Simple Predictive Model of Early Failure among Patients Undergoing First-Time Arteriovenous Fistula Creation.

BACKGROUND

Native arteriovenous fistulas (AVFs) have a high 1 year failure rate leading to a need for secondary procedures. We set out to create a predictive model of early failure in patients undergoing first-time AVF creation, to identify failure-associated factors and stratify initial failure risk.

METHODS

The Vascular Study Group of New England (VSGNE) (2010-2014) was queried to identify patients undergoing first-time AVF creation. Patients with early (within 3 months postoperation) AVF failure (EF) or no failure (NF) were compared, failure being defined as any AVF that could not be used for dialysis. A multivariate logistic regression predictive model of EF based on perioperative clinical variables was created. Backward elimination with alpha level of 0.2 was used to create a parsimonious model.

RESULTS

We identified 376 first-time AVF patients with follow-up data available in VSGNE. EF rate was 17.5%. Patients in the EF group had lower rates of hypertension (80.3% vs. 93.2%, P = 0.003) and diabetes (47.0% vs. 61.3%, P = 0.039). EF patients were also more likely to have radial artery inflow (57.6% vs. 38.4%, P = 0.011) and have forearm cephalic vein outflow (57.6% vs. 36.5%, P = 0.008). Additionally, the EF group was noted to have significantly smaller mean diameters of target artery (3.1 ± 0.9 vs. 3.6 ± 1.1, P = 0.002) and vein (3.1 ± 0.7 vs. 3.6 ± 0.9, P < 0.001). Multivariate analyses revealed that hypertension, diabetes, and vein larger than 3 mm were protective of EF (P < 0.05). The discriminating ability of this model was good (C-statistic = 0.731) and the model fits the data well (Hosmer-Lemeshow P = 0.149). β-estimates of significant factors were used to create a point system and assign probabilities of EF.

CONCLUSIONS

We developed a simple model that robustly predicts first-time AVF EF and suggests that anatomical and clinical factors directly affect early AVF outcomes. The risk score has the potential to be used in clinical settings to stratify risk and make informed follow-up plans for AVF patients.