Cost‐Effectiveness of Ultra‐Early Tranexamic Acid as Add‐On to Standard Care After Subarachnoid Hemorrhage (ULTRA Trial)

ABSTRACT

Objective

To evaluate the cost‐effectiveness and cost–utility of adding ultra‐early and short‐term administration of tranexamic acid (TXA) to standard care in patients with subarachnoid hemorrhage (SAH).

Materials and Methods

An economic evaluation was performed alongside the ultra‐early tranexamic acid after subarachnoid hemorrhage (ULTRA) trial. The main outcomes were the incremental cost‐effectiveness ratio (ICER), expressed as costs per one‐point increase in modified Rankin scale (mRS) score, and the incremental cost–utility ratio (ICUR), expressed as costs per quality‐adjusted life‐year (QALY). Cost‐effectiveness acceptability curves (CEACs) were visualized with varying ICER cut‐offs. Bootstrapping techniques and sensitivity analyses were performed to account for uncertainty.

Results

The ULTRA trial included 955 patients, with 480 assigned to the TXA group and 475 to the control group. The mean mRS score was 3.4 (95% CI: 3.2–3.5) in the TXA group and 3.2 (95% CI: 3.0–3.4) in the control group. The mean QALY was 0.26 (95% CI: 0.24–0.28) in the TXA group and 0.28 (95% CI: 0.26–0.30) in the control group. Mean costs were €62,180 (95% CI: 57,589–66,913) in the TXA group and €58,624 (95% CI: 53,693–63,955) in the control group. The probability of treatment with TXA being cost‐effective ranged from 4% to 16% for mRS and from 8% to 16% for QALYs.

Conclusions

Ultra‐early and short‐term administration of TXA to patients with SAH is not cost‐effective. Therefore, we recommend against using TXA for this patient group.

Trial Registration: Netherlands Trial Register: NTR3272. ClinicalTrials.gov identifier: NCT02684812

1. Introduction

Subarachnoid hemorrhage (SAH) accounts for 5% of strokes and is characterized by a high morbidity and case fatality rate (approximately 25% have a favorable outcome, and case fatality in Europe is estimated at 44%) [1, 2]. The global incidence of SAH is 7.9 per 100,000 person‐years [3]. SAH often impacts individuals who are relatively young, with the incidence peaking between 50 and 60 years [1]. Therefore, the economic and societal burden of SAH entails both financial strains on healthcare systems and the loss of productive life‐years [4]. To the best of our knowledge, recent real‐world data on the costs of SAH are very scarce and no cost‐effectiveness or cost–utility analysis has been conducted in conjunction with an RCT [5, 6, 7, 8, 9].

The ultra‐early tranexamic acid after subarachnoid hemorrhage (ULTRA) trial was a multicenter randomized open‐label trial comparing SAH patients treated with ultra‐early and short‐term administration of tranexamic acid (TXA) prior to aneurysm occlusion as add‐on to standard care to SAH patients treated with standard care without additional TXA [10]. The primary aim was to investigate whether treatment with TXA improves clinical outcome, measured by the modified Rankin scale (mRS). At 6‐month follow‐up, the study demonstrated that TXA did not improve clinical outcome. Although there is no demonstrated advantage of TXA in terms of clinical outcome, it is important to study the impact of the intervention on societal and direct medical costs from a cost‐effectiveness perspective. TXA may still offer cost savings or reduced resource utilization, making it potentially worthwhile from an economic standpoint. A cost‐effectiveness analysis can provide valuable insights into the trade‐offs between the cost of treatment and the outcomes achieved, which is crucial for decision making in healthcare policy and resource allocation.

The aim of this study was to assess the cost‐effectiveness and cost–utility of ultra‐early and short‐term TXA in addition to standard care versus standard care alone in the ULTRA trial from a societal perspective up to 6 months after inclusion [10].

2. Methods

2.1. Study Design and Participants

This economic evaluation was performed alongside the multicenter, randomized, controlled, open‐label ULTRA trial, which was approved by the ethics committee of the Amsterdam UMC, location Academic Medical Center (AMC), and registered in the Netherlands Trial Register (number NTR3272) and ClinicalTrials.gov (number NCT02684812). The rationale and design of this trial and the 6‐month follow‐up have been reported elsewhere [10, 11]. In summary, 955 patients from eight SAH treatment centers and 16 referring hospitals met the inclusion criteria (age ≥ 18 years, SAH proven by CT within 24 h after the last hemorrhage). In all, 480 patients were randomized to the treatment group: standard care according to the national guidelines and additional administration of an intravenous bolus of 1 g TXA directly followed by continuous intravenous infusion of 1 g TXA every 8 h until a maximum of 24 h or until aneurysm treatment. To the control group, 475 patients were randomized: standard care [12, 13]. Patients or their legal representatives gave written informed consent. The primary endpoint of the ULTRA trial was clinical outcome assessed with the mRS at 6 months after randomization. The CHEERS guidelines and checklist were used as guidance for this analyses [14].

2.2. Economic Evaluation and Effect Measures

We performed an economic evaluation from a societal perspective with a time horizon of 6 months. The analyses were conducted according to the guidelines for health economic analysis issued by the Dutch Healthcare Institute [15, 16]. Because of the 6 months study follow‐up, no discounting was used for either costs or effects.

Two analyses were performed in this economic evaluation: first, a cost‐effectiveness analysis, in which we analyzed the mRS [17] as the effectiveness outcome, and second, a cost–utility analysis, in which we analyzed quality‐adjusted life‐years (QALYs) derived from the EQ‐5D‐3L questionnaire [18, 19].

The mRS is an ordinal score, where 0 indicates no disability and 6 represents death. For this cost‐effectiveness analysis, the mean was chosen over the median as it provides a measure of central tendency that reflects the expected values of outcomes, aligning with standard practices in economic evaluations and facilitating interpretation and comparison given that most cost‐effectiveness thresholds (e.g., cost per QALY gained) and decision‐making guidelines are based on mean values. A distribution of the mRS scores is provided in Table S1 and Figures S1 and S2. The mRS was assessed at baseline and 6‐month follow‐up by a trained research nurse through a standardized telephone interview [11]. The research nurses were masked to treatment allocation.

The EQ‐5D‐3L questionnaire consists of five items: mobility, self‐care, usual activities, pain/discomfort, and anxiety/depression. Each item has three levels: (1) no problems, (2) some problems, and (3) extreme problems. The ranking of the items can be converted into a simple, generic index value that ranges from −0.329 to 1 in the Dutch value set, where 1 represents optimal health‐related quality of life. The health utility of death was set to zero. Health utilities were derived using Dutch tariffs [20]. We calculated the QALYs over the 6‐month follow‐up period by extrapolating the health utilities measured at 3 months over the period from randomization to 3 months plus the health utilities measured at 6 months over the period from 3 to 6 months.

Definitions of complications (i.e., hydrocephalus, rebleeding) have been reported in the appendix of the ULTRA trial [10].

2.3. Resource Use and Unit Costs

Three information sources were used to gather data on healthcare resources: electronic patient data management systems, case report forms, and a version of the Dutch Health and Labour Questionnaire [15], adjusted for the study population by Westendorp et al. [21], during follow‐up at 3 and 6 months (with a recall period of 3 months). In this questionnaire, patients or caregivers are asked to specify their healthcare providers, visit frequency, medication use, home adjustments, medical devices, home care, and travel expenses. All relevant healthcare costs, costs associated with home and informal care, out‐of‐pocket expenses by patients, and costs resulting from productivity loss were assessed and are outlined in Table S2. Direct medical costs included primary hospital admission (ward and ICU stay), aneurysm treatment if applicable, all additional procedures such as (neuro)surgical (re)interventions (i.e., lumbar puncture or ventricular drain), diagnostic imaging, laboratory tests, microbiological cultures, transfusions, medication, readmissions, in‐hospital and outpatient consultation visits (with a medical specialist, psychologist, occupational physician, general practitioner, physiotherapist, occupational therapist, speech therapist, dietitian or social worker), emergency department visits, ambulance costs, rehabilitation, and nursing home care. Direct nonmedical costs included data on out‐of‐pocket costs (i.e., travel expenses), over‐the‐counter medication, costs for medical devices and/or home adjustments, and costs associated with home and informal home care assessed with the costing questionnaires (at 3 and 6 months).

In 2024, the Dutch National Health Care Institute published an updated guideline for economic evaluations in health care [22]. Unit costs of used resources were based on the prices provided by the newest version of the Dutch cost manual (DCM) [22], the Dutch Healthcare Authority (NZA) [23], and the weighted average of prices derived from the hospital ledgers from the eight treatment centers if costs were not described in the DCM or NZA price lists. All medication costs were derived from the National Health Care Institute website with an official listing of drugs with prices [24]. Travel costs were determined based on average distances traveled and parking fees outlined in the DCM [22]. Prices for unpaid home care were based on shadow prices for domestic care. The friction cost method was used to estimate indirect nonmedical costs as loss of productivity. Total costs per patient were calculated by multiplying resources used by associated unit costs. Costs were expressed in Euros and price indexed to 2022 based on the Dutch consumer price index [25].

2.4. Statistical Analysis

All patients were analyzed according to the intention‐to‐treat principle. Additionally, we did per‐protocol and as‐treated analyses, which are reported in Table S3 and Figures S3–S10. Cost differences were analyzed by the independent T test or Mann–Whitney U test, depending on the distribution of the data.

We calculated incremental cost‐effectiveness ratios (ICERs) by dividing the difference in mean total costs by the difference in outcomes. We bootstrapped the costs and outcomes for 5000 samples with replacement to obtain the 95% confidence intervals (95% CI) of skewed costs and effects. Next, we visualized the bootstrapping results on a cost‐effectiveness plane, which is divided into four quadrants representing the possible conclusions: since a higher mRS score indicates worse clinical outcome, northeast = TXA is less effective than control and also more expensive, “inferior”; southeast = TXA is less effective and less expensive; northwest = TXA is more effective and more expensive; and southwest = TXA is more effective and less expensive, “dominant.” In case of QALYs, northeast = TXA is more effective than control and also more expensive; southeast = TXA is more effective and less expensive, “dominant”; northwest = TXA is less effective and more expensive, “inferior”; and southwest = TXA is less effective and less expensive. This approach allows for an assessment of TXA's effectiveness and costs relative to the control. When the quadrants of the cost‐effectiveness plane are crossed, the numerical values for the ICERs and 95% CIs have conflicting meanings. Therefore, we constructed cost‐effectiveness acceptability curves (CEACs) with varying ICER cut‐offs on the bootstraps following Fenwick et al. [26] In CEACs, instead of considering one value such as the ICER, a full range of monetary values from null to €100,000 is considered. These threshold monetary values are the “willingness‐to‐pay” (WTP): the maximum budget we would be willing to pay to gain one unit in the outcome. The purpose of the CEACs is to show the proportion of bootstrap samples in which a certain ICER (or lower, i.e., better) was found, thereby reflecting the uncertainty around cost‐effectiveness even when bootstrap samples cross all four quadrants.

All analyses were performed in SPSS version 24.0 (IBM, Armonk, New York, USA), R version 4.2.0 (R Foundation for Statistical Computing, Vienna, Austria), and RStudio (Posit, Boston, MA, USA).

2.5. Missing Data

If the costing questionnaires were not completed, no volume data regarding the use of out‐of‐hospital resources, travel expenses, and productivity losses were available. Hence, cost calculations for these patients could not be determined. For mRS, we only used complete data (mRS assessed at 6 months). For QALYs, if one of the two assessments for the EQ‐5D‐3L questionnaire was available, we set the missing assessment to the value of the available assessment. If neither EQ‐5D‐3L assessment was available, we excluded the patient from the analysis.

2.6. Sensitivity Analysis

To determine the robustness of the calculated costs, multiple sensitivity analyses were performed. The first sensitivity analysis was performed by calculating cost‐effectiveness from a healthcare perspective, taking into account only direct healthcare costs. In the second sensitivity analysis, we assessed the impact of imputation by including cases with two completed EQ‐5D‐3L questionnaires only. Finally, exploratory subgroup analyses were performed for patients with aneurysmal SAH.

3. Results

Between July 2013 and July 2019, a total of 955 patients were enrolled in the ULTRA trial. Within this group, 945 (99%) had complete mRS data at 6 months and 837 (88%) had complete EQ‐5D‐3L data at 3, 6 months, or both, allowing us to calculate the QALYs (Figure 1).

FIGURE 1.

Flowchart of study patients. mRS, modified rankin scale; TXA, tranexamic acid.

Patient characteristics of the 945 patients with complete mRS are shown in Table 1.

TABLE 1.

Patient characteristics of the ULTRA trial patients included in the cost‐effectiveness analyses.

	Tranexamic acid (n = 475)	Control (n = 470)
Male sex, n (%)	146 (30.7)	161 (34.3)
Age in years, mean (SD)	59 (12.6)	58 (12.3)
WFNS score [27], n (%) a
1	168 (35.4)	187 (39.8)
2	94 (19.8)	93 (19.8)
3	23 (4.8)	16 (3.4)
4	92 (19.4)	93 (19.8)
5	88 (18.5)	78 (16.6)
Type of SAH, n (%)
Aneurysmal	406 (86.2)	399 (85.3)
Nonaneurysmal	65 (13.8)	69 (14.7)
Fisher grading scale [28], n (%)
2	35 (7.4)	20 (4.3)
3	125 (26.3)	149 (31.7)
4	315 (66.3)	301 (64.0)
Hydrocephalus, n (%)	289 (60.8)	257 (54.7)
Rebleed, n (%)	48 (10.1)	66 (14.0)
DCI, n (%)	106 (22.3)	104 (22.1)
Length of stay, mean (SD)
Neurosurgery ward	15 (14.6)	14 (14.5)
ICU	7 (7.6)	6 (8.2)
Discharge destination, n (%)
Home	139 (37.5)	161 (43.6)
Rehabilitation unit	59 (15.9)	68 (18.4)
Nursing home	42 (11.3)	34 (9.2)
Mortality at 6 months, n (%)	128 (26.9)	114 (24.3)

Abbreviations: DCI, delayed cerebral ischemia; ICU, intensive care unit; mRS, modified Rankin scale; SAH, subarachnoid hemorrhage; SD, standard deviation; WFNS, World Federation of Neurosurgical Surgeons.

^a WFNS was missing in 13 patients.

3.1. Outcomes and Costs

A summary of unit costs of major resources can be found in Table S2. The cost of TXA was €2,97 per 1 g infusion, totaling €11,87 if all four doses are administered. Resource use and calculated costs are shown in Table 2. The mean costs were €62,180 (95% CI: 57,589–66,913) in the TXA group and €58,624 (95% CI: 53,693–63,955) in the control group, with a mean difference of €3556 (95% CI: −3419 to 10,545). The average cost of TXA per patient was €7,21. Table 2 shows that ICU admissions, neurosurgical ward stays, ventricular drainage procedures, and subsequent readmissions were mainly responsible for an increase in costs.

TABLE 2.

Resource use and costs per treatment group.

	Tranexamic acid		Control		Mean difference a
	Total units	Total costs (€)	Total units	Total costs (€)
Hospital admission
Neurosurgical ward	7062	4.381.433,56	6763	4.195.926,82	296,57 (−856,25–1449,39)
Intensive care unit	3094	8.128.457,00	2781	7.306.153,65	1567,53 (−1090,34–4225,41)
Readmissions	134	915.766,84	110	747.178,13	338,19 (−517,61–1193,99)
Diagnostic procedures
CT	1646	242.618,97	1412	208.127,82	67,95 (11,35–124,55)
MRI	689	168.599,09	566	138.500,77	60,26 (−14,09–134,61)
DSA	717	610.510,09	639	544.095,63	127,64 (−135,61–390,88)
Laboratory tests	—	249.490,71	—	220.992,76	55,05 (−6,27–116,36)
Microbiological cultures	4598	111.470,78	4176	101.240,17	19,27 (−22,77–61,31)
Therapeutic procedures
Endovascular coiling	183	2.162.988,88	175	2.070.946,80	−14,37 (−448,45–419,71)
Neurosurgical clipping	89	591.665,94	96	637.684,40	5,39 (−193,06–203,83)
External ventricular drain	225	754.402,72	206	670.580,20	97,66 (−216,00–411,31)
Ventricular shunt	87	440.291,63	61	324.875,38	−265,00 (−988,25–458,24)
Craniotomy	9	28.456,72	7	22.133,00	0,00 (0,00–0,00)
Craniectomy	26	55.060,28	22	44.863,93	78,43 (−93,48–250,35)
Clinical consultations
Neurologist	367	89.371,84	348	84.744,96	7,07 (−48,23–62,38)
Physiotherapist	3911	157.691,52	3674	148.135,68	15,45 (−38,15–69,05)
Occupational therapist	1179	47.537,28	1177	47.456,64	−1,32 (−22,22–19,57)
Speech therapist	636	23.309,40	658	24.115,70	−2,46 (−19,79–14,88)
Institutional care
Rehabilitation center	5948	4.864.982,61	5796	4.740.658,93	155,56 (−2945,66–3256,79)
Day care rehabilitation center	2435	870.313,09	2478	885.682,05	−52,19 (−825,04–720,65)
Nursing home	1341	374.653,19	1088	303.969,17	142,00 (−428,24–712,24)
Day care nursing home	402	32.256,82	464	37.231,75	−11,31 (−106,56–83,95)
Other costs
Emergency room	64	15.907,34	41	10.190,62	11,81 (−1,70–25,31)
General practitioner	459	13.650,66	434	12.907,16	1,28 (−5,68–8,24)
Outpatient clinic	1745	245.086,61	2096	288.092,42	−96,99 (−194,26–0,28)
Medication	—	573.158,54	—	578.907,47	−38,45 (−299,55–222,65)
Productivity loss	14,752	626.362,18	14,102	598.757,09	44,90 (−471,29–561,09)
Formal home care
Nursing	94	6.791,89	179	12.933,53	−13,22 (−66,01–39,58)
Caretaking	96	5.325,32	151	8.376,27	−6,61 (35,73–22,51)
Domestic help	356	11.235,55	89	2.808,85	17,68 (−9,26–44,62)
Informal home care	3076	55.711,82	2715	49.173,46	12,67 (−34,42–59,75)
Travel expenses	—	86.367,16	—	8.8621,27	−6,73 (−84,03–70,57)
Out‐of‐pocket	—	12.760,23	—	11.934,70	1,47 (−17,92–20,86)
Total healthcare costs (direct costs)	—	28.715.707,85	—	26.760.231,39
Total societal costs (direct and indirect costs)	—	29.552.027,62	—	27.564.232,25

Abbreviations: CT, computed tomography; DSA, digital subtraction angiography; MRI, magnetic resonance imaging.

^a Mean difference including lower and upper limits of the 95% CI.

3.2. Cost‐Effectiveness and Cost–Utility Analyses

3.2.1. mRS Scores

The mean mRS score was 3.4 (95% CI: 3.2–3.5) in the TXA group and 3.2 (95% CI: 3.0–3.4) in the control group, with a mean difference of 0.2 (95% CI: −0.1 to 0.4). In Figure 2, we visualized the bootstrap samples in the cost‐effectiveness plane. Of bootstrap samples, 78.6% appeared in the northeast (TXA is less effective than control and also more expensive, “inferior”), 14.3% in the southeast (TXA is less effective and less expensive), 5.8% in the northwest (TXA is more effective and more expensive), and 1.3% in the southwest (TXA is more effective and less expensive, “dominant”), which means that TXA was on average inferior to the control, and no ICER is applicable.

FIGURE 2.

Cost‐effectiveness plane showing differences between the TXA and control group in costs and mRS score in all 5000 bootstrap samples. Note: Since a higher mRS score indicates worse clinical outcome, northeast = TXA is less effective than control and also more expensive, “inferior”; southeast = TXA is less effective and less expensive; northwest = TXA is more effective and more expensive; and southwest = TXA is more effective and less expensive, “dominant.”

The CEAC in Figure 3 provides more information on the uncertainty surrounding the cost‐effectiveness for a range of WTP. We found that at a WTP of zero, that is, one point in mRS is valuated at nothing at all, the probability that TXA was cost‐effective as compared to control was 16%. This decreased as the WTP increased to 4% at a WTP of €100,000.

FIGURE 3.

Cost‐effectiveness acceptability curve for mRS. Cost‐effectiveness acceptability curve showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for mRS (y‐axis) when compared to a range of threshold monetary values (x‐axis).

3.2.2. QALYs

The mean QALY was 0.26 (95% CI: 0.24–0.28) in the TXA group and 0.28 (95% CI: 0.26–0.30) in the control group, with a mean difference of −0.02 (95% CI: −0.04 to 0.01). The cost‐effectiveness plane is shown in Figure 4. Of bootstrap samples, 8.8% appeared in the northeast (TXA is more effective than control and also more expensive), 1.8% in the southeast (TXA is more effective and less expensive, “dominant”), 75.2% in the northwest (TXA is less effective and more expensive, “inferior”), and 14.2% in the southwest (TXA is less effective and less expensive), indicating that TXA was on average inferior to control, and no ICER was applicable.

FIGURE 4.

Cost‐effectiveness plane showing differences between the TXA and control group in costs and EQ‐5D‐3L based difference in QALYs in all 5000 bootstrap samples. Note: Northeast = TXA is more effective than control and also more expensive; southeast = TXA is more effective and less expensive, “dominant”; northwest = TXA is less effective and more expensive, “inferior”; and southwest = TXA is less effective and less expensive.

The CEAC in Figure 5 provides more information on the uncertainty surrounding the cost–utility. We found that at a WTP of zero, that is, a QALY is valuated at nothing at all, the probability that TXA was cost‐effective as compared to control was 16%. This decreased as the WTP increased to 8% at a WTP of €100,000.

FIGURE 5.

Cost‐effectiveness acceptability curve for QALY. Cost‐effectiveness acceptability curve showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for QALYs (y‐axis) when compared to a range of threshold monetary values (x‐axis).

3.3. Sensitivity Analysis

Calculating cost‐effectiveness from a healthcare perspective (taking direct medical costs only) or excluding the imputed cases did not change the results compared to the primary analysis. Furthermore, stratification for aneurysmal SAH did not increase the probability of TXA being cost‐effective (Table 3). Cost‐effectiveness planes and CEACs can be found in Figures S11–S20.

TABLE 3.

Summary of results for the sensitivity analyses. For comparison, the base case analysis is at the top.

Sensitivity analysis	Sample size	Incremental effect	Incremental costs (€)	ICER (€)	Dominance (TXA more expensive and less effective) (%)	Probability of cost‐effectiveness at €20,000
Base case analysis	945 837	mRS: 0.19 QALYs: −0.02	3556 3900	N/A N/A	78% 75%	6% 14%
Direct medical costs only	945 837	mRS: 0.18 QALYs: −0.02	3512 3910	N/A N/A	79% 76%	5% 14%
Complete case analysis	660	QALYs: −0.02	5317	N/A	77%	9.2%
Stratification by aneurysmal SAH	805 715	mRS: 0.16 QALYs: −0.02	734 640	N/A N/A	52% 46%	20% 41%

Abbreviations: ICER, incremental cost‐effectiveness ratio; mRS, modified Rankin scale; N/A, not applicable; QALY, quality‐adjusted life‐year; SAH, subarachnoid hemorrhage; TXA, tranexamic acid.

4. Discussion

The primary conclusion of the ULTRA trial was that ultra‐early and short‐term administration of TXA does not improve clinical outcome in SAH patients at 6‐month follow‐up. This economic evaluation of patients included in the ULTRA trial not only confirms that ultra‐early and short‐term administration of TXA in addition to standard care of SAH patients is ineffective, but also shows that it is more expensive. At 6‐month follow‐up, the overall societal costs of SAH patients in the TXA group were €29.552.027 compared to €27.564.232 in the control group. This means the costs in the TXA group were approximately €3556 higher per patient, which is not attributable to the costs of TXA itself since these costs are approximately €12 per patient. Resources that seem to play a key role in this economic difference are ICU admissions, with additional elevated costs associated with admission to the neurosurgical ward, surgical procedures, and subsequent readmissions, all of which were slightly more prevalent in the TXA group. The underlying cause for the prolonged ICU admissions in the TXA group remains unclear. However, multiple analyses indicate a trend toward worse outcomes in the TXA group, which could explain the increased costs [29, 30]. The exact reason for the poorer outcomes in the TXA group is not fully understood, but one hypothesis is that TXA may hinder the recovery from early brain injury [10, 29, 30]. Results from the ULTRA trial revealed a higher incidence of seizures in the TXA group compared to the control group (12% vs. 8%; odds ratio 1.52, 95% CI: 1.00–2.33) [10]. Especially patients with poor‐grade SAH experienced a significantly higher incidence of seizures in the TXA group compared to the control group [30]. Patients who experienced seizures may have required longer ICU stays or more frequent ICU admissions. Since we observed only slight differences in medication use, antiseizure medication does not seem to be a key contributor. Surgical procedure costs were also higher in the TXA group, which may be due to the more frequent need for external ventricular drains or permanent shunts in this group (48% vs. 41% and 14% vs. 10%, respectively). A recent study by Tjerkstra et al. suggested that this could be attributed to reduced breakdown of blood clots in the subarachnoid space, which may hinder cerebrospinal fluid absorption and result in hydrocephalus [30].

4.1. Sensitivity Analyses

Despite some minor differences from the main analysis in terms of the probability of being cost‐effective, TXA was generally not cost‐effective, demonstrating the robustness of the results.

In the main analysis, we used data imputation if one or both EQ‐5D‐3L scores were. missing. Very similar results were found for the complete case analysis, also implying robust results. Moreover, the mean EQ‐5D‐3L index value within this cohort is relatively low. In another study presently under review, a median index of 0.78 was discovered at 6 months after SAH; this is likely due to this study's assignment of a health utility value of zero to death.

4.2. Strengths and Limitations

A major strength of this study is its randomized multicenter design with a large study population of over 900 patients. Furthermore, this is the first economic evaluation of TXA in SAH patients using real‐world cost data. The study was conducted from a societal perspective and includes a 6‐month follow‐up period, accounting for the high costs associated with rehabilitation and nursing home admissions. All resource use was documented prospectively, and healthcare consumption was valued at the level of individual units rather than diagnosis‐related groups. Missing values were imputed relatively conservatively, by setting the missing EQ‐5D‐3L score to the available assessment value if only one was available, thereby avoiding overestimation of intervention effects common with multiple imputations [31]. Finally, the pragmatic design of the ULTRA trial reflects everyday clinical practice.

Our study has several limitations. First, while we accounted for productivity loss due to absenteeism from work, we did not include costs associated with impaired productivity while at work. Additionally, the EQ‐5D‐3L is a self‐reported questionnaire, which may have introduced measurement bias through socially desirable answers and recall bias. Another concern is the response rate to the EQ‐5D questionnaire during follow‐up, which ranged from 76% to 81% among SAH survivors. Nevertheless, this response rate remains relatively high for this population, as such percentages are typically lower in similar studies. Moreover, the ULTRA trial was not specifically powered to detect differences in cost‐related or patient‐reported outcomes. Furthermore, the study was open label, potentially biasing patients or physicians regarding treatment allocation, which could influence outcomes and costs. Finally, resource use and costs were based on Dutch reference prices, limiting the generalizability of our findings to other countries where SAH treatment practices may significantly differ.

5. Conclusion

Our study demonstrates that ultra‐early and short‐term administration of TXA in addition to standard care in SAH patients results in an excess societal cost of 2 million euros compared to standard care. In addition to previous studies showing no benefit in functional outcome, this study concludes that TXA is also not cost‐effective. Therefore, TXA should not be administered to SAH patients.

Author Contributions

W.V.A., D.V.E., R.P.O., and N.D.E. conceived this study. M.G.E., R.P.O., B.C.O., GRI, W.V.A., and D.V.E. developed the ULTRA protocol. The ULTRA investigators, recruited patients and contributed to data collection. N.D.E. and R.E.E. performed the analyses. N.D.E. drafted the manuscript. R.E.E., R.P.O., D.V.E., and W.V.A. helped draft the manuscript. All authors reviewed and approved the final manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1. Distribution of mRS scores.

Figure S1. Distribution of mRS at baseline.

Figure S2. Distribution of mRS at 6‐month follow‐up.

Table S2. Unit costs of major resources used.

Table S3. Summary of results for the as‐treated and per‐protocol analyses. For comparison, the base case analysis is at the top.

Figure S3. Cost‐effectiveness plane showing differences between the TXA and control group in costs and mRS score based on an as‐treated protocol.

Figure S4. Cost‐effectiveness acceptability curve based on an as‐treated protocol showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for mRS (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S5. Cost‐effectiveness plane showing differences between the TXA and control group in costs and EQ‐5D‐3L based QALYs based on an as‐treated protocol.

Figure S6. Cost‐effectiveness acceptability curve based on an as‐treated protocol showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for QALYs (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S7. Cost‐effectiveness plane showing differences between the TXA and control group in costs and mRS score based on a per‐protocol analysis.

Figure S8. Cost‐effectiveness acceptability curve based on a per‐protocol analysis showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for mRS (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S9. Cost‐effectiveness plane showing differences between the TXA and control group in costs and EQ‐5D‐3L based QALYs based on a per‐protocol analysis.

Figure S10. Cost‐effectiveness acceptability curve based on a per‐protocol analysis showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for QALYs (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S11. Cost‐effectiveness plane showing differences between the TXA and control group in costs and mRS score from a healthcare perspective (direct medical costs only).

Figure S12. Cost‐effectiveness acceptability curve from a healthcare perspective (direct medical costs only) showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for mRS (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S13. Cost‐effectiveness plane showing differences between the TXA and control group in costs and EQ‐5D‐3L based QALYs from a healthcare perspective (direct medical costs only).

Figure S14. Cost‐effectiveness acceptability curve from a healthcare perspective (direct medical costs only) showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for QALYs (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S15. Cost‐effectiveness plane showing differences between the TXA and control group in costs and EQ‐5D‐3L based QALYs for patients with complete cases only.

Figure S16. Cost‐effectiveness acceptability curve for complete cases only showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for QALYs (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S17. Cost‐effectiveness plane showing differences between the TXA and control group in costs and mRS score for patients with aneurysmal SAH.

Figure S18. Cost‐effectiveness acceptability curve for patients with aneurysmal SAH showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for mRS (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Figure S19. Cost‐effectiveness plane showing differences between the TXA and control group in costs and EQ‐5D‐3L based QALYs for patients with aneurysmal SAH.

Figure S20. Cost‐effectiveness acceptability curve for patients with aneurysmal SAH showing the proportion of bootstrap samples that were found in TXA to be cost‐effective for QALYs (y‐axis) when compared to a range of threshold monetary values (x‐axis).

Denneman N., Post R., van Eekelen R., et al., “Cost‐Effectiveness of Ultra‐Early Tranexamic Acid as Add‐On to Standard Care After Subarachnoid Hemorrhage (ULTRA Trial),” European Journal of Neurology 32, no. 8 (2025): e70208, 10.1111/ene.70208.

Data Availability Statement

The authors have reported all relevant data used to conduct the research. One author (NDE) had full access to all the data in the study and takes responsibility for its integrity and the data analysis. All data requests should be submitted to the corresponding author (NDE) for consideration. Access to anonymized data may be granted following review.