Comparison of Ultrasonography, CBCT, Transgingival Probing, Colour‐Coded and Periodontal Probe Transparency With Histological Gingival Thickness: A Diagnostic Accuracy Study Revisiting Thick Versus Thin Gingiva

ABSTRACT

Aim

To assess the reliability of ultrasonographic, cone beam computed tomographic (CBCT), probe transparency and transgingival probing (TGP) methods in evaluating gingival thickness (GT), compared with the gold standard histological assessment.

Methods

Sixteen fresh cadaver heads with intact gingivae were used. The sequence for GT measurement included CBCT, ultrasonography, probe transparency, TGP and histology. Both stainless steel periodontal probe and colour‐coded probes were used for transparency. TGP involved a calibrated endodontic spreader, and histologic samples served as a comparative standard. Primary outcomes evaluated accuracy in GT measurement, while secondary outcomes assessed agreement among methods and established an optimal threshold for thin versus thick gingiva.

Results

One hundred and fifteen teeth were examined, yielding a mean GT of 1.34 mm histologically. US and CBCT underestimated GT (means of 1.25 mm and 1.13 mm, respectively), while TGP overestimated (1.51 mm). Correlations (r = 0.88–0.98) and ICC values (0.73–0.95) indicated strong inter‐method agreement. Regression models significantly estimated histological GT from US, CBCT and TGP. A new 1.18 mm cut‐off, based on histology, improved diagnostic accuracy over the traditional 1 mm threshold.

Conclusions

While histology remains the GT reference standard, US, CBCT and TGP achieved clinically acceptable accuracy. US showed the highest agreement with histology, followed by TGP and CBCT. The study supports US as the most practical non‐invasive tool, although CBCT and TGP remain viable options. Further clinical validation is recommended, acknowledging the limitations of cadaveric models in reflecting in vivo conditions.

1. Introduction

Gingival thickness (GT) is a key component of the periodontal phenotype that plays a pivotal role in the health and function of both teeth and dental implants (De Rouck et al. 2009; Jepsen et al. 2018). Moreover, the periodontal soft tissue phenotype, including GT, is crucial for the outcomes of periodontal and implant therapies, orthodontic treatments and prosthetic rehabilitations (Jepsen et al. 2018; Wang et al. 2022). Therefore, assessing this key component at various stages of treatment planning and execution is essential for optimal outcomes. In the 2017 World Workshop, a GT cut‐off point of 1 mm was established, defining thin (< 1 mm) and thick (≥ 1 mm) gingiva (Jepsen et al. 2018), while it was also emphasised that a thin gingiva possesses a greater risk of developing gingival recessions. A recent longitudinal study demonstrated that, in the presence of a band of keratinized tissue (KT) ≥ 1.5 mm, a GT of at least 1.46 mm was the main factor associated with the long‐term stability of the gingival margin over 10 years (Barootchi et al. 2022). This underscores the importance of accurate quantitative assessment of GT not only before treatments but also during and after to ensure optimal outcomes.

Several methods and tools have been developed and employed over the years to assess GT (Wang et al. 2022; Kloukos et al. 2018). Among the quantitative methods, transgingival probing (TGP), cone beam computed tomography (CBCT) (either alone or combined with 3D intraoral scans; Couso‐Queiruga et al. 2021) and ultrasonography (US) are widely used (Wang et al. 2022; Fan et al. 2023). Additionally, qualitative assessments of GT in clinical practice are performed using probe transparency methods, either with regular periodontal probes (PPB) or specialised colour‐coded probes (CCP), to classify gingiva as thin or thick (De Rouck et al. 2009; Uysal et al. 2024). A meta‐analysis by Fan et al. (2023) found no significant difference between CBCT measurements and direct assessment (mean difference: −0.07 mm), or between CBCT and TGP (mean difference: −0.09 mm). Similarly, another meta‐analysis showed no significant difference between CBCT and TGP (mean difference: 0.10 mm) (de Freitas Silva et al. 2023). However, both studies rated the certainty of evidence as low to very low, highlighting the need for further research to validate the reliability of GT quantification methods compared with direct clinical measurements.

When it comes to qualitative assessments, different studies have established various thresholds for defining thin versus thick gingiva. Nonetheless, all of them have used TGP measurements as the gold standard to determine these cut‐offs. Guliyev et al. (2024) reported substantial agreement between PPB and CCP, defining 0.8 mm as the threshold for distinguishing thin from thick gingiva. This same threshold was also identified by an earlier study by Frost et al. (2015), which assessed 306 maxillary anterior teeth. However, that study was unable to establish a new threshold with acceptable reliability. More recent studies have tested the accuracy of PPB and CCP, with both reporting high agreement between the two methods (Uysal et al. 2024; Guliyev et al. 2024; da Costa et al. 2023). However, it was also emphasised that both methods carry a significant risk of misclassifying gingival phenotype with the 1 mm cut‐off value.

Despite the availability of these methods, it should be emphasised that the gold standard for assessing true soft tissue thickness remains histology. However, only two studies (Ferry et al. 2022; Gonçalves Motta et al. 2017) have incorporated histological assessment. While these studies provided valuable insights into the agreement between CBCT, TGP and histological assessment, neither established a cut‐off value based on histological GT. Given these limitations in the current literature—both in terms of the accuracy of various diagnostic methods (CBCT, US, TGP, CCP and PPB) compared with the gold standard of histology, and the absence of a histologically based cut‐off threshold for thin versus thick gingiva—this study aimed to test the reliability of these tools while revisiting the cut‐off value through histological assessment.

2. Materials and Methods

2.1. Study Design

This study was conceptualised as a diagnostic accuracy cadaveric experiment. The study was conducted with full adherence to the latest EQUATOR guidelines, specifically the STARD (Standards for Reporting of Diagnostic Accuracy Studies) (Bossuyt et al. 2015) statement. The experiment was exempted by the University of Michigan Institutional Review Board (IRB) under the application number HUM00168533.

2.2. Study Specimens and Eligibility Criteria

The study consisted of 16 fresh, non‐embalmed human cadaver heads. To minimise any structural damage to the intra‐oral tissues, all specimens were kept frozen under a controlled temperature of −20°C without formalin fixation following harvesting from the donors. Immediately prior to utilisation for the research experiment purpose, the specimens thawed to room temperature. Thawing was conducted at a controlled temperature of 2°C over an extended period (approximately 48 h) to ensure gradual and uniform thawing (Klop et al. 2017). Measurements were completed within an 8‐h window after thawing to avoid the onset of significant autolytic changes, such as nuclear vacuolation and eosinophilia, which have been observed to occur beyond this timeframe (Yadav et al. 2015; Mahalakshmi et al. 2016). The following inclusion criteria were considered: 1—fully or partially dentate maxilla and/or mandible; 2—probing pocket depth ≤ 4 mm; 3—absence or ≤ 3 mm midfacial gingival recession; 4—intact gingivae (without mechanical injury or trauma due to handling the specimens); 5—presence of at least 2 mm band of keratinized mucosa. The specimens were excluded in case of: 1—Complete edentulism; 2—teeth with the absence of neighbouring teeth; 3—GR > 3 mm; 4—probing pocket depth > 4 mm; 5—third molars; 6—less than 2 mm keratinized mucosa width; and 7—presence of gingival pigmentation.

2.3. Clinical Measurements and Data Collection

Each cadaver head was accompanied by an information sheet regarding the subject's age and gender. These data were collected prior to other measurements. The sequence of data collection was as follows: 1—CBCT; 2—ultrasonographic scan; 3 and 4—probe transparency (both [PPB] and [CCP]); 5—TGP; and 6—histology sample collection. This sequence was defined to minimise tissue damage/distortion prior to CBCT and US scans and concluded by taking biopsy samples from the gingivae. A detailed description of each measurement technique is described below.

2.3.1. US Scans

Ultrasound scans were taken at 2 mm apical to the free gingival margin (FGM) using a ZS3 device with a 24 MHz transducer and specific settings for optimal periodontal imaging (Figure 1a). The gingival thickness (GT) was measured perpendicular to a reference line parallel to the root surface, extending 2 mm apically from the FGM, with measurements recorded in DICOM format for analysis (Figure 1b,c). A detailed description of US imaging is provided in Appendix.

FIGURE 1.

Clinical data collection and scanning. (a) Ultrasonographic assessment using a miniaturized probe. (b) Mid‐facial ultrasound scan of one of the included sites. (c) Annotated midfacial scan with soft tissue demarcated in pink colour. (d) Cone beam computed tomographic gingival thickness assessment. (e) Stainless steel probe transparency. (f–h) Colour‐coded probe visibility (Aslan et al. 2021): Initially, a white probe was inserted into the gingival sulcus with less than 30 g of pressure; if the probe's colour was visible through the tissue, the phenotype was classified as thin. If not, a green probe was used under the same conditions, with visibility indicating a medium‐thickness phenotype. If the Green probe was also not visible, the blue probe was utilised, with its visibility marking a thick phenotype. The absence of visibility even with the blue probe classified the tissue as very thick. (i–k) Transgingival probing method and assessment of measured thickness under microscope.

2.3.2. CBCT Scans (Figure 1d)

The description of CBCT scans is provided in Appendix.

2.3.3. Probe Transparency

Two transparency methods to assess gingival phenotype were employed. The first of which was PPB (stainless steel probe) visibility. This assessment consisted of using a stainless steel PPB (UNC 15, Hu‐Friedy, Chicago, IL, USA). Briefly, the probe was inserted into the mid‐facial gingival sulcus and the visibility of the probe was assessed and recorded (as a binary outcome: Yes/No). Based on the visibility of the probe through gingiva, the tooth site was categorised as either thin or thick phenotype (Figure 1e).

2.3.4. Colour‐Coded Probe Visibility

The second transparency technique used was utilising CCPs (Colorvue Biotype Probes, Hu‐Friedy) (Aslan et al. 2021). To assess the gingival phenotype, one calibrated examiner (H.S.) employed a sequential probing method using CCPs with increasing visibility thresholds (Figure 1f–h).

2.3.5. Transgingival Probing

Gingival thickness (GT) was clinically measured using a no. 25 endodontic spreader with a rubber stopper for marking. The spreader was inserted 2 mm below the gingival margin until it contacted a solid surface, with measurements taken under 16× magnification (Figure 1i–k). The calibration and measurement process, including validation steps to ensure accuracy, is detailed further in Appendix.

2.3.6. Histological Assessment

As the final step in data collection, histological samples were obtained to serve as the gold‐standard measurement for GT. To ensure that the histological assessment accurately corresponded with the TGP site, a standardised procedure was followed. A black marker was used to create a half‐circle reference point 2 mm below the gingival margin, precisely aligned with the transgingival probe location. A 6‐mm diameter trephine drill (Long Trephine Drills Kit, Salvin Dental Specialties LLC, Charlotte, NC, USA) was then used to extract a biopsy sample that included both gingival and underlying hard tissue (Figure 2). The inclusion of hard tissue was a deliberate step to capture and analyse the full thickness of the gingiva at the exact probing site, thereby minimising any risk of missing tissue or measurement bias. All samples were fixed in 10% neutral buffered formalin for 1 week, followed by demineralisation for 10 days using Immunocal (Statlab, McKinney, TX, USA), with the solution refreshed every 24 h. To ensure accurate measurement at the mid‐point of the sample, each specimen was carefully bisected along the central axis. This ensured that the section observed under the microscope represented the centre of the sample. Histologic slides (4 μm thick) were then prepared using paraffin embedding and haematoxylin–eosin staining. Two independent investigators with expertise in histology (A.A. and P.G.‐F.) measured the histologic distance 2 mm below the marginal gingiva and reported the mean value (Figure 2).

FIGURE 2.

Histological sample and quantification of gingival thickness. At this precise point, GT was measured in a buccolingual direction, from the outer surface of the gingiva to the hard tissue (either the root surface or buccal bone), perpendicular to the tissue layers. GM: gingival margin; STT: soft tissue thickness.

2.3.7. Examiner Calibration

Clinical data collection, including ultrasonographic scan acquisition and probe transparency, was conducted by a single examiner (H.S.). Separate examiners performed CBCT (P.H.), TGP (P.N.) and histological thickness measurements (A.A. and P.G.‐F.). To ensure measurement reliability, a comprehensive calibration process was implemented.

The calibration included an initial training session followed by multiple rounds of reproducibility testing. Data from 10 teeth were used as calibration dataset (non‐study sites). Each examiner conducted measurements on the same set of samples across different sessions, separated by several days, to assess intra‐examiner reliability. The inter‐examiner agreement was evaluated through blinded measurements on a separate set of samples, with comparisons made against the reference examiner (H.S.). Intra‐class correlation coefficients (ICC) were calculated for both inter‐examiner and intra‐examiner reliability, with all ICC values exceeding 0.80, confirming high consistency and excellent agreement among the examiners.

2.3.8. Sample Size Calculation

Sample size calculations were performed to ensure sufficient power for both the continuous and categorical assessments of GT. For continuous measures (histology, US, CBCT, TGP), we aimed to assess agreement using the ICC, based on Walter et al. (1998) suggested approach, with an expected agreement of 80% (ICC = 0.80), a significance level of 0.05 and 90% power. The ICC(2,k) model for absolute agreement among the four methods was applied, resulting in a required sample size of 115 teeth. For the categorical methods (PPB and CCP transparency), we evaluated diagnostic performance using ROC analysis with a threshold of 1 mm for classifying thin (≤ 1 mm) versus thick (> 1 mm) gingiva. For the CCP, categories were simplified by merging ‘thin’ and ‘medium’ into a single ‘thin’ category, and ‘thick’ and ‘very thick’ into a ‘thick’ category, aligning with the binary classification used in the PPB method. With expected sensitivity and specificity of 80%, a sample size of 85 teeth was calculated. All calculations were conducted using G*Power 3.1 (Heinrich Heine University, Düsseldorf, Germany) and RStudio (Version 1.4.1717; RStudio, PBC, Boston, MA, USA).

2.4. Study Outcomes and Focused Questions

2.4.1. Primary Outcome

The primary outcome of this study was to evaluate the accuracy of US, CBCT and TGP in measuring GT compared with the gold standard histological thickness. Additionally, the study assessed the accuracy of PPB and CCP in categorising gingiva as thin or thick, using histological thickness categorised as < 1 mm or ≥ 1 mm for comparison.

2.4.2. Secondary Outcomes

Secondary outcomes included the level of agreement among US, CBCT and TGP for quantifying GT, assessing their reliability and consistency in clinical practice. Additionally, a new threshold for classifying gingiva as thin or thick was proposed based on histology, and this was compared with the traditional 1 mm cut‐off.

2.5. Statistical Analysis

The description is provided in Appendix.

3. Results

3.1. Descriptive Results of Included Samples

Initially, 125 samples (teeth) underwent full sequence of data collection; however, 10 teeth were excluded due to various reasons (inability to quantify the TGP thickness [n = 2], damaged tissue in histological processing [n = 2] and unidentifiable CBCT soft tissue [n = 6]). Finally, 115 samples were included in the study. Table 1 (top) presents the characteristics of included cadavers and teeth, along with the descriptive results of the study. A total of 16 cadavers were included (10 male and 6 female). Among the included teeth, 63 (54.7%) and 52 (45.3%) were from maxillary and mandibular arches respectively. Moreover, the distribution of type of teeth included were incisors: 43 (37.4%), canines: 20 (17.4%), premolars: 31 (26.9%) and molars 21 (18.3%). The mean GT from histological samples was 1.34 ± 0.65 mm, while the mean GT from US, CBCT and TGP was 1.25 ± 0.55, 1.13 ± 0.54 and 1.51 ± 0.65 mm (Figure 3a). When it comes to the probe visibility methods, PPB indicated 45 (39.1%) thin and 70 (60.9%) thick sites. However, CCP assessment showed 41 (35.6%) thin, 14 (12.2%) medium, 14 (12.2%) thick and 46 (40%) thick GT among included teeth.

TABLE 1.

(Top) Characteristics of the cadavers and sites included along with the mean and standard deviations (SD) of gingival thickness (in mm) measured with each method; (Bottom) Pearson's correlation coefficients (r), p‐values and 95% confidence intervals for relationships between gingival thickness measurement techniques (ultrasonography, transgingival probing, CBCT and histology).

	Count	GT assessment method
		Histology	US	CBCT	TGP	SS probe		Colour coded probe
						Thin, N (%)	Thick, N (%)	Thin, N (%)	Medium, N (%)	Thick, N (%)	Very thick, N (%)
Subjects
Total (N)	16
Male (%)	10 (62.5)	1.32 ± 0.67	1.23 ± 0.57	1.11 ± 0.56	1.48 ± 0.66	26 (35.6)	47 (64.4)	21 (31.54)	10 (13.7)	12 (16.4)	28 (38.36)
Female (%)	6 (37.5)	1.37 ± 0.62	1.28 ± 0.53	1.16 ± 0.51	1.55 ± 0.62	19 (45.2)	23 (54.8)	18 (42.9)	4 (9.5)	2 (4.8)	18 (42.8)
Teeth
Total (N)	115	1.34 ± 0.65	1.25 ± 0.55	1.13 ± 0.54	1.51 ± 0.65	45 (39.1)	70 (60.9)	41 (40)	14 (12.2)	14 (12.2)	46 (35.6)
Arch
Maxilla (%)	63 (54.7)	1.37 ± 0.63	1.30 ± 0.51	1.17 ± 0.52	1.53 ± 0.62	25 (39.7)	38 (60.3)	20 (31.7)	9 (14.3)	9 (14.3)	25 (39.7)
Mandible (%)	52 (45.3)	1.30 ± 0.67	1.19 ± 0.60	1.08 ± 0.57	1.47 ± 0.68	20 (38.5)	32 (61.5)	21 (40.4)	5 (9.6)	5 (9.6)	21 (40.4)
Type
Incisor (%)	43 (37.4)	1.12 ± 0.43	1.08 ± 0.39	0.98 ± 0.40	1.28 ± 0.40	23 (53.5)	20 (46.5)	20 (46.5)	8 (18.6)	7 (16.3)	8 (18.6)
Canine (%)	20 (17.4)	0.81 ± 0.47	0.80 ± 0.37	0.78 ± 0.36	0.98 ± 0.47	14 (70)	6 (30)	12 (60)	4 (20)	3 (15)	1 (5)
Premolar (%)	31 (26.9)	1.48 ± 0.57	1.36 ± 0.47	1.16 ± 0.51	1.66 ± 0.54	8 (25.8)	23 (74.2)	9 (29.03)	2 (6.45)	3 (9.67)	17 (54.8)
Molar (%)	21 (18.3)	2.07 ± 0.57	1.86 ± 0.53	1.72 ± 0.54	2.24 ± 0.63	0 (0)	21 (100)	0 (0)	0 (0)	1 (4.8)	20 (95.2)

Variables	Pearson's correlation				Mixed‐models ICC a
	r	95% CI lower	95% CI upper	p value	ICC	p value	95% CI lower	95% CI upper
US versus TGP	0.96	0.94	0.97	< 0.001	0.87	0.01	0.76	0.96
US versus CBCT	0.88	0.83	0.91	< 0.001	0.86	< 0.001	0.76	0.91
US versus histology	0.97	0.96	0.98	< 0.001	0.95	< 0.001	0.89	0.97
TGP versus CBCT	0.89	0.84	0.92	< 0.001	0.73	0.02	0.61	0.9
TGP versus histology	0.98	0.97	0.99	< 0.001	0.95	0.008	0.90	0.98
CBCT versus Histology	0.89	0.85	0.92	< 0.001	0.83	< 0.001	0.55	0.92

Abbreviations: CBCT: cone beam computed tomography; CI: confidence interval; N: number; r: Pearson's rho/correlation coefficient; SS: stainless steel; TGP: transgingival probing; US: ultrasonographic.

^a The ICC model was performed under a mixed‐models approach where subject IDs were entered as random effects, to account for between‐subject variability.

FIGURE 3.

(a) Violin plot illustrating the distribution of GT measurements across four different assessment methods: US, TGP, CBCT and histological assessment. The width of each violin represents the probability density of the data at various thickness levels, showing the distribution and spread of GT for each method. The solid line within each violin indicates the median GT, while the black bars represent the interquartile range. (b) Heatmap of the correlation coefficients between GT measurements using four different methods: CBCT, US, TGP and histological assessment. Higher correlation values (closer to 1) indicate stronger agreement between the methods, with histology serving as the gold standard. (c) Scatter plots showing the correlations between GT measurements with corresponding Pearson correlation coefficients (r) and p values. Note that the r values were rounded up compared to the exact numbers reported in Table 1.

3.2. Agreement Between Different GT Measurement Methods

The comparison of various measurement techniques using Pearson's correlation and mixed‐models ICC demonstrated strong linear relationships and varying degrees of agreement between the methods (Table 1 [bottom], Figure 3b,c). Pearson's correlation coefficients indicated very strong correlations across all pairs, with values ranging from 0.88 to 0.98. Specifically, the correlation between US and TGP was particularly high (r = 0.96, 95% CI [0.94, 0.97], p < 0.001), as was the correlation between US and histology (r = 0.97, 95% CI [0.96, 0.98], p < 0.001) and TGP and histology (r = 0.98, 95% CI [0.97, 0.99], p < 0.001). Similarly, strong correlations were observed between CBCT and histology (r = 0.89, 95% CI [0.85, 0.92], p < 0.001), and TGP and CBCT (r = 0.89, 95% CI [0.84, 0.92], p < 0.001). The lowest correlation, although still strong, was found between US and CBCT (r = 0.88, 95% CI [0.83, 0.91], p < 0.001). The mixed‐models ICCs indicated moderate to excellent agreement between the measurement techniques, with ICC values ranging from 0.73 to 0.95. The agreement was highest between US and histology (ICC = 0.95, 95% CI [0.89, 0.97], p < 0.001) and between TGP and histology (ICC = 0.95, 95% CI [0.90, 0.98], p = 0.008). US and TGP also demonstrated a high degree of agreement (ICC = 0.87, 95% CI [0.76, 0.96], p = 0.01). However, the agreement was more moderate between TGP and CBCT (ICC = 0.73, 95% CI [0.61, 0.90], p = 0.02), while CBCT and histology showed strong agreement (ICC = 0.83, 95% CI [0.55, 0.92], p < 0.001). Similarly, the agreement between US and CBCT was also strong (ICC = 0.86, 95% CI [0.76, 0.91], p < 0.001).

3.2.1. Bland–Altman Analysis

The Bland–Altman analysis was conducted to assess the agreement and bias levels between various measurement methods (Figure 4a–f). The comparison between histology and CBCT showed a mean difference of 0.125 (underestimation by CBCT), with 95% limits of agreement ranging from −0.351 to 0.601. Similarly, histology compared to US exhibited a mean difference of 0.032, with limits of −0.191 to 0.255. When histology was compared with TGP, a mean difference of −0.085 (overestimation) was observed, with limits of agreement from −0.165 to −0.005. The TGP versus CBCT comparison revealed a mean difference of 0.210, with limits of agreement extending from −0.257 to 0.677. The TGP versus US comparison demonstrated a mean difference of 0.117, with limits of agreement from −0.101 to 0.336. Lastly, when US was compared to CBCT, the analysis showed a mean difference of −0.093 (underestimation by CBCT), with limits of agreement ranging from −0.470 to 0.285.

FIGURE 4.

Bland–Altman plots with histograms showing the agreement between GT measurements across different methods: H versus CBCT, H versus US, H versus TGP, TGP versus US, TGP versus CBCT, and US versus CBCT. The red dashed line indicates the mean difference (Mean Diff), and the dotted lines represent the 95% limits of agreement. Histograms display the distribution of GT measurements for each method.

3.3. Mixed‐Models Regression for Estimating Histological Gingival Thickness From Alternative Methods

The results of mixed‐effects linear regression models revealed significant associations between each measurement method (US, CBCT and TGP) and histological thickness (Table A1). The regression model showed that for every 1 mm increase in the US measurement, the histological thickness increased by 1.11 mm (Estimate: 1.11, SE: 0.03, p < 0.001). The second model indicated that each 1 mm increase in the CBCT measurement corresponded to a 0.99 mm increase in histological thickness (Estimate: 0.99, SE: 0.06, p < 0.001). Lastly, TGP measurement resulted in a 0.97 mm increase in each unit increase in histological thickness (Estimate: 0.97, SE: 0.02, p < 0.001). Based on the regression models following formulae can be defined to calculate histological thickness based on available measurements from other methods:

• Using US: Histological Thickness = −0.06 + (1.11 × US)

• Using CBCT: Histological Thickness = 0.23 + (0.99 × CBCT)

• Using TGP: Histological Thickness = −0.13 + (0.97 × TGP)

3.3.1. Probe Visibility Methods

Table 2 presents the results of qualitative GT assessment techniques. Phenotype assessment with PPB identified 45 sites as thin (0.80 ± 0.29 mm) and 70 sites as thick (1.69 ± 0.57 mm) gingiva. Considering the 1 mm cut‐off value (Jepsen et al. 2018), this resulted in 62 true‐positive and 33 true‐negative sites, leading to 88.6% positive predictive value (PPV) and 73.3% negative predictive value (NPV). The sensitivity and specificity accuracy for PPB were 83.8%, 80.5% and 82.6%, which resulted in an AUC of 0.82 (95% CIs 0.75–0.89). Moreover, CCP assessment classified 55 sites (0.78 ± 0.26 mm) as thin and 60 sites (1.78 ± 0.51 mm) as thick gingiva. This resulted in 60 TP, 41 TN, 100% of PPV and 74.5% of NPV. The ROC analysis indicated sensitivity, specificity and accuracy of 100%, 81.1% and 87.8%, respectively, while AUC being 0.90 (95% CIs 0.86–0.95). Lastly, the Kappa statistics assessing the agreement between the two methods of probe transparency indicated a substantial agreement (Cohen's Kappa = 0.78, SE: 0.06, p < 0.001, z score: 12).

TABLE 2.

The results of ROC analysis on qualitative gingival thickness assessment methods (periodontal probe and colour‐coded probes). Note that in order to convert the four categories of colour‐coded probes to a binary outcome, thin and moderate categories were merged into thin and thick and very thick categories into thick gingiva.

Method of assessment	Descriptive analysis		ROC analysis a
	Mean ± SD (SE)	Total (N)	TP/TN	FP/FN	PPV/NPV	Sensitivity	Specificity	Accuracy	AUC (95% CIs)
Periodontal probe		115
Thin	0.80 ± 0.29 (0.04)	45	62/33	8/12	88.6%/73.3%	83.8%	80.5%	82.6%	0.82 (0.75–0.89)
Thick	1.69 ± 0.57 (0.06)	70
CCP (merged)
Thin	0.78 ± 0.26 (0.03)	55	60/41	0/14	100%/74.5%	100%	81.1%	87.8%	0.90 (0.86–0.95)
Thick	1.78 ± 0.51 (0.06)	60

	Kappa statistics
	Cohen's Kappa	SE	p value	z score	Interpretation b
Periodontal probe versus CCP	0.78	0.06	< 0.001	12.00	Substantial agreement b

Abbreviations: AUC: area under curve; CCP: colour‐coded probe; CIs: confidence intervals; N: number; NPV: negative predictive value; PPV: positive predictive value; ROC: receiver operating characteristic; SD: standard deviation; SE: standard error.

^a Model adjusted for within‐subject variability (Subject ID).

^b Kappa values ≤ 0 as indicating no agreement and 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial and 0.81–1.00 as almost perfect agreement.

3.4. Definition of a New Cut‐Off Point for Thin Versus Thick Gingivae

Testing different cut‐off points resulted in varying ROC curves, sensitivity, specificity and Youden's index values. The visualised ROC curve (Figure 5a and Table A2) included a range of cut‐off points from 0.6 to 1.3 mm. The visualisation indicated that compared with the previous 1.00 mm cut‐off (AUC = 0.80, Sensitivity = 73%, Specificity = 88%, Youden's index = 0.62), the 1.18 mm cut‐off demonstrated improved diagnostic performance with a higher AUC of 0.88, sensitivity of 96%, specificity of 80% and Youden's index of 0.76. Therefore, based on histological thickness and PPB visibility, the cut‐off value of 1.18 was suggested to serve as the new clinical reference (Figure 5b).

FIGURE 5.

(a) ROC curves with 95% confidence intervals for different gingival thickness cut‐off points, displaying the AUC values for each threshold. (b) Sensitivity, specificity and Youden's index plotted against various cut‐off values for gingival thickness, with the selected cut‐off point of 1.18 mm highlighted in red.

4. Discussion

4.1. Main Findings

This diagnostic accuracy study tested all commonly available GT assessment methods in clinical practice and research compared with the gold standard assessment of GT through histology. To the best of the authors' knowledge, this is the first study that included histological data as the reference point (in addition to TGP) to test the diagnostic accuracy of other techniques for measuring soft tissue thickness. Likewise, another significance of the study was providing the agreement and degree of bias between all other GT quantification methods (US, CBCT and TGP), which can serve as a reference for future research.

When comparing the accuracy of GT quantification methods (US, CBCT and TGP) compared to histology, US (r = 0.97, ICC: 0.95, p < 0.001) and TGP (r = 0.98, ICC: 0.95, p = 0.008) outperformed CBCT (r = 0.89, ICC: 0.83, p < 0.001). Moreover, the Bland–Altman analysis revealed that US and CBCT tend to slightly underestimate the GT compared with histology (mean difference of 0.03 and 0.12 mm respectively), while TGP tended to slightly overestimate (mean difference: −0.08 mm). This mainly indicates, despite various reliability range, all three techniques fit into an acceptable range of accuracy compared to the gold standard when considering clinical implications. These results were mainly in accordance with the previous studies. The only available histological assessment was performed by Ferry et al. (2022), in an in vitro study on 72 teeth. While the landmark definition to assess GT and CBCT acquisition settings in that study was slightly different from that of ours, their TGP and CBCT assessment also showed an overestimation (0.22 mm) and underestimation (0.23 mm) respectively compared with histology. Nonetheless, the ICC values in that study revealed considerably lower values (0.58 and 0.54 for TGP and CBCT, respectively) compared with this study. Overall, it can be postulated that while slight biases were observed across the comparisons, the methods generally showed good agreement within clinically acceptable ranges, with CBCT displaying slightly more variability in its measurements compared to the other techniques. Despite lack of histological assessment, several studies, with a wide range of publication timespan, have attempted to assess the diagnostic accuracy of the other techniques including US, CBCT, TGP and probe transparency (Wang et al. 2022; Kloukos et al. 2018; Fan et al. 2023; Uysal et al. 2024; de Freitas Silva et al. 2023; Guliyev et al. 2024; Frost et al. 2015; da Costa et al. 2023; Ferry et al. 2022). Among the very limited number of studies where histological assessment was performed, Gonçalves Motta et al. (2017) compared PPB visibility versus histological GT. The results of their study on 10 sites indicated an average of 1.93 ± 0.19 mm GT in thick and 1.4 ± 0.16 mm GT in thin sites, with a statistically significant difference (p = 0.01). These results were slightly higher than the GT determined in our study (based on PPB transparency). Nonetheless, it should be emphasised that the sample size in that study included only five sites per group, which makes it difficult to perform a proper comparison with our cohort.

In the present study, the diagnostic accuracy of the PPB and the CCP for identifying thin and thick gingival phenotypes was meticulously assessed. Our findings demonstrated that while both PPB and CCP are effective in identifying gingival phenotypes, CCP exhibited a marginally superior diagnostic accuracy with an AUC of 0.90, compared with 0.82 for PPB. This aligns with the results of the study by Frost et al. (2015), who reported that traditional probe visibility methods have limited accuracy, particularly when using a 1 mm threshold, which only achieved an AUC of 0.43. Interestingly, Frost et al. (2015) noted an improved accuracy (AUC of 0.67) when the threshold was adjusted to 0.8 mm, suggesting that the specific GT threshold plays a critical role in the diagnostic performance of these tools. Nevertheless, their assessment was based on TGP and digital photographs to determine the visibility of the PPB through the gingival tissue. In contrast, our study performed the same analysis using histologically validated GT measurements as the gold standard. When applying a 1.18 mm threshold in our study, we achieved an improved diagnostic accuracy compared with 0.8 mm threshold with an AUC of 0.88 (vs. 0.69 for 0.8 mm threshold), surpassing the accuracy reported by Frost et al. (AUC: 0.67) for the 0.8 mm threshold. This highlights the enhanced precision of our approach, which may better capture the nuances of gingival tissue variability and improve clinical decision‐making. Additionally, a recent study by Guliyev et al. (2024), compared TGP, CCP and PPB. Aligned with our results, their study indicated a substantial agreement (Kappa = 0.726, p < 0.001) between PPB and CCP. Also, PPB showed an overall accuracy of 85% (Kappa = 0.538, p < 0.001), however, in contrast to our study, Guliyev et al. (2024) used TGP as the gold standard reference for actual GT values.

4.2. Clinical Implications and Practice Guideline

It should be emphasised that despite serving as the gold standard measurement, obtaining histological samples from patients for diagnostic purposes is not feasible. This mainly highlights the importance of assessment of the accuracy of TGP, CBCT and US measurements, because these methods are more widely used and available in clinical practice and research. Therefore, the accuracy of the aforementioned methods along with their association with histological GT would provide crucial implications for clinical practice. Based on this, in this study, not only the diagnostic accuracy of the GT quantification methods was assessed, but also a mixed‐models linear regression model was built to carry out a prediction model for histological GT based on the available tools in clinical practice. The results of this regression model indicated strong associations between US (Estimate: 1.11, SE: 0.02, p < 0.001), CBCT (Estimate: 0.99, SE: 0.05, p < 0.001) and TGP (Estimate: 0.97, SE: 0.01, p < 0.001) with histology. Based on these results, three GT prediction formulae were generated (see Section 3), using which the clinicians would be able to predict histological GT by obtaining CBCT, US or TGP thickness clinically.

Additionally, to address the qualitative GT assessment methods (thin/thick), we tested a wide range of possible cut‐off points while entering the histological thickness and PPB visibility into the model as references. Herein, we aimed to first define the most reliable and accurate cut‐off value (1.18 mm) and, secondarily, evaluate the diagnostic performance of this cut‐off. The results indicated that when using 1.18 mm as the threshold, the method demonstrated a high sensitivity of 0.96, meaning that 96% of gingiva classified as thick by probe invisibility indeed had a histological thickness of ≥ 1.18 mm. The specificity was 0.80, indicating that 80% of gingiva classified as thin by probe visibility were correctly identified as having a thickness of < 1.18 mm. The overall accuracy of the method was 0.86, reflecting the correct classification in 86% of cases. The area under the curve (AUC) was 0.88, underscoring the model's strong ability to differentiate between thin and thick gingiva. Additionally, the Youden's index, calculated at 0.76, further highlights the optimal balance between sensitivity and specificity. These findings affirm that the 1.18 mm cut‐off is a reliable and accurate metric for distinguishing between thin and thick gingiva using probe visibility as a qualitative assessment method. However, while this threshold provides an evidence‐based refinement over the arbitrary 1 mm consensus, it is essential to emphasise that its primary value lies in research rather than direct clinical application. Given the potential for measurement errors and inherent limitations of the cadaveric model, the clinical detectability of a 0.18 mm difference remains uncertain, and future studies are necessary to validate this cut‐off in vivo to confirm its broader applicability. Lastly, it is crucial to acknowledge the limitations of each gingival thickness assessment method in clinical practice. While CBCT provides detailed imaging, it exposes patients to radiation, making repeated use a concern. TGP, although accurate, requires local anaesthesia, adding an invasive element to the procedure. Ultrasonography, despite its potential, is not yet widely accessible in dental clinics and demands a relatively high learning curve, limiting its current widespread application.

4.3. Limitations

The description is provided in Appendix.

5. Conclusions

Within its limitations, this diagnostic accuracy study demonstrates that transgingival probing, ultrasonography and CBCT are all reliable methods for quantifying GT compared with the ‘gold standard’ histology. The transgingival probing method showed a slight overestimation of GT, while ultrasonography and CBCT tend to slightly underestimate it. In qualitative assessments, the periodontal probe and colour‐coded probe showed excellent agreement in distinguishing between thin and thick gingiva, but with some limitations in accuracy. Furthermore, a new cut‐off threshold of 1.18 mm discriminating thin versus thick gingiva was established based on histological GT and probe visibility. The limitations of using cadaveric tissues must be acknowledged, with caution in translating these results into clinical practice.

Author Contributions

Hamoun Sabri: conceptualisation, data collection, data analysis, visualisation, manuscript writing, final approval. Paolo Nava: data collection, visualisation, manuscript writing, final approval. Parham Hazrati: data collection, manuscript writing, final approval. Abdusalam Alrmali: data collection, histological analysis, final approval. Pablo Galindo‐Fernandez: data collection, histological analysis, final approval. Muhammad H. A. Saleh: conceptualisation, critical review and final approval. Javier Calatrava: data collection, final approval. Shayan Barootchi: conceptualisation, critical review, final approval. Lorenzo Tavelli: conceptualisation, manuscript writing, critical review, final approval. Hom‐Lay Wang: conceptualisation, critical review and final approval.

Conflicts of Interest

The authors declare no conflicts of interest.

Methods

Ultrasonographic Scan Acquisition

A commercially available ultrasonographic (US) imaging device (ZS3, Mindray, Mountain View, CA, USA) coupled with a prototype transducer of 24 MHz imaging frequency (32 mm axial image resolution) and toothbrush‐sized (~30 × 18 × 12 mm) probe (Figure 1a) were used to acquire the images. Superficial modes were selected, including ‘small parts examination’, ‘superficial scanning’, ‘15‐mm field depth’, and ‘spatial compounding’, for optimal magnification and delineation and superior imaging resolution of periodontal structures on the display during the scanning procedure. Ultrasound gel and a gel‐based stand‐off pad (Aquasonic, Parker Inc., Fairfield, NJ, USA) were applied to the probe head to achieve acoustic coupling to the periodontal tissue area of interest. B‐mode scans generating 2‐dimensional images of the apico‐coronal and facio‐palatal dimensions were acquired at the mid‐facial site of the teeth and saved in DICOM format. The US B‐mode images consisted of grey‐scale pixels, in which the degree of pixel brightness (i.e., letter B) signifies the intensity of the received ultrasound echo in any tissue location. This method allowed for visualisation of spatial relations (soft‐hard tissue boundaries and various tooth structures), their quantification and characterisation of soft tissues as a result of backscatter changes. The DICOM files were imported into a software (Horos v.4.0.0, Horos Project) where a single pre‐calibrated operator (H.S.) performed the measurement of GT. GT was measured 2 mm apically to the FGM (Robin et al. 2011) by first drawing a reference line starting from FGM that was parallel to the root surface, then extending 2 mm apically. GT line was drawn perpendicular to the reference line, starting from the root surface and ending on the soft tissue surface. The displayed length of GT line was then recorded (Figure 1b,c).

Cone Beam Computed Tomography Scans

Prior to clinical data collection, a cone beam computed tomography (CBCT) scan was obtained from cadaver heads. In order to prevent overlapping of lip and cheek with gingiva and periodontal soft tissue, a plastic lip retractor was placed during the scan to separate the soft tissues of the lips and cheek from the gingiva. These scans were taken at University of Michigan School of Dentistry using a commercially available CBCT scanner (3D Accuitomo 170; J Morita, Kyoto, Japan). The settings for exposure were 5 mA and 90 kVp for 17.5 s. The field of view (FOV) was set at 140 × 100 mm, with a nominal spatial resolution of 0.27 mm.

Images were then converted into Digital Imaging and Communications in Medicine (DICOM) files and subsequently imported into a commercially available software package for CBCT measurements (Blue Sky Bio, Grayslake, IL, USA). The measurements were conducted on the multi‐planar reformation (MPR) images using non‐orthogonal sagittal cuts, with a 1‐mm slice interval and 1‐mm slice thickness. The gingival thickness was assessed on the mid‐cross‐section slice of each tooth. Initially, the gingival margin was demarcated by drawing a tangent line to the free gingival margin. From this tangent line, a perpendicular line of 2 mm in length was drawn. The gingival thickness was then measured from the bone or tooth surface to the external part of the gingiva at the endpoint of this 2 mm line (Figure 1d).

Transgingival Probing

A #25 endodontic finger spreader file (DENTSPLY Maillefer, CanadaWoodbridge, ON) was used to clinically measure GT. A rubber stopper was placed on the spreader to mark the outer surface of the gingiva. A reference point was established and measured with periodontal probe, located 2 mm below the gingival margin on the mid‐facial aspect of each tooth. The endodontic spreader was then inserted into the gingiva, perpendicular to the facial plane, until it contacted a solid surface (either tooth or buccal bone). The rubber stopper adapted to the outer surface of the gingiva using small‐tip tissue forceps, before the spreader was removed. To measure the actual thickness of the gingiva, the spreader was positioned alongside a periodontal probe for calibration purposes. Next, under 16× magnification of an operating microscope (OPMI pico, Carl Zeiss, Germany) a digital screenshot was captured with the probe and spreader next to each other (Figure 1i–k). The JPEG file of this photo was imported to a commercially available photo processing software (ImageJ program, NIH, Bethesda, MD, USA). The measurement ruler of the software was calibrated using the 1 mm markings on the periodontal probe. The distance from the tip of the spreader to the rubber stopper was then measured, providing the gingival thickness at each site. To ensure the accuracy of GT measurements and mitigate potential errors, an immediate validation process was implemented. After taking three consecutive readings of transgingival probing (TGP) at each site, the data was reviewed in real time for consistency. If a significant discrepancy was detected, the measurement was immediately retaken, with the median value of all measurements used to account for any potential outliers caused by issues such as rubber stopper movement. All probe transparency and TGP assessments were performed by one pre‐calibrated examiner (H.S.).

Statistical Analysis

Descriptive statistics were calculated as percentages for categorical variables and means with standard deviations (SDs) for continuous variables. All statistical analyses were performed by one author (H.S.) using RStudio software (Version 1.4.1717; RStudio, PBC). Agreement between continuous measurements (US, CBCT, TGP) and histological thickness was assessed using ICC with mixed‐effects model to account for the clustering of data within the same cadaver heads. Furthermore, a Pearson's correlation model was used to test the correlation coefficient (reported as Pearson's rho [r]) among the different measurement methods. Additionally, a Bland–Altman analysis was conducted to visualise and quantify the differences between each method and histology as the gold standard by calculating the mean difference (bias) and 95% limits of agreement. For categorical outcomes, receiver operating characteristic (ROC) curve analysis and area under the curve (AUC) were used to evaluate the diagnostic accuracy of PPB and CCP visibility in classifying gingiva as thin (< 1 mm) or thick (≥ 1 mm) (Jepsen et al. 2018), with adjustments for clustering effects using a mixed‐effects model. The level of agreement between CCP and PPB transparency was also tested using Cohen's Kappa statistics. Additionally, to establish a more precise cut‐off point, derived from histological data, between thin and thick gingivae, an ROC curve analysis was performed on various potential thresholds (ranging from 0.6 to 1.30 mm). Sensitivity, specificity and Youden's index (calculated as Youden's index = [Sensitivity % + Specificity % − 100]) were computed for each threshold to identify the optimal cut‐off. Youden's index is a measure that maximises the effectiveness of a diagnostic test by balancing sensitivity and specificity; a higher Youden's index indicates a better performance of the test. Lastly, a mixed‐models linear regression was used to assess systematic biases and predict histological thickness from other methods. The following formula was used to build prediction models for histological thickness based on other quantitative methods (CBCT, US and TGP):

The standard formula for a linear regression model is Y = β ₀  + β ₁ X, where:

• Y is the dependent variable (in this case, Histological Thickness).

• X is the independent variable (e.g., US, CBCT or TGP).

• β ₀ is the intercept (the constant term).

• β ₁ is the slope (the coefficient for the independent variable).

The irr (Gamer et al. 2019), blandr (Caldwell 2022), pROC (Robin et al. 2011) and lme4 (Bates et al. 2015) packages in RStudio were used to perform the statistical analyses, with significance set at p < 0.05.

Discussion

Limitations

The limitations of the current study should also be acknowledged. First, the readers should bear in mind that the assessment of histological and/or other methods on cadaveric samples has inherent limitations. Nevertheless, we ensured that the use of cadaveric samples adhered to the highest quality guidelines and recommendations, as outlined in the latest forensic medicine studies and cadaver study guidelines. The freezing, thawing and measurement phases were all meticulously conducted in accordance with these recommendations, as previously mentioned in the methods section. Due to ethical considerations, performing a human study with the same methodology is not feasible. Secondly, while the findings regarding the 1.18 mm cut‐off for distinguishing between thin and thick gingiva are methodologically sound, it is important to note that the results were derived from a cadaveric model. Caution is needed when extrapolating this threshold to clinical (in vivo) settings, as physiological factors such as tissue hydration and blood flow may influence gingival thickness measurements. Further clinical studies are necessary to validate the reproducibility of our findings in live patient scenarios.

Moreover, it is worth mentioning that other GT assessment methods have also been tested in recent studies. Nik‐Azis et al. (2023) compared TGP versus direct calliper measurement of GT, their results indicated a mean difference of −0.07 mm between the two methods with a good agreement level (r = 0.229, p = 0.03). As a more novel technique, Schwarz et al. (2023) tested the accuracy of a dental magnetic resonance imaging (MRI) device, compared with CBCT (superimposed with 3D intraoral scans). The mean difference between these two methods was 0.17 ± 0.27 mm (p < 0.001). Moreover, the comparison of CBCT versus PPB visibility in their study resulted in 47.02% agreement. Additionally, the exclusion of specimens with GR > 3 mm and the requirement for at least 2 mm of keratinized tissue (KT) may slightly reduce the external validity of the study. These criteria ensured methodological consistency and focus on relatively periodontally healthy tissues, but caution is needed when applying the results to individuals with more advanced gingival recession or reduced keratinized tissue. Moreover, readers should keep in mind that, while our study focused on the periodontal phenotype (GT), the results may differ when peri‐implant soft tissue is considered, and further studies are needed to validate our findings in this context. Therefore, applying the outcomes of our study to peri‐implant soft tissue should be done with caution. Lastly, while this study provides valuable insights into the GT assessment, it should be highlighted that re‐conducting the same assessments with devices other than that used in this study (CBCT and US scanners), may result in different results. Thus, the generalisability of the results of this study should be done with caution.

TABLE A1.

Linear mixed‐effects regression model analysis on the outcome of histological GT and prediction models by ultrasonographic, CBCT and transgingival thicknesses. Note that all three models were conducted by assigning the subject IDs as random effects and predictor variables as fixed effects.

Predictor	Random effects	Fixed effects
	Variance (SD)	Estimate (SE)	p value
Model 1 (ultrasonography)
Subject ID	0.004 (0.06)	—	—
Intercept	—	−0.05 (0.03)	0.158
US	—	1.11 (0.02)	< 0.001
Model 2 (CBCT)
Subject ID	0.01 (0.1)	—	—
Intercept	—	0.22 (0.07)	0.006
CBCT	—	0.99 (0.05)	< 0.001
Model 3 (transgingival probe)
Subject ID	0.03 (0.06)	—	—
Intercept	—	−0.13 (0.03)	< 0.001
TGP	—	0.97 (0.01)	< 0.001

Note: Significant p values are marked in bold.

TABLE A2.

ROC analysis on multiple suggested cut‐offs for defining thin versus thick gingivae, based on periodontal probe visibility and histological thickness. Note that the most accurate measure is that of the highest Youden's index, taking into account both sensitivity and specificity. Note that the 1.18 value was selected based on the highest value in Youden's index presented in the ROC curve (see Figure 5).

Cut‐off (mm)	Accuracy	AUC	Youden's index	Specificity	Sensitivity
0.80	0.74	0.69	0.39	0.94	0.44
0.90	0.76	0.72	0.45	0.91	0.53
1.00	0.82	0.80	0.62	0.88	0.73
1.10	0.84	0.83	0.67	0.85	0.82
1.18	0.86	0.88	0.76 a	0.80	0.96
1.20	0.85	0.87	0.74	0.77	0.97
1.30	0.80	0.83	0.67	0.70	0.97

Abbreviations: AUC: area under curve; FPR: false‐positive rate; mm: millimetres; TPR: true‐positive rate.

^a Highest Youden's index, indicating the most accurate cut‐off point.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.