Share this article on:

Intrathecal Midazolam I: A Cohort Study Investigating Safety

Tucker, Adam P. MB, ChB, FANZCA, PhD; Lai, Cindy MB, BS, FANZCA; Nadeson, Raymond PhD; Goodchild, Colin S. MA, MB, BChir, PhD, FANZCA, FFPMANZCA

Section Editor(s): COUSINS, MICHAEL J.

doi: 10.1213/01.ANE.0000087075.14589.F5
PAIN MEDICINE: Research Report

Despite conflicting evidence regarding the safety of intrathecal midazolam from animal investigations, its clinical use is increasing. We investigated the potential of intrathecal midazolam to produce symptomatology suggestive of neurological damage. This study compared two cohorts of patients who received intrathecal anesthesia with or without intrathecal midazolam (2 mg). Eighteen risk factors were evaluated with respect to symptoms representing potential neurological complications. The definitions of these symptoms were made wide to maximize the chance of counting patients with neurological sequelae after intrathecal injections. Eleven-hundred patients were followed up prospectively during the first postoperative week by a hospital chart review and 1 mo later by a mailed questionnaire. Symptoms suggestive of neurological impairment, including motor or sensory changes and bladder or bowel dysfunction, were investigated. Intrathecal midazolam was not associated with an increased risk of neurologic symptoms. In contrast, neurologic symptoms were found to be increased by age >70 yr (relative risk, 8.72) and the occurrence of a blood-stained spinal tap (relative risk, 8.07). The administration of intrathecal midazolam, 2 mg, did not increase the occurrence of neurologic or urologic symptoms, as suggested by some preclinical animal experimentation.

IMPLICATIONS: Intrathecal midazolam provides segmental analgesia, but conflicting animal studies have cast doubts on its safety. This investigation studied the effect of intrathecal midazolam by observing two cohorts of patients. In clinical practice, intrathecal midazolam (2 mg) did not increase adverse neurological symptoms compared with conventional therapies.

Department of Anaesthesia, Monash Medical Centre, and Department of Anaesthesia, Monash University, Victoria, Australia

Accepted for publication June 20, 2003.

Address correspondence and reprint requests to Adam P. Tucker, MB, ChB, FANZCA, Department of Anaesthesia, Monash Medical Centre, 246 Clayton Rd., Victoria 3168, Australia. Address e-mail to

Editor's note: Please refer to the editorials by Cousins and Miller (pp. 1507–8) and Yaksh and Allen (pp. 1509–11) and the articles by Tucker et al. (pp. 1521–7), Johansen et al. (pp. 1528–35), and Yaksh and Allen (pp. 1536–45) in this issue.

Animal studies have shown that segmental analgesia is produced by intrathecal injections of midazolam (1). They have also shown that its action is mediated at γ-aminobutyric acid receptors (2). Despite intrathecal injections of midazolam having shown promise in human case reports and clinical trials (3,4), this has not been adopted widely in clinical practice because of concerns of possible neurotoxicity (5). Midazolam's potential for neurotoxicity has been studied in several animal species, including rats, cats, and rabbits. Four studies have found neurotoxicity associated with intrathecal midazolam (6–9), whereas six others have reported no adverse effects associated with its use (10–15). Differences in the methods and procedures used by these studies—such as untreated hypotension, hypotonic midazolam, and delay in postmortem tissue fixation—may explain the inconsistent findings. However, taken together, these studies have identified a dose of midazolam that is not associated with histopathology. Clinical investigation is required to determine the relevance of these animal studies to human physiology, because the technique is currently being used clinically in a number of centers.

The clinical study of adverse events resulting from regional anesthesia has been hindered by the infrequent incidence of these complications. Large surveys have shown that severe forms of neurological injury, such as cauda equina syndrome and arachnoiditis, occur rarely after spinal anesthesia (16). The large number of patients that must be studied to establish a difference between therapies has resulted in few sufficiently statistically powered studies being completed (17). Recent attention has been focused on transient neurological syndrome, which is a transient neurological dysfunction found in patients after intrathecal lidocaine (18); it represents a subtle form of neurotoxicity that is not revealed by histopathology (19). Most importantly, because transient neurological syndrome is common, it may be detected in a relatively small number of patients by using an assessment of symptoms such as that reported here.

Data from preclinical animal investigations, which suggest that intrathecal midazolam causes severe adverse effects, indicate that midazolam-induced neurological damage is common. The incidence of histopathology associated with intrathecal midazolam was reported to be 30% in an investigation conducted by Malinovsky et al. (8). The incidence in other investigations was either described as or implied to be 100% (6,7,9). Thus, if these histopathologic studies are predictive of neurological sequelae in humans, a small sample of patients may be sufficient to demonstrate an association between the clinical use of midazolam and subsequent forms of neurological damage.

In the light of these preclinical investigations and the lack of safety data for intrathecal midazolam in humans, this study sought to monitor two cohorts of patients who received intrathecal medications—either with or without intrathecal midazolam—for signs and symptoms of neurological complications.

Back to Top | Article Outline


After institutional ethics committee approval and informed patient consent were obtained, patients who received an intrathecal local anesthetic for a surgical procedure, without a concurrent general anesthetic, were enrolled in this study. Patients were enrolled in the study after the anesthetic had been given and the surgery had been performed. The choice of anesthetic technique was made by the anesthesiologist concerned in consultation with the patient and was independent of the study. The operator recorded details concerning the intrathecal injection that were thought to be risk factors for subsequent neurological damage. These included factors such as the size and type of spinal needle used, the number of attempts at intrathecal injection, and the occurrence of paraesthesia or blood in the cerebral spinal fluid. In addition, patient and surgical factors were also recorded. For each of these categories, one value or result was chosen as a reference point with which to compare others regarding the occurrence of adverse neurological symptoms. These are shown in Tables 1 and 2. Thus, ASA status I patients were compared with ASA status II, III, or IV patients, and patients with lumbar punctures at L3-4 and L4-5 were grouped together (low lumbar) and compared with patients with high lumbar punctures (L1-2 and L2-3) for adverse sequelae. Orthopedic procedures were chosen arbitrarily as the reference category with which to compare surgical procedures for the risk of adverse neurological events. Because one or two attempts at lumbar puncture were thought to carry a lesser risk of neurological damage, the results from cases reporting the number of attempts were compared with cases in which a larger number of attempts were made. As far as drugs administered in the intrathecal injection are concerned, apart from a variety of opioids, midazolam is in common use by this route, with several hundred cases performed per year. Bupivacaine, which is the most common drug used and which has a well accepted safety record, was chosen as the reference for comparison of the incidence of complications associated with the use of other drugs, including midazolam.

Postoperative complications that were considered to suggest neurological damage were collected by two methods: a review of the hospital charts concerning the first postoperative week and a questionnaire sent to the patients by mail 1 mo after surgery. No day-surgery cases were included in this study. Some patients were discharged on the fourth or fifth postoperative day, and some stayed longer than 1 wk. Thus, all patient charts were examined for all patients from the first postoperative day until the time of discharge or up to 1 wk. The questionnaire used for the 1-mo follow-up is shown in Table 3. No attempt was made to separate persistent effects of the spinal anesthetic. Data were collected from the first postoperative day, ie, 12–24 h after the administration of the spinal anesthetic, when the effects could be expected to have disappeared. Any persistent effects, whether caused by slow resolution of local anesthetic effects or neurological dysfunction from other causes, were scored as possible neurotoxic effects. If no reply to the questionnaire was received from the postal contact, the answers were obtained by a structured telephone interview. Complications during the first 7 days after surgery were predefined: back pain represented any documentation of pain present in the region of the back or radiating below the hip; numbness or weakness was represented by decreased sensation or strength of the trunk or lower limb; bladder dysfunction was represented by urinary incontinence or retention or treatment for such, including bladder catheterization; and bowel dysfunction was represented by incontinence of feces or constipation. Symptoms sought by the postoperative questionnaire included the following: the presence of back or leg pain, leg numbness, or weakness; urinary incontinence or difficulty voiding; fecal incontinence or difficulty voiding; and numbness or a burning sensation around the anal or genital area.

An association between the use of intrathecal midazolam and each risk factor was tested using the Pearson χ2 test and the Yates contingency correction test, as appropriate, and the association between the risk factors (including midazolam) and complications suggesting neurological damage was assessed with a χ2 test or Fisher's exact test. Risk factors shown to be associated with complications at the level of P = 0.10 were entered into logistic regression analyses. These analyses also included the use of intrathecal midazolam. Separate multiple logistic regression models were developed to determine the effects of these factors on the presence of complications by using SPSS (Version 10.0; SPSS Inc., Chicago, IL). The relative risk was calculated for each factor in the presence of the other factors in the model. The factors that were associated at a level of P ≤ 0.05 were considered to have a significant association with specific complications.

Back to Top | Article Outline


This study followed up 1100 patients, 547 of whom received intrathecal midazolam and 553 of whom did not. Eighteen risk factors thought to be important for adverse neurological sequelae were studied in addition to the use of intrathecal midazolam. The intrathecal midazolam and nonmidazolam cohorts were compared with respect to 17 of these risk factors (Table 4). Age exhibited a bimodal distribution and was similar in both cohorts, as judged by the Mann-Whitney U-test (P = 0.16). The median and 25th–75th percentiles for the midazolam and nonmidazolam cohorts were 72 and 71 yr and 54–78 yr and 59–79 yr, respectively. Seventy years of age was chosen as the cutoff point for age as a risk factor. When intrathecal midazolam was administered, the preparation used was isotonic, acidic (pH 3.5), and preservative free (Hypnovel; Roche). A dose of 2 mg was administered in all cases.

The occurrence of each complication showed that bladder and bowel dysfunction each occurred in 18% of the sample in the early postoperative period (Table 5). The other complications occurred less often in the early and late postoperative period (<10% of the sample). Only 18 patients did not complete the 1-mo follow-up questionnaire.

The univariate association of each risk factor with complications is shown in Tables 1 and 2. The logistic models for complications identified in the hospital chart are shown in Table 5, and the postoperative questionnaire is shown in Table 6. These models did not find that intrathecal midazolam was associated with an increased risk of complications.

An age of 70 yr or older increased the incidence of radiating leg pain in the early postoperative period (relative risk, 8.72; 95% confidence interval, 1.08–71.14; P = 0.04;Table 5). However, this association did not persist after 1 mo. Leg numbness 1 mo after surgery occurred most often in patients in whom blood-stained cerebral spinal fluid was detected by the anesthesiologist (relative risk, 8.07; 95% confidence interval, 2.24–29.04; P < 0.01;Table 6).

Acute bladder dysfunction occurred more often with orthopedic surgery (relative risk for nonorthopedic specialties, 0.33; 95% confidence interval, 0.14–0.77; P = 0.01;Table 5). One month later, urinary incontinence was associated with sex alone, with men carrying a decreased risk (relative risk, 0.3; 95% confidence interval, 0.15–0.61; P < 0.01;Table 6). The incidence of acute bowel dysfunction was less frequent in men (relative risk, 0.6; 95% confidence interval, 0.43–0.85; P < 0.01;Table 5) and in patients aged older than 70 yr (relative risk, 0.62; 95% confidence interval, 0.44–0.88; P < 0.01;Table 5), but these were not associated with any of the studied risk factors 1 mo later.

Back to Top | Article Outline


The major finding of this study is that neurological complications were increased most profoundly by the demographic and surgical factors; the exception to this rule was traumatic spinal injections that resulted in blood-stained cerebral spinal fluid. Furthermore, this study found that 2 mg of midazolam given intrathecally did not increase the occurrence of symptoms suggestive of neurological damage compared with conventional therapies.

Although animal toxicity studies have yielded conflicting results, this study is consistent with those studies that found that intrathecal midazolam did not cause histopathology and differed from those that suggested it did. Previous authors have stated that the primary outcome of preclinical trials is the determination of the dose of a trial drug that is not associated with adverse outcomes, because even naturally occurring peptides and amino acids that are important to the normal functioning of the nervous system can result in neurotoxic effects when administered in large concentrations (20). Whereas preclinical animal trials have demonstrated histopathology associated with the use of intrathecal midazolam, none has done so by using the equivalent dose used in this study (0.02–0.05 mg/kg). In addition, animal studies have found that this dose has not been associated with the alteration of behavioral measures other than those consistent with antinociception, and this study confirms these findings in humans.

It would seem that some measures used in animal studies are more sensitive than others in demonstrating potential neurotoxicity. Histological investigations have suggested that lidocaine is toxic only at large concentrations but not at concentrations used clinically (19). However, the smaller concentrations that are used clinically have been found to result in altered neurophysiology (21) and abnormal clinical symptoms in humans (18). The difficulty with animal studies, therefore, is the translation of their findings into the equivalent human situation. At present, animal histopathologic models appear to only crudely approximate their clinical corollary. The primary limitation of animal studies may be the inability of the animal subjects to elaborate on the subjective aspects of pain and other symptoms and, therefore, provide more detailed information. Furthermore, the lack of a standard animal model has resulted in many competing techniques that have provided a diverse and conflicting body of knowledge. In contrast, patients are able to provide extensive and relevant information regarding the subjective aspects of their experience.

The question of how to translate the findings from preclinical animal investigations to the human condition has not been fully resolved. Editorial comment has stated that several basic questions regarding the findings from animal investigations must be answered before the relevance of animal histological and other findings is known (22). The implication of this dialog is that the animal models used to screen for spinal cord neurotoxicity require scrutiny. Neurotoxicity has been cited as the cause of a number of clinical syndromes, including cauda equina syndrome, adhesive arachnoiditis, transverse myelitis, and paralysis (23–25). Whereas these syndromes represent the clinical markers of spinal pathophysiology, the animal literature most often describes neurotoxicity in terms of surrogate markers, such as histopathology, a decrease in spinal cord blood flow, a change in neurophysiology, or various behavioral measures (26). Investigators have recommended none of these tests as a “gold standard” for detecting drug-induced neurotoxicity, nor is it likely that any of these tests will be validated formally for their ability to predict human pathophysiology. This has led to the situation in which numerous authors have recommended that a broad range of tests be conducted in the preclinical investigation of spinally-administered drugs. Because of these uncertainties, the presence of clinical signs and symptoms in humans remains the most relevant and undisputed marker of spinal cord neurotoxicity.

Eleven-hundred patients may not be a large enough sample to demonstrate a slightly increased risk of neurological injury with intrathecal midazolam overall. However, 1100 patients does represent a sufficient sample size with which to determine whether adverse symptoms occur in a similar proportion of individuals to that of the animals exhibiting histopathology after intrathecal midazolam (30%–100%) in the articles that reported midazolam-mediated neurotoxicity (6–9).

The review of the patients occurred one month after surgery. The choice of one month was arbitrary, but it was chosen to detect a symptom that may have persisted beyond a reasonable time for recovery after the intrathecal injection and yet also to represent a delay sufficiently long to detect a progressive pathology. In all cases, the histopathology associated with intrathecal midazolam in preclinical animal studies was found to occur between hours and days after administration.

This investigation was an observational study of clinical practice and its outcome. As such, it did not control for all of the sources of bias that may have occurred in the selection of patients who received intrathecal midazolam. The cohort methodology is well suited to identify risk factors for adverse outcomes by following up a group of patients over time. The association of specific risk factors with known adverse effects can be identified and the basis established for future investigations using randomized trials. Therefore, future investigations studying the efficacy of midazolam by using a double-blinded, randomized, and controlled methodology should include the follow-up of patients for adverse effects.

In summary, this study sought to identify whether intrathecal midazolam was associated with neuropathologic symptoms with a similar frequency to that of the occurrence of histopathology in animals. It was found that in contrast with the studies that reported histopathology in animals, no adverse effects were found in association with intrathecal midazolam when assessed by symptoms in humans. This is consistent with the administration of a dose of midazolam, approximately 0.03 mg/kg, that is less than that associated with both histopathology and behavioral changes in previous animal investigations. Therefore, despite the animal literature describing a range of histopathology occurring often with intrathecal midazolam, a safe dose of midazolam has been identified. When this dose of midazolam has been administered to patients and the patients have been monitored, no increased adverse effects have been associated with its use compared with conventional therapies.

The authors gratefully acknowledge Maureen O'Flaherty for her assistance with data collection, Associate Professor Paul Myles for his assistance with the statistical analysis, and Tattersalls for their financial support of the Monash University Department of Anesthesia.

Back to Top | Article Outline


1. Goodchild CS, Serrao JM. Intrathecal midazolam in the rat: evidence for spinally-mediated analgesia. Br J Anaesth 1987;59:1563–70.
2. Edwards M, Serrao JM, Gent JP, Goodchild CS. On the mechanism by which midazolam causes spinally mediated analgesia. Anesthesiology 1990;73:273–7.
3. Goodchild CS, Noble J. The effects of intrathecal midazolam on sympathetic nervous system reflexes in man: a pilot study. Br J Clin Pharmacol 1987;23:279–85.
4. Yanez A, Peleteiro R, Camba MA. [Intrathecal administration of morphine, midazolam, and their combination in 4 patients with chronic pain]. Rev Esp Anestesiol Reanim 1992;39:40–2.
5. Goresky GV. The clinical utility of epidural midazolam for inguinal hernia repair in children. Can J Anaesth 1995;42:755–7.
6. Bozkurt P, Tunali Y, Kaya G, Okar I. Histological changes following epidural injection of midazolam in the neonatal rabbit. Paediatr Anaesth 1997;7:385–9.
7. Erdine S, Yucel A, Ozyalcin S, et al. Neurotoxicity of midazolam in the rabbit. Pain 1999;80:419–23.
8. Malinovsky JM, Cozian A, Lepage JY, et al. Ketamine and midazolam neurotoxicity in the rabbit. Anesthesiology 1991;75:91–7.
9. Svensson BA, Welin M, Gordh T Jr, Westman J. Chronic sub-arachnoid midazolam (Dormicum) in the rat: morphologic evidence of spinal cord neurotoxicity. Reg Anesth 1995;20:426–34.
10. Auroy P, Schoeffler P, Maillot C, et al. [Intrathecal tolerability of midazolam: histological study.] Ann Fr Anesth Reanim 1988;7:81–2.
11. Bahar M, Cohen ML, Grinshpoon Y, et al. An investigation of the possible neurotoxic effects of intrathecal midazolam combined with fentanyl in the rat. Eur J Anaesthesiol 1998;15:695–701.
12. Schoeffler P, Auroy P, Bazin JE, et al. Subarachnoid midazolam: histologic study in rats and report of its effect on chronic pain in humans. Reg Anesth 1991;16:329–32.
13. Schwieger IM, Jorge-Costa M, Pizzolato GP, et al. Intrathecal midazolam reduces isoflurane MAC and increases the apnoeic threshold in rats. Can J Anaesth 1994;41:144–8.
14. Serrao JM, MacKenzie JM, Goodchild CS, Gent JP. Intrathecal midazolam in the rat: an investigation of possible neurotoxic effects. Eur J Pharmacol 1990;7:115–22.
15. Nishiyama T, Matsukawa T, Hanaoka K. Acute phase histopathological study of spinally administered midazolam in cats. Anesth Analg 1999;89:717–20.
16. Auroy Y, Narchi P, Messiah A, et al. Serious complications related to regional anesthesia: results of a prospective survey in France. Anesthesiology 1997;87:479–86.
17. Hopwood MB. Statistics: can we prove an association for a rare complication? Reg Anesth 1993;18:428–33.
18. Hampl KF, Schneider MC, Ummenhofer W, Drewe J. Transient neurologic symptoms after spinal anesthesia. Anesth Analg 1995;81:1148–53.
19. Ready LB, Plumer MH, Haschke RH, et al. Neurotoxicity of intrathecal local anesthetics in rabbits. Anesthesiology 1985;63:364–70.
20. Yaksh TL, Collins JG. Studies in animals should precede human use of spinally administered drugs. Anesthesiology 1989;70:4–6.
21. Bainton CR, Strichartz GR. Concentration dependence of lidocaine-induced irreversible conduction loss in frog nerve. Anesthesiology 1994;81:657–67.
22. Coombs DW, Fratkin JD. Neurotoxicology of spinal agents. Anesthesiology 1987;66:724–6.
23. Horlocker TT, McGregor DG, Matsushige DK, et al. Neurologic complications of 603 consecutive continuous spinal anesthetics using macrocatheter and microcatheter techniques: Perioperative Outcomes Group. Anesth Analg 1997;84:1063–70.
24. Lee DS, Bui T, Ferrarese J, Richardson PK. Cauda equina syndrome after incidental total spinal anesthesia with 2% lidocaine. J Clin Anesth 1998;10:66–9.
25. Loo CC, Irestedt L. Cauda equina syndrome after spinal anaesthesia with hyperbaric 5% lignocaine: a review of six cases of cauda equina syndrome reported to the Swedish Pharmaceutical Insurance 1993–1997. Acta Anaesthesiol Scand 1999;43:371–9.
26. Hodgson PS, Neal JM, Pollock JE, Liu SS. The neurotoxicity of drugs given intrathecally (spinal). Anesth Analg 1999;88:797–809.
© 2004 International Anesthesia Research Society