The best-fit multiple regression model for risk of CD in children with type 1 diabetes
| Genetic factor . | OR (95%CI) . | Wald P . |
|---|---|---|
| HLA-DQB1*02-DQA1*05 | ||
| Double dose* | 4.45 (1.75–11.3) | 0.002 |
| Single dose* | 2.04 (1.11–3.76) | 0.022 |
| TNF-α −308A | 1.87 (1.10–3.20) | 0.021 |
| IL-1α −889T | 0.58 (0.37–0.93) | 0.023 |
| Genetic factor . | OR (95%CI) . | Wald P . |
|---|---|---|
| HLA-DQB1*02-DQA1*05 | ||
| Double dose* | 4.45 (1.75–11.3) | 0.002 |
| Single dose* | 2.04 (1.11–3.76) | 0.022 |
| TNF-α −308A | 1.87 (1.10–3.20) | 0.021 |
| IL-1α −889T | 0.58 (0.37–0.93) | 0.023 |
The variables associated with CD in the univariate model were used to build a step-up multiple model. The predictor variables included two indicator variables showing a double and single dose of HLA-DQB1*02-DQA1*05 and dichotomous variables showing the phenotypic positivity for a given allele of a cytokine SNP. The adequacy of inclusion of predictor variables into the model was tested using the log-likelihood ratio test. No interaction terms improved the model. The effect of HLA-DQB1*0302-DQA1*03 observed in the univariate analysis was lost after adjusting for HLA-DQB1*02-DQA1*05.
Two indicator variables compare the levels of risk relative to an absence of HLA-DQB1*02-DQA1*05.