Kelong Shi, Jiang Xu, Yongxiang Tang, MingXu Ou, GuChao Zhou
Intraoperative DSA-derived lesion length showed a strong, markedly non-linear association with AVF re-failure after PTA and carried nearly all of the discriminative ability of the model; the apparent effect of composite morphology was substantially attenuated once length was modelled appropriately, indicating that morphology acts in part as a surrogate for lesion extent. Because vessel diameter, stenosis severity, local calcification and procedural parameters were not available, these findings are hypothesis-generating and require prospective external validation with comprehensive lesion characterisation before any clinical implementation.
BACKGROUND: Risk stratification for arteriovenous fistula (AVF) re-failure after percutaneous transluminal angioplasty (PTA) relies on systemic comorbidity markers. These markers are costly and inconsistently predictive in resource-limited settings. We hypothesized that intraoperative digital subtraction angiography (DSA) variables, lesion length and lesion morphology, could stratify the risk of re-failure without additional testing.
METHODS: This single-center retrospective derivation study included 243 patients with successful PTA using uniform Mustang™ balloons. We used Fine-Gray subdistribution hazard models with death and kidney transplantation as competing events. Missing data were handled by multiple imputation. Model discrimination and calibration were assessed by optimism-corrected C-index via 500 bootstrap resamples and calibration plots. Lesion length was modelled with a restricted cubic spline (3 knots); lesion location was analysed as four mutually exclusive categories with anastomotic/peri-anastomotic stenosis as the reference. Calibration was assessed in quintiles of predicted 1-year cumulative incidence.
RESULTS: During follow-up, 103 patients (42.4%) experienced re-failure and 24 (9.9%) had competing events. Lesion length showed a strong non-linear association with re-failure (test for non-linearity, P < 0.001). Relative to the median length of 1.75 cm, the subdistribution hazard ratio (SHR) was 0.25 (95% CI, 0.20-0.32) at 1 cm, 1.46 (95% CI, 1.37-1.56) at 2 cm and 4.31 (95% CI, 3.37-5.51) at 3 cm. Once lesion length was modelled as a spline, the effect of composite morphology was attenuated by 35%, from 2.81 (95% CI, 1.70-4.66) under a linear specification to 1.83 (95% CI, 1.13-2.97) (P = 0.014). The final model achieved an optimism-corrected C-index of 0.79 at 1 year and 0.80 at 2 years. Age, diabetes and coronary artery calcium score showed no significant association.
CONCLUSION: Intraoperative DSA-derived lesion length showed a strong, markedly non-linear association with AVF re-failure after PTA and carried nearly all of the discriminative ability of the model; the apparent effect of composite morphology was substantially attenuated once length was modelled appropriately, indicating that morphology acts in part as a surrogate for lesion extent. Because vessel diameter, stenosis severity, local calcification and procedural parameters were not available, these findings are hypothesis-generating and require prospective external validation with comprehensive lesion characterisation before any clinical implementation.