- Email: [email protected]
Longitudinal Impedance Variability in Patients with Chronically Implanted DBS Devices
Brain Stimulation 6 (2013) 746e751
Contents lists available at SciVerse ScienceDirect
Brain Stimulation journal homepage: www.brainstimjrnl.com
Longitudinal Impedance Variability in Patients with Chronically Implanted DBS Devices Tyler Cheung a, Miriam Nuño a, Matilde Hoffman a, Maya Katz b, Camilla Kilbane c, Ron Alterman d, Michele Tagliati a, * a
Department of Neurology, Cedars-Sinai Medical Center, 8730 Alden Drive, Thalians E-238, Los Angeles, CA 90048, USA Department of Neurology, UCSF, San Francisco, CA, USA c Department of Neurology, Stanford University School of Medicine, Palo Alto, CA, USA d Department of Neurosurgery, Beth Israel Deaconess Medical Center, Boston, MA, USA b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 4 December 2012 Received in revised form 9 March 2013 Accepted 20 March 2013 Available online 22 April 2013
Background: Deep brain stimulation (DBS) is an effective therapy for advanced movement disorders, but its optimal use is still controversial. One factor that could play a role in the proper delivery of therapeutic stimulation by current DBS devices is the variability of the impedance at the interface between the electrode surface and surrounding tissue. Objective: To analyze variability and trends in the impedance of chronically-implanted DBS electrodes in subjects with movement disorders. Methods: We reviewed impedance values from medical records of DBS patients at an academic tertiarycare movement disorders center. The standard deviation of data recorded within individual subjects and single contacts were used as measures of longitudinal impedance variability. A generalized linear mixed model (GLMM) determined if a number of effects had significant influences on impedance. Results: We analyzed 2863 impedance measurements from 94 subjects. Median variability, for subjects with follow-up from 6 months to 5 years (n ¼ 77), was 194 U for individual subjects and 141 U for individual contacts, with a range spanning from 18 to over 600 U. The GLMM, incorporating all subjects (n ¼ 94), identified time, electrical activity, implanted target, contact position on the electrode and side of implantation as significant predictors of impedance. Age and disease duration at surgery, gender or ethnicity were not significant predictors. Conclusions: Our analysis suggests that a significant amount of impedance variability can be expected in chronically implanted DBS electrodes and indicates a number of factors with possible predictive value. Further studies are needed to link impedance characteristics to clinical outcomes. Ó 2013 Elsevier Inc. All rights reserved.
Keywords: Impedance Deep brain stimulation Movement disorders Globus pallidus Subthalamic nucleus
Introduction Deep brain stimulation (DBS) is an effective therapy for medication-refractory movement disorders including advanced Funding: This study was funded in part by the Parkinson Alliance. Financial disclosures: Tyler Cheung, MD, Maya Katz, MD, and Camilla Kilbane, MD, were beneficiaries of unrestricted fellowships from Medtronic, Inc. Ron L. Alterman, MD, has received consultation fees from Medtronic, Inc., unrelated to the conduct of this study. Michele Tagliati, MD, has received speaker honoraria from Medtronic, Inc., and consultation fees from St. Jude Medical, Inc. (formerly Advanced Neuromodulation Systems), Abbott Laboratories, Merz and Impax Laboratories, Inc., unrelated to the conduct of this study. Miriam Nuño, PhD, and Matilde Hoffman have nothing to disclose. This manuscript has been presented as an abstract at the 2012 International Congress of the Movement Disorders Society in Dublin, Ireland, but has not been previously submitted or published in any other journal. * Corresponding author. E-mail address: [email protected] (M. Tagliati). 1935-861X/$ e see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.brs.2013.03.010
Parkinson’s disease (PD), essential tremor (ET) and dystonia [1e5]. However, DBS mechanisms of action are not completely understood and, as a consequence, the optimal use of this therapy remains a source of controversy. One area of interest is the consistency of the therapeutic stimulation delivered by currently available neurostimulators, with particular attention on the electrical impedance of these implanted systems. Impedance, the measure of the opposition to current in an oscillatory circuit, is fundamentally related to voltage and frequency, two DBS parameters crucial to determining therapeutic effect. Currently available devices, controlled via constant-voltage designs, do not automatically adjust the amount of therapeutic stimulation to account for impedance fluctuations. Therefore, large changes in impedance could potentially affect clinical outcomes. Studies of impedance in computational and animal models have demonstrated changes over the time period following implantation, or activation of stimulation, particularly between the
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
interface between the metal electrode surface and surrounding tissue [6e9]. The behavior of impedance in human subjects is much more difficult to study and therefore less well characterized. Only a few studies have indicated trends that include decreased impedance in electrically active versus inactive electrode contacts and a wide range of differences between individual subjects [10,11]. These studies focused on impedance values derived from patients’ customized “therapeutic” impedances. While clinically relevant, these measures are difficult to compare both longitudinally and across a given sample of subjects due to differing follow-up times and wide variety of stimulation parameters (i.e., voltage, pulse widths, frequency, and electrode contact configuration). In order to fill some of these gaps, we designed a study to characterize the range of impedance variability in measurements within individual subjects, as well as within individual electrode contacts in each subject, over time. In addition, we investigated a number of variables, such as time, electrical activity, anatomic location and other demographic data that may have an influence on impedance. We took advantage of the standardized impedance measurement capability built into current generation devices. These measurements, routinely employed in the course of clinical care to detect hardware malfunction such as breaks or shorts in the circuitry, are performed at pre-defined and uniform parameters not necessarily reflective of therapeutic stimulation. Nevertheless, they automatically control for variables such as voltage, pulse duration, frequency, and electrode contact configuration, making them more amenable to repeated measures and more conducive to robust statistical analysis. We present a retrospective analysis of such data collected after surgical implantation in a large cohort of DBS patients. Methods Subjects We retrospectively reviewed the medical records of all patients treated with DBS in an academic tertiary-care movement disorders center between 2005 and 2010. The institutional review board approved all data collection methods and activities. Subjects were selected based on the criteria of having received DBS implantation and at least 1 follow-up visit where standardized single contact impedances were measured. DBS implantations were performed according to surgical practices previously described [12,13], using Model No. 3387 S electrodes and Soletra neurostimulators (Medtronic, Inc., Minneapolis, MN). Impedances were measured using the built-in function of the Soletra, accessed and read via a handheld DBS programming device (nVision model, Medtronic, Inc., Minneapolis, MN). Measurements were conducted at single monopolar configuration for each of the 4 contacts of the quadripolar 3387 electrode, using a standard setting of 3.0 V, 210 ms pulse width, and a frequency of 30 Hz. In addition to impedance, collected data included diagnosis, age and disease duration prior to surgery, gender, ethnicity, the side and implanted target, and the electrically active contact configuration at each follow-up visit. The data were collected at the following post-operative “milestones”: baseline (i.e., initial programming), 1 month, 3 months, 6 months, 1 year, and 2 years. In addition, data from a “long term,” (w5 year) follow up visit were also recorded. Available visits in the database were grouped as close to these milestones as possible. Impedance readings from re-implanted electrodes were discarded to minimize potential confounds from prior tissue trauma and inconsistency of anatomical location. Therapeutic impedances relative to the subject’s customized stimulation parameters were not analyzed.
747
Within-subject and within-contact variability In order to calculate how much impedance variability might occur across the electrodes implanted in an individual subject, as well as within each individual contact on a quadripolar electrode over time, we divided impedance readings by individual subject and further stratified them by individual contact (n ¼ 4) on each electrode. Subjects whose data were included in the distributions had a minimum follow-up period of 6 months, up to a maximum of 60 months. We calculated the standard deviation for each of these groups, thus obtaining two sets of standard deviations, describing the longitudinal variation in impedance for 1) individual subjects and 2) individual contacts. For each of these two data sets, we created a probability distribution and calculated its median and quartiles values, as well as kurtosis and skew. Evaluating multivariate effects influencing impedance over time In order to investigate the effects on impedance from the variables captured in our dataset, we incorporated them into a generalized linear mixed model (GLMM). Unlike standard regressions, in which all observations are independent of one another, the GLMM takes into account the fact that each subject contributes multiple repeated measures, creating correlated observations. This modeling approach also considers unbalanced longitudinal data (i.e., the differing number of subjects present at each follow-up time-point) as observed in this study, enabling the use of all values in our dataset. The GLMM allows estimation of how the covariates differ, how their effects change over cycles, and forms a basis for exploring interactions between predictors. We elected to treat time as a categorical variable, given the structure of the data which featured visits heavily clustered around the seven time points defined as “milestone visits.” We adjusted for unequally spaced repeated measures via a time-series-type of covariate structure widely available in GLMM modeling. In this study we assumed a spatial power law covariance structure. Effects were fitted to the GLMM in a stepwise fashion, and those that did not improve the Akaike Information Criterion (AIC) were discarded to prevent overfitting. Correlations among impedance measurements made on the same patient were modeled via mixed effects through the specification of a covariance structure. Effects were considered to be significant when P < 0.05 as determined by Type III tests for sum of squares. The results from these models are discussed in terms of least square (LS) means when evaluating fixed effects such as implanted target, electrical activity, side of implant, and contact position. LS means estimate the marginal means over a particular population and an approximate t-test to assess potential differences between the means being assessed. We incorporated significant effects along with relevant 2-way interaction terms into an overall multivariate model. Statistical analysis was conducted primarily using SAS version 9.1 for Windows (SAS Institute Inc., Cary, NC). Supplementary analysis was performed using R version 2.15.1 and the ggplot2 library [14,15]. Results Demographics We were able to tabulate 2863 impedance measurements in 94 patients. Twelve measurements were discarded as indicative of an “open circuit,” defined as an impedance value greater than 2000 U with a battery current draw below 11 mA. All subjects
748
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
Table 1 Demographics. Gender Female Male Age at surgery, years (SD) Parkinson’s disease Primary dystonia Essential tremor Chorea acanthocytosis Diagnosis Parkinson’s disease Dystonia Essential tremor Chorea acanthocytosis Anatomic location Subthalamic nucleus Globus pallidus pars interna Thalamus Ethnicity African American Asian Caucasian Hispanic Other Disease duration at surgery, months (SD) Parkinson’s disease Primary dystonia Essential tremor Chorea acanthocytosis
35 59 46 (22) 66 (8) 33 (18) 76 49 34 58 1 1 35 58 1 1 2 75 3 13 152 (83) 246 (174) 156 12
with Parkinson’s disease were implanted in the subthalamic nucleus (STN), while subjects with dystonia were implanted in the globus pallidus pars interna (GPi), with the exception of one subject implanted in STN. The subject with tremor was implanted in the ventral intermediate (VIM) nucleus of the thalamus, while the subject with chorea-acanthocytosis was also implanted in GPi. Further demographic data and the compositional background of our cohort are shown in Table 1, while post-operative follow-up times are show in Table 2. Visit 7 was less well attended and more variable than the other milestone visits. The ranges of impedances per visit across our entire cohort, as well as the overall range of impedances per subject, are visualized in Figs. 1 and 2.
Figure 1. The range of impedance measurements for individual subjects across time. Box plot illustrating the range of impedances for all individual subjects (n ¼ 94) throughout all time points, sorted by median impedance per subject. Boxes indicate the range between the 25th and 75th percentiles (interquartile range), with the median (50th percentile) as the horizontal line within the box. Whiskers represent the overall range of values. Filled circles indicate outliers, defined by Tukey’s criteria as values outside of 1.5 times the interquartile range past the upper or lower box hinges.
147 U, respectively, while those for individual contacts were 204 U and 94 U. While the within-subject distribution showed only a mild right skew and leptokurtosis (Pearson kurtosis 0.39, skewness 0.72), these characteristics were much more pronounced for the withincontact distribution (Pearson kurtosis 3.18, skewness 1.57). Analysis of effects on impedance over time We utilized all 94 subjects and 2851 impedance measurements for our GLMM analysis looking at significant predictors of impedance. Univariate analysis of individual effects and their interaction over time addressed the role of implantation side, implanted target, contact activity and position in the implanted lead. Electrically active contacts had significantly lower LS mean impedance than non-active contacts (P < 0.0001; Fig. 4, top) with a significant interaction with time (P < 0.0001). The impedance of subthalamic nucleus (STN) electrodes started out higher than those in globus
Within-subject and within-contact variability For this portion of our analysis, 77 subjects and 536 electrode contacts, representing 2556 impedance measurements, met the inclusion criteria of a follow-up period from at least 6 months up to 5 years. We calculated the standard deviation for impedance measurements within each subject, and within each contact. Probability distributions were constructed for each of these two conditions, and are shown in Fig. 3. Median impedance variability was 194 U for individual subjects, and 141 U for individual contacts. The upper and lower quartiles for individual subjects were 241 U and
Table 2 Designated milestone time points and respective number of subjects. Visit number
Designated follow-up time (months)
Number of subjects
1 2 3 4 5 6 7
0 1 3 6 12 24 60
82 61 54 53 55 49 29
Figure 2. The range of impedance measurements across subjects at each follow-up visit. Box plot depicting the range of impedances across all subjects at each milestone follow-up time point. Boxes indicate the range between the 25th and 75th percentiles (interquartile range), with the median (50th percentile) as the horizontal line within the box. Whiskers represent the overall range of values. Filled circles indicate outliers, defined by Tukey’s criteria as values outside of 1.5 times the interquartile range past the upper or lower box hinges.
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
Figure 3. The distribution of impedance variability within individual subjects and contacts. Longitudinal impedance variability, as depicted by probability distribution of standard deviations of measurements within individual subjects (n ¼ 77, top), and individual contacts (n ¼ 538, bottom). Solid vertical lines depict median values, while dotted vertical lines represent the 25th and 75th quartiles. The probability density function of these distributions is represented by the y axis, and indicates the likelihood of a random variable (in this case, impedance variability) to take on the corresponding value on the x axis. Top: distribution describing the set of standard deviations of impedance measurements within subjects (including all contacts on the quadripolar electrode), representing the range of variability experienced across all contacts of a given DBS electrode for our study population. Median variability was 194 U with an interquartile range of 147e241 U. Bottom: the distribution of standard deviations for the impedance measurements of each independent contact on each electrode in each subject, representing the range of variability from repeated measurements on a single electrode contact surface. Median variability was 141 U with an interquartile range of 94e204 U. The right skew and peaked, or leptokurtic, shape of these curves implies that most subjects experience a level of variability clustered around the median of these distributions, but a sizeable subset experienced variability substantially higher than the rest of the group.
pallidus pars interna (GPi), but then subsequently decreased at a faster rate, particularly after 1 year, with LS means showing a significant difference between targets (P ¼ 0.03; Fig. 4, bottom) as well as a significant interaction with time (P < 0.0001). The middle contact positions (contacts 1 and 2) showed consistently lower impedance than the most dorsal and ventral contacts (P < 0.0001), in the absence of a significant interaction with time. Side was significant (P < 0.0001), with the left sided electrodes slightly but consistently lower than the right side, also without significant interaction with time. Age and disease duration at surgery, gender and ethnicity were not significant. Disease diagnosis correlated almost exactly to implanted target and was excluded as a duplicate effect. We fitted the significant effects into a multivariate analysis. Results, shown in Table 3, revealed that time, as indicated by milestone follow-up visit, as well as whether or not a contact was active, contact position along the electrode, and the side and target of implantation were significant predictors of impedance.
Discussion Our analysis is the first attempt to study DBS impedance longitudinally, using standardized repeated measurements over a period of 6e60 months, and provides insight into DBS impedance
749
Figure 4. The effect of electrical activity and stereotactic target on impedance, and their interaction with time. This figure highlights two significant predictors of impedance, electrical activity and stereotactic target, and their interactions with time, from our generalized linear mixed model, described in Table 3. Data points represent least-square mean impedance values calculated for the specified effects, while balancing for the other covariates in the GLMM. Error bars represent standard error of the LS means. Top: Effect of electrical activity on impedance over time. Electrically active contacts exhibited significantly lower impedance as compared to inactive ones, and was a significant predictor in our generalized linear mixed model (P < 0.0001). Electrical activity also showed a significant interaction with time (P ¼ 0.0004), visible here as the differences in the slopes between the two lines. Bottom: Effect of implanted target on impedance over time, a significant predictor in our GLMM (P < 0.0001). Subthalamic (STN) electrodes started out with higher mean impedances but declined at a faster rate than electrodes targeting the globus pallidus pars interna (GPi) (P < 0.0001).
variability over time in a large sample of individuals. In addition, we identified a number of factors that were predictors of impedance behavior, including the duration of stimulation therapy (i.e., time), activity status of the contact, the implanted target, the position of the contacts on the implanted electrode and, to a certain extent, the side of implantation. The stability of DBS impedance over time is of clinical interest due to its ramifications on the reliability and consistency of therapeutic stimulation delivered with currently available devices that are based on constant voltage technology. We found a fairly large range of impedance variability across all subjects (Figs. 1 and 2), consistent with data previously reported [11]. In particular, outliers in these cross-sectional plots appear to be predominantly above the interquartile range, suggesting a tendency for extreme values, should they occur, to be due to an increase, as opposed to a decrease, in impedance. In addition to our initial cross-sectional survey, we measured impedance variability within individual subjects and electrode contacts over time, as depicted in our probability distributions in Fig. 3. These distributions exhibited a right skew with long upper tails, which led to positive Pearson kurtosis and skewness values (normal Gaussian distributions having zero kurtosis and skew). This effect was relatively mild for the within-subject distribution, likely due to the variability being calculated across a relatively heterogeneous group of electrode contacts within a given individual. However, in the within-contact distribution, representing the variability of repeated measures over a single unique electrode contact, this effect was notably more pronounced. While the median variability for impedance was 141 U, the rightward skew indicated that a sizable subset experienced significantly higher variability, with the upper quartile of electrode contacts displaying a range between 200 and 600 U. These results indicate impedance
750
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
Table 3 Generalized linear mixed model (GLMM) results, showing significant predictors of impedance. Main effect Time (per visit)a Side (ref: left)b Right Contact (ref: 2)b 1 3 0 Stereotactic target (ref: GPi)b STN Electrical activity (ref: active)b Non-active Two-way interactions Side stereotactic target (increased from GPi to STN) Right Left Contact electrical activity (increases from active to non-active) 2 1 0 3 Time (visit) electrical activity (decreases with visit) Active Non-active Time (visit) stereotactic target (decreases with visit) STN GPi
Effect size (LS means difference, in ohms)
F value
P value
22.2
12.7 29.9
<0.0001 <0.0001
19.6
<0.0001
4.8
<0.0001
168.4
<0.0001
6.9
0.009
6.4
0.0003
3.3
0.0004
7.99
<0.0001
28 19 90 117 86 150
106.2 49.6
106.6 107.3 133.5 134.4
18.8 26.2
54.4 5.8
GPi ¼ globus pallidus pars interna; STN ¼ subthalamic nucleus. a Average impedance decrease per visit. b Increase with respect to reference variable.
variability is not negligible, and that impedance stability over time cannot be taken for granted in subjects undergoing DBS therapy. In fact, average impedance values showed distinct, predictable temporal trends. Over the first month of therapy, active contacts showed a sizeable decrease in impedance, as compared to a slight increase in inactive contacts (Fig. 4, top). A gradual impedance increase was observed also for the active contacts later on. After the first year of stimulation, however, a gradual impedance decline was seen for all contacts, and the overall effect as seen in our GLMM model was a net decrease of 22 U per visit over the course of the seven milestone visits. Other factors, including the implanted target, contact position along the electrode and implanted side of the brain also were predictors of impedance in our population. Differences in local anatomy and tissue characteristics could contribute to the effect of implanted target, although the virtually exact correlation with disease diagnosis (i.e., PD patients received only STN electrodes, while GPi electrodes were all implanted in dystonic patients) could have played a role. The significant effect of contact position may be due to the fact that the inner contacts [1 and 2] happened to be activated more frequently than the outer leads (0 and 3), although these two factors did not completely co-correlate in the GLMM (Table 3). Differences in the chronic therapeutic stimulation settings required by PD and dystonia patients could also play a role, although this data was not tracked in our analysis. It is not immediately clear to us what role the side of implantation may have in predicting DBS impedance variability. The behavior of tissue impedance, especially with relation to time, has attracted research interest due to its potential to affect the amount of therapeutic stimulation delivered by DBS.
Computational models have quantified the dependency of the volume of neural tissue activation from DBS on impedance [6]. Measurements in animal models of DBS revealed significant impedance changes in implanted neurostimulator electrodes [7,9], generally describing an increase over the first few days of stimulation with a corresponding decrease in peak cathodic voltage delivered to tissue [7e9,16,17]. Studies in animals and humans have also recorded decreases in impedance in electrically active contacts [7,9,10,17], theorized to be an effect of a “cleaning” polarization of the electrode surface [7]. The observation of a fibrous sheath or tissue encapsulation layer around chronically implanted electrodes [18,19] suggest that at least some of the time-based variability of impedance may be related to a tissue response to implantation. This response may account for some of the earlier post-implantation increases in impedance observed in our data and previous studies, and may account for the higher proportion of outliers above, as opposed to below, our cross sectional impedance ranges (Figs. 1 and 2). Clinical evidence in human subjects have been more limited, due to the wide variety of treatment conditions and customized, patient-specific stimulation parameters employed at the bedside. A study of 32 patients implanted for tremor reported an average increase in therapeutic impedance over the first three months of stimulation, as well as an increase in the voltage of stimulation required to control tremor over that time period [20]. A study of 24 subjects implanted with pallidal DBS for dystonia described similar differences between electrically active vs. inactive electrode contacts that appeared temporally correlated to changes in stimulation parameters [10]. An analysis of 20 pairs of consecutive impedance measurements from 16 subjects showed significant cross-sectional variability across patients, but was unable to show statistically significant within-subject variability over time, presumably because of the difficulty in finding consistent therapeutic stimulation settings across multiple visits and patients [11]. Our study design did not track motor scores or other clinical outcome measures, and therefore cannot draw any direct conclusions with regards to the clinical relevance of impedance variability on the efficacy or stability of long-term DBS therapy for movement disorders. However, we can make some inferences and estimations based on the ranges of impedance variability and LS-means encountered in our data. The concept of total electrical energy delivered to tissue per second, or TEED1 s, has been used as an estimate of the “amount” of stimulation for a given patient with DBS [21,22]. This is represented by the following equation:
TEED1 s ¼
Voltage2 frequency pulse width 1s Impedance
Assuming a range of 141 U based on the median from our within-contact probability distribution (Fig. 3, bottom), at a fairly typical clinical setting of 3.0 V, 60 ms pulse width, and 130 Hz, a change in impedance from 400 U to 541 U would require an increase to 3.5 V at the higher impedance in order to deliver the same TEED1 s to tissue. At an increase from 1000 U to 1141 U, this would imply an increase from 3.0 V to 3.2 V. Similarly, taking for example the decrease in impedance from 1222 U to 1056 U as measured by LS means in electrically active contacts between Visit 1 and Visit 2 would require a reduction in voltage to 2.8 V in order to deliver an equivalent TEED1 s. Practitioners experienced in DBS management can easily adjust for these differences in the clinic. However, the possibility of such variability may account for the emergence of side effects or the reemergence of parkinsonian or dystonic symptoms at home not observed in the clinic, and may necessitate extra adjustments or follow-up visits by patients in order to main a stable level of
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
therapeutic benefit. New generations of DBS device designs may have the capability to correct for these fluctuations. DBS pulse generators designed to maintain constant current delivery across electrode contacts maintain a consistent peak cathodic voltage delivered to tissue, and are less vulnerable to variations in electrode impedance [8]. The latest generation of DBS devices approved for clinical use (Activa family, Medtronic, Inc., Minneapolis, MN), while still using traditional voltage control by default, are equipped with a limited set of constant-current capabilities, while upcoming devices designed exclusively for constant-current stimulation appear to be effective for Parkinson’s disease in a recent randomized multi-center controlled trial [23]. A number of limitations must be acknowledged. First, our measurements depended on the function of the implanted Soletra model neurostimulators, which derive impedance values from a test waveform generated by its output capacitor. This function was designed to detect gross hardware malfunctions and was never intended for precise scientific measurement of tissue or system impedances. A study testing this impedance measurement function using circuits with known impedances between 560 and 1800 U demonstrated reasonable reliability with typically identical repeated measures. However, it revealed notable limitations on accuracy that typically underestimated known impedances by 5e10%. This accuracy appeared to worsen with declining battery charge [24], potentially affecting the predictive value of the data points at 5 years or Visit 7, where many Soletra units in our center are typically nearing hardware end-of-life. In addition, the 2000 U upper limit of measurement might have also contributed to underestimation of impedance variability. Finally, due to the retrospective nature of this study, we cannot exclude the possibility of an unobserved effect that might have had a strong influence on impedance. In summary, this study offers descriptive statistics and group analysis of trends, adding to the small but growing body of scientific evidence describing the behavior of impedance in implanted DBS devices in human subjects. While not matching the precision of controlled laboratory measurements in in vitro and in-vivo animal models, our data, collected in a large population of patients, supports the notion of significant DBS impedance variability over time. Further studies are required to better characterize such variability, particularly at several years post-operatively, and to link these findings with clinical outcomes. References [1] Deuschl G, Schade-Brittinger C, Krack P, Volkmann J, Schäfer H, Bötzel K, et al. A randomized trial of deep-brain stimulation for Parkinson’s disease. N Engl J Med 2006;355(9):896e908. [2] Vidailhet M, Vercueil L, Houeto J-L, Krystkowiak P, Benabid AL, Cornu P, et al. Bilateral deep-brain stimulation of the globus pallidus in primary generalized dystonia. N Engl J Med 2005;352(5):459e67.
751
[3] Weaver FM, Follett K, Stern M, Hur K, Harris C, Marks WJ, et al. Bilateral deep brain stimulation vs best medical therapy for patients with advanced Parkinson disease: a randomized controlled trial. JAMA 2009;301(1):63e73. [4] Weaver FM, Follett KA, Stern M, Luo P, Harris CL, Hur K, et al. Randomized trial of deep brain stimulation for Parkinson disease: thirty-six-month outcomes. Neurology 2012;79(1):55e65. [5] Kumar R, Lozano AM, Sime E, Lang AE. Long-term follow-up of thalamic deep brain stimulation for essential and parkinsonian tremor. Neurology 2003; 61(11):1601e4. [6] Butson CR, Maks CB, McIntyre CC. Sources and effects of electrode impedance during deep brain stimulation. Clin Neurophysiol 2006;117(2):447e54. [7] Lempka SF, Miocinovic S, Johnson MD, Vitek JL, McIntyre CC. In vivo impedance spectroscopy of deep brain stimulation electrodes. J Neural Eng 2009; 6(4):046001. [8] Lempka SF, Johnson MD, Miocinovic S, Vitek JL, McIntyre CC. Current-controlled deep brain stimulation reduces in vivo voltage fluctuations observed during voltage-controlled stimulation. Clin Neurophysiol 2010;121(12):2128e33. [9] Steinke GK, Moffitt MA, Carcieri S, Jackson A, Makous J, Stoffregen W, et al. Acute impedance changes of DBS system & corresponding stimulation estimates. Mov Disord 2012;27(S1):S86. [10] Hemm S, Vayssiere N, Mennessier G, Cif L, Zanca M, Ravel P, et al. Evolution of brain impedance in dystonic patients treated by GPI electrical stimulation. Neuromodulation 2004;7(2):67e75. [11] Sillay KA, Chen JC, Montgomery EB. Long-term measurement of therapeutic electrode impedance in deep brain stimulation. Neuromodulation 2010;13(3): 195e200. [12] Tagliati M, Shils J, Sun C, Alterman R. Deep brain stimulation for dystonia. Expert Rev Med Devices 2004;1(1):33e41. [13] Shils JL, Tagliati M, Alterman RL. Neurophysiological monitoring during neurosurgery for movement disorders. In: Deletis V, Shils JL, editors. Neurophysiology in neurosurgery: a modern intraoperative approach. San Diego, CA: Academic Press; 2002. p. 393e436. [14] R Core Team. R: a language and environment for statistical computing [Internet]. R Foundation for Statistical Computing. Available from: http:// www.R-project.org; 2012. [15] Wickham H. ggplot2: elegant graphics for data analysis. New York: Springer; 2009. [16] Grill WM, Mortimer JT. Electrical properties of implant encapsulation tissue. Ann Biomed Eng 1994;22(1):23e33. [17] Johnson MD, Otto KJ, Kipke DR. Repeated voltage biasing improves unit recordings by reducing resistive tissue impedances. IEEE Trans Neural Syst Rehabil Eng 2005;13(2):160e5. [18] Haberler C, Alesch F, Mazal PR, Pilz P, Jellinger K, Pinter MM, et al. No tissue damage by chronic deep brain stimulation in Parkinson’s disease. Ann Neurol 2000;48(3):372e6. [19] Moss J, Ryder T, Aziz TZ, Graeber MB, Bain PG. Electron microscopy of tissue adherent to explanted electrodes in dystonia and Parkinson’s disease. Brain 2004;127(Pt 12):2755e63. [20] Benabid AL, Pollak P, Gervason C, Hoffmann D, Limousin P, Gay E, et al. Longterm suppression of tremor by chronic stimulation of the ventral intermediate thalamic nucleus. Lancet 1991;337(8738):403e6. [21] Moro E, Lang AE, Strafella AP, Poon YY, Arango PM, Dagher A, et al. Bilateral globus pallidus stimulation for Huntington’s disease. Ann Neurol 2004; 56(2):290e4. [22] Koss AM, Alterman RL, Tagliati M, Shils JL. Calculating total electrical energy delivered by deep brain stimulation systems. Ann Neurol 2005;58(1):168 [author reply168e9]. [23] Okun MS, Gallo BV, Mandybur G, Jagid J, Foote KD, Revilla FJ, et al. Subthalamic deep brain stimulation with a constant-current device in Parkinson’s disease: an open-label randomised controlled trial. Lancet Neurol 2012;11(2):140e9. [24] Patel NV, Oza CS, Wong S, Danish SF, Hargreaves EL. Impedance accuracy and reliability of Medtronic Soletra and Activa PC neuromodulation models as a function of battery charge. New Orleans, LA: Society for Neuroscience; 2012.
Contents lists available at SciVerse ScienceDirect
Brain Stimulation journal homepage: www.brainstimjrnl.com
Longitudinal Impedance Variability in Patients with Chronically Implanted DBS Devices Tyler Cheung a, Miriam Nuño a, Matilde Hoffman a, Maya Katz b, Camilla Kilbane c, Ron Alterman d, Michele Tagliati a, * a
Department of Neurology, Cedars-Sinai Medical Center, 8730 Alden Drive, Thalians E-238, Los Angeles, CA 90048, USA Department of Neurology, UCSF, San Francisco, CA, USA c Department of Neurology, Stanford University School of Medicine, Palo Alto, CA, USA d Department of Neurosurgery, Beth Israel Deaconess Medical Center, Boston, MA, USA b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 4 December 2012 Received in revised form 9 March 2013 Accepted 20 March 2013 Available online 22 April 2013
Background: Deep brain stimulation (DBS) is an effective therapy for advanced movement disorders, but its optimal use is still controversial. One factor that could play a role in the proper delivery of therapeutic stimulation by current DBS devices is the variability of the impedance at the interface between the electrode surface and surrounding tissue. Objective: To analyze variability and trends in the impedance of chronically-implanted DBS electrodes in subjects with movement disorders. Methods: We reviewed impedance values from medical records of DBS patients at an academic tertiarycare movement disorders center. The standard deviation of data recorded within individual subjects and single contacts were used as measures of longitudinal impedance variability. A generalized linear mixed model (GLMM) determined if a number of effects had significant influences on impedance. Results: We analyzed 2863 impedance measurements from 94 subjects. Median variability, for subjects with follow-up from 6 months to 5 years (n ¼ 77), was 194 U for individual subjects and 141 U for individual contacts, with a range spanning from 18 to over 600 U. The GLMM, incorporating all subjects (n ¼ 94), identified time, electrical activity, implanted target, contact position on the electrode and side of implantation as significant predictors of impedance. Age and disease duration at surgery, gender or ethnicity were not significant predictors. Conclusions: Our analysis suggests that a significant amount of impedance variability can be expected in chronically implanted DBS electrodes and indicates a number of factors with possible predictive value. Further studies are needed to link impedance characteristics to clinical outcomes. Ó 2013 Elsevier Inc. All rights reserved.
Keywords: Impedance Deep brain stimulation Movement disorders Globus pallidus Subthalamic nucleus
Introduction Deep brain stimulation (DBS) is an effective therapy for medication-refractory movement disorders including advanced Funding: This study was funded in part by the Parkinson Alliance. Financial disclosures: Tyler Cheung, MD, Maya Katz, MD, and Camilla Kilbane, MD, were beneficiaries of unrestricted fellowships from Medtronic, Inc. Ron L. Alterman, MD, has received consultation fees from Medtronic, Inc., unrelated to the conduct of this study. Michele Tagliati, MD, has received speaker honoraria from Medtronic, Inc., and consultation fees from St. Jude Medical, Inc. (formerly Advanced Neuromodulation Systems), Abbott Laboratories, Merz and Impax Laboratories, Inc., unrelated to the conduct of this study. Miriam Nuño, PhD, and Matilde Hoffman have nothing to disclose. This manuscript has been presented as an abstract at the 2012 International Congress of the Movement Disorders Society in Dublin, Ireland, but has not been previously submitted or published in any other journal. * Corresponding author. E-mail address: [email protected] (M. Tagliati). 1935-861X/$ e see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.brs.2013.03.010
Parkinson’s disease (PD), essential tremor (ET) and dystonia [1e5]. However, DBS mechanisms of action are not completely understood and, as a consequence, the optimal use of this therapy remains a source of controversy. One area of interest is the consistency of the therapeutic stimulation delivered by currently available neurostimulators, with particular attention on the electrical impedance of these implanted systems. Impedance, the measure of the opposition to current in an oscillatory circuit, is fundamentally related to voltage and frequency, two DBS parameters crucial to determining therapeutic effect. Currently available devices, controlled via constant-voltage designs, do not automatically adjust the amount of therapeutic stimulation to account for impedance fluctuations. Therefore, large changes in impedance could potentially affect clinical outcomes. Studies of impedance in computational and animal models have demonstrated changes over the time period following implantation, or activation of stimulation, particularly between the
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
interface between the metal electrode surface and surrounding tissue [6e9]. The behavior of impedance in human subjects is much more difficult to study and therefore less well characterized. Only a few studies have indicated trends that include decreased impedance in electrically active versus inactive electrode contacts and a wide range of differences between individual subjects [10,11]. These studies focused on impedance values derived from patients’ customized “therapeutic” impedances. While clinically relevant, these measures are difficult to compare both longitudinally and across a given sample of subjects due to differing follow-up times and wide variety of stimulation parameters (i.e., voltage, pulse widths, frequency, and electrode contact configuration). In order to fill some of these gaps, we designed a study to characterize the range of impedance variability in measurements within individual subjects, as well as within individual electrode contacts in each subject, over time. In addition, we investigated a number of variables, such as time, electrical activity, anatomic location and other demographic data that may have an influence on impedance. We took advantage of the standardized impedance measurement capability built into current generation devices. These measurements, routinely employed in the course of clinical care to detect hardware malfunction such as breaks or shorts in the circuitry, are performed at pre-defined and uniform parameters not necessarily reflective of therapeutic stimulation. Nevertheless, they automatically control for variables such as voltage, pulse duration, frequency, and electrode contact configuration, making them more amenable to repeated measures and more conducive to robust statistical analysis. We present a retrospective analysis of such data collected after surgical implantation in a large cohort of DBS patients. Methods Subjects We retrospectively reviewed the medical records of all patients treated with DBS in an academic tertiary-care movement disorders center between 2005 and 2010. The institutional review board approved all data collection methods and activities. Subjects were selected based on the criteria of having received DBS implantation and at least 1 follow-up visit where standardized single contact impedances were measured. DBS implantations were performed according to surgical practices previously described [12,13], using Model No. 3387 S electrodes and Soletra neurostimulators (Medtronic, Inc., Minneapolis, MN). Impedances were measured using the built-in function of the Soletra, accessed and read via a handheld DBS programming device (nVision model, Medtronic, Inc., Minneapolis, MN). Measurements were conducted at single monopolar configuration for each of the 4 contacts of the quadripolar 3387 electrode, using a standard setting of 3.0 V, 210 ms pulse width, and a frequency of 30 Hz. In addition to impedance, collected data included diagnosis, age and disease duration prior to surgery, gender, ethnicity, the side and implanted target, and the electrically active contact configuration at each follow-up visit. The data were collected at the following post-operative “milestones”: baseline (i.e., initial programming), 1 month, 3 months, 6 months, 1 year, and 2 years. In addition, data from a “long term,” (w5 year) follow up visit were also recorded. Available visits in the database were grouped as close to these milestones as possible. Impedance readings from re-implanted electrodes were discarded to minimize potential confounds from prior tissue trauma and inconsistency of anatomical location. Therapeutic impedances relative to the subject’s customized stimulation parameters were not analyzed.
747
Within-subject and within-contact variability In order to calculate how much impedance variability might occur across the electrodes implanted in an individual subject, as well as within each individual contact on a quadripolar electrode over time, we divided impedance readings by individual subject and further stratified them by individual contact (n ¼ 4) on each electrode. Subjects whose data were included in the distributions had a minimum follow-up period of 6 months, up to a maximum of 60 months. We calculated the standard deviation for each of these groups, thus obtaining two sets of standard deviations, describing the longitudinal variation in impedance for 1) individual subjects and 2) individual contacts. For each of these two data sets, we created a probability distribution and calculated its median and quartiles values, as well as kurtosis and skew. Evaluating multivariate effects influencing impedance over time In order to investigate the effects on impedance from the variables captured in our dataset, we incorporated them into a generalized linear mixed model (GLMM). Unlike standard regressions, in which all observations are independent of one another, the GLMM takes into account the fact that each subject contributes multiple repeated measures, creating correlated observations. This modeling approach also considers unbalanced longitudinal data (i.e., the differing number of subjects present at each follow-up time-point) as observed in this study, enabling the use of all values in our dataset. The GLMM allows estimation of how the covariates differ, how their effects change over cycles, and forms a basis for exploring interactions between predictors. We elected to treat time as a categorical variable, given the structure of the data which featured visits heavily clustered around the seven time points defined as “milestone visits.” We adjusted for unequally spaced repeated measures via a time-series-type of covariate structure widely available in GLMM modeling. In this study we assumed a spatial power law covariance structure. Effects were fitted to the GLMM in a stepwise fashion, and those that did not improve the Akaike Information Criterion (AIC) were discarded to prevent overfitting. Correlations among impedance measurements made on the same patient were modeled via mixed effects through the specification of a covariance structure. Effects were considered to be significant when P < 0.05 as determined by Type III tests for sum of squares. The results from these models are discussed in terms of least square (LS) means when evaluating fixed effects such as implanted target, electrical activity, side of implant, and contact position. LS means estimate the marginal means over a particular population and an approximate t-test to assess potential differences between the means being assessed. We incorporated significant effects along with relevant 2-way interaction terms into an overall multivariate model. Statistical analysis was conducted primarily using SAS version 9.1 for Windows (SAS Institute Inc., Cary, NC). Supplementary analysis was performed using R version 2.15.1 and the ggplot2 library [14,15]. Results Demographics We were able to tabulate 2863 impedance measurements in 94 patients. Twelve measurements were discarded as indicative of an “open circuit,” defined as an impedance value greater than 2000 U with a battery current draw below 11 mA. All subjects
748
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
Table 1 Demographics. Gender Female Male Age at surgery, years (SD) Parkinson’s disease Primary dystonia Essential tremor Chorea acanthocytosis Diagnosis Parkinson’s disease Dystonia Essential tremor Chorea acanthocytosis Anatomic location Subthalamic nucleus Globus pallidus pars interna Thalamus Ethnicity African American Asian Caucasian Hispanic Other Disease duration at surgery, months (SD) Parkinson’s disease Primary dystonia Essential tremor Chorea acanthocytosis
35 59 46 (22) 66 (8) 33 (18) 76 49 34 58 1 1 35 58 1 1 2 75 3 13 152 (83) 246 (174) 156 12
with Parkinson’s disease were implanted in the subthalamic nucleus (STN), while subjects with dystonia were implanted in the globus pallidus pars interna (GPi), with the exception of one subject implanted in STN. The subject with tremor was implanted in the ventral intermediate (VIM) nucleus of the thalamus, while the subject with chorea-acanthocytosis was also implanted in GPi. Further demographic data and the compositional background of our cohort are shown in Table 1, while post-operative follow-up times are show in Table 2. Visit 7 was less well attended and more variable than the other milestone visits. The ranges of impedances per visit across our entire cohort, as well as the overall range of impedances per subject, are visualized in Figs. 1 and 2.
Figure 1. The range of impedance measurements for individual subjects across time. Box plot illustrating the range of impedances for all individual subjects (n ¼ 94) throughout all time points, sorted by median impedance per subject. Boxes indicate the range between the 25th and 75th percentiles (interquartile range), with the median (50th percentile) as the horizontal line within the box. Whiskers represent the overall range of values. Filled circles indicate outliers, defined by Tukey’s criteria as values outside of 1.5 times the interquartile range past the upper or lower box hinges.
147 U, respectively, while those for individual contacts were 204 U and 94 U. While the within-subject distribution showed only a mild right skew and leptokurtosis (Pearson kurtosis 0.39, skewness 0.72), these characteristics were much more pronounced for the withincontact distribution (Pearson kurtosis 3.18, skewness 1.57). Analysis of effects on impedance over time We utilized all 94 subjects and 2851 impedance measurements for our GLMM analysis looking at significant predictors of impedance. Univariate analysis of individual effects and their interaction over time addressed the role of implantation side, implanted target, contact activity and position in the implanted lead. Electrically active contacts had significantly lower LS mean impedance than non-active contacts (P < 0.0001; Fig. 4, top) with a significant interaction with time (P < 0.0001). The impedance of subthalamic nucleus (STN) electrodes started out higher than those in globus
Within-subject and within-contact variability For this portion of our analysis, 77 subjects and 536 electrode contacts, representing 2556 impedance measurements, met the inclusion criteria of a follow-up period from at least 6 months up to 5 years. We calculated the standard deviation for impedance measurements within each subject, and within each contact. Probability distributions were constructed for each of these two conditions, and are shown in Fig. 3. Median impedance variability was 194 U for individual subjects, and 141 U for individual contacts. The upper and lower quartiles for individual subjects were 241 U and
Table 2 Designated milestone time points and respective number of subjects. Visit number
Designated follow-up time (months)
Number of subjects
1 2 3 4 5 6 7
0 1 3 6 12 24 60
82 61 54 53 55 49 29
Figure 2. The range of impedance measurements across subjects at each follow-up visit. Box plot depicting the range of impedances across all subjects at each milestone follow-up time point. Boxes indicate the range between the 25th and 75th percentiles (interquartile range), with the median (50th percentile) as the horizontal line within the box. Whiskers represent the overall range of values. Filled circles indicate outliers, defined by Tukey’s criteria as values outside of 1.5 times the interquartile range past the upper or lower box hinges.
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
Figure 3. The distribution of impedance variability within individual subjects and contacts. Longitudinal impedance variability, as depicted by probability distribution of standard deviations of measurements within individual subjects (n ¼ 77, top), and individual contacts (n ¼ 538, bottom). Solid vertical lines depict median values, while dotted vertical lines represent the 25th and 75th quartiles. The probability density function of these distributions is represented by the y axis, and indicates the likelihood of a random variable (in this case, impedance variability) to take on the corresponding value on the x axis. Top: distribution describing the set of standard deviations of impedance measurements within subjects (including all contacts on the quadripolar electrode), representing the range of variability experienced across all contacts of a given DBS electrode for our study population. Median variability was 194 U with an interquartile range of 147e241 U. Bottom: the distribution of standard deviations for the impedance measurements of each independent contact on each electrode in each subject, representing the range of variability from repeated measurements on a single electrode contact surface. Median variability was 141 U with an interquartile range of 94e204 U. The right skew and peaked, or leptokurtic, shape of these curves implies that most subjects experience a level of variability clustered around the median of these distributions, but a sizeable subset experienced variability substantially higher than the rest of the group.
pallidus pars interna (GPi), but then subsequently decreased at a faster rate, particularly after 1 year, with LS means showing a significant difference between targets (P ¼ 0.03; Fig. 4, bottom) as well as a significant interaction with time (P < 0.0001). The middle contact positions (contacts 1 and 2) showed consistently lower impedance than the most dorsal and ventral contacts (P < 0.0001), in the absence of a significant interaction with time. Side was significant (P < 0.0001), with the left sided electrodes slightly but consistently lower than the right side, also without significant interaction with time. Age and disease duration at surgery, gender and ethnicity were not significant. Disease diagnosis correlated almost exactly to implanted target and was excluded as a duplicate effect. We fitted the significant effects into a multivariate analysis. Results, shown in Table 3, revealed that time, as indicated by milestone follow-up visit, as well as whether or not a contact was active, contact position along the electrode, and the side and target of implantation were significant predictors of impedance.
Discussion Our analysis is the first attempt to study DBS impedance longitudinally, using standardized repeated measurements over a period of 6e60 months, and provides insight into DBS impedance
749
Figure 4. The effect of electrical activity and stereotactic target on impedance, and their interaction with time. This figure highlights two significant predictors of impedance, electrical activity and stereotactic target, and their interactions with time, from our generalized linear mixed model, described in Table 3. Data points represent least-square mean impedance values calculated for the specified effects, while balancing for the other covariates in the GLMM. Error bars represent standard error of the LS means. Top: Effect of electrical activity on impedance over time. Electrically active contacts exhibited significantly lower impedance as compared to inactive ones, and was a significant predictor in our generalized linear mixed model (P < 0.0001). Electrical activity also showed a significant interaction with time (P ¼ 0.0004), visible here as the differences in the slopes between the two lines. Bottom: Effect of implanted target on impedance over time, a significant predictor in our GLMM (P < 0.0001). Subthalamic (STN) electrodes started out with higher mean impedances but declined at a faster rate than electrodes targeting the globus pallidus pars interna (GPi) (P < 0.0001).
variability over time in a large sample of individuals. In addition, we identified a number of factors that were predictors of impedance behavior, including the duration of stimulation therapy (i.e., time), activity status of the contact, the implanted target, the position of the contacts on the implanted electrode and, to a certain extent, the side of implantation. The stability of DBS impedance over time is of clinical interest due to its ramifications on the reliability and consistency of therapeutic stimulation delivered with currently available devices that are based on constant voltage technology. We found a fairly large range of impedance variability across all subjects (Figs. 1 and 2), consistent with data previously reported [11]. In particular, outliers in these cross-sectional plots appear to be predominantly above the interquartile range, suggesting a tendency for extreme values, should they occur, to be due to an increase, as opposed to a decrease, in impedance. In addition to our initial cross-sectional survey, we measured impedance variability within individual subjects and electrode contacts over time, as depicted in our probability distributions in Fig. 3. These distributions exhibited a right skew with long upper tails, which led to positive Pearson kurtosis and skewness values (normal Gaussian distributions having zero kurtosis and skew). This effect was relatively mild for the within-subject distribution, likely due to the variability being calculated across a relatively heterogeneous group of electrode contacts within a given individual. However, in the within-contact distribution, representing the variability of repeated measures over a single unique electrode contact, this effect was notably more pronounced. While the median variability for impedance was 141 U, the rightward skew indicated that a sizable subset experienced significantly higher variability, with the upper quartile of electrode contacts displaying a range between 200 and 600 U. These results indicate impedance
750
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
Table 3 Generalized linear mixed model (GLMM) results, showing significant predictors of impedance. Main effect Time (per visit)a Side (ref: left)b Right Contact (ref: 2)b 1 3 0 Stereotactic target (ref: GPi)b STN Electrical activity (ref: active)b Non-active Two-way interactions Side stereotactic target (increased from GPi to STN) Right Left Contact electrical activity (increases from active to non-active) 2 1 0 3 Time (visit) electrical activity (decreases with visit) Active Non-active Time (visit) stereotactic target (decreases with visit) STN GPi
Effect size (LS means difference, in ohms)
F value
P value
22.2
12.7 29.9
<0.0001 <0.0001
19.6
<0.0001
4.8
<0.0001
168.4
<0.0001
6.9
0.009
6.4
0.0003
3.3
0.0004
7.99
<0.0001
28 19 90 117 86 150
106.2 49.6
106.6 107.3 133.5 134.4
18.8 26.2
54.4 5.8
GPi ¼ globus pallidus pars interna; STN ¼ subthalamic nucleus. a Average impedance decrease per visit. b Increase with respect to reference variable.
variability is not negligible, and that impedance stability over time cannot be taken for granted in subjects undergoing DBS therapy. In fact, average impedance values showed distinct, predictable temporal trends. Over the first month of therapy, active contacts showed a sizeable decrease in impedance, as compared to a slight increase in inactive contacts (Fig. 4, top). A gradual impedance increase was observed also for the active contacts later on. After the first year of stimulation, however, a gradual impedance decline was seen for all contacts, and the overall effect as seen in our GLMM model was a net decrease of 22 U per visit over the course of the seven milestone visits. Other factors, including the implanted target, contact position along the electrode and implanted side of the brain also were predictors of impedance in our population. Differences in local anatomy and tissue characteristics could contribute to the effect of implanted target, although the virtually exact correlation with disease diagnosis (i.e., PD patients received only STN electrodes, while GPi electrodes were all implanted in dystonic patients) could have played a role. The significant effect of contact position may be due to the fact that the inner contacts [1 and 2] happened to be activated more frequently than the outer leads (0 and 3), although these two factors did not completely co-correlate in the GLMM (Table 3). Differences in the chronic therapeutic stimulation settings required by PD and dystonia patients could also play a role, although this data was not tracked in our analysis. It is not immediately clear to us what role the side of implantation may have in predicting DBS impedance variability. The behavior of tissue impedance, especially with relation to time, has attracted research interest due to its potential to affect the amount of therapeutic stimulation delivered by DBS.
Computational models have quantified the dependency of the volume of neural tissue activation from DBS on impedance [6]. Measurements in animal models of DBS revealed significant impedance changes in implanted neurostimulator electrodes [7,9], generally describing an increase over the first few days of stimulation with a corresponding decrease in peak cathodic voltage delivered to tissue [7e9,16,17]. Studies in animals and humans have also recorded decreases in impedance in electrically active contacts [7,9,10,17], theorized to be an effect of a “cleaning” polarization of the electrode surface [7]. The observation of a fibrous sheath or tissue encapsulation layer around chronically implanted electrodes [18,19] suggest that at least some of the time-based variability of impedance may be related to a tissue response to implantation. This response may account for some of the earlier post-implantation increases in impedance observed in our data and previous studies, and may account for the higher proportion of outliers above, as opposed to below, our cross sectional impedance ranges (Figs. 1 and 2). Clinical evidence in human subjects have been more limited, due to the wide variety of treatment conditions and customized, patient-specific stimulation parameters employed at the bedside. A study of 32 patients implanted for tremor reported an average increase in therapeutic impedance over the first three months of stimulation, as well as an increase in the voltage of stimulation required to control tremor over that time period [20]. A study of 24 subjects implanted with pallidal DBS for dystonia described similar differences between electrically active vs. inactive electrode contacts that appeared temporally correlated to changes in stimulation parameters [10]. An analysis of 20 pairs of consecutive impedance measurements from 16 subjects showed significant cross-sectional variability across patients, but was unable to show statistically significant within-subject variability over time, presumably because of the difficulty in finding consistent therapeutic stimulation settings across multiple visits and patients [11]. Our study design did not track motor scores or other clinical outcome measures, and therefore cannot draw any direct conclusions with regards to the clinical relevance of impedance variability on the efficacy or stability of long-term DBS therapy for movement disorders. However, we can make some inferences and estimations based on the ranges of impedance variability and LS-means encountered in our data. The concept of total electrical energy delivered to tissue per second, or TEED1 s, has been used as an estimate of the “amount” of stimulation for a given patient with DBS [21,22]. This is represented by the following equation:
TEED1 s ¼
Voltage2 frequency pulse width 1s Impedance
Assuming a range of 141 U based on the median from our within-contact probability distribution (Fig. 3, bottom), at a fairly typical clinical setting of 3.0 V, 60 ms pulse width, and 130 Hz, a change in impedance from 400 U to 541 U would require an increase to 3.5 V at the higher impedance in order to deliver the same TEED1 s to tissue. At an increase from 1000 U to 1141 U, this would imply an increase from 3.0 V to 3.2 V. Similarly, taking for example the decrease in impedance from 1222 U to 1056 U as measured by LS means in electrically active contacts between Visit 1 and Visit 2 would require a reduction in voltage to 2.8 V in order to deliver an equivalent TEED1 s. Practitioners experienced in DBS management can easily adjust for these differences in the clinic. However, the possibility of such variability may account for the emergence of side effects or the reemergence of parkinsonian or dystonic symptoms at home not observed in the clinic, and may necessitate extra adjustments or follow-up visits by patients in order to main a stable level of
T. Cheung et al. / Brain Stimulation 6 (2013) 746e751
therapeutic benefit. New generations of DBS device designs may have the capability to correct for these fluctuations. DBS pulse generators designed to maintain constant current delivery across electrode contacts maintain a consistent peak cathodic voltage delivered to tissue, and are less vulnerable to variations in electrode impedance [8]. The latest generation of DBS devices approved for clinical use (Activa family, Medtronic, Inc., Minneapolis, MN), while still using traditional voltage control by default, are equipped with a limited set of constant-current capabilities, while upcoming devices designed exclusively for constant-current stimulation appear to be effective for Parkinson’s disease in a recent randomized multi-center controlled trial [23]. A number of limitations must be acknowledged. First, our measurements depended on the function of the implanted Soletra model neurostimulators, which derive impedance values from a test waveform generated by its output capacitor. This function was designed to detect gross hardware malfunctions and was never intended for precise scientific measurement of tissue or system impedances. A study testing this impedance measurement function using circuits with known impedances between 560 and 1800 U demonstrated reasonable reliability with typically identical repeated measures. However, it revealed notable limitations on accuracy that typically underestimated known impedances by 5e10%. This accuracy appeared to worsen with declining battery charge [24], potentially affecting the predictive value of the data points at 5 years or Visit 7, where many Soletra units in our center are typically nearing hardware end-of-life. In addition, the 2000 U upper limit of measurement might have also contributed to underestimation of impedance variability. Finally, due to the retrospective nature of this study, we cannot exclude the possibility of an unobserved effect that might have had a strong influence on impedance. In summary, this study offers descriptive statistics and group analysis of trends, adding to the small but growing body of scientific evidence describing the behavior of impedance in implanted DBS devices in human subjects. While not matching the precision of controlled laboratory measurements in in vitro and in-vivo animal models, our data, collected in a large population of patients, supports the notion of significant DBS impedance variability over time. Further studies are required to better characterize such variability, particularly at several years post-operatively, and to link these findings with clinical outcomes. References [1] Deuschl G, Schade-Brittinger C, Krack P, Volkmann J, Schäfer H, Bötzel K, et al. A randomized trial of deep-brain stimulation for Parkinson’s disease. N Engl J Med 2006;355(9):896e908. [2] Vidailhet M, Vercueil L, Houeto J-L, Krystkowiak P, Benabid AL, Cornu P, et al. Bilateral deep-brain stimulation of the globus pallidus in primary generalized dystonia. N Engl J Med 2005;352(5):459e67.
751
[3] Weaver FM, Follett K, Stern M, Hur K, Harris C, Marks WJ, et al. Bilateral deep brain stimulation vs best medical therapy for patients with advanced Parkinson disease: a randomized controlled trial. JAMA 2009;301(1):63e73. [4] Weaver FM, Follett KA, Stern M, Luo P, Harris CL, Hur K, et al. Randomized trial of deep brain stimulation for Parkinson disease: thirty-six-month outcomes. Neurology 2012;79(1):55e65. [5] Kumar R, Lozano AM, Sime E, Lang AE. Long-term follow-up of thalamic deep brain stimulation for essential and parkinsonian tremor. Neurology 2003; 61(11):1601e4. [6] Butson CR, Maks CB, McIntyre CC. Sources and effects of electrode impedance during deep brain stimulation. Clin Neurophysiol 2006;117(2):447e54. [7] Lempka SF, Miocinovic S, Johnson MD, Vitek JL, McIntyre CC. In vivo impedance spectroscopy of deep brain stimulation electrodes. J Neural Eng 2009; 6(4):046001. [8] Lempka SF, Johnson MD, Miocinovic S, Vitek JL, McIntyre CC. Current-controlled deep brain stimulation reduces in vivo voltage fluctuations observed during voltage-controlled stimulation. Clin Neurophysiol 2010;121(12):2128e33. [9] Steinke GK, Moffitt MA, Carcieri S, Jackson A, Makous J, Stoffregen W, et al. Acute impedance changes of DBS system & corresponding stimulation estimates. Mov Disord 2012;27(S1):S86. [10] Hemm S, Vayssiere N, Mennessier G, Cif L, Zanca M, Ravel P, et al. Evolution of brain impedance in dystonic patients treated by GPI electrical stimulation. Neuromodulation 2004;7(2):67e75. [11] Sillay KA, Chen JC, Montgomery EB. Long-term measurement of therapeutic electrode impedance in deep brain stimulation. Neuromodulation 2010;13(3): 195e200. [12] Tagliati M, Shils J, Sun C, Alterman R. Deep brain stimulation for dystonia. Expert Rev Med Devices 2004;1(1):33e41. [13] Shils JL, Tagliati M, Alterman RL. Neurophysiological monitoring during neurosurgery for movement disorders. In: Deletis V, Shils JL, editors. Neurophysiology in neurosurgery: a modern intraoperative approach. San Diego, CA: Academic Press; 2002. p. 393e436. [14] R Core Team. R: a language and environment for statistical computing [Internet]. R Foundation for Statistical Computing. Available from: http:// www.R-project.org; 2012. [15] Wickham H. ggplot2: elegant graphics for data analysis. New York: Springer; 2009. [16] Grill WM, Mortimer JT. Electrical properties of implant encapsulation tissue. Ann Biomed Eng 1994;22(1):23e33. [17] Johnson MD, Otto KJ, Kipke DR. Repeated voltage biasing improves unit recordings by reducing resistive tissue impedances. IEEE Trans Neural Syst Rehabil Eng 2005;13(2):160e5. [18] Haberler C, Alesch F, Mazal PR, Pilz P, Jellinger K, Pinter MM, et al. No tissue damage by chronic deep brain stimulation in Parkinson’s disease. Ann Neurol 2000;48(3):372e6. [19] Moss J, Ryder T, Aziz TZ, Graeber MB, Bain PG. Electron microscopy of tissue adherent to explanted electrodes in dystonia and Parkinson’s disease. Brain 2004;127(Pt 12):2755e63. [20] Benabid AL, Pollak P, Gervason C, Hoffmann D, Limousin P, Gay E, et al. Longterm suppression of tremor by chronic stimulation of the ventral intermediate thalamic nucleus. Lancet 1991;337(8738):403e6. [21] Moro E, Lang AE, Strafella AP, Poon YY, Arango PM, Dagher A, et al. Bilateral globus pallidus stimulation for Huntington’s disease. Ann Neurol 2004; 56(2):290e4. [22] Koss AM, Alterman RL, Tagliati M, Shils JL. Calculating total electrical energy delivered by deep brain stimulation systems. Ann Neurol 2005;58(1):168 [author reply168e9]. [23] Okun MS, Gallo BV, Mandybur G, Jagid J, Foote KD, Revilla FJ, et al. Subthalamic deep brain stimulation with a constant-current device in Parkinson’s disease: an open-label randomised controlled trial. Lancet Neurol 2012;11(2):140e9. [24] Patel NV, Oza CS, Wong S, Danish SF, Hargreaves EL. Impedance accuracy and reliability of Medtronic Soletra and Activa PC neuromodulation models as a function of battery charge. New Orleans, LA: Society for Neuroscience; 2012.