Digital assessment of the fetal alcohol syndrome facial phenotype: reliability and agreement study

PURPOSE: To examine the three facial features of fetal alcohol syndrome (FAS) in a cohort of Australian Aboriginal children from two-dimensional digital facial photographs to: (1) assess intrarater and inter-rater reliability; (2) identify the racial norms with the best fit for this population; and (3) assess agreement with clinician direct measures. METHODS: Photographs and clinical data for 106 Aboriginal children (aged 7.4–9.6 years) were sourced from the Lililwan Project. Fifty-eight per cent had a confirmed prenatal alcohol exposure and 13 (12%) met the Canadian 2005 criteria for FAS/partial FAS. Photographs were analysed using the FAS Facial Photographic Analysis Software to generate the mean PFL three-point ABC-Score, five-point lip and philtrum ranks and four-point face rank in accordance with the 4-Digit Diagnostic Code. Intrarater and inter-rater reliability of digital ratings was examined in two assessors. Caucasian or African American racial norms for PFL and lip thickness were assessed for best fit; and agreement between digital and direct measurement methods was assessed. RESULTS: Reliability of digital measures was substantial within (kappa: 0.70–1.00) and between assessors (kappa: 0.64–0.89). Clinician and digital ratings showed moderate agreement (kappa: 0.47–0.58). Caucasian PFL norms and the African American Lip-Philtrum Guide 2 provided the best fit for this cohort. CONCLUSION: In an Aboriginal cohort with a high rate of FAS, assessment of facial dysmorphology using digital methods showed substantial inter- and intrarater reliability. Digital measurement of features has high reliability and until data are available from a larger population of Aboriginal children, the African American Lip-Philtrum Guide 2 and Caucasian (Strömland) PFL norms provide the best fit for Australian Aboriginal children.