This was a cross-sectional analysis of PWHIV receiving care at participating hospital trusts using Climate-HIV, an electronic patient record system used to manage PWHIV, conducted from December 2012 through March 2018. Climate-HIV is updated in real time and contains patient information recorded by the multidisciplinary team, including referral letters, diagnostic information, and drug histories. Patient information is collected and manually entered by the specialist pharmacist during the clinic appointment. For this study, data were included from North Middlesex University Hospital NHS Trust, Nottingham University Hospitals NHS Trust, Birmingham Heartlands Hospital, and Homerton University Hospital NHS Foundation Trust.
Potential interactions between ARV medications and guideline-recommended therapies for other conditions were reported [e.g. guideline treatment recommendations for diabetes include metformin, which the University of Liverpool DDI tool indicates has no interaction with the ARV medication efavirenz (EFV)]. The proportion of individuals with documented health conditions as recorded by their physician in Climate-HIV was also reported.
All PWHIV included in the study had a recorded diagnosis of HIV-1; were aged ≥ 18 years at the index date; were registered at the clinic (or had a record within the 18 months before the index date); were prescribed an ARV regimen within the year before the index date; and had a current ARV prescription in February 2018. The study period included data from the time of patient record entry into Climate-HIV. The index date was the most recent record in Climate-HIV and was required to be within 18 months before data extraction. The most recent record rather than the most recent ARV prescription was used in the analyses, allowing inclusion of potential concomitant non-ARV therapies prescribed after or added to the last ARV medication.
Assessment of DDIs
All 'current' medications/drug treatments were included in the analysis of DDIs. Drug–drug interactions between ARV and non-ARV medications were determined using University of Liverpool HIV interactions database and categorized as red (do not co-administer), amber (potential interaction; may be managed by dose adjustment or monitoring), yellow (potential interaction likely to be of weak intensity; unlikely to require dose adjustment or additional monitoring), and green (no interaction expected).
Drug–drug interaction charts by ARV class and comedication were cross-tabulated against medications observed at the patient and population level using R version 3.5.1(The R Foundation for Statistical Computing, Vienna, Austria).
Patient characteristics, including sex, ethnicity, and age, as well as time from HIV diagnosis to index date, were summarized using descriptive statistics [median and interquartile range (IQR); Table 1].
Patient- and Interaction-level DDIs. The number of non-ARV therapies at baseline was described, in addition to the number of concomitant therapies [category, mean (SD), and median (IQR)]. The number and type of health conditions were similarly determined [n (%), mean (SD), and median (IQR)]. The number of patients without any recorded non-ARV medications was also determined. For patients with non-ARV medications, DDIs with ARV medications were identified, and the analyses were stratified by age group (< 50 and ≥ 50 years). The proportions of the study population with red, amber, yellow, and green interactions were reported, as well as the proportion of interactions for each regimen of each colour.
A stepwise logistic regression model was derived to determine which factors increased the odds of a red or amber interaction based on the observed ARV/non-ARV medications in the overall population. Because of the low frequency of red interactions observed in the study population, red and amber interactions were given an equal weighting in the model. Age (< 50 and ≥ 50 years), sex, current ARV regimen, number of concomitant drugs, and the presence of any comorbidity or concurrent health condition were included in the baseline model. Stepwise selection was then used to choose other variables for selection. Initially, individual comorbidities/health conditions and ethnicity were also included; however, these variables were removed from the final model because of under-reporting and collinearity, respectively. Aside from ethnicity, no other variables exhibited collinearity. The model with the lowest Akaike information criterion was selected.
Theoretical DDIs. To evaluate the propensity for DDIs in PWHIV when adhering to clinical guidelines for treating common comorbidities/health conditions, we first identified those occurring at high prevalence in the UK Climate-HIV database as follows: type 2 diabetes, hypertension, hyperlipidaemia, chronic obstructive pulmonary disease, asthma, chronic kidney disease, hepatitis B, hepatitis C, malignancy, tuberculosis, cardiovascular disease, peripheral neuropathy, osteoporosis, and mental health disorders (depression, anxiety, bipolar disorder, schizophrenia, mania, and psychosis). Next, we selected recommended first-line comedications based on adherence to UK National Institute for Health and Care Excellence (NICE) treatment guidelines. Finally, we evaluated the likelihood of DDIs with the ARV regimens currently used in our study population, using the University of Liverpool DDI tool. For the high-prevalence comorbidities, the proportion of possible interactions [n (%) of red, amber, yellow, and green interactions for each condition] with the observed medication groups was determined. This analysis represented theoretical DDIs encountered if neither the ARV regimen nor treatment for comorbidities was changed.
HIV Medicine. 2020;21(8):471-480. © 2020 Blackwell Publishing