1 NEW RESEARCH 59

2 60
3 61
4
5 Lithium Versus Other Mood-Stabilizing Medications in a 62
63
6
7 Q1Q2Longitudinal Study of Bipolar Youth 64
65
8 66
9 Q3 Danella M. Hafeman, MD, PhD, Brian Rooks, PhD, John Merranko, MA, Fangzi Liao, MS, 67
68
10 Mary Kay Gill, MSN, Tina R. Goldstein, PhD, Rasim Diler, MD, Neal Ryan, MD,
11
12 Benjamin I. Goldstein, MD, David A. Axelson, MD, Michael Strober, PhD, Martin Keller, MD, 69
70
13 Jeffrey Hunt, MD, Heather Hower, MSW, Lauren M. Weinstock, PhD, 71
14 Shirley Yen, PhD, Boris Birmaher, MD 72
15 73
16 74
17 75
18 Q4 Objective: Lithium is the mainstay for bipolar disorder (BD) treatment in adults, but evidence in youths is limited. We used data from the Course
76
19 and Outcome of Bipolar Youth (COBY) study to assess whether lithium versus other mood-stabilizing medication (OMS) was associated with
20 77 78
improved outcomes, including mood symptoms and suicidality.
21 Method: COBY is a naturalistic, longitudinal study of 413 youths, 7 to 17.11 years old at intake, with BD. At each visit, medication exposure,
22 79
80
23
24 psychiatric symptoms, and psychosocial function over the preceding follow-up period were assessed using the Adolescent Longitudinal Interval Follow- 81
Up Evaluation. Using mixed models, we determined whether participants taking lithium versus OMS (but not lithium) differed regarding mood 82
25
symptoms, suicidality, psychosocial function, hospitalization, aggression, and substance use. 83
26
Results: A total of 340 participants contributed 2,638 six-month follow-up periods (886 lithium, 1,752 OMS), over a mean follow-up of 10 years.
27
84
28 During lithium (versus OMS) follow-up periods, participants were older, less likely to have lifetime anxiety, and less likely to be on
29 antidepressants (p 85
30 values<.005). After covariate adjustment, the lithium group (versus OMS) had half as many suicide attempts¼(p .03),
31 fewer depressive symptoms (p 86
32 .004), less psychosocial impairment (p¼ .003), and less aggression (p .0004). Similar findings were observed in the subgroup of follow-up periods in
33 87
34 which participants were <18 years old. 88
35 Conclusion: Findings are consistent with adult studies, showing that lithium is associated with decreased suicidality, less depression, and better
36 89
37 psychosocial functioning. Given the paucity of evidence regarding lithium in children and adolescents, these findings have important clinical impli- 90
38 cations for the pharmacological management of youths with BD. 91
39 92
40 Key words: bipolar disorder, child and adolescent, lithium, mood stabilizers, suicidality
93
J Am Acad Child Adolesc Psychiatry 2019;-(-):-–-. 94
95
96
97
98
41 n impressive body of literature supports lithium medication rates
42

43
A as an effective maintenance treatment for
bipolar
disorder (BD) in adults. Meta-analyses of
50

51
(OMS) for the prevention of mood episodes in
adults with
BD.4,5 In the United Kingdom, lithium was
of
58

ran- associated with

44 domized controlled trials (RCTs) in adults with BD 52 a longer time to treatment failure (ie,
53 show
augmentation or
45 that lithium compared to placebo prevents manic
46 54 switching medications due to poor response)
episodes than olanza-
47 and, to a lesser extent, depressive episodes.1-3 Large 55 pine, quetiapine, or valproate.4 In a recent study
obser- assessing
48 vational studies using electronic health records have 56 rehospitalization in a study of electronic
found health records in
49 lithium to be superior to other mood-stabilizing 57 Finland, lithium was associated with lower
FLA 5.6.0 DTD ■ JAAC2770_proof ■ 26 August 2019 ■ 3:36 pm ■
rehospitalization compared to other medications, including 99 116
oral antipsychotics and valproate.5 A recent systematic re- view 100
found that observational studies provide real-world evidence 101
that lithium is superior to other maintenance medications in 102
103
adults with BD, and is associated with less rehospitalization, 104
fewer treatment failures (defined as medication switches or 105
augmentation), and fewer relapses.6 Lithium is also one of the few 106
medications in psychiatry that may help prevent suicide (in both 107
unipolar and bipolar depression). A meta-analysis of RCTs in 108
adults with bipolar or unipolar depression found that lithium 109
110
(compared to placebo) was associated with a substantial 111
reduction in suicide deaths (odds ratio 0.13; p 112
.01) and, impres- sively, all-cause 113 ¼
mortality (odds ratio 0.38; p .04).7 A recent 114 ¼
observational study using linked Swedish national 115

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 1

Volume - / Number - / - 2019

FLA 5.6.0 DTD ■ JAAC2770_proof ■ 26 August 2019 ■ 3:36 pm ■

 HAFEMAN et al.

registers found that lithium was associated with a 14% draws, and possible toxicity, have led to significant ques-
117 reduction in suicide attempts or deaths; valproate was not tions about the role for lithium in children and adolescents. 176
118 associated with any change in suicide rates.8 To our Using longitudinal data from the Course and Outcome 177
119
120 knowledge, only one small study has assessed the Bipolar Youth (COBY) study, we conducted a post hoc of 178
179
121 relation- ship between lithium and suicidal ideation. In analysis to assess whether exposure to lithium (versus 180
122 a group of severely depressed, suicidal adults other mood-stabilizing medications) is associated with 181
123 randomized to lith- ium citalopram (n 40) versus differences in suicidality, mood symptoms, mood 182
124
placebo
þ citalopram ¼(n 40), lithium did not
þ decrease episodes (ie, re- currences of [hypo]mania or 183
125
¼
suicidal ideation over a 4-week period; however, depression), psychosocial function, hospitalizations, and 184
126 185
127 improvement was observed in the small (n 11) aggression. Although RCTs are the gold standard to
186
128 subgroup
¼ with therapeutic lithium levels.9 assess the effects of medication, this rich longitudinal 187
129 the However,
use of lithium in children,
compared adolescents,
to evidence and transitional
in adults, support the effects
dataset of medication
allows a real-worldover a longer
evaluation of duration of 188
follow-up
130 age for
adults remains sparse. Although a recent systematic re- than would be feasible with an RCT. 189
131 view did find evidence for safety and efficacy of 190

132 191
133 adolescents, sample sizes are small, follow-up times METHOD 192
134 are short, and studies comparing lithium to OMS are Study Subjects 193
135 largely negative.10 A review of RCTs in children and Methodology used in the COBY study has previously 194
136 adolescents with acute mania found only 4 relevant trials 195
137 been described in detail.19 Briefly, this ongoing,
and concluded that, although there is some evidence that naturalistic study recruited 413 children and adolescents
196
138 197
139 lithium might be superior to placebo in this population, 7 to 17.11 years old who met criteria at intake for
additional studies with larger samples are needed.11 The 198
140 DSM-IV bipolar I (BD-I), BD-II, or operationally 199
141 largest study to assess the anti-manic effects
defined BD not otherwise specified (NOS)20,21 according 200
142 (Treatment of Early Age Mania [TEAM]) compared 201
143 to the Schedule for Affective Disor- ders and
lithium, divalproex, and risperidone in children and 202
144 Schizophrenia for School-Aged Children–Present and
adolescents 6 to 15 years of age who were currently in a 203
145 Lifetime Version (K-SADS-PL22; see “Methods” in 204
146 manic episode. This study found that lithium and
Supplemental Material, available online). Participants 205
147 divalproex were less effective than risperidone in
were recruited primarily from outpatient clinics (68%), 206
148 treating mania,12 but findings differed greatly according to
149
inpatient units (15%), or advertisements (14%) at the 3 207
site and comorbidity (eg, attention-deficit/hyperactivity 208
150 study sites (University of Pittsburgh Medical Center
disorder [ADHD]).13 In addition, the mean length of 209
151 [UPMC], Uni- versity of California Los Angeles 210
manic episode in this study was almost 5 years, indicating
152 [UCLA], and Brown University). Participants were 211
153 an absence of episodic mood disturbances that may be
enrolled regardless of current mood state or treatment 212
154 more likely to respond to lithium,14 and the average age
status. Participants with schizo- phrenia, IQ less than 213
155 of the subjects was only 10.1 years. More recently, 214
156 70, autism, and mood disorders sec- ondary to
an RCT of lithium in children and adolescents 7 to 17 215
157 substances, medications, or medical conditions were
years old, and in a current manic episode, found that 216
158 excluded. 217
159 lithium can be effective for mania and/or mixed episodes
218
160 in youth. Compared to the placebo group, youths treated Procedures 219
161 with lithium¼showed a 5- point greater decrease in the
162
Each study site’s institutional review board approved 220
Young Mania Rating Scale (Cohen’s d 0.53, p .03); 221
163 the study before enrollment of any participant. Youths
there were no differences between the lithium and and their parents or primary caretakers provided written
222
164 223
165 placebo groups on the Children’s Depression Rating informed assent and consent, respectively, after receiving
Scale, although these participants were not selected for 224
166 a complete description of the study procedures. 225
167 depression.15 There is also some evidence that lithium 226
168 can be helpful for aggression, particularly in the context of Diagnostic Assessment. At intake, current and 227
169 Scale (MRS) and Depression Rating Scalebased
(DRS)on
of 228
BD.17,18disorder, in children and adolescents,16 and may
conduct lifetime psychiatric disorders were assessed
the
youth and parent interviews, using the KSADS-PL. The
improve substance use outcomes in youths with
170 To our knowledge, there have not been studies to KSADS-PL were also administered to better characterize 229
Mania Rating

173 mood
171 174
whether lithium might prevent suicidal behaviors, episo
172
de spitalizations, or aggression in youths with BD. Although
s, adult data are robust, the lack of pediatric data, in mood symptoms. Trained research staff conducted all as- 230
ho sessments; cases were presented to child psychiatrists or 231
232
psychologists, who confirmed diagnoses. K-SADS- PL k 233
175 conjunction with significant potential side effects, coefficients for psychiatric disorders were at least 0.8; Q5 234
blood
2 www.jaacap.org Journal of the American Academy of Child & Adolescent
Psychiatry Volume - / Number
- / - 2019
 LITHIUM IN YOUTHS WITH BIPOLAR DISORDER

intraclass correlation coefficients for the K-MRS and the took other mood -stabilizing medication, but not
235 K- SADS-P depression section were at least 0.95. lithium, for more than 75% of the weeks, that is, 294
236 antimanic anti- convulsants, first- and second- 295
237 Illness Severity. During each follow-up (scheduled 296
generation antipsychotics, and/or lamotrigine (see Table
238 approximately every 6 months), the Longitudinal Interval 297
S1, available online, for spe- cific medications). For
239 Follow-up Evaluation (LIFE)23 was used to assess 298
240 follow-up periods longer than 6 months (eg, due to a 299
psychi- atric symptoms.-The LIFE Psychiatric Status
241 late or missed assessment), we included only the most 300
Rating (PSR) scale was used to assess week-by-week
242 recent 6 months of data. Except where noted, analyses 301
243 fluctuations (since the last follow-up) in mood, anxiety, 302
discussed below used 6-month blocks as the unit of
244 substance use, and other psychiatric symptoms; 303
analysis.
245 hospitalizations were also assessed using a comparable 304
The primary analysis focused on the relationship
246 tool. Beginning in August 2005 (approximately 5 years 305
247 be- tween lithium versus OMS and the following 306
after study launch), the PSR was also used to assess
248 clinical out- comes: suicide attempts and suicidal 307
suicidal ideation. Mood symptoms and sui- cidal ideation
249 ideation (in the absence of an attempt); threshold and 308
250 were rated weekly on a scale of 1 (no symp-- toms) to 6 309
subthreshold depression; threshold and subthreshold
251 (extreme symptoms). The LIFE Psychosocial 310
(hypo)mania; psy- chosocial functioning; hospitalization;
252 Functioning (PSF) tool was used to assess monthly 311
253 aggression; and sub- stance use disorder. As in previous 312
fluctu- ations in psychosocial functioning, including
254 papers, any self-injurious behavior that included 313
work/school, interpersonal, recreation, and satisfaction
255 definite intent (intent score of 3 or more) and/or mild 314
256
domains. The DRS and MRS were also administered 315
lethality (medical threat score of 3 or more) was
257 during each follow- up, providing detailed information 316
considered a suicide attempt.26 There has been one
258 about mood symptoms and suicidality during the worst 317
suicide death in this cohort study. Suicidal ideation was
259 week of the past month. 318
260 since the previous visit was recorded using the PSR considered positive
any time if the
during suicidal interval.
a follow-up ideation (The
itemPSR
on the
for 319
(PSR- Medication. At each follow-up, weekly medication KDRS and/or the PSR was rated as a 3 (mild) or suicidal
more
261 medication). The participant reported the medication
dosage ideation
at was collected starting in August 2005; prior to this 320
that

263 322
264 average weekly dosage was also collected, but not informa- KDRS only in the past month.) We used PSR-
323
265 tion about the number of days that they took the depression and PSR-mania both to assess for 324
266 medica- tion or about the blood levels. Thus, a low threshold mood epi- sodes (based on DSM-IV criteria) 325
267 dose could
average reflect the prescribed dose or could be due
weekly symptoms
and (mean mood
subthreshold scores across each follow-up period). 326
to
268 poor adherence, possibilities that may have very Because mania scores were heavily right-skewed, these 327
different were
269 implications for prognosis. For the purposes of this log transformed. For aggression, we used the 10-item 328
analysis,
270
271 because of difficulties with interpretation of dosage data, we included
aggressionitems related
subscale from to
theboth
BCSverbal
(range,and physical
0-30), aggres-
which 329
defined exposure according to whether the participant took 330
272 331
273 the medication during each week (yes/no), regardless of sion. Substance use was defined as meeting threshold 332
274
276 dosage.
293 criteria for abuse of or dependence on alcohol or drugs 333
parti
277 Q6 of Social Status measure was used to quantify cipan
278
socioeco- nomic status (SES)24 at baseline. The t
279
280 Behavior Control Scale (BCS), a parent-report form took
281 administered at all follow- ups, was used to assess lithiu
282 aggression and other externalizing behaviors over the m for
283 past week.25 more
284
than
285
286 Statistical Analysis 75%
287 For the current analysis, we included all 6-month follow- of
288
up periods where the participant reported taking the
289
290 lithium or another mood-stabilizing medication more week
than 75% of the follow-up weeks. Six-month follow-up s,
291
292 periods were classi- fied as “lithium blocks” if the regar
FLA 5.6.0 DTD ■ JAAC2770_proof ■ 26 August 2019 ■ 3:37 pm ■
dless confounding. Thus, we ran a second model adding time-
of varying covariates that could, 336
sion) in R for each outcome to determine which
other 337
covariates to adjust for in the final model (see 338
medic
Supplement 1, available online, for details). Because of 339
a-
the complexities of longitu- dinal analysis, particularly 340
tions. 341
in a naturalistic study, many time-varying covariates
Blocks 342
may be influenced by previous lithium exposure, and
were 343
adjustment for these covariates can induce bias.27 To 344
classi
address this, we ran two models for each outcome. 345
fied as
First, we adjusted only for variables that clearly 346
“OMS
preceded (and could not be influenced by) lithium 347
blocks 348
expo- sure, either prior to or during COBY study
” if 349
follow-up (eg, age, age at bipolar onset, family history of
partici 350
mania, race, SES, sex, and site). This model avoids 351
pants
overadjustment, but also may not account for important 352

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 3

Volume - / Number - / - 2019

FLA 5.6.0 DTD ■ JAAC2770_proof ■ 26 August 2019 ■ 3:37 pm ■

 HAFEMAN et al.

at least theoretically, be influenced by previous exposure demographic factors predicted lithium exposure over
353 to lithium; these include bipolar subtype, lifetime follow- up (using participant as the level of analysis). We 412
354 disorders (including anxiety, disruptive behavioral, and also tested whether associations between lithium and 413
355 substance use disorders up to and including the clinical outcomes were found in this subset, after 414
356 415
357 relevant assessment), lifetime psychosis, and number of adjusting for the relevant intake clinical variables and
In SAS 9.4 (SAS Institute, Cary, NC), mixed models 416
previous hospitalizations. Lasso-selected
Finally, weintake covariates.
conducted a supplemental analysis to test 417
were
358
359 implemented using selected variables. the relationship between taking lithium (versus OMS) 418
360 To test additional questions of interest, we ran during the 3 months prior to a mood episode and the 419
several
361 supplemental analyses. First, we tested observed relation- severity of the mood episode (eg, suicidal ideation, 420

362
suicide
ships in youths (less than18 years old) only. Second, to attempt, hospitalization, and polarity). Specifically, we 421
363 422
assess whether effects were specific to lithium or were also tested whether lithium versus OMS for at least 5 weeks 423
364
365 found with (and perhaps explained by) other medication during the 8-week “exposure period” (defined as 12 424
368 number of medications into models that showed a for this analysis are given in the Supplemental 427
369 428
signifi- Material
370 429
371
cant effect of lithium. Third, we conducted sensitivity available online. 430
372 an- alyses blocks
removing to test with
(1) whether observed
both lithium andeffects persisted
OMS exposure RESULTS 431
(ie, after
373 comparing lithium-only versus OMS-only) and (2) Demographics 432
whether
375 434
376 versus BD-NOS). Fourth, although the above analyses up periods (886 lithium blocks, 1752 OMS blocks).
377 435
were primarily between-subject, we also tested whether During 65% (574/886) of lithium blocks, participants 436
effects were also on
378 remained when including only those participants who an additional mood stabilizer (ie, antipsychotic, lamotrigine, Q7
437

381 of analysis). Fifth, a subset of the sample had never been up reported an older age of bipolar onset
¼ (p .05) and 440
382 441
383
on lithium at the time of intake, defined as no past had an earlier intake ¼date (p .02) than those who 442
384 lithium exposure and no lithium exposure over the took only non-lithium mood stabilizers (Table 1). 443
385 first follow-up period. In this subset, we tested which Compared to OMS blocks, participants exposed to 444
386 clinical and lithium were slightly older 445
387 Age
TABLE 1 Demographic Characteristics of Individuals Who Contributed to Lithium Blocks Over Follow-up ¼
Compared to
388 of
Those Who Contributed Only to Other Mood-Stabilizing Medication Blocks
389 BD
390 Variable Lithium ± OMS (n ¼ 139) OMS Only (n ¼ 201) Statistic p Onse
391 Q9 Age at Intake, y (SD) 12.6 (3.2) 12.3 (3.3) t ¼ 1.06 t
392 .29 (SD)
393 Intake Date (SD) 14 April 2003 (1.3 y) 02 August 2003 (1.1 y) t ¼ e2.27
394 (4.1)
.02
395 Total follow-up, y (SD) 10.7 (2.9) 10.1 (3.5) t ¼ 1.86
396 (3.9)
.06
397 Sex (% female) 70 (50%) 90 (56%) c2 [1.03 2.00
398
.31 05
399
400
Site c2 [ 3.73 Fami
401 .16 ly
402 Brown 35 (25%) 69 (34%) Histo
403 UCLA 25 (18%) 37 (18%) ry of
UMPC 79 (57%) 95 (47%) Mani
% Nonwhite Ethnicity 20 (14%) 33 (16%) c2 [ 0.26 a
.61
Socioeconomic Status (SD) 3.5 (1.2) 3.5 (1.1) t ¼ e0.10 46/1
.92 29
(41% 446 455
) 447 456
448 457
64/1 449 458
450 459
88
451 460
(59%
452 461
) 462
379 453
contributed both lithium and OMS blocks over follow-up and/or antimanic anticonvulsant) for more than 75% of 438
0.09 454
380
77 (a within-subject comparison, using participant as the unit follow-up weeks. Participants who took lithium over follow- 439
404 Lithium years per 3.2 (2.4) [0.5,10] — — 463
405 participant (SD) [min, — 464
406 465
max]
407 1.2 (1.8) [0, 8] 3.5 (2.9) [0.5, 12] — 466
OMS Years per
Participant —
(SD) [min, max]
408 467
409 Note: BD ¼ bipolar disorder; Brown ¼ Brown University; max ¼ maximum; min ¼ minimum; OMS ¼ other mood-stabilizing medication; UCLA ¼ 468
410 University of California Los Angeles; UMPC ¼ University of Pittsburgh Medical Center. 469
411 470

4 www.jaacap.org Journal of the American Academy of Child & Adolescent

Psychiatry Volume - / Number
- / - 2019
 LITHIUM IN YOUTHS WITH BIPOLAR DISORDER

TABLE 2 Time-Varying Characteristics of Youths During Lithium Versus Other Mood-Stabilizing Medication Blocks
471 530
472 Variable Lithium Blocks (n ¼ 886) OMS Blocks (n ¼ 1752) Statistic p 531
473 Age, y (SD) 17.4 (4.8) 16.9 (4.7) t ¼ e2.86 .004 532
474 Mean Follow-up Date (SD) 16 November 2007 (3.4 y) 29 August 2008 (3.5 y) t ¼ e4.66 533
475 BD Type <.0001 F [ 1.25 .29 534
476 BD I 722 (82%) 1,203 (69%) 535
477 BD II 50 (6%) 195 (11%) 536
478 BD NOS 113 (13%) 353 (20%) 537

479 538
480 Lifetime Hospitalizations (SD) 2.22 (2.98) 2.06 (3.34) t ¼ e1.61 .20
539
481 Lifetime DBD (%) 624 (71%) 1,394 (80%) t ¼ e2.65
540
482 .008
541
483 Lifetime Anxiety (%) 482 (54%) 1,207 (69%) t ¼ e4.68 542
484 <.0001 543
485 Lifetime Psychosis (%) 380 (43%) 508 (29%) t ¼ 1.95 .05 544
486
Lifetime Substance Abuse (%) 89 (10%) 127 (7%) t ¼ e1.15 545
.25 Current Medications (>75% of
follow-up period)
487 Non-lithium Mood- 574 (65%) 1,752 (100%) — 546
488 Stabilizing Medication — 547
(%)
489 Antipsychotics (%) 496 (56%) 1,295 (74%) t ¼ e5.30 548
490 <.0001 549
491 Antimanic Anticonvulsants (%) 114 (13%) 573 (33%) t ¼ e5.43 550
492 551
<.0001
493 552
Lamotrigine (%) 68 (8%) 390 (22%) t ¼ e5.61
<.0001
Antidepressants (%) 223 (25%) 527 (30%) t ¼ e2.61
.009
494 Stimulants (%) 203 (23%) 627 (36%) t ¼ e0.75 .45 553
495 Note: DBD ¼ disruptive behavioral disorder; NOS ¼ not otherwise specified; OMS ¼ other mood-stabilizing medication. 554
496 555
497 556
498 557
499
500 .003), assessed at an earlier follow-up date (p < .
(p ¼ on antidepressants (p¼ .009) (Table 2). The findings that 558
559
501 0001), less likely to have a history of disruptive individuals with lithium had an earlier intake date,
¼ 560
502 behavioral disorder (p .008) and anxiety (p < .0001), and earlier follow-up dates, is consistent with an 561
503 and more likely to have¼a history of psychosis (p . observed decrease in the prescribing rate of lithium over 562
504 05). During lithium blocks (versus OMS blocks), the course of follow-up (Figure S1, available online). Q10
563
509
505 568
510 participants were also less likely to be
Outcome Variable Lithium OMS OR/b (CI) p
564
569
506 565
511 Suicidal Ideation and Attempts 570
507
512 566
571
TABLE 3Attempt,
Suicide Relationship Between Lithium Versus Other
n (%) Mood-Stabilizing 63
17 (2%) Medication
(4%) Exposure and
0.51Outcome Variables, Adjusting
(0.28, 0.95) .03
508 567
513 for Lasso-Selected Characteristics 56 (6%) 171 (10%) 0.69 (0.45, 1.07) 572
514 Suicidal Ideation, n (%) .10 573
515 Mood Symptoms 574
516 Threshold Depression, n (%) 212 (24%) 497 (28%) 0.76 (0.55, 1.05) .10 575
517 Mean PSR Depression Score (SD) 2.0 (1.1) 2.3 (1.2) e0.19 (e0.32, e0.06) .004 576
518 Threshold Mania/Hypomania, n 164 (19%) 248 (14%) 1.35 (0.95, 1.91) .09 577
(%)
519 Mean PSR (Hypo)mania Score (SD) 1.9 (1.8) 1.9 (1.9) 0.03 (e0.04, 0.08) .47 578
520 Psychosocial Functioning 579
521 Worst PSF (lower is better) 10.2 (10.0) 11.1 (11.2) e0.55 (e0.90, e0.19) .003 580
522 Other Outcomes 581
523 Hospitalization, n (%) 71 (8%) 139 (8%) 1.03 (0.71, 1.51) 582
.87
524 583
525 Aggression score (BCS) 4.7 (4.7) 5.9 (4.7) e1.1 (e1.7, e0.5) .0004
Substance Use Disorder, n (%) 45 (5%) 105 (6%) 0.65 (0.33, 1.28) 584
526 .21 585
527 Note: BCS ¼ Behavior Control Scale; OMS ¼ other mood-stabilizing medication; OR ¼ odds ratio; PSF ¼ Psychosocial Functioning; PSR ¼ Psychiatric Q11
528 586
Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 5
Volume - / Number - / - 2019
 HAFEMAN et al.

TABLE 4 Relationship Between Lithium Versus Other Mood-Stabilizing Medication and Outcome Variables Where p ::; .10 in
589 648
3, Adjusted for Additional Lasso-Selected Time-Varying Covariates and in Youths <18 Years of Age
590 649
591 Symptom Baseline Covariates þ Time-Varying Covariates Youths <18 y of 650
592 651
Age
593 652
Suicide Attempts 0.51 (0.28, 0.95) 0.47 (0.25, 0.88) 0.54 (0.26, 1.1)
594 653
Suicidal Ideation 0.69 (0.45, 1.07) 0.66 (0.43, 1.01) 0.56 (0.32, 0.96)
595 654
596 Threshold Depression 0.76 (0.55, 1.05) 0.76 (0.55, 1.05) 0.64 (0.44, 0.93)
Mean PSR Depression e0.19 (e0.32, e0.06) e0.16 (e0.29, e0.03) e0.23 (e0.39,e0.07) 655
597 656
598 Threshold (Hypo)mania 1.35 (0.95, 1.91) 1.27 (0.90, 1.80) 1.31 (0.88, 1.94)
657
Worst PSF e0.55 (e0.90, e0.19) e0.46 (e0.81, e0.11) e0.72 (e1.13, e0.31)
Aggression score (BCS) e1.1 (e1.7, e0.5) e1.0 (e1.6, e0.38) e1.1 (e1.8, e0.5)
599 658
600 Note: BCS ¼ Behavior Control Scale; PSF ¼ Psychosocial Functioning; PSR ¼ Psychiatric Status Rating. 659
601 660
602 661
662
603 Primary Analysis: Lithium Versus OMS and Clinical Supplemental Analyses: Sensitivity Analyses Q8
604 663
605 Outcomes Adjustment for Specific Medications. Adjustment 664
606 Variables selected by the Lasso regression are listed in specific
for types of mood stabilizers, antidepressants, and total 665
607 Table S2 (available online). After adjusting for intake number of medication classes did not diminish the associ- 666

608 covariates and age, youths in the lithium versus OMS ations between lithium (versus OMS) and significantly 667
609 668
groups were half as likely to have a suicide attempt
¼ (p associated clinical outcomes; in addition, the effects of
610 669
611 .03) and had lower mean PSR-depression
¼ scores
¼ (p . other medications were generally not significant, and if 670
612 004), better psychosocial function (p .003), and less signifi- cant, were in the opposite direction from that of 671
613 aggression
¼ (p .0004) (Table 3). Adjusting for additional lithium (eg, associated with more depression, decreased 672
614 time- varying covariates did not alter the point functioning) (Table 5). 673
615 674
estimates or con- clusions (Table 4). Findings were
616 675
similar when only considering blocks in which the Lithium-Only Versus OMS-Only. This analysis
617 676
618 participants were less than 18 years old. Although the excluded blocks in which participants were exposed to 677
619 lithium relationship
inverse and suicide attempts
between was no longer significant, and OMS.
both lithiumComparing 256 lithium-only blocks 678
the (contrib-
620 inverse relationship between lithium and both suicidal uted by 60 participants) and 1,732 OMS-only 679
blocks
621 ideation and threshold depression became significant in 680
(contributed by 265 participants), findings generally
this remained significant. An exception was suicide attempts, 681
622
subgroup (Table 4).
623 682
624 683
625 684
626 TABLE 5 Adjustment for Other Specific Types of Mood-Stabilizing Medication, Antidepressants, and Total Number of 685
627 Medication Classes Q12 686

628 687
629 Medication Suicide Attempt Mean PSR Depression Worst PSFa Aggression (BCS) 688
630 Lithium 0.51 (0.28, 0.95)* e0.19 (0.07)** e0.55 (0.18)** e1.1 (0.32)*** 689
631 Lithium, lamotrigine 0.54 (0.29, 1.00) e0.19 (0.07)** e0.58 (0.18)** e1.1 (0.32)*** 690
632 1.4 (0.8, 2.5) e0.02 (0.07) e0.29 (0.19) 0.37 (0.33) 691
633 Lithium, atypical 0.54 (0.29, 1.00)* e0.17 (0.07)** e0.47 (0.18)* e1.0 (0.31)** 692
antipsychotics
634 1.4 (0.8, 2.5) 0.13 (0.06)* 0.54 (0.16)*** 0.66 (0.30) 693
635 Lithium, anticonvulsants 0.52 (0.28, 0.98)* e0.17 (0.07)** e0.49 (0.18)** e1.2 (0.32)*** 694
636 695
1.2 (.7, 2.1) 0.13 (0.07) 0.36 (0.19) e0.36 (0.35)
696
637 0.52 (0.28, 0.97)* e0.17 (0.07)** e0.53 (0.18)** e1.2 (0.31)*** 697
638 Lithium, antidepressants
639 1.4 (0.9, 2.4) 0.20 (0.06)*** 0.33 (0.16)* e0.58 (0.29)* 698
640 Lithium, number of 0.49 (0.27, 0.91)* e0.24 (0.07)*** e0.66 (0.18)*** e1.0 (0.32)** 699
medication
641 classesb 2.0 (1.2, 3.4)** 0.32 (0.05)*** 0.68 (0.14)*** e0.59 (0.23)* 700
642 701
643 Note: BCS ¼Interval
Longitudinal Behavioral ControlEvaluation
Follow-Up Scale; PSF ¼ Adolescent
Psychiatric Longitudinal
Status Ratings. Interval Follow-Up Evaluation Psychosocial Functioning; PSR ¼ Adolescent 702
644 a
Lower is better. 703
b FLA 5.6.0 DTD ■ JAAC2770_proof ■ 26 August 2019 ■ 3:37 pm ■ 704
645 Dichotomized at median of 2.1 current medication classes.
 LITHIUM IN YOUTHS WITH BIPOLAR DISORDER

TABLE 6 Clinical Predictors at Intake of Lithium Exposure and bipolar subtype, we found that no participant who
707 (Lithium, n ¼ 23; Other Mood-Stabilizing Medication, had BD-II during the first follow-up period was later 766
708 n ¼ 143), Adjusting for Demographic Predictors in started on lithium; thus, we limited the analysis to 767
709 (ie, Assessment Date, Age of Bipolar Disorder Onset)a participants with BD-I or BD-NOS (n 23, n 143). 768
710 769
Variable OR (CI) p Despite limited power¼ to detect associations, individuals
711 770
712 Suicidal Ideation 3.2 (1.1, 9.1) .03 who would later be prescribed lithium were 3 times more 771
713 Suicide Attempts 2.1 (0.2, 21.6) likely to have suicidal ideation during the first 6- 772
714 .53 month follow-up (p .03) (Table 6). Within this initially
¼ 773
715 Threshold Depression 1.2 (0.5, 3.0) 774
lithium-free subset, associ- ations found between
716 .70 775
Threshold (Hypo)mania 1.8 (0.7, 4.7)
lithium and clinical outcomes were similar in
717 776
718 .20 magnitude to those found in the entire sample, but 777
719 PSF Worst 1.1 (0.9, 1.3) were not significant because of limited power (Table S5, 778
720 .37 available online). An exception to this was depressive 779
721 Aggression (BCS) 1.0 (0.9, 1.1) symptoms, which was not inversely associated with 780
722 .70 lithium in this subgroup. 781
723 BD Type (BD I vs. BD NOS) 1.2 (0.5, 3.1) 782
.65 Site-by-Time Interactions. From Figure S1, available
Lifetime Hospitalizations 0.8 (0.5, 1.2) on-
.26
Lifetime DBD 0.5 (0.2, 1.5)
.22
Lifetime Anxiety 2.2 (0.8, 5.9)
.12
Lifetime Psychosis 0.9 (0.3, 2.8)
.83
724 Antidepressant 1.4 (0.5, 3.9) line, we noticed that, between the years of 2003 and 783
725 (during first follow- .4 784
726 7
2010, there was a significant decrease in the
up period) 785
727 percentage of par- ticipants on lithium versus OMS,
Stimulant 786
728 driven entirely by the UPMC site. This significant site- 787
(during first 1.0 (0.4, 2.1)
.9 by-time interaction (p ¼
follow-up
6
period)
729 757 Note: BCS Behavior Control Scale; BD bipolar disorder; BD NOS Predi
730 758 ¼ ¼ bipolar disorder not otherwise specified; DBD disruptive behavioral ctors
731 759 ¼ ¼ disorder; OR odds ratio; PSF Psychosocial Functioning. of
a
732 760 ¼ Substance abuse could not be assessed because of the low Lithi
733 761 frequency of this predictor at baseline (n ¼ 11).
um
734 762 Expo
763
735 who would later be exposed to lithium and 156 who lithium use (as compared to other, non-lithium mood-sta- 822
sure.
would
736
764
737 not. Testing the relationship between later starting bilizing medications) was associated with fewer suicide To
823
738 asse
which did not show a significant difference between
739 ss
lithium-only versus OMS-only; however, there were
740 the
741 only six attempts in the lithium-only group, thus
degr
742 limiting inference (Table S3, available online).
743
ee to
744 Bipolar Subtype. The effect of lithium versus OMS whic
745 was similar in youths with threshold (BD-I/II) versus h
746 sub- threshold BD (BD-NOS), with no significant asso
747 interactions (Table S4, available online). ciati
748 ons
749
750
Within-Subject Analysis may
751 A within-subject analysis, including only individuals who have
752 contributed both lithium and OMS blocks over follow- been
753 ¼ up (n 65), yielded similar effect sizes; however, several attri
754 as- sociations were not significant due to limited buta
755
power (Table S5, available online). ble
756
FLA 5.6.0 DTD ■ JAAC2770_proof ■ 26 August 2019 ■ 3:37 pm ■
to nfounding by indication), we tested whether our .006) presented the opportunity to test whether site- 788
(or outcome variables predicted later lithium exposure. specific temporal changes in lithium prescriptions would 789
ma Subgroup for this analysis included 23 individuals be asso- ciated with site-by-time interactions in the 790
sk clinical outcomes of interest. Indeed, both psychosocial 791
792
ed function and depressive symptoms showed a similar 793
by) (inverse) pattern 794
pr (interaction p values<.001), with UPMC worsening over 795
e- time relative to other sites (Figure S2 and Figure 796
exi S3, available online). The patterns for suicidal ideation 797
and 798
sti
799
ng attempt were similar, although interactions were not
800
cli sig- nificant (p values >.2; Figure S4 and Figure S5, 801
nic available online). There were no site-by-time 802
al interactions with the aggression outcome (Figure S6, 803
dif available online). 804
805
fer Episode Severity. Finally, we assessed whether 806
en lithium (versus OMS) during the 3 months prior to the 807
ce onset of a mood episode was associated with 808
s indicators of severity within the episode (Table S6, 809
bet 810
available online). Consistent with the primary analyses, 811
we we found that individuals taking lithium (versus OMS) 812
en had better psychosocial
¼ functioning during episodes (p 813
lit .006), and a less marked decline in 814
hi functioning during the episode (p < .0001). There was 815
u no significant effect on length of episode or 816
m 817
hospitalization. Use
818
ve of LI (versus OMS) was not significantly related to 819
rs suicidal ideation or attempts in this analysis, 820
us although, given the rarity of events, there was limited 821
O power to detect differences.
M
S DISCUSSION
(ie
In a longitudinal analysis of youths and emerging
,
adults with BD, after controlling for confounders, we
co
found that

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 7

Volume - / Number - / - 2019

FLA 5.6.0 DTD ■ JAAC2770_proof ■ 26 August 2019 ■ 3:37 pm ■

 HAFEMAN et al.

attempts, fewer subthreshold depression symptoms, better option for youths with BD. However, when considering
825 psychosocial function, and less parent-reported optimal treatment for BD in youths, these potential 884
826 aggression. All findings were of similar magnitude (and benefits should be weighed against the possible risks of 885
827 most were also significant) when only follow-up periods long-term lithium exposure (eg, effects on kidney and 886
828 before age 18 years were included in the analysis; in 887
thyroid func- tion) and the fact that it can be lethal
829 888
830 fact, in the subset of youths <18 years old, lithium was in overdose. 889
also associated with less This study has limitations that should be
832
831 threshold depression and less suicidal ideation. The finding when interpreting results. First, as discussed above,
considered 891
890
833 regarding suicide attempts was especially striking, as suicidal 892
834 ideation during the first follow-up period increased the 893
founding by indication can be problematic in an
835 894
836
likelihood of later taking lithium. Results were specific observa- tional study. We had the benefit of assessing 895
837 to lithium; other medication classes did not show these predictors of lithium exposure in a subset of the sample, 896
838 posi- tive effects. Comparing lithium-only versus OMS- and findings from this analysis showed that participants 897
839 only yielded similar effect sizes on most clinical later prescribed lithium had higher levels of clinical 898
840 measures, although numbers were insufficient to make symptomatology; thus, it is unlikely that such factors 899
841 900
an inference about suicide attempts in this sensitivity explained observed beneficial ef- fects of lithium,
842 901
843 analysis. Associa- tions between lithium and clinical although they could have masked or decreased 902
844 outcomes did further
inference was not differ in participants
supported with threshold
by supplemental site- observed
starting effects.
lithium. WeAlso, this was ajust
also conducted a subset within-
supplemental of the 903
by- versus subthreshold BD. Our sample; participants already on lithium at the beginning
845 time interactions, which indicated that site-dependent subject
of analysis
the study thathave
might showed
hadsimilar results,
different making between-
characteristics prior 904
846 changes in prescribing patterns (ie, decrease in lithium subject confounding a less likely explanation of observed 905
847 to
prescribing in one site) were mirrored by predicted changes effects. Of note, the observational nature of this study 906
848 907
in rates of depressive symptoms and psychosocial function. limits our ability to assess other effects of medication (eg, 908
849
the effect
850 Overall, findings are consistent with previous literature, of antidepressants or polypharmacy), as it is difficult to 909
tease

853 912
854 suicide attempts, and ameliorate aggression. 6,8,16,28 Inter-
only every 6 months, so temporally fine-grained recall
913
855 estingly, in our sample, lithium did not appear to be of medication dosage and symptomatology is 914
856 su- perior to OMS regarding (hypo)mania outcomes. limited. Although we assessed average weekly dose, 915
857 somewhat
This is inconsistent with the literature in adults, collect
we diddata
not regarding daily dose or blood levels; thus, it 916
which is
858 has generally found lithium to be even more protective possible that some individuals were not on therapeutic 917
for doses
859 manic episodes than for depressive episodes.2 It is of lithium or other mood-stabilizing medications. We 918
possible also
860 919
862 that our null finding reflects that the relative benefits of did not collect data about adherence to medication.
861 920
lithium for (hypo)manic symptoms might not be as clear Third, we did not collect data regarding side effects or 921
in tolerability
866 925
867 is also possible that confounding by indication up (using the KDRS); thus, some follow-up periods
926
868 might explain the lack of observed effect (comparing with suicidal ideation were likely misclassified. This 927
869 lithium versus OMS). Albeit not significant, youths measure- ment error would most likely have led to an 928
870 manic
with episode during the first follow-up period were
a hypo- of the negative association between lithium (versus
underestimate 929
almost OMS)
871 twice as likely to be prescribed lithium later in follow- and suicidal ideation. 930
up;
872 thus, this observation should be interpreted with Despite these limitations, the COBY study provides a 931
caution.
873
This observational study provides evidence for effec- unique opportunity to assess the effect of medications in 932
874 933
tiveness in this population of youths with BD, as opposed to a larger sample, with much longer follow-up, and 934
875
arguably
876 efficacy in a controlled setting (ie, RCT); although more generalizable to our patient population, than an RCT. 935
efficacy
881 940
882 these effects, these findings provide evidence that improving psychosocial function in youths and emerging 941
883 lithium should be considered as a viable and important adults. Future work should focus on (1) elucidating 942
treatment the
8 www.jaacap.org Journal of the American Academy of Child & Adolescent
Psychiatry Volume - / Number
- / - 2019
 LITHIUM IN YOUTHS WITH BIPOLAR DISORDER

2018;20:583-593.
mechanisms by which lithium improves outcomes in 12. Geller B, Luby JL, Joshi P, et al. A randomized controlled trial of risperidone, Dis
943 BD, so that novel therapeutics can be developed to target lithium, or divalproex sodium for initial treatment of bipolar I disorder, manic or
clos
mixed phase, in children and adolescents. Arch Gen Psychiatry. 2012;69:515- ure:
944 these mechanisms with fewer side effects, and (2) 528. Dr.
945 identifying who will best respond to lithium, to optimize
- 13. Vitiello B, Riddle MA, Yenokyan G, et al. Treatment moderators and Haf
946 predictors of outcome in the Treatment of Early Age Mania (TEAM) study. J Am em
the risk benefit ratio. Acad Child Adolesc Psychiatry. 2012;51:867-878. an
947 has
14. Kleindienst N, Engel R, Greil W. Which clinical factors predict response to
948 prophylactic rec
949 Accepted July 23, 2019.
lithium? A systematic review for bipolar disorders. Bipolar Disord. 2005;7:404-417. eiv
ed
950
Drs. Hafeman, Merranko, T.R. Goldstein, Diler, Ryan, Birmaher, gra
951 and Messrs. Merranko and Liao, and Ms. Gill are with the nts
952 University of Pittsburgh, PA. Dr. Rooks is with the University of fro
953 Rochester Medical Center, New York. Dr. B.I. Goldstein is with m
Sunnybrook Health Sciences Centre, University of Toronto, NIM
954 Ontario, Canada. Dr. Axelson is with Nationwide Children’s Hospital H
955 fornia,
and Los Angeles.
The Ohio Drs. Keller,
State University, Hunt, Weinstock,
Columbus, OH. Dr. and Yen,isand
Strober Ms. Hower are Institute of Justice. Dr. Yen has received research support from NIMH,
with and
956 University of Cali- the
Klin
957 gen
with Brown University, Providence, RI.
958 stei
959 This research was supported by National Institute of Mental n
Q13 Health (NIMH) grants MH59929 (PI Boris Birmaher, MD), Thir
960 MH59977 (PI Michael Strober, PhD), MH59691 (PI Martin Keller, d
961 MD, and Shirley Yen, PhD), and MH110421 (PI Danella Hafeman, Gen
962 MD, PhD). erat
ion
963 Dr. Rooks and Mr. Merranko served as the statistical experts for
Fou
964 this research.
nda
965 The authors thank the families who participated in the Course tion
and Outcome of Bipolar Youth (COBY) study, the COBY .
966
research team, and the faculty and staff of the Child and Dr.
967 Adolescent Bipolar Spectrum Services clinic at the University T.
968 of Pittsburgh. The authors also thank David Brent, MD, at the Gol
969 University of Pittsburgh, for helpful analytic suggestions. dst
ein
970 has
971 rec
972 eiv
REFERENCES ed
973 gra
1. Geddes JR, Burgess S, Hawton K, Jamison K, Goodwin GM. Long-term lithium
974 therapy for bipolar disorder: systematic review and meta-analysis of randomized nts
975 controlled trials. Am J Psychiatry. 2004;161:217-222. fro
m
976 2. Severus E, Taylor MJ, Sauer C, et al. Lithium for prevention of mood episodes in
NIM
977 bipolar disorders: systematic review and meta-analysis. Int J Bipolar Disord.
H,
2014;2:15.
978 3. Miura T, Noma H, Furukawa TA, et al. Comparative efficacy and tolerability of
the
Am
979 phar- macological treatments in the maintenance treatment of bipolar disorder: a eric
980 systematic review and network meta-analysis. Lancet Psychiatry. 2014;1:351- an
981 359. Fou
4. Hayes JF, Louise M, Kate W, GJ R, Michael K, OD PJ. Lithium vs. valproate nda
982 vs. olanzapine vs. quetiapine as maintenance monotherapy for bipolar disorder: a tion
983 population- based UK cohort study using electronic health records. World Psychiatry. for
984 2016;15:53-58. Sui
985 5. L€ahteenvuo M, Tanskanen A, Taipale H, et al. Real-world effectiveness of cid
pharmacologic treatments for the prevention of rehospitalization in a Finnish e
986 Pre
nationwide cohort of patients with bipolar disorder. JAMA Psychiatry.
987 2018;75:347-355. ven
988 6. Kessing LV, Bauer M, Nolen WA, Severus E, Goodwin GM, Geddes J. tion
(AF
989 Effectiveness of maintenance therapy of lithium vs other mood stabilizers in
SP),
monotherapy and in com- binations: a systematic review of evidence from
990 and
observational studies. Bipolar Disord. 2018; Feb 14. doi: 10.1111/bdi.12623.
991 [Epub ahead of print].
the
Brai
992 7. Cipriani A, Hawton K, Stockton S, Geddes JR. Lithium in the prevention of n
993 suicide in mood disorders: updated systematic review and meta-analysis. BMJ. and
994 2013;346. Beh
8. Song Jie, Sj€olander Arvid, Joas Erik, et al. Suicidal behavior during lithium and avi
995 valproate treatment: a within-individual 8-year prospective study of 50,000 patients or
996 with bipolar disorder. Am J Psychiatry. 2017;174:795-802. Fou
997 9. Khan A, Khan SR, Hobus J, et al. Differential pattern of response in mood symptoms nda
998 and suicide risk measures in severely ill depressed patients assigned to citalopram with tion
placebo or citalopram combined with lithium: role of lithium levels. J Psychiatr Res. and
999 2011;45:1489-1496. roy
1000 10. Amerio A, Ossola P, Scagnelli F, et al. Safety and efficacy of lithium in alti
1001 children and adolescents: a systematic review in bipolar illness. Eur Psychiatry. es
2018;54:85-97. fro
11. Duffy A, Heffer N, Goodday SM, et al. Efficacy and tolerability of lithium for the m
Guil
treat- ment of acute mania in children with bipolar disorder: a systematic review: a
ford
report from the ISBD-IGSLi joint task force on lithium treatment. Bipolar Disord.
Pre
s o n Canada, the Canadian Institutes for Health Research, the Heart Eur J Clin Invest. 2017;47:617-621.
s. f and Stroke Foundation, NIMH, the Sunnybrook Hospital
D Department of Psychiatry, and the University of Toronto
r. Department of Psychiatry. Dr. Axelson has received grants from
1002
D H NIMH and personal fees from Janssen Research and 1003
e
il
a
Development, LLC, and UpToDate. Dr. Strober has received 1004
e research support from NIMH and as the Resnick Endowed Chair
r l in Eating Disorders at the University of California, Los Angeles.
1005
h t Dr. Keller has received research support from NIMH. Dr. Hunt 1006
h
a has received honorarium from Wiley Publishers and has received 1007
s support from the NIMH. Dr. Weinstock has received funding from 1008
r ( the National Center for Complementary and Integrative Health
e N (NCCIH), NIMH, the NIH Office of Behavioral and Social 1009
c I Sciences Research, and the National 1010
ei H 1011
v ) and AFSP and has served as a consultant at Janssen
e . Global Services. Dr. Birmaher has received grants from 1012
d NIMH and royalties from Random House, UpToDate, and 1013
r Lippincott, Williams and Wilkins. Ms. Hower has received
e D funding from NIMH and honoraria from the US Department of
s r De- fense. Dr. Rooks, Mr. Merranko, Ms. Liao, and Ms. Gill
1015
e . report no biomedical financial interests or potential conflicts 1016
a of interest. 1017
rc
h
B Correspondence to Danella M. Hafeman, MD, PhD, 3811 1018
.
s O’Hara Street, Pittsburgh, PA 15213; e-mail: 1019
hafemand@upmc.edu
u 1020
p G
o
0890-8567/$36.00/ª2019 American Academy of Child and 1021
p Adolescent Psychiatry
o l 1022
rt d https://doi.org/10.1016/j.jaac.2019.06.013 1023
fr s
t
1024
o
m e 1025
N i 1026
I n 1027
M 15. Findling RL, Robb A, McNamara NK, et al. Lithium in the acute treatment of
H 1028
h bipolar i disorder: a double-blind, placebo-controlled study. Pediatrics.
. 2015;136:885-894. 1029
a
D 16. Malone RP, Delaney MA, Luebbert JF, Cater J, Campbell M. A double-blind 1030
s
r.
R
placebo- controlled study of lithium in hospitalized aggressive children and 1031
adolescents with conduct disorder. Arch Gen Psychiatry. 2000;57:649-654. 1032
y r
17. Geller B, Cooper TB, Sun K, et al. Double-blind and placebo-controlled study of
a e
lithium for adolescent bipolar disorders with secondary substance dependency. J Am
1033
n c
h Acad Child Adolesc Psychiatry. 1998;37:171-178. 1034
e
a i 18. Goldstein BI, Strober M, Axelson D, et al. Predictors of first-onset substance use 1035
s v dis- orders during the prospective course of bipolar spectrum disorders in 1036
r e adolescents. J Am Acad Child Adolesc Psychiatry. 2013;52:1026-1037.
1037
e d 19. Birmaher B, Axelson D, Goldstein B, et al. Four-year longitudinal course of children
c and adolescents with bipolar spectrum disorders: the Course and Outcome of Bipolar 1038
ei Youth (COBY) study. Am J Psychiatry. 2009;166:795-804. 1039
v r 20. Birmaher B, Axelson D, Strober M, et al. Clinical course of children and adolescents
e 1040
e with bipolar spectrum disorders. Arch Gen Psychiatry. 2006;63:175-183.
d s 1041
21. Axelson D, Birmaher B, Strober M, et al. Phenomenology of children and
g e 1042
a adolescents with bipolar spectrum disorders. Arch Gen Psychiatry.
r 2006;63:1139-1148. 1043
a r
n c 22. Kaufman J, Birmaher B, Brent D, et al. Schedule for Affective Disorders and 1044
h Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL): 1045
ts
initial reliability and validity data. J Am Acad Child Adolesc Psychiatry. 1997;36:
fr
980-988. 1046
o s
m 23. Keller MB, Lavori PW, Friedman B, et al. The Longitudinal Interval Follow-up 1047
u
t p
Eval- uation: a comprehensive method for assessing outcome in prospective 1048
h longitudinal studies. Arch Gen Psychiatry. 1987;44:540-548. 1049
p
e o 24. Hollingshead AB. Four-Factor Index of Social Status. New Haven, CT: Yale 1050
N r University, Department of Sociology; 1975.
a t 25. Kolko DJ, Baumann BL, Bukstein OG, Brown EJ. Internalizing symptoms and
1051
ti affective reactivity in relation to the severity of aggression in clinically referred, 1052
o behavior- disordered children. J Child Family Stud. 2007;16:745-759. 1053
n f 26. Goldstein TR, Ha W, Axelson DA, et al. Predictors of prospectively examined
al r 1054
suicide attempts among youth with bipolar disorder. Arch Gen Psychiatry.
In o
2012;69:1113-1122. 1055
st m
27. Schisterman EF, Cole SR, Platt RW. Overadjustment bias and unnecessary 1056
it
u
adjustment in epidemiologic studies. Epidemiology. 2009;20:488-495. 1057
B 28. Mu€ller-Oerlinghausen B, Lewitzka U. Lithium reduces pathological aggression
t
r
1058
and suicidality: a mini-review. Neuropsychobiology. 2010;62:43-49.
e 1059
a 29. James JE. Reviving Cochrane’s contribution to evidence-based medicine: bridging the
s
i gap between evidence of efficacy and evidence of effectiveness and cost-effectiveness. 1060

Journal of the American Academy of Child & Adolescent Psychiatry www.jaacap.org 9

Volume - / Number - / - 2019