Early thrombolysis in ischemic stroke patients reduces the area of cerebral infarction and improves the functional outcome after three months.1 The benefit of such treatment decreases on a continuous basis with the time elapsed from symptoms onset2; administration therefore should be carried out as soon as possible.

In Spain, different autonomous regions have developed plans for the administration of thrombolysis in ischemic stroke. These strategies often disadvise thrombolysis in centers lacking a Stroke Unit, and recommend transferring the patient to a reference center.3–6 Unfortunately, this recommendation is not based on the best available evidence but on the opinions of experts, which are hampered by potential conflicting professional interests, and fails to take into consideration the added delay in treatment implied by patient transfer.

Quality evidence has recently become available on the efficacy of thrombolysis according to the time from symptoms onset,2 thereby allowing a rational approach to the problem, from the patient perspective.

The present study involves decision analysis to evaluate the relative effectiveness of transfer to a reference center vs on-site treatment of ischemic stroke in patients amenable to thrombolytic treatment.

The analysis was based on a decision tree with two alternatives: on-site treatment vs referral to a reference hospital (Fig. 1). The primary clinical outcome (terminal nodes of the decision tree) was the recording of a favorable neurological outcome (defined as a score of 0–1 on the modified Rankin scale) after three months (Table 1).

Figure 1.

Decision tree. T1: Symptoms-treatment decision time: Triangular (135/45/270). T2: Secondary transfer time: Triangular (60/30/120).

Po: A priori probability of a favorable neurological outcome: Triangular (0.34/0.25/0.45). Px: Penalization due to lack of a Stroke Unit: Triangular (0.3/0/0.5). In the case of on-site treatment, the odds ratio is penalized by the variable Px. In the case of treatment in a reference center, the delay is penalized by the transfer time.

(0.08MB).

The association between the efficacy of thrombolytic treatment and the time elapsed from symptoms onset was based on the study published by Lees et al.,2 a metaanalysis of 8 randomized clinical trials estimating the effectiveness (odds ratio) of thrombolysis according to the time to treatment. To this effect, we adjusted a geometric function to the data of the mentioned metaanalysis, based on linearized models. The resulting function showed an excellent data fit (r2=0.9787):

With the purpose of evaluating the possible impact of the lack of a Stroke Unit, the OR in the case of on-site thrombolysis was penalized by a fictitious variable (Px), according to the following expressions:

The OR values were transformed into relative risk (RR) values by means of the following expression7:

The practical meaning of these expressions is shown in Table 2.

The number needed to treat (NNT) was calculated as follows:

The impact of treatment delay and of penalization due to the lack of a Stroke Unit was assessed by means of sensitivity analysis. To this effect, the expected value of each strategy was recalculated for different values of these variables, checking whether the optimum strategy changed or not.8 However, traditional sensitivity analysis is not very practical for jointly assessing the effect of more than two variables. As a result, the deterministic sensitivity analysis was completed by a Monte Carlo simulation9,10–a probabilistic sensitivity analysis taking into account all the variables subject to uncertainty simultaneously. To this effect, instead of directly including the values of the variables in the decision tree, it is assumed that each variable included in the model has a probability distribution. In our case the variables were: the baseline prevalence of a favorable neurological outcome without thrombolytic treatment, the interval from symptoms onset to treatment, the transfer time to the reference center, and the penalization due to the lack of a Stroke Unit. These variables in turn were transformed into triangular distributions with most frequent/minimum/maximum values of 0.34/0.25/0.45; 135/30/160; 60/45/120; and 0.3/0/0.5, respectively. These four distributions were randomly sampled in 10,000 simulations, registering the number of cases in which each alternative was of choice. The end result offers a compact measure of the uncertainty of the results associated to the probabilistic nature of the input variables.

The baseline scenario assumed a prevalence of favorable neurological outcomes in the control group (i.e., not subjected to thrombolysis) of 34%, an interval from symptoms onset of 135min (similar to the data of the SIT-MOST study11), and an additional delay due to transfer to the reference center of 60min. Under these conditions, the expected prevalence of favorable neurological outcomes was 45.3% in the on-site treatment group vs 41.3% in the patients transferred to the reference center (number needed to treat [NNT]=25).

The relative benefit of on-site treatment was greater during the first hours following symptoms onset (Fig. 2). Accordingly, the NNT varied from 10 (45min after symptoms onset) to 44 (270min).

Figure 2.

Sensitivity analysis of the variable symptoms-treatment decision time. The prevalence of a favorable outcome decreases as the time from symptoms onset to thrombolysis treatment decision increases. In the considered time range, the expected prevalence of favorable outcomes is always greater with the on-site treatment strategy. The difference between the two strategies is greater at the start of the clinical condition, and decreases over time.

(0.17MB).

The sensitivity analysis showed that within the time range considered, when the penalization of on-site treatment is under 17%, the treatment of choice is on-site thrombolysis.

In the Monte Carlo analysis, biased against on-site treatment through a 30% reduction in effectiveness, on-site therapy was found to be superior to patient referral in 77.2% of the cases.

Many studies show Stroke Units to improve the clinical outcome of ischemic stroke patients compared with conventional treatment in the ward.12–14 Unfortunately, however, the number of such Units is limited, and sometimes transfer to the reference center cannot be made in an acceptable period of time. Under such conditions it is necessary to decide between patient transfer to a Stroke Unit (with the associated additional delay) or on-site treatment in any of its variants (admission to Intensive Care,15 telemedicine-assisted thrombolysis,16,17 thrombolysis and referral–“drip and ship”18–etc.).

In the absence of randomized clinical trials establishing face-to-face comparisons of these two strategies, our study involves a decision analysis based on the best available evidence regarding the impact of a delay in thrombolysis upon the clinical outcome, and biased in favor of patient transfer to a Stroke Unit. According to the results obtained, the treatment of choice in a baseline situation similar to that described in different epidemiological studies11,13 is on-site patient treatment. Transfer to another center would only be justified when the effectiveness of on-site thrombolytic treatment is taken to be far less effective than that administered in a hospital with a Stroke Unit.

This latter assumption does not appear reasonable, since it contradicts the evidence derived from observational studies that support the following conclusions: (1) thrombolysis in patients without signs of bleeding in the computed axial tomography study is safe19; (2) following a brief training intervention, the clinical results of thrombolysis obtained in centers with physicians not specialized in stroke are comparable to those reported by the published clinical trials11; (3) the neurological outcome is not associated to initial care provided by a neurologist20; (4) the bleeding complications of thrombolysis increase when the door-to-needle interval exceeds 60min.21

These results would advise urgent modification of some of the Spanish regional plans, authorizing “stroke teams” to administer fibrinolytic agents on-site after establishing a correct indication, guaranteed through adequate training and accreditation of the clinicians and/or remote assistance provided by an expert.

The data obtained also point to the need for the health authorities to be sufficiently flexible to adapt to the local realities and to take all the available resources into account–independently of the limits imposed by the specialty as such.

The authors declare no conflicts of interest.