The success of combination antiretroviral therapy (ART) in reducing the incidence of both AIDS and death in people with HIV  means that it is now feasible that people with HIV may ultimately receive ART for several decades. However, despite the continued improvement in ART efficacy, toxicities and other issues mean that individual drugs in an otherwise successful regimen may be discontinued. A number of studies [2–14] have investigated the rate of discontinuation of ART regimens, with estimates that up to half of patients discontinue one or more drugs in their initial ART regimen within 1 year [8,9,13].
These high rates of discontinuation of all or part of the initial ART regimen may in part be driven by the toxicities that are known to occur in the early stages of ART use. These discontinuation rates may however not reflect the long-term experience of patients who achieve and maintain viral suppression. To our knowledge, there have been no previous studies of regimen modification in such patients. As lifelong use of ART requires tolerable and durable regimens, it is important to study the durability of regimens in patients who achieve and maintain viral suppression.
In this study, we investigated the durability of certain regimens in patients seen at the Royal Free Clinic who started ART after January 2000. We considered two outcomes of interest in virally suppressed patients who had never experienced any previous virological failure: the rate of regimen change and the rate of toxicity-related regimen change. It is known that the rate of virological failure in such patients is low [15–18], but it is unclear whether the rate of treatment switching for other reasons is such that long-term use will ultimately be sustainable.
Patients included in this study attended the Royal Free Hospital, London, UK. Information on these patients is collected prospectively as patients attend for care, and includes information on demographics, start and stop dates and reasons for stopping all antiretroviral drugs, all AIDS diagnoses, in-patient visits, and Pneumocystis jirovecii pneumonia (PJP) prophylaxis treatment. A 100% audit of the information is conducted every 12 months by a trained research assistant. Laboratory data are transferred directly from the appropriate department electronically .
The analysis included all ART-naive patients who started ART during 2000–2005, achieved a viral load of less than 50 copies/ml within 180 days of starting ART, had a CD4 cell count and viral load measurement in the year before starting ART, had a CD4 measurement between the start of ART and viral suppression, had at least one follow-up CD4 cell count or viral load measurement after becoming virally suppressed, and were on an eligible regimen at some point when virally suppressed. An eligible regimen was defined to be a regimen consisting of either lamivudine (3TC) or emtricitabine (FTC), with a second nucleoside or nucleotide reverse transcriptase inhibitor (NRTI), and a third drug which was either a nonnucleoside reverse transcriptase inhibitor (NNRTI) or a protease inhibitor (PI). The study was restricted to these regimens as they were the most common regimen types received during the study period. Each patient was followed from the date of viral suppression until the date of the last CD4 cell count or viral load measurement, with follow-up on regimens not of the form described above excluded from the analysis and with all follow-up censored on 1 March 2006. Patient follow-up was also censored in the event of virological failure, defined to be two consecutive viral load measurements above 200 copies/ml, with the date of virological failure defined as the date of the first of these measurements. We chose this value of 200 copies/ml, rather than the higher value of 400 copies/ml that would be more commonly used because of the concern that some switches due to virological failure may be made by clinicians before a level of above 400 copies/ml is reached and we particularly wished to study only switches which were not due to virological failure. Although the actual value could not be known, an undetectable viral load on an assay with a lower limit of detection of 400 copies/ml was considered to be above 200 copies/ml for the purpose of identifying virological failure. Of 78 patients with follow-up censored because of virological failure, eight (10%) were due at least in part to an undetectable viral load on as assay with a lower limit of detection of 400 copies/ml. In patients with a viral load measurement less than 50 copies/ml, the proportion of subsequent viral loads measured on an assay with a lower limit of detection of 400 copies/ml was low.
CD4 cell counts were performed using standard flow cytometry techniques and plasma HIV-1 RNA measured by a variety of commercially available methods. During the study period, the majority of samples were analyzed using ultrasensitive assays with a lower limit of detection of 50 copies/ml. However, a small number, particularly pretreatment samples, may have been analyzed using assays with lower limits of detection of 400 copies/ml.
Both outcomes of interest were rates of treatment change, with a maximum of one treatment change counted in any 14-day period. Dose and formulation changes were ignored, as were switches between 3TC and FTC. Person-years of follow-up and treatment changes occurring when not on an eligible regimen were not counted.
Where possible, a single reason was assigned to each treatment change. This was derived from the reasons for drug discontinuation in the original dataset; up to three reasons are recorded for each drug discontinuation. Only the primary reason given for each drug discontinuation was considered, and these were categorized into the following groups: virological failure; resistance failure; CD4 failure; toxicities; patient choice; poor compliance; pregnancy related; study change; rationalization; and other reasons. If a treatment change did not involve any drug discontinuation, so if drugs were added to the regimen, then no reason could be assigned to the treatment change. If the change involved the discontinuation of one drug, the primary reason for discontinuation of that drug was defined to be the reason for the treatment change. If more than one drug was discontinued in the treatment change, the primary reason for discontinuation of each drug was obtained, rationalization discounted where another reason was present, and duplicates removed. If one reason remained, this was defined to be the reason for the treatment change. If more than one reason remained at this stage, each case was considered individually to determine the treatment change reason. For treatment changes where toxicity was mentioned, an attempt was made to identify a single specific toxicity using a similar process, although some treatment changes had more than one associated toxicity.
Factors associated with the rate of treatment change, and the rate of treatment change due to toxicities, in virally suppressed patients were investigated using Poisson regression models, Candidate time updated covariates were the second NRTI [zidovudine (ZDV), didanosine (DDI), stavudine (D4T), abacavir (ABA), tenofovir (TDF)], the third drug [efavirenz (EFV), nevirapine (NEV), lopinavir (LOP), saquinavir (SAQ), other PI (OTH)], CD4 cell count, time from viral suppression, regimen changes since ART initiation (none, at least one), Hepatitis C coinfection status (yes, no, unknown), Hepatitis B coinfection status (yes, no, unknown), and age. Candidate fixed covariates were ethnicity (white, black African, other) and a variable incorporating both gender and risk group (heterosexual men, heterosexual women, homosexual men, other). Whether the patient was on 3TC or FTC was not considered as a possible covariate, as it was assumed that most changes from 3TC or FTC would be secondary to other changes made at the same time, for example, it is likely that a change from combivir (3TC and ZDV) to truvada (FTC and TDF) would be driven by a desire to replace ZDV with TDF.
Factors found to be associated with the rate of treatment change (or the rate of toxicity-related treatment change) with a P value less than 0.2 in a univariable model were considered as potential covariates in the multivariable models. Factors with an adjusted P value less than 0.05 were retained in the multivariate models. All P values are two sided. Analyses were performed using SAS version 9.1. (SAS Institute Inc., Cary, North Carolina, USA).
There were 508 patients included in the analysis, providing 912 person-years of eligible follow-up. Patient characteristics and the number of person-years of follow-up on specific drugs are given in Table 1. A total of 78 (15.4%) patients had follow-up censored due to virological failure, and 883 (96.8%) person-years of follow-up were during a period in which the most recent viral load was less than 50 copies/ml.
Most follow-up was spent on a 3TC-containing regimen (782 person-years, 85.7% of total follow-up). The most used second NRTIs were ZDV (419 person-years, 45.9%) and TDF (368 person-years, 40.4%). Most follow-up was spent on an NNRTI based regimen, with 398 person-years (43.6%) on EFV and 169 person-years (18.5%) on NEV. A total of 240 person-years (26.3%) was spent on LOP.
During the study period, there were a total of 357 treatment changes. Reasons were available for 279 (78.2%) treatment changes, with toxicities being the most common reason (140 treatment changes, 50.2%). The distribution of reasons for treatment changes is shown in Fig. 1a. Patient choice was the assigned reason for 49 (17.6%) treatment changes, and poor compliance for 10 (3.6%) treatment changes. Two treatment changes were attributed to virological failure, despite the fact that follow-up was censored when a patient experienced virological failure defined by two consecutive values more than 200 copies/ml. The most recent viral loads for these two patients at the time of the treatment changes were 491 copies/ml and less than 50 copies/ml.
Single specific toxicities were identified for 136 of the 140 toxicity-related treatment changes, and two toxicities were identified for each of the remaining four toxicity-related treatment changes, giving a total of 144 reported toxicities. The distribution of toxicities associated with treatment changes is shown in Fig. 1b. Those most frequently reported were central nervous system (CNS) effects (33, 22.9%) and lipodystrophy (28, 19.4%). All 33 reports of CNS effects were in patients with current third drug EFV.
The overall rate [95% confidence interval (CI)] of treatment change was 39.1 (35.1–43.2) per 100 person-years. In a Poisson regression model, factors found to be associated with the rate of treatment change were current second NRTI, current third drug, risk group, time from viral suppression, and current CD4 cell count (Table 2). Being on ABA or TDF was associated with a lower rate of treatment change compared with being on ZDV [incidence rate ratio (IRR) = 0.29, 95% CI 0.12–0.67; IRR = 0.61, 95% CI 0.48–0.79 respectively], and being on D4T was associated with a higher rate of treatment change than being on ZDV (IRR = 1.67, 95% CI 1.28–2.17). Being on LOP or SAQ was associated with a higher rate of treatment change than being on EFV (IRR = 1.53, 95% CI 1.21–1.94; IRR = 1.75, 95% CI 1.04–2.95 respectively). The rate of treatment change was significantly lower in heterosexual men than in the other combined sex and risk groups (heterosexual women IRR = 1.60, 95% CI 1.13–2.26; homosexual men IRR = 1.54, 95% CI 1.11–2.12; other IRR = 2.09, 95% CI 1.00–4.37). Increased time from viral suppression was associated with a lower rate of treatment change (per twofold increase in time from viral suppression: IRR = 0.90, 95% CI 0.85–0.96). Higher CD4 cell count was associated with an increased rate of treatment change (per 100 cells/μl increase: IRR = 1.06, 95% CI 1.02–1.11). Having previously experienced a treatment change at any point since the start of ART was associated with a lower rate of treatment change while virally suppressed, but this effect was not found to be significant when adjusted for other covariates in the model. Current viral load was not found to be associated with the rate of treatment change (per log10 increase: unadjusted IRR = 1.02, 95% CI 0.70–1.47). Other factors considered but not found to be associated with the rate of treatment change were ethnicity, pre-ART AIDS, current age, and current Hepatitis B and C status.
The overall rate (95% CI) of treatment change attributed to toxicities was 15.4 (12.8–17.9) per 100 person-years. In a Poisson regression model, factors found to be associated with this rate were current second NRTI, current third drug, risk group, and current age (Table 2). Being on D4T was associated with a higher rate of toxicity-related treatment change than being on ZDV (IRR = 2.04, 95% CI 1.28–3.26), and being on TDF was associated with a lower rate of toxicity-related treatment change than being on ZDV (IRR = 0.46, 95% CI 0.29–0.73). Being on LOP was associated with a higher rate of toxicity-related treatment change than being on EFV, and being on ATV was associated with a lower rate of toxicity-related treatment change than being on EFV (IRR = 1.55, 95% CI 1.04–2.31; IRR = 0.23, 95% CI 0.06–0.91 respectively). The rate of toxicity-related treatment change was significantly higher in homosexual men and in patients in the ‘other’ risk group than in heterosexual men (IRR = 2.26, 95% CI 1.37–3.73; IRR = 3.28, 95% CI 1.09–9.84 respectively). The rate of toxicity-related treatment change increased with age (for every 10 years increase: IRR = 1.28, 95% CI 1.04–1.57). Other factors found to be univariately associated with the rate of toxicity-related treatment change were ethnicity and previous treatment changes since ART initiation, but these factors were not found to be significant when adjusted for the other covariates in the model. Factors considered but not found to be associated with the rate of toxicity-related treatment change were pre-ART AIDS, current CD4 cell count, current viral load, and current Hepatitis B and C status.
When the models were fitted separately according to current calendar year strata (2000–2003, 2004–2006), there was no suggestion of differences in the models in the two periods other than that which would be expected due to some drugs not being in use in the earlier period.
We performed a sensitivity analysis using a cut off for virological failure of 400 copies/ml rather than 200 copies/ml. Here, there were 368 treatment changes in 926 person-years (141 due to toxicities), follow-up for 66 patients (13.0%) was censored due to virological failure, and four treatment changes were attributed to virological failure. Current viral load was again not a significant predictor of the rate of treatment change (unadjusted IRR = 1.13, 95% CI 0.84–1.52). The resulting multivariate models for the rate of treatment change and for the rate of treatment change due to toxicities were consistent with those previously obtained.
The present study considered the rate of treatment change in a group of previously ART-naive patients who achieved viral suppression within 6 months of starting ART and who had never previously experienced virological failure, during the time spent on a regimen consisting of either 3TC or FTC with a second NRTI and a third drug that was either a PI or NNRTI. Studying this group allowed us to evaluate treatment changes occurring in successfully responding patients during a more settled period some time after ART initiation, omitting the switching that may take place in the period immediately after starting treatment. We observed that the overall rate (95% CI) of treatment change was 39.1 (35.1–43.2) per 100 person-years, and the overall rate (95% CI) of treatment change for toxicity reasons was 15.4 (12.8–17.9) per 100 person-years.
Factors found to be associated with both the rate of treatment change and the rate of toxicity-related treatment change were the specific second NRTI and third drug included in the regimen, and the combined gender and risk group. Time from viral suppression and most recent CD4 cell count were found to be associated with the overall rate of treatment change, and age was found to be associated with the rate of toxicity-related treatment change.
To our knowledge, this is the first study to evaluate factors associated with the rate of treatment change in patients who have become virally suppressed and who have not yet experienced treatment failure. Hence, results from studies of ART modification and discontinuation in patients who may have not yet experienced a durable response to an ART regimen are not directly comparable. However, it is of interest to consider whether the factors found here to be associated with the rate of treatment change have been found in previous studies to be associated with the incidence of discontinuation of any part of the initial ART regimen. The observation that D4T as second NRTI was associated with higher rate of treatment change (overall and due to toxicities) than ZDV is consistent with the finding in a previous study that the nucleoside pair 3TC and D4T was associated with a higher incidence of discontinuation than the nucleoside pair 3TC and ZDV of the initial ART regimen due to toxicity, poor compliance, or patient choice . It is important to note that in resource-poor settings the incidence of haematological toxicities from ZDV is likely to be higher [20,21], so the relative toxicity of D4T compared with ZDV would be expected to be different.
As this analysis does not consider early treatment changes, the results may appear to be biased in favour of drugs that are associated with early toxicities. For example, hypersensitivity associated with ABA presents soon after ART initiation, whereas neuropathy associated with D4T occurs following a greater amount of time on ART .
We found that risk group was a significant predictor of both rates of treatment change, with homosexual men having a higher rate of change than heterosexual men. This is difficult to interpret, but a possible explanation is that this group is more aware of available alternative treatments and more proactive at seeking them out, or they find or see more proactive physicians. The breakdown of reasons for treatment change by risk group does not support this, as the proportion of treatment changes due to patient choice in homosexual men was 16%, compared with 22% of treatment changes in heterosexual men.
A European study  found that having been on ART for at least 6 months was associated with a reduced incidence of modifying the initial ART regimen for toxicity-related reasons, compared with having been on ART for less than 6 months. The same study found that being on ART for at least 12 months was associated with an increased incidence of stopping the original ART regimen due to treatment failure. In our study, the overall rate of treatment change was found to be lower in patients who had been virally suppressed for longer and not yet experienced virological failure. This may be a positive indicator as it suggests that patients who have responded to treatment may be successfully treated with the same regimen for an extended period of time. However, the time from viral suppression was not found to be a significant predictor of treatment change due to toxicities, so this should be interpreted with caution.
A limitation of this analysis is that the study considered only a single treatment centre, which is large and urban. The follow-up in this analysis extends back to 2000, so some regimens represented may not be representative of those currently used at the treatment centre. It is also important to note that we did not exclude those patients who were enrolled in clinical trials. However, the number of treatment changes attributed to a study change was small (14, 5.0%). Further limitations were the small amount of follow-up available for some drugs (ABA, DDI, D4T, SAQ, ATV), and that reasons were unknown for 78 of the 357 treatment changes.
Although we ignored switches between 3TC and FTC on the assumption that they were secondary to other nucleoside switches (for example, switches from ZDV to TDF), there is some suggestion that FTC carries a higher risk of headache so some switches may have in fact been due to this apparent toxicity.
Most PI use involved the use of ritonavir in addition to the named PI (94%). Ritonavir use has been associated with increased lipid levels compared with other PIs , and has been associated with an increased risk of regimen modification or discontinuation . In this study, it was not possible to differentiate between the effect of the main PI and the effect of ritonavir on treatment change.
As not all toxicities will result in regimen modification, the true incidence of toxicities will be underestimated by toxicity-related treatment discontinuations. Toxicities are more likely to result in a discontinuation where an alternative drug is available. With the availability of new drugs within the original three classes and of new classes of drugs, there will be more alternatives in the event of toxicities, and so the rate of drug discontinuation due to toxicities may increase in the future.
The true rate of treatment change caused by drug toxicities can be thought of as lying somewhere between the two rates of treatment change considered in this study. Treatment changes recorded as being due to reasons other than specific toxicities, such as patient or physician choice, may in fact be driven by toxicities. In patients who have never experienced virological failure, the rate of treatment change due to toxicities is low, with certain regimens associated with an even lower rate of change. This suggests that so long as virological failure is avoided, some regimens are so far proving to be sufficiently stable to suggest that very long-term use is potentially feasible.
1. Mocroft A, Ledergerber B, Katlama C, Kirk O, Reiss P, Monforte AD, et al
. Decline in the AIDS and death rates in the EuroSIDA study: an observational study. Lancet 2003; 362:22–29.
2. Ahdieh-Grant L, Tarwater PM, Schneider ME, Anastos K, Cohen M, Khalsa A, et al
. Factors and temporal trends associated with highly active antiretroviral therapy discontinuation in the women's interagency HIV study. JAIDS 2005; 38:500–503.
3. Chen RY, Westfall AO, Mugavero MJ, Cloud GA, Raper JL, Chatham AG, et al
. Duration of highly active antiretroviral therapy regimens. Clin Infect Dis 2003; 37:714–722.
4. Hansel A, Bucher HC, Nuesch R, Battegay H. Reasons for discontinuation of first highly active antiretroviral therapy in a cohort of proteinase inhibitor-naive HIV-infected patients. JAIDS 2001; 26:191–193.
5. Hooshyar D, Napravnik S, Miller WC, Eron JJ. Effect of hepatitis C coinfection on discontinuation and modification of initial HAART in primary HIV care. AIDS 2006; 20:575–583.
6. Mocroft A, Youle M, Moore A, Sabin CA, Madge S, Lepri AC, et al
. Reasons for modification and discontinuation of antiretrovirals: results from a single treatment centre. AIDS 2001; 15:185–194.
7. Mocroft A, Rockstroh J, Soriano V, Ledergerber B, Kirk O, Vinogradova E, et al
. Are specific antiretrovirals associated with an increased risk of discontinuation due to toxicities or patient/physician choice in patients with hepatitis C virus coinfection? Antivir Ther 2005; 10:779–790.
8. Monforte AD, Lepri AC, Rezza G, Pezzotti P, Antinori A, Phillips AN, et al
. Insights into the reasons for discontinuation of the first highly active antiretroviral therapy (HAART) regimen in a cohort of antiretroviral naive patients. AIDS 2000; 14:499–507.
9. O'Brien ME, Clark RA, Besch CL, Myers L, Kissinger P. Patterns and correlates of discontinuation of the initial HAART regimen in an urban outpatient cohort. JAIDS 2003; 34:407–414.
10. Spire B, Carrieri P, Garzot MA, L'Henaff M, Obadia Y. Factors associated with efavirenz discontinuation in a large community-based sample of patients. AIDS Care 2004; 16:558–564.
11. van Roon EN, Verzijl JM, Juttmann JR, Lenderink AW, Blans MJ, Egberts ACG. Incidence of discontinuation of highly active antiretroviral combination therapy (HAART) and its determinants. JAIDS 1999; 20:290–294.
12. Yuan Y, L'Italien G, Mukherjee J, Iloeje UH. Determinants of discontinuation of initial highly active antiretroviral therapy regimens in a US HIV-infected patient cohort. HIV Med 2006; 7:156–162.
13. Mocroft A, Phillips AN, Soriano V, Rockstroh J, Blaxhult A, Katlama C, et al
. Reasons for stopping antiretrovirals used in an initial highly active antiretroviral regimen: increased incidence of stopping due to toxicity or patient/physician choice in patients with hepatitis C coinfection. Aids Res Hum Retroviruses 2005; 21:743–752.
14. Li XH, Margolick JB, Conover CS, Badri S, Riddler SA, Witt MD, et al
. Interruption and discontinuation of highly active antiretroviral therapy in the multicenter AIDS cohort study. JAIDS 2005; 38:320–328.
15. Smith CJ, Phillips AN, Hill T, Fisher M, Gazzard B, Porter K, et al
. The rate of viral rebound after attainment of an HIV load <50 copies/ml according to specific antiretroviral drugs in use: results from a multicenter cohort study. J Infect Dis 2005; 192:1387–1397.
16. Lampe FC, Gatell JM, Staszewski S, Johnson MA, Pradier C, Gill MJ, et al
. Changes over time in risk of initial virological failure of combination antiretroviral therapy – A multicohort analysis, 1996 to 2002. Arch Intern Med 2006; 166:521–528.
17. Benzie AA, Bansi LK, Sabin CA, Portsmouth S, Hill T, Johnson M, et al
. Increased duration of viral suppression is associated with lower viral rebound rates in patients with previous treatment failures. AIDS 2007; 21:1423–1430.
18. Sungkanuparph S, Groger RK, Overton T, Fraser VJ, Powderly WG. Persistent low-level viraemia and virological failure in HIV-1-infected patients treated with highly active antiretroviral therapy. HIV Med 2006; 7:437–441.
19. Sabin CA, Lampe FC, Chaloner C, Madge SJ, Lipman MCI, Youle M, et al
. An audit of antiretroviral treatment use in HIV-infected patients in a London clinic: the limitations of observational databases when auditing antiretroviral treatment use. HIV Med 2003; 4:87–93.
20. Ssali F, Stohr W, Munderi P, Reid A, Walker AS, Gibb DM, et al
. Prevalence, incidence and predictors of severe anaemia with zidovudine-containing regimens in African adults with HIV infection within the DART trial. Antivir Ther 2006; 11:741–749.
21. Moh R, Danel C, Sorho S, Sauvageot D, Anzian A, Minga A, et al
. Haematological changes in adults receiving a zidovudine-containing HAART regimen in combination with cotrimoxazole in Cote d'Ivoire. Antivir Ther 2005; 10:615–624.
22. Carr A, Cooper DA. Adverse effects of antiretroviral therapy. Lancet 2000; 356:1423–1430.
23. Fontas E, van Leth F, Sabin CA, Friis-Moller N, Rickenbach M, Monforte AD, et al
. Lipid profiles in HIV-infected patients receiving combination antiretroviral therapy: are different antiretroviral drugs associated with different lipid profiles? J Infect Dis 2004; 189:1056–1074.