Background: Infants with multiple malformations are important in birth defect epidemiology and malformation monitoring because human teratogens have often caused complex malformations. Various methods for the analysis of multimalformed infants have been tried.
Method: By using data from four large registries of congenital malformations, 5256 infants were identified with two or more among 73 selected malformations. Pairwise associations between malformations were detected by multiple logistic regression analyses, and putative confounders (programme, maternal age, autopsy, etc.) were controlled for. For each significant pairwise association, further analyses were performed in order to find associations with a possible third malformation.
Results: The importance of controlling for several confounders was demonstrated. Several well-known associations were found, which supports the technique used. The interpretation of three-way associations was discussed. Results from the present study were compared with those obtained using some other methods.
Conclusions: Different confounders can cause biased associations. The method presented in the paper takes this into consideration and is therefore more likely than previously used techniques to give unbiased information on the clustering of different malformations among multimalformed infants.