A meta-analysis of six prospective studies of falling in Parkinson's disease.

Recurrent falls are a disabling feature of Parkinson's disease (PD). We have estimated the incidence of falling over a prospective 3 month follow-up from a large sample size, identified predictors for falling for PD patients repeated this analysis for patients without prior falls, and examined the risk of falling with increasing disease severity. We pooled six prospective studies of falling in PD (n = 473), and examined the predictive power of variables that were common to most studies. The 3-month fall rate was 46% (95% confidence interval: 38-54%). Interestingly, even among subjects without prior falls, this fall rate was 21% (12-35%). The best predictor of falling was two or more falls in the previous year (sensitivity 68%; specificity 81%). The risk of falling rose as UPDRS increased, to about a 60% chance of falling for UPDRS values 25 to 35, but remained at this level thereafter with a tendency to taper off towards later disease stages. These results confirm the high frequency of falling in PD, as almost 50% of patients fell during a short period of only 3 months. The strongest predictor of falling was prior falls in the preceding year, but even subjects without any prior falls had a considerable risk of sustaining future falls. Disease severity was not a good predictor of falls, possibly due to the complex U-shaped relation with falls. Early identification of the very first fall therefore remains difficult, and new prediction methods must be developed.