Three-Dimensional Surface Texture Characterization of In Situ Simulated Erosive Tooth Wear

Abstract

This in situ erosive tooth wear (ETW) study tested enamel 3-dimensional (3D) surface texture outcomes for the detection and differentiation of ETW lesions simulated in clinically relevant conditions. Twenty participants enrolled in this 3-arm crossover intraoral ETW simulation and wore their own partial denture for 14 d holding 2 human enamel specimens (per arm). In each arm, participants were assigned to 1 of 3 different dental erosion protocols: severe (lemon juice/pH 2.5), moderate (grapefruit juice/pH 3.5), and no erosion (bottled drinking water, control). Enamel specimens were evaluated by white-light scanning confocal profilometry for 3D surface texture and surface loss (ETW model validation). Individual point clouds were analyzed using standard dental microwear texture characterization protocols for surface roughness and anisotropy. Fractal complexity (Asfc), texture aspect ratio (Str), and arithmetical mean height (Sa) values were generated at baseline, 7 d, and 14 d. Data were analyzed by analysis of variance models suitable for the crossover design with repeated measurements, and correlation coefficients were used to examine the relationship between outcomes. Asfc and Sa differentiated ETW severity (no erosion < moderate < severe, P < 0.001) at days 7 and 14. Asfc and Sa were lower at baseline compared to days 7 and 14 (P < 0.001) for moderate and severe challenges. Asfc increased from day 7 to 14 (P = 0.042) for the severe challenge. For Str, ETW severity did not have a significant effect overall (P = 0.15). Asfc and Sa were highly positively correlated (r = 0.89, P < 0.001), while Asfc and Sa were not correlated overall with Str (r < 0.1, P ≥ 0.25). Enamel surface loss increased with ETW severity (no erosion < moderate < severe, P < 0.001) at days 7 and 14, validating the ETW simulation model. Complexity (Asfc) and roughness (Sa) outcomes were able to detect and differentiate ETW levels, with Asfc being able to monitor the progression of severe lesions. No clear characterization of ETW lesions could be provided by the anisotropy (Str) parameter.

Introduction

Erosive tooth wear (ETW) is a dental condition clinically characterized by tooth structure loss as the result of dental erosion and/or dental erosion-abrasion wear processes (Schlueter et al. 2020). The prevalence of ETW is high worldwide (Jaeggi and Lussi 2014). Specifically, in the United States, it affects approximately 46% of teenagers (McGuire et al. 2009) and 80% of adults (Okunseri et al. 2015), as reported by the National Health and Nutrition Examination Survey. Despite these alarming numbers, dental professionals are largely unaware of this condition (Goldfarb et al. 2020), and specific diagnostic rules and evidence-based management guidelines are needed but yet to be established. Currently, the clinical assessment and monitoring of ETW is performed by visual examination using subjective indices (Ganss and Lussi 2014). This traditional approach limits ETW lesion detection to mostly advanced stages, wherein considerable destruction of the tooth has already occurred, resulting in pain and irreversible changes in dental form, function, and esthetics. In these circumstances, the required restorative treatments are complex and costly (Peutzfeldt et al. 2014). Thus, there is a clear need for objective methods to early detect and differentiate the severity of ETW lesions, allowing the implementation of personalized and evidence-based management plans focused specifically on preventive measures.

In this study, we hypothesized that objective 3-dimensional (3D) enamel surface texture parameters of roughness, complexity, and anisotropy can detect and differentiate ETW lesions of distinct severity levels. They have been successfully used in paleontological and neontological dental research to identify and differentiate the dietary habits of human ancestors and nonhuman primates (Ungar 2019) and other mammals (Ungar 2015). Moreover, our previous in vitro study has shown the potential of these surface texture parameters for the early identification and differentiation of simulated ETW lesion types (dental erosion, abrasion, and erosion-abrasion) (Hara et al. 2016). However, it is unknown whether these promising findings can be translated into clinical application. Therefore, herein we test the study hypothesis using an established intraoral ETW experimental model (Hara et al. 2014). The ability of these 3D parameters of enamel surface complexity and anisotropy to identify ETW lesions early and differentiate their severity level can lead to the future development of innovative technologies and tools for the objective clinical diagnosis and monitoring of ETW. This study aimed specifically to test whether these parameters can be used to 1) detect and differentiate simulated ETW lesions of distinct severities and 2) monitor the development of these lesions longitudinally.

Materials and Methods

Ethics and Inclusion/Exclusion Criteria

The study protocol was reviewed and approved by the local institutional review board (#1810761388). All subjects signed a written informed consent form before screening. Twenty adult volunteers (15 women, 5 men), 41 to 84 y of age, met the following inclusion criteria: ability to wear a mandibular partial denture for 24 h per day, except for study procedures; no evidence of severe ETW, active caries, or periodontal disease; stimulated and unstimulated salivary flow rates of ≥0.8 and ≥0.2 mL/min, respectively; and the ability to comply with the study procedures. Exclusion criteria were any medical condition potentially interfering with the subject’s health, use of medication that could lead to a reaction with the test materials, use of denture cleansing solutions, and unwillingness to comply with the study procedures.

Study Design

This study consisted of a 3-arm crossover intraoral ETW simulation, based on a previously developed model (Hara et al. 2014). The Figure provides the study flowchart. Twenty participants (n = 20) were randomly assigned into the 3 different ETW simulation protocols (1 for each arm): severe (s-ETW), specimen exposure to lemon juice; moderate (m-ETW), specimen exposure to grapefruit juice; and no erosion control (non-ETW), specimen exposure to bottled water. With this design, every subject simulated the 3 ETW severity levels. During each arm, subjects had 2 specimens (replicas) placed into their partial dentures (Fig.) and performed dental erosion and toothbrushing abrasion simulation procedures for 14 d. Impressions were taken from the surface of the specimens at baseline and days 7 and 14. These were then evaluated for enamel surface texture parameters (Hara et al. 2016) (Table 1). Optical profilometry was used to evaluate enamel surface loss and verify the validity of the ETW model used.

Figure.

Study design and in situ erosive tooth wear (ETW) model adopted in the study. (A) Flowchart of the crossover experimental design. (B) Partial denture loaded with the dental specimens.

Table 1.

Studied Surface Texture Parameters for Erosive Tooth Wear Detection and Differentiation.

Parameter	Abbreviation	Definition	Interpretation (meaning of higher values)
Area-scale fractal complexity	Asfc	Measure of the steepest part of a slope of relative area over scale of observation	More complex surface, with more pits and scratches of different sizes overlaying each other
Three-dimensional average surface roughness	Sa	Measure of the difference in height of each point compared with the arithmetical mean for the surface	Rougher surface at point cloud scale
Anisotropy of relief	Str	Measure of the uniformity of surface texture or anisotropy of the enamel surface	Surface texture does not depend on direction

Specimen Preparation

Details of specimen preparation can be found elsewhere (Hara et al. 2014). Briefly, 120 human enamel slabs (3 mm in diameter and 2 mm thick) were cut from previously extracted permanent molars, flattened/polished, selected based on enamel surface quality and microhardness (336–444 KgF/mm²), and further processed to tightly fit into stainless steel holders. After completion, specimens presented with a 3-mm round flat and polished surface encompassing the enamel surface surrounded by the metal references (for the optical profilometry). Before insertion into the subjects’ dentures, specimens were sterilized by ethylene oxide gas.

Clinical Phase

Before the experimental period, subjects received oral soft and hard tissue exams and dental prophylaxis, and they had their dentures cleaned. The 20 subjects had 2 enamel specimens cemented into 2 box-shape hollows made on the vestibular surfaces of adjacent molars of their partial denture. Impressions of the specimens were taken (baseline) using a polyvinylsiloxane material (President Jet Regular Body; Coltene Whaledent), dispensed into custom plastic trays, followed by labeling, disinfection, and storage. Previous to the impression, specimens were gently brushed with a manual toothbrush (Oral-B 40; Procter & Gamble) for 10 s under tap water. During the 14-d experimental period, subjects wore the partial dentures for 24 h/d, except during cleaning and experimental procedures. The dental erosion simulation was performed 4 times a day (after breakfast, lunch, and dinner, and before bedtime). The severe ETW (s-ETW) group used pure lemon juice (Organic Pure Lemon Juice; Santa Cruz Natural), the moderate ETW (m-ETW) group used pure grapefruit juice (100% White Grapefruit Juice; Ocean Spray Cranberries), and the non-ETW (negative control) group used bottled water (Drinking Water; Kroger Co.). All test solutions were provided to subjects weekly in sufficient amounts and kept in their refrigerators, being removed only as needed for the erosive challenge. Toothbrushing was performed on specimens twice daily with a standard toothbrush (Oral-B 40) and toothpaste (1,100 ppm fluoride, as NaF; Crest Cavity Protection; Procter & Gamble) after breakfast and before bedtime (daily procedures in Table 2).

Table 2.

Daily Experimental Procedures.

Procedure	Breakfast	Lunch	Dinner	Bedtime
1. Brush partial denture (PD) with tap water	×			×
2. Place PD in 30 mL of test solution (5 min)	×	×	×	×
3. Rinse PD with tap water (~10 s)	×	×	×	×
4. Place PD in mouth, brush 20 s to create slurry	×			×
5. Swish slurry in mouth for ~5 s	×			×
6. Spit, remove PD, and brush specimens (~10 s)	×			×
7. Rinse PD with tap water (~10 s)	×			×
8. Brush natural teeth, swish with tap water	×			×
9. Place PD back in mouth	×	×	×	×
10. Wait 30 min before eating or drinking	×	×	×	×

Training on the study procedures was provided to subjects before starting each arm of the study until they were able to perform them proficiently without supervision by the clinical study personnel. After 7 and 14 d, subjects returned for follow-up visits and had their ability to perform the study procedures reviewed. Impressions of the specimens were taken after gently brushing them with water, as described for baseline. Subjects were instructed not to perform the erosion simulation and specimen brushing procedure after breakfast on days 7 and 14, and thus the last erosive and abrasive challenges were performed the night before the impressions were taken. At the 14-d visit, the specimens were removed and the hollows in the denture tooth filled with temporary restorative material (Coe-Soft; GC America, Inc.). Oral soft tissue examination was performed and compliance to the procedures checked at each visit. All procedures described were repeated for each arm of the study. A washout period of 1 wk was implemented between arms. At the end of the study, the hollows in the partial denture were repaired (Tetric EvoCeram Bulk Fill; Ivoclar Vivadent AG).

Surface Texture Analysis

Enamel specimen impressions were analyzed by white-light scanning confocal profilometry (Neox; Sensofar USA LCC). As previously described (Hara et al. 2016), the planimetric work envelope for each sample was 242 × 181 µm² with a lateral point spacing of 0.17 μm in both x and y directions and a vertical resolution <2 nm measured in 3 locations. Measurements were performed on the central area of the specimen surface, and specimens were evaluated in blind and random conditions by a trained examiner. Point clouds for each surface were examined by microwear texture analysis. Area-scale fractal complexity (Asfc) (Brown et al. 2018) and ISO 25178 standards for texture aspect ratio (Str) and arithmetical mean height (Sa) were calculated (Table 1) to characterize scale-sensitive complexity, anisotropy, and roughness of each surface, respectively. Analyses were conducted using MountainsMap 8 (Digital Surf). These variables have been shown to consistently reveal aspects of surface texture of value for distinguishing dental wear types (Ungar et al. 2003; Schulz et al. 2013; Hara et al. 2016).

Optical Profilometry

Enamel surface loss was analyzed on the specimens’ impressions by a white-light optical profilometer (Proscan 2000; Scantron) with an accuracy of 0.1%, a precision (SD) of ±0.06 mm, and a detection limit of <0.3 mm (Steiner-Oliveira et al. 2010). The step size was set at 0.02 mm and the number of steps at 200 in both x and y directions. A surface area of 4 × 4 mm² was scanned, comprising the metal reference and enamel surface. Using dedicated software (ProScan 2000 V2.1.1.10; Scantron), the depth of the enamel-treated area was calculated in relation to the reference metal surfaces. A 3-point height tool was used, which allowed the selection of a 1 × 1-mm² area at the center of the enamel and two 1 × 0.5-mm² areas on the adjacent metal surfaces. Enamel surface loss was calculated for each specimen by subtracting the depth at baseline from the depth at each of the other days.

Statistical Analysis

The effects of ETW severity (s-ETW, m-ETW, non-ETW) and time (0, 7, 14 d) on surface texture parameters (Asfc, Sa, and Str) and enamel surface loss (µm) were analyzed using an analysis of variance (ANOVA) model suitable for a 3-arm, 3-ETW severity crossover design with repeated days within specimens. The models for both enamel surface texture and enamel surface loss accounted for correlation within each subject due to the crossover design, accounted for correlation between specimens within the same ETW severity for each subject with different correlations allowed for each ETW severity, and allowed different variances and correlations between days within a specimen for each severity. Correlation coefficients and scatterplots were used to examine the relationships between the measurements. Analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) at a 5% significance level.

Results

In order to satisfy the requirements of normal distribution and homoscedasticity, a natural logarithm transformation was applied to the Sa and Str data. Ranks of the data were used for the enamel surface loss and Asfc analyses. Means (standard errors) and 95% confidence intervals for the means are summarized in Tables 3 and 4.

Table 3.

Means (SEs) and 95% CIs for the Means of the Tested Outcomes.

		Asfc a		Sa b		Str b
ETW	Day	Mean (SE)c	CI	Mean (SE)c	CI	Mean (SE)c	CI
No	0	0.83 (0.07)A,a	0.67–0.98	39.6 (3.1)A,a	32.7–46.5	0.57 (0.03)A,a	0.51–0.63
	7	1.48 (0.42)A,a	0.64–2.33	67.0 (13.3)A,a	40.1–93.9	0.64 (0.03)AB,a	0.59–0.70
	14	0.95 (0.08)A,a	0.79–1.11	57.1 (5.1)A,a	46.8–67.5	0.62 (0.03)A,a	0.56–0.68
Moderate	0	0.73 (0.09)A,a	0.54–0.91	33.0 (2.5)A,a	27.6–38.4	0.58 (0.03)A,a	0.53–0.64
	7	4.74 (1.49)B,b	1.72–7.75	157.5 (19.7)B,b	117.7–197.4	0.58 (0.03)A,a	0.53–0.64
	14	6.30 (2.18)B,b	1.90–10.70	186.0 (23.0)B,b	139.6–232.5	0.59 (0.03)A,a	0.53–0.65
Severe	0	0.76 (0.05)A,a	0.65–0.87	40.7 (4.5)A,a	31.4–49.9	0.58 (0.03)A,a	0.51–0.64
	7	34.23 (3.32)C,b	27.50–40.96	564.6 (39.9)C,b	483.8–645.5	0.68 (0.03)B,b	0.62–0.74
	14	38.31 (4.00)C,c	30.22–46.41	601.9 (37.9)C,b	525.2–678.7	0.60 (0.03)A,ab	0.54–0.66

ETW, erosive tooth wear.

^aStatistical analysis done on ranks.

^bStatistical analysis done on natural logarithm-transformed data.

^cETW severity comparison within days: means followed by different capital letters differ significantly (P < 0.05); day comparison within ETW severity: means followed by different lowercase letters differ significantly (P < 0.05). Each parameter was analyzed independently.

Table 4.

Means (SEs) and 95% CIs for the Means of Surface Loss.^a

ETW	Days	Mean (SE)b	CIc	Minimumc	Maximum
No	7	0.85 (0.40)A,a	0.03, 1.67	−1.63	11.14
	14	0.62 (0.40)A,a	−0.25, 1.49	−1.67	7.96
Moderate	7	5.75 (1.68)A,b	2.31, 9.19	−8.64	29.17
	14	18.80 (6.50)B,b	5.66, 31.95	−1.01	174.93
Severe	7	72.11 (40.84)A,c	−13.33, 157.54	2.10	847.11
	14	98.78 (40.84)B,c	13.35, 184.22	7.85	893.15

ETW, erosive tooth wear.

^aStatistical analysis done on ranks.

^bETW severity comparison within days: means followed by different capital letters differ significantly (P < 0.05); day comparison within ETW severity: means followed by different lowercase letters differ significantly (P < 0.05). Each parameter was analyzed independently.

^cNegative values represent enamel surface deposition or convex surface.

Area-Scale Fractal Complexity (Asfc)

There was a significant interaction between ETW severity and time for Asfc (P < 0.001). Within days 7 (D7) and 14 (D14), Asfc increased with ETW severity (non-ETW < m-ETW < s-ETW; P < 0.001). For baseline (D0), there were no significant differences among ETW severities (P = 0.21). When comparing days within each ETW severity, no differences were shown for non-ETW (D0 = D7 = D14; P = 0.17), while differences were observed for m-ETW (D0 < D7 = D14, P < 0.001) and s-ETW (D0 < D7 < D14; P < 0.001 for D0 versus D7 and D14; P = 0.042 for D7 versus D14, Appendix Fig.).

Three-Dimensional Average Surface Roughness (Sa)

There was a significant interaction between ETW and time for Sa (P < 0.001). Within days 7 (D7) and 14 (D14), Sa increased with ETW severity (non-ETW < m-ETW < s-ETW; P < 0.001). For baseline (D0), there were no significant differences among ETW severities (P = 0.27). When comparing days within each ETW severity, no differences were shown for non-ETW (D0 = D7 = D14; P = 0.27), while differences were observed for m-ETW and s-ETW (D0 < D7 = D14; P < 0.001 for D0 versus D7 and D14).

Anisotropy of Relief (Str)

There was no significant interaction between ETW severity and time for Str (P = 0.15). ETW did not have a significant effect on Str overall (P = 0.15) or for D0 (P = 0.89) or D14 (P = 0.78); however, for D7, m-ETW < s-ETW (P = 0.007). Within ETW severities, there was significant time effect only at s-ETW with D0 < D7 (P = 0.006).

Enamel Surface Loss

There was a significant interaction between ETW severity and time for enamel surface loss (P < 0.001). Within days 7 (D7) and 14 (D14), enamel loss increased with ETW severity (non-ETW < m-ETW < s-ETW; P < 0.001). When comparing days within each ETW severity, no differences were shown between D7 and D14 for non-ETW (P = 0.34) and m-ETW (P = 0.91), while difference was observed for s-ETW (P < 0.001).

Correlations

Asfc and Sa were highly positively correlated (r = 0.89, P < 0.001), while Asfc and Sa were not correlated with Str (r < 0.1, P = 0.45 and P = 0.27, respectively). Enamel surface loss was significantly correlated with Asfc and Sa measurements (Appendix Table).

Discussion

The in situ model used considered the main biological (saliva, acquired pellicle, soft tissues), chemical (remineralization), and behavioral (dietary acids, toothbrushing) factors associated with ETW. The dental erosion challenge was performed extraorally for safety reasons, and the protocols using lemon and grapefruit juices were chosen based on a previous in vitro test (unpublished data), showing their ability to yield a relatively wide range of enamel surface loss, as confirmed by the surface loss data. This created a favorable scenario for the test of the 3D enamel surface micromorphology parameters. It can be inferred that the lesions simulated in this study would represent initial ETW lesions clinically, characterized by loss of the enamel surface texture with possible initial bulk loss.

Our findings support the hypothesis that 3D surface texture parameters could identify the ETW lesions and differentiate their severity levels. Overall, better performance was related to Asfc and Sa, which were able to identify and quantify the very early changes on enamel surface. These findings corroborate those from our previous in vitro study (Hara et al. 2016) and build on the previously reported assumption that the enamel surfaces of ETW lesions show wear features of different sizes overlaying each other (Scott et al. 2005), which further increase in complexity in more severe lesions. Similar findings were also reported in another in vitro study (Austin et al. 2016), where a comparable approach was used to determine the optimal area scale for 3D surface roughness (Sa) analysis, aiming to best highlight and differentiate enamel surface features.

In addition to lesion detection and differentiation, the studied parameters were also tested for monitoring ETW progression. Sa did not show a significant increase in terms of enamel surface roughness from days 7 to 14 for any of the simulated lesions (although a nonsignificant trend could be observed). On the other hand, Asfc showed significant increase of complexity from 7 to 14 d but only for the s-ETW lesions (Appendix Fig.). Although an increase in enamel surface complexity and roughness is expected as the lesion progresses, it can be speculated that it reaches a plateau where bulk surface loss starts and little to no change is observed in the surface characteristics (Ganss et al. 2014). Therefore, we were encouraged by the Asfc results. The better ability of Asfc to monitor s-ETW lesion progression compared to Sa may be related to scale-specific morphological changes that likely happened on the enamel surface of these lesions.

Asfc differs fundamentally from Sa as it uses a scale-sensitive approach (Brown et al. 2018), which considers that surface texture pattern of the ETW lesions (especially for s-ETW) varies with scale of observation. It can be speculated that the acid action resulted in enamel surface morphology changes due to the overall bulk loss at a gross scale, as well as finer change in the microstructure of the enamel surface due to the increase of porosity. A scale-sensitive approach is often used in studies of microscopic dental wear specifically because enamel itself is organized in a scalar manner (e.g., enamel tissue comprises packed rods, rods are an amalgam of crystallites, etc.), so it is assumed that wear processes differ at varying scales too. Another contributing aspect in this scenario would be toothbrushing abrasion, as it can also significantly change the enamel surface complexity at different microscopic scales, especially if it had been previously demineralized by the exposure to the erosive acids as done in this study.

Contrasting with the complexity and roughness findings, the anisotropy parameter (Str) did not evince the ability to differentiate the severity levels of the simulated ETW lesions overall, with the exception at D7, when s-ETW lesions were less anisotropic than m-ETW lesions. Similarly, the anisotropy of the s-ETW lesions was lower only at D7 compared to the baseline, with no clear trend pattern from baseline to D14. Lack of stronger response was also previously observed in our in vitro study, using a similar anisotropy (epLsar) parameter (Hara et al. 2016). The expectation was that the Str could quantify the development of structured wear patterns on the enamel surface such as microgrooves, resulting from the unidirectional abrasive forces generated by the back-and-forth movement of the toothbrushes, especially when acting on previously eroded enamel surfaces (m- and s-ETW groups) and considering that toothbrushing was performed with a medium-abrasive toothpaste. It can be argued that the in situ experimental model used was biased toward dental erosion (acid action), rather than dental abrasion (toothbrushing action). In fact, the acid impact becomes evident when considering the significant differences observed among ETW severities, in terms of enamel complexity, roughness, and surface loss. Despite these limited results, the value of the Str parameter should not be underestimated, since different patterns of ETW progression clinically may involve directional abrasive forces, such as excessive toothbrushing or other abrasive challenges (attrition). Different simulation models or clinical protocols may be able to better explore and show the performance of Str for identifying and monitoring ETW progression.

This study presented limitations, including the relatively short length of simulation at 14 d and the use of flattened and polished specimens. Both limitations reflect practical compromises adopted in order to be able to conduct such a controlled study, as normally done in ETW in situ experimental models (West et al. 2011). It is possible that they could have affected the dynamics of the ETW lesion development. An in vitro evaluation has shown that the dynamics of ETW lesion development may differ on natural surfaces, with slower progression observed at earlier stages when compared to polished enamel (Mylonas et al. 2019). Therefore, further validation of the studied 3D parameters in longitudinal ETW clinical studies is warranted. An important aspect is that the enamel surface analysis was conducted in a laboratory setting using impressions of the specimens. Although this approach can be done in a reproducible and noninvasive manner, further developments in this area should focus on the development of tools (handheld device) that could be potentially used in a dental clinical setting. Similar tools have been developed for the analysis of less specific enamel surface characteristics, such as light reflection, with promising results (Carvalho et al. 2016).

The importance of the 3D surface parameters tested for the objective identification and differentiation of early lesions cannot be overstated. In a recent online survey investigating dentist awareness of ETW lesions, we have demonstrated that dental practitioners failed to identify the presence of early ETW lesions in approximately 86% of the cases, in contrast to the 6% of failed identified cases of early caries lesions (Goldfarb et al. 2020). We have attributed this difficulty mostly to the limitations imposed by the visual clinical examination, which is subjective and relies on the ability of the dentist to identify the micromorphological changes on the enamel surface. Therefore, the tested parameters represent a fundamental step toward the development of quantitative tools that could objectively identify ETW lesions, especially at earlier stages. Considering the limitations of this study, we conclude that the 3D parameters of surface micromorphology complexity (Asfc) and roughness (Sa) provided an objective characterization of the enamel surface at subclinical levels and objectively identified and differentiated ETW lesions of different severities. In addition, Asfc could monitor the progression of the severe ETW lesions. Conversely, Str was not able to identify elements of anisotropy on the simulated ETW lesions.

Author Contributions

A.T. Hara, contributed to conception, design, and data interpretation, drafted and critically revised the manuscript; D. Elkington-Stauss, contributed to data acquisition, analysis, and interpretation, critically revised the manuscript; P.S. Ungar, contributed to design, data analysis, and interpretation, critically revised the manuscript; F. Lippert, contributed to design and data interpretation, critically revised the manuscript; G.J. Eckert, contributed to data analysis, and interpretation, critically revised the manuscript; D.T. Zero, contributed to data interpretation, critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.