Incidence, predictors and reasons for initial regimen modifications in patients on antiretroviral therapy in Witbank, South Africa, 2003-2017

Background Antiretroviral therapy (ART) is associated with unpleasant adverse effects that may require modification of regimens. ART modifications may lead to poor treatment outcomes. We determined the incidence, reasons and predictors for modification of initial ART regimens. Methods We retrospectively analysed data from Witbank pharmacovigilance sentinel site, South Africa. Censoring targeted the first incident of ART modification from the initial regimen. We included human immunodeficiency virus (HIV)-infected patients on ART, aged more than 18 years. We used the Cox-proportional hazard model to identify predictors for changing initial ART regimens. Results Among 2, 045 eligible patients, 38% (n=783) had their initial ART regimens changed. The overall incidence rate of ART modification was 10.0 per 100 person-years within a follow-up period of 7 794.6 person-years (PYs). Reasons for changing were adverse drug reactions (ADRs) (60%), prescriber’s decisions (37%), drug toxicity (26%) and treatment failure (12%). The most commonly changed regimens were stavudine (68%) and zidovudine (44%) based regimens. Stavudine-based regimen had the highest changing rate of 13.6 per 100 PYs compared to zidovudine (8.0 per 100 PYs) and tenofovir (6.5 per 100 PYs). Using tenofovir as reference, stavudine (aHR 2.3; 95% CI 1.8-2.9; p<0.001) and zidovudine (aHR 1.5; 95% CI 1.2-3.2; p<0.001) based regimens were significantly associated with regimen modifications. The predictors for changing ART regimens included drug toxicity (aHR 2.6; 95% CI 2.1-3.1), ADRs (aHR 2.1; 95% CI 1.3-3.2), treatment failure (aHR 2.0; 95% CI 1.5-2.4), baseline cd4 count of ≥200 (aHR 1.7; 95% CI 1.3-2.1) and initiation regimens (stavudine and zidovudine). Conclusion The findings were suggestive of a moderate incidence of initial ART regimen changing. Patients on stavudine and zidovudine based regimens changed primarily due to ADRs and drug toxicity. We recommended that clinicians should consider changing patients who are still on stavudine-containing regimens, however, changing should be individualized.


Human immunodeficiency virus (HIV) infection is a retroviral infection that causes acquired
The major drivers for regimen modifications world-wide included treatment failure and adverse drug 71 reactions (ADRs) [17][18][19][20]. With improvement in ART over years, the factors associated with 72 modifying regimens changed from merely treatment failure and drug resistance to several factors like 73 non-adherence, drug interactions and drug related toxicity [14]. However, drug toxicity still appears 74 to be the main reason for changing ART regimens. Most toxicity was found with stavudine (d4T) and 75 zidovudine (AZT) containing regimens [3,12]. The regimens can be switched for convenience, 76 tolerability, simplification of treatment, potential new drug interactions, pregnancy and plans for 77 pregnancy, elimination of food restrictions, virological failure and drug toxicities [1, 8-9, 11, 18].

79
Patients on d4T, didanosine or AZT-containing regimens may also have been changed even when 80 they were doing well and were satisfied with their treatment to prevent the long-term toxicity 81 associated with these drugs [3,13]. The factors to be considered before a regimen is modified, include  Other studies conducted in South Africa, have reported higher rates of treatment modification in 88 stavudine and zidovudine containing regimens compared to tenofovir-containing regimens [20,22].

89
In order to ensure antiretroviral drug safety, it is essential to understand drug effects, prescribing 90 patterns, and regimen modification patterns (and reasons). Identification of risk factors for early 91 treatment modification is essential to inform/guide treatment policies and guidelines. The aim of the 92 study was to determine the incidence of, predictors and reasons for changing of initial antiretroviral 93 treatment regimens.

96
Study design and setting 97 We conducted a retrospective cohort study which entailed analysis of secondary data collected from   The study excluded patients less than 19 years of age, those without recorded initiation date, missing 127 regimen on initiation, registered in 2018 and ART changing date earlier than initiation date. The study    We defined treatment discontinuation as stopping any antiretroviral drug or regimen temporarily or 158 permanently due to toxicity. Regimen changing/modification/switching was defined as changing the 159 ART regimen from one standardised regimen to another or to a non-standardised (individualised) 160 regimen. Substitution of a drug was defined as replacement of one or more drugs within the regimen 161 either standardised or non-standardised to another more potent drug. 167 Data analysis 169 We described categorical variables using proportions and frequencies and compared them using 170 Pearson's chi square test. We calculated the frequency of initial ART modification and related 171 characteristics. We calculated proportions to determine the frequency of adverse effects of different 172 treatment regimens, and related characteristics. The Fisher's exact test was used to test for statistical 173 significance of differences in frequency of ADRs between drug regimens. We used Cox proportional 174 hazard model to determine predictors for changing ART regimens. We used a manual forward step-   Characteristics of the study population 203 Of the 2,045 eligible patients, 783 (38%) had their initial ART regimens changed (Fig 1) The frequencies of regimen changing for d4T-based, AZT-based, and non-standardised regimens 227 were 68% (n=490), 44% (n=93), and 34% (n=21), respectively. Majority of patients were on TDF-228 based regimens, however, the frequency of changing (17%, n=179) was lower than all other regimens.

229
Females accounted for 66% (n=521) of the patients who had their initial ART regimens changed.  (Table 2). There were 8% (n=158) who had their treatment discontinued 234 temporarily. Forty-one percent (n=64) of those who had their ART discontinued temporarily also had 235 their regimens changed.

246
Kaplan-Meier (KM) survival plots for the length of time after ART initiation until occurrence of the 247 first regimen change were presented for the different backbone NRTIs groups (Fig 2). There was a 248 significant difference in survival times between the backbone NRTIs groups (log rank test=0.001).

249
Patients on TDF-based regimens had a better survival on the initial ART regimen, than those on d4T 250 and AZT based regimens. Patients on d4T based regimens had the shortest survival time. Reasons for changing ART 255 The main reason for initial ART regimen changing was ADRs 60%, followed by prescriber's decision 256 37%, drug toxicity 26% and treatment failure 15%. Stavudine-based regimens accounted for the 257 majority of ART regimen changes due to toxicity (86%, n=176) and ADRs (81%, n=267).

258
Approximately half of the patients with treatment failure (58%) were initiated on d4T-based 259 regimens. Among the pregnant women who had their regimens changed, 59% (n=20) were on d4T-260 based regimens. Patients on tenofovir-based regimens were changed from a non-combined drugs to 261 a fixed-dose combination (FDC) pill. Guidelines related (7.5%) and unavailability of ART (2.6%) 262 were the other reasons for ART regimen changing (Table 3).  Kaplan-Meier plots of the probability of changing initial ART regimen show that patients who did 288 not experience ADRs had better survival rates on the initial ART regimen than those who experienced 289 ADRs (Fig 3).

290
Predictors of ART changing 291 There was evident increase in ART regimen changing attributable to ADRs, drug toxicity, treatment 292 failure and prescribed initiation regimens. We analysed the time to the first incidence on initial ART 293 regimen changing using Kaplan-Meier plots, stratified by backbone first-line antiretroviral regimens 294 (Fig 2). Using TDF-based regimens as reference, the results from Cox proportional hazard analysis and initiation regimens that was included in the Cox proportional hazard model (Table 5).  This study reports the occurrence of and the reasons for initial ART regimen changing in a The majority of the patients were initiated on TDF-based (51%) regimens followed by d4T-based 321 (35% AZT and TDF base regimens were 13.0, 8.0 and 6.5 respectively. We observed that TDF was more 347 tolerated than d4T and AZT. The significant difference in changing rates were further observed on 348 the KM plot that indicated the higher rate of changing the initial ART regimens in stavudine

352
With the improvement in ART over years, the reasons for regimen modification changed from merely 354 treatment failure and drug resistance to several factors [Günthard]. There are other studies that found ADRs playing the major role in ART changing [29][30]. In a 379 multivariate analysis, there were increased hazard ratios for regimen changes related to d4T, AZT 380 and non-standardised regimens. These findings are in line with other previous studies that have 381 reported that d4T is one of the most intolerable drugs, and TDF as the most tolerable [22,25]. The