Table 7.
Systematic Review of PD Cluster Analysis Research.
| Reference (in year order) | # of Patients | Patient Variables Included in Clustering | Reduction, Preprocessing & Clustering Methods | Patient Clusters / Subgroups / Domains |
|---|---|---|---|---|
| Graham and Sagar (1999) | 176 | •Disease duration ?Age at onset ?Alternate Finger Tapping Test (A-FTT) ?UPDRS - Motor, ?Duration: time to dyskinesias, falls and fluctuations ?UPDRS daily activities ?Beck depression inventory (BDI) ?National Adult Reading Test (NART) ?UPDRS mentation ?Blessed dementia scale ?Digit ordering ?Letter fluency ?Cambridge Neuropsychological Test Automated Battery (CANTAB) |
Preprocessing: Normalized Clustering: K-means |
A total of 5 clusters as follows: Mean Duration: 5 years 1.Good motor control, No cognitive impairment 2.Good motor control, Executive cognitive deficits 3.Older onset age, Poor motor control and complications, Mild cognitive impairment Mean Duration: 14 years 4.Poor motor control, No cognitive impairment 5.Poor motor control, Moderately severe cognitive impairment |
| Gasparoli et al. (2002) | 103 | •UPDRS - Motor ?Motor fluctuations presence ?Dyskinesia presence |
Clustering: Unknown | 1.Slow course, Earlier onset age, Lateralization of parkinsonian signs, Rest tremor presence, Gait disturbance absent (61%) 2.Rapid progression, Older age onset, Absence of lateralization of parkinsonian signs, Predominance of bradykinesia-rigidity and gait disturbance (39%) |
| Dujardin et al. (2004) | 44 | •UPDRS - Motor, ?Stroop Colour and Word Test (SCWT) ?Alternating word fluency ?Grober and Buschke test (G&B test) ?Mattis Dementia Rating Scale (MDRS) ?Semantic word fluency |
Reduction: Principle Component Analysis (PCA) reduction to eight variables, Preprocessing: Standardized Clustering: K-means |
1.No cognitive deficit, Less severe motor symptoms (26 patients) 2.Reduced cognitive efficiency, Exacerbated subcorticofrontal syndrome, Severe motor dysfunction (16 patients) |
| Lewis et al. (2005) | 120 | •Age at onset ?Progression: Total - UPDRS / disease duration ?Dopaminergic therapy ?Motor phenotype: Tremor / non-tremor UPDRS ?Mini Mental State Examination (MMSE) ?NART ?Pattern recognition memory (PRM) ?Tower of London (TOL) ?BDI |
Preprocessing: Standardized Clustering: K-means K = 2 - 5 |
1.Young onset 2.Tremor dominant 3.Non-tremor dominant, Cognitive impairment, Mild depression 4.Rapid progression, No cognitive impairment |
| Schrag et al. (2006) | 124 | •Age at onset ?Current age ?Dementia presence ?Fluctuations presence ?Dyskinesias presence ?HY ?Progression: HY per year |
Clustering: K-means K = 2 - 5 |
1.Young (mean) onset, Higher depression scores, Taking higher levodopa doses 2.Older onset, Rapid progression, Less often dyskinesias and fluctuations |
| Post et al. (2008) | 131 | •Age ?Age of disease onset ?Progression: UPDRS - Motor / disease duration ?Levodopa responsive PD symptoms ?Levodopa non-responsive PD symptoms ?MMSE ?Hospital Anxiety and depressive symptoms (HADS) |
Preprocessing: Standardized Clustering: K-means K = 2 & 3 |
1.Young onset with slow progression 2.Intermediate age onset, More anxiety and depression; 3.Oldest onset, More motor impairment, Higher progression rate |
| Reijnders et al. (2009) | 346 (two cohorts, 224, 122) | •Tremor ?Hypokinesia / rigidity ?Postural-Instability-Gait-Difficulty (PIGD) ?L-dopa complications, ?Apathy and Hallucinations (Both from UPDRS) ?Age at onset ?Cognition: MMSE ?Depressive symptoms: Montgomery - Asberg Depression Rating Scale (MADRS) ?Progression: UPDRS - Total / disease duration |
Clustering: K-means |
1.Rapid progression 2.Young onset with motor complications 3.Non-tremor dominant and psychopathology 4.Tremor dominant |
| van Rooden et al. (2011) | Dutch cohort PROPARK1344, PROPARK2 276, Spanish cohort (ELEP) 357 |
•Motor Complications: Scales for Outcomes in PD (SCOPA) - Motor ?Cognitive functioning: SCOPA - COG ?Autonomic symptoms: SCOPA - AUT ?Psychiatric symptoms: SCOPA - PC ?Nighttime sleep problems and excessive daytime sleepiness: SCOPA - Sleep ?Depressive symptoms: PROPARK: BDI ELEP: HADS ?Postural-instability-gait-difficulty (PIGD) ?Freezing during on, speech, and swallowing (FOSS) ?Levodopa dose equivalent (LDE) |
Preprocessing: Linear regression residual values for disease duration, Transformation to z-scores Clustering: Model based |
1.Mild severity all domains, Younger onset age (largest cluster) 2.Severe, frequent motor complications, Moderately severe sleep and depressive symptoms, Youngest age onset, Large portion of women 3.Intermediate severity in nondopaminergic domains, Higher onset age 4.Severely affected most domains, except mild tremor, High onset age |
| Liu et al. (2011) | 138 | •Age ?Age at disease onset ?Disease duration ?H&Y ?Tremor, Rigidity, ?Hypokinesia, PIGD, ?L-dopa complications (all from UPDRS) ?Motor phenotype: Tremor / non-tremor UPDRS ?Cognitive function: MMSE ?Depression: HAM-D ?Pittsburg Sleep Quality Index (PSQI) ?Constipation ?Fatigue severity scale (FSS) ?UPDRS - Total / disease duration, ?UPDRS - Motor / disease duration, ?UPDRS Activities of Daily Living (ADL) / disease duration |
Preprocessing: Standardized Clustering: K-means K = 3 - 5 |
1.Non-tremor dominant 2.Rapid disease progression 3.Young onset 4.Tremor dominant (largest cluster) |
| Erro et al. (2013) | 100 | •Motor disability ?Tremor ?Bradykinesia ?Axial (all from UPDRS) ?Time of onset ?Progression: UPDRS / duration ?L-Dopa equivalent daily doses (LEDDs) ?MMSE, ?Frontal Assessment Battery (FAB) ?HADS ?Depression subscale (HADS-D) ?Anxiety subscale (HADS-A) ?Non Motor Symptoms Questionnaire (NMSQuest) |
Clustering: K-means Gower method K = 3 - 6 |
1.Benign pure motor 2.Benign mixed motor-non-motor 3.Non-motor dominant 4.Motor dominant |
| Lawton et al. (2015) | Oxford Parkinson Disease Centre (OPDC) Discovery Cohort 769 |
•MDS - UPDRS Parts I and III ?Sniffin’ Sticks 16-item odour identification Test ?Big Five Inventory (BFI) extraversion scale ?Epworth Sleepiness Scale (ESS) ?Rapid eye movement sleep behavior disorder questionnaire (RBDSQ) ?Leeds scales for the self-assessment of anxiety and depression ?BDI, ?Questionnaire for Impulsive-Compulsive ?Disorders in Parkinson’s Disease (QUIP) ?Honolulu Asia Aging Study Constipation Questionnaire ?Montreal Cognitive Assessment (MoCA) ?Phonemic and semantic verbal fluency ?Purdue Peg-board Test ?Timed Up and Go Test ?Flamingo Balance Test ?Orthostatic blood pressure measurement |
Reduction: Confirmatory factor analysis Clustering: K-means |
1.Mild motor and non-motor disease 2.Poor posture, gait, cognition, smell and postural hypotension, 3.Severe tremor 4.Poor psychological well-being, RBD and sleep 5.Severe motor, non-motor and cognitive disease, with poor psychological well-being |
| Ma et al. (2015) | 1,510 | •Age of onset ?Disease progression, ?Stage (HY) ?Motor evaluating scores: UPDRS III, tremor, hypokinesia, rigidity, PIGD ?Non-motor evaluating scores: cognition: MMSE depression: HAM-D sleep disorder: PSQI constipation scores |
Preprocessing: Standardized Clustering: K-means K = 3 - 5 |
1.Non-tremor dominant (NTD, n=469, 31.1%); 2.Rapid progression with late onset (RDP-LO, n=67; 4.4%); 3.Benign pure motor characteristics (BPM, n = 778; 51.5%), without non-motor disturbances; 4.Tremor dominant with slow progression (TD-SP, n = 196; 13.0%). |
| Fereshtehnejad et al. (2015) | 113 | •UPDRS Parts II, III ?Orthostatic hypotension ?Mild cognitive impairment (MCI) ?RBDQ ?Depression ?Anxiety |
Preprocessing: Transformation to z-scores Clustering: Two-step |
1.Mainly motor/slow progression 2.Diffuse/malignant 3.Intermediate |
| Erro et al. (2016) | Parkinson’s Progressive Marker Initiative (PPMI) 398 | •Gender ?Age at onset ?MoCA ?Geriatric Depression Scale (GDS) ?State-Trait Anxiety Inventory for Adults (STAIT) ?University of Pennsylvania Smell Identification Test (UPSIT) ?RBDSQ ?SCOPA - AUT ?Tremor, bradykinesia, motor, rigidity, axial, apathy, fatigue, hallucinations, pain presence (all from MDS - UPDRS) |
Clustering: K-means Gower method K = 3 - 6 |
1.Lowest motor and non-motor burden 2.Motor disability 3.Motor disability, apathy and hallucinations |
| van Balkom et al. (2016) | 226 | •UPDRS Part III ?MMSE ?SCWT ?Trail Making Task (TMT) ?Wechsler Adult Intelligence Scale (WAIS)-III ?Dutch version of the Rey Auditory Verbal Learning Test ?Beck Anxiety Inventory (BAI) |
BAI, UPDRS-III, MMSE transformed to z-scores Clustering: Hierarchical |
1.Young-age (59.9 years) Mildly affected, (N ¼ 86) 2.Old-age (72.3 years), Severe motor and non-motor symptoms (N ¼ 15) 3.(age 64.7 years) Mild motor symptoms, Below-average executive functioning and affective; symptoms (N ¼ 46) 4.(age 64.8 years), Severe motor symptoms, affective symptoms and below-average verbal memory (N ¼ 79) |
| Fereshtehnejad et al. (2017) | PPMI 421 | •Age ?Genetic risk ?Orthostatic systolic ?blood pressure drops ?MDS - UPDRS Parts II and III ?Tremor, PIGD (MDS - UPDRS) ?ESS ?GDS ?STAIT ?QUIP ?RBDSQ ?SCOPA-AUT ?UPSIT ?Visuospatial, speed / attention, memory, executive function cognitive |
Reduction & Preprocessing: Normative values, Composite indicators for redundant variables, Transformation to z-scores Clustering: Hierarchical |
1.Mild motor-predominant (composite motor and all three non-motor scores below the 75th percentile) (223) 2.Diffuse malignant (composite motor score plus either ≥1/3 non-motor score >75th percentile, or all three non-motor scores >75th percentile) (52) 3.Intermediate (146) |
| Mu et al. (2017) | 904 2 Cohorts as shown below: 1. Non-Motor Symptoms Scale (NMSS) 2. Non-Motor International Longitudinal Study (NILS) |
•HY ?NMSS ?Clinical Impression of Severity Index for PD (CISI-PD) ?SCOPA - Motor |
Clustering: K-means |
1.Mildly affected all domains (largest) 2.Severely affected non-motor domains 3.Severely affected motor domains 4.Severely affected all domains, except tremors |
| Lawton et al. (2018) | Tracking cohort 1,601 Discovery cohort 944 |
•MDS - UPDRS ?Big Five Inventory ?ESS ?RBDSQ ?HADS ?Questionnaire for Impulsive-Compulsive Disorders in PD ?Honolulu Asia Aging Study Constipation Questionnaire ?MoCA ?Orthostatic blood pressure measurement ?Sniffin’ 16 odour identification scores |
Reduction: Factor analysis Clustering: K-means |
1.Fast motor progression, Symmetrical motor disease, Poor olfaction, cognition and postural hypotension, Highest age at diagnosis 2.Mild motor and non-motor disease, Intermediate motor progression, Youngest onset, Large portion of women 3.Severe motor disease, Poor psychological well-being, Poor sleep, Intermediate motor progression 4.Slow motor progression, Tremor-dominant, Unilateral disease |
| Krishnagopal et al. (2020) | PPMI 194 | •Gender ?Age ?MDS - UPDRS ?Benton Judgement of Life Orientation (JLO) ?Symbol Digit Modalities Test (SDMT) ?MoCA ?Hopkins Verbal Learning Test (HVLT) ?Letter number sequencing ?Semantic Fluency Test ?Schwab and England (S&E) ?ADL ?RBDQ ?ESS ?SCOPA - AUT ?GDS ?STAIT |
Preprocessing: Transformation Clustering: Trajectory Profile Clustering (TPC) |
1.Mixed subtype, Severe impairment of motor and autonomic function, mental health and sleep, good cognition, at baseline and over time, Young and predominantly female 2.Mild subtype, Milder impairment in all domains (motor, cognitive, autonomic and mental) at baseline and over time 3.Severe subtype, Worse impairment in all domains, Older than average, Predominantly male |
| Belvisi et al. (2021) | 100 | •Primary motor cortex (M1): Transcranial Magnetic Stimulation (TMS) ?Motor: fast index finger abduction ?Sensory: Somatosensory Temporal Discrimination Threshold (STDT) |
Clustering: Agglomerative Hierarchical |
1.Mild motor predominant (younger, milder motor, nonmotor symptom severity) n = 76 2.Diffuse malignant (more severe motor, nonmotor manifestations) n = 24 |