Portal de Periódicos da CAPES

The Neurotoxicity of Drugs Given Intrathecally (Spinal)

1999; Lippincott Williams & Wilkins; Volume: 88; Issue: 4 Linguagem: Inglês

10.1213/00000539-199904000-00023

1526-7598

Peter S. Hodgson, Joseph M. Neal, Julia E. Pollock, Spencer S. Liu,

Pain Mechanisms and Treatments

A growing understanding of the neuropharmacology of spinal cord processing of nociceptive input has led to intense interest in the use of spinal drugs in anesthesia and pain management. The direct application of receptor-specific therapeutics at the spinal cord can potentially interrupt specific pain pathways and limit systemic side effects, but this practice also carries the inherent risk of injury to the central nervous system. Thus, the neurotoxicity of spinal drugs is a central safety issue. Spinal cord or nerve root toxicity may manifest itself as histologic, physiologic, or behavioral/clinical derangements after exposure to a spinal drug. Neurohistopathology is broadly classified as neural injury, gliosis, or damage to the myelin sheath, and it also describes inflammatory changes and involvement of the arachnoid cell layers. Physiologic neurotoxicity of spinal drugs includes changes in spinal cord blood flow, disruption of the blood-brain barrier, and changes in the electrophysiology of impulse conduction. Behavioral and clinical signs of neurotoxicity include pain, motor and sensory deficits, and bowel and bladder dysfunction. Ideally, a complete roster of histological, physiologic, and behavioral testing would be performed on spinal drugs in several animal species, followed by safety trials in humans before widespread use. In practice, drugs have taken a variety of roads from conception to application, and often without safety data. In this article, we review available neurotoxicity data on drugs that have a clinical application, classified as spinal local anesthetics, spinal analgesics, or spinal adjuvants. Spinal Local Anesthetics The 100-yr history of spinal local anesthetic use in humans has typically involved self-experimentation, followed by widespread application with little or no controlled testing for neurotoxicity. Bier and Hildebrant [1] initially performed spinal anesthesia with cocaine on themselves in 1898, and essentially all of the earliest local anesthetics for spinal anesthesia were introduced in this fashion without toxicity studies. Despite a long history of clinical use, recent interest in neurotoxicity has arisen due to concerns over reports of cauda equina syndrome and transient neurologic symptoms (TNS) from spinal local anesthetics. We review animal data that have been used to assess the neurotoxicity of local anesthetics and summarize data available from human studies. In 1985, Ready et al. [2] evaluated the neurotoxic effects of single injections of local anesthetics in rabbits. They reported that spinal cord histopathology remained normal and that persistent neurologic deficits were not seen with clinically used concentrations of tetracaine, lidocaine, bupivacaine, or chlorprocaine. However, histopathologic changes and neurologic deficits did occur with higher concentrations of tetracaine (1%) and lidocaine (8%). In this model, extensive neurologic impairment was not necessarily accompanied by equally extensive lesions in the spinal cord and nerve roots, thus demonstrating the need for multiple models to fully assess neurotoxicity. Recent studies have used desheathed peripheral nerve models, designed to mimic unprotected nerve roots in the cauda equina, to further assess electrophysiologic neurotoxicity of clinically relevant concentrations of local anesthetics [3-5]. These models demonstrate that clinically used concentrations of 5% lidocaine and 0.5% tetracaine cause irreversible conduction block, whereas 1.5% lidocaine, 0.75% bupivacaine, and 0.06% tetracaine do not. Electrophysiologic toxicity of lidocaine in these models is concentration-dependent (Figure 1) beginning at 40 mM (approximately 1%) with irreversible ablation of the compound action potential at 80 mM (approximately 2%). Kanai et al. [4] subsequently demonstrated that generation of action potentials was more vulnerable than maintenance of resting membrane potential and that irreversible ablation of action and resting membrane potential by lidocaine seems to be both concentration- and time-dependent.Figure 1: The nonreversible effect of 40 mM lidocaine on the compound action potential (CAP) of frog sciatic nerve. Lidocaine was applied to a stable nerve preparation for 15 min, then washed with frog Ringer's solution for 2 h. Tracings represent CAPs in response to stimuli (1-Hz stimulus = heavy line, 40-Hz stimulus = thin line). Lidocaine 40 mM completely ablated the CAP when applied to the nerve. The 1-Hz CAP response began to return after 10-15 min of washing and reached a new level in 45 min, where it was stable for the subsequent 2 h of observation. The recovered 1-Hz CAP is only 65% of the original. Reprinted with the publisher's permission from Holman SJ. Anesthesiology: hyperbaric dye solution distribution. Anesthesiology 1997;86:969.Effects of local anesthetics on spinal cord blood flow seem benign. Spinal administration of bupivacaine, lidocaine, mepivacaine, and tetracaine causes vasodilation and increase spinal cord blood flow [6,7], whereas ropivacaine causes concentration-dependent vasoconstriction and reduction in spinal cord blood flow [6]. However, the effects of lidocaine on blood flow of in vitro peripheral nerve models are more concerning. Myers et al. [8] applied solutions of isotonic sodium chloride solution, 1% and 2% lidocaine with and without epinephrine, and epinephrine alone to isolated rat sciatic nerve and measured changes in blood flow with a laser Doppler flow probe. Blood flow was significantly depressed for all solutions except isotonic sodium chloride solution. Epinephrine by itself significantly reduced nerve blood flow, and, when added to local anesthetic solutions, it reduced blood flow to a greater extent than the reduction caused by local anesthetics alone. Although experimental studies in animals have provided ample evidence that some local anesthetics in clinically relevant concentrations can injure nerve tissue, the exact mechanisms of injury are unclear. Recent work on neuronal cell lines has attempted to determine the mechanism of local anesthetic neurotoxicity. Johnson and Uhl [9] have shown that direct application of 2.5%-5.0% lidocaine caused a >3-fold increase in intracellular calcium and up to a 20% incidence of cell death during 60 min of exposure in the neuronal cell line. They postulated that the mechanism of neurotoxicity was not likely from sodium channel blockade, because such a block would not lead to an increase in cytoplasmic calcium. Subsequent work in this model determined that 0.5% and 1.0% lidocaine, as well as 0.625% bupivacaine, lead to transient, moderate increases in calcium, probably from the endoplasmic reticulum, without cell death [10]. Thus, several different laboratory models have proven that all local anesthetics can be neurotoxic but that lidocaine and tetracaine are potentially more neurotoxic than bupivacaine (Table 1).Table 1: Local Anesthetic ToxicityDespite the knowledge that all local anesthetics can be neurotoxic in the laboratory model, large-scale surveys of the complications of spinal anesthesia attest to the relative safety of spinal local anesthetics in humans (Table 2). Retrospective [11], prospective [12], and historical studies [13-15] report 0%-0.7% incidence of postoperative neurologic injury in patients undergoing spinal anesthesia. Although lacking a denominator, information from closed-claims databases corroborate these findings [16,17]. Thus, the neurotoxic potential of spinally administered local anesthetics has not manifested itself in large-scale studies.Table 2: Large Epidemiologic Studies of the Neurologic Complications of Spinal AnesthesiaThere are few nonepidemiologic clinical studies evaluating the potential neurotoxicity of local anesthetics, and all have focused on electrophysiologic variables after spinal anesthesia. Somatosensory evoked potentials, monosynaptic H-reflex [18], and cutaneous current perception thresholds [19] have been used to evaluate recovery after spinal anesthesia. These measurements have shown complete return to baseline activity after 5% lidocaine spinal anesthesia in very small study populations. Histopathologic or other physiologic data in humans are lacking; thus, information from controlled studies in humans is essentially not available. Lidocaine Controversy about the use of single-injection spinal lidocaine began in 1993 when Schneider et al. [20] published four cases of short-lived neurologic symptoms after spinal anesthesia with 5% hyperbaric lidocaine. This was the first report to question the potential for neurotoxicity with standard clinical doses and concentrations of lidocaine after single-injection spinal anesthesia. Subsequent prospective, randomized studies reveal a 4%-33% incidence of TNS after lidocaine spinal anesthesia (Table 3) [21,22]. This incidence varies with the type of surgical procedure and is unaffected by baricity or the dilution of lidocaine to 0.5%. Contemporary reports of cauda equina syndrome after continuous lidocaine spinal anesthesia and the potential concentration-dependent neurotoxicity of lidocaine have led several authors to label TNS as a manifestation of subclinical neurotoxicity.Table 3: Incidence of Transient Neurologic Symptoms (TNS) with Spinal Anesthesia in Prospective Randomized StudiesAs previously discussed, laboratory work in both intrathecal and desheathed peripheral nerve models has proven that the concentration of lidocaine is a critical factor in neurotoxicity. Because concentrations of lidocaine <40 mM (approximately 1.0%) are not neurotoxic to desheathed peripheral nerve, such dilute concentrations of spinal lidocaine should not cause TNS if the syndrome is caused by subclinical concentration-dependent neurotoxicity. We recently examined whether spinal lidocaine concentrations of <1.0% might therefore decrease the incidence of TNS [21]. Patients undergoing knee arthroscopy were randomized to receive 50 mg of hyperbaric lidocaine as either a 2.0%, 1.0%, or 0.5% solution. There was no difference in the incidence of TNS (18%) among the three groups. The high incidence of TNS with lidocaine concentrations 968 surgical, obstetrical, and chronic pain patients) with no clinical evidence of neurotoxicity [36]. Although no histopathologic or physiologic studies have been reported, clonidine seems to be a safe spinal drug in humans. Acetylcholine Esterase Inhibitors Neostigmine indirectly produces a muscarinic agonist effect by inhibiting acetycholinesterase and has been shown to cause analgesia in animal and human experiments. Neurohistopathological analysis of rats and dogs after long-term intrathecal neostigmine (with and without paraben preservatives) administration reveal no spinal cord toxicity [37,38]. Neostigmine does not affect spinal cord blood flow in sheep [37], and there were no behavioral changes suggesting neurotoxicity reported in the above animal studies. Phase I safety assessments in human volunteers have been performed for both preservative-free (50-750 [micro sign]g) and paraben-containing hyperbaric preparations (10-100 [micro sign]g) of spinal neostigmine without clinical evidence of neurotoxicity [40,41]. Although the human experience with spinal neostigmine is limited, clinical trials performed thus far have not reported any evidence of neurologic sequelae. gamma-Amino Butyric Acid Agonists (Table 5) Midazolam. Unlike other benzodiazepines, midazolam is soluble in an aqueous solution when buffered to approximately pH 3.5. At physiologic pH, midazolam becomes lipophilic, facilitating tissue penetration. These characteristics have made midazolam the most extensively studied spinal benzodiazepine. Neurotoxicity studies in animals have yielded conflicting results. Four initial rat studies with intrathecal catheter implantation using 0.15 mg/kg for 15 days (two studies) or isolated exposures to 0.1-0.3 mg/kg of midazolam (two studies) prepared in saline solution showed no neurotoxic reactions on light or electron microscopy [42]. However, a subsequent study in rabbits after a single 0.1 mg/kg intrathecal injection of midazolam reported that three of nine animals (33%) showed spinal cord histopathologic changes 8 days after exposure [43]. The diffuse nature of the histopathologic abnormalities that uncharacteristically extended from cervical to lumbar sections, the presence of significant and persistent diastolic hypotension in the treatment group, and the delayed postmortem tissue fixation all suggest a possible systemic source of artifact in the three affected animals. To investigate these contrasting results, a state-of-the-art study on the rat was performed using light microscopy, electron microscopy, cell morphometry, and transcardial tissue fixation after daily intrathecal administration of approximately 0.3 mg/kg midazolam for 20 days [44]. The spinal cords showed strong evidence of neuronal death and cellular abnormalities, even on light microscopy, in most midazolam-treated rats. Of note, a hypotonic commercial preparation of midazolam was used in contrast to the isotonic saline preparation used in all previous reports. Hypotonicity results in permanent nerve injury in isolated nerve preparations [33] and has been implicated in the neurotoxicity of spinal sufentanil in sheep [45]. Although hypotonicity may be the etiology of the reported abnormalities, intrinsic neurotoxicity of spinal midazolam is a consideration.Table 5: Comparative Animal and Human Toxicity Data of Common Spinal AdjuvantsNo animal studies on spinal cord blood flow or electrophysiology have been reported. The effect of midazolam on blood-brain barrier integrity was investigated in the above-described study on rabbits and showed compromise in three of nine animals [43]. Despite some histopathologic evidence of neurotoxicity, no significant behavioral abnormalities have been reported in any of the animal neurotoxicity studies of intrathecal midazolam. There are no histologic or physiologic studies of humans exposed to spinal midazolam, although there are seven small reports of intrathecal midazolam for anesthesia and pain management. Within this limited human experience, there are no reports of clinical neurologic deficit, even after prolonged continuous intrathecal use in four patients with chronic benign pain syndromes [46]. Midazolam neurotoxicity is controversial. Although the commercial midazolam solution seems to be neurotoxic in rats, hypotonicity of the solution may be culpable, rather than the drug itself. Midazolam in saline does not seems to be neurotoxic in the rat, but it may be toxic in the rabbit. There are insufficient studies in humans to determine the risk of neurotoxicity with spinal midazolam. Baclofen Spinal spasticity, which is thought to result from disinhibition of motor horn cells after upper motor neuron damage, can be treated with baclofen. Baclofen is a stable analog of GABA and interacts primarily with the inhibitory GABA-B receptors in lamina 2 of the dorsal horn (Figure 1). Twenty-eight dogs exposed to chronic spinal delivery of either saline, clinical, or supraclinical doses of baclofen showed no neurohistopathologic changes. Cats exposed to intrathecal baclofen for 8 days likewise showed no abnormal spinal cord histopathology [47]. No studies on spinal cord blood flow, blood-brain barrier effects, or electrophysiology have been published with regards to intrathecal baclofen. Neither dogs, cats, nor monkeys showed behavioral evidence of neurotoxicity [25]. There are no published human studies evaluating histopathologic or physiologic neurotoxicity of baclofen. Spinal baclofen was infused in seven subjects for 3-22 mo without clinical evidence of neurotoxicity [48]. An extensive review of cases involving intrathecal baclofen overdose did not note any long-term sequelae [49]. Based on animal studies and considerable clinical experience, spinal baclofen is not likely to cause neurologic damage. N-Methyl-D-Aspartate Antagonists Ketamine. Animal studies examining the neurotoxicity of ketamine generally support its safety, with the exception of some poorly explained findings. Studies performed in monkeys, baboons, and rabbits after single-dose intrathecal ketamine injections (0.3-0.6 mg/kg) with and without benzethonium chloride preservative did not uncover histopathologic evidence of neurotoxicity [50,51]. In contrast, two other rat studies reported histopathologic evidence of neurotoxicity with spinal ketamine. One study found vacuolization of the ganglion cells in posterior nerve roots in 3 of 33 rats that died immediately on injection of preservative-free ketamine [52]. However, it is difficult to ascribe these findings to ketamine neurotoxicity given the uncertain circumstances of demise. Similarly, another rat study with single intrathecal injections of 2.5 mg of ketamine hydrochloride with the preservative benzethonium chloride reported that two of six rats had radicular demyelination injury at sites distant from the injections on histological examination [53]. However, more than half of the animals in the ketamine treatment group either died after rapid injection (two rats) or had single hindlimb paralysis after injection (four rats), which puts the validity of the results in question. No animal studies have been published on the effect of spinal ketamine on spinal cord blood flow or electrophysiology. Rabbits showed a pattern of blood-brain barrier compromise differing significantly from that of saline-injected control rabbits after single 3-mg injections of ketamine with chlorbutanol [43] but not with preservative-free ketamine [50]. The neurotoxicity of spinal ketamine is largely untested in humans, despite a few case series reporting ketamine use for spinal anesthesia and analgesia [54]. Taken together, the rat, rabbit, and primate studies with intrathecal ketamine support its safety if used without a preservative. A small preliminary human experience suggests that the anesthetic is well tolerated. However, the commercially available preparation of ketamine contains an untested preservative (benzethonium chloride) and cannot be recommended for intrathecal use in humans. Amitriptyline. Preclinical animal testing of intrathecal amitriptyline is limited to a physiologic assessment in adult sheep [55]. A range of intrathecal doses (0.25, 1, or 5 mg) with a maximal dose representing approximately 25-50 times the anticipated human dose did not reduce spinal cord blood flow. No behavioral abnormalities were reported except for transient agitation with the injection of 5 mg of amitriptyline. The injection of 5 mg of amitriptyline into the cervical intrathecal space sedated the animals for 10-60 min, whereas 10 mg produced intense sedation and seizure, followed by death from uncertain causes in one animal. Long-term exposure and neurohistopathologic studies are reported to be in progress [55]. The determination of amitriptyline neurotoxicity awaits further investigation. Somatostatin In rats, intrathecal somatostatin (SST) produced marked, dose-dependent neurotoxic responses at or near doses required to produce analgesia [56] with similar findings in cats and mice [57]. SST administered in similar doses to guinea pigs, in contrast, did not provoke significant neurohistopathological changes [58]. Spinal SST in rats had significant vasoconstrictive effects in the spinal cord and brain, leading to reduced blood flow, increased vascular permeability, and compromised blood-brain barrier [59]. No spinal cord blood flow changes were seen in the guinea pig. Significant behavioral changes were noted in the rat, cat, and mouse experiments, but not in the guinea pig experiments. Species differences in the neurotoxic susceptibility to SST may explain these differences, but there is evidence for neurotoxicity in several animal species. In humans, spinal SST has been offered to terminally ill patients. Four patients with intractable cancer pain were studied after daily injections of spinal SST. Postmortem histopathology was undertaken on the spinal cords of two patients. One showed moderate degeneration of some dorsal roots within the cauda equina, whereas the other demonstrated no histopathologic changes. Clinical signs of neurotoxicity were not discussed [60]. In summary, SST has been shown to be neurotoxic in rats, mice, and cats at doses comparable to those that confer analgesia. In humans, relatively small doses of SST have been anecdotally reported to variably relieve pain in the absence of overt neurologic sequelae. Based on this demonstrated record of neurotoxicity in several animal species, spinal SST should be considered a last-line analgesic only in the terminally ill patient. Nonsteroidal Antiinflammatory Drugs Prostaglandins are involved in the spinal cord facilitation of pain processing, and spinal nonsteroidal antiinflammatory drugs (NSAIDs) can abolish wind-up behavior in animals [61]. Although the analgesic efficacy of spinal ketorolac has been established, no animal neurotoxicity studies have been published [61]. However, the 10% alcohol solvent used for commercial preparations is potentially neurotoxic, and commercial ketorolac should not be used spinally. Lysine acetylsalicylic acid (L-ASA) dosed intrathecally has been shown to be antinociceptive and has been tested in the rat for neurotoxicity with conflicting results. Although limited technically by a >50% spinal cord trauma rate related to needle puncture, one rat study reported radicular demyelination injury in one of the seven undamaged rats (14%) who received intrathecal L-ASA [53]. In contrast, neither histopathologic abnormalities nor a persistent decrease in spinal cord blood flow were seen in another rat study with large daily doses of L-ASA for at least 14 days [62]. The initial study described aggressive behavioral changes in rats after the spinal injection of L-ASA, leading to death from combat. The subsequent study did not report behavioral findings. Although the second report suppor