Cost-effectiveness of conventional and endovascular repair of abdominal aortic aneurysms: Results of a randomized trial

Study design and patients 

A detailed description of the design and methods of the DREAM trial has been presented elsewhere.19 In brief, patients diagnosed with an AAA of ≥5 cm in diameter who were considered suitable for both techniques were randomly assigned to endovascular repair or open repair, after giving written informed consent. The study excluded patients requiring emergency AAA repair and patients with inflammatory aneurysms, anatomic variations, connective tissue disease, previous organ transplantation, and a life expectancy of <2 years. The study was performed according to the principles of the Declaration of Helsinki. The Institutional Review Boards of all participating hospitals approved the protocol. The present analyses are limited to patients with complete 1-year data on costs and effects by fall 2005.

End points 

Complications were classified and graded according to the reporting standards of the Ad Hoc Committee for Standardized Reporting Practices in Vascular Surgery of The Society for Vascular Surgery/International Society for Cardiovascular Surgery.20, 21 A committee blinded to the treatment allocation evaluated the clinical outcomes. Three severity grades—mild, moderate, and severe—were distinguished. Only severe events were considered for the present analyses.

Quality of life 

HRQOL was assessed with the Dutch language version of the EuroQol 5 Dimensions questionnaire (EQ-5D) using the summary score based on the tariffs by Dolan.22 The EQ-5D questionnaire is a multiattribute instrument reflecting the societal preferences and values for the five domains discerned and their mutual interdependence. Value judgments have been derived previously, independent of potential (economic) cost consequences for current patients. At different points in time, utility scores were obtained by means of EQ-5D. The EQ-5D results in a single numeric score that represents the societal valuation of the actual health status. A value of 1 represents optimal health, and 0 represents death.

Questionnaires were sent to patients at baseline (upon randomization), at 3 and 6 weeks, and at 3, 6, and 12 months postoperatively. Using linear interpolation for the periods between measurements, we calculated the quality-adjusted survival time in terms of quality-adjusted life-years (QALY) by determining the individual area under the curve. These results were previously reported.13 For the present analyses, the results of the remaining patients that completed their 1-year follow-up were added. We used simple imputation based on clinical outcome to obtain a “complete” data set. Baseline and other (outcome) characteristics of the patients with missing data were no different from those with complete data.

Costs 

Medical costs associated with treatment and follow-up until 1 year after inclusion were assessed in Euros. Costs per patient were calculated by multiplying individual resource use with unit costs. The volume of the resource use during admission and follow-up (out-patient visits) was recorded in the case record forms and completed by means of patient diaries. The latter were specifically designed to capture additional resource use such as consultations of a physiotherapist or family physicians, home care, medication used, and other admissions, such as for rehabilitation. Additional examinations specifically performed for study purposes and not serving any clinical purpose, such as additional computed tomography angiography follow-up in the open repair group, were not included in the analysis. Where unavailable from existing sources, unit costs were determined in two academic and two peripheral hospitals.

Notably, the costs of the devices were based on the actual purchasing prices (inclusive of value added taxes and shipping). Personnel costs comprised all individuals involved, inclusive of specialists and operating room staff. A weighted mean of costs per item was calculated by using the ratio of the patients treated in academic and peripheral settings.

We based our estimates of average costs per hospital day (ICU and ward) on a study previously performed in the Netherlands.23 Costs of outpatient visits, visits to family physicians, and home care were calculated from unit costs reported by the Dutch Costing Manual issued by the National Health Insurance Council.24 Costs of medication were estimated using the Dutch Formulary (Pharmacotherapeutic Compass 2003) and included the pharmacist’s charges. Costs of rarely performed interventions were based on national tariffs (College Tarieven Gezondheidszorg (CTG), http://www.ctg-zaio.nl). Where necessary, unit costs were adjusted to 2003 values according to National Health Service Costs Index issued by the National Bureau for Statistics Netherlands.

We also assessed costs due to losses in productivity associated with sick leave and travel as well as other private costs incurred by patients and their families. Time costs were valued using the friction cost method.25 The Health and Labour Questionnaire26 was used to assess actual sick leave.

Cost-effectiveness 

For the economic evaluation, the estimates of QALY were chosen as the primary measure of effect. The balance between costs and effects of endovascular repair compared with open repair was expressed in terms of incremental costs per QALY gained using a societal perspective. The incremental cost-effectiveness ratio was calculated by dividing the difference in costs incurred during a 1-year period by the difference in QALYs. The incremental costs per year without severe complications and per life-year gained were estimated as secondary outcome measures.

The time horizon considered for the current analyses was 1 year. Accordingly, the time preference for any of the outcomes was considered negligible, thus obviating discounting of costs and effects.

Statistical analysis 

Patients were classified according to the original treatment allocated for all analyses (as randomized). Mortality and complication rates at 1 year were compared between the two trial arms using the χ2 test.

Because the distribution of costs across individuals tends to be skewed, calculating conventional confidence intervals (CI) is hampered because these are based on the assumption of normal distribution. Furthermore, the concept of a 95% CI pertaining to a ratio of incremental costs and effects is problematic as such. A negative incremental cost-effectiveness ratio may imply a negative cost difference (cost savings) and positive health effects, or a positive cost difference (extra costs) and negative health effects. The first would obviously be an outcome leading to immediate implementation, whereas the latter would call for immediate halt to the experimental method. Likewise, positive incremental cost-effectiveness ratios may be attained by positive cost differences and positive health effects, or negative cost differences and negative health effects. These situations might also have quite different implications; therefore, a single point estimate of average incremental cost-effectiveness with pertaining 95% CI a priori is a useless concept.

Accordingly, it has become customary in health economics to depict the joint distributions of incremental costs and effects with their pertaining uncertainty in a “cost-effectiveness plane,” where incremental costs are on the y-axis and incremental effects are on the x-axis. To evaluate the joint uncertainty in trial data while maintaining the underlying correlation between costs and effects, a bootstrap procedure may be applied.27 Patients are randomly drawn from the original trial data set, without being excluded for subsequent sampling, until the original number of participants is reached. This process is typically repeated several 1000 times, and for each trial replicate, the incremental costs and effects are calculated, to be depicted in the cost-effectiveness plane. Then, nonparametric 95% CI for the incremental costs can be calculated, and the density distribution of the incremental cost-effectiveness ratio appears in the cost-effectiveness plane.

Thus, our analysis presents costs and health outcomes as the means per patient inclusive of 95% CIs. From the original data set, 10,000 bootstrap replicates were drawn and the incremental costs for each replicate were plotted against the incremental effects representing the uncertainty surrounding the incremental cost-effectiveness ratio.

Characteristics of the patients and treatment assignments 

Details of the clinical study results were previously reported.1, 19 Briefly, between November 2000 and December 2003, 351 patients were randomly assigned to endovascular repair or open repair. Six patients did not undergo aneurysm repair after randomization: four declined treatment (1 assigned to endovascular repair and 3 to open repair), and one patient in each group died. Five patients were lost to follow-up in the first year. This resulted in a study group of 340 patients, with 170 open repairs and 170 endovascular repairs. Five crossovers occurred: one patient assigned to endovascular repair had open repair, and four patients assigned to open repair had endovascular repair.

Baseline characteristics of patients and aneurysms are given in Table I. The two groups had similar demographic characteristics, comorbid conditions, cardiovascular risk profile, American Society of Anesthesiologists classifications, and aneurysm characteristics.

Table I. Baseline characteristics⁎

	Characteristic⁎	OR (n = 170)	EVAR (n = 170)
	Age, y	69.4 ± 6.8	70.7 ± 6.6
	Male sex	154 (91)	158 (93)
	SVS/ISCVS risk-factor score, %†
	Diabetes mellitus	8.8	10.0
	Tobacco use	55.3	64.7
	Hypertension	54.7	58.2
	Hyperlipidemia	53.7	47.3
	Carotid disease	15.3	13.5
	Cardiac disease	46.5	41.2
	Renal disease	7.6	7.6
	Pulmonary disease	18.2	27.1
	Total†	4.5 ± 2.5	4.4 ± 2.5
	FEV1, L/s	2.6 ± 0.7	2.5 ± 0.7
	Body mass index, kg/m2	26.5 ± 4.0	26.3 ± 3.4
	ASA class
	I—healthy status	41 (24)	37 (22)
	II—mild systemic disease	105 (62)	119 (70)
	III—severe systemic disease	24 (14)	14 (8)
	Maximum AAA diameter, mm	60.0 ± 8.5	60.6 ± 9.0

OR, Open repair; EVAR, endovascular repair; SVS/ISCS, Society for Vascular Surgery/International Society for Cardiovascular Surgery; FEV1, forced expiratory volume in 1 second; ASA, American Society of Anesthesiologists; AAA, Abdominal aortic aneurysm.

⁎ Continuous data are presented as means ± SD, and categoric data as number (%) and percentages (because of rounding not all percentages total 100). † Mild, moderate, or severe risk-factor SVS/ISCS score ranges for each of eight domains from 0 (no risk factors) to 3 (severe risk factors).20 Total scores can range from 0 to 24, with higher scores indicating more risk factors.

End points 

In the first postoperative year, one or more severe complications occurred in 33 patients in the endovascular repair group and in 37 in the open repair group (P = .6). Death (all causes) occurred in 10 patients in the endovascular repair group and 12 in the open repair group (P = .7). Aneurysm related events occurred in two patients in the endovascular repair group and nine in the open repair group (P = .03). Detailed results were previously reported.3

Quality of life 

A marginal and nonsignificant benefit of open repair compared with endovascular repair was observed in QALYs, defined as the primary health outcome for the economic evaluation 1 year after randomization. Patients who underwent endovascular repair experienced 0.72 QALYs (95% CI, 0.29 to 1.14), and patients that underwent open repair experienced 0.73 QALYs (95% CI, 0.27 to 1.19); that is, a nonsignificant 0.01 QALY loss (95% CI, −0.038 to 0.058) was observed after endovascular repair (Fig 1).

View Large Image Download to PowerPoint

Fig 1. Utility score over time in the first postoperative year after endovascular repair (gray line) and open repair (dashed line) with 95% confidence intervals. Area under the curve equals quality-adjusted life years.

Costs 

The estimates of the main unit costs are presented in Table II. Actual averages of total costs per patient according to treatment at 1 year are presented in Table III. The estimated difference, as estimated with bootstrap replicates of the trial, amounted to €4300 per patient (95% CI, €2770 to €5830) in favor of open repair (€13,886 for open repair vs €18,179 for endovascular repair). The greater part of the cost difference was attributable to the costs of the endoprosthesis, which was only partly compensated for by savings in admission costs ensuing from shorter lengths of stay in the ICU and on the regular ward. The costs of surveillance of the endovascular repair group also contributed to the difference. Only 24 patients reported paid employment, resulting in only minimal costs associated with loss in productivity (Table III). Focusing on these 24 patients, the mean costs due to sick leave were €3820 in 11 patients (95% CI €1509 to €6131) in the endovascular repair group and €4,662 for 13 patients (95% CI, €2511 to €6812) in the open repair group.

Table II. Main unit costs based on cost study

	Variable	Unit costs (€)
	Preoperative
	Workup, inclusive examinations and visits	493
	Intraoperative
	Operation room per hour	15
	Anesthesia, excluding personnel
	OR: university hospital	436
	OR: general hospital	316
	EVAR: university hospital	206
	EVAR: general hospital	75
	Inpatient hospital days⁎
	Ward: university hospital	330
	Ward: general hospital	240
	Medium care unit	735
	Intensive care unit	1140
	Investigations
	Angiography	662
	CT angiography	183
	Duplex scanning	84

OR, Open repair; EVAR, endovascular repair; CT, computed tomography.

⁎ From Oostenbrink, et al.23

Table III. Average costs per patient per treatment group in the first postoperative year

	Variable	OR (n = 170)		EVAR (n = 170)
		Costs (€)	95% CI	Cost (€)	95% CI
	Direct costs in hospital	11,975	10,674-13,277	14,915	14,283-15,546
	Operation	5672	5416-5928	12,679	12,292-13,066
	Use of operating room	540	533-547	308	293-323
	Operating room personnel, inclusive of specialists	4256	4093-4420	3775	3579-3970
	Conventional prosthesis	488	477-500	9	0-19
	Endovascular prosthesis	193	2-385	7857	7572-8143
	Other materials	195	131-258	730	635-824
	Admission (ward + ICU)	6011	4828-7195	2136	1705-2567
	Additional costs⁎	291	108-475	100	44-155
	Direct costs during 1-year follow-up	1651	1334-1969	3618	3301-4160
	Standard follow-up	266	249-283	829	807-850
	Complications	291	265-317	1320	971-1669
	Sick leave	356	120-593	247	56-438
	Outpatient health care†	729	494-964	1295	983-1607
	Total‡	13,627	12,297-14,957	18,595	17,835-19,335
	Estimate total direct costs using bootstrap‡	13,886		18,179

OR, Open repair; EVAR, endovascular repair; CI, confidence interval; ICU, intensive care unit.

⁎ Investigations, medications, etc. † Open repair, n = 149; EVAR, n = 156. ‡ The actual trial-based absolute difference in costs is almost €5000. An accurate estimate of the absolute difference and pertaining 95% CI was obtained using 10,000 bootstrap replications of the trial. The resulting estimate of the absolute difference was €4,393 (95% CI, €2,770 to 5,830: €13,886 versus €18,179).

Cost-effectiveness 

In terms of QALYs, open repair appeared to be the dominant strategy; that is, costs were lower and better results were observed, albeit marginal and nonsignificant. The incremental cost-effectiveness ratio in terms of costs per year without severe complications was €76,100 and was €171,500 for costs per life-year gained.

Results of the bootstrapping are depicted in cost-effectiveness planes in Fig 2. The QALYs gained panel (Fig 2, A) shows that 65% of replicates fall into the left upper quadrant, indicating that we can be 65% certain that open repair yields a marginally better outcome in terms of QALYs at 1 year. Note that all replicates lie above the x-axis, indicating that with 100% certainty endovascular repair is more costly. The oblique line indicates a (rather arbitrary) societal willingness to pay a threshold of €25,000 per QALY. All replicates lie above this threshold, indicating that with 100% certainty open repair can be considered the preferred strategy at a societal €25,000 per QALY willingness to pay threshold. Note that even if a threshold of €50,000 per QALY gained would be considered acceptable, still >99% of estimates would lie above this threshold. Fig 2, B (event-free years gained) and Fig 2, C (life-years gained) show that, respectively, 95% and 85% of the replicates fell into in the right upper quadrant. This indicates that with high certainty that endovascular repair had favorable health outcomes but against higher costs.

View Large Image Download to PowerPoint

Fig 2. Cost-effectiveness planes for (A) quality-adjusted life years, (B) event-free years gained, and (C) life-years gained for costs of endovascular repair (ER) and open repair (OR). The cost-effectiveness plane consists of four quadrants. A dot to the right of the y-axis means that ER yields a better outcome, whereas a dot on the left side means the OR yields a better outcome. Likewise, a dot above the x-axis means that the costs of ER are higher, whereas a dot below the x-axis implies that OR is more expensive. The lower right and upper left quadrant would indicate that both costs and effects are favorable (dominant) for ER and OR, respectively. The percentages represent the proportion of bootstrap replication in which the clinical outcome is favorable for open repair. A, The oblique line indicates a societal willingness to pay a threshold of €25,000 per quality-adjusted life year.