Share this article on:

Preclinical Insights into the Implementation of Intrathecal Midazolam: A Cautionary Tale

Yaksh, Tony L. PhD; Allen, Jeffrey W. PhD

doi: 10.1213/01.ANE.0000121768.79904.7F

Department of Anesthesiology, University of California, San Diego

Accepted for publication January 28, 2004.

Address correspondence and reprint requests to Tony L. Yaksh, PhD, Department of Anesthesiology, University of California-San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0818. Address e-mail to

Editor's note: Please refer to the editorial by Cousins and Miller (pp. 1507–8) and the articles by Tucker et al. (pp. 1512–20 and 1521–7), Johansen et al. (pp. 1528–35), and Yaksh and Allen (pp. 1536–45) in this issue.

In the present issue of the journal, there are four papers focusing on intrathecal (IT) midazolam: a preclinical assessment of safety in sheep and pigs (1), a population evaluation of postoperative side effects in humans receiving perioperative IT midazolam in combination with local anesthetics and fentanyl (2), a human study examining IT midazolam and fentanyl in labor (3), and a review related to the issues pertinent to the development of the use of spinal midazolam (4). These papers reflect a continuing process that began almost 20 years ago.

As reviewed (4), insights in the late 1970s regarding the spinal actions of γ-aminobutyric acid and the pharmacology of the benzodiazepine receptors led to the initial preclinical work suggesting the activity of spinal midazolam in regulating spasticity and pain processing. These insights, given the zeitgeist of the late 1970s, formed by the appreciation of the clinical benefits arising from the spinal delivery of drugs such as opiates and baclofen provided a foundation for considering the spinal delivery of midazolam as a therapeutic approach to manage pain and spasticity.

Tempering the enthusiasm for IT therapy, spinal drugs have long been known for having the potential for producing local injury leading to functional deficits. Accordingly, even early work emphasized that the initial delivery of any drug into the human spinal canal must be preceded by some assessments of safety and the assertion of no toxicity. What justifies this emphasis on preclinical work? First, the preclinical models have been reliable in predicting the analgesic efficacy, physiological effects, and pharmacology of IT delivered drugs ranging from opiates through blockers of N-type calcium channels in humans (5,6). Second, they provide a readout of any potential changes in function that are deleterious. Third, they allow a direct examination of the drug-exposed spinal tissue. Ethical issues aside, the first two aims could be achieved by careful titration of doses in humans. Obtaining normal patient spinal tissue for histopathology poses obvious limitations. Why is the third component so important? The principal concern of a preclinical spinal safety assessment is that the drug therapy induces no deleterious changes in spinal morphology. Whereas persistent changes in behavior and function may indicate underlying events, a lack of change in behavior does not exclude underlying tissue pathology. Normal function is at best only a surrogate marker for the absence of underlying tissue effect. Thus, whereas the lack of deleterious or irreversible effects of a treatment upon autonomic and behavioral function is a required aspect of safety assessment, it is not sufficient. Evolving deficits may not be revealed by functional indices for an extended period of time, whereas histological examination demonstrates a continuing event.

How then does the evolution of IT midazolam compare with that of other spinal drugs. IT morphine, frequently used in humans since 1978, initially underwent extensive behavioral investigations in species ranging from the rat through the primate. In all instances, the bolus delivery resulted in predictable, reversible behavioral effects. Although anecdotal comments on tissue pathology were made, it was only later that IT pathology was systematically examined after single dosing in primates (7) and repeated dosing in rats, cats (8), and dogs (9). In each case, the studies provided a convergent assertion that bolus delivery was without evident toxicity. In contrast, the first spinal delivery of many drugs such as clonidine, d-ala2-d-leu5-enkephalin, and neostigmine were preceded not only by a host of preclinical studies on behavioral and pharmacological characteristics, but also by systematic studies assessing the histopathological effects of the drug (6). Again, in each instance, although the study paradigms may have been limited, the absence of an irreversible functional deterioration and the absence of tissue toxicity provided convergent support for the assertion of safety. In each case, this histologically based assertion of a lack of toxicity was the point of departure for obtaining the appropriate institutional approvals for clinical trials.

Considering the development of midazolam, the two initial clinical studies (10,11) were supported by an initial study examining spinal histology in the feline model (10) and the absence of untoward physiological consequences in the anesthetized dog (12). As the feline data were not widely disseminated, this series of modestly described histological studies had little practical impact. The concern for issues of safety prompted the rabbit studies by Malinovsky et al. (13). Whereas the Malinovsky et al. model was indeed different with its percutaneous cisternal delivery, it demonstrated distinct differences between neuraxial vehicle and midazolam at concentrations that were used in humans after a single injection. It is here that the clinical development of IT midazolam deviated from that with other spinally-developed drugs. Taken at face value, the continued delivery of midazolam in the several clinical studies between 1991 and 1998, in face of the IT rabbit studies displaying histopathological toxicology, required a clear disregard for the literature. Before the rat IT study by Behar et al. (14) in 1997, human work was continued with 8 additional human reports and no additional preclinical safety data (4). One might argue that with the Behar et al. study, all was well. Then an additional rabbit study was published (15) that essentially confirmed the previous Malinovsky study. Despite these convergent data by different groups arguing for toxicity, clinical studies continued with no further substantiation for the assertion of safety.

It is argued by Tucker et al. (2) that the discrepancies between different preclinical studies (rat versus rabbit) and the lack of effect with the human experience reflect on the inability of the preclinical work to predict the human condition. With all due respect, we reject this statement. First, there is no ignoring the fact that the initial preclinical behavioral data provided the enabling information for all subsequent clinical work, even to the first dose used in the 1987 paper (approximately 1 mg), e.g., the spinal effects of these drugs are well predicted by the preclinical models. Second, in contrast to the rabbit data, on what basis can the authors dismiss the possibility of spinal changes in their clinical patients? They never looked!

These issues have now been readdressed with the study by Johansen et al. (1). We deem the Johansen et al. study to be a robust assessment of potential midazolam toxicity for several reasons. (a) It undertook chronic exposure of the spinal cord for an extended period of time; (b) It used multiple doses; (c) It used the maximally available concentrations of preservative-free midazolam (5 mg/mL); (d) It used two large species (pig and sheep); and (e) The sheep model, as used by these investigators, has previously been shown to display toxicity with other treatments, e.g., large concentrations of IT morphine (16). Accordingly, we suggest that these results in the rat, and now in the sheep and pig, support a likely acceptable safety profile for IT midazolam, within the range of the variables that were investigated. We still believe that the rabbit studies argue for watchful consideration (17).

This present outcome with IT midazolam will doubtless be subject to several interpretations (4). The clinical investigators who have published with IT midazolam will most likely conclude that their approach of attending to the preclinical behavioral data that show no irreversible effects is adequate for the first human experience. If there is no evident physiological problem then proceed with deliberate haste. However, there is no doubt a group of interested observers who would suggest that the present scenario with midazolam suggests an example of good fortune, and perhaps good intentions, rather than good logic. Current experience provides us with a number of examples where the appropriate preclinical safety study has revealed unexpected effects, even with drugs for which we have significant clinical experience. The issue of lidocaine radiculopathies with small-bore catheters (18) and the appearance of granulomas with large concentrations of continuously infused morphine (19) are two relevant examples. In both cases, when the appropriate questions were asked and the preclinical models properly used, the human consequences were indeed observed. We find the morphine issue particularly telling because systematic preclinical safety studies in 1992 were accomplished as cited above, but these studies used repeated bolus delivery to assess safety for acute use, as in postoperative pain and not for chronic delivery (9). In the human study, the chronic infusion of yet larger concentrations became a paradigm that led to the IT granuloma. Importantly, the preclinical safety evaluations using chronic spinal delivery of these larger concentrations provide confirming data on the phenomena (20). We believe that these models will now likely provide mechanistic insights to this problem. The message here is that safety studies only address issues that they are designed to test. Seemingly minor changes in practice (e.g., large concentration and slow infusion rates do not equal small concentrations and rapid infusion rates) or technique (bolus versus infusion) can produce distinct consequences.

With regard to midazolam, much remains to be considered. Thus, the safety data have been with midazolam alone, but all manner of combinations of bupivacaine, fentanyl, buprenorphine, and the like have been used. The effects of these combinations remain to be assessed. Whereas combinations are used with other drugs, it should be stressed that few adjuvants pose the potential problems of midazolam, with a solubility that is strongly influenced by pH values and perhaps the counter ion of the additive.

It is for these many reasons that we plead with clinicians to exercise due restraint in efforts such as reported in the Tucker et al. papers (2,3) when the clinical enthusiasm may outweigh reasonable precautions. We plead with human studies committees and journal editors to cast suspect studies aside. Despite all of the pleading found in editorial after editorial over the years, little seems to change (21). We suspect that the only effective restraint will be that exerted by the clinical peers of those who undertake such efforts as those reported here and elsewhere with drugs that have not undergone appropriate evaluation. When asked to participate in a trial or to review a trial proposal, it is fair to ask, “And what are the safety data for this formulation?” The cynic may say that the real restraint will evolve after an unfortunate event that occurs when the novel untested drug results in a bad outcome. We truly respect our clinical colleagues and it is the intent of our comments to prevent this tragic and perhaps avoidable scenario from ever occurring. IT safety is dependent upon many factors. We should not treat these issues of therapy development by choosing to wear a blindfold in a room full of dangerous objects.

Back to Top | Article Outline


1. Johansen MJ, Gradert TL, Satterfield WC, et al. Safety and efficacy of continuous intrathecal midazolam infusion in the sheep model. Anesth Analg 2004;98:1528–35.
2. Tucker AP, Lai C, Nadeson R, Goodchild CS. Intrathecal midazolam. I. A cohort study investigating safety. Anesth Analg 2004;98:1512–20.
3. Tucker AP, Mezzatesta J, Nadeson R, Goodchild CS. Intrathecal midazolam. II. Combination with intrathecal fentanyl for labour pain. Anesth Analg 2004;98:1521–7.
4. Yaksh TL, Allen JW. The use of intrathecal midazolam in humans: a case study of process. Anesth Analg 2004;98:1536–45.
5. Yaksh TL. Spinal systems and pain processing: development of novel analgesic drugs with mechanistically defined models. Trends Pharmacol Sci 1999;20:329–37.
6. Yaksh TL, Rathbun ML, Provencher JC. Preclinical safety evaluation for spinal drugs. In: Yaksh TL, ed. Spinal drug delivery. Amsterdam: Elsevier Science B.V., 1999:417–37.
7. Abouleish E, Barmada MA, Nemoto EM, et al. Acute and chronic effects of intrathecal morphine in monkeys. Br J Anaesth 1981;53:1027–32.
8. Yaksh TL, Noueihed RY, Durant PAC. Studies of the pharmacology and pathology of intrathecally administered 4-anilinopiperidine analogues and morphine in rat and cat. Anesthesiology 1986;64:54–66.
9. Sabbe MB, Grafe MR, Mjanger E, et al. Spinal delivery of sufentanil, alfentanil and morphine in dogs: physiologic and toxicologic investigations. Anesthesiology 1994;81:899–920.
10. Muller H, Gerlach H, Boldt J, et al. Spasticity treatment with spinal morphine or midazolam:in vitro experiments, animal studies and clinical studies on compatibility and effectiveness. Anaesthesist 1986;35:306–16.
11. 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.
12. Whitwam JG, Niv D, Loh L, Jack RD. Depression of nociceptive reflexes by intrathecal benzodiazepine in dogs. Lancet 1982;2:1465.
13. Malinovsky JM, Cozian A, Lepage JY, et al. Ketamine and midazolam neurotoxicity in the rabbit. Anesthesiology 1991;75:91–7.
14. Bahar M, Cohen ML, Grinshpon Y, Chanimov M. Spinal anaesthesia with midazolam in the rat. Can J Anaesth 1997;44:208–15.
15. Erdine S, Yucel A, Ozyalcin S, et al. Neurotoxicity of midazolam in the rabbit. Pain 1999;80:419–23.
16. Gradert TL, Baze WB, Satterfield WC, et al. Safety of chronic intrathecal morphine infusion in a sheep model. Anesthesiology 2003;99:188–98.
17. Malinovsky JM. Is intrathecal midazolam safe? Can J Anaesth 1997;44:1321–2.
18. Johnson ME. Potential neurotoxicity of spinal anesthesia with lidocaine. Mayo Clin Proc 2000;75:921–32.
19. Yaksh TL, Hassenbusch S, Burchiel K, et al. Inflammatory masses associated with intrathecal drug infusion: a review of preclinical evidence and human data. Pain Med 2002;3:300–12.
20. Follett KA. Intrathecal analgesia and catheter-tip inflammatory masses. Anesthesiology (Editorial Views) 2003;99:5–6.
21. Eisenach JC, Yaksh TL. Safety in numbers: how do we study toxicity of spinal analgesics? Anesthesiology 2002;97:1047–9.
© 2004 International Anesthesia Research Society