Long-term functional decline following upper limb botulinum toxin injections in children with unilateral cerebral palsy: A preliminary retrospective 5-year follow-up case-control study Authors: Anne Tabard-Fougère, Alice Bonnefoy-Mazure, Sophie Louis, Stephane Armand, Geraldo De Coulon, Oscar Vazquez Published: 2026 ABSTRACT PurposeTo compare long-term upper extremity functional evolution in children with spastic cerebral palsy (CP) receiving upper limb botulinum toxin A (BoNT-A) injections versus physical therapy (PT) alone.MethodThis retrospective cohort study included children aged 5–12 years at baseline with ≥2 Assisting Hand Assessment (AHA) evaluations over ≥3 years. The primary outcome was AHA score change. Secondary outcomes included changes in maximal wrist extension and forearm supination during functional tasks, measured from standardized videos by three blinded assessors. Changes were classified using minimal clinically important difference (MCID) thresholds: 5 units (AHA), 20° (wrist extension), and 30° (supination).ResultsEleven patients with unilateral spastic CP (BoNT-A+PT, n=5; PT alone, n=6) were included. All BoNT-A+PT patients demonstrasted AHA deterioration (Δ = –3.6 ± 2.7), versus improvements in 5/6 PT patients (Δ = +9.0 ± 6.4). Wrist extension declined in all BoNT-A+PT patients (Δ = –53° ± 19°) versus 1/5 PT patients (Δ = –4° ± 8°). Supination deteriorated in 4/5 cases of BoNT-A+PT patients (Δ = –30° ± 7°) but remained stable in all PT patients (Δ = +2° ± 9°).ConclusionUpper limb BoNT-A was associated with systematic long-term functional decline, while PT alone preserved or improved function.Significance of StudyThese findings suggest that BoNT-A treatment may compromise long-term motor control and functional capacity, potentially influencing treatment algorithms in children with spastic cerebral palsy.Level of evidenceIII (retrospective cohort study). Introduction Unilateral cerebral palsy (CP) affects upper limb function and bimanual activities, significantly impacting children’s daily life and participation. The Assisting Hand Assessment (AHA) is the gold standard for measuring spontaneous use of the assisting hand in bimanual tasks, providing a validated measure of bimanual hand use in children with unilateral CP.1,2 The natural history of hand function in unilateral CP has been well characterized.1–3 Children with Manual Ability Classification System (MACS) level I reach approximately 90% of their developmental limit by age 3 years. 1 Longitudinal studies have demonstrated that developmental trajectories remain stable from 7 to 18 years across all MACS levels, with children maintaining their relative functional capacity through adolescence. 2 This stable natural history provides a baseline against which treatment effects can be evaluated. Botulinum toxin type A (BoNT-A) is widely used to reduce upper limb spasticity in children with CP,4,5 with the goal of improving motor function and preventing contractures. However, existing studies show heterogeneous results: a systematic review found that only 3 of 7 studies using AHA as an outcome measure demonstrated improvement following BoNT-A treatment. 4 Importantly, most studies have short follow-up periods of 12 months or less, limiting our understanding of long-term effects and whether initial gains are maintained or lost over time. Emerging concerns have been raised regarding the potential adverse effects of repeated BoNT-A injections. Recent studies suggest that repeated injections may cause permanent muscle changes including atrophy and fibrosis. 6 Muscle volume reduction has been documented up to 12 months post-injection, and histopathological changes including neurogenic atrophy and type 1 fiber loss, have been correlated with the number of injections received.7–9 These structural changes raise questions about whether BoNT-A may paradoxically compromise long-term motor function despite short-term spasticity reduction. Despite widespread use of BoNT-A for upper limb spasticity, there is a significant gap in the literature regarding long-term effects (>3 years) on bimanual function. No studies have compared long-term functional trajectories in children receiving BoNT-A versus physical therapy alone using validated outcome measures such as the AHA. This study therefor aims to compare the long-term functional evolution of the upper limb in children with unilateral CP treated with BoNT-A in addition to physical therapy (PT) versus those receiving PT alone. The primary outcome was the change in bimanual performance measured by the AHA, with a minimum follow-up interval of three years between evaluations. Secondary outcomes included changes in maximal wrist extension, and forearm supination, assessed by blinded raters using subjective visual estimation from standardized video recordings. The hypothesis was that BoNT-A induced muscle atrophy and structural changes (including fibrosis and neurogenic alterations) disrupt forearm muscle function, leading to progressive loss of wrist extension and forearm supination, and ultimately a decline in bimanual hand function over time. Methods Study design and setting This retrospective cohort study was conducted at a tertiary university hospital from 2010 to 2022 and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. The study was approved by the institutional ethics committee and all parents and participants gave written consent. Population Eligibility criteria Children were eligible for inclusion if they had: (1) unilateral spastic CP; (2) age at baseline from 5 to 12 years old; (3) at least two AHA assessments with a minimum three-year interval; and (4) consistent follow-up at our institution. Exclusion criteria Upper limb surgery during the study period, other neuromuscular conditions, or incomplete medical records. Sources and methods for participants selection All eligible patients meeting inclusion criteria during the study period were identified through retrospective chart review. Patients were followed longitudinally with standardized assessments at baseline and follow-up visits (minimum three-year interval). Follow-up duration was determined by the availability of repeated AHA evaluations meeting the minimum interval criterion. Patients were stratified into two groups according to treatment history:• BoNT-A+PT group: patients who received upper limb BoNT-A injections in combination with physical therapy• PT group: patients who received physical therapy without BoNT-A injections • BoNT-A+PT group: patients who received upper limb BoNT-A injections in combination with physical therapy • BoNT-A+PT group: patients who received upper limb BoNT-A injections in combination with physical therapy • PT group: patients who received physical therapy without BoNT-A injections • PT group: patients who received physical therapy without BoNT-A injections Interventions All patients received standardized PT consisting of one session per week throughout the study period, with one group was additionally treated with BoNT-A. Each session lasted approximately 45-60 minutes and typically included stretching, range of motion exercises, functional task practice, and parent education. However, we did not systematically document the specific content or focus of individual therapy sessions (e.g., emphasis on stretching, strengthening, or functional training), which may have varied between therapists and over time. Orthosis use was documented for each patient. In the BoNT-A group, injections were performed by one single experienced physicians onabotulinumtoxinA (Botox®, Allergan). Target muscles included the flexor carpi ulnaris (FCU), pronator teres (RP), flexor carpi radialis (FCR), biceps, and adductor pollicis, based on individual clinical assessment. Dosing followed our institutional standardized protocol of 1–1.5 U/kg per muscle (with typical total doses of 25–30 U per muscle), depending on size of the muscle, body weight and severity of spasticity. Rationale for treatment allocation Until 2010, upper limb botulinum toxin injections were routinely practiced in our clinic for children with unilateral CP presenting with spasticity, regardless of severity level. Following the emergence of evidence suggesting potential long-term adverse effects on muscle structure and function, our institutional standard of care evolved, leading to a progressive reduction in BoNT-A injections.8,10 Accordingly, patient allocation to the BoNT-A+PT group or PT group was primarily determined by treatment era rather than by baseline disease severity. While clinical judgment regarding the presence of spasticity influenced individual treatment decisions within each era, the temporal shift in institutional practice patterns means that treatment allocation was driven predominantly by when a patient presented for care rather than how severely affected they were. This temporal shift in clinical practice reduces the risk of selection bias, whereby more severely affected patients might otherwise been preferentially selected for BoNT-A injections. AHA assessment AHA evaluations were performed by specialized ergotherapists using standardized video recordings at baseline and follow-up (minimum three-year interval). The AHA assesses how effectively children with unilateral CP use their affected hand during activities that require bimanual handling, with scores ranging from 0 to 100 units. It is a standardized, criterion-referenced test for children aged 18 months to 18 years. 11 The AHA scores were calculated by specialized ergotherapists trained in the assessment method. Range of motion assessment Maximal wrist extension and forearm supination were visually assessed from standardized videos by three evaluators blinded to group allocation (two senior surgeons, one physician). Evaluators independently estimated maximal angles achieved during AHA evaluations. Outcomes measures Baseline demographic and clinical characteristics including age, sex, CP classification, baseline functional status, and orthosis use were reported to describe the population. The primary outcome was the change in AHA score (0-100 units) between baseline and follow-up. The secondary outcomes were the change in maximal wrist extension and forearm supination angles. Preservation of wrist extension was defined as maximal angle ≥ 0°, and preservation of forearm supination was defined as maximal angle ≥ 0°. Changes between baseline and follow-up (delta values: Δ) were calculated for all outcomes. Minimal clinically important difference (MCID) thresholds were defined as 5 units for AHA, 11 and using the smallest detectable change (SDC) derived from inter-rater reliability analyses for the maximal wrist extension and forearm supination angles. Patients were classified as improved (Δ ≥ +MCID), stable (Δ within ± MCID), or declined (Δ ≤ −MCID). Bias Several potential sources of bias were considered. Selection bias may have occurred as treatment allocation was not randomized, with BoNT-A decisions based on clinical judgment and treatment era. However, the temporal shift in practice patterns (routine use before 2010, progressive reduction thereafter) reduces confounding by indication, as allocation was determined primarily by era rather than baseline severity. Assessment bias was minimized through blinding of video-based evaluators to treatment group allocation. Information bias may have occurred due to the retrospective design and potential incomplete data capture. Differential follow-up between groups (median 7 years vs 4.5 years) may have influenced outcome trajectories and represents a potential source of bias. These limitations were addressed through transparent reporting and conservative interpretation of findings. Study size No formal sample size calculation was performed given the retrospective observational design. All eligible patients meeting inclusion criteria during the study period were included to maximize statistical power while acknowledging the exploratory nature of this analysis. Statistical analysis Analyses were performed using R software (version 4.4.1, R Foundation for Statistical Computing, Vienna, Austria) and the RStudio interface (version 2024.09.0, Posit software, PBC). The level of significance was set at p < 0.05, and we reported 95% confidence intervals (95%CI) and effect sizes (ES). The normality of continuous variables was tested using Q–Q plots and the Shapiro–Wilk test. Complete case analysis was performed given the small sample size. Descriptive statistics were calculated for all variables. Continuous variables with non-normal distributions were reported as median [interquartile range] or mean ± standard deviation [range], as appropriate. Between-group comparisons were performed using unpaired Mann-Whitney U tests for continuous outcomes and Fisher’s exact test for categorical outcomes, reported as n (%). Inter-rater raliability for video-based assessment (visually) of wrist extension and forearm supination was analyzed using intraclass correlation coefficients (ICC) for wrist extension and supination, calculated using a two-way random effects model with single measures and absolute agreement using the icc function from the irr package in R software. Standard error of measurement (SEM = SD × √(1 − ICC)) and smallest detectable changes (SDC= SEM × 1.96 × √2) were calculated to assess measurement precision. The SDC was used as the MCID threshold for ROM outcomes (20° for wrist extension, 30° for forearm supination). ICC values were interpreted as: < 0.50 poor, 0.50–0.75 moderate, 0.75–0.90 good, and > 0.90 excellent. Results Population flow and recruitment A total of 30 patients were identified from the institutional database during the study period (2010–2022). Of these, 19 were excluded: 10 had no AHA evaluation available, 5 had only a single AHA evaluation without follow-up video, 2 had a follow-up interval of less than three years, 1 had incomplete medical records, and 1 had insufficient outcome data. The final study sample comprised 11/30 (37%) patients (BoNT-A+PT, n=5; PT alone, n=6). Among these 11 included patients, none were lost to follow-up. Descriptive data All patients had unilateral spastic cerebral palsy (hemiplegia) and were classified as MACS level I. Baseline characteristics are summarized in Table 1. Median age was 7 years [IQR: 5-7] in the BoNT-A+PT group versus 9 years [IQR: 7-10] in the PT group (p=0.095). Median follow-up duration was 7 years [IQR: 5-8] in the BoNT-A+PT group and 4.5 years [IQR: 3-5] in the PT group (p=0.140). Orthosis use was more frequent in the BoNT-A+PT group (4/5, 80%) than in the PT group (2/6, 33%), although this difference was not statistically significant (p=0.347). No patients were lost to follow-up during the study period. All patients had unilateral spastic cerebral palsy (hemiplegia) and were classified as MACS level I, indicating comparable baseline functional profiles between groups.Table 1.Population characteristics.​BoNT-A+PT n=5PT n=6p-value95%CIESFollow-up duration7.0 [5.0; 8.0]4.5 [3.2; 5.0]0.140-1 to 60.326Sex, F/M3/5 (40%)0/6 (0%)0.122-0.1 to 1.00.719Baseline age7.0 [5.0; 7.0]9 [7.2; 10.8]0.095-6 to 10.395Orthosis use, n (%)4/5 (80%)2/6 (33%)0.347-0.2 to 1.00.266 Baseline measurements AHA38 [35; 39]58 [53; 59]0.082-25 to 260.419Max wrist extension0 [0; 5]10 [10; 10]0.066-40 to 00.454Max supination-20 [-20; -10]0 [0; 8]0.134-55 to 250.334Data are presented as median [interquartile range] for continuous variables and n (%) for categorical variables. Between-group comparisons were performed using Mann-Whitney U test for continuous variables and Fisher’s exact test for categorical variables. AHA: Assisting Hand Assessment (score range 0–100); BoNT-A: botulinum toxin type A; PT: Physical Therapy; ES: effect size; 95% CI: 95% Confidence Interval for the difference between groups; F: female; M: male. Table 1.Population characteristics.​BoNT-A+PT n=5PT n=6p-value95%CIESFollow-up duration7.0 [5.0; 8.0]4.5 [3.2; 5.0]0.140-1 to 60.326Sex, F/M3/5 (40%)0/6 (0%)0.122-0.1 to 1.00.719Baseline age7.0 [5.0; 7.0]9 [7.2; 10.8]0.095-6 to 10.395Orthosis use, n (%)4/5 (80%)2/6 (33%)0.347-0.2 to 1.00.266 Baseline measurements AHA38 [35; 39]58 [53; 59]0.082-25 to 260.419Max wrist extension0 [0; 5]10 [10; 10]0.066-40 to 00.454Max supination-20 [-20; -10]0 [0; 8]0.134-55 to 250.334Data are presented as median [interquartile range] for continuous variables and n (%) for categorical variables. Between-group comparisons were performed using Mann-Whitney U test for continuous variables and Fisher’s exact test for categorical variables. AHA: Assisting Hand Assessment (score range 0–100); BoNT-A: botulinum toxin type A; PT: Physical Therapy; ES: effect size; 95% CI: 95% Confidence Interval for the difference between groups; F: female; M: male. Population characteristics. Data are presented as median [interquartile range] for continuous variables and n (%) for categorical variables. Between-group comparisons were performed using Mann-Whitney U test for continuous variables and Fisher’s exact test for categorical variables. AHA: Assisting Hand Assessment (score range 0–100); BoNT-A: botulinum toxin type A; PT: Physical Therapy; ES: effect size; 95% CI: 95% Confidence Interval for the difference between groups; F: female; M: male. BoNT-A injection details In the BoNT-A+PT group, each child received a single injection event. Individual patient details are presented in Table 2. Injections were administered at a median age of 8 years (range: 5-10), with a median interval from injection to final follow-up of 6.6 years (range: 2-14). Target muscles included FCU (5/5), pronator teres (5/5), biceps (2/5), and adductor pollicis (2/5), FCR (1/5) (Table 2). Notably, all patients received injections to both FCU and pronator teres, the primary deforming forces affecting wrist extension and forearm supination. No adverse effects were reported.Table 2.BoNT-A injection details.Patient IDAge (year)Time from injection to final FU (years)Muscles injectedDoseFirst AHAInjectionLast AHAPAT_00375127FCU, Pronator teres, Biceps, Adductor pollicis1.5 U/kg/muscle (25 U/muscle)PAT_00689112FCU, Pronator teres, Biceps1.5 U/kg/muscle (30 U/muscle)PAT_00878157FCU, Pronator teres1.5 U/kg/muscle (25 U/muscle)PAT_009581614FCU, Pronator teres1.5 U/kg/muscle (25 U/muscle)PAT_010510133FCU, Pronator teres, Adductor pollicis, FCR1 U/kg/muscle (25 U/muscle)BoNT-A: botulinum toxin type A; FU: follow-up; FCU: flexor carpi ulnaris; FCR: flexor carpi radialis; U: units. Dosing was standardized at 1–1.5 U/kg per muscle. No adverse effects were reported. Table 2.BoNT-A injection details.Patient IDAge (year)Time from injection to final FU (years)Muscles injectedDoseFirst AHAInjectionLast AHAPAT_00375127FCU, Pronator teres, Biceps, Adductor pollicis1.5 U/kg/muscle (25 U/muscle)PAT_00689112FCU, Pronator teres, Biceps1.5 U/kg/muscle (30 U/muscle)PAT_00878157FCU, Pronator teres1.5 U/kg/muscle (25 U/muscle)PAT_009581614FCU, Pronator teres1.5 U/kg/muscle (25 U/muscle)PAT_010510133FCU, Pronator teres, Adductor pollicis, FCR1 U/kg/muscle (25 U/muscle)BoNT-A: botulinum toxin type A; FU: follow-up; FCU: flexor carpi ulnaris; FCR: flexor carpi radialis; U: units. Dosing was standardized at 1–1.5 U/kg per muscle. No adverse effects were reported. BoNT-A injection details. BoNT-A: botulinum toxin type A; FU: follow-up; FCU: flexor carpi ulnaris; FCR: flexor carpi radialis; U: units. Dosing was standardized at 1–1.5 U/kg per muscle. No adverse effects were reported. AHA score evolution As illustrated Figure 1(a), using the MCID threshold of 5 points for the AHA, the majority of patients in the PT group (5/6, 83%) improved, whereas no patients in the BoNT-A+PT group met the threshold for improvement (0/5). Decline was observed in 1/5 (20%) of BoNT-A+PT-treated patients and in none of the PT group. The remaining patients were classified as stable: 4/5 (80%) in the BoNT-A+PT group and 1/6 (17%) in the PT group.Figure 1.Evolution of functional outcomes between baseline and follow-up in patients with botulinum toxin type A and physical therapy (BoNT-A+PT) group (n=5) and those with PT alone (n=6). (a) Assisting Hand Assessment (AHA) score evolution; (b) Maximal wrist extension evolution. (c) Maximal forearm supination evolution. Individual changes are color-coded based on minimal clinically important difference (MCID, AHA: 5, wrist extension: 20°, forearm supination: 30°) thresholds, green = improvement, red = decline, grey = stable. Evolution of functional outcomes between baseline and follow-up in patients with botulinum toxin type A and physical therapy (BoNT-A+PT) group (n=5) and those with PT alone (n=6). (a) Assisting Hand Assessment (AHA) score evolution; (b) Maximal wrist extension evolution. (c) Maximal forearm supination evolution. Individual changes are color-coded based on minimal clinically important difference (MCID, AHA: 5, wrist extension: 20°, forearm supination: 30°) thresholds, green = improvement, red = decline, grey = stable. These MCID-based findings were consistent with the observed magnitude of change. All BoNT-A+PT patients showed AHA deterioration (Δ: mean=-3.6 ± 2.7; range: -8.0 to -1.0), while all patients in PT group (n=6) showed AHA improvement (Δ: mean +9.0 ± 6.4; range=0.0 to +18.0). Range of motion: Inter-rater reliability Inter-rater reliability for subjective video-based assessment, conducted by three blinded evaluators, was excellent for wrist extension (ICC = 0.90, SEM = 9.84°, SDC = 19.28°) and good for forearm supination (ICC = 0.78, SEM = 15.44°, SDC = 30.26°). These SDC values were used to define MCID thresholds (20° for wrist extension, 30° for forearm supination). Range of motion: Wrist extension As illustrated Figure 1(b), at baseline, wrist extension (maximal angle ≥ 0°) was preserved in 4/5 patients in the BoNT-A+PT group and in all 6 patients in the PT group. At follow-up, wrist extension was preserved in none of the BoNT-A+PT patients (0/5) compared with 5 of 6 patients in PT group. Based on MCID thresholds, decline occurred in all BoNT-A+PT patients (5/5, 100%) and in 1/6 (17%) in PT group. No patient in either group improved, and stability was observed exclusively in the PT group (5/6, 83%). The magnitude of change confirmed these findings: the BoNT-A+PT group showed marked deterioration in wrist extension (Δ: mean −53° ± 19°, median −50°, range −75° to −30°), whereas the PT group demonstrated minimal change (Δ: mean −4° ± 8+, median 0°, range −20° to 0°). Range of motion: Forearm supination As illustrated Figure 1(c), at baseline, forearm supination (maximal angle ≥ 0°) was preserved in 1/5 patients in the BoNT-A+PT group versus 5/6 patients in the PT group. At follow-up, supination was preserved in none of the BoNT-A+PT group (0/5) versus 4/6 patients in the PT group. Based on MCID thresholds, decline was documented in 4/5 (80%) of BoNT-A+PT patients and in none of the PT patients. The remaining BoNT-A+PT patient (1/5, 20%) was classified as stable, as were all PT patients (6/6, 100%). No improvement occurred in either group. The magnitude of change was consistent with these classifications: the BoNT-A+PT group showed substantial deterioration in forearm supination (Δ: mean −30° ± 7°, median −30°, range −40° to −20°), while the PT group remained essentially unchanged (Δ: mean +2° ± 9°, median 0°, range −5° to +20°). Discussion Key results This retrospective cohort study compared long-term upper limb functional evolution in children with unilateral CP (MACS 1) treated with BoNT-A+PT versus PT alone. The principal findings were threefold: (1) all children in the BoNT-A+PT group demonstrated systematic deterioration in bimanual performance (AHA), with no patient meeting the threshold for improvement, (2) marked loss of wrist extension (mean Δ = −53°) and forearm supination (mean Δ = −30°) occurred exclusively in the BoNT-A+PT group; and (3) children treated with PT alone showed functional improvement or stability over a similar follow-up. These findings were consistent across primary and secondary outcomes, with significant between-group differences despite the small sample. Comparison with other studies These results align with and extend Klingels et al. 12 In their cohort (81 uCP, mixed MACS levels, longitudinal follow-up), children receiving BoNT-A showed greater AHA decline compared with non-injected children, although most were MACS II–III. Our homogeneous MACS I sample reduces baseline severity bias and suggests that repeated injections may contribute to decline independently, even in children expected to improve.1–3 The loss of joint mobility observed in our study provides additional evidence of long-term harm not previously reported in this population. Conversely, Lidman et al. 13 (n=10, MACS I–III, 4-year follow-up) reported maintained AHA gains at 4 years (+7.5 AHA units [98% CI 2–16]), with 8/10 participants improving by ≥5 AHA units. First, their intervention combined BoNT-A with intensive multimodal rehabilitation (constraint-induced movement therapy, goal-directed training, and structured bracing), whereas our BoNT-A+PT group received only standardized weekly physical therapy. This intensive rehabilitation may have counteracted the deleterious effects of BoNT-A on muscle structure. Second, their study included children across MACS levels I–III, making direct comparison with our homogeneous MACS I population difficult. Third, cumulative dose and injection cycles may have differed. While repeated injections are associated with progressive muscle changes including atrophy and fibrosis,7,8,14 our patients received only a single injection event yet still demonstrated progressive deterioration over a median 7-year follow-up, suggesting that even a single injection during critical periods of musculoskeletal development may trigger long-term structural changes. Notably, all patients received injections to both FCU and pronator teres, the primary deforming forces affecting wrist extension and forearm supination, which may explain the consistent pattern of deterioration observed across all BoNT-A+PT patients despite variation in secondary target muscles. Importantly, the authors observed that significant AHA improvement at 1 year was no longer present at 4 years, suggesting initial benefit followed by attenuation — a trajectory that, if extended beyond 4 years, may ultimately lead to the deterioration observed in our cohort. When examining the broader literature on upper limb BoNT-A in children with CP, the evidence for functional improvement remains inconsistent. A recent systematic review by Klein et al. 4 found improvement in only 3 of 7 AHA studies, with no reports of systematic deterioration. However, most studies had follow-up ≤12 months, likely insufficient to detect long-term trends. In contrast, longer follow-up studies (>5 years), including ours and Klingels et al., 12 suggest a trajectory from initial improvement to later decline, indicating that short-term benefits may be transient or even reversible. Interpretation in context of natural history The improvement observed in our PT group (mean Δ AHA = +9.0) is consistent with the natural history of hand function in unilateral CP. Longitudinal studies have demonstrated that children with unilateral CP follow a “stable limit” developmental pattern, reaching 90% of their maximum AHA performance between 30 months and 8 years of age, with subsequent stability through adolescence. Declines in upper-limb ROM typically occur only after 9 years of age and primarily in children with MACS II-III, not MACS I.1,2 All patients in both groups had comparable baseline functional profiles (MACS level I), supporting the premise that the groups were similar at baseline despite the non-randomized design. Klingels et al. 12 reported AHA decline after age 9 exclusively in children with MACS II-III, not in MACS I, which aligns with our PT group findings. In contrast, the systematic deterioration observed in the BoNT-A+PT group deviates markedly from this expected natural history. Because both groups shared identical MACS levels at baseline, the divergent trajectories are more plausibly attributable to treatment exposure, specifically repeated BoNT-A injections, rather than pre-existing differences in severity. Possible mechanisms Several biological mechanisms may explain our observations. Persistent muscle atrophy has been documented up to one year following BoNT-A injection, with incomplete recovery of muscle volume.7,8 Peeters et al. 15 (n=15 children, lower limb BoNT-A) demonstrated medial gastrocnemius volume reduction of 5.2% and semitendinosus reduction of 16.2% at 8-10 weeks post-injection. De Beukelaer et al. 16 (n=15 children, lower limb BoNT-A) reported a 17% decrease in mid-belly cross-sectional area normalized to skeletal growth at 6 months post-BoNT-A, with a 5% increase in echo-intensity, suggesting altered muscle composition. Lambrechts et al. 8 (n=22 children, lower limb BoNT-A, 1-year follow-up) extended these findings to one year, showing a declining trend in muscle volume at 3, 6, and 12 months post-injection, with a reduced growth rate compared to untreated controls. These documented muscle volume reductions of 20-40% provide biological plausibility for the substantial ROM deterioration observed in our study. Histologically, Valentine et al. 9 (n=20 patients, upper and lower limb injections) reported neurogenic atrophy in 60% of patients 4 months to 3 years post-BoNT-A, with type 1 fiber loss strongly correlated with injection number (r=0.89, p<0.001). Other studies confirmed persistent fiber alterations, increased variability, and reduced capillarization up to 6 months. 17 These changes may accumulate with repeated injections and become permanent; accordingly, The Lancet Neurology 6 warned that frequent BoNT-A use can lead to lasting muscle atrophy and fibrosis. Importantly, our patients received only a single injection event, yet demonstrated progressive deterioration over a median 7-year follow-up, suggesting that even a single injection may trigger irreversible structural changes when administered during critical periods of musculoskeletal development. These cumulative structural alterations may explain the progressive loss of joint range of motion observed in our BoNT-A+PT group. The prounounced deterioration in wrist extension (mean change -53°) and forearm supination (mean change -30°) far exceeds the SDC thresholds (20° and 30° respectively), confirming true deterioration rather than measurement error. This magnitude of change is biologically plausible given the documented 20-40% muscle volume reductions and histopathological changes described above, and suggests structural changes beyond transient spasticity reduction and may instead reflect irreversible muscle remodeling after repeated injections. The complete loss of preserved wrist extension (from 4/5 patients at baseline to 0/5 at follow-up) and forearm supination (from 1/5 to 0/5) in the BoNT-A+PT group, contrasted with maintained range of motion in the PT group (wrist extension: 6/6 to 5/6; supination: 5/6 to 4/6), provides compelling evidence that these observed changes exceed what would be expected from the natural progression of spasticity alone despite comparable baseline functional status (MACS I). Clinical implications The striking contrast between groups in our study, with 100% of BoNT-A+PT-patients showing deterioration in AHA and range of motion versus 83-100% improvement/stability in the PT group, raises important questions about the long-term benefit-risk ratio of upper limb BoNT-A injections in children with relatively preserved function (MACS I). As highlighted in recent literature, treatment success after BoNT-A injections varies widely, and better characterization of treatment response predictors is urgently needed. 6 Our findings suggest that MACS level I children with unilateral CP may represent a subgroup at particular risk of long-term functional deterioration following repeated BoNT-A injections, possibly because their baseline muscle structure is better preserved and therefore more vulnerable to toxin-induced atrophy. Van Heest et al. 18 demonstrated that tendon transfer surgery in upper extremity CP produced superior functional outcomes compared to BoNT-A injections or ongoing therapy alone, with sustained improvements in both ROM and functional hand use. Combined with our observation of long-term deterioration following even a single BoNT-A injection, these findings suggest that surgical intervention may be preferable to BoNT-A in appropriately selected patients. We intentionally excluded patients who underwent upper limb surgery to isolate the effects of BoNT-A versus PT alone and compare trajectories against expected natural history. For children with MACS I demonstrating progressive functional decline, early consideration of surgical options may be warranted rather than BoNT-A injections. Limitations This study has several limitations that should be considered when interpreting the results. First, regarding study design and potential biases, the small sample size (n=11) limits statistical power and generalizability. The retrospective design introduces potential selection bias; children receiving BoNT-A may have had more severe spasticity initially. Indeed, there was a trend toward lower baseline values in the BoNT-A+PT group compared to the PT group for AHA (median 38 vs. 58, p=0.082), wrist extension (median 0° vs. 10°, p=0.066), and forearm supination (median -20° vs. 0°, p=0.134), although none reached statistical significance. However, the risk of selection bias is mitigated by several factors. First, treatment allocation was primarily determined by the era of clinical management rather than by baseline disease severity. Until 2010, BoNT-A injections were routinely indicated in our institutional clinical practice; subsequent changes in the standard of care led to reduced use of injections. As result, group assignment was primarly determined by this temporal shift in clinical practice rather than by individual patient characteristics. Second, although there was a trend toward lower baseline AHA scores in the BoNT-A+PT group, one patient in the PT group presented a severely impaired baseline AHA score (<20 units) yet demonstrated improvement at follow-up, suggesting that even children with significant baseline deficits can improve without BoNT-A intervention. This observation further supports the hypothesis that the deterioration observed in the BoNT-A+PT group is not solely attributable to baseline severity. Second, regarding potential confounders, the longer follow-up duration in the BoNT-A+PT group (6.8 vs. 4.2 years) could theoretically allow more time for functional deterioration to occur. However, this would not explain the consistent improvement observed in the PT group over their respective follow-up period. Moreover, according to established natural history data, children with MACS I typically show stable AHA scores from age 7 to 18 years. This makes it unlikely that longer follow-up duration alone accounts for the deterioration observed in the BoNT-A+PT group. 2 The younger age of the BoNT-A+PT group (median 7 vs. 9 years), although not statistically significant different (p=0.095), may have contributed to greater functional decline due to progressive spasticity during periods of active musculoskeletal growth. However, this age difference would be expected to favor the BoNT-A+PT group, as younger children with MACS I typically show continued functional improvement or stability according to natural history studies. Consequently, the deterioration observed in this group is particularly concerning and unlikely to be explained by age alone.1–3 Thirdly, regarding measurement limitations, the lack of standardized timing for AHA assessments, the absence of blinding for treatment allocation, and the use of video-based ROM assessment rather than direct goniometry represent important constraints. However, inter-rater reliability was good to excellent (ICC 0.78–0.90), and we used rigorous SDC-based thresholds derived from these reliability data. The SDC values were used as MCID threshold for wrist extension (20°) and forearm supination (30°) to ensure that only changes exceeding measurement error were classified as clinically meaningful. While these SDC values were derived from inter-rater reliability rather than test-retest data and may therefore be conservative, the magnitude of observed deterioration in the BoNT-A+PT group (mean wrist extension: −53°; supination: −30°) far exceeds these thresholds, confirming true deterioration rather than measurement error. Additionally, convergent findings across multiple independent outcomes (AHA, wrist extension, and supination all deteriorating in the BoNT-A+PT group) strengthen confidence in the validity of our findings despite the absence of direct goniometric measurement. Furthermore, while PT frequency was standardized at one weekly sessions for all patients, we did not systematically document the specific therapeutic content (e.g., emphasis on stretching, strengthening, or functional training), which may have varied between patients, therapists, and over time. Finally, we did not systematically assess selective motor control (SMC), which may influence susceptibility to progressive deformity following BoNT-A injection. Baseline SMC deficits in wrist extensors and forearm supinators could predispose patients to greater functional decline, though this hypothesis caonnot be evaluated with our available data. Strengths Despite these limitations, our study has notable strengths. The long follow-up (mean 5.5 years, range 3-11 years) exceeds that of most previously published studies in this field, allowing for robust assessment of long-term outcomes. The use of AHA, a validated gold-standard measure for bimanual performance, together with blinded evaluation of the range of motion, further strengthens our findings. Additionally, the homogeneity of our population, with all aprticipants classified MACS level I, minimizes confounding related to baseline functional severity and allows a more direct and meaningful comparison between treatment groups. Generalizablility These findings are derived from a single tertiary center and a small, homogeneous sample of children with unilateral spastic CP classified as MACS level I. While the homogeneity of the population strengthens internal validity, it limits the generalizability of these results to children with more severe functional impairment (MACS II–III), those with bilateral CP, or those receiving different rehabilitation protocols. Future directions Prospective randomized studies with long-term follow-up are needed to confirm these findings. Systematic surveillance of BoNT-A-treated patients should include functional outcomes beyond immediate post-injection assessments.Future research should focus on identifying subgroups of patients at increased risk of deterioration and optimizing treatment protocols regarding dose, frequency, and target muscle selection. Given the emerging evidence of potential long-term harm, futher studies should specifically investigate whether MACS level I children represent a population in whom BoNT-A should be used with particular caution or avoided altogether. Additionally, research is required to determine whether a cumulative dose threshold exists beyond which irreversible muscle damage may occur, and whether strategies such as limiting the number of injections or extending intervals between treatments could mitigate long-term adverse effects. Conclusion In this retrospective cohort study of children with unilateral spastic CP (MACS I), upper limb BoNT-A injections combined with PT were associated with a systematic deterioration of bimanual performance, as well as a marked loss of wrist extension and forearm supination over a mean follow-up of 5.5 years. In contrast, children receiving PT alone demonstrated functional improvement or stability over time. These findings should be interpreted with caution given the small sample size, retrospective design, and potential confounders. Future prospective studies with larger sample sizes and long term follow-up are needed to confirm these findings. Until such data are available, clinicians should carefully balance the potential short-term benefits of BoNT-A injections against possible long-term functional risks when treating upper limb spasticity in children with relatively preserved manual function.