Is the assessment of bone age by the Greulich-Pyle method reliable at forensic age estimation for Turkish children?

BACKGROUND: Estimation of age is an important task for forensic experts especially in developing countries where birth records are often not well maintained. In this study, we investigated whether or not the Greulich-Pyle (G-P) method is sufficient at forensic age estimation for Turkish children.
METHODS: Plain radiographies of left hands and wrists of 492 (241 (49.0%) female, 251 (51.0%) male) healthy children between 11 and 18 years of age for girls and 11 and 19 years of age for boys were taken. Mean chronological ages (CA) were compared with mean skeletal ages according to G-P atlas for each gender and in the second step the differences those for each age group were determined. The children were Caucasian and had a low-middle socioeconomic status in this study population. The paired sample t test was used to indicate the difference between G-P (mean skeletal age according to G-P atlas) and CA (mean chronological age). In order to indicate the relation Pearson correlation coefficient was used. Modeling the relationship between G-P and CA linear regression was used. The analyses were done under SPSS 11.5.
RESULTS: The G-P compared to their CA. The CA was 14.52+/-2.18 S.D. years, S.E.: 0.14 (median: 14.47, range: 11.07-18.92 years) whereas G-P was 15.06+/-2.31 S.D. years, S.E.: 0.15 (median: 15.00, range: 10-18 years) for girls. The difference between the two parameters was statistically significant (p<0.001). The CA was 15.28+/-2.41 S.D. years, S.E.: 0.15 (median: 15.09, S.E. range 11.13-19.94 years) and G-P was 15.41+/-2.92 S.D. years, S.E.: 0.18 (median: 15.60, range 9-19 years) for boys. The difference was not statistically significant (p>0.05). There was a high correlation (Pearson r=0.882, p<0.001) for girls and (Pearson r=0.900, p<0.001) for boys. The determination coefficient (R2) is equal to 0.778 for girls and 0.81 for boys. The regression model also tested by ANOVA and it is found significant (p<0.01) for both genders. According to age groups G-P was advanced (0.17-1.1 year) almost for all ages and differences were significant at 11, 12, 14, 16 ages for girls. G-P was delayed at 11-14 ages (0.01-0.58 year) but not significant except for 13 years and G-P were significantly advanced in 15-17 ages (0.88-0.98 years) but then delayed in 18-19 years of age (0.02-0.48) for boys. The difference's standard deviation at 12, 13, 15, 16 years of ages for girls and between 12 and 16 and 18 years of ages for boys were more than 1 year.
CONCLUSION: When the results of the previous study and this study are compared, the test method could be used technically by pediatricians and other clinicians, but it is even more important for ethically unacceptable errors to disappear, especially in cases involving the possible criminal liability of the supposed minor. So standard deviation at 12, 15 years of ages for girls and 12, 15, 18 years of ages for boys were more than 1 year. But it is not known that the other methods whether more useful or not than this method. For the time being unless any other methods will be proved more useful, we have to use this method cautiously for possible criminal liability cases in forensic age diagnosis.