Main findings
This is, to our knowledge, the first study to investigate the relation between CPPopt and CBF in SAH patients. CPPopt could be calculated in adjunct to at least one Xe-CT scan in 78% of the patients and in 60% of all Xe-CT scans performed, using data from a 4-h time window. The fact that CPPopt could not be calculated in all patients is of course a limitation, but the numbers are reasonably high to be of use clinically. ΔCPP was negatively correlated with CBF% <10 and CBF% <20, meaning that patients with actual CPP < calculated CPPopt had higher numbers of low-flow regions (Table
3, Fig
1). The effect was not attributed to the CPP level per se, as CPP itself was not correlated with any of the CBF parameters in any of the time windows (Tables
2 and
3). This suggests that CPP in SAH patients should be individualized according to the status of autoregulation and that low CPP in relation to the calculated optimum should be avoided. The findings were consistent in both the early (day 0–3) and late (day 4–14) time windows, but were more clear when data from all monitoring times were used, possibly because of the larger sample size. In the two time windows with smaller sample size, actual CPP above CPPopt was associated with a lower number of regions with CBF <10 ml/100 g/min. This is clinically important since CBF 10 ml/100 g/min is very close to ischemic levels [
19,
30].
The correlations found were only weak to moderate, and CBF is of course determined by many factors other than the difference between CPP and CPPopt. CPPopt however seems to be important to consider in the clinical setting since it provides an available measure that may be possible to use to individualize treatment according to the status of autoregulation.
It appears from the results of this study that keeping the actual CPP at the level of calculated CPPopt or even higher reduces the risk of regional hypoperfusion at critical CBF levels (<10 ml/100 g/min), while actual CPP levels below the calculated CPPopt increase the risk. The lack of association between CPP values per se and CBF may of course be a statistical artifact (type II error), but suggests that CPP need not be high, as long as it is high enough for the individual patients. It should also be noted that the dispersion of CPP values was not great (Table
1), and if low CPP values were more common, there would probably have been a correlation between CPP per se and CBF.
Since the present study was not an outcome study, the effect of the relation between CPP and CPPopt on clinical outcome is not known. Hypoperfusion after SAH has however been associated with unfavorable outcome, and optimization of CBF is probably of great importance in the clinical management [
25].
Methodological issues
When calculating CPPopt, a few methodological aspects are important to consider. One is craniospinal compliance. In situations with high craniospinal compliance (large capacity to compensate for added intracranial volume), the relation between changes in cerebral blood volume (vasodilatation and vasoconstriction) and ICP may be less pronounced. Changes in MAP will then not be reflected in ICP changes regardless of the status of pressure autoregulation, and PRx will be less reliable. This may be the case after decompressive craniectomy or when CSF is drained via an open ventricular catheter drainage system. In the ventricular drainage system used at our institution, however, a small rubber valve at the outflow tube keeps the CSF from flowing freely, and the oscillations of the ICP curve are still visible even with an open drain. There is evidence that PRx calculations are valid with open ventricular drains as long as the ICP curve has a normal configuration [
1]. Furthermore, differences found in PRx calculated with open/closed ventricular drains are probably small and clinically insignificant [
17]. In this study, we therefore opted to use PRx data also from time periods with open ventricular drains.
Another aspect is the time window used to calculate CPPopt. CPPopt calculated from long time periods (i.e., data from all monitoring time) produces robust associations with outcome [
26]. The status of autoregulation, and therefore optimal CPP, may however be different at different time points [
27]. This approach is therefore not applicable if CPPopt is to be used for clinical decision-making. In this study we used a 4-h moving time window for CPPopt calculations. This approach has proved feasible in TBI and may be valuable in the clinical setting for continuous monitoring and customization of treatment [
1].
A third methodological aspect is the algorithm used to calculate CPPopt. Different methods have been described [
1,
26,
31]. In the original report by Steiner et al., a U-shaped curve was required, and CPPopt was taken as the lowest point on that curve [
26]. This method allowed for estimation of CPPopt in 60% of patients, which is in accordance with the results of this study. Other approaches have allowed different curve shapes as well or simply taken CPPopt as the lowest PRx in a given CPP interval [
1,
31]. In this study, CPPopt was taken as either the lowest PRx on a U-shaped curve or the lowest PRx on an ascending or descending convex curve. In case of an ascending or descending convex curve, it was assumed that the range of CPP was too small to form a U-shaped optimization curve, and a U-shaped curve would be found with a wider CPP range.
Another consideration is how vasospasm may affect PRx. Vascular constriction at different levels of the vascular tree may affect autoregulation in different ways. The calculation of PRx may also be affected in an unpredictable way, and the PRx values may be invalid with both false high and low values making interpretation difficult [
18]. Calcium antagonists (Nimodipine) are routinely used in SAH and improve outcome [
13]. However, its impact on autoregulation is not clear. Calcium antagonists have a direct effect on smooth muscle cells and in one study on healthy subjects have been shown to affect CBF and flow velocity in intracerebral arteries [
28]. Another study on SAH patients found that Nimodipine affected one autoregulatory index (ORx) but not PRx [
8]. A limitation of the present study is that vasospasm was not assessed routinely with transcranial Doppler or angiography. The effect of vasospasm on the measurement of CPPopt could therefore not be evaluated. However, it is unlikely that vasospasm occurred during the early period day 0–3 and the results were consistent for all periods.
Retrospective studies, such as the present, are susceptible to bias. For example, the inclusion of subjects may be different if subjects are recruited continuously by predetermined criteria. Confounding is another common problem in retrospective studies, i.e., the effects are caused by unknown factors that are not measured. Since the data in this study were collected systematically in a prospective fashion and not known to the investigators at the time of study design, some of the inherent limitations of a retrospective study are overcome.
Finally, studies with a small sample size, such as the present study, are common in clinical research and have some caveats regarding significance testing. A reduced ability of detecting a true effect is intuitive, but there is also an increased risk of significant findings being false positives [
6,
23]. Increasing the statistical power by increasing the sample size decreases these risks. In clinical research, however, this is not always possible. The present study is to be considered hypothesis generating rather than hypothesis testing. The statistical methods were chosen to be as robust as possible, and we do not make any claims about effect sizes. The results must still be interpreted with some caution.