Long-term adherence to antiretroviral therapy in a South African adult patient cohort: a retrospective study

Currently, an estimated 36.7 million people are living with HIV (PLWH) worldwide [1]. South Africa has the highest adult HIV prevalence (18%) in sub-Saharan Africa and the largest adult population of PLWH in the world, estimated at approximately 7.5 million people [2, 3]. In September 2016, universal test and treat policy was incorporated into the South African national antiretroviral therapy (ART) guidelines and the country’s government funded national antiretroviral (ARV) programme is now the largest globally, providing HIV care and ART to over 4 million people [4, 5].

Access to ART is only one aspect of an effective HIV management programme. Optimal adherence to ART is essential and early studies reported that ≥95% adherence to ART was required to achieve and maintain viral suppression [6, 7]. Although recent studies have shown that virologic suppression may still be achieved with < 95% adherence levels, this is dependent on the ART regimen, duration of treatment and previous ART exposure [8, 9]. Furthermore, repeated adherence levels of < 100% and treatment interruptions are associated with an increased risk of nucleoside reverse transcriptase (NRTI) and non-nucleoside reverse transcriptase (NNRTI) resistance which form the backbone of current first line ART regimens in South Africa [10‐12]. Poor adherence may lead to a number of adverse consequences on both individual and public HIV healthcare. Firstly, poorly adherent patients who acquire resistance to first line NNRTI-based regimens are switched to second and third line ARV regimens which are costlier, have a higher pill burden, increased dosing frequency and often have less tolerable side effects [13, 14]. Access to second line, third line and salvage treatment regimens are also limited or may be unavailable in developing countries [15]. Secondly, resistant HIV strains can be transmitted to others resulting in primary resistance to first line treatment regimens in newly infected ART-naïve patients or acquired resistance in infected patients already on ART [16, 17]. Thirdly, treatment failure due to non-adherence is associated with a greater risk of progression to Acquired Immune Deficiency Syndrome (AIDS) and mortality [18, 19]. In addition, hospital admissions due to HIV disease progression and opportunistic infections further add to the burden of public health costs in resource limited countries [20, 21]. Thus, it is still recommended that ≥95% adherence be maintained to achieve optimal viral suppression and prevent HIV resistance [22, 23].

Short term ART adherence rates (less than 2 years from treatment initiation) in South African patients from multiple urban HIV clinics is reported to range from 63 to 88% [24‐27], however, there is a paucity of information on long-term adherence in this population. As ART access continues to expand with the implementation of test and treat, our public health facilities face the challenge of retaining an increasing number of PLWH in life time care with limited healthcare budgets and insufficient human resources [28]. Initial studies in African test and treat patient cohorts have reported > 80% adherence levels in the first year post-ART initiation [29, 30]. Monitoring and determining long term adherence patterns can provide insight into observed treatment outcomes and assist in developing adherence support and intervention strategies that would be relevant to resource-constrained countries implementing a test and treat policy.

A total of 270 patients met the inclusion criteria for this analysis (Fig. 1). The median age was 34 years (IQR:29–40) and 54.8% were female (Table 1). At ART initiation the median CD4+ cell count was 145 cells/mm³ (IQR:76–249) and median viral load was 141,000 copies/mL (IQR:37757–386,000). Most patients (254/270) were classified as WHO stage 3 at baseline. A third of the cohort had completed secondary school, 60% were in full or part-time employment and more than half of all patients had disclosed their HIV status. The median duration on ART from initiation to completion of follow up was 70 months (IQR: 64–78).

The time period between patient exit from SAPiT and enrolment in TRuTH (where no pill count data available) was a median of 13 months (IQR: 5–21). The number of patients for whom pill count data was available in each 6 month period is shown in Table 2. Mean adherence was ≥95% for each year on ART. The majority of the cohort (94.8%) had an overall mean ART adherence of ≥95%. Of the 14 patients with sub-optimal adherence, nine had adherence estimates in the 90–95% range, two in the 80–90% range, two in the 70–80% range, and only one had less than 70% adherence. Mean adherence peaked (99.3%) at 18–24 months after ART initiation and remained above 98% from 32 months onward.

In the 6 months post-ART initiation, 91.8% of patients had optimal adherence. Half of all patients had at least one sub-optimal adherence measurement (< 95%) in the first 6 months after starting ART, less than 20% between the first and sixth year and 23% after 6 years on ART (Fig. 2). In the TRuTH study, patients with optimal adherence by pill count and an undetectable viral load (n = 267) were provided with an extended 3-month ART supply at a total of 1061 study visits. Overall mean adherence for patients who received a 3-month ART package was 99.5%.

Patient variables included in the univariable and multivariable GEE analyses are shown in Table 3. In the multivariable model, time since ART initiation (adjusted odds ratio [aOR] =1.10; 95%CI: 1.05–1.16), female gender (aOR = 1.33, 95%CI:1.05–1.67) and being a primary breadwinner (aOR = 1.32, 95%CI 1.02–1.72) were significantly associated with achieving optimal adherence.

Viral suppression rates were 87.4% 6 months after ART initiation and thereafter increased and remained above 92% throughout follow up (Fig. 2). Eleven patients had a viral load > 400 copies/mL after 5 years and 94% of patients were virologically suppressed at the end of the follow up period. Pill count showed high sensitivity in predicting viral suppression but had poor specificity when the viral load was detectable (Table 4).

Thirty-three patients experienced treatment failure with approximately half of all treatment failures (n = 17) occurred within 2 years after ART initiation. The cumulative probability of treatment failure was 10.7% at 5 years after ART initiation (Fig. 3). Adherence and viral load measurements prior to and after switching to a second line boosted LPV/r regimen was available for 21 of the treatment failure patients. The number of patients with optimal adherence improved and there was a significant increase in viral load suppression at 6 and 12 months after switching to the second line regimen (Table 5).

Of the 21 patients diagnosed and initiated on treatment for recurrent TB in the TRuTH study, three exited the study whilst on TB treatment and one did not complete their course of TB treatment. In the remaining 17 patients, there was a decline in optimal adherence and viral load suppression when patients were on both ART and TB treatment, but this was not statistically significant (Table 5).

Our study found overall high adherence to ART, determined by pill count, in this South African cohort followed up over a period of more than 5 years. Treatment outcomes were successful on first and second line treatment regimens, with 93.9% of patients virologically suppressed at study completion and only 12% experiencing treatment failure over the entire study period. The likelihood of achieving optimal adherence also improved with time on ART. Adherence during the first year of treatment was higher (98%) than those reported in several African studies, where adherence estimates ranged from 72 to 94% [25, 37‐41]. Over the long term, mean adherence was maintained ≥95% for each year of follow up. Long-term adherence studies in Botswana, Senegal and Nigeria also found comparably high adherence levels [42‐45]. Intensive pre-ART education and counselling sessions, an ongoing adherence support programme post-ART initiation, provision of pillboxes and use of a once-daily ART regimen, support measures that have been shown to assist adherence, may have played a role in helping to attain high adherence [46‐49]. Additionally, enteric coated ddI was used instead of stavudine (D4T) in our first line ART regimen due its better side-effect profile, thereby avoiding non-adherence and discontinuation associated with stavudine toxicity observed elsewhere [50, 51].

Female patients were 33% more likely to have optimal adherence compared to males in our cohort. Although gender has not been consistently found to influence ART adherence, HIV infected women in developing countries tend to demonstrate better health seeking behaviours than men which may be motivated by their role as the primary care giver in their families [52, 53]. Being a primary breadwinner was also positively associated with optimal adherence which could indicate that the responsibility of being the main income earner may have been an adherence facilitator in this group of patients.

Viral suppression rates exceeded 90% for each year throughout the observation period. Similar long-term virologic outcomes have been reported in other low- and middle-income regions [51, 54‐58]. The proportion of patients failing treatment was highest in the 2 years after ART initiation, thereafter, the number of treatment failures declined with each subsequent year on ART. Interestingly, non-adherence by pill count was most frequent in the first 6 months of ART in our cohort. Poor adherence in the initial months after ART initiation has been shown to increase the risk of virologic failure in later years and these early adherence lapses may be the reason for the higher number of treatment failures initially observed thus highlighting the importance of prioritizing adherence monitoring and support interventions to ensure optimal adherence from ART initiation [25, 44, 59‐61].

Sensitivity was high, but specificity was low for the use of pill count data as a proxy for viral load outcomes. Whilst some studies have demonstrated an association between pharmacy adherence measures and virologic outcomes others have found poor agreement between the two measures [39, 40, 62‐65]. We used an adherence threshold of ≥95%, however, viral suppression has been reported in patients with 85–94% adherence on long-term ART [66]. Furthermore, while pill counts provide a more objective method of adherence assessment, patient manipulation is possible as non-adherent patients may discard or leave behind medication not taken and return the desired number of pills to demonstrate good adherence [67, 68]. Although plasma viral suppression is possible with sub-optimal adherence, low-level viral replication continues in the plasma and reservoirs such as the central nervous system and genital tract, the long-term effects of which have not been fully elucidated [8, 69]. Therefore, high adherence to ART remains essential to ensuring the best clinical outcomes.

Patients who were provided with an extended 3-month ART supply in the TRuTH study were able to maintain high adherence levels over the extended time between clinic visits. This finding provides reassurance that stable patients can maintain optimal adherence when provided with an extended ART supply and patients who successfully adhere to treatment in the first 2 years are ideal candidates for more than 1 month’s supply of treatment. In an overburdened public health care setting, such as South Africa, this practice can ease clinic patient load and reduce the need for monthly clinic visits which impact negatively on patient finances and employment prospects.

High pill burden, regimen complexity and dosing more than once a day are known to hinder adherence [70]. Despite the additional pill burden and dosing frequency with switching from a once daily NNRTI-based regimen to a twice daily LPV/r regimen, with almost double the number of pills to be taken, adherence was not impaired and viral load suppression improved significantly in our patients who failed first-line treatment. The use of a boosted LPV/r regimen, provision of enhanced adherence counselling to patients on the implications of treatment failure with reinforcement of the importance of good adherence and enlisting a treatment supporter, where possible, may account for the improved outcomes [71‐73].

HIV and TB co-infection rates are highest in South Africa [74]. Concurrent treatment of both diseases reduces morbidity and mortality; however, patients must cope with the additional pill burden and overlapping side effects when taking ART and TB treatment which can lead to impaired or selective adherence to either ART or TB medication [32, 75, 76]. We found no significant changes in ART adherence or viral suppression in our patients on concurrent ART and TB treatment. This is commensurate with findings from other South African studies [77].

There are several limitations to our study. This was a retrospective study of ART adherence conducted at a single research site and adherence behaviour may have been influenced by participation in a clinical trial, where patients were routinely monitored and received extensive adherence support and counselling. Factors associated with adherence behaviour were limited to available baseline patient and clinical data. In addition, our patients were initiated on ART at lower CD4 counts with advanced HIV disease and sicker patients may be more motivated to adhere to treatment [78]. Hence, adherence in this cohort may not be generalizable to a ‘test and treat’ or public health patient setting. Assessment of adherence by pill count may be prone to bias as non-adherent patients familiar with the pill count assessment may ‘count pills’ and return the required number of pills to show good adherence. An accurate adherence calculation is not possible for patients who did not bring any remaining pills back, reported lost pills and/or pills left at home. The exclusion of patients that were lost to follow up, patients with missing pill count data for more 6 consecutive months, and the lack of pill count data in the time period between the SAPiT and TRuTH studies, could have resulted in an overestimation of adherence in our cohort.

Optimal long-term adherence rates observed in this cohort support the evidence of high adherence reported in other long-term African patient cohorts. Pill count should not be used as a standalone measure of adherence as it was not a good predictor of viral failure and viral load measurement remains the benchmark for monitoring treatment response. However, adherence during the 6 months post-ART initiation period has been shown to impact long-term treatment outcomes and, per in-country treatment guidelines, viral load is only measured 6 months after ART initiation and thereafter annually. Therefore, we recommend adherence monitoring using pill count as a quick, simple adherence measure at each visit in this critical 6 month period to identify patients with early adherence challenges and allow for timeous intervention.

Provision of adherence support measures which may have assisted patient adherence included individual and group counselling, provision of pill boxes, treatment supporters and tracking of patients who missed their clinic appointment date. Pharmacists were also able to identify patients experiencing adherence challenges when conducting pill counts and could either provide or refer patients for appropriate adherence counselling. We also demonstrated that providing stable patients with a 3-month ART supply was feasible as optimal adherence was still maintained over this extended period. This has the potential to reduce clinic workloads and patient time and costs associated with monthly clinic visits and is currently being investigated further in a randomized control trial [79]. Several studies are currently being undertaken in sub-Saharan Africa to monitor long term treatment outcomes in tests and treat populations [80]. Further research on long term adherence and adherence support strategies in these test and treat cohorts would help to determine if these factors could support optimal long term adherence to ART.

In summary, our study found optimal long-term adherence with good treatment outcomes in a structured ART programme with close adherence monitoring and support. With the implementation of a test and treat policy and rapid scale up of ART provision, it is imperative that patient adherence is monitored early on from ART initiation and then continually assessed and supported, to ensure a successful ART programme in South Africa.