’ follow-up, which may have been insufficient to find differences in younger adolescents with low use levels. None of the potential effect modifiers examined (study, population, intervention, and control characteristics) appeared to explain variability in effect sizes (Figure 6), and there was no evidence of a small studies effect.

Figure 6.

Summary of sensitivity analyses of primary drug outcome (KQ2) for general prevention trials: Results of meta-analyses for subgroups of studies with the indicated characteristics

CI = Confidence interval; DL = DerSimonian and Laird; REML = Restricted maximum likelihood; USA = United States of America

Pooled effects for alcohol and tobacco use both showed statistically significant, but very small pooled effect sizes (alcohol pooled SMD=−0.11 [95% CI, −0.16 to −0.06], 22 effects [from 21 studies], n=11,438, I²=4.9%; tobacco pooled SMD=−0.09 [95% CI, −0.15 to −0.03], 15 studies, n=8366, I²=35.0%, Table 1). Other behavioral outcomes were sparsely reported with most trials finding no differences between groups.

KQ3. What are the harms of to prevent drug use in children, adolescents, and adults?

Only 1 of the included trials (a Family Spirit trial) directly reported on harms.²⁶ The authors stated that the proportion of adverse events and serious adverse events was similar between groups after accounting for increased contact time within the intervention group, but did not provide detailed data.²⁶ As mentioned above, 2 general prevention trials reported increased illicit drug use in intervention groups over the control groups.^38,40Additionally, 7 other trials reported increases in illicit drug, alcohol, or tobacco use that were not statistically significant, with point estimates for the SMD larger than 0.20 or an OR of 2.0 or greater.^{25,28,36,45,48,56,57}

DISCUSSION

This evidence review included 28 studies, with findings that were inconsistent for illicit and nonmedical drug use, and effects ranging from clearly beneficial to possibly harmful (with the evidence summarized in Table 2). While some interventions were associated with reduced illicit and nonmedical drug use, they tended to either target a relatively narrow population (e.g., 10 to 14 year-old Asian-American girls, truant youth, 8th grade girls in foster care) with limited generalizability to other populations, or to have not had their results replicated.

Table 2.

Summary of evidence among all 28 included trials (n=17,482) of interventions to prevent illicit and nonmedical drug use in children, adolescents, and young adults, by key question

Key question	No. of Studies, (No. of Observations)	Summary of findings	Consistency/ precision	Other limitations	Strength of evidence	Applicability
KQ1 (Health and social/legal outcomes)	19 (9042)	No single health, social, or legal outcome was widely reported. Family functioning was improved in 3 computer-based general prevention trials among middle school-aged girls and their mothers; isolated group differences were found for delinquency (in 2 of 5 trials), global functioning (in 1 trial), and consequences of drug use (in 2 of 3 trials) in general prevention trials. Group differences were rarely found for a variety of mental health scales (9 general prevention trials, 3 Family Spirit trials).	Inconsistent, imprecise	Wide variety of instruments used; specific outcomes rarely reported by more than 4 trials; many trials limited to narrow demographic or risk groups	Low evidence of small to no benefit	14 conducted in the U.S., 8 limited to females, including 3 that were limited to pregnant American Indians recruited through the Indian Health Service; 4 additional trials conducted in U.S. primary care settings
KQ2 (Behavioral outcomes)	28 (17,482)	Although some general prevention interventions were effective in reducing nonmedical and illicit drug use and other behavioral outcomes, the effects were very wide ranging and the pooled effect for drug use was not statistically significant (pooled SMD=−0.08 [95% CI, −0.16 to 0.01], k=23 [from 22 studies], n=11,932, I2=58.2%). Pooled estimates showed very small beneficial effects on alcohol use (SMD=−0.11 [95% CI, −0.16 to −0.06], k=22 [from 21 studies], n=11,438, I2=4.9%) and tobacco use SMD=−0.09 [95% CI, −0.15 to −0.03], k=15, n=8366, I2=35.0%). Of the 3 Family Spirit intervention trials among pregnant adolescent Native Americans, only the largest and best-quality trial found reductions in drug use and only at long-term (38-month) follow-up.	Inconsistent, Imprecise	Heterogeneity in outcomes reported; only 6 were rating as good quality; 10 trials had less than 12 months’ follow-up, which may be insufficient to find differences younger adolescents with low use levels; many trials limited to narrow demographic or risk groups	Low evidence of small to no benefit	21 of 28 trials conducted in the U.S., 14 of which included >50% racial or ethnic minority participants; primarily targeting adolescents (vs young adults); only 11 trials were conducted in or recruited from health care settings, including 3 that were limited to pregnant Native American youth recruited through the Indian Health Service
KQ3 (Harms)	Reported:1 (322) Paradoxical findings: 2 (1925)	One Family Spirit trial found no differences in adverse events or serious adverse events, after controlling for contact time. In addition, 2 general prevention trials reported statistically significant increases in drug use outcomes, and others reported statistically nonsignificant increases in drug, alcohol, or tobacco use.	Consistency NA, imprecise	Only directly reported in 1 trial, and raw proportions were not provided, nor details of how they adjusted for contact time.	Insufficient	Trial directly reporting harms limited to pregnant Native Americans; trials showing statistically significant harmful drug outcomes conducted in Sweden and the U.S.

Abbreviations: CI = Confidence interval; EPC = Evidence-based Practice Center; k = number of studies; KQ = Key Question; NA = Not applicable; SMD = Standard mean difference; US = United States

The previous USPSTF review on this topic¹⁶ concluded that there was inadequate evidence to determine whether preventive interventions were effective in reducing the likelihood of illicit drug use, based on 6 trials, all of which were also included in the current review.^{23,35,48,50,51,57} The current review added newly published literature, including 22 additional studies, and expanded the scope of this topic to include trials that were deemed feasible for implementation in a health care system (i.e., clinicians and related staff in the primary care setting could have the skills necessary to deliver the intervention, or could refer to others in the health system with the necessary skills) even if the study was conducted in the community or other non-health care settings. Despite this expansion of scope, the strength of evidence was low that primary care-relevant interventions could prevent illicit and nonmedical drug use in children, adolescents and young adults. This was due to the inconsistency in effects, the relatively narrow target populations for most of the interventions that showed a benefit, and the lack of benefit among studies conducted in U.S.-based primary care settings, which were primarily limited to low-dose interventions. One RCT was published after the end of our search window, and although it demonstrated promising results for a pediatrics-based screening and brief intervention program among adolescents, it was very unlikely to affect the results or conclusions of our review. In this trial, there was a reduction in time to first cannabis use among youth who had reported alcohol or cannabis use in the previous 12 months (HR=0.62, 95% CI 0.41 to 0.94), although the effect was slightly smaller and not statistically significant among youth with no baseline alcohol or drug use (HR=0.76, 95% CI 0.44 to 1.32).¹⁷

Most of the studies of interventions to prevent illicit and nonmedical drug use have been conducted in school classrooms or after school settings, and other reviews have found these school-based prevention program to be effective in reducing illicit drug use, including some approaches that would likely be feasible to implement in a healthcare setting.⁶⁰ Two interventions included in this review that have been primarily studied in school settings are Familias Unidas and the Strengthening Families Program. Familias Unidas⁶¹ is a family-based preventive intervention to improve family communication, positive parenting, and parental monitoring to reduce risky substance use and sexual behaviors in Hispanic adolescents. Participants were generally recruited from middle schools, and sessions occurred outside of school hours. The intervention generally included 8 group sessions for parents and 4 family visits that included the adolescents.⁶² The intervention has shown reductions in illicit drug, alcohol, and cigarette use, as well as improvements in family functioning.^63,64 The online version of Familias Unidas that was tested in this review reported lower frequency of illicit drug use in the intervention group, and a between-group difference in the proportion using illicit drugs that was similar in magnitude to that seen in school-based studies.³³

In contrast, the 3 trials in this review that either directly implemented or were based on the Strengthening Families Program 10–14 (SFP10–14) did not prove to be effective outside of school settings.^24,36,38 The Strengthening Families Program is a widely studied intervention designed for high-risk families; several versions exist for different age groups (eg, preschool, elementary, early teens, and high school).⁶⁵ The program consists of 14 sessions and includes training in parenting skills, family life skills, and children’s social skills; can be implemented in various settings (e.g., schools, community centers, drug courts); and has been adapted to be culturally sensitive.⁶⁵ A 10-year follow-up (n=446 families) of an RCT originally conducted in Iowa in 1993 found a long-term reduction in substance use (27.5% of SFP participants had initiated illicit substance use by age 21 versus 38.3% control group [β=−0.14, P <0.001]).^66,67 The beneficial effects of this program appear to have emerged between the 18 and 30-month assessments in most studies, and since the SFP studies included in this review followed participants for a maximum of 24 months, the lack of benefit could have been due to insufficient follow-up rather than the setting. Nevertheless, the fact that the success of this program did not clearly translate to the health care setting illustrates the importance of testing the feasibility and effectiveness of prevention programs in health care settings before recommending their full-scale implementation.

There are important limitations in the research that should be considered. First, reporting of health, social, and legal outcomes was sparse and heterogeneous, limiting the conclusions for these important outcomes. In addition, drug use outcomes were very heterogeneous, making it difficult to draw overall conclusions. Second, no evidence was found that included children younger than age 10 years, likely because a very long follow-up window would be needed to determine the effect on illicit drug use. Additionally, very limited evidence on young adults was found, since this literature primarily focused on the reduction of use in those who were regular users, hazardous users, or who had a likely substance use disorder, which was excluded from the review. These types of secondary prevention trials are included in the USPSTF review on screening and interventions for drug misuse.⁶⁸

Third, most of the interventions were studied by the teams who developed the intervention and have not been replicated by independent researchers. This may be especially important in this field in which outcomes are measured by self-report and are subject to social desirability effects. In general, studies are needed that replicate in a healthcare context, and that further refine and broadly implement some of the effective interventions described in this review. These include the clinician training and QI intervention,⁴⁹ the computer-based Familias Unidas intervention,³³ and some of the computer-based interventions for adolescent girls.^50,51,53 It would also be valuable to conduct a trial in a health care setting of the full in-person version of the Familias Unidas intervention and other interventions proven to be effective in schools. It is important to continue to explore the influence of context and mechanisms of change.

Limitations

This review has several limitations. First, trials that did not report a drug use outcome were excluded, so studies that only included intermediate outcomes such as child development or school functioning were excluded, even if substance abuse prevention might have been a long-term aim. Second, interventions that did not explicitly address prevention of illicit and nonmedical drugs were excluded, although some broad prevention or resilience interventions may be effective in preventing illicit drug use. Third, since trials targeting alcohol and tobacco use (without targeting illicit drugs use) were not comprehensively included, the results for the alcohol and tobacco outcomes do not represent all available evidence on these topics.

Conclusions

The evidence for behavioral counseling interventions to prevent initiation of illicit and non-medical drug use among adolescents and young adults was inconsistent and imprecise, with some interventions associated with reduction in use, and others associated with no benefit or increased use. Health, social, and legal outcomes were sparsely reported and few showed improvements.