Assessing the construct validity of nutrient profiling models for restricting the marketing of foods to children in South Africa

In an effort to combat childhood obesity the WHO has called on governments to restrict the marketing of unhealthy foods to children. Regulators have turned to nutrient profiling (NP) to provide the evidence for their decisions. This has resulted in the development of NP models, of which the validity of only a few have been established. The aim of this study was to assess the construct validity of various NP models for the purpose of restricting the marketing of unhealthy foods to children in South Africa by comparing the classification of foods by the models to the ranking of the same foods by registered dietitians. Six current NP models were identified, then a representative food database of 120 foods was developed and each individual food was classified by each of the six models. Lastly dietitians were recruited to rank the healthfulness of the same 120 foods. Dietitians allowed 24% of the included foods for marketing to children, whereas the percentage of foods allowed by the included models ranged from 7 to 47%. Majority of pairwise comparisons between the NP models and dietitians yielded ҡ statistics >0.6, indicating substantial agreement. An almost perfect pairwise agreement was found between dietitians and the WHO Regional Office for Europe model. The included NP models displayed good construct validity by agreeing with dietitians on what are ‘less healthy’ foods, thus the foods not allowed for marketing. The findings of this study contributes to the process of establishing validity of NP models.