Evaluation of cefuroxime concentration in the intrathecal and extrathecal compartments of the lumbar spine—an experimental study in pigs

Optimal antibiotic prophylaxis is crucial to prevent postoperative infection in spinal surgery. Sufficient time above the minimal inhibitory concentration (fT > MIC) for relevant bacteria in target tissues is required for cefuroxime. We assessed cefuroxime concentrations and fT > MIC of 4 μg·ml−1 for Staphylococcus aureus in the intrathecal (spinal cord and cerebrospinal fluid, CSF) and extrathecal (epidural space) compartments of the lumbar spine.


| INTRODUCTION
Surgical site infection (SSI) following spinal surgery occurs with an incidence of around 3%, as reported in a large systemic review, and is frequently caused by Staphylococcus aureus (Zhou et al., 2020).
Spinal SSI is usually seen in the extrathecal tissues, presented as spondylodiscitis or epidural abscess/empyema (Babic & Simpfendorfer, 2017;Darouiche, 2006), but occurs occasionally in the intrathecal tissue compartments, that is, the cerebrospinal fluid (CSF) and spinal cord.
The bacteria are derived from either haematogenous spread or direct inoculation (Lener et al., 2018). Incidental durotomy is a common complication during lumbar spine surgery with an incidence ranging between 0.2% and 20%, increasing with age and revision (Alluri et al., 2020;Hassanzadeh et al., 2021). Durotomy exposes the intrathecal compartments for direct inoculation of bacteria and is associated with an increased rate of SSI (Li et al., 2019). However, postoperative bacterial infections of the intrathecal compartments are rare but represent serious conditions associated with significant morbidity and mortality (Chan & Gold, 1998;Greenlee, 2003;Iwasaki et al., 2011;Lener et al., 2018;Lin et al., 2014;Martin & Yuan, 1996;Oordt-Speets et al., 2018). To minimize the risk of SSI, perioperative prophylactic antibiotics with both good blood-spinal and blood-CSF barrier penetration and good penetration to relevant extrathecal tissues are essential (Laban & O'Neill, 2009).
Cefuroxime is a second-generation cephalosporin effective against a wide range of Gram-positive and negative bacteria, including S. aureus, and is commonly used as perioperative antibiotic prophylaxis in spinal surgery (Shaffer et al., 2013). Guidelines on surgical prophylaxis recommend that antibiotics reach plasma and tissue concentrations above the minimal inhibitory concentration (MIC) of relevant bacteria for the duration of the procedure as a minimum (Bratzler & Houck, 2005). For cefuroxime, which bactericidal effect is best related to the time with active concentrations above the MIC (fT > MIC) (Craig, 2003), a redose is recommended after 3-4 h in long-lasting surgical procedures to maintain concentrations above relevant MICs (Bratzler & Houck, 2005). For wild-type S. aureus, the clinical breakpoint MIC of unbound cefuroxime is 4 μg ml À1 , according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST, n.d.).
Only the unbound fraction of an antibiotic is considered microbiologically active (Drusano, 2004). Microdialysis ( Figure 1) is a wellestablished method, which, in case of antibiotics, allows for continuous sampling of the unbound fraction of antibiotic in the extracellular fluid, simultaneously from several target tissues, with high temporal resolution (Joukhadar & Müller, 2005;Müller, 2002). The method has previously been used in the cervical spine to investigate cefuroxime concentrations in extrathecal tissues (Hanberg et al., 2016;Hanberg, Bue, Jørgensen, et al., 2020) and in brain tissue (Hosmann et al., 2018). No studies have previously investigated fT > MIC of cefuroxime in the intrathecal compartments of the lumbar spine using microdialysis. This experimental study, using pigs, aimed to assess fT > MIC of 4 μgÁ ml À1 in the intrathecal (spinal cord and CSF) and extrathecal (epidural space) compartments in the lumbar spine, following administration of a single dose of 1500 mg cefuroxime.  (Lilley et al., 2020). With reference to reduction, refinement and replacement (the 3Rs) (Tannenbaum & Bennett, 2015), the pigs used in this study were also included in studies assessing cefuroxime concentrations in the anterior and posterior vertebral column of the spine and inside a cannulated pedicle screw .

What is already known
• Optimal antibiotic prophylaxis is crucial to prevent postoperative infection in spinal surgery.
• Cefuroxime's bactericidal effects are best related to time above minimal inhibitory concentration (fT > MIC).

What does this study add
• fT > MIC of cefuroxime (4 μgÁml À1 ) is shorter in intrathecal than in extrathecal compartments.
• A single dose of 1500 mg cefuroxime seems inadequate as intrathecal lumbar spine surgery prophylaxis.

What is the clinical significance
• The blood-spinal and blood-cerebrospinal fluid barriers seem to compromise cefuroxime penetration.
• Alternative dosing regimens should be considered to increase fT > MIC in intrathecal compartments.
Because this explorative study is the first to investigate cefuroxime concentrations in the specified compartments, there are currently no data to predict sample size nor make appropriate power calculations. However, earlier studies investigating cefuroxime concentrations used a similar number of animals Knudsen et al., 2021).

| Study design
This is a prospective experimental in vivo porcine study with a repeated-measures design.

| Surgical procedure
Eight female pigs (74-77 kg, mixed Duroc and Danish Landrace-Yorkshire; provided by the Aarhus University Experimental Animal Farm, Aarhus, Denmark) were sedated with intramuscular Zoletil mix (tiletamin 2.5 mg ml À1 , zolazepam 2.5 mg ml À1 , butorphanol 1 mg ml À1 , ketamine 10 mg ml À1 , and xylazine 2 mg ml À1 ), intubated and mechanically ventilated, and anaesthetized by continuous infusion of propofol (4 mgÁkg À1 Áh À1 ) and fentanyl (0.008 mgÁkg À1 Áh À1 ) and kept physiologically stable, in terms of temperature, pH, CO 2 , glucose, potassium, and sodium in arterial blood. For blood sampling and pressure monitoring, an intravascular 7-French sheath was positioned in the right common jugular vein by ultrasound-guided modified Seldinger technique and in the left common carotid artery after surgical exposure.
The pigs were placed in prone position, and the vertebral arch, facet joints, and spinous and transverse processes were surgically exposed from L3 to L4 by a posterior approach. Laminectomy was performed at these levels, removing the spinous process, the posterior part of the vertebral arch, and ligamentum flavum.
Microdialysis catheters from μ-Dialysis AB, Stockholm, Sweden (63 Microdialysis Catheter), with a 10-mm membrane length and 20-kDa molecular weight cut-off were placed in the spinal cord, CSF, and epidural space ( Figure 2). With an angle of 45 to the sagittal plane in caudal direction, a catheter was inserted into the spinal cord by visual guiding using a splitable introducer. Using a 1.6-mm needle, the dura and arachnoid mater was perforated at the level of the L3-L4 intervertebral disc in caudal direction, allowing a catheter to be placed in the CSF in the subarachnoid space. During insertion of the introducer, CSF leakage was observed briefly to determine correct placement. No visible CSF leakage was observed after catheter placement. Another catheter was placed in the epidural space by advancing the tip into the fatty tissue beneath L5. Catheters were fixated to paravertebral muscle using endo clips. A senior spine surgeon performed all the surgical procedures. To reduce risk of potential bias, the surgeon did not take part in the subsequent data collection.

| Calibration and sampling procedures
Because the concentration of cefuroxime in the dialysate only will represent a fraction of the absolute tissue concentrations (relative recovery), individual calibration of the microdialysis catheters is imperative when absolute tissue concentrations are evaluated. Calibration was carried out as continuous retrodialysis by drug with an internal standard, using 5-μgÁml À1 meropenem mixed with isotonic 0.9% sodium chloride solution. This calibration method has previously been thoroughly investigated and applied in both in vitro and in vivo studies (Hanberg, Bue, Jørgensen, et al., 2020). All catheters were calibrated individually. The 107 Microdialysis Pump (μ-Dialysis AB, Stockholm, Sweden) produced a flow rate of 1 μlÁmin À1 .
Following catheter placement, a 30-min period of tissue equilibration was allowed. Cefuroxime (1500 mg) was dissolved in 0.9% NaCl F I G U R E 1 The principle of microdialysis: Microdialysis is a dynamic sampling method, which allows quantification of the unbound fraction of antibiotics in the extracellular fluid simultaneously from multiple target tissues. At the tip of a catheter, a semipermeable membrane allows for diffusion of antibiotics following the concentration gradient. The catheter is constantly perfused with the perfusate, resembling the ionic composition (0.9% NaCl) of the extracellular fluid that contains the substance of interest. The blue and red dots are an illustrative example of the diffusion across the membrane of both the internal calibrator (meropenem) and the analyte (cefuroxime) within the extracellular fluid.
(100 ml) and administered intravenously over 10 min. Time 0 of the 8-h sampling period was defined as the initiation of the infusion. From 0 to 240 min, dialysates were collected at 30-min intervals and from 240 to 480 min at 60-min intervals. In total, 12 samples were collected from each compartment during the 8-h dosing interval. The measured dialysate concentration was applied to the midpoint of the sampling intervals in the data analysis. Venous blood samples were collected at the midpoint of the 12 microdialysis sample intervals.
Finally, the animals were killed with pentobarbital.
Venous blood samples were instantly cooled (5 C) for a maximum of 8 h and centrifuged at 2500 Â g for 10 min, and plasma aliquots were isolated. Dialysates and plasma were stored at À80 C until analysis.

| Chemical analyses
The chemical analyses were performed by an author without involvement in data collection or analysis. Plasma samples and dialysates were prepared for protein precipitation with in-house reagents. First, however, using a Hamilton STARlet workstation, plasma samples were ultrafiltrated by pipetting 300-μL plasma to a 30-kDa multi-well filter plate (Pall Corporation), placing it on top of a 700-μL collection plate and centrifuging the solution for 10 min at 1520 Â g to obtain ultrafiltrate containing the unbound fraction of cefuroxime. Plasma samples that initially were measured to have concentrations above 10 μgÁml À1 (highest calibrator) were diluted after filtration. Ultrafiltrate from the plasma samples and dialysate was then prepared for protein precipitation by mixing 5-μl dialysate or plasma ultrafiltrate (from the lower collection plate), calibrator or control for 15 min at 37 C with 45-μl internal standard (0.04-μgÁ ml À1 meropenem-d6 and 0.04-μgÁ ml À1 cefuroxime-d3 in ammonium acetate 5 mM in water, pH 7.0). Subsequently, the mixed solutions (from both the plasma ultrafiltrate and dialysate samples) were precipitated by adding 150-μl acetonitrile followed by 4-min mixing at room temperature. After centrifugation for 20 min at 1520 Â g, 100 μl of supernatant was transferred to a new microtitre plate and mixed with 200-μl ammonium acetate (5 mM in water, pH 7.0).
The concentrations of unbound cefuroxime and meropenem in dialysate and unbound cefuroxime in plasma were then measured by Waters Acquity two-dimensional ultra-high-performance liquid chromatography with Xevo TQ-S tandem mass spectrometer operated in electrospray positive mode. The mobile phases consisted of 30-μM NH 4 F in water (mobile phase A) and 30-μM NH 4 F in MeOH with 10% isopropanol (mobile phase B); 10 μl of the prepared samples was then injected on a Waters XBridge C8, 3.5-μm trapping column (2.1 Â 50 mm) at 5% mobile phase B where the analytes were trapped on the column and unretained interferents were eluted to waste.
After 1 min, the trap column was switched into series with a Phenomenex Kinetex 2.6-μm biphenyl analytical column (150 Â 3.0 mm). The analytes were eluted from the trap column and refocused on the analytical column at 20% mobile phase B. Gradient elution from the analytical column was initially 20% mobile phase B for 2.5 min, linear gradient to 100% mobile phase B at 3.5 min, 100% mobile phase B from 3.5-4.4 min and then re-equilibration at 20% mobile phase B from 4.45 to 4.5 min. During the analytical gradient run, the trap column was switched back at 2.5 min and then washed with 100% mobile phase B before being re-equilibrated at 5% mobile phase B to be ready for the next sample.  Bias was assessed by analysing three samples prepared independently from the calibrators and spiked with cefuroxime and meropenem at 0.050, 5.00, and 10.0 μgÁml À1 . Mean recovery for all three samples and two compounds was within 90-110%, measured in 10 runs. The lower limit of quantification was 0.01 and 0.05 μgÁml À1 for cefuroxime and meropenem, respectively. The intermediate precision for the internal controls at four levels, determined over 43 runs, for cefuroxime was 14.2% (target 0.010 μgÁml À1 ), 9.6% (target 0.050 μgÁml À1 ), 2.6% (target 5.00 μgÁml À1 ), and 3.9% (target 10.00 μgÁml À1 ), whereas it was 16.6% (target 0.050 μgÁml À1 ), 3.9% (target 5.00 μg ml À1 ), and 5.6% (target 10.00 μgÁml À1 ) for meropenem at three levels, determined over 36 runs. The chemical analysis has previously been applied in studies examining cefuroxime concentrations Tøstesen et al., 2022).

| Data and statistical analysis
Relative recovery was calculated for each catheter and expressed as mean ± SD. The relative recovery was determined by measuring the loss of the internal standard from the perfusate. When the concentration of internal standard in the perfusate (C in ) is known and the concentration of internal standard in the dialysate (C out ) is measured, the loss (C in À C out ) can be used to calculate the relative recovery as CinÀCout Cin . Absolute tissue concentrations can then be calculated as Cout relative recovery .
Using a non-compartmental approach, pharmacokinetic parameters were estimated, including the area under the drug concentrationtime curve from Time 0 to the last measured value at 480 min (AUC 0-480 ), half-life (T ½ ), peak drug concentration (C max ), and time to C max (T max ) for each compartment for each pig. AUC 0-480 was calculated using the linear-log trapezoidal method. T ½ was calculated as ln 2 ð Þ λeq , where λ eq is the elimination rate constant, estimated by linear regression of the logarithmic transformed concentration on time, based on points between C max and the last measured value. The maximum concentration recorded was defined as C max , enabling calculation of T max .
Tissue AUC 0-480 to plasma AUC 0-480 ratio (AUC tissue /AUC plasma ) in percentage was calculated for each compartment as a measure of tissue penetration. fT > MIC for the S. aureus clinical breakpoint MIC of 4 μg ml À1 (primary endpoint) and MICs for bacteria with lower susceptibility (0.5, 1, and 2 μg ml À1 ) as reference were calculated from unbound tissue concentrations using linear interpolation.
Results are presented as means with 95% confidence interval (CI).
The pharmacokinetic parameters and fT > MIC were obtained in all four compartments in each pig, and mixed-effects models for repeated measures with compartment as fixed effect and pig as random effect were applied to analyse differences in fT > MIC and the pharmacokinetic variables (AUC 0-480 , C max , T max , T ½ , and AUC tissue / AUC plasma ) between each of the four compartments (spinal cord, CSF, epidural space, and plasma). Models were fitted using restricted maximum likelihood (Rubin, 1976), and the correction method of Kenward and Roger (2009) was applied to reduce small-sample bias. Homoscedasticity of residuals was checked by plotting them as a function of predicted values. Normality of residuals and random effects were checked by inspection of quantile plots of residuals and best linear unbiased predictors, respectively. For AUC, T ½ , C max , and T max , logarithmic transformation improved normality and was analysed correspondingly. Post hoc pairwise comparisons of the pharmacokinetic parameters and fT > MIC between compartments were made using two-tailed pairwise t tests, interpreted at a statistical significance level of 0.05. Analyses were performed using Stata, Version 16.1 (StataCorp LLC, College Station, TX, USA). The statistical analysis itself was not blinded, but data were analysed by following a pre-study specified strategy with confirmatory statistical modelling and testing

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in www.guidetopharmacology.org and are permanently archived in the Concise Guide to PHARMACOLOGY 2021/22 (Alexander et al., 2021).

| RESULTS
All pigs completed the study. Mean relative recovery ± SD was 55% ± 7% for the spinal cord, 74% ± 14% for the CSF, and 59% ± 8% for the epidural space. Mean linear concentration-time curves for each compartment are shown in Figure 3a

| fT > MIC
fT > MIC for the spinal cord, CSF, epidural space, and plasma is shown in Table 1. For MIC of 4 μgÁml À1 , fT > MIC was significantly shorter in the intrathecal compartments (58 min in the spinal cord and 0 min in the CSF) than in the extrathecal compartment (115 min in the epidural space) and plasma (123 min), which was also true for MIC of 2 μgÁml À1 . For MICs of 0.5 and 1 μg ml À1 , fT > MIC was significantly shorter in the CSF than in the spinal cord, epidural space, and plasma.
The cefuroxime concentration in the CSF failed to reach a concentration of 4 μg ml À1 in any of the pigs. There were no differences in fT > MIC between the epidural space and plasma for the investigated MICs.

| Pharmacokinetics
Mean key pharmacokinetic parameters of cefuroxime for each compartment are presented in Table 2. Penetration into the intrathecal compartments was 32% for the spinal cord and 9% for CSF, and was significantly lower than the penetration of 63% for the epidural space. AUC 0-480 and C max were statistically significantly lower in the intrathecal compartments than in the extrathecal compartment and plasma, with the lowest AUC 0-480 and C max in CSF. T ½ was significantly longer in the CSF than in the spinal cord, epidural space, and plasma. For pharmacokinetic comparison between the epidural space and plasma, only C max was statistically significantly lower in the epidural space.

| DISCUSSION
This study is the first to present descriptive data on intrathecal con- penetration of compounds into the CSF and the extracellular fluid of the spinal cord is affected by numerous factors, including molecular size, lipophilicity, ionization, plasma protein binding, transport mechanism, and eventual metabolism of the compound. Also, alterations of the barriers by inflammation or infection play a role in the ability to pass these barriers (Nau et al., 2010). In terms of drug penetration, this process is complex and not fully understood.
Nevertheless, the physiochemical properties are important determinants of drug penetration. Cefuroxime is a hydrophilic (pK a 2.96) compound with a protein binding of 30%-50% and a molecular weight of 424 gÁmol À1 (Rimmler et al., 2019;Wishart et al., 2018).
This study does not describe the complex mechanism of cefuroxime penetration through the blood-spinal and blood-CSF barriers, and the specific actions of transport through the barriers are, to the authors' knowledge, not known. However, the molecular size and ionization of cefuroxime could reduce penetration to both the spinal cord and CSF because of repulsion due to the negative charge of the blood-brain barrier (Avdeef, 2011). Active transporters, including organic anion transporters and peptide transporters, are located at the choroid plexus, whereas P-glycoprotein is an abundant transporter protein at the blood-brain barrier. The transporters represent *P < 0.05, significantly different from the epidural space. **P < 0.05, significantly different from plasma. ***P < 0.05, significantly different from the spinal cord. ****P < 0.05, significantly different from the CSF. *****Because a negative time value is illogical, these lower values of their respective confidence intervals should be interpreted as 0 (zero).
T A B L E 2 Key pharmacokinetic parameters of all eight pigs, presented as mean values (with 95% confidence intervals). Abbreviations: AUC 0-480 , area under the drug concentration-time curve from Time 0 to the last measured value at 480 min; AUC tissue /AUC plasma , tissue AUC 0-480 to plasma AUC 0-480 ratio; C max , peak drug concentration; CSF, cerebrospinal fluid; T ½ , half-life; T max , time to peak drug concentration. *P < 0.05, significantly different from the epidural space. **P < 0.05, significantly different from plasma. ***P < 0.05, significantly different from the spinal cord. ****P < 0.05, significantly different from the CSF. an outward transport system, potentially decreasing the unbound concentration of cefuroxime in the CSF and spinal tissue (Nau et al., 2010). However, this is merely speculative because no studies have investigated the transport mechanisms of cefuroxime in the blood-spinal and blood-CSF barriers. Interestingly, local drug concentration differences within the CSF have previously been found, probably due to the constant directional flow of the CSF, originating from both the choroid plexus and the extracellular fluid of the central nervous system (CNS), which might pose a problem when comparing CSF regions (Shapiro et al., 1975). This might also in part explain the low concentration, as well as the prolonged T ½ , of cefuroxime in the CSF.
The present study showed poor penetration of cefuroxime into the CSF (9%) and spinal cord (32%), suggesting different penetration mechanisms of the two barriers. By means of lumbar puncture, multiple earlier studies found a similar low penetration of cefuroxime into the CSF in uninflamed meninges but higher in the presence of inflammation, owing to a disruption of the intercellular junctions and an increased blood flow (de Lange & Danhof, 2002;Kossmann et al., 1996;Nau et al., 2010;Sullins & Abdel-Rahman, 2013). Furthermore, the blood flow to the spinal cord might have an effect on tissue partitioning of drugs; that is, increased flow may result in increased penetration, although this may not apply to cefuroxime, given its hydrophilicity. Studies have shown that cerebral blood flow has a modest (if any) effect on drug transport across the blood-brain barrier, and animal studies suggest that regulation of spinal cord blood flow is similar to that of brain (Saleh & de Lange, 2021). However, internal variations of blood flow exist within the spinal cord, including a high segmental flow in the cervicothoracic junction (C5-T2) and in the lumbar segments (L1-L5) (Høy et al., 1994). Further complexity may be seen in anaesthetic situations, although it has been suggested that anaesthesia does not affect the spinal cord blood flow when mean arterial pressure, PaO 2 and PaCO 2 are kept constant (Høy et al., 1994). However, studies investigating CSF shunting in the awake state and in sleep-promoting states, as during anaesthesia, have shown a possible effect of brain activity on drug distribution (Mestre et al., 2020). Further studies should investigate the effects of anaesthesia, of an ongoing infection or of inflammation, on cefuroxime penetration.
In contrast to the numerous studies investigating cefuroxime concentrations in the CSF, only two studies have investigated cefuroxime concentrations in the extracellular fluid of the brain (Hosmann et al., 2018;Tsai et al., 1999). Dose recommendations have been developed using CSF concentrations as a surrogate marker for concentrations in the brain (de Lange & Danhof, 2002;Nau et al., 2010).
However, as illustrated by our results, the CSF and brain or spinal tissue should not be considered as one homogenous compartment. How substances are exchanged between the CSF and the extracellular spinal tissue compartment is complex and not fully understood (Nau et al., 2010 To reach sufficient concentrations in the CSF, early administration due to the prolonged T max as well as higher doses or alternative dosing regimens should be considered. Spinal surgery is generally long-lasting and thus requires a longer fT > MIC. A double dose of 1500 mg cefuroxime (administered with a 4-h interval) has been found superior to a single dose of 3000 mg in terms of fT > MIC in a cervical porcine model (Hanberg, Bue, Jørgensen, et al., 2020). Continuous infusion has demonstrated higher therapeutic effectiveness (Kasiakou et al., 2005) and superior coverage in terms of fT > MIC in bone and subcutaneous tissue (Tøttrup et al., 2019). Finally, a cefuroxime-loaded lipid nanoemulsion delivery system has been proposed to increase penetration into the CNS (Harun et al., 2018).
Whether re-dosing, continuous administration, or alternative delivery systems of cefuroxime during spinal surgery would sufficiently increase fT > MIC in intrathecal compartments remains to be investigated.

| Study limitations
Even though pigs have been used in previous microdialysis studies assessing cefuroxime concentrations in the cancellous bone of the cervical spine (Hanberg et al., 2016;Hanberg, Bue, Jørgensen, et al., 2020) and have a high resemblance in terms of anatomy and physiology (Busscher et al., 2010), the translational potential of our findings can be challenged. Earlier studies have shown that cefuroxime concentrations tend to be lower in pigs than in humans, possibly as a result of the age of the pigs (usually juveniles), as well as blood flow and cefuroxime metabolism being different from those in patients . Also, differences in body mass and relative compartment size may result in tissue concentration differences, because large individuals are likely to have greater distribution volumes than pigs.
This study simulated a perioperative prophylactic setup imitating a common surgical spine procedure on healthy tissue. However, for practical reasons, cefuroxime had to be administered after surgery and catheter placement, as opposed to a clinical setting where administration should be prior to surgical incision (Stulberg et al., 2010). It cannot be ruled out that the local tissue inflammatory reaction, caused by the catheter placement and surgery, influenced the local tissue concentrations. However, catheter placement in intrathecal compartments may reflect trauma extent and local tissue reaction with an incidental durotomy (Hamberger et al., 1991). A similar experimental setup in a clinical setting with application of the method in humans would be difficult and has ethical issues, thereby supporting the use of an animal model in the present study.
In conclusion, this study found that an intravenous single-dose of 1500 mg cefuroxime provided shorter fT > MIC (4 μg ml À1 ) in the intrathecal compartments, than in the extrathecal compartment of the lumbar spine and plasma with an fT > MIC of approximately 1 h in the spinal cord and twice as long in the extrathecal compartment and plasma. CSF failed to reach a concentration of 4 μgÁml À1 . Furthermore, a lower intrathecal penetration to the spinal cord and CSF than an extrathecal penetration in the epidural space of the lumbar spine was found, which suggests a compromising effect of the blood-spinal and especially the blood-CSF barrier. This effect seems to be particu-