Parkinson's ex machina

Last week Anna talked about the need to ensure the 'human' is kept at the centre of drug design for Parkinson's treatments - "focussing on human cells, tissues and subjects". Today I also want to discuss the role of humans in Parkinson's and the role of machines.

Neurology (and neurologists) take pride from being a bastion of clinical medicine - the heart and soul of our practice is listening to our patient and observing their every movement. In some ways neurology has not moved on much (in spirit) from the historic view of 'physicians' portrayed so poignantly by Sir Luke Fields. Movement disorders, and the diagnosis and management of Parkinson's are perfect examples of this human approach, with no equipment other than a tendon hammer necessary to make the diagnosis.

taken from https://www.tate.org.uk/art/artworks/fildes-the-doctor-n01522 with thanks

However, today's world has moved on leaps and bounds, and medicine and neurology are desperately trying to keep pace. In the 130 years since this scene was painted, our world has changed almost unrecognisably (although a Google image search of 'doctor' would make anyone believe that it is still delivered primarily by middle aged white men - which thankfully is less true now than ever).

It is old hat to say that technology continues at a blistering pace. Miniaturisation  has made yesterday's science fiction into the everyday. 'The Fantastic Voyage' described a vessel that was shrunk and inserted into a person's body to save them - well thrombectomy for stroke, stunting for heart attacks, and a video camera so small you can swallow it to get diagnostic pictures throughout the gut are all in routine clinical practice in the NHS today.

The Fantastic Voyage (1966) - picture from https://www.britannica.com/topic/Fantastic-Voyage

High street technology has revolutionised our lives. There is more processing power in my 4 year old iPhone than Armstrong, Aldrin and Collins had on Apollo 11. There is now an enormous market in wearable technology that will give you a beat-by-beat analysis of your heart rate, count your steps in a day, your calorie output, measure your power in watts on a bike, feet climbed on the stairs and just about any other metric you can imagine.

Just a small number of the dizzying array of activity tracking watches available on the high street (picture from  Groupon.co.uk)

All this brings me onto a review I have read, "Optimizing clinical assessments in Parkinson's disease through the use of wearable sensors and data driven modelling"

The article offers a review on some of the hardware software and statistical approaches to understanding the data. At its heart is the understanding that a neurologist sees their patient for perhaps 3 or 4 15-20 appointments in a year. That clearly leaves 8,759 hours per year where the neurologist has no ability to observe their patient move. Wearable technology brings us a (relatively) non-intrusive window into the 'real lives' of our patients. With the vast amounts of data that this can bring, we need very sophisticated algorithms to understand it, and we need to understand the computer science and mathematical constructs that underlie those algorithms.

Going further than mere observation is prediction - the holy grail in healthcare, and our clearly stated aim in PREDICT-PD. This article concludes by cautioning us about the variation in quality and output of this technology, but offering us the hope that these devices and underlying data analysis will improve clinical decision making, provide more personalised care, and better understand and highlight the differences in the manifestations of what we currently call 'Parkinson's disease', leading to better trials, treatments and one day - cure.

RNR

Plain English - LRRK2 G2019S Parkinson's disease with more benign phenotype than idiopathic

This research study compared patients with Parkinson's who carry the commonest gene mutation associated with the disease (called LRRK2), to patients with Parkinson's who do not carry this mutation. They found that on average patients with the mutation tended to have a milder form of Parkinson's (i.e. they had lower motor impairment scores and better scores on memory and other cognitive tests). 

Even though mutations in LRRK2 are the commonest genetic cause of Parkinson's, they are still pretty rare overall and affect about 1% of all patients with Parkinson's. This proportion is higher in North Africans (such as those in this study).

These findings in many ways match ones clinical impression as well. I would say on average patients with LRRK2 Parkinson's tend to have milder symptoms and a lower burden of non motor symptoms. These features may also be less prominent in the phase of the disease before diagnosis compared with Parkinson's that is not related to LRRK2. Of course, this is all 'on average'. On an individual basis it can be very tricky to choose between patients that do or don't have an underlying genetic cause, unless they are particular young when it starts or have a strong family history....

- Alastair Noyce

Acta Neurol Scand. 2018 Jul 10. doi: 10.1111/ane.12996. [Epub ahead of print]
Ben Romdhan S, Farhat N, Nasri A, Lesage S, Hdiji O, Ben Djebara M, Landoulsi Z, Stevanin G, Brice A, Damak M, Gouider R, Mhiri C.

https://onlinelibrary.wiley.com/doi/abs/10.1111/ane.12996

OBJECTIVES: The LRRK2-G2019S mutation is the most common cause of Parkinson's disease (PD) in North Africa. G2019S-PD has been described as similar to idiopathic with minor clinical differences. The aim of this study was to determine the G2019S-related phenotype and to investigate gender and gene dosage effects on clinical features of G2019S carriers.

PATIENTS AND METHODS: The G2019S mutation was screened in 250 Tunisian patients with PD. Twenty-four patients carrying mutations in other PD genes were excluded. Logistic regression models were used to compare clinical features between the studied groups.

RESULTS: G2019S carriers (107 cases) and non-carriers (119 cases) were similar in disease duration, levodopa doses, and gender and phenotype distributions. However, carriers had a younger age at examination, higher level of education, and were more likely to report family history of PD and to develop PD at earlier age (P = 0.017). Adjusted for age, sex, disease duration, levodopa-equivalent dose and educational level, MMSE scores remained significantly higher (adjust P = 0.019) and UPDRS-III scores were lower (adjust P = 0.012) in the G2019S carriers than non-carriers. Demographic characteristics of men and women with G2019S mutation were similar, but men had higher level of education, better cognition (adjust P-value for educational level = 0.042) and less tendency towards depression than females (adjust P = 0.046). Furthermore, PD phenotype did not differ between the homozygous and heterozygous G2019S carriers.

CONCLUSION: In this study, G2019S carriers had a more benign phenotype than non-carriers. Cognitive impairment and depression were less common in G2019S male carriers compared with females. In addition, we found that LRRK2 gene dosage does not influence the severity of PD.

Using Smartphones and Machine Learning to Quantify Parkinson Disease Severity: The Mobile Parkinson Disease Score

Smart phone technology and app-based approaches for measuring PD symptoms and signs have been around for some time and there are seemingly innumerable companies and groups that all suggest that their approach is the best...

The results of this recent study published in JAMA Neurology do however look really impressive. I have always felt that given the heterogeneity of PD (motor and non-motor), a good measuring device will take account of multiple domains. This app measures voice, gait, balance, reaction time and finger tapping (which you might argue are exclusively motor)... but the results suggest that it does measure these really well. 

The hardest thing is to develop a tool that captures fluctuation well and objectively. Some purpose-built devices do (the PKG for example), but it is impressive to see these kind of results through utilisation of the standard hardware that comes with a smart phone. Furthermore the objective response to dopaminergic therapy is substantial and apparently clinically meaningful. 

Will we see this app used in clinical trials in the coming years... I expect we might!

- Alastair Noyce

JAMA Neurol. 2018 Mar 26. doi: 10.1001/jamaneurol.2018.0809. [Epub ahead of print]
Zhan A, Mohan S, Tarolli C, Schneider RB, Adams JL, Sharma S, Elson MJ, Spear KL, Glidden AM, Little MA, Terzis A, Dorsey ER, Saria S.

https://jamanetwork.com/journals/jamaneurology/fullarticle/2676504

IMPORTANCE: Current Parkinson disease (PD) measures are subjective, rater-dependent, and assessed in clinic. Smartphones can measure PD features, yet no smartphone-derived rating score exists to assess motor symptom severity in real-world settings.

OBJECTIVES: To develop an objective measure of PD severity and test construct validity by evaluating the ability of the measure to capture intraday symptom fluctuations, correlate with current standard PD outcome measures, and respond to dopaminergic therapy.

DESIGN, SETTING, AND PARTICIPANTS: This observational study assessed individuals with PD who remotely completed 5 tasks (voice, finger tapping, gait, balance, and reaction time) on the smartphone application. We used a novel machine-learning-based approach to generate a mobile Parkinson disease score (mPDS) that objectively weighs features derived from each smartphone activity (eg, stride length from the gait activity) and is scaled from 0 to 100 (where higher scores indicate greater severity). Individuals with and without PD additionally completed standard in-person assessments of PD with smartphone assessments during a period of 6 months.

MAIN OUTCOMES AND MEASURES: Ability of the mPDS to detect intraday symptom fluctuations, the correlation between the mPDS and standard measures, and the ability of the mPDS to respond to dopaminergic medication.

RESULTS: The mPDS was derived from 6148 smartphone activity assessments from 129 individuals (mean [SD] age, 58.7 [8.6] years; 56 [43.4%] women). Gait features contributed most to the total mPDS (33.4%). In addition, 23 individuals with PD (mean [SD] age, 64.6 [11.5] years; 11 [48%] women) and 17 without PD (mean [SD] age 54.2 [16.5] years; 12 [71%] women) completed in-clinic assessments. The mPDS detected symptom fluctuations with a mean (SD) intraday change of 13.9 (10.3) points on a scale of 0 to 100. The measure correlated well with the Movement Disorder Society Unified Parkinson Disease's Rating Scale total (r = 0.81; P < .001) and part III only (r = 0.88; P < .001), the Timed Up and Go assessment (r = 0.72; P = .002), and the Hoehn and Yahr stage (r = 0.91; P < .001). The mPDS improved by a mean (SD) of 16.3 (5.6) points in response to dopaminergic therapy.

CONCLUSIONS AND RELEVANCE: Using a novel machine-learning approach, we created and demonstrated construct validity of an objective PD severity score derived from smartphone assessments. This score complements standard PD measures by providing frequent, objective, real-world assessments that could enhance clinical care and evaluation of novel therapeutics.

Keep it simple..

Predictive medicine is becoming increasingly complex, with predictive models which incorporate both clinical information and a huge range of biomarkers. Biomarkers commonly used in neurodegenerative disease include molecules in bodily fluids such as saliva, blood and spinal fluid and genetic data from these sources, scans for both structural aspects of the brain and molecular studies which can quantify the amount of protein build up in the brain. These biomarkers vary in how invasive and risky they are to collect and while they may help us understand the disease process, many people are uncomfortable with going through an invasive procedure such as lumbar puncture. 

The basis of the current PREDICT-PD model is to use participant survey to collect clinical information to predict risk. The study below, from a team in Quebec, also uses a clinical approach. They looked at a group of people with Parkinsons' disease and used an 8 item scoring scale including age, evidence of cognitive impairment at baseline, motor features such as freezing or falls, drop in blood pressure on standing and features of rapid eye movement sleep behaviour disorder. 

Interestingly, the predictive value of this model seems to be as good as studies which include biomarker values such as amyloid protein levels in the spinal fluid. Biomarker studies in this area will continue to be important, especially as they hold the hope of elucidating the molecular processes which can be interrupted to disrupt the progression of disease. But this is a good reminder of the power that can be harnessed from information which is available simply by asking questions of and examining our patients. 

-Anna

https://www.ncbi.nlm.nih.gov/pubmed/29582054

Office-Based Screening for Dementia in Parkinson Disease: The Montreal Parkinson Risk of Dementia Scale in 4 Longitudinal Cohorts.

IMPORTANCE:
Parkinson disease dementia dramatically increases mortality rates, patient expenditures, hospitalization risk, and caregiver burden. Currently, predicting Parkinson disease dementia risk is difficult, particularly in an office-based setting, without extensive biomarker testing.

OBJECTIVE:
To appraise the predictive validity of the Montreal Parkinson Risk of Dementia Scale, an office-based screening tool consisting of 8 items that are simply assessed.

DESIGN, SETTING, AND PARTICIPANTS:
This multicenter study (Montreal, Canada; Tottori, Japan; and Parkinson Progression Markers Initiative sites) used 4 diverse Parkinson disease cohorts with a prospective 4.4-year follow-up. A total of 717 patients with Parkinson disease were recruited between May 2005 and June 2016. Of these, 607 were dementia-free at baseline and followed-up for 1 year or more and so were included. The association of individual baseline scale variables with eventual dementia risk was calculated. Participants were then randomly split into cohorts to investigate weighting and determine the scale's optimal cutoff point. Receiver operating characteristic curves were calculated and correlations with selected biomarkers were investigated.

MAIN OUTCOMES AND MEASURES:
Dementia, as defined by Movement Disorder Society level I criteria.

RESULTS:
Of the 607 patients (mean [SD] age, 63.4 [10.1]; 376 men [62%]), 70 (11.5%) converted to dementia. All 8 items of the Montreal Parkinson Risk of Dementia Scale independently predicted dementia development at the 5% significance level. The annual conversion rate to dementia in the high-risk group (score, >5) was 14.9% compared with 5.8% in the intermediate group (score, 4-5) and 0.6% in the low-risk group (score, 0-3). The weighting procedure conferred no significant advantage. Overall predictive validity by the area under the receiver operating characteristic curve was 0.877 (95% CI, 0.829-0.924) across all cohorts. A cutoff of 4 or greater yielded a sensitivity of 77.1% (95% CI, 65.6-86.3) and a specificity of 87.2% (95% CI, 84.1-89.9), with a positive predictive value (as of 4.4 years) of 43.90% (95% CI, 37.76-50.24) and a negative predictive value of 96.70% (95% CI, 95.01-97.85). Positive and negative likelihood ratios were 5.94 (95% CI, 4.08-8.65) and 0.26 (95% CI, 0.17-0.40), respectively. Scale results correlated with markers of Alzheimer pathology and neuropsychological test results.

CONCLUSIONS AND RELEVANCE:
Despite its simplicity, the Montreal Parkinson Risk of Dementia Scale demonstrated predictive validity equal or greater to previously described algorithms using biomarker assessments. Future studies using head-to-head comparisons or refinement of weighting would be of interest.