Study Population and Design
A retrospective, observational cohort study was performed of PLWH treated between April 1, 2010, and April 1, 2020, in the University Medical Center Utrecht (Utrecht, the Netherlands), a tertiary hospital. PLWH were identified through a laboratory database of all PLWH with VL measurements in the aforementioned period and were considered eligible if they were aged 18 years or older; had received cART consisting of 2 NRTIs and an NNRTI, a PI, or an INSTI; and had achieved virologic suppression (plasma VL <50 copies/mL). PLWH with elite controller status, HIV-2, or a registered objection to data research were excluded together with those without VL measurements after suppression because this precluded us from evaluating the occurrence of blips. When PLWH switched to a different eligible treatment course during the study period, each treatment course was analyzed separately. During every evaluable treatment course, ≥1 VL measurements must have been performed after achievement of virologic suppression. Treatment courses in which PLWH had detectable viremia and were deemed nonadherent to their cART, as based on the documented conclusion made by the treating physician in the medical charts, were excluded from the analysis. When no remarks regarding treatment adherence were made, the treatment course was not excluded. When nonadherence existed only in a specific and well-documented period, the follow-up was split into a treatment course before the period of nonadherence and a course after this period. Follow-up of treatment courses ended on regimen switch, cART discontinuation, the last VL recorded before loss to follow-up/ending of the study period, or the occurrence of a virological end point [persistent low-level viremia (PLLV)/virologic failure]. However, when the last VL before ending of the study period was 50–499 copies/mL, subsequent measurements up to September 1, 2020, were examined to determine whether this viremia was transient and thus considered a blip. The Medical Research Ethics Committee Utrecht judged that the study met with the criteria to be exempt from formal review.
Measurements and Case Definitions
The electronic medical records of all PLWH were reviewed to extract information on demographics, biochemical data, and therapy adherence. When multiple potential HIV transmission routes existed, the source with the highest risk of transmission was collected (eg, intravenous drug use and not sexual transmission in men who have sex with men, when both factors were present). For pretreatment CD4+ count and VL, the last available values were collected up to 1 year preceding course initiation. For PLWH with missing nadir CD4+ counts, the lowest available CD4+ count was used. VLs were determined using the Roche COBAS TaqMan v2.0 during the entire study period.
The primary study outcome was defined as the difference in blip incidence when comparing PLWH treated with different cART anchors. A blip was defined as a VL of 50–499 copies/mL, preceded and followed by a VL < 50 copies/mL without a change in anchor. Dose modifications were not considered a change in therapy. Multiple measurements within 30 days meeting the blip definition were considered a single blip where the highest VL was recorded. Virologic failure was defined as 2 consecutive VLs ≥ 200 copies/mL more than 30 days apart or a single VL ≥ 500 copies/mL.[19,20] Two consecutive measurements ≥ 50 copies/mL more than 30 days apart not meeting the failure criteria were considered PLLV. In PLWH demonstrating a single VL of 50–499 copies/mL before loss to follow-up, no distinction between a blip and either PLLV or virologic failure could be made, and they were censored at their last undetectable VL. Similarly, treatment courses in which PLWH demonstrated a single VL of 50–499 copies/mL immediately before anchor switch were censored at the last undetectable VL because these VLs did not meet the blip criterion without a "change in anchor." In PLWH experiencing blips, the scheduling of extra visits and/or laboratory testing because of the blip were recorded.
The Fisher exact test and the Pearson χ2 analysis were used to compare categorical variables, and the Mann–Whitney U test was used for continuous variables. CD4+ counts were square root transformed to normalize their distribution for analysis. Multiple imputation was performed to minimize potential bias resulting from missing variables. The variables that served as predictors for multiple imputation were sex, transmission route, calendar year of diagnosis, age at diagnosis, time since ART initiation, Fiebig stage at ART initiation, treatment period, pretreatment and lowest available CD4+ count, pretreatment and zenith VL, anchor, and the number of blips. Multiple imputation was used to fill in data for all variables with missing data that were included in the model described further. Virologic end points (ie, blips, PLLV, and virologic failure) were never imputed. Twenty-five imputed data sets were constructed using the SPSS automatic method. Associations with blip incidence were investigated using negative binomial regression models with an offset for the differential follow-up duration of the individual treatment courses. Generalized estimating equations with an exchangeable correlation structure were used to account for participant clustering. Factors generally considered to be of interest in blip incidence (Fiebig stage, VL test frequency,[5,15] pre-ART VL,[5,7] and pre-ART CD4+ count) and all other variables with P < 0.20 in unadjusted analyses were included in the multivariable model. Zenith VL and lowest available CD4+ count, and not the pretreatment values, were selected for the multivariable model because these factors were expected to have the highest influence on the viral reservoir as a potential source for blips. Blips were categorized as low level (50–199 copies/mL) or moderate level (200–499 copies/mL) to explore their potential relationship with PLLV and virologic failure. Results were expressed as incidence rate ratios (IRRs) and their 95% confidence interval (CI) with 2-tailed P values < 0.05 being considered statistically significant. All statistical analyses were performed using IBM SPSS Statistics v24.
J Acquir Immune Defic Syndr. 2022;89(5):575-582. © 2022 Lippincott Williams & Wilkins