For Which Children with ADHD is TBR Neurofeedback Effective? Comorbidity as a Moderator

We examined psychiatric comorbidities moderation of a 2-site double-blind randomized clinical trial of theta/beta-ratio (TBR) neurofeedback (NF) for attention deficit hyperactivity disorder (ADHD). Seven-to-ten-year-olds with ADHD received either NF (n = 84) or Control (n = 58) for 38 treatments. Outcome was change in parent-/teacher-rated inattention from baseline to end-of-treatment (acute effect), and 13-month-follow-up. Seventy percent had at least one comorbidity: oppositional defiant disorder (ODD) (50%), specific phobias (27%), generalized anxiety (23%), separation anxiety (16%). Comorbidities were grouped into anxiety alone (20%), ODD alone (23%), neither (30%), or both (27%). Comorbidity (p = 0.043) moderated acute effect; those with anxiety-alone responded better to Control than to TBR NF (d = − 0.79, CI − 1.55– − 0.04), and the other groups showed a slightly better response to TBR NF than to Control (d = 0.22 ~ 0.31, CI − 0.3–0.98). At 13-months, ODD-alone group responded better to NF than Control (d = 0.74, CI 0.05–1.43). TBR NF is not indicated for ADHD with comorbid anxiety but may benefit ADHD with ODD. Clinical Trials Identifier: NCT02251743, date of registration: 09/17/2014

. Common comorbidities included a quarter with learning disorders, half with disruptive behavior disorder (DBD) such as oppositional defiant disorder (ODD) and conduct disorder (CD), a third with anxiety (Reale et al., 2017;Sonnack & Brenneman, 2014, MTA Cooperative Group, 1999a. Thus, comorbidity is highly prevalent in ADHD and adds to impairment. Growing concern about health care expenses has elicited research looking specifically into the economic burden of ADHD. The 2011 Medical Expenditure Survey determined that children 3-17 years old with ADHD spent 58.4% more on medical care than their non-ADHD counterparts (Gupte-Singh et al., 2017). In another study, 14-17-year-olds with ADHD had direct and indirect costs five times those of controls ($15,036 vs. $2,848) over the course of development (Zhao et al., 2019). Given the economic burden of ADHD treatments, there is a need for the benefit of emerging treatments to outweigh the costs. Precision medicine offers personalized treatments based on an individual's unique physiological, genetic, and other biological factors (Keizer, 2019), thus limiting expenditures to treatments likely to benefit a given individual.
The most widely used treatments for children with ADHD are medication management and behavioral therapy. For 6-11-year-olds with ADHD, about 2/3 have tried medication and half have received behavioral treatment (Danielson et al., 2017). Among children receiving methylphenidate for ADHD, around 42% do not achieve remission, and even combined with behavioral treatment 32% do not . A possible impediment to ADHD treatment effectiveness is the presence of comorbidities.
Certain treatments may be more or less effective in the presence of certain comorbidities. In a double-blind placebo-controlled trial of methylphenidate for 43 children with ADHD, children with anxiety did not respond as well to methylphenidate as youth without anxiety (Pliszka, 1989). In contrast, in the same study, youth with comorbid ODD or CD responded as well to methylphenidate as youth without either comorbidity (Pliszka, 1989). The Multimodal Treatment Study of Children with ADHD (MTA), comparing four treatments (medication, behavioral treatment, the combination, and community comparison treatment as usual), found that children with combined-type ADHD and anxiety responded better to behavioral treatment than those without anxiety (MTA Cooperative Group, 1999b). In another MTA analysis, those with combined-type ADHD and anxiety but not ODD/CD were likely to respond equally well to the MTA medication and behavioral treatments while those with ADHD-only or ADHD with ODD/CD responded best to MTA medication (with or without behavioral treatment) and those with both comorbidities (anxiety and ODD/ CD) responding optimally to combined treatment . Another study on community-based treatments found that children with ADHD and a mental health comorbidity, particularly anxiety and depression, responded less effectively to ADHD medication (Al Ghriwati et al., 2017). In a randomized controlled trial of a collaborative school-home-based social skills training versus treatment as usual, teacher-rated anxiety comorbidity was associated with greater improvement for children receiving the social skills training (Morgan et al., 2020). While it is widely documented that mental health comorbidity is associated with social and educational outcomes for children with ADHD, and impacts medication treatment outcomes (Al Ghriwati et al., 2017;Jensen et al., 2001;Pliszka, 1989;Roy et al., 2016;the MTA Cooperative Group, 1999b), less is known about how comorbidities are associated with non-medication treatment response. More information is needed about effects of comorbidity on treatment outcomes for children with ADHD in order to make the best clinical decisions for each child.
An increasing number of parents, physicians, and psychologists are calling for additional evidence-based treatments, particularly for patients with comorbidities, those experiencing side effects, and those seeking a long-term solution to their symptoms (Pellow et al., 2011). While FDA-approved medications have been proven effective for up to 2 years (MTA Cooperative Group, 2004), side-effects can lead to termination of the medication (Fredriksen et al., 2014), highlighting the need for alternatives to pharmacotherapy in these patients. Charach (2020), citing the growthsuppression side effect, called for evidence-based nonpharmacologic treatments. Psychiatric concurrent diagnoses with ADHD, such as anxiety, complicate the diagnosis, treatment, and presentation of ADHD regarding inhibitory dysfunction, working memory, and sleep problems, resulting in recommendation for adjunctive psychosocial therapy for these patients (D'Agati et al., 2019). In the MTA long-term follow-up, baseline comorbid anxiety was a strong predictor of ADHD persistence into adulthood (Arnold et al., 2019). In a meta-analysis of 17 longitudinal studies of childhood ADHD persistence into adulthood, major depressive disorder and CD were associated with higher likelihood of persistence (Caye et al., 2016). A Swedish study examined newly diagnosed adults and found 49% to have a comorbid diagnosis (Ahnemark et al., 2018). There appears to be a need for personalized, evidence-based therapies that can be used in the half of patients with two or more diagnoses.
An alternative treatment for ADHD is neurofeedback (NF), which applies operant conditioning principles to brain electrical activity as quantified using an electroencephalogram (EEG). EEG is a potential biomarker allowing a precision medicine approach to ADHD (Olbrich et al., 2015). In 2016, 1 out of 10 children with ADHD have received neurofeedback (Danielson et al., 2017). The research on the 1 3 efficacy of neurofeedback for ADHD has varied in its scientific rigor (Arns et al., 2014) and has mixed results from previous studies and meta-analyses (Cortese et al., 2018;Van Doren et al., 2019). The International Collaborative ADHD Neurofeedback (ICAN) Study was a 2-site, double-blind, randomized clinical trial (Clinical Trials Registration: Clini-calTrials.gov, Identifier: NCT02251743) to compare active neurofeedback to control treatment of identical appearance, frequency, and duration; the only difference between treatments was deliberate reinforcement of lower theta-beta ratio. The primary outcome (parent-and teacher-rated inattention) did not show a significant difference between treatments at the end of treatment (The Collaborative Neurofeedback Group, 2020). There were clinically meaningful effects as evidenced by remission rates and lower medication usage for neurofeedback than the control group at 13-month follow-up (The Collaborative Neurofeedback Group, 2020). It is possible that two treatments showing no overall difference in a heterogeneous sample may contain identifiable subsamples defined by baseline factors (moderators) that respond well to one of the treatments. We here examined whether comorbid mental health diagnoses (i.e., anxiety or ODD) moderated treatment effects, i.e., identified subpopulations of children with ADHD in which the two treatments (NF versus Control) had different effects.

Participants
This study has been approved by the appropriate ethics committee (i.e., the Institutional Review Boards at both sites where data was collected) and therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. 142 children aged 7-10 who met DSM-5 criteria for ADHD inattentive or combined presentation on structured clinical interview, had item mean ≥ 1.5 standard deviation above norms on both parent-and teacher-rated inattention while off medication, a Theta-Beta Ratio (TBR) ≥ 4.5, and IQ ≥ 80 participated in a double-blind randomized trial of Neurofeedback (NF) vs. Control (see Fig. 1). For full exclusion criteria see The Collaborative Neurofeedback Group (The Collaborative Neurofeedback Group, 2013, 2020. At each of two sites, participants were randomized to NF vs. Control treatment of identical appearance, intensity, and duration in a 3:2 ratio, stratified on site and on current ADHD medication. The two treatments differed only in deliberate training down of TBR based on the child's own EEG in the active NF vs based on a pre-recorded EEG from another child's neurofeedback session in the Control. Treatment adherence was excellent. Specifically, 137 (93%) of the 142 participants completed the protocol-prescribed number of treatments (The Collaborative Neurofeedback Group, 2020). Within-session treatment fidelity was 98.7% as rated by trainers and 93.2% rated by a recognized expert in neurofeedback who was blinded to treatment condition (The Collaborative Neurofeedback Group, 2020).

Measures
The following measures were given as part of the assessment battery.

Comorbidity Assessment
Children's Interview for Psychiatric Syndromes -child (ChIPS) and -parent (P-ChIPS), a standardized diagnostic interview, was used to assess DSM ADHD and comorbidities (Weller et al., 1999a and b). ChIPS and P-ChIPS are structured interviews with 15 sections that screen for 20 DSM-IV psychiatric diagnoses. ChIPS and P-ChIPS have good validity and reliability (Weller et al., 2000;Younget al., 2016). Comorbid diagnoses were given if either the child or parent endorsed disorder criteria. These were classified into four groups: anxiety only, ODD only, both anxiety and ODD, and neither comorbidity.

Outcome Assessment
Conners-3rd edition: Parent and Teacher Report Long Version (C-3:P/T; Conners, 2008), was used to assess DSM ADHD symptoms. C-3:P/T was completed at screen; baseline; mid-treatment, end of treatment, and follow-up at 13 months. Parents and teachers rated the severity of child ADHD symptoms on a 4-point scale. C-3:P/T has good internal consistency and high test-retest reliability (Conners et al., 1998). Parent and teacher inattention ratings were averaged for the primary outcome. The specific outcome measures here considered were the change from baseline to end of treatment, and the change from end of treatment to 13-month follow-up.
Parents also completed a demographics form at screen with sections on child and family information and school information, including information about gender, race, ethnicity, caregiver employment information, annual household income, and children's sleep.

Ethical Considerations
Consent and assent were obtained using IRB-approved procedures and documents.

Procedures
This study utilizes data from the screen/baseline, endpoint, and 13-month follow-up visits. During the screening visit, participants underwent a physical examination and clinical tests (e.g., intellectual ability, clinical interview of mental health history with parent and child separately). Parents also completed questionnaires about medical history, psychological health history, demographics, and ADHD symptoms. After passing screen and completing the baseline assessment, participants were randomized into the NF or Control conditions. All participants did 19 sessions of NF, and then completed a mid-point evaluation visit, done blinded to treatment. Those who did not show a 10% improvement in the average of parent and teacher ratings on the C-3 at the midpoint assessment exited their study treatment and were treated clinically (The Collaborative Neurofeedback Group, 2013), but were retained for assessments. Those who showed a 10% improvement (n = 114) in the average C-3:P/T continued for another 19 sessions. Regardless of whether they continued in treatment, all participants had an end-point visit (still blinded), with baseline assessments repeated. Fig. 1 Consort Chart. Total of 144 eligible subjects were randomized to receive Neurofeedback treatment or Placebo in a 3:2 ratio. Two subjects (one from each group) withdrew consent, at sessions 3 and 9 for logistical reasons, without any assessment of outcomes after baseline. Thus 142 subjects were included in the primary analysis (modi-fied ITT population): 58 received control treatment (41%) and 84 active NF; 28 had 19 sessions and exited per protocol; 10 had 20-37 sessions; and 104 completed all sessions. All 142 subjects were expected to participate in the evaluation at treatment end (session 40) and all Follow-ups

Statistical Analysis
The comorbid disorder profile of each participant was classified as none, anxiety only, ODD only, or both, based on the disorder diagnosis from the ChIPS and P-ChIPS. Chi-square analyses were used to examine whether there were baseline differences between the four groups. The primary outcome was the change in DSM-5 parent-and teacher-rated Conners (C-3:P/T) inattentive symptom severity (AN) in the trajectory between baseline and end of treatment (acute), and between end of treatment and 13-month follow-up (delayed). Linear mixed model for repeated measures was used to estimate the inattention changes over time for subjects who received NF or Control treatment while accounting for the association of the measures at different time points from the same informant (source) for each subject. The comorbidity and its interaction (comorbidity x treatment x time) were included in the model as fixed effects along with the treatment and treatment x time interaction. Data analyses were conducted using SAS v9.4 (SAS, 2014). Sensitivity analysis for missing data was also carried out without data imputation, assuming missing at random.

Results
Of the total sample, 79 (55.63%) had combined presentation and 48 (33.80%) had inattentive ADHD presentation at baseline.
Overall, 70% of participants had at least one comorbid diagnosis. Table 1 shows the comorbidity results from the combined ChIPS and P-ChIPS. The most prominent comorbidity was ODD (50%), followed by specific phobias (26.8%), generalized anxiety disorder (22.5%), and separation anxiety (16.2%). These comorbidities were then grouped into anxiety only, ODD only, neither, or both. Table 1 also shows the percentages for each comorbidity category broken down by randomized groups. Table 2 provides demographic comparisons by comorbidity.
There is a significant interaction of comorbidity X treatment X time (p = 0.043). See Table 3. Figure 2 demonstrates the treatment effect (NF vs Control) for the four comorbidity groups at treatment end and 13-month follow-up. The acute effects were different between the anxiety only group and the other three groups pooled (p = 0.012). At the 13-month follow-up, the ODD-alone group had an effect size of d = 0.74 favoring NF, likely due to the Controls regressing almost to  Children with anxiety only had significantly worse inattention symptoms from baseline to treatment end with NF than with control treatment, compared to both, neither, and ODD-only groups, which had nonsignificantly better attention with NF. By 13 months follow-up, youth with comorbid ODD only had significantly better improvement in inattention symptoms with NF than with control treatment, while treatment differences for youth with both, neither, or anxiety-only comorbidities were not significant. There were no statistically significant differences between treatment end and 13-month follow-up.*p < 0.05 Because children with comorbid Anxiety alone fared so much worse with NF than Control treatment, we wondered how excluding them from the analyses would change the outcome for the remaining children. We conducted the same analysis as for our primary hypothesis (The Collaborative Neurofeedback Group, 2020) after excluding the 28 with comorbid Anxiety alone. For the remaining 114 children, there were effect sizes slightly favoring NF over control of d = 0.21, 95% CI (− 0.1, 0.52) at treatment end and d = 0.31, 95% CI: (-0.07, 0.69) at 13-month follow-up.

Discussion
Similar to previous research (Reale et al., 2017;Sonnack & Brenneman, 2014;Ahnemark et al., 2018;MTA Cooperative Group, 1999a), 70% of youth in this study had at least one comorbid mental health diagnosis. In line with previous research examining whether comorbidity moderates ADHD treatments (i.e., methylphenidate and behavioral therapy; Pliszka, 1989;Jensen, 2001; The MTA Cooperative Group, 1999b), we also found differences in treatment response to neurofeedback versus control treatment. Specifically, we found that comorbid anxiety and ODD moderated treatment outcome for children with a primary rigorous diagnosis of ADHD who participated in a double-blind randomized controlled trial of neurofeedback.
This study demonstrated that comorbidity moderates improvement of children's inattentive symptoms with neurofeedback vs. control treatment. Specifically, youth with anxiety comorbidity alone had less improvement of inattentive symptoms with neurofeedback than control treatment at both treatment end and at 13-month follow-up. In contrast, children with ODD had more inattentive improvement with neurofeedback than control treatment, especially at 13-month follow-up. It should be noted that the confidence intervals were quite wide and the sample size of the anxiety only youth was small. Thus, tentatively, it appears that the neurofeedback delivered in this TBR protocol was contraindicated for the limited number of youth with only comorbid anxiety and ADHD, although it appears beneficial for youth with ADHD and ODD in this study. More research with larger sample sizes of comorbid anxiety only and ODD only is needed to draw a more conclusive finding for the impact of comorbidity on TBR neurofeedback.
TBR neurofeedback was significantly worse than control treatment response in the presence of anxiety alone, making inattention symptoms, the primary treatment focus, show less improvement than they might have without the neurofeedback in this study. This is clearly seen by examination of Fig. 1 where there is mild but incomplete "recovery" after treatment end at 13-month follow-up for children with comorbid anxiety symptoms when they were no longer exposed to NF. ICAN carefully attended to participant safety with pharma-like AE tracking (The Collaborative Neurofeedback Group, 2020). This is another important safety finding with implications for clinical guidance, particularly if upheld in other studies.
Excluding youth with comorbid anxiety yielded a treatment difference slightly favoring neurofeedback of d = 0.21 at treatment end and d = 0.31 at 13-month follow-up for the remaining 114 children (d = 0.74 for ODD alone at 13-month follow-up). This analysis was exploratory, but there appears to be a delayed benefit for those with ODD comorbidity.
Several studies offer neurofeedback protocols for anxiety (Kerson et al., 2009;Martins-Mourao & Kerson, 2016;Nusslock et al., 2018), and none indicate TBR NF training. This is likely because the brain areas and functional patterns (e.g., alpha, sensory-motor rhythm and high beta waves; Marzbani et al., 2016;Thompson & Thompson, 2006) that are typically targeted in anxiety are incongruent with the TBR protocol. Research to date on EEG differences suggests higher slow wave/fast wave ratios in ADHD are not typical in anxiety (Putman et al., 2010). ADHD-like behaviors can be symptoms of anxiety and when anxiety is the primary diagnosis, it is better to attend to its underlying brain patterns. However, we thought we had prevented this complication by requiring a TBR > 4.5 for entry to ICAN. It should  Fig. 2 Positive Inattention Treatment Effect Suggests More Improvement with NF than Control. Change in treatment-effect difference between neurofeedback and control treatment on composite parentand teacher-rated inattention mean item score by comorbid disorder type. Moderator interaction significant at p = 0.04. Positive difference favors neurofeedback; negative difference (lower line) favors control treatment. NF Neurofeedback. ODD oppositional-defiant disorder. Both both ODD and anxiety. FU Follow-up. Children with ADHD and anxiety alone (bottom line) did significantly worse with NF than with control treatment. In contrast, those with ODD-alone (top line) improved more with NF than controls from baseline to 13-month F/U. This was largely due to controls with ODD initially getting better but regressing to baseline by 13-month FU while those who received NF maintained their improvement (see Online Appendix).
1 3 be noted that there are no previous systematic studies that have investigated TBR NFB in anxiety.
In alignment with a recent ADHD medication study that found youth with comorbid anxiety fared worse than youth without (Al Ghriwati et al., 2017), youth with comorbid anxiety alone fared worse than those with no comorbidity in this neurofeedback study. Similar to the MTA's finding that children with comorbid DBD responded well to medication Greenhill et al., 2001), youth with only ODD fared well with neurofeedback compared to youth with no comorbidity and with anxiety. Thus, it appears the two treatments have similar effects vis-à-vis comorbidity.
Our results of opposing effects for anxiety and ODD comorbidities also may be meaningful for transdiagnostic research domain criteria (RDoC) considerations. Specifically, one important diagnostic factor being considered is a move toward a continuum of anxiety to ODD symptoms within the transdiagnostic general psychopathology dimension, or p-factor (Caspi et al., 2014). Our results lend some limited support for a continuum of anxiety and ODD symptomology. In this vein, neurofeedback may be improving the ODD end of the dimension more.
The findings, which are exploratory, are limited by our small sample size, particularly of anxiety-only comorbidity. We were underpowered with regard to the moderator analyses we conducted, which were exploratory. Future studies exploring this topic should counterbalance by anxiety disorders to ensure adequate power to examine these effects for each anxiety disorder separately. Another limitation is that we did not include a no treatment control condition, to which we would have been able to compare the Neurofeedback and Control groups. This should be considered in future studies of TBR neurofeedback. Another limitation of the present study was not systematically measuring verbal positive reinforcement and exposure to boredom. We also cannot generalize to all neurofeedback treatments, as we only tested theta-beta ratio downtraining. Another limitation is the relatively low percentage of black, indigenous, youth of color represented in our study. Future clinical trials need to include more diverse youth with regard to race and ethnicity.
In conclusion, comorbidities seem to moderate the effectiveness of TBR neurofeedback compared to control treatment for ADHD inattentiveness. Youth with comorbid anxiety alone, in contrast to other comorbidity profiles, had significantly worse ADHD inattentive symptoms with NF than with the control condition by end of treatment and should not have TBR NF. On the other hand, TBR NF seems to benefit those with ODD comorbidity, especially at delayed follow-up.

Supplementary Information
The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s10484-022-09575-x. Author Contributions MERR wrote the method, discussion, and Table 1 and supervised RB, SC, and KH's writing of the introduction. RB, SC, and KH wrote the introduction. YT prepared Figure 1. JP co-wrote the results and created Figure 2 and Table 3. CS and RdB created Table 2. LEA and NGJ co-wrote results. CK and MA assisted MERR in writing and revising the discussion. All authors reviewed the manuscript.
Data Availability Data used in this publication are available in the NIMH Data Archive.

Declarations
Conflicts of interests Dr. Arnold has received research funding from Curemark, Forest, Lilly, Neuropharm, Novartis, Noven, Shire, Supernus, Roche, and YoungLiving (as well as NIH and Autism Speaks), has consulted with Gowlings, Neuropharm, Organon, Pfizer, Sigma Tau, Shire, Tris Pharma, and Waypoint, and been on advisory boards for Arbor, Ironshore, Novartis, Noven, Otsuka, Pfizer, Roche, Seaside Therapeutics, Sigma Tau, Shire. Dr. Arns is (unpaid) director and owner of Research Institute Brainclinics, a minority shareholder in neuroCare Group (Munich, Germany), and a co-inventor on 4 patent applications related to EEG, neuromodulation and psychophysiology, but receives no royalties related to these patents; Research Institute Brainclinics received research funding from Brain Resource (Sydney, Australia), UrgoTech and neuroCare Group (Munich, Germany), and equipment support from Deymed, neuroConn and Magventure. Dr. Barterian  Informed Consent All research participants provided informed consent to participate in the study, and consented to allowing the research team to publish using the data collected as part of their participation in the research study.