Table 3.
Statistical summary of all baseline additive and interactive STR (CNTN1-[T]N) models with polygenic scores (PGS) for suicidal thoughts and behaviors.
| Model specifications | Effect (Odds ratio) | SE | Z-score | p value |
|---|---|---|---|---|
| Baseline additive$ | 1.052 | 2.499 | 0.421 | 0.674 |
| Baseline additive$ + TSH PGS | 1.049 | 2.619 | 0.402 | 0.688 |
| Baseline additive$ + TLNWL PGS | 1.054 | 2.401 | 0.439 | 0.661 |
| Baseline interactive# | 1.997 | 0.523 | 3.804 | 1.43×10−4 |
| Baseline interactive# + TSH PGS | 1.965 | 0.523 | 3.759 | 1.70×10−4 |
| Baseline interactive# + TLNWL PGS | 1.960 | 0.523 | 3.745 | 1.81×10−4 |
| Baseline interactive#,@ -by-TSH PGS | 2.014 | 0.544 | 3.700 | 2.17×10−4 |
| Baseline interactive#,@ -by-TLNWL PGS | 1.97 | 0.529 | 3.728 | 1.93×10−4 |
Each “additive” model treated the STR and developmental stage as independent items in the generalized linear model while each “interactive” model considered STR-by-DevStage effects. When PGS were added to the models, their interaction was with respect to STR-by-DevStage. TSH, “thoughts of self-harm”; TLNWL, “thought life was not worth living.” $Baseline additive model: SUI009 ~ STR + DevStage + covariates. #Baseline interactive model: SUI009 ~ STR*DevStage + covariates. @Reported effect estimate corresponds to TR-by-DevStage in the presence of PGS as a third interacting variable; in other words, we report STR-by-DevStage, not STR-by-DevStage-by-PGS in this table. All GLM results are shown in Supplementary Table S5.