Results

Table 2 presents, for each group, the median, mean, and range obtained for each score.

As shown in table 2, these bivariate analyses did not allow isolation of a set of variables differentiating the two subject groups, as the distributions of only one variable (the dynamic organisation part of the motor dynamic sequences task) were significantly different (p=0.03). A multivariate analysis was then conducted.

• Download figure
• Open in new tab
• Download powerpoint

Projection of members of both groups on the discriminant score. PD=Parkinson’s disease.

The results of the factorial discriminant analysis showed a relatively moderate quality of discrimination with a square correlation ratioR ²=0.42. However, our aim was not to completely separate both populations but to look for those, among the first degree relatives, whose cognitive profile could be considered to be different from those of the healthy controls.

The median value of discriminant score was 27.4 (mean=27.6, range 25.7–29.4) in the first degree relatives and 25.9 (mean=25.9, range 23.9–27.5) in the healthy controls. The projection of the persons of both groups on the discriminant score, as represented in the figure, clearly showed two populations among the first degree relatives.

By examining this projection, a threshold (discrimant score=28) could be identified. When the discriminant score was below this threshold, the healthy controls and the relatives populations overlapped. When the discriminant score was higher than this threshold, there were no more healthy controls but a group all from the population of first degree relatives. Examination of the correlations between the observed variables and the discriminant score allowed identification of some variables which best explained such typology. They were: the score on the dynamic organisation part of the motor dynamic sequences task (Spearman correlation coefficientr=−0.44), the “L” fluency (r=−0.34), and the percentage of “L” (r=0.33) and “M” perseverative namings (r=0.33). The better the scores on the dynamic organisation part of the motor dynamic sequences and the higher the “L” fluency, the higher was the probability for a person to belong to the healthy controls and the “non-deviant” first degree relatives. The higher the percentage of “L” and “M” perseverative namings, the higher was the probability for a person to belong to the deviant first degree relatives.

Other variables were also correlated with the discriminant score. However, they were less useful for separating the two populations. A stepwise discriminant analysis was performed to simplify the discriminant score by removing some variables. The square correlation ratio of this analysis was low. This was not surprising because the bivariate analyses could only isolate one variable and the factorial discriminant analysis failed to completely separate both populations. Consequently, all variables were used for the following analyses.

To explain the observed typology, the performances of the two groups as shown by the factorial discriminant analysis were examined.

To standardise the scores on each variable, the raw scores were converted to z scores (raw score−mean/SD), on the basis of the scores from the healthy control group.32 This was done in such a way that positive scores always indicated performance better than the control mean. The sum of these z scores was computed to obtain a global executive score. This global executive score included all tests except the Mattis dementia rating scale and the VIQ, which served as exclusion criteria and did not provide an evaluation of the executive function. This method of analysis gives equal weight to each component of the summary score. Similar methods of data reduction and analysis have been reported by other authors.33 34 In the healthy control group, the normality assumption of the global executive score was not rejected (Shapiro-Wilk test, p=0.55). Consequently, those whose global executive score was 2 SD below the mean score of the healthy control group were identified as impaired persons. By this procedure, eight first degree relatives were isolated. Among them, five were identified as “deviant” by factorial discriminant analysis. The three others were just at the limit of the cut off threshold. An analysis of the individual scores of these subjects disclosed that they performed globally worse than the other subjects on all task variables. However, the factorial discriminant analysis also considered as deviant 11 other people. Among them, four were comparable with the previous subjects. Although their global executive score was not below 2 SD of the mean score of the healthy control population, they performed globally at a lower level on all task variables. The seven other subjects cannot be considered as impaired on the basis of their global executive score. The analysis of their individual scores showed that all of them, except one, presented a particular cognitive profile. All had mean or even high scores on all task variables, except on those highly correlated with the discriminant score. All had low performance on the dynamic organisation part of the motor dynamic sequences, low “L” fluency, and made more “L” and “M” perseverative namings. One subject showed a different profile: he only made more perseverative namings on the “M” word fluency test.

Thus, according to the factorial discriminant analysis, among the 41 first degree relatives, 15 could be considered as cognitively deviant. Among them, nine presented with a global executive impairment. The six others only showed a limited executive dysfunction.

Discussion

The present study showed that among first degree relatives of patients with Parkinson’s disease, some manifest executive dysfunction comparable with that typically associated with early Parkinson’s disease.

Such impairment cannot be related to global cognitive deterioration. Indeed, all the subjects who performed below the 20th percentile of the reference population on the Mattis dementia rating scale22and whose VIQ was below 85 were excluded from the present study. Moreover, examination of table 1 clearly shows that most of the included subjects performed largely above these exclusion criteria. In addition, as the same criteria were applied to healthy controls, if the cognitive impairment shown by the subjects was related to a global cognitive deterioration, controls could also be discriminated by the factorial discriminant analysis. Now, the discriminant score showed a clear limit between the healthy controls and the deviant first degree relatives.

An alternative hypothesis is that the typology found could be explained by age. Such an argument can also clearly be discarded. Indeed, among those considered as deviant by factorial discriminant analysis, age varied between 24 and 64 years (mean 40.76 (SD 14.32), range 24–64). This subgroup was then representative of the initial sample. Moreover, as first degree relatives and healthy controls were matched for age and education, if age was an explanatory factor the factorial discriminant analysis could also discriminate healthy subjects as deviant. This was not the case.

The discriminated people thus manifested specific cognitive impairment in tasks evaluating executive function. Such impairment has largely been considered as reflecting dysfunction of the striatofrontal system.25 35-38 It can then be suggested that, although they are clinically normal, some people considered as being at risk for developing parkinsonism, already present dysfunction of the processes depending on the striatofrontal loop. As the impairment found typically resembles those seen in Parkinson’s disease, even early in the course of the disease,12-15 25 39 40 can such cognitive signs indicate preclinical parkinsonism? The discriminated persons were distributed into two categories: nine subjects who manifested a generalised dysexecutive syndrome and performed below the healthy controls on all tasks and six subjects who manifested poor performance only on some particular tasks. In the first category, given the diversity of the tasks administered, it can be argued that the impairment concerns most of the cognitive operations dependent on executive function and might correspond to a preclinical stage of the disease. These people represent 22% of the initial sample. This proportion is then very close to that described by Picciniet al 11 when measuring striatal¹⁸F dopa uptake. In the second category, consideration of the processes involved in the impaired tasks showed that all required abilities to alternate between categories of response. Moreover, this shifting was not dependent on an external signal but required internal control. It can then be argued that the supervisory attentional system of the model of attentional control of Norman and Shallice41 is largely involved in such tasks. It can then in turn be hypothesised that the subjects of the second category presented some “fragility” of their supervisory attentional system involving impairment in a limited number of tasks highly dependent on the supervisory attentional system. As the impaired tasks also required complex motor production, an alternative explanation of the findings could be that these subjects present a breakdown in motor and articulatory processes. This would be in agreement with the results of Puccini et al showing reduced putaminal¹⁸F dopa uptake in subjects at risk of developing Parkinson’s disease.11 However, the predictive value of suchimpairment has to be determined by a longitudinal survey of these subjects. Indeed, the duration of the preclinical period of Parkinson’s disease is variable. On the basis of extrapolation of pathological and PET data, it was estimated to vary from 3 to 5 years42 43 to 40 to 50 years.44 45Moreover, the high interpatient variability in the clinical rate of progression of the disease46 further enhances the variability of the preclinical phase. Among the 15 deviant first degree relatives, none had actually developed Parkinson’s disease and four presented minimal motor signs at clinical examination. Unilateral mild slowing or reduction in amplitude of finger taps was found in two of them and slight rigidity activated by mirror movement was found in the two others. The two people with unilateral mild slowing or reduction in amplitude of finger taps belong to the second category of deviant subjects, although the two others belong to the category of globally impaired subjects. It is then possible that both categories of people identified here represent different rates of evolution of the preclinical period. However, the relevance of finding executive dysfunction in clinically normal subjects still has to be determined, as has the predictive value of the association of isolated motor signs and executive dysfunction.

The present study thus shows that cognitive impairment without obvious motor disturbances could represent a preclinical form of Parkinson’s disease in relatives of patients with familial disease. The nigrostriatal dysfunction shown by PET9-11 can then induce a dysexecutive syndrome before the identification of motor signs. A longitudinal survey of these first degree relatives of patients with Parkinson’s disease is in progress to ascertain the predictive value of these cognitive changes.

Moreover, the discovery of cognitive dysfunction in first degree relatives of some patients with apparently sporadic Parkinson’s disease is interesting. Indeed, the evidence of neuropsychological abnormalities close to those found here could then be interpreted as a phenotypic variant of the disease, demonstrating that familial Parkinson’s disease is more frequent than usually estimated.