The spinal delivery of the cholinesterase inhibitor neostigmine yields analgesia in rats and augments the analgesic effects of alpha 2 agonists in sheep. To assess its activity in humans, preclinical toxicology studies to define its safety were required in two species.
Rats with chronic intrathecal catheters received daily injections of saline (vehicle) or 5 micrograms/10 microliters or 10 micrograms/10 microliters neostigmine HCl (n = 6/group) for 4 days and were observed for general behavior and nociception (52.5 degrees C hot plate). On day 6, rats were anesthetized and submitted to whole body perfusion/fixation. For dog studies, male beagles were prepared following rigid aseptic precautions with catheters passed from the cisterna magna to the lumbar intrathecal space. Catheters were connected to an external vest-mounted pump. Based on preliminary studies, ten implanted dogs were randomly assigned to receive infusions of neostigmine for 28 days (4 mg/4 ml/day; n = 6) or saline (4 ml/day; n = 4). At 28 days, dogs were anesthetized, cisternal cerebrospinal fluid was obtained, and dogs were submitted to perfusion-fixation. Rat and dog spinal cords were embedded, sectioned, stained, and assessed by the pathologist without knowledge of treatment.
In rats, neostigmine produced a dose-dependent increase in hot plate latency, and no tolerance was observed. Mild tremor was observed but was not debilitating. Histopathology revealed a mild fibrotic reaction to the catheter with mixed signs of moderate, acute, and chronic inflammation with no differences between saline or drug groups. In dogs, neostigmine had no effect on blood pressure or on the skin twitch response but produced bradycardia and an increase in muscle tone. At sacrifice, cerebrospinal fluid protein, specific gravity, and glucose were elevated in both saline and neostigmine groups. Histopathology displayed a local reaction to the spinal catheter and a mixed acute and chronic inflammatory reaction. No group differences were observed. These results suggest that, at the neostigmine concentration of 1 mg/ml in the rat and dog and in doses up to 4 mg/day in the dog, there is no evidence of spinal tissue toxicity that can be attributed to the drug. This result, observed in two species, suggests that intrathecal neostigmine given in this manner is without distinguishable toxicity in these two models.
Key words: Animal: dog; rat; sheep. Neostigmine: blood flow; intrathecal. Toxicology: neostigmine; spinal.
INTRATHECALLY administered carbachol will yield a dose-dependent antinociception in rats [1–4] and cats,  with the effects being uniformly antagonized by muscarinic antagonists. Cholinergic binding is found in the spinal dorsal horn, notably in the regions of the substantia gelatinosa. [5–11] Rhizotomies significantly attenuate muscarinic binding in the spinal dorsal gray matter, [12,13] suggesting a possible presynaptic effect on sensory neurons. Such mechanisms have been posited for a variety of spinal receptor classes, such as those of the mu and delta opioid and alpha2adrenoceptor type, known to selectively alter pain behavior ().
Studies on spinal cholinergic systems have revealed significant levels of choline acetyltransferase and acetylcholinesterase, [15–17] and this activity appears likely to be associated with local interneuronal systems.  Given the rapidity with which acetylcholine is inactivated by cholinesterase, it was a reasonable hypothesis that blockade of this enzyme might serve to produce a significant analgesia by itself or to augment the activity of other antinociceptive systems. In this regard, it has been found that the systemic [19–21] or spinal delivery of cholinesterase inhibitors such as neostigmine or edrophonium will produce a powerful analgesia in the rat,  depress nerve stimulation evoked somatosensory reflexes,  and synergize with the antinociceptive effects of spinal alpha2agonists in the sheep  and rats. [22,25] This latter effect is of interest, because it has been reported that spinal alpha2agonists may serve to release acetylcholine in the spinal cord.  Of equal importance, it was observed that neostigmine would prevent the hypotension that was otherwise observed by the action of spinal alpha2agonists.  These joint benefits accruing from the effects of spinal cholinesterase inhibition suggest that it might serve a potential role in clinical pain therapy. Because the site of action is spinal and because cholinesterase inhibitors can have undesired effects after systemic delivery, the appropriate route of delivery in humans would be by the spinal route.
Before initiating clinical trials, it is of particular importance to define the safety of such agents delivered by this route. To address these safety issues, we initiated a series of investigations to define the effects on behavior and spinal histopathology of neostigmine methylsulfate given by bolus intrathecally over 4 days in rats and by chronic intrathecal infusion over 28 days in dogs. These studies, carried out in well defined animal models over a range of doses, can provide information on the safety of this agent given by the spinal route.
These safety studies were carried out according to protocols approved by the Institutional Animal Care and Use Committee at the University of California, San Diego. There are two components to these studies: rat and dog.
Animals. Adult Sprague-Dawley male rats (aged 6–8 months, weight 275–350 gram; Harlan Industries, Indianapolis, IN) were employed in these studies. They were maintained after surgery in individual microfilter cages with ad libitum food and water on a 12 h/12 h light/dark cycle.
Surgical Preparation. Specially prepared lumbar intrathecal catheters were placed to permit repeated spinal drug delivery. Rats were anesthetized (50/50 oxygen/air) with 3–5% halothane. Each rat was placed in a stereotaxic head holder and fitted with a face mask delivering 1–3% halothane in an air/oxygen mixture. The surgical field at the dorsal head and neck was clipped and surgically prepared with betadine and alcohol. A skin incision was made, and muscles were retracted to expose the cisternal membrane. A small incision was made in the membrane and a spinal catheter passed caudally 9 cm to the lumbar level. The exposed end of the catheter was tunneled forward and externalized on the top of the skull. The incision was sutured, and the catheter was flushed with 10 micro liter of saline and then plugged with a steel pin. On anesthetic recovery, if the rat showed a deficit of fore or hindlimb function or detectable asymmetry of the spinal axis, it was sacrificed.
Catheter Preparation. A 12-cm length of PE-10 (polyethylene, ID 0.28 mm, OD 0.61 mm, Intramedic, Becton Dickinson) tubing was used. A loose overhand knot, which does not occlude the tubing, was thrown so that lengths of 3 cm and 7 cm, external and indwelling ends, respectively, leave the knot. The knot was dipped in dental acrylic to prevent tightening or loosening. The indwelling catheter end was evenly stretched by 75%, which results in a significant reduction in the diameter and stiffness of the tubing, left 2 h, and then cut to 9 cm from the knot.
Injection Protocol. Drugs were prepared in sterile saline (USP, 0.9%) to be delivered in a volume of 10 micro liter. After each injection, a second 10 micro liter of sterile saline was injected to flush the catheter. Injections were delivered over 30 s using a hand-driven syringe pump.
Treatment Paradigm. The rat study was performed in two phases: dose ranging and repeated fixed dosing.
Phase 1: Dose Ranging. To establish the maximum dose that is acutely tolerated by the animals, sequential rats received single bolus doses of drug, and the behavioral toxicity was assessed (see below). In the absence of toxicity, the subsequent rat received a 0.5 log unit increase in dose. If, in a given rat, toxicity was observed, the next rat received a 0.5 log unit lower dose. This paradigm was repeated until four rats had been observed to have toxicity at a given dose. This method, as described by Dixon and Massy,  was employed to rapidly approximate the upper toxic dose. The highest dose that can be used (i.e., as defined by solubility in saline) or the highest dose compatible with lack of 24-r morbidity was thus defined as the high dose (HD24). A drug dose was defined as toxic in the following conditions:(1) death at any time within 24 h of injection; or (2) presence at 0–2 h or development by 24 h of any of the following: loss of hindlimb motor function (either due to spasticity or flaccidity); asymmetric body posture; loss of righting reflex; prominent, spontaneous squeaking or signs of protracted agitation persisting after the acute for a period of longer than 3 min after drug injection; or any syndrome that produced symptoms requiring sacrifice in the opinion of the observer. Based on these criterion, the HD24dose was found to be 1 mg/ml in 10 micro liter.
Phase 2: Repeated Fixed Dosing. Rats were prepared with chronic intrathecal catheters as described and randomly assigned to one of the following treatment groups. Based on the dose ranging study outlined above, the groups were: saline (n = 6), neostigmine (5 micro gram/10 micro liter), and neostigmine (10 micro gram/10 micro liter). Each group received as a single bolus injection on 4 sequential days. On day 6, the animals were sacrificed. The timing of the treatment sequence is summarized in Table 1. Immediately before and then 20 min after each daily injection, animals were assessed for general behavior, appearance, weight, and function as described below in Table 2. The timing of the postinjection testing was arranged to coincide with the time of peak antinociceptive effect as defined in previous work with intrathecal neostigmine. .
Test Drug and Solutions. Neostigmine methylsulfate (International Medication Systems, El Monte, CA) was provided in 1-mg/ml glass ampules. Dilutions of this drug were made with an aseptic technique using sterile saline (0.9%, USP). Solutions were filtered through 0.2-micro sterile bacterial filters (Gelman Sciences, Ann Arbor, MI).
Clinical Observations. General Behavior. All animals were assessed once daily for changes in behavior, which include a measure of arousal, motor coordination, and motor tone (AMCMT).
A description of the scoring procedures for AMCMT is given in Table 2. In addition, specific behavioral descriptors were applied as warranted and scored on basis of 0 (none, normal) to +3 (maximum). In the current study, specific descriptors included tremor, rigidity, and salivation. For analysis, the sum of the individual daily scores for each descriptor was calculated for each rat over the four daily injection periods and divided by 12 (the maximum possible score that could be observed over the 4-day period) to give the percent of the maximum possible score.
Analgesic Testing. Thermal analgesia was assessed using the hot plate test. The animal was placed on a 52.5 degrees Celsius metal surface (20 x 30 cm surrounded by a 45-cm high plastic wall). The latency to licking of the hindpaws was measured. If no response occurred in 60 s, the animal was removed and that time assigned as the response. Data are expressed as percent of the maximum possible effect (%MPE), where %MPE = 100 x [post-drug response - pre-drug response]/[cutoff time (60 s)- pre-drug response]).
Rat Sacrifice. On completion of the study on day 11, the animal was sacrificed. After induction of a deep anesthetic state (pentobarbital sodium, 35 mg/kg, intraperitoneally), the left ventricle was cannulated and the rat perfused with 150–200 ml of heparinized saline (1,000 micro liter), followed by perfusion with 100–150 ml of 10% formalin using an infusion pump set to deliver at a pressure of approximately 120 mmHg. The vertebral column from the lower sacral to the cervical area was removed. Excess tissue was removed, leaving the vertebral column intact. The spinal column was fixed in 10% formalin for at least 1 week and then decalcified overnight in 1% hydrochloric acid (S/P Decalcifying Solution, Baxter). Representative cross-sections of the spinal column with spinal cord were taken from thoracic (with catheter) and lumbar (below catheter) regions, as well as at the catheter tip. The blocks were further decalcified overnight, then embedded in paraffin, sectioned at a thickness of 6–7 micro, and stained with hematoxylin and eosin. Coded sections from all animals were concurrently examined by the histopathologist, who was unaware of drug treatment. Particular attention was given to the presence or absence of fibrosis or other reactions around the catheter; dural thickening or other reaction; inflammation in the epidural space, leptomeninges/subarachnoid space, or spinal cord parenchyma; and other damage to the spinal cord parenchyma or nerve roots, such as neuronophagia, microglial nodules, demyelination, or gliosis. All tissue samples were graded and assigned scores. The degree of chronic and/or acute inflammation and fibrosis was graded as normal (0), mild (+), moderate (++), or severe (+++).
Animals. Fifteen adult male beagles (aged 12–20 months, weight 13–17 kg; Harlan Industries, Ridglan Farms, Mt. Horeb, WI) were employed in these investigations. Each animal was housed alone in individual dog runs with freely available food and water on a 12 h/12 h light/dark cycle.
Surgical Preparation. After receipt, the animals were conditioned for a minimum of 5 days, and during this time, they were adapted to experimental protocols and handling and the placement of a nylon vest. To be able to place the spinal catheter, the dog was sedated (atropine, 0.04 mg/kg, and xylazine-Rompun, 1–2 mg/kg, intramuscularly), given an intramuscular injection of penicillin G, and procaine (20,000 units/kg), and brought to an anesthetic depth by mask administration of halothane (3–5%), and then the trachea was intubated. The dog was maintained under spontaneous ventilation with 1–2% halothane and 50% N sub 2 O/50% Oxygen2. The dogs were continuously monitored for oxygen saturation; inspired and end-tidal values of halothane, carbon dioxide, nitrous oxide, and oxygen; and heart rate. Surgical areas on the back of the neck and head were shaved and prepared with alcohol and a betadine scrub, and the dog was placed in a stereotaxic head holder. After draping and using sterile technique, the cisterna magna was exposed, and a small incision (1–2 mm) was made. The intrathecal catheter was inserted and passed caudally a distance of 40 cm, to a level corresponding approximately to the L3-L4 segment. Presence of the catheter in the intrathecal space at the time of surgery was confirmed by the free withdrawal of cerebrospinal fluid (CSF). A small stainless steel screw was placed in the skull and the catheter tied to the screw. The catheter was tunneled subcutaneously and caudally to exit on the upper left back at the level of the scapula. The incision was closed in layers with 3–0 Vicryl suture. On closure of incisions, the halothane was turned off and the animal allowed to recover under observation. Incision areas were treated with a topical antibiotic. Butorphanol tartrate (Torbugesic, 0.2 mg/kg, intramuscularly) was administered for postoperative pain medication. At this time, the catheter was connected to the infusion pump placed into a vest side pocket, and an infusion of sterile saline (2 ml/day) was initiated.
Catheter Preparation. The catheter was constructed from polyethylene tubing (PE-50, 0.97 mm OD, Intramedic, Becton Dickinson) stretched by 30%, making the nominal diameter 0.6 mm.
Infusion Protocol. A nylon vest (Alice King Chatham Medical Arts, Hawthorne, CA) was placed on each dog after surgery and an infusion pump (CADD-1 pump; Pharmacia Deltec, St. Paul, MI; model 5100 HFX) installed in the vest pocket, where it was connected to the externalized end of the catheter. All pumps were checked for rate of delivery before and after the infusion period and checked daily to confirm volume delivered.
Treatment Paradigm. The dog portion of the study was performed in two phases: dose ranging and continued fixed dose.
Phase 1: Dose Ranging. For the dose-ranging study, using three dogs, the infusion pumps were loaded on day 3 after surgery with a low drug concentration, which was 0.1 times that found to produce measurable effects in rats (this was found to be 1 mg/ml), and 2 ml/day was delivered. At 4-day intervals, the concentration of drug was raised by half a log unit. When the highest concentration (1 mg/ml) was reached, the volume of infusion was raised by 1 ml every 4th day. This was continued through 28 days. Thus, the animals received 0.2 mg/2 ml, 0.6 mg/2 ml, 2 mg/2 ml, 3 mg/3 ml, 4 mg/4 ml, and 5 mg/5 ml. At the 5 mg/5 ml dose, significant tremor, reduced blood pressure, and salivation were noted. Based on these results, it was decided that a dose delivery rate of 1 mg/ml drug could be given in a volume of 4 mg/24 h in the subsequent fixed-dose paradigm.
Phase 2: 28-day Fixed Dose. For the 28-day infusion study, after catheter placement, dogs were randomly assigned to one of two groups to receive saline (vehicle, n = 4) or neostigmine (4 mg/4 ml, n = 6). After 28 days of infusion of the assigned drug, dogs were sacrificed. Food intake and general behavior were assessed daily. Respiration rate, heart rate, and blood pressure were taken presurgery and on days -5, 2, 9, 10, 15, 20, and 28. The cisternal CSF clinical chemistry and concentrations of the target drug were assessed on the day of surgery, at the time of catheter placement, and on day 28, the day of sacrifice, by percutaneous puncture.
Clinical Observations. Dogs were observed twice each day. The following observations were made periodically.
General Behavior. Assessment of arousal, motor coordination, and motor tone were made as described in Table 2.
Thermal Nociceptive Reflexes. The thermally evoked skin twitch was examined using a probe with a 1 cm2surface area maintained at 62.5 plus/minus 1 degree Celsius. The probe was applied to shaven areas at the thoracic or lumbar level of the back. Typically, this application results in a brisk contraction of the local musculature within 1–3 s. On the appearance of this response, the probe was removed and the latency to respond recorded. Failure to respond within 10 s was caused to remove the probe and assign that value as the latency.
Respiratory Rate, Heart Rate, and Blood Pressure. Respiratory rates were measured by observation of chest expansion. Heart rates and blood pressures were assessed by the use of a Dinamap 8100 automatic pressure monitor (Critikon) with a pediatric cuff placed at the shaven base of the tail.
Body Temperature. Temperature was assessed by the placement of a rectal thermometer with a digital readout.
CSF Chemistry. After anesthetization on the day of sacrifice, a percutaneous puncture of the cisterna was made using a 22-G (1.5-inch) spinal needle, and a CSF sample was taken for assessment of specific gravity, erythrocyte count, leukocyte count, glucose, and protein. Plasma samples were taken concurrently.
Measurement of Neostigmine. The concentrations of neostigmine were measured in cisternal CSF and plasma samples using an HPLC separation coupled with electrochemical detection, according to methods published by Morris et al.  Absolute assay sensitivity was 1 pg/ml, with an interassay coefficient of variation at 20 pg/ml of 13%.
Dog Histopathology. Twenty-eight days after initiation of drug infusion, the dogs were sacrificed. After induction of a deep anesthetic state (sodium pentobarbital, 35 mg/kg, intravenously), the animal was manually ventilated to maintain adequate oxygenation. A percutaneous puncture of the cisterna magna was performed and CSF withdrawn for analysis. The chest was opened and a large-bore cannula placed in the aortic arch, through which was perfused 4 1 of saline followed by 4 1 of 10% formalin using an infusion pump set to deliver at a pressure of approximately 120 mmHg. After fixation, the dura was exposed by an extensive laminectomy of the spinal canal and the lower brainstem, being careful to leave the catheter and dura undisturbed. Dye was injected through the catheter to determine its integrity, visualize the position of the intrathecal catheter, and determine the spread of the dye around the catheter. The catheter was examined and photographed. The spinal cord was removed in four blocks (cervical, thoracic, caudal, and rostral from the catheter tip), taking care to keep the dura intact, and placed in formalin. After fixation, tissue blocks were embedded in paraffin and then prepared and examined as described above for the rat studies.
All data are presented as mean plus/minus SD for quantitative data. Comparison of changes over time were assessed using a one-way or two-way analysis of variance (ANOVA) as necessary. Statistical analyses were accomplished with Systat (Version 5.2). For ordinal data (such as rigidity indexes or presence or absence of salivation) or histopathologic comparisons across groups, data were analyzed using nonparametric test rank-order methods. In the rat studies, where there were three dose groups (saline 5 and 10 micro gram), comparisons across doses were established using the Jonkheere test for ranked data.  For statistical comparison of changes across time in the dog studies, mean data were computed for the preinfusion observations (days-8 to -1, 1 to 4, and 20 to 28). Days 20–28 and repeated-measures analyses were employed in the comparison across intervals.
Physiologic Effects. Antinociception. The intrathecal injection of saline over 4 days was without effect on measured sensory (hot plate, Figure 1) or motor endpoint (Figure 2). In contrast, the intrathecal injection of neostigmine resulted in an elevation in the hot plate response latency after each injection (two-way ANOVA across doses, F = 83.6, P < 0.0001). Individual comparisons revealed that the increase in the response latencies were dose-dependent such that 10 micro gram neostigmine > 5 micro gram neostigmine > saline (P < 0.001). Comparisons of preinjection latencies across the 4 days for each group revealed no differences, indicating that rats always returned to baseline before the next injection. The magnitude of the antinociception observed after each neostigmine injection for each of the 4 days showed no evidence of abatement (one-way repeated measures ANOVA of post-neostigmine responses across days: P > 0.20 for each treatment group).
Motor. Examination of motor tone revealed that the intrathecal injection of neostigmine resulted in an increase in the incidence of motor tone (Figure 2). This change in tone was evidenced by an increased stiffness of the chest wall (rigidity) and the perception of tremor by the observer when the paw was gently withdrawn or the chest wall compressed (tremor). The onset of the increased tone was within 5–10 min after injection. There was always a complete return to normal motor tone before the subsequent injection 24 h later. Comparison of the cumulated motor tone scores for each animal across the 4 days revealed the increase was dose-dependent (Jonckheere comparison of 10 micro gram neostigmine > 5 micro gram neostigmine > saline, tremor critical value 3.63, P < 0.001, rigidity critical value 4.33, P < 0.001).
Salivation. Salivation was noted, and as with motor tone, it occurred with a short latency, persisted over the 4-day interval, and was dose-dependent (Jonckheere comparison of 10 micro gram neostigmine > 5 micro gram neostigmine > saline (tremor critical value 2.26, P < 0.05). Salivation always resolved before the next injection.
Toxicologic Indexes. All rats survived the respective treatments to the scheduled sacrifice at 2 days after the last injection. Measurement of body weight indicated that there was no statistically significant change in body weight between days 0 and 6 for any group. However, although there were no differences between groups between days 0 and 6, after the high dose of neostigmine, this group was significantly reduced as compared to saline (Table 3).
Histopathology. At the time the tissue blocks were prepared, catheters were identified in the intrathecal space of all rats. In the histologic sections, no catheter site was seen in the thoracic sections in three rats, and in one rat, no catheter site was identified in the region of the catheter tip. Multiple sections were made in these animals, suggesting that little reaction was present around the catheter, allowing it to fall away during processing for histology.
In the remaining rats, the catheter site was identified within the arachnoid or subarachnoid space as a rim of fibrin, fibroblasts, and inflammatory cells of variable thickness, surrounding a clear space (catheters dissolved during processing;Figure 3). The inflammatory infiltrate associated with the catheter was predominantly mononuclear (most lymphocytes) with occasional polymorphonuclear cells and multinucleated giant cells. The proportion of various types of inflammatory cells was similar in all rats. There were variable amounts of inflammatory cells in the subarachnoid space not immediately adjacent to the catheter. Inflammatory cells were seen in the intrathecal nerve roots in 10 of the 18 rats; this was present in all groups (5 in the 10 micro gram group, 2 in the 5-micro gram group, and 3 in the saline group). Vasculitis of the leptomeningeal vessels was seen in three rats, one in each treatment group. Dural fibrosis and thickening adjacent to the catheter site was seen in six (one in the 10-micro gram group, two in the 5-micro gram and three in the saline group), and dural chronic inflammation was seen in seven (two in the 10-micro gram group, three in the 5-micro gram group, and two in the saline group). Epidural fibrosis and/or chronic inflammation was seen in seven rats (three in the 10-micro gram group, two in the 5-micro gram group, and two in the saline group). There was erosion of the vertebral bone adjacent to the catheter site in four rats (one in the 10-micro gram group, two in the 5-micro gram group, and one in the saline group).
The degree of inflammation and fibrosis at the catheter site near the tip and above the tip (with catheter) was graded as described in Methods. A summary of this data is shown in Table 4. These two grades were combined to give a total score for the reaction around the catheter and were ranked. The Jonckheere test for ranked data showed no difference in the reaction around the catheter for the three treatment groups (saline vs. 5 micro gram vs. 10 micro gram).
Almost all of the rats showed indentation of the spinal cord by the catheter site. The diameter of the catheter was typically one-fifth the width of the spinal cord. In four rats, the leptomeningeal inflammatory infiltrate extended into the perivascular spaces of the spinal cord, and in one, the infiltrate extended into the spinal cord parenchyma. In one animal (10 micro gram), there was focal white matter vacuolization beneath the catheter site, but similar vacuolization was seen in other areas of the periphery of the spinal cord, and the appearance is most consistent with fixation artifact. Despite the evident indentation by the catheter, there was no evidence of neuronal injury, gliosis, or demyelination in rats that received even the highest dose of neostigmine.
Physiologic Effects. Antinociception. The baseline skin twitch latency before the initiation of infusion was 4.5 plus/minus 1.6 and 5.7 plus/minus 1.0 s for saline and neostigmine groups, respectively. Continuous intrathecal infusion of saline at 4 m/day or neostigmine in the dose of 4 mg/4 ml/day did not result in a significant increase in the skin twitch latency over the 28-day period, even during the acute interval immediately after initiation of infusion. Thus, comparing the mean skin twitch response latency measured during the preinfusion with that observed on days 1–4 (5.0 plus/minus 1.6 and 6.0 plus/minus 0.7 s for saline and neostigmine, respectively) revealed no statistically significant effect for neostigmine as compared to the mean preinfusion response or as compared to saline (P > 0.20).
Heart Rate. There was no persistent change in the heart rate of dogs administered intrathecal saline, although there was a modest but statistically significant decline over time after the initiation of the infusion of neostigmine (Figure 4and Table 5). The bradycardic effect was observed within 2–3 days after the initiation of infusion and was sustained over the 28 days of blood pressure: Although modest variations in mean, systolic, and diastolic pressure were noted over the 28-day period in dogs administered saline or neostigmine (Figure 5), no differences were considered physiologically significant.
Respiration. There were no systematic changes in the respiration rate of dogs administered saline or neostigmine, although there was a tendency for saline dogs to show a slower rate than the neostigmine dogs. These differences were slight and not considered behaviorally significant (Table 5).
General Behavior. Mild somnolence (arousal score -1) was observed in four of the six neostigmine dogs at the outset of neostigmine infusion, but this had disappeared by the 4th-8th day. Motor function as defined by hindlimb placing and stepping reflexes, strength of withdrawal, and ability to ambulate was essentially unimpaired at all doses of all dogs (motor coordination score 0). Motor tone was observed to show a modest increase during the period of the infusion of neostigmine, but this was limited to four of the six neostigmine dogs (motor tone + 1) and was not considered debilitating.
Body Temperature. As indicated in Table 6, modest changes were observed in both saline and neostigmine animals over the 28-day interval, but these differences were not believed to be clinically significant.
Level of Neostigmine in the Plasma and Cisternal CSF. At the time of sacrifice, the levels of neostigmine in the cisternal CSF were measured. The plasma levels were below assay sensitivity. In the cisternal CSF, the levels of neostigmine at the time of sacrifice were found to be 38 plus/minus 18 ng/ml (micro plus/minus SD, n = 5; one animal not included because of sample contamination).
Toxicologic Indexes. All animals survived the treatment and completed the 28-day treatment schedule. There was no indication of gastrointestinal dysfunction. No vomiting was observed in any of the animals.
Mean body weights (kg) before the initiation of infusion were not different for the drug and saline groups (Table 6). Over the 28-day infusion interval, all dogs showed modest losses in body weight with no difference between groups.
CSF Pathology. Values measured in each dog are summarized by each treatment group in Table 7. There were no treatment-dependent changes in CSF protein or specific gravity, whereas glucose levels displayed a small but statistically significant increase. Microscopic examination of the CSF revealed no notable pathology in any group. The elevated levels of erythrocytes in several of the animals was interpreted as the result of sample contamination during the percutaneous sampling.
Histopathology. Summary of histopathologic examination by dog is presented in Table 8. At the time of sacrifice, all catheters in the dogs were observed to lie within the intrathecal space. At histology, the catheter site was identified in the leptomeninges or subarachnoid space as a rim of well defined fibrous tissue of varying thickness (Figure 6). The fibrous rim was composed of fibroblasts and well established collagen. In six animals, there was granulation tissue with neovascularization at the periphery of the reaction around the catheter. The inflammatory reaction associated with the catheter was much less than in the rats and composed almost entirely of lymphocytes and plasma cells. In only two animals (one neostigmine, one saline) were small numbers of polymorphonuclear cells seen around the catheter site. Several animals had small collections of hemosiderin-filled macrophages near the inner surface of the dura. The number of inflammatory cells in the subarachnoid space away from the catheter was also less in the dogs than in the rats. In one dog (neostigmine), there was a blood clot in the subarachnoid space next to the catheter. Intrathecal nerve root inflammation was seen in only one dog (neostigmine). Leptomeningeal vasculitis was seen in four dogs (one saline, three neostigmine). Dural fibrosis and thickening adjacent to the catheter site was seen in all dogs, and most (9 of 13) showed focal chronic inflammation in the dura. Epidural inflammation was seen in only two dogs (neostigmine, one phase I, one phase II).
Vertebral bone changes were not seen grossly in the dogs, and the bone was not evaluated histologically.
Indentation of the spinal cord was less frequent in the dogs than in the rats and, when present, was of a lesser degree (Figure 6). Nine dogs had minor indentation of the spinal cord adjacent to the catheter site. There was no extension of the leptomeningeal inflammatory infiltrate into the perivascular spaces of the spinal cord in the dogs. One dog (saline) had focal white matter vacuolization in the lateral column of the spinal cord, which was beneath the catheter site. Intact axons could be seen within the vacuoles, suggesting that this represented focal myelin edema or degeneration. There was no evidence of neuronal injury or gliosis in any dogs.
The degrees of inflammation and fibrosis around the catheter and inflammation in the subarachnoid space away from the catheter were graded as described above. These grades were combined to give a total score for the reaction to the catheter and ranked. Comparisons of saline versus neostigmine dogs showed no differences in the reaction around the catheter in sections with catheter or at the catheter tip.
The current work was designed to define the safety of spinally delivered neostigmine before assessing the activity of this agent in humans.
Behavioral Effects of Intrathecal Neostigmine
The spinal delivery of neostigmine resulted in a dose-dependent increase in the hot plate response latency and the appearance of tremor and salivation in the rat. All effects were reversible over time. It is believed that these actions reflect on the enhancement of spinal cholinergic activity secondary to the inhibition of cholinesterase (see Introduction). The effects are reversed by spinal muscarinic antagonists, mimicked by spinal muscarinic agonists, and are reversible by muscarinic antagonists. [1,5,22] In the current study, four repeated injections of neostigmine failed to show any diminution of acute activity over the time course of the four repeated tests, suggesting a possible failure to see any tachyphylaxis over this interval of exposure. In the dog, infusions of neostigmine resulted in a mild tremor activity and bradycardia. Unlike the rat, there was no change in the single nociceptive endpoint, the thermally evoked skin twitch. Detweiller et al.  also noted in sheep little change in the nociceptive endpoints in that model when neostigmine is given alone. In sheep, spinal neostigmine enhanced antinociception produced by the alpha2agonist, clonidine  while inhibiting clonidine-induced hypotension.  Antinociceptive synergy also has been demonstrated in the rat between cholinesterase inhibitors and alpha sub 2 agonist. [22,25] Such interactions have not been studied in the dog model.
Histopathology after Spinal Neostigmine
In both the rat and the dog, the control animals displayed a distinct reaction believed to be evoked by the presence of the catheter. The histologic appearance of the catheter site in the dogs differed from that in the rats. There are two major factors likely contributing to this difference:(1) the longer period of catheter placement in the dogs (31 day vs. 11 day) and (2) the relatively smaller size of the catheter relative to the spinal cord and intrathecal space in the dogs. Thus, in the rat, but not in the dog, there was a mild distortion of the spinal parenchyma reflecting the relatively large size of the spinal catheter to the respective intrathecal space. In both species, the catheter alone resulted in the evolution of a mild local response that consisted of a mixed acute and chronic inflammatory reaction. Such a reaction has been reported in dogs implanted with similar catheters for this interval. [30,31].
Examination of the histopathology in the rat and dog revealed no systematic differences in the response evoked by the vehicle or by the drug. Further, examination of the cisternal CSF taken after 28 days of infusion revealed that there were no differences between the drug and vehicle group. The analysis of CSF obtained from the cisternal membrane at the time of sacrifice is of particular importance because levels of protein, specific gravity, and cytology are measures of acute and chronic inflammatory responses.  Previous studies using intrathecal catheters have indicated that the 28-day implant results in a modest but statistically significant increase in protein and glucose concentrations that are comparable to those observed in the current study. [30,31] Failure to see differences between the drug and vehicle groups provides further support for the lack of a proinflammatory reaction that was contributed by the spinally infused neostigmine.
Recent studies have examined the effects of intrathecal neostigmine on spinal cord blood flow in the sheep model using colored microspheres. In concentrations employed in the current study, neostigmine did not alter spinal cord flow. (a).
The lack of apparent behavioral dysfunction, the lack of any specific evidence of tissue injury in two species, and the absence of an effect on spinal cord blood flow suggests the relative safety of intrathecally delivered neostigmine within the range of doses and exposure provided by the drug treatment paradigm.
Preclinical Models of Spinal Toxicity
Given the apparent absence of toxicologic signs, an important question would be how sensitive are the test models to displaying toxicity, and given the dose of agent employed, how robustly were the test systems challenged?
Rat. Extensive use has been made of the rat with a chronically placed intrathecal catheter for defining the toxicity of different agents. Consideration of the published literature indicates that the spinal delivery of a number of agents has been shown to produce both behaviorally and histopathologically defined indexes of tissue injury. Thus, spinally administered dynorphin, [33,34] somatostatin, [33,35,36] ICI 174864  and DPDPE-TFA (Malmberg and Yaksh, unpublished observations) have been shown to produce significant signs of neuronal damage after even a single dose of the respective agent. In contrast, the spinal delivery of multiple intrathecal doses of DADL (7 daily boluses).  DPDPE HCl (4 daily boluses; Malmberg and Yaksh, unpublished data), clonidine, guanfacine (14 daily bolus).  R-phenylisopropyl adenosine (14 daily bolus injections),  sufentanil, alfentanil, morphine,  ketamine, MK801, dextrorphan (4 daily bolus: Yaksh, Quint, Marsala, unpublished), 3-(2-carboxypiperazin-4-yl) propyl-1-phosphonic acid, kynurenic acid,  carbachol (14 daily bolus),  and droperidol  have been shown to be without effect on the histologic appearance of the rat spinal cord. While the studies cited cannot exclude the likelihood that a novel agent deemed safe in that model will not show toxicity later in other paradigms, the positive results observed with even the single-dose regimens and the lack of at least behavioral signs of toxicity with repeated or continued dosing using a wide variety of agents, provides supportive evidence that the rodent model is able to detect toxicity with some but not all agents. Importantly, agents in which there have been negative signs of toxicity include a wide range of agents that have been given in humans, including clonidine, morphine, alfentanil, and sufentanil. Aside from evident morphologic changes, the incidence of toxicity after spinal drug delivery frequently appears to correlate with irreversible hindlimb motor dysfunction, and this likely provides a clear index of focal neurotoxicity. Thus, after single spinal intrathecal injections, dose-related, irreversible hindlimb dysfunction has been observed with somatostatin, [33,35,36] Dynorphin A-(1–17), (2–17), (1–18), (1–7), and (3–8), [34,45,46] ICI 174864.  These effects are typically associated with a general debilitation of the animal and a variety of changes such as in the micturition response, suggestive of autonomic dysfunction. In contrast, other spinal agents, such as arginine vasopressin, may produce reversible weakness which is without effect on neuronal appearance.  Together, the above comments suggest that multiple bolus dose spinal delivery in the rat can serve as an important index of local drug toxicity.
Dog. The dog has been an alternate model for spinal toxicity testing. To date, a variety of agents have been examined in the dog model, including morphine (28-day intrathecal bolus),  sufentanil (28-day intrathecal bolus),  alfentanil (28-day intrathecal bolus),  bupivacaine (continuous, 3–14 weeks),  baclofen (28-day intrathecal continuous infusion),  and clonidine (28-day epidural continuous infusion).  Typically, based on behavior, physiology, CSF chemistry, and systematic light level microscopy, these studies have not observed evidence of histopathology. In our experience, the single exception has been the observation that DPDPE-TFA (10-day daily bolus) resulted in a dose-dependent inflammatory reaction. Although the mechanisms of this are not known, it coincided with the effects of the same agent in the rat (see above). Not withstanding this single exception in which spinal toxicity was observed in the dog, it might be argued that either the dog is insensitive or the dose exposure has been inadequate. Alternately, it should be noted that, because of the expense of the dog, the rational procedure of assessing the safety of the candidate agent in a smaller model (such as the rat) might preclude the further examination of agents with questionable safety in the more complex and costly model.
Factors Defining the Robustness of the Drug Challenge
We have argued that local tissue toxicity is defined by the local drug concentration that is sustained around the catheter tip.  Although total doses delivered may be small, the concentrations in the vicinity of the tip of the spinal catheter may be high. Although there are no systematic data, the role of concentration in this regard has been demonstrated in humans with 5% lidocaine. Under normal circumstances, there is a significant dilution and a long history of apparent safety in humans. In contrast, the use of small-bore catheters that preclude barbotage and redistribution [50,51] apparently led to an increased incidence of radiculopathies,  a finding in concert with preclinical studies of drugs placed near nerves. [53–55] Thus, a legitimate challenge is to deliver the highest dose possible in a small volume (to achieve a high local concentration) or provide a continued delivery, with the highest concentration available. In the current study, the highest formulated concentration available is 1 mg/ml. In the rat, this dose delivered in 10 micro liter resulted in a significant behavioral effect. Higher doses resulted in significant tremor and precluded its routine use. In the dog, dose-ranging studies revealed that the animal could readily tolerate 1 mg/ml in a volume up to 4 ml/day. Thus, these doses were employed. An important question related to the degree to which the spinal cord is exposed to the drug. After bolus delivery, as in the rat, the initial concentration undergoes an acute dilution in the local CSF and then drug is cleared, either into tissue and thence the vasculature or by rostral bulk flow. [56,57] In the dog, the continued infusion would be considered to result in a persistent drug presence. To the degree that the infusion rate is high compared to the local rate at which new fluid reaches the spinal infusion site, then the concentration of drug at the site of infusion will tend to be near the concentration of the infusate. Based on the supposition that the volume of dilution is relative to the volume of the local CSF space, we have estimated that the effective local volume of distribution space in the human lumbar cord is approximately 2–3 times that of the dog. In the current study, we expect that the 4 mg/4 ml/day local distributes into a volume that is one-half to one-third that of humans. Thus, while hypothetical, if there is a comparable clearance, we consider it likely that the effective concentration in dogs would mirror a dose that is in excess of 8–12 mg in humans. Although the dose to be required in humans is not known, neostigmine has been delivered spinally in spinally injured humans to evoke ejaculation and doses considerably less than these have been reported. [58,59] This suggests that the dog infusion model with its 28-day continuous exposure should provide a robust assessment of the apparent safety of doses of neostigmine that are considerably in excess of that which should be required in humans.
In summary, we believe that absence of toxicity in these two preclinical models over dose ranges chosen on the basis of extremes in tolerable concentrations provides strong evidence for the presumption of safety and suggests their applicability for the initiation of phase 1 safety assessments in humans.
The authors thank Dr. Chuanyo Tong, Bowman Gray School of Medicine, for the performance of neostigmine analysis, and Dr. R. Sharma and Dr. K. Miller, Department of Anesthesia, University of California, San Francisco, for help in setting up the neostigmine assay and preliminary analysis. The authors also thank Dr. Bruce Burris, Department of Pathology, University of California, San Diego, for the preparation of the histopathology.