BACKGROUND:

Adolescent marijuana prevalence has not increased since 2005 despite a substantial decrease in the percentage of adolescents who believe marijuana use leads to great risk of harm. This finding calls into question the long-standing, inverse connection between marijuana prevalence and perceived risk of use, a connection central to many arguments opposing marijuana legalization. We tested 2 hypotheses for why marijuana prevalence did not increase after 2005: (1) decreases in adolescent use of cigarettes and alcohol reduced risk for marijuana use and counteracted the expected risk in marijuana prevalence, and/or (2) perceived risk of harm now plays a smaller role in marijuana use.

METHODS:

Data came from the annual, nationally-representative Monitoring the Future study from 1991 to 2016, in which 1 100 000 US students in eighth, 10th, and 12th grade were surveyed.

RESULTS:

The entire sample was stratified into 3 mutually exclusive and exhaustive groups on the basis of cigarette and alcohol use. Within each of the 3 groups, marijuana prevalence increased from 2005 to 2016. Paradoxically, when the 3 groups were combined into 1 analysis pool, overall marijuana prevalence did not increase. The seeming paradox results from a decline in the percentage of adolescents who used cigarettes; as this group grew smaller, so too did its disproportionately large contribution to overall marijuana prevalence. Perceived risk of harm from marijuana remained a strong indicator of use throughout 2005 to 2016.

CONCLUSIONS:

Perceived risk of marijuana remains tightly associated with use, and adolescent marijuana prevalence today would be at or near record highs if cigarette use had not declined since 2005, according to study projections.

What’s Known on This Subject:

The absence of an increase in marijuana use among adolescents in recent years has been a surprise. An increase was expected as a result of a substantial decline in levels of perceived risk of harm from marijuana use among adolescents.

What This Study Adds:

In this study, we show that declines in adolescent use of cigarettes have kept levels of adolescent marijuana use from increasing over the past decade. Without these declines, we project adolescent marijuana use today would be at record high levels.

The absence of an increase in marijuana use among adolescents in recent years has been a surprise and has substantial implications for both the field and particularly the current debate on marijuana legalization. An increase was expected as a result of a substantial decline in levels of perceived risk of harm from marijuana use among adolescents.1 For many decades, perceived risk and marijuana prevalence have closely tracked each other in mirror image, with decreases in perceived risk accompanying future increases in marijuana prevalence2 as well as the inverse.3 A disconnect between these 2 measures has emerged in recent years, as marijuana prevalence has not increased while levels of perceived risk have dropped.1,4 

Figure 1 presents trends in adolescent marijuana prevalence and perceived risk of harm from marijuana use. Marijuana use levels changed little from 2005 to 2016, although perceived risk declined substantially during this time. As Fig 1 demonstrates, this marks a break from the mirror-image connection of these 2 factors that is apparent in the earlier years when changes in perceived risk usually preceded a reverse change in annual prevalence a year later. If the connection between the 2 variables had continued into 2005 and afterward, marijuana prevalence would be expected to have increased by approximately the same proportion that perceived risk had declined in each grade: by ∼40% in 12th grade, 30% in 10th grade, and 20% in eighth grade.

FIGURE 1

Perceived risk of regular marijuana use and past 12 month marijuana prevalence, observed by grade and year. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 1. Note: In 12th grade, 5 out of the 6 randomly distributed forms contained the question on perceived risk of marijuana use, and sample size for this variable is five-sixths, or 83%, of the total sample size. In eighth and 10th grade, the question on perceived risk was randomly distributed to 90% of the sample.

FIGURE 1

Perceived risk of regular marijuana use and past 12 month marijuana prevalence, observed by grade and year. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 1. Note: In 12th grade, 5 out of the 6 randomly distributed forms contained the question on perceived risk of marijuana use, and sample size for this variable is five-sixths, or 83%, of the total sample size. In eighth and 10th grade, the question on perceived risk was randomly distributed to 90% of the sample.

Close modal

One prominent argument against legalizing marijuana centers on a close connection between these 2 factors. It is posited that legalization is a public health danger because it will reduce perceived risk of marijuana use among adolescents in all US states, who will interpret the growing wave of legalization as a sign that the state considers marijuana use to be safe and legitimate.5 Proponents of this argument then predict that a national decline in perceived risk will lead to national increases in marijuana prevalence, given their strong connection historically, and increases in associated health problems.6 For example, the American Academy of Pediatrics presented a figure similar to this article’s Fig 1 in support of its stance against marijuana legalization,7,8 although at the time of publication the disconnect was not yet apparent. Some have suggested that the recent disconnect between these 2 factors considerably weakens the argument against marijuana legalization.9,10 

We tested 2 hypotheses for the surprising lack of increase in marijuana prevalence since 2005, when the disconnect with perceived risk first began. The first hypothesis points to the substantial decline in adolescent cigarette and alcohol use over the past decade that has brought their prevalence to historic lows.1 This decline should work to counter any increases in marijuana prevalence according to both the “gateway”11,12 and “liability”13,14 perspectives. Those promoting the gateway perspective build on the observation that drug use typically follows a sequence in which use of cigarettes and alcohol precedes use of marijuana; consequently, fewer youth initiating this sequence could result in fewer youth ending up as marijuana users. Researchers working along these lines posit that youth who do not use cigarettes or drink alcohol are less likely to experience associated processes that promote marijuana use, such as exposure to drug-using peer networks15 and the “priming” of the brain’s reward system for substance use.16 

Proponents of the liability perspective posit that declines in levels of cigarette and alcohol use are a marker for declines in marijuana use because trends in all 3 substances should move together, to the extent that their use is driven by a subset of adolescents who have a heightened tendency for general polydrug use or, more generally, deviant behavior. Whether resulting from gateway or liability processes, the declining prevalence of cigarettes and alcohol may have lowered adolescent risk for marijuana use and offset the expected increases in marijuana prevalence.

The second hypothesis is that the connection between perceived risk and marijuana use has weakened since 2005, as indicated by a relative risk association that has attenuated over the past decade.

In sum:

  • Hypothesis 1: Expected increases in marijuana use since 2005 have been offset as a result of the concurrent decline in the percentage of youth who use cigarettes and alcohol.

  • Hypothesis 2: Marijuana use has not increased since 2005 because the association of perceived risk of harm and marijuana use has weakened.

Data came from the annual Monitoring the Future study, which uses self-administered questionnaires in school classrooms to survey US adolescents. The project has been approved by the University of Michigan Institutional Review Board. Independent nationally representative samples of eighth, 10th, and 12th grade students were surveyed each year from 1991 to 2016. The cumulative sample size is 428 194 in eighth grade, 392 195 in 10th grade, and 366 406 in 12th grade. Student response rates averaged 90%, 87%, and 83% in eighth, 10th, and 12th grades, respectively, for the combined time period 1991–2016. The great majority of nonresponse is because of student absence.

Any past 12-month marijuana use is based on the question “On how many occasions (if any) have you used marijuana during the last 12 months?” and is coded 1 for respondents who reported at least 1 occasion and 0 otherwise. Perceived great risk of harm in regular marijuana smoking is based on response to the question “How much do you think people risk harming themselves (physically or in other ways) if they smoke marijuana regularly: (1) no risk, (2) slight risk, (3) moderate risk, (4) great risk, and (5) can’t say, drug unfamiliar.” It is coded 1 for respondents who respond “great risk” and 0 otherwise, with category (5) treated as missing data (∼3% in each grade). Ever smoked is based on the question “Have you ever smoked cigarettes” and coded 0 for respondents who report “never” and 1 for those who answered “Once or twice,” “Occasionally but not regularly,” “Regularly in the past,” or “Regularly now.” Had an alcoholic drink but never smoked is coded 1 for adolescents who responded to the question “On how many occasions (if any) have you had any alcoholic beverage to drink – more than just a few sips – in your lifetime” by marking any of the categories “1–2,” “3–5,” “6–9,” “10–19,” “20–39,” or “40 or more,” and, in addition, reported never smoking.

The Monitoring the Future survey consists of multiple forms with both form-specific questions and “core” questions that appear on all forms. The questions on use of marijuana, cigarettes, and alcohol appear on all forms. In 12th grade, 5 out of the 6 randomly distributed forms contained the question on perceived risk of marijuana use, and sample size for this variable is five-sixths, or 83%, of the total sample size. In eighth and 10th grade, the question on perceived risk was on forms that were randomly distributed to 90% of the sample. All data available were used in analyses; when presenting data on associations with perceived risk of marijuana use, we used the sample that received both questions, otherwise we used the full sample.

In the 2 study hypotheses, different empirical findings are predicted. A classic case of confounding is predicted in hypothesis 1, in which the expected marijuana increase is present when the analysis is stratified into subgroups defined by substance use, such as ever smoked or ever drank alcohol. In hypothesis 2, it is predicted that the association of perceived risk and marijuana prevalence has attenuated, which we evaluate by calculating the relative risk for these 2 factors in every year.

In our analysis, we present means and relative risk estimates by using binomial regression with a log link17 and take into account sample weighting and clustering. In our analysis, we used multiple imputation to handle missing data and used the chained equations algorithm18 with 20 imputed data sets in Stata MP 12 (StataCorp, College Station, TX).19 

The analyses also include estimates of a hypothetical standard population, and project estimates of overall marijuana prevalence in 2006 and beyond, assuming the size of the cigarette and/or alcohol groups had stayed at 2005 levels. Details of this method are noted in Supplemental Tables 4–6.

The final analyses exclude cases with imputed values for the main variable of marijuana use in the last 12 months (4% or less in all grades). All variables in the analysis have item-specific missing values of 3% or less.

In Fig 2, we present trends in adolescent use of cigarettes and alcohol. An overall decline in the use of these substances since 2005 is a first requirement for hypothesis 1, and a decline is present for cigarette use. Since 2005, the prevalence of having ever smoked a cigarette declined by 43%, 55%, and 62% among 12th, 10th, and eighth grade students, respectively. In all 3 grades, this is now at an historic low, at 28%, 18%, and 10% in 12th through eighth grade.

FIGURE 2

Percentage of respondents who have used cigarettes and alcohol, by grade and year. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 2.

FIGURE 2

Percentage of respondents who have used cigarettes and alcohol, by grade and year. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 2.

Close modal

Also supporting a decline in the percentage of adolescents who use cigarettes and alcohol is the growing percentage who had never used either. Since 2005, this group grew by 62% among 12th and 10th grade students, to 35% and 54%, respectively. Among eighth grade students, this group grew by 37% since 2005 and now consists of 3 out of every 4 students.

Trends were less pronounced for the size of the group that used only alcohol and not cigarettes. Since 2005, this group grew slightly larger among 12th graders, to 36% in 2016 as compared with a level of 29% in 2005. In 10th grade, the size of this group in 2016 did not differ from its size in 2005, at 29% and 28%, respectively. In eighth grade, the size of the group declined and in 2016 was 16% as compared with 20% in 2005.

In Fig 3, we present trends in marijuana prevalence for the entire sample of adolescents, stratified into 3 mutually exclusive and exhaustive groups defined by cigarette and alcohol lifetime prevalence. In hypothesis 1, it was predicted that marijuana prevalence increased in each of the groups since 2005, a prediction that is strongly supported. In all 3 grades, marijuana prevalence significantly increased for (1) youth who had ever smoked, (2) youth who had never smoked and never drank alcohol, and (3) youth who had drank alcohol but never smoked. The increase in prevalence levels in 2016 as compared with 2005 are each statistically significant as indicated by nonoverlapping 95% confidence intervals for all 3 groups in all 3 grades. Data on estimates and SEs for specific years are presented in Supplemental Table 3.

FIGURE 3

Marijuana prevalence in the past 12 months by cigarette and alcohol status, year, and grade. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 3.

FIGURE 3

Marijuana prevalence in the past 12 months by cigarette and alcohol status, year, and grade. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 3.

Close modal

In Fig 4, we project trends in marijuana prevalence if levels of cigarette and alcohol use were both hypothetically frozen at 2005 levels. The projected prevalence level for marijuana use in 2016 is the highest level ever recorded during the study period (1991–2016) among 12th grade students. In 10th grade and eighth grade, the highest levels of projected marijuana prevalence ever recorded during the study period occur during the past 4 years.

FIGURE 4

Marijuana prevalence in past 12 months: observed and projected if cigarette and alcohol prevalence frozen at 2005 levels, by year and grade. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 4.

FIGURE 4

Marijuana prevalence in past 12 months: observed and projected if cigarette and alcohol prevalence frozen at 2005 levels, by year and grade. Note: detailed information on the Figure’s estimates and SEs is provided in Supplemental Table 4.

Close modal

In analyses supplemental to those for Fig 4, we considered the relative influence of the 2 cigarette and alcohol use groups on the projections (Supplemental Tables 4–6). The results reveal that almost all of the projected increase in marijuana prevalence results from holding cigarette prevalence at 2005 levels and that marijuana prevalence in 2016 would be little changed if only the alcohol group were held constant.

In our analyses, we also examined potential trends in perceived risk as an indicator of marijuana use. In hypothesis 2, it was predicted that the strength of the indicator had attenuated since 2005, a prediction not supported by the study results. In each grade, the association did not significantly change, as indicated by overlapping 95% confidence intervals for the 2016 and 2005 estimates. The relative risk estimate from 2005 to 2016 varied between a small window of 0.250 and 0.299 in 12th grade, between 0.215 and 0.259 in 10th grade, and between 0.137 and 0.160 in eighth grade. In Supplemental Table 7 we present details of these estimates.

We set out with this study to examine the surprising lack of increase in marijuana prevalence among adolescents since 2005. An increase was expected as a result of substantial declines in the perceived risk of marijuana use among adolescents, a factor that had strongly tracked inversely with increases in marijuana prevalence until 2005.

In support of the first study hypothesis, we found that since 2005, marijuana use increased in each of 3 groups on the basis of use of cigarettes and alcohol. Specifically, the analysis focused on adolescents who had (1) ever smoked, (2) never smoked nor ever drank alcohol, and (3) had drank alcohol but never smoked. Increases in marijuana prevalence from 2005 to 2016 were statistically significant for each of these 3 groups in all 3 grades.

This finding at first seems somewhat paradoxical. If increases in marijuana prevalence since 2005 are present when the entire sample is divided into 3 separate groups, why is no increase present when the 3 groups are combined? The answer lies in the decreasing number of adolescents who have ever used cigarettes or alcohol. As more and more adolescents refrained from using these substances, they no longer had the associated high levels of marijuana use. For example, in all grades marijuana use was highest among adolescents who had ever smoked a cigarette, and for this group marijuana use increased substantially since 2005. During this same period, this group decreased dramatically in size, and by 2016 it was less than half its 2005 size among eighth and 10th grade students and more than 40% smaller in 12th grade. Consequently, in the overall sample, the increases in marijuana use among adolescents who had ever smoked a cigarette were offset by both the diminishing numbers of these adolescents as well as the growing numbers of adolescents who had never smoked a cigarette, a group that has consistently had much lower levels of marijuana use.

In sum, the results support the first study hypothesis and show that expected increases in marijuana use since 2005 have been offset by a concurrent, marked decline in adolescent use of cigarettes and alcohol. It logically follows that levels of marijuana use today would be substantially higher had cigarettes and alcohol use not declined since 2005. To assess just how much higher they would be, we projected marijuana use in 2006 and later by using 2005 as a standard population. That is, for each year we took the marijuana prevalence for the study’s 3 groupings of cigarette and/or alcohol use and averaged them per the size of these groups as they were in 2005 to produce an overall marijuana prevalence. These projections indicate that today marijuana use would be at or near the highest levels ever recorded since 1991 in each grade if levels of cigarette and alcohol use had remained at 2005 levels. In addition, this projected increase results almost entirely from the decline in cigarette use.

The results do not support the second study hypothesis that perceived risk of marijuana as an indicator of marijuana use has attenuated since 2005. In each grade, the association of these 2 variables did not significantly differ in 2016 as compared with 2005 and varied little in the intervening period.

We note 3 implications of these findings. First, the lack of an overall increase in adolescent marijuana prevalence over the past decade is not evidence that adolescent beliefs about potential harm from marijuana no longer play a role in marijuana use as some have suggested. A major reason adolescent marijuana use has not increased is that substantial progress has been made in the reduction of cigarette use among adolescents, and without this offsetting influence, marijuana use today would likely be at or near the highest levels in more than 2 decades. These results underscore the importance of continued efforts to reduce adolescent substance use and contradict claims that such efforts are unnecessary.

Second, the study results inform the current debate about marijuana legalization. The findings reveal that the link remains strong between adolescent marijuana use and adolescent beliefs about the risks of marijuana. This link plays a central role in arguments against marijuana legalization, which some fear will lead to increased adolescent marijuana prevalence by reducing perceived risk of marijuana use among adolescents. Our results reveal that this argument is not easily dismissed. At the same time, the results reveal that policies and programs proven to reduce adolescent cigarette use as well as polydrug use can be effective tools to also reduce adolescent marijuana use at the population level.

Finally, the study results support adolescent attitudes toward marijuana as an important target for efforts that are aimed at the reduction of marijuana use. Educational programs and media campaigns that educate youth cohorts about risks of marijuana use (especially heavy use) have been and will continue to be important strategies to prevent youth marijuana use. Much still remains unknown about the risks of marijuana use,20 and when evidence for specific, negative effects accumulates, it can be effective in preventing adolescent marijuana use if it is accurately and effectively conveyed to youth. Information about marijuana risks delivered by health professionals to adolescents is particularly influential.21 

We note that the sample does not contain adolescents who have dropped out of school. For the purposes of this study, we expect that inclusion of dropouts would not change the main study results or substantive conclusions because the findings replicate for eighth and 10th grade students, who have low levels of school dropout. Similarly, school absenteeism also holds the potential to affect the study results, but replication of the study findings in eighth grade, in which absenteeism is the lowest, suggests this influence does not alter the study’s main conclusions.

We also note that the psychological and social characteristics of the study’s 3 cigarette and alcohol groupings may have changed from 2005 to 2016. An important avenue for future work is to document these potential changes, with emphasis on any role they have played in the increase of marijuana prevalence, either directly or through their interaction with changes in perceived risk of marijuana use.

Adolescent marijuana prevalence today would be at or near record highs if cigarette use had not declined since 2005, according to study projections. Marijuana use among adolescents continues to track in mirror image with their beliefs about the risk of marijuana use.

Dr Miech conceptualized the study, drafted the initial manuscript, and performed all analyses; Dr Johnston supervised data collection, contributed to the hypotheses, reviewed and approved the manuscript as submitted, and made substantive intellectual contributions to it; Dr O’Malley supervised data collection, reviewed the manuscript, and made substantive intellectual contributions to it; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: Supported by National Institute on Drug Abuse, grants R01 DA 001411 and R01 DA 016575. Funded by the National Institutes of Health (NIH).

We thank Steven Herringa, PhD, for his guidance in the calculation of SEs for prevalence projections in which estimates from a complex sample design are used.

1
Miech
RA
,
Johnston
L
,
O’Malley
PM
,
Bachman
JG
,
Schulenberg
JE
.
Monitoring the Future National Survey Results on Drug Use: 1975-2015: Secondary School Students
.
Ann Arbor, MI
:
Institute for Social Research, The University of Michigan
;
2016
2
Bachman
JG
,
Johnson
LD
,
O’Malley
PM
.
Explaining recent increases in students’ marijuana use: impacts of perceived risks and disapproval, 1976 through 1996.
Am J Public Health
.
1998
;
88
(
6
):
887
892
[PubMed]
3
Bachman
JG
,
Johnston
LD
,
O’Malley
PM
,
Humphrey
RH
.
Explaining the recent decline in marijuana use: differentiating the effects of perceived risks, disapproval, and general lifestyle factors.
J Health Soc Behav
.
1988
;
29
(
1
):
92
112
[PubMed]
4
Johnston
LD
,
O’Malley
PM
,
Miech
RA
,
Bachman
JG
,
Schulenberg
JE
.
Monitoring the Future National Survey Results on Drug Use: 1975-2015: Overview of Key Findings on Adolescent Drug Use
.
Ann Arbor, MI
:
Institute for Social Research, The University of Michigan
;
2016
5
DuPont
RL
,
Voth
EA
.
Drug legalization, harm reduction, and drug policy.
Ann Intern Med
.
1995
;
123
(
6
):
461
465
[PubMed]
6
Volkow
ND
,
Baler
RD
,
Compton
WM
,
Weiss
SRB
.
Adverse health effects of marijuana use.
N Engl J Med
.
2014
;
370
(
23
):
2219
2227
[PubMed]
7
Joffe
A
,
Yancy
WS
;
American Academy of Pediatrics Committee on Substance Abuse
;
American Academy of Pediatrics Committee on Adolescence
.
Legalization of marijuana: potential impact on youth.
Pediatrics
.
2004
;
113
(
6
). Available at: www.pediatrics.org/cgi/content/full/113/6/e632
[PubMed]
8
Joffe
A
;
American Academy of Pediatrics Committee on Substance Abuse
;
American Academy of Pediatrics Committee on Adolescence
.
Legalization of marijuana: potential impact on youth.
Pediatrics
.
2004
;
113
(
6
). Available at: www.pediatrics.org/cgi/content/full/113/6/1825
[PubMed]
9
Ingraham
C
.
The case for marijuana legalization just got stronger.
Washington Post.
2015
10
Sullum
J
.
As fear and intolerance of marijuana declined, so did adolescent use.
Forbes
.
2016
11
Kandel
DB
, ed.
Stages and Pathways of Drug Involvement: Examining the Gateway Hypothesis
.
New York, NY
:
Cambridge University Press
;
2002
12
Keyes
KM
,
Hamilton
A
,
Kandel
DB
.
Birth cohorts analysis of adolescent cigarette smoking and subsequent marijuana and cocaine use.
Am J Public Health
.
2016
;
106
(
6
):
1143
1149
[PubMed]
13
Vanyukov
MM
,
Tarter
RE
,
Kirillova
GP
, et al
.
Common liability to addiction and “gateway hypothesis”: theoretical, empirical and evolutionary perspective.
Drug Alcohol Depend
.
2012
;
123
(
suppl 1
):
S3
S17
[PubMed]
14
Jessor
R
,
Jessor
SL
.
Problem Behavior and Psychosocial Development: A Longitudinal Study of Youth
.
New York, NY
:
Academic Press
;
1977
15
Kosterman
R
,
Hawkins
JD
,
Guo
J
,
Catalano
RF
,
Abbott
RD
.
The dynamics of alcohol and marijuana initiation: patterns and predictors of first use in adolescence.
Am J Public Health
.
2000
;
90
(
3
):
360
366
[PubMed]
16
Kandel
ER
,
Kandel
DB
.
Shattuck Lecture. A molecular basis for nicotine as a gateway drug.
N Engl J Med
.
2014
;
371
(
10
):
932
943
[PubMed]
17
Lovasi
GS
,
Underhill
LJ
,
Jack
D
,
Richards
C
,
Weiss
C
,
Rundle
A
.
At odds: concerns raised by using odds ratios for continuous or common dichotomous outcomes in research on physical activity and obesity.
Open Epidemiol J
.
2012
;
5
:
13
17
[PubMed]
18
Raghunathan
TE
,
Lepkowski
JM
,
Van Hoewyk
J
,
Solenberger
P
.
A multivariate technique for multiply imputing missing values using a sequence of regression models.
Surv Methodol
.
2001
;
27
(
1
):
85
95
19
Stata Statistical Software: Release 12.0 [computer program].
College Station, TX
:
StataCorp LP
;
2011
20
National Academies of Sciences, Engineering, and Medicine
.
The Health Effects of Cannabis and Cannabinoids: The Current State of Evidence and Recommendations for Research
.
Washington, DC
:
The National Academies Press
;
2017
21
D’Amico
EJ
,
Miles
JN
,
Stern
SA
,
Meredith
LS
.
Brief motivational interviewing for teens at risk of substance use consequences: a randomized pilot study in a primary care clinic.
J Subst Abuse Treat
.
2008
;
35
(
1
):
53
61
[PubMed]
22
Kish
L
.
Survey Sampling
.
New York, NY
:
John Wiley & Sons
;
1965

Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2017-3164.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data