Introduction

Bonding of brackets to enamel has been a criticalissue in orthodontics since the introduction ofdirect bonding, because of the biomechanicalimportance of a stable bracketadhesive inter-face to transfer the loads generated from theengagement of an activated archwire to thetooth. As new bonding agents were introduced,research focused on this area and the resultantpublication rate of papers on bonding increasedconsiderably. This is illustrated in the steadilyincreasing number of bonding papers appearingin the leading orthodontic journals.

In spite of the vast amount of informationpresented in hundreds of articles during the lastdecade, there is a remarkable lack of consensusregarding clinical bond strength values. More-over, no reliable protocol for estimating the invitro strength provided by orthodontic bonding

systems has been described. The necessity for sucha fundamental reference in restorative dentistryled to the establishment of the ASC MD156(Accredited Standards Committee 156 forMedical Devices) Task Group in 1990 sponsoredby the American Dental Association Council onDental Materials, Instruments, and Equipment.The main purpose of that task group was toexamine the clinical significance of in vitro bondstrength tests of composite resin restorativematerials bonded to tooth structure, includingenamel and dentine. The report (Sderholm,1991) published from this collaborative projectemphasized the clinical inapplicability of mostresearch methodologies employed in this area.Suggestions were made for overcoming problemsassociated with the standardization of various in vitro screening tests, thereby facilitating propercomparison among studies published fromdifferent research groups.

European Journal of Orthodontics 22 (2000) 1323 2000 European Orthodontic Society

The inappropriateness of conventional orthodontic

bond strength assessment protocols

T. Eliades* and W. A. Brantley** *Biomaterials Science Unit, University of Manchester Dental School, UK and **The Ohio State University, College of Dentistry, Columbus, OH, USA

SUMMARY The purpose of this article is to examine the soundness of conventional ortho-dontic bonding assessment methods. A classification of bond strength studies is proposedwith the testing environment (in vivo, in vitro, and ex vivo), loading mode (shear, tensile,and torsion), and bonding substrate (enamel, restorative, and prosthetic materials) servingas discriminating variables. Inconsistencies throughout the various stages of researchprotocols are analysed. These include the following: tooth selection, storage, and prepar-ation; bonding; testing; and data analysis with regard to the clinical applicability of thereported information, as well as the scientific integrity of the testing procedure.

Contradictory models may partially account for the considerable variability noted forreported bond strength values of different orthodontic bonding systems. Such discrep-ancies may also explain the conflicting evidence reported on the failure characteristics ofthe components of the bonding system in different trials examining the efficacy of nominallyidentical materials.

A novel approach to study the fatigue life of materials is proposed to understand theprocesses occurring prior to bond failure. Mock research data manipulation is also utilizedto illustrate the correct statistical treatment of findings, and recommendations for futureresearch are made to ensure scientific soundness and clinical applicability of data.

The purpose of this paper is to provide anoverview of the protocols currently available onorthodontic bonding assessment, examine theirscientific integrity, and reveal their underlyingclinical importance and impact on the extrapo-lation of conclusions to in vivo conditions. Each ofthe important variables at the different stages ofthe experimental procedures will be considered.

Classification of studies on orthodontic bondingevaluation

In general, orthodontic bond strength assessmentand the associated study of failure patterns mayfirst be classified according to the test environ-ment (Table 1) as:

(1) in vitro tests, usually performed with the aid of a mechanical testing machine or bysimulation of clinically applied debondingprocedures in the laboratory, where thefailure mode of the systems evaluated aredetermined by microscopic examination;

(2) measurements of the failure rates of bracketsin vivo, usually during the full course of treat-ment, with the bracket type and failure sitefrequency being the parameters examined;

(3) ex vivo studies utilizing finite elementanalysis modelling of the stress distributionsin the components of the enamel-adhesive-bracket system.

A second characterization of bond strengthtests can be made according to whether themode of load application is shear, tension, ortorsion. Use of shear loading has been verypopular due to the relative simplicity of theexperimental configuration and the presumablyincreased reliability of simulating debonding thatoccurs during treatment. The tension or torsionloading modes have been considered by manyinvestigators as less relevant to clinical practiceand have attracted much less interest.

A third classification of bond strength tests isbased upon the bonding surface of the substrate;enamel (Eliades et al., 1991), composite resinveneers (Kao et al., 1995), porcelain (Winchester,1991), and amalgam surfaces (Gross et al., 1997)have been employed. The emergence of thelatter three categories arises from the increasedemphasis on adult orthodontics during the lasttwo decades, which led to the necessity ofbonding aesthetic orthodontic attachments torestored teeth.

Orthodontic bond strength assessment: protocol stages

Tooth selection, storage, and preparation

A variety of teeth have been used in orthodonticbonding experiments, including upper centralincisors, premolars, and lower incisors; this has

14 T. ELIADES AND W. A. BRANTLEY

Table 1 Classification of orthodontic bond strength tests.

Environment Loading mode Substrate

In vitroMechanical testing machine;debonding with pliers,simulating clinical debonding procedures

In vivoAnalysis of rate and site ofbracket failure during thecourse of treatment

Ex vivoFinite element analysis modelling

ShearMovement of bar againstmounted rig (testing machine)

TensileWire dislodging loop (testing machine)

TorsionCustomized wrenches (testing machine)

EnamelFluoridated, normal

Composite resin veneersConventional orthophosphoricacid-etching using variousconcentration gradients andtimesFluoric acid treatment

PorcelainRoughened or unalteredSilane-treated

Amalgam

contributed to the lack of comparable resultsamong trials performed in different laboratories.While premolar extraction may be an integralpart of orthodontic therapy, facilitating thecollection of those teeth, premolar crown con-tour variations (Taylor, 1978) may complicatethe effort to have substrate surface consistency.

On the other hand, upper and lower incisorsare mostly retrieved from periodontally involveddentitions. The use of such teeth introduces thecomplicating factor of the age of the averageperiodontal patient, since the fluoride content inthe outermost surface layers has been docu-mented to change with age (Weatherell et al.,1972); perhaps etching patterns vary accordingly,although no evidence regarding this parameterhas been presented. In addition, possibleadsorption of inorganic or proteinaceous species,as well as the consequences of various thera-peutic procedures and pharmaceutical agentsadministered to these patients, may modify thereactivity of the enamel surface layers with anundetermined impact on etching patterns.

An overview of bond strength experimentalprotocols reveals a wide range of storage timeperiods extending from 24 hours to 5 years(Williams and Svare, 1985). When these varyingstorage times are combined with the use ofmiscellaneous storage media (Rueggeberg, 1990)that have employed different concentrations ofthymol, saline, aqueous chloramine, and formalin,it becomes very difficult to draw conclusionsfrom such studies. The variability in results has prompted examination of the influence ofpost-extraction time and storage conditions onbond strength. The majority of the investigationswere not able to demonstrate any additionalsignificant effect when teeth were stored for timeperiods greater than 20 minutes (Sderholm,1991). However, bond strength values providedby teeth stored in formalin were reported to betwice as much as those of their saline-storedcounterparts (Kimura et al., 1985), although noinformation was provided about the statisticalanalysis used in that study. In one of the fewarticles published on the effect of storagemedium on enamel, Mhlemann (1964) showedthat enamel specimens stored in physiologicsaline were softer than corresponding specimens

stored in water. Linden (1968) subsequentlyexamined the structure of enamel in extractedteeth following miscellaneous storage condi-tions and found only minor colour differences.Silverstone (1967) suggested the avoidance offormaldehyde, because its strong acidity follow-ing oxidation to formic acid may affect the pH ofstorage media. Even though an animal model(Pashley et al., 1988) has reproduced in vitro bondstrength values acquired from in vivo tests, otherin vitro studies (Jemt et al., 1986), that examinedthe bond strength between glass ionomers andenamel, yielded values twice as high as in vivomeasurements under nominally identical experi-mental conditions.

The fact that most of the more recentlypublished research on adhesion in the dentalscientific literature deals with dentinal bondstrength, which is critical for the survival ofrestorations, emphasizes the lack of data onmodelling of orthodontic bonding conditions invitro. Extrapolation of the results of dentinalbonding studies to enamel bonding must beviewed with scepticism because of the highlyorganic content of dentine, which might bealtered by ionic or enzymatic storage environ-ments, in contrast to the highly inorganic enamel,which seems to be unaffected (Rueggeberg,1990). Thus, it is plausible that extraction timeand storage media have little, if any, influence onadhesive bond strength to enamel. Most criticalreviews (Rueggeberg, 1990; Sderholm, 1991)suggest that a storage time of 6 months may beused for normalization purposes among miscel-laneous experimental protocols.

Often, experimental treatment of collectedteeth include levelling of the prospective enamelsurfaces by grinding in an attempt to standardizethe topographic variants of the substrate(Eliades et al., 1991). The argument supportingthis notion relates to the incongruities found inthe profile contour and convexity of the labialenamel surface, particularly those of premolars;the latter induces a variable pertinent toadaptation of the adhesive layer to the toothcrown, inevitably modifying the composite resinthickness. Although this procedure is obviouslyinappropriate to clinical conditions, its majorflaw is the profound alteration of the substratum.

BOND STRENGTH TESTS 15

Apparently, surface layers of enamel possessproperties dissimilar to those found in deeperzones, due to the higher fluoride content of the outermost 10 m m layer (Jenkins, 1978). Inaddition, grinding of the enamel surfaces is per-formed ad libitum, using stones or silica discs ofvarying roughness, while the duration of thisprocess is determined by visual inspection, beingthus highly subjective (Schneider et al., 1981).Therefore, not only is there failure in con-structing a simulated clinical analogue, but thismethod also introduces a variability in enamelcondition that precludes comparing results fromdifferent studies.

An interesting aspect associated with theroutine practice of examining the bond strengthof various materials bonded to extracted teethreceived attention during the 1980s. Whilealarming levels of bacterial contamination in thedental operatory during cavity preparation hadpreviously been noted (Larato et al., 1966), 20 years elapsed before the potentially hazard-ous nature of manual utilization and especiallygrinding of extracted teeth was documented(Pagniano et al., 1985). Rueggeberg (1990) indi-cated that extracted teeth stored in alcohol orformalin disinfectants contain substantial numbersof micro-organisms that are capable of colonizingthe surroundings of the laboratory via aerosolspreading induced by the preparation of teethwith air turbines. Staphylococci, Pseudomonas,Shingella, Enterobacter, Klebsiella, and Proteuswere found to be the prevailing species. How-ever, qualitative aspects of bacterial coloniesdetected varied considerably, depending uponthe microflora of the oral cavity from which theteeth were collected, and cross-contaminationeventually occurred from specimen storage ingroups. Investigations concluded that it is doubt-ful that commonly-used storage media possessany bactericidal activity, thus necessitating auto-claving of extracted teeth.

Use of this procedure gave rise to studiesfocusing on the impact of sterilization on thebond strength and enamel structure alterations.The consensus from the limited data available is that autoclaving at 127C for 20 minutes,followed by storage in 1 per cent sodium hypo-chloride, does not seem to alter measured values

of bond strength or the enamel morphologyobserved with the scanning electron microscope(Shaffer et al., 1985). Nonetheless, the ASC MD156 Task Group report (Sderholm, 1991) sug-gested that minimal guidelines, including the useof gloves, masks, and protective eyewear must beapplied whenever animal or human body partsare handled.

Bonding

In general, orthodontic bonding to enamel mayinvolve a combination of the following:

(1) penetration of the initially fluid material intothe etched enamel and formation of resintags after polymerization;

(2) development of strongly bonded surfaceprecipitates, which serve as a substrate towhich a resin can be mechanically retainedor chemically bonded (Causton andJohnson, 1982);

(3) chemical bonding to the calcium ion of the hydroxyapatite principal constituent of enamel, which is employed in manyapproaches involving polycarboxylate orpolyphosphate ionic binding (Smith andCartz, 1973).

Evidence obtained in the last decade suggeststhat both phosphoric acid concentration andetching time may be significantly reduced with-out notable effects on bond strength (Barkmeieret al., 1987; Wang et al., 1994).

The procedure of adhesive application to thebracket base has raised the issues of the quan-titative aspects of adhesive and force utilizationduring bonding. The methods in published studies(Eliades et al., 1991) involve either the applicationof a standardized quantity of adhesive or the useof an undetermined amount of composite resin.Even though the first approach may normalizevariables related to adhesive paste application,allowing for the estimation of reference materialproperties such as degree of conversion andmonomer leaching, it lacks the essential elementof simulating the typical clinical procedureemployed by orthodontists. A method proposedto overcome this deficiency (Eliades et al., 1995)

16 T. ELIADES AND W. A. BRANTLEY

combines components from both approaches byhaving multiple pilot trials involving applicationof an adhesive to bracket bases for bonding by a trained orthodontist. This approach allows anestimate of the weight range of the adhesiveused, which represents a standardized baselineamount for application to the bracket bases.

A similar concern has been expressed aboutforce application during bracketadhesive attach-ment to enamel. In the majority of protocols thisis achieved by manual, free-handed applicationof an undetermined amount of pressure to thebracket. Efforts to adjust the pressure by apply-ing a fixed load to the bracket will yield moreconsistent results (Kao et al., 1995). However, aserious problem will arise if the amount of forceutilized results in significantly thinner adhesivelayers, with unknown effects on the materialproperties.

In spite of the presumed appropriateness ofsimulating the in vivo milieu in laboratory test-ing, it is worth noting that the oral environmentcontains a number of parameters that are impos-sible to reconstruct in an ex vivo model. Some ofthese factors are the stresses arising from anactivated archwire coupled with occlusal loads,extreme pH and temperature variations, and thepresence of complex oral microflora and theirby-products. This latter factor has been foundcapable of inducing substantial alterations in thestructure and surface properties of restorativematerials, orthodontic adhesives (Matasa, 1995),and archwires exposed in the oral cavity (Oshidaet al., 1992). In particular, orthodontic adhesivedegradation induced by microbial attack duringtreatment has been recently described by Matasa(1995), who examined retrieved brackets intendedfor recycling.

Testing

The mode of load application and the instru-mental configuration for bond strength testinghave been investigated by Katona and colleagues(Katona and Chen, 1994; Katona and Moore,1994; Katona, 1997). Finite element analysis hasestablished that the stress distribution withinthe adhesive layer, and the stresses generated in the brackets and enamel during testing are

inhomogeneous, contradicting the uniform stressassumption that has been prevalent in the majorityof in vitro experimental protocols. Evidenceemphasizing the inappropriateness of comparingresults derived from different loading modes(shear, tension, and torsion) was presented, and it was shown that the maximum stressesdeveloped in the orthodontic bonding systemunder tensile loading may be five times greaterthan the reported average stress (Katona, 1997).Hence, traditional bond strength studies sub-stantially under-estimate the probability ofsystem failure. Moreover, failure analyses thatare intended to provide inferences about thestrength of individual components of the bond-ing system based on their prospective interfacialfracture characteristics, should be questioned(Eliades et al., 1993). This is because the site offailure may arise from crack initiation caused byhigher stresses compared with other areas, whichis not taken into consideration in the traditionalassumption of homogeneous stress.

The validity of comparing results of similarstudies is affected by the experimental test con-figuration, as analysed by Fox et al. (1994). Theapplied force may generate moments of variousmagnitudes, depending upon the distance of thepoint of force application from the bracket base surface. This parameter may complicatethe extrapolation of conclusions regarding theanticipated failure events (Van Noort et al., 1989).

In summary, some critical aspects of ortho-dontic bond strength protocols that affect theoutcome of research trials may include thefollowing:

1. The cross-head speed of the loading plate inshear testing is usually set at 0.5 mm/min forconsistency (Eliades et al., 1991; Kao et al.,1995), although this value lacks correspond-ence to clinical conditions. In vivo debondingincidents are expected to occur at muchhigher impact velocity, where viscoelasticbehaviour of the adhesive, which may beimportant at low cross-head speeds, is largelyabsent.

2. In debonding procedures where the bracket ispulled with the use of a wire loop, the loopharness adaptation and frictional resistance

BOND STRENGTH TESTS 17

may complicate interpretation of the results.Katona and Chen (1994) proposed that long,thin wires should be used in such an experi-mental model.

3. Bracket design may contribute to misalign-ment of load application, making the bondingsystem prone to failure, depending on the stressgradients generated. It has also been foundthat variability exists among manufacturerswith respect to wing design or dimensions forbrackets with a nominally identical prescrip-tion (Katona, 1997). This variability poses a substantial problem for the comparison ofstudies evaluating bracket bond strength.

Another previously unstudied factor affectingthe survival of orthodontic bonding may be thepropagation of fatigue damage in the adhesivecomponent of the enamel-adhesive-bracketsystem. The five major stages of fatigue failure(Suresh, 1991) include:

(1) microstructural changes initiating nucleationof permanent damage;

(2) microscopic crack formation;(3) growth of flaws to yield macroscopic cracks;(4) stable propagation of macrocracks;(5) structural instability leading to failure.

A number of variables relevant to theenvironmental conditions, as well as the mech-anical properties and structural configuration ofthe bonding system members, have a dominantrole in determining the rate of crack propagationand the progression of failure. In general, researchin this field employs two major approaches instudying fatigue phenomena (Suresh, 1991):

(1) total-life approach where the objective is tocharacterize the cyclic stress or strain rangerequired for initiation of a dominant crack inan initially uncracked specimen and prop-agation of this flaw until failure is reached;

(2) defect-tolerant approach that is based uponthe premise that all engineering componentsare inherently flawed. Therefore, if the extentof microdefects present in an as-receivedspecimen can be characterized, the numberof fatigue cycles or time to propagate the

dominant crack from its initial size to acritical dimension leading to failure can bedetermined, thereby yielding the fatigue life.

In new research studies on fatigue in ortho-dontic bonding systems, post-mortem analyseswould be required to provide evidence about the potentially complex interactions among thecomponents of the system. The scarcity of suchevidence may be attributed to the multiplicity ofthe materials in the system, the complex mech-anical behaviour at the diverse interfaces, andthe anticipated subtle microscopic character ofthe fatigue process.

The crack nucleation and propagation thatdetermine fatigue life have been found to dependupon the testing environment (Hertzberg andManson, 1980). For polymers, the rate and modeof microscopic failure progression are affectedby the molecular structure, the nature of thecyclic loading conditions and the type ofdeformation (elastic, linear, or non-linear visco-elastic). Since it is probable that the detailedstages of the fatigue response cannot be detectedin the laboratory for an orthodontic bondingsystem, the location, and description of flaws willprobably be confined to the site of terminal orcatastrophic failure.

The clinical implications of these conjecturedfatigue processes in orthodontic bonding systemsalso remain to be elucidated and it is doubtfulthat the sensitive methods required to investigatethese phenomena will be developed in the nearfuture. The necessity of introducing a bracket fortesting applications may partially lift the burdenof hypothetical inferences and approximationsexisting within some models (Katona, 1997)currently employed to study orthodontic bondstrength. These models generally limit theapplicability of the research findings for clinic-ally orientated orthodontists, who will disregardreported observations unless they have evidentpractical significance.

Fox et al. (1994) recently presented an extensivecritique of 60 publications on orthodontic bondstrength testing. Upon reviewing 22 articlesinvestigating the bond strength of a well-knowncommercial product, they found that variations

18 T. ELIADES AND W. A. BRANTLEY

in tooth type, storage conditions, method ofdebonding, analysis of the results, and theselection of other products for comparisonresulted in none of the studies having the samemethodology. Consequently, despite the largenumber of previous publications, Fox et al.(1994) concluded that the bond strength of thiscommercial product had not been properlystudied and they proposed a detailed protocolfor bond strength evaluation.

Data analysis and presentation

A review of the literature on orthodontic bondstrength testing reveals some basic inconsistenciesin the use of units and the statistical analysis ofdata. A hypothetical research project to inves-tigate the shear bond strength for three brands of ceramic brackets to upper enamel incisorsurfaces, using a chemically-cured orthodonticadhesive, may serve to elucidate some aspects of these issues. The data used in this example are based upon a previously published study(Eliades et al., 1991). The sample size for eachgroup of brackets tested was 10. The mean andstandard deviation of the debonding force(measured in Newtons) for each bracket type islisted in the first row of Table 2. Assumingrectangular bracket bases, the calculated surfaceareas in m2 (using the SI system of units) are provided in the second row. The third rowpresents the resultant mean bond strength valuesexpressed in the stress units of N/m2 or MPa(ignoring the previously discussed effects of stressconcentrations). A scanning electron micro-scopic investigation of the bracket base surfaceswas also performed to seek correlation of the

bond strength values with the morphological andstructural features of the bases. Representativephotomicrographs are provided in Figures 13for brackets X, Z and Y, respectively.

Table 2 reveals an interesting discrepancybetween the statistically significant differencesfound for debonding forces and mean bondstrengths. One source of this discrepancy is thatthe effective surface area of the bracket base incontact with the adhesive is far from rectangular(Figures 13). Furthermore, substantial variationsin the load distribution patterns are expectedamong the three bracket types because ofdifferences in overall morphology, as well as inthe interfacial characteristics of the bracketadhesive complex. The previous assumption of auniform load distribution among the interfacesinvolved must be rejected. Moreover, significantfluctuations in adhesive thickness have been

BOND STRENGTH TESTS 19

Table 2 Results (mean SD) from the shear bond strength test of three brackets bonded to enamel (n = 10)*.

Measured variable Bracket X Bracket Y Bracket Z

Debonding force (N) 17.9 0.9 [A] 15.2 1.0 [B] 14.1 0.6 [B]Surface area of base (m2) 1.6 106 1.4 106 1.8 106Mean bond strength (MPa) 11.4 [A] 10.9 [A] 7.8 [B]

*Hypothetical model using arbitrarily defined force values. The bracket base area was calculated from the dimensions ofrepresentative upper incisor ceramic brackets, assuming a simple rectangular geometry.Means with same letters are not significantly different at the a = 0.05 level, using the Tukey multiple range test.

Figure 1 SEM photomicrograph of the base of a poly-crystalline ceramic bracket (X). Note the dramatic increaseof surface area through the projection of crystal-like forma-tions. (Original magnification 100.)

noted between smooth bracket bases that con-tribute to a homogeneous load application andrough bracket bases, where crystal-like formationresults in the retention of the adhesive (Eliadeset al., 1991). This effect may depend upon therheological properties of the adhesive, and the size of the pores or grooves formed in thebracket base. The presence of these variablesmay validate the argument that clinicians shouldnot be concerned with the expression of bondstrength values in terms of stress, because this

may be irrelevant to the actual force at which thesystem fails in vivo. Moreover, some authorshave provided evidence supporting the independ-ence of bond strength variations from the nom-inal area and mesh size for 14 types of bracketbases (Dickinson and Powers, 1980). Thus, in theforegoing example, which represents the averagestyle of papers published in this field, the actualcontact area of the bracket base cannot beaccurately estimated to allow for the propertransformation of units from force to stress, andthere is little reliability in projecting laboratoryresults to clinical conditions.

An important final consideration about theuse of units has been described by Katona(1997), who pointed out the potential confusionin reporting torsional strength, which is expressedin N/m as the quotient of torque (Nm) and area(m2). The confusion arises because the units forshear strength are N/m2, and torsion correspondsto a state of shear stress (Popov, 1968). Whentorsional loading is involved, the polar momentof inertia must be considered, which describesthe distribution of the cross-section area aboutthe axis of twisting.

The statistical treatment of data in Table 2employed one-way analysis of variance (ANOVA),followed by the Tukey multiple range test. Thedata for the debonding force and mean bondstrength were subjected to separate statisticalanalyses and the results are provided in Table 3.Assuming that all other testing variables havethe same effect upon all three samples and that these groups are normally distributed, it is shown that two different pairwise multiplecomparison tests (Tukey, Duncan) can yielddifferent results about which specific groups aresignificantly different. The discrepancy noted inTable 3 may not be limited to simply stating thedifferences between the two methods of statisticalanalysis, since authors often feel that theyshould provide substantiation of the reportedinformation. In this example, the researcher whoused the multiple range test (Duncan) showingmore significant differences between the threegroups would probably attempt to correlatedifferences in mean debonding force to bracketbase features observed in the SEM photomicro-graphs, making inferences about the effect of the

20 T. ELIADES AND W. A. BRANTLEY

Figure 2 SEM photomicrograph of the base of a poly-crystalline ceramic bracket (Y) presenting less morpho-logical variability compared with bracket X. (Originalmagnification 100.)

Figure 3 SEM photomicrograph of the base of a single-crystal ceramic bracket (Z) showing a uniform, relativelysmooth surface. (Original magnification 100.)

specific base design. Thus, a series of argumentspertinent to microstructural and morphologicalfeatures of the bracket base surfaces may beformed in accordance with the results observed;in this case, the proposed theoretical justificationmay lack a scientific basis.

The appropriate sample size has been a matterof dispute and has served as a criterion of the soundness of research. This is attributed tothe likelihood that sample sizes of less than 10specimens per group may not follow a normaldistribution. Normally distributed samples, rand-omness of individual sampling and homogeneityof variances are fundamental assumptions forthe use of ANOVA (Sokal and Rohlf, 1995). The publication of studies reporting mean bondstrength values derived from groups containingless than 10 specimens has resulted in strongcriticism (Fox et al., 1994) and some authorsadvocated that the minimum sample size pergroup should be of the order of 30. Rather thanset a specific sample size a priori, the appropriatestatistical approach is to perform a poweranalysis (Sokal and Rohlf, 1995) on the results ofpilot experiments to establish the correct samplesize that meets a and b levels previouslyestablished in the research protocol.

Finally, the use of a Weibull survival analysishas been proposed (Fox et al., 1994) for bondstrength studies. This analysis does not requirenormally distributed samples and also focuses on the tail of the distribution that contains thesmaller values, thereby providing more emphasison the safety of the bonding system performance.

Future research directions

As the state of our current knowledge onbonding advances rapidly, the clarification of thepurpose of the bond strength research protocolsand the precise definition of the objectives ofrelevant research may assist investigators in thefield to arrive at clinically meaningful conclu-sions. This will help orthodontists to efficientlyhandle this clinical stage, while, at the same time,provoke the manufacturing of materials that willmeet the clinical demands.

The simulation of clinical conditions is a taskthat is not seen to be attainable in the nearfuture. The fact that the failure pattern of con-trolled debonding procedures occurring in a setup involving the use of a testing machine bearslittle comparison with the topography of failureoccurring in vivo (Katona, 1997), distorts theapplicability of the information provided. Apotential solution might be the construction ofdebonding pliers for testing purposes, which willpresumably direct the applied loads according tothe manufacturers suggestion, while providing,through a strain gauge mechanism, a quantitativescale of the magnitude of the force applied. Thiswill facilitate a reference that may serve as adiscriminating variable for comparison purposesamong various adhesivebracket systems. How-ever, this approach limits the effectiveness of thetrial to the debonding incidents occurring underthe given loading pattern, and cannot provide aninsight of the wide array of tensile or torqueloads transferred to the system, emanating frommastication or engagement of heavy rectangulararchwires on fully prescribed, pre-adjusted brackets.

Additional approaches in resolving the issue of soundness of in vitro derived data, mayinvolve the challenging area of fatigue of theadhesive-bracket system. This may be accom-plished through protocols aimed at exploring the total-life tolerance of the system to a low-magnitude, cyclic, mechanical stress, rather thanmeasuring its strength to a sudden, powerfulimpact, such as that occurring during standardstrength tests. This is because the system may failat considerably lower values, as it was notconstructed to resist high impact loading, whichis supposedly absent in vivo.

BOND STRENGTH TESTS 21

Table 3 Significant differences among results(mean SD) of the debonding forces (N) in Table 2,using two different multiple pairwise comparisontests.

Bracket Mean SD Tukey Duncan type grouping grouping

X 17.9 0.9 A AY 15.2 1.0 B BZ 14.1 0.6 B C

ANOVA F value: 42.9.Means with same letters are not significantly different at the a = 0.05 level.

Another route, currently followed by someinvestigators, is the finite element analysis, whichhas already clarified some issues pertaining to research methodology and interpretation offindings relevant to bond strength experimentalconfiguration. The introduction of a bracket forstrictly testing applications proposed by Katona(1997), points at the direction of standardizationof research and elimination of the exceedinglyincreased variability of the findings noted amongdifferent studies investigating identical materials.

While it is true that certain aspects of physicaland chemical adhesive properties may be clarifiedby in vitro approaches, the actual performance ofthe system can only be illustrated in the environ-ment where it was intended to function. There-fore, clinical studies focusing on the failure rateof appliances in a controlled environment underidentical conditions with respect to malocclusion,appliance prescription and slot dimensions, statusof patients, as well as applied mechanics, maytarget the examined variable. However, thesetrials are laborious, requiring extended monitor-ing, which may be problematic in a private office.On the other hand, the multiplicity of variablesexisting in an academic environment pertainingto the educational scope of the programmeinvolving the exposure of students to a widevariety of appliances and mechanics, as well asthe treatment of a range of malocclusions, makethe construction of similar research protocolsnearly infeasible.

The clear definition of research objectivesimplemented by the construction of a standard-ized bracket analogue is intended for solelytesting applications. The controlled testing envir-onment with respect to the application of forcesimposing measurable biomechanical effects onthe bracketadhesive system, and the appropriatestatistical design, may enhance the integrity and repeatability of research findings. Thismight be an integral part of conducting soundinvestigations.

Finally, it seems that the underlying difficultyin carrying out these tests relates to the challengeof defining the goals of research; an essentialstage of this is the elimination of some apparentdeficiencies existing in research protocol design.Performing investigations with vaguely defined

objectives and employing obviously erroneousmethodologies indiscriminately, inevitably con-fines the purpose of our research efforts.

Address for correspondence

Dr Theodore Eliades57 Agnoston Hiroon StreetNea Ionia 14231AthensGreece

References

Barkmeier W W, Gwinnett A J, Shaffer S E 1987 Effects ofreduced acid concentration and etching time on bondstrength and enamel morphology. Journal of ClinicalOrthodontics 21: 395398

Causton B E, Johnson N W 1982 Improvement ofpolycarboxylate adhesion to dentine by the use of a newcalcifying solution. An in vitro study. British DentalJournal 152: 911

Dickinson P T, Powers J M 1980 Evaluation of fourteendirect-bonding orthodontic bases. American Journal ofOrthodontics 78: 630639

Eliades T, Viazis A D, Eliades G 1991 Bonding of ceramicbrackets to enamel: morphologic and structural consid-erations. American Journal of Orthodontics and Dento-facial Orthopedics 99: 369375

Eliades T, Viazis A D, Lekka M 1993 Failure mode analysisof ceramic brackets bonded to enamel. American Journalof Orthodontics and Dentofacial Orthopedics 104: 2126

Eliades T, Eliades G, Brantley W A, Johnston W M 1995Polymerization efficiency of chemically cured and visiblelight-cured orthodontic adhesives: degree of cure. AmericanJournal of Orthodontics and Dentofacial Orthopedics108: 294301

Fox N A, McCabe J F, Buckley J G 1994 A critique of bondstrength testing in orthodontics. British Journal of Ortho-dontics 21: 3343

Gross M W, Foley T F, Mamandras A H 1997 Directbonding to Adlloy-treated amalgam. American Journal ofOrthodontics and Dentofacial Orthopedics 112: 252258

Hertzberg R W, Manson J A 1980 Fatigue of engineeringplastics. Academic Press, New York

Jemt T, Stlblad P A, ilo G 1986 Adhesion of poly-carboxylate-based dental cements on enamel: an in vivostudy. Journal of Dental Research 65: 885887

Jenkins G N 1978 The physiology of the mouth. BlackwellScientific Publications, Oxford

Kao E C, Eliades T, Rezvan E, Johnston W M 1995Torsional bond strength and failure pattern of ceramicbrackets bonded to composite resin laminate veneers.European Journal of Orthodontics 17: 533540

22 T. ELIADES AND W. A. BRANTLEY

Katona T R 1997 A comparison of the stresses developed intension, shear peel, and torsion strength testing of directbonded orthodontic brackets. American Journal of Ortho-dontics and Dentofacial Orthopedics 112: 244251

Katona T R, Chen J 1994 Engineering and experimentalanalyses of the tensile loads applied during strengthtesting of direct bonded orthodontic brackets. AmericanJournal of Orthodontics and Dentofacial Orthopedics106: 167174

Katona T R, Moore B K 1994 The effects of load misalign-ment on tensile load testing of direct bonded orthodonticbracketsa finite element model. American Journal ofOrthodontics and Dentofacial Orthopedics 105: 543551

Kimura S, Shimizu T, Fujii B 1985 Influence of dentin onbonding of composite resin, Part 1effect of fresh dentinand storage conditions. Dental Materials Journal 4: 6880

Larato D C, Ruskin P F, Martin A, Delanko R 1966 Effectof a dental air turbine drill on the bacterial counts in air.Journal of Prosthetic Dentistry 16: 758765

Linden L A 1968 Microscopic observations of fluid flowthrough enamel in vitro. Odontologisk Revy 19: 349365

Matasa C G 1995 Microbial attack of orthodontic adhesives.American Journal of Orthodontics and DentofacialOrthopedics 108: 132141

Mhlemann H R 1964 Storage medium and enamelhardness. Helvetica Odontologica Acta 8: 112117

Oshida Y, Sachdeva R C, Miyazaki S 1992 Microanalyticalcharacterization and surface modification of TiNi ortho-dontic archwires. Biomedical Materials and Engineering2: 5169

Pagniano R P, Scheid R C, Rosen S, Beck F M 1985Reducing airborne microbes in the preclinical dentallaboratory. Journal of Dental Education 50: 234235

Pashley E L, Tao L, Mackert J R, Pashley D H 1988Comparison of in vivo vs. in vitro bonding of compositeresin to the dentin of canine teeth. Journal of DentalResearch 67: 467470

Popov E P 1968 Introduction to mechanics of solids.Prentice-Hall, Englewood Cliffs, NJ, USA

Rueggeberg F A 1990 Substrate for adhesion testing totooth structurereview of the literature. Dental Materials7: 210

Schneider P M, Messer L B, Douglas W H 1981 The effectof enamel surface reduction in vitro on the bonding ofcomposite resin to permanent human enamel. Journal ofDental Research 60: 895900

Shaffer S E, Barkmeier W W, Gwinnett A J 1985 Effect ofdisinfection/sterilization on in vitro enamel bonding.Journal of Dental Education 49: 658659

Silverstone L 1967 The histopathology of enamel lesionsproduced in vitro and their relation to enamel caries,Vol.1. PhD Thesis, University of Bristol, UK

Smith D C, Cartz L 1973 Crystalline interface formed bypolyacrylic acid and tooth enamel. Journal of DentalResearch 52: 1155 (Abstract)

Sderholm K-J M 1991 Correlation of in vivo and in vitroperformance of adhesive restorative materials: a reportof the ASC MD 156 task group on test methods for theadhesion of restorative materials. Dental Materials 7:7483

Sokal R R, Rohlf F J 1981 Biometry. The principles andpractice of statistics in biological research, 3rd edn.Freeman, New York

Suresh S 1991 Fatigue of materials. Cambridge Solid StateScience Series, Cambridge

Taylor R M S 1978 Variation in morphology of teeth.Anthropologic and forensic aspects. Thomas Springfield,IL, USA

Van Noort R, Noroozi S, Howard I C, Cardew G 1989 A critique of bond strength measurements. Journal ofDentistry 17: 6167

Wang W N, Yeh C L, Fang B D, Sun K T, Arvystas M G1994 Effect of H3PO4 concentration on bond strength.Angle Orthodontist 64: 377382

Weatherell J A, Robinson C, Hallsworth A S 1972 Changesin the fluoride concentration of the labial enamel surfacewith age. Caries Research 6: 312324

Williams V D, Svare C W 1985 The effect of five-year stor-age prior to bonding on enamel/composite bond strength.Journal of Dental Research 64: 151154

Winchester L 1991 Direct orthodontic bonding to porcelain:an in vitro study. British Journal of Orthodontics 18:299308

BOND STRENGTH TESTS 23