Review of Existing Models to Predict Reductions in Neural Tube Defects Due to Folic Acid Fortification and Model Results Using Data from Cameroon

Several models have been developed to predict the effects of folic acid fortification programs on prevention of neural tube defects (NTDs), but each relies on different assumptions and data inputs. We identified and reviewed 7 models that predict the effects of folic acid intake or status on NTD risk. We applied 4 of these models [the original and a modified version of the Lives Saved Tool (LiST) and models developed by Arth et al. and Wald et al.] to predict the effect of folic acid fortification of wheat flour on reduction of NTDs using national survey data from Cameroon. The estimated percentage of NTDs averted due to fortified wheat flour (5.0 μg folic acid/g flour) varied by predictive model, with a 21–31% reduction in LiST to 83% in Arth's model, and 15% in Wald's model. As the simulated fortification level was increased from 1.0 to 7.0 μg folic acid/g flour, the pattern of change in estimated numbers of NTDs averted differed due to different model assumptions: the number of NTDs averted increased and then reached a plateau in the modified LiST model (as would be expected in real-world conditions), increased sharply in Arth's model, and increased continuously in Wald's model. This wide variation in predicted effects, and implausible results in some cases, undermines the models’ utility for users of model outputs. Concurrent collection of dietary and biomarker data, including plasma and RBC folate concentrations, and NTD outcomes, is necessary to validate these models and monitor change in folic acid intake, folate-related biomarkers, and reduced NTD risk due to fortification. In the meantime, models based on erythrocyte folate concentration are recommended, based on biological plausibility and consistency with empirical evidence. Where erythrocyte folate data are unavailable, sensitivity analyses (using several models) could be conducted to examine the range of possible outcomes.