Medical charts of children and young people with acquired, genetic or idiopathic dystonia who underwent RP at the Bambino Gesù Children’s Hospital between 2011 and 2023 have been retrospectively reviewed using a structured proforma. Clinical data were retrieved from their medical files and video recordings, when available.

Patients with disabling dystonia refractory to medical treatment were considered eligible to surgical treatment. Medically refractory dystonia was defined after failed attempts of treatment with ≥2 medications or in case of failure to abate SD after appropriate medical management, in accordance with ongoing expert recommendations (7). According to expert recommendations (8), the factors underpinning the decision to opt for RP instead of DBS included: (i) high risk of hardware-related complications because of extreme dystonic posturing, poor nutritional status, small body dimensions; (ii) life-threatening, medically refractory SD requiring urgent treatment and preventing time-consuming programming sessions; (iii) severe pre-existing neurological deficits including severe speech and feeding disturbances, which mitigate the concerns for bilateral RP-induced side effects; (iv) limited willingness or feasibility of attending the programming center of the patient and their caregiver; and (v) previous complications from an implanted DBS system, when the severity of the condition precluded the possibility of waiting for the time required to perform a new implantation.

Some of the patients included in this study have already been described in previous publications from our group (9, 10).

As part of the preoperative work-up, to visualize basal ganglia and vascular structures, patients underwent brain MRI with 3D T1-, T2-, inversion recovery and gadolinium-enhanced T1-weighted images. A preoperative volumetric CT scan was performed in all patients. A full video-taped neurological examination was performed before the surgery, except in cases where the severity of the clinical conditions, the urgency of the procedure, or the state of sedation in the intensive care unit did not allow for a meaningful or feasible recording.

All surgeries were performed under general anesthesia, through a frameless stereotactic neuronavigation robotic system (ROSA ONE^® Brain, Zimmer Biomet, Warsaw, IN, United States).

Over the study period, anesthesia protocols changed and were progressively adapted according to evolving clinical practice. In general, anesthesia induction was achieved using one or more hypnotic agents (propofol, midazolam, and/or sevoflurane) combined with an opioid analgesic (fentanyl or sufentanil). Anesthesia maintenance was achieved with intravenous anesthesia using propofol or midazolam as hypnotic agents and remifentanil or morphine as analgesics. When available, continuous monitoring of anesthesia depth was performed using bispectral index (BIS™) monitoring (Medtronic plc, Dublin, Ireland).

The target was localized to the mid portion of the GPi, and was preoperatively identified by visual direct targeting through volumetric inversion recovery, T2, FLAIR and post-contrast T1-weighted brain MRI sequences merged with brain CT.

Intraoperative neurophysiological monitoring was individualized and protocols varied during the study period. Microelectrode recordings (MER) and/or macrostimulation under surface EMG and/or flash visual evoked potentials (fVEPs) and Transcranial Electric Motor Evoked Potentials (tcMEPs) were performed (Leadpoint MER acquisition System – Medtronic, Inc.; Cascade Elite and Iomax systems – Cadwell Industries, Inc.), to detect the nucleus and to avoid the involvement of relevant surrounding structures by the lesioning (i.e., internal capsule and optic tract), respectively.

MER recording was performed along the planned trajectory from 10 mm above to the target.

The ablation was performed through a radiofrequency generator (Cosman RFG-1; Cosman Medical Inc., Burlington, MA, United States) on the target and along a trajectory of 8 mm above the target, at 40 V for 60 s. A second and a third trajectory (2 mm posterior, 2 mm lateral to the first target) was used to perform lesioning. The standard entry point for the procedure was located in close proximity to the coronal suture, approximately 20 mm lateral to the midline. This trajectory allows the exploration of the nucleus along the final 10 mm of the trajectory, before entering the target. Alternative trajectories (e.g., a more lateral entry point or a transventricular approach) were used in patients with distorted brain anatomy, resulting from severe cerebral atrophy and/or ventricular enlargement.

All patients performed brain imaging (CT scan or MRI) within 24 h from surgery. The decision to perform CT or MRI was guided by the patient’s clinical status, immediate availability, the clinical suspicion of intraoperative complications and the possible presence of a DBS system contralateral to the lesion. For patients who underwent immediate CT scan, brain MRI was postponed to the following days. Late brain MRI (≥30 days after surgery) was performed according to clinical judgement, based on the patients’ clinical status.

No standardized protocol for postoperative clinical evaluation was implemented, in line with the aim of minimizing logistical burdens and patient discomfort in this frail cohort.

As outcome measure after surgery, the Clinical Global Impression–Improvement scale (CGI-I) was retrospectively rated by two child neurologists based on medical records, integrating caregivers’ impressions of improvement or worsening, findings from neurological examinations, and changes in medical treatment, including medication escalation or de-escalation. In patients treated for ongoing SD, resolution was defined as the discontinuation of intravenous sedative therapy, and the interval between surgery and SD resolution was calculated accordingly. For patients with previous episodes of SD, recurrence of further SD episodes was annotated and employed as an outcome measure. Recurrent SD was defined as the occurrence of a new episode of dystonic exacerbation requiring admission to an intensive care or high-dependency unit for intravenous sedative treatment.

During the study period, 18 patients underwent 21 surgical procedures. Sixteen patients performed bilateral simultaneous RP. Two patients, who had been previously implanted with GPi-DBS and in which one of the DBS electrodes was removed because of a hardware infection, underwent unilateral RP. In both cases, RP was performed using a standard surgical technique after removal of the infected electrode during a previous procedure. In one patient, the contralateral DBS system was left in place. In the other patient, the DBS system had been explanted bilaterally; however, RP was performed unilaterally because the contralateral GPi had been previously damaged by a perielectrodal intracranial hemorrhage at the time of DBS implantation, rendering it an unsuitable target for RP. In both patients, ablative surgery was chosen over DBS reimplantation because conservative management of the hardware infection failed to avoid DBS explanation, and delayed DBS reimplantation was considered unsuitable given the severity of the underlying dystonia and previous SD episodes.

Three patients with previous bilateral surgery underwent a second RP during the follow up period due to SD recurrence. Ten patients had acquired dystonia (Dyskinetic cerebral palsy), 5 patients had definite genetic dystonia, and 3 patients had dystonia of unknown etiology (but likely genetic) (Table 1).

At the time of surgery, 12 patients were receiving enteral feeding through gastrostomy, jejunostomy or nasogastric tube, 4 had a tracheostomy for respiratory failure and 3 were already on intrathecal baclofen pump treatment.

Mean age at first surgery was 11.97 ± 5.87 years. Indication for surgery included: urgent intervention for ongoing, medically refractory status dystonicus (7 patients); semi-urgent surgery for previous medically managed SD with residual severe dystonia requiring priority surgery to avoid SD recurrence (7 patients); and elective surgery for severe medically refractory dystonia without previous SD (4 patients). In the three patients undergoing a second pallidotomy, mean age at reintervention for second lesion surgery was 15.23 years (range 10.1–23.5 years), with a mean interval between surgeries of 4.5 years (range 3.3–5.3 years). In two of them, second-step RP was performed during an ongoing, medication-refractory SD. The third patient was operated after the resolution of a previous, medically managed SD. Mean time between surgeries was 4,52 years (range 3.34–5.33 years).

During anesthesia, BIS™ monitoring was recorded and documented in 13 out of 21 surgeries. During anesthesia maintenance and surgery (including intraoperative recordings) BIS was maintained between 35 and 50.

In 14 procedures, MER was performed to detect GPi activity and improve target localization. In 9 patients, a spiking pattern compatible with GPi activity (i.e., high activity, rapidly firing neurons) was documented within the targeted area. In the remaining 5 cases, GPi electrical activity was not detectable or doubtful. Notably, in all of these 5 patients, preoperative MRI showed pallidal signal abnormalities. Motor evoked potentials were performed on 10 patients, and visual evoked potentials on 8 patients.

Three patients had been referred for surgery from other Institutions and were monitored only for the immediate post-operative period (i.e., equal or less than 3 months). The mean follow-up time, for the other 14 patients, was 6.62 ± 3.65 years, with two patients missing the scheduled follow-up assessments at 6 and 12 months. In total, 16 patients were evaluated at 3 months post-surgery, 13 at 6 months and 12 months and 15 had a last follow-up more than 1 year after surgery (Figure 1). During the follow-up period, two patients died for medical complications unrelated to pallidotomy.

In the first post-operative day, MRI was performed in 8 surgeries and CT scan in 13. For patients whose MRI was postponed, the exam was performed 13.33 ± 10.11 days later (range 2–34 days). One patient did not perform postoperative MRI due to the presence of a contralateral DBS system. Two out of three patients undergoing double pallidotomy did not perform any MRI scan after the second surgery. Early MRI (within the first 2 postoperative days) displayed limited hemorrhagic foci within the lesion, in all subjects. In 6 out of 8 patients performing MRI within 30 days after surgery, a limited hemorrhagic focus within the lesion was still evident (Figure 3). In addition to intralesional bleeding, one patient experienced a frontal intraparenchymal hemorrhage with a slight midline shift, two patients had asymptomatic subdural bleeds, and one patient developed a bilateral frontal pneumocephalus with unilateral pneumoventricle (Figure 4). In 14 surgeries, the first postoperative imaging detected extrapallidal extension of the lesion or perilesional oedema; involving the putamen (6 cases), internal capsule (3 cases) or both (3 cases).

In 8 cases, a second MRI was performed after a variable time delay (37.25 ± 44.39 months, range 1–139 days). In 5 cases, extrapallidal extension of the necrotic lesion was detected: to the putamen (1 case), internal capsule (2 cases), both (1 case) or in the white matter along the surgical trajectory (1 case).

In all patients undergoing late MRI, RP lesion was still visible on both sides (Figure 5), including in the three patients who underwent re-pallidotomy for lesion extension.

Overall, adverse events were reported after 9 surgeries, 3 of which were procedure-related, while 6 were due to medical complications. The adverse events related to the surgical procedure included: motor apraxia, cognitive decline and behavioral changes succeeding frontal hemorrhage in 1 patient, and hemiplegia associated with global akinesia in 2 other patients. Adverse events related to the post-surgical management included: respiratory distress (3 patients, 2 of which required tracheostomy due to extubation failure), post-surgical sepsis (2 patients), ventilator associated pneumonia (1 patient) and deep vein thrombosis (1 patient).

From our cohort, RP emerges as an effective therapy for medication-refractory SD and severe dystonia. All patients undergoing the first surgery for medication-refractory SD were successfully discharged from ICU. The time needed for SD resolution after surgery was highly heterogenous, with many patients requiring weeks of intravenous sedative treatment after surgery. Nevertheless, RP facilitated the tapering of sedative therapy and enabled successful transfer from the ICU to lower-intensity care units. These findings are in line with a previous report suggesting that RP is effective for SD in various forms of dystonia (12).

In two patients with relapsing medication-refractory SD after a previous successful management of SD by RP, a second surgery to extend the first lesion was effective to abate a SD only in one patient.

A positive effect on dystonia was found in the first months after surgery in most patients, regardless of the indication for surgery. These findings suggest that, in the short term, RP is effective in relieving dystonic symptoms. The therapeutic effects were not immediately visible but took several days to weeks to become detectable; some patients suffered transient exacerbations of dystonia in the early postoperative course.

After the first 6 months after surgery, an overall trend toward dystonia reemergence was observed. Similarly, SD recurred after RP in a significant proportion of patients with previous SD episodes (7 out of 14 patients). This observation is in line with previous reports of long-term follow up after unilateral RP for hemidystonia (13), where some patients were found to worsen over time. Several possible, non-mutually exclusive explanations may account for the long-term loss of therapeutic effect and increased risk for SD in a significant portion of the patients. First, it is likely that a progressive disease course contributed to dystonia worsening in some patients with an underlying neurodegenerative condition. Nevertheless, a gradual loss of improvement was observed also in patients with static diseases, such as dyskinetic cerebral palsy, with 5 out of 10 patients with acquired dystonia experiencing SD relapses after pallidotomy. An inadequately placed or sized lesion is a second possible explanation, as distorted brain anatomy in this population often compromises GPi targeting by reducing imaging contrast and landmark reliability and increasing susceptibility to brain shift. Inconsistency in postoperative imaging timing prevent us from a formal analysis of lesion size and location. Nevertheless, in all the patients performing late postoperative scans lesion was still clearly visible on MRI, including patients with SD relapse. This finding suggests that dystonia reemergence is not due to lesion shrinkage, differently from what has been observed for tremor after MRI-guided focused ultrasound (MRIgFUS) thalamotomy (14).

Another contributing mechanism could be the reemergence of dystonia through alternative networks, such as the cerebello-thalamic network. This hypothesis is supported by multiple lines of evidence suggesting that dystonia may arise from the dysfunction of a broad network including the motor and somatosensory cortices, the cerebellum, thalamus and basal ganglia (15), enabling network reorganization after ablative surgery. A similar hypothesis has been argued to explain the reemergence of Parkinson’s Disease-related tremor after MRIgFUS thalamotomy. In this case, a possible propagation of tremor to a pallidothalamic network has been postulated (16). In the three patients from our cohort who underwent lesion extension through second-stage pallidotomy, surgery was less effective compared to first-stage surgery, possibly supporting the hypothesis of a shift in the underlying dystonia network organization. Notably, in a subset of patients with gradual worsening of dystonia after transient improvement, intraoperative MER had failed to identify pallidal neuronal activity. While this could reflect deep anesthesia or suboptimal localization, the concomitant presence of pallidal MRI hyperintensities may indicate markedly reduced neuronal activity, suggesting that preserved GPi functional integrity could influence RP outcome. Accordingly, preoperative multimodal assessment with 18F-FDG PET and structural/functional MRI connectivity may help predict surgical outcome (17, 18).

Regardless of the mechanisms of dystonia recurrence after surgery, the individual outcome remains difficult to predict. Therefore, patients and families should be informed that the initial surgical benefit may diminish over time. In our cohort, no clear-cut differences in terms of outcome were observed based on dystonia etiology (genetic, idiopathic or CP), possibly due to the limited size of our sample and the lack of a standardized post-operative assessment.

Moreover, advances in the understanding of genetic basis of dystonia and the identification of genetic predictors of DBS response have prompted a re-evaluation of the traditional exclusion criteria for neuromodulation. For instance, GNAO1-related dystonia is now known to respond dramatically to DBS (14), and, in the present day, it would be considered first line invasive treatment, even in fragile patients. Unfortunately, many patients with genetic dystonia in our cohort underwent surgery prior to the completion of their diagnosis, or before data on DBS responsiveness of their specific condition became available. This is particularly relevant for patients with (yet) undetermined dystonia at the time of surgical evaluation. When genetic diagnosis is unknown, DBS should not be excluded solely based on the severity of their clinical picture, especially in absence of overt brain abnormalities on MRI.

Although the overall rate of postoperative adverse events was substantial, the majority were medical complications, likely reflecting prolonged hospitalization and the generally poor clinical condition of the patients. Small intralesional hemorrhages and perilesional edema extending into surrounding structures were frequently observed on early postoperative scans. These may reflect an increased risk of lesion expansion from thermal dispersion, pneumocephalus, hemorrhage, and brain shift in patients with complex or distorted brain anatomy, particularly when transventricular trajectories are employed. Nevertheless, adverse events directly related to surgery only occurred in 3 out of 21 patients and were generally mild. One patient experienced cognitive decline and motor apraxia after frontal hemorrhage, whereas two patients experienced hemiplegia and hypokinesia that improved over time and did not significantly impact their functional status. Considering that most of the patients had previous neurological deficits with absent speech and severe feeding difficulties, no overt worsening of dysarthria and dysphonia was reported. However, in two patients requiring postoperative tracheostomy, worsening of the bulbar function, with increased drooling and ineffective airway clearance, may have contributed to extubation failure.

Our results should be interpreted in light of some limitations. We did not perform any standardized assessment of dystonia severity in the follow up; our retrospective evaluation of treatment outcome may suffer from reporting, observer and loss-at-follow up biases. Our cohort was highly heterogenous in terms of dystonia etiology and indication for surgery. Furthermore, only a subgroup of our cohort performed a tardive postoperative brain MRI, impeding evaluation of exact size and location of the lesions and their correlation with clinical outcome.