Introduction

Studying youth at high risk of developing bipolar disorder may extend our understanding of neurobiological factors associated with the vulnerability to this illness. Magnetic resonance spectroscopy (MRS) is a noninvasive neuroimaging technique that provides in vivo measurement of specific biochemicals in localized brain regions and previous work has identified neurochemical changes in bipolar patients thought to be involved in bipolar symptoms ¹. However, few efforts have been made to evaluate whether these chemicals are associated with the vulnerability of bipolar disorder. In the present study, we used support vector machine (SVM) ² to characterize the potential of neurochemicals in predicting symptomatic converters in youth offspring with bipolar parents.

Methods

Thirty-eight offsprings of bipolar disorder parents and 19 offsprings of healthy parents were recruited from the Cincinnati Children’s Hospital Medical Center and the University of Cincinnati Medical Center. Their baseline MRS scans were performed on a Varian 4T whole-body scanner, and a 1H transverse electromagnetic (TEM) volume head coil was used as a transmitter and receiver. Three ROIs (8 cc in volume) were positioned as done before ³, one in the anterior cingulate cortex (ACC) and one each in the left and right ventral lateral prefrontal cortex (LVLPFC and RVLPFC). Metabolite levels were determined by analyzing spectra using LCModel (Linear Combination of Model spectra) with the water reference in unsuppressed-water spectra ⁴. The concentrations of glutamate (Glu), myo-inositol (mI), choline (Cho), N-acetyl aspartate (NAA) and creatine (Cr) were examined within each ROI. The bipolar offsprings were later divided into two subgroups according to the presence or absence of lifetime psychopathology (hereafter called symptomatic bipolar offspring (n=19, mean age: 13.05±1.89 years, 5 females) and healthy bipolar offspring (n=19, mean age: 13.15±2.19 years, 6 females)). SVM as implemented in the PROBID software package (http://www.brainmap.co.uk/probid.htm) was employed and a linear kernel SVM was adopted to classify between symptomatic and healthy bipolar offspring based on their baseline metabolites. Statistical significance of classification accuracy for each modality was set at p < 0.005 after permutation testing (1000 times).

Results

The two subgroups of bipolar offsprings were well-matched in age, sex, education and parental socioeconomic status. SVM allowed the classification of the two groups with each metabolite across all the ROIs: Cho, accuracy=76% (p=0.001); Cr, accuracy=71% (p=0.006); mI, accuracy=71% (p=0.006); Glu, accuracy=63% (p=0.063); NAA, accuracy=58% (p=0.195). Receiver operating characteristic (ROC) curves and weights of each region for each metabolite during the classification were presented in Figure 1. Then, the discriminating ability of each ROI with all the metabolites were examined and findings were: left VLPFC, accuracy= 76.5% (p=0.001); right VMPFC, accuracy= 68.5% (p=0.012); ACC, accuracy= 52.5% (p=0.409). Please see Figure 2 for details. From findings above, left VLPFC with all the metabolites was found with significant differentiating ability in the classification, especially the measures of mI (0.682), Cho (0.480) and Cr (0.542). In the final step, we just examined the measures of mI, Cr and Cho in the left VLPFC for the differentiation, and the finding achieved the highest accuracy of 79.0% (sensitivity=0.84, specificity=0.74, p=0.001). Detailed comparisons of each metabolite in each ROI between symptomatic and healthy bipolar offspring and healthy controls were presented in Figure 3.

Discussion

To our best knowledge, the current study is the first to examine the capability of SVM with brain chemicals in predicting symptomatic converters in youth offspring with bipolar parents. By identifying the inter-group differences in neuro-metabolites, the present study suggests that chemicals of prefrontal regions are related to vulnerability of bipolar disorder, which may act as predictors for development of disorder among people as high risks, especially the mI, Cr and Cho in the left VLPFC. The abnormal metabolites were mainly on the left, further indicating a potential lateralization underlying the pathogenesis of this disorder. Anyhow, future studies integrating other objective imaging modalities may help to develop a better tool for identifying the vulnerability of bipolar disorder.

Conclusion

By presenting some specific differential chemicals which could be used to predict the symptomatic converters of youth offspring with familial history at the individual level, this study provides evidence that some metabolites are associated with the vulnerability of bipolar disorder and can possibly act as a diagnostic aid in identifying converters of bipolar youth offspring.

Acknowledgements

This study was supported by National Institute of Mental Health (NIMH) Grant (Grant No. 5R01MH080973 (DelBello)).