Preoperative glycated hemoglobin predicts postoperative plantar pressure improvement after posterior cervical laminoplasty in patients with cervical spondylotic myelopathy and diabetes mellitus

Abstract

Background

Diabetes mellitus (DM) may adversely affect neurological recovery after cervical decompression, but its relationship with postoperative plantar pressure restoration in cervical spondylotic myelopathy (CSM) remains insufficiently defined.

Objective

To evaluate whether preoperative glycated hemoglobin (HbA1c) is associated with postoperative improvement in plantar pressure distribution after posterior single-door laminoplasty in patients with CSM and DM.

Methods

This retrospective single-center study evaluated diabetic patients with CSM treated with posterior single-door laminoplasty and followed for at least 24 months. Dynamic plantar pressure assessment was performed preoperatively and at final follow-up, with medial foot pressure (MFP) and lateral foot pressure (LFP) used as the primary biomechanical outcomes. Linear and logistic regression analyses were used to examine the association between preoperative HbA1c and plantar pressure recovery after adjustment for selected clinical and radiological variables. Receiver operating characteristic (ROC) analysis was performed as an exploratory assessment of discrimination.

Results

Postoperatively, plantar loading shifted toward a more physiological distribution, with reduced MFP and increased LFP. Higher preoperative HbA1c was independently associated with smaller improvements in both ΔMFP2 and ΔLFP2 at 2 years. Radiological parameters, including canal narrowing ratio and modified K-line interval, showed associations with outcome in univariable analyses, but their effects were attenuated after adjustment. HbA1c demonstrated fair discriminatory ability for unfavorable plantar pressure recovery, with AUC values of 0.72 for ΔMFP2 and 0.68 for ΔLFP2.

Conclusions

In diabetic patients with CSM undergoing posterior laminoplasty, poorer preoperative glycemic control was associated with less favorable improvement in plantar pressure distribution at long-term follow-up. HbA1c may be useful as one component of preoperative risk stratification within a laminoplasty cohort, but its discriminatory performance was only fair and should not be interpreted as a stand-alone treatment threshold.

Introduction

Cervical spondylotic myelopathy (CSM) is the most common cause of non-traumatic spinal cord dysfunction in adults and frequently manifests with lower-extremity incoordination and gait disturbance [1, 2]. In recent years, quantitative gait assessment has gained increasing interest as an adjunct to conventional neurological evaluation, because it may detect locomotor abnormalities not fully captured by routine clinical scales. Among these methods, plantar pressure analysis provides an objective description of weight-transfer asymmetry and abnormal medial-lateral loading during stance and gait in patients with CSM [3–7].

Diabetes mellitus (DM) may further complicate postoperative recovery in this population. Chronic hyperglycemia is associated with microvascular dysfunction, impaired neural repair, and peripheral neuropathic change, all of which may reduce the capacity for functional restoration after spinal cord decompression [8–10]. Previous clinical studies have suggested that diabetic patients, particularly those with poor glycemic control, may experience less favorable recovery after cervical spine surgery; however, most reports have focused on complications, Japanese Orthopaedic Association (JOA) scores, or general patient-reported outcomes rather than objective postoperative gait-related metrics [11–17].

Posterior single-door laminoplasty remains a widely used motion-preserving procedure for multilevel CSM. Because surgical approach itself may influence postoperative recovery patterns, combining laminoplasty, anterior decompression, and laminectomy-fusion in a single analysis may obscure procedure-specific prognostic relationships. The present study therefore focused on a relatively homogeneous laminoplasty cohort in order to examine prognostic factors for postoperative plantar pressure recovery within this specific surgical setting [1, 2, 18].

The aim of this study was to investigate the association between preoperative HbA1c and long-term improvement in plantar pressure distribution in diabetic patients with CSM undergoing posterior single-door laminoplasty. We also explored whether this association persisted after adjustment for selected radiological indicators of cervical cord compression. We hypothesized that higher preoperative HbA1c would be associated with less favorable postoperative restoration of medial-lateral plantar loading balance.

Methods

Study design and patient selection

This retrospective observational study included consecutive patients with cervical spondylotic myelopathy (CSM) and type 2 diabetes mellitus (DM) who underwent posterior single-door laminoplasty at our institution between January 2018 and December 2022. The institutional ethics committee approved the study protocol, and all patients provided informed consent for clinical data use.

This retrospective study was designed as an analysis within a single surgical cohort rather than a comparison across operative techniques. Patients treated with anterior decompression, posterior laminectomy with fusion, or combined approaches were not included, because these procedures differ substantially in indication, alignment correction, segmental stabilization, and postoperative rehabilitation, all of which could independently influence gait-related outcomes. By restricting the cohort to posterior single-door laminoplasty, we aimed to reduce procedure-related heterogeneity and to evaluate the prognostic value of preoperative HbA1c within a more uniform surgical setting. Accordingly, the present findings should be interpreted as applicable to laminoplasty-treated diabetic CSM patients and not as evidence that laminoplasty is superior to other surgical strategies.

Patients were eligible if they:

1. had a confirmed diagnosis of CSM based on clinical manifestations and MRI evidence of cervical spinal cord compression;

2. had DM diagnosed according to ADA criteria;

3. underwent single-door laminoplasty without additional anterior or combined procedures;

4. completed ≥ 24 months of postoperative follow-up, including plantar pressure testing and radiological assessment.

Exclusion criteria included:

1. prior cervical spine surgery;

2. coexisting neurological disease potentially affecting gait (e.g., Parkinson’s disease, stroke, neuropathies unrelated to DM);

3. lower-limb deformity, amputation, ulceration, or conditions interfering with plantar pressure measurement;

4. inadequate imaging quality;

5. incomplete laboratory, radiological, or gait data.

The screening procedures and reasons for exclusion are summarized in the study flowchart (Fig. 1). Among the 641 excluded patients, the main reasons were: prior cervical spine surgery or combined anterior–posterior procedures (n = 46), coexisting neurological disorders affecting gait (n = 33), severe lower-limb deformities/amputation/ulceration interfering with plantar pressure assessment (n = 9), and poor image quality, incomplete data, or follow-up < 24 months (n = 553). Each excluded patient was assigned a single primary exclusion reason to avoid double counting.

Fig. 1.

Patient flowchart. Flowchart illustrating patient identification, eligibility assessment, and final cohort formation. Among 732 screened patients with cervical spondylotic myelopathy, 91 diabetic patients met the inclusion criteria, underwent posterior single-door laminoplasty, and had complete data for the primary analyses. Of the 641 excluded patients, reasons and counts are provided in the flowchart. The same final analytic cohort was used for descriptive, regression, and ROC analyses

Surgical procedure

All surgeries were performed by the same senior spine team using a standardized posterior single-door laminoplasty technique. A midline posterior approach was used to expose laminae and facet joints. A hinge was created on the contralateral side, and the ipsilateral lamina was opened to decompress the spinal cord. An open-door plate or suture anchor system was placed to maintain laminar elevation. No facet joint violation or supplemental fusion was performed. Patients underwent routine postoperative rehabilitation emphasizing cervical ROM restoration and gait training.

Clinical evaluation

Neurological status was assessed using the Japanese Orthopaedic Association (JOA) score for cervical myelopathy preoperatively and at final follow-up. Diabetes-related variables included duration of diabetes, fasting glucose, insulin treatment status, and preoperative HbA1c measured within 1 week before surgery.

Because of the retrospective nature of the study, additional performance-based gait metrics, such as gait speed, Timed Up and Go testing, formal balance testing, fall events, and patient-reported functional outcome measures, were not uniformly available and were therefore not included in the primary analysis. Likewise, nerve conduction studies were not routinely performed as part of the perioperative assessment in all patients; consequently, electrophysiological grading of diabetic peripheral neuropathy could not be incorporated into the present study. To mitigate the potential confounding effect of severe peripheral neuropathy on plantar pressure outcomes, we applied strict exclusion criteria for conditions that could substantially distort plantar loading, including severe lower-limb deformities, amputation, and active ulceration. In addition, medical records were reviewed for clinical documentation suggestive of major gait-limiting peripheral neuropathy (e.g., marked distal sensory loss, severe balance impairment attributable to neuropathy, or non-spinal neurological conditions), and such cases were excluded when identified. Nevertheless, because standardized neuropathy grading and electrophysiological confirmation were not available for all patients, residual confounding by subclinical or variably documented diabetic peripheral neuropathy cannot be fully excluded.

The JOA recovery rate was calculated as:

JOA recovery rate (%) = (postoperative JOA − preoperative JOA) / (17 − preoperative JOA) × 100.

Recovery1 and Recovery2 denote the JOA recovery rates at 6 months and 2 years, respectively.

Plantar pressure change indices were defined as absolute changes (unit: N/kg). To ensure that higher values consistently reflected better recovery, we defined:

ΔMFP1 = MFP_pre − MFP_6mo, and ΔMFP2 = MFP_pre − MFP_2 year;

ΔLFP1 = LFP_6mo − LFP_pre, and ΔLFP2 = LFP_2 year − LFP_pre.

Under these definitions, larger ΔMFP indicates a greater reduction of pathological medial loading, and larger ΔLFP indicates a greater restoration of lateral loading, both representing more favorable plantar pressure recovery.

Radiological assessment

Standardized cervical radiographs (neutral, flexion, extension) and MRI scans were acquired preoperatively and at follow-up. Radiological measurements included:

• C2–7 Cobb angle: overall cervical lordosis on lateral radiograph.

• C2–7 sagittal vertical axis (SVA): horizontal distance between the C2 plumb line and the posterior superior corner of C7.

• Cervical range of motion (ROM): difference between flexion and extension Cobb angles.

• Pavlov ratio: sagittal canal diameter/vertebral body diameter at the most stenotic level.

• Canal narrowing ratio (CNR): ratio of the anteroposterior spinal cord diameter at the compression level to that at the C2 level on mid-sagittal MRI.

• mK-line interval (mK-line INT): perpendicular distance between the maximal anterior spinal cord margin and the modified K-line on sagittal T2-weighted MRI, reflecting the spatial reserve available for posterior shift after decompression.

Measurements were performed by two trained observers blinded to clinical outcomes. The mean value of the two measurements was used in analyses.

Conventional MRI was used to confirm the diagnosis and to derive CNR and mK-line interval measurements. Although intramedullary T2-weighted signal abnormality was reviewed during routine imaging assessment, its severity was not graded in a sufficiently standardized manner across the retrospective cohort to permit reliable inclusion in the multivariable models.

Plantar pressure measurement

Dynamic plantar pressure assessment was performed using a pressure-sensitive platform (RSscan International, Belgium) placed in the middle of a 6-m walkway. Before data acquisition, the system was calibrated according to the manufacturer’s instructions. Patients walked barefoot at a self-selected comfortable speed after familiarization trials. A mid-gait protocol was used to minimize the influence of gait initiation and deceleration. A valid trial was defined as a natural uninterrupted pass with complete foot contact on the platform and without obvious targeting behavior. Three valid trials were recorded for each patient, and the averaged values were used for analysis in order to reduce within-session variability.

Medial foot pressure (MFP) was defined as the mean peak load recorded over the medial forefoot and medial midfoot regions, whereas lateral foot pressure (LFP) was defined as the mean peak load over the lateral forefoot and lateral midfoot regions. Plantar loading values were normalized to body weight and are reported as N/kg, thereby allowing comparisons across patients with different body mass. Preoperative and postoperative measurements were processed using the same regional mask definitions and analysis workflow.

Because this was a retrospective clinical study, formal test-retest reliability statistics for the platform were not independently recalculated for the present cohort. To improve intra-session reproducibility, however, all measurements were obtained using a standardized acquisition protocol and averaged across repeated valid trials.

Statistical analysis

Continuous variables were expressed as mean ± standard deviation, and categorical variables as frequencies and percentages. Univariable linear regression was first used to examine the associations between candidate preoperative variables and plantar pressure improvement. To avoid model overfitting in view of the sample size, multivariable models were kept parsimonious and included variables with biological relevance and/or a univariable p value < 0.10. Logistic regression was then used to identify factors associated with poor recovery, defined a priori as the lowest quartile of improvement in ΔMFP2 or ΔLFP2. For consistency, ROC analyses were performed using the same binary endpoint (poor recovery defined as the lowest quartile of ΔMFP2 or ΔLFP2), with “poor recovery” treated as the positive class.

Receiver operating characteristic (ROC) analysis was performed as an exploratory assessment of discrimination. The resulting AUC values and Youden-derived cut-off points were interpreted as hypothesis-generating rather than definitive clinical decision thresholds. All statistical analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY), and two-sided p values < 0.05 were considered statistically significant.

Results

Patient characteristics

After application of the eligibility criteria, 91 patients constituted the final analytic cohort, and this same cohort was used for the descriptive analyses, regression models, and ROC analyses. The mean age was 61.4 ± 7.9 years, and 52 patients (57.1%) were male. The mean BMI was 25.8 ± 3.4 kg/m²; 29 patients (31.9%) were current smokers and 44 (48.4%) had hypertension. The mean duration of diabetes was 9.2 ± 4.6 years, the mean preoperative HbA1c level was 7.1 ± 1.2%, and 27 patients (29.7%) were receiving insulin therapy. (Table 1).

Table 1.

Baseline demographic, clinical, radiological, and plantar pressure characteristics of included patients

Category	Variables	Mean ± SD / n (%)
Demographics	Age (years)	61.4 ± 7.9
	Sex (male/female)	52 / 39
	BMI (kg/m²)	25.8 ± 3.4
	Smoking (yes/no)	29 / 62
	Hypertension (yes/no)	44 / 47
Diabetes-related indices	Diabetes duration (years)	9.2 ± 4.6
	HbA1c (%)	7.1 ± 1.2
	Insulin therapy (yes/no)	27 / 64
Clinical outcomes	Preoperative JOA score	9.0 ± 1.5
	Postoperative JOA score (2 years)	14.4 ± 1.6
Radiological parameters	C2–7 Cobb angle (°)	13.2 ± 8.1
	Canal narrowing ratio	0.38 ± 0.09
	mK-line INT (mm)	2.4 ± 1.1
	C2–7 SVA (mm)	23.8 ± 10.6
	Cervical ROM (°)	36.7 ± 12.3
	Pavlov ratio	0.79 ± 0.12
Plantar pressure parameters	Preoperative MFP (N/kg)	45.12 ± 6.21
	Postoperative MFP at 2 years (N/kg)	36.04 ± 4.82
	Preoperative LFP (N/kg)	55.28 ± 5.36
	Postoperative LFP at 2 years (N/kg)	58.63 ± 4.74

Values are presented as mean ± SD or n (%)

Baseline radiological evaluation showed a mean C2–7 Cobb angle of 13.2 ± 8.1°, C2–7 SVA of 23.8 ± 10.6 mm, cervical ROM of 36.7 ± 12.3°, a canal narrowing ratio (CNR) of 0.38 ± 0.09, and a Pavlov ratio of 0.79 ± 0.12. (Table 1).

Clinical and radiological outcomes

The mean preoperative JOA score was 9.0 ± 1.5 and improved to 14.4 ± 1.6 at 2-year follow-up, corresponding to an approximate recovery rate of 67%.

No patient experienced neurological deterioration after surgery, and there were no cases of implant failure or reoperation during the follow-up period (data not shown).

Radiological alignment was generally preserved after laminoplasty, without obvious instability-related deterioration during follow-up.

Changes in plantar pressure after surgery

Preoperatively, patients demonstrated excessive medial plantar loading and relatively reduced lateral loading. The mean preoperative medial foot pressure (MFP) was 45.12 ± 6.21 N/kg, whereas lateral foot pressure (LFP) averaged 55.28 ± 5.36 N/kg. At 2 years postoperatively, MFP decreased to 36.04 ± 4.82 N/kg and LFP increased to 58.63 ± 4.74 N/kg, indicating a redistribution of plantar loading toward a more physiological pattern. (Table 1).

These quantitative changes were consistent with the plantar pressure maps, which showed a shift from concentrated medial forefoot loading preoperatively to a more balanced and posteriorly shifted pressure pattern after surgery (Fig. 2).

Fig. 2.

Representative plantar pressure distribution maps before and after surgery. Heatmaps illustrate the transition from preoperative excessive medial foot loading to a more physiologic, laterally balanced plantar pressure pattern at 2-year follow-up. Postoperative images demonstrate reduced medial forefoot burden and improved overall load symmetry, consistent with clinical improvement following posterior cervical decompression

Association between HbA1c and plantar pressure improvement

At 6 months, none of the examined variables—including HbA1c—showed a significant association with early changes in MFP or LFP. BMI showed borderline associations with both ΔMFP1 and ΔLFP1, whereas preoperative JOA, CNR, and mK-line INT showed borderline associations with ΔMFP1 only (Table 2).

Table 2.

Linear regression analysis of 6-month plantar pressure improvement (ΔMFP1 and ΔLFP1)

Predictors	ΔMFP1 (T, p, 95% CI)	ΔLFP1 (T, p, 95% CI)
Age	0.68, 0.50, (–0.006, 0.012)	0.63, 0.53, (–0.005, 0.010)
Sex	–0.89, 0.37, (–0.151, 0.064)	–0.91, 0.36, (–0.144, 0.055)
BMI	–1.73, 0.088*, (–0.024, 0.002)	–1.68, 0.097*, (–0.023, 0.001)
Smoking	0.51, 0.61, (–0.074, 0.124)	0.47, 0.64, (–0.072, 0.118)
Hypertension	–0.33, 0.74, (–0.102, 0.073)	–0.29, 0.77, (–0.097, 0.069)
HbA1c	–0.49, 0.63, (–0.067, 0.042)	–0.45, 0.66, (–0.064, 0.040)
Preoperative JOA	1.81, 0.076*, (–0.001, 0.027)	1.63, 0.109, (–0.001, 0.024)
CNR	1.74, 0.086*, (–0.008, 0.102)	1.59, 0.117, (–0.010, 0.086)
mK-line INT	1.69, 0.095*, (–0.001, 0.026)	1.61, 0.111, (–0.002, 0.025)
C2–7 Cobb	–0.71, 0.48, (–0.012, 0.006)	–0.64, 0.52, (–0.010, 0.005)
SVA	0.32, 0.75, (–0.006, 0.008)	0.28, 0.78, (–0.006, 0.008)
ROM	0.37, 0.71, (–0.004, 0.006)	0.34, 0.73, (–0.003, 0.005)
Pavlov ratio	–0.28, 0.78, (–0.226, 0.163)	–0.25, 0.80, (–0.209, 0.157)
Multivariable model:
Predictors	ΔMFP1	ΔLFP1
CNR	1.42, 0.16	1.30, 0.20
mK-line INT	1.37, 0.18	1.29, 0.21
BMI	–1.21, 0.23	–1.16, 0.25
Preoperative JOA	1.17, 0.26	—

Univariable analysis includes all predictors; variables with *p < 0.10 were entered into multivariable regression

At 2 years, higher preoperative HbA1c was associated with smaller improvements in both ΔMFP2 and ΔLFP2 in univariable linear regression (Table 3).

Table 3.

Linear regression analysis of 2-year plantar pressure improvement (ΔMFP2, ΔLFP2)

Predictor	β (ΔMFP2)	t (ΔMFP2)	p (ΔMFP2)	β (ΔLFP2)	t (ΔLFP2)	p (ΔLFP2)
Age	-0.179	-1.72	0.089	-0.170	-1.63	0.104
Sex	-0.060	-0.57	0.570	-0.065	-0.61	0.540
BMI	-0.038	-0.36	0.720	-0.036	-0.34	0.740
Smoking	-0.117	-1.11	0.270	-0.110	-1.04	0.300
Hypertension	0.105	1.00	0.320	0.103	0.98	0.330
HbA1c	-0.297	-2.93	0.004	-0.257	-2.51	0.015
Preoperative JOA	0.191	1.84	0.071	0.185	1.78	0.080
CNR	0.209	2.02	0.046	0.193	1.86	0.067
mK-line INT	0.237	2.30	0.024	0.218	2.11	0.038
C2–7 Cobb	0.073	0.69	0.490	0.068	0.64	0.530
SVA	-0.036	-0.34	0.730	-0.039	-0.37	0.710
ROM	-0.025	-0.24	0.810	-0.030	-0.28	0.780
Pavlov ratio	-0.176	-1.69	0.097	-0.159	-1.52	0.134
Multivariable Model:
Predictor	Partial r (ΔMFP2)*	t (adj)	p (adj)	Partial r (ΔLFP2)**	t (adj)	p (adj)
HbA1c	-0.269	-2.58	0.012	-0.242	-2.31	0.027
CNR	0.185	1.74	0.086	0.171	1.61	0.113
mK-line INT	0.210	1.98	0.052	0.194	1.83	0.072
Preoperative JOA	0.130	1.21	0.230	0.124	1.16	0.250
Pavlov ratio	-0.152	-1.42	0.160	—	—	—

* ΔMFP2 model df = 85 (n = 91; 5 predictors)

** ΔLFP2 model df = 86 (n = 91; 4 predictors)

*p < 0.10

β and partial r were derived from t statistics using: effect size = t / √(t² + df)

These findings were consistent with the scatter plots, which showed moderate negative correlations between HbA1c and ΔMFP2 (r = − 0.42, p < 0.01) and between HbA1c and ΔLFP2 (r = − 0.38, p = 0.015) (Fig. 3).

Fig. 3.

Scatter plots showing associations between preoperative HbA1c and plantar pressure recovery. A Scatter plot demonstrating the negative correlation between preoperative HbA1c and improvement in medial foot pressure (ΔMFP2) at 2 years. B Scatter plot showing a similar negative relationship between HbA1c and improvement in lateral foot pressure (ΔLFP2). Each dot represents one patient. Regression lines with 95% confidence bands illustrate the inverse association between glycemic burden and postoperative gait restoration

In multivariable models adjusting for age, BMI, preoperative JOA, CNR, mK-line INT, and Pavlov ratio, HbA1c remained independently associated with ΔMFP2 and ΔLFP2 (Table 3).

Contribution of CNR and mK-line INT

Radiological indices reflecting the severity and configuration of spinal cord compression also showed meaningful associations with outcome. In univariable analysis, lower CNR and smaller mK-line INT were associated with reduced improvement in ΔMFP2, whereas for ΔLFP2, mK-line INT reached statistical significance and CNR showed a borderline association (Table 3).

CNR and mK-line INT were likewise associated with JOA recovery at 2 years (T = 2.01, p = 0.047 and T = 2.10, p = 0.039, respectively; Table 4).

Table 4.

Linear regression analysis of neurological recovery (JOA Recovery1 and Recovery2)

Predictor	Recovery1 (T, p, 95% CI)	Recovery2 (T, p, 95% CI)
Age	–0.91, 0.37	–1.78, 0.081*
Sex	0.62, 0.54	0.66, 0.51
BMI	–1.25, 0.22	–1.70, 0.093*
Smoking	–0.75, 0.45	–1.18, 0.24
Hypertension	–0.40, 0.69	–0.94, 0.35
HbA1c	–1.10, 0.28	–2.41, 0.020
Preoperative MFP	0.72, 0.47	1.38, 0.18
Preoperative LFP	–0.55, 0.59	–1.05, 0.30
C2–7 Cobb	0.66, 0.51	0.92, 0.36
SVA	–0.79, 0.43	–1.21, 0.23
CNR	1.91, 0.060*	2.01, 0.047
mK-line INT	1.76, 0.083*	2.10, 0.039
Pavlov ratio	–0.82, 0.41	–1.33, 0.19
Multivariable model (Recovery2):
Predictors	T	p
Age	–1.39	0.17
BMI	–1.26	0.22
HbA1c	–2.09	0.042
CNR	1.79	0.078
mK-line INT	1.92	0.059
Pavlov ratio	–1.41	0.16

In multivariable analysis, the effects of CNR and mK-line INT on plantar pressure and JOA recovery showed a persistent trend but did not reach conventional significance after adjustment for HbA1c and other covariates (p values between 0.05 and 0.10; Tables 3 and 4).

These findings suggest that radiological severity and cord decompression geometry modulate outcome, but their influence is partly mediated or overshadowed by systemic metabolic status.

Predictors of poor plantar pressure recovery

Binary logistic regression identified HbA1c as the main risk factor for poor plantar pressure improvement. For ΔMFP2, higher HbA1c increased the odds of being in the lowest quartile of improvement (OR 3.58, 95% CI 1.46–8.85, p = 0.006 in univariable analysis; adjusted OR 3.21, 95% CI 1.24–8.12, p = 0.015 in multivariable analysis; Table 5).

Table 5.

Binary logistic regression for risk factors of poor ΔMFP2

Predictor	OR (95% CI)	p
Age	1.78 (0.74–4.32)	0.19
Sex	1.11 (0.46–2.65)	0.81
BMI	1.29 (0.54–3.16)	0.57
Smoking	2.01 (0.82–4.92)	0.12
Hypertension	1.26 (0.51–3.09)	0.61
HbA1c	3.58 (1.46–8.85)	0.006
Preoperative JOA	1.69 (0.70–4.12)	0.23
CNR	0.54 (0.24–1.23)	0.14
mK-line INT	0.49 (0.21–1.17)	0.11
Pavlov ratio	2.02 (0.85–4.80)	0.10
Multivariable model:
Predictor	OR (95% CI)	p
HbA1c	3.21 (1.24–8.12)	0.015
CNR	0.63 (0.25–1.55)	0.31
mK-line INT	0.58 (0.23–1.45)	0.24
Pavlov ratio	1.83 (0.74–4.48)	0.19

Poor recovery defined as ΔMFP2 below the 25th percentile

Similarly, for ΔLFP2, HbA1c remained an independent risk factor for poor recovery (univariable OR 2.98, 95% CI 1.21–7.35, p = 0.019; adjusted OR 2.84, 95% CI 1.12–7.26, p = 0.027; Table 6).

Table 6.

Binary logistic regression for risk factors of poor ΔLFP2

Predictor	OR (95% CI)	p
Age	1.92 (0.81–4.63)	0.14
Sex	1.11 (0.46–2.65)	0.48
BMI	1.36 (0.57–3.28)	0.57
Smoking	1.95 (0.80–4.79)	0.087*
Hypertension	1.31 (0.54–3.15)	0.54
HbA1c	2.98 (1.21–7.35)	0.019
Preoperative JOA	1.78 (0.73–4.35)	0.20
CNR	0.59 (0.26–1.37)	0.22
mK-line INT	0.55 (0.24–1.29)	0.17
Pavlov ratio	1.89 (0.78–4.51)	0.12
Multivariable model:
Predictor	OR (95% CI)	p
HbA1c	2.84 (1.12–7.26)	0.027
CNR	1.80 (0.72–4.58)	0.20
mK-line INT	0.66 (0.27–1.63)	0.37
Smoking	1.80 (0.72–4.58)	0.20

Smoking status showed a trend toward association with poor ΔLFP2 but did not persist in adjusted models. CNR and mK-line INT exhibited ORs < 1, indicating a possible protective effect with better canal morphology, but these did not reach statistical significance after adjustment.

ROC analysis for HbA1c threshold

ROC analysis showed that preoperative HbA1c had fair, rather than strong, discriminatory ability for identifying poor long-term plantar pressure recovery (Fig. 4). For ΔMFP2, the AUC was 0.72 (95% CI 0.60–0.84, p = 0.006), and the Youden-derived cut-off was 6.8%, with a sensitivity of 72.0% and a specificity of 74.6%. For ΔLFP2, the AUC was 0.68 (95% CI 0.55–0.81, p = 0.017), with a corresponding cut-off of 6.7%, sensitivity of 68.4%, and specificity of 70.1% (Table 7). These thresholds should be interpreted as exploratory reference values rather than definitive clinical decision points.

Fig. 4.

ROC curves evaluating the exploratory discriminatory value of HbA1c for poor plantar pressure recovery. A ROC curve for identifying poor recovery of ΔMFP2 (defined as the lowest quartile of ΔMFP2 improvement). B ROC curve for identifying poor recovery of ΔLFP2 (defined as the lowest quartile of ΔLFP2 improvement). Youden-derived cut-off values were 6.8% for ΔMFP2 and 6.7% for ΔLFP2; however, these values should be interpreted cautiously given the fair, rather than strong, discriminatory performance

Table 7.

ROC analysis for predictive ability of preoperative HbA1c

Parameter	ΔMFP2	ΔLFP2
AUC (95% CI)	0.72 (0.60–0.84)	0.68 (0.55–0.81)
Optimal cut-off	6.8%	6.7%
Sensitivity (%)	72.0	68.4
Specificity (%)	74.6	70.1
p-value	0.006	0.017

Outcome: poor plantar pressure recovery at 2 years (defined as the lowest quartile of ΔMFP2 or ΔLFP2 improvement)

Taken together, these findings suggest that preoperative glycemic status is independently associated with the degree of postoperative plantar pressure recovery after laminoplasty in diabetic patients with CSM. However, recovery remained multifactorial, and HbA1c should be interpreted as one prognostic variable within a broader clinical and radiological context.

Discussion

In this cohort of diabetic patients with cervical spondylotic myelopathy (CSM) treated by posterior single-door laminoplasty, we found that preoperative glycemic control—quantified by HbA1c—was independently associated with long-term normalization of plantar pressure distribution. Higher HbA1c levels were related to smaller reductions in pathological medial foot loading and less recovery of lateral loading at 2 years, even after adjustment for age, body mass index, baseline neurological status, and radiological indices of canal compromise. Canal narrowing ratio (CNR) and the modified K-line interval (mK-line INT) also showed significant or near-significant relationships with plantar pressure and JOA recovery, suggesting that both systemic metabolic status and local spinal canal morphology influence functional gait restoration after decompression.

An important point is that the present study was intentionally restricted to a laminoplasty cohort. Surgical approach selection in CSM depends on multiple factors, including the number and location of compressive levels, sagittal alignment, instability, and the need for direct ventral decompression. Accordingly, the current data should not be extrapolated to anterior procedures or posterior laminectomy with fusion. A true comparison of operative strategies would require a separate design with matched cohorts and procedure-specific adjustment for baseline anatomical differences.

Relationship between diabetes, HbA1c, and surgical outcome in CSM

The detrimental impact of diabetes on outcomes after cervical spine surgery has been reported in several large series [11–14]. Machino et al. observed that diabetic status was associated with inferior JOA recovery after laminoplasty in more than 500 patients with CSM, and a subsequent meta-analysis confirmed that diabetes modestly worsens functional outcome and complication risk in this population [11, 15]. Previous studies further showed that preoperative HbA1c levels correlate with recovery rate after CSM surgery, implicating chronic glycemic burden rather than the binary presence of diabetes [15–17]. Recent work across broader spine populations indicates that elevated preoperative HbA1c is associated with poorer patient-reported outcomes, delayed achievement of minimal clinically important difference (MCID), and higher readmission or reoperation rates after both anterior and posterior cervical procedures [15–17].

Our data extend these observations from global neck-related disability to an objective biomechanical endpoint—plantar pressure distribution—which directly reflects integration of long-tract spinal cord function, proprioception, and distal neuromuscular control. Previous studies focused mainly on JOA scores, pain, and generic PROMs after laminoplasty in diabetics, with inconsistent conclusions regarding whether diabetes alters long-term satisfaction. By demonstrating that higher HbA1c is linked to persistent medial overloading and insufficient restoration of lateral support during gait, our results suggest that even when subjective neck symptoms improve, subtle deficits in lower-limb control may remain in poorly controlled diabetics.

The HbA1c thresholds identified by our ROC analysis (≈ 6.7–6.8%) are broadly consistent with prior literature suggesting that poorer preoperative glycemic control is associated with less favorable postoperative outcomes after spine surgery [15, 17]. However, these thresholds should be interpreted cautiously, because the AUC values in the present study indicate only fair discriminatory performance. Khoylyan et al. recently found that HbA1c ≥ 7.3% correlated with poorer subjective outcomes after single-level ACDF [17]. A systematic review also concluded that elevated preoperative HbA1c is a robust predictor of adverse events and suboptimal recovery following spine surgery [15]. The ROC-derived HbA1c values of approximately 6.7–6.8% should be interpreted with caution. The corresponding AUC values indicate only fair discriminatory performance, and these thresholds are therefore better viewed as exploratory reference points than as definitive clinical cut-offs. In practice, HbA1c should complement—rather than replace—assessment of neurological severity, imaging findings, overall medical condition, and postoperative rehabilitation potential.

Diabetic peripheral neuropathy is a key potential confounder for plantar pressure outcomes, particularly in long-standing diabetes. Although we excluded overt non-diabetic neurological disorders affecting gait and applied strict exclusion criteria for lower-limb conditions that could markedly distort plantar loading, we cannot definitively rule out the contribution of subclinical or variably documented diabetic neuropathy in this retrospective cohort. Nerve conduction studies and standardized neuropathy grading were not routinely available; therefore, the relative contributions of peripheral nerve dysfunction versus spinal cord recovery could not be fully disentangled. Future prospective studies incorporating electrophysiological testing and validated neuropathy scoring are needed to clarify these mechanisms.

Mechanistic links between hyperglycemia, spinal cord function, and gait

Several mechanisms may explain why poor glycemic control blunts plantar pressure improvement after laminoplasty. Diabetes induces widespread microangiopathy in the spinal cord and dorsal root ganglia, leading to capillary basement membrane thickening, endothelial dysfunction, and impaired perfusion [8, 19]. Experimental models show that hyperglycemia causes structural damage to spinal cord axons and oligodendrocytes and alters nociceptive processing in the dorsal horn. Clinically, diabetic patients with spinal pathology exhibit higher rates of polyneuropathy, impaired proprioception, and delayed nerve conduction, all of which may limit the capacity of decompressed tracts to normalize complex gait patterns [8–10, 19].

In CSM specifically, gait disturbance arises from a combination of corticospinal tract injury, dorsal column dysfunction, and compensatory changes in lower-extremity kinematics. Recent gait-laboratory and plantar-pressure studies have demonstrated that CSM patients walk with prolonged stance and double-support phases, shorter stride length, increased knee valgus, and a characteristic shift toward higher medial foot loading and reduced lateral loading compared with asymptomatic controls [3–6]. Wearable sensor studies have confirmed that these abnormalities persist even in moderate disease and correlate with standard myelopathy scales. Our findings that laminoplasty reduces medial pressure and enhances lateral support are in line with these reports and with recent evidence that both anterior and posterior decompression can gradually normalize plantar pressure over time [3, 6, 20].

However, chronic hyperglycemia may restrict this plasticity. Diabetic polyneuropathy preferentially affects distal sensory fibers, potentially limiting afferent feedback from the sole during stance and push-off, while microvascular injury within the decompressed cord may cap the degree of remyelination and synaptic reorganization achievable postoperatively. This could explain why we observed a stronger association between HbA1c and plantar pressure improvement at 2 years than at 6 months: early after surgery, mechanical decompression dominates, but longer-term adaptation of gait depends more heavily on neural recovery, which is sensitive to cumulative glycemic burden.

Plantar pressure should also be interpreted as an objective biomechanical surrogate rather than a comprehensive measure of ambulatory recovery. Because gait speed, Timed Up and Go performance, formal balance testing, fall events, quality-of-life measures, and patient-reported outcome measures were not uniformly available in this retrospective cohort, the present study cannot establish that the observed pressure redistribution necessarily translates into improved real-world ambulation or lower fall risk. Future prospective studies should integrate plantar pressure analysis with broader functional outcome assessment.

Influence of canal morphology: CNR and mK-line interval

Beyond systemic factors, our results support the growing body of evidence that local canal morphology influences surgical outcome in CSM [18, 21]. Congenital or acquired canal stenosis, quantified by indices such as Pavlov ratio or canal narrowing ratio, has long been recognized as a risk factor for myelopathy and as a determinant of postoperative recovery [18]. Patients with a narrower canal tend to have more severe cord compression, greater T2-signal change, and less reserve for posterior cord drift after laminoplasty.

The K-line and its derivatives have emerged as practical MRI-based tools to describe the relationship between cervical alignment, ossified structures, and the spinal cord [21]. Several studies have shown that K-line negativity, reduced K-line distance, or unfavorable K-line tilt are associated with worse neurological recovery or neck function after laminoplasty or posterior fusion [18, 21]. In our cohort, smaller mK-line INT and lower CNR were linked to poorer improvement in plantar pressure and JOA scores in univariable analyses, although their independent effects diminished after adjustment for HbA1c and other covariates. These trends align with prior work and suggest that inadequate posterior shift of the cord, due to severe ventral compression or kyphosis, may limit the benefit of decompression on long-tract function controlling lower-limb coordination.

The fact that mK-line INT and CNR did not remain statistically significant in fully adjusted models does not necessarily negate their importance. Rather, it may reflect limited statistical power and inter-correlation between radiological indices and baseline neurological severity. Furthermore, our primary outcome—plantar pressure—may be particularly sensitive to subclinical neuropathy and systemic microvascular disease, where HbA1c exerts a stronger influence than morphology alone.

Clinical implications

From a clinical standpoint, our findings highlight two complementary dimensions of preoperative risk stratification in diabetic CSM patients: systemic metabolic control and local canal configuration. First, the consistent association between HbA1c and both plantar mechanics and JOA recovery suggests that HbA1c should be incorporated into preoperative counseling and perioperative optimization protocols. From a clinical perspective, the present findings support careful metabolic optimization and realistic perioperative counselling in diabetic patients undergoing laminoplasty for CSM. They do not justify using a single HbA1c threshold to deny, postpone, or mandate a specific surgical procedure. Rather, elevated HbA1c may identify patients who warrant closer glucose management and more attentive postoperative gait-oriented rehabilitation.

Second, evaluation of CNR, mK-line INT, and related alignment parameters may help identify patients in whom posterior laminoplasty is less likely to achieve optimal functional recovery. Whether alternative surgical strategies would provide superior gait-related outcomes in such patients requires dedicated comparative studies.

Finally, our use of quantitative plantar pressure analysis underscores the value of objective gait metrics as complementary outcomes to traditional neurological scales [3, 7]. As shown in recent gait-laboratory and wearable-sensor studies, plantar pressure and spatiotemporal parameters can reveal subtle improvements or residual deficits that are not fully captured by JOA or mJOA scores. Incorporating such measures into routine follow-up could refine rehabilitation programs and facilitate earlier detection of unsatisfactory recovery, particularly in high-risk diabetic patients.

Limitations and future directions

This study has several limitations. First, its retrospective single-center design introduces the possibility of selection bias and residual confounding. Second, although preoperative HbA1c was available for all analyzed patients, longitudinal postoperative glycemic control, the differential effects of insulin-based versus non-insulin treatment strategies, and adherence to postoperative rehabilitation were not recorded in sufficient detail for robust adjustment. Third, diabetic peripheral neuropathy was not systematically graded with nerve conduction studies, and its contribution to plantar pressure patterns therefore could not be fully separated from that of cervical cord dysfunction. Fourth, while radiological severity was partially captured by CNR and mK-line interval, standardized incorporation of mJOA subclassification and MRI signal-change grading was not feasible in this retrospective cohort. Fifth, plantar pressure is an objective biomechanical endpoint but does not fully represent functional ambulation; gait speed, balance testing, fall data, quality-of-life measures, and patient-reported outcomes were not uniformly available. Finally, because all included patients underwent posterior single-door laminoplasty, the present study does not permit conclusions regarding the comparative effects of laminoplasty, anterior decompression, or laminectomy with fusion on postoperative gait recovery.

Conclusions

In diabetic patients with cervical spondylotic myelopathy treated with posterior single-door laminoplasty, higher preoperative HbA1c was independently associated with less favorable long-term improvement in plantar pressure distribution. This association persisted after adjustment for selected clinical and radiological variables, but the discriminatory ability of HbA1c alone was only fair. Preoperative glycemic status should therefore be regarded as one component of perioperative risk stratification rather than a stand-alone surgical decision threshold. Prospective multicenter studies incorporating standardized neuropathy assessment, postoperative metabolic control, and broader functional gait outcomes are warranted.

Authors’ contributions

Conception and design: Yang Liu and Yong Hu; Acquisition of data: Yang Liu and Zichuan Wu; Data analysis and interpretation: Yang Liu and Zichuan Wu; Statistical analysis: Xiaoyang Sun; Manuscript Preparation: Zichuan Wu, Xiaoyang Sun; Manuscript revision and modification: Yang Liu.

Funding

This work is funded by the Shanghai Municipal Education Commission Foundation (Grant No. 2023 Science and Technology 05–60) and Ningbo Clinical Research Center for Orthopedics, Sports Medicine & Rehabilitation (2024L004).

Data availability

All the data of the manuscript are presented in the paper.

Declarations

Ethics approval and consent to participate

Approval for the current study protocol was obtained from the ethics committees of Shanghai Changzheng Hospital (2021SL044). The norms on which the study is based are in accordance with the “Declaration of Helsinki”. Informed consent was explicitly obtained from all study participants for this study.

Consent for publication

Each participant expressed agreement to publish. Relevant data about the participants’ medical images and clinical details were obtained for this study with the consent of all participants.

Competing interests

The authors declare no competing interests.