A Neglected Topic in Neuroscience: Replicability of fMRI Results With Specific Reference to ANOREXIA NERVOSA

Abstract

Functional magnetic resonance imaging (fMRI) studies report impaired functional correlates of cognition and emotion in mental disorders. The validity of preexisting studies needs to be confirmed through replication studies, which there is a lack of. So far, most replication studies have been conducted on non-patients (NP) and primarily investigated cognitive and motor tasks. To fill this gap, we conducted the first fMRI replication study to investigate brain function using disease-related food stimuli in patients with anorexia nervosa (AN). Using fMRI, we investigated 31 AN patients and 27 NP for increased amygdala and reduced midcingulate activation when viewing food and non-food stimuli, as reported by the original study (11AN, 11NP; Joos et al., 2011). Similar to the previous study, we observed in the within group comparisons (food>non-food) a frontoinsular activation for both groups. Although in AN the recorded activation clustered more prominently and extended into the cingulate cortex. In the between-group comparisons, the increased amygdala and reduced midcingulate activation could not be replicated. Instead, AN showed a higher activation of the cingulate cortices, the pre-/postcentral gyrus and the inferior parietal lobe. Unlike in the initial study, no significant differences between NP>AN could be observed. The inconsistency of results and the non-replication of the study could have several reasons, such as high inter-individual variance of functional correlates of emotion processing, as well as intra-individual variances and the smaller group size of the initial study. These results underline the importance of replication for assessing the reliability and validity of results from fMRI research.

Keywords: anorexia nervosa; food; functional magnetic resonance imaging (fMRI); neurobiology; replicability.

Figure 1

Clinical characteristics of anorexia nervosa (AN) and non-patient (NP), study sample 2011 compared to 2020. BDI-II, Becks Depression-Inventary-2; BMI, Body-Mass-Index; EDI-drive for thinness, Eating Disorder Inventory; kg, kilograms; m², square meter; *p = 0.014. For further clinical characteristics of the replication study see Table 1 .

Figure 2

Within group and between group contrasts of the replication study compared to results of Joos et al., 2011 (A) Cerebral activation of the within group contrast of anorexia nervosa and non-patients for food>non-food (p _{uncorrected <}0.001, k=10, for visualization purposes). Results of Joos et al., 2020 (one-sample t-test) compared to those of Joos et al., 2011 (full-factorial). All slices at MNI coordinates (0, 45, 0) where chosen as in Joos et al., 2011 for a good comparison. Color bars represent the t-scores (white/yellow = high, red = low). Maps from Joos et al., 2011 with kind permission of Elsevier. (B) Left Fig. Rain cloud plot of the contrast estimates (food>non-food) in the non-significant midcingulate cortices (MCC) of the replication study (MNI: x=6, y=-28, z=35, with 3mm radius), next to the MCC which derived from the NP>AN contrast reported by Joos et al., 2011 (MNI: x=9, y=-33, z=47). Right Fig: T-maps of the second level analysis (t-test AN(food>non-food) >NP (food>non-food)) according to the current recommendations (cluste-defining threshold p<0.001). Slices where chosen at non-significant peak cluster activity in the right MCC AN>NP (MNI coordinates x=6, y=-28, z=35); p_uncorrected <0.001, k>0) Color bars represent the t-scores (white/yellow = high, red = low).