We read with interest the report by Lewis et al¹ on clinical heterogeneity of Parkinson’s disease based on a data driven approach. In their cohort of patients with Parkinson’s disease they identified different subgroups within their population, including those with relatively early onset, those with a tremor dominant subtype, those with non-tremor dominant subtype with cognitive impairment and mild depression, and those with rapid progression without cognitive impairment. We undertook a similar analysis—which was originally conducted before the publication by Lewis et al—in a previous community based prevalence study in the London area, the details of which are described elsewhere.² The clinical features of 124 patients with previously proposed clinical subtypes of Parkinson’s disease were compared, and a cluster analysis entering all variables that had differed between these groups undertaken. Age at onset, current age, presence of dementia, fluctuations, dyskinesia, and rate of progression all differentiated patients within this sample of individuals with Parkinson’s disease, whereas sex, predominance type, symptoms at onset, family history, and occurrence of depression or hallucinations were not useful discriminators. K-means cluster analysis with two to five cluster solutions, however, revealed only two meaningful subgroups of patients (table 1): (1) patients with young mean onset and current age, who had higher depression scores and were taking higher doses of levodopa; (2) patients with older onset who had more rapid progression, but less often had motor fluctuations (p<0.001) and dyskinesias (p<0.05). The three cluster solution (table 2) showed that in the older subgroup there were two further subgroups: (a) patients with higher age of onset and lower cognitive scores, with an apparently more rapid disease progression on a relatively low dose of levodopa; and (b) patients with slightly lower age, no cognitive impairment, and slower disease progression who were also on a higher dose of levodopa. The latter group more often had motor fluctuations (p<0.001) and hallucinations (p<0.05). Thus in our sample we could not confirm the existence of a distinct subgroup of patients with tremor dominant presentation, or a subgroup without cognitive impairment but with rapid disease progression. However, our results confirm the finding of a young age at onset as a phenotype distinct from older onset disease, which in our slightly older sample was also associated with more rapid progression. Older patients who were on higher doses of levodopa had a slower decline of function, most probably because of the effect of treatment, but they also had higher rates of complications.

As Lewis et al discuss, cluster analysis is very dependent on the variables that are entered into the analysis and the populations studied, and necessitates cautious interpretation of such data. For example, the use of different numerators to calculate the rate of clinical progression (UPDRS v Hoehn and Yahr) or the detail and number of neuropsychological tests (greater in the study by Lewis et al) may have influenced the results of the cluster analysis. In addition, our population was from a community based prevalence study with a higher average age (mean (SD): 72 (10.9) v 64 (SD 9.3) years). These differences may at least in part explain the difference in the two studies. Another explanation for the lack of a clear subgroup with tremor dominant Parkinson’s disease may be that we excluded those in whom the diagnosis was uncertain as they had only tremor (four cases). However, the finding that patients with a younger age of onset and an older age of onset appear to belong to different subgroups in two different samples supports the validity and robustness of this result. The data confirm studies with univariate analysis from previous studies, including the slower disease progression and higher rate of motor fluctuations and dyskinesias in younger onset patients. However, it should also be noted that the levodopa dose in both studies was highest in the younger onset group. The findings of different characteristics in young onset patients are also supported by the increasing evidence from genetic and epidemiological studies showing a stronger genetic component in younger onset Parkinson’s disease than in older onset cases. While the younger onset group in both studies was older than “young onset” defined as onset before age 40 years, the different clinical behaviour may at least partly be influenced by genetic contributions in the younger onset group, or, alternatively, by superimposed aging in the older onset phenotype. The finding of different clinical subgroups within Parkinson’s disease, particularly related to age of onset, by this study and that by Lewis et al therefore warrants further study into the genetic or age related causes of these subtypes.