(2) Relations of biochemical measures to group features, symptoms and attention
Group features (relations to tryptophan metabolism and the cytokines)
Across all children tryptophan levels increased with BMI and age (R2 0.14, F(1,54) = 8.69: β +0.37, p = 0.005): this was evident in both the ADHD and ADHDmed groups (R2 0.22/0.60, F(1,19) = 5.33/8.16: β + 0.47/+0.96, p = 0.03/0.002, respectively). 3HK levels also tended to increase in controls (R2 0.20, F(1,19) = 4.69, β +0.45, p = 0.04). IQ was unrelated, but low SES was predicted by lower tryptophan and increasing kynurenate levels (R2 0.15, F(2,53) = 4.70: β + 0.33/-0.34, p = 0.018/0.015, respectively). This was supported by modest associations with increased tryptophan breakdown and lower neuroprotection ratio-B (β -0.3/-0.28, p = 0.02/0.03). Lastly, across all children (and within each group), those with an experience of allergy tended to be predicted by low levels of 5-HIAA (R2 0.07, F(1,54) = 3.93: β -0.26, p = 0.05).
Group features were only marginally predicted by the levels of a few proinflammatory cytokines. Across all subjects increasing TNF-α and decreasing IFN-γ levels tended to predict increasing age (R2 0.084, F(2,53) = 0.09: β +0.28/-0.31, p = 0.077/0.049, respectively). BMI also tended to be predicted by increasing TNF-α and IL-6 levels (β +0.32/+0.45, p = 0.03/0.001, respectively).
Across subjects allergic experience tended to relate to decreasing proinflammatory TNF-α but increasing antiinflammatory IL-10 levels (R2 0.16, F(2,53) = 5.0, p = 0.01: β -0.27/ + 0.31, p = 0.039/0.017, respectively). In the ADHD group, in addition to the TNF-α decrease, (β -0.63, p = 0.03) increasing antiinflammatory IL-13 levels predicted response (β +0.56, p = 0.01). In contrast, for the controls alone the increase in IL-10 was masked by decreases of IL-13 in the prediction of allergy severity (R2 0.42, F(2,18) = 6.6, p = 0.007: β + 0.33/-0.13, p = 0.08/0.01, respectively).
Tryptophan levels increased with age and BMI, but lower SES status facilitated metabolism in the kynurenine pathway. Some proinflammatory cytokine levels decreased (IFN-γ) or increased with BMI (TNF-α, IL-6). Independent of diagnosis, children with allergies tended to show lower levels of 5-HT metabolism. Most children with experience of allergy showed decreases of the proinflammatory TNF-α. However, in contrast to the controls who showed decreases of the anti-inflammatory IL-13, the ADHD group showed increased IL-13 levels.
Symptom ratings: relations of tryptophan metabolism
In general, symptoms were not associated clearly with tryptophan metabolism. Partial correlations accounting for age, BMI and SES suggested that inattentive symptoms were related to tryptophan levels across all children (r = +0.32, p = 0.026). Alone for controls anxiety related to 5-HIAA levels (r = 0.52, p = 0.038) and inattentive ratings with 3HK (r = +0.49, p = 0.037). Only the latter result (3HK) approached the threshold for significance in the regression (R2 0.19, F(1,19) = 4.5: β +0.44, p = 0.048).
Symptom ratings: relations of cytokine measures
Overall, considering total symptom ratings in the ADHD group the best predictors were increasing levels of IL-13 and IL-16 along with decreasing levels of S100B (R2 0.57, F(4,16) = 5.21, p = 0.007: β +0.59, +0.51, -0.54, p = 0.024, 0.019, 0.013, respectively) with a minor contribution from TNF-α.
Partial correlations accounting for age, BMI and SES pointed to associations for pro-inflammatory cytokines with two symptom groupings, namely the oppositional and hyperactive-impulsive ratings. In the ADHD group decreasing IL-2 (r -0.64, p = 0.026) and TNF-α (r -0.62, p = 0.018: figure ) correlated with increased ratings of opposition. Regression confirmed the IL-2 result (R2 0.41, F(2,18) = 6.2, p = 0.009: β -0.55, p = 0.007: figure ) and implied a modest predictive role for increasing levels of S100B (β +0.37 p = 0.057). Of interest is that in the controls another proinflammatory cytokine (IL-6) correlated with oppositional ratings (r +0.48, p = 0.05): the predictive role was confirmed in the regression (R2 0.30, F(1,19) = 7.99, p = 0.01: β +0.54, p = 0.01).
For children with ADHD, ratings of oppositional behaviour vs. levels of TNF-α (linear regression).
For children with ADHD, ratings of oppositional behaviour vs. levels of IL-2 (linear regression).
The association of hyperactive-impulsive ratings in the ADHD group with IL-16 (r +0.66, p = 0.015) was confirmed by the regression (R2 0.52, F(2,18) = 9.82, p = 0.001: β +0.80, p = 0.000) with modest support from decreasing levels of S100B (β -0.42, p = 0.035). Anxiety ratings were without relationships to the cytokines. But, intriguingly ratings of inattention in the ADHD group tended to relate positively to levels of the antiinflammatory IL-13, yet negatively to IL-13 in the controls (R2 0.33/0.45, F(2/3, 18/17) = 4.5/4.6, p = 0.026/0.015: β +0.44/-0.67, p = 0.03/0.007, respectively: figure ). The ADHD result was modestly supported by increasing IL-6 levels (β +0.38) and the control result by increasing S100B (β + 0.64).
For children with ADHD, ratings of inattention vs. levels of IL-13 (linear regression).
There were modest associations of tryptophan levels and its metabolism to the potentially toxic 3HK with inattention, and 5-HT metabolism to anxiety. Proinflammatory cytokines were associated with oppositional symptoms: decreases of IL-2 in the ADHD group contrasted with increases of IL-6 in the controls. In general, ADHD symptoms were associated with increases of the antiinflammatory IL-13 (inattention), increases of IL-16 (hyperactive-impulsive symptoms) and with decreases of S100B (total symptoms).
CPT performance: relations of tryptophan metabolism
Across all subjects motor activity (duration, microevents) showed a very modest tendency to increase with lower levels of 3HK (R2 0.07, F(1,54) = 4.0/3.4: β-0.26/-0.24, p = 0.05/0.07, respectively), but this was not evident in the partial correlations and single group analyses.
In control children faster RTs were associated with increasing kynurenine levels (R2 0.32, F(1,19) = 9.13: β -0.57, p = 0.007: figure ). This confirmed the partial correlation (r = -0.52, p = 0.026). Unexpectedly, the opposite relationship appeared in the ADHDmed group (r = + 0.93, p = 0.000: R2 0.48, F(1,12) = 10.93: β +.69, p = 0.006) with no association evident in the non-medicated cases. Across all subjects, there was a partial correlation for the coefficient of RT variance with tryptophan levels (r+0.30, p = 0.035), yet a predictive role for tryptophan was not confirmed in the regression. Interestingly, the decrease of tryptophan (and variability) in controls (vs. the ADHD group, figure ) tended to be reflected by increases of 5-HIAA (r = -0.45, R2 0.17, F(1,19) = 3.95: β -0.42, p = 0.06). But, no associations were evident in the ADHD groups. The interpretation of the RT associations in terms of 5-HT metabolism is supported by the partial correlations of the turnover indices. While tryptophan availability was positively related to RT (r+0.36, p = 0.012), the breakdown to kynurenine was negatively associated (r-0.31, p = 0.03).
For healthy control children, reaction time (RT) on the CPT task vs. levels of kynurenine (linear regression).
The error analysis suggested that omissions were positively associated with tryptophan levels across subjects (r = +0.38 p = 0.006: R2 0.14, F(1,54) = 9.03: β +0.38, p = 0.004). This positive association between tryptophan levels and omission errors was evident as a trend in the ADHDmed group where the relationship to commission errors was negative (R2 0.35/0.54, F(1,12)/(2,11) = 6.37/6.42: β +0.59/-0.92, p = 0.027/0.004, respectively). This negative trend with commission errors is upheld by their association with the tryptophan breakdown index (R2 0.49, F(1,12) = 11.8: β +0.70, p = 0.005).
CPT performance: relations of cytokine measures
Considering motor activity across all subjects, the partial correlations pointed to a positive association of the duration measure with IL-16 levels (r+0.30, p = 0.038). This was confirmed by the regression (R2 0.08, F(1,54) = 4.56: β +0.28, p = 0.037). The motor associations appeared to be driven by IL-16 in the ADHD group for both duration and microevent measures (R2 0.55/0.47, F(3,17) = 6.9/5.0, p = 0.003/0.01: β +0.55/+0.47, p = 0.007/0.028, respectively). [The other degrees of freedom here refer to minor contributions from decreases of the pro-inflammatory IL-6 and TNF-α.] In contrast, the effect of medication introduced a potential modest antiinflammatory influence (e.g. IL-10 with microevents, R2 0.35, F(1,12) = 6.3: β +0.59, p = 0.027).
Several analyses indicated minor negative associations of RT with cytokines (e.g. IL-2, S100B) but only the effect of medication showed that decreasing RTs related to increasing levels of IL-2 (ADHDmed: β -0.73, p = 0.003: figure ). This contrasts with RT variability, where across subjects the coefficient of variance was influenced in opposite ways by TNF-α and IFN-γ (R2 0.18, F(2,53) = 6.1, p = 0.004: β -0.46/+0.04, p = 0.002/0.007, respectively: figure ). While these two relationships were not driven by processes related to diagnosis, the association with TNF-α was most marked in the medicated ADHD group (β -0.70, p = 0.007).
For children with ADHD on medication the reaction time vs. levels of IL-2 (linear regression).
For all subjects, the coefficient of variance (CV) vs. levels of TNF-α (linear regression).
In the error analyses errors of commission were predicted by levels of IL-16 (r+0.42, p = 0.003: R2 0.17, F(1,54) = 10.75: β +0.41, p = 0.002: figure ). But, as with variability (above) this was most apparent in the ADHDmed group (β +0.75, p = 0.000). In this group alone, again as with variability, decreases of proinflammatory TNF-α (β-0.36, p = 0.016) and IL-6 (β-0.39, p = 0.032) also tended to predict errors of commission. For omission errors only the negative associations of IL-13 (ADHD: β-0.47, p = 0.03)) and IL-16 (ADHDmed: β-0.63, p = 0.004) were of note.
For all subjects, commission errors on the continuous performance test vs. levels of IL-16 (linear regression).
Measures of motor activity were not related to tryptophan metabolism but were positively predicted by IL-16 levels. In contrast, shorter RTs were associated with increasing kynurenine levels in controls, but by decreasing levels in the ADHDmed group. Here, the cytokines were without any major influence. Interestingly, decreased RT variability in controls was reflected by modest increases of catabolism to 5-HIAA. With regard to the cytokines, across subjects, levels of variability reflected the opposing influences of TNF-α and IFN-γ. In the error analysis, while the build up of tryptophan predicted omission errors, its breakdown predicted errors of commission. But only the errors of commission were predicted by IL-16 levels. Medication may moderate interactions of pro-and antiinflammatory cytokines in motor activity (IL-2/RT, IL-10/microevents) and the influences of proinflammatory TNF-α, IL-6 (variable RT) and IL-16 (impulsivity).