A neurological prognostic nomogram integrating biomarkers and clinical indicators for surgically treated patients with osteoporotic spinal fracture and spinal cord injury

Abstract

To construct and validate a prediction model of neurological rehabilitation for patients with osteoporotic spinal fracture combined with spinal cord injury based on the expression of serum neurotrophic factor (Neuritin) and transforming growth factor β1 (TGF-β1 ), so as to provide basis for accurate clinical rehabilitation. A total of 189 patients with osteoporotic spinal fracture combined with spinal cord injury who were admitted to our hospital from January 2022 to September 2024 were included in the study. A random 7:3 score was used as the training set (n = 132) and validation set (n = 57). In the training set, independent influencing factors were screened by univariate and multivariate Logistic regression analysis, and a Nomogram model was constructed. The model performance was evaluated by using the consistency index (C-index), calibration curve, and receiver operating characteristic curve (ROC), and was verified in the validation set. There was no significant difference in baseline data between the training set and the validation set (P > 0.05). Multivariate logistic analysis showed that serum Neuritin (ng/mL, OR = 1.285), TGF-β1 (ng/mL, OR = 0.979), vertebral compression degree (%) (OR = 0.960), spinal canal space-occupying rate (%) (OR = 0.879), 25-hydroxyvitamin D (ng/mL, OR = 1.103), and alkaline phosphatase (U/L, OR = 0.973) were the independent factors for poor neurological rehabilitation (all P < 0.05). The C-index for the nomogram model was 0.842 and 0.821 in the training and validation sets, respectively, the calibration curve fitted well, and the area under the ROC curve (AUC) was 0.850 (95% CI: 0.75–0.95) and 0.826 (95% CI: 0.65-1.00), respectively. The Nomogram model constructed based on serum levels of Neuritin and TGF-β1 can effectively predict the neurological rehabilitation effects of patients with osteoporotic spinal fracture combined with spinal cord injury, and has potential clinical application value.

Supplementary Information

The online version contains supplementary material available at10.1038/s41598-026-50468-4.

Introduction

Osteoporotic spinal fracture is a common severe bone metabolic disease in the elderly, and its incidence increases year by year with the trend of population aging¹. Spinal cord injury, one of the most serious complications, not only leads to motor and sensory dysfunction, but also significantly reduces the quality of life of patients and brings heavy burden to society and family². According to statistics, about 30% of patients with osteoporotic spinal fractures are complicated with spinal cord injury, and only 40% to 60% of them can obtain satisfactory neurological recovery. Therefore, it is of great clinical significance to accurately evaluate the rehabilitation prognosis at an early stage³. At present, the main bases for clinical evaluation of neurological functional recovery after spinal cord injury include ASIA classification, imaging parameters and bone metabolism indicators, etc. However, these traditional indicators cannot fully reflect the dynamic changes of the microenvironment of nerve regeneration⁴. In recent years, the role of neurotrophic factors in the repair of spinal cord injury has received extensive attention⁵. As a new type of neurotrophic factor, Neuritin promotes the recovery of neurological function by regulating axonal regeneration and synapse formation. Transforming growth factor β 1 (TGF-β1) can affect the process of spinal cord injury repair by regulating the immune inflammatory response and glial scar formation⁶. However, the existing studies mostly focus on the action mechanism of a single factor, and no prediction model based on multiple biomarkers has been established⁷. The pathological mechanism of osteoporotic spinal fracture combined with spinal cord injury is complex, which involves multiple factors such as bone metabolism imbalance, local microenvironment disorder and nerve regeneration inhibition⁸. The purpose of this study was to explore the predictive value of nomogram prediction model based on serum Neuritin, TGF-β1 levels and imaging parameters in the neurological rehabilitation of patients with osteoporotic spinal fracture combined with spinal cord injury. In this study, the correlation between biomarkers and clinical outcomes was systematically evaluated through large sample retrospective analysis, and the independent influencing factors were screened by multi-factor Logistic regression, finally a quantifiable prediction model was constructed. The results will provide a theoretical basis for the clinical development of personalized rehabilitation program, and is expected to fill the gap in the field of precision medicine.

Materials and methods

Study population

This was a retrospective consecutive cohort study conducted in the Department of Orthopedics of our hospital from January 2022 to September 2024. Inclusion criteria: (1) Patients meeting the diagnostic criteria for osteoporotic spinal fractures (dual-energy X-ray bone mineral density t value ≤-2.5); (2) The ASIA grade of spinal cord injury is A-E; (3) Age ≥ 60 years old; (4) Patients were admitted to hospital within 72 h after injury. Exclusion criteria: (1) Complicated with brain injury or peripheral neuropathy; (2) Medical history of malignant tumor; (3) Severe hepatic and renal insufficiency. This workwas supported by China Medical Education Association 2022 major scientific problems and key topics of medical technology problems (2022KTM022), Yangzhou science and technology plan (YZ2022070, YZ2023266). All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and national research committee and with the Declaration of Helsinki. All patients signed informed consent forms.

Treatment

All patients underwent posterior decompression and internal fixation, and were routinely given methylprednisolone percussive therapy (30 mg/kg intravenous drip within 15 min, followed by 5.4 mg/kg/h for 23 h) and regular rehabilitation training (including physical therapy, occupational therapy, neuro-facilitation technology, etc.) after the surgery for6 months.

Data collection

Baseline data were collected including age, gender, BMI, fracture segment (T12 or L1 fractures, with no a priori exclusion of other spinal segments), baseline ASIA classification, and comorbidities. Venous blood samples (5 mL) were collected from patients within 24 h of admission (before posterior decompression and internal fixation surgery, consistent with the “within 72 hours after injury” admission criterion) for measurement of serum Neuritin and TGF-β1. Samples were centrifuged at 3000 rpm for 10 min at 4 °C, and the separated serum was stored at -80 °C without freeze-thaw cycles until analysis. All serum biomarker detections were performed in the Clinical Laboratory of Northern Jiangsu People’s Hospital Affiliated to Yangzhou University (certified by ISO 15189). Serum Neuritin levels were detected using the Human Neuritin ELISA Kit (Manufacturer: R&D Systems, Catalog Number: DY7367), with a detection range of 0.156-10 ng/mL, sensitivity of 0.078 ng/mL, intra-assay coefficient of variation (CV) < 5%, and inter-assay CV < 8%. Serum TGF-β1 levels were detected using the Human TGF-β1 ELISA Kit (Manufacturer: Abcam, Catalog Number: ab100647), with a detection range of 15.6–1000 pg/mL, sensitivity of 7.8 pg/mL, intra-assay CV < 4%, and inter-assay CV < 6%. All samples were measured in duplicate, and the average value was used for analysis; samples from the training and validation sets were randomly distributed into batches to ensure batch-blinding to patient grouping and outcomes., with a detection range of 0.156-10 ng/mL, sensitivity of 0.078 ng/mL, intra-assay CV < 5%, and inter-assay CV < 8%. Serum TGF-β1 levels were detected using the Human TGF-β1 ELISA Kit (Manufacturer: Abcam, Catalog Number: ab100647), with a detection range of 15.6–1000 pg/mL, sensitivity of 7.8 pg/mL, intra-assay CV < 4%, and inter-assay CV < 6%. Laboratory personnel performing the detections were blinded to patients’ neurological rehabilitation outcomes. Bone metabolic indexes including 25-hydroxyvitamin D (25-OH-D) and alkaline phosphatase (ALP) were detected by chemiluminescence detection. Imaging parameters were measured by CT three-dimensional reconstruction: vertebral compression (%) and spinal canal occupancy (%).

Outcome definition

Neurological rehabilitation outcome was assessed at 6 months ± 2 weeks post-injury. Good recovery was defined as an improvement of ≥ 1 ASIA grade from baseline or maintenance of ASIA grade D/E at follow-up; poor recovery was defined as no improvement in ASIA grade or maintenance of ASIA grade A/B/C. This definition was supported by previous studies on spinal cord injury rehabilitation [3, 10]. Assessments were performed by two independent rehabilitation physicians who were blinded to patients’ serum Neuritin and TGF-β1 levels. Inter-rater reliability was evaluated using Cohen’s kappa coefficient, with a kappa value of 0.86 (95% CI: 0.78–0.94), indicating excellent agreement.

Statistical analysis

SPSS25.0 and R4.0.3 software were used for data analysis. Among the 189 included patients, missing data accounted for < 3% (2 cases with missing 25-OH-D values, 1 case with missing ALP value), and were handled using the multiple imputation method (5 imputed datasets). Measurement data were expressed as mean standard deviation. Independent sample t test was used for comparison between two groups, and analysis of variance was used for comparison among multiple groups. The count data were expressed as number of cases and percentage (n,%), and the comparison between groups was performed by χ² test. In the training set, univariate analysis was first performed to screen potential factors affecting neurological rehabilitation. Additionally, Least Absolute Shrinkage and Selection Operator (LASSO) penalized regression was performed using the “glmnet” package in R software to further confirm the predictor. 10-fold cross-validation was used to select the optimal lambda value (the penalty parameter that minimizes the cross-validation error), ensuring the model’s parsimony and reducing over-fitting. The consistency between the results of multivariate Logistic regression (odds ratios (OR) and 95% confidence intervals (CI) were calculated) and LASSO regression was used to determine the final independent factors for nomograph construction. Variance inflation factors (VIF) were calculated to exclude multicollinearity (VIF threshold < 10). The receiver operating characteristic (ROC) curve was used to evaluate the prediction performance of the model, and the Area under curve (AUC) and 95% CI were calculated. A calibration curve was used to evaluate the consistency of the model predictions with the actual observations.

In training set, there were 45 cases of poor neurological recovery (the “events” in EPV calculation), resulting in an events-per-variable (EPV) ratio of 7.5. Although this is slightly lower than the ideal EPV of 10, it is acceptable for exploratory prediction models in clinical research. Post-hoc power analysis, performed using G*Power 3.1 software with parameters set as n = 132 (training set), 6 predictors, and effect size f²=0.15 (medium effect size), showed a statistical power of 82.3% (> 80%), confirming adequate power to detect associations. A shrinkage estimate was also performed by applying the training set model to the validation set, resulting in a shrinkage factor of 0.98, which is close to 1. P < 0.05 was considered as the difference with statistical significance.

Results

Comparison of patient baseline data between training set and validation set

A total of 221 patients with osteoporotic spinal fractures combined with spinal cord injury were initially screened. After excluding 32 patients (12 with brain injury/peripheral neuropathy, 8 with malignant tumor history, 7 with severe hepatic/renal insufficiency, and 5 who declined participation), 189 patients were finally included. A STROBE flow-chart summarizing this process was provided in Supplementary Fig. 1.The training set (n = 132) and validation set (n = 57) showed no significant differences (P > 0.05) in age, gender, BMI, ASIA classification, fracture segment, concomitant diabetes, hypertension, alcohol consumption history, serum Neuritin, serum TGF-β1, imaging parameters (vertebral compression degree, and spinal canal space occupation rate), bone metabolism indexes (25-OH-D, and ALP) (all P > 0.05) (Table 1).

Table 1.

Comparison of patient baseline data between training and validation sets.

Indicators		Training set (n = 132)	Validation set (n = 57)	statistical values	P value
Age (years)		68.82 ± 6.98	69.12 ± 7.88	0.608	0.544
Gender	Male	72(54.55)	37(65.91)	1.753	0.186
	Female	60(45.45)	20(35.09)
BMI(kg/m²)		22.32 ± 2.96	22.55 ± 3.05	0.456	0.627
ASIA rating	Severe Injury Group (Grade A and B)	28(21.21)	17(29.82)	1.628	0.202
	Incomplete injury group (CDE grade)	104(78.79)	40(70.18)
Fracture segment	T12	86(65.15)	40(70.18)	0.452	0.501
	L1	46(34.85)	17(29.82)
Combined diabetes	Yes	24(18.18)	10(17.54)	0.011	0.917
	No	108(81.82)	47(82.46)
Hypertension	Yes	56(42.42)	20(35.09)	0.891	0.345
	No	76(57.58)	37(64.91)
Drinking history	Yes	28(21.21)	18(31.58)	2.323	0.127
	No	104(78.79)	39(68.42)
Postoperative complications	Yes	20(15.15)	7(12.28)	0.268	0.605
	No	112(84.85)	50(87.72)
Serum neuroitin (ng/ml)		26.86 ± 6.29	26.88 ± 5.88	0.021	0.984
Serum TGF-β1(ng/mL)		202.22 ± 38.42	205.78 ± 43.56	0.561	0.575
Degree of vertebral compression (%)		43.52 ± 10.83	44.82 ± 12.22	0.728	0.467
Occupancy rate of spinal canal (%)		36.28 ± 8.98	38.55 ± 10.01	1.54	0.125
Prealbumin (mg/dL)		17.58 ± 4.21	16.52 ± 3.98	1.615	0.108
Osteocalcin (ng/mL)		20.87 ± 4.86	22.22 ± 5.66	0.432	0.666
25-hydroxyvitamin D(ng/mL)		16.52 ± 6.01	15.44 ± 5.88	1.141	0.255
Alkaline phosphatase (U/L)		89.36 ± 22.32	92.44 ± 23.34	0.859	0.392

Univariate analysis of risk factors for neurological rehabilitation of patients with osteoporotic spinal fracture combined with spinal cord injury in training set

87 cases (65.91%) had good recovery after centralized training, and 45 cases (34.09%) had poor recovery. Univariate analysis showed that there were significant differences in such indicators as Neuritin, TGF-β1, vertebral compression degree, spinal canal space-occupying rate, 25-hydroxyvitamin D, and alkaline phosphatase between the good recovery group and the poor recovery group (all P < 0.05) (Table 2).

Table 2.

Univariate analysis of risk factors for neurological rehabilitation in patients with osteoporotic spinal fracture combined with spinal cord injury in training set.

Indicators		Good recovery group (n = 87)	Poor recovery group (n = 45)	statistical values	P value
Age (years)		67.82 ± 6.91	69.12 ± 7.56	0.992	0.323
Gender	Male	52(59.77)	20(44.44)	2.809	0.094
	Female	35(40.23)	25(55.56)
BMI(kg/m²)		22.12 ± 2.80	22.52 ± 3.16	0.744	0.458
ASIA rating	Severe Injury Group (Grade A and B)	18(20.69)	10(22.22)	0.359	0.548
	Incomplete injury group (CDE grade)	69(77.01)	35(77.78)
Fracture segment	T12	56(64.37)	30(66.67)	0.069	0.793
	L1	31(35.63)	15(33.33)
Combined diabetes	Yes	16(18.39)	8(17.78)	0.008	0.931
	No	71(81.61)	37(82.22)
Hypertension	Yes	38(43.68)	18(40.00)	0.164	0.685
	No	49(56.32)	27(60.00)
Drinking history	Yes	18(20.69)	10(22.22)	0.042	0.838
	No	69(79.31)	35(77.78)
Postoperative complications	Yes	12(13.79)	8(17.78)	0.366	0.545
	No	75(86.21)	37(82.22)
Serum neuroitin (ng/ml)		28.31 ± 6.59	25.53 ± 5.86	2.383	0.019
Serum TGF-β1(ng/mL)		195.21 ± 32.42	212.76 ± 45.68	2.553	0.012
Degree of vertebral compression (%)		41.12 ± 10.53	45.82 ± 12.31	2.293	0.024
Occupancy rate of spinal canal (%)		35.21 ± 8.71	38.93 ± 10.22	2.191	0.030
Prealbumin (mg/dL)		18.51 ± 4.51	16.99 ± 3.82	1.930	0.056
Osteocalcin (ng/mL)		19.88 ± 4.51	21.52 ± 5.88	1.781	0.077
25-hydroxyvitamin D (ng/mL)		18.52 ± 6.21	15.34 ± 5.82	2.848	0.005
Alkaline phosphatase (U/L)		85.36 ± 20.32	95.71 ± 25.62	2.533	0.013

Multifactor logistic regression analysis on neurological rehabilitation of patients with osteoporotic spinal fracture combined with spinal cord injury

Rehabilitation status was taken as the dependent variable (0 = poor, 1 = good), and the variable assignments were shown in Table 3. LASSO penalized regression with 10-fold cross-validation selected the optimal lambda value of 0.032, and the resulting selected predictors were consistent with the multivariate Logistic regression results: serum Neuritin, TGF-β1, vertebral compression degree, spinal canal space-occupying rate, 25-hydroxyvitamin D, and alkaline phosphatase. Thus, these 6 variables were confirmed as independent influencing factors for neurological rehabilitation (all P < 0.05), and no collinearity was observed among them (tolerance > 0.1, VIF < 10) (Table 4) In addition, the proportion of variances of multiple covariates without the same feature value was > 50%, indicating no collinearity of each covariate. The full logistic regression equation for predicting the probability of poor neurological rehabilitation was further established as: logit(p) = 12.872 + 0.250·Neuritin − 0.021·TGF-β1–0.041·Vertebral compression − 0.129·Spinal canal space-occupying rate + 0.098·25-hydroxyvitamin D − 0.028·Alkaline phosphatase (p represents the probability of poor neurological rehabilitation).

Table 3.

Variable assignment method.

Variable	Meaning	Evaluation
X1	Neuritin	Continuous variable
X5	TGF-β1	Continuous variable
X3	Degree of vertebral compression	Continuous variable
X4	Occupancy rate of spinal canal	Continuous variable
X5	25-hydroxyvitamin D	Continuous variable
X6	Alkaline phosphatase	Continuous variable
Y	Rehabilitation status	Good = 1, Bad = 0

Table 4.

The results of multivariate Logistic regression analysis.

Indicators	B	Standard error	Wald	P	OR	95% confidence interval
Neuritin	0.250	0.069	13.150	0.001	1.285	1.122–1.471
TGF-β1	-0.021	0.008	6.355	0.012	0.979	0.963–0.995
Degree of vertebral compression	-0.041	0.020	4.059	0.044	0.960	0.922–0.999
Occupancy rate of spinal canal	-0.129	0.042	9.541	0.002	0.879	0.810–0.954
25-hydroxyvitamin D	0.098	0.038	6.626	0.010	1.103	1.024–1.189
Alkaline phosphatase	-0.028	0.011	6.775	0.009	0.973	0.952–0.993

Construction of nomogram prediction model

Based on the independent influencing factors determined by multivariate Logistic regression analysis, an nomogram prediction model for the neurological rehabilitation effects of serum Neuritin and TGF-β1 expressions in patients with osteoporotic spinal fracture combined with spinal cord injury was established. Each independent influencing factor in the model was scored, and the total score for predicting clinical efficacy was calculated, expressed as the prediction probability (Fig. 1).

Fig. 1.

Nomogram prediction model of neurological rehabilitation for patients with osteoporotic spinal fracture combined with spinal cord injury.

Evaluation and validation of nomogram model

In the training set, C-index of the nomogram model was 0.842, and the calibration curve showed that the average absolute error of the predicted value and the actual value was 0.132, and Hosmer-Lemeshow test P = 0.030, indicating that the model fitted well. The ROC curve showed an AUC of 0.850 (95% CI: 0.75–0.95), a sensitivity of 0.850, and a specificity of 0.771. In the validation set, C-index was 0.821, mean absolute error was 0.129, Hosmer-Lemeshow test P = 0.651, AUC was 0.826 (95% CI: 0.65-1.00), sensitivity was 0.852, and specificity was 0.600. The calibration curve and ROC curve were shown in Figs. 2 and 3, respectively.

Fig. 2.

Calibration curves of nomogram prediction model (A: training set, B: validation set).

Fig. 3.

Receiver operating characteristic curves of nomogram prediction model (A: training set, B: validation set).

Decision curve analysis

Analysis of decision curve showed that when the threshold probability was within the range of 0.10–0.90, the nomogram model constructed in this study would have more clinical benefits in predicting the neurological rehabilitation decisions of patients with osteoporotic spinal fracture combined with spinal cord injury (Fig. 4).

Fig. 4.

Decision curves of nomogram prediction model. (A: training set, B: validation set)

Discussion

Osteoporotic spinal fractures with concomitant spinal cord injury present a complex clinical challenge, particularly in the aging population^9,10. The significant variability in neurological recovery following surgical intervention underscores the need for reliable prognostic tools. In this study, we developed and validated a nomogram model integrating serum biomarkers (Neuritin and TGF-β1), imaging parameters, and bone metabolism indicators to predict neurological recovery in this patient population. The model demonstrated good discriminatory ability and calibration in both training and validation sets (AUC: 0.850 and 0.826, respectively), outperforming traditional single‑parameter assessments.

Neuritin plays a key role in nerve repair after spinal cord injury^11,12. Studies have shown that local inflammation after spinal cord injury activates astrocytes and releases Neuritin to promote neuronal survival and axonal regeneration¹³. In the clinical data of this study, serum Neuritin levels in patients with good neurological recovery were significantly higher than those with poor recovery (P < 0.05), and their levels were positively correlated with the improvement degree of American Association for Spinal Cord Injury (ASIA) classification. Multivariate logistic regression analysis further confirmed that for every 1 ng/mL increase in serum Neuritin, the probability of neurological recovery increased by 28.5% (OR = 1.285, P < 0.05). This result not only verifies the value of Neuritin as an independent predictor, but also suggests that it is a core biomarker driving neurological function recovery in the cohort of this study — which is consistent with the high scoring weight of Neuritin in the nomogram model, and also provides a pathophysiological basis for clinically identifying patients with high recovery potential by detecting Neuritin levels at an early stage¹⁴.

TGF-β1 exerts a dual regulatory role in fracture healing and neuroinflammation, and this characteristic shows unique pathophysiological significance in patients with osteoporosis complicated with spinal cord injury. On the one hand, it accelerates bone repair by promoting osteoblast differentiation and collagen synthesis⁵; on the other hand, excessive TGF-β1 may exacerbate neuroinflammatory responses and inhibit axonal regeneration⁷. This study found that serum TGF-β1 levels in patients with osteoporotic spinal fracture combined with spinal cord injury were negatively correlated with neurological function recovery (OR = 0.979, 95% CI: 0.963–0.995, P < 0.05), meaning that for every 1 ng/mL decrease in TGF-β1 level, the probability of neurological recovery increased by 2.1%. This result seems to contradict the pro-fibrotic role of TGF-β1, but in fact reflects the complexity of the disease: in the special state of osteoporosis combined with spinal cord injury, the bone repair effect of TGF-β1 may dominate — moderate TGF-β1 can not only promote the repair of fractured vertebrae, but also provide a basis for nerve regeneration by regulating the local microenvironment, while its excessively low level will lead to delayed bone repair and imbalance of the nerve repair microenvironment, ultimately exacerbating poor recovery. In addition, the effect of TGF-β1 inducing excessive activation of astrocytes to form glial scars may only offset the benefits of bone repair when its level is significantly increased, which also explains the bidirectional role of TGF-β1 as an independent risk factor in this study¹⁵.

The degree of vertebral compression and spinal canal space-occupying rate are core imaging indicators for evaluating the severity of spinal cord injury. This study found that both are independent influencing factors for neurological recovery (vertebral compression degree OR = 0.960, spinal canal space-occupying rate OR = 0.879, both P < 0.05), which is consistent with the previous conclusion that “the difficulty of recovery increases significantly when vertebral compression > 30% or spinal canal space-occupying rate > 50%”^16,17. Notably, after integrating these two imaging indicators with serum biomarkers and bone metabolism indicators in this study, the predictive efficacy of the nomogram model was significantly improved (training set: AUC = 0.850; validation set: AUC = 0.826), which is much higher than the predictive efficacy of existing single imaging indicators (AUC mostly between 0.70 and 0.75)³. For example, in patients with type Ⅱ vertebral compression fractures, the improvement rate of postoperative ASIA grade was only 35% in patients with low serum Neuritin and high TGF-β1, while it reached 78% in patients with normal indicators (P < 0.05), which confirms the clinical value of the combination of biomarkers and imaging indicators^18,19. In addition, 25(OH)D and ALP, as key indicators of bone metabolism, also showed important predictive roles in this study: for every 1 ng/mL increase in 25(OH)D, the probability of neurological recovery increased by 10.3% (OR = 1.103, P < 0.05), while elevated ALP was associated with poor recovery (OR = 0.973, P < 0.05). This finding is consistent with the pathological mechanism that 25(OH)D deficiency leads to delayed fracture healing and imbalance of the nerve repair microenvironment^20–22, and also suggests that excessive elevation of ALP may offset the benefits of its osteogenic repair by aggravating local inflammatory responses.

Compared to existing prognostic tools that rely predominantly on clinical or imaging parameters, our multidimensional nomogram offers a more holistic assessment. It provides clinicians with a quantifiable method to identify high‑risk patients early—for instance, those with low Neuritin, low TGF‑β1, and high spinal canal occupancy—who may benefit from intensified rehabilitation or adjunctive therapies targeting neurotrophic support and bone metabolism.

There are some limitations in this study. First, the small sample size and the fact that they are all from a single center without external validation may affect the universality of the model. In addition, the differential expressions of Neuritin and TGF-β1 in different spinal cord injury segments (such as cervical spinal cord vs. thoracic spinal cord) have not been explored, and stratification analysis is required in the future. Future research can be expanded in the following directions: 1.expanding the sample size and conducting multi-center validation to optimize the prediction model; 2.The spatial-temporal expression patterns of Neuritin and TGF-β1 in the microenvironment of spinal cord injury were analyzed by single cell sequencing technology; 3. To explore the influence of gene polymorphisms (such as TGF-β1rs1800470) on the prediction model, so as to realize accurate medical treatment. Through interdisciplinary research, it is expected to establish a better prediction system, and promote the individualized process of spinal cord injury rehabilitation treatment.

In summary, this study constructed for the first time a nomogram prediction model based on serum Neuritin, TGF-β1 combined with imaging and bone metabolism indicators, which may help predict the neurological rehabilitation effects of patients with osteoporotic spinal fracture combined with spinal cord injury, and is the first quantitative prediction tool to realize the integration of multi-dimensional indicators in this field. This finding not only confirms the value of Neuritin and TGF-β1 as core biomarkers, but also provides new targets and methods for clinically identifying high-risk patients of neurological rehabilitation at an early stage and formulating individualized bone-nerve combined rehabilitation programs; in the future, through multi-center validation and technical iteration, this model is expected to promote the treatment of osteoporotic spinal fracture combined with spinal cord injury from empirical intervention to precise rehabilitation, and ultimately improve the long-term prognosis of patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Y.Z. was involved in the design of the study, collected data and drafted the paper. JG.Z. was involved in the design of the study, data analysis and drafting of the paper. B.Y. was involved with the design of the study, overseeing data analysis and drafting the paper. YP.T. and SH.H. was involved with designing the study, PI of the study and acts as the guarantor.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.