International Journal of Pediatric Otorhinolaryngology (2006) 70, 1423—1429

www.elsevier.com/locate/ijporl

Lateral graft type 1 tympanoplasty using AlloDermW

for tympanic membrane reconstructionin children§

Philip Lai, Evan Jon Propst, Blake Croll Papsin *

Department of Otolaryngology — Head and Neck Surgery, The Hospital for Sick Children,Toronto, Ont., Canada

Received 27 January 2006; received in revised form 20 February 2006; accepted 20 February 2006

KEYWORDSPediatric;Tympanoplasty;Tympanic membraneperforation;AlloDerm

Summary

Objectives: To describe the lateral graft type 1 tympanoplasty technique usingAlloDerm1 for tympanic membrane reconstruction in children and to compare itssurgical and audiometric outcomes with the traditional underlay type 1 tympano-plasty.Methods: The records of 34 consecutive children undergoing type 1 tympanoplastybetween 2004 and 2005 were reviewed; 18 received lateral graft tympanoplasty withAlloDerm1 and 16 received underlay tympanoplasty (8 AlloDerm1 and 8 temporalisfascia). Pre- and post-surgical audiograms, speech reception threshold, closure rateand complication rate were evaluated using one-way and repeated measures ANOVAs.Results: Children who underwent lateral graft type 1 tympanoplasty pre-operativelyhad larger tympanic membrane perforations, worse pure tone averages, air bone gapsand speech reception thresholds as compared with children undergoing underlay type1 tympanoplasty (P < 0.001). Pure tone averages and air bone gaps improved sig-nificantly with surgery in both lateral and underlay type 1 tympanoplasty groups(P < 0.05), with both groups achieving comparable postoperative audiometric out-comes (P > 0.01). The lateral graft group demonstrated a higher perforation closurerate (94%) as compared with both underlay groups (88%). Complication rates werevirtually non-existent.Conclusions: Despite larger perforations and worse pre-operative audiometricscores, children who underwent lateral graft type 1 tympanoplasty achieved compar-able postoperative audiometric results and perforation closure rates as compared

§ Presented at the Society for Ear, Nose and Throat Advances in Children (SENTAC) Annual Meeting, December 1—4, 2005, Baltimore,Maryland.* Corresponding author at: Cochlear Implant Program, Department of Otolaryngology, 6th Floor, Elm Wing, The Hospital for Sick

Children, 555 University Avenue, Toronto, Ont., Canada M5G 1X8. Tel.: +1 416 813 7259; fax: +1 416 813 5036.E-mail address: blake.papsin@utoronto.ca (B.C. Papsin).

0165-5876/$ — see front matter # 2006 Elsevier Ireland Ltd. All rights reserved.doi:10.1016/j.ijporl.2006.02.012

1424 P. Lai et al.

with children who underwent underlay type 1 tympanoplasty. Results suggest thatlateral graft type 1 tympanoplasty using AlloDermW is effective for tympanic mem-brane reconstruction in children and should be used when temporalis fascia is notavailable or the extent of the perforation limits its use.# 2006 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Total tympanic membrane reconstruction using alateral graft is a surgical option that can be used inchildren with very large perforations in the pre-sence of granular inflammation or tympanosclerosisor in cases where there has been surgical failureafter traditional underlay grafts. Temporalis fasciahas traditionally been used for lateral graft recon-struction, accepting the donor site morbidity andoften insufficient tissue available in the case ofrevision surgery. The use of AlloDermW (LifeCellCorporation, Branchburg, NJ), an acellular dermalmatrix processed from cadaveric skin, has beendescribed for total tympanic membrane recon-struction primarily in adults [1]. We describe herethe lateral graft type 1 tympanoplasty techniqueusing AlloDermW for tympanic membrane recon-struction in children and compare its surgical andaudiometric outcomes with the traditional under-lay type 1 tympanoplasty.

Tympanic membrane reconstruction in childrenwith perforations that do not heal spontaneously isessential to allow optimal speech and languagedevelopment, decrease the possible risk of choles-teatoma formation, and to allow the child to fullyparticipate in water sports. The traditional method,and first surgical choice of tympanic membranereconstruction, is the underlay technique, whichinvolves placement of a graft medial to any tympa-nic membrane remnant and the malleus. Most smallto medium sized, dry, posterior perforations can betreated using the underlay technique. Conversely,the lateral technique involves placing a graft lateralto the bony annulus and any remaining fibrous mid-dle layer of the tympanic membrane after removalof the squamous layer. Lateral type 1 tympanoplastyfor total tympanic membrane reconstruction is apossible surgical option for large anterior or mar-ginal perforations, in cases of extensive tympano-sclerosis, subtotal perforations with refractorygranular inflammation and finally in children whohave unsuccessfully undergone one or moreattempts to close complicated perforations usingthe underlay graft technique. The traditional lateralgraft technique involves elevation of a free anteriorcanal wall skin graft, removal of any residual annu-lus and a generous canalplasty to recreate theanterior angle and prevent lateralization [1].

Temporalis fascia has traditionally been the graft-ing material of choice because it is autologous and iseasily accessible in the surgical field. Unfortunately,patients who are candidates for a lateral graft forrevision tympanoplasty often no longer have enoughnative donor tissueavailable, necessitating theuseofan alternative grafting material. There is evidencethat AlloDermW (LifeCell Corporation) that acts as ascaffold for fibroblast and endothelial cell in-growth[2], is a suitable substitute for tympanic membranereconstruction. The main advantages of using Allo-DermW are prevention of donor site morbidity, it isimmunologically inert [3,4], it is stronger than fascia[5,6] and there is a reduced operation time because agraft does not need to be harvested [7].

Several studies have demonstrated the efficacy ofAlloDermW in closing small tympanic membrane per-forations using the underlay technique [4,7—10].Fishman et al. [1] evaluated the use of AlloDermW

for total tympanic membrane reconstruction primar-ily in adults and demonstrated a high closure rate anda significantly shortenedhealing time.Thepurposeofthis study was to describe the lateral graft type 1tympanoplasty technique using AlloDermW for tym-panic membrane reconstruction exclusively in chil-dren and to compare its surgical and audiometricoutcomes with the underlay type 1 tympanoplastytechnique using AlloDermW and temporalis fascia.

2. Methods

2.1. Subjects

Subjects were selected from the patient populationof the Department of Otolaryngology, The Hospitalfor Sick Children, which is an academic teachinghospital drawing referrals primarily from otolaryn-gologists. This project was approved by The Hospitalfor Sick Children Ethics Review Board which adheresto the ‘‘Tri-Council Policy Statement: Ethical Con-duct for Research Involving Humans’’. A databasekept by the senior otolaryngologist (B.C.P.) wassearched for children who underwent type 1 tym-panoplasty between 2004 and 2005. Patients whohad concomitant mastoidectomy or ossiculoplastywere not included in this study. The records of 34consecutive patients were reviewed: 18 receivedlateral graft type 1 tympanoplasty with AlloDermW

Lateral graft type 1 tympanoplasty using AlloDermW 1425

and 16 received underlay graft type 1 tympano-plasty (8 AlloDerm and 8 temporalis fascia). Thelateral graft group comprised 18 children (5 malesand 13 females) with a mean age of 10.84 � 3.09years. Ten of these subjects had previously unsuc-cessfully undergone tympanoplasty, and 16 hadperforations occupying greater than 50% of thetympanic membrane area. One subject had exten-sive tympanosclerosis, three had cholesteatomaand four had bilateral tympanicmembrane perfora-tions. The mean duration of follow up in this groupwas 0.54 � 0.50 years. The underlay type 1 tympa-noplasty group comprised 16 children (6 males and10 females) with a mean age of 13.03 � 2.58 years.Ten of these subjects had previously unsuccessfullyundergone tympanoplasty, and 15 had perforationsoccupying less than 50% of the tympanic membranearea. Two subjects had tympanosclerosis and onehad bilateral tympanic membrane perforations.The mean duration of follow up in the underlaytype 1 tympanoplasty group was 0.44 � 0.27 years.Apart from perforation size, these groups were notsignificantly different after Bonferroni correction(t-tests; P > 0.01).

2.2. Lateral graft technique usingAlloDermW

The technique used in children is very similar to thatdescribed by Fishman et al. [1] and only the sig-nificant differences will be described here. Theanterior and posterior walls of the external auditorycanal are infiltrated with lidocaine with 1:10,000epinephrine to improve hemostasis and make raisingthe tympanomeatal flaps easier. Great care is takennot to lacerate the anterior canal wall with the auralspeculum because the canal wall skin will later beraised as a pedicled flap in the technique described.

The Koerner’s flap is designed to include poster-ior canal skin down to the annulus from 6 to 12o’clock. The tympanic membrane remnant is thenreflected anteriorly and included in the anteriorcanal skin flap, which is left pedicled antero-lat-erally. The anterior flap is dissected using the roundknife in a retrograde fashion from medial to lateraluntil the flap is lateral to the bony-cartilaginousjunction. At this position it can be safely retractedout of the way when the drill is used to enlarge andsmooth the canal anteriorly.

The middle ear is not packed with GelfoamW

(Pharmacia & Upjohn Company Kalamazoo, MI).Rather, a slit is made in the superior aspect of theAlloDermW to allow the malleus handle to comelateral to the graft inferior to the insertion ofthe tensor tympani tendon. This ensures thatany remaining squamous epithelium on the malleus

handle remains lateral to the neo-tympanic mem-brane. We use 6—12/1000 in. thick AlloDermW andtrim the edges to decrease the amount of redun-dant graft placed anteriorly, under the base of theanterior pedicled flap. A suction canula is insertedinto the middle ear with the suction turned offusing a Hough-Cadogan foot pedal suction controlW

(Cadogen Engineering, Oklahoma city, OK). Suctionis then applied in order to medialize the graft andallow the annulus to be seen in relief through thethin AlloDermW graft. The AlloDermW is then takenoff the bony wall in quadrants (anterior, inferior,posterior and superior) and TisseelW (Baxter, Deer-field, IL) is applied to directly medialize the graftand attach it at the level of the annulus. Super-iorly, the AlloDermW from the posterior flap isoverlayed lateral to the superior portion of theexposed malleus by approximately 1 mm and isheld in place with Tisseel.

The anterior canal skin (and tympanic membraneremnant) pedicled graft is overlayed onto the Allo-DermW anteriorly. The tympanic membrane remnantis often divided so that maximal coverage of thecanal can be obtained. We rely on the AlloDermW toprovide the coverage for the neo-tympanic mem-brane. The anterior flap and the Koerner’s flap areoften shrunken at the end of the procedure andneither is capable of reaching the annulus. The goalof the flaps is to provide a source of living cells forepithelialization of the AlloDermW flap. This tech-nique does not require additional skin grafts.

A circular piece of GelfilmW (Pharmacia & UpjohnCompany Kalamazoo) is then place over the neo-tympanicmembrane andGelfoamW is packed circum-ferentially against the walls from the neo-tympanicmembrane laterally to the level of the flaps. A Mer-ocelW post-op pack (Medtronic Xomed, Jacksonville,FL) is placed into thecanalandCiprodexW (Bayer Inc.,Toronto, ON) otic drops are used to expand thesponge and hold the flaps and AlloDermW graft inplace. The sponge is removed 1 week later and oticantibiotic drops are used for 2 weeks after surgery.This technique does not require silicon stenting.

2.3. Surgical and audiometric evaluation

Patients were evaluated pre-operatively by thesenior author for tympanic membrane perforationsize and additional tympanic membrane findingssuch as tympanosclerosis or cholesteatoma, andpresence of bilateral tympanic membrane perfora-tions was noted. Children were examined usingotoscopy or endoscopy, and microscopy whererequired. Audiometric assessment included puretone average and air bone gap calculated as theaverage of values obtained at 500, 1000, 2000,

1426 P. Lai et al.

Table 1 Lateral graft type 1 tympanoplasty using AlloDermW vs. underlay type 1 tympanoplasty

Characteristics Lateral graft (n = 18) Underlay groups (n = 16) Difference

N Mean Standard deviation N Mean Standard deviation P-value

Pre-op PTA (dB) 18 43.3 13.2 16 21.3 12.5 0.000Pre-op ABG (dB) 18 31.9 12.5 16 11.9 12.2 0.000Pre-op SRT (dB) 16 39.1 13.7 13 18.9 6.8 0.000Post-op PTA (dB) 18 31.9 15.1 16 19.4 14.2 0.018Post-op ABG (dB) 18 19.2 17.9 16 10.7 14.9 0.143Post-op SRT (dB) 14 31.1 12.1 13 20.4 15.5 0.056PTA change (dB) 18 �11.4 13.4 16 �2.0 7.2 0.015ABG change (dB) 18 �12.7 20.2 16 �1.2 10.3 0.043SRT change (dB) 13 �8.5 13.0 10 �1.5 8.8 0.161Closure (yes:no) 17:1 16:2 0.528

PTA, pure tone average; ABG, air bone gap; SRT, speech reception threshold.

4000 Hz and speech reception threshold. Intra-operatively, tympanic membrane perforation sizewas again evaluated by the senior author. Childrenwith ossicular anomalies were excluded from thestudy. Postoperatively, tympanic membrane status,pure tone average, air bone gap and speech recep-tion threshold were evaluated 6 weeks after surgeryand every 6 months thereafter. Data was analyzedusing SPSS Version 11.1 (SPSS Inc.) using indepen-dent samples t-tests for continuous data, non-para-metric Mann—Whitney tests for categorical data andrepeated measures ANOVAs to compare pre-opera-tive versus postoperative measures.

Fig. 1 Effect of surgery on estimated marginal meanpure tone average for lateral and underlay type 1 tympa-noplasty groups.

3. Results

3.1. Underlay type 1 tympanoplasty

Pre- and postoperative surgical and audiometricoutcomes were compared for eight patients whoreceived underlay type 1 tympanoplasty using Allo-DermW and eight patients who underwent underlaytype 1 tympanoplasty using temporalis fascia. Therewere no significant differences across groups in anyof the measures evaluated, suggesting that bothAlloDermW and temporalis fascia can be used toreconstruct partial tympanic membrane perfora-tions using the underlay technique. Since therewere no statistically significant differences acrossAlloDermW and fascia underlay groups, these groupswere collapsed into one general underlay group forcomparison with the lateral graft group.

3.2. Lateral graft versus underlay type 1tympanoplasty

Results for the lateral graft and underlay groups areillustrated inTable 1. Childrenwhounderwent lateral

graft type 1 tympanoplasty pre-operatively had lar-ger tympanic membrane perforations, worse puretone averages, air bone gaps and speech receptionthresholds, as compared with children undergoingunderlay type 1 tympanoplasty (t-tests, P < 0.001).Postoperatively, there were no significant differ-ences across lateral graft and underlay type 1 tym-panoplasty groupswith respect to pure tone average,air bone gap or speech reception threshold afterBonferroni correction (t-tests; P > 0.01).

Figs. 1—3 illustrate estimated marginal meanchanges in audiometric scores with surgery.Figs. 1 and 2 demonstrate a significant improvementin pure tone average and air bone gap with surgery inboth the lateral and underlay type 1 tympanoplastygroups (ANOVA, P < 0.05). There was a significantinteraction of group and surgery on pure tone aver-age and air bone gap, signifying that relativeimprovements in pure tone average and air bonegap were greater in the lateral graft group as com-pared with the underlay type 1 tympanoplasty group

Lateral graft type 1 tympanoplasty using AlloDermW 1427

Fig. 2 Effect of surgery on estimated marginal mean airbone gap for lateral and underlay type 1 tympanoplastygroups.

(ANOVA, P < 0.05). Fig. 3 illustrates the trendtoward improvement in speech reception thresholdwith surgery in both groups, however, this findingdid not attain significance (ANOVA, P > 0.05). Therewas no interaction effect of group and surgery onspeech reception threshold, suggesting that bothlateral graft and underlay type 1 tympanoplastygroups improved in the same fashion (ANOVA,P > 0.05).

The perforation closure rate was higher in thelateral graft type 1 tympanoplasty group as com-pared with the underlay type 1 tympanoplasty group(94% versus 88%, respectively), although this findingdid not attain significance (Mann—Whitney,P > 0.05). The only non-surgical complication in thisstudy occurred in a patient who developed otitismedia after undergoing lateral graft type 1 tympa-

Fig. 3 Effect of surgery on estimated marginal meanspeech reception threshold for lateral and underlay type 1tympanoplasty groups.

noplasty, which resolved with insertion of a ventila-tion tube.

4. Discussion

Although most tympanic membrane perforationsbegin in childhood, there have been relativelyfew recent studies on new techniques of performingtype 1 tympanoplasty in children. Previous studiesfocused on reconstruction of small tympanic mem-brane perforations with an underlay technique.Total tympanic membrane reconstruction using Allo-DermW has been described in adults [1], but has notbeen as extensively studied for use in children. Inthis manuscript, we described the lateral graft type1 tympanoplasty technique using AlloDermW mod-ified for tympanic membrane reconstruction in chil-dren and compared its surgical and audiometricoutcomes with the traditional underlay type 1 tym-panoplasty technique using AlloDermW and tempor-alis fascia.

Pre-operatively, children who underwent lateralgraft type 1 tympanoplasty had larger tympanicmembrane perforations and worse pre-operativepure tone averages, air bone gaps and speech recep-tion thresholds as compared with those receivingunderlay type 1 tympanoplasty. The lateral grafttype 1 tympanoplasty involves removal of the squa-mous layer of the tympanic membrane and is typi-cally reserved for reconstruction of larger tympanicmembrane perforations. Worse pre-operative puretone averages, air bone gaps and speech receptionthresholds in the lateral graft type 1 tympanoplastygroup were likely due to the greater number ofpeople with large tympanic membrane perforationsin this group as compared with the underlay type 1tympanoplasty group.

Post-operatively, there were no significant differ-ences across lateral graft and underlay type 1 tym-panoplasty groups with respect to pure toneaverage, air bone gap, or speech reception thresh-old after Bonferroni correction (t-tests; P > 0.01).Mean postoperative pure tone averages, air bonegaps and speech reception thresholds were similarto those previously obtained primarily in adults[1,11]. These results suggest that reconstructionof large tympanic membrane perforations using alateral graft technique and AlloDermW can achievecomparable outcomes to those obtained afterunderlay tympanoplasty.

As depicted in Figs. 1 and 2, there was a signifi-cant improvement in pure tone average and air bonegap with surgery in both the lateral and underlaytype 1 tympanoplasty groups. Improvement in puretone average and air bone gap following type 1

1428 P. Lai et al.

tympanoplasty has been demonstrated previously[1,4,7,12]. In the present investigation, a greaterrelative improvement in pure tone average and airbone gap was noted in the lateral graft group ascompared with the underlay type 1 tympanoplastygroup. These results needs to be taken in the con-text of the fact that the lateral graft group hadworse audiometric recordings before surgery, andtherefore had more room for improvement beforereaching a plateau, as compared with the underlaytype 1 tympanoplasty group. Success following lat-eral graft type 1 tympanoplasty in this study mayalso be due to minimized blunting of the graft,decreased lateralization and the inherent durabilityof AlloDermW.

The tympanic membrane perforation closure ratein this study was 94% in the lateral graft group and88% in each of the underlay groups, for an overallclosure rate of 91%. This is comparable to previouslyreported closure rates for pediatric tympanic mem-brane perforations [13,14]. The trend seen in ourstudy for perforation closure rates to be higher afterlateral graft type 1 tympanoplasty as compared withunderlay type 1 tympanoplasty have been reportedpreviously [13]. This may be due to greater exposureduring surgery allowing for greater precision whileworking at the anterior drum margin or to theinherent durability of AlloDermW used in the lateralgraft group in the present study. High perforationclosure rates were found in the present studydespite the fact that 47% of operations were revisiontympanoplasties. Previous studies have also foundthat success rates for revision type 1 tympanoplastycould be similar to those achieved with primarysurgery [15,16].

There is debate about the optimal age at which atympanic membrane perforation should be safelyreconstructed. In a recent survey of otologists in theUnited Kingdom, most reported that they would notperform tympanoplasty in children younger than 10years of age because they have an increased inci-dence of otitis media and Eustachian tube dysfunc-tion which could lead to poorer results [17]. In thepresent study, we investigated children as young as7.2 years and found young age not to be an adversefactor for success after type 1 tympanoplasty.

The results reported must be taken in the contextthat the lateral graft type 1 tympanoplastydescribed in this manuscript is, not even in ourhands, the first choice for repair of type 1 tympanicmembrane perforations in children. In the Canadianhealth care setting, ‘‘routine’’ type 1 tympanicmembrane perforations are not commonly referredto an academic center. This technique has evolvedto allow us to obtain acceptable results in thesemore complicated perforations, and our results sup-

port our offering this procedure to patients referredwith complicated perforations or previously failedrepairs.

5. Conclusions

This study provides a description of the lateral grafttype 1 tympanoplasty technique using AlloDermW

used in a purely pediatric population. Results sug-gest that AlloDermW may be used for type 1 tympa-nic membrane reconstruction in children as asubstitute when temporalis fascia is not availableor desirable. Despite having larger perforations andworse pre-operative audiometric scores, childrenwho underwent lateral graft type 1 tympanoplastyachieved comparable postoperative audiometricresults and perforation closure rates as comparedwith children who underwent underlay type 1 tym-panoplasty. Success following type 1 tympanoplastyin the present study was realized even for youngchildren. A surgeon treating complicated tympanicmembrane perforations in children should considerthe lateral graft type 1 tympanoplasty using Allo-Derm1 as a surgical treatment option in theseuncommon cases.

References

[1] A.J. Fishman, M.S. Marrinan, T.C. Huang, S.J. Kanowitz,Total tympanic membrane reconstruction: AlloDerm versustemporalis fascia, Otolaryngol. Head Neck Surg. 132 (6)(2005) 906—915.

[2] LifeCell Corporation. AlloDerm: applying the power ofregenerative medicine, 2004. World wide web URL:http://www.lifecell.com.

[3] A.K. Abbas, A.H. Lichtman, J.S. Pober, The major histocom-patibility complex, in: A.K. Abbas, A.H. Lichtman, J.S. Pober(Eds.), Cellular and Molecular Immunology, WB Saunders,Philadelphia, 1991, pp. 99—113.

[4] J.E. Benecke Jr., Tympanic membrane grafting with Allo-Derm, Laryngoscope 111 (9) (2001) 1525—1527.

[5] A.P. Sclafani, S.A. McCormick, R. Cocker, Biophysical andmicroscopic analysis of homologous dermal and fascial mate-rials for facial aesthetic and reconstructive uses, Arch.Facial Plast. Surg. 4 (3) (2002) 164—171.

[6] T.J. Downey, A.L. Champeaux, A.B. Silva, AlloDerm tympa-noplasty of tympanic membrane perforations, Am. J. Otlar-yngol. 24 (1) (2003) 6—13.

[7] J.D. Vos, M.D. Latev, R.F. Labadie, et al., Use of AlloDerm intype I tympanoplasty: a comparison with native tissuegrafts, Laryngoscope 115 (9) (2005) 1599—1602.

[8] D. Saadat, M. Ng, S. Vadapalli, U.K. Sinha, Office myringo-plasty with AlloDerm, Laryngoscope 111 (1) (2001) 181—184.

[9] J.N. Fayad, T. Baino, S.C. Parisier, Preliminary results withthe use of AlloDerm in chronic otitis media, Laryngoscope113 (7) (2003) 1228—1230.

[10] A.M. Youssef, Use of acellular human dermal allograft intympanoplasty, Laryngoscope 109 (11) (1999) 1832—1833.

Lateral graft type 1 tympanoplasty using AlloDermW 1429

[11] W.P. Potsic, M.R. Winawer, R.R. Marsh, Tympanoplasty forthe anterior—superior perforation in children, Am. J. Otol.17 (1) (1996) 115—118.

[12] G.B. Singh, T.S. Sidhu, A. Sharma, N. Singh, Tympanplastytype I in children: an evaluative study, Int. J. Pediatr.Otorhinolaryngol. 69 (8) (2005) 1071—1076.

[13] F.M. Rizer, Overlay versus underlay tympanoplasty. Part I:historical review of the literature, Laryngoscope 107 (12Pt2)(1997) 1—25.

[14] T. Lau, M. Tos, Tympanoplasties in children: an analysis oflate results, Am. J. Otol. 7 (1) (1986) 55—59.

[15] A.G. Gibb, S.K. Chang, Myringoplasty (a review of 365operations), J. Laryngol. Otol. 96 (10) (1982) 915—930.

[16] P. Packer, A. Mackendrick, M. Solar, What’s best in myringo-plasty: underlay or overlay, dura or fascia? J. Laryngol. Otol.96 (1) (1982) 25—41.

[17] J.L. Lancaster, Z.G. Makura, G. Porter, M. McCormick, Pedia-tric tympanoplasty, J. Laryngol. Otol. 113 (7) (1999) 628—632.