A recent comparison of in-hospital mortality following rupture of an abdominal aortic aneurysm in England and the USA suggests that too few patients in England are being selected for aneurysm repair, too few are treated in high-volume centres and too few are offered endovascular (EVAR) versus open repair, writes Michael Sweeting.1
Therefore, to reduce the mortality from ruptured aneurysms, to provide equity of access to aneurysm repair (endovascular and open surgery) and to avoid aneurysm repair in those cases where it was futile, we aimed to develop a “point-of-care” risk score, which is readily applicable in the emergency department and focuses on 48-hour mortality.
We used data from the IMPROVE randomised controlled trial,2 a multicentre UK trial of 613 patients who were randomised to either an EVAR strategy or open repair at presentation at an emergency department. Due to the number of patients recruited and the careful collection of patient characteristics, the data from this trial provides a unique opportunity to develop a comprehensive risk score for predicting short-term (48-hour) mortality in a group of ruptured aneurysm patients.
The risk factors considered in the prediction of 48-hour mortality risk are based on those that could be measured easily and quickly and are restricted to those collected within the IMPROVE trial, as well two further randomised trials that were used for external validation—the AJAX trial3 and the ECAR trial.4 Furthermore, this is the first ruptured abdominal aortic aneurysm risk score to incorporate easily measured morphological variables from a computed tomography scan. The risk factors considered are age, sex, admission haemoglobin, creatinine and systolic blood pressure, volume of IV fluids given before arrival in theatre, acute ischaemia, loss of consciousness, aneurysm diameter, aneurysm neck diameter, length and proximal angle.
Five hundred and thirty-six patients with a final diagnosis of ruptured aneurysm are included in the analysis, of which 135 (25%) died within 48 hours of randomisation. This represents the largest group of ruptured abdominal aortic aneurysm patients ever studied for developing a risk score. EVAR was commenced in 182 patients, open repair in 319 and no operation was performed in 35 patients, either because the patient did not reach the operating theatre alive or because a conscious decision to palliate was made.
After variable selection, the risk factors age, sex, haemoglobin, creatinine, systolic blood pressure, acute ischaemia, neck length and neck angle were all significant predictors of 48-hour mortality. A non-linear association was found with increasing age. A simplified “bedside” risk score was obtained by taking dichotomies of continuous variables at clinically relevant cut-points and assigning integer scores for being above or below each cut-point.
The model was well-calibrated with observed risk. The highest decile of predicted risk was associated with a 60% mortality rate suggesting that the score was not adequate to make decisions regarding treatment futility in high-risk patients. The discriminative performance of the risk score outperformed other published risk scores when evaluated on the IMPROVE data and gave comparable performance on the external ECAR trial data. When evaluating the risk score on the AJAX trial the performance was poorer and similar to previously published scores. A simplified risk-score without morphological factors did not significantly detract from the score’s predictive capabilities.
In conclusion, the new risk score performs well compared with previously published risk scores and could be a useful point-of-care device for triaging high-risk patients to specialist vascular centres.
Michael Sweeting, University of Cambridge, Cambridge, UK
- Karthikesalingam A, Holt PJ, Vidal-Diez A, et al. The Lancet 2014; 383:963–969
- IMPROVE trial investigators. BMJ 2014; 348:f7661
- Reimerink JJ, Hoornweg LL, Vahl AC, et al. Ann Surg 2013; 258:248–56
- Desgranges P, Kobeiter H, Katsahian S et al. Eur J Vasc Endovasc Surg 2015; 50(3):303–10