Rare Outcome Analysis in Longitudinal Real-World Data: Methodological Insights From Switching HIV Treatment to 2 Versus 3-Drug Regimen in Sweden

OBJECTIVES: Analyzing longitudinal HIV outcomes in patients who switch treatment poses an extra challenge to real-world data (RWD) analysis due to selection bias and the rarity of the outcomes, in addition to common issues like missing data and small patient numbers in long-term follow-up. This study used a combination of statistical methods to assess the long-term outcomes of switching to the 2-drug regimen dolutegravir and lamivudine (DTG/3TC) versus various 3-drug regimens (3DR), in Sweden.

METHODS: Data from the Swedish National Quality Registry for HIV (InfCareHIV) included all antiretroviral therapy-experienced people living with HIV, switching to either DTG/3TC or 3DR between 07/2019-05/2023. The study aimed to determine the associations between treatment, demographic, and clinical variables on virological failure (VF) at several time points over a 3.5-year period. A logistic generalized estimating equations (GEE) model was employed to handle correlated longitudinal data. Cox-regression was used to evaluate the overall time to VF. Sensitivity analyses improved robustness of the findings by using penalization for rare outcomes.

RESULTS: Over 3.5 years 1125 individuals switched to DTG/3TC and 1336 to 3DR. The number of VF was low with the highest incidence at any one time being 2 in DTG/3TC and 12 in 3DR across all timepoints. In the GEE modelling, no significant difference in VF between DTG/3TC and 3DR was found. Older age (p=0.005) and higher baseline CD4 count (p<0.001) were significantly associated with decreased VF risk, while viral blips post-switch and suboptimal adherence increased VF risk (p<0.001). Results from Cox-regression and sensitivity analyses supported the findings of the primary analysis.

CONCLUSIONS: The findings support growing evidence of DTG/3TC as a viable simplification strategy for virologically suppressed people living with HIV in routine clinical care. Methodological strategies effectively addressed the challenges of rare outcome analysis. However, controlling for unmeasured variables affecting treatment selection remains a challenge.