In this paper we have introduced the problem of participants’ symptomatic variability in between-subject studies of PD. We revealed the responses to dopaminergic treatment in the basal ganglia, in accordance with previous fMRI work from Kraft et al. 
, Holiga et al. 
and a positron emission tomography study by Feigin et al. 
. Furthermore, we demonstrated the results of two distinct approaches revealing this variability by employing the UPDRS-III scores, which remarkably influenced the activation patterns.
A fair number of functional imaging studies examining motor circuitry of PD 
embodied the UPDRS-III in statistical analyses to reveal correlations between the clinical presentation of the disease and the blood-oxygen level dependent (BOLD) signal amplitude (). Correspondingly and noticeably, the UPDRS-III ‘in’
approach presented here identified a strong relationship between the total UPDRS-III score, midline subscore and BOLD responses in the basal ganglia regardless of the treatment condition. This finding alone should motivate the inclusion of the scores in future motor fMRI studies. The majority of previous studies did not consider the heterogeneity of PD symptoms in the analyses at all () 
. In addition to examining correlations, we also took advantage of the UPDRS-III scores by means of weeding out variability in the measured data originating from symptomatic deviations within/between subjects (UPDRS-III ‘out’
). This way we were able to equalize particular clinical symptoms between the investigated patients intra/inter-individually, thus ‘simulated homogeneity’ with respect to a certain symptom. To our knowledge, this and our previous fMRI study 
are the only studies which considered using UPDRS-III in the analyses to account for the symptomatic variability of PD. Here, akinesia and rigidity were demonstrated as primarily responsive to levodopa treatment, as documented by the scores (, bottom bar-plot). Thus, incorporating them quantitatively in statistical prediction of the group response achieved the uppermost sensitive activity pattern representing the patients’ response to levodopa.
The summary of fMRI studies investigating motor deficits in Parkinson’s disease.
Several interesting findings which weren’t obvious when using the conventional analysis emerged when evaluating UPDRS-III models. As we used the hemibody specific symptom scores in all UPDRS-III models except in the laterally unspecific total UPDRS-III and midline score, we systematically equalized the differential lateral involvement of all symptoms in our sample. Note the particular heterogeneity of lateral involvement of PD in our sample (Table S1
). Accounting for the general hemibody score () delivered a significant difference between the treatment conditions in the left basal ganglia. Therefore we might speculate that some of the symptoms reacting to levodopa were expressed more in the contralateral right side of the body. In case the statistical power is sufficient, this might eventually suggest splitting the study group further according to the lateral dominance of PD symptoms, or explore the results in more detail using more specific subscores.
In the present patient’s group, equalizing the akinesia for each hemibody revealed laterally unspecific and more sensitive levodopa modulation of activity in BG, confirming that our patients were mainly affected by akinesia, and had improvement in both body parts when treated with levodopa (). When assessing the results of both UPDRS-III approaches accounting for rigidity, we observed an overlap between the results of both UPDRS-III ‘out’
and UPDRS-III ’in’
approaches in the left basal ganglia (). We may conclude that this area particularly reflects rigidity and simultaneously exhibits sensitivity to levodopa treatment. The asymmetry in basal ganglia activation observed with the ON-OFF medication contrast when considering the hemibody score might be explained by left/right asymmetry in the rigidity score. This is in agreement with the previously observed higher synaptic dopamine increase in the more affected hemisphere when levodopa was administered 
. Higher activation in the left basal ganglia may then reflect a higher reaction to treatment because of higher expression of rigidity on the right side extremities. Indeed, six patients in our study had rigidity expressed predominantly on the right hemibody, three patients on the left side and for three patients it was manifested symmetrically. All this is particularly interesting, because with a conventional approach alone the sensitivity would be considerably lower and the model would never reveal relationships between various symptoms, laterality or the effects of treatment.
Since PD is considerably heterogeneous, we advocate systematically checking for scores and subscores unquestionably related to the investigated sample using both proposed approaches. This might reveal activity patterns specific to the individual aspects of the disease and potentially lead to unforeseen findings due to increased sensitivity, or suggest further dividing the investigated group of patients in subgroups and analysing them separately. In our case, all patients were categorized as akinetic-rigid, therefore akinesia and rigidity subscore delivered the most sensitive group response. Hence, the proper choice of regressors strongly depends on the population sample studied. It is beneficial to study the outcome of the clinical measures and its variance separately and select the proper UPDRS-III ‘in’ and/or UPDRS ‘out’ approach using the particular score or subscore accordingly to research question asked. Moreover, statistical limitations regarding the proper covariate choice must also be considered with this type of analyses. With the exception of one pair, we observed a high degree of correlation between all scores. In a potential multi-score design involving several correlated regressors, besides the reduced degrees of freedom, this would lead to inefficient parameter estimates with high variance, and the incapability to correctly attribute the effect of a particular score to the model fit.
This work is aimed at underlining the strong relationship between the BOLD response and the clinical severity of the disease, but also the importance of considering the intra/inter-subject variability, even in a pre-unified group of PD patients. Because clinical heterogeneity is not clearly defined and is still a matter of debate 
, we advocate using the proposed approaches as leverage for prospective studies involving any group of PD participants for personalizing the statistical evaluations considering the various clinometric involvement. Depending on the research question asked, suitable aspects of UPDRS-III scores can be selected and incorporated in analyses when using UPDRS-III ‘in’
or UPDRS-III ‘out’
approaches, to obtain more reliable statistical inferences allowing for unbiased comparisons of results between studies.