Injection of local anesthetic is routinely given for pain control and is generally considered safe. Neurological side effects do occasionally develop. Reported adverse drug reactions include prolonged anesthesia and temporary or permanent neurosensory disturbances—such as anesthesia, hypoesthesia, paresthesia, dysesthesia, allodynia, ageusia, hypogeusia, and dysgeusia—indicating mechanical or toxic nerve injury.1–3 Postmarketing reports indicate a higher risk of permanent neurosensory disturbance after injection of 4% formulations in comparison with alternative weaker formulations in current use. 1,3–7 Both type and concentration of anesthetic solution seem to influence the occurrence of sensory disturbances in humans3,7 and the magnitude of tissue reactions in animals and in vitro experiments.8–11
Kalichman et al.8 compared the neurotoxic potentials of varying concentrations of local anesthetics of both amide type and ester type after perineural injection in the rat sciatic nerve. Nerve damage was assessed at the light microscopic level by estimating the degree of nerve damage (i.e., axonal or myelin degeneration) and edema. Degenerative features increased with increased concentration of local anesthetic. Kroin et al.9 studied the neurotoxicity of lidocaine in concentrations 1%, 2%, and 4%. Infusion of 4% lidocaine led to loss of all conductive function and considerable nerve fiber degeneration, whereas infusion of 1% lidocaine did not elicit any neurotoxic reaction. In agreement with these studies, Cornelius10 and Cornelius et al.12 demonstrated that the neurotoxic injury in the rat sciatic nerve was more severe after intraneural injection of articaine 4% than after articaine 2% and lidocaine 2%. The effects of injection after 3 weeks were evaluated by means of electrophysiology and histology. Preservatives, vasoconstrictors, and mechanical needle injury did not produce any discernible changes in nerve function or nerve structure. Unfortunately, this study,10 an approved German PhD thesis, was never published in full in a journal.
The present experimental study was conducted in the same manner as the study performed by Cornelius10 to substantiate the relative roles of needle penetration with intraneural injection of saline versus articaine in 2% and 4% concentrations.
Animals and Test Substances
Thirty female Wistar rats (median weight 253 g, range 232 to 279 g; Taconic, Lille Skensved, Denmark) were used in the study. Each rat received an intraneural injection of 50 μL of the test substance (saline, articaine 2%, or articaine 4%) in the right sciatic nerve. The left sciatic nerve served as control. Each test substance was aspirated into 10 syringes, and the resulting 30 syringes were randomly numbered 1 through 30 in a sequence unknown to the investigators. The test substances used were Ultracaine D-S 1:200.000 (Aventis Pharma, Frankfurt am Main, Germany), containing articaine hydrochloride 40 mg/mL and epinephrine 0.006 mg/mL (articaine 4%); Ultracaine Suprarenin (Aventis Pharma, Frankfurt am Main, Germany), containing articaine hydrochloride 20 mg/mL and epinephrine 0.006 mg/mL (articaine 2%); and physiological saline (Nycomed, Roskilde, Denmark), pH 4.5 to 7.5. The study was approved by the Animal Experiments Inspectorate, Denmark, No. 2006/561 to 1148, and veterinary supervision was provided by the Department of Experimental Medicine, University of Copenhagen. All experimental procedures in this study were practiced and tested on 14 pilot animals before the 30 test animals were included.
Anesthesia, Surgical Intervention and Injection
The animals were anesthetized with a subcutaneous injection of fentanyl/fluanisone-midazolam (0.3 mL per 100 g body weight of a mixture of Hypnorm [Vetapharma, Leeds, UK], and Dormicum [Roche, Basel, Switzerland]; Hayton et al.13). The right thigh was shaved and disinfected with iodine, and the sciatic nerve was exposed through a 2-cm skin incision after separation of the biceps femoris muscle. The nerve was exposed by blunt dissection as far proximal as feasible. The injection site was at the medial intersection of the nerve with the ileo-femoral ligament (Fig. 1). A 1-mL syringe fitted with a simple, fine-threaded screw arrangement and a 30-gauge needle delivered a precise dose of the test substance. A pilot test of the precision of the injected volume of 10 consecutive trial boluses weighed on a precision scale showed a mean weight of 49.8 mg ∼ μL (SD = 0.58 and coefficient of variation (SD/mean) = 0.012). After injection, the wound was closed in layers with Vicryl 6 to 0 (Ethicon, Johnson & Johnson, St.-Stevens-Woluwe, Belgium). Postsurgical pain control was obtained with Rimadyl Vet (50 mg/mL, Pfizer, NY) with a dose of 0.1 mL/kg.
Lumbar-evoked electrospinograms, i.e., spinal cord field potentials, were elicited from stimulation of the experimental leg and the control leg. In the first session the immediate pharmacological effect of the injection on the evoked electrospinograms was examined after 1, 2, and 5 minutes. The rats were reexamined after 3 weeks to assess a possible effect of the injections. The anesthetized rats were placed in prone position and strapped with moderately stretched hindlegs (Fig. 1). A stimulus (0.1-ms square wave) to the distal segments of the sciatic nerve was delivered at 3-Hz repetition rate (120 to 140 stimulations) with a median intensity of 1.0 mA (range 0.5 to 5.8 mA; Keypoint EMG/NCS/EP workstation version 3.26; Dantec Medical, Skovlunde, Denmark) with 2 electrodes (disposable subdermal needle electrodes, model 017K025, length 12 mm and diameter 0.3 mm; Viasys Healthcare, Madison, WI) placed between the ankle joint and the Achilles' tendon with a 5-mm interelectrode distance. Potentials were acquired with needle electrodes (disposable subdermal needle electrodes, model 017K025, length 12 mm and diameter 0.3 mm; Viasys Healthcare, Madison, WI) with the recording electrode placed subcutaneously above the spinous process of the L5 and the reference electrode contralateral to the stimulation side in the gluteus maximus muscle close to the hip bone. The ground electrode (model 9013S0735, 19.5 cm; Alpine Biomed, Skovlunde, Denmark) was, for practical reasons, placed under the head and forelegs, not to interfere with the surgical, injection, and recording procedures. The stimulation and recording set-up was according to Cornelius.10 When shifting recording from one side to the other, the recording electrode and the ground electrode stayed in place and the placement of the reference electrode and the stimulation electrodes were changed. The electrospinogram was averaged (120 to 130 individual responses) and displayed (10-ms window, 1 to 5 μV/Division), and the intensity of the stimuli was increased until maximal amplitude of the evoked response was obtained with only weak movements of the hindpaw of the stimulated side. The amplitude was calculated using the integrated algorithms of the Keypoint workstation. To avoid disturbance from the stimulus artifact, we measured the amplitude (in microvolts) from the most negative component of the response after the stimulus to the following most positive component or, if this was not well defined, to the baseline in the last 5 ms of the window (Fig. 2). Both in the first session and at the 3-week follow-up, the recordings were obtained from both the right, experimental, leg and the left, control, leg. In the first session the spinal cord field potentials were obtained from the experimental side in the rats before and 1, 2, and 5 minutes after the injection.
Fixation and Preparation of Tissues, Stereological Methods, and Statistical Analysis
The animals were killed by perfusion fixation immediately after the recordings of lumbar-evoked electrospinograms at the 3-week follow-up. After a 15-second initial perfusion with 0.2 M phosphate buffer through the ascending aorta, the animals were perfused for 10 minutes with phosphate-buffered 4% glutaraldehyde while their legs were stretched. Perfusion was made by a Masterflex easy-load pump (Cole-Parmer Instrument Company, London, UK) with a flow of 56 mL/min. The torso with hindlegs was stored in the same fixative at 4°C until a 2- to 4-mm long segment from the injection site (medial intersection with the ileo-femoral ligament; Fig. 2) was dissected from both sciatic nerves. After 3 rinses in 0.15 M phosphate buffer (pH 7.4), the specimens were postfixed in 2% OsO4 in 0.12 M sodium cacodylate buffer (pH 7.2) for 2 hours. The specimens were subsequently dehydrated in ethanol, transferred to propylene oxide, and embedded in EPON resin according to standard procedures. Semithin (1 μm) sections were cut perpendicular to the long axis of the nerve segment with a Reichert Ultracut S microtome (Leica, Herlev, Denmark), stained with toluidine blue and coverslipped with Pertex mounting medium. One nerve segment was excluded from the study for technical reasons.
The stereological methods used in the present study have been described in detail by Larsen.14 Briefly, to obtain unbiased number and size estimates from a sample of axons, all axons regardless of size, shape, orientation, and location must have an equal probability of being sampled. The unbiased counting frame ensures that all axons have an equal probability of being sampled, and systematic, uniform random sampling of frames ensures that all locations within the nerve cross-section are equally represented. Estimates of the total number of myelinated axons were obtained by the fractionator technique. With this technique the myelinated axons in a known fraction, F, of the nerve cross-section are counted with systematic, randomly sampled unbiased counting frames. Estimates of the total number of myelinated axons, N, are obtained as the number of axons counted, ΣQ, divided by the sampling fraction, F:
Size estimates were performed on the nerve fibers uniformly sampled by the 2-dimensional (2D)-fractionator. The areas of myelinated axons, a, were estimated with the 2D-nucleator, where the distances from an approximately central point in the axon to the profile boundary, ℓ, are measured in a predetermined number of isotropic directions:
A square root transformation of myelinated axon areas was performed to correct for right-skewed size distributions. The transformation was chosen as
because it may be regarded as the equivalent circle diameter, i.e., the diameter a circle of equal area would have, and is thus comparable to studies that report axon diameter.
Myelin thickness was measured at a boundary-weighted random point. The random point was found by placing a grid of parallel 2D IUR (isotropic uniform random) lines randomly on the axon profile. The myelin thickness was measured at the lower-most right intersection point between the line grid and the axon profile. The g rate (i.e., the axon diameter divided by the fiber diameter) was calculated from the equivalent circle diameter and the measured myelin thickness. The cross-sectional area of the nerve trunk was estimated by averaging 5 estimates obtained by the 4-way 2D nucleator. The practical analysis of myelinated axons was performed on video images of microscopic fields merged with a graphic representation of an unbiased counting frame and a 2D IUR line grid using the C.A.S.T.-GRID software (Olympus Denmark, Ballerup, Denmark). The stage of the microscope was controlled by stepping motors that moved the section in precise steps of length dx and dy along the X-Y coordinates. Myelinated axons were counted and measured at a final magnification of ×5125 using a ×100 UPLAN oil immersion objective (NA = 1.35). The a(frame) was 88.7 μm2, dx and dy were 65 μm, i.e., the sampling fraction was
Four isotropic directions were chosen for the nucleator measurements. On average, 187 (range 161 to 233) myelinated axons were counted and measured per nerve cross-section. Assuming a Poisson distribution of the axons, the coefficient of error on the number estimates is the inverse of the square root of the number of axons counted:
Differences in amplitudes of the electrospinogram at the 3-week follow-up after the injections of saline and 2% and 4% articaine were investigated using ordinary least square regression and analysis of variance (ANOVA). The analysis included the body weight in the first session before intervention, and the amplitudes of the lumbar electrospinograms in the first session before intervention as explanatory factors.
ANOVA P values were reported, and if the overall effect of the treatment group was significant, post hoc comparisons were reported between articaine 2% and saline and between articaine 4% and saline.
One-way ANOVA was used to compare the 3 groups, and in case of significance, post hoc unpaired t tests were also used for saline-treated versus 2% articaine–treated groups, saline-treated versus 4% articaine-treated groups, and 2% articaine–treated versus 4% articaine–treated groups. The analyses were performed with Microsoft Office Excel 2003 and R (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was considered at P ≤ 0.05.
In general the clinical course was uneventful, and most animals gained weight during the observation period. One animal from the 2% articaine group had a superficial wound infection that responded favorably to antibiotic treatment. As a clinical effect, the affected hindleg in animals with a severe depression of electrospinogram amplitude appeared atrophic and less forceful. This feature was not foreseen, and therefore not registered systematically.
The immediate pharmacological effect in each animal is illustrated in Fig. 3 in which the evoked lumbar responses 1 to 5 minutes after the injection clearly differed between articaine and saline groups. Marked effects of the injections after 3 weeks were also present. At the follow-up the least square regression analysis of the evoked-lumbar electrospinogram in the experimental side showed a significant treatment-group effect on the amplitudes of the lumbar responses (ANOVA P value = 0.0022). The effects of baseline body weight and baseline amplitude (recorded in the first session) were not statistically significant. Between-groups comparison showed that injection of articaine 2% (P value = 0.0295) and injection of articaine 4% (P value = 0.0006) both significantly decreased the amplitude in comparison with injection of saline (Table 1and Fig. 4). A similar analysis of the control side amplitudes (left legs) showed no significant treatment-group effect (ANOVA P value = 0.6924).
Qualitative and Quantitative Structural Changes
Micrographs of nerve cross-sections are shown in Fig. 5, and the quantitative histological data on the myelinated axons and the cross-sectional areas of the nerve trunks are presented in Fig. 6, and corresponding P values are given in Table 2. Both the cross-sectional area of the nerve trunk and the total number of myelinated axons were comparable in all treatment groups, though the ANOVA P value was in trend (P value = 0.065) for an increased number of myelinated axons in the groups treated with articaine. The mean axon area, the mean equivalent circle diameter, and the mean myelin sheath thickness were significantly reduced in the animals that received articaine 4% when compared with the saline-treated animals, and reduced, but not significantly, in the animals that received articaine 2% when compared with the saline-treated animals. The g ratio was unaffected by either of the treatments. The mean axon area and the mean equivalent circle diameter were both significantly smaller in the 4% articaine–treated animals when compared with the 2% articaine–treated animals. The individual size distributions (for which size is chosen as the equivalent circle diameter) of the myelinated axons are shown in Fig. 7. It is clear that the large myelinated axons disappear in almost all animals that were treated with articaine 4% and to a lesser extent in the animals that were treated with articaine 2%. The 2 articaine-treated groups instead had supernumerary small myelinated axons.
Opinions on the causative mechanism leading to neurosensory disturbance in humans have been divided, needle penetration trauma and neurotoxic reaction to the injected substance being potential options. The neurotoxicity of articaine 4% has been denied as a potential mechanism behind an apparent overrepresentation of 4% formulations of local anesthetic-associated neurosensory disturbance in humans.15 A recent combined clinical and registry study on data from the Danish Medicines Agency's national database on adverse drug reactions indicated that neurotoxicity of the injected substance is the causative factor rather than the needle penetration.16 Thus, an experimental study to examine the effects of needle penetration with injection of saline solution versus needle penetration with injection of a local anesthetic in different concentrations was deemed appropriate. A group with needle penetration alone (without injection) was not included because it would be incompatible with the blinding. Thus, saline without preservatives and without vasoconstrictors was chosen as the most probable inert substance for the control group. Therefore, the mechanical trauma of needle penetration and the influence of injection of an equal volume of a commercially available formulation of 2% and 4% articaine (including preservatives) were tested against saline.
The rat sciatic nerve model has proved valid in a number of previous studies similar to ours,9,12,17,18 some of which with slightly different aims. The sensory axons constitute about 70% of the sciatic nerve fibers, and the sensory neurons are located in lumbar ganglia L3 to 6.18
In the present study we found a significant concentration- related suppression of the amplitude in the lumbar-evoked electrospinogram in the articaine groups 3 weeks after injections. Because the amplitudes of the electrospinograms in the articaine groups were significantly lower than those in the saline group, the suppression could not have been caused by damage from the needle penetration alone but rather by the injected substances. Electrophysiological findings with such reduction in the amplitude of the evoked responses as seen in the articaine groups might be associated with axonal loss in humans.19 In addition, it was not possible to elicit a lumbar-evoked response in one third of the animals in the group treated with articaine 4%, whereas responses were elicited in every animal in the articaine 2% and the saline group. These results compare favorably with the in vitro study by Werdehausen et al.11 and the experimental studies of Cornelius et al.12 Werdehausen et al.11 found that all local anesthetics were neurotoxic in a concentration-dependent manner and even in clinically used concentrations. In a study similar to ours, Cornelius et al.12 demonstrated extinction of somatosensory evoked-lumbar responses in 9 of 10 animals with 4% articaine and 1 of 10 with 2% articaine. Likewise, Kroin et al.9 demonstrated severe neurotoxic reaction to lidocaine 4% in the rat tibial nerve.
Lumbar-evoked responses have proven suitable as a method to express variations of the chemical action of different concentrations and formulations of local anesthetics.8,12 Considering the limited number of animals in our study, a dependable baseline response level was essential. This was achieved through averaging of a large number of recordings of evoked lumbar responses from the sciatic nerve, as well as the randomization of the test substances and blinding of the investigators. An anesthetic effect was quickly established after injections of the test substances in the first session, shown as reduced amplitude in the lumbar-evoked electrospinogram in the articaine groups. The quick onset after 1 minute was precipitated by the intraneural injection of the drug. Delivery of a precise volume of test substances was attempted. However, in a number of cases some spillage was inevitable, which may account for some variability. The technique may also have caused minor trauma, resulting in a missing response at 1 minute in 1 saline rat and in 1 articaine 2% rat20,21 (Fig. 3). Fried et al.17 showed limited histological damage in the close vicinity of the needle lesion when using a microneurography electrode in the rat sciatic nerve. These lesions underwent spontaneous healing during the subsequent weeks.
We also found that the total number of axons were slightly higher (9% and 7%, respectively) in the groups treated with articaine when than those in the saline-treated group, but the difference did not reach significance. After the code was broken the sections were inspected (but not blinded), and remnants of degenerating axons were present in the rats treated with articaine 4%. It was also our impression that more small-diameter myelinated axons than usual were surrounded by a nucleated Schwann cell in the 2 groups treated with articaine. These small-diameter axons associated with the Schwann cell nucleus may represent regenerated axons,10 which can explain the apparent paradox that the groups treated with articaine had (insignificantly) more axons than did the saline-treated animals. The mean axon area, the equivalent circle diameter, and the myelin sheath thickness were significantly smaller in the animals treated with articaine 4% when compared with both the saline-treated animals and the articaine 2%–treated animals, but the g ratio was the same in all groups. Size distributions showed that the animals treated with articaine had virtually no large myelinated axons but instead had many small-diameter axons. This shift towards smaller axon diameters may be reflective either of shrinkage of existing axons or of the combined effect of lost large-diameter axons and the appearance of new regenerating axons with small diameter. A short-term study is needed to establish whether articaine 4% injection can elicit significant degeneration of large myelinated axons and a subsequent regeneration with supernumerary regenerating axons.
Despite species-related differences, our results are in accord with epidemiologic studies on humans in demonstrating a significant difference between 2% and 4% formulations of local anesthetics.7,16 Considering the relative robustness of the rat sciatic nerve, more dramatic changes might affect humans and other higher species. A recent study on reports of paresthesia involving dental local anesthetics during the period from November 1997 through August 2008 from the United States Food and Drug Administration Adverse Event Reporting System7 demonstrated significant overrepresentation of neurosensory disturbance associated with 4% solutions of prilocaine and articaine used in dentistry. Also in Europe, clinical and registry studies on data from the Danish Medicines Agency's national registry on adverse drug reactions associated with local anesthetics show overrepresentation of adverse drug reactions with articaine 4%.3,16,22
The observed degenerative changes (stereology) and the reduced amplitude of the lumbar-evoked electrospinograms (electrophysiology) after 3 weeks indicate a concentration-dependent neurotoxic effect of intraneural injection of 50 μL of commercially available 2% and 4% formulations of an articaine-based local anesthetic. Needle penetration with injection of 50 μL of saline had no significant effect on nerve conduction and histomorphology.
Name: Søren Hillerup, DDS, PhD, Dr. Odont.
Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.
Attestation: Søren Hillerup has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.
Name: Merete Bakke, DDS, PhD, Dr.Odont.
Contribution: This author helped conduct the study, analyze the data, and write the manuscript.
Attestation: Merete Bakke has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Jytte Overgaard Larsen, MD, PhD.
Contribution: This author helped conduct the study, analyze the data, and write the manuscript.
Attestation: Jytte Overgaard Larsen has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Carsten Eckhardt Thomsen, MScEE, PhD.
Contribution: This author helped conduct the study, analyze the data, and write the manuscript.
Attestation: Carsten Eckhardt Thomsen has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Thomas Alexander Gerds, Dr.Rer.Nat.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Thomas Alexander Gerds has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Prof. C. P. Cornelius, MD, DDS, Lab technician J. S. Lykkeaa, Consultant T. Dalager, MD, Prof. R. H. Jensen, MD, Dr. C. J. Bundgaard, DVM, and Prof. M. Lauritzen, MD, are cordially acknowledged for their kind assistance.
1. Haas DA, Lennon D. A 21 year retrospective study of reports of paresthesia following local anaesthetic administration. J Can Dent Assoc 1995;61:319–30
2. Haas DA. Articaine and paraesthesia: epidemiological studies. J Am College of Dentists 2006;73:5–10
3. Hillerup S, Jensen R. Nerve injury caused by mandibular block analgesia. Int J Oral Maxillofac Surg 2006;35:437–43
4. Haas DA. An update on local anesthetics in dentistry. J Can Dent Assoc 68: 546–51
5. McFarlane D. Paraesthesia following local anaesthetic solution. Dispatch 2005;19:26
6. Gaffen AS, Haas DA. Retrospective review of voluntary reports of nonsurgical paresthesia in dentistry. J Can Dent Assoc 2009;75:579
7. Garisto GA, Gaffen AS, Lawrence HP, Tenenbaum HC, Haas DA. Occurrence of paresthesia after dental local anesthetic administration in the United States. J Am Dent Assoc 2010;141:836–44
8. Kalichman MW, Moorhouse DF, Powell HC, Myers RR. Relative neural toxicity of local anesthetics. J Neuropathol Exp Neurol 1993;52:234–40
9. Kroin J, Penn R, Levy F, Kerns J. Effect of repetitive lidocaine infusion on peripheral nerve. Exp Neurol 1986;94:166–73
10. Cornelius CP. Nerveninjektionsschäden durch Lokalanaesthetika. Experimentelle Untersuchungen zur Neurotoxizität und Longitudinalausbreitung [thesis]. Tübingen, Germany: University of Tübingen, 1997
11. Werdehausen R, Fazeli S, Braun S, Hermanns H, Essmann F, Hollmann MW, Bauer I, Stevens MF. Apoptosis induction by different local anaesthetics in a neuroblastoma cell line. Br J Anaesth 2009;103:711–8
12. Cornelius CP, Roser M, Wiethölter H, Wolburg H. Nerve injection injuries due to local anaesthetics. Experimental work. J Cranio Max Fac Surg 2000;28(Suppl. 3):134–5
13. Hayton SM, Kriss A, Muller DP. Comparison of the effects of four anaesthetic agents on somatosensory evoked potentials in the rat. Lab Anim 1999;33:243–51
14. Larsen JO. Stereology of nerve cross sections. J Neurosci Methods 1998;85:107–18
15. Malamed SF. Letter to the editor. Int J Oral Maxillofac Surg 2006;35:876–7
16. Hillerup S, Jensen RH, Ersbøll BK. Trigeminal nerve injury associated with injection of local anesthetics. Needle lesion or neurotoxicity? J Am Dent Assoc (in press)
17. Fried K, Frisen J, Mozart M. De- and regeneration of axons after minor lesions in the rat sciatic nerve. Effects of microneurography electrode penetrations. Pain 1989;36:93–102
18. Schmalbruch H. Fiber composition of the rat sciatic nerve. Anat Rec 1986;215:71–81
19. Tankisi H, Pugdahl K, Fuglsang-Frederiksen A, Johnsen B, de Carvalho M, Fawcett PR, Labarre-Vila A, Liguori R, Nix WA, Schofield IS. Pathophysiology inferred from electrodiagnostic nerve tests and classification of polyneuropathies. Suggested guidelines. Clin Neurophysiol 2005;116:1571–80
20. Rice ASC, McMahon SB. Peripheral nerve injury caused by injection needles: influence of bevel configuration, studied in a rat model. Brit J Anaesthesia 1992;69:433–8
21. Rice AS, McMahon SB, Wall PD. The electrophysiological consequences of electrode impalement of peripheral nerves in the rat. Brain Res 1993;631:221–6
© 2011 International Anesthesia Research Society
22. Hillerup S, Jensen R. Nerveskader opstået ved lokalanalgesi i tandlægepraksis. Tandlaegebladet 2006;110:556–67