Debugging Adaptive Deep Brain Stimulation for Parkinson's Disease

The review by Little and Brown1 on adaptive deep brain stimulation (aDBS) was very informative. However, a fundamental conundrum that challenges medicine and is likely a not infrequent cause of failure is the difficulty in caring for the individual patient through extrapolation from aggregate statistical descriptors of research studies.2 Often, the problem is intractable. However, in the case of adaptive aDBS, we have an opportunity. We can focus on the detection algorithms and ask what is the specificity and sensitivity. Because importantly, we need to know the positive and negative predictive values, which requires knowledge of the prior probabilities of the signal to be detected. For example, if 15% of patients with Parkinson's disease do not have meaningfully increased beta power, the detection algorithm will have, at the minimum, a 15% false-negative rate. It may be that the other 85% of patients will do well with aDBS, and the aggregate clinical benefit in research studies will be positive. But the physician addressing the individual patient does not know a priori whether the patient is among the 15% who will be false negatives. Unfortunately, prior probabilities are difficult to determine because most published studies of beta oscillations pool individuals' data. Reporting of other analyses such as logistic regression and the receiver operator curve characteristics of the detection algorithm may be helpful. The second half of the question is the consequent therapeutic stimulation. It is not clear that the dynamics of the therapeutic response have been adequately considered relative to the timing and duration of stimulation. We can take, as a metaphor, drug pharmacokinetics where the dosing intervals are greatly affected by the drug half-life, which affects wash-in and washout periods. With drugs, any dosing interval less than the half-life likely is therapeutically equivalent. Dosing intervals greater than the half-life likely will be less therapeutic. Further, it may take multiple doses for the pharmacological effect to reach steady-state benefit unless one loads the patient. It is not likely that a loading dose of aDBS is feasible; thus, the time it takes DBS to reach steady-state effect is critical. For example, DBS in cycling mode at 500 ms on and 500 ms off was nearly as effective as continuous DBS, whereas cycling at 100 ms on and off was not as effective, even though the same number of stimulation pulses at the same parameters were given.3 However, a caution, the study was not designed to address the therapeutic kinetics of DBS. The issues of DBS therapeutic kinetics analogous to pharmacokinetics are important. We know some disabilities have significant latencies to gain and loss of benefit, and hence differences in wash-in and washout effects. For many disabilities, we do not know the functional half-life of DBS. If the wash-in period for a DBS effect is on the order of many tens of seconds or more, what does that mean for aDBS? If the washout is on the order of many tens of seconds or more, what are the implications for aDBS?

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.