Predictive nomogram of cage nonunion after anterior cervical discectomy and fusion: A retrospective study in a spine surgery center.

The cage nonunion may cause serious consequences, including recurrent pain, radiculopathy, and kyphotic deformity. The risk factors for nonunion following anterior cervical discectomy and fusion (ACDF) are controversial. The aim of the study is to investigate the risk factors for nonunion in cervical spondylotic cases after ACDF. We enrolled 58 and 692 cases in the nonunion and union group respectively and followed up the cases at least 6 months. Patient demographic information, surgical details, cervical sagittal parameters, and the serum vitamin D level were collected. A logistic regression was performed to determine the independent predictors for nonunion, which were used for establishing a nomogram. In order to estimate the reliability and the net benefit of nomogram, we applied a receiver operating characteristic curve analysis, calibration curves and plotted decision curves. Using the multivariate logistic regression, we found that age (odds ratio [OR] = 1.16, P < .001), smoking (OR = 3.41, P = .001), angle of C2 to C7 (OR = 1.53, P < .001), number of operated levels (2 levels, OR = 0.42, P = .04; 3 levels, OR = 1.32, P = .54), and serum vitamin D (OR = 0.81, P < .001) were all significant predictors of nonunion (Table 3). The area under the curve of the model training cohort and validation cohort was 0.89 and 0.87, respectively. The calibration curves showed that the predicted outcome fitted well to the observed outcome in the training cohort (P = .102,) and validation cohort (P = .125). The decision curves showed the nomogram had more benefits than the All or None scheme if the threshold probability is >10% and <100% in training cohort and validation cohort. We found that age, smoking, angle of C2 to C7, number of operated levels, and serum vitamin D were all significant predictors of nonunion.

1. Introduction

Cervical spondylosis is a common lesion for the elderly, which is often caused by disc herniation. Anterior cervical discectomy and fusion (ACDF) was the most common therapy in the treatment of cervical spondylosis. The satisfactory neurological and radiological outcomes were compulsory for successful ACDF surgery.[ Additionally, radiological evidence of cage union is the essential condition for a good long-term prognosis. Although the cage nonunion is not very high, it may cause serious consequences, including recurrent pain, radiculopathy, and kyphotic deformity.[ Several studies showed that implant material, age, low bone mineral density, smoking status, low level of vitamin D and multilevel ACDF were risk factors for cage nonunion.[ However, its risk factors are controversial (Table S1, Supplemental Digital Content 1, http://links.lww.com/MD/H389).

The clinical prediction model, known as translational medicine, is an emerging method and process of facilitating medical advances efficiently from the scientist to the clinician.[ The risk factors for cage nonunion are controversial. For that purpose, we recommend construct a nomogram for predicting cage nonunion based the database in a spine surgery center.

The aim of the study is to investigate the risk factors for cage nonunion in cervical spondylotic cases after ACDF. In our study, we used logical regression to construct a nomogram for predicting cage nonunion based the database in a spine surgery center.

2. Methods

2.1. Patient

From January 1, 2016 till October 31, 2020 in the Department of Osteology, Bin Zhou People’s Hospital, we enrolled a total of consecutive 857 patients with cervical spondylotic disease who matched the following criteria. The study was approved by the ethical committee of Bin Zhou People’s Hospital. All subjects were followed up at least 6 months.

The inclusion criteria: 18 < ages; cervical spondylosis diagnosed by magnetic resonance imaging; clinical symptoms are consistent with the imaging; cases were underwent ACDF.

The exclusion criteria: subjects with acute spinal cord injury or inflammation; poor general physical state and other severe primary diseases; inability to give informed consent; patients with unknown information or loss to follow-up (Fig. 1).

Figure 1.

Flow chart for cervical spondylosis patients in training and validation cohorts.

2.2. Treatment

The ACDF technique was performed by the senior author with a standard cervical anterior approach. With general anesthesia, the patient was placed in a supine position with the neck slightly extended. We made a transverse anterior cervical incision and cut the platysma sharply. The disc was removed with angled curettes and pituitary rongeurs of various sizes. We prefilled the bone substitute into an IMPIX C + polyetheretherketone cervical spacer and placed it in the intervertebral space. The plate and screws (Medtronic, America) was used to secure the inter-somatic spacer (Fig. 2).

Figure 2.

The pre and post operation images. Before operation, the MRI showed the cervical disc herniation in C4/5 (A). The X-rays showed the plate and screws post operation (B). During the operation, the disc was removed with angled curettes and pituitary rongeurs of various sizes (C). MRI = magnetic resonance imaging.

2.3. Image assessment

Before ACDF, all patients had X-rays of the posteroanterior, lateral, and dynamic lateral over flexional and over extensive views. The radiological variables, including the angle from C2-C7, C2-C7 ROM, C2-C7 sagittal vertical axis, T1 slope, were taken into account.

2.4. Outcome assessment

We used a visual analogue scale (VAS) to assess the pain of patients. The patient’ functional status was assessed using the Neck Disability Index (NDI) questionnaire and modified Japanese Orthopedic Association (mJOA) scale.

2.5. Data collection

Patient demographic information, including age, sex, body mass index (BMI), history of smoking, alcohol, hypertension, diabetes, and osteoporosis were recorded. Surgical details including number of operated levels, surgical duration and blood loss were noted. The serum vitamin D levels were assessed in all patients before surgery.

Normal ranges for vitamin D in our clinical laboratory were 30 to 100 ng/mL.

2.6. Assessment of fusion

At the 6-month follow-up, CT was performed to evaluate fusion. We defined radiological fusion if we find the presence of bone trabeculation, the absence of indications of instrumentation loosening or breaking. The patients were separated into 2 groups based on the CT results: nonunion and union group.

2.7. Statistical analysis

Categorical and continuous data were reported as percentages and mean standard deviations, respectively. To compare groups, we employed the t test or the Mann–Whitney U test for categorical data, and the chi-square test or Fisher exact tests for continuous variables. To develop a well-reliable nomogram model predicting the risk of nonunion, our nomogram was built and validated using the training and validation cohort at 7:3 ratio (Fig. 1). The variance inflation factors and tolerance were used for analyzing multicollinearity between variables. We utilized logistic regression procedure to analyze the risk factors for nonunion. The plotted decision curves, calibration curves and receiver operating characteristic curve were used to verify the reliability of nomogram. If the P value was <.05, we viewed statistically significant. Programs R (The R Foundation; http://www.r-project.org; version 3.4.3) was used to analyze the data.

3. Result

3.1. Demography

Finally, we enrolled 750 patients in this study, 58 in the nonunion group, 692 in the union group. In terms of sex, BMI, history of alcohol, hypertension and diabetes, VAS, NDI, mJOA, C2–C7 ROM, C2–C7 SVA, T1 slope, surgical duration and blood loss, there were no significant differences between the 2 groups (P > .05). The mean age in the nonunion group was 62.9 ± 7.0 years, which was significantly higher than that in the union group (P < .001). There were 39 (67.2%) cases with smoking history and 34 (58.6%) cases with osteoporosis in the nonunion group, which was higher significantly than that in union group (P < .05). The mean angle of C2 to C7 in the nonunion group was 18.2° ± 2.3°, which was significantly higher than that in the union group (P < .001). The mean serum vitamin D in the nonunion group was 21.3 ± 6.8 ng/mL, which was significantly lower than that in the union group (P < .001). In the nonunion group, the percentage of 2 and 3 operated levels was higher than that in the union group (P < .001) (Table 1).

Table 1

Clinical summary in the nonunion and union group.

Items	Nonunion group(n = 58)	Union group(n = 692)	P
Age (yr)	62.9 ± 7.0	57.2 ± 5.7	<.001
Sex (male, %)	28 (48.3%)	330 (47.7%)	.93
BMI (kg/m2)	24.0 ± 1.9	23.9 ± 2.4	.91
Smoking (n, %)	39 (67.2%)	245 (35.4%)	<.001
Alcohol (n, %)	20 (34.5%)	228 (32.9%)	.81
Hypertension (n, %)	22 (37.9%)	226 (32.7%)	.41
Diabetes (n, %)	12 (20.7%)	129 (18.6%)	.70
Osteoporosis (n, %)	34 (58.6%)	252 (36.4%)	.001
VAS	5.3 ± 1.6	4.9 ± 1.8	.13
NDI	32.6 ± 9.4	30.6 ± 10.9	.18
mJOA	11.9 ± 3.1	11.2 ± 3.7	.12
Angle of C2 to C7 (°)	18.2 ± 2.3	15.9 ± 2.2	<.001
C2–C7 ROM (°)	31.6 ± 6.7	31.7 ± 6.9	.98
C2–C7 SVA (mm)	17.9 ± 3.1	17.3 ± 3.7	.13
T1 slope (°)	22.4 ± 2.9	22.7 ± 3.5	.47
Number of operated levels			<.001
1	17 (29.3%)	406 (58.7%)
2	21 (36.2%)	150 (21.7%)
3	20 (34.5%)	136 (19.7%)
Surgical duration (min)	88.3 ± 8.6	87.5 ± 9.9	.57
Blood loss (mL)	80.2 ± 7.3	79.2 ± 7.8	.30
Serum vitamin D (ng/mL)	21.3 ± 6.8	27.8 ± 4.9	<.001

Angle of C2 to C7 = the angle formed by the inferior endplates of C2 and C7 in lateral radiographs, BMI = body mass index, C2–C7 ROM = the sum of the C2–7 Cobb angle during flexion and extension lateral radiographs, C2–C7 SVA = distance from the posterosuperior corner of C7 and the vertical line from the center of the C2 body, mJOA = modified Japanese Orthopedic Association, NDI = neck disability index, T1 slope = the angle between a horizontal line and the superior endplate of T1 on lateral radiograph, VAS = visual analogue scale.

We enrolled 525 and 225 cases into the training cohort and the validation cohort. There was no difference in various indicators between the 2 cohorts (P > .05, Table 2).

Table 2

Characteristics of patients in the training and validation cohorts.

Items	Training cohort(N = 525)	Validation cohort(N = 225)	P
Age (yr)	57.8 ± 5.9	57.4 ± 6.3	.44
Sex (male, %)	246 (46.9%)	112 (49.8%)	.46
BMI (kg/m2)	24.1 ± 2.3	23.8 ± 2.5	.21
Smoking (n, %)	191 (36.4%)	93 (41.3%)	.20
Alcohol (n, %)	185 (35.2%)	63 (28.0%)	.07
Hypertension (n, %)	169 (32.2%)	79 (35.1%)	.44
Diabetes (n, %)	100 (19.0%)	41 (18.2%)	.79
Osteoporosis (n, %)	203 (38.7%)	83 (36.9%)	.65
VAS	5.0 ± 1.8	4.9 ± 1.9	.79
NDI	30.8 ± 10.6	30.6 ± 11.4	.78
mJOA	11.3 ± 3.5	11.2 ± 3.8	.76
Angle of C2 to C7 (°)	16.0 ± 2.3	16.3 ± 2.5	.12
C2–C7 ROM (°)	31.5 ± 6.7	32.2 ± 7.0	.19
C2–C7 SVA (mm)	17.4 ± 3.5	17.3 ± 3.8	.78
T1 slope (°)	22.8 ± 3.4	22.3 ± 3.4	.11
Number of levels operated			.71
1	291 (55.4%)	132 (58.7%)
2	122 (23.2%)	49 (21.8%)
3	112 (21.3%)	44 (19.6%)
Surgical duration (min)	87.8 ± 9.8	87.1 ± 9.8	.34
Blood loss (mL)	79.3 ± 8.0	79.1 ± 7.5	.75
Serum vitamin D (ng/mL)	27.3 ± 5.2	27.1 ± 5.5	.50
Outcome			.44
Nonunion	487 (92.8%)	205 (91.1%)
Union	38 (7.2%)	20 (8.9%)

BMI = body mass index, mJOA = modified Japanese Orthopedic Association, NDI = neck disability index, ROM = range of motion, SVA = sagittal vertical axis, VAS = visual analogue scale.

3.2. Nomogram construction

We used the univariate logistic regression to analyze the association between sex, age, BMI, history of smoking, alcohol, hypertension, osteoporosis and diabetes, VAS, NDI, mJOA, angle of C2 to C7, C2–C7 ROM, C2–C7 SVA, T1 slope, number of operated levels, surgical duration and blood loss, serum vitamin D and nonunion. We found that age (odds ratio [OR] = 1.17, P < .001), smoking (OR = 3.75, P < .001), osteoporosis (OR = 0.40, P = .001), angle of C2 to C7 (OR = 1.55, P < .001), number of operated levels (2 levels, OR = 0.29, P < .001; 3 levels, OR = 0.95, P = .88), and serum vitamin D (OR = 0.79, P < .001) were all significant predictors of nonunion (Table 3).

Table 3

Univariate and multivariate logistic regression model for predicting nonunion after ACDF.

Variables	Univariate analysis		Multivariate analysis
	OR (95% CI)	P value	OR (95% CI)	P value
Age (yr)	1.17 (1.12, 1.23)	<.001	1.16 (1.08, 1.24)	<.001
Sex				.44
Female	1		1
Male	0.98 (0.57, 1.67)	.93	1.32 (0.65, 2.68)
BMI (kg/m2)	1.01 (0.89, 1.13)	.91	1.01 (0.78, 1.31)	.94
Smoking		<.001		.001
No	1		1
Yes	3.75 (2.12, 6.62)		3.41 (1.64, 7.10)
Alcohol (n, %)		.81		.44
No	1		1
Yes	0.93 (0.53, 1.64)		1.33 (0.64, 2.76)
Hypertension (n, %)		.41		.81
No	1		1
Yes	0.79 (0.46, 1.38)		1.09 (0.53, 2.27)
Diabetes (n, %)		.70		.64
No	1		1
Yes	0.88 (0.45, 1.71)		1.23 (0.52, 2.92)
Osteoporosis (n, %)		.001		.09
No	1		1
Yes	0.40 (0.23, 0.69)		1.81 (0.90, 3.64)
VAS	1.11 (0.85, 1.28)	.19	1.18 (0.96, 1.45)	.12
NDI	1.02 (0.99, 1.04)	.17	1.01 (0.92, 1.05)	.18
mJOA	1.05 (0.98, 1.13)	.18	1.02 (0.97, 1.12)	.16
Angle of C2 to C7 (°)	1.55 (1.36, 1.77)	<.001	1.53 (1.28, 1.83)	<.001
C2–C7 ROM (°)	0.99 (0.96, 1.04)	.98	0.98 (0.94, 1.05)	.95
C2–C7 SVA (mm)	1.03 (0.97, 1.14)	.16	1.05 (0.98, 1.15)	.17
T1 slope (°)	0.97 (0.89, 1.05)	.47	0.94 (0.81, 1.09)	.46
Number of operated levels
1	1		1
2	0.29 (0.15, 0.56)	<.001	0.42 (0.18, 0.98)	.04
3	0.95 (0.49, 1.83)	.88	1.32 (0.54, 3.22)	.54
Surgical duration (min)	1.01 (0.98, 1.04)	.57	1.02 (0.98, 1.07)	.28
Blood loss (mL)	1.02 (0.98, 1.05)	.33	1.03 (0.97, 1.08)	.25
Serum vitamin D (ng/mL)	0.79 (0.75, 0.84)	<.001	0.81 (0.75, 0.87)	<.001

ACDF = anterior cervical discectomy and fusion, BMI = body mass index, CI = confidence interval, mJOA = modified Japanese Orthopedic Association, NDI = neck disability index, OR = odds ratio, ROM = range of motion, SVA = sagittal vertical axis, VAS = visual analogue scale.

Using the multivariate logistic regression, we found that age (OR = 1.16, P < .001), smoking (OR = 3.41, P = .001), angle of C2 to C7 (OR = 1.53, P < .001), number of operated levels (2 levels, OR = 0.42, P = .04; 3 levels, OR = 1.32, P = .54), and serum vitamin D (OR = 0.81, P < .001) were all significant predictors of nonunion (Table 3). Additionally, we found the tolerance was greater than 0.1 and variance inflation factors were <10 for the predictors, which indicated there was no collinearity among these variables (Table S2, Supplemental Digital Content 2, http://links.lww.com/MD/H390). We established the nomogram for predicting nonunion on the basis of the above risk factors (Fig. 3).

Figure 3.

The nomogram to predict the nonunion of in patients after ACDF. Based on the risk factors selected, we developed a nomogram to predict the probability of nonunion based on the logistic model. ACDF = anterior cervical discectomy and fusion.

3.3. Nomogram validation

The area under the curve of the model training cohort and validation cohort were 0.89 (Fig. 4A) and 0.87 (Fig. 4B), respectively, which indicated favorable discrimination. The calibration curves confirmed that the observed outcome fitted nicely to the predicted outcome in the training cohort (P = .102, Fig. 4C) and validation cohort (P = .125, Fig. 4D). The decision curve showed that if the threshold probabilities in the training cohort and validation cohort were >10% and <100%, the nomogram had more advantages than the All or None scheme (Fig. 4E and F). The area under the curve, accuracy, specificity, sensitivity, PLR, NLR, DOR were 0.89, 0.84, 0.74, 0.93, 3.58, 0.09, 39.78, respectively (Table 4).

Figure 4.

Nomogram validation. The AUC of the model training cohort and validation cohort were 0.89 (A) and 0.87 (B) respectively, which indicated favorable discrimination. The calibration curves showed that the predicted outcome fitted well to the observed outcome in the training cohort (P = .102, C) and validation cohort (P = .125, D). The decision curves showed the nomogram had more benefits than the All or None scheme if the threshold probability is >10% and <100% in training cohort and validation cohort (E and F). AUC = area under the curve.

Table 4

Performance of the nomogram in predicting nonunion after ACDF.

Performance parameter	AUC	Accuracy	Specificity	Sensitivity	PLR	NLR	DOR
Nomogram	0.89	0.84	0.74	0.93	3.58	0.09	39.78

ACDF = anterior cervical discectomy and fusion, AUC = area under the curve, DOR = diagnostic odds ratio, NLR = negative likelihood ratio, PLR = positive likelihood ratio.

4. Discussion

ACDF is a most appropriate choice to deal with cervical spondylosis, which can promote immediate stability and relief of symptoms. Previous researchers showed that the proportion of nonunion in cages after ACDF was <10%.[ Nonunion after spinal fusion procedures can result in recurrent pain, radiculopathy, and kyphotic deformity. However, its risk factors are controversial.[ There was little article regarding comprehensive risk factors, including demographic information, surgical details, sagittal parameters, and serum vitamin D level, for cage nonunion after ACDF. The aim of the study is to investigate the risk factors for cage nonunion in cervical spondylotic cases after ACDF and construct a nomogram based the database in a spine surgery center. In the study, there were 7.7% cases with cage nonunion, which is similar to previous studies. We demonstrated that age, smoking, angle of C2 to C7, number of operated levels, and serum vitamin D were all significant predictors of nonunion.

Several studies have shown that older age and a history of smoking are risk factors for cage nonunion with ACDF.[ Our data is consistent with previous studies which showed the average age of the nonunion group was 62.9 years and the age of the union group was 57.2 years. We also found that cases with the history of smoking had significantly higher rate of nonunion comparing with no smoking (OR = 3.41). Smoking is believed to contribute to osteoporosis with the destruction of osteoblast.[ Furthermore, no difference in BMI and osteoporosis was found between 2 groups in our study, which is opposite to the findings in previous studies. An increase in the number of operated levels has also been found to be associated with nonunion.[ Yang confirmed that cases with 2-level ACDF were with higher rate of cage nonunion compared with 1 level.[ However, Bao and Van found no statistical difference between 1-level and 2-level ACDF.[ In our study, we found that the rate of nonunion in cases with 3-level ACDF was 1.32-fold compared with 1-level.

Recently, several studies reported the cervical sagittal parameters, including angle of C2-C7, T1 slope, C2 - C7 ROM and C2-C7 SVA, in predicting clinical and functional outcomes after ACDF.[ Roguski et al conducted a study to investigate the role of cervical sagittal balance on outcome following spinal surgery and demonstrated that it can predict outcomes.[ Bao et al discovered that the angle of C2–C7 and T1 slope was all risk factors for nonunion of cage following ACDF.[ In our study, we found that the angle of C2 to C7 (OR = 1.53, P < .001) was a significant predictor of nonunion. While, there was no significant difference between nonunion and union group in C2–C7 ROM, T1 slope and C2–C7 SVA. Patwardhan et al demonstrated that the strain increases with increasing joint angle of the 2 bones of the joint.[ As cases with excessive C2–C7 angle, it could cause a massive quantity of shear pressure paralleled to the intervertebral space. It is nicely recognized that higher shear pressure is not really helpful to cage union.

As previous studies showed that vitamin D stability may have an effect on the metabolic milieu for bone fusion. Vijay et al found that 30% of patients undergoing elective spinal fusion had vitamin D deficiency, and 38.9% had insufficient levels of vitamin D.[ It was proven that the vitamin D contributed to the consolidation of bone in a rat model, in which there was a direct relationship between levels of vitamin D and bone fusion. Vijay et al determined that lower vitamin D level played an important role in a lower rate of and a longer time to fusion.[ In our study, we also found that serum vitamin D (OR = 0.79, P < .001) was significant predictor of nonunion. These findings indicate it is necessary to proceed a prospective study to analyze the role of vitamin D supplementation on spinal fusion.

5. Conclusion

We found that age, smoking, angle of C2 to C7, number of operated levels, and serum vitamin D were all significant predictors of nonunion.

Author contributions

Conceptualization: Changliang Liu.

Data curation: Kai Zhou.

Formal analysis: Kai Zhou.

Investigation: Longfei Ji.

Methodology: Longfei Ji.

Project administration: Longfei Ji.

Resources: Shuwei Pang.

Software: Shuwei Pang.

Supervision: Shuwei Pang.

Writing – original draft: You Tang.

Writing – review & editing: Changliang Liu.