Reliability of a systematic methodology to estimate added sugars content of foods when applied to a recent Australian food composition database

• We applied a 10-step method to estimate added sugars to a food composition database.
• Added sugar content was estimated for 5740 foods by two researchers independently.
• Mean difference was 0.2 g/100 g, correlation 0.98.
• Kappa statistic (0.76) indicated good inter-rater agreement on steps chosen.
• Showed good reliability and transferability to other food composition databases.

This study aims to assess the reliability of a previously published method for estimation of added sugars content of foods when applied to the latest Australian food composition database, AUSNUT2011-2013. Two researchers independently applied a previously published, 10-step, stepwise methodology to estimate added sugars content of 5740 foods in AUSNUT2011-2013. The added sugars content of individual foods estimated by the two researchers was compared using paired sample t-tests and Pearson's correlations. Kappa statistics were used to assess the consistency between two researchers in steps chosen. Overall 4126 foods (72% of all foods in AUSNUT2011-2013) were assigned an estimated value based on objective criteria (Steps 1–6), and 1614 (28%) were assigned a subjectively estimated value (Steps 7–10). While statistically significant, the mean difference between the values estimated by the two researchers was small (0.7 g; SD 4.6 g). The kappa statistic (0.761; p < 0.001) indicated good inter-researchers agreement in steps chosen. Excellent correlations were observed between the two sets of values where the two researchers used the same step. In conclusion, the method has good reliability when applied to the new AUSNUT2011-2013. Further studies to examine its reliability when applied to food composition databases from other countries is warranted.