Current issues in spinal anesthesia.

Abstract

Received from the Departments of Anesthesiology, Virginia Mason Medical Center and the University of Washington, Seattle, Washington.SPINAL anesthesia has enjoyed a long history of success and recently celebrated a centennial anniversary. 1Anesthesiologists master spinal anesthesia early during training with achievement of competence (> 90% technical success rate) after only 40–70 supervised attempts. 2,3The ease and long history of spinal anesthesia may give the impression that it is a simple technique with little sophistication. However, much has been learned recently regarding the anatomy, physiology, pharmacology, and applications of spinal anesthesia. This review article focuses on what is new, interesting, and clinically relevant for this simple and popular technique.Many anatomic structures important for spinal anesthesia have only recently been investigated. The arachnoid membrane is a structure of obvious interest, as spinal agents must be delivered within its confines. The arachnoid membrane is composed of overlapping layers of epithelial cells connected by tight junctions. 4This anatomic arrangement allows the arachnoid membrane, not the dura, to function as the principal meningeal barrier (90% of resistance) to materials crossing in and out of the cerebrospinal fluid (CSF). 4A functional proof of the arachnoid’s importance as gatekeeper to the CSF is that spinal CSF resides in the subarachnoid and not subdural space. The arachnoid membrane serves not only as a passive container of CSF but also actively processes and transports agents attempting to cross J1te meninges. Recent studies demonstrated that metabolic enzymes are expressed in the arachnoid that can affect agents (e.g. , epinephrine) 5and neurotransmitters important for spinal anesthesia (e.g. , acetylcholine). 6–8Active transport of compounds across the arachnoid membrane occurs in the area of the neural root cuffs. 8Here, unidirectional transport of materials from the CSF into the epidural space occurs and may contribute to clearance of spinal anesthesia agents. Another potential clinical consideration of the lamellar structure of the arachnoid is easy separation of the arachnoid membrane from the dura during spinal puncture. This mechanical arrangement allows easy subdural deposition of spinal agents despite the free return of CSF during spinal injection, which may help to explain individual effects of spinal anesthesia. 9After injection of spinal anesthetics, dilution with the CSF occurs before arrival at effector sites in the central nervous system. Thus, individual variation in lumbosacral volumes of CSF and distribution within this volume will affect spinal anesthesia. Recent use of magnetic resonance imaging (MRI) demonstrates great variability between individuals in volume of lumbosacral CSF, with a range of 28–81 ml. 10Interestingly, obese individuals have substantially less CSF (∼10 ml less), which is partly caused by compression of the neural foramina. Clinical correlation between volume of lumbosacral CSF and spinal anesthesia with hyperbaric lidocaine and isobaric bupivacaine is excellent, with CSF accounting for 80% of the variability for peak block height and regression of sensory and motor block (fig. 1). 11Unfortunately, volume of lumbosacral CSF does not correlate with external physical measurements other than weight (r = 0.4, P < 0.05); therefore, volume cannot be easily estimated from physical examination. 11Other important considerations include the observation on MRI that the CSF is not a “still lake” of fluid but vigorously oscillates with arterial pulsations. 9These wavelike movements may be another important factor in distribution and clearance of spinal agents and may influence neurotoxicity from exposure to concentrated agents (see Transient Neurologic Symptoms–Neurotoxicity). The target sites of spinal anesthetics are the spinal nerve roots and spinal cord. In a similar fashion to volume of CSF, individual variability in anatomy of spinal nerve roots may also explain variability in spinal anesthesia. 12,13Recent autopsy and microscopic studies have observed great interindividual variability in size of human nerve roots. For example, the range of the posterior nerve root of L5 is 2.3–7.7 mm3. Other interesting anatomic findings are the relatively larger size of dorsal nerve roots, compared with ventral, with packaging into easily separable strands. 13Although a larger dorsal nerve root would seem more impenetrable to local anesthetics, the separation of the dorsal root into component bundles creates a much larger surface area for local anesthetic penetration than the single smaller ventral nerve root. This anatomic finding may help explain the relative ease of sensory versus motor block.Finally, recent microscopic and endoscopic examination of the subarachnoid space reveals the presence of numerous membranes surrounding nerve roots and ligaments within the arachnoid that potentially compartmentalize spinal CSF. 9These partitions may help to concentrate local anesthetics near nerve roots and augment spinal anesthesia but could also impede communication of CSF between dorsal and ventral nerve roots, thus again explaining relative difficulty of achieving motor block.Mild perioperative hypothermia is associated with an increased incidence of myocardial ischemia, cardiac morbidity, wound infection, blood loss, and transfusion requirements. 14Both general and regional anesthesia impair temperature homeostasis to a similar degree, 15,16and careful monitoring and active maintenance of temperature is a simple means to prevent morbidity.The effects of spinal anesthesia on temperature homeostasis have been well studied, and there are three main mechanisms causing core hypothermia. 14,15,17,18The first is redistribution of central heat to the periphery caused by vasodilation from sympathetic block. This effect is maximal during the first 30–60 min, causes a decrease in core temperature of approximately 1–2°C, and depends on extent of sensory block and patient age (fig. 2). 19The second mechanism is loss of thermoregulation characterized by reduced shivering and vasoconstriction thresholds during spinal anesthesia. This abnormal tolerance for hypothermia occurs because of subjective warmth exceeding the actual surface temperature increase from sympathectomy. This exaggerated sense of warmth is proportional to extent of sensory and sympathetic block 15and decreases thresholds for shivering and vasoconstriction. Thus, hypothermia may occur during spinal anesthesia without a conscious perception of cold. 20Finally, with loss of thermoregulatory vasoconstriction below the level of the sympathetic block, there is increased heat loss from vasodilation. Spinal anesthesia will predictably cause core hypothermia within 30–60 min, and patients should be monitored and actively warmed if needed. 14Unfortunately, a recent survey of practicing members of the American Society of Anesthesiologists revealed that only 33% of practitioners use temperature monitoring during regional anesthesia. 18Furthermore, temperature was most commonly monitored on a surface site such as the forehead and not at accessible core temperature sites during regional anesthesia (e.g. , tympanic membrane). These surface sites provide inherently inaccurate estimates of core temperature during regional anesthesia because of the aforementioned redistribution of core heat, compensatory vasoconstriction above the level of spinal anesthesia, and influence of ambient temperature. 16If hypothermia develops, patients should be rewarmed with forced air heating. Spinal anesthesia accelerates rewarming compared with general anesthesia because of the residual sympathetic block and vasodilation. 15The most common serious side effects from spinal anesthesia are hypotension and bradycardia, 21,22and closed claims surveys of 40,000–550,000 spinal anesthetics indicate an incidence of cardiac arrest from 0.04–1/10,000. 23,24Large surveillance studies typically observed incidences of hypotension around 33% and bradycardia around 13% in nonobstetric populations. 21,22Risk factors for hypotension in nonobstetric populations include block height T5 or greater (odds ratio [OR], 3.8), age 40 yr or greater (OR, 2.5), baseline systolic blood pressure less than 120 mmHg (OR, 2.4), and spinal puncture above L3–L4 (OR, 1.8). Risk factors for development of bradycardia in nonobstetric populations include baseline heart rate less than 60 beats/min (OR, 4.9), American Society of Anesthesiologists physical status I (OR, 3.5), use of β blockers (OR, 2.9), prolonged PR interval on electrocardiogram (OR, 3.2), and block height T5 or greater (OR, 1.7). 21,25Analysis of closed claims for cardiac arrest during spinal anesthesia indicated that administration of sedation to produce a sleep-like state without spontaneous verbalization and lack of early administration of epinephrine were common patterns of management in cases of cardiac arrest. 26Cardiovascular effects of spinal anesthesia typically include a decrease in arterial blood pressure and central venous pressure with only minor decreases in heart rate, stroke volume, or cardiac output even in patients with poor left ventricular function (ejection fraction < 50%;fig. 3). 27,28Typical preservation of cardiac output during spinal anesthesia allows maintenance of oxygen delivery to vital organs such as the brain, as demonstrated by lack of change in jugular bulb oxygen saturation. 29The decrease in sympathetic activity and motor block also leads to a decrease in total body oxygen consumption that correlates with extent of spinal anesthesia. 30Hypotension occurs from decreases in systemic vascular resistance and central venous pressure from sympathetic block with vasodilation and redistribution of central blood volume to lower extremities and splanchnic beds. 27,28,31,32This sympathetic block is rarely complete, and some preservation of sympathetic reflexes to stressful challenge typically occurs. 33Sudden bradycardia can occur from shift in cardiac autonomic balance toward the parasympathetic system, as evidenced in spectral analysis of heart rate variability, 34from activation of left ventricular mechanoreceptors from a sudden decrease in left ventricular volume (Bezold Jarisch reflex), 35or from increases in baroreflex activity. 36Various prophylactic and rescue regimens have been advocated for hemodynamic disturbances with emphasis on prevention of hypotension. Studies are difficult to interpret because of different definitions of hypotension and different patient populations (elderly, pregnant, surgical). 28Prophylactic measures include prehydration with crystalloid or colloid or administration of vasoactive agents. On the whole, prehydration of crystalloid (250–2,000 ml) appears to temporarily increase preload and cardiac output without consistently increasing arterial blood pressure or preventing hypotension. 22,32,37–40Pharmacokinetics of crystalloid explain its poor efficacy, as crystalloid is quickly redistributed from the intravascular to the extravascular space. 41Administration of large volumes (> 1 l) of crystalloid does not appear to confer additional benefit over small volumes (250 ml) 38and may be detrimental to patients with limited cardiopulmonary reserve. Prehydration with colloid (≥ 500 ml) appears to be more effective than crystalloid at maintaining arterial blood pressure and perhaps decreasing incidence of hypotension depending on definition and population. 39The greater effectiveness of colloid is a result of greater effect for increasing central venous pressure and cardiac output caused by slower redistribution out of the intravascular space (fig. 4). 41In contrast to prophylaxis, treatment of hypotension during spinal anesthesia will be effective with crystalloid or colloid because of changes in kinetics induced by spinal anesthesia 42and intravascular hypovolemia. 43Both clinical scenarios alter kinetics of cystalloid and colloid to allow retention within the intravascular space. Prophylactic administration of pharmacologic agents may be more effective than prehydration for prevention of hypotension. 44α-Adrenergic agonists (e.g. , metaraminol, phenylphrine) reliably increase arterial blood pressure by increasing systemic vascular resistance; however, heart rate and cardiac output may decrease because of increased afterload. 22,32,45Mixed α- and β-adrenergic agents (e.g. , ephedrine, epinephrine) are also effective for increasing arterial blood pressure and preventing hypotension but act by primarily increasing heart rate and cardiac output with a smaller increase in systemic vascular resistance. 28These different physiologic mechanisms for α-versus mixed α- and β-adrenergic agents also occur in treatment of hypotension during spinal anesthesia (fig. 5). 31Thus, initial treatment can be tailored to α only for patients with hypotension and mixed α and β for patients with both hypotension and bradycardia. A potential means for prophylaxis of hypotension is by manipulation of spinal anesthesia to achieve a predominantly unilateral block. Unilaterality can be maintained if the patient remains in a lateral position for surgery; however, eventual turning of the patient into a supine position results in partial redistribution to bilateral anesthesia. 46Unilaterality can be maximized by using a side port spinal needle (e.g. , Whitacre) 47and a small dose of local anesthetic, 48and by keeping the patient in the lateral position for 6–20 min. 46Concentration of anesthetic solution 49and speed of injection 50are minor factors for unilaterality. 51With such optimization of unilaterality and decreased extent of sympathetic block, hypotension has been reported to decrease from 22–53 to 5–7%. 44,52There has been a recent convergence in mechanisms of general and spinal anesthesia. Minimum alveolar concentration, a traditional measure of inhalational agent potency for depth of anesthesia, appears to have a primary mechanism in the spinal cord. 53In contrast, central neuraxial anesthesia may have direct effects on suppression of consciousness, and multiple studies have observed that patients appear drowsy after spinal anesthesia despite lack of sedative medications. 54,55Correspondingly, both spinal and epidural anesthesia reduce the hypnotic requirements of midazolam, isoflurane, sevoflurane, and thiopental in surgical patients and laboratory studies. 56Possible mechanisms for decreased consciousness during spinal anesthesia include rostral spread of local anesthetics or decrease in reticular activating system activity caused by interruption of afferent input. 54Animal models support the latter, as spinal anesthesia in rats decreases hypnotic requirements of thiopental without detection of local anesthetic in the brain or cervical spinal cord. 57In humans, degree of sedation caused by spinal anesthesia is related to peak block height, with greater sedation observed with greater block heights. 58This finding again indirectly supports the hypothesis that greater loss of afferent input from extension of spinal anesthesia increasingly suppresses consciousness. Time of maximal sedation with spinal anesthesia in volunteers shows a biphasic distribution, with one peak occurring during peak spinal block (∼30 min after injection) and a second peak occurring later, approximately 1 h after injection. 54Mechanisms for the second peak in sedation are unclear and may include late rostral spread of local anesthetic into the brain or psychological relief over regression of spinal anesthesia. Clinical relevance for these observations is the decreased need for pharmacologic sedatives with the use of spinal anesthesia.Injection of local anesthetics into the spinal CSF allows access to sites of action both within the spinal cord and the peripheral nerve roots. 28The traditional concept of spinal anesthesia causing complete conduction block is simplistic, as studies with somatosensory evoked potentials demonstrate little change in amplitudes or latencies after induction of dense spinal or epidural anesthesia. 59There are multiple potential actions of local anesthetics within the spinal cord at different sites. For example, within the dorsal and ventral horns, local anesthetics can exert sodium channel block and inhibit generation and propagation of electrical activity. 60Other spinal cord neuronal ion channels, such as calcium channels, are also important for afferent and efferent neural activity. Spinal administration of N-type calcium channel blockers results in hyperpolarization of cell membranes, resistance to electrical stimulation from nociceptive afferents, and intense analgesia. 61Local anesthetics may have similar actions on neural calcium channels, which may contribute to analgesic actions of central neuraxially administered local anesthetics. 62Multiple neurotransmitters are involved in nociceptive transmission in the dorsal horn of the spinal cord. 63Substance P is an important neurotransmitter that modulates nociception from C fibers and is released from presynaptic terminals of dorsal root ganglion cells. Administration of local anesthetics in concentrations that occur after spinal and epidural anesthesia inhibits release of substance P and inhibits binding of substance P to its receptor in the central neuraxis in a noncompetitive fashion. 64Other inhibitory neurotransmitters that may be important for nociceptive processing in the spinal cord, such as γ-aminobutyric acid, are also affected by local anesthetics. Local anesthetics can potentiate the effects of γ-aminobutyric acid by preventing uptake and clearance. 65These studies suggest spinal anesthesia may be partially mediated via complex interactions at neural synapses in addition to ion channel blockade and may explain the ability of spinal anesthesia to reduce central temporal summation in humans. 66Although spinal local anesthetics can block sodium channels and electrical conduction in spinal nerve roots, other mechanisms may also come into play. It is theorized that a large part of the sensory information transmitted via peripheral nerves is carried via coding of electrical signals in after-potentials and after-oscillations. 67,68Evidence for this theory is found in studies demonstrating loss of sensory nerve function after incomplete local anesthetic blockade. For example, sensation of temperature of the skin can be lost despite unimpeded conduction of small fibers. 69Furthermore, a surgical depth of epidural and spinal anesthesia can be obtained with only minor changes in somatosensory evoked potentials from the anesthetized area. 59,70Previous studies have demonstrated that application of subblocking concentrations of local anesthetic will suppress normally occurring after-potentials and after-oscillations without significantly affecting action potential conduction. 69Thus, disruption of coding of electrical information by local anesthetics may be a primary mechanism for block of spinal nerve roots during spinal anesthesia.Introduction of small-gauge pencil-point spinal needles has reduced the risk of postdural puncture headache (PDPH) to approximately 1%, 71and use of spinal anesthesia for ambulatory surgery has become more popular. The ideal spinal anesthetic would combine rapid and adequate surgical anesthesia with rapid achievement of discharge criteria such as ambulation and urination. The most important determinant of both successful surgical anesthesia and time until recovery is dose of local anesthetic. 72,73Neither volume of injectate nor concentration of solution within a 10-fold range (0.5–5% lidocaine) have significant effects. 74,75Unfortunately, selection of dose for ambulatory spinal anesthesia will inherently result in variable individual patient response. Both pharmacokinetics and pharmacodymanics of individual patients are highly variable and are not easily predicted by individual patient demographics (e.g. , age, height). 76However, ambulatory spinal anesthesia can be designed to provide similar discharge times (∼202 min) as general anesthesia (∼185 min). 77Further acceleration of patient discharge may or may not improve efficiency depending on institutional staffing, compensation of staff, and patient volume. 78Determination of appropriate patient discharge criteria for ambulatory anesthesia and surgery are evolving. 79Ability to void is a common discharge criterion that may delay patient discharge after resolution of spinal anesthesia. 79,80Recent evidence suggests that patients at low risk (e.g. , nonpelvic surgery, no history of urinary retention) for urinary retention do not need to void before being discharged home. 79,81,82High-risk patients may be monitored and managed optimally by bladder ultrasound determination of urine volume and need for catheterization. 82,83Spinal lidocaine has been a popular choice for ambulatory spinal anesthesia, and recent studies have examined dose–response effects of lidocaine on anesthesia and recovery (table 1). 84,85Although lidocaine has enjoyed a long history of safety and popularity since its introduction in 1945, it has come under recent scrutiny because of transient neurologic symptoms (TNS). TNS is clearly associated with use of spinal lidocaine, with an approximate incidence of 20% in the ambulatory setting (table 2). 86–88Concern over the potential for neurologic injury and for patient comfort has led to interest in alternative spinal local anesthetics. An ideal replacement for lidocaine should posses clinical characteristics suitable for ambulatory anesthesia (fast, successful anesthesia with rapid recovery) and less risk for TNS. Bupivacaine has been the most studied alternative to lidocaine. TNS is virtually absent in all clinical studies with spinal bupivacaine (0–1%;table 2). 86–88Recent dose–response data on clinical anesthetic characteristics for spinal bupivacaine (table 1) indicate that small doses can be used for ambulatory anesthesia. 72,89It is particularly important to select small doses of bupivacaine (≤ 10 mg) to avoid prolonged detrusor block, inability to void, and excessively prolonged time until discharge as compared with equipotent doses of lidocaine. 80Mepivacaine has been used for spinal anesthesia since the 1960s. Clinical anesthetic characteristics are similar to lidocaine, with an approximate potency of 1.3:1 (table 1). 90,91Reported risk of TNS with mepivacaine is highly variable. Small-scale studies (60–75 patients) report a low incidence of TNS (0–8%), whereas larger studies (200+ patients) report incidences of approximately 30% (table 2). 90,92It seems mepivacaine has similar clinical characteristics as lidocaine for spinal anesthesia but likely shares the same risk of TNS.Ropivacaine is a new local anesthetic released in the United States in 1996. It is a lipid-soluble agent that is approximately 50–60% as potent as spinal bupivacaine. Like bupivacaine, there is little risk of TNS with use of spinal ropivacaine (0–1% incidence;table 2). 93,94The decreased potency of ropivacaine offers the potential for more rapid recovery and better suitability as an outpatient spinal anesthetic. However, dose–response data indicate that equipotent doses of ropivacaine will have similar recovery times as bupivacaine (table 1). 93,94Thus, ropivacaine in equipotent doses (2:1) will be virtually indistinguishable from bupivacaine for clinical anesthesia and risk of TNS without any obvious advantages.Procaine was the first synthesized local anesthetic and has been used for spinal anesthesia since the early 1900s. Procaine has suitable clinical characteristics for brief spinal anesthesia but was supplanted by lidocaine because of more reliable anesthesia and fewer side effects. For unclear reasons, procaine carries a higher risk of nausea than other spinal local anesthetics (OR, 3:1). 21No studies have adequately determined dose–response data 95for spinal procaine, and very few have compared it with lidocaine. A recent prospective, randomized, double-blind study compared 100 mg hyperbaric procaine to 50 mg hyperbaric lidocaine (2:1 ratio) for ambulatory knee arthroscopy. 96This study observed a higher anesthetic failure rate with the procaine (17%vs. 3%), a higher incidence of nausea (17%vs. 3%), and a 30-min longer time until readiness for discharge. Although a larger dose of procaine would probably increase anesthetic success, a larger dose would likely further increase the greater risk of nausea and prolonged recovery. The risk of TNS was much less with procaine versus lidocaine (6%vs. 24%). Thus, recent data on procaine spinal anesthesia are not encouraging, as it appears to be less reliable for surgical anesthesia than lidocaine while having a slower recovery (table 1). Risk of TNS is less than lidocaine but probably greater than bupivacaine (table 2).Prilocaine is unavailable in the United States for central neuraxial use. It is an amide local anesthetic with pharmacologic properties similar to lidocaine. There are no dose–response data with prilocaine to allow determination of optimal doses for ambulatory spinal anesthesia, nor are there formal potencies to allow comparison with lidocaine. Recent studies suggest that prilocaine is approximately equipotent to lidocaine within a dose range of 40–70 mg 87,97and thus may have suitable clinical characteristics for ambulatory spinal anesthesia (table 1). Risk of TNS appears to be minimal with spinal prilocaine (0–1%;table 2). 87,97Prilocaine could be a suitable agent for ambulatory spinal anesthesia with fast recovery properties and low risk of TNS.Both anesthetic success and especially time until readiness for discharge are dependent on dose of local anesthetic. There has been recent interest in using analgesic additives to spinal local anesthetics to decrease the dose of local anesthetic for faster recovery while maintaining or improving anesthetic success. The optimal analgesic additive would increase anesthetic success while sparing local anesthetic and decreasing time until discharge. Multiple analgesics are active in the spinal cord and could potentially be used as spinal anesthesia additives. 63However, analgesic activity (dose response, effects on acute vs. chronic pain) and neurotoxicity have not been fully evaluated for the multitude of known analgesics. Thus, only reasonably well-investigated agents are discussed in the following sections (table 3). Both epinephrine and phenlyephrine have a long history as additives to local anesthetics. Both agents will intensify and prolong sensory and motor anesthesia 98–100and allow use of lower doses of local anesthetic in a dose-dependent fashion (0.1–0.6 mg). 101Vasoconstrictors may act by a combination of decreased clearance of spinal local anesthetic via vasoconstriction and direct analgesic effects on spinal cord α-adrenergic receptors. Unfortunately, their usefulness for ambulatory spinal anesthesia is limited by their propensity to prolong recovery from sensory and motor block and ability to urinate to a disproportionate degree as compared with anesthetic benefit (table 3). 100For example, addition of 0.2 mg epinephrine to 60 mg 2% isobaric lidocaine in patients undergoing outpatient knee arthroscopy prolonged sensory block by 90 min but prolonged recovery milestones and time to discharge by 106 min. 102Similar effects are observed with bupivacaine 98and procaine. 95Use of epinephrine is not associated with increased risk of TNS 88but has been associated with a case report of cauda equina syndrome. 103Use of phenylephrine has been implicated as a risk for TNS (10-fold increase). 99Thus, use of vasoconstrictors are safe and effective for prolonging and intensifying spinal anesthesia but are ill advised for ambulatory surgery because of delay in patient recovery and potential increased risk of TNS.Opioids were the first clinically used selective spinal analgesics after the discovery of opioid receptors in the spinal cord. 104Intrathecal opioids selectively decrease nociceptive afferent input from Aδ and C fibers without affecting dorsal root axons or somatosensory evoked potentials. 104Hydrophilic opioids such as morphine provide excellent selective spinal analgesia because of small volume of distribution and slow clearance from the spinal cord. 8However, slow spinal cord penetration and prolonged duration in CSF caused by hydrophilicity also results in slow onset (> 30 min), prolonged duration of action (6+ h), and risk of delayed respiratory depression from rostral spread in CSF. Lipophilic opioids have a more favorable clinical profile of fast onset (minutes), modest duration (1–4 h), and little risk of delayed respiratory depression. 104Fentanyl and sufentanil are the most commonly used spinal lipophilic opioids. Clinical studies suggest that intrathecal administration of sufentanil may produce selective spinal analgesia; however, laboratory studies suggest that systemic uptake followed by supraspinal analgesia may be the dominant mechanism of action. Because of the extreme lipid solubility of sufentanil, it has a very large volume of distribution in the spinal cord with rapid clearance into the spinal cord vasculature and epidural space in pig models. 8This laboratory finding implies that very little spinal sufentanil is available for interaction with spinal cord opioid receptors because of sequestration in lipid soluble white matter and systemic redistribution. Further studies are needed to determine the dominant mechanism of action of spinal sufentanil and whether spinal administration is rational.Fentanyl is less lipid soluble and will maintain modest spinal selectivity when injected intrathecally. 7,8Dose–response data indicate that spinal fentanyl alone provides dose-dependent analgesia with a minimally effective dose of approximately 10 μg. 105Risk of early respiratory depression is also dose-dependent, wit