The sample size of the present study (n = 16) was determined based on the commonly used range for similar in vitro studies in the field of prosthodontic materials science (Yoshida, 2020; Mirt et al., 2024), combined with the laboratory’s conditions and experimental feasibility. To assess the statistical power of the sample size, a post-hoc power analysis (using G*Power 3.1 software) was performed. The analysis is based on the observed effect sizes derived from this study (specifically, Cohen’s *f* = 10.21 for the material principal effect, which is well above the conventional large-effect threshold). The probability of error in the setting α was 0.05, the total sample size was N = 256. The analysis showed that the statistical power (1 – β) of this study was >0.999 for the detection of the main effect of the material. This indicates that the current sample size provides extremely adequate testing power.

Among the four restorative materials, the titanium and ceramic blocks were custom-made to the dimensions of 8 × 8 × 5 mm, while the composite resin (3M Z350) and glass ionomer cement (Fuji IX) were prepared using polytetrafluoroethylene (PTFE) molds (8 × 8 × 5 mm, with five cavities). For the resin composite, an incremental filling technique was employed, in which each layer was ≤2 mm thick, and light-cured using a dental light-curing unit (Yihuojia, Liechtenstein) at an intensity of 1,200 mW/cm², according to the manufacturer’s instructions. A glass slide was used to flatten the surface before final curing. The glass ionomer cement was mixed according to the manufacturer’s guidelines, and allowed to self-cure in the molds at room temperature. For each sample, four rectangular sides (8 mm × 5 mm) were defined as experimental surfaces. Specimens with visible bubbles under magnification were discarded. Ultimately, 16 valid specimens were obtained for each material, yielding a total of 64 experimental surfaces (four per specimen).

The specimens for each material were randomly allocated into four treatment groups. Four specimens were assigned to each treatment group per material, and with each specimen contributing four experimental surfaces, this resulted in 16 surfaces per group per material. The surface of each specimen was marked with a pen. For each group, the following surface treatments were applied until the markings were completely removed under visual inspection.

Group A (ultrasonic): the treatment was performed using an ultrasonic scaler (Woodpecker UDS-J, China) with a sweeping motion for 30 s;

The experimental design is schematically illustrated in Figure 1.

For each group, the Ra value was measured after surface treatment using a profilometer (Siderixin TR230, China) in contact mode. The measurement parameters were set, as follows: probe tip diameter, 2 μm; pressure, 0.75 mN; feed rate, 1 mm/s. For each sample, three measurements were taken on the central surface area, and the average value was recorded as the final Ra.

The central thickness of each specimen was measured using an electronic micrometer (Mitutoyo, Japan) before and after treatment. Three measurements were taken and averaged. The volume loss (μm) was calculated as the difference between the pre- and post-treatment thickness values.

Following the surface treatment, the samples were ultrasonically cleaned, dried, and sputter-coated with gold. Then, the surface microtopography (TESCAN MIRA3 LMH, Czech Republic) was examined by scanning electron microscopy (SEM) at an acceleration voltage of 15 kV and a working distance of 15 mm.

Statistical analysis was performed using the SPSS 25.0 software (SPSS Corporation, Chicago, IL, United States). Quantitative data were expressed in mean ± standard deviation (x ± s). In order to comprehensively evaluate the impact of material type and treatment method on volume loss and Ra, the present study employed a multi-level statistical strategy. Initially, two-way ANOVA was conducted to test the main effects and interactions between materials and treatment methods. Then, the effect size (partial η²) was calculated for significant effects. If the interactions were significant, a simple effect analysis was further performed (i.e., one-way ANOVA was conducted for each material separately). Next, post-hoc multiple comparisons were performed using the Tukey HSD test when significant differences were observed, in order to clarify the specific differences among treatment groups. The post-hoc comparison results were labeled with superscript letters in the relevant tables. In addition, paired-sample t-test was used to compare the changes in Ra before and after treatment within each group.

We acknowledge that measurements obtained from multiple surfaces of the same specimen are not completely independent. In this study, the main statistical analyses (e.g., two-way ANOVA and subsequent comparisons) were designed to examine differences between different ‘material-treatment’ combinations. Therefore, we use the average of the measurements of the 16 individual specimens under each combination as the basic unit of analysis. This strategy effectively handles potential within-sample correlations at the between-group-comparison level. In addition, the effect size observed in this study is extremely large (e.g., the partial η² > 0.99 for the main effect of the material). The post-hoc power analysis shows that the statistical power was close to 1.000, indicating that the main conclusions are extremely robust and not subject to the substantive effects of potential minor correlations at the measurement level.

In order to comprehensively evaluate the influence of material type and surface treatment on the experimental results, two-way analysis of variance was initially carried out. For volume loss, the main effects of material type (F = 8369.234, p < 0.001, ηp² = 0.991) and surface treatment method (F = 78.472, p < 0.001) were extremely significant. More importantly, there was a significant interaction between the two (F = 10.097, p < 0.001), indicating that the impact of different surface treatments on volume loss varied with the material type (Table 1). Furthermore, the model had an extremely high degree of fit (R ² > 0.99). For post-treatment Ra, the main effects of material type (F = 15637.941, p < 0.001, ηp² = 0.995) and treatment method (F = 21.942, p < 0.001) were both highly significant. Furthermore, a significant interaction effect between these two factors was observed (F = 5.931, p < 0.001) (Table 2). This confirms that the improvement of surface treatment on roughness is highly dependent on the specific restorative material. The simple effect analysis revealed that the sandblasting treatment was generally better than the mechanical treatment in volume loss, and that the effect of sandblasting on roughness differed according to the material. That is, some materials were sensitive to the sandblasting treatment, while some materials were tolerant to the sandblasting treatment. The marginal means plots in Figures 2, 3 visually illustrate the above interaction. The curves that represented the different treatments were clearly non-parallel and intersecting, clearly indicating that there is no “one-size-fits-all” treatment that is optimal for all materials.

On the basis of confirming the presence of significant interactions, the specific effects of various treatments on the Ra of each material were further analyzed by paired sample t-test (Table 3).

The paired t-test results (Table 3) indicated that the Ra was significantly altered (p < 0.05) following treatment in all groups, except for the titanium specimens in Group A (p = 0.394). Furthermore, the ultrasonic treatment (group A) significantly increased the Ra value of glass ionomer (from 3.758 ± 0.104 to 3.956 ± 0.152, p < 0.001). In contrast, both sandblasting methods (Groups B and C) significantly reduced the Ra across all materials.

One-way analysis of variance was conducted to further explore the differences in effects across treatment methods within specific materials.

The SEM results revealed distinct microstructural alterations that resulted from the different surface treatments (Figure 6). The untreated surfaces (Group E) had an inherent material morphology, serving as the baseline for comparison. Surfaces subjected to ultrasonic treatment (Group A) had minor, uneven deposits, suggesting incomplete cleaning. In contrast, both sandblasting groups (Groups B and C) had significantly improved surface uniformity and flatness, with original irregularities effectively removed to form a homogeneously abraded surface morphology. This observation is consistent with the quantitative roughness data, confirming the efficacy of sandblasting. Although the polished group (Group D) achieved general smoothness, distinct directional scratches were evident on the glass ionomer and resin surfaces, indicating that mechanical polishing can induce new micro-defects in comparatively softer materials. In summary, sandblasting produced the most favorable surface topography, while ultrasonic and polishing methods were associated with residual debris and scratching, respectively.

The present study confirmed the significant interaction between the type of restorative material and the surface treatment method (p < 0.001), indicating that there is no universal optimal treatment scheme. The difference in the response of materials to surface treatment stems from the matching relationship between its inherent properties and the treatment mechanism. Sandblasting treatment can generally reduce the Ra value (Petersilka, 2000). However, the specific effect varies with the material: erythritol particles with moderate hardness and regular morphology can achieve the uniform micro-cutting of titanium materials, and obtain the best smoothing effect (ΔRa = −0.295 ± 0.181) (Atagün and Kalyoncuoğlu, 2025). Furthermore, sodium bicarbonate particles with higher hardness and irregular shape can level the surface of softer glass ionomer more effectively (ΔRa = −0.199 ± 0.045) (Amin et al., 2021). The negative effect of ultrasonic treatment on glass ionomer cement (ΔRa = +0.197 ± 0.137) revealed that its porous matrix was prone to packing shedding or microcrack propagation under high-frequency oscillation (Mensi et al., 2024; Mohammadi et al., 2023). The volume loss data further confirmed the role of the material’s intrinsic properties. Ceramic materials have extremely high hardness and dense microstructures, thereby exhibiting minimal volume loss (14–17 μm) and optimal wear resistance in all treatments (Martins et al., 2023). In contrast, due to its low hardness and multiphase structure, glass ionomer cement is prone to plastic deformation and interfacial failure under mechanical stress, resulting in the largest volume loss (139–153 μm) (Worthington et al., 2021). Although abrasive blasting removes the material, this generally leads to less volume loss, when compared to ultrasonic or polishing. This suggests that sandblasting may preferentially remove the prefabricated defect layer of the surface, and form a denser and more durable surface transition layer through uniform micro-deformation or micro-forging (Sekino et al., 2020). In contrast, ultrasound can cause subsurface damage, while the directional scratches left by polishing on softer materials can become new stress concentration points and wear initiators (Okumuş et al., 2023).

The present study correlated the in vitro experimental data with the key evaluation indicators of long-term clinical performance of restorations, providing an empirical basis for clinical decision-making. There is a broad consensus that increased Ra promotes the attachment and maturation of dental plaque biofilms, thereby elevating the risk of secondary caries, marginal discoloration, and peri-implantitis, and may also affect the long-term retention of dental restorations (Alnazzawi et al., 2025; Al Hatem et al., 2025). Therefore, achieving and maintaining a low Ra value is an important goal for clinical maintenance. The results of the present study show that sandblasting treatment has clear advantages in reducing the surface Ra value of different materials. For example, the surface Ra value of ceramics, titanium, and composite resins after sandblasting can be reduced to less than 0.3 μm. In contrast, the surface Ra value of ultrasonically treated glass ionomer cement (approximately 3.96 μm) is much higher than this threshold, suggesting that the ultrasonic maintenance of the material may be detrimental to long-term oral health, and potentially shorten its lifespan (Eram et al., 2024).

In terms of material preservation, sandblasting technology shows a better balance. Although there is a controllable loss of micron-sized material, it is significantly lower than that of polishing and ultrasonic treatments, which can introduce microscopic scratches or cause subsurface damage (Harman and Murchie, 2025). For materials with excellent intrinsic properties such as ceramics, the extremely low volume loss (14–17 μm) combined with the ideal smooth surface after sandblasting support its application in areas where high durability is required, such as the posterior tooth area (Al-Akhali et al., 2025; Schubert et al., 2019). For glass ionomer cement, despite its relatively large volume loss, sodium bicarbonate blasting, rather than ultrasonic treatment, is a more reasonable clinical compromise between “controlled material removal” and “avoiding severe surface roughening”, which may help extend its service time in low-stress zones (Sekino et al., 2020).

In summary, clinicians may be able to achieve more predictable maintenance outcomes by considering the material properties when selecting a restoration surface treatment method, and prioritizing sandblasting techniques that better match the material (Saraç Atagün et al., 2025; Janiszewska-Olszowska et al., 2020).