Figure 3. Interaction between riboflavin intake and iron intake at baseline in relation to anemia at follow up stratified by gender.

Marginsplot syntax was used to make the plot after logistic regression adjusting for age, smoking, energy intake, dietary patterns, education, income, BMI, and hypertension at baseline. Anemia status at baseline was also adjusted in the model. The values of iron intake represent the median intake in each quartile of iron intake. The values represent the adjusted probability of anemia at follow-up. P for interaction between riboflavin and iron intake was 0.016.