Cost-Effectiveness of Two Decision Strategies for Shunt Use During Carotid Endarterectomy

Carotid endarterectomy (CEA) is a prophylactic intervention to prevent future ischemic events in patients with a symptomatic carotid stenosis. However, patients involved are exposed to a perioperative stroke or death risk of approximately 3%, of which one-third occurs intraoperatively due to embolization or cerebral ischemia during cross-clamping [1‐4]. The use of a shunt might reduce cerebral ischemia by maintaining ipsilateral flow but is still debated since it has only been shown necessary in 10–14% of patients undergoing CEA under local or regional anesthesia, which can be considered as reference standard. [5‐7]. Moreover, shunting itself is associated with complications too, including atheromatous or air emboli, arterial dissection, and acute arterial occlusion [8‐11]. Therefore, many surgeons advocate selective shunting, instead of routine shunting, in those patients at high risk of cerebral ischemia. Methods frequently used to evaluate cerebral perfusion during cross-clamping and therewith the need for selective shunting include computerized electroencephalography (EEG), transcranial Doppler (TCD), stump pressure measurement (SPM), and neurologic examination when CEA is performed under regional or local anesthesia. None of these methods has been proved to be superior with regard to intraoperative stroke risk reduction [9]. These methods do, however, differ in labor intensity and might consequently be associated with different costs and/or cost-effectiveness.

Considering the increasing costs of healthcare and decreasing health resources, costs might be taken into account when deciding which strategy to adopt. The primary aim of this study was to assess the cost-effectiveness of two modalities used to assess cerebral perfusion after cross-clamping: SPM versus combined EEG and TCD.

This study shows that SPM was less costly without a significant difference in adverse events in terms of stroke or death rate compared to EEG/TCD. The mean adjusted difference in total admission costs for CEA was €2053 and largely attributable to the perioperative neurophysiologic assessments when using EEG/TCD. Perioperative morbidity and mortality in our cohorts were comparable to those found in other CEA studies. We consider our cost-effectiveness analysis of interest. The different modalities used in the participating hospitals have been chosen for reasons in the past. SPM is less frequently used nowadays since it does not provide any information on cerebral perfusion following cross-clamping and neither accounts for the occurrence of cerebral ischemia during plaque removal and arterial reconstruction. EEG/TCD does, on the other hand, provide this information, but no benefit in terms of perioperative stroke risk reduction can be found in the literature. EEG/TCD is, however, associated with higher costs.

A reason to adopt a selective shunting strategy instead of routine shunting can be the fact that shunts are not necessarily benign and have been associated with complications [17‐20]. The shunt rate in the SPM group was quite high, probably due to the chosen systolic cutoff point of 50 mmHg as indicative for shunt insertion. Yet, the optimal stump pressure cutoff point is still controversial. Several studies attempted to determine an optimal stump pressure threshold in patients undergoing CEA under regional anesthesia and found increasing accuracy by lowering the threshold from 50 to 40 mmHg. Unnecessary shunt use decreased by 20–25%, and the erroneously non-shunted rate was kept between one and three percent [7, 8, 21]. The threshold for shunt insertion in hospital A may therefore have resulted in the unnecessary high shunt rate. Since three out of five intraoperative strokes in this study occurred in shunted patients, the threshold for shunt insertion was lowered to 40 mmHg after the study. It is, however, not certain that the intraoperative stroke in those three shunted patients was due to the shunt placement itself. The actual need for a shunt might indicate that these patients were prone for ischemia anyway Moreover, in search for the ideal stump pressure, a certain number of false-positive outcomes (unnecessary use of a shunt) have to be accepted in order to keep the false negative (erroneously not-shunted patients) as low as possible.

On the contrary, the shunt rate found in our EEG/TCD group approaches those found in studies where CEA was performed in awake patients, which is considered as reference standard [5, 7, 21]. EEG and TCD allow for continuous monitoring and can therefore, in contrast to SPM, also detect a malfunctioning shunt. Moreover, TCD provides information about the occurrence of microemboli, allowing adaptation of surgical technique and handling and might also be useful in the early postoperative phase to predict cerebral hyperperfusion syndrome (CHS) and upcoming thrombotic stroke [22‐24]. This shunt rate is, however, not necessarily worth pursuing, since low false-positive rates usually come at the expense of the sensitivity. Only few studies have determined the accuracy of EEG recordings in awake patients (reference standard). These studies show positive predictive values (PPVs) of EEG detecting true neurologic deterioration ranging from 40.9 to 90.0%. Thus, if these patients had undergone surgery under general anesthesia, a shunt would have been placed unnecessarily in 10.0–59.1% of the patients. Negative predictive values were found ranging from 94.4 to 99.2%, meaning that in 0.8–5.6% of the patients, EEG would have failed to detect neurologic deficit and a shunt would have been wrongly withheld [7, 25‐27]. PPVs and NPVs of TCD as sole modality in detecting cerebral ischemia range from 19 to 75% and 97 to 99%, respectively, depending on criteria used as indicative for shunting [28‐30]. These varying figures might reflect differences in subjective interpretations of EEG tracings and/or TCD recordings. Both techniques therefore require well-trained personnel, are time-consuming, and thus are costly.

The mean hospital costs for CEA found in our series are comparable to those found in other studies. Recently, Buisman et al. [31] determined hospital resource use and costs for ischemic stroke and TIA in the Netherlands. Costs were estimated at € 6836 ± 2862, with surgery and hospitalization (average 4.8 inpatient days) as main determinants, accounting for 51 and 34% of the total costs, respectively, similar to our series.

It is important to note that a variety of methods other than those we studied are available to determine the need for a shunt. Moreover, some surgeons prefer the routine use of a shunt, and others perform CEA under loco-regional anesthesia making neuromonitoring unnecessary. In the Netherlands, a selective shunting strategy is used in the large majority of CEAs (>90%). EEG is most frequently used (43%), followed by EEG combined with TCD (40%). SPM is only used in 1.6% of all CEAs [4]. A recent meta-analysis could not demonstrate a clinical benefit of one strategy above the other in terms of 30-day death or stroke rate [9]. Both routine shunting as well as selective shunting, whatever modality used, seem to be acceptable.

There are several limitations to our study. First, due to its retrospective character, this study had to rely on completeness of existing data registries. Several patients were excluded due to missing data (Fig. 1). Furthermore, cost analysis is dependent on the accuracy of registration of resources used during the admission. Under-registration can lead to erroneous lower costs, but this may have occurred in both groups. Moreover, since there was a rather large difference in costs between the two groups, unequal under-registration is not likely to affect the outcome of the study. Under-registration or differences in definition of certain complications might also explain the differences found in non-fatal strokes, deaths, nerve injuries, and bleeding. Second, the results may have been confounded by the fact that the two strategies were performed in different hospitals, although both the participating hospitals are very comparable midsize teaching hospitals using similar guidelines and standards. Third, both cohorts consist of relatively small numbers of patients, with even smaller numbers of adverse events. The sample size is too small to rule out a type II error for the stroke or death outcome between the SPM and EEG/TCD cohorts. Therefore, we can only make cautious statements regarding the true effect of both strategies in terms of stroke or death rate. The rates found in our series do, however, correspond with those found in the literature. Fourth, as discussed earlier, the cutoff point for shunt insertion in the SPM cohort was quite high. While excessive shunt use is beneficial for training purposes, it may increase the intraoperative stroke risk. We do not know whether more neurologic events would have occurred if a lower value had been used, which, in case, would affect the results of the cost-effectiveness analysis. Finally, this study only addresses SPM and EEG/TCD as decision-making modality. There are, however, many more strategies advocated. Inclusion of other modalities too would be of value when looking for the most cost-effective strategy.

In conclusion, although this study is limited by its small sample size and retrospective nature, the primary clinical outcomes found are comparable to those in the literature. There is, however, a significant difference in admission between both strategies. Therefore, SPM might, although nowadays virtually abandoned for reasons, still be considered as a modality to indicate the need for shunting from a cost-effectiveness point of view.