Efficacy of Nonpharmacological Treatments on Comorbid Internalizing Symptoms of Adults With Attention-Deficit/Hyperactivity Disorder: A Meta-Analytic Review

Participants, Interventions, Comparators, Outcomes, and Studies Included

Systematic Review Protocol

In accomplishing this systematic review, no previous protocol was carried out.

Search Strategy

A systematic literature search was carried out in Scopus, PsycINFO and MEDLINE (via PubMed) without a date limit. The following terms were used in the title and abstract: adult*; ADHD OR attention deficit hyperactivity OR attention deficit disorder; psychosocial t* OR skills train* OR CBT OR cognitive behavio* OR DBT OR dialectical behavio* OR neurofeedback OR MBCT OR mindful*. In addition, reference lists of relevant articles found were screened. The last search was performed on October 20, 2018.

Study Selection and Data Extraction

First, duplicates were removed, and a screening search on the basis of title and abstract was performed by the first author. Then, the masked articles were selected independently by two researchers, with an excellent agreement (κ = 0.81), and solving inconsistencies by consensus.

Data were extracted independently and coded in a pilot-tested spreadsheet by two researchers, with an excellent agreement (κ = 0.88 for RCTs; κ = 0.98 for pretest–posttest studies). A consensus was reached in cases where there were disagreements. Coding covered all items of study characteristics described in the Cochrane Handbook (Higgins & Green, 2011). The codebook can be requested from the corresponding author.

Data Analysis

Comprehensive Meta Analysis (Version 3.3.070) and Review Manager (Version 5.3) software were used for data analysis. Raw data can be accessed in the following link: https://www.dropbox.com/sh/0la6u15y9jnq3c5/AAByTErq7aqsnGknXpgrEphsa?dl=0. Posttreatment and long-term reduction of comorbid anxiety and depression symptoms, problems with self-esteem and ED and improvement of QoL were considered the primary outcomes. The ES for within-subject (pretest to posttest and follow-up) single-group studies was defined as the average pretest-follow-up change, divided by the pretest standard deviation, and with a correction factor for small sample sizes (Morris, 2000). The ES index for between-group (nonpharmacological treatments vs. control groups) outcomes was computed as the difference between the average pretest–posttest and follow-up change of the experimental and control groups, divided by a pooled estimate of the pretest standard deviations of the two groups, also with a correction factor for small sample sizes (Morris, 2008). Sampling variances of the ESs were estimated following Morris’s proposal (Morris, 2000, 2008). In these calculations, the Pearson correlation coefficient between the pretest and posttest/follow-up scores must be known. As studies did not report the Pearson correlation coefficients, these were imputed following Rosenthal’s recommendation (Rosenthal, 1991) by using r = 0.7. ES values of 0.2, 0.5, and 0.8 were considered small, moderate, and large, respectively (Cohen, 1988).

ES estimates of the individual studies, weighted by their inverse variance, were synthesized for each outcome. A random-effect model was selected because of the expectation of a high heterogeneity between studies and it was assumed that the true effect size may vary from study to study (Borenstein, Hedges, Higgins, & Rothstein, 2010). A 95% confidence interval (CI) was computed. Interpreting the values of 25%, 50%, and 75% as reflecting low, moderate, and high heterogeneity, respectively; the I² index was used for the assessment of the ES consistency (Higgins, Thompson, Deeks, & Altman, 2003). This index describes the percentage of total variation across studies that is due to heterogeneity, rather than chance.

The first author, supervised by the second, assessed the internal validity of each study using The Cochrane Collaboration’s tool for assessing risk of bias (Higgins et al., 2011). Discrepancies were decided by consensus. Blinding of the therapists could not be assessed, as the studies used nonpharmacological treatments. To ascertain the influence of this variable on ES estimates, sensitivity analyses were performed, but studies were not excluded for presenting a high risk of bias evaluation.

We conducted several analyses to assess the publication bias. First, the funnel plots were examined visually. Second, the trim-and-fill method was performed (Duval & Tweedie, 2000), which trims the asymmetric results on the right side and fills the plot with their imputed counterparts on the left, estimating a new corrected ES. Third, the Egger test was carried out to test the asymmetry of the funnel plots (Egger, Smith, Schneider, & Minder, 1997). This test evaluates bias predicting the standardized effect (ES divided by the standard error) with the precision of each ES (the inverse of the standard error). Finally, the fail-safe N (Rosenthal, 1979), which is the number of additional studies with a null effect needed to make the meta-analysis nonsignificant (raising the p value above .05), was also calculated.

The impact on ES estimates of some clinically and demographically relevant variables, such as therapy type, therapy setting, percentage of participants under medication, mean age of participants, and percentage of male participants, was assessed. These variables were selected because some of them were assessed in previous meta-analytic reviews and because of their clinical significance. In addition, because of the suspected heterogeneity in the methodological quality of the included studies, some relevant variables for this issue, such as the risk of bias assessment, outcome measure source, control group type, and follow-up length, were included in the moderator analyses as well. These analyses were performed with chi-square tests for categorical variables and with meta-regressions for continuous variables. To increase the statistical power of these analyses, all therapies were merged for each outcome, following the procedures of previous meta-analytic reviews in the field (Knouse et al., 2017). Due to the importance of an adequate comparison group, additional sensitivity analyses were made within each nonpharmacological intervention, isolating the effect of those studies that compared the treatment with an active control group, where positive results were found.

Study Selection

From 1,054 records, 20 RCTs, three controlled pretest–posttest, and 12 uncontrolled single-group pretest–posttest studies were identified and included in the analyses (see Figure 1 and Tables 1 and 2). All studies included were published in peer review journals. A total of 587 studies were excluded after a screening on the basis of the title and abstract. Four controlled studies were excluded from between-group but included in within-subject meta-analyses, because they were not randomized (Bramham et al., 2009; Bueno et al., 2015; Wiggins et al., 1999), and because all groups received a specific therapy for ADHD (Cherkasova et al., 2016). Thus, finally, up to 18 comparisons were available for between-group outcomes, and up to 33 treatment groups were included in the within-subject meta-analyses. For the studies by Philipsen et al. (2015) and Schönenberg et al. (2017), each treatment group was compared with the equivalent control group, based on medication status and comparing NFB versus sham NFB, respectively. A total of 18 studies were excluded because they did not measure any of the outcomes considered in our review (Bachmann et al., 2018; Barth, Mayer, Strehl, Fallgatter, & Ehlis, 2017; Cowley, Holmström, Juurmaa, Kovarskis, & Krause, 2016; Hirvikoski, Waaler, Lindström, Bölte, & Jokinen, 2015; Janssen, de Vries, Hepark, & Speckens, 2017; Janssen et al., 2018; Kaiser, 1997; LaCount, Hartung, Shelton, Clapp, & Clapp, 2015; Schoenberg et al., 2014; Solanto, Marks, Mitchell, Wasserstein, & Kofman, 2008; Thompson & Thompson, 1998; Van der Oord et al., 2018; Virta, Hiltunen, Mattsson, & Kallio, 2015; Wang et al., 2016; Weiss et al., 2012; White & Shah, 2006; Zilverstand et al., 2017; Zwart & Kallemeyn, 2001); four because there were insufficient data in the article (Cole et al., 2016; Hirvikoski et al., 2011; Huang et al., 2019; Weiss & Hechtman, 2006); two for not including all participants who were at least 18 years old (Vidal-Estrada et al., 2015; Zylowska et al., 2008); two for including participants with severe substance abuse (Bihlar Muld, Jokinen, Bölte, & Hirvikoski, 2016) and personality disorders (Prada et al., 2015); and one because the treatment was self-applied (Moëll, Kollberg, Nasri, Lindefors, & Kaldo, 2015) or a formal diagnostic assessment for ADHD was not made (Wymbs & Molina, 2015).

OPEN IN VIEWER

Figure 1.

PRISMA 2009 flow diagram.

OPEN IN VIEWER

Table 1.

Summary of Characteristics of Included Randomized Controlled Trials.

	Sample (n) (treated/controls)			Participants (treated/controls)		Treatment				Outcomes (and measure source)
	Randomized	Posttreatment assessed	Follow-up assessed	% Males	% on ADHD meds	Therapy	Setting	Sessions	Control	Depression	Anxiety	Self-esteem	QoL	ED	Follow-up length (months)
Dittner, Hodsoll, Rimes, Russell, and Chalder (2018)	30/30	27/19	27/19	79.3/60	63.3/86.7	CBT	Individual	16	TAU	SR	SR	—	—	—	3
Emilsson et al. (2011)	27/27	17/17	8/13	63	100	CBT	Combined	15	TAU	SR	SR	—	—	SR	3
Fleming, McMahon, Moran, Peterson, and Dreessen (2015)	17/16	16/16	16/16	58.8/56.2	70.6/18.8	DBT	Combined	8	Wait-list	SR	SR	—	SR	—	3
Gu, Xu, and Zhu (2017)	30/26	28/26	28/26	57.1/53.8	71.4/72.9	MBT	Individual	6	Wait-list	SR	SR	—	—	—	3
Hepark et al. (2015)	55/48	41/42	—	62/46	60/54	MBT	Group	12	Wait-list	SR	SR	—	—	SR	—
Hoxhaj et al. (2018)	41/40	39/36	32/32	43.9/52.5	0	MBT	Group	12	Active	SR	SR	—	—	—	8
Mitchell et al. (2013)	11/11	11/9	—	45.5/33.3	54.5/55.6	MBT	Group	8	Wait-list	—	—	—	—	SR	—
Pettersson, Söderström, Edlund-Söderström, and Nilsson (2014)	14/18	13/15	11/0	42.9/22.2	42.9/50	CBT	Group	10	Wait-list	SR	SR	—	—	—	—
Philipsen et al. (2015)	107/110	91/95	69/70	51.5/50.5	100	DBT	Group	12	Active	SR	—	IE/SR	—	—	12
	109/107	86/80	59/45	54.7/43.7	0										12
Safren et al. (2005)	16/15	16/15	—	45.2	100	CBT	Individual	15	TAU	IE/SR	IE/SR	—	—	—	—
Salomone et al. (2015)	24/27	18/19	15/14	66.6/74.1	33.3/22.2	NFB	Individual	N.E.	Active	SR	SR	SR	—	—	3
Schönenberg et al. (2017)	38/39	35/36	34/36	51/61	16/11	NFB	Individual	30	Active	SR	SR	—	—	—
Solanto et al. (2010)	45/43	41/40	—	29/39	42.2/46.5	CBT	Group	12	Active	SR	IE	SR	—	—	—
Stern et al. (2016)	34/26	23/13	—	44 /42	38.2/38.5	CT	Individual	∼12 weeks	Active	—	—	—	SR	—	—
Stevenson, Whitmont, Bornholt, Livesey, and Stevenson (2002)	22/21	22/21	—	72.7/61.9	55.9	CBT	Group	8	Wait-list	—	—	SR	—	—	3
Vidal-Estrada et al. (2013)	15/17	12/16	—	60 /35.2	100	CBT	Group	11	Active	SR	SR	—	SR	—	—
Virta et al. (2010b)	10/10	10/10	—	30 /40	50/70	CBT	Individual	10	Wait-list	SR	—	—	SR	—	—
	9/10	9/10		77.7/40	55.5/70	CT		20
Virta et al., (2010a)	9/10	9/10		33.3/40	77.7/70	Hypn		10
S. Young et al. (2015)	48/47	35/39	27/32	37.5/32.9	100	CBT	Combined	15	TAU	SR	SR	—	SR	SR	3

Note. ADHD = attention-deficit/hyperactivity disorder; QoL = quality of life; ED = emotional dysregulation; CBT = cognitive-behavioral therapy; TAU = treatment as usual; DBT = dialectical behavior therapy; MBTs = mindfulness-based therapies; NFB = neurofeedback; CT = cognitive training; Hypn = hypnotherapy; SR = self-reported; IE = independent evaluator-reported.

OPEN IN VIEWER

Table 2.

Summary of Characteristics of Included Pre–Post Studies.

	Sample (n)			Participants			Treatment			Outcomes
	Pretreatment assessed	Posttreatment assessed	Follow-up assessed	Mean age	% Males	% on ADHD meds	Therapy	Setting	Sessions	Depression	Anxiety	Self-esteem	QoL	ED	Follow-up length (months)
Anastopolous and King (2015)	43	40	—	20.3	62.8	N/A	CBT	Combined	10+14	SR	SR	SR	—	—	—
Arns, Drinkenburg, and Leon Kenemans (2012)	21	12	—	29.95	61.9	42.85	NFB	Individual	33	SR	—	—	—	—	—
Bramham et al. (2009)	61	41	—	32.9	65.6	96.7	CBT	Group	4	SR	SR	SR	—	—	—
Bueno et al. (2015)	24 a	21 a	—	31.2	52	66.6	MBT	Group	8	SR	SR	—	SR	—	—
Cherkasova et al. (2016)	46	42	34	33.69	58.6	0	CBT	Combined	12	N/A	N/A	SR	—	—	6
	42	34	26	35.57	80.9	100
Hepark et al. (2015)	11	9	—	N/A	63.6	72.8	MBT	Group	10	SR	SR	SR	—	—	—
Hesslinger et al. (2002)	8	8	—	31.9	62.5	37.5	DBT	Group	13	SR	SR	—	—	—	—
Mayer, Wyckoff, Schulz, and Strehl (2012)	10	10	—	28.4	40	20	NFB	Individual	15	SR	SR	—	—	—	—
Mayer, Blume, Wyckoff, Brokmeier, and Strehl (2016)	25	24	21	34.29	51.7	N/A	NFB	Individual	30	SR	SR	—	—	—	6
Morgensterns, Alfredsson, and Hirvikoski (2016)	98	72	58	37.4	31.6	74.7	DBT	Group	14	SR	SR	—	SR	SR	3
Nasri et al. (2017)	18	18	18	39.7	55.5	72	CBT+DBT	Group	14	SR	SR	—	SR	—	6
Philipsen et al. (2007)	72	66	—	35.8	59.7	66.7	DBT	Group	13	SR	SR	—	—	—	—
Ramsay and Rostain (2011)	5	5	—	36.6	80	0	CBT	Individual	16	IE/SR	IE/SR	—	—	—	—
Rostain and Ramsay (2006)	74	43	—	30.8	74.4	100	CBT	Individual	16	IE/SR	IE/SR	—	—	—	—
Schönenberg et al. (2017) b	38	38	37	38.21	58	11	CBT	Group	12	SR	SR
Virta et al. (2008)	29	29	25	31	48.4	66	CBT	Group	10-11	SR	—	—	—	—	6
Wiggins, Singh, Getz, and Hutchins (1999)	9	9	—	42	11.11	N/A	PsyED	Group	4	—	—	SR	—	—	—

Note. ADHD = attention-deficit/hyperactivity disorder; QoL = quality of life; ED = emotional dysregulation; CBT = cognitive-behavioral therapy; NFB = neurofeedback; MBTs = mindfulness-based therapies; DBT = dialectical behavior therapy; SR = self-reported; IE = independent evaluator-reported.

a

In the meditation ADHD group.

b

In the CBT group.

Results for Each Therapy

CBT

CBT for ADHD studies obtained a small-to-moderate ES on depression symptoms after treatment, both compared with control groups (standardized mean difference [SMD] = 0.27; 95% CI = [0.02, 0.52]; I² = 52%) and moderate-to-large on within-subject outcome (SMD = 0.54; 95% CI = [0.35, 0.72]; I² = 72%; see Supplementary Table 2, Figure 2, and Supplementary Figure 1). This effect increased to moderate-to-large at follow-up assessment (SMD = 0.52; 95% CI = [0.24, 0.80]; I² = 0%, for between-group outcome; SMD = 0.72; 95% CI = [0.48, 0.96]; I² = 42%, for within-subject outcome). In the same way, effect on anxiety symptoms was small-to-moderate after treatment for both between-group (SMD = 0.31; 95% CI = [0.01, 0.62]); I² = 64%; see Supplementary Figure 2) and within-subject outcomes (SMD = 0.49; 95% CI = [0.25, 0.72]; I² = 82%), increasing to moderate-to-large at follow-up (SMD = 0.73; 95% CI = [0.45, 1.02]; I² = 0%, for between-group outcome; SMD = 0.74; 95% CI = [0.28, 1.21]; I² = 84%, for within-subject outcome). Sensitivity analyses showed that between-group ES became null when only active controlled studies were considered, both for posttreatment depression (SMD = −0.15; 95% CI = [−0.47, 0.16]; I² = 12%) and anxiety (SMD = 0.05; 95% CI = [−0.61, 0.72]).

OPEN IN VIEWER

Figure 2.

Summary of the results on the main outcomes for each therapy.

No significant effect was found on between-group self-esteem outcomes, but within-subject ES increased from small-to-moderate at posttreatment (SMD = 0.62; 95% CI = [0.31, 0.92]); I² = 80%) to large at follow-up (SMD = 1.404; 95% CI = [0.45, 1.64]; I² = 74%; see Supplementary Figure 3). Posttreatment within-subject ES on QoL was small-to-moderate (SMD = 0.28; 95% CI = [0.09, 0.48]; I² = 80%), which increased to moderate-to-large at follow-up (SMD = 0.57; 95% CI = [0.22, 0.92]), whereas a small-to-moderate effect was found when compared with TAU controls at follow-up (SMD = 0.39; 95% CI = [0.01, 0.78]; see Supplementary Figure 4). Likewise, in comparison with TAU control groups, treatment effect on ED increased from null up to moderate-to-large at follow-up (SMD = 0.64; 95% CI = [0.30, 0.98]; I² = 0%; see Supplementary Figure 5), whereas after treatment within-subject effect rose from small-to-moderate (SMD = 0.38; 95% CI = [0.16, 0.60]; I² = 0%) to moderate-to-large (SMD = 0.73; 95% CI = [0.42, 1.03]; I² = 80%) at follow-up.

Regarding publication bias, trim-and-fill method adjusted the effect on between-group posttreatment anxiety and depression, with a CI including zero after trimming one and two studies, respectively (adjusted SMD = 0.19; 95% CI = [−0.13, 0.53] for anxiety; adjusted SMD = 0.13; 95% CI = [−0.10, 0.37] for depression). On the contrary, Egger test was nonsignificant for these outcomes (bias = 1.83; 95% CI = [−3.35, 7.03]; p = .42 for depression; bias = −1.70; 95% CI = [−10.65, 7.18]; p = .63 for anxiety). No publication bias indicators were identified in the remaining outcomes.

Risk of Bias Assessment

All outcomes were rated with a high risk of bias (see Figure 3 and Supplementary Figures 6 to 9). Only one study in each of the following outcomes was rated with an unclear risk of bias: anxiety and depression symptoms (Safren et al., 2005), ED (Mitchell et al., 2013), and self-esteem (Philipsen et al., 2015).

OPEN IN VIEWER

Figure 3.

Risk of bias assessment summary for depression symptoms outcomes.

Moderator Analyses

ES on ADHD core symptoms

Results of the meta-regressions that could be computed for this covariate are shown in Table 3.

OPEN IN VIEWER

Table 3.

Meta-Regressions for ADHD Symptoms ES Covariate.

Outcome	Data	Posttreatment				Follow-up
		Coefficient	SE	95% CI	p value	Coefficient	SE	95% CI	p value
Depression symptoms	B-G all therapies	0.27	0.07	[0.13, 0.42]	<.01	0.31	0.13	[0.04, 0.57]	.02
	W-S all therapies	0.26	0.09	[0.09, 0.43]	<.01	0.31	0.17	[−0.30, 0.65]	.07
	B-G CBT	0.49	0.16	[0.17, 0.82]	<.01	Insufficient studies
	W-S CBT	0.28	0.14	[0.01, 0.56]	.04	0.22	0.26	[−0.30, 0.74]	.40
	W-S DBT	0.29	0.19	[−0.08, 0.66]	.12	0.18	0.50	[−0.80, 1.17]	.72
	W-S MBT	0.26	0.20	[−013, 0.66]	.19	Insufficient studies
	W-S NFB	0.49	0.41	[−0.31, 1.30]	.23	Insufficient studies
Anxiety symptoms	B-G all therapies	0.54	0.12	[0.30, 0.79]	<.01	0.57	0.19	[0.19, 0.95]	< .01
	W-S all therapies	0.31	0.14	[0.04, 0.58]	.02	0.60	0.25	[0.10, 1.09]	.02
	B-G CBT	0.59	0.19	[0.22, 0.96]	<.01	Insufficient studies
	W-S CBT	0.11	0.17	[−0.22, 0.45]	.50	0.49	0.49	[−0.47, 1.46]	.31
	W-S MBT	0.51	0.41	[−0.29, 1.32]	.21	Insufficient studies
Self-esteem	B-G all therapies	0.41	0.11	[0.20, 0.62]	<.01	Insufficient studies
	W-S all therapies	0.41	0.14	[0.13, 0.70]	<.01	−0.04	0.64	[−1.31, 1.22]	.95
	W-S CBT	0.32	0.29	[−0.16, 0.81]	.19	Insufficient studies
QoL	B-G all therapies	0.33	0.26	[−0.17, 0.84]	.20	Insufficient studies
	W-S all therapies	0.17	0.21	[−0.24, 0.58]	.40	Insufficient studies
	W-S CBT	−0.17	0.18	[−0.53, 0.20]	.37	Insufficient studies
ED	B-G all therapies	0.26	0.45	[−0.63, 1.16]	.56	Insufficient studies
	W-S all therapies	−0.11	0.21	[−0.55, 0.30]	.58	Insufficient studies

Note. ADHD = attention-deficit/hyperactivity disorder; CBT = cognitive-behavioral therapy; DBT = dialectical behavior therapy; MBTs = mindfulness-based therapies; NFB = neurofeedback; QoL = quality of life; ED = emotional dysregulation; B-G = between-group; W-S = within-subject. Significant results are boldfaced.

When all therapies were merged, higher ES on core symptoms significantly predicted positively the efficacy on depression (coefficient = 0.27, 95% CI = [0.13, 0.42]; p < .01 for between-group; coefficient = 0.31, 95% CI = [0.04, 0.57]; p < .01 for within-subject), anxiety (coefficient = 0.49, 95% CI = [0.17, 0.82]; p < .01 for between-group; coefficient = 0.31, 95% CI = [0.04, 0.58]; p = .02 for within-subject), and self-esteem (coefficient = 0.41, 95% CI = [0.20, 0.62]; p < .01 for between-group; coefficient = 0.41, 95% CI = [0.13, 0.70]; p < .01 for within-subject) at posttreatment. At follow-up, this covariate also positively predicted efficacy on between-group depression (coefficient = 0.31, 95% CI = [0.04, 0.57]; p < .01) and anxiety (coefficient = 0.57, 95% CI = [0.19, 0.95]; p < .01), and within-subject anxiety (coefficient = 0.60, 95% CI = [0.10, 1.09]; p = .02). Regressions were nonsignificant for QoL and ED outcomes.

For CBT, efficacy on between-group depression (coefficient = 0.49, 95% CI = [0.17, 0.82]; p < .01) and anxiety (coefficient = 0.59, 95% CI = [0.22, 0.96]; p < .01) was predicted positively by the ES on ADHD symptoms. For follow-up and within-subject outcomes, regressions were nonsignificant.

Effect on ADHD symptomatology was not a predictor for within-subject efficacy of MBT, DBT, and NFB on depression, nor for MBT on anxiety symptoms.

Findings on Nonpharmacological Intervention Efficacy on Comorbid Internalizing Symptoms and the Core Symptoms Improvement as Moderator

Our results indicated that CBT for ADHD improved comorbid depression and anxiety, both when compared with waiting list and TAU control groups and also considering within-subject change. These findings are consistent with previous adult ADHD meta-analyses (Jensen et al., 2016; Lopez et al., 2018), and also with those assessing the efficacy of CBT for internalizing disorders in adults without ADHD, with a similar effect size (Carpenter et al., 2018; Lepping et al., 2017). Effect size on comorbid internalizing symptoms was also similar to those reported on posttreatment core ADHD symptoms (Knouse et al., 2017), but lower than those obtained in the long term (López-Pinar et al., 2018). In addition, CBT for ADHD was significantly more effective than the other therapies in treating comorbid depression and anxiety, particularly in the long term. A possible explanation for these findings might be that CBT programs for adult ADHD include therapy ingredients that have been shown to be effective in the treatment of depression and anxiety disorders, such as cognitive therapy (Cuijpers, Cristea, Karyotaki, Reijnders, & Huibers, 2016) or problem-solving skills (Cuijpers, de Wit, Kleiboer, Karyotaki, & Ebert, 2018), so the treatment could be specific not only for core symptoms but also for comorbidity. On the contrary, results of the meta-regressions showed that ES on core symptoms significantly predicted the between-group efficacy on depression and anxiety. This might suggest that these symptoms were secondary to ADHD, and that CBT improvement of the inattention and hyperactivity/impulsivity symptoms led to the decrease in the associated symptomatology. CBT for ADHD also improved the other comorbid conditions, such as self-esteem (only when within-subject data were considered), and ED. This might be caused by the change to more functional beliefs about themselves, which is an essential therapeutic mechanism in CBT. This cognitive change could in turn improve the expectation of being able to deal with emotional situations (Hirsch et al., 2018), effectively enhancing emotional self-regulation (Bigman, Mauss, Gross, & Tamir, 2016). CBT for adult ADHD also improved QoL in comparison with TAU groups, especially in the long term. Contrary to expected, the reduction in ADHD symptoms did not lead to an improvement of QoL. Thus, this finding might indicate that CBT for ADHD was also specific with this variable. An interesting finding is that the effectiveness of CBT increased at follow-up in all outcomes, and the observed heterogeneity decreased. As suggested by CBT models of adult ADHD, the learning of adaptive coping strategies (both behavioral and cognitive) can reduce the functional impairment caused by neurobiological deficits (Safren et al., 2004). Because CBT efficacy on ADHD core symptoms has proved to be stable for at least a year (López-Pinar et al., 2018), the long-term integration of adaptive coping skills and the subsequent reduction of chronic functional impairment could also result in an improvement in comorbid internalizing conditions. This might have a greater impact as time goes by, once the psychological framework created by the therapy is well established.

Despite sharing several key therapeutic components with CBT (e.g., planning skills, cognitive strategies, among others), results showed that when compared with active control groups, DBT for ADHD only improved the posttreatment QoL, which was not sustained at follow-up. Regarding within-subject change, a small effect was found on self-esteem (only at follow-up) and comorbid depression, whereas the moderate-to-large posttreatment effect on QoL decreased to small at follow-up. These results are in line with those found in individuals with depression disorders without ADHD (Valentine, Bankoff, Poulin, Reidler, & Pantalone, 2015). Unexpectedly, no effect was observed on comorbid anxiety symptoms, despite the fact DBT skills predicted symptom improvement in the clinical population without ADHD (Webb, Beard, Kertz, Hsu, & Björgvinsson, 2016).

Some of the proposed MBT therapeutic working pathways are emotion regulation and somatic awareness (Hölzel et al., 2011). This seems a potential explanation for the improvement observed in posttreatment ED in our study, when these interventions were compared with waiting list control groups. Results also showed a within-subject improvement in this outcome, in addition to anxiety and depression, but gains were not sustained at follow-up. These improvements were not predicted by the treatment effect on ADHD core symptoms. However, no effect was found on depression and anxiety when MBT for ADHD were compared with control groups. Effect of MBT on internalizing comorbidity was inferior to that reported for ADHD core symptoms (Cairncross & Miller, 2016). These poor results were an unexpected finding, as mindfulness-based cognitive therapy (the intervention used by most studies in MBT subgroup) was initially developed for treating recurrent depression (Segal, Williams, & Teasdale, 2002), and MBTs have shown a small-to-moderate effect on anxiety and depression symptoms on the clinical and nonclinical population (Blanck et al., 2018). In contrast, these interventions were the most effective in improving posttreatment QoL when within-subject data were considered, although this finding was based only on one small uncontrolled trial (Bueno et al., 2015).

No effect was found on ADHD comorbid anxiety and depression when NFB was compared with control groups. These results are consistent with those obtained in individuals with a primary diagnosis of depression and anxiety disorders without ADHD (Begemann, Florisse, van Lutterveld, Kooyman, & Sommer, 2016) and with studies assessing the efficacy on ADHD symptoms (Schönenberg et al., 2017). On the contrary, regarding within-subject change, there were improvements in depression, anxiety and low self-esteem, which increased in the long term. Two potential NFB mechanisms were proposed by Mayer et al. (2016): First, intervention improves not only core ADHD symptoms but also comorbid symptoms, and second, reduction of core ADHD symptoms leads to improvements in comorbidity. The direction of the relationship is unknown and needs further investigation, although there are some studies preliminarily suggesting that NFB training could be a promising intervention for depression (Choi et al., 2010) and anxiety disorders (Mennella, Patron, & Palomba, 2017) in the clinical population, but presenting serious methodologic limitations (Begemann et al., 2016). Moreover, meta-regression results also supported the first hypothesis.

Results showed that CT, PsyEd, and hypnotherapy were not effective in treating comorbid internalizing symptoms in adults with ADHD. These results differ from those obtained in the clinical population without ADHD, in which some evidence of improvement of internalizing symptoms has been found after CT (Motter et al., 2016), PsyEd (Donker, Griffiths, Cuijpers, & Christensen, 2009), and hypnotherapy (Golden, 2012) interventions. However, more research is needed to evaluate the usefulness of these interventions, because these findings were based on a small number of trials presenting some methodological shortcomings. Thus, findings concerning these therapies need to be interpreted with caution with regard to adults with ADHD.

Results of meta-regressions (when all interventions were merged) supported the hypothesis that comorbid depression, anxiety, and self-esteem improvements are a secondary gain from the reduction of ADHD symptomatology. Nonetheless, it seems that the treatment is also specific for QoL and ED. These results could have important clinical implications regarding the sequencing of treatments. We know that not treating the symptoms of ADHD is a risk factor for the development of internalizing comorbid problems (Biederman et al., 2011), and our results support that an effective treatment of ADHD in adults could prevent the development of some of these problems and, therefore, alleviate the impairment. Similarly, because the presence of ADHD symptoms is often overlooked in the treatment of anxiety and depression patients, these symptoms should also be considered to improve the evolution of this population (Barkley & Brown, 2008).

Findings of the Moderator Analyses

Given the importance of having an appropriate comparison group, a possible limitation of the findings is that when CBT studies using an active control group were isolated, ES became nonsignificant for depression and anxiety posttreatment outcomes. This might be an indicator of the nonspecificity of CBT for improving these comorbid symptoms, but it has to be interpreted with caution, because it is based on only two studies with different placebo groups (Solanto et al., 2010; Vidal-Estrada et al., 2013). Another possible explanation is, as some authors have suggested, that placebo psychotherapy groups are not fully “inert” (Wampold et al., 2016). When all therapies were merged, therapy efficacy on depression and self-esteem after treatment was significantly lower in comparison with active control groups, which is in line with previous adult ADHD meta-analyses (Knouse et al., 2017; Lopez et al., 2018; López-Pinar et al., 2018), which could mean that treatments were not specifically effective for those outcomes. On the contrary, results supported the specificity of the treatment in improving these outcomes in the long term, and, in addition, in the rest of posttreatment outcomes.

Another clinically relevant finding was that treatment given in an individual setting was more effective on self-esteem, depression, and anxiety comorbid symptoms. This is consistent with the literature about treatment of anxiety and depression disorders (Carpenter et al., 2018). A possible explanation for this is that the ability to fit the specific needs of each case of individual therapy might be more specifically effective for comorbid internalizing symptoms than the support of a group. In addition, the alliance, which is positively related to the outcome of a psychological treatment (Horvath, Del Re, Flückiger, & Symonds, 2011), is significantly stronger in individual relative to group CBT (Mörtberg, 2014).

One important result supporting the robustness of the findings was that no significant difference was found between self-reported and independent-assessed measures in the vast majority of outcomes. This finding is consistent with those reported by previous meta-analytic reviews (Jensen et al., 2016). However, this result must be interpreted with caution because of the small number of studies which included independent-assessed measures.

Another significant result supporting the stability of the gains on depression comorbid symptoms was that follow-up length did not predict the effect on these outcomes. Notwithstanding, the results suggest that the effect on anxiety faded as time passes, differing from those obtained by CBT for anxiety disorders, which reported improvements that were maintained up to 1 year at follow-up (DiMauro, Domingues, Fernandez, & Tolin, 2013).

A higher proportion of males and younger participants predicted better results in some of anxiety, self-esteem, and QoL outcomes, as was suggested by previous literature (Edvinsson et al., 2013; Müller et al., 2017; Piñeiro-Dieguez et al., 2016; Solberg et al., 2018). These findings might be explained because anxiety disorders (Bandelow & Michaelis, 2015) and low self-esteem (Bleidorn et al., 2016) are more prevalent in women and because of the increase with age of the comorbid internalizing problems severity in adults with ADHD (Bramham et al., 2012). However, these results are not consistent with other studies that have found that older people tend to be more satisfied with their lives (Mercier, Péladeau, & Tempier, 1998) and have a higher self-esteem (Bleidorn et al., 2016).

As expected, a higher percentage of participants on ADHD medication predicted better results in depression, anxiety, and self-esteem, which supports the effectiveness of pharmacotherapy on the comorbid internalizing symptoms and is consistent with previous findings in adults with ADHD (Lopez et al., 2018). In addition, shorter treatments predicted worst results in QoL and longer interventions predicted better results in depression after treatment, but not at follow-up. This is consistent with the good-enough-level model about treatment duration. This model states that progressively less gains become acceptable as treatment takes longer, and then longer duration predicts poorer outcomes (Barkham et al., 2006).

Finally, another important finding was that publication bias indicators were found only in posttreatment CBT anxiety and depression outcomes, disappearing at follow-up. The rest of outcomes for this and the other interventions were free from publication bias, which supports the validity of the results and is consistent with previous meta-analyses in the field (Lopez et al., 2018).

Limitations and Future Research

There are some limitations to the study that should be considered when interpreting the results. First, a key limitation of this review is the small number of studies evaluating some interventions (e.g., PsyEd, CT, and hypnotherapy), so the generalizability of the results concerning these psychotherapies is subject to certain limitations and more research about them is needed to ascertain their efficacy on comorbid internalizing symptoms. Second, differences in the designs of the studies and sample size might be responsible for differential efficacy and, therefore, limit the direct comparison between the included interventions. Third, the high risk of bias on most outcomes, mainly due to the use of self-reported and unblinded measures alone, might threaten the validity of results. Fourth, the small number of studies in some subgroups could limit the statistical power of their moderator analyses. Fifth, the merger of all therapy modalities to carry out the moderator analysis may limit the findings regarding how these variables could influence the effect for each of the interventions. Sixth, the small number of studies comparing the treatment with active control groups might limit findings regarding the specificity of the psychotherapies assessed. Seventh, the inclusion of uncontrolled studies could raise issues of internal validity, as this hinders the control of other variables that could influence the observed effect. Finally, the high heterogeneity observed in some of the posttreatment outcomes might limit the generalization of the findings.

To improve the validity of the results, future research should address the lack of independent and blinded assessment of outcome measures, the use of larger samples and comparison groups (preferably an active control), the inclusion of a follow-up assessment, and the assessment of the attrition, including intent-to-treat analyses.

Moreover, as DBT seems a promising intervention for ED in the clinical population (Neacsiu, Eberle, Kramer, Wiesmann, & Linehan, 2014), it could be relevant to study its effect on this variable in adults with ADHD. In addition, further research should be undertaken to investigate the potential benefits of adding well-established therapy components such as behavioral activation (Ekers et al., 2014) to the treatment of comorbid depression in adults with ADHD.