Subjects
The study sample consisted of 31 women with PMDD and 31 asymptomatic controls recruited by advertisement in local newspapers, on a student website for clinical trials, via social media platforms, and by posters at out-patient clinics. Women were eligible for inclusion if they were aged 18–45 years, had regular menstrual cycles (25–31 days), used non-hormonal contraception, fulfilled PMDD diagnostic criteria according to the Diagnostic and Statistical Manual of Mental Disorders, 5th Ed (DSM-V) (PMDD group), were otherwise essentially healthy (PMDD and control group), and provided informed consent. Current use of steroid hormones, psychotropic or anti-depressant medication, significant somatic or psychiatric conditions, drug or alcohol abuse, pregnancy and contraindications for MRI were grounds for exclusion. Prior to entering the study, we required wash-out periods of three months for psychotropic drugs (like selective serotonin re-uptake inhibitors and benzodiazepines) or alternative medicines with potential effects on mood, and one month for hormonal contraceptives. Potential participants were screened for psychiatric conditions (past and current) by the investigator using the Mini International Neuropsychiatric Interview questionnaire [34]. We included women with past major depressive episodes as long they had been in remission for more than two years. All participants completed prospective daily ratings of PMDD symptoms for a minimum of two menstrual cycles using the Daily Record of Severity of Problems (DRSP), a validated diagnostic tool for PMDD [35]. The DRSP was implemented via an ad-hoc web platform and daily text reminders were sent encouraging participants to log their symptoms. The study was approved by the Regional Ethical Review Board in Umeå (2016-111-31M, 2017-266-32M).
Pmdd Diagnosis
PMDD was diagnosed using the algorithm developed by Endicott et al. [35]. The criteria were as follows: 1) no average daily symptom score greater than 3 (“mild”) during the mid-follicular phase (days + 6 to + 10 after the onset of menses), 2) during the late-luteal phase (days − 5 to -1 prior to the onset of menses), at least two days with ratings \(\ge\)4 (“moderate”) on a minimum of one “core” mood symptom (depressed mood, anxiety, affective lability, irritability) and on at least five symptoms overall, 3) symptoms in the late-luteal phase interfered with daily functioning, which was defined as ratings of \(\ge\)4 for two days on at least one impairment item (interference with work/school, social activities, relationships). A diagnosis of PMDD was given if the above criteria were met for two consecutive menstrual cycles. Women included in the control group had to be asymptomatic across the entire menstrual cycle, i.e. no mean ratings >3 during either the mid-follicular phase or the late-luteal phase.
Study Design
Participants were scanned once in the asymptomatic mid-follicular phase (menstrual cycle day + 5 to + 11), and once in the late-luteal phase (menstrual cycle day − 8 to -1). Luteal phase testing coincided with peak PMDD symptoms [4]. Prior to each scanning session, a blood sample was drawn, and serum was frozen at -80o within 30 min from collection for further analysis of ovarian hormones and neurosteroids. Ovulation was confirmed if serum progesterone concentrations were within 2 standard deviations from the progesterone standard curve during the corresponding luteal phase day [14]. To avoid test order effects, the menstrual cycle phase in which participants underwent their first scanning session was counterbalanced within each group: 44.4% of controls and 51.7% of women with PMDD underwent their first scan in the mid-follicular phase.
Steroid Analyses
Serum concentrations of ALLO and ISO were analyzed by LabLytica, Uppsala, Sweden. In a first step, serum samples were extracted using liquid-liquid extraction in a hexane/ether solvent phase. They were then derivatized using 3-aminooxypropyl (trimethyl) ammonium bromide and quantified using ultra-high performance liquid chromatography mass spectrometry (UPLC-MS/MS). The samples were compared against a freshly prepared calibration curve in surrogate matrix (water) to determine their concentrations. The limit of quantification (LLOQ) for ALLO was 0.2 nM and for ISO 0.1 nM. Serum concentrations of progesterone and estradiol were analyzed by the central hospital laboratory at Norrlands University Hospital, Umeå, Sweden.
Experimental Paradigm
The emotional discrimination task used in this study has been previously described [36]. Participants were presented with Ekman faces displaying expressions of anger or fear (emotion task) and vertical or horizontal ellipses (sensorimotor control task). Participants were instructed to select one of two images matching the emotion or orientation of a target image by pressing a button with the right index finger. Emotion and sensorimotor control task trials were presented in blocks of six, in which stimuli were presented for 4 s, interspaced with a fixation cross (2 s for the sensorimotor control task and a randomly selected duration of 2, 4 or 6 s for the emotion task). Emotional content and sex of the individuals depicted were balanced across trials, as was the orientation of shapes. The entire paradigm consisted of four blocks of faces (24 trials) and five blocks of shapes (30 trials). Accuracy and reaction times were registered for each trial.
Image Acquisition
Magnetic resonance (MR) images were acquired using a 3.0T Discovery MR750 (General Electric, Madison, WI, USA) scanner available through the Umeå Center for Functional Brain Imaging (UFBI). The scanner was equipped with a 32-channel head coil. The stimulus presentation software E-prime (Psychology Software Tools, Sharpsburg, PA, USA) was used for paradigm handling and viewed through a tilted mirror attached to the head coil. fMRI images were acquired with a gradient echo planar imaging sequence [37 transaxial slices; thickness, 3.4 mm; gap; 0.5 mm, repetition time (TR), 2000 ms; echo time (TE), 30 ms; flip angle, 80°; field of view, 25 × 25 cm; 200 volumes; duration, 07:00 min]. High-resolution T1-weighted structural images were collected with a 3D fast spoiled gradient echo sequence (176 transaxial slices; thickness, 1 mm; TR, 8.2 ms; TE, 3.2 ms; flip angle, 12°; field of view, 25 × 25 cm; duration, 08:11 min). A fieldmap was acquired prior to the fMRI images, for use in controlling for magnetic field (B0) inhomogeneities [46 transaxial slices; thickness; 4 mm; gap; 0 mm; repetition time (TR), 800 ms; flip angle, 10°; field of view, 25.6×25.6 cm; duration, 01:05 min]. All sequences were acquired in the A/P (anterior-to-posterior) frequency-encoding direction.
Fmri Data Preprocessing And Analysis
Image processing was conducted using the Oxford Centre for Functional Magnetic Resonance Imaging of the Brain (FMRIB)’s Software Library (FSL), version 6.00 [37]. Preprocessing steps included motion correction (reference image = middle volume), correction for B0 inhomogeneities, slice timing correction, and spatial smoothing with a 5-mm full-width at half maximum (FWHM) Gaussian kernel. A rigid body registration with FMRIB’s Linear Image Registration Tool (FLIRT) [38, 39], 6 degrees of freedom (DOF), was used to co-register functional images to individual structural T1 images. Spatial normalization into Montreal Neurological Institute (MNI) space was performed by applying an initial registration with FLIRT, 12 DOF affine transformations, followed by a non-linear transformation using FMRIB’s Non-linear Image Registration Tool (FNIRT) [40] with a warp-resolution of 8mm, and resulting in 2×2×2 mm3 voxels.
First-level temporal modelling within a general linear model (GLM) framework was performed with FSL Expert Analysis Tool (FEAT) to generate single subject 3D maps of parameter estimates the contrast of interest [Faces > Shapes]. Design matrices were convolved with the default gamma hemodynamic response (HRF) function. Six motion parameters estimated from the spatial realignment were added to the model as covariates of no interest, and frames corrupted by large movements detected by FSL’s Motion Outliers tool (default metric = refrms) were removed from the analyses. Outliers were defined as falling outside the boxplot cut-off of 75th percentile + 1.5 x interquartile range. A high-pass filter (cut-off = 90s) was applied to attenuate the lowest frequency components (linear scanner drift).
Statistical Analyses
Group differences in emotion-induced brain activity in the amygdala and at the whole brain level
FSL’s non-parametric permutation testing Randomise tool [41] was used for statistical inference. Voxel-wise analyses were conducted both with a small volume correction (SVC) for the right and left amygdala and at the whole-brain level. The region-of-interest (ROI) mask for the left and right amygdalae was defined using the Harvard-Oxford Subcortical Structural Atlas (thresholded at 80% probability). In order to assess group x phase interactions for the contrast of interest [Faces > Shapes] without violating the assumption of exchangeability relied on by permutation tests, individual differences between follicular and luteal scans were first computed for each subject before testing for group differences using an unpaired two-sample t-test. Separate paired and unpaired two-sample t-tests were then conducted to detect effects of menstrual cycle phase within each group, and group differences within each menstrual cycle phase, respectively. Additional analyses assessing the influence of prior psychiatric history on the findings were conducted, as this variable tended to differ between groups (see Results, Table 1).
Table 1
Baseline characteristics of women with PMDD and controls.
| PMDD (N = 29) Mean (SD) or N (%) | Controls (N = 27) Mean (SD) or N (%) |
Demographics |
Age (years) | 28.5 (6.1) | 28.3 (5.7) |
BMI | 23.7 (3.2) | 24.4 (4.1) |
Menstrual cycle length (days) | 27.7 (1.9) | 28.6 (1.9) |
Psychiatric history | 9 (31.0) | 3 (11.1) |
Depression | 8 (27.6) | 3 (11.1) |
Eating disorder | 1 (3.45) | 0 |
Parous | 9 (31.0) | 7 (25.9) |
DRSP ratings | Mid-follicular | Late-luteal | Mid-follicular | Late-luteal |
Total symptom score | 25.5 (3.5) | 56.7 (14.3) | 24.2 (3.3) | 24.4 (3.0) * |
Depression score | 3.5 (0.6) | 8.3 (2.8) | 3.4 (0.7) | 3.5 (0.6) * |
Anxiety score | 1.2 (0.3) | 2.8 (1.0) | 1.1 (0.2) | 1.1 (0.2) * |
Emotion lability score | 2.3 (0.4) | 9.3 (3.1) | 2.2 (0.4) | 3.4 (0.4) * |
Irritability score | 2.4 (0.4) | 5.3 (2.2) | 2.4 (0.6) | 2.9 (0.4) * |
Steroids | Mid-follicular Mean (IQR) or N (%) | Late-luteal Mean (IQR) or N (%) | Mid-follicular Mean (IQR) or N (%) | Late-luteal Mean (IQR) or N (%) |
Test day | + 8.0 (1.9) | -4.4 (2.0) | + 7.7 (1.4) | -3.9 (1.7) |
Progesterone (nmol/L) | 0.6 (0.5) | 23.8 (17.5) | 0.7 (0.5) | 23.4 (20.4) |
Estradiol (pmol/L) | 309 (201) | 424 (238) | 263 (162) | 432 (156) |
ALLO (nmol/L) | 0.337 (0.098) | 2.110 (1.420) | 0.419 (0.247) | 2.208 (1.060) |
Missing | 6 (20.7) | 0 | 3 (11.1) | 2 (7.4) |
ISO (nmol/L) | 0.127 (0.020) | 0.742 (0.484) | 0.137 (0.044) | 0.844 (0.652) |
Missing† | 21 (72.4) | 0 | 16 (59.3) | 3 (11.1) |
ISO/ALLO (nmol/L) | 0.367 (0.093) | 0.349 (0.108) | 0.271 (0.079) | 0.358 (0.153) |
Missing | 21 (72.4) | 0 | 16 (59.3) | 3 (11.1) |
Total symptom scores are the mean summed ratings for all 21 symptom items of the DRSP scale (minimum = 21, maximum = 126) over days + 5 to + 11 for the mid-follicular phase, and days − 8 to -1 for the late-luteal phase. The depression scores include the DRSP items “depressed”, “hopeless” and “guilty” (minimum = 3, maximum = 18); anxiety scores correspond to the item “anxious” (minimum = 1; maximum = 6); emotion lability scores include the items “mood swings” and “easily hurt” (minimum = 2; maximum = 12); and irritability scores include the items “irritable” and “conflicts” (minimum = 2, maximum = 12). Differences between groups were assessed using Mann-Whitney U-tests for continuous variables, and Fisher’s exact tests for categorical variables. *Significant group difference at p < 0.05. Abbreviations: ALLO, Allopregnanolone; BMI, body mass index; DRSP, Daily Record of Severity of Problems; IQR, Interquartile Range; ISO, Isoallopregnanolone; PMDD, Premenstrual Dysphoric Disorder; SD, Standard Deviation. |
†The large proportion of missing values for serum ISO was due to the lower detection limit of the method (see Materials and Methods, Steroid analysis). |
Group differences in associations between emotion-induced brain activity and serum steroid levels
To test whether the linear relationship between brain activity and serum steroid levels differed between women with PMDD and controls, a two-group with continuous covariate interaction analysis was conducted voxel-wise using Randomise. Analyses were restricted to voxels within a mask combining clusters showing increased recruitment during task [Faces > Shapes] for both the PMDD and control group across the menstrual cycle (Table S1). Steroid covariates were mean centered across all subjects and split into two separate regressors according to group before being included in the model, allowing for the detection of group differences in slope between the dependent variable (brain activity) and steroid levels. Group-by-steroid level interaction effects on brain activity were investigated for progesterone, estradiol and ALLO in both menstrual cycle phases. Analyses of interaction effects for ISO and ISO/ALLO were restricted to the late-luteal phase due to the high number of values below LLOQ in the mid-follicular phase (72.4% in the PMDD group and 59.3% in controls).
Associations between emotion-induced brain activity and symptom severity in women with PMDD
In order to evaluate whether significant associations existed between brain activity measures and symptom severity, the linear relationship between brain activity and DRSP scores was tested voxel-wise using Randomise within a mask combining brain regions exhibiting increased task-related activation [Faces > Shapes] in PMDD women during the late-luteal phase (Table S2). Specifically, the total 21 symptom DRSP score and core PMDD symptoms (depression, irritability, affective lability and anxiety) DRSP scores were investigated. DRSP scores were averaged over days − 8 to -1 to reflect the late-luteal phase scanning window. DRSP scores were derived from ratings recorded over the screening period for the study.
For all permutation testing using Randomise, a total of 5000 permutations were performed to build up null distributions to test against. Results were considered significant at p < 0.05 Family Wise Error (FWE) corrected using the threshold-free cluster enhancement (TFCE) method [42]. Trend-level results were defined as 0.05 < p < 0.1 for FWE-corrected statistics. We report t-statistics and cluster results (anatomical location, location in functional networks according to Yeo et al. [43], size, and local maximum).