Triple-drug class antiretroviral treatment failure:
Time to triple drug class antiretroviral treatment failure after initiation of
HAART : Results from the EuroSIDA study group
A Mocroft1, B Ledergerber2, JP Viard3, S Staszewski4, M Murphy5, A Chiesi6, A Horban7, A-B E Hansen8, AN
Phillips1, JD Lundgren9 for the EuroSIDA study group*.
1Royal Free Centre for HIV Medicine and Dept Primary Care and Population Sciences, Royal Free and
University College Medical School, London, UK; 2University Hospital, Zurich, Switzerland; 3Hôpital Necker-
Enfants Malades, Paris, France; 4JW Goethe University Hospital, Frankfurt, Germany; 5St Bartholmews
Hospital, London, UK; 6Istituto Superiore di Sanita, Rome, Italy; 7Centrum Diagnostyki i Terapii AIDS,
Warsaw, Poland; 8Rigshospitalet, Copenhagen, Denmark; 9Hvidovre Hospital, Copenhagen, Denmark.
*Members of the study group are listed in the appendix.
Royal Free Centre for HIV Medicine and Dept Primary Care and Population Sciences
Royal Free and University College London Medical Schools
Royal Free Campus, Rowland Hill St, London, NW3 2PF
Objectives: To determine the prevalence, incidence of and time to triple class failure (TCF) and the factors
related to it after initiation of HAART.
Design and Methods: Observational longitudinal study of 3538 patients starting HAART from the pan-
European study, EuroSIDA. Patients were followed from the date of starting HAART (baseline) until TCF.
The incidence of TCF after starting HAART among both treatment experienced (TE) and treatment naïve
(TN) patients according to calendar year and years of exposure to HAART were determined, as were factors
associated with TCF in both patient groups, using Cox proportional hazards models.
Results: 469 patients (13.3%) failed 3 drug classes; of these, 400/2430 (16.5%) were TE and 69/1108 were
TN (6.2%). At 6 years after baseline, 24.1% of TE patients were estimated to have TCF (95% confidence
interval (CI) 21.6 – 26.6) compared to 11.9% of TN patients (95% CI 8.6 – 15.2) while the prevalence of TCF
among patients under follow-up at/after 2002 was 16.1% in TE patients and 5.5% in TN patients. Among TN
patients, there was an increase in the incidence of TCF with increasing time from baseline from 1.2 per 100
PYFU (95% CI 0.7 – 1.7) in the first 2 years after baseline to 4.7 per 100 PYFU (95% CI 2.1 – 8.9) at or after
5 years from baseline (33% increase per year, 95% CI 12 – 58%, p = 0.030), similar to the rate seen in TE
patients treated for the same period of time (5.4 per 100 PYFU, 95% CI 3.5 – 7.3). TE patients who also
started 2 new nucleosides at baseline had a 49% reduced risk of TCF (95% CI 23 – 66%, p = 0.0012),
while each additional 12 months cumulative exposure to nucleosides prior to baseline was associated with a
6% increased risk of TCF (95% CI 2 – 11%, p = 0.0016).
Conclusion: We found a low rate of TCF among patients starting HAART, particularly among TN naïve
patients, although the rate has increased progressively over time since starting HAART. Despite the influx of
new patients, the prevalence of TCF is increasing significantly over calendar time. The long term
consequences of TCF on the durability of HAART and how best to manage this situation deserves further
One of the goals of highly active antiretroviral therapy (HAART) is to reduce the viral load to below the limit of
detection to reduce the chances of further viral evolution under therapeutic selection pressures. Despite an
initial good response to HAART the viral load may rebound in some patients. This might be related to the
potentially serious adverse events, the emergence of drug resistant viruses, and the difficulties of
maintaining long-term adherence (1-2). Patients with rebounding viral load are typically switched to a
second-line or salvage regimen, where the response is usually poorer than when patients first start HAART
(3-6). Such salvage regimens often contain a different class of antiretrovirals, so an initial protease-inhibitor
(PI) containing HAART regimen may be switched to a non-nucleoside reverse transcriptase inhibitor (NNRTI)
containing HAART regimen or vice versa. Once patients have been exposed to the three main classes of
antiretrovirals their treatment options are limited because of cross resistance; it is usually not possible to
keep the viral load in such patients at sufficiently low levels and eventually this may lead to decreasing CD4
lymphocyte counts and clinical disease progression (7-8). While there is considerable evidence from both
observational studies and clinical trials of the response to both first-line and second-line HAART regimens
(9-11), there is relatively little known about the time to triple drug class failure (TCF) or the factors related to
The aims of this study were therefore to describe the time to triple drug class drug exposure, the incidence of
virologic TCF among 3538 patients from across Europe starting HAART and to describe the factors
The EuroSIDA study is a prospective, European study of patients with HIV-1 infection in 72 centres across
Europe (including Israel - see appendix) and now including Argentina. Details of the study have been
published (12). In brief, Centres provided data on consecutive patients seen in the outpatient clinic from 2
May 1994 until a predefined number of patients was enrolled from each centre. This cohort of 3118 patients
was defined as the EuroSIDA I cohort. Enrolment of a second cohort of 1365 patients began in December
1995. In April 1997 a further 2839 patients were recruited and was defined as the EuroSIDA III cohort.
Cohort IV, 1225 patients, was enrolled from April 1999, and a fifth cohort, Cohort V, 1256 patients, was
recruited from September 2001. At recruitment, in addition to demographic and clinical information, a
complete antiretroviral history is collected, together with the 4 most recent CD4 counts and viral load
measurements. At each follow-up visit, details on all CD4 lymphocyte counts measured since last follow-up
and viral load measurements were collected, as was the date of starting and stopping each antiretroviral
drug and the use of drugs for prophylaxis against opportunistic infections. Dates of diagnosis of all AIDS
defining illnesses have also been recorded using the 1993 clinical definition of AIDS from the Centers for
Disease Control. Members of the coordinating office visited all centres to ensure correct patient selection
and that accurate data was provided by checking the information provided against case-notes for a
HAART was defined as a minimum of 3 antiretroviral drugs, of which at least one was a protease inhibitor
(PI), non-nucleoside reverse transcriptase inhibitor (NNRTI) or abacavir. All patients starting HAART for the
first time were eligible for inclusion. Patients who had previously taken a PI or NNRTI which was not
included in a HAART regimen were excluded. The baseline date was taken as the date of starting HAART.
Patients without a CD4 count or viral load measured in the 6 months before baseline were excluded, as were
patients with no prospective follow-up. Patients who started HAART prior to recruitment to EuroSIDA were
only included if they had a minimum of 2 viral loads measured in each year prior to recruitment to try to
ensure that patients had not failed three drug classes prior to the start of prospective follow-up.
Virological failure of a drug class was defined in the same way as for the PLATO (P
ptions) collaborative study (7) . In brief, virological failure of a drug class was defined as a viral load >
1000 copies/ml for a total of at least 4 months after starting the drug class whilst still being on the same drug
class. Failure of any class could occur when it was used alone as monotherapy or as a component of dual,
triple or more intensive regimen, providing viral load measures were available. Patients were classified as a
TCF on the first date they had failed nucleosides, PI’s and NNRTI’s. For example, a patient starting HAART
with 2 nucleosides and a PI and with a viral load > 1000 copies/ml for at least 4 months would be defined at
that time as failing both nucleosides and PI’s.
The prevalence of TCF at each time point was defined as the proportion of patients with TCF divided by the
total number at risk. Patients were removed from this analysis at the date of their last viral load
measurement. The incidence of TCF was defined as the number of triple-class failures divided by the
person-years of follow-up (PYFU) and was stratified according to time since starting HAART, calendar year
and use of antiretrovirals prior to starting HAART (i.e., treatment experienced (TE) compared to treatment
naive patients). Patient follow-up was measured from the date of starting HAART until the date of TCF.
Patients who did not have TCF were censored at the date of their last viral load measurement. Trends over
time were tested using Poisson regression.
Time from starting HAART to exposure to three drug classes and to TCF was analysed using Kaplan-Meier
survival curves. The factors associated with TCF were determined using Cox proportional hazards models.
All Cox models were stratified by centre and performed separately for TN and TE patients. Factors that were
significant in univariate analyses (p < 0.1) were included in multivariate analyses. Variables in univariate
analyses included gender, exposure group, race, hepatitis B and C status, prior AIDS diagnosis, age, date of
starting antiretroviral therapy, CD4 and viral load. Treatment variables included whether patients were TN at
baseline, HAART regimen started and the number of antiretrovirals patients were taking at baseline. CD4
and viral load were included both as continuous and categorical variables. In TE patients the additional
variables considered included time since starting antiretrovirals, total exposure time to nucleosides (i.e., the
sum of exposure to each of the individual nucleosides), the cumulative number of nucleosides ever taken,
the number of new (i.e., never previously taken) nucleosides started at HAART and what prior treatment
strategy patients had taken (monotherapy, dual therapy or both).
A further Cox model was constructed which redefined the baseline date to be 1 January 1987, and patient
follow-up was left-censored until the date of starting HAART. This analysis allowed changes in the rate of
TCF over time since starting HAART to be formally tested after adjustment for the other factors related to
Poisson regression was performed using STATA (version 7) and all other analyses were performed using
SAS (version 8.2, Cary NC, USA). All tests of significance were 2-sided.
Of 9803 patients enrolled in EuroSIDA, 3873 had not started HAART and 2209 had no CD4 or viral load
measured in the 6 months prior to starting HAART. A further 183 patients were excluded because they had
less than 2 viral loads measured per year prior to recruitment to EuroSIDA; thus 3538 patients satisfied the
inclusion criterion, and are described in Table 1. The majority of the patients were male (2747, 77.6%),
belonged to the homosexual exposure group (1644, 46.5%) and were of White ethnic origin (2978, 84.2%).
832 patients (23.5%) had a diagnosis of AIDS at or before the date of starting HAART. 1192 patients
(33.7%) started HAART prior to recruitment in EuroSIDA, a median time of 5 months before recruitment (IQR
3 – 10 months). The median CD4 count at starting HAART was 221/mm3 (IQR 110 – 343/mm3) and
median viral load was 4.43 log10copies/ml (IQR 3.62 – 5.08 log10copies/ml). There was a median of 15 viral
loads per patient measured during follow-up (IQR 9 – 21) at a median time apart of 3 months (IQR 2 – 4
months). Median follow-up after starting HAART was 50 months (IQR 29 – 64 months), with a total of 13,614
2430 patients (68.7%) had prior antiretroviral treatment at baseline. Patients first started antiretrovirals a
median time of 34 months before starting HAART (IQR 15 – 60 months). A large proportion of the TE
patients had taken both monotherapy and dual therapy (1417, 58.3%), 787 patients had taken dual therapy
alone (32.4%) and 226 had only taken monotherapy (8.3%). The median number of antiretrovirals patients
had been exposed to prior to baseline was 3 (IQR 2 –4). The median cumulative exposure time (i.e., the
sum of exposure to each of the nucleosides taken) to nucleosides prior to starting HAART was 48 months
(IQR 23 – 81 months). Almost half the patients starting HAART did not start any new nucleosides (i.e.,
nucleosides to which they had never previously been exposed) at the date of starting HAART (1115, 45.9%),
while 811 patients started 1 (33.4%) and 504 started two new nucleosides (20.7%).
The Kaplan-Meier estimate of the median time to exposure to 3 classes of antiretrovirals was 47 months
(95% confidence interval (CI) 45 – 50 months), and was significantly shorter among patients who were TE
(41 months, 95% CI 39 – 43 months) compared to TN patients (61 months, 95% CI 56 – 66 months, p <
Prevalence and incidence of TCF and calendar time
469 patients (13.3%) experienced TCF after starting HAART; of these, 400/2430 (16.5%) were TE and
69/1108 were TN (6.2%). Figure 1 shows the prevalence of TCF over calendar time. The prevalence
at/after 2002 was 16.1% in TE patients and 5.5% among TN patients. The majority of patients in this study
started HAART prior to 1999, as shown in Figure 2. There was a low incidence of TCF during 2001,
particularly in TN patients. Using Poisson regression, over time the incidence of TCF has increased, at an
estimated 29% per year in TN patients (95% CI 9 – 53%, p = 0.003). The change over time was not linear
for TE patients; there was a significantly decreased incidence of TCF during 1999 (rate ratio (RR) 0.42, 95%
CI 0.32 – 0.57, p < 0.0001) and during 2001 (RR 0.57, 95% CI 0.41-0,73, p = 0.001) compared to 1999, but
no significant differences in the rates in other calendar years (p > 0.05).
Figure 3 shows the time from start of HAART to TCF, stratified by prior antiretroviral treatment. At 6 years
after starting HAART, 24.1% of TE patients were estimated to have failed three classes of antiretrovirals
(95% CI 21. 6 – 26.6) compared to 11.9% of TN patients (95% CI 8.6 – 15.2, p < 0.0001, log-rank test).
During follow-up, just over half of the patients who started HAART were exposed to three classes of
antiretrovirals (1816 patients, 51.3%). At 3 years after starting the third class, 32.3% of TE patients were
estimated to have TCF (95% CI 29.5 – 35.1) compared to 17.2% of TN patients (95% CI 13.2 – 21.2, p <
Among TN patients, the overall incidence of TCF was 1.7 per 100 PYFU (95% CI 1.3 – 2.1), compared to 4.2
per 100 PYFU (95% CI 3.8 – 4.6) among TE patients, a rate 2.4 times higher (95% CI 1.88 – 3.13, p <
0.0001, Poisson regression). Figure 4 shows the incidence of triple-class failure according to time since
starting HAART, separately for TN and TE patients. Among TN patients, there was an increase in the
incidence of failure with increasing time from baseline from 1.2 per 100 PYFU (95% CI 0.7 – 1.7) in the first 2
years after baseline to 4.7 per 100 PYFU (95% CI 2.1 – 8.9) at or after 5 years from baseline (33% increase
per year, 95% CI 12 – 58%, p = 0.030), similar to the rate seen in TE patients treated for the same period of
time (5.4 per 100 PYFU, 95% CI 3.5 – 7.3).
Is the rate of triple-class failure increasing over time since starting HAART?
Given the pattern of TCF seen in Figure 2 and 4, for TN patients a continuous variable was used for time
since starting HAART while for TE patients a categorical variable was fitted with between 2-3 years as the
reference category. Among TN patients, after adjustment for CD4 and viral load at starting HAART there
was a 41% increased risk of TCF with each extra year since starting HAART (relative hazard (RH) 1.41, 95%
CI 1.06 – 1.88, p = 0.018). In TE patients, after adjustment for CD4 and viral load at starting HAART,
number of new nucleosides started and the total cumulative exposure to nucleosides prior to HAART
patients were at a significantly decreased risk of TCF in the first 2 years after starting HAART (RH 0.42; 95%
CI 0.31 – 0.57, p < 0.0001) compared to patients 2-3 years after starting HAART. After this time, there were
no significant differences in the risk of TCF with increasing time from starting HAART. Similar results were
seen when the first 2 years of follow-up after starting HAART (where there were few TCFs) were excluded
Factors associated with TCF in TN patients
Table 2 shows the univariate and multivariate factors associated with TCF among TN patients. In the first
multivariate model, both CD4 and viral load were included as continuous variables. There was no
statistically significant relationship between year of starting HAART and risk of TCF in multivariate analyses.
Patients with a higher CD4 count at starting HAART were at a decreased risk of TCF, while patients with a
higher viral load at starting HAART were at an increased risk. For example, patients with a CD4 of below
200/mm3 at starting HAART had a 55% increased risk of TCF compared to patients with a CD4 of 350/mm3
or higher (RH 1.55; 95% CI 0.74 – 3.22, p = 0.24), although this was not statistically significant and patients
with an intermediate CD4 count of 200 – 349/mm3 had no increase in risk.
Factors associated with TCF in TE patients
Among TE patients, the total number of ARV’s taken prior to starting HAART, number of new drugs started
and the cumulative exposure to ARV’s prior to HAART were significantly associated with TCF in univariate
analyses and were therefore included in multivariate analyses. The results are shown in Table 3.
There was no statistically significant relationship between year of starting HAART and risk of TCF in
multivariate analyses. After adjustment, patients with a higher viral load at starting HAART were at an
increased risk of failing three drug classes (RH 1.51; 95% CI 1.34 – 1.70, p < 0.0001), while patients with a
higher CD4 count at starting HAART were at a reduced risk of TCF (RH 0.94; 95% CI 0.88 – 1.00, p = 0.067)
although this did not quite reach statistical significance. Patients who started 2 new nucleosides at the date
of starting HAART had almost a 50% reduced risk of TCF (RH 0.51, 95% CI 0.34 – 0.77, p = 0.0012), while
each additional 12 months cumulative exposure to nucleosides prior to starting HAART was associated with
a 6% increased risk of TCF (95% CI 2 – 11%, p = 0.0016). Similar results were seen when categorical
variables were used for CD4 and viral load. Compared to patients with a CD4 at starting HAART of 350/mm3
or higher, patients with a CD4 count of less than 200/mm3 at starting HAART had a significantly increased
risk of TCF (RH 1.59; 95% CI 1,14 – 2.20, p = 0.0060), while those with a CD4 count of 200-349/mm3 had an
increased risk of TCF, which was marginally statistically significant (RH 1.31; 95% CI 0.94 – 1.83, p = 0.11).
For viral load, there was no clear line where the risk of TCF changed but a steadily increasing risk of TCF as
In 2002, 1 in 20 patients who were treatment naïve and 1 in 6 who were treatment experienced had triple
drug class failure after starting HAART. The incidence of TCF was considerably lower among TN patients
but was significantly increasing with increasing time since starting HAART, whereas the incidence among TE
patients was stable after an initially low incidence in the first 2 years after starting HAART. The World
Health Organisation projects that 3 million people with HIV in the developing world will be on treatment by
the end of 2005 (13). As the rate of TCF will be comparable in the developing world to that reported here,
and as the population with extended exposure to HAART in the developed world increases, the number of
patients with TCF worldwide will continue to increase in the coming years.
At 6 years after starting HAART just over 24% of TE patients and 12% of TN patients were estimated to have
TCF. In some small but detailed studies with more extended follow-up, around 30% of patients who were
initially TN have virologic failure of their first HAART regimen at 2-3 years after starting HAART (14-15) while
from larger observational studies between 20-40% of patients are estimated to have virologic failure of their
first HAART regimen, with, on average, 2-3 years follow-up (16-17). The definition of virologic failure differs
between studies, as do the patients included (TN, TE or a mix of both), but those patients who fail their initial
HAART regimen usually change to a regimen containing a different class of drug and are therefore at
immediate risk of TCF. Our estimates of TCF in TN and TE patients at 6 years after starting HAART appear
consistent with the above estimates, coupled with those suggesting that a further 20-50% of patients fail
subsequent second-line regimens (3-6), and the fact that there is a substantial number of patients who
interrupt treatment (Olsen, manuscript in preparation).
Several studies have shown that TE patients have a poorer virologic response to HAART (18-20). The time
between TCF and subsequent clinical progression is currently unknown, and in general, to date, there has
been no difference in risk of clinical progression between TE and TN patients (21-23). However, it may take
several years from TCF to subsequent clinical progression and with extended follow-up of patients with
exposure to HAART and TCF differences in clinical progression may become apparent. To date, cross-
resistance to antiretroviral classes is irreversible, thus the prevalence of patients living with TCF will increase
over time both within a study population such as EuroSIDA and within the population of patients with HIV as
more patients are exposed to more drug classes. For the individual, this may lead to a poorer prognosis
over time, and the potential for transmission of resistant virus to others (24). For the clinics, it may lead to
increased costs due to more intensive diagnostic tests, the use of more expensive regimens such as
enfuvirtude (25), and the use of more drugs in each regimen.
The incidence of TCF was increasing with more extended exposure to HAART among TN patients, but
remained fairly steady among TE patients, after an initially low rate. The low rate of TCF among TN patients
during 2001 may be partly explained by the comparatively high number of TN patients starting HAART in this
time period, as patients recently starting HAART were comparatively over-represented. By 5 years after
starting HAART the incidence of TCF in TN patients was approaching the rate seen in TE patients, although
the confidence intervals around the estimate were comparatively wide. The initially low and then steady rate
of TCF in TE patients may simply reflect that, in the first years after starting HAART, patients with rapid TCF
had more resistance, were least able to adhere to the new regimens, or differences in early treatment
guidelines (26). Once such patients have TCF, the remaining TE patients may fail at a rate similar to that
seen in TN patients and the curves will increase with extended exposure to HAART at the same rate.
The factors related to failing three drug classes in this study were similar to those related to failing either an
initial or second-line HAART regimen. Thus in both TN and experienced patients, a higher viral load at
starting HAART was associated with a higher risk of TCF, as seen for patients starting HAART or a second-
line HAART regimen (5, 27). Among TE patients adding new nucleosides to the initial HAART regimen
resulted in a lower risk of TCF, as reported from studies looking at other aspects of treatment failure (4-5, 7,
18, 21, 27-31). In addition, a higher cumulative exposure to nucleosides prior to starting HAART rather than
the number of nucleosides taken was associated with TCF, suggesting resistance may be accumulating over
time. Although a number of studies have shown a significant correlation between drug resistance and
virologic response (32-33), the role of resistance testing remains unclear (2). In patients with extensive prior
treatment and multiple treatment failures, the interpretation of resistance tests is difficult and other factors,
such as treatment history, adherence and toxicities need to be taken into account (34-36).
Results from other studies of the virologic, immunologic or clinical benefits of starting HAART at higher CD4
counts vary (37-41). In this study, when CD4 was included as a continuous variable, patients with higher
CD4 counts at starting HAART had a reduced risk of TCF in both TN and TE patients, and the increased risk
of TCF at lower CD4 counts was comparable to that of TE patients. However, the results were not so clear
when the CD4 count was categorised. In general, TN or TE patients starting HAART with a CD4 of below
200/mm3 were at an increased risk of TCF compared to patients with a CD4 count of 350/mm3 or above.
This was not statistically significant in TN patients, as found in a larger study of treatment naïve patients (39),
and this highlights that the decision to start therapy is a complicated one which depends on many different
factors. The CD4 count at starting HAART may reflect previous adherence in TE patients, and thus may be
acting as a marker for future adherence after patients start HAART.
Patients starting HAART with 4 or more drugs had a similar risk of TCF as those starting HAART with 3
drugs. Patients starting HAART with a boosted PI regimen would be categorised as starting 4 drugs, but a
further analysis which included specific regimen started showed similar results. In addition, calendar date of
starting HAART was not associated with the risk of TCF, and this was explained by the increasing CD4 count
and decreasing viral load over time at the date of starting HAART. Some patients in this analysis started
HAART back in 1997, and there have been considerable changes over time in the regimens used, the way
HAART is started and in managing toxicities. It is possible that further improvements over the next five years
make TCF less likely than that reported here. However, given that patients have up to 6 years follow-up,
those who started HAART earlier in the study will have swapped regimens over time, thus these results
represent an estimate of TCF among a heterogeneous clinic population where there are many factors
involved in which regimen to start, and the virologic threshold required for treatment failure.
There are several limitations of this study
FUSSPILZ Niemand redet gerne darüber, aber viele tragen ihn mit sich herum – der Fusspilz. Der Fusspilz ist die weitaus am meisten verbreitetste Pilzinfektion. Fusspilz ist ebenso hartnäckig wie unangenehm. Er schleicht sich klammheimlich zwischen die Zehen, unter die Nägel und macht sich an der Fusssohle breit. Fast ein Viertel der Bevölkerung der Schweiz leidet darunter. Der Fusspilz
Publicado en M. Lluch, (ed.): Bases antropológicas y culturales de la formación universitaria , Eunsa, Pamplona (en prensa) Erotismo y pornografía Jaime Nubiola " A man is known by the company his mind keeps ". Thomas B. Aldrich, Ponkapog Papers , 1903, 40. 1. Introducción En primer lugar quiero agradecer muy vivamente la invitación de D. Miguel Lluch para impartir est