Adverse Drug Reactions to Integrase Strand Transfer Inhibitors

Katherine J. Lepik; Benita Yip; Ana C. Ulloa; Lu Wang; Junine Toy; Linda Akagi; Viviane Dias Lima; Silvia Guillemi; Julio S.G. Montaner; Rolando Barrios


AIDS. 2018;32(7):903-912. 

In This Article


Study Design and Setting

The study cohort included HIV-1-infected persons in British Columbia, Canada, who received ART through the publicly-funded British Columbia Centre for Excellence in HIV/AIDS (BC-CfE) Drug Treatment Program (DTP).[20] Persons aged at least 19 years were included if they received their first prescription for raltegravir, elvitegravir co-formulated with cobicistat (elvitegravir-cobicistat), or dolutegravir between 1 January 2012 and 31 December 2014, and were excluded if the INSTI was initiated during a clinical trial or outside British Columbia. Both ART-experienced and naive persons were included, and individuals could contribute data for more than one INSTI. Baseline demographic and clinical variables, ART dispensing, and pharmacovigilance data were abstracted from BC-CfE databases. Ethical approval was obtained from the University of British Columbia, Providence Healthcare Research Ethics Board.

Follow-up and Censoring

Adverse drug reaction data collection continued until 31 December 2016, providing at least 2 years follow-up opportunity. Persons were followed until they experienced an INSTI ADR resulting in INSTI discontinuation, or were censored at the earliest occurrence of INSTI discontinuation for other reason, patient relocation outside British Columbia, death, or 24 months after INSTI start. Changes to INSTI tablet formulation or dose were not considered discontinuations.

For each person, INSTI treatment status at 24 months was summarized as ongoing, stopped due to INSTI ADR, stopped for other reason, or person deceased, moved, or lost to follow-up.

Primary Outcome: Adverse Drug Reaction

The primary outcome was any INSTI ADR resulting in INSTI discontinuation. ADR reports were abstracted from the BC-CfE Pharmacovigilance database. This longitudinal, population-based surveillance program collects clinician-reported antiretroviral ADRs through a system integrated with ART prescribing, and maintains individual ADR histories for DTP participants.[21] ADR attribution to specific antiretrovirals is recorded as documented by the clinician.

To apply consistent ADR evaluation criteria, a causality review of INSTI ADR reports was undertaken by study co-authors who were not involved in Pharmacovigilance data collection. De-identified ADR records, blinded to INSTI drug, cobicistat, and year were independently reviewed by two HIV clinic pharmacists (J.T., L.A.). Reviewers applied WHO causality assessment criteria[22] to classify each ADR as 'possibly' (WHO categories 'possible' or higher) or 'unlikely' causally associated with the INSTI exposure. Discordant assessments were resolved by consensus.

Adverse drug reaction reports were qualified as events in this study if the reporter implicated the INSTI, and if the ADR resulted in INSTI discontinuation within 2 years after start. ADR reports were censored as nonevents if causality was unclassifiable due to insufficient information, or if the ADR was assessed as 'unlikely' to be associated with the INSTI.

Other Outcomes

For those with ADR-related INSTI discontinuation, occurrence of ART interruption (defined as clinician-documented interruption and/or plasma viral load rebound >1000 copies/ml at the time of ADR report) and next ART regimen were summarized descriptively. Post-ADR re-exposure to the same INSTI and switches to a different INSTI were assessed for tolerability up to May 2017. Other reasons for stopping the INSTIs were summarized according to predefined categories in the DTP database.

Exposure and Confounder Variables

The main exposure variable was INSTI drug. Baseline variables that could potentially influence ADR occurrence or reporting were summarized by descriptive statistics and considered as adjustment variables in the confounder model. These included age, biological sex, hepatitis C antibody status, and prior ART exposure at the time of INSTI start. Plasma viral load and CD4+ cell count values were measured within 6 months prior to INSTI start, with missing values imputed (methods described elsewhere[23]). For treatment-experienced persons, the reasons for switching to the INSTI, and antiretroviral ADR history were also recorded.

Antiretrovirals prescribed concurrently with the INSTI were categorized as either two nucleoside (-tide) reverse transcriptase inhibitors (NtRTIs): tenofovir disoproxil fumarate-emtricitabine/lamivudine, or abacavir-lamivudine, or as 'other antiretroviral combination.' To account for the possible confounding effect of simultaneous, multiple antiretroviral changes on ADR assessment, it was also noted whether or not the INSTI regimen included at least one other new antiretroviral. The pharmacokinetic enhancer cobicistat was only available in co-formulation with elvitegravir during the study period, and was not counted as a new antiretroviral in elvitegravir-containing regimens.

To adjust for the potential influence of INSTI availability on ART prescribing and ADR-related therapy changes, persons were classified according to INSTI start-time period, based on INSTI availability: Period 1 (raltegravir) 1 January 2012 to 30 April 2013, period 2 (raltegravir, elvitegravir-cobicistat) 1 May 2013 to 31 December 2013, and period 3 (raltegravir, elvitegravir-cobicistat, dolutegravir) 1 January 2014 to 31 December 2014.

Statistical Analyses

Baseline clinical and demographic variables were summarized by descriptive statistics and compared across INSTI categories using Pearson's chi-square or Fisher's exact test for categorical, and Kruskal–Wallis test for continuous variables.

For each INSTI, ADRs were summarized by symptom detail and symptom class. Multiple symptoms within a class were counted once per class, per person. ADR symptom classes were compared across INSTIs for all ADRs and for late (second year) ADRs using Pearson's chi-square or Fisher's exact test.

Crude INSTI ADR rates and 95% confidence intervals (95% CIs, Poisson method) were calculated per 100 person-years follow-up. Median (25th–75th percentile, Q1-Q3) follow-up time and time to INSTI ADR were calculated to censor date, and to date of ADR-related INSTI discontinuation, respectively. Time from INSTI start to ADR-related discontinuation was plotted by the Kaplan–Meier method, with the log-rank test used to test for a statistically significant difference (P < 0.05) between INSTIs. Poisson regression was used to explore associations between baseline variables and INSTI ADR rates.

Cox proportional-hazards regression was used to estimate the hazard ratio for ADR-related INSTI discontinuation. To allow comparison among all three INSTIs, both raltegravir and dolutegravir were assigned as the INSTI reference category. The robust sandwich covariance matrix estimate for clustered data structure was used to account for persons who contributed data for more than one INSTI.[24] Potential confounder variables were included in the full model and sequentially removed using a change-in-estimate strategy, whereby variables were removed if their exclusion resulted in a maximum change of less than 5% in the estimated INSTI coefficients in the reduced model.[25,26] Age and sex were included in the model a priori. Covariates were tested for co-linearity. Fulfillment of the proportional hazards assumption was assessed by visual inspection of Kaplan–Meier plots and by the weighted residuals score test.[27] Statistical analyses were performed using SAS v9.4 (Cary, North Carolina, USA) and R v.3.3.2 (Vienna, Austria).