Four patients (three males, mean ± SD 52 ± 19 years) were recruited using the National Spinal Cord Injury Register. The inclusion criteria were cervical incomplete SCI (International Standards for Neurological Classification of Spinal Cord Injury and American Spinal Cord Injury Impairment Scale (AIS) B, C, D) (2) at the chronic stage (over 18 months) and an age range of 18 to 70 years. The exclusion criteria were concomitant brain injury, epilepsy, high intracranial pressure, metal implants in the head area, implemented electrical devices, and pregnancy. Patients continued their conventional rehabilitation program and medication during the study. Detailed information on patient characteristics is presented in Table 1. The Helsinki University Hospital Regional Committee on Medical Research Ethics approved the study. The entire study was performed in accordance with the Declaration of Helsinki. All participants provided written informed consent for participation.

Patients selected the hand to be stimulated; all chose the dominant right hand. Patients were allowed to choose the stimulated hand to increase their motivation and commitment to the study to support their individual needs, and to enhance the possible positive effects on daily living. The contralateral hand was not stimulated. Based on earlier repetitive TMS studies (30) and previous high-PAS studies (16), PAS was administered 5 days a week for the first 2 weeks to maximize the induction of response to PAS and 3 times a week subsequently. One session included separate stimulations for three nerves (median [MN], ulnar [UN], and radial nerves [RN]), performed in alternating order. Patients were evaluated by an experienced physiotherapist before and once per month after starting high-PAS. Each patient was evaluated by the same physiotherapist, except for the last follow-ups of patient 2 and the pre-evaluation of patient 3. The physiotherapists were blinded to the setup, i.e., they did not know which side was stimulated. AT performed stimulation on patients 1–2, and KH performed stimulation on patients 3–4.

The primary outcome measure was the Medical Research Council (MRC) manual muscle test. The secondary outcomes were functional tests on hand dexterity (Box and Block Test and Nine-Hole Peg Test), hand and finger dynamometry scores, American Spinal Injury Association (ASIA) sensory and motor scores, Modified Ashworth Scale (MAS) spasticity score, International Spinal Cord Injury Pain Basic Data Set (ISCIPBDS), World Health Organization Quality-of-Life (WHOQOL) assessment, and the Spinal Cord Independence Measure (SCIM). High-PAS sessions continued as long as the MRC score level or other outcomes (or both) improved. If the MRC score decreased or did not increase further from one evaluation to another, stimulations were terminated. However, secondary outcomes and patient opinion were also considered. Stimulation timelines of each patient are summarized in Table 2. Follow-up was performed at 1 month, 3 months, 6 months, and 12 months after the last high-PAS session.

TMS was performed using a figure-of-8 coil of an eXimia magnetic stimulator with an MRI-based navigation system (Nexstim Ltd., Helsinki, Finland). Cortical representations of the abductor digiti minimi (ADM), abductor pollicis brevis (APB), and extensor digitorum communis (EDC) muscles were mapped. The site where the TMS elicited the largest MEPs in the muscle innervated by the nerve to be stimulated was chosen as the TMS “hotspot” for each muscle–nerve pair individually. All three hotspots were stored in the TMS device, and the same hotspots were used throughout the stimulations. Thereafter, the resting motor threshold (RMT) of the muscles was identified as a minimum TMS intensity eliciting at least 50 μV in 5/10 stimuli (31). For interstimulus interval (ISI) calculation, MEP latency was measured from an average of 15 MEPs elicited by TMS of 100% stimulator output.

The MN, UN, and RN were stimulated. Self-adhering surface electrodes (Neuroline 720, AMBU A/S, Ballerup, Denmark) were used for electrical stimulation and F-response measurement. Two stimulation electrodes were placed in the middle of the inner wrist for MN, in Guyon’s canal for UN, and laterally above the elbow for RN (32). During RN stimulation, electrodes were gently pressed against the humerus to enable effective nerve activation. The same electrode montage was used for the F-response measurement and PNS in high-PAS (Figure 1). Two recording electrodes were placed at a belly–tendon montage over APB, ADM, and EDC during MN, UN, and RN F-response measurements, respectively.

PNS was delivered, and F-responses were measured using a Keypoint device (Natus Medical Inc., Pleasanton, CA, United States). An experienced clinical neurophysiologist measured F-responses at supramaximal intensity with a 0.2-ms pulse for determining minimum latency out of 10 responses (33), which was used for ISI calculation. Thereafter, minimum intensity-evoking F-responses [1 F-response out of 10 stimuli (34)] were determined with a 1-ms pulse stimulation, and this intensity and pulse length were used for PNS in high-PAS. F-responses from MN were not detected in patient 3. ISI and PNS onset intensity were set to be the same as for UN (15 mA). After 2 months of PAS, the intensity was gradually increased to 19 mA, which was still tolerated by the patient.

ISI between TMS and PNS was set to synchronize arrivals of a single TMS pulse and the first pulse of the PNS train in the corticomotoneuronal synapses in the spinal cord (10). This was calculated individually for each patient by the formula [F latency – MEP latency] (see Shulga et al. (35) for a detailed description).

During high-PAS, a single TMS pulse was delivered at 100% of the maximum stimulator output (MSO). PNS was delivered as a 100-Hz train consisting of six 1-ms square wave pulses. Two hundred forty PAS sequences of one target muscle were triggered through Presentation^® software (Neurobehavioral Systems Inc., Albany, United States) at 0.2 Hz. The session duration for one nerve was 20 min. Patients voluntarily activated the target muscles with each PAS (little finger abduction/flexion of IV and V fingers for ADM, abduction and flexion at the carpometacarpal joint of the thumb for APB, and wrist and finger extension for EDC). The patients were advised to slightly contract the muscle just before the TMS pulse to draw their attention to the stimulated hand and decrease motor thresholds (16). The three target nerves were stimulated in alternate order during each PAS session. Details on the stimulation parameters of the patients are presented in Table 2. Due to occasional initial discomfort, PNS intensity was gradually adjusted to the target level during the first 10–15 pulses of stimulation in the first stimulation sessions. Habituation to the PNS intensity occurred during the first stimulation sessions. All patients tolerated 100% MSO TMS intensity without adverse events. During hand representation area stimulation, slight face sensations, twitches, or involuntary movements of the arm might have occurred.

MRC, ASIA motor and sensory scores, hand dynamometry, hand dexterity tests, pain score, spasticity score, and SCIM were evaluated at each evaluation point (baseline and pre, end of PAS, and 1-month, 3-month, 6-month, and 12-month follow-ups). Both hands were assessed. WHOQOL assessment was performed before, immediately after long-term PAS, and at 12 months of the follow-up. To detect intra-individual variability in the tests, two baseline assessments were performed before PAS administration, separated by a minimum of 1 month and a maximum of 7 months. These evaluations were performed on days when the patients had their usual general health to preclude confounding factors. Patient 1 had the evaluation after 5 months of stimulation postponed for a few weeks due to elevated blood pressure and related unwellness. The rescheduled evaluation was used as the “End of PAS” as no further improvement was detected in MRC, and stimulations were discontinued after this evaluation. Patient 1 also missed the 12-month follow-up because of temporary health deterioration unrelated to this study.

Voluntary muscular strength was assessed for each muscle separately with the MRC scale as follows: 0 = no contraction, 1 = flicker or trace of contraction, 2 = active movement, with gravity eliminated, 3 = active movement against gravity, 4 = active movement against gravity and resistance, and 5 = normal power (36). The scapula and shoulder muscles were tested in addition to the muscles innervated by MN, UN, and RN and analyzed separately.

Strength in the key hand muscles (C2–C8, T1) and sensory score in hands (C2–C8, T1–T2) were assessed according to the ASIA examination sheet (2). The sensory test consisted of light touch (LT) and pin prick (PP) tests. Grip strength was measured using an Exacta™ Hydraulic Hand Dynamometer (North Coast Medline, Inc., United States). The dynamometer was adjusted to the size of the patient’s hand. Pinch dynamometry was performed using a Baseline^® Mechanical Pinch Gauge (Fabrication Enterprises Inc., United States) for key pinch, tip pinch, and palmar pinch. The patient had three attempts for each test, and the best result in kg was used for analysis (37).

Unilateral gross manual dexterity for both hands was evaluated using the Box and Block Test (BBT) (38). The result reflects the number of transferred blocks in 1 min. Unilateral finger dexterity was measured using the Nine-Hole Peg Test (9-HPT) (39). The result was expressed as the number of successful placements performed in 50 s. Hand spasticity was assessed using the Modified Ashworth Scale (MAS) (40). MAS scores range from 0 to 4, where 0 represents normal muscle tone and higher scores represent spasticity.

SCIM is a self-reported evaluation of daily activity divided into the following three domains: self-care, respiration and sphincter management, and mobility (41). SCIM scores range from 0 to 100, with higher scores indicating better daily performance. Pain was measured using the ISCIPBDS (42). The overall ISCIPBDS score (range 0 to 36 points, higher scores indicate worse pain) was used in this study. The WHOQOL-BREF is a quality-of-life assessment divided into four domains, namely, physical, psychological, social, and environmental domains (43). The total raw score of all domains was used; a higher score indicates a better quality of life.

Subjective reports and verbal feedback were collected throughout the study. The patients were asked to verbally describe changes they experienced in their hand functions in daily life. Adverse events and other (not study-related) events causing changes in patients’ status and wellbeing were also collected.

A motor point (MP) integrity test of forearm muscles with electrical stimulation was conducted as described in Bersch et al. (28). MP was defined as the site that produces a maximal selective and defined muscle response at the lowest stimulation intensity (44). Five muscles innervated by the radial nerve (extensor carpi ulnaris [ECU], extensor carpi radialis [ECR], extensor digitorum communis [EDC], extensor pollicis longus [EPL], and abductor pollicis longus [APL]) on the dorsal aspect of the forearm and three muscles innervated by the median nerve (pronator teres [PRT], flexor digitorum profundus [FDP], and flexor pollicis longus [FPL]) on the palmar aspect of the forearm were tested using a Mettler 212 Digi-Tens (Mettler Electronics Corp., United States) electrical stimulator. One self-adhesive electrode (Everyway Medical Instruments Co., Ltd., Taiwan) was attached to the epicondyle as a reference electrode. A pen electrode with a diameter of 0.8 cm (ZMI Electronics Ltd., Taiwan) was used for stimulation. The MPs of selected muscles were detected with a pen electrode using a cartography system and calculated stimulation points (28) (Supplementary file). In addition, palmar interossei I, II, and III (IOD1-3) muscles were tested. The IOD1 MP was defined as the midpoint of the first and second metacarpophalangeal joints of the thumb (45). IOD2 MP was searched manually from the base between the index finger and middle finger and IOD3 MP between the middle finger and ring finger. The test pulse width was 260 μs with a frequency of 35 Hz. The contraction was elicited in the muscles by stimulation intensity between 20 and 50 mA. The results were categorized into innervated, partially denervated, and denervated according to muscle response (28).

MRC was pre-registered at clincaltrials.com as the primary outcome measure at timepoints before PAS, end of PAS, and 1-month follow-up. MRC results were analyzed using IBM SPSS Statistics 28.0. Non-parametric tests were used as the data were non-normally distributed in the Kolmogorov–Smirnov and Shapiro–Wilk tests. MRC data from three timepoints (pre, end of PAS, and 1-month follow-up) were analyzed using Friedman’s two-way ANOVA for related samples. A post-hoc test was performed using the Dunn–Bonferroni test. The alpha level was set at 0.05/3 = 0.0167 (Bonferroni correction) to control multiple comparisons. The results from the right (stimulated) hand and left (unstimulated) hand were analyzed separately (n = 4 hands). The same statistical analysis was also performed for MRC results from all timepoints (pre, end of PAS, and 1-, 3-, 6-, and 12-month follow-ups); the alpha level was set at 0.05/15 = 0.0033 and without results from 12-month follow up; the alpha level was set at 0.05/10 = 0.005. The mean of baseline and pre-measurements was calculated to control day-to-day variation. Muscles with a maximum score of 5 in pre-evaluation were removed from the analysis as it was not possible to detect further improvement in these muscles with MRC. The mean MRC score of all included muscles was calculated for each evaluation, and these values were used for analysis. In addition, changes between pre and other evaluations were calculated. The correlation between MRC changes from pre to end of PAS and MP integrity test was calculated using Pearson’s product–moment correlation. All muscles with a maximum score pre-evaluated to 5 were also removed from this analysis. Due to the small number of patients, the results from other measurements that were not pre-registered at clincaltrials.com are presented using descriptive methods. All data are presented as mean ± standard error.

MRC scores did not differ between baseline and pre-evaluations (p = 1.00; Wilcoxon signed-rank test in both hands). The mean of these measurements was calculated and used for analysis; the timepoint “Pre” later in the text refers to this value. MRC scores of stimulated muscles differed significantly between the three timepoints (pre, end of PAS, and 1-month follow-up) in Friedman’s test in stimulated (n = 4) [X²(2)=6.5, p = 0.039] and unstimulated hands (n = 4) [X²(2)=6.5, p = 0.039]. The post-hoc test indicated that MRC scores increased significantly from pre to end of PAS (stimulated +0.77 ± 0.1 points, p = 0.013, unstimulated +0.71 ± 0.14 points, p = 0.013; Dunn–Bonferroni test) but not from pre to 1-month follow-up (stimulated +0.73 ± 0.14 points, p = 0.077, unstimulated +0.63 ± 0.07 points, p = 0.077). MRCs between the end of PAS and 1-month follow-up did not differ (p = 0.480).

MRC scores between all six timepoints (pre, end of PAS, and 1-, 3-, 6-, and 12-month follow-ups) did not differ significantly in the stimulated [X²(5)=9.515, p = 0.090] or unstimulated hand [X²(5)=9.515, p = 0.090]. This result was affected by the dropout of one patient after 6 months of follow-up. When the 12-month follow-up was excluded from the analysis and all four patients included, Friedman’s test revealed a significant difference between the timepoints in the stimulated [X²(4)=12.101, p = 0.017] but not in the unstimulated hand [X²(4)=9.455, p = 0.051]. The post-hoc test showed a significant difference between the pre and 6-month follow-up in the stimulated hand (+0.9 ± 0.1 points, p = 0.002; Dunn–Bonferroni test) and almost a significant difference between the pre and 3-month follow-up (+0.78 ± 0.1, p = 0.01). MRC scores did not differ between the other timepoints. The individual development of MRC mean scores (Figure 2) showed that despite the lack of statistical significance at every timepoint, the MRC mean scores remained above the pre-level throughout the follow-up time up to 12 months in all three patients.

MRC scores of non-stimulated scapula and shoulder area muscles increased from before to End of PAS, especially in patients who had strength deficiencies in these muscles. Scores for patient 1 increased from 4.2/4.4 to 5 and for patient 4 from 3.6 to 4.9 in the right hand and from 3.5 to 5 in the left hand. Patient 2 had maximum test scores for the right arm and patient 3 for both arms already in the pre-test, so detecting further improvement was not possible.

The duration of PAS required to achieve the plateau of MRC varied between patients (range 14–25 weeks, mean 18.5 ± 2.4 weeks). The development of MRC scores for each patient and hand during the stimulation period, including the most notable phases of increase and plateau, is presented in Figure 3.

From pre to end of PAS, BBT scores increased in all patients on average by 6 ± 1.8 points (15%) in the stimulated hand and 5.1 ± 1 points (12%) in the unstimulated hand. BBT scores remained above the pre-level up to the 12-month follow-up (Figure 4). 9-HPT scores of the stimulated hand did not increase from the pre-test to the end of PAS; three of four patients had between 17 and 18 points already in the pre-test (18 points is the maximum). The unstimulated hand 9-HPT score decreased slightly from pre to end of PAS, and the results varied between different evaluations. Group-level results are presented in Table 3.

Hand dynamometry and palmar pinch scores increased slightly from pre to end of PAS and remained increased up to the 6-month follow-up. Key pinch increased slightly in the stimulated hand and tip pinch in the unstimulated hand. Patient 3 was unable to execute hand or pinch dynamometry in pre-evaluation but had a slight improvement in the key pinch and palmar pinch during the intervention. Group-level results are presented in Table 3.

SCIM’s overall score increased on average by 3.75 ± 2.46 points from pre to end of PAS. The change in overall score during intervention and during follow-up is presented in Figure 5.

ASIA motor score increased by 3 ± 1.4 points from pre to end of PAS in the stimulated but not in the unstimulated hand. ASIA sensory scores did not show notable change; the stimulated hand PP score increased by 3.3 ± 1.9 points from pre to end of PAS, but the increase was 1.5 ± 1.7 points after 6 months. Spasticity did not change according to the MAS score in the stimulated or the unstimulated hand. Group-level test results are presented in Table 3.

According to ISCIPBDS, neuropathic pain did not change after high-PAS (Table 3). One of four patients had continuous pain (ISCIPBDS score over 20/36 points in every evaluation), one patient did not have pain at all, and pain scores varied throughout the intervention for two patients. The mean WHOQOL total score decreased slightly from pre to end of PAS but increased from pre to 12-month follow-up (Table 3). In WHOQOL sub-questions related to quality of life and the ability to perform daily activities, quality of life increased from 3.75 ± 0.25 to 4.3 ± 0.3 and the ability to perform daily activities from 3.75 ± 0.25 to 4 from pre to end of PAS.

MP integrity test results, including data from patient 5 from Rodionov et al. (27), are presented in the Supplementary file. The MP integrity test revealed that patients 1 and 4 had only fully innervated muscles and patients 2 and 5 had individual completely or partially denervated muscles in the stimulated right hand. Patient 3 had partial or complete denervation in all flexor/ palmar side muscles, including interossei 1–3 in his right hand. In the left hand, patients 1 and 3 had individual completely or partially denervated muscles. MP integrity test results were combined with MRC score changes (from pre to end of PAS) of corresponding muscles. Thirty-eight right-hand muscles and 39 left-hand muscles from all patients were analyzed. MRC and MP integrity test results correlated positively (r = 0.515, p ≤ 0.001) in the right-stimulated hand but not in the left-unstimulated hand (0.267, p = 0.101; Pearson’s correlation coefficient).

Subjective reports on the long-term effects of high-PAS were collected from patients throughout the study. The most common improvement was that the stimulated hand felt stronger and more flexible. Examples of improved finger movements from patients 1 and 4 can be seen in Supplementary Videos. Patients also reported a tighter grip and improvements in their handwriting, ability to grasp objects, and ability to make a fist more easily. No adverse events were reported during the study.