A Nomogram to Predict Intra-Spinal Canal Cement Leakage Among Elderly Patients with Spine Metastases: An Internal-Validated Model

Figures & data

Table 1 Patient’s Demographic and Clinical Characteristics

Characteristics	Patients (n=163)
Age (years)
≥65 and <70	28.22% (46/163)
≥70 and <80	47.24% (77/163)
≥80	24.54% (40/163)
Sex
Male	60.74% (99/163)
Female	39.26% (64/163)
Primary cancer types
Slow growth	32.52% (53/163)
Moderate growth	12.88% (21/163)
Rapid growth	54.60% (89/163)
Preoperative treatments
Topical treatments	35.58% (58/163)
Systematic treatments	33.74% (55/163)
No treatment	30.67% (50/163)
Number of treated vertebrae levels
1	40.49% (66/163)
2	22.09% (36/163)
3	16.56% (27/163)
≥4	20.86% (34/163)
Vertebrae collapse
No collapse	58.28% (95/163)
Less than 50%	26.38% (43/163)
More than 50%	15.34% (25/163)
Cortical osteolytic destruction in posterior wall
Yes	37.42% (61/163)
No	62.58% (102/163)
Vertebral endplate fracture
Yes	17.18% (28/163)
No	82.82% (135/163)
The Bilsky scale
0	80.37% (131/163)
1	10.43% (17/163)
2	7.98% (13/163)
3	1.23% (2/163)
Appearance of spine metastases
Mixed lesions	25.15% (41/163)
Osteolytic lesions	74.85% (122/163)
Load-bearing lines of spine
Normal	84.66% (138/163)
Abnormal	15.34% (25/163)
Intra-spinal canal cement leakage
Yes	9.82% (16/163)
No	90.18% (147/163)

Figure 1 An 80-year-old man with spine metastases treated with PVP and suffered from intra-spinal canal cement leakage (before surgery). (A) Preoperative anteroposterior thoracic vertebral radiography showed vertebral collapse at T10 and L1; (B) Preoperative anteroposterior lumbar vertebral radiography revealed vertebral collapse at L1; (C) Preoperative lateral thoracic vertebral radiography showed vertebral collapse at T10 and L1; (D). Preoperative lateral lumbar vertebral radiography showed vertebral collapse at L1; (E) Preoperative sagittal thoracic vertebral MRI showed multiple spine metastases and vertebral collapse at T10 and L1; (F) Preoperative sagittal lumbar vertebral MRI showed multiple spine metastases and vertebral collapse at T10 and L1; (G) Preoperative transversal T10 MRI showed deformation of the dural sac, but no spinal cord or nerve root compression; (H) Preoperative transversal MRI at T11 showed metastatic lesion; (I) Preoperative transversal MRI at T12 showed metastatic lesion; (J) Preoperative transversal MRI at L1 showed metastatic lesion; (K) Preoperative transversal MRI at L2 showed metastatic lesion; (L) Preoperative transversal MRI at L5 showed metastatic lesion; (M) Postoperative anteroposterior lumbar vertebral radiography revealed bone cement at T10, T11, T12, L1, L2, and L5; (N) Postoperative lateral vertebral radiography showed bone cement at T10, T11, T12, L1, L2, and L5; (O) Postoperative transversal CT at T10 showed distribution of bone cement; (P) Postoperative transversal CT at T11 showed intra-spinal canal cement leakage; (Q) Postoperative transversal CT at T12 showed intra-spinal canal cement leakage; (R) Postoperative transversal CT at L1 showed intra-spinal canal cement leakage.

Table 2 Multivariate Logistic Regression Analyses of Potential Risk Variables for Predicting Intra-Spinal Canal Cement Leakage in Patients with Spine Metastases Treated with Percutaneous Vertebroplasty in the Training Group

Potential Risk Variables	Rates	Univariate Logistic Regression		Multivariate Logistic Regression
		OR (95% CI)	P	OR (95% CI)	P
Age (years)
≥65 and <70	10.00% (3/30)	0.66 (0.24–1.79)	0.41	0.60 (0.17–2.17)	0.43
≥70 and <80	9.09% (4/44)
≥80	3.85% (1/26)
Sex
Male	4.55% (3/66)	3.62 (0.81–16.19)	0.09	9.43 (1.00–89.69)	0.05
Female	14.71% (5/34)
Primary cancer types
Slow growth	8.82% (3/34)	1.11 (0.50–2.51)	0.79	1.45 (0.29–7.16)	0.65
Moderate growth	0.00% (0/15)
Rapid growth	9.80% (5/51)
Preoperative treatments
Topical treatments	5.56% (2/36)	1.15 (0.46–2.83)	0.77	1.55 (0.47–5.10)	0.47
Systematic treatments	11.11% (4/36)
No treatment	7.14% (2/28)
Number of treated vertebrae levels
1	10.26% (4/39)	0.67 (0.33–1.38)	0.28	0.54 (0.23–1.30)	0.17
2	12.50% (3/24)
3	0.00% (0/16)
≥4	4.76% (1/21)
Vertebrae collapse
No collapse	10.00% (6/60)	0.69 (0.22–2.18)	0.53	0.28 (0.03–3.23)	0.31
Less than 50%	3.70% (1/27)
More than 50%	7.69% (1/13)
Cortical osteolytic destruction in posterior wall
Yes	15.79% (6/38)	5.63 (1.07–29.49)	0.04	12.35 (1.35–113.36)	0.03
No	3.23% (2/62)
Vertebral endplate fracture
Yes	5.88% (1/17)	0.68 (0.08–5.91)	0.73	1.71 (0.04–79.45)	0.78
No	8.43% (7/83)
The Bilsky scale
0	8.33% (7/84)	1.10 (0.33–3.72)	0.87	0.74 (0.11–4.99)	0.76
1	0.00% (0/11)
2	25.00% (1/4)
3	0.00% (0/1)
Appearance of spine metastases
Mixed lesions	6.90% (2/29)	1.25 (0.24–6.57)	0.80	0.12 (0.00–4.55)	0.25
Osteolytic lesions	8.45% (6/71)
Load-bearing lines of spine
Normal	6.82% (6/88)	2.73 (0.49–15.42)	0.25	20.91 (1.03–423.34)	0.05
Abnormal	16.67% (2/12)

Abbreviations: OR, odds ratio; CI, confidential interval.

Table 3 The Nomogram to Calculate the Probability of Intra-Spinal Canal Cement Leaks in Patients with Spine Metastases

Included Variables	Scores	Estimates^a
Intercept		−5.53
Sex
Male	1	1.27
Female	2
Cortical osteolytic destruction in posterior wall
Yes	1	1.80
No	0
Load-bearing lines of spine
Normal	0	1.43
Abnormal	1

Notes: ^aIndicates the estimates were obtained from the multivariate logistic analysis of the three significant variables. A formula was developed as follows: P(Y=1) = /(1+); x1 indicates gender; x2 indicates cortical osteolytic destruction in posterior wall; x3 indicates load-bearing lines of spine. P(Y=1) indicates the predicted probability of occurrence of intra-spinal canal cement leakage. An example was shown as follows: If a female (x1=2) patients with cortical osteolytic destruction in posterior wall (x2=1) and normal load-bearing lines of spine (x3=0) treated with percutaneous vertebroplasty, then the predicted probability of occurrence of intra-spinal canal cement leakage was P(Y=1) = /(1+) = 23.33%.

Table 4 Predictive Performance of the Nomogram

Formula	Discrimination Ability			Calibration Ability
	AUROC	Slope	95% CI	Chi-Square Value	P^a
Training group	0.75	0.16	0.09–0.22	1.13	0.77
Validation group	0.64	0.07	−0.01–0.15	0.56	0.97
Both groups	0.69	0.12	0.07–0.17	2.12	0.71

Notes: ^aIndicates P value was obtained from the goodness-of-fit test.

Abbreviations: AUROC, area under the receiver operating characteristic curve; CI, confidential interval.

Figure 2 The area under the receiver operating characteristic (AUROC) curve for the nomogram: (A) the training group; (B) the validation group; (C) the both group.

Figure 3 Box plots of predicted probabilities for the nomogram: (A) the training group; (B) the validation group; (C) the both group combined. Discrimination slope was the difference between the mean predicted probability with (positive events) and without (negative events) intra-spinal canal cement leakage. Black solid dots indicate the mean value of the predicted probability of intra-spinal canal cement leakage.

Figure 4 Plotting deciles of the predicted probability of intra-spinal canal cement leakage against the observed proportions for the nomogram: (A) the training group; (B) the validation group; (C) the both groups combined. The x-axis is the predicted risk and the y-axis is the actual risk. The grey dotted lines indicate a perfect prediction by an ideal model. The solid black lines indicate the performance of the nomogram, and a closer fit to the diagonal dotted lines indicate a better prediction.

Table 5 Subgroup Analysis of Patients with Spine Metastases According to the Predicted Probability of Intra-Spinal Canal Cement Leaks

Groups	Patients (n=163)	Predicted Probability	Actual Probability	P^a
Low risk group	128	0.00–9.99%	7.03%	˂0.01
Medium risk group	26	10.00–24.99%	11.54%
High risk group	9	≥25.00%	44.44%

Notes: ^aIndicates P was obtained from the chi-square test for actual probability.