Preoperative indicators of clinical outcome following stereotaxic pallidotomy

Posteroventral pallidotomy has been shown to improve akinetic symptoms as well as levodopa-induced dyskinesias in patients with advanced Parkinson's disease (PD).^1-4 In experimental animal models, parkinsonian signs have been associated with excessive pallidofugal inhibitory outflow, with concomitant suppression of the cortico-striato-pallido-thalamo-cortical (CSPTC) motor loop.^5,6 Analogously, in keeping with autoradiographic studies in animal models,^7,8 PET with¹⁸ F-fluorodeoxyglucose (FDG) in PD patients has revealed abnormal expression of a metabolic brain network characterized by relative increases in lentiform and thalamic glucose metabolism covarying with decreased motor cortical metabolic activity.^9-12 Indeed, recent studies have revealed postoperative increases in the activity of premotor and supplementary motor areas following pallidotomy, suggesting modulation of CSPTC circuits by this surgery.^13-15

Although pallidotomy is a promising treatment for PD, it is unknown which clinical features are associated with favorable clinical outcomes. We previously noted a significant correlation between preoperative lentiform metabolism and operative outcome in 10 patients with idiopathic PD undergoing unilateral pallidotomy.¹⁵ Based upon this finding, we suggested that preoperative FDG/PET may be useful in selecting optimal candidates for surgery. In this study we assessed the utility of FDG/PET and quantitative clinical motor performance indices as potential pre-operative indicators of surgical success. To this end, we studied 22 subsequent PD patients undergoing unilateral pallidotomy. We also examined the relationship of clinical outcome with intraoperative measurements of spontaneous pallidal neuronal activity and with postoperative MRI measurements of lesion size and location.

Materials and methods. Patients. We studied 22 patients with moderately advanced Parkinson's disease (13 men and nine women; mean age 57.2 ± 6.8 years; mean Hoehn and Yahr[H&Y] score 3.6 ± 0.6; mean disease duration 12.7 ± 5.1 years). A diagnosis of PD was made if the patient had "pure" parkinsonism without a history of known causative factors such as encephalitis or neuroleptic treatment and did not have dementia, supranuclear gaze abnormalities, or ataxia. In all patients a convincing response to levodopa was evident at the time of diagnosis. In all patients, family histories were negative for neurodegenerative illnesses.

Patients were considered for unilateral pallidotomy because of (1) bradykinesia and rigidity-predominant PD, and (2) severe response fluctuations, or disabling on-state dyskinesias, or both.¹⁶ The hemisphere to undergo surgery was determined clinically as that contralateral to the more affected limbs. The clinical characteristics of these patients appear in the table. (This cohort of PD patients was entirely different from that reported by us previously.¹⁵)

All patients were assessed according to the Core Assessment Program for Intracerebral Transplantation protocol (CAPIT).¹⁷ Quantitative assessment by CAPIT consisted of three timed tests, pronation-supination, hand-arm movement between two points, and finger dexterity. Preoperative measurements of mean limb CAPIT scores were obtained 12 hours off medications and again in the best on state following a therapeutic dose of dopaminergic medication. We computed a preoperative levodopa responsiveness index (LRI), defined as (off - on) / off× 100%, in which on and off refer to summed limb CAPIT scores measured in the respective pharmacologic states defined above.

CAPIT assessment was repeated 3 months following pallidotomy. Surgical outcome was determined as the percent clinical improvement (PCI), defined as(pre - post) / pre × 100%, where pre and post refer to preoperative and postoperative summed limb CAPIT scores measured in the off state. PCI measures were obtained separately for limbs ipsilateral and contralateral to the surgical lesion.

Magnetic resonance imaging. MRI before and immediately (<12 hours) following surgery was obtained for all patients on a 1.5-T scanner. Patients were imaged with T1- and T2-weighted sequences. T1 images were acquired in approximately 6 minutes with a 3D magnetization prepared rapid gradient echo sequence (MPRAGE)¹⁸ with matrix size 128× 180 × 256 giving 1- to 1.5-mm resolution in each dimension. We also acquired T2-weighted images in approximately 3 minutes with a whole-brain multisection fast spin echo (FSE) sequence (TR = 2,500 msec, TE = 90 msec, 3 mm section thickness, 250 × 256 matrix size, in-plane resolution ≈ 0.8 × 0.8 mm). These images were used to verify normal basal ganglia signal preoperatively and for postoperative assessment of lesion localization and volume.^19,20

Positron emission tomography. All 22 patients underwent quantitative FDG/PET imaging prior to surgery (mean interval 28 ± 24 days) as described previously.^10,15 The patients were awake during PET imaging; all antiparkinsonian medications were withdrawn at least 12 hours beforehand. Patients were imaged using the Advance tomograph at North Shore University Hospital (General Electric Medical Systems, Milwaukee, WI). The performance characteristics of this instrument have been described elsewhere.²⁰ This 18-ring bismuth germanate whole-body tomograph produces 35 sections with an axial field of view of 15 cm and a resolution of 4.2 mm (FWHM) in all directions. Ethical permission for these studies was obtained from the Institute Review Boards of North Shore University Hospital, Manhasset, NY, and the Hospital for Joint Diseases, New York, NY.

In each PET study patients were positioned in the imager using a Laitinen stereoadapter²¹ with three-dimensional laser alignment with reference to the orbitomeatal line. All studies were performed with the subject's eyes open in a dimly lit room and minimal auditory stimulation. The time course of ¹⁸F radioactivity was determined by sampling radial arterial blood. We calculated global and regional cerebral metabolic rates for glucose (GMR and rCMRGlc, respectively) in all FDG/PET studies on a pixel by pixel basis employing the autoradiographic method.^10,22,23

All images were analyzed using standardized regions of interest (ROIs) as described by us previously.^10,11 ROI analysis was performed on 256 × 256 PET reconstructions using a SUN microcomputer(SUN Microsystems, Mountain View, CA) with Scan/VP Software.²⁴ We defined the following motor system ROIs in the operated hemisphere: cerebellum, caudate and lentiform nuclei, thalamus, and the lateral frontal and paracentral cortices (corresponding respectively to the lateral premotor and supplementary motor areas[SMA]).¹⁰ To reduce intersubject variability, regional metabolic measurements were normalized by global hemispheric values(rCMRGlc/GMR).

Pallidotomy and single-cell recording. Stereotaxic pallidotomy was subsequently performed as described by us previously.¹⁶ Spontaneous pallidal single unit activity was recorded intraoperatively in an awake medication-free state.²⁵ A protective cannula, fixed to the stereotaxic frame, guided the microelectrode through a twist drill hole along the distance to the point 15 mm above the target point. A microdrive was employed to provide micrometer-graded extrusion of the microelectrode tip from the cannula. One to three electrode trajectories were performed for each patient. Single unit activity was recorded extracellularly with a high impedance tungsten-tip disposable microelectrode. Investigators blinded to the operative outcome analyzed spontaneous spike activity off line and calculated firing rate/time histograms, and interspike and interburst interval histograms, for a total of 978 internal pallidal (GPi) cells with an average of 44.5 ± 17.2 cells per patient. Only the spikes that were clearly separable from the background noise at a ratio greater than 2:1 were analyzed. The average spontaneous activity (spikes/second) for GPi cells was measured for each patient.

Lesion assessment. We assessed lesion location and volume using both T1- and T2-weighted MRI images. On postoperative MRI, the lesion was circular, comprising two concentric internal zones and a peripheral rim of irregular perilesional edema. The inner zone was characterized by isointensity on T2-weighted images. The outer zone, corresponding to the probable permanent lesion,¹⁹ was characterized by low intensity on T1-weighted images and high intensity in T2-weighted images. In the acute-phase MRI images, perilesional edema was minimal and was visualized on T2-weighted images.

In all cases, we measured lesion location with respect to the mid AC-PC point. Additionally, because of intersubject variability in brain size and AC-PC length, we compared lesion location across subjects by standardizing the individual MRI lesion coordinates by scaling with respect to the length of the AC-PC line.²⁶ Lesion volume was computed utilizing the cross-sectional area of the outer zone on parallel transverse planes and measurements of the length of the lesion cylinder.

Data analysis. Clinical correlation. We correlated the surgical outcome measure (PCI) with the following clinical variables: 1) age and duration of illness, 2) preoperative disease severity assessed by CAPIT, and 3) the preoperative LRI.

Metabolic correlation. We correlated the PCI surgical outcome measure with preoperative rCMRGlc/GMR values measured in the following ROIs in the operated hemisphere: cerebellum, caudate and lentiform nuclei, thalamus, and lateral premotor cortex and SMA.

Physiologic correlation. We correlated intraoperative measurements of spontaneous GPi firing rates with preoperative CAPIT scores as well as with the PCI measures of surgical outcome.

Anatomic correlation. We used multiple linear regression to correlate standardized lesion coordinates and lesion volume with PCI surgical outcome measures.

All correlations were reported as Pearson product moment correlation coefficients. The results of the correlational analyses were considered significant for p < 0.05. All statistical analyses were performed using SAS (SAS Institute, Cary, NC).

Results. Surgical outcome measures. Preoperative limb CAPIT scores and PCI surgical outcome measures are presented in the table. Preoperative CAPIT limb scores were 75.3 ± 30.5 contralateral to the operated side and 59.8 ± 13.9 ipsilaterally. Mean CAPIT scores were significantly reduced postoperatively in both contralateral (52.8 ± 29.0, p < 0.005) and ipsilateral limbs (49.9 ± 14.6, p < 0.005). Mean PCI surgical outcome measures were 30.9 ± 15.5% for the contralateral limbs and 18.1 ± 14.3% for the ipsilateral limbs.

Correlations with surgical outcome. Clinical correlation. Correlations between PCI for either body side and subject age or disease duration were not statistically significant. Additionally, preoperative off state CAPIT scores did not correlate significantly with PCI for either body side. Nonetheless, the preoperative levodopa responsiveness index (LRI) for the contralateral limbs correlated significantly with operative improvement on the body side (r = 0.60, p< 0.005; figure 1).

Metabolic correlation. Preoperative off state CAPIT scores and LRI measures did not correlate significantly with glucose metabolism in any of the motor regions assessed with preoperative FDG/PET. Nonetheless, ipsilateral lentiform metabolism correlated significantly with contralateral PCI measures (r = 0.71, p < 0.0005; figure 2A). Ipsilateral lentiform rCMRGlc also correlated significantly with the ipsilateral percent CAPIT improvement (r = 0.60, p < 0.005;figure 2B). Regional glucose metabolism in the other sampled motor regions did not correlate significantly with PCI. Examples of preoperative FDG/PET images from patients with optimal and suboptimal operative responses (patients 2 and 19, respectively) are presented in figure 3.

Physiologic correlation. The average spontaneous activity of GPi neurons was 101.0 ± 21.8 spikes per second (range 60-134). Spontaneous GPi firing rates did not correlate significantly with preoperative CAPIT ratings or with PCI values obtained for either body side.

Anatomic correlation. In all patients with ventral end of the lesion cylinder was localized to the posteromedial globus pallidus on postoperative MRI. The location of the ventral end of the lesion was x = 20.2 ± 1.7 mm (range 16.5-23.3 mm), y = 0.6± 1.6 mm (range -2.4-3.5 mm), z = -5.6 ± 1.4 mm(range -7.6 to -2.3 mm). When these coordinates were standardized to an AC-PC length of 23 mm for between-subject comparison, the ventral end of the lesion was measured as x = 19.6 ± 1.4 mm (range 17.3-21.8 mm), y = 0.7 ± 1.6 mm (range -1.9-2.7 mm), z = -4.4± 1.2 mm (range -6.0 to -1.7 mm). The average volume of the lesion was 262 ± 76 mm³. We found no significant correlation between clinical outcome and MRI measurements of lesion position (actual or standardized) and volume.

Prediction of surgical outcome. To evaluate the utility of preoperative clinical-pharmacologic assessment of FDG/PET as a predictor of surgical response, we performed linear regression analysis using either contralateral limb LRI or ipsilateral lentiform rCMRGlc/GMR as predictive variables, and contralateral PCI as the outcome variable. The best single predictor was preoperative ipsilateral lentiform metabolism, accounting for 50.4% of the outcome variance [F(1,20) = 22.74, p < 0.001], as compared with LRI, which accounted for only 36.5% of the variance [F(1,20) = 13.47, p = 0.002]. In combining preoperative FDG/PET and LRI together in a whole model analysis to predict PCI, both variables accounted for a total of 71% of the variance in the outcome measure (PCI = 88.6 lentiform rCMRGlc/GMR + 36.5 LRI - 96.0; p < 0.001). This suggests that both preoperative measures are individually predictive of outcome, albeit FDG/PET is more accurate. Nonetheless, the use of both PET and LRI preoperative measures in combination can afford a superior indicator of postoperative motor improvement.

Discussion. Our findings indicate that the clinical outcome of unilateral pallidotomy for PD can be predicted accurately by the quantification of several clinical and imaging indicators that are amenable to measurement prior to surgery. In this study we confirmed our earlier observation that preoperative rates of lentiform glucose utilization measured with FDG/PET can be highly predictive of outcome.¹⁵ Moreover, quantitative assessment of the levodopa response preoperatively can serve as an additional predictor of operative benefit. By contrast, preoperative measures of disease severity as well as intraoperative determination of spontaneous pallidal neuronal activity were uncorrelated with surgical outcome.

In an earlier FDG/PET study of 10 PD patients undergoing pallidotomy, we noted that preoperative measures of lentiform metabolism accounted for approximately 50% of patient variability in motor improvement up to 6 months following pallidotomy.¹⁵ In the current study of 22 subsequent PD patients imaged on a higher-resolution tomograph, we found that preoperative FDG/PET predicted clinical outcome with a nearly identical degree of accuracy. These findings support our contention that optimal pallidotomy candidates have functionally overactive lentiform nuclei. Indeed, the two patients with no operative improvement in CAPIT ratings also had the lowest preoperative rates of lentiform glucose metabolism measured in the study cohort.

As postulated previously, preoperative lentiform metabolism may define a functional set point for the CSPTC motor loop.¹⁵ Our data indicate that approximately half of the individual variability in clinical outcome is determined by the position of this metabolic set point prior to surgery: patients with greater preoperative lentiform glucose metabolism tend to have better outcomes than those beginning with comparatively lower set points. Moreover, in accordance with our previous findings, we found that measures of lentiform glucose metabolism and quantitative surgical outcome are each uncorrelated with objective clinical indices of motor dysfunction obtained prior to surgery.^9,10,15 Thus, while highly correlated with preoperative FDG/PET, the surgical outcome measure is not a simple manifestation of the severity of clinical manifestations in the preoperative state.

The pathophysiologic basis for elevations in resting-state lentiform metabolism in PD has been discussed previously.^9,10,15 Elevations in local rates of glucose utilization in both pallidal segments have been noted in experimental animal models of parkinsonism.^7,8 It has been postulated that nigrostriatal dopamine cell loss gives rise to increases in the activity of inhibitory afferents to the external pallidum from the putamen as well as increases in excitatory afferents to the GPi from the subthalamic nucleus(STN). Because rates of glucose metabolism are determined mainly by afferent synaptic activity,^27-29 pallidal hypermetabolism is an expected consequence of nigrostriatal dopaminergic dysfunction in both experimental animal models of parkinsonism as well as in classical PD in humans.¹⁰ Indeed, we have used FDG/PET in PD patients with regional covariance analysis to demonstrate that the lentiform nucleus is a functional constituent of a spatially distributed metabolic brain network involving other elements of the motor CSPTC loop.^9,10,15 The current findings suggest that the level of activity of pallidal inputs from the putamen and STN defines a preoperative lentiform metabolic baseline for pallidotomy. To the extent that these inputs are overactive in PD, they represent a major determinant of the metabolic activity of the surgical target, and consequently the degree of functional correction that can occur following the procedure.

The results of our previous comparative study of preoperative and postoperative PET in pallidotomy suggested that, apart from the preoperative metabolic set point, the remaining intersubject variability in surgical outcome is determined largely by individual differences in the operative modulation of CSPTC motor circuits.¹⁵ The extent of network modulation can be determined explicitly only with postoperative FDG/PET measurements.¹⁵ Nonetheless, we propose that the dynamic range of the CSPTC motor loop with pallidal suppression can be clinically estimated using a levodopa challenge. Prior PET studies in experimental animal models demonstrated reductions in pallidal hypermetabolism with levodopa infusion.^7,8 These findings have been supported by comparable FDG/PET experiments in PD patients.³⁰ To the extent that the severity of clinical manifestations is correlated with the expression of abnormal CSPTC metabolic networks,^9,10,15,31 the measured clinical benefit with levodopa administration may simply provide an indicator of the dynamic range of network modulation that can occur with either pharmacologic or surgical pallidal suppression. In this vein, we found that preoperative measurements of lentiform metabolism with FDG/PET reproducibly predict 50% of operative improvement in off-state CAPIT scores. The preoperative levodopa responsiveness index, though simpler to measure, predicted approximately 36% of individual differences in outcome. Together, however, both preoperative measures predicted approximately 70% of the pallidotomy response. Thus, preoperative FDG/PET measurements of the functional set point of the CSPTC network combined with an independent clinicopharmacologic estimate of the patient's individual capacity for network modulation can provide useful and complementary criteria for patient selection.

Our findings suggest that an accurate prediction of pallidotomy outcome can be achieved by preoperative quantitative assessments of lentiform metabolism and levodopa responsiveness. However, these measures can be construed as meaningful indicators of outcome only if the pallidum itself is anatomically normal before surgery, and if the pallidotomy lesion is placed accurately. Indeed, pallidal MRI signal was universally normal preoperatively and postoperative MRI confirmed target localization in GPi in all cases. In our patient population, we did not find a correlation between operative outcome and the position and size of the surgical lesions. However, this may have been a consequence of the limited intersubject variability in the lesion parameters noted in our patient cohort. In this series pallidotomy lesions were made at 80 °C for 60 seconds during a single trajectory with a 1.1-mm diameter tip electrode.¹⁶ Lesion volume was thus relatively consistent across patients and did not correlate with individual differences in postoperative motor improvement: We note, however, that lesion volumes determined in the immediate postoperative period may overestimate ultimate lesion size.¹⁹ Therefore, longitudinal MRI volumetrics and correlations with clinical outcome will be needed to address this issue conclusively.

Similarly, the issue of optimal lesion localization is unresolved. A recent study³² indicated general accord on the desired coordinates of the pallidal target. Moreover, the use of microelectrode recording allows for the tailoring of the pallidal target for individual patients.^26,33 We found that in our patient group pallidal lesions were comparable in location across subjects and agreed well with the proposed target. We also note that with the average 6-mm diameter of the pallidal lesions, there was apt to be considerable overlap in lesion position across patients. Therefore, given the limited variability in target location, it is not surprising that no association between lesion coordinates and clinical outcome could be discerned in our data. Indeed, we found that lesions of comparable position in the posteroventral pallidum were associated with striking individual differences in postoperative motor improvement.

In summary, our findings suggest that the clinical outcome of pallidotomy for PD is dependent on the following criteria: 1) The pallidum should be anatomically intact: Patients with coexistent vascular lesions of the basal ganglia may not be optimal surgical candidates. Similarly, this procedure may not be beneficial for patients with parkinsonism secondary to other forms of pallidal pathology such as manganese intoxication.³⁴ 2) The pallidum should be functionally overactive: Patients with preoperative evidence of lentiform hypermetabolism in FDG/PET are likely to be good candidates. Patients with abnormally low glucose utilization in this region, as occurs in atypical forms of parkinsonism such as striatonigral degeneration or multiple systems atrophy, are less likely to benefit from pallidal ablation.³⁵ 3) The pallidum should be functionally suppressible: Patients who are noted preoperatively to have a high degree of levodopa responsiveness may be good candidates. The measurement of the clinical response to levodopa in a surgical candidate can provide an estimate of the individual's capacity to reduce baseline pallidal overactivity and to modulate the expression of CSPTC motor networks. Patients with blunted levodopa responses, as in atypical parkinsonism, are therefore unlikely to have significant benefit from pallidotomy.

Our results support the use of FDG/PET, preferably in combination with the quantitative assessment of levodopa responsiveness, as appropriate preoperative screening procedures for selecting optimal candidates for pallidotomy. Nonetheless, the ultimate utility of these screening measures can be determined only through the study of large cohorts of patients followed for extended periods after surgery.

Acknowledgments

We thank Dr. Abdel Belaklef and Mr. Claude Margouleff for technical support, Dr. Robert Dahl and Mr. Ralph Matacchieri for cyclotron support, Dr. Thomas Chaly for radiochemistry support, and Ms. Lauren Moran for manuscript preparation.