Ultradian hydrocortisone replacement alters neuronal processing, emotional ambiguity, affect and fatigue in adrenal insufficiency: The PULSES trial

Abstract

Background: Primary adrenal insufficiency (PAI) mortality and morbidity remain unacceptably high, possibly arising as glucocorticoid replacement does not replicate natural physiology. A pulsatile subcutaneous pump can closely replicate cortisol's circadian and ultradian rhythm.

Objectives: To assess the effect of pump therapy on quality of life, mood, functional neuroimaging, behavioural/cognitive responses, sleep and metabolism.

Methods: A 6-week randomised, crossover, double-blinded and placebo-controlled feasibility study of usual dose hydrocortisone in PAI administered as either pulsed subcutaneous or standard care in Bristol, United Kingdom (ISRCTN67193733). Participants were stratified by adrenal insufficiency type. All participants who received study drugs are included in the analysis. The primary outcome, the facial expression recognition task (FERT), occurred at week 6.

Results: Between December 2014 and 2017, 22 participants were recruited - 20 completed both arms, and 21 were analysed. The pump was well-tolerated. No change was seen in the FERT primary outcome; however, there were subjective improvements in fatigue and mood. Additionally, functional magnetic resonance imaging revealed differential neural processing to emotional cues and visual stimulation. Region of interest analysis identified the left amygdala and insula, key glucocorticoid-sensitive regions involved in emotional ambiguity. FERT post hoc analysis confirmed this response. There were four serious adverse events (AE): three intercurrent illnesses requiring hospitalisation (1/3, 33.3% pump) and a planned procedure (1/1, 100% pump). There was a small number of expected AEs: infusion site bruising/itching (3/5, 60% pump), intercurrent illness requiring extra (3/7, 42% pump) and no extra (4/6, 66% pump) steroid.

Conclusions: These findings support the administration of hormone therapy that mimics physiology.

Keywords: fMRI; glucocorticoid replacement therapy; primary adrenal insufficiency; ultradian.

Fig. 1

Trial profile. Local institutional, NHS Health Research Authority and Medicines Healthcare Regulatory Approval was obtained. The study was conducted in accordance with the principles of the Declarations of Helsinki and International Conference on Harmonisation E6 guidelines for Good Clinical Practice (GCP).

Fig. 2

Emotional Test Battery treatment effect. Note: The top panel shows the difference in mean facial expression recognition task (FERT) and Emotional Categorisation Task (ECAT) accuracy, and EREC relative recall (blue circles) between the hydrocortisone administration groups (pulsatile minus oral). The middle panel shows the difference in FERT and ECAT time to respond on the logarithmic scale (blue circles) between the hydrocortisone administration groups (pulsatile over oral). The bottom panel shows the difference in mean Faces Dot Probe Task (FDOT) vigilance score (blue circle) between the hydrocortisone administration groups (pulsatile minus oral). The red line depicts the 95% confidence interval (CI). GMR, geometric mean ratio; MD, mean difference.

Fig. 3

(a) Neural processing differences between the oral and pulsatile hydrocortisone replacement groups, under the same conditions of emotional stimulation (emotionally valenced face presentation). Whole‐brain analysis revealed a cluster of brain regions (containing 778 voxels, highlighted in blue, coronal view), corresponding to parts of the left superior frontal gyrus, precentral gyrus, supplementary motor cortex, anterior cingulate and paracingulate (based on the Harvard‐Oxford cortical structural atlas), showing a reduced (compared to baseline) per cent blood oxygen level dependent (BOLD) signal response to emotional stimulation in the pulsatile treatment group compared to the oral one (p < 0.0001). (b) Region of interest (ROI) analysis revealed similar, between‐treatment group differences in the per cent BOLD signal responses to emotional stimulation for parts (highlighted in yellow and green, respectively, coronal view) of the left amygdala (Lamy, blue) and left insula (LIns, blue). (c) Post hoc analysis of variance showed a comparatively reduced accuracy of the participants in the oral treatment to correctly identify highly ambiguous, negatively valenced faces (*). Data presented as group mean ± SD. FERT, facial expression recognition task; MAX z‐value, the maximum z‐statistic contained in the cluster; MAX X/Y/Z (mm), the location of the voxel with the maximum z‐statistic, given as X/Y/Z coordinate values in MNI152 standard space (mm); COG X/Y/Z (mm), the location of the centre of gravity for the cluster (i.e. a weighted average of the coordinates by the intensities within the cluster).

Fig. 4

(a) Neural processing differences between the oral and pulsatile hydrocortisone replacement groups, under the same conditions of visual stimulation (flashing checkerboard). Whole‐brain analysis revealed a cluster of brain regions (containing 1107 voxels, highlighted in orange–yellow, sagittal view), corresponding to parts of the posterior cingulate and the precuneous (based on the Harvard‐Oxford cortical structural atlas), which show a reduced (compared to baseline) per cent blood oxygen level dependent (BOLD) signal response to visual stimulation in the pulsatile treatment group compared to the oral one (p < 0.0001). (b) Evolution of positive and negative affect index scores throughout the study period (at baseline, during the 6‐week study treatment period and at follow‐up, after the end of each treatment arm) for participants receiving either the oral or the pulsatile hydrocortisone replacement therapy. (c) The boxplots summarize the distribution of the correlation values (connection strengths) in the two groups – pulsatile and oral treatment – for a particular node‐pair, either the middle temporal gyri with the right occipital fusiform gyrus (upper pair of thumbnails) or the left inferior frontal gyrus with the subcallosal cortex (lower pair of thumbnails). The thickness of the blue‐coloured bar joining each pair of thumbnails indicates the strength of the overall group‐average connection, whereas the blue colour itself indicates that this connection is negative (i.e. the nodes anti‐correlate to each other on average). This anti‐correlation is comparatively stronger in the oral treatment group. Data presented as mean ± SD. IQR, interquartile range (range between 25th and 75th percentile of the data), MAX z‐value, the maximum z‐statistic contained in the cluster, MAX X/Y/Z (mm), the location of the voxel with the maximum z‐statistic, given as X/Y/Z coordinate values in MNI152 standard space (mm); COG X/Y/Z (mm), the location of the centre of gravity for the cluster (i.e. a weighted average of the coordinates by the intensities within the cluster).

Fig. 5

Secondary outcome measures treatment effects. Note: The figure shows the difference in mean response for the secondary outcomes (blue circles) between the hydrocortisone administration groups (pulsatile over oral). The red line depicts the 95% confidence interval (CI). N‐Back response time, quality of sleep, awakening from sleep, behaviour following wakefulness from the Leeds Sleep questionnaire (LSEQ), Pittsburgh Sleep Quality Index (PSQI), Identity‐Consequence Fatigue Scale (ICFS), Negative Affect Score and Chalder fatigue score (CFS) were reversed to help interpretation. MD, mean difference.