Validation of data provided by the deuterium oxide dose-to-mother technique for better evaluation of breastfeeding practice
We describe a simple method to validate data collected from a study using the deuterium oxide dose-to-the-mother technique for breastfeeding evaluation. We used human milk intake calculation spreadsheets (n=180). The calculation was performed by fitting the deuterium enrichment data to a model for water turnover in the mother and in the baby. We assumed that the validity of the results is as high as the square root mean square error (SRMSE) between calculated and fitted data is low. Based on the original spreadsheets that fitted well with the model (n=87), we developed a simple prediction of the SRMSE and we used it as cut-off to check, correct (by removing enrichment data) and validate or remove the other spreadsheets. We found a cut-off dependent on the measured enrichment (E_m) that was And the mean SRMSE (90%CI) of the fitted sheets was 23.37 mg.kg-1 (22.01 mg.kg-1, 24.73 mg.kg-1) with a maximum of 38.96 mg.kg-1. After correction we noticed that the number of enrichments removed per file varied from 1 to 4. We observed within the corrected spreadsheets a significant reduction (p≤0.0001, n=53) of the SRMSE (90%CI) from 49.78 mg.kg-1 (46.35 mg.kg-1, 53.20mg.kg-1) before correction to 25.88 mg.kg-1 (24.13 mg.kg-1, 27.64 mg.kg-1) after correction. We also observed that after correction, the mean difference (90%CI) of HM respectively non-HM that was 29.34 mg.kg-1 (21.71 mg.kg-1, 36.97 mg.kg-1) respectively 24.13 mg.kg-1 (17.4 mg.kg-1, 30.79 mg.kg-1) was strongly (p≤0.0001, n=53) different from zero. Therefore, the correction is very important to optimizing the results. Keywords: Breastfeeding; Deuterium; Excel spreadsheet; Square root mean square error; Validation