Correlation between dental age and chronological age of Taiwanese children on panoramic X-ray images by different evaluation methods

Abstract

Background:

Accurate age estimation is vital in forensic medicine and clinical dentistry. The Demirjian method is commonly used for this purpose, but its applicability to Asian individuals is understudied. The present study evaluated the accuracy of dental age (DA) estimation for Taiwanese children by using the Demirjian, Willems, and modified Demirjian methods, comparing these estimates to those of chronological age (CA) based on panoramic X-ray images.

Methods:

A total of 232 Taiwanese children aged between 5 and 12 years underwent panoramic X-ray scans. Their permanent teeth were assessed using the Demirjian, Willems, and modified Demirjian methods to estimate DA. Regression analysis was used to determine the correlation between CA and DA, with linear regression equations established using SPSS statistics software to identify differences.

Results:

The Willems method had the lowest mean absolute error and the smallest mean difference between DA and CA among the 3 age estimation methods analyzed (p < 0.001). The R² value for the difference between DA and CA was 0.831 for the Willems method, 0.813 for the Demirjian method, and 0.782 for the modified Demirjian method (p < 0.001). The Willems method had the highest correlation with CA, with the linear equation for Taiwanese children being CA = 0.822 × DA + 1.093. For comparison, for the Demirjian method, the equation was CA = 0.894 × DA + 0.165.

Conclusion:

The effectiveness of predictive methods varies across ethnicities, and therefore, region-specific formulas are required. For Taiwanese children, the Willems method predicts CA most accurately. This study contributes to the fields of legal medicine and clinical dentistry by demonstrating the accuracy of DA in predicting CA.

1. INTRODUCTION

Accurately predicting age is crucial in both clinical dentistry and forensic science. Physicians often utilize molecular pathology techniques to develop personalized treatments on the basis of an individual’s predicted age.¹ In clinical pediatric dentistry, age prediction is useful for evaluating the growth and development of patients, diagnosing precocious development or orthodontic malocclusion, and determining whether medical intervention is necessary. Age prediction is also crucial in forensic science research, assisting in solving legal cases by identifying the age of individuals from skeletal remains.²

Most age estimation methods are invasive, have long processing times, utilize expensive equipment, and must be performed by experienced pathologists.³ However, a less invasive and less expensive method is available that involves calculating age on the basis of tooth development.⁴ This method is viable because dentition is solid, undergoes few structural modifications, and is relatively unaffected by the external environment.⁵ Additionally, after death, dentition is more resistant to decomposition than other parts of the body. Furthermore, dentition is relatively stable over time,⁶ rendering it a useful indicator for estimating maturity and predicting the age of children and teenagers.⁷

Scholars have developed several methods for estimating the dental age (DA) of children and teenagers,⁸ including the Cameriere, Willems, and Demirjian methods. Other methods, such as the Kvaal method, are used for estimating the DA of adults.⁹

In 1973, Demirjian, Tanner, and Goldstein developed the Demirjian method for estimating chronological age (CA) from DA; they examined panoramic X-ray images of French-Canadian children aged between 2 and 20 years and divided dentition development into eight stages (A-H), with a score assigned to each stage. The total score of all tested teeth is the dental maturity score (DMS), which is subsequently converted to a tooth age by using a standard table.¹⁰

In 2001, Willems et al modified the Demirjian method by creating a new table to predict the age of a White Belgian population. The modified method can be used to directly calculate the DA without DMS conversion.¹¹ Acharya¹² modified the original Demirjian method to create the modified Demirjian method. They studied the DA of Indian individuals between the ages of 7 and 25 years and observed that the modified Demirjian method required adjustment before use in other populations.¹²

Multiple studies have compared the Demirjian and Willems methods and have determined that the Willems method is more accurate,¹³ but these studies have rarely focused on children in Asian regions. Therefore, the present study analyzed the developmental stages of the dentition of Taiwanese children and compared the accuracy of the Demirjian, Willems, and modified Demirjian methods. This study also established a regression equation using dentition as an indicator of age, providing reference data for Taiwanese children aged between 5 and 12 years.

2. METHODS

2.1. Clinical data collection

This study was conducted in the Pediatric Dentistry Department of China Medical University Hospital in Taiwan, the imaging databases of which were accessed to obtain panoramic X-ray images (Fig. 1). A total of 232 patients aged between 5 and 12 years were included: 119 (mean CA: 7.78 years, SD: 1.74 years) boys and 113 (mean CA: 7.8 years, SD: 1.4 years) girls. This study was approved by the Ethics Committee of China Medical University Hospital (Approval Number: CMUH111-REC2-206).

Fig. 1.

Panoramic X-ray image.

The inclusion criteria were (1) being a child aged 5 to 12 years in the tooth replacement stage, (2) having complete digital X-ray images in the hospital database, (3) having no permanent teeth, (4) having no systemic disease that may affect growth, and (5) having no history of orthodontic treatment. The exclusion criteria were (1) having had surgery on the upper and lower jaws, (2) having a history of jaw trauma, (3) missing teeth on both sides, (4) having abnormal development, (5) having severe malocclusion, or (6) having unclear X-ray images.

2.2. Demirjian method

The age prediction method was proposed by Demirjian et al,¹⁴ which remains the most widely used age estimation method. On the basis of observations of the developmental status of the 7 permanent teeth on the left side of the mandible (central incisors, lateral incisors, canines, first premolars, second premolars, first molars, and second molars), each tooth position is assigned a developmental stage (grades A-H; Fig. 2).

Fig. 2.

Correlation between CA and DA using the Demirjian method. R² = 0.813 p < 0.001. CA = chronological age; DA = dental age.

The maturity conversion tables (Tables 1 and 2) established by Demirjian et al¹⁴ are used to convert the developmental stages of each tooth position into a maturity score. The total maturity score is calculated as the sum of the maturity scores for the 7 teeth. The maturity score ranges from 0 to 100. On the basis of the sex and maturity score comparison tables (Tables 3 and 4), the patient’s DA is calculated.

Table 1.

Maturity score from the Demirjian method according to individual tooth development stage (boys)¹⁴

Stage	31	32	33	34	35	36	37
0					0.0		0.0
A					1.7		2.1
B				0.0	3.1		3.5
C		0.0	0.0	3.4	5.4	0.0	5.9
D	0.0	3.2	3.5	7.0	9.7	8.0	10.1
E	1.9	5.2	7.9	11.0	12.0	9.6	12.5
F	4.1	7.8	10.0	12.3	12.8	12.3	13.2
G	8.2	11.7	11.0	12.7	13.2	17.0	13.6
H	11.8	13.7	11.9	13.5	14.4	19.3	15.4

Table 2.

Maturity score from the Demirjian method according to individual tooth development stage (girls)¹⁴

Stage	31	32	33	34	35	36	37
0					0.0		0.0
A					1.8		2.7
B				0.0	3.4		3.9
C		0.0	0.0	3.7	6.5	0.0	6.9
D	0.0	3.2	3.8	7.5	10.6	4.5	11.1
E	2.4	5.6	7.3	11.8	12.7	6.2	13.5
F	5.1	8.0	10.3	13.1	13.5	9.0	14.2
G	9.3	12.2	11.6	13.4	13.8	14.0	14.5
H	12.9	14.2	12.4	14.1	14.6	16.2	15.6

Table 3.

Demirjian method maturity score DA conversion scale (boys)¹⁴

Age	Score	Age	Score	Age	Score
5.0	25.4	8.0	71.6	11.0	92.0
0.1	26.2	0.1	73.5	0.1	92.2
0.2	27.0	0.2	75.1	0.2	92.5
0.3	27.8	0.3	76.4	0.3	92.7
0.4	28.6	0.4	77.7	0.4	92.9
0.5	29.5	0.5	79.0	0.5	93.1
0.6	30.3	0.6	80.2	0.6	93.3
0.7	31.1	0.7	81.2	0.7	93.5
0.8	31.8	0.8	82.0	0.8	93.7
0.9	32.6	0.9	82.8	0.9	93.9
6.0	33.6	9.0	83.6	12.0	94.0
0.1	34.7	0.1	84.3	0.1	94.2
0.2	35.8	0.2	85.0	0.2	94.4
0.3	36.9	0.3	85.6	0.3	94.5
0.4	38.0	0.4	86.2	0.4	94.6
0.5	39.2	0.5	86.7	0.5	94.8
0.6	40.6	0.6	87.2	0.6	95.0
0.7	42.0	0.7	87.7	0.7	95.1
0.8	43.6	0.8	88.2	0.8	95.2
0.9	45.1	0.9	88.6	0.9	95.4
7.0	46.7	10.0	89.0	13.0	95.6
0.1	48.3	0.1	89.3	0.1	95.7
0.2	50.0	0.2	89.7	0.2	95.8
0.3	52.0	0.3	90.0	0.3	95.9
0.4	54.3	0.4	90.3	0.4	96.0
0.5	56.8	0.5	90.6	0.5	96.1
0.6	59.6	0.6	91.0	0.6	96.2
0.7	62.5	0.7	91.3	0.7	96.3
0.8	66.0	0.8	91.6	0.8	96.4
0.9	69.0	0.9	91.8	0.9	96.5

DA = dental age.

Table 4.

Demirjian method maturity score DA conversion scale (girls)¹⁴

Age	Score	Age	Score	Age	Score
5.0	28.9	8.0	78.8	11.0	94.5
0.1	29.7	0.1	80.2	0.1	94.7
0.2	30.5	0.2	81.2	0.2	94.9
0.3	31.3	0.3	82.2	0.3	95.1
0.4	32.1	0.4	83.1	0.4	95.3
0.5	33.0	0.5	84.0	0.5	95.4
0.6	34.0	0.6	84.8	0.6	95.6
0.7	35.0	0.7	85.3	0.7	95.8
0.8	36.0	0.8	86.1	0.8	96.0
0.9	37.0	0.9	86.7	0.9	96.2
6.0	38.0	9.0	87.2	12.0	96.3
0.1	39.1	0.1	87.8	0.1	96.4
0.2	40.2	0.2	88.3	0.2	96.5
0.3	41.3	0.3	88.8	0.3	96.6
0.4	42.5	0.4	89.3	0.4	96.7
0.5	43.9	0.5	89.8	0.5	96.8
0.6	45.2	0.6	90.2	0.6	96.9
0.7	46.7	0.7	90.7	0.7	97.0
0.8	48.0	0.8	91.1	0.8	97.1
0.9	49.5	0.9	91.4	0.9	97.2
7.0	51.0	10.0	91.8	13.0	97.3
0.1	52.9	0.1	92.1	0.1	97.4
0.2	55.5	0.2	92.3	0.2	97.5
0.3	57.8	0.3	92.6	0.3	97.6
0.4	61.0	0.4	92.9	0.4	97.7
0.5	65.0	0.5	93.2	0.5	97.8
0.6	68.0	0.6	93.5	0.6	98.0
0.7	71.8	0.7	93.7	0.7	98.1
0.8	75.0	0.8	94.0	0.8	98.2
0.9	77.0	0.9	94.2	0.9	98.3

DA = dental age.

2.3. Willems method

The Willems method was proposed in 2001¹⁵ as a modified version of the Demirjian method that was originally applied to White Belgian individuals. As with the Demirjian method, the individual developmental status of the 7 permanent teeth on the left side of the mandible (central incisors, lateral incisors, canines, first premolars, second premolars, first molars, and second molars) is assessed. Additionally, as with the Demirjian method, each tooth position is assigned a developmental stage grade from A to H.

Maturity conversion tables (Tables 5 and 6) were proposed for converting the developmental stages of boys and girls at each tooth position into a maturity score. The sum of the maturity scores for a patient’s 7 teeth represents their DA.

Table 5.

Maturity score from the Willems method according to individual tooth development stage (boys)¹⁵

Stage	31	32	33	34	35	36	37
A				0.15	0.08		0.18
B				0.56	0.05		0.48
C	1.68	0.55		0.75	0.12		0.71
D	1.49	0.63	0.04	1.11	0.27	0.69	0.8
E	1.5	0.74	0.31	1.48	0.33	1.14	1.31
F	1.86	1.08	0.47	2.03	0.45	1.6	2
G	2.07	1.32	1.09	2.43	0.4	1.95	2.48
H	2.19	1.64	1.9	2.83	1.15	2.15	4.17

Table 6.

Maturity score from the Willems method according to individual tooth development stage (girls)¹⁵

Stage	31	32	33	34	35	36	37
A				−0.95	−0.19		0.14
B				−0.15	0.01		0.11
C	1.83		0.6	0.16	0.27		0.21
D	2.19	0.29	0.54	0.41	0.17	0.62	0.32
E	2.34	0.32	0.62	0.6	0.35	0.9	0.66
F	2.82	0.49	1.08	1.27	0.35	1.56	1.28
G	3.19	0.79	1.72	1.58	0.55	1.82	2.09
H	3.14	0.7	2	2.19	1.51	2.21	4.04

2.4. Modified Demirjian method

Acharya¹² developed the Demirjian 8-teeth method (the modified Demirjian method) for use in the Indian population. In this method, a developmental stage ranging from 0 to 9 is assigned to the 8 permanent teeth on the left side of the mandible (central incisors, lateral incisors, canines, first premolars, second premolars, first molars, second molars, and third molars) on the basis of the individual developmental status of the teeth.¹⁶ The difference between the Demirjian method and the modified Demirjian method is that the modified Demirjian method includes third molars.

In the modified Demirjian method, the modified maturity score conversion tables¹⁷ (Tables 7 and 8) of Chaillet and Demirjian are used to convert the developmental stage of each tooth into a maturity score. The total maturity score is calculated as the sum of the maturity scores of the 8 deciduous teeth. The maturity score ranges between 0 and 100. On the basis of the patient’s sex and maturity score, the patient’s DA is estimated using the same tables as those in the Demirjian method (Tables 3 and 4).¹⁴

Table 7.

Maturity score using the modified Demirjian method for individual tooth development stage (boys)¹⁷

Stage	31	32	33	34	35	36	37	38
0							1.7	6.19
1					1.69		2.98	7.64
2				1.70	2.27		3.41	8.28
3			1.70	1.98	3.41		4.74	8.86
4			2.67	3.52	3.41		4.88	9.89
5	2.31	2.55	4.34	5.19	5.59	2.13	6.69	11.17
6	4.35	4.71	6.14	6.47	6.96	3.73	7.89	12.25
7	5.16	5.75	7.59	8.18	8.68	4.94	9.08	13.66
8	6.56	6.97	9.52	9.84	10.64	7	11.13	14.07
9	10.68	10.91	12.56	12.57	13.11	11.22	13.63	15.32

Table 8.

Maturity score using the modified Demirjian method for individual tooth development stage (girls)¹⁷

Stage	31	32	33	34	35	36	37	38
0								6.40
1							2.57	7.74
2					2.43			8.92
3				2.56	3.43		2.65	9.31
4			2.55	3.54	3.83		4.1	10.22
5	2.58	2.65	3.15	5.09	5.75	2.58	6.51	11.04
6	3.1	4.54	5.4	6.31	6.81	3.25	8	12.65
7	5.02	5.40	7.19	8.09	8.7	4.25	9.13	13.77
8	6.66	7.02	9.22	9.82	10.8	6.88	11	14.45
9	10.61	10.89	11.99	12.29	12.79	10.94	13.84	16.65

2.5. Statistical analysis

SPSS (SPSS, Chicago, IL) was used to analyze the data. Pearson correlation analysis was applied to examine the correlations between the DAs and CAs estimated using the Demirjian, Willems, and modified Demirjian methods. The R² value was calculated for each method. The p value in Pearson correlation was also utilized to assess the significance of the correlation analysis and to determine whether the correlation coefficient was significant. A p value threshold of 0.01 was used for the Pearson correlations.

The paired t test was used to compare significant differences between the CA and the DA of the deciduous teeth estimated using each method. The significance level (p value) was set at α = 0.05.

2.6. Establishment of linear regression equations

The following two models were established for age estimation for forensic medicine and clinical dentistry use:

1. Forensic perspective: A linear regression equation was established on the basis of observations of the developmental status of the deciduous teeth to estimate the CA of an individual, where the independent variable (X-axis) was set to DA and the dependent variable (Y-axis) was set to CA.

2. Dental perspective: A linear regression equation was established on the basis of observations of the CA of children for predicting the developmental status of the deciduous teeth of a typical school-age child, where the independent variable (X-axis) was set to CA and the dependent variable (Y-axis) was set to DA.

3. RESULTS

3.1. Accuracy of estimated age

The average CA for all patients was 7.79 ± 1.58 years. The average predicted DA was 8.53 ± 1.6 years for the Demirjian method, 8.15 ± 1.76 years for the Willems method, and 7.2 ± 0.69 years for the modified Demirjian method (Table 9). The estimated mean absolute error (MAE) was the smallest for the Willems method, which had an MAE of 0.63 years (Table 9).

Table 9.

CA and predicted DA of the patients using various methods

	Demirjian method		Williem method		Modified Demirjian method
	Mean	SD	Mean	SD	Mean	SD
Chronological age	7.79	1.58	7.79	1.58	7.79	1.58
Dental age	8.53	1.60	8.15	1.76	7.20	0.69
MAE	0.82	0.61	0.63	0.51	0.91	0.76
Mean diff of age (DA-CA)	0.74		0.36		−0.59
p	<0.001		<0.001		<0.001

CA = chronological age; DA = dental age; MAE = mean absolute error.

Significant differences between CA and the DA for each of the three methods were determined using a paired t test. Significant differences were observed between the CA and DA (p < 0.001) for the three methods (Table 9). Specifically, the Demirjian and Willems methods overestimated the patients’ DA, whereas the modified Demirjian method underestimated DA. The Willems method had the smallest mean difference of age, with an average overestimation of DA by 0.36 years (Table 9).

3.2. Correlation between DA and CA

The correlations between the CA and the DA of Taiwanese children estimated using the Demirjian, Willems, and modified Demirjian methods were analyzed using Pearson correlation coefficients. The results illustrated in Fig. 2 reveal that the R² value for the difference between the DA and CA predicted by the Demirjian method was 0.813. By contrast, the results illustrated in Fig. 3 reveal that the R² value for the difference between DA and CA predicted by the Willems method was 0.831, indicating the highest correlation. The results displayed in Fig. 4 reveal that the R² value for the difference between DA and CA predicted by the modified Demirjian method was 0.782. All p values were <0.001, indicating significant correlations between DA and CA for all three methods.

Fig. 3.

Correlation between CA and DA using the Willems method. R² = 0.831 p < 0.001. CA = chronological age; DA = dental age.

Fig. 4.

Correlation between CA and DA using the modified Demirjian method. R² = 0.782 p < 0.001 CA = chronological age; DA = dental age.

3.3. Linear equations

The linear regression curves for the CA and DA of the Taiwanese children were established using SPSS. The following linear equations were created on the basis of the Willems method, with consideration of both forensic and dental perspectives (Fig. 5 and Fig. 6):

Fig. 5.

The DA to CA equation constructed using the Willems method and developed for forensic medicine perspective, allows for the derivation of CA from a child’s DA obtained from the morphology of a tooth. CA = chronological age; DA = dental age.

Fig. 6.

The equation from CA to DA, constructed using the Willems method and developed from a clinical dentistry perspective. This equation allows for the estimation of the age of a tooth should be, based on a patient’s CA. CA = chronological age; DA = dental age.

$$\text{CA}=0.822\times \text{DA}+1.093$$

$$\text{DA}=1.012\times \text{CA}+0.267$$

The linear regression curves for the Taiwanese children were established using SPSS. The following linear equations were created on the basis of the Demirjian method, with consideration of both forensic and dental perspectives (Fig. 7 and Fig. 8):

Fig. 7.

The DA to CA equation constructed using the Demirjian method and developed for forensic medicine perspective. CA = chronological age; DA = dental age.

Fig. 8.

The CA to DA equation constructed using the Demirjian method and developed for clinical dentistry perspective. CA = chronological age; DA = dental age.

$$\text{CA}=0.894\times \text{DA}+0.165$$

$$\text{DA}=0.909\times \text{CA}+1.447$$

4. DISCUSSION

4.1. Correlation between DA and CA for various methods

In the present study, the correlation between DA and CA for the Willems method had an R² value of 0.831, indicating a high correlation. Validation of the model revealed that the Willems method was the most accurate for estimating the age of Taiwanese children. The results of this study demonstrated that the Demirjian method significantly overestimated the CA of Taiwanese children, with an average overestimation of 0.74 years (p < 0.05). The average absolute error was 0.82 years. These results are consistent with those of Yan et al,¹⁸ who examined the results of 26 studies and determined that the Demirjian method significantly overestimated CA. Furthermore, a study conducted on individuals of East Turkish descent revealed an overestimation of between 0.4 and 1.3 years for boys and between 0.2 and 1.9 years for girls when the Demirjian method was used.¹⁹ Additionally, in a study conducted in Malaysia, the estimation of DA by using the Willems method produced results closer to the CA than those of the Demirjian method. In that study, the Demirjian method overestimated the CA of boys by an average of 0.7 years and that of girls by an average of 0.6 years. In the same study, the Willems method only slightly overestimated the CA of boys (by 0.2 years) and girls (by 0.1 years), indicating that the Willems method more accurately estimated the CA from DA in Malaysian children.²⁰

The Willems method improves on the Demirjian method but uses the same classification of developmental stages for permanent teeth. However, the Willems method is simpler, rendering it superior in clinical settings.

This study revealed the lowest correlation between DA and CA for the modified Demirjian method. The modified Demirjian method was created because the original Demirjian method excluded the third molars for age prediction beyond 16 years of age; thus, the original method was unsuitable for forensic medicine because the remaining permanent teeth, excluding the third molars, have already fully developed.²¹ Including the third molars in assessments may provide the only reliable radiographic parameter for individuals aged between 16 and 23 years.²² However, the patients in the present study were children aged between 5 and 12 years, rendering the inclusion of third molars unnecessary for age prediction.

4.2. Clinical significance

In this study, the Demirjian method overestimated the age of Taiwanese boys by an average of 0.98 years and that of Taiwanese girls by an average of 0.49 years. By contrast, the Willems method overestimated the age of Taiwanese boys by an average of 0.6 years and that of Taiwanese girls by an average of 0.11 years.

In clinical practice, radiation therapy and chemotherapy may cause irreversible damage to dentition, facial bones, and muscles during childhood.²³ Therefore, the findings of this study—specifically, those regarding the superior accuracy of the Willems method—can be applied to clinically assess the degree of dental developmental abnormality caused by cancer therapy.²⁴ This information can assist dentists and pediatric patients with early detection of delayed or accelerated dental development.

The present study used a sufficiently large sample size and a sufficiently accurate predictive method to ensure the accuracy of the results. However, this study has some limitations. First, the sample was exclusively drawn from Taiwanese children. Consequently, the linear equations developed to predict dental and chronological ages are likely only applicable to children in Taiwan. Nevertheless, the equations are relatively accurate for Taiwanese children. Second, this study excluded some methods for estimating age on the basis of dentition, such as the Haavikko, Cameriere, and Nolla methods. Future studies must include alternative methods to identify a suitable age prediction approach for Taiwanese children.

This study reached the following conclusions: (1) The DA calculated using the Demirjian and Willems methods overestimates the CA of Taiwanese children. However, the DA calculated using the modified Demirjian method underestimates the CA of Taiwanese children. (2) The MAE for age estimation using the Willems method was the smallest among those of the three methods examined. The R² value of the Willems method for the estimated DA and CA was 0.831, indicating it was the most accurate of the three examined methods. (3) The linear equations using the Willems method for estimating the ages of Taiwanese children are CA = 0.822 × DA + 1.093 or DA = 1.012 × CA + 0.267. (4) The linear equations using the Demirjian method for estimating the ages of Taiwanese children are CA = 0.894 × DA + 0.165 or DA = 0.909 × CA + 1.447.

ACKNOWLEDGMENTS

This study was supported by academic and government organizations: China Medical University Hospital (DMR-113-025) and National Science and Technology Council (MOST 111-2221-E-039-008-MY2) in Taiwan.