Table 1. Model parameters and data sources.
| Outcome | Source | Statistical model | Starting | Value (95% CI) | Dimension | Risk |
| Time to virologic failure | ||||||
| First-line ART; second-line ART withimmediate switch | Cohorts | Parametric Weibull | 3 months fromART start | 0.47 (0.43–0.50) | Shape | 5.6% fail by 1 year after ART start |
| 3.30 (2.77–3.95) | Scale (100 years) | |||||
| Resistance penalty | [11] | *) | n/a | 0.05 (0.00–0.20) | Decrease in ART efficacy | n/a |
| Time to immunologic failure | ||||||
| After virologic failure | Cohorts | Parametric exponential | Virologic failure | 0.08 (0.06–0.10) | Rate (years−1) | 7.6% fail by 1 year after virologic failure |
| Before virologic failure | Cohorts | Parametric Weibull | 3 months fromART start | 0.22 (0.20–0.25) | Shape | 3.0% fail by 1 year after ART start |
| 5.46 (3.14–9.51) | Scale (106 years) | |||||
| Time to death and LTFU | ||||||
| Non-HIV related mortality, men | ASSA2008 [25] | No specific model**) | Birth | 67 | Median (years) | 21% die by age of 50 |
| Non-HIV related mortality, women | ASSA2008 [25] | No specific model**) | Birth | 72 | Median (years) | 13% die by age of 50 |
| HIV-related observed mortality | Cohorts and ASSA2008 [25] ***) | Double Weibull****) | ART start | 0.92 (0.92–0.92) | Shape 1 | 8.4% have died 1 year after ART start |
| 0.30 (0.30–0.30) | Scale 1 (years) | |||||
| 1.00 (1.00–1.00) | Shape 2 | |||||
| 124.25 (121.27–127.31) | Scale 2 (years) | |||||
| 0.06 (0.06–0.06) | Weight (1st component) | |||||
| LTFU | Cohorts | Double Weibull****) | ART start | 0.94 (0.94–0.94) | Shape 1 | 4.2% LTFU 1 year after ART start |
| 1.00 (1.00–1.00) | Scale 1 (years) | |||||
| 25.45(25.45–25.45) | Shape 2 | |||||
| 66.19(66.19–66.19) | Scale 2 (years) | |||||
| 0.07 (0.07–0.07) | Weight (1st component) | |||||
| Extra hazard after immunologic failure | Cohorts | Cox regression | Immunologic failure | 1.75 (1.15–2.67) | HR, constant over time | n/a |
| Extra hazard after virologic failure | Cohorts | Cox regression | Virologic failure | 1.07 (0.98–1.18) | HR per 3 months on failing ART | n/a |
| Observed delay in switching | ||||||
| After virologic failure | Cohorts | Parametric exponential | Virologic failure | 0.75 (0.63–0.89) | Rate (years−1) | 53% switched 1 year after virologic failure |
| After immunologic failure | Cohorts*****) | Parametric exponential | Immunologic failure | 0.06 (0.05–0.08) | Rate (years−1) | 6% switched 1 year after immunologic failure |
Distributions of times to event were assumed to be exponential, Weibull or double Weibull, based on the cohort data. Cohort data are from the Khayelitsha and Gugulethu ART programmes in Cape Town, South Africa, unless otherwise specified.
CI, confidence interval; ART, antiretroviral therapy; HR, hazard ratio; ASSA, Actuarial Society of South Africa; LTFU, loss to follow-up; n/a, not applicable.
*
)Relative decrease in second-line efficacy per year spent on failing first-line ART.
**
)Age-specific mortality rates.
***
)Non-HIV related mortality estimated from the ASSA2008 model deducted from cohort data on all-cause mortality.
****
)Weighted sum of two Weibull distributions.
*****
)Data from Ministry of Health-Centre for Infectious Disease Research in Zambia.