A critical examination of the dental literature for wear publications and wear testing methods since the early 1970 s has resulted in two statements that form the basis of this article. When discussing wear data in 1984, Lambrechts and co-workers stated that the ‘experimental results are best characterised as inconclusive’ [1] while in 1987 Roulet stated that ‘there are almost as many wear testing devices as there are scientists who are interested in wear’ [2].

Critical evaluation of a scientific or clinical manuscript relies on the level of evidence offered, where evidence is defined as ‘anything that establishes a fact or gives reason for believing something’ [3]. First and foremost, when assessing what constitutes evidence in the dental literature, one must always look towards the clinical data. Clinical data can be used to predict the survival probability of dental restorations although – and this is critically important – they cannot establish the underlying reason for clinical failure. The early clinical in vivo wear studies compared various resin-based composite (RBC) materials under direct observation at regular time intervals [4], [5] or indirect assessment of tooth replicas over time [6]. The major reasons for the lack of in vivo studies include the significant shortcomings of both the observation indices [4], [5] and the tooth replica [6] techniques. In the United States Public Health Service (USPHS) technique reported by Cvar and Ryge [4], two or more independent clinical observers directly evaluated the wear of restorations after regular time intervals and rated them as either clinically ideal (Alpha), acceptable (Bravo) or unacceptable (Charlie) [7]. One of the drawbacks of the USPHS technique was that clinical observers were subjective in their assessment and often disagreed over ratings such that restorations nominated as Bravo at one visit, could be re-assessed as Alpha at a later visit [7] when wear would be expected to increase with time and not decrease. In addition, it was difficult to achieve the continuity of clinical observers for subsequent patient visits, notwithstanding patient attendance, expense and compliance issues [7]. The indirect technique reported by Leinfelder et al. [6] quantitatively assessed restorative wear indirectly by comparing stone cast replicas of teeth containing worn restorations to six standard casts corresponding to specific wear depths in 0.1 mm increments. The technique required positive and negative impressions of the restored teeth to be taken. However, both dental impression and stone materials are susceptible to dimensional instability on setting, thereby affecting the accuracy and precision of the replicas and as a result the confidence in the in vivo restorative wear data [8]. This approach took two years to achieve the same findings as the direct (USPHS) method [4], [5], which took four years, highlighting a significant saving in the costs associated with the clinical trial. It was, however, susceptible to inter-evaluator error of 0.05 mm (50%), since the standard casts used 0.1 mm increments. In all experimentation the resolution of the measurement tool should be an order of magnitude better than the critical discriminatory level (i.e. lower in this instance), namely 0.005 mm (or better) here [9], again highlighting the difficulties with the interpretation of in vivo wear data. Unfortunately, the ADA specification for the maximum wear a RBC should encounter in a year is 0.05 mm, which unfortunately is exactly the inter-evaluator error in the Leinfelder et al. [6] technique. In summary, the imprecision of the clinical data values for both the direct (USPHS [4], [5]) and indirect (Leinfelder et al. [6]) in vivo methods routinely employed to assess the wear resistance of restorative materials renders the clinical data compromised at best and useless in reality [9].

Secondly when assessing what constitutes evidence in the dental literature we must look at the laboratory data. Since the first RBC wear studies published in 1975 and 1976 [10], [11], [12], [13], there has been considerable interest in the wear performance of RBCs shown in the dental literature. These early (pre-1980) studies were simplistic in nature and focused on the ability of commercial RBC materials to take a smooth polish (when measured by an electronic roughness gauge) and the relative loss in smoothness due to tooth-brushing [10]. The first laboratory abrasive wear resistance studies focused on sliding samples across silicon carbide abrasive paper [11], [12], [13] in an effort to rank commercial and experimental RBCs with dental amalgam by wear facet depth [11]. These measured abrasion rates, determined with the original pin-on plate wear test apparatus [12], were compared with tensile strength and hardness values, but with the findings indicating no correlation or relationship. Such abrasion rates, and neither hardness nor coefficient of friction showed any correlation with the wear of tooth tissue or the dental materials examined [13]. A critically assessment of the early laboratory data showed a focus on finding a correlation between simplistic laboratory abrasive wear resistance studies and established materials science laboratory techniques, but with no actual correlation identified.

The final approach when assessing what constitutes evidence in the dental literature is to look at studies that attempt to simulate the clinical situation, namely replicating the masticatory cycle in the mouth in the form of an oral wear simulator. The abrasive wear resistance studies determined with the original pin-on-plate wear test apparatus [12] would form the basis for such an explosion of oral wear simulators that less than a decade later Roulet would state that ‘there are almost as many wear testing devices as there are scientists who are interested in wear’ [2]. The most positive aspect of oral wear simulators would be for the initial screening of potential RBC restoratives prior to placement on the market, thereby eliminating the need for expensive clinical studies on substandard materials. However, we know that the wear depth measurements achieved from the wear facets produced by a variety of oral wear simulators ‘are best characterised as inconclusive’ [1].

“The First Three Questions” was the title of a guest editorial in the Australian Dental Journal in 1996 authored by Brian W Darvell [14]. The author suggested that during discussions of proposed projects (regardless of the topic), three questions emerge that can be employed successfully in any discipline, for any project, regardless of the motivation [14]. The three questions are:

What do you really want to know?

What do you want to measure? and,

What are you going to measure? [14].

In line with “The First Three Questions”, what are the answers available to us based on the evidence available in the clinical data, laboratory and oral wear simulator data reported in the dental literature as appraised above.

What do you really want to know? The restorative material loss that occurs in wear service?;

What do you want to measure? Material loss in wear service as a function of depth or volume?; and

What are you going to measure? Essentially nothing valuable as the evidence available highlights major issues with the clinical data reproducibility, laboratory data correlations and oral wear simulator data. As a result, the data acquisition variables in the x- and y-planes need to be examined to consider how they can influence the accuracy and precision of the laboratory wear measurements.

In an IADR symposium presented in Baltimore in 2005 and entitled ‘Intra-oral Restorative Materials: Wear: Rethinking the Current Approaches, the symposium chair Jack Ferracane stated that ‘the assessment of wear in clinical studies, the prediction of wear for new materials based on in vitro test methods, and the refinement of methods for quantitating wear remain as important concerns for dental researchers’ [15]. The motivation came from a manuscript by Söderholm and Richards [16] published in 1998 entitled the ‘Wear resistance of composites: a solved problem? They said that ‘based on some clinical data, we can conclude that under some conditions, occlusal wear of posterior composites remains a clinical problem, although not as bad as it was 10 years ago’ [16]. Between 1995 and 2004, from a PubMed literature search for ‘dental composite’, 5970 articles were published in the dental literature [15]. Then employing ‘dental composite (wear OR abrasion)’ for the same period found 551 publications, some 9% of the total, and 53% of all dental wear papers written since 1970 [15]. A critical assessment of these 551 publications prompted Ferracane [15] to conclude that ‘there are many questions to address when performing in vitro wear evaluations, and to date, little has actually been accomplished to standardize test methods or data reporting’. The four critical questions Ferracane [15] suggested needed addressing are considered in turn.

Wear is a tribological process resulting in the loss of material due to the interaction of opposing surfaces [17]. From a tribology perspective, there are four fundamental wear mechanisms that can exist, namely adhesion, abrasion, fatigue or corrosion [17]. Adhesive wear is that due to localised bonding between contacting solid surfaces and leading to material transfer between those two surfaces or loss from either surface [18]. Abrasive wear is wear due to hard particles or hard protuberances forced against and along a solid surface [18] and involves microscale cutting and ploughing processes. Fatigue wear of a solid surface is caused by fracture arising from material fatigue [18]. It is observed with rolling contacts rather than the sliding of surfaces and caused by asperities with very repeated high local stress, with or without lubrication. Corrosive wear is occurs where chemical or electrochemical reaction with the environment is significant [18]. This may result from the interaction with chemicals which have a softening effect on the surface so that the surface is worn by an opposing surface. However, within dentistry, terms are used to describe the clinical manifestations of wear rather than the fundamental wear mechanisms that are operative [19]. Focus is thus on surface loss at sites of occlusal contact (attrition), surface loss at non-contacting sites (abrasion) and surface loss attributed to chemical effects (erosion) [19]. Two-body abrasion is where the cutting asperities are fixed on one or both surfaces [18] and three-body abrasion is where the abrasive particles are essentially loose particles [18]. In the masticatory cycle three main events occur: the preparatory phase where the food bolus is not in contact with the teeth during mastication, the crushing phase where a three-body interaction occurs between the teeth and food bolus until tooth-tooth contact occurs [8], and the gliding phase which occurs on tooth-tooth contact and results in the masticatory force being concentrated in that area of occlusal contact and where both two- and three-body wear processes are operative [8]. In short, no single wear simulator can exactly mimic two- and three-body abrasion and three-body attrition, the clinical manifestations of wear in service [18].

Wear theory suggests that the frictional coefficient (µ) is a function of the frictional force (F) divided by the force normal to the surface at the point of contact, in the present context the occlusal load (W) from the maxillary element [8], [20]. The coefficients of friction and wear are parameters of a tribological system and describes the state of contact of material bodies within that system. The magnitude of F is dependent upon the interaction of the contacting surfaces, given that all surfaces in practice, no matter how fine the degree of lustre, carry microscopic asperities that provide the actual contact. When surface asperities come into contact, the actual (true) contact area (At) is markedly less than the nominal contact area (An) [8], [20]. At first sight, it would seem impossible to make a quantitative statement about the magnitude of At without having information about the circumstances of contact, namely the size and shape of the apparent (nominal) area of contact, the surface roughness of the mating materials and the manner in which the surfaces are placed together. Fortunately, we can make a simple limit analysis and calculate a minimum value for At, (assuming plastic deformation). Therefore, the largest compressive stress that a contacting asperity could withstand without plastic deformation is governed by the hardness (H) of the softer material, consequently, the value of At can be given by W/H [8], [20].

Therefore, Archard’s wear equation [21] can be re-written asVL=KWHwhere V is the volume loss of material due to wear (mm3), L is the total lateral sliding distance (mm) and K is a dimensionless wear coefficient. Wear is here proportional to W (the occlusal load) and L (the distance travelled) but is not dependent on the apparent surface area. Note that while μ does not appear in Eq. (1), it is implicit in the value of K itself. Wear volume was considered to be a measure of the ‘work done’ due to the interaction of opposing surfaces [8], such that DeLong recommended the use of volumetric wear as the measure of choice for reporting wear data [8].

Siegward Heintze in his publication, ‘How to quantify and validate wear simulation devices and methods’ [22] focused critically on the wear test set-up required to ‘fulfil the prerequisites’ to simulate wear operative in the oral environment. The author detailed the ‘Quantification of a wear simulating device’ in terms of ‘force, force profile, distance, contact time, sliding movement and clearance’. Additionally, the ‘Wear influencing factors’, namely, the ‘surface of the specimens, flat specimen versus standardised filling, number of specimens, storage protocol prior to testing, loading force, size and shape of the stylus, stylus material, sliding, descent and lifting speed, thermocycling, number of cycles and abrasive medium’ [22] were also discussed critically.

Paul Lambrechts and co-workers [23] in their article ‘How to simulate wear? Overview of existing methods’ reported on the number of two-body wear simulators designed and employed to replicate clinical wear mechanisms. The list was extensive and limited to the ‘capsule, compule concept’, two-body abrasion single-pass sliding, two-body wear rotating countercycle, Taber abraser, two-body machine sliding wear, pin-on-disc tribometer, abrasive disc, oscillatory wear test, modified polisher (two-body) and fretting test (oscillating friction and wear test rig) [23]. The interpretation of the wear data was made impossible since ‘key information on the force, frequency displacement, number of cycles, lubricant, hardness, Poisson ratio, and elastic modulus of the counterbody, machine running-in period, force of friction, force-displacement loop with coefficient of friction and displaced energy were often missing or not considered’ [23].

Three-body wear simulators can reproduce more closely the oral environment by including environmental factors but the degree of success in replicating clinical wear is questionable. In short, no single wear simulator can replicate the clinical manifestations of wear (abrasion and attrition) occurring in the oral environment. Innovations were studied extensively on three-body wear simulators including the Academisch Centrum for Tandheelkunde Amsterdam (ACTA) wear machine, Oregon Health Science University (OHSU) oral wear simulator, University of Alabama wear simulator – a four station Leinfelder-type three-body wear device, Zurich computer-controlled masticator, BIOMAT wear simulator, Minnesota MTS wear simulator and the Willytec Munich and Muc3 [23]. Detailed information regarding the components (stylus, medium), set-up (movement) and protocol (force, loading frequency, number of cycles) for each of the three-body oral wear simulators listed above along with the complicating factors for laboratory simulation were critically discussed by Lambrechts and co-workers, who concluded solemnly that ‘in vitro models cannot replicate the oral environment with all its biological variations. Extrapolation to the oral environment is impossible to calculate. Only trends and indications as to the true extent of wear can be obtained’ [23].

In an attempt to validate wear simulation methods, Heintze and co-workers reported on the results of a round robin test using five [24] and six [25] three-body wear simulators but the results ‘varied tremendously’ [24], [25] between individual test centres. Wear-influencing factors were critically evaluated and discussed as well as concerns around the reproducibility of the test results from these devices. Comparisons of in vivo clinical wear data [4], [5], [6] with the laboratory findings from a round robin test, using Spearman correlation, the OHSU abrasion wear method gave the best result (r = 0.86, p = 0.001) [26], with the OHSU attrition wear method second best (r = 0.66, p = 0.02) [26]. The Alabama localised wear method had a negative correlation (r = −0.14, p = 0.64) [26] with all other methods (ACTA, Ivoclar method – both vertical and volumetric, Zurich and Munich) producing correlation coefficients from 0.29 to 0.44 [26]. However, they suggested that the results should be ‘interpreted with care and reservation’ [24], [25]. The correlations were further weakened due to the ‘semi-quantitative methods for assessing the wear in vivo on pooled data from clinical trials’ and the problems with using mean maximum wear depth data [8]. Critically, the reproducibility of the wear depth measurements for the OHSU abrasive wear (20 N) and OHSU attrition wear (70–90 N) methods between different test centres was limited, with differences of 33–56% and 31–78%, respectively [27], [28], [29] for the wear depth measurements for three RBC materials [22]. These figures suggest serious difficulties with using mean wear depth measurements, emphasizing the need for ‘care and reservation’.

Compared with the direct USPHS test [4], [5] or the indirect Leinfelder et al. [6] methods which have been employed extensively in clinical studies to provide in vivo data, three-dimensional (3D) scanning is quantitative and accurate, although the disadvantages include the costs of the specialised hardware and software [8]. High accuracy of measurement, i.e. how well does the measured value represent the ‘true’ value, and high precision, effectively the repeatability of the measuring system [8], are necessities for any wear measurement. The potential of computerised 3D mapping techniques, such as the Michigan computer graphic technique [30] and 3D profilometry techniques [31], [32], were championed in 2005 by Delong [8] as revolutionary methods that could produce consistent in vivo wear measurements for dental restoratives, stating that ‘no matter what system is used to measure wear, it is mandatory that the accuracy and precision of the system is known’.