Antidepressant Use in a 3- to 12-Year Follow-up of Anxious Youth: Results from the CAMELS Trial

Abstract

The current study explored whether patient characteristics predicted patterns of antidepressant use (i.e., never used, single episode of use, or two or more episodes) in a naturalistic follow-up. Participants in the Child/Adolescent Multimodal (CAMS) Extended Long-term Study. (n=318) indicated medication use over the course of eight follow-up visits, 3–12 years after receiving treatment in CAMS. 40.6% of participants reported never using an antidepressant during follow-up, 41.4% reported a single episode of antidepressant use, and 18.0% reported multiple episodes of antidepressant use. Greater baseline anxiety severity marginally predicted a single episode of antidepressant use; baseline depression severity predicted multiple episodes of use. Reasons for discontinuing antidepressants included perceived ineffectiveness (31.8%), side effects (25.5%), and improvement in symptoms (18.5%).Exploratory analyses examined predictors of medication use. Findings suggest that antidepressant use is common among anxious youth, as is discontinuation of antidepressant use. Clinical implications and future directions are discussed.

Anxiety disorders are prevalent in children and adolescents, affecting 10–20% of youth [1, 2]. Untreated anxiety disorders are associated with an increased risk of comorbid depression disorders and with reduced academic achievement, occupational functioning, and quality of life [3–7]. Fortunately, interventions are available to treat anxiety disorders in youth, including psychopharmacology [8]. Antidepressants have demonstrated short-term (1–2 years; [9–12] and long-term (3–7 years; [13–20] efficacy, and guidelines recommend antidepressants, particularly Selective Serotonin Reuptake Inhibitors (SSRIs), as a pharmacological intervention for youth anxiety disorders [21].

Little is known about long-term antidepressant utilization in youth, and recommendations on length of use are largely anecdotal. Some practitioners recommend that anxious youth remain on an antidepressant for 6–9 months (25) while others recommend a year [22, 23], after which point youth should try a medication-free period to evaluate symptoms, restarting an antidepressant only if needed [23]. Data collected by the Center for Disease Control indicates that, among individuals 12 years and older on an antidepressant, 68% reported having taken it for two years and 25% for 10 years, regardless of diagnosis [24]. However, there is little research exploring long term patterns of medication use in youth, or the reasons families offer for discontinuing medication. Although mental health providers are frequently asked by parents about the length and correlates of medication use, providers are currently unable to provide evidence about how long anxious youth may need to remain on antidepressants [25, 26].

Studies examining medication use in youth anxiety report varying results. One study found that only 55% of youth who started on an antidepressant continued to refill the medication after 6 months, 34% continued for one year, and 17% remained on an antidepressant for two years [27]. A second found that 73% of patients diagnosed with anxiety were prescribed medication at least once in a 5-year period, and 50% remained on this medication for 5 years [28], although this finding is based on a small (n=108) convenience sample in a single medical center. However, these studies did not examine the pattern of medication use during the follow- up years (e.g., multiple instances of medication use vs. a single instance of medication use) or reasons for discontinuing. Little is known about how many anxious youth attempt multiple trials of antidepressants, and why individuals elect to discontinue medication use.

Antidepressants are also recommended treatment for depressive disorders [29], which are highly comorbid with anxiety [30]. Evidence suggests that anxious youth with a comorbid internalizing disorder (e.g., depression) are less likely to experience remission of anxiety following treatment [31]. Some studies suggest that successful anxiety treatment reduces the likelihood of future depressive episodes [32], although this finding is inconsistent [4, 33]. Comorbid anxiety has been found to predict recurrence of major depressive disorder following treatment of depression [34], suggesting that these disorders often co-occur. It is likely that comorbid depressive symptoms may increase utilization of antidepressants following treatment for anxiety, but research has not explored this question.

Thus, although use of antidepressants is efficacious for the treatment of anxiety disorders in youth, little is known about antidepressant use over a period of years. The present study reports on long-term patterns of antidepressant use among youth who participated in the Child/Adolescent Multimodal (CAMS) Extended Long-term Study (CAMELS; 13) and examines medication use patterns during the naturalistic follow-up period. We examined whether medication use patterns varied based on demographic factors, clinical diagnoses at CAMS baseline, symptoms of depression and anxiety, CAMS treatment condition, and the use of therapeutic services during the CAMELS follow-up period.

Methods

The present study used data from CAMELS [13], the follow-up of the Child/Adolescent Anxiety Multimodal Study (CAMS). At the start of CAMS, youth were 7–17 years and met diagnostic criteria for an anxiety disorder; youth with major depression were excluded. Participants were randomized to medication (sertraline), cognitive behavioral therapy (Coping Cat [35]), the combination, or pill placebo ([9]. See Compton and colleagues [36] for methods and Kendall et al [37] for patient characteristics. After 12 weeks, non-responders (defined as a Clinical Global Impression-Improvement score of greater than 2) in the placebo condition received the treatment condition of their choice. Thus, nearly all participants in CAMS received some form of active treatment by the end of their study involvement.

CAMELS recruitment occurred between 2011 and 2015. Of the 488 CAMS youth, 319 completed at least one CAMELS visit. Participants completed their first CAMELS visit an average of 6.51 years (3–12 years; SD=1.65) after CAMS randomization (14). Assessments occurred approximately every six months (M=0.52 years, SD=0.02) for up to eight visits; participants completed an average of 5.03 visits (SD=2.80). Institutional Review Board approval was obtained at all sites; informed consent was obtained from all individual participants included in the study. See Ginsburg and colleagues [13] for details. To be in the present study, participants had to complete the Anxiety Disorders Interview Schedule Supplemental Services Form (ADIS- SSF; described below). All but one CAMELS participant (N=318) provided this information and were included. In keeping with previous analyses done in this dataset, data from the last visit were excluded due to missing data.

Measures

Socioeconomic Status (SES).

SES was measured using the Hollingshead Index at CAMS baseline [38], using ratings of parental occupation and education level. Occupation is weighted more heavily than education. Data from both parents were collected in 63.5% of the sample (then averaged). The resulting SES scores were dichotomized into low (1–3) and high (4–5).

Mood and Feelings Questionnaire (MFQ).

CAMS baseline depressive symptoms were assessed using the MFQ [39], a 33-item parent-report measure derived from DSM-III criteria. Higher MFQ scores indicate greater severity of depressive symptoms. The MFQ has shown internal consistency (α = 0.75 to 0.78; [40–42], with internal consistency in the current sample of α = 0.91.

Pediatric Anxiety Rating Scale (PARS).

CAMS baseline anxiety was assessed using the PARS [43], a clinician-administered assessment of severity and impairment (higher scores indicate more severe anxiety). The PARS has internal consistency (α=0.64), inter-rater reliability (r=0.97), retest reliability (r=0.55), and convergent validity (r=0.22 – 0.61; [43]. PARS raters were blind to initial CAMS treatment assignment and outcome. In the current sample, the PARS had internal consistency of α = 0.68.

Anxiety Disorders Interview Schedule Supplemental Services Form (ADIS-SSF).

The ADIS-SSF assesses interim mental health service use between CAMELS visits [44]. Research assistants administered the ADIS-SSF at each CAMELS visit, and evaluated the use of psychiatric hospitalization, psychotropic medication, and various forms of psychotherapy, as well as dose, duration, response, compliance, and discontinuation pertaining to services used since the prior visit. Participants were asked, “Have you received any medication for mental health since your last evaluation?” Those who endorsed were asked follow-up questions evaluating the type of medication, dose, duration of use, and medication indication (e.g. depression, anxiety, attention deficit hyperactivity disorder). This study did not assess total duration of medication use across visits due to the fact that the ADIS-SSF did not capture detailed information about instances in which medications were changed multiple times within the reporting period.

Analytic Plan

An antidepressant use variable was created for participants who completed the ADIS-SSF at any time point (N=318) to indicate whether the participant took an antidepressant¹⁰, discontinued an antidepressant, or did not take any antidepressants since the last visit. Using this variable, we conducted descriptive analyses to examine 1) the patterns of medication use at each of the visits (i.e., percentage who took an antidepressant, did not take an antidepressant, discontinued an antidepressant since the last study visit) and 2) participants’ reasons for discontinuing antidepressants, collapsing across all time points.

Additional exploratory analyses were conducted to further examined distinct patterns of medication use during naturalistic follow-up. For participants who completed three or more visits, the antidepressant use variable was used to categorize participants into one of three medication use groups: 1) Never Used antidepressants, 2) Single Episode of antidepressant use, and 3) Cyclers (two or more episodes of antidepressant use). The Never Used participants did not report any antidepressant use throughout CAMELS. Single Episode participants indicated at any CAMELS visit that they had taken an antidepressant since the previous visit and either remained on an antidepressant throughout remainder of the follow-up period or stopped entirely and did not resume taking an antidepressant at another time point. Cyclers indicated that they had taken an antidepressant at some point during CAMELS, discontinued antidepressant use, and then resumed an antidepressant at one or more other points. Of note, medication use groups captured what participants chose to do during the naturalistic CAMELS follow-up, and did not encompass medication status during CAMS. Thus, a participant who was randomly assigned to receive medication during CAMS and immediately discontinued at the end of the trial would be categorized in the Never Used group, as this participant never reported taking an antidepressant at any point during CAMELS follow-up. Additionally, a participant who reported no antidepressant use during CAMELS but began an antidepressant immediately after their last CAMELS follow-up visit would be in the Never Used group. As a result, these groups were intended to be descriptive of use during the follow-up period, but not predictive of future use.

Analyses were restricted to participants that attended three or more CAMELS visits to allow sufficient observations to determine patterns of medication use (i.e., if only two observations were present, then a participant could not be categorized as a cycler), and to be in keeping with previous analyses [14]. To address concerns regarding external validity, descriptive analyses were run to examine the participants who extended less than three follow-up sessions. Participants who attended 3 or more visits (n=256) were significantly more likely to be of higher SES, (χ²(1) = 8.10, p= .004) and treatment non-responders (χ² (1) = 5.55, p = .02) than those who attended less than three visits; participants with three or more visits did not differ on any other variable examined.¹¹

Following the creation of medication use groups (i.e., Never Used, Single Episode, Cyclers) exploratory analyses examined whether any baseline characteristics were associated with group membership. Specifically, a multinomial logistic regression examined whether demographic (age, gender, SES, race) and clinical (pre-CAMS PARS total score, pre-CAMS MFQ total score, CAMS treatment condition, CAMS responder status, number of CAMS comorbid diagnoses, and psychotherapy use since CAMS) variables predicted group membership, using the Never Used group as the reference group. A dichotomized race variable was created using White as the reference group; all other races were collapsed into one category due to the small samples, consistent with previous studies using the CAMS sample (e.g., Compton et al., 2014).

Results

Analyses were conducted in SPSS Version 24. One participant did not complete the ADIS-SSF at any CAMELS visits and was excluded from subsequent analyses. The average age in the current sample was 10.63 (SD=2.80), 55.3% of participants were female (n=176) and 81.4% were White (n= 259). CAMELS participants (N=319; 65.4% of the original CAMS sample) have previously been found to be more likely to be female, of non-Hispanic ethnicity, and of higher SES than non-participants (for details see [14].

Patterns of antidepressant use were relatively stable across visits. Across the entire follow-up, over half of the sample reported taking an antidepressant at least one time. At each visit, approximately one third of participants reported that they were presently taking an antidepressant. Five percent of participants at each visit reported discontinuing an antidepressant since the last visit (see Figure 1). Of those who took an antidepressant and reported a reason for discontinuing (n=198 instances across visits), participants most frequently attributed discontinuation to medication ineffectiveness (n=63, 31.8% of responses) and side effects (n=51, 25.5% of responses). Only 18.5% (n=37) of participants reported stopping because the medication was no longer needed (Figure 2).

Figure 1.

Antidepressant use across all CAMELS assessment time points.

Note: Reporting interval for Time 1 differed from the other follow-up points because it covers the interval between end of CAMS and beginning of CAMELS

Figure 2.

Participants’ reported reasons for stopping and changing medications during CAMELS

Additional analyses examined patterns of medication use during naturalistic follow-up (see Table 1 for additional descriptive information for medication use groups). Using the classification described above (Never Used, Single Episode, Cyclers), 40.6% of participants reported that they had not used any antidepressant during the follow-up period (Never Used n=104), 41.4% reported one episode of antidepressant use (Single Episode n=106), and 18.0% of participants reported more than one episode of antidepressant use during the follow-up (Cyclers n = 46).

Table 1.

Sample characteristics by antidepressant use group

	Never n=104		One Episode n=106		Cyclers n=46		Total N=256
	Mean/n	SD/%	Mean/n	SD/%	Mean/n	SD/%	Mean/n	SD/%
Age	10.54	2.86	10.42	2.51	10.91	3.12	10.56	2.77
Gender (female)	54	51.9%	62	58.8%	29	63%	145	56.6%
Race/Ethnicity
White	81	77.9%	96	90.6%	36	78.3%	213	83.2%
Black	13	12.5%	5	4.7%	3	6.5%	21	8.2%
Other	7	6.7%	1	0.9%	4	8.7%	12	4.7%
Asian	1	1.0%	3	2.8%	3	6.5%	7	2.7%
American Indian	2	1.9%	1	0.9%	0	0.0%	3	1.2%
SES
Low SES	18	17.3%	13	12.3%	4	8.7%	35	13.7%
High SES	86	82.7%	93	87.7%	42	91.3%	221	86.3%
CAMS Treatment Condition
SSRI only	24	23.1%	34	32.1%	15	32.6%	73	28.5%
CBT	41	39.4%	24	22.6%	9	19.6%	74	28.9%
SSRI+CBT	23	22.1%	32	30.2%	12	26.1%	67	26.2%
Placebo	16	15.4%	16	15.1%	10	21.7%	42	16.4%
Attended Psychotherapy since CAMS	37	35.6%	92	86.8%	43	93.5%	172	67.2%
Comorbid Diagnoses
0	62	45.6%	55	41.4%	21	42.9%	138	43.4%
1	49	36.0%	54	40.6%	14	28.6%	117	36.8%
2 or more	25	18.4%	24	18.0%	14	28.6%	63	19.8%
PARS Score	18.23	4.15	19.62	4.22	19.72	3.89	19.08	4.18
MFQ Total	11.62	9.24	14.42	10.03	17.63	9.91	13.86	9.90
Responder	66	65.3%	65	64.4%	25	56.8%	156	63.4%

Note: percentages indicate percentage of participants within that column (e.g., percentage of responders among people who never took medication)

The exploratory multinomial logistic regression model examining predictors of medication use group (Table 2) was a good fit, χ²(462) =479.66, p=.28, and the variables included in the model significantly predicted medication use group membership, χ²(26)=118.78, p<.001. SES, gender, age, CAMS treatment condition, number of comorbid diagnoses, responder status, and PARS scores were not significant predictors of membership in either the Single Episode or Cycler groups relative to the Never Used group. Likelihood ratio tests showed that dichotomized race, χ²(2)=9.04, p=.01, MFQ total score, χ²(2)=11.99, p=.002, and psychotherapy use, χ²(2)=72.84, p<.001, were significant predictors of group membership. Specifically, if participants identified as non-White, the multinomial log-odds for being in the Single Episode compared to Never Used group decreased by 1.30, B=−1.30, SE=0.49, p=.008. In addition, for every one unit increase in CAMS pre-treatment MFQ total score, the multinomial log-odds for being in the Cycler group compared to the Never Used group increased by 0.08 units, holding all other variables in the model constant, B=0.08, SE=0.02, p=.001. Finally, participants who did not engage in psychotherapy after CAMS were less likely to have used medication during that time period. Specifically, the multinomial log-odds for being in the Single Episode compared to the Never Used group decreased by −2.61, B=−2.61, SE=0.41, p<.001. Similarly, for those who did not engage in psychotherapy, the multinomial log-odds for being in the Cycler group compared to the Never Used group decreased by 3.55, B=−3.55, SE=0.68, p<.001. Table 2 shows the association these variables and group membership.

Table 2.

Association between sociodemographic and clinical characteristics and antidepressant use group membership

Predictor	WaldX2	df	ρ	OR	95% Cl
Comparing single episode to never used
Female	.00	1	.994	1.00	.50–2.03
Age	.55	1	.459	.95	.84–1.08
Low SES	.07	1	.788	.87	.30–2.47
Non-Caucasian	7.07	1	.008	.27	.11–71
Non-Responder	.33	1	.569	1..27	.56–2.87
MFQ	2.61	1	.107	1.03	.99–108
PARS	1.70	1	.192	1.06	.97–1.16
Comparing cycler to never used
Female	.35	1	.556	1.31	.54–3.18
Age	.07	1	.787	.98	.83–1.15
Low SES	.47	1	.491	.60	.14–2.56
Non-Caucasian	.13	1	.716	.29	.29–2.33
Non-Responder	1.30	1	.254	1.80	.66–4.91
MFQ	10.95	1	.001	1.08	1.03–1.14
PARS	.32	1	.575	1.033	.92–.834

Note: Odd’s Ratio (OR); Confidence Interval (CI); Socioeconomic Status (SES) scores were dichotomized into low (1–3) and high (4–5) using the Hollingshead Index; Mood and Feeling Questionnaire total score at CAMS pre-treatment (MFQ); Pediatric Anxiety Rating Scale total score at CAMS pre-treatment (PARS); Non-responder is defined as receiving a Clinical Global Impression- Improvement (CGI-I) score of greater than 2 at week 12 of CAMS.

Discussion

This study examined the patterns of antidepressant use over a 3−12 year follow-up period among youth previously treated for anxiety. Over half (59.4%) of the current sample reported one or more episodes of antidepressant use, while a substantial minority (40.6%) did not report antidepressant use at any point over the follow-up. Among those who attended more than three visits, 41.4% reported a single episode of antidepressant use and 18.0% reported two or more episodes of use. These results are largely consistent with the few studies that have examined long-term medication use among youth with anxiety disorders, and offer a valuable contribution to our understanding of medication use in this population by being the first to examine patterns of long-term antidepressant use.

Our analyses indicated that approximately one third of youth were taking an antidepressant at any single visit during CAMELS. This highlights that, although the majority of participants reported trying an antidepressant at some point during CAMELS, at every assessment point the majority of participants were not currently taking an antidepressant. This pattern remained consistent across the follow-up period, with the largest number of discontinuations reported between the end of CAMS and the first CAMELS follow-up appointment (which represented a longer period of time than between any two CAMELS assessments). This finding suggests fairly consistent rates of both use and discontinuation of antidepressant over the course of many years, and may indicate that consistent long-term use of antidepressants is relatively rare in this population.

Ineffectiveness was the most common reason reported for discontinuing an antidepressant, although this was based on verbal report from families rather than based on objective measures of symptoms. In addition, as information regarding dosage of antidepressant medication was not collected, it is possible some participants discontinued medication due to perceived ineffectiveness before reaching the recommended therapeutic dose. Approximately one quarter of participants stopped taking antidepressants because of reported side effects, although those who stopped an antidepressant due to side effects may have switched to a different antidepressant rather than end medication altogether, which we were unable to examine in the current dataset.¹² Less than one fifth of participants reported discontinuing because they felt medication was no longer needed, and it is possible some had benefited from additional therapeutic services (e.g., psychotherapy). Together, this suggests that symptom remission may be a relatively rare reason for discontinuing medication use.

Exploratory analyses explored what baseline variables were predictive of medication use group. Findings demonstrated that more severe pretreatment anxiety was marginally associated with a single episode of antidepressant use, and greater pretreatment severity of depressive symptoms significantly predicted membership in the “Cyclers” group. Taken together, these findings suggest initial symptom severity may be an indicator of more persistent antidepressant use. Participants who did not accessed psychotherapeutic services during the follow-up period were also less likely to have utilized antidepressants during that time. This finding may indicate that those participants with higher acuity and/or higher engagement in services are more likely to utilize medication, and these factors should be examined in future research. Race was also associated with medication use group in the exploratory analyses, with White participants more likely to take medication. This finding is consistent with previous research that found that non- Hispanic White children had higher use of psychotropic medication more broadly [45]. Due to the small samples within racial groups, follow-up pairwise comparisons were not possible. Given the substantial heterogeneity across and within racial and ethnic groups, as well as the non- representative racial distribution of our sample, more research is needed to more carefully consider the influence of race and ethnicity on medication use.

These results have clinical implications. Prescribing providers should continue to discuss both the potential benefits and limitations of medications, as side effects and perceived ineffectiveness appear to be common barriers to consistent use. Providers may wish to incorporate discussions of medication use into conversations about treatment maintenance, particularly with those clients who present with more severe pre-treatment anxiety symptoms and/or depressive symptoms. It may be especially important to consider the cyclical nature of depression for those with notable depressive symptomology, as our findings suggest this may be associated with a more cyclical use of medication over time. Future studies should explore potential interactions between medication use, symptom severity, comorbidity, and ongoing access to mental health services.

Among the strengths of this study were the large sample size and the long follow-up period. However, limitations merit consideration. First, CAMS itself only included participants with a primary diagnosis of generalized anxiety disorder, separation anxiety disorder, or social phobia. Any clinical interpretations based on the results of the current study, therefore, should not extend beyond youth without at least one of these diagnoses. Additionally, nearly all participants had previously received sertraline, CBT, or both interventions as a part of their participation in CAMS. Although no association was found between CAMS treatment conditions and antidepressant use during CAMELS, it is possible that patterns of medication use would differ among those who had never received any form of prior treatment. Future studies could explore treatment-naïve youth and length of medication use in populations that have not chosen to participate in a long-term follow-up to a randomized treatment study. As youth with major depression were excluded from CAMS, additional research is therefore needed to examine medication use in anxious youth with and without depression. CAMS also excluded youth with other significant psychological concerns (e.g., substance use, unmedicated ADHD, combined type, substance use disorders, psychosis, pervasive developmental disorders, active safety concerns, etc.,) and future research should aim to replicate findings in a more representative sample. In addition, specific timing of initiation and cessation of antidepressant use was not collected in the current dataset, limiting our ability to answer additional longitudinal questions.

Additional research is needed to answer the question of exactly how long youth with anxiety disorders remain on antidepressants. Future studies should also aim to utilize objective measures of medication use such as insurance claims or prescription data, as well as examining other psychological and pharmacological treatments youth may have received concurrent with or as an alternative to antidepressants (e.g., CBT, benzodiazepines, etc). Finally, attrition in our dataset biases the sample and thus may not be representative of the larger population. Thus, current findings are descriptive of the current sample and should be used to generate hypotheses in future research.

Summary

This study is one of the first to report patterns of antidepressant use over time for anxious youth, and, to our knowledge, the only study that includes reasons for discontinuation of medication use among anxious youth. Specifically, 3–12 years following CBT and/or medication treatment for youth anxiety, a majority of anxious youth elected to use antidepressants as a treatment option for at least some portion of time. Additional findings reveal that the majority of participants discontinued antidepressant use, most commonly because of perceived ineffectiveness. Exploratory analyses suggest that more severe pretreatment anxiety may associated with a single episode of antidepressant use, while more severe pretreatment depressive symptoms may be associated with multiple cycles of antidepressant use. In addition, participants who were non-white and who did not utilize psychotherapy were less likely to utilize medication. These findings provide useful information for providers when talking with families of anxious youth about antidepressants as well as guidance for future research.