Skip to main content
. Author manuscript; available in PMC: 2019 Apr 21.
Published in final edited form as: Ann Intern Med. 2018 Jun 26;169(3):146–155. doi: 10.7326/M17-2252

Table 2.

Adjusted general estimating equation and Cox regression models for depression and suicidal ideation

All participants Excluding those with the outcome at baseline
Depression Suicidal ideation Depression Suicidal ideation
EFV NVP EFV NVP EFV NVP EFV NVP




Proportion of participants ever with outcome* (%) 61/305 (20.0%) 125/389 (32.1%) 19/305 (6.2%) 47/389 (12.1%) 17/206 (8.3%) 52/260 (20.0%) 14/284 (4.9%) 35/363 (9.6%)
Proportion of person-visits with outcome (%) 105/1522 (6.9%) 228/2332 (9.8%) 24/1522 (1.6%) 62/2332 (2.7%) 24/1014 (2.4%) 86/1586 (5.4%) 16/1405 (1.1%) 41/2195 (1.9%)
Events/PY 49/430.0 142/598.4 15/478.4 52/699.6 15/313.3 57/444.2 11/451.5 39/662.4
Crude IR (events/100 PY) 11.4 23.7 3.1 7.4 4.8 12.8 2.4 5.9
Crude IR difference (95% CI) −12.3 (−17.4, −7.3) −4.2 (−6.8, −1.5) −8.0 (−12.2, −3.9) −3.5 (−5.8, −1.1)
Crude IRR, (exact 95% CI) 0.48 (0.34, 0.67), P<0.001 0.42 (0.22, 0.76), P=0.002 0.37 (0.20, 0.67), P<0.001 0.41 (0.19, 0.82), P=0.006
Adjusted risk of outcome at a visit§ (95% CI) 6.6% (4.7, 8.5) 9.7% (8.2, 11.2) 1.6% (0.7, 2.5%) 2.5% (1.8, 3.2%) 2.1% (0.8, 3.5) 5.0% (3.7, 6.4) 1.0% (0.3, 1.7) 1.9% (1.2, 2.5)
Adjusted risk difference at a visit§ (95% CI) −3.1% (−5.8, −0.4), P=0.023 −0.91% (−2.1, 0.3), P=0.143 −2.9% (−5.0, −0.1), P=0.007 −0.9% (−2.0, 0.2), P=0.097
GEE logistic regression aOR (95% CI) 0.62 (0.40, 0.96), P=0.032 0.61 (0.30, 1.25), P=0.178 0.40 (0.18,0.87), P=0.020 0.50 (0.20,1.25), P=0.138
Cox regression aHR (95% CI) 0.56 (0.35, 0.89), P=0.013 0.47 (0.21, 1.07), P=0.072 0.38 (0.17, 0.83), P=0.015 0.47 (0.18, 1.22), P=0.121
Weighted marginal structural model aOR** (95% CI) 0.42 (0.21, 0.81), P=0.010 0.54 (0.25, 1.19), P=0.126 0.43 (0.16, 1.21), P=0.112 0.37 (0.15, 0.89), P=0.026

ART = antiretroviral therapy; EFV = efavirenz; NVP = nevirapine; PY = person-year; IR = incidence rate; CI = confidence interval; IRR = incidence rate ratio; GEE = generalized estimating equations; aOR = adjusted odds ratio; aHR = adjusted hazard ratio

*

Comparing participants in ever-EFV vs only-NVP groups

Attributed events and person-time accounting for time-varying exposure; corresponding measures that follow also account for time-varying exposure

All measures of effect and association comparing efavirenz use to nevirapine use

§

Adjusted risk and risk differences based on GEE model estimates

In GEE model, efavirenz use was treated as a time-varying exposure and adjusted for the following time-dependent covariates: age, sex, marital status, educational attainment, household asset ownership, CD4+ T-lymphocyte count, viral suppression, ART duration, health status, and heavy drinking; and the following time-independent covariates: year of enrollment, probable depression at enrollment, suicidal ideation at enrollment, and tuberculosis co-infection.

In Cox model, efavirenz use treated as a time-varying exposure and adjusted for the following baseline, time-independent covariates: age, sex, marital status, educational attainment, household asset ownership, CD4+ T-lymphocyte count, viral suppression, RT duration, health status, heavy drinking, year of enrollment, probable depression, suicidal ideation, and tuberculosis co-infection

**

See methods appendix for details on model specification