Common use of Statistical Methods Clause in Contracts

Statistical Methods. In the database congenital malformations are classified through a standard coding system by organ system into specific categories or into non-specific categories if no details are known. Eight different organ systems are distinguished for which there are 51 specified and 20 unspecified categories of congenital malformations. Logistic regression models were used to study the relationship between maternal ethnicity and congenital malformations. The overall relationship between ethnic group and the total prevalence of congenital malformations, the total prevalence within the eight organ systems and the prevalence of some specific congenital malformations was determined with the Likelihood Ratio Test (LRT). If this test was significant it showed the existence of an overall relationship between ethnicity and congenital malformations. Thereafter, the individual significance of the calculated odds ratios (ORs) expressing the observed risk differences in prevalence between the different ethnic groups and the Dutch group, used as reference group, was studied. Because maternal age is related to the ethnic group and to the occurrence of certain congenital malformations, we calculated the ORs both unadjusted and adjusted for the age of the mother. Because the prevalence of some malformations was low, even in this 5-year birth cohort, not all could be tested. We decided that the predicted number of malformations had to be at least 5 in each ethnic group to perform a worthwhile and clinically significant test. Therefore, from the 51 specific malformations registered in the linked national database only the following 15 were analysed for possible differences between ethnic groups: neural tube defects (NTD); congenital malformations of the ears; ventricular septal defect; single umbilical artery; cleft lip with/without cleft palate; cleft palate without cleft lip; intestinal/anorectal atresia; hypospadias and/or epispadias; undescended testes; polydactyly; syndactyly; deformities of the foot without NTD; Down’s syndrome; other chromosomal malformations; and multiple malformations. Many comparisons were performed to test for a possible ethnic difference in prevalence of any congenital malformations. To avoid chance findings resulting from to multiple testing we applied a Bonferroni correction in which the usual critical value of 0.05 is adapted to a lower one depending on the number of tests performed. For example, to determine in which of the ethnic groups a possible difference in overall prevalence of malformations exists, the critical value used was 0.05/6=0.008. To test whether there is any significant relationship between ethnicity and one of the eight organ systems the critical value used for the LRT’s was 0.05/8=0.006. Significant organ systems were selected for further examination to determine which ethnic group differed significantly from the Dutch reference group using a critical value of 0.05/8*6=0.001. For the analyses of the specific malformations similar adaptations of the critical value were performed depending on the number of malformations tested within each organ system. This article will only focus on the significant observations where the calculated P-values are below the Bonferroni corrected critical values. The Bonferroni correction is conservative. Therefore, the overall risk of stating a possible relationship of any malformation with ethnicity, while in reality it does not exist, is kept at least below 5%. All statistical analyses were performed in SPSS, version 11.