Influence of Lumbosacral Cerebrospinal Fluid Density, Velocity, and Volume on Extent and Duration of Plain Bupivacaine Spinal Anesthesia

The current study was conducted at the Self Defense Force Hanshin Hospital in Hyogo, Japan, and was approved by the Hospital Ethics Committee. Written informed consent was obtained from each patient before participation in the study. There were 41 patients with American Society of Anesthesiologists physical status class I who were scheduled to undergo orthopedic lower limb surgery with a thigh tourniquet and plain bupivacaine spinal anesthesia. Apart from the usual contraindications to spinal anesthesia, patients with obvious spinal postural abnormalities (kyphosis) or neurologic disturbances were excluded from the study.

Low thoracic and lumbosacral axial MRIs for the measurement of CSF volume and velocity were obtained using an MRI system (Excel Art; Toshiba, Tokyo, Japan) operating at 1.5 T. Posterior–anterior lumbar spine radiographs were used for the identification of the spinal level, marking the intersection of a line joining the iliac crests. MRIs and radiographs were obtained a few days before anesthesia. Low thoracic, lumbar, and spinal axial MRIs were obtained at 8-mm increments with a fast-spin echo sequence, which highlights CSF, using the modified method reported by Hogan et al.  11Briefly, technical specifications included a 10,000-ms repetition time, a 160-ms echo time, a 15-cm field of view, a 160 × 256-image matrix, and 3-mm slices at 5-mm intervals. The dural sac and spinal cord areas were determined by a radiologist (J. H.), who was blinded to the purpose of this study, for each image using a digital analysis system that came equipped with the MRI system. The area of the sac minus the area of the cord constituted the area of the CSF and roots; this area multiplied by 8 mm resulted in the CSF/root volume. The level of the disk between the eleventh and twelfth thoracic vertebrae was determined, and the CSF/root volume was measured caudal from this site. Although this calculated volume includes the roots, it is hereafter referred to as the CSF volume (fig. 1, A and B).

Measurements of CSF velocity at the L3–L4 level were derived from phase-contrast MRI by scans oriented approximately perpendicular to the CSF direction of flow, using the modified method reported by Greitz et al.  5Briefly, technical specifications included a 30-ms repetition time, a 16-ms echo time, a 128 × 256-image matrix, and 5-mm slices (fig. 2A). Electrocardiography triggering was used for prospective gating of the acquisition. Cardiac gating produced a series of phase-contrast images at different cardiac phases. The CSF velocity (CSFV) profile at this site shows systolic (CSFV^systolic) and diastolic (CSFV^diastolic) components. From these phase-contrast images, a blinded investigator (J. H.) measured peak CSFV^systolic, CSFV^diastolic, and average CSFV at the L3–L4 level (fig. 2, B and C).

Thirty minutes before transfer to the operating room, all patients received an intramuscular injection of atropine (0.5 mg). After placement of standard noninvasive monitoring devices, a lumbar puncture was performed using a 25-gauge pencil-type needle (Pencan®; B Braun, Tokyo, Japan) at the L3–L4 level, with the patient in the lateral decubitus position with the operated side up. The L3–L4 level interspace was identified by counting the spines of the vertebrae and palpation of the iliac crest. After obtaining free CSF reflux, 2 ml CSF was sampled, immediately stored in closed tubes to avoid evaporation, and frozen at −80°C. Spinal injection of 3 ml (15 mg) plain bupivacaine, 0.5% (4-ml vial, 0.5% Marcaine®; AstraZeneca, Osaka, Japan), with the needle aperture directed toward the nondependent (operated) side was then performed over 15 s, and the patient was immediately turned to the supine position and remained horizontal until the end of the operation. The extent of the sensory block was assessed by pinprick (23-gauge) on the skin on the midline from top to bottom of the body up to T12 and on the nondependent side thereafter. For practical reasons, motor block was only evaluated on the nonoperated side by the previously described modified Bromage scale (0 = able to move hip, knee, ankle, and toes; 1 = unable to move hip, able to move knee, ankle, and toes; 2 = unable to move hip and knee, able to move ankle and toes; 3 = unable to move hip, knee, and ankle, able to move toes; 4 = unable to move hip, knee, ankle, and toes). 4Hemodynamic data (mean arterial pressure, heart rate) were also recorded. Data sampling was performed every 5 min for the first 30 min after spinal injection and then every 15 min until the end of the observation period, which was defined as regression of the sensory block level to L5. After complete motor block was obtained, patients were encouraged to urinate every 15 min. Time to urinate was recorded. These data were recorded by the nurses of the operating room and the orthopedic ward in charge of the patient, who were unaware of the purpose of the study, and who were instructed to accurately report the sensory block level and the degree of the motor block. Acetated Ringer’s solution, 5 ml/kg, was administered at 1 h before spinal anesthesia, and 1 ml · kg⁻¹· h⁻¹was administered during and after anesthesia. Ephedrine (5 mg) was administered intravenously when mean arterial pressure decreased by more than 30% of the baseline value. A decrease in heart rate to less than 45 beats/min was treated with 0.5 mg intravenous atropine. The relation between peak sensory block level and maximal decrease in mean arterial pressure during the 60 min after spinal injection was calculated.

Cerebrospinal fluid density with a resolution of up to four digits at 37°C was measured using a volumetric method. 12All samples were warmed to 37.0°C in a water bath before weight measurements with an electronic balance, Ribror R AEU-210 (Shimadzu, Tokyo, Japan). These measurements were performed in triplicate for each sample, and mean values were calculated. The limit of detection of the balance used in this study was 0.0001 g/ml.

The patient sample size of the current study was determined by power analysis (α= 0.05, β= 0.20) to reveal a significant correlation coefficient. Power analysis indicated that 37 patients were required to obtain a significant correlation coefficient, assuming that the two variables were continuous data and the correlation coefficient between the CSF volume and onset time of complete motor block was 0.45, which was based on a preliminary study. Continuous data were expressed as mean ± SD, and discrete data were expressed as medians with ranges. Statistical correlation coefficients (ρ) were assessed by Spearman rank correlation among the CSF variables such as volume and density; peak sensory block level to pin prick; time until development of peak sensory level to pin prick and complete motor block; time required for peak sensory block level to regress across two segments; time required for pin prick analgesia to regress to the L1, L2, L3, L4, and L5 dermatomes; time until complete motor recovery; time until spontaneous voiding; and maximal decrease in mean arterial pressure. A P  value of less than 0.05 was considered significant. In addition, multiple linear regression analysis was used to examine the relative importance of patient and CSF variables to the aforementioned measures of spinal anesthesia. Age, height, weight, body mass index (BMI), CSF density, CSF volume, peak CSFV^systolic, CSFV^diastolic, and average CSFV were considered independent variables. Multicolinearity among the variables can hinder the interpretation of results; therefore, forward and backward stepwise selections were used to identify the independently associated variables. For adding and deleting variables, the F ratio criterion was 4.0, which is the squared value obtained from a t  test for the hypothesis that the coefficient of the variable in question equals zero. 13 

Demographic and CSF characteristics of all patients are presented in table 1. The volume of lumbosacral CSF ranged from 20.5 to 61.6 ml. Although the CSF velocity profile in the lumbar spine at the L3–L4 level was more irregular than previously reported in the cervical and thoracic region, 5–9the systolic and diastolic components are still shown (fig. 2D). The caudal systolic flows started at 90–420 ms. The caudal systolic and cephalad diastolic velocities varied from 0 to 2.7 cm/s and 0.1 to 1.6 cm/s, respectively. The average CSF velocity was 0.1 cm/s, which was caudally directed.

None of the patients needed general anesthesia or adjuvant anesthesia as a result of unsuccessful spinal anesthesia. All patients developed complete motor block on the nonoperated side. No patient required urinary catheterization for urinary retention. Only one patient was given intravenous ephedrine (10 mg) because of a decrease in mean arterial pressure of more than 30% from the baseline value. The main spinal anesthetic and hemodynamic data are summarized in table 2.

Correlations between the values of CSF characteristics and the measures of spinal anesthesia, and among the measures of spinal anesthesia, assessed by Spearman rank correlation, are presented in table 3. There was a positive correlation between CSF density and peak sensory block level (ρ= 0.33, P = 0.034;fig. 3A). There was no correlation between CSF density and other measures of spinal anesthesia (table 3). Lumbosacral CSF volumes inversely correlated with the peak sensory block level (ρ=−0.65, P < 0.0001;fig. 3B) and positively correlated with the onset of motor block (ρ= 0.42, P = 0.008;fig. 4A). There was no significant correlation between peak CSFV^systolicand measures of sensory and motor anesthesia. Average CSFV correlated with onset of complete motor block (ρ= 0.38, P = 0.015;fig. 4B). There was a significant inverse correlation between peak CSFV^diastolicand the duration of motor blockade (ρ=−0.44, P = 0.005;fig. 5).

Cerebrospinal fluid volume positively correlated with time required for the sensory level to regress from the peak block level across two segments (ρ= 0.31, P = 0.048;fig. 6A), whereas CSF volume inversely correlated with regression to L1 (ρ=−0.35, P = 0.026;fig. 6B), regression to L2 (ρ=−0.33, P = 0.033;fig. 6C), and time to spontaneous voiding (ρ=−0.41, P = 0.010;table 3). There were significant inverse correlations between peak sensory block level and onset time of the sensory block (ρ= 0.35, P = 0.029) and between peak level and onset of the motor block (ρ= 0.44, P = 0.006;table 3). Peak sensory block level inversely correlated with time required for the sensory block level to regress from the peak block level across two segments (ρ= 0.56, P = 0.0004;fig. 6Dand table 3) and the decrease in mean arterial pressure from baseline (ρ= 0.43, P = 0.007;table 3). In contrast, the peak sensory block level positively correlated with BMI (ρ=−0.33, P = 0.035) and time required for regression to L1 (ρ=−0.53, P = 0.001;fig. 6E), L2 (ρ=−0.56, P = 0.0004;fig. 6F), and L3 (ρ=−0.49, P = 0.002;table 3). Average CSFV correlated with regression to L1 (ρ=−0.32, P = 0.045) and L2 (ρ=−0.33, P = 0.038;table 3).

There was no significant multiple linear regression coefficient between each variable and the time of regression to L3 and L4 as dependent variables. The significant multiple linear regression coefficients are shown in table 4. Multiple regression analysis revealed that weight and CSF volume significantly contributed to the peak sensory block level (predicted peak sensory block level = 4.48 − 0.08 × weight + 0.18 × CSF volume; R  ²= 0.46, P < 0.0001;table 4). CSF volume was the only significant predictive variable for time to spontaneous voiding (R  ²= 0.15, P = 0.007). BMI and peak CSFV^systolicwere significant predictive variables for the time required for the sensory block level to regress from the peak block level across two segments (R  ²= 0.29, P = 0.001). BMI and average CSFV were significant predictive variables for the time required for the regression of the block to L2 (R  ²= 0.15, P = 0.016) and for the complete motor block onset time (R  ²= 0.29, P = 0.001). BMI was the only significant predictive variable for onset time of the sensory block (R  ²= 0.30, P = 0.001) and for the decrease in mean arterial pressure (R  ²= 0.25, P = 0.001). Peak CSFV^diastolicwas the only significant predictive variable for the duration of motor block (R  ²= 0.23, P = 0.001). Although average CSFV and peak CSFV^diastolicwere the only significant predictive variables for the time required for regression of the block to L1 and L2, respectively, each R  ²was quite low (table 4).

Many studies have investigated the factors that influence the extent of plain local anesthetic solutions because of the inability to predict the peak sensory block level of spinal anesthesia with plain anesthetic solutions. These studies reported that dose of local anesthetics, 14,15orientation of spinal needle aperture, 16or site of injection influence the extent of plain local anesthetics, 17whereas speed of injection 18or patient demographic data 19did not. In the current study, we injected an identical dose of plain bupivacaine with the needle aperture directed toward the nondependent (operated) side. Therefore, it is unlikely that the dose, site of injection, or technique influenced the results of the current study.

The lumbosacral CSF volume in the current study, 41.7 ml, was comparable with that obtained in previous studies, 36.3–53.7 ml. 3,20The MRI technique used in this study (3-mm-thick slices, two-dimensional fast-spin echo) was similar to that described by Hogan et al.  11Although this method was validated in a previous study by measurement of known volumes and reliable duplication with repeated measurement, 11Lee et al.  20recently reported a new method using three-dimensional fast-spin echo, in which slice spatial resolution (1-mm sagittal partitions) was improved. Although the CSF volume measured by these MRI techniques included the cauda equina, this contributes to only approximately 15% of the total volume of the dural sac at this level. 21 

After spinal anesthetic solution is injected in the CSF, the dilution of spinal anesthetic solutions by CSF occurs before arrival at effect sites in the central nervous system. Consistent with the results obtained by Carpenter et al. , 3who demonstrated a strong relation between CSF volume and peak sensory block level after the administration of hyperbaric lidocaine (r =−0.91), there was a significant relation between CSF volume and the peak sensory block level in the current study (ρ=−0.65, P < 0.0001). In addition, CSF volume correlated with the onset time of the motor blockade, which is not consistent with the findings of Carpenter et al.  3These results suggest that CSF volume is the important factor influencing the extent of plain bupivacaine. Further, it is also generally believed that after dilution occurs, the distribution of hyperbaric solutions away from the site of injection is governed mainly by the effects of gravity, whereas plain solutions are not. 2The results of the current study confirmed those of the Schiffer et al. , 4who reported that higher densities are associated with a higher cephalad level. Taken together with the results of CSF volume, the current study indicated that the lower the CSF volume is and the higher the CSF density is, the higher the peak sensory block levels for plain bupivacaine solutions are.

After an intrathecal injection of a local anesthetic, plain local anesthetic is more likely to remain where it was injected, and a concentration gradient develops with a progressive decrease in anesthetic concentration away from the injection site. 22Therefore, more cephalad afferent nerves within the thecal sac would encounter increasingly dilute concentrations of plain bupivacaine, resulting in an earlier ebb of the peak sensory level. Indeed, the time required for the sensory block level to regress from the peak block level across two segments inversely correlated with the peak sensory level in the current study, i.e. , the higher the level of the spinal block is, the shorter time to regress across two segments is. On the other hand, in the lumbar region where local anesthetic remains, the local anesthetic would be more diluted by a larger lumbosacral CSF volume, resulting in a shorter duration of lumbar sensory anesthesia. The results of previous studies by Carpenter et al.  3and Urmey et al.  16can be explained by this mechanism. Carpenter et al.  3demonstrated that the lower peak sensory block levels are and the greater the CSF volume is, the shorter the duration of tolerance of transcutaneous electronic stimulation at the ankle (L5–S1) is. Urmey et al.  16reported that orientating the aperture of the needle directly caudally prolonged the duration of sensory anesthesia and motor block and time to urinate. They speculated that this prolongation is due to a greater amount of the local anesthetic being concentrated near the lower lumbar and sacral nerve roots where the injection is directed caudally. We demonstrated that the lower sensory block level is associated with a greater CSF volume and less time required for regression to L1–L3, and CSF volume inversely correlates with the regression of the block to L1 and L2 and time to spontaneous voiding. Therefore, this mechanism might explain the recovery profile of the current study. The lack of a correlation between CSF volume and the regression to L3–L5 might be related to the high degree of interindividual variability at these segments. Recent autopsy and microscopic studies indicate that as the root size at the segments of L3–S2 increases, the interindividual variability in root size also increases. For example, the posterior L5 area, the most variable root, ranges from 2.2 to 7.4 mm³. 21 

Time to regression of the block to L1, L2, and L3 correlated with peak sensory level, i.e. , the higher the level of the spinal block was, the more time was required for regression of the block to L1, L2, and L3 in the current study. These correlations between the peak sensory block level and time of regression to L1–L3 contradict the results obtained by Schiffer et al. , 4who demonstrated an inverse correlation between peak sensory block level and the time of regression to L4, i.e. , higher cephalad levels are associated with a shorter time for the block to regress to L4. It is difficult to explain this discrepancy. The protocols of the two studies were identical except for the measurement of sensory block. The extent of sensory block was assessed by pinprick in the current study, whereas Schiffer et al.  4used ether drops. However, the difference in the measurement of the sensory block does not account for the discrepancy because loss of temperature discrimination occurs one or two dermatomes higher, but its onset and regression closely parallel those of the sensory level assessed by pin prick. 19Further study is required.

On the basis of the assumption that the direction of flow of the 500 ml CSF produced daily is mainly intracranial from the site of production in the choroid plexus to the principal sites of vascular absorption through intracranial arachnoid villi and that downward flow of CSF through the spinal subarachnoid space is minimal, involving perhaps less than 10% of the 500 ml CSF produced, the subarachnoid space is classically portrayed as a “limited pool” of CSF gently bathing the delicate neuraxial structures. 2,10However, recent MRI observations reveal that CSF vigorously oscillates with arterial pulsations. 5–9The amplitude of CSF movement is approximately 0.9 cm/beat in the cervical (C6–C7) CSF and approximately 0.4 cm/beat at the thoracolumbar junction, with minimal movement in the distal lumbosacral sac. The net flow from the site of production of CSF to the absorption site at the C2 level is 0.02 ml/beat, which is only 2% of the CSF pulsatile flow. 5Oscillatory CSF pulsation is a possible but unexplored mechanism for local anesthetic distribution after subarachnoid injection. 10The CSF pulsations might be an important factor in the distribution and clearance of spinal agents because the epidural venous plexus flow, which probably has a significant role in the absorption and generation of CSF pulsations, is closely coupled to CSF flow. 7Because CSF oscillation is also amplified with increased intraabdominal pressure, this effect might be an important means by which the anesthetic effect is extended in pregnant and obese patients. 10 

In the current study, the mean systolic and diastolic velocities were 1.1 (0–2.7) and 0.8 (0.1–1.6) cm/s, respectively, similar to those obtained in previous reports. Greitz et al.  5reported that the systolic velocities varied from 0 to 2.0 cm/s at the L3 level. Enzmann et al.  6reported that the mean systolic and diastolic velocities just anterior to the conus were both 1.0 cm/s. Although there was no significant relation between peak CSFV^systolicand measures of sensory and motor anesthesia, there was a significant relation between peak CSFV^diastolicand the duration of motor blockade and between average CSFV and duration of sensory blockade. These findings suggest that CSFV influences the extent and duration of plain bupivacaine spinal anesthesia. However, the CSF velocity findings in the current study must be interpreted cautiously. We used electrocardiography triggering for prospective gating of the acquisition in the current study. The disadvantage of prospective gating is that the acquisition is stopped within approximately 200 ms of the next R wave for accurate direction of the next trigger. Therefore, the diastolic phase is not fully recorded, 23,24and the diastolic and average velocities in the current study might be inaccurate. In addition, there might be other methodologic inaccuracies, such as the influence of arterial venous waveform amplitude. Improved flow analysis with new MRI technology will reveal greater detail regarding the patterns of CSF motion and factors that influence the extent of the movement. 23,24 

Other potential study limitations include the following: Although we assumed that the spinal puncture was performed at the L3–L4 level, we must consider the possible influence of the injection site because of the uncertainty of the injection level. 25,26The possible influence of spinal puncture on the peak sensory level was discussed in the article by Schiffer et al.  4In the current study, 2 ml CSF was sampled in all patients. Although 5 ml CSF, but not 3 ml, removed before spinal injection has an impact on the spread of spinal anesthesia with plain bupivacaine, 27the effect of 2 ml CSF removal on the spread and duration of spinal anesthesia might not have been equivalent in all patients because there was great interindividual variability of CSF volume with a threefold difference. 3Finally, there are inaccuracies in the volumetric method used in the current study 12compared with the well-established accuracy (± 0.00001 g/ml) and high level of precision of the mechanical oscillation technique for the measurement of density. 28,29Schiffer et al.  4, who demonstrated a highly significant relation between CSF density and peak block level, measured CSF density with a resolution of up to six digits using a densitometer, whereas we measured only four digits using a less accurate volumetric method. In addition, the temperature control during CSF density measurements in the current study was also less precise (37.0°C). Strict temperature control (37.00°C) during CSF density measurements is required for increased accuracy. 28,29Therefore, reduced accuracy and precision of mass, volume, and temperature measurements limit the findings of CSF density in the current study.

Multiple linear regression analysis in the study of physiologic systems provides a powerful statistical tool for sorting out the relations among several variables. Although all selected predictive values must ideally be uncorrelated, there are no completely independent variables among age, height, weight, BMI, CSF density, CSF volume, peak CSFV^systolic, CSFV^diastolic, and average CSFV. Rather, it is important to select parameters that produce an R  ²close to 1.0 in multiple linear regressions. Multicolinearity among variables makes regression analyses difficult to interpret. Therefore, we used forward and backward selection to identify the most useful variables for predicting the extent and duration of plain bupivacaine spinal anesthesia. 13,30Although CSF density correlated with the peak sensory block level by Spearman rank correlation (table 3and fig. 3A), according to multiple linear regression modeling, it was not a predictor of the peak sensory block level (table 4), possibly because of the weak correlation between CSF density and the peak block level because of inaccuracies in the CSF density measurement methods.

Although our findings provide valuable insight into the mechanism of plain bupivacaine spinal anesthesia, they will not be helpful for guiding clinical practice. For example, the R  ²of the peak sensory block level was 0.46, which means that only 46% of the peak sensory block level can be explained with our selected parameters of weight and CSF volume. Further, CSF volume is not always measured before anesthesia. On the other hand, BMI can routinely be calculated before anesthesia. However, multiple linear regression analysis indicates that the correlation between BMI and onset time of sensory block is statistically significant (P = 0.001) but poorly predictive (R  ²= 0.30). Therefore, our results cannot be expected to improve the ability to predict the spread or duration of plain bupivacaine spinal anesthesia in clinical practice, although they provide a better explanation of the unpredictability of extent and duration of plain bupivacaine spinal anesthesia.

In summary, the lower the CSF volume is with a higher CSF density, the higher the peak sensory block level is for plain bupivacaine solutions, indicating that CSF density and volume are main factors that influence the spread of spinal anesthesia with plain bupivacaine. The current study also indicates that CSF volume is associated with the duration of the sensory block. The results of the current study also suggest that CSF velocity influences the extent and duration of plain bupivacaine spinal anesthesia.

The authors thank the surgeons and nurses in the operating room and orthopedic wards of Self Defense Force Hanshin Hospital (Hyogo, Japan) for their cooperation. They also thank the workers in the Departments of Radiology of their hospitals.