Patients’ Characteristics
After screening, a total of 826 male patients involved in our study. Among them, a total of 140 patients were defined as OP, and other 686 patients were defined as non-OP according to the diagnosis criteria. All patients have completed relevant examination. The cohort consisted of 165 patients with hypertension, 79 patients with diabetes, 92 patients with fracture history, 101 patients with cardiovascular disease, 49 patients with cerebrovascular disease, 27 patients with tumor, 18 patients with chronic renal failure, 17 patients with thyroid disease, 6 patients with rheumatism, 30 patients with neuropathy. All data of patients including demographic and disease features in the two group are shown in Table 1.
Table 1. Differences between demographic characteristics of the male patients. Categorical parameters were shown as percentage
Parameters
|
Non-OP (n = 686)
|
OP (n = 140)
|
P-values
|
Age (years)
|
|
|
< 0.001
|
40-59
|
268 (39.07%)
|
24 (17.14%)
|
|
60-74
|
319 (46.50%)
|
54 (38.57%)
|
|
≥ 75
|
99 (14.43%)
|
62 (44.29%)
|
|
BMI (Kg/m2)
|
|
|
< 0.001
|
< 18.5
|
19 (2.77%)
|
35 (25%)
|
|
18.5-24
|
302 (44.02%)
|
81 (57.86%)
|
|
> 24
|
365 (53.21%)
|
24 (17.14%)
|
|
Hypertension
|
|
|
0.358
|
Yes
|
141 (20.55%)
|
24 (17.14%)
|
|
No
|
545 (79.45%)
|
116 (82.86%)
|
|
Diabetes mellitus
|
|
|
0.089
|
Yes
|
71 (10.35%)
|
8 (5.71%)
|
|
No
|
615 (89.65%)
|
132 (94.29%)
|
|
Fracture history
|
|
|
< 0.001
|
Yes
|
56 (8.16%)
|
36 (25.71%)
|
|
No
|
630 (91.84%)
|
104 (74.29%)
|
|
Cardiovascular disease
|
|
|
< 0.001
|
Yes
|
71 (11.54%)
|
30 (21.43%)
|
|
No
|
615 (89.65%)
|
110 (78.57%)
|
|
Cerebrovascular disease
|
|
|
< 0.001
|
Yes
|
29 (4.23%)
|
20 (14.29%)
|
|
No
|
657 (95.77%)
|
120 (85.71%)
|
|
Tumor
|
|
|
0.458
|
Yes
|
21 (3.06%)
|
6 (4.29%)
|
|
No
|
665 (96.94%)
|
134 (95.71%)
|
|
CRF
|
|
|
0.216
|
Yes
|
13 (1.90%)
|
5 (3.57%)
|
|
No
|
673 (98.10%)
|
135 (96.43%)
|
|
Thyroid disease
|
|
|
0.042
|
Yes
|
11 (1.60%)
|
6 (4.29%)
|
|
No
|
675 (98.40%)
|
134 (95.71%)
|
|
Rheumatism
|
|
|
1.000
|
Yes
|
5 (0.73%)
|
1 (0.71%)
|
|
No
|
681 (99.27%)
|
139 (99.29%)
|
|
Neuropathy
|
|
|
0.148
|
Yes
|
22 (3.21%)
|
8 (5.71%)
|
|
No
|
664 (96.79%)
|
132 (94.29%)
|
|
ESR (mm/h)
|
|
|
< 0.001
|
≤ 20
|
600 (87.46%)
|
103 (73.57%)
|
|
> 20
|
86 (12.54%)
|
37 (26.43%)
|
|
WBC (109/L)
|
|
|
< 0.001
|
< 4
4-10
|
17 (2.48%)
617 (89.94%)
|
6 (4.29%)
110 (78.57%)
|
|
> 10
|
38 (5.54%)
|
24 (17.14%)
|
|
CRP (mg/L)
|
|
|
< 0.001
|
≤ 8
|
567 (82.65%)
|
93 (66.43%)
|
|
> 8
|
119 (17.35%)
|
47 (33.57%)
|
|
Abbreviations: BMI, body mass index; CRF: chronic renal failure; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; WBC, white blood cell count. P < 0.05 (two-sided) was considered statistically significant.
Cut-Off Value of Systemic Inflammatory Markers
The appropriate cut-off value for SII, NLR, LMR, and PLR were 725.00 (area under the curve, 0.601; P < 0.001), 2.52 (area under the curve, 0.595; P < 0.001), 3.23 (area under the curve, 0.626; P < 0.001), 147.80 (area under the curve, 0.600; P < 0.001), respectively (Figure 1). As shown in Table 2, the systemic inflammatory markers (SII, NLR, LMR, and PLR) were higher in the OP group than in the non-OP group (all P value < 0.001).
Table 2. Comparison of the systemic inflammatory markers of male patients. Categorical parameters were shown as percentage
Parameters
|
Non-OP (n = 686)
|
OP (n = 140)
|
P-values
|
SII
|
|
|
< 0.001
|
≤ 725.00
|
555 (80.90%)
|
80 (57.14%)
|
|
> 725.00
|
131 (19.10%)
|
60 (42.86%)
|
|
NLR
|
|
|
< 0.001
|
≤ 2.52
|
472 (68.80%)
|
66 (47.14%)
|
|
> 2.52
|
214 (31.20%)
|
74 (52.86%)
|
|
LMR
|
|
|
< 0.001
|
≤ 3.23
|
202 (29.45%)
|
73 (52.14%)
|
|
> 3.23
|
484 (70.55%)
|
67 (47.86%)
|
|
PLR
|
|
|
< 0.001
|
≤ 147.80
|
525 (76.53%)
|
76 (54.29%)
|
|
> 147.80
|
161 (23.47%)
|
64 (45.71%)
|
|
Abbreviations: LMR, lymphocyte-to-monocyte ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; SII, systemic immune-inflammation index. P < 0.05 (two-sided) was considered statistically significant.
Feature Selection
The variables in present study included demographic, disease, and blood biochemical characteristic. We selected 10 characteristic variables from 19 variables through LASSO regression model with nonzero coefficients (Figure 2 A and B). These variables included age, body mass index (BMI), cardiovascular diseases, cerebrovascular disease, neuropathy, thyroid disease, fracture history, SII, PLR, CRP (Table 3).
Development of an Individualized Prediction Model
We employed the feature variables selected from the LASSO regression model to build the prediction model. The predictive model was construct using multiple logistic regression method (Table 3), and the predictive model constructed by the predictors mentioned above was displayed as a nomogram (Figure 3).
Table 3. Prediction factors for osteoporosis in Chinese male patients.
Parameters
|
β
|
Wald
|
P
|
OR
|
95%CI
|
Intercept
|
-0.585
|
-1.551
|
0.121
|
0.557
|
0.264-1.168
|
Age (years)
|
|
|
|
|
|
40-59
|
|
|
|
|
Reference
|
60-74
|
0.416
|
1.480
|
0.139
|
1.516
|
0.882-2.668
|
≥ 75
|
1.118
|
3.565
|
<0.001
|
3.058
|
1.664-5.711
|
BMI (Kg/m2)
|
|
|
|
|
|
< 18.5
|
|
|
|
|
Reference
|
18.5-24
|
-1.762
|
-5.192
|
<0.001
|
0.172
|
0.087-0.331
|
> 24
|
-2.995
|
-7.823
|
<0.001
|
0.050
|
0.023-0.104
|
Cardiovascular disease
|
0.295
|
1.019
|
0.308
|
1.344
|
0.752-2.348
|
Cerebrovascular disease
|
1.053
|
2.907
|
0.004
|
2.865
|
1.392-5.794
|
Neuropathy
|
0.821
|
1.654
|
0.098
|
2.273
|
0.818-5.830
|
Thyroid disease
|
0.803
|
1.219
|
0.223
|
2.231
|
0.581-7.818
|
Fracture history
|
0.829
|
2.724
|
0.006
|
2.292
|
1.255-4.151
|
SII
|
|
|
|
|
|
≤ 725
|
|
|
|
|
Reference
|
> 725
|
0.402
|
1.298
|
0.194
|
1.495
|
0.811-2.741
|
PLR
|
|
|
|
|
|
≤ 147.8
|
|
|
|
|
Reference
|
> 147.8
|
0.156
|
0.534
|
0.594
|
1.168
|
0.653-2.054
|
CRP
|
|
|
|
|
|
≤ 8
|
|
|
|
|
Reference
|
> 8
|
0.041
|
0.144
|
0.886
|
1.041
|
0.592-1.796
|
Abbreviations: BMI, body mass index; CRP, C-reactive protein; PLR, platelet-to-lymphocyte ratio; SII, systemic immune-inflammation index. β is the regression coefficient. P < 0.05 (two-sided) was considered statistically significant.
Apparent Performance of the Osteoporosis Risk Nomogram in the Cohort
The area under the ROC (AUC) of the risk nomogram was 0.822 (CI: 0.798-0.846) (Figure 4), which suggested that the model had adequate predictive potential and performance. The calibration curve of the osteoporosis risk nomogram for the prediction of osteoporosis risk in male based on features such as SII and PLR illustrated good agreement in this cohort (Figure 5 A). The C-index for this nomogram was 0.822 (CI: 0.798-0.846) for the cohort, and was confirm to be 0.8048 through bootstrapping validation, which suggested this prediction model had a good discrimination. In the osteoporosis risk nomogram, apparent performance addressed a good prediction capability.
Clinical Use
The decision curve analysis (DCA) for the male osteoporosis risk nomogram is shown in Figure 5 B. The DCA showed that if the threshold probability of the DCA >3% and <76%, using the risk nomogram to predict the risk of OP for male population was beneficial in clinical work. Within this range, net benefit was comparable with several overlaps.