General data and baseline characteristics
In Table 1, we conclude general data of case and control group. Difference of weight, tobacco smoking, alcohol intake, N-MID (ng/mL), T-score were not statistically significant (56 vs 57, p = 0.093; 3 vs 2, p = 0.767; 8 vs 5, p = 0.548; 12.79 vs 12.37, p = 0.649; -1.9 vs -1.9, p = 0.434; -1.6 vs -1.5, p = 0.877; -1.6 vs -1.7, p = 0.395), And individuals in the fracture group were older and taller than those in control group (70 vs 72, p = 0.032) (154.24 vs 162.49, p < 0.01), a higher prevalence of hypertension(18 vs 35, p < 0.01) and Diabetes(18 vs 29, p = 0.02) was observed among patients who experienced fractures after a fall, In terms of BTMs, the concentration of β-CTX in case group was higher than that in control group (275.9 vs 383.3; p < 0.01), while concentration of 25(OH)D was significantly lower in case group (27.32 vs 17.15, p < 0.01).
Table 1
Baseline data of indicators and clinical data
Characteristic | Control group (n = 125) | Fracture group (n = 109) | P value |
---|
Age (years) | 70 (67–74) | 72 (68–79) | 0.032 |
Height (cm) | 154.24 ± 6.22 | 162.49 ± 7.56 | < 0.01 |
Weight (kg) | 56 (51–62) | 57 (52–65) | 0.093 |
BMI (kg/m2) | 23.5(21.6–25.3) | 21.8(20.2–24.0) | < 0.01 |
Tobacco smoking | 3(2.40%) | 2(1.83%) | 0.767 |
Alcohol intake | 8(6.40%) | 5(4.59%) | 0.548 |
Hypertension | 18(14.40%) | 35(32.11%) | < 0.01 |
Diabetes | 18(14.40%) | 29(26.60%) | 0.02 |
β-CTX (ng/L) | 257.9 (149.1-400.5) | 383.3 (237.3-602.2) | < 0.01 |
N-MID (ng/mL) | 12.79 (9.68–17.59) | 12.37 (9.40-16.88) | 0.649 |
25(OH)D3 (ng/mL) | 27.32 (20.94–36.29) | 17.15 (12.40–21.80) | < 0.01 |
TotaHl hip T-score | -1.9 (-2.3–1.2) | -1.9 (-2.4–1.4) | 0.434 |
Femoral neck T-score | -1.6 (-2.1–0.9) | -1.5 (-2–1.1) | 0.877 |
Total T-score | -1.6 (-2.1–1) | -1.7 (-2.1–1.1) | 0.359 |
Abbreviation: BMI, Body Mass Index; β- CTX, C-terminal telopeptide of type I collagen; N-MID, N-terminal mid-fragment of osteocalcin; 25(OH)D,25-Hydroxyvitamin D |
Logistic regression analysis
Logistic regression analysis proves that height (crude OR = 1.203, 95%CI = 1.14–1.268; P < 0.01), weight (crude OR = 1.034, 95%CI = 1.002–1.066; P = 0.035), BMI (crude OR = 0.882, 95%CI = 0.813–0.957; P < 0.01), hypertension (crude OR = 2.812, 95%CI = 1.481–5.338; P < 0.01), β-CTX (crude OR = 1.002, 95%CI = 1.001–1.003; P < 0.01), 25(OH)D (crude OR = 0.883, 95%CI = 0.851–0.917; P < 0.01) are associated with the risk of hip fractures for elderly women without Osteoporosis(T-score≥-2.5). After adjustment for confounding factors, In all case groups, β-CTX (adjusted OR = 1.004, 95%=1.002–1.006, P < 0.01) and N-MID (adjusted OR = 0.816, 95%=0.743–0.897, P < 0.01) was significantly higher than that in the control group, 25(OH)D was significantly higher than that in the control group (adjusted OR = 0.884, 95%=0.841–0.928, P < 0.01), height remained a risk factor (adjusted OR = 1.245, 95%=1.058–1.487, P < 0.01). The confounding factors include age, weight, BMI, hypertension, tobacco smoking, femoral neck T-score, intertrochanteric T-score and total hip T-score (Table 2).
Table 2
Logistic regression of each indicator
| Model 1 (crude) | | | Model 2 (adjusted) | |
---|
| OR (95% CI) | P value | | OR (95% CI) | P value |
Age (years) | 1.021 (0.988–1.055) | 0.219 | | 1.023 (0.97–1.079) | 0.399 |
Height (cm) | 1.203 (1.14–1.268) | < 0.01 | | 1.254 (1.058–1.487) | p < 0.01 |
Weight (kg) | 1.034 (1.002–1.066) | 0.035 | | 0.989 (0.794–1.233) | 0.925 |
BMI (kg/m2) | 0.882 (0.813–0.957) | < 0.01 | | 0.945 (0.549–1.629) | 0.839 |
Hypertension | 2.812 (1.481–5.338) | < 0.01 | | 1.974 (0.751–5.186) | 0.168 |
β-CTX | 1.002 (1.001–1.003) | < 0.01 | | 1.004 (1.002–1.006) | p < 0.01 |
N-MID | 0.985 (0.944–1.027) | 0.476 | | 0.816 (0.743–0.897) | p < 0.01 |
25(OH)D | 0.883 (0.851–0.917) | < 0.01 | | 0.884 (0.841–0.928) | p < 0.01 |
Confounding sets include: Age, Weight, BMI, hypertension, Tobacco smoking, Femoral neck T-score, Intertrochanteric T-score, Total hip T-score |
Abbreviation: β-CTX, C-terminal telopeptide of type I collagen; N-MID, N-terminal mid-fragment of osteocalcin; 25(OH)D, 25-Hydroxyvitamin D |
The bolded values indicate a significance level of p < 0.05. |
The most accurate serum predictive indicator
We calculated the area under the curve (AUC) and determined the optimal cutoff value of N-MID (AUC = 0.477), β-CTX (AUC = 0.655), 25(OH)D (AUC = 0.804) (Fig. 2a), 25(OH)D exhibited the highest predict value (AUC = 0.85), with a cutoff value established at 24.07 ng/mL (Fig. 2b). And we compare the AUC curves of serum indicators using the DeLong test and found that 25(OH)D demonstrated significant disparities when compared to β-CTX and N-MID (P < 0.01, P < 0.01) (Table 3). This indicates that levels of 25(OH)D below 24.07 ng/mL, while not diagnostic of vitamin D deficiency, are still highly indicative of an increased risk of hip fracture.
Table 3
Pairwise comparison of ROC curves
serum indicator | Z statistic | P value |
---|
β-CTX ~ N-MID | 1.988 | P = 0.047 |
N-MID ~ 25(OH)D | 5.509 | P < 0.01 |
β-CTX ~ 25(OH)D | 3.632 | P < 0.01 |
Abbreviation: β-CTX, C-terminal telopeptide of type I collagen; N-MID, N-terminal mid-fragment of osteocalcin; 25(OH)D, 25-Hydroxyvitamin D |
Combined Indicators in assessing elderly hip fractures
We developed seven predictive models based on height, BTMs, and 25(OH)D (Model 1: Height, β-CTX; Model 2: Height, N-MID; Model 3: Height, 25(OH)D; Model 4: Height, β-CTX and N-MID; Model 5: Height, β-CTX and 25(OH)D; Model 6: Height, N-MID and 25(OH)D; Model 7: Height, β-CTX, N-MID and 25(OH)D) and calculated the the area under the curve.
Model 7 exhibited the highest area under the curve (AUC) value at 0.902, surpassing the AUC values of Model 1, Model 2, and Model 4 (0.822, 0.817, 0.848, respectively) (Fig. 2c). Subsequently, Delong's test was subsequently employed to compare whether there was a significant difference in the area under the ROC curve (AUC) between the models (Table 4). Model 3 demonstrated significant disparities when compared to Models 1 and 2 (P = 0.02, P = 0.01), Model 6 showed no significant differences compared to Model 3 (P = 0.09), while a significant distinction was observed between Model 7 and Models 3, 4, and 5. (P = 0.01, P = 0.03, P < 0.01) (Fig. 2c) (Fig. 2d), indicating that Model 7 is the best predictive model with an area under the curve of 0.902. The calculation formula is as follows: y = log0.199*height + 0.003*β-CTX-0.182*N-MID-0.121*25(OH)D, and a result exceeding 0.571 indicates a high risk of hip fracture.
Table 4
Pairwise comparison of ROC curves
model | Z statistic | P value |
---|
model1 ~ model2 | 0.382 | P = 0.70 |
model2 ~ model3 | 2.418 | P = 0.02 |
model1 ~ model3 | 2.445 | P = 0.01 |
model3 ~ model6 | 1.683 | P = 0.09 |
model3 ~ model7 | 2.44 | P = 0.01 |
model7 ~ model4 | 2.241 | P = 0.03 |
model7 ~ model5 | 2.876 | P < 0.01 |
model7 ~ model6 | 1.728 | P = 0.08 |
Model 1: height, β-CTX; Model 2: height, N-MID; Model 3: height, 25(OH)D; Model 4: height, β-CTX and N-MID; Model 5: height, β-CTX and 25(OH)D; Model 6: height, N-MID and 25(OH)D; Model 7: height, β-CTX, N-MID and 25(OH)D |