Table 3.

Associations Between Genotype level and Suicidality, Adjusted for Genetic Ancestry and Other Covariates^a

Analysis Period, Adjustment, and Efavirenz Metabolizer Measure	Estimated HR (95% CI)	Confidence Limit Ratio (Precision)	Wald P value
Efavirenz-exposed analysis
Self-identified race/ethnicity (n = 1831; 41 events)
10 level	1.12 (.97–1.29)	1.3	.13
12 level	1.11 (.97–1.28)	1.3	.12
3 level	1.48 (.85–2.56)	3.0	.17
Principal components (n = 1384; 40 events)
10 level	1.10 (.96–1.27)	1.3	.17
12 level	1.08 (.96– 1.23)	1.3	.21
3 level	1.42 (.84– 2.41)	2.9	.19
On-treatment analysis
Self-identified race/ethnicity (n = 1831; 34 events)
10 level	1.17 (1.01–1.36)	1.3	.04
12 level	1.16 (1.01–1.33)	1.3	.04
3 level	1.87 (1.05–3.31)	3.2	.03
Principal components (n = 1384; 33 events)
10 level	1.15 (.99–1.32)	1.3	.06
12 level	1.12 (.99–1.27)	1.3	.08
3 level	1.74 (1.02–2.97)	2.9	.04

Abbreviations: CI, confidence interval; HR, hazard ratio.

^aEach weighted Cox model to estimate the association between genotype level and a suicidality outcome was adjusted either for self-identified race/ethnicity or for genetic ancestry using 4 principal components. Each analysis also adjusted for psychiatric history or psychoactive medication, injection drug use history, sex, age category, and body weight category. In the multivariable adjusted analyses the events to covariates ratio is low, with 33–41 events, and covariates using 12 degrees of freedom. Results adjusted only for race/ethnicity or principal components 1–4 were similar.