Objective: To (1) evaluate which factors are associated with increasing falls in Progressive Supranuclear Palsy (PSP), unrelated to freezing; and (2) subsequently develop a predictive model to identify those at risk of increasing falls using clinical parameters amenable to treatment, to develop a treatment plan to reduce the risk of future falls.

Background: Falls in PSP occur early and increase in frequency and severity over time, causing significant morbidity and mortality. Few studies have evaluated the factors contributing to these falls, or ways to predict or help prevent them. Moreover, no predictive models to identify patients at risk are currently available, and little research has investigated measures that could prevent future falls in PSP.

Methods: Using comprehensive clinical data, 339 PSP patients were divided into two groups based on falls within the previous month: Infrequent Fallers ("IF"; n = 118) with no or rare falls, and Frequent Fallers ("FF"; n = 221) who fell occasionally to multiple times a day. 38 clinical parameters were analyzed to determine their relationship with an increasing risk of falls. A multivariate logistic regression model and Neural Network model to identify FF or IF were developed and compared. Both models were reduced to include only clinical parameters amenable to intervention (postural stability/body sway, neck rigidity or dystonia, arising from a chair, sitting down, gait, eyelid dysfunction, and modified turning).

Results: 25 of 38 clinical parameters analyzed were significantly correlated with FF (Table 1).

Table 1: Univariate regression analysis of 38 clinical parameters to identify those significantly associated with increasing falls

Parameter	Actionable (Yes/No)	P-value
Dementia Rating Scale
Initiation/Perseveration	N	NS
Attention	N	NS
Frontal Assessment Battery
Conflicting instructions	N	NS
Go-No-Go Task	N	NS
PSPRS
Bradyphrenia	N	NS
Emotional Incontinence	N	NS
Voluntary Upward Saccades	N	NS
Voluntary Downward saccades	N	NS
Voluntary Left saccades	N	0.0009
Voluntary Right saccades	N	0.0006
Eyelid dysfunction	Y	0.0032
Limb rigidity	N	0.0459
Limb dystonia	Y	NS
Toe tapping	N	0.0053
Neck rigidity or dystonia	Y	0.0067
Arising from chair	Y	<0.0001
Gait	Y	<0.0001
Postural Stability	Y	0.0083
Sitting down	Y	<0.0001
*Ocular	N	0.0126
*Limb	N	0.0038
*Gait/Midline	N	<0.0001
*Total	N	<0.0001
UPDRS
Walking	Y	<0.0001
Rigidity-RLE	N	NS
Rigidity-LLE	N	NS
Rigidity-Neck	N	0.0087
Right leg agility	N	NS
Left leg agility	N	0.02
Arising from a chair	Y	<0.0001
Postural stability-Body sway	Y	0.0006
Body bradykinesia and hypokinesia	N	<0.0001
*Motor	N	<0.0001
*Total	N	<0.0001
Novel
Modified turning	Y	<0.0001
Posture (Hyperextended)	Y	NS
Staging
PSP Staging System: Gait/Stability	N	<0.0001
Modified Hoehn & Yahr	N	<0.0001

*Item score comprising the sum of individual parameter subscores. Novel = Clinical Parameter developed for this study. NS = Not Significant (P-value >0.05). “Actionable” = clinical parameters amenable to treatment. The best multivariate logistic regression model was able to correctly classify 92% of FF but only identified 26% of IF. However, the Neural Network model predicted FF and IF with 100% accuracy after validation. Table 2 provides an example of how improvement of specific clinical parameters would be expected to reduce fall risk based on the neural network predictive model.

Table 2: Example of a Treatment Plan Using 3 Clinical Parameters Amenable to Therapeutic Intervention

	Arising from a Chair (UPDRS)	Eyelid Dysfunction (PSPRS)	Modified Turning	Prob (0)	Prob (1)	Likelihood of Fall	Risk Reduction
Current	3	3	3	0.405	0.595	1	*
Plan	2	1	2	0.817	0.183	0	41%

The clinical parameters included can be treated to significantly reduce the risk of future falls. Prob (0) = the likelihood of being an infrequent faller. Prob (1) = the likelihood of being a frequent faller.

Conclusions: Certain clinical parameters are associated with a higher frequency of falls, which help identify PSP patients at risk for increasing falls and improve our understanding of why these falls occur. The neural network model correctly identified FF with high accuracy, and warrants validation in a prospective study. The ability to create a treatment plan based on symptoms amenable to treatment could help reduce the high morbidity and mortality caused by falls in PSP.

CurePSP 2015 International Symposium.

« Back to 2016 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/understanding-predicting-and-preventing-falls-in-progressive-supranuclear-palsy/