Background: In Africa, for people with HIV on a dolutegravir-based regimen with a viral load of more than 1000 copies per mL despite enhanced adherence counselling, the appropriate course of action is uncertain. We aimed to evaluate the predicted effects of alternative antiretroviral regimen switching options in this population, including consideration of cost-effectiveness.

Methods: We used an existing individual-based model to simulate risk and experience of HIV in 100 000 adults alive between 1989 and 2076. Using sampling of parameter values, we created 1000 setting-scenarios, reflecting the uncertainty in assumptions and a range of settings similar to those seen in eastern, central, southern, and western Africa. For each setting-scenario, we predicted the outcomes from the three alternative policies for people with sustained viral load non-suppression on a dolutegravir-containing regimen from 2026: a switch to a protease inhibitor-based regimen (switch policy), a switch to a protease inhibitor-based regimen only if HIV drug resistance testing beforehand shows integrase inhibitor resistance (resistance test policy), and no switch with no HIV drug resistance test (no switch policy). We considered predicted outcomes over 10-year and 50-year periods from 2026, used a 3% discount rate, and a cost-effectiveness threshold of US$500 per disability-adjusted life-year (DALY) averted. Ritonavir-boosted darunavir costs $210 per year, and dolutegravir less than $20. We assumed a cost of HIV drug resistance testing of $200 and considered variations around this. For comparing policies, we calculated net DALYs, which account for the health consequences of differences in costs and provide a measure of the impact of a policy on overall population burden of disease.

Findings: Across setting-scenarios, there was a mean of 14 480 deaths per year (95% CI 13 750-15 210) over 50 years with a mean annual discounted cost of $103·2 million (95·8-106·5) with the switch policy in the context of having scaled to a setting with an adult population of 10 million in 2024. Compared with the switch policy, the no switch policy was predicted to lead to an overall increased number of DALYs incurred (mean 4400 per year, 95% CI 3200-5500), although it resulted in the lowest overall cost, with a difference in annual discounted costs of $5·1 million (95% CI 4·6-5·6) lower than the switch policy. The resistance test policy led to a similar risk of death and DALYs to the switch policy at a lower overall cost (difference in annual discounted costs $3·5 million per year, 95% CI 3·1-3·9), leading to 6900 (95% CI 5500-8200) fewer net DALYs per year. Net DALYs for the resistance test versus no switch policies were similar (-1000 net DALYs, 95% CI 400 to -2300). The incremental cost-effectiveness ratio when comparing the resistance test policy with the no switch policy was $376 per DALY averted; the switch policy was dominated.

Interpretation: Introduction of HIV drug resistance testing for people with sustained viral load non-suppression on dolutegravir-based antiretroviral therapy is likely to be cost-effective. We suggest that exploratory planning for increased access and scale-up of high-quality, low-cost drug resistance testing for the region is undertaken.

Funding: Gates Foundation as part of the HIV Modelling Consortium.