ADVERSE drug reactions are a common cause of injury and death in hospitalized patients. 1However, in many cases, serious adverse reactions are rare and are not recognized until a new drug has been in use for many years and after a huge patient exposure. 2Such is the case with neurotoxicity from intrathecal injection of local anesthetics; these drugs have been in use for many decades and have been given to tens of millions of patients. Nevertheless, their toxic potential is a current medical issue.

The current issue of Anesthesiology includes a report 3that examines the relative neurotoxicity of lidocaine and prilocaine, an alternative to lidocaine for brief spinal anesthesia. In the review process of this manuscript, concern was raised that the way the authors induced neurotoxicity (by slow infusion of drug through a small-bore intrathecal catheter with the tip deliberately placed in the cauda equina) bore no resemblance to modern clinical practice, and, therefore, the results were of questionable clinical relevance. Why did the authors do this, and are their results relevant to clinical practice? How do we go about assessing the toxicity of intrathecal drugs such as these?

Cauda equina syndrome is perhaps the most serious toxic complication of a spinal anesthetic. It is a rare event that occurs with an incidence perhaps as low as 1:100,000 following a single intrathecal injection of lidocaine. Nevertheless, considerable efforts are still being made to better understand the mechanisms of lidocaine neurotoxicity in an effort to reduce this risk. Speculation has implicated factors that include dose, concentration, hyperbaricity, and speed of injection, at least as associated with slow delivery through microspinal catheters. A common mechanism by which these factors are hypothesized to relate to toxicity is maldistribution of a drug within the thecal sac. In this, high concentrations of a drug are in prolonged contact with the sacral nerve roots, which are unprotected by sheaths. Lidocaine is of considerable concern in this regard because it has been marketed in a concentration (5%) well above that shown to be neurotoxic. 4 

Many practitioners tacitly accept this theory of maldistribution and inject no more than 60 mg lidocaine, using nondextrose-containing nearly isobaric solutions, avoid epinephrine, dilute 5% hyperbaric lidocaine to at least 2.5% with cerebrospinal fluid or preservative-free saline, and avoid injection through a spinal catheter. Dilution of a drug before injection and avoidance of lidocaine injection through spinal catheters or small-bore spinal needles were suggested by the Food and Drug Administration in a labeling change to the 5% lidocaine formulation after 1995. Unfortunately, cases of cauda equina syndrome have also been reported after intrathecal injection of small doses of lidocaine, nondextrose-containing lidocaine, or lower concentrations of lidocaine (2%). Without knowledge of the denominator to accompany the numerator of these several case reports, one is uncertain whether the manipulations listed previously truly reduce the risk of neurotoxicity.

Returning to the current laboratory study, 3what are we to make of the contribution of a paradigm using the slow infusion of undiluted lidocaine to our understanding of clinical neurotoxicity? Figure 1demonstrates the dilemma. If we assume that the incidence of serious toxicity is similar to that seen in humans, we would need to perform “routine” spinal anesthesia in nearly 300,000 animals to have a reasonable statistical likelihood of observing just one case of neurotoxicity. Ethical, financial, and practical considerations render this approach unacceptable. In other words, it is not feasible to examine this issue using a “clinically relevant” experimental protocol. Therefore, some alternative is needed. If one understood the mechanism of neurotoxicity, one might mimic local anesthetic neurotoxicity in cell culture. This is not the case with neurotoxicity from intrathecal lidocaine; we do not understand the cellular or vascular mechanism and do not know with any certainty that we can meaningfully reproduce this phenomenon in cell culture. As such, investigators, as in the current case, 3manipulate the in vivo  experimental conditions, usually by increasing drug dose, concentration, or time of exposure, to increase the likelihood of observing toxicity in or at a more feasible incidence for study, such as 50% (fig. 1).

Fig. 1. Relation between incidence of a problem (toxicity) and number of subjects required to study in order to have an 80% probability of observing at least one case of the problem. Neurotoxicity from intrathecal lidocaine in clinical use may be as rare as 1:100,000, meaning more than 300,000 subjects must be studied to have an 80% probability of observing a case. By increasing dose, concentration, or time of exposure, researchers increase the incidence to 1:2, making it feasible to study the phenomenon.

Fig. 1. Relation between incidence of a problem (toxicity) and number of subjects required to study in order to have an 80% probability of observing at least one case of the problem. Neurotoxicity from intrathecal lidocaine in clinical use may be as rare as 1:100,000, meaning more than 300,000 subjects must be studied to have an 80% probability of observing a case. By increasing dose, concentration, or time of exposure, researchers increase the incidence to 1:2, making it feasible to study the phenomenon.

Close modal

Does this “high-dose” approach allow us to say anything meaningful about the mechanism of a rare clinical problem that occurs following lower dose drug exposure? As one often hears from colleagues in the hallway, “What does toxicity mean if it only happens while the animals were swimming in a solution of drug?” It is broadly believed that drug effect is positively correlated with concentration. We know that lidocaine blocks sodium channels. We can devise a study to assess whether a persistent blockade of sodium channels is deleterious. Yet, it is also clear that these molecules may have a plethora of actions, with the number of target mechanisms increasing as drug concentrations increase. Therefore, at higher concentrations lidocaine may interact with membrane lipids or alter the blood–brain barrier, leading to other events. So, we may tailor the study to determine the effects of lidocaine at concentrations higher than those required to block sodium channels. Here, the issue is not whether lidocaine is a local anesthetic, but at what concentrations “other things” happen.

From a practical standpoint, the greater the separation of the sodium channel blocking concentration from the concentration that produces other effects (e.g ., breeching the blood–brain barrier), the greater the therapeutic ratio and the “safer” the drug. In addition, we believe that drug effects have mechanisms that evolve over time. For local anesthetics, this time is short and the reversal of the sodium channel blockade disappears with the removal of the drug. Other effects (e.g ., an effect on blood–brain barrier) may also occur short-term, but the consequences of that breeching may require days or even weeks to evolve (and to disappear). If extended exposure has no effect, the greater the likelihood that a shorter exposure will be without consequence. These lines of reasoning are consistent with the current thinking about spinal drug safety evaluation. In our animal studies of clonidine, 5neostigmine, 6and adenosine, 7we used drug exposures up to 100 times those used in humans up to 28 days and observed no toxicity. These observations provide reassurance that the bolus delivery of these agents at a fraction of those concentrations is likely to be clinically “safe.” On the other hand, if we begin to approach those higher concentrations or change the protocol to increase the interval of exposure (e.g ., alter formulation with hyperbaricity, initiate continuous infusion, or alter formulation to change drug clearance or metabolism), the rigor of our assertions of safety will be reduced. In other words, the assertion of an absence of toxicity is dependent on the conditions in which the drug was tested and those in which the drug is used.

In the present case, if one accepts the hypothesis that maldistribution is the underlying phenomenon that allows lidocaine neurotoxicity to occur, then exposing the rat spinal cord to concentrations of lidocaine no greater than those in the commercial formulation (5%) during conditions that mimic maldistribution (e.g ., continued local infusion) is logical and appropriate. Importantly, the use of such a model permits comparison across drug molecules in a given class (e.g ., prilocaine). If the toxicity reflects sodium channel blockade, one would predict that there would be no difference in the safety ratio defined in the rat, with continuous cauda equina infusion. On the other hand, if toxicity is dependent on another property (undefined) of the molecule that does not reflect on the sodium channel blocking properties, then it is possible that the other agent would have a different safety ratio (may be better, may be worse). This then serves to develop predictions as to possible underlying mechanisms. One may then consider the human experience (if it exists) with the drug and assess the issue of toxicity. That experience then serves to validate (or not) the preclinical model and provide insights into the underlying covariate.

One example of this regards morphine. Case reports of intrathecal granuloma concentrations have become prevalent, with a likely covariate being the use of moderate doses with low infusion rates requiring high concentrations (25–50 mg/ml). 8Until recently, the only systematic laboratory safety studies with intrathecal morphine were those carried out with 28-day intrathecal bolus deliveries with maximum concentrations of 10 mg/ml in dogs. 9These studies were specifically targeted at considering the possible safety of acute bolus delivery. Recently, we completed studies showing that in dogs high concentrations of morphine (12 mg/ml) delivered intrathecally for 28 days will also reliably result in aseptic intrathecal granulomas (personal communication between Eisenach and Yaksh, June 20, 2002). These observations (unfortunately) provide validation for the dog model and striking support for the need to examine the limits of the concentration profile in a preclinical model prior to implementation in humans.

So, how do we extrapolate the parameters of the surrogate (laboratory model) to that of the human condition? As noted, safety in humans will always initially be proposed within the context of the laboratory model in which it was studied and the assumed validity of this model. At the minimum, tissue exposure to drugs in animals should be tailored as closely as possible to that presumed to occur in humans. In the case of maldistribution of lidocaine, at least one component should include the continued infusion of the concentration of the commercial formulation itself. In this case, given comparable exposure profiles in the same animal model, a first-order prediction is that the relative safety of two drugs in the surrogate expressed as a fraction of their therapeutic dose might predict the same ratio in humans. Hence, if two drugs, A and B, have equal therapeutic activity in the rat and A is more “toxic” than B in that model, we would predict that A would be more toxic than B in humans. If A had already been tried in humans, we might tentatively suggest that B at its therapeutic dose would be safer than A. This rationale would not permit us to say how much safer one drug is and it certainly would not permit us to say that B was without toxicity.

What is a clinician to do? First, experience is a great teacher. As indicated by the recent observations that major adverse events, leading to black box warnings by the Food and Drug Administration, occurred more than 7 yr after approval in half of the cases of drugs approved since 1975, and that drug withdrawals from the market by the Food and Drug Administration occurred more than 2 yr after approval in half the cases over the same time period, one can only conclude that a drug is safe after much patient exposure. Approaches such as that used by Kishimoto et al ., 3while not solving this problem, allow these numbers to be reduced in preclinical studies in order to understand what went wrong, why, and how we can make our care even safer. To the extent that the present practice with intrathecal lidocaine has changed, that preclinical influence has already been felt. Second, one can have little reason to believe that drugs used outside of those dose, concentrations, or delivery protocols (e.g ., with or without additives), which increase drug exposure, will be safe. Moving beyond the end of the dose–effect curve places us in a terra incognita . The approach used by these authors in this case follows logically from the presumed underlying factor, which allows lidocaine neurotoxicity to be manifest. Short of another few decades of human experience, this approach may be the only one to further our understanding of risk factors and mechanisms of local anesthetic-induced neurotoxicity and to hopefully improve the safety of spinal anesthesia. These considerations provide weight to the dictum of Paracelsus that there is no safe drug, only safe doses or concentrations.

Lazarou J, Pomeranz BH, Corey PN: Incidence of adverse drug reactions in hospitalized patients: A meta- analysis of prospective studies. JAMA 1998; 279: 1200–5
Lasser KE, Allen PD, Woolhandler SJ, Himmelstein DU, Wolfe SM, Bor DH: Timing of new black box warnings and withdrawals for prescription medications. JAMA 2002; 287: 2215–20
Kishimoto T, Bollen AW, Drasner K: Comparative spinal neurotoxicity of prilocaine and lidocaine. A nesthesiology 2002; 97: 1250–3
Bainton CR, Strichartz GR: Concentration dependence of lidocaine-induced irreversible conduction loss in frog nerve. A nesthesiology 1994; 81: 657–67
Yaksh TL, Rathbun M, Jage J, Mirzai T, Grafe M, Hiles RA: Pharmacology and toxicology of chronically infused epidural clonidine HCl in dogs. Fundam Appl Toxicol 1994; 23: 319–35
Yaksh TL, Grafe MR, Malkmus S, Rathbun ML, Eisenach JC: Studies on the safety of chronically administered intrathecal neostigmine methylsulfate in rats and dogs. A nesthesiology 1995; 82: 412–27
Chiari A, Yaksh TL, Myers RR, Provencher J, Moore L, Lee CS, Eisenach JC: Preclinical toxicity screening of intrathecal adenosine in rats and dogs. A nesthesiology 1999; 91: 824–32
Coffey RJ, Burchiel K: Inflammatory mass lesions associated with intrathecal drug infusion catheters: Report and observations on 41 patients. Neurosurgery 2002; 50: 78–86
Sabbe MB, Grafe MR, Mjanger E, Tiseo PJ, Hill HF, Yaksh TL: Spinal delivery of sufentanil, alfentanil, and morphine in dogs: Physiologic and toxicologic investigations. A nesthesiology 1994; 81: 899–920