This study was a retrospective analysis of a series of craniocervical dystonia patients who underwent DBS at the Fixel Institute for Neurological Diseases, University of Florida. Diagnosis of craniocervical dystonia (blepharospasm, oromandibular dystonia and cervical dystonia) in these patients was established by movement disorders trained neurologists. Patients had been selected for DBS following comprehensive interdisciplinary evaluation and careful risk–benefit assessment [9]. DBS electrodes and the latest pulse generators available at the time of scheduling of DBS were used. Detailed information about surgical procedures and electrode measurements has been described previously [10]. The lateral portion of the sensorimotor posteroventral GPi was targeted to balance the ability to deliver higher stimulation amplitude with capsular side effects with input insights gained from awake surgery with microelectrode recording.

Preoperative clinical data for these patients were available in our database up to 12 months prior to DBS surgery. Postoperative follow-up at our center is scheduled monthly for the first six to 12 months, followed by visits at 12 months and annually thereafter. All patients included in this study had data entered in our database from one or more postoperative follow-up visits.

Data was retrospectively retrieved from the INFORM database, and electronic medical records at the University of Florida. Information was collected using a standardized proforma, with independent verification by two investigators to ensure reliability. INFORM is a clinical-research database that provides information on demographics, clinical, and functional characteristics of patients with movement disorders at the University of Florida. Clinical information extracted included demographic data, disease characteristics and longitudinal outcome data extracted at prespecified intervals: prior to surgery (baseline or T0) and postoperatively at 3–6 months (T1), 1–2 years (T2), 3–5 years (T3), and the most recent follow-up (T4). Baseline clinical data were derived from records within the 12 months preceding DBS implantation, with dystonia severity assessed under standard treatment conditions during routine clinical evaluations. To accommodate variability in clinical follow-up inherent to routine care, time windows were defined in advance and data were permitted from within 3 months of baseline and T1 visit and within 6 months of the desired timepoint for the T2, T3, T4 visits. For each interval, the clinical encounter closest to the target time point was selected for analysis. If no encounter occurred within the specified window, the corresponding data point was considered missing.

Outcome measures: The Unified Dystonia Rating Scale (UDRS) was administered by fellowship-trained movement disorders specialists at baseline and follow-up intervals, with scores documented in our database. The UDRS evaluates dystonia in 14 body regions such as eyes/upper face, lower face, jaw/tongue, larynx, neck, trunk, and limbs (proximal and distal) by rating both severity (amplitude/intensity) and duration (proportion of time present) of dystonic movements or postures. Severity and duration scores from individual regions are summed to generate a total score (maximal total score of UDRS is 112), with higher scores indicating more severe and widespread dystonia [11]. In our study, in addition to total UDRS scores, we obtained body region–specific UDRS scores categorized into five subdomains: upper face (eyes), lower face (including jaw, tongue, larynx), neck, upper limb (arm and/or hand), and trunk/lower limb (leg and/or foot). For each body region, dystonia duration and motor severity scores were summed to generate regional scores, and the total UDRS score was calculated as the sum across all regions. For the purposes of our study, the primary body regions of interest were the upper face (max score 16), lower face (max score 48), and neck (max score 16) which were analyzed separately in addition to the total UDRS scores.

HRQoL outcomes were assessed at baseline and follow-up time points using the patient-reported Short Form Health Survey-36 (SF-36). The eight SF-36 domains were evaluated and summarized into the total score, Physical Component Summary (PCS), and Mental Component Summary (MCS) [12]. The SF-36 total score was calculated using previously established methods. In addition, non-motor symptoms were assessed using the Beck Depression Inventory (BDI, maximal total score of 63 with higher scores indicating worse depression), Beck Anxiety Inventory (BAI, maximal total score of 63 with higher scores indicating worse anxiety), and the Apathy Scale (AS, maximal score of 42 with higher scores indicating more apathy) described in previous publication [13].

Descriptive data for demographics and baseline characteristics were summarized as median values with ranges. Median values and interquartile range (IQR) were calculated at each time point for total and individual body region UDRS scores. Paired comparisons between baseline and each postoperative time point were conducted using the Wilcoxon signed-rank test among patients with available paired measurements. The % change in UDRS total score (UDRS% change) after surgery was calculated with the following formula: (preoperative UDRS – postoperative UDRS)/preoperative UDRS × 100. While no specific threshold has been universally established for a clinically meaningful change in UDRS scores, improvements of approximately 25%–30% are generally considered indicative of a DBS responder [14]. While total UDRS could be interpretable as a quasi-continuous measure only when multiple ordinal items are summed across body regions, reporting percentage changes at the level of an individual body region score (ordinal rather than continuous) could be potentially misleading. For example, a score of 4 on the Eye item does not represent “twice as severe” as a score of 2 and calculating percentage changes for a single item is therefore not meaningful. Thus, for individual body regions, we considered meaningful improvement (minimally clinically important difference or MCID) as follows: eyes and upper face (≥2-point drop, representing ≥25% improvement of max 8 points for this body region), neck (≥2-point drop, representing ≥25% improvement of 8 points), and lower face (lower face, jaw, tongue, and larynx; ≥ 6-point drop representing ≥25% improvement of 24 points for these body regions). We examined the number of individuals achieving these thresholds of improvement in each body region at each of the follow-up time points. Percentage change in SF-36 PCS and MCS was calculated as the difference between the postoperative and preoperative scores, divided by the preoperative (baseline) score, and multiplied by 100 (PCSΔ and MCSΔ). Change in individual domains of SF-36, BDI, BAI and AS was noted.

We further examined the longitudinal DBS related improvement in dystonia using a linear mixed-effects model with UDRS score treated as a continuous outcome and a subject-specific random intercept. A similar approach was used for determining longitudinal change in HRQol and change in each domain was estimated by independently modeling domain score as a continuous outcome with a subject-specific random intercept. We included baseline factors such as age at onset for dystonia, duration of dystonia before DBS, number of body areas affected (eyes, upper face, jaw, tongue, lower face, neck) and sex as covariates in each model. Sensitivity analyses were performed using random patient specific slopes, quadratic fit, and exclusion of variables including disease duration and number of body regions affected to ensure model viability. Visual inspection of residuals and fitted value plots were performed and no flagrant violations of normality of homoscedasticity were appreciated. Confidence intervals were obtained using parametric bootstrapping. To account for multiple comparisons, p-values were adjusted using Benjamini–Hochberg false discovery rate (FDR) technique. Statistical significance was assessed at a two-sided FDR-adjusted α level of 0.05.

Statistical analysis was performed with SPSS version 26 and RStudio: Integrated Development Environment for R (Posit Software, PBC, Boston, MA.). Coding for some aspects of statistical modeling was developed with assistance from OpenAI sources.

The study was approved by the Institutional Review Board of the University of Florida (IRB No. 201702279), with a waiver of informed consent due to the use of de-identified data. The UF INFORM database was independently approved by the Institutional Review Board of the University of Florida (IRB No. 201501166) and all subjects previously provided informed consent in accordance with the Declaration of Helsinki. The datasets analyzed during the current study are not publicly available because of institutional regulations and patient privacy considerations but are available from the corresponding author on request and with appropriate approvals.

We identified 19 patients (13 females). The median (range) age of this cohort at first surgery was 66 years (48–83 years) and mean (SD) was 64.8 (8.9) years. The median (range) age at dystonia onset was 59 years (40–72 years) and mean (SD) was 57 (9.3) years. The median (range) duration of follow-up after DBS was 6 years (1–24 years) and mean (SD) was 7.8 (6.1) years. Eight patients remained actively followed at our center. Six patients relocated to another state, two patients were deceased, two discontinued follow-ups for unknown reasons, and one patient had documented travel difficulties that prevented continued follow-up (Table 1). Two of these individuals had exposure to neuroleptic medications prior to onset of symptoms. None of these individuals developed dystonia secondary to structural brain lesions or neurodegenerative diseases. There was no history of prior ablative surgery (e.g., pallidotomy or thalamotomy) for dystonia before DBS implantation or a revision surgery or hardware explantation due to infection or device failure within a year of DBS. Three patients (15.8%) had undergone blepharoplasty, 2 (10.5%) prior eyelid myectomy, and 1 (5.3%) a sling procedure. There was no reported history of tremors in addition to dystonia.

15 patients (78.9%) were using oral medications at baseline, including 11 (57.9%) on benzodiazepines, 2 (10.5%) on baclofen, and 4 (21.1%) on trihexyphenidyl. At the most recent follow-up, 9 patients (47.4%) continued to use oral agents for dystonia. Four patients (21%) were not taking oral medications prior to surgery, whereas eight patients (42%) were not requiring oral medications at the most recent follow-up after DBS. 14 patients (73.6%) received BoNT approximately every 12 weeks prior to surgery, and 11 (57.9%) continued BoNT therapy at the last follow-up (Table 1). Additionally, DBS programming data are summarized with the most recent settings displayed in Table 2.

Among 19 subjects, the median baseline UDRS score was 19 (IQR: 9–20.5). Relative to baseline, median UDRS scores improved at T1 (median 9.5, IQR: 6–19.8, n = 16; p = 0.04), at T2 (median 8.5, IQR: 4–13, n = 16; p = 0.02), at T3 (median 5.5, IQR: 2–12.5, n = 12; p = 0.02), and at T4 (median 6.5, IQR: 4.3–11, n = 6; p = 0.04). The percentage improvements in UDRS scores (median of individual % improvements) was 34.5% at T1, 51.0% at T2, 33.3% at T3, and 43.4% at T4. UDRS total scores demonstrated significant improvement across all follow-up time points (all p-values FDR-adjusted). Full data, including median and IQRs, are available across all time points in Table 3. Figure 1 shows UDRS data (mean) from individuals with available baseline and corresponding follow-up time points.

Regarding regional changes in UDRS scores (Table 3), in the eyes and upper face region, 7/16 patients (43.8%) at T1, 12/16 (75.0%) at T2, 9/12 (75.0%) at T3, and 5/6 (83.3%) at T4 achieved a ≥ 2-point reduction. In the lower face subdomain, 6/16 patients (37.5%) at T1, 7/16 (43.8%) at T2, 6/12 (50%) at T3, and 3/6 (50%) at T4 achieved a ≥ 6-point reduction. For the neck region, 5/16 patients (31.3%) at T1, 9/16 (56.3%) at T2, 7/12 (58.3%) at T3, and 3/6 (50%) at T4 achieved a ≥ 2-point reduction. Given the small sample size, statistical analysis was not performed for the change in individual body regions. Additionally, among the 9 of 19 patients who continued follow-up in our clinic and were clinically interviewed, 6 reported subjective improvements. These included reduced involuntary movements; improved eyelid control and head movements; improved mastication; decreased pain and tightness in the jaw and neck muscles; and enhanced performance of daily activities (e.g., eating, personal care, and driving), all of which were meaningful to them and contributed to improved quality-of-life.

Longer duration of DBS therapy was associated with a significant reduction in UDRS score, indicating improvement over time (β = −0.10 points per month, p = 0.00018 [FDR corrected], Supplementary Figure S1). This association remained robust after adjustment for sex, disease duration, age at onset, and body areas involvement, and in sensitivity analyses excluding disease duration and number of body areas affected. Our model showed that number of body areas affected predicts a higher UDRS score (β = 4.86, p = 0.0043) as did longer duration of disease prior to DBS (β = 0.43634, p = 0.039), but the effects of other co-variates on modeled UDRS score were not significant. Because some patients may initially have benefits with programming that plateau over time, and we were interested in understanding the longer-term effects of pallidal DBS in this context, we performed quadratic modelling to undervalue this initial programming effect and clarify longer term benefit. While this disclosed a mildly better fit based on likelihood ratio testing with ANOVA (χ² = 4.78, p = 0.028), the linear term remained significant, and the primary conclusions were unchanged.

With regards to the HRQoL, median Physical Component Summary (PCS) scores were 52.5 (IQR: 36.9–62.3) at baseline, 52.3 (IQR: 47.6–60.8) at T1, 53.3 (IQR: 34.8–69.3) at T2, 51.3 (IQR: 44.8–66.3) at T3 and 63.5 (IQR: 44.3–68.1) at T4. Median Mental Component Summary (MCS) scores were 55.8 (IQR: 33.6–74.3) at baseline, 60.5 (IQR: 38.9–72) at T1, 50.5 (IQR: 37.5–73.3) at T2, 61.0 (IQR: 48.6–73.6) at T3 and 66.6 (IQR: 60.8–80.3) at T4. Individual domain scores are shown in Figure 2.

Following FDR-adjusted analysis, the physical and mental component summary scores did not differ significantly across time points. Likewise, statistical comparisons across the eight individual domains of quality-of-life did not yield any statistically significant differences. However, social functioning demonstrated a directional trend toward improvement based on median values.

In independent mixed effect linear regression models with domains of the MOS-36 as the dependent variable, a longer duration of DBS therapy was associated with improvement in multiple HRQol domains in unadjusted analyses. The Role Physical domain (β = 2.63 points per month, p = 0.011), Vitality domain (β = 2.021 points per month, p = 0.049), and Social functioning domain (β = 3.01, p = 0.0042) improved over time, while Physical Functioning, Bodily Pain, General Health, Role Emotional and Mental Health domains did not vary significantly with time. Notably, baseline UDRS and number of body regions explained variance in the physical functioning and role emotional domains in unadjusted analyses. After correction for multiple comparisons, social functioning and role physical functioning domain scores (p = 0.034 and 0.044 respectively) remained significantly correlated, though other domains did not.

The median BDI scores were 8.5 (IQR 4–11.3) at baseline, 9 (IQR 5.8–11.5) at T1, 8.5 (IQR 4–11.5) at T2, 10.5 (IQR 6.0–14.3) at T3, and 7.0 (IQR: 5.0–7.5) at T4. Median BAI scores were 6 (IQR 1.0–8.5) at baseline and 6 (IQR 4.0–10.0) at T1, 7.0 (IQR 2.0–12.0) at T2, 8.0 (IQR 2.5–12.0) at T3, and 2 (IQR 1.0–3.0) at T4, respectively. Median (apathy) AS scores were 9.0 (IQR: 8.0–10.8) at baseline and 13.0 (IQR 10.0–18.0), 12 (IQR 7.5–16.0), 11 (IQR 6.3–16.3), and 13.0 (IQR 8.0–14.0) at T1 to T4, respectively.

In Wilcoxon signed-rank paired analyses comparing each follow-up time point to baseline, none of these comparisons reached the threshold for statistical significance after FDR-adjustment.