“Well, Jane, it just goes to show you, it’s always something—if it ain’t one thing, it’s another.” – Roseanne Roseannadanna (Gilda Radner) Saturday Night Live, NBC television
There is great diversity between surgical approaches for Deep Brain Stimulation (DBS) lead implantation; however, there is no “smoking gun” evidence for one approach over another. However, it is important to understand what is meant by “smoking gun” evidence. One of the main concerns is if there potentially could be “smoking gun” evidence potentially, and would it ever be an actuality. Unlike scientists who can hold questions in abeyance, awaiting more data, the immediate needs of patients press physicians to offer the best healthcare they can in that moment.
As is often the case, the decision as to what is best for the patient, is generally made on intangibles and trade-offs, requiring value judgements. The trade-offs follow the Roseanne Roseannadana Law, which states, “it’s always something—if it ain’t one thing, it’s another.” Following the examination of possible options, a decision is made based on consequences of each possible option. Each option may confer a different set of advantages and disadvantages, making comparison between them impossible by computers. Thus, in the absence of Randomized Control Trials, the decision based on the advantages and disadvantages may turn on a set of fundamental considerations common to all approaches. Focusing on the set of fundamental considerations provides great economy of intellectual effort, in that the potentially infinite number of variations in surgical approaches still all turn on a manageable set of fundamental considerations. It is hard to know a priori what the best route is when deciding the optimal surgical approach. However, it is relatively epistemically safe to say that failure to consider the fundamental set of considerations, should they exist, is less likely to result in the best practice. Indeed, the failure to consider the fundamental concerns likely leads to a false sense of confidence in the option pursued, based on the Fallacy of Limited Alternatives.
The effect of choice on the decision method
The first and perhaps most important decision is the choice of an array, such as the five-microelectrode “Ben-Gun” array versus the single-microelectrode-at-a-time approach. This question centers on whether the surgeon would make a second pass with the array (or some subset of the array) by moving the array to a new location. Generally, neurophysiologists using a single-microelectrode-at-a-time approach fully expect the electrode trajectory may have to be moved. Note, the answer to the question need not be dichotomous, yes or no; merely the hesitancy to move the array is a de facto and at the least a partial “no” answer. The degree to which the surgeon is unwilling to relocate the array, whether in principle or in practice, necessitates a decision as to which among the microelectrodes is the best of the options dictated by the microelectrodes used. It is very important to note that this is not the same question as “where is the best trajectory”. It is possible, none of the options conferred by the array contain the best possible trajectory, the probability to be discussed latter. Thus, it is not the best trajectory that is offered to the patient, instead it is the best of the choices limited by the methods the surgeon uses.
A consensus appears to have emerged among intraoperative neurophysiologists (regardless who is playing that role), that at least in the case targeting the subthalamic nucleus (STN), the best trajectory is one that contains at least 5 mm of STN trajectory containing neurons whose activities are sensori-motor driven, no adverse effects with stimulation, and some improvement in symptoms. The last criterion is not hard and fast. The question should arise as to the basis for such a consensus. Clearly, there is no Evidence Based Medicine Level I empirical data to “prove” the criteria. Rather, the criteria is based on a set of heuristics. First, it is believed, generally, that stimulation of the sensori-motor region of the STN is important for the motoric benefits of DBS. Second, it was thought that if the trajectory contained at least 5 mm of sensori-motor STN, it would mean any subsequent move, typically involving a move of at least 2 mm, would result in a trajectory containing less than 5 mm of sensori-motor STN and hence, potentially resulting in less clinical benefit. This is based on the anatomy of the STN. The third criteria is to avoid placing the permanent therapeutic DBS lead in a trajectory that has a high probability of producing adverse effects that would limit the clinical utility post-operatively. Typically, microstimulation is done, for reasons of interpretable spatial resolution, which is not the same as therapeutic stimulation. However, if there were adverse effects with microstimulation, such as persistent paresthesias or tonic contraction, then there would be an overwhelming probability that attempting therapeutic DBS in the same trajectory would not be tolerated by the patient.
In the case of targeting the ventral intermediate nucleus of the thalamus (Vim) and the globus pallidus interna (GPi), there probably is less consensus (it is hard to determine the degree of consensus as there is minimal discussion of the issue). One approach defines the best trajectory as locating the structures one does not want to stimulate through the permanent DBS lead. In the case of Vim, one does not want to stimulate the caudal region of the ventral caudal thalamus (Vc) as this can cause intolerable paresthesias that prevent effective clinical stimulation. In the case of GPi, the structures to be avoided are the corticospinal and corticobulbar tracts in the posterior limb of the internal capsule; stimulation of which can lead to tonic muscle contractions. Thus, one criteria for the best trajectory is one that is 2-3 mm anterior to the anterior border of the posterior Vc in Vim DBS and to the anterior border of the posterior limb of the internal capsule in GPi DBS. This distance is increased to 3-4 mm in the case of dystonia, as often high DBS voltages or currents are necessary for clinical effect.
An additional criterion in the case of targeting Vim and GPi is that the trajectory is in the appropriate motor homuncular representation of the target nuclei. In the case of Vim, typically this means avoiding the head representation in order to lessen the risk of subsequent speech, language and swallowing problems with post-operative treatment. In the case of GPi, it means a trajectory through the homuncular representation most appropriate to the patient’s needs. For example, because the spatial distance between the leg homuncular representation and the head homuncular representation is much greater than the effective radius of the volume of tissue activation in typical DBS, placing the DBS lead in the leg representation is unlikely to benefit a patient with cervical dystonia.
There is no reason to believe that the criteria for best trajectory a priori should be different for physicians using microelectrode arrays compared to physicians using a single microelectrode at a time. Thus, the question is whether a single pass with fixed spatial array of microelectrodes has a sufficient probability of identifying the best trajectory as defined above and not “best” in the sense of relative to the other trajectories available. To be sure, it may well be that using a fixed array that is unlikely to be moved to allow a second pass may have a lower probability of identifying the best trajectory. This may be acceptable if there are corresponding advantages of using the array as compared to the single electrode at a time approach that might require repeated penetrations.
The user of a single-electrode-at-a-time approach faces a question that a physician who uses a microelectrode array does not. The user of the single electrode is confronted with the question of where to move the single electrode if the previous trajectory does not meet the criteria for best trajectory. As a move of the multiple electrode array typically is not done, the user of the array simply choose among the options available.
The critical question is how many moves are required and over what spatial distance is optimal for a single microelectrode. In one study of 144 cases of STN DBS targeting, the criteria for best trajectory was found with only a single pass of the microelectrode in 100 penetrations (Montgomery, E. B. Jr. Microelectrode targeting of the subthalamic nucleus for deep brain stimulation surgery. Mov Disord. 2012 Sep 15;27(11):1387-91. doi: 10.1002/mds.25000. Epub 2012 Apr 16. PubMed PMID: 22508394). Of the remaining 44 cases, 30 (68%) required only a single additional penetration, three penetrations in 11 (25%) and four penetrations in 2 (5%). The average number of penetrations of the brain using the single microelectrode approach was 1.4. This means that on average those using a 5-microelectrode array are exposing their patient to 3.6 more penetrations in order to find the best trajectory. Further, the “best” for a microelectrode array is not the same and arguably less accurate, than the best for a single-microelectrode-at-a-time approach. In none of the cases, did any of the patients require 5 penetrations when compared to the minimum case using a five electrode array. Clearly, the use of the five electrode array has a high probability of subjecting the patient to a greater number of penetrations of the brain than using a single-microelectrode-at-a-time approach. However, this is not to say that there are no other compensating advantages of the five microelectrode array.
Can the array cover the volume of potential optimal trajectories?
One argument has advanced in favor of the five microelectrode array, in terms of greater “efficiency”. Unfortunately, there is no Evidence Based Medicine level I (or lesser levels) empirical data to support the claim. Rather, the claim for greater efficiency lies in a rationale that the five electrode array can cover a greater region of the brain in a more economical period of time. Thus there are two components, efficiency spatially and efficiency in time.
Considering the issue of spatial efficiency, the nature of the space of the target needs to be considered. As both the arrays and the single microelectrodes are moved a large distance in the axis of the electrodes, both would “sample” the same vertical space (defined as parallel to the long axis of the microelectrode). Therefore, the question is the nature of the space in the plane orthogonal (at 90o or right angles to the vertical), in other words, the cross-sectional area. The issue seems to rest on the notion that the five electrode array covers an area in the shape of a square with the four outer microelectrodes forming the corners of the square. If one assumes that the sampling radius of a microelectrode tip is on the order of 250 μm, then using the typical Ben-Gun array with 2 mm offsets, would have a cross-sectional sampling area of 11.1 mm2. This is less than the 32.2 mm2 found necessary to contain 99% of the best trajectories in the study cited above. Thus, the cross-sectional area covered by the Ben-Gun array would cover only 34% of the cross-sectional area that is needed to have a 99% confidence of including the best trajectory.
The notion that the five electrode array “samples” a large sample area is an illusion. Assuming a “sampling” radius of 250 μm, the actual area sampled is 0.2 mm2, which would be a total of 1 mm2 for the five electrode array. The actual cross-sectional area sampled by the five electrodes is 0.6%, which is needed to have a 99% confidence of including the best trajectory.
The argument could be made that at least the cross-sectional area surveyed by the five microelectrode array is 5 times that of a single microelectrode. That is true, however, the single microelectrode can be moved anywhere in the 99% confidence cross-sectional area and thus, in principle, could cover the entire 99% confidence area if necessary. A minimum of 164 penetrations could cover the entire 99% confidence area, whereas a fixed array not allowing for movement of the array cannot cover more than 0.6% of the 99% confidence area. It is indeed fortunate, that the algorithms available to the user of the single-microelectrode-at-a-time approach are so robust that in nearly every case, less than 4 moves from the original trajectory are necessary.
The spatial analysis described above is sobering, but it would be a mistake to read too much into it. Clearly, the surgeons and neurophysiologists using the five electrode arrays provide post-operative benefit to their patients, at least in a manner that is not clearly inferior to those using a single-microelectrode-at-a-time approach. This is not an inconsequential defense of the practice of using the five microelectrode array. Whether the use of the five microelectrode array results in a small degree of worst outcomes is unknown and attempting to answer that question is highly problematic for many reasons. Yet, those using or advocating for the use of the five microelectrode arrays must realize that there is little justification based on spatial considerations. Also, it is important to recognize that a small difference in outcomes, even if not detectable because of practical considerations, could have a serious impact when applied to the many thousands of patients undergoing DBS.
Is using the five channel array really faster than multiple sequential penetrations using a single microelectrode?
Another attractive perception of using five microelectrode arrays is that of temporal efficiency, that is being able to utilize less operating room time. Certainly it seems reasonable that advancing five electrodes at one time should make “quicker time of it,” in comparison to advancing one microelectrode and then move and advance it again, and so on. However, will such a perception survive scrutiny? This would only be true if advancing the five electrodes at a time took the same (or less) time than advancing a single microelectrode. But is this the case?
If one assumes that the criteria for best trajectory is the same, regardless of the numbers of microelectrodes simultaneously advanced and if it is assumed that the purpose of each microelectrode trajectory is to obtain the information necessary on which to apply the criteria, then each microelectrode needs to obtain the same information regardless of how many microelectrodes are advanced simultaneously. This means that the five microelectrode array cannot be advanced any faster than a single microelectrode. However, if the information in the five trajectories of the five microelectrode arrays could be processed in parallel (meaning the information in each trajectory can be independently and simultaneously assessed), then clearly, the five microelectrode array will acquire more information per unit time compared to using a single microelectrode at a time.
However, it is highly unlikely, given current technology, that each channel of information in the five microelectrode array can be assessed independently and simultaneously. Consider sensori-motor testing to assess whether the trajectory is within the sensori-motor region, which is the most time demanding and largely determines the time spent in neurophysiological monitoring. Current technology allows physicians to play the microelectrode recordings through an audio monitor as the patient is manipulated in order to determine whether or not there is a change in the firing pattern of the neuron, correlated with the manipulation of the patient. This makes it virtually impossible to assess two independent microelectrode recording sites simultaneously. Technological advancements may potentially test for sensori-motor driving simultaneously in an arbitrarily large number of microelectrode recording sites by using automated analyses.
The current technologies involving multiple microelectrode arrays are an “attention dividing” task. Still the neurophysiologist must attend to each channel and stop the advance of all microelectrodes every time neuronal activities are discovered in any one of the channels. It is arguable that this provides an increase in time efficiency. The only time saving advantage comes not from the actual recordings of neuronal activities but the time necessary to physically move the single microelectrode to a new location; even this on average would happen 0.7 times during any single surgery.
In this author’s experience, each microelectrode trajectory for a single-microelectrode-at-a-time approach typically takes 30 minutes; which means, just based on the recording times, exclusive of the physical moves, the typical recording time would be approximately 42 minutes for the STN based on an average of 1.4 penetrations. In cases involving the five microelectrode array approach, the typical recording times take far more than 40 minutes, in this author’s experience. Thus, it is not clear that the five microelectrode array method is much more time efficient than the single-microelectrode-at-a-time approach.
What can be said about the relative safety of the multi-electrode array and the single-microelectrode-at-a-time approach? Some proponents of the multi electrode array claim that the risk of intracerebral hemorrhage is the same as those reported in the literature, presumably reflecting the risk of hemorrhage attributable to the use of a single-microelectrode-at-a-time approach. The question is what confidence does one have with such a claim?
First, most claims are based on not finding a statistically significant difference in the rate of intracerebral hemorrhage compared to rates published in the literature. However, failing to demonstrate a difference using standard statistical hypothesis testing is not appropriate as it may be a type II error related to the variance in the measures and small sample sizes, even with effect sizes that would be meaningful. The better approach is to decide first, what is a meaningful increase in hemorrhage rate, determine the sample size necessary to demonstrate a meaningful increase in hemorrhage rate and then apply the statistical tests.
Unfortunately, it is not as easy as the above paragraph would suggest. There are a number of challenges to the statistical analyses. In most cases, a hemorrhage is a rare event. The result is a highly asymmetric or skewed distribution that makes use of traditional statistical analyses problematic as the distribution often violates the necessary assumptions of most tests. Fundamental to most statistical analyses is the Large Number Theorem which holds that any statistic will converge to some value, assumed to be the true value, with increasing sample size. However, because of the rare dichotomous nature of hemorrhage events, it is hard for the series to converge on a stable statistic. In other words, greater number of observations are required. For example, is there really a difference between 1 hemorrhage in 100 cases using one method and 2 hemorrhages in 100 cases using an alternative method, even though the rate has doubled? However, assume that 900 additional cases are conducted with each method. Now there are 100 hemorrhages with the first method and 200 hemorrhages with the second method and most would claim there is a significant difference. Thus, it might well be that in the first 100 cases, the first method had less risk than the second, but that would not be discernable with any degree of confidence. Certainly, after the first 1000 cases, the proponent of the second method could claim that the hemorrhage rate was no greater than the first method and it would be hard to refute that claim.
In a large meta-analysis of stereotactic surgery related to brain penetrations, the average rate of intracerebral hemorrhage was 1.57% (95% CI, 1.26-1.95%) (Kimmelman J, Duckworth K, Ramsay T, Voss T, Ravina B, Emborg ME. Risk of surgical delivery to deep nuclei: a meta-analysis. Mov Disord. 2011 Jul;26(8):1415-21. doi: 10.1002/mds.23770. Epub 2011 May 14. PubMed PMID: 21574186). This suggests that the average intracerebral hemorrhage rate would be 7.9% compared to on average 2.2% for the single-microelectrode-at-a-time technique, assuming an average number of penetrations per case at 1.4. This represents nearly a four times increased risk of intracerebral hemorrhage. One might think that this would be readily detectable, but again, the overall risk is still very small and the issue of the statistics of rare events still holds.
From a logical perspective, what would it mean to hold that the risk of intracerebral hemorrhage with the five microelectrode array is less than five times the risk of a single-microelectrode-at-a-time, or four times for an average 1.4 penetrations with the single-microelectrode-at-a-time approach? If the probability of a hemorrhage with each microelectrode penetration is independent of another, then the total probability for a five microelectrode array would be just the sum of the probabilities for each microelectrode, which would be 7.9%. For the total probability to be less, then the probabilities are not independent, which means that the risk of hemorrhage with one microelectrode would have to be alter the risk of hemorrhage with any of the other microelectrodes. In this case, the probability of a hemorrhage for any other microelectrode would be reduced. It is not at all clear how the risk of hemorrhage in one of the microelectrodes in the five-microelectrode array would reduce the risk of hemorrhage from any other microelectrode. Unfortunately, the empiric data published is of no help. Thus, it would seem that the “better part of valor” would be to assume that the risk of intracerebral hemorrhage is higher with the five-microelectrode array, even if it is not apparent from published studies.
This author has worked with surgeons who have used both approaches. It is this author’s clinical experience (with all the usual caveats admitted) that the trauma to the brain, evidenced by post-operative MRI scan, was much greater following the use of the five microelectrode array compared to the use of the single-microelectrode-at-a-time approach. It is important to recognize that the passage of the microelectrodes exert shear forces on the surrounding brain tissue and that these forces are propagated radially from the surface of the microelectrode. The concern is that the shear forces from adjacent microelectrodes would summate and increase tissue damage. An example is the bed of nails trick where if a person was to lay on a single upright nail, the nail would pierce the body. However, if there is an array of nails that are closely aligned, the skin resists penetration by the nails. In this case, increased resistance of the brain to the movement of the microelectrode with the multiple microelectrode array would tend to increase the force necessary to advance the electrode and thereby, increase the shear forces in the surrounding brain tissue increasing the risk for trauma.
At the FDA expert advisory panel that met in 2000 to review the application for approval of DBS of the STN and GPi, an issue arose about whether the FDA should regulate or at least require instruction as to the surgical methods of DBS lead implantation. Those speaking on behalf of the manufacturer argued against this, stating that oversight for best practices should be the purview of hospitals and clinics that credentialed surgeons to perform DBS lead implantation. That was perhaps naïve. Nonetheless, the FDA does not regulate the manner by which DBS leads are implanted. The diversity of the different approaches means that the typical notion of the standard of care being what similarly situated physicians do, derived from medical malpractice law, allows a wide latitude, perhaps too wide. This ultimately means that the neurosurgeon must look to her own sense of ethics and values when deciding upon a method of DBS lead implantation. However, this should not be interpreted as free license, but rather a responsibility to act as wisely as possible.