Kunming Locomotor Training: neurological, functional, and autonomic outcomes in complete spinal cord injury

Abstract

Purpose of review

Recovery after spinal cord injury (SCI) is variable, and the contribution of locomotor training to neurological and functional outcomes remains debated. This review summarizes post-SCI locomotor recovery patterns, compares training modalities, and presents recovery findings from the Kunming Locomotor Training (KLT) program, one of the largest reported series of patients with initial complete (AIS A) injuries.

Recent findings

Several months of intensive task-specific overground training yielded substantial gains in neurological, locomotor, and autonomic outcomes. In a retrospective cohort of 485 AIS A patients, 47% improved their AIS grade, while nearly all showed some locomotor recovery measured by the Kunming Locomotor Scale (KLS). A ≥4-point KLS gain strongly predicted AIS conversion (sensitivity 83%, specificity 82%). Recovery probability was highest with lower thoracic and lumbar injuries. Improvements, including ambulation with assistive devices, and significant bladder and bowel recovery occurred even without AIS change. These outcomes reinforce the importance of active, patient-driven training.

Summary

KLT experience indicates that prolonged, intensive overground locomotor training can promote neurological and functional recovery in individuals with initially complete injuries. KLS provides a sensitive measure of functional progress. These findings underscore the clinical value of task-specific stepping and provide a new benchmark for evaluating sustained rehabilitation strategies and research into post-SCI recovery.

STUDY HIGHLIGHT

Study design: Retrospective analysis of 485 patients with initially complete SCI (AIS A) treated with the Kunming Locomotor Training (KLT) program.

Neurological outcomes: 47% converted to higher AIS grades; KLS gains strongly correlated with AIS improvement (r ≈ 0.80).

Predictive model: A ≥4-point KLS increase optimally predicted AIS conversion (sensitivity 83%, specificity 82%).

Functional outcomes: Locomotor improvements were observed even without AIS change; over one-third reached KLS level 4 or higher.

Autonomic outcomes: By discharge, 89.7% achieved bladder scores ≥9 and 72.6% achieved bowel scores ≥8.

Mechanisms: Upright, task-specific stepping may drive neural plasticity while also conferring cardiovascular, bone, and metabolic benefits.

Comparisons: Unlike robotic or treadmill-based training, KLT emphasizes active, patient-driven stepping with high therapist engagement.

Clinical implications: KLS should be considered both as a motor outcome and a prognostic marker; a 4-point gain may serve as a stratification trigger for adaptive rehabilitation or clinical trial design.

INTRODUCTION

Neurological recovery potential after spinal cord injury (SCI) is variable and difficult to predict. Standard classification measures, used as outcomes, such as the ASIA Impairment Scale (AIS), capture categorical improvements but may overlook other clinically significant gains. The Kunming Locomotor Training (KLT) program, developed over a two-decade period, provides structured, high-intensity overground gait training that emphasizes task-specific stepping, standing, and balance.

The value of locomotor rehabilitation after SCI has been well established [1]. The advent of body-weight supported locomotor training allowed individuals with severe SCI to participate more fully in upright locomotor exercises compared to complex bracing systems [2]. In parallel, health benefits of standing and upright exercises after SCI, including their effects on bones and metabolism, have been documented [3,4]. Debate emerged over the best training methods [5,6^▪], with initial enthusiasm for robotic or treadmill-based approaches shifting toward evidence supporting overground methods that permit greater variability in step cycles [7,8]. More recently, several exoskeleton systems have been tested to assist with upright walking [9,10]. In the US and Canada, inpatient rehabilitation duration is typically limited, often leaving insufficient time for intensive locomotor training. However, at Kunming Tongren Hospital, inpatient programs can last up to a year, allowing for more sustained and high-intensity interventions [11].

The KLT protocol was empirically developed through iterative clinical practice. It utilizes simple, low-cost assistive devices – such as walking platforms, four-post walkers, crutches, and canes – together with experienced therapists, progressively increasing lower-limb loading and stepping independence. The program follows a progressive adaptive principle: specific deficits are identified, targeted assistance is provided, and support is gradually withdrawn as recovery occurs. The Kunming Locomotor Scale (KLS) was created to document these stages, analogous to the WISCI scale but adapted to the Chinese clinical setting.

Prior reports have documented neurological recovery in AIS A participants [12^▪,13–15], but quantitative predictors and functional correlates have not been systematically analyzed. In this study, we tested three hypotheses: that KLS change predicts AIS improvement, that there exists a clinically relevant KLS threshold for stratifying recovery, and that neurological improvement is associated with injury level. 

Box 1.

no caption available

METHODS

Study population

We retrospectively analyzed 485 patients with SCI admitted to the Kunming Tongren Hospital rehabilitation program with AIS A classification upon admission. Patients ranged in age from 3 to 76  years (mean 35.1 years); 401 were male (82.7%) and 84 were female (17.3%). Regarding the time from injury to the start of rehabilitation: 58.6% (284 cases) began within 15  days after injury, 16.3% (79 cases) between 15 and 90  days, 2.5% (12 cases) between 90 and 180  days, 5.2% (25 cases) between 180 and 365  days, 15.1% (73 cases) from 1 to 10  years, and 2.5% (12 cases) more than 10  years postinjury. Injury distribution included 118 cervical (24.3%), 295 thoracic (60.8%), and 72 lumbar (14.9%) injuries. All patients underwent standardized neurological examinations according to ASIA criteria to confirm the diagnosis of initial complete SCI (AIS A). An ISNCSCI assessment was completed before patients started walking training, with subsequent assessments conducted every 3  months thereafter. For patients in the acute phase, the first ISNCSCI assessment was conducted upon admission, prior to surgery, with a second ISNCSCI completed 15  days after surgery, before initiating the Kunming Walking Training. After the training began, ISCNSCI assessments were conducted every 3  months. The ISNCSCI assessments for each patient were conducted by their attending physician to ensure the consistency of the evaluation results. All examiners had undergone ISNCSCI training. Here, we report the change from before KLT initiation to the one-year endpoint. The Kunming walking program costs are reimbursed if individuals belong to industrial and commercial insurance programs in China.

Kunming Locomotor Training protocol

KLT is grounded in two key principles: weight-bearing modulation, with patients’ lower limbs in direct ground contact through assistive devices to enable progressive loading; and active leg propulsion, emphasizing patient-driven stepping and balance to retrain spinal gait sequences.

The training program was empirically developed over a two-decade period and follows a progressive, adaptive principle: the identification of specific deficits, provision of targeted assistance, and systematic reduction of supports as function returns. Safety measures include therapist manual stabilization, knee bracing to prevent falls, and guided limb advancement when necessary. This approach has been described as a function–compensation–progression paradigm.

Training was delivered for up to 4–6  h per day, 6  days a week, typically in two sessions, depending on tolerance and other rehabilitation activities.

Training stages

• (1)Initial weight-bearing reduction (KLS 2–3): Minimal weight-bearing standing and gravity shift exercises, supported by walking frames and therapists to establish trunk stability and symmetry.
• (2)Progressive weight-bearing (KLS 4): Gradual increases in weight-bearing (20%→50%→80%) with assisted knee-locking and stepping. Posterior knee support was provided when voluntary control was insufficient.
• (3)Functional walking (KLS 5–10): Transition to four-point walkers, crutches, and canes, with near-complete weight-bearing and independent knee locking. Training emphasized gait quality, endurance, and adaptability to real-world terrains (slopes, thresholds, stairs).

Outcome measures

KLS: A 10-grade ordinal assessment scale designed for the KLT protocol (Table 1). Grade 1 indicates inability to stand; Grades 2–8 correspond to successive stages of assisted walking with walkers/crutches. Grade 9 denotes unsupported walking with observable instability; Grade 10 represents stable unsupported ambulation (Fig. 1).

Table 1.

The Kunming Locomotor Scale

Stage	Description
1	Unable to stand
2	Stands with wheeled arm support walker and assistance
3	Walk with assistance of pulling the arm support walker, aligning trunk and hips, facilitating movements
4	Walk with wheeled arm support walker, with assistance to lock knees
5	Walk with wheeled arm support walker, no assistance
6	Walk with 4-point walker, no assistance
7	Walk with crutches, no assistance
8	Walk with cane, no assistance
9	Walk unstably without aid or assistance
10	Walk stably without aid or assistance

The 10 ordinal grades of KLS are shown with their descriptors.

FIGURE 1.

The stages in the Kunming Locomotor Training program illustrated. The stages of the KLS are illustrated with typical walking aids and modes of assistance.

ASIA Impairment Scale (AIS): Assessed using standardized ASIA examination worksheets at admission and discharge [16,17].

SCIM III bladder and bowel subscores: The bladder score includes seven graded elements (0, 3, 6, 9, 11, 13, 15 points) from indwelling catheter use to continence without external drainage. The bowel score includes four elements (0, 5, 8, 10 points) from irregular movements to independent regular movements without accidents [18,19].

Statistical analysis

Primary analyses: Logistic regression was used to examine predictors of AIS improvement, controlling for covariates (age, sex, injury level, and time to admission). ROC analysis identified the KLS change cutoff most predictive of AIS conversion.

Secondary analyses: Regression models examined the effects of KLS change, AIS conversion, and injury level on bladder and bowel outcomes. Interaction models tested for moderation by neurological level (KLS × level; AIS × level).

Robustness checks, including Huber regression, bootstrap resampling, and outlier trimming, were performed to assess the stability of the estimates.

RESULTS

A total of 485 patients were analyzed. Demographics and injury distributions are summarized above (Table 2, Fig. 2). Regarding the time from injury to admission, the median time was 11  days, range, 0.5  days to 22.0  years; 71% were admitted within 30  days of injury, but only 2.7% were admitted within 24 h. There were 14 patients <18 years of age, comprising 2.9% of the total cohort.

Table 2.

Distribution of patients by region of injury

AIS range	No. of the patients	Ratio
C1–C8	118	24.33%
T1–T12	295	60.82%
L1–L4	72	14.85%

The distribution of the cohort is shown by the cervical, thoracic, and lumbar level groupings.

FIGURE 2.

The distribution of neurological levels at admission. The number of patients at each neurological level of injury is shown.

CHARACTERISTICS OF THE STUDY POPULATION (Table 2)

Neurological and functional recovery

By discharge, 229 patients (47%) improved their AIS grade: 105 (21.6%) converted to AIS B, 79 (16.3%) to AIS C, 37 (7.6%) to AIS D, and 8 (1.7%) to AIS E. The majority (256 cases, 52.8%) remained AIS A (Fig. 4).

FIGURE 4.

AIS grades at discharge. The final AIS grade at discharge by percentages.

KLS improvement was nearly universal, with patients showing gains ranging from 1 to 9 points (mean 3.96). The distribution at discharge was: KLS 3 (14.4%), KLS 4 (36.7%), KLS 5 (17.3%), KLS 6 (12.6%), KLS 7 (11.8%), KLS 8 (3.5%), KLS 9 (1.9%), and KLS 10 (1.9%). Notably, 36.7% achieved KLS level 4 or higher (Table 3, Fig. 3).

Table 3.

KLS scores at discharge as a percentage of the total

KLS	No. of the patients	Ratio
3	70	14.43%
4	178	36.70%
5	84	17.32%
6	61	12.58%
7	57	11.75%
8	17	3.51%
9	9	1.86%
10	9	1.86%

The KLS scores at discharge are shown as a percentage of the total.

KLS, Kunming Locomotor Scale.

FIGURE 3.

KLS scores at discharge. Bar graph of the number of patients with each final KLS score. KLS, Kunming Locomotor Scale.

KLS and AIS changes demonstrated a strong correlation (r = 0.80, R² = 0.64). Greater AIS improvements corresponded with larger KLS gains: patients with no AIS change averaged 3.02 KLS points, while those improving by 1, 2, 3, and 4 AIS grades gained 3.99, 5.43, 6.59, and 8.50 KLS points, respectively (Fig. 5).

FIGURE 5.

Mean KLS change per AIS grade change. The KLS changes obtained per change in the AIS. KLS, Kunming Locomotor Scale.

Predictive modeling

Logistic regression confirmed that each additional KLS point increased the odds of AIS improvement [odds ratio (OR) 1.22, 95% confidence interval (CI) 1.15–1.29].

ROC analysis identified a 4-point KLS gain as the optimal cutoff (AUC 0.88, sensitivity 83%, specificity 82%). Patients below this threshold had a <30% probability of AIS improvement, those at 4 points had a ~60% probability, and those with ≥5 points had a>85% probability.

Supplemental predictive models confirmed this threshold, with consistent results across bootstrap and robust regression analyses.

Injury level effects

AIS improvement rates and KLS recovery varied significantly across levels, with more caudal injuries showing greater gains. Patients at or below T10 had a substantially higher probability of AIS conversion and greater KLS gains compared with those with more rostral injuries.

Age and AIS conversion

No clear trend for the influence of older age on AIS conversion was apparent until the age of 60. The average conversion rate below age 60 was 47.3%, and above 60, it was 42.9%, with only 14 subjects at or above 60. For those patients below the age of 18, the conversion rate was higher at 57.1%.

Autonomic outcomes

Bladder scores showed significant improvement, with distribution at discharge: grade 6 (0.4%), grade 9 (51.8%), grade 11 (25.4%), grade 13 (14.9%), and grade 15 (7.6%). At admission, 65.6% had grade 3, and 32.2% had grade 6. Change score analysis confirmed that bladder improvement was nearly universal, with a mean gain of 6.4 points (median 6; range 0–10); 99.6% of participants showed positive change. Bowel scores similarly improved, with discharge distribution: grade 5 (27.4%), grade 8 (69.1%), and grade 10 (3.5%). At admission, 45.0% had grade 0, 45.2% had grade 5, and 9.9% had grade 8. However, bowel change was more limited (mean +4.2 points; median 3; range 0–10), and 23.3% of participants had no measurable improvement. This restricted distribution – particularly the clustering of values at zero or near ceiling – likely explains the weaker correlations observed between bladder and bowel outcomes and with locomotor measures. Multivariable regression demonstrated that KLS change predicted both bladder and bowel improvements (per +5 KLS ≈ +2.3 bladder points; +1.7 bowel points). AIS improvement was associated with bladder gains, and a more caudal injury level strongly predicted bowel improvements. KLS × level interaction showed that the impact of KLS recovery on these SCIM outcomes was larger at thoracic and lumbar injury levels than for cervical levels.

To further assess whether these associations reflected absolute outcome levels vs. recovery dynamics, we repeated the analysis using change scores (discharge – admission) (Fig. 6). In this analysis, correlations were substantially attenuated (e.g., ΔKLS–ΔAIS ρ = 0.75, ΔKLS–Δbladder ρ = 0.35, ΔKLS–Δbowel ρ = 0.26), particularly for bladder and bowel, consistent with the restricted variability and ceiling effects observed in these subscores. Thus, while final discharge values showed strong cross-domain associations, change-score correlations highlighted that improvements in bladder and bowel function followed a more constrained and partly independent trajectory relative to locomotor recovery.

FIGURE 6.

Spearman correlation matrices of locomotor, neurological, and autonomic outcomes at final discharge and for total change. Spearman's rank correlations (ρ) are shown for Kunming Locomotor Scale (KLS), ASIA Impairment Scale (AIS), SCIM III bladder subscore, and SCIM III bowel subscore at discharge (N = 485). Values in each cell indicate the correlation coefficient, significance stars, and corresponding P-values. Strong associations were observed between KLS and AIS (ρ = 0.75, P < 0.001) and between AIS and bladder outcomes (ρ = 0.77, P < 0.001). Correlations between locomotor measures (KLS, AIS) and bowel outcomes were weaker but still significant. All reported associations were highly significant after nonparametric testing.

Additional improvements

Patients demonstrated multifaceted neurological improvements, including partial recovery of autonomic nervous function (reappearance of sweating below injury level), enhanced excretory control, and partial restoration of sexual function. However, no consistent caudal shift of sensory level was observed.

Robustness checks

Robust regression, bootstrap confidence intervals, and sensitivity analyses confirmed the stability of results. Forest plots demonstrated consistent positive effects of KLS change across models.

DISCUSSION

This extensive, single-center retrospective analysis provides new quantitative evidence linking locomotor recovery, measured by the KLS, with neurological conversion on the AIS and autonomic outcomes in patients admitted with initially complete (AIS A) SCI. Nearly half of the cohort converted to higher AIS grades by discharge, and KLS gains strongly correlated with neurological and functional improvements. Importantly, meaningful locomotor and autonomic recovery was also observed in patients without AIS conversion, underscoring the clinical relevance of functional gains that conventional neurological measures may not capture.

Key findings and clinical implications

Four significant findings emerge. First, 47% of patients improved their AIS grade, an unusually high rate for initially complete injuries. Second, the KLS change was nearly universal and strongly correlated with AIS improvement (r ≈ 0.80), with a ≥4-point KLS gain representing a clinically meaningful threshold that optimally discriminated sustained neurological recovery. Third, functional gains, including the ability to ambulate with assistive devices, were observed even without AIS conversion. Finally, substantial improvements in bladder and bowel function were documented, highly valued outcomes for patients and families. These findings support KLS both as a clinically relevant outcome and as a potential stratification tool for future trials.

Comparison to other reports of recovery rates

In the past, recovery from complete injuries was regarded as exceptional; however, increasingly it has been reported that some individuals, especially with lower thoracic and lumbar level injuries, do experience AIS conversion [12^▪,14,20]. Early surgery may be a contributor to increased conversion rates, with reports of up to 46.1% conversion, similar to the findings of this report [21].

Based on our clinical observations, the probability of AIS conversion is related to the intensity of the initial spinal trauma [22] and injury severity observed on MRI [23]. According to our protocol, the earliest ASIA grade conversion after starting KLT training is observed during the re-examination at the third month. We also observed subsequent AIS conversions at the one-year program discharge examination.

The KLS improvement is associated with rehabilitation duration. On average, it takes 1–3  months for initially complete patients to reach KLS level 4. Many patients remain at this level because progressing from KLS level 4 to 5 requires achievement of voluntary knee joint locking, which requires greater neural control and muscle strength. This is a key obstacle that is relatively difficult to overcome in this patient population, and it often requires 4–6  months to reach KLS level 5. Overall, the higher the KLS level that patients aim to achieve, the longer the rehabilitation period required. In our prior study, which enrolled only chronically injured subjects, we observed that some outcomes, such as the Ashworth scale, plateaued at 4  months, whereas others, such as the SCIM-III score, showed improvement throughout the one-year treatment duration [24].

Mechanistic considerations

KLT emphasizes progressive weight-bearing and active stepping, which may retrain spinal circuits through several mechanisms. Standing and walking provide multisystem benefits, including cardiovascular conditioning, bone loading, and proprioceptive activation, which counteract the systemic deconditioning that follows SCI [25,26]. Weight-bearing activity engages residual spinal segmental circuits, including central pattern generators, which contribute to rhythmic locomotor output [27,28]. Sensory feedback from muscle, tendon, and joint receptors shapes activation patterns and drives adaptive changes in spinal and supraspinal networks [29].

Emerging evidence also supports metabolic adaptations with locomotor training. Six months of intensive training can reduce glucose and insulin AUCs by 15–33%, increase GLUT-4 expression, and shift muscle fibers toward greater glucose efficiency [30]. Cardiovascular benefits, including improved efficiency sustained over 24  weeks, have been demonstrated in randomized pilot studies [31]. Together, these findings suggest that upright, task-specific walking provides both neural and systemic benefits.

Comparisons with technology-based approaches

The KLT approach contrasts with technology-driven methods, such as body-weight-supported treadmill training or robotic/exoskeleton-assisted gait. These modalities standardize assistance and can be helpful for safety and reproducibility; however, they can limit step variability with the aim of stabilized balance [32] and reducing the workload of therapists [33]. Numerous robotic devices have been reported to yield rehabilitation gains [34]. KLT emphasizes active, patient-driven stepping, requiring voluntary intention and focused effort with every movement, while integrating proprioceptive, tactile, and visual sensory feedback. This aligns with principles of task-specific training and may reinforce neural pathways critical for community walking [35]. The low-tech, therapist-guided nature of KLT may allow broader scalability than robotic devices, although its success depends heavily on expertise developed at Kunming over decades. Whether this would be cost-effective in other settings is unclear.

Kunming Locomotor Scale change as a potential prognostic marker

AIS grade captures only categorical changes and is determined during a supine examination. It is insensitive to short-term neurological improvements that do not cross a classification boundary. In this study, KLS gains – especially those reaching or exceeding four points – were linked to a significant increase in the likelihood of AIS improvement. This indicates that KLS, as a more continuous and functional measure, might detect early recovery trends that AIS alone cannot identify within typical rehabilitation timeframes. This dataset is limited by the 3-month interval for each ISNCSCI exam, but the KLS is detected continuously. Reaching KLS level 4 more quickly may indicate a greater potential for recovery than achieving the same milestone at a later stage. Therefore, the time to KLS change could be considered a potential prognostic marker of neurological recovery potential from an initially complete state, but this would require further validation through prospective studies. If validated, the time to achieve KLS thresholds could serve as adaptive triggers in clinical trials or as a stratification factor, helping guide therapy intensity and patient enrollment in interventional studies.

Limitations and future directions

The retrospective design and single-center cohort limit causal inference and generalizability. Further, we report only baseline and endpoint AIS grades. Other differences in patient-specific factors, therapist expertise, and patient motivation may influence outcomes. While robustness checks confirmed the stability of our models, prospective studies are needed to validate the predictive relationship between KLS and AIS improvement, as well as to evaluate the durability of observed locomotor and autonomic gains beyond discharge. Further mechanistic work, including neurophysiological, urological, and imaging biomarkers, may help clarify who will be most responsive to prolonged weight-bearing locomotor training.

CONCLUSION

Intensive task-specific locomotor training can yield substantial improvements in motor, neurological, and autonomic function in patients who are initially neurologically complete. KLS gains, especially ≥4 points, strongly predict AIS conversion and functional recovery. Injury level shapes both motor and autonomic outcomes, with enhanced recovery in injuries at or below T10. These findings support the use of KLS as both a clinical and research tool and strengthen the rationale for intensive overground locomotor training programs.

Acknowledgements

None.

Financial support and sponsorship

None.

Conflicts of interest

There are no conflicts of interest.

REFERENCES AND RECOMMENDED READING

Papers of particular interest, published within the annual period of review, have been highlighted as:

• ▪ of special interest

• ▪▪ of outstanding interest