Analysis of correlation and difference between degree of bone healing and observation index
The ICC of the average CT value and the median wall thickness measured by the two doctors was 0.81 and 0.86, respectively, showing high repeatability. We carried out correlation analysis and one-way ANOVA of the degree of bone healing and R2, R4, and R5. R2, R4, and R5 did not follow a normal distribution. Correlation analysis showed a correlation coefficient of 0.060 between the degree of bone healing and R4 (p=0.529), indicating no correlation. The correlation coefficient of the degree of healing with R2 and R5 was 0.654 (P<0.001) and 0.542 (P<0.001), respectively. The median R2 in groups A, B, and C was 0.907, 0.799, and 0.667, respectively, with a statistically significant difference among them (p<0.05). Pairwise comparison showed that the median R2 in group A was significantly greater than that in groups B and C and that the median R2 in group B was significantly greater than that in group C (Fig. 2).
The median R4 in groups A, B, and C was 0.933, 0.932, and 0.932, respectively. There was no significant difference among the three groups (Fig. 3).
The median R5 in groups A, B, and C was 0.850, 0.735, and 0.568, respectively, with a statistically significant difference among them (p<0.05). Pairwise comparison showed that the median R5 in group A was significantly greater than that in groups B and C and that the median R5 in group B was significantly greater than that in group C (Fig. 4).
ROC curve analysis results
To clarify the diagnostic efficacy of R2 and R5 in determining the degree of healing and the critical diagnostic value, the patients were first divided into the bone healing group (group A) and the nonunion group (groups B and C). The ordinate represents the true positive (sensitivity), and the abscissa represents the false positive (1-specificity). The ROC curve for predicting fracture healing was drawn with different R2 and R5 values as diagnostic cutoff points. The area under the curve of R2 and R5 was 0.858 (95% CI: 0.788~0.929) (p<0.001) and 0.784 (95% CI: 0.700-0.868) (p<0.001), respectively (Fig. 5). R2 was more effective in diagnosing bone healing than R5. The best diagnostic node was then selected according to the maximum Youden index method. The best value of R2 for the diagnosis of bone healing was 0.836 (sensitivity, 0.796, specificity, 0.841), and the maximum Youden index was 0.637. The best value of R5 for the diagnosis of bone healing was 0.796 (sensitivity, 0.653, specificity, 0.175), and the maximum Youden index was 0.478.
We then divided all patients into the nonunion group (group C) and the bone healing group (groups A and B) for ROC curve analysis to determine the diagnostic efficacy of R2 and R5 for nonunion. The area under the curve of R2 and R5 was 0.831 (95% CI: 0.727~0.934) (p<0.001) and 0.781 (95% CI: 0.671-0.892) (p<0.001), respectively (Fig. 6). R2 was more effective in diagnosing bone nonunion than R5. The best diagnostic value of R2 for bone nonunion was 0.743 (sensitivity, 0.849, specificity, 0.308), and the maximum Youden index was 0.541. The best diagnostic value of R5 for bone nonunion was 0.582 (sensitivity, 0.953, specificity, 0.423), and the maximum Youden index was 0.530.
Typical case reports
Case 1: A patient with poor bone healing achieved bone healing after periodic observation.
A 25-year-old male patient with a right femoral shaft fracture underwent intramedullary nail fixation. X-ray and CT examinations at 6 months after surgery showed poor bone healing. X-ray and CT examinations at 9, 15, and 23 months after surgery all indicated poor bone healing, but the differential ratios R2 and R4 continued to increase (Fig. 8). The imaging examination at 35 months postoperatively indicated bone healing, and the internal fixation was removed. No complications were observed 1 year after the operation. The follow-up imaging data are shown in Fig. 7.
The CT data at each follow-up after the operation were analyzed, and the corresponding R2 values were obtained. As the healing process progressed, R2 gradually increased (table 2).
Table. 2 R2, R4 and R5 data at different time points in case 1
Postoperative time (months)
|
6
|
9
|
15
|
23
|
35
|
R2
|
0.57
|
0.66
|
0.75
|
0.86
|
0.91
|
R4
|
0.96
|
0.89
|
0.98
|
0.94
|
0.93
|
R5
|
0.55
|
0.59
|
0.74
|
0.81
|
0.85
|
Case 2: A patient with poor bone healing continued to be observed, and no growth was observed, but bone healing was achieved after bone graft surgery.
A 37-year-old male patient with a right femoral shaft fracture underwent open reduction and internal fixation. Imaging examinations at 6, 7, and 10 months (Fig. 9) after surgery all showed that no obvious increasing trend in R2. Combined with the CT findings, bone nonunion was considered. Iliac bone grafting was performed at 10 months after the fixation operation. At 17 months (7 months after bone grafting) and 19 months (9 months after bone grafting) (Fig. 9) after the fixation surgery, bone healing was considered by physical and imaging examinations.
R2 still did not reach the healing standard at 7 months after internal fixation, and R2 did not change much during the observation period. However, R2 gradually increased after the bone graft surgery at 10 months after the fixation surgery (table 3, Fig. 10).
Table. 3 R2, R4, and R5 at different time points in case 2
Postoperative time (months)
|
6
|
7
|
10
|
17
|
19
|
R2
|
0.64
|
0.73
|
0.74
|
0.83
|
0.92
|
R4
|
0.96
|
0.93
|
0.95
|
0.89
|
0.94
|
R5
|
0.61
|
0.68
|
0.70
|
0.74
|
0.86
|