External validation of the Glasgow Aneurysm Score to predict outcome in elective open abdominal aortic aneurysm repair

Methods 

Glasgow Aneurysm Score and other variables 

The GAS is calculated using the formula: risk score = age in years + (7 points for myocardial disease) + (10 points for cerebrovascular disease) + (14 points for renal disease). Myocardial disease refers to previously documented myocardial infarction or ongoing angina pectoris, or both. Cerebrovascular disease comprises all grades of stroke including transient ischemic attack. Renal disease is defined as a serum level of urea >20 mmol/L, a serum level of creatinine >150 μmol/L at the time of surgery, a history of chronic or acute renal failure, or a combination of these.1

In addition to the GAS, we included several other variables that might also predict prognosis after elective surgery (Table I). Furthermore, the risk factors of age, cardiac disease, cerebrovascular disease, and renal disease are components of the GAS and have also been analyzed individually to have a better understanding of the GAS.

Table I. Demographic characteristics in 229 patients (1994 to 2003)

	Examined risk factors	GAS component	N	%
	Sex, male		193	84.3
	Age at operation (yrs) (range)	Yes	69 (23-89)
	ASA classification
	I		4	1.7
	II		57	24.9
	III		138	60.3
	IV		30	13.1
	Clamp
	Infrarenal		193	84.3
	Suprarenal		36	15.7
	No β-blockers (peri-op)		132	57.6
	Cardiac disease	Yes	83	36.2
	Cerebrovascular disease	Yes	30	13.1
	Renal disease	Yes	14	6.1
	Diabetes mellitus		17	7.4
	COPD		28	12.2
	Mean BMI (kg/m2) (range)		25.6 (15.9-36.4)
	History of smoking		206	90
	GAS in points, mean (range)		73 (23-112)
	Outcome end points
	In-hospital mortality		5	2.2
	Major morbidity		30	13.1

GAS, Glasgow Aneurysm Score; ASA, American Society of Anesthesiologists; COPD, chronic obstructive pulmonary disease; BMI, body mass index.

End points 

Two end points were chosen: in-hospital mortality and major morbidity. We defined major morbidity as postoperative complications that are irreversible or need admittance at the intensive care unit or reoperation (Table II).

Table II. Major morbidity

	Major morbidity	Frequency n (%)
	Reoperation
	For peripheral ischemia	9(3.9)
	For bowel ischemia	6(2.6)
	Renal failure requiring dialysis	3(1.3)
	Pulmonary embolism	0(0.0)
	Clinically relevant myocardial infarction	6(2.6)
	Cardiac arrhythmia requiring resuscitation	4(1.7)
	Stroke	0(0.0)
	ARDS	1(0.4)
	LV failure requiring ICU readmittance	5(2.2)
	Trash-foot	3(1.3)
	Infection of prosthesis	1(0.4)

ARDS, Adult respiratory distress syndrome; ICU, intensive care unit; LV, left ventricular.

Results 

During the study period, 229 consecutive patients underwent elective open repair of an infrarenal or juxtarenal aneurysm. All were included in this study. The demographic characteristics of this population are shown in Table I. Not included were 68 patients with endovascular repair.

In-hospital mortality 

Five patients (2.2%; 95% confidence interval [CI], 0.7% to 5.0%) died after surgery (Table I). With univariable analysis, four factors had a possible predicting value (Table III). After multivariable analysis, the GAS (P = .021), suprarenal clamping (P = .017), and a history of smoking (P = .026) proved to be independent predictors of death (Table IV). Among the individual components of the GAS (eg, age), no significant differences could be found.

Table III. Univariable analysis: possible predicting factors (in bold) for the in-hospital mortality and major morbidity

	Examined variables	Mortality (P)	RR⁎ (95% CI)	Major morbidity (P)	RR⁎ (95% CI)
	GAS	0.027		0.011
	Age at time of operation	0.163		0.038
	Cardiac disease	0.059	7.0(0.80-61.9)	0.043	2.0(1.03-3.91)
	Cerebrovascular disease	0.508	1.7(0.19-14.3)	0.775	0.7(0.24-2.28)
	Renal disease	0.273	3.8(0.46-32.1)	0.094	2.4(1.08-5.94)
	Sex	0.578	0.8(0.09-6.52)	0.278	0.5(0.24-1.35)
	ASA classification	0.794		0.722
	Body mass index	0.389		0.342
	Suprarenal clamp	0.029	8.3(1.42-46.4)	0.278	0.6(0.28-1.32)
	Diabetes mellitus	1.000	0.0(NA)	1.000	0.9(0.23-3.43)
	COPD	1.000	0.0(NA)	0.548	0.5(0.13-2.04)
	No β-blockers (peri-op)	0.165	0.2(0.02-1.62)	0.693	0.8(0.43-1.64)
	History of smoking	0.080	0.2(0.03-0.95)	0.516	0.7(0.28-1.90)

RR, Risk ratios; CI, confidence intervals; GAS, Glasgow Aneurysm Score; ASA, American Society of Anesthesiologists; COPD, chronic obstructive pulmonary disease; NA, not accessible.

⁎ The risk ratios are given for the dichotomous variables.

Table IV. Multivariable analysis: predicting factors for in-hospital mortality and major morbidity

	Predicting factors	P	RR (95% CI)
	In-hospital mortality
	Suprarenal clamp	.017	10.0(1.54-84.6)
	History of smoking	.026	0.09(0.01-0.74)
	Major morbidity
	Glasgow Aneurysm Score	.029	1.3(1.09-1.43)
	Age at time of operation	.046	1.2(1.01-1.38)
	Cardiac disease	.032	5.0(1.14-16.7)
	Renal disease	.041	2.8(1.26-37.9)

Analysis of the ROC curve (Fig 1) showed that the GAS had an area under the curve of 0.839 for predicting the postoperative in-hospital mortality. According to the ROC curve, the best cutoff point for the GAS was 76.5. At this cut-off point the false-negatives were reduced to a minimum (sensitivity, 100%; specificity, 67%). Further characteristics of this test can be found in Table V.

View Large Image Download to PowerPoint

Fig 1. According to the receiver operating characteristic (ROC) curve, the best cutoff point is for the Glasgow Aneurysm Score (GAS) for mortality is 76.5.

Table V. Glasgow Aneurysm Score characteristics for in-hospital mortality and major morbidity⁎

	Test characteristics cutoff point: 76.5	In-hospital mortality value (95% CI)	Major morbidity value (95% CI)
	Sensitivity	1.00(0.52-1.00)	0.63(0.46-0.78)
	Specificity	0.67(0.61-0.73)	0.70(0.64-0.76)
	Positive predictive value	0.06(0.03-0.15)	0.24(0.16-0.35)
	Negative predictive value	1.00(0.97-1.00)	0.93(0.87-0.96)
	Positive likelihood ratio	3.07(2.07-3.81)	2.14(1.51-3.02)
	Negative likelihood ratio	0.00(0.00-0.76)	0.52(0.32-0.84)

CI, Confidence interval.

⁎ Patients with ≥77 points are categorized in the high-risk group.

Major morbidity 

Thirty patients (13.1%) had one or more major complications (Table I). With univariable analysis, four factors (Table III) had a possible predicting value. Multivariable analysis (Table IV) showed the GAS (P = .029) was a predicting scoring system. Furthermore, three of its four individual components—age (P = .046), cardiac disease (P = .032), and renal disease (P = .041)—were also independent predictors of major morbidity.

The GAS had an area under the curve of 0.661 (Fig 2) in predicting postoperative major morbidity after elective AAA repair. The best cutoff value for the GAS was 76.5 (sensitivity, 63%; specificity, 70%). Patients with a risk score of ≥77 points belonged to the high-risk group. In Table V, the characteristics of the test and the 95% CIs are summarized.

View Large Image Download to PowerPoint

Fig 2. According to the receiver operating characteristic (ROC) curve, the best cutoff point for the Glasgow Aneurysm Score (GAS) for morbidity is 76.5.

Discussion 

Our study showed the GAS had a predictive value for prognosis after elective open AAA repair. Suprarenal clamping also was predictive for mortality. Three components of the GAS, namely age, cardiac disease and renal disease, also independently proved to have a predictive value for major morbidity. The end points in our study were mortality and major morbidity (Table I, Table II). Minor morbidity was not chosen as an end point because the GAS was mainly developed to prevent more severe complications as they have a higher negative impact on patients’ lives.

According to our study results, patients are divided into low-risk and high-risk groups with a cutoff point to differentiate at a GAS of 77 (Table VI). This fairly corresponds with the original study of Biancari4 that classified patients with a score of 79 points in the high-risk group. As such, the GAS is shown to be externally valid; however, the GAS has limitations for clinical decision-making. Striking are the low positive predictive value and positive likelihood ratio, meaning that the system is not very helpful for the individual high-risk patient. Choosing another cutoff value does not contribute to a better test because it will lead to more false-negatives and only slightly improves the positive likelihood ratio.

Table VI. Prognosis in the two risk groups⁎

	Risk group (n)	In-hospital mortality n (%)	Major morbidity n (%)
	Low-risk, 151	0(0.0)	11(7.3)
	High-risk, 78	5(6.4)	19(24.4)

⁎ Patients with ≥77 points belong to the high-risk group.

On the other hand, the GAS turned out to have a good negative predictive value and the negative likelihood ratio is also acceptable, meaning that patients with a large AAA who belong to the low-risk group are not likely to die or develop a major complication if open AAA repair is done. Other studies draw similar conclusions: when patients have few or no risk factors the postoperative mortality after elective repair is low, from 2.7% to 3.0% vs 7.8% to 9.0% in patients with more risk factors.4, 5

Nesi et al10 also describe a low positive predictive value and the high negative predictive value of this scoring system for postoperative mortality and morbidity. Their study compared the GAS with four other scoring systems for the prediction of outcome and concluded that the GAS was the most useful for a quick risk stratification during initial patient consultations and is the easiest to use in routine clinical practice. Samy et al9 also recommended use of the GAS in clinical practice because the test meets the criteria for a useful surgical score in that the test is simple, easy to calculate, and does not require an experienced physician. These two studies described the pros of the test but not the likelihood ratios and predictive values in this context.

The GAS also has some other limitations: it does not account for the protective effects of coronary artery bypass grafting or percutaneous transluminal coronary angioplasty for cardiac disease, carotid endarterectomy for cerebrovascular disease, and the use of medications in general. Nonetheless, the GAS was examined in several European studies1, 4, 10 and all emphasized its validity in predicting outcome after open repair. As for endovascular repair, the Endovascular Aneurysm Repair 2 (EVAR2) trial showed that this type of surgery is accompanied with a high mortality and morbidity rate in high-risk patients and has no survival benefit compared with no intervention.11 Therefore, assessment of the validity of the GAS for such patients can be useful. A recent report of Biancari et al12 made clear that the GAS also predicts the postoperative mortality after endovascular repair.

From the other 12 examined parameters, suprarenal clamping was a predictor of postoperative mortality (Table VII). Green et al13 also concluded that suprarenal clamping was associated with a higher mortality rate. El-Sabrout et al14 mentioned a comparable percentage for mortality after suprarenal clamping (5.7%), but concluded that there was no significant difference between infrarenal and suprarenal clamping because some variables (eg, age) differed among the examined groups in their study. Other reports showed an insignificant increase in the mortality rate and stated that suprarenal clamping could be performed at low risk.15, 16 However, all studies that have examined this possible risk factor, including the current one, included a small number of patients (n < 60) with suprarenal clamping.

Table VII. Patients with suprarenal clamping have a significant higher chance of dying, those in the smoking group have not

	Risk factor	Total N	Postoperative death N (%)
	Clamping
	Infrarenal	193	2(1.0)
	Suprarenal	36	3(8.3)
	History of smoking
	No	23	2(8.7)
	Yes	206	3(1.5)

Our study found a history of smoking was also significant after multivariable analysis for postoperative mortality (Table 4). The mortality rate was lower in the smoking group (Table VII), however, which would assign a protective effect for smoking. This is not conceivable and is merely a result of a high percentage of smokers (90.0%) and a low number of events (5 deaths, 2.2%), as shown in Table I. Besides, we could not differentiate in the quantity and duration characteristics of the smoking habits of the patients with our data, which influenced the results. This is one of the limitations of this study. Brady et al5 examined a population of 820 patients and concluded that smoking was not associated with a poorer outcome regarding mortality. Another limitation of our study remains the relatively small number of patients included and the retrospective design.

As for major morbidity, the factors of age, cardiac disease, and renal disease—three of the four components of the GAS—independently proved to have a predictive value (Table IV). Thus, the significance of the GAS cannot exclusively be ascribed to one separate parameter: it is, with exception of the cerebrovascular component, formed by all.

Conclusion 

The GAS has been documented to have external validity. The scoring system is particularly useful in low-risk patients because of its negative predictive value. These patients have a good prognosis and this can be communicated to the patient. It also helps to identify high-risk patients, although some limitations remain with the individual risk assessment as a result of the low predictive value. Because the test is not useful for clinical decision-making for every individual patient, we do not recommend that patients be selected for elective surgery solely on the GAS. The usefulness of the GAS is thus mainly as an additional tool to help the surgeon to reach a decision about patients who have a large aneurysm but who also have a complicated risk profile. We have incorporated the calculation of the GAS score in our electronic patient chart system and consider its calculation helpful in the risk communication with the patient.

Author contributions