Body temperature affects cerebral hemodynamics in acutely brain injured patients: an observational transcranial color-coded duplex sonography study

Data were collected at two times: with fever (defined as a T_core ≥38.4°C [9]) and after defervescence (a temperature reduction of at least 0.7°C) induced by antipyretic drugs, acetaminophen or diclofenac (Table 1). Eight patients had ICP monitoring and we calculated the cerebral perfusion pressure (CPP) as:

where MAP is mean arterial pressure. In six patients the internal jugular bulb was cannulated for intermittent determination of oxygen saturation in the jugular vein (S_jO₂).

Patients in this group suffered an unexpected drop in T_core during general anesthesia. Data were collected immediately when they entered the NICU and when the temperature target, a T_core increase ≥0.7°C, was achieved by active warming. We used a convective air warming system connected to a full-body blanket (WarmTouch™ Convective Air Warming System, Covidien, Boulder, CO, USA). ICP and S_jO₂ were not monitored in this group. All patients had uneventful post-surgery recovery.

We used the iE33 xMATRIX Echography System (Koninklijke Philips Electronics NV, Amsterdam, The Netherlands) for TCCDS, measuring CBF-V and pulsatility index (PI) in the middle cerebral artery (MCA), at the M1 tract, the most proximal to the internal carotid artery bifurcation. The measurements were always taken at the same depth and insonation angle, with very accurate angle correction, in order to ensure they were comparable. We insonated the MCA bilaterally in 21 patients, and on the right side only in 3 patients, and on the left side in 2 patients, because there was no adequate controlateral insonation window. We insonated the MCA at a mean depth of 54 ± 3 mm on the right and 54 ± 7 mm on the left; the angle was 22 ± 19° on the right and 27 ± 13° on the left.

We used two statistical software packages: Prism 5.00 (GraphPad Software, San Diego, CA USA) and SPSS 7.0 (IBM Software, Armonk, NY, USA). All data were normally distributed (checked with D'Agostino-Pearson omnibus normality test) so they are all summarized as mean ± standard deviation.

Our analysis consisted of two steps: (1) first, the paired t-test was used to compare TCCDS measurements before and after defervescence or warming; (2) then, we built a general linear model for repeated measures, in order to verify the effect of three independent variables on CBF-V changes: insonation side (right or left), method used to modify temperature (defervescence or re-warming) and temperature itself (T_core). A probability (P-value) <0.05 was accepted as significant.

We studied patients on average 7 ± 6 days after brain injury. The mean T_core fell from 39.1 ± 0.4 to 37.9 ± 0.5°C. Cerebral monitoring showed a reduction in ICP from 16 ± 8 to 12 ± 6 mmHg (P = 0.003) (Figure 1A) although CPP did not change (Figure 1B). CBF-V decreased: mean velocity from 75 ± 26 (95% CI 65, 85) to 70 ± 22 cm/s (95% CI 61, 79) (P = 0.04) and systolic peak velocity from 144 ± 54 (95% CI 123, 166) to 126 ± 41 (95% CI 110, 142) cm/s (P = 0.002). The PI dropped from 1.36 ± 0.33 (95% CI 1.23, 1.50) to 1.16 ± 0.26 (95% CI 1.05, 1.26) (P = 0.0005) (Figure 2A).

Systemic monitoring showed a reduction in HR from 86 ± 17 to 80 ± 16 bpm (P = 0.006) and MAP from 91 ± 14 to 85 ± 15 mmHg (P = 0.02); other parameters were stable (Table 3).

On average, the T_core rose from 35.5 ± 0.8 to 36.3 ± 0.7°C. CBF-V increased: mean velocity from 36 ± 10 (95% CI 31, 41) to 39 ± 13 (95% CI 33, 45) cm/s (P = 0.04) and systolic peak velocity from 59 ± 18 (95% CI 51, 67) to 68 ± 20 (95% CI 58, 77) cm/s (P = 0.0004). The PI rose from 0.98 ± 0.14 (95% CI 0.91, 1.04) to 1.09 ± 0.22 (95% CI 0.98, 1.19) (P = 0.02) (Figure 2B).

Systemic monitoring showed an increase in MAP from 82 ± 9 to 91 ± 15 mmHg (P = 0.01) and a reduction in arterial lactate from 1.3 ± 0.8 to 1.1 ± 0.7 (P = 0.01); other parameters were stable (Table 3).

The general linear model for repeated measures analysis confirmed that the only independent variable that explained CBF-V changes was T_core (P = 0.01); the other two independent variables (insonation side and method used to modify temperature) could not predict changes in CBF-V, by Greenhouse-Geisser, Huynh-Feldt or lower-bound analysis (Figure 3).

CBF-V decreased during defervescence in patients with acute brain damage, and rose during re-warming in a T_core range from 33.7 to 39.9°C. The PI also dropped during defervescence and rose during re-warming. CBF-V varied despite constant CPP, which remained stable because the MAP reduction was concurrent to the decrease in ICP during defervescence. SjO₂, which reflects the balance between cerebral metabolic rate and delivery of oxygen, did not change during defervescence, suggesting that the relationship between metabolism and CBF was preserved. A more complex but similar explanation can be proposed for the changes in PI. There are reports of a linear relationship between the PI and ICP [10],[11] but from our data we could not distinguish whether changes in PI correlated more with T_core or ICP.

There is little information about how temperature affects cerebral hemodynamics in the clinical setting. Published data mainly focus on hypothermia and the most widely investigated setting is cardiac surgery, where a reduction in CBF-V during cooling and a parallel increase when temperature returned to baseline has been observed [12],[13]. In one study 19 patients with chronic hepatitis C virus infection were subjected to experimental therapy with extracorporeal whole-body hyperthermia at 41.8°C for 120 minutes under propofol anesthesia; there was a decrease in cerebral oxygen extraction and an increase in CBF-V [14]. These findings take the same direction as our results, but our brain-injured patients had spontaneous episodes of fever, not passive hyperthermia. Hyperthermia and fever follow different pathways, so studies on hyperthermia cannot be directly translated into the fever setting.

Findings in healthy volunteers seem to contrast with our findings. CBF-V rose when T_core was lowered by an external system in healthy volunteers; this might be explained by concomitant changes in several other parameters, like arterial carbon dioxide partial pressure, which is the most widely investigated factor that influences CBF-V [15]. Some authors [16] also investigated the effect of exercise-induced hyperthermia on CBF-V and their results too appear to contrast with ours: CBF-V fell as temperature rose. Prolonged exercise causes hyperventilation, hence hypocapnia, which can reduce CBF and consequently CBF-V. We do not know how the other parameters change.

Changes in temperature affect cerebral hemodynamics, and potentially ICP. CBF-V, which estimates CBF, increases when temperature rises. Cerebral blood volume (CBV) is likely to increase as well. In patients with reduced intracranial compliance, small changes in CBV can be very important in terms of the effect on ICP. As temperature affects CBF-V, it must be taken into consideration for the interpretation of TCCDS data, as well as arterial partial pressure of carbon dioxide (P_aCO₂) [17]-[22], sedatives/analgesics [23],[24] and induced hypertension [25], which are already known to change CBF-V. Moreover, changes in the PI parallel temperature and ICP; if this is the case, it can hardly serve as a useful basis for non-invasively estimating ICP, as previously proposed [10],[11].

Our study has some limitations. First of all TCCDS does not directly quantify CBF but only estimates it. Autoregulation changes, which are of obvious importance, could not be studied. In addition, our analysis involved a limited number of patients. Even if we tried to change nothing other than the temperature, in some cases there were small changes in certain parameters. We also could not measure T_brain but only T_core, which generally underestimates T_brain.