Predictors of effective bilateral subthalamic nucleus stimulation for PD
Erwin BMontgomery, Departments of Neurology and Neuroscience Cleveland OHMONTGOE@ccf.org
Submitted October 29, 2002
Charles et al. make an important contribution with their article
describing possible predictors of deep brain stimulation (DBS) of the
subthalamic nucleus efficacy for Parkinson disease. [1] While DBS is
highly effective and FDA approved, the procedure has considerable risks.
Effective predictors could favorably shift the risk to benefit ratio.
Unfortunately, the analysis performed is of limited value and potentially
misleading.
A more appropriate analysis would be to report the area under the
Receiver-Operator Characteristic curve, which relates the specificity and
sensitivity of the tests to age and levodopa responsiveness. The goal of
any predictive task not only is to avoid surgery for those patients not
likely to benefit but also to avoid withholding surgery from those that
would. Visual inspection of the data represented in the graphs provides
little confidence that either age or levodopa responsiveness will have
sufficient specificity and sensitivity to be an effective predictor that
can be used for patient selection.
In addition, the study of predictors was limited to a retrospective
correlational analysis. Correlation is a mathematically optimizing
procedure which will find a correlation even if spurious. [2] Thus, it
remains unclear how generalized are the regression analyses performed.
That is why it is so important to apply the predictive regression
equations in a prospective manner. Often, dividing the sample population
into two groups, the first to develop the regression equations and the
second to prospectively test those equations, can do this. The large
majority of times, the specificity and sensitivity of predictors protectors fall when
tested prospectively.
References:
1. Charles PD, Van Blercom N, Krack P, et al. Predictors of effective
bilateral subthalamic nucleus stimulation for PD, Neurology 2002;59:932-
934.
2. Wasson JH, Sox HC, Neff RK, Goldman L. Clinical prediction rules:
applications and methodological standards. N Engl J Med 1985;313:793-799.
Charles et al. make an important contribution with their article describing possible predictors of deep brain stimulation (DBS) of the subthalamic nucleus efficacy for Parkinson disease. [1] While DBS is highly effective and FDA approved, the procedure has considerable risks. Effective predictors could favorably shift the risk to benefit ratio. Unfortunately, the analysis performed is of limited value and potentially misleading.
A more appropriate analysis would be to report the area under the Receiver-Operator Characteristic curve, which relates the specificity and sensitivity of the tests to age and levodopa responsiveness. The goal of any predictive task not only is to avoid surgery for those patients not likely to benefit but also to avoid withholding surgery from those that would. Visual inspection of the data represented in the graphs provides little confidence that either age or levodopa responsiveness will have sufficient specificity and sensitivity to be an effective predictor that can be used for patient selection.
In addition, the study of predictors was limited to a retrospective correlational analysis. Correlation is a mathematically optimizing procedure which will find a correlation even if spurious. [2] Thus, it remains unclear how generalized are the regression analyses performed. That is why it is so important to apply the predictive regression equations in a prospective manner. Often, dividing the sample population into two groups, the first to develop the regression equations and the second to prospectively test those equations, can do this. The large majority of times, the specificity and sensitivity of predictors protectors fall when tested prospectively.
References:
1. Charles PD, Van Blercom N, Krack P, et al. Predictors of effective bilateral subthalamic nucleus stimulation for PD, Neurology 2002;59:932- 934.
2. Wasson JH, Sox HC, Neff RK, Goldman L. Clinical prediction rules: applications and methodological standards. N Engl J Med 1985;313:793-799.