In the study period, 2593 cases of primary joint replacement were recorded in our database. A total of 1782 cases received ultrasound examination of VTE-3m were enrolled in the present study, 811 cases did not receive examination were excluded. The mean age of the study population was 65.4 (standard deviation 12.3) years-old, and 72.8% were females. The gross incidence of VTE-3m was 10.5% (187 cases). The distribution of every parameter by patients with and without VTE-3m is shown in Table 1. Age (p < 0.001, t-test), gender (p < 0.001, chi-square test), previous history of hypertension, thrombotic disease and anticoagulants usage (p = 0.017, < 0.001 and < 0.001, respectively, chi-square test), VTE pre-surg (p < 0.001, chi-square test), VTE post-surg (p < 0.001, chi-square test) and OT-season (p = 0.022, chi-square test) were correlated with the existence of VTE-3m. The incidence of VTE-3m was significantly higher when the surgery was performed in autumn (spring 8.4%, summer 10.7%, autumn 14.0%, and winter 9.1%). Confounder effect of gender was also tested. The difference between gender were observed in the type of operation (p < 0.001), previous history of smoking (p < 0.001), tumor (p = 0.002) and hypertension (p = 0.039), and VTE-presurg (p = 0.025), VTE-postsurg (p < 0.001) and VTE-3m (p < 0.001).
Before risk factors were further identified by using logistic regression analysis, 38 cases were omitted due to missing data. Final analyses were performed in 1744 patients. VTE pre-surg was excluded from logistic regression analysis to avoid collinearity (VTE pre-surg was significantly correlated with VTE post-surg (p < 0.001, chi-square test)). The second reason to exclude VTE pre-surg from the analysis was the high missing data rate (9.9%) caused by the limited pre-surgical preparation time. The use of anticoagulants before surgery was also excluded because it was significantly correlated with the previous history of the thrombotic disease (p = 0.012, chi-square test). Previous history of hypertension was also excluded because it was found not significant (p = 0.973, logistic regression test) by running preemptive logistic regression analysis. Finally, five robust risk factors (age, gender, previous history of thrombotic disease, VTE post-surg, and OT-season) were identified by using logistic regression analysis (Table 2). The goodness-of-fit of this statistic model was performed by Hosmer and Lemeshow test ( p = 0.863), the area of ROC is 0.788 (p < 0.001, 95%CI 0.752 - 0.825).
The results showed that VTE post-surg was the strong risk factor of VTE-3m, odds of having VTE-3m increased over 7 times in those patients who had lower limb VTE one-day after surgery (OR = 7.633, 95% CI 5.410-10.769, p < 0.001).
The OT season also affected the incidence of VTE-3m. Arthroplasty surgeries performed in autumn increased 1.9 times odds of having VTE-3m compared to surgeries performed in spring (OR = 1.874, 95% CI 1.241-2.829, p = 0.003). However, the risk of VTE was not increased when surgery was performed in summer and winter (p = 0.191 and 0.857, respectively).
Female gender was also a risk factor of VTE-3m; odds of having VTE-3m increased 82.8% in females compared to males (OR = 1.828, 95% CI 1.164-2.873, P = 0.009).
Odds of having VTE-3m also significantly increased with age, with increased 2.3% each yearly (OR = 1.023, 95% CI 1.006-1.041, P = 0.007).
Previous history of thrombotic disease could increase the odds of incidence of VTE-3m with 59.2% (OR = 1.592, 95% CI 1.044-2.427, P = 0.031).
A forest plot (Figure 2) was provided for better visualization of the logistic regression results.