Objective To investigate the relationships of CXC motif chemokine ligand 10 (CXCL10) and CC motif chemokine ligand 11 (CCL11) in the peripheral blood with type 2 diabetic osteoporosis (T2DOP), and to construct a predictive model.

Methods A total of 260 patients with type 2 diabetes mellitus (T2DM) were prospectively selected as the T2DM group. They were divided into T2DOP group (n=82) and non-T2DOP group (n=178) according to the occurrence of T2DOP. Additionally, 68 healthy volunteers with physical examinations in the same period were selected as control group. Enzyme-linked immunosorbent assay was used to detect the levels of CXCL10 and CCL11 in the peripheral blood as well as bone metabolism indicatorsβ-C-terminal telopeptide of type Ⅰ collagen (β-CTX), N-terminal propeptide of type Ⅰ procollagen (PINP). Pearson correlation analysis was used to explore the correlation of CXCL10 and CCL11 in peripheral blood with bone metabolism indicator levels in patients with T2DM. Multifactor non-conditional Logistic regression analysis was used to explore the influencing factors of T2DOP and construct a predictive model. The Hosmer-Lemeshow (H-L) test was used to assess the goodness of fit. The receiver operating characteristic (ROC) curve analysis was used to evaluate the predictive value of each indicator and the predictive model for T2DOP, and decision curve analysis and Bootstrap resampling were used for internal validation.

Results Compared with the control group, the levels of CXCL10 and CCL11 in the peripheral blood as well as β-CTX in the T2DM group were significantly increased, while the level of PⅠNP was significantly decreased (P < 0.05). Pearson correlation analysis showed that CXCL10 and CCL11 in the peripheral blood in patients with T2DM were positively correlated with β-CTX level (r=0.786, 0.816, P < 0.001) and negatively correlated with PⅠNP level (r=-0.675, -0.716, P < 0.001). Compared with the non-T2DOP group, the T2DOP group had significantly increased age and the ratio of diabetic nephropathy, prolonged duration of diabetes, decreased body mass index (BMI) and PⅠNP levels, and increased fasting blood glucose, glycated hemoglobin (HbA1c), β-CTX, CXCL10, and CCL11 levels (P < 0.05). Non-conditional Logistic regression analysis showed that advanced age, long duration of diabetes, high HbA1c, high CXCL10, and high CCL11 were independent risk factors for T2DOP (P < 0.05), while high BMI was an independent protective factor (P < 0.05). The ROC curve showed that the area under the curve (AUC) of the predictive model for predicting T2DOP was 0.919, which was significantly greater than the AUCs of 0.643, 0.742, 0.654, 0.715, 0.759 and 0.741 respectively for age, duration of diabetes, BMI, HbA1c, CXCL10 and CCL11 alone (Z=7.468, 5.400, 7.415, 6.365, 5.242, 5.800, P < 0.001). After internal validation, the decision curve of the predictive model was higher than the two extreme curves. After 1, 000 times of Bootstrap resampling internal validations, the concordance index of the predictive model was 0.919(95%CI, 0.914 to 0.923).

Conclusion Increased levels of CXCL10 and CCL11 in the peripheral blood are associated with T2DOP, and the predictive model constructed based on these factors has a high predictive value for T2DOP.