Consecutive cohorts were combined from three different studies of the same principal investigator in which the same inclusion and exclusion criteria, imaging protocols, and PET system were used, and a total of three investigators for the clinical scales were involved. Based on power calculations using previous findings on DAT, a sample of 40 patients should be able to measure a correlation coefficient of 0.4 with power 0.73 (α = 0.05). The studies were approved by the Swedish Ethical Review Authority, the Radiation Safety Committee of the Karolinska University Hospital, and the Swedish Medicinal Product Agency (EudraCT 2011-0020050, EudraCT 2017-003327-29, and 2017-001585-19). Written informed consent was given by the participants. The study was conducted in accordance with the ethical standards of the Declaration of Helsinki.

Demographic, clinical, and baseline PET data of the first cohort of PD patients (n = 20) have been previously reported [21]. The second cohort (n = 27) includes a group of eleven PD patients that were examined in a previously reported test–retest study [15] and sixteen PD patients that were prospectively recruited.

Patients with idiopathic PD, age 45–80, and H&Y stage < 3, were recruited at two local outpatient clinics for movement disorders and through advertisement at the Swedish Parkinson Foundation. Cognitive impairment was excluded through Mini-Mental State Exam (MMSE) ≥ 27; psychiatric comorbidity through structured psychiatric interview; clinically significant physical comorbidity through physical examination, ECG, routine urine and blood tests; significant intracranial deviations through structural MRI. Additional file 2: Table S1 shows summarized exclusion criteria and rules of conduct minimizing possible confounding factors during DAT measurements.

Symptom duration was approximated from the reported onset of motor symptoms. The modified Hoehn and Yahr staging scale was used. In the first cohort, Unified Parkinson’s Disease Rating Scale motor score (UPDRS-III) had been performed at the screening visit while on Parkinson medication for practical reasons. In the remaining patients, the Movement Disorder Society—Unified Parkinson’s Disease Rating Scale motor score (MDS-UPDRS-III) was used and assessed while in practically defined ‘OFF’ state on the PET day, i.e. at least 12 h off dopaminergic medication, or 24 h in case of the longer-lasting Parkinsonian medication. This last improvement in the protocol was implemented to be able to perform correlation analyses between clinical measures and DAT availability in ‘OFF’ state, considering the effect of Parkinsonian medication on the motor scores.

Healthy controls (HC) were recruited and sex and age-matched (± 5 years) to the included patients with Parkinson’s disease, to correct on a group level for age- and sex-related effects on DAT availability when comparing disease-related differences to healthy controls.

T1- and T2-weighted images were acquired with a 3 Tesla MRI system (General Electric, Discovery MR750). MRI acquisition was done as previously described [15, 21], see Additional file 1 for details. The T1-weighted images were used for co-registration with PET images and delineation of the regions of interest (ROI); the T2-weighed images were used to exclude clinically significant pathology.

All PD subjects included were in practically defined ‘OFF’ state during PET acquisition. An individually made plaster helmet was used for head fixation in the PET camera. [18F]FE-PE2I was produced in-house as previously described [22]. PET acquisition and reconstruction were done as previously described [15]. In short, list mode PET data were acquired for 93 min with a high-resolution research tomograph (HRRT, Siemens Molecular Imaging). A 6-min transmission scan using a 137Cs source was performed before radioligand administration for attenuation correction. PET data were reconstructed into 37 frames of increasing duration (8 × 10, 5 × 20, 4 × 30, 4 × 60, 4 × 180, 12 × 360 ms) using 3D OP OSEM with 10 iterations and 16 subsets, including modelling of the point spread function, with a transaxial resolution of ~ 2 mm [23]. The first two minutes of data were used as reference image for frame-to-frame realignment. Frame-to-frame realignment was performed using statistical parametric mapping (SPM, version 12). Realignment plots were evaluated visually to exclude excessive head motion. In the case of translations of more than 3 mm in the z-axis, additional motion correction was performed.

The analysis plan was pre-registered at https://aspredicted.org/f3rq6.pdf. Three-dimensional (3D) regions of interest were automatically delineated on the T1-weighted MR image and co-registered to PET. The main striatal regions striatum, caudate nucleus, putamen, and accumbens area (in this manuscript referred to as ventral striatum; VS) were delineated using Freesurfer. An FSL template (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases/striatumconn) was used for the delineation of the sensorimotor striatum (Fig. 1), and an in-house developed template [21] was used for the delineation of the substantia nigra.

3D Functional subdivisions of striatum using the FSL-template overlaid on T1-weighted MRI sequences, with the sensorimotor striatum displayed in red. Figure previously published as Fig. 1 in Kerstens et al., Clin Transl Imaging 2020. Republication permitted under CC BY licensing

DAT availability (BPND) was estimated with wavelet-aided parametric imaging (t* = 27 min), using the Logan graphical analysis and cerebellum as reference region, as described earlier [15]. BPND (binding potential) is calculated as distribution volume ratio (DVR) minus 1: DVR-1.

ROI sides were assigned as belonging to the more or less affected hemisphere based on the individual’s BPND in the putamen sides, with the lower putamen BPND being assigned as the more affected hemisphere (maff), and the higher BPND as the less affected hemisphere (laff). These sides were cross-checked with the laterality of the MDS-UPDRS-III scores.

Statistical analysis was performed with R version 3.6.2. The normality of distribution was tested with the Shapiro–Wilk test and density plots. Testing for the difference in variance between the PD and control group was done with Levene’s test. Identification of statistically significant outliers was tested with Grubb’s test, combined with boxplots for visual assessment. The difference between the mean BPND of control subjects and PD subjects was tested with an unpaired, one-sided t test. The calculation of the effect size of BPND differences between controls and PD patients was done with Cohen's ds, with the formula: (mean(controls)-mean(PD))/SD(controls and PD).

The asymmetry index (AI) in PD was calculated per ROI as:

$$\left( }_}}} } - }_}}} }} \right)/\left( }\;}_}}} }\;}\;}} \right)*00.$$

As reference asymmetry index, the upper limit of the 95% CI of the control group was used. The 95% CI was calculated as follows:

$$}\left( }\;}\;}\left[ }} \right]} \right)\, \pm \,qt\left( }\, = \,\left( } \right)} \right)*\left( }\left( }\;}\;}\left[ }} \right]} \right)/}\left( n \right)} \right)} \right),$$

$$}\;t\, = \,0.\;(t\, = \,}\, + \,\left( - }} \right)/).\, }\, = \,0.\;}\;n\, = \,.$$

which takes the sample distribution into account instead of assuming a normal distribution.

Simple linear regression was used for the correlation analysis between DAT availability and clinical measures, as described in the pre-registration. In the case of categorical or ordinal variables (such as H&Y stage), a generalized linear model (GLM) was used. LEDD (levodopa equivalent daily dose), age, and sex were tested as covariates.

A priori hypotheses had a p value threshold of < 0.05 for statistical significance; other comparisons were adjusted for multiple comparisons with Meff-correction [24, 25].

Data are available upon reasonable request.