Administered in large doses, alfentanil produces limbic system (hippocampus, amygdala, septal nucleus, cingulate and entorhinal cortices, dentate gyrus, and association areas) hypermetabolism, seizure, and histopathologic alterations (i.e., brain damage) in rats . We previously demonstrated that the neurotoxicity associated with administration of a large dose (i.e., approximately equivalent to a "stress-free" anesthetizing dose) of alfentanil is also shared by fentanyl and sufentanil . In these prior studies, two doses of opioids were administered: a dose thought to be appropriate for basal analgesia in rats, and a dose 10- to 20-fold larger to simulate a cardiac anesthesia dosing scheme used in humans. Whether even larger doses of opioids are required to produce brain damage in these models is unknown, although rats receiving small doses (i.e., <or=to 400 micro g/kg) occasionally sustain limbic system injury .
The objectives of this investigation were to determine: (a) the relationship between opioid dose and severity and frequency of brain damage, and (b) whether, and to what extent, opioid neurotoxicity depends on electrographic activation.
Approval of the institutional animal care and use committee was obtained for all the experiments. Forty male Sprague-Dawley rats weighing 335-440 g were allowed access to food and water overnight. Anesthesia was induced with 4% halothane in 70% N2 O and 30% O2, the trachea was intubated with a 14-gauge Teflon catheter, and 0.6 mg of pancuronium bromide was injected intramuscularly and then intravenously as often as needed to maintain neuromuscular block. The lungs were mechanically ventilated (Analytical Specialties Ventilator, St. Louis, MO) for the duration of the opioid administration and up to 3 h afterward, and inspired and expired O2 and CO2 were continuously monitored by infrared spectrometry (Hewlett Packard, Waltham, MA). Mechanical ventilation was adjusted to maintain PaCO2 approximately 35 mm Hg. Biparietal stainless steel electroencephalogram (EEG) needle electrodes and rectal temperature probe were inserted. A femoral vein and artery were cannulated and arterial blood pressure was continually monitored. Anesthesia during surgery consisted of 1% halothane in O2/N2 O 30%/70% and after surgery sedation was maintained with O2/N2 O 30%/70% without halothane for 1 h. After 1 h, each rat received 1 mL of 8.4% sodium bicarbonate and was randomly assigned to one of eight groups: Control = 0.9% normal saline solution (NSS) loading dose (LD) 4 mL/kg maintenance dose (MD) 4.0 mL [centered dot] kg-1 [centered dot] h-1 NSS; 50-2 = fentanyl LD 50 micro g/kg, MD 2.0 micro g [centered dot] kg-1 [centered dot] min-1; 100-4 = fentanyl LD 100 micro g/kg, MD 4.0 micro g [centered dot] kg-1 [centered dot] min-1; 200-8 = fentanyl LD 200 micro g/kg, MD 8.0 micro g [centered dot] kg-1 [centered dot] min-1; 400-16 = fentanyl LD 400 micro g/kg, MD 16.0 micro g [centered dot] kg-1 [centered dot] min-1; 800-32 = fentanyl LD 800 micro g/kg, MD 32.0 micro g [centered dot] kg-1 [centered dot] min-1; 1600-64 = fentanyl LD 1600 micro g/kg, MD 64.0 micro g [centered dot] kg-1 [centered dot] min-1; and 3200-128 = fentanyl LD 3200 micro g/kg, MD 128.0 micro g/kg.
In each group (n = 5), the LD was administered over 1-2 min. In all groups receiving opioid, the fentanyl was lyophilized before the experiment and reconstituted in NSS to permit a LD of 4 mL/kg and MD of 4 mL [centered dot] kg-1 [centered dot] h-1. During the infusions, opioid-treated rats received O2/N2 30%/70% and control animals received 0.4% halothane in O2/N2 O 30%/70%. After 2 h of anesthetic or saline, all rats received 100% O2 and underwent tracheal extubation when alert. They were placed in an O2-infused enclosure for 1 h and then moved to a cage overnight.
Seven days after fentanyl infusion, the rats were anesthetized with halothane and underwent cerebral perfusion fixation with 10% neutral buffered formalin via thoracotomy with left ventricular cannulation. The calvaria were removed and the exposed brains were further immersion-fixed in neutral buffered formalin for at least 24 h before their removal. The brain was sliced at approximately 3-mm intervals and the slices were embedded in paraffin by routine techniques. Six-micrometer sections of each brain block were prepared, stained with hematoxylin and eosin, and evaluated by light microscopy. Fifty brain areas were examined. Degrees of neuropathologic alterations within a given anatomic region were scored based on subjective assessment of number and distribution of eosinophilic neurons and severity of tissue edema. Scores, assigned by a pathologist who was unaware of the experimental treatment, are: 0 = none; 1 = minimal; 2 = mild; 3 = moderate; 4 = marked; and 5 = severe. These grades are also semiquantitative in that they reflect the approximate number of degenerating neurons present in the affected neuroanatomic areas as follows: 1, less than 5%; 2, 6%-20%; 3, 21%-50%; 4, 51%-75%; and 5, 76%-100% dead neurons. An overall neuropathologic score was calculated for each rat by summating the pathologic scores for all brain areas examined.
Arterial blood pH, PO2, PCO2, and glucose were determined immediately before opioid infusion, five times during the 2-h opioid infusion, and 30 min after infusion.
EEG data were digitized and recorded using Labtech Notebook[R] (Laboratory Technologies Corp., Wilmington, MA) and underwent post-hoc Fourier processing. Total power was derived over the 30 min immediately before opioid infusion, during the 2 h of opioid infusion, and over the 15 min immediately after the end of opioid infusion. Using time-stamped EEG data, average power spectra were calculated for each group.
Arterial fentanyl concentrations were measured by radioimmunoassay using rabbit antifentanyl antibody. Arterial blood gases and pH measurements were made on 100-micro L samples with a Corning 178 pH/blood gas ionizer (CIBA Corning Diagnostics Corporation, Medfield, MA). Whole arterial blood glucose assays were performed on 25-micro L samples with a Yellow Springs Instruments, Model 23A glucose analyzer (Yellow Springs, OH).
Physiologic, EEG data, and plasma fentanyl concentrations were analyzed by analysis of variance (ANOVA), followed by post-hoc least significant difference testing where P < 0.05 on ANOVA. Neuropathologic scores observed for each brain area were tabulated. Due to constraints of performing multiple comparisons over many groups and brain areas, these data did not undergo statistical analysis. Rather, for each rat the pathologic grades for all brain areas examined were summated, resulting in an overall neuropathologic score for each rat. This variable was analyzed by Kruskal-Wallis ANOVA across all groups, followed by one-tailed Wilcoxon's signed rank test, comparing treatment groups with the control group, if P < 0.05 on ANOVA. Spearman correlation coefficients were calculated for the relationships between plasma fentanyl concentration, power on spectral analysis of EEG, and overall neuropathologic grade.
Physiologic variables measured during baseline and opioid infusion in rats are listed in Table 1. During the baseline period the only difference between groups was with the 200-8 group where PaO2 was lower (110 +/- 10 mm Hg) than in the control group. The lowest PaO2 in the 200-8 group was 99 mm Hg. With the LD of fentanyl, lowest mean arterial pressure was 99 +/- 4 mm Hg, occurring in the 100-4 group. However, there were no statistically significant between-group differences in post-LD maximum or minimum blood pressures. With opioid infusion several statistically significant differences from the control group occurred in PaO2, mean arterial pressure, rectal temperature, and blood glucose. Opioid infusion was associated with moderate increases in blood pressure throughout the infusion period. After 30 min of opioid infusion, blood glucose was higher than the control group in the fentanyl larger dose groups, although this difference was not present by the end of the infusion. In addition, PaO2 continued to be lower in the 200-8 group than in the control group with the lowest individual PaO2 93 mm Hg. By the end of the opioid infusion, there was a moderate increase in rectal temperature in the 100-4 and 200-8 groups, compared to the control group.
Fentanyl concentration increased progressively with each sequential increase in infusion rate (Table 2). Significant differences from the control group occurred in the 200-8, 800-32, 1600-64, and 3200-128 groups.
Fentanyl administration was associated with activation of EEG in rats, with power generally increasing with larger doses (P = 0.01 on ANOVA) during opioid infusion, and correlating with the plasma fentanyl concentrations (P < 0.001) (Figure 1). EEG power normalized to baseline was significantly greater in the 400-16 and 1600-64 groups compared to the control group during fentanyl infusion (P < 0.05) (Figure 2). Total power during (P < 0.01) and after (P < 0.001) opioid infusion correlated with the overall neuropathologic score.
There were no lesions observed in the brains of any of the control or 200-8 group rats (Table 3 and Figure 3). Statistically significant differences from the control group's overall neuropathologic score occurred in the 400-16, 800-32, 1600-64, and 3200-28 groups. Plasma fentanyl concentrations correlated with the overall neuropathologic score (P = 0.03 and P = 0.02 for samples drawn after 15 and 120 min of fentanyl infusion, respectively).
One of the rats in the 50-2 group had foci of mild eosinophilic neuron degeneration and neuropil vacuolation within the cerebellar cortex, but this neuroanatomic area was not affected in any other rat from any group. One of the five rats in the 100-4 group had mild to moderate degrees of neuron degeneration within at least four limbic system regions. Several rats in each of the 400-16 and larger dose groups also had evidence of neuropathologic change within the limbic system, with the greatest degree of effect being evident within the 800-32 group.
The primary histopathologic alteration seen in the fentanyl-dosed rats was one of acute eosinophilic neuron degeneration. On low magnification, the degenerating (necrotic) neurons were characterized by nuclear pyknosis and by the presence of perineuronal retraction spaces (Figure 4A, Figure 4B, and Figure 4C). In areas of marked neuron degeneration, secondary vacuolation of the neuropil, malacia, or hemorrhage were occasionally present (Figure 4B). On higher power examination, the necrotic neurons were found to have brightly stained eosinophilic cytoplasm (Figure 5B). Lesions were primarily confined to the limbic system and its associated areas and were most consistently present within the piriform and entorhinal cortices and within the hippocampus. For the hippocampus, the large pyramidal neurons (Figure 5A and Figure 5B) were more frequently affected than the granular neurons of the dentate gyrus (Figure 6A and Figure 6B), but both areas were affected in some animals.
Previous studies have indicated that opioids can produce limbic system hypermetabolism and brain damage [1-4]. However, in each study the neurotoxic doses of opioid were rather large, equal to or exceeding what might be considered "cardiac" or full-anesthetizing doses in rats. Therefore, it has become important to determine whether opioids can produce brain damage at smaller doses which might be considered to mimic in rats the dose ranges used in humans. Moreover, it is helpful to ascertain whether there might be a submaximal dose associated with the most extensive brain damage, as such information may give hints regarding the mechanism of neurotoxicity.
Our data confirm that fentanyl produces brain lesions in rats across a range of doses. Moreover, because the rats in our experiments underwent perfusion-fixation seven days after opioid infusion, it is likely that the lesions were not a transient abnormality. Kissin et al.  assessed fentanyl effects in rats over a range of end points which model basal analgesia, hypnosis, and stress-free anesthesia. Their data indicate 50% effective dose values for fentanyl in Sprague-Dawley rats as follows: tail clamp 10 micro g/kg, loss of righting reflex 25 micro g/kg, and lack of tachycardia to tail clamp 366 micro g/kg. Our 400-16 group corresponds to their largest dose group. Two of the five rats in this group sustained marked to severe damage in some brain areas (Table 3). Plasma fentanyl levels were similar at the beginning and end of the infusion, ensuring that a stress-free dose was maintained throughout the infusion. If Kissin et al.'s largest dose group corresponds to a stress-free large-dose opioid paradigm commonly used in humans, our data may indicate that opioid neurotoxicity can occur in rats at doses comparable to those used clinically.
Opioids produce seizures in both in vivo and in vitro preparations [1-4,6-9]. However, there have been disparate observations. For example, in one investigation, intracerebroventricular administration of morphine produced seizure activity, which was abolished by intravenous administration of morphine . Opioids are thought to produce seizures by inhibition of inhibitory neurons producing, in effect, disinhibition of glutamatergic excitatory neurons [6,8]. In addition, there are some reports of micro opioids having glutamatergic effects through other mechanisms . At the membrane level, however, opioids appear to have a uniform hyperpolarizing effect on neuronal membranes . There is a suggestion in our data that there may be dose-related disparate effects of opioids. Inspection of Figure 3 and Table 3 suggests a nonlinear relationship, with possibly a decrease in opioid neurotoxity at the largest doses. Perhaps opioid inhibitory effects may first occur in the interneurons, thereby disinhibiting glutamatergic cells, followed by hyperpolarization of all cells with larger doses and thus, inhibition of the glutamatergic cells.
Many anesthetics have both proconvulsant and anticonvulsant properties . It is thus congruent with these observations, therefore, that fentanyl also may have dose-related variable effects on seizure activity. Taken altogether, studies of these various anesthetics suggest that the brain exists in a delicate equilibrium between excitation and inhibition and, depending on the circumstances, any anesthetic may alter the equilibrium to elicit either excitation, manifest as seizure activity (possibly risking brain injury if sustained), or inhibition, manifest as EEG suppression and a hypnotic state.
Kofke et al.  in a previous study, reported that large-dose alfentanil produced (a) epileptiform activity on scalp and hippocampal EEG recordings, (b) increased glucose utilization in the ventral hippocampus, and (c) limbic system lesions with particular susceptibility in amygdala. Our neuropathologic and scalp EEG observations with progressively increasing fentanyl doses are qualitatively similar to those in this earlier alfentanil report. Given the prior report of fentanyl-induced hippocampal hypermetabolism in rats  and our pathologic data, it seems likely that limbic system epileptiform activity and hypermetabolism occurred in the rats we studied. It is unclear whether the epileptiform activity and the hypermetabolism, either jointly or singly, are essential components of the pathogenesis of opioid-induced brain damage. We have recently observed phenytoin-mediated amelioration of opioid-induced brain damage, but without complete ablation of epileptiform activity on scalp EEG .
We chose to use total power as an electrographic quantitative indicator of activation with concurrent qualitative observation of unprocessed EEG. This is a procedure which has not undergone scrutiny to determine its specificity for activation due to seizure versus increased power due to high-amplitude slowing. It is thus possible that nonspecific, nonepileptiform slowing contributed to the statistically significant electrographic effects we observed.
Fentanyl-treated rats, particularly those receiving larger doses, exhibited higher blood glucose concentration and blood pressure. This has been discussed in a prior report . Prior studies indicate no effect of glucose on seizure-induced brain damage [13,14]. It is thus unlikely that the modest hyperglycemia we observed with fentanyl administration contributed to brain damage. If hypermetabolism is important in the pathogenesis of opioid neurotoxicity, increased blood glucose should be protective. However, this is a complex notion, as we observed a protective effect with blood pressure reduction . The protective effect we observed in this earlier study was likely due in part to ganglionic blockade with its attendant catecholamine-reducing side effects .
It is noteworthy that another substance, domoic acid, produces a pattern of brain injury in rats [16,17], similar to that seen with opioids. Additionally, it has been given to subhuman primates , and has been ingested by humans [19,20] with a pattern of injury quite similar to that seen in rats. Such observations indicate that it is possible that a limbic system convulsant, which activates the limbic system in rats, can have similar effects in humans. It is thus of interest that fentanyl administered to monkeys in preliminary experiments increased qualitative glucose utilization in the temporal lobes, similar to that seen in rodents . Epileptiform activity has been observed with opioids administered to both epileptic [22,23] and nonepileptic  humans. It is unclear whether such intermittent spiking activity, such as these reports describe, is deleterious.
In many animal models, sustained seizures without hypoxemia produce brain damage [1,2,25,26]. Evidence suggests that sustained seizures also produce brain damage in humans . However, it is unknown whether such sustained seizure activity can occur in opioid-treated humans, or whether it would be neurotoxic if it did occur in humans. Thus, although there is ample circumstantial evidence that some deleterious opioid effects seen in animals occur in primates, at this time the clinical relevance of these observations can only be a source of speculation.
The authors gratefully acknowledge the assistance of Margaret Morriello, Cynthia Hladik, and Carol Campbell in manuscript preparation and Francie Siegfried for editorial review.
1. Kofke WA, Garman RH, Tom WC, et al. Alfentanil-induced hypermetabolism, seizure, and histopathology in rat brain. Anesth Analg 1992;75:953-64.
2. Kofke WA, Garman RH, Janosky J, Rose ME. Opioid neurotoxicity: neuropathologic effects in rats of different fentanyl congeners and the effects of hexamethonium-induced normotension. Anesth Analg 1996;83:141-6.
3. Young ML, Smith DA, Greenburg J, et al. Effects of sufentanil on regional cerebral glucose utilization in rats. Anesthesiology 1984;61:564-8.
4. Tommasino C, Maekawa T, Shapiro HM, et al. Fentanyl-induced seizures activate subcortical brain metabolism. Anesthesiology 1984;60:283-90.
5. Kissin I, Kerr C, Smith LR. Assessment of anaesthetic action of morphine and fentanyl in rats. Can Anaesth Soc J 1983;30:623-8.
6. Zieglgansberger W, French ED, Siggins GR, Bloom FE. Opioid peptides may excite hippocampal pyramidal neurons by inhibiting adjacent inhibitory interneurons. Science 1979;205:415-7.
7. Madison DV, Nicoll RA. Enkephalin hyperpolarizes interneurones in the rat hippocampus. J Physiol (Lond) 1988;398:123-30.
8. Caudle RM, Chavkin C. Mu opioid receptor activation reduces inhibitory postsynaptic potentials in hippocampal CA3 pyramidal cells of rat and guinea pig. J Pharmacol Exp Ther 1990;252:1361-69.
9. Urca G, Frenk H. Systemic morphine blocks the seizures induced by intracerebroventricular (i.c.v.) injections of opiates and opioid peptides. Brain Res 1982;246:121-6.
10. Chen L, Huang L-YM. Sustained potentiation of NMDA receptor-mediated glutamate responses through activation of protein kinase C by a mu-opioid. Neuron 1991;7:319-26.
11. Modica PA, Tempelhoff R, White PF. Pro- and anticonvulsant effects of anesthetics. Anesth Analg 1990;70(Part I):303-15;70(Part II):433-44.
12. Sinz EH, Kofke WA, Garman R. Phenytoin, midazolam, and naloxone protect against fentanyl-induced brain damage in rats [abstract]. Anesthesiology 1995;83:A1269.
13. Kofke WA, Ahdab-Barmada M, Rose M, et al. Substantia nigra damage after fluorothyl-induced seizures in rats worsens after post-seizure recovery: no exacerbation with hyperglycaemia. Neurol Res 1993;15:333-8.
14. Swan JH, Meldrum BS, Simon RP. Hyperglycemia does not augment neuronal damage in experimental status epilepticus. Neurology 1986;36:1351-4.
15. Werner C, Hoffman WE, Thomas C, et al. Ganglionic blockade improves neurologic outcome from incomplete ischemia in rats: partial reversal by exogenous catecholamines. Anesthesiology 1990;73:923-9.
16. Tryphonas L, Truelove J, Nera E, Iverson F. Acute neurotoxicity of domoic acid in the rat. Toxicol Pathol 1990;18:1-9.
17. Tryphonas L, Iverson F. Neuropathology of excitatory neurotoxins: the domoic acid model. Toxicol Pathol 1990;18:165-9.
18. Tryphonas L, Truelove J, Iverson F. Acute parenteral neurotoxicity of domoic acid in cynomolgus monkeys (M. fascicularis). Toxicol Pathol 1990;18:297-303.
19. Perl TM, Bedard L, Kosatsky T, et al. An outbreak of toxic encephalopathy caused by eating mussels contaminated with domoic acid. N Engl J Med 1990;322:1775-80.
20. Teitelbaum JS, Zatorre RJ, Carpenter S, et al. Neurologic sequelae of domoic acid intoxication due to the ingestion of contaminated mussels. N Engl J Med 1990;322:1781-7.
21. Kofke WA, Mintun M, Nemoto E, et al. Opioid neurotoxicity: preliminary studies in monkeys undergoing positron emission tomographic (PET) assessment of regional glucose utilization [abstract]. J Neurosurg Anesthesiol 1994;6:323.
22. Tempelhoff R, Modica PA, Bernardo KL, Edwards I. Fentanyl-induced electrocorticographic seizures in patients with complex partial epilepsy. J Neurosurg 1992;77:201-8.
23. Cascino GD, So EL, Sharbrough FW, et al. Alfentanil-induced epileptiform activity in patients with partial epilepsy. J Clin Neurophysiol 1993;10:520-5.
24. Kearse LA Jr, Koski G, Husain MV, et al. Epileptiform activity during opioid anesthesia. Electroencephalogr Clin Neurophysiol 1993;87:374-9.
25. O'Connell BK, Towfighi J, Kofke WA, Hawkins RA. Neuronal lesions in mercaptopropionic acid-induced status epilepticus. Acta Neuropathol (Berl) 1988;77:47-54.
26. Nevander G, Ingvar M, Auer R, Siesjo BK. Status epilepticus in well-oxygenated rats causes neuronal necrosis. Ann Neurol 1985;18:281-90.
27. Corsellis JA, Bruton CJ. Neuropathology of status epilepticus in humans. Adv Neurol 1983;34:129-39.