Beyond the label "major depressive disorder"-detailed characterization of study population matters for EEG-biomarker research

Abstract

Introduction: Major Depressive Disorder (MDD) is a prevalent, multi-faceted psychiatric disorder influenced by a plethora of physiological and environmental factors. Neuroimaging biomarkers such as diagnosis support systems based on electroencephalography (EEG) recordings have the potential to substantially improve its diagnostic procedure. Research on these biomarkers, however, provides inconsistent findings regarding the robustness of specific markers. One potential source of these contradictions that is frequently neglected may arise from the variability in study populations.

Methods: This study systematically reviews 66 original studies from the last 5 years that investigate resting-state EEG-biomarker for MDD detection or diagnosis. The study populations are compared regarding demographic factors, diagnostic procedures and medication, as well as neuropsychological characteristics. Furthermore, we investigate the impact these factors have on the biomarkers, if they were included in the analysis. Finally, we provide further insights into the impact of diagnostic choices and the heterogeneity of a study population based on exploratory analyses in two publicly available data sets.

Results: We find indeed a large variability in the study populations with respect to all factors included in the review. Furthermore, these factors are often neglected in analyses even though the studies that include them tend to find effects.

Discussion: In light of the variability in diagnostic procedures and heterogeneity in neuropsychological characteristics of the study populations, we advocate for more differentiated target variables in biomarker research then simply MDD and healthy control. Furthermore, the study populations need to be more extensively described and analyses need to include this information in order to provide comparable findings.

Keywords: biomarker; diagnosis; electroencephalography; label; major depressive disorder; study population.

Figure 1

Basic characteristics of the studies. (A) Overall number of participants for studies with own data (boxplot: n = 31) and overlaid the population sizes of the public data sets. (B) Percentages of MDD patients relative to the sum of MDD patients and HC (boxplot: n = 31) and the percentages of female participants in the two clinical sub-groups (boxplots: n = 27) for the studies with own data. Overlaid are the respective information for the public data sets. Note that for CAV the diagnostic label is based on BDI score, for NOW based on diagnosis, for DIA all participants are included in (A) but the diagnostic label in (B) is based on diagnosis and BDI score (n = 50), and for WRO all participants are included in (A) but the diagnostic label in (B) is based on BDI score (n = 86) The information about MODMA, CAV, MUMTAZ, NOW, DIA, and WRO is based on the public data itself, the information about other three data sets is based on the studies included in the review. Note that information about gender was missing in four studies with own data and in the study using TDBRAIN.

Figure 2

Age distributions of participants in studies with own data collection (n = 26) and in public data sets. Color coding for the publicly available data sets: reddish colors mark MDD patients, blueish colors mark HC. Diagnosis is defined analogue to Figure 1B. For study abbreviations see chapter 3.3. Note that information about age was missing in five studies with own data and in the study using TDBRAIN.

Figure 3

Distributions of the two most frequently used neuropsychological tests. (A) HAMD-17 (n = 11/4 for MDD/HC) (B) BDI (n = 9/9 for own data), both versions of the test. The dashed lines demarcate the cut-offs between severity categories for each of the tests. Color-coding and definition of diagnosis groups analogue to Figure 2.

Figure 4

Relationship between the BDI and the STAI (A) and the BDI and the HAMD-17 scores (B) for the CAV data set. Grouping of participants is based on the diagnostic procedure (see text).

Figure 5

Clustering the MODMA data set based on depression (PHQ-9), anxiety (GAD-7), and sleep (PSQI) scores with hierarchical clustering (A, B) and k-means clustering (C, D). Colors in each row correspond to the same groups but colors across rows cannot be interpreted. Note that the light blue circles are intentionally drawn smaller in order to show the overlap to the dark blue circles.