87 Validation of a multicenter risk model for implant loss following implant-based reconstruction Figure 1. Flow-chart of in- and excluded patients Risk model validation The validation cohort consisted of 3769 reconstructions and implant loss occurred after 307 reconstructions (8.1%). Patient and surgery characteristics stratified for implant loss are summarized in Table 3. There were active smokers in 486 (12.9%) reconstructions and obese patients (BMI>30) in 401 (10.6%) reconstructions. A nipple sparing procedure was performed in 1126 (29.9%) reconstructions and a definite implant was directly placed in 832 (22.1%) reconstructions. This resulted in no risk factors for 1764 reconstructions, one risk factor for 1480 reconstructions, two risk factors for 485 reconstructions, three risk factors for 39 reconstructions and four risk factors for one reconstruction. The observed implant loss rates for each number of risk factors are presented in Table 4. The predicted probabilities for each risk factor combination were extracted and compared to the observed probabilities of the validation cohort. This comparison was visualized in a calibration plot (Figure 2). A substantial agreement in probabilities was observed from 0.0 to 0.13, as the reference line lies within the CI of four out of five data points. However, the rest of the predicted and observed probabilities did not match, indicating a poor agreement. Association between risk factors and implant loss in current cohort The associations between risk factors and implant loss were determined in the current cohort using univariable logistic regression. Obesity and active smoking status were significantly associated with implant loss (OR: 1.499 (1.072-2.094), P=0.019 and OR: 1.772 (1.315-2.387), P<0.001 respectively). A nipple preserving procedure and DTI reconstruction were not significantly related to implant loss (OR: 1.005 (0.799-1.295), P=0.971 and OR: 0.984 (0.742-1.305), P=0.984 respectively). These results are summarized in Table 5. 6
RkJQdWJsaXNoZXIy MTk4NDMw