A comparison of the validity of the Demirjian, Willems, Nolla and Häävikko methods in determination of chronological age of 5-15 year-old Indian children.

BACKGROUND: Demirjian's method has been the most popular and extensively tested radiographic method of age estimation. More recently, Willems' method has been reported to be a better predictor of age. Nolla's and Häävikko's methods have been used to a lesser extent. Very few studies have compared all four methods in non-Indian and Indian populations. Most Indian research is limited by inadequate sample sizes, age structures and grouping and different approaches to statistical analysis.

AIM: The present study aimed to evaluate and compare the validity of the Demirjian, Willems, Nolla and Häävikko methods in determination of chronological age of 5 to 15 year-old Indian children.

DESIGN: In this cross-sectional observational study, four methods were compared for validity in estimating the age of 1200 Indian children aged 5-15 years.

RESULTS: Demirjian's method overestimated age by +0.24 ± 0.80, +0.11 ± 0.81 and +0.19 ± 0.80 years in boys, girls and the total sample, respectively. With Willems' method, overestimations of +0.09 ± 0.80, +0.08 ± 0.80 and +0.09 ± 0.80 years were obtained in boys, girls and the total sample, respectively. Nolla's method underestimated age by -0.13 ± 0.80, -0.30 ± 0.82 and -0.20 ± 0.81 years in boys, girls and the total sample, respectively. Häävikko's method underestimated age by -0.17 ± 0.80, -0.29 ± 0.83 and -0.22 ± 0.82 years in boys, girls and the total sample, respectively. Statistically significant differences were observed between dental and chronological ages with all methods (p < 0.001). Significant gender-based differences were observed with all methods except Willems' (p < 0.05). Gender-specific regression formulae were derived for all methods.

CONCLUSION: Willems' method most accurately estimated age, followed by Demirjian's, Nolla's and Häävikko's methods. All four methods could be applicable for estimating age in the present population, mean prediction errors being lower than 0.30 years (3.6 months).