Appendix Table 1.
Sensitivity analyses for GEE and Cox models
| Analysis | Depression | Suicidal ideation | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GEE | Cox | GEE | Cox | |||||||||
|
AOR (95% CI) |
P-value | N |
AHR (95% CI) |
P- value |
N |
AOR (95% CI) |
P-value | N |
AHR (95% CI) |
P- value |
N | |
| Primary | 0.62 (0.40–0.96) |
0.032 | 694 | 0.56 (0.35–0.89) |
0.013 | 654 | 0.61 (0.30–1.25) |
0.178 | 694 | 0.47 (0.21–1.07) |
0.072 | 654 |
| Excluding baseline probable depression | 0.40 (0.18–0.87) |
0.020 | 466 | 0.38 (0.17, 0.83) |
0.015 | 473 | -- | -- | -- | -- | -- | -- |
| Excluding baseline suicidal ideation | -- | -- | -- | -- | -- | - | 0.50 (0.20–1.25) |
0.138 | 648 | 0.47 (0.18, 1.22) |
0.121 | 609 |
| Excluding some EFV use* | 0.55 (0.35–0.88) |
0.012 | 692 | 0.55 (0.34–0.89) |
0.015 | 643 | 0.55 (0.26–1.15) |
0.111 | 692 | 0.39 (0.16–0.91) |
0.029 | 645 |
| Excluding those with tuberculosis co-infection | 0.65 (0.41–1.04) |
0.072 | 660 | 0.61 (0.38–0.98) |
0.040 | 622 | 0.71 (0.34–1.47) |
0.36 | 660 | 0.56 (0.25–1.28) |
0.168 | 622 |
| Excluding pregnant women, restricted to females | 0.49 (0.29–0.83) |
0.008 | 394 | 0.58 (0.33–0.99) |
0.048 | 378 | 0.41 (0.15–1.11) |
0.078 | 394 | 0.34 (0.12–1.03) |
0.056 | 378 |
| Missing visits imputed as positive for depression or suicidal ideation† | 0.54 (0.42–0.71) |
< 0.001 | 694 | -- | -- | -- | 0.55 (0.41–0.73) |
< 0.001 | 694 | -- | -- | -- |
| Missing visits imputed as negative for depression or suicidal ideation | 0.73 (0.48, 1.12) |
0.153 | 694 | -- | -- | -- | 0.69 (0.34, 1.41) |
0.310 | 694 | -- | -- | -- |
| Lost to follow-up‡ coded as positive for depression or suicidal ideation at last visit | 0.61 (0.40–0.91) |
0.015 | 694 | 0.58 (0.38, 0.87) |
0.008 | 654 | 0.62 (0.34–1.12) |
0.114 | 694 | 0.53 (0.30, 0.93) |
0.026 | 654 |
| Lost to follow-up coded as negative for depression or suicidal ideation at last visit§ | 0.62 (0.40, 0.97) |
0.038 | 694 | -- | -- | -- | 0.51 (0.20–1.27) | 0.149 | 694 | -- | -- | -- |
| Limiting analysis to 1 year after ART initiation | 0.52 (0.31–0.87) |
0.013 | 693 | 0.52 (0.31–0.87) |
0.013 | 652 | 0.53 (0.21–1.34) |
0.181 | 693 | 0.52 (0.20–1.36) |
0.184 | 652 |
| Extending analysis to full follow-up period of study | 0.73 (0.51–1.03) |
0074 | 699 | 0.77 (0.53–1.11) |
0.162 | 658 | 0.52 (0.28–0.96) |
0.037 | 699 | 0.44 (0.23–0.84) |
0.013 | 658 |
| Including a spline term for ART duration∥ | 0.64 (0.41–0.98) |
0.042 | 694 | -- | -- | -- | 0.57 (0.28–1.15) |
0.117 | 694 | -- | -- | -- |
| Including term for internalized stigma score¶ | 0.58 (0.35–0.96) |
0.035 | 640 | 0.50 (0.26–0.94) |
0.032 | 380 | 0.60 (0.27–1.33) |
0.21 | 640 | 0.44 (0.17–1.16) |
0.096 | 380 |
GEE = generalized estimating equation; AOR = adjusted odds ratio; AHR = adjusted hazard ratio; N = sample size; EFV = efavirenz; ART = antiretroviral therapy
Some EFV use is defined as a prescription of EFV in the previous 90 days but not meeting the criteria for EFV exposure (receipt of an EFV prescription in the 60 out of previous 90 days and previous 7 days)
Missing visits were filled in between observed visits for the 694 participants in the primary analysis. 1 missed visit was also added after the last recorded visit if the last recorded visit was >3.5 months before the end of the two year study. Exposure and covariate values for missing visits were imputed using a last observation carried forward approach.
Lost to follow-up was defined as a participant whose last visit within two years was 550 days or fewer after ART initiation
This sensitivity analysis was not done for the Cox model, as the analysis would be the same as that of the primary Cox model
ART duration was fit as a natural cubic spline with knots at the 5th, 25th, 50th, 75th, and 95th percentiles
The Internalized AIDS-related stigma score (40) was defined as a continuous covariate (range 0–6).