Additive or sequential nucleoside analogue therapy compared with continued
zidovudine monotherapy in human immunodeficiency virus-infected patients
with advanced disease does not prolong survival: an observational study
van Leeuwen, R.; Bonsel, G.J.; Reiss, P.; Danner, S.A.; Lange, J.M.A.
DOI
10.1086/516466
Publication date
1997
Published in
The Journal of Infectious Diseases
Link to publication
Citation for published version (APA):
van Leeuwen, R., Bonsel, G. J., Reiss, P., Danner, S. A., & Lange, J. M. A. (1997). Additive
or sequential nucleoside analogue therapy compared with continued zidovudine monotherapy
in human immunodeficiency virus-infected patients with advanced disease does not prolong
survival: an observational study. The Journal of Infectious Diseases, 175, 1344-1351.
https://doi.org/10.1086/516466
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You
Additive or Sequential Nucleoside Analogue Therapy Compared with Continued
Zidovudine Monotherapy in Human Immunodeficiency Virus – Infected Patients
with Advanced Disease Does Not Prolong Survival: An Observational Study
Remko van Leeuwen, Gouke J. Bonsel, Peter Reiss, Academic Medical Center, University of Amsterdam, Department of Medicine, Division of Infectious Diseases, Tropical Medicine, and AIDS, Sven A. Danner, and Joep M. A. Lange
National AIDS Therapy Evaluation Centre, and Department of Clinical Epidemiology and Statistics, Amsterdam, Netherlands
To study the effect of sequential or additive use of zalcitabine or didanosine on survival in 308 human immunodeficiency virus – infected patients with advanced disease treated with zidovudine, an observational study using time-dependent Cox proportional hazards models was done. Changing to sequential or additive therapy was based on deterioration of a patient’s health status, a significant drop in CD4 cell count, or intolerance for zidovudine. The median CD4 cell count at baseline was 1101 106/L; 42% of patients had AIDS. The median count before a change in therapy was 501 106
/L. Additive or sequential treatment was associated with an increased risk for death (relative hazard, 1.59; 95% confidence interval [CI], 1.01 – 2.49; and 1.58; 95% CI, 1.10 – 2.37, respectively). Adjustment of the models for prognostic factors failed to substantially affect this observation. Possibly the lack of benefit in this study is because patients switched therapy at advanced stages, whereas the switch may be more effective in early disease.
Zidovudine treatment has considerable benefit in patients outcome and interpretation of the results in patients who had received prior antiretroviral treatment [5, 6]. While AIDS Clini-infected with the human immunodeficiency virus (HIV), yet
monotherapy with the drug does not prevent ultimate HIV- cal Trials Group (ACTG) protocol 175 showed a clinical benefit for patients receiving zidovudine-didanosine treatment and sig-related disease progression and death. Additive or sequential
use of other antiretroviral agents has increased the options nificant improvements in CD4 cell parameters for patients in all combination arms, the European-Australian Delta-2 study for antiretroviral treatment. Combining drugs may provide the
advantage of synergistic activity of drugs, while the emergence showed a less-pronounced benefit of combination therapy in patients previously treated with zidovudine. The most obvious of resistant viruses may be delayed (reviewed in [1, 2]) and
different cellular reservoirs of the virus may be targeted [3, 4]. difference between the studies is the difference in the baseline CD4 cell count, which was just above 3001 106/L in ACTG A combination of antiretroviral agents may be administered to
antiretroviral-naive patients or as additive treatment to patients 175 and just below 2001 106/L in Delta-2 [5, 6].
While there have been many reports of positive effects on who were previously treated with zidovudine monotherapy.
Sequential therapy, that is, switching to alternative drugs, may immunologic and virologic parameters following first-line combination, additive, or sequential therapy with various anti-also provide advantages over monotherapy, while toxicities
may be reduced compared with that seen in combination treat- retroviral agents [7 – 11], only few of them have addressed whether such treatment ultimately results in a more durable ment.
Recently, two large prospective studies that addressed the clinical benefit as judged by parameters such as disease pro-gression and survival.
impact of combination and additive treatment on the clinical
outcome of HIV infection were completed. Although these As no other prospective studies are expected to generate esti-mates of the clinical benefits of combination and additive therapy, studies were consistent with respect to the effect of combination
therapy in previously untreated patients, showing a marked there is a clear need for additional data on the optimal timing of these strategies. We analyzed data from an observational database benefit for patients receiving either a combination of
zidovu-dine and zalcitabine or zidovuzidovu-dine and didanosine over zidovu- of patients attending a large referral hospital in Amsterdam, many of whom had severe symptomatic disease and changed therapy dine monotherapy, there was a difference in the overall study
because it was believed that they were refractory or intolerant to monotherapy with zidovudine. In this population with advanced disease, we determined the effects on survival of adding
didano-Received 10 June 1996; revised 6 January 1997.
sine or zalcitabine to therapy with zidovudine or switching to
Reprints or correspondence: Dr. Remko van Leeuwen, National AIDS
Ther-apy Evaluation Centre, Academic Medical Centre, University of Amsterdam, monotherapy with these drugs as alternative treatment strategies Room F5-160, P.O. Box 22700, 1100 DE Amsterdam, Netherlands.
to continuing zidovudine monotherapy. Since this is not a random-The Journal of Infectious Diseases 1997; 175:1344 – 51
ized study, we applied specific survival techniques analyses that
q 1997 by The University of Chicago. All rights reserved.
tients in whom zidovudine treatment was not effective (doses up
Methods
to 2.25 mg/day).
Patients. We included all HIV-1 – infected patients attending Laboratory methods. Blood was usually sampled in the morn-the outpatient clinic of morn-the Academic Medical Centre in Amster- ing for the determination of the CD4 cell counts, which were dam, who started treatment with zidovudine monotherapy between measured by use of peripheral blood mononuclear cells isolated 1 January 1990 and 31 August 1994. Homosexual contacts were from heparinized venous blood by flow cytometry. This was qual-the predominant risk factor for HIV infection in this group. The ity controlled and performed according to standard methods. The hospital policy for prescribing antiretroviral treatment was based usual interval for measuring CD4 cells was every 3 months; other on the consensus guidelines of the Dutch AIDS treatment special- laboratory measures were usually sampled on each visit.
ists (reviewed in [12]). Data collection. The observational database that was used for
Zidovudine was started in asymptomatic patients with CD4 cell this study was derived by combining several computer files with countsõ300 1 106
/L or in patients with CD4 cell countsõ400 laboratory and clinical information on these patients. The data have 1 106
/L and additional laboratory indicators of a poor prognosis been computerized both for purposes of patient care (in particular or symptoms of HIV disease. Patients could either start with zido- laboratory data) as well as for several research and surveillance vudine monotherapy or participate in one of the clinical trials in purposes (in particular such data as antiretroviral treatment and which they might receive double-blinded or open-label experimen- time of death). The final database included all routine laboratory tal antiretroviral agents. The latter group of patients, who did not markers from these patients, as well as all AIDS-related illnesses start with zidovudine monotherapy, was excluded from the analy- and the date of death or last follow-up. Most data were collected sis. However, patients who initially started zidovudine monother- prospectively. However, as hospital patients are often referred from apy and were subsequently included in one of these studies were other health care settings, chart abstraction to define the patients’ not excluded from the analysis to avoid selection bias for evaluat- baseline status was allowed.
ing the effect of subsequent treatment. The 1987 AIDS case defini- Statistical analysis. Homogeneity of baseline characteristics tion for adults of the Centers for Disease Control and Prevention was established using the x2
test for categorical variables and the (CDC) [13] was used. As this definition requires patients to be at Wilcoxon rank sum test or the Kruskal-Wallis test for continuous
least 13 years of age, younger patients were excluded. variables.
Changes in antiretroviral treatment were usually based on a Cox proportional hazards models were fitted using the PROC deterioration of a patient’s health status, a significant drop in CD4 PHREG program of SAS software (release 6.03 for DOS; SAS cell count, or intolerance for zidovudine. Zidovudine was either Institute, Cary, NC). The primary outcome analyzed was survival, continued with the addition of didanosine or zalcitabine (additive defined as time from the start of zidovudine therapy (baseline) treatment) or discontinued and followed by the subsequent admin- until death. Subjects were censored at the date of their last follow-istration of didanosine or zalcitabine only (sequential therapy). up if alive at the end of the follow-up period, which was 31 For the purpose of the analysis, treatment groups were defined as December 1994. The significance of parameter estimates derived follows: patients who started additive therapy, patients who started from the proportional hazards models were tested with the Wald sequential therapy, and patients who continued zidovudine statistic [15].
monotherapy. Once patients switched to either additive or sequential therapy,
The use of acyclovir was limited to short-term treatments for continued use of the same treatment was presumed until death or herpes simplex virus infections only. Prophylaxis for Pneumocystis censoring. This method has been applied in other proportional carinii pneumonia (PCP) was strongly encouraged in all patients hazards modeling in HIV studies over the past few years [16 – 18], with CD4 cell countsõ200 1 106
/L. Preferably, a once-daily dose because of its advantage of ignoring changes in therapy that occur of cotrimoxazole was used, while monthly pentamidine inhalations after the initiation of a therapy, which could bias estimates of were available for those patients who were cotrimoxazole-intoler- effectiveness. The disadvantage of using this method is that poten-ant. A previous study has shown that in the Netherlands, compli- tial benefits can be masked or, alternatively, over-estimated if a
ance with this regimen is very good [14]. therapy change occurs more frequently in one of the study arms.
The usual starting dosage of zidovudine during the study period Therefore, the number of patients who discontinued the antiretrovi-was 200 mg three times daily or 250 mg twice daily. Didanosine ral treatment within each study arm was tabulated and tested. No became available in August 1990: at first within a double-blinded evidence was found for a selective discontinuation in one of the randomized study comparing two different dosages of the drug study arms.
(750 vs. 200 mg/day; didanosine-ALPHA study), from October The use of Cox proportional hazards models with updated mea-1991 onwards in a parallel-track program (dosages dependent on surements of the covariables provides a means to update the prog-body weight, 375 – 750 mg/day), and from 1993 as a registered nosis of a patient on the basis of each new observation. Thus, the drug (standard dosage, 200 mg twice daily). Within the didanosine- timing of sequential or additive treatments is taken into account, ALPHA study, patients were required to be zidovudine-intolerant; which controls for survival bias: For example, a patient who sur-within the other programs, the drug could also be administered vives for a longer period may be more likely to receive (new) to patients whose condition was deteriorating during zidovudine treatment compared with the likelihood for a patient who has a monotherapy. Zalcitabine became available in the fall of 1991, shorter survival (reviewed in [19]). Also, the inclusion of CD4 initially for a short period through a community-based HIV treat- cell counts and other factors as time-dependent variables adjusts ment organization and later within an official parallel-track pro- for these prognostic factors at each time the treatment groups are compared. The method of complete time-dependent adjustment gram as monotherapy or in combination with zidovudine in
pa-might, however, mask potential benefits of a switch in therapy, in mined significant (Põ .15) were retained in the model and were subsequently used to adjust the models that included the treatment case this benefit could be explained by changes induced by
treat-ment in the variables that are used for adjusttreat-ment of the models. variables. The relatively high P value for inclusion of variables in the intermediate models was chosen so as not to miss potentially Therefore, additional models were constructed that adjusted until
the change in therapy but were unadjusted thereafter. important parameters that might be less significant because of the power of the study. Since the CD4 cell count has been well-The following covariates with updated measurements were
in-cluded in the models at monthly intervals from (and including) established as a predictive marker, we decided to include this variable in all multivariate models. Departures from the propor-the time of propor-the first use of zidovudine: propor-the presence of Kaposi’s
sarcoma without other AIDS-defining conditions (dichotomous: tionality assumption of the models were examined by multiplying the covariates with the log-transformed follow-up time (the log yes or no), presence of all other AIDS-defining conditions (i.e.,
including Kaposi’s sarcoma simultaneously presenting with other scale was used because of the skewed distribution of the follow-up time) and testing whether these interaction variables had a AIDS-defining events, dichotomous: yes or no). The treatment
variables were examined in two ways: One model included the significant contribution to the models. use of zalcitabine either as sequential or additive therapy
(dichoto-mous: yes or no) and the use of didanosine also as additive or
Results
sequential therapy (dichotomous: yes or no). Another model exam-ined the use of additive therapy of zidovudine with either
didano-Three hundred eight patients started zidovudine treatment
sine or zalcitabine (dichotomous: yes or no) or the use of sequential
between 1 January 1990 and 31 August 1994. At baseline (the
therapy with either didanosine or zalcitabine following zidovudine
initiation of zidovudine therapy), CD4 cell counts, hematologic
therapy (dichotomous: yes or no). To adjust for potential use of
parameters, and the number of patients with AIDS did not
other antiretroviral medication, the use of other experimental or
significantly differ between the groups (table 1). Patients
re-blinded antiretroviral drugs was examined (dichotomous: yes or
ceiving sequential therapy (nÅ 66) had received zidovudine
no).
Since the indication for PCP prophylaxis was having a CD4 cell monotherapy for a shorter period than had the patients receiving
countõ200 1 106
/L, by including the CD4 cell count as time- additive therapy (nÅ 43) (P Å .01, log rank test). The patients
dependent variable, we have controlled for the use of PCP prophy- receiving additive therapy had a slightly higher median CD4 laxis. We did not control the models for CDC class B symptoms, cell count prior to the change in therapy: 701 106/L (90%
as a vast majority of patients were symptomatic at baseline. reference range, 10 – 480) versus 401 106/L (90% reference
Given the average frequency at which CD4 cell counts were
range, 0 – 360) for those receiving sequential therapy (PÅ .36,
sampled in these patients, laboratory values were included in the
Wilcoxon rank sum test). The number of deaths within the
model at 3-monthly intervals. If more than one value was available
study was 174 (table 1).
within this period, the first known value was included in the model.
The effects of potential prognostic laboratory parameters
The variables were coded as follows: hemoglobin level
(continu-were first estimated in unadjusted time-fixed (data not shown)
ous: per mmol/L increment), leukocyte count (continuous: per 109
/
and time-dependent univariate models (table 2), which showed
L increment), platelet count (continuous: per 109
/L increment),
CD4 cell count (continuous: per 1001 106/L increment). The age that the probability of survival was predicted by the CD4 cell
of the patient at baseline (continuous: per 10 years increment) and count, hemoglobin levels, the leukocyte count, and the platelet
the timing of the start of zidovudine monotherapy (dichotomous: count. We examined the hematologic parameters separately in before 1992 [about halfway through the study period] or from a time-dependent multivariate model to investigate their mutual 1992 onwards) were examined as time-fixed covariables, given the predictive independence, which showed that the hemoglobin deterministic nature of their changes over time (reviewed in [20]).
level was the primary predictive parameter for survival.
To investigate the influence of the duration of the preceding period
The clinical parameters were also first fitted in unadjusted
of zidovudine use for the outcome with sequential or additive
time-fixed and time-dependent models. The diagnosis of AIDS
therapy, three time-dependent dummy variables were fitted
repre-and the age of the patient were both predictive for survival,
senting categories of time since starting zidovudine (õ12 months,
while the year of inclusion in the study was not (table 2). Since
12 – 24 months, and§24 months).
it is known that survival is influenced by the initial
AIDS-To exclude the possibility of introducing a bias by a difference
between the study arms in number of patients who permanently defining event [21 – 23], we analyzed in a bivariate model the
discontinued their antiretroviral treatment at a considerable time patients who were diagnosed with Kaposi’s sarcoma separately before death, we tabulated the number of patients with such a from those who presented with other AIDS-defining events. In premature discontinuation. No difference was found in the number this time-dependent model, a diagnosis of Kaposi’s sarcoma of patients (x2, PÅ .66) that discontinued therapy prematurely or
was only marginally significantly predictive for survival, while
in the average time of discontinuation prior to death (Wilcoxon
the predictive value of a diagnosis of any other AIDS-defining
rank sum test, PÅ .39).
event was improved. When we adjusted this bivariate model
A stepwise procedure was performed to determine the relative
for the CD4 cell count, a diagnosis of Kaposi’s sarcoma was
importance of the laboratory and clinical covariates. The first step
no longer found predictive for survival (relative hazard, 1.37;
in the development of a final model was to identify the most
Table 1. Demographics of patient population in study of additive or sequential therapy with didanosine (ddI) or zalcitabine (ddC) versus zidovudine monotherapy.
Continued
zidovudine Sequential
monotherapy Additive therapy therapy P
No. of patients 199 43 66 —
No. receiving ddI therapy — 10 43 —
No. receiving ddC therapy — 33 23 —
No. receiving other antiretroviral therapy 7 9 9 .001* No. with AIDS-defining condition (%)
Kaposi’s sarcoma only 15 (8) 3 (7) 7 (11) .70*
Other illnesses 72 (36) 9 (21) 22 (33) .16*
Age, years 37 (26 – 57) 40 (27 – 51) 38 (28 – 54) .58†
CD4 cell count1 106
/L 130 (10 – 470) 130 (10 – 470) 100 (0 – 500) .09†
Hemoglobin concentration, mmol/L 8.0 (6.0 – 9.9) 8.3 (6.2 – 10.2) 8.0 (6.6 – 9.6) .47†
Platelet count1 109/L 177 (55 – 314) 215 (95 – 455) 201 (116 – 329) .17†
White blood cell count1 109/L 4.2 (2.2 – 8.9) 3.1 (2.5 – 9.5) 4.0 (1.3 – 7.4) .15†
No. of deaths 106 25 43 .23*
NOTE. Age and blood analysis data are given as median (90% reference range). * x2test.
†
Kruskal-Wallis test.
A stepwise procedure to define statistically significant pre- use of either additive or sequential therapies versus zidovudine alone (table 4) in unadjusted time-dependent models. The use dictors for survival resulted in a time-dependent model that
included the following variables: CD4 cell count, the hemoglo- of both treatments was associated with increased probability of death. The other model included the use of didanosine or bin level, a diagnosis of AIDS (excluding those patients who
presented with Kaposi’s sarcoma only), and the age of the zalcitabine compared with the continued use of zidovudine monotherapy. Again, these changes in therapy were also associ-patient (table 3). The same variables were found when we
constructed a time-fixed model in a stepwise procedure. The ated with an increased probability of death.
To assess whether this poor prognosis of a change in therapy predictive value of the parameters in the final model was stable
when individual variables were deleted from the model. may have been confounded by other factors, we examined different models that controlled for several prognostic vari-The therapeutic variables were examined without and with
adjustments of the models. First, we compared the effect of the ables. We first considered the adjustment for CD4 cell counts in time-dependent models. Both in the model that examined the use of either zalcitabine or didanosine and in the model Table 2. Unadjusted time-dependent proportional hazards mod- that assessed the use of either additive or sequential therapy, els for survival for laboratory and clinical parameters. the inclusion of the CD4 cell count decreased the relative
haz-Relative 95% hazard confidence
Table 3. Multivariate time-dependent proportional hazards models
Variable for death interval P
of laboratory and clinical parameters that significantly predict sur-vival.
CD4 cell count (/1001 106/L
increase) 0.22 0.13 – 0.35 õ.0001
95% Hemoglobin concentration
Relative hazard confidence (/mmol/L increase) 0.67 0.60 – 0.74 õ.0001
Variable for death interval P
Platelet count (/109
/L increase) 0.77 0.63 – 0.93 .009 White blood cell count
CD4 cell count (/1001 (/109 /L increase) 0.81 0.74 – 0.90 õ.0001 106 /L increase) 0.34 0.21 – 0.55 õ.0001 AIDS (yes) 4.09 2.81 – 5.95 õ.0001 Hemoglobin concentration KS only (yes) 1.99 1.15 – 3.43 .01 (/mmol/L increase) 0.74 0.67 – 0.84 õ.0001 Other AIDS-defining events (yes) 5.05 3.45 – 7.42 õ.0001
AIDS* (yes) 1.98 1.38 – 2.85 .0002
Age (/10 years increment) 1.41 1.20 – 1.66 õ.0001
Age†
(/10 years increment) 1.34 1.13 – 1.60 .001 Year of inclusion in study
(1992 or later) 0.94 0.67 – 1.32 .73
* Excluding AIDS diagnoses based on Kaposi’s sarcoma only. Other experimental therapy (yes) 0.57 0.27 – 1.17 .12 †
Table 4. Multivariate time-dependent hazards models for survival for treatment parameters, adjusted for laboratory and clinical parameters.
Adjusted for CD4 cell
Adjusted for CD4 cell count, hemoglobin, age,* Adjusted for CD4 cell count, count for complete and AIDS†
for complete hemoglobin, age,* and Unadjusted models study period study period AIDS†
until therapy switch
RH (95% CI) P RH (95% CI) P RH (95% CI) P RH (95% CI) P
Additive therapy 1.59 (1.01 – 2.49) .04 1.53 (0.97 – 2.40) .07 1.72 (1.07 – 2.77) .03 2.10 (1.27 – 3.48) .004 Sequential therapy 1.58 (1.10 – 2.37) .01 1.36 (0.94 – 1.97) .10 1.41 (0.97 – 2.06) .08 1.62 (1.04 – 2.53) .03 ddI (additive/sequential) 1.39 (0.93 – 2.10) .11 1.27 (0.84 – 1.92) .26 1.32 (0.86 – 2.03) .20 1.24 (0.27 – 2.13) .44 ddC (additive/sequential) 1.80 (1.22 – 2.65) .003 1.58 (1.07 – 2.32) .02 1.72 (1.14 – 2.60) .01 2.06 (1.35 – 3.14) .0008
NOTE. All patients with CD4 cell counts£200 1 106
/L were offered Pneumocystis carinii pneumonia prophylaxis; acyclovir was used intermittently for herpes simplex virus infections only. These two factors were therefore not used to adjust models. RH, relative hazard for death; CI, confidence interval.
* Fitted as time-fixed variate.
†Excluding AIDS diagnoses based on Kaposi’s sarcoma only.
ard for death compared with that in the unadjusted models. clinical status or laboratory indicators, which are associated with a poor survival. Adjustment of the models for the CD4 Subsequent adjustment in time-dependent models for
hemoglo-cell count resulted in lower estimates of the risk for death bin levels, age, and the presence of AIDS-defining conditions
after switching therapy compared with the estimates from the yielded models with similar relative hazards.
unadjusted models, but no survival differences were detected Although the adjustment for several prognostic variables is
between the study groups. Nevertheless, a potential benefit necessary to deal with survivor bias, it might mask potential
of a switch in therapy may have been overshadowed by a benefits of a switch in therapy, if this benefit could be explained
deterioration of HIV disease in our patients, while additive and by increases in, for example, the CD4 cell count. To study this
sequential treatment strategies might well be effective in other in more detail, we considered time-dependent models that were
populations. Alternatively, increased toxicity of the combina-adjusted for the same set of prognostic variables until the
tion of drugs in late-stage patients may also have contributed change in therapy and were unadjusted thereafter (table 4,
right-to our findings. The absence of therapeutic benefit in late-stage hand column). By use of this method, we were not able to
patients is in accordance with the recent demonstrations of the demonstrate any additional benefit of the therapy changes.
highly dynamic nature of HIV replication [24 – 26], resulting Finally, we assessed whether the effect of alternative
regi-in regi-increasregi-ing amounts of viremia durregi-ing subsequent stages of mens would depend on the length of the preceding period of
the infection [27, 28]. There are studies that provide support zidovudine monotherapy. This time-dependent model did not
for these hypotheses. reveal any consistent or significant trend for an effect of the
A study reported by Graham et al. [18], who analyzed data duration of zidovudine treatment (data not shown). Similarly,
for subjects within the Multicenter AIDS Cohort Study when we assessed whether other (potential) HIV treatments
(MACS), showed that by use of time-dependent proportional might have biased our findings, we did not find that the use of
hazards models adjusted for CD4 cell counts, platelet counts, these alternative treatments was predictive for survival (relative
hemoglobin levels, HIV symptoms, PCP prophylaxis, and hazard, 1.07; 95% CI, 0.50 – 2.31; PÅ .85).
acyclovir use, a small decrease in the probability of death could be detected in patients who had changed zidovudine monotherapy to sequential or additive therapy with either
zalci-Discussion
tabine or didanosine. While this finding was statistically sig-The main conclusion of the study is that within this patient nificant, the absolute magnitude of this benefit was only esti-population, no survival benefit of additive therapy with di- mated at a maximum of 3 – 6 months over continued zidovudine danosine or zalcitabine or for switching to monotherapy with monotherapy.
these drugs could be demonstrated in patients who had initially A possible explanation for the seemingly different results in been treated with zidovudine monotherapy. In fact, unadjusted our study and the MACS analysis is the difference in baseline and partially adjusted models showed a small increased risk disease stage and the level of CD4 cell depletion at the start for death after switching therapy. Since an increased drop in of zidovudine monotherapy. The median CD4 cell count for CD4 cell counts was observed immediately before the switch the different treatment groups in the MACS analysis at baseline in therapy, this increased risk is likely to reflect that changes ranged from 284 to 3421 106/L, and all subjects were selected before the development of an AIDS-defining event. Although in therapy were often based on deterioration of a patient’s
our hospital protocol allowed patients to start antiretroviral subjects who entered the study with a prior diagnosis of AIDS [30]. Similarly, a smaller study comparing sequential therapy therapy at CD4 cell counts below 300 – 4001 106/L, the median
CD4 cell count prior to the start of zidovudine therapy was with didanosine to continued zidovudine in patients who had used zidovudine monotherapy for at least 6 months showed a only 1101 106/L, while 128 (42%) of the patients were
diag-nosed with AIDS before the start of zidovudine therapy. This clinical benefit on progression to new AIDS-defining events or death for patients switching to didanosine; this benefit was difference between the hospital guidelines for starting treatment
and the actual CD4 cell counts of patients at entry into the most apparent among patients with CD4 cell counts at entry of §100 1 106/L [31]. Finally, the long-term follow-up of study reflects a referral bias: A hospital population selects for
patients with later stages of HIV infection. Also, many patients study ACTG 019 on the effects of zidovudine monotherapy in asymptomatic patients suggested a more durable response in with higher CD4 cell counts did not start zidovudine
monother-apy but were included in antiretroviral drug studies. Other dif- patients with CD4 cell countsú300 1 106/L [32].
These studies provide cumulative evidence that successful ferences between both studies, such as the length of the study
period, the frequency of follow-up visits within the study, the treatment of HIV infection may depend not only on the number of drugs that are used but also on the level of immunodeficiency number of study sites, differences with respect to the clinical
care of HIV-infected patients (e.g., the type of PCP prophylaxis at which such treatment is initiated. Recent data on the changes in HIV RNA load during monotherapy with different nucleo-used in the study period), and regional differences in CD4 cell
measurements [29], may all have contributed to the different side analogues (including zidovudine) show that substantial inhibition of the virus is seen only within a narrow treatment clinical and statistical findings of both studies.
It should be appreciated that the character of both studies window following the start of treatment [33 – 35]. Therefore, a direct combination of drugs might be more effective than addi-was observational. The estimates that are provided by this study
may be biased; therefore, the interpretation of our findings tive or sequential strategies, as is also suggested by studies that show that a combination of zidovudine and lamivudine in naive should be conservative. Yet, our study, as well as the analyses
of the MACS cohort, is of key importance for the interpretation patients [36, 37] confers superior benefits on CD4 cell parame-ters compared with the addition of lamivudine in zidovudine-of the different outcomes zidovudine-of two recently completed studies zidovudine-of
combination treatment (ACTG 175 and Delta) [5, 6], as no experienced patients [38, 39]. Preliminary findings from these studies have also suggested that initial suppression of viral other prospective studies are expected to further address the
clinical outcome of combination therapy. The ACTG 175 study replication by use of the specific combination of zidovudine and lamivudine is more potent than the effects caused by the showed a statistically significant clinical benefit for zidovudine/
didanosine treatment in patients who had received prior antiret- combinations used in our analysis [39].
There may have been differential biases that could not be roviral treatment but not for zidovudine/zalcitabine treatment,
although it might be argued that the point estimates of the successfully modeled either in this study or in the MACS study [18]. These could be responsible for the different findings of relative risks for both combination arms suggests a benefit over
zidovudine monotherapy and that their confidence intervals are both studies. Another possible explanation for the lack of bene-fit in our study is that patients in this study switched therapy largely overlapping. The European-Australian Delta study, on
the other hand, demonstrated only a modest benefit for either at advanced stages, whereas this switch may be more effective in early disease. This explanation will have important clinical of the combination arms [5, 6].
The most obvious difference between both studies is the implications, as successful improvement of the effects of anti-retroviral therapy, especially in, but probably not restricted to, baseline CD4 cell count at the start of additive therapy. The
average CD4 cell count in the ACTG 175 study was 3381 advanced HIV disease, will depend on a powerful combination of drugs [40, 41]. Preliminary results from studies with triple-106/L, whereas in the Delta study it was just below 2001 106/
L. Another indication that patients with less advanced disease combination therapy including protease inhibitors have shown that even in late-stage and pretreated patients, successful anti-may benefit more from combination treatment comes from
ACTG 155, which compared a combination of zidovudine and retroviral treatment has become possible [42] (reviewed in [43]). Long-term efficacy studies on these regimens will be zalcitabine with either drug alone. This study did not
demon-strate overall differences in progression rates. Nevertheless, a observational studies; for ethical reasons a randomized control arm can no longer be included. The methodology that is used subgroup analysis within this study suggested a clinical benefit
in patients with CD4 cell countsú150 1 106/L [8]. within our study may therefore be an important tool for the analyses of these and future studies.
The above explanation for the differing outcome of antiret-roviral treatment depending on the level of immunodeficiency is supported by other prospective studies. Study ACTG 116B/
Acknowledgments 117 showed a clinical benefit of sequential monotherapy with
didanosine over continued monotherapy with zidovudine, but We gratefully acknowledge Menno de Jong, Gerrit-Jan Wever-ling, and Jan Tijssen for their thoughtful review of the manuscript.
sis of data in the Multicenter AIDS Cohort Study. Ann Intern Med
We also acknowledge the contributions of Sandra van Velzen,
1994; 121:100 – 8.
Petra Hollak, Joyce Kramp, Karin Rock, Lilian Marsman, Nadine
17. Lundgren JD, Phillips AN, Pedersen C, et al. Comparison of long-term
Pakker, Antoon Hekman, Arie Verboon, Jan Dols, Geert
Haver-prognosis of patients with AIDS treated and not treated with zidovudine.
kamp, Karel Koll, and Jan Karel Eeftinck Schattenkerk, who
par-JAMA 1994; 271:1088 – 92 (comment in: JAMA 1994; 271:1121 – 2,
ticipated in the collection, computerization, and retrieval of these
JAMA1994; 272:1001 – 2).
data.
18. Graham NM, Hoover DR, Park LP, et al. Survival in HIV-infected patients who have received zidovudine: comparison of combination therapy with sequential monotherapy and continued zidovudine monotherapy. Ann References
Intern Med1996; 124:1031 – 8. 1. Richman DD. Resistance of clinical isolates of human immunodeficiency
19. Glesby MJ, Hoover DR. Survivor treatment selection bias in observational virus to antiretroviral agents. Antimicrob Agents Chemother1993; 37:
studies: examples from the AIDS literature. Ann Inter Med1996; 124: 1207 – 13.
999 – 1005. 2. Boucher CA. Clinical significance of zidovudine-resistant human
immuno-20. Altman DG, De Stavola BL. Practical problems in fitting a proportional deficiency viruses. Res Virol1992; 143:134 – 6.
hazards model to data with updated measurements of the covariates. 3. Watson AJ, Wilburn LM. Inhibition of HIV infection of resting peripheral
Stat Med1994; 13:301 – 41. blood lymphocytes by nucleosides. AIDS Res Hum Retroviruses1992;
21. Ammann AJ, Abrams D, Conant M, et al. Acquired immune dysfunction in 8:1221 – 7.
homosexual men: immunologic profiles. Clin Immunol Immunopathol 4. Gao WY, Shirasaka T, Johns DG, Broder S, Mitsuya H. Differential
phos-1983; 27:315 – 25. phorylation of azidothymidine, dideoxycytidine, and dideoxyinosine in
22. van Leeuwen R, van den Hurk PJ, Mulder JW, et al. Opportunistic resting and activated peripheral blood mononuclear cells. J Clin Invest
diseases as measure of immunodeficiency in HIV infection. In: Ja-1993; 91:2326 – 33.
nossy G, Autran B, Miedema F, eds. Immunodeficiency in HIV 5. Delta Coordinating Committee. Delta: a randomised double-blind
con-infection and AIDS. EC/FERS/MRC workshop on immunodefi-trolled trial comparing combinations of zidovudine plus didanosine or
ciency in HIV-1 infections (Windsor, UK). Basel, Switzerland: zalcitabine with zidovudine alone in HIV infected patients. Lancet1996;
Karger,1992:54 – 63. 348:283 – 91.
23. Lundgren JD, Pedersen C, Clumeck N, et al. Survival differences in Euro-6. Hammer SM, Katzenstein DA, Hughes MD, et al. A trial comparing
pean patients with AIDS, 1979 – 89. BMJ1994; 308:1068 – 73. nucleoside monotherapy with combination therapy in HIV-infected
24. Ho DD, Neumann AU, Perelson AS, Chen W, Leonard JM, Markowitz adults with CD4 cell counts from 200 to 500 per cubic millimeter. N
M. Rapid turnover of plasma virions and CD4 lymphocytes in HIV-1 Engl J Med1996; 335:1081 – 90.
infection. Nature1995; 373:123 – 6. 7. Meng TC, Fischl MA, Boota AM, et al. Combination therapy with
zidovu-25. Wei X, Ghosh SK, Taylor ME, et al. Viral dynamics in human dine and dideoxycytidine in patients with advanced human
immunode-ficiency virus type 1 infection. Nature1995; 373:117 – 22. ficiency virus infection. A phase I/II study. Ann Intern Med1992; 116:
26. Coffin JM. HIV population dynamics in vivo: implications for genetic 13 – 20 (comment in: Ann Intern Med1992; 116:85 – 6).
variation, pathogenesis, and therapy. Science1995; 267:483 – 9. 8. Fischl MA, Stanley K, Collier AC, et al. Combination and monotherapy
27. Ho DD, Moudgil T, Alam M. Quantitation of human immunodeficiency with zidovudine and zalcitabine in patients with advanced HIV disease.
virus type 1 in the blood of infected persons. N Engl J Med1989; 321: Ann Intern Med1995; 122:24 – 32 (comment in: Ann Intern Med 1995;
1621 – 5 (comment in: N Engl J Med1989; 321:1673 – 5). 123:395).
28. Piatak M Jr, Saag MS, Yang LC, et al. High levels of HIV-1 in plasma 9. Yarchoan R, Lietzau JA, Nguyen BY, et al. A randomized pilot study of
during all stages of infection determined by competitive PCR. Science alternating or simultaneous zidovudine and didanosine therapy in
pa-1993; 259:1749 – 54 (comment in: Science pa-1993; 262:1585 – 6). tients with symptomatic human immunodeficiency virus infection. J
29. Veugelers PJ, Schechter MT, Tindall B, et al. Differences in time from HIV Infect Dis1994; 169:9 – 17 (erratum: J Infect Dis 1994; 170:260).
seroconversion to CD4/lymphocyte end-points and AIDS in cohorts of
10. Schooley RT, Ramirez-Ronda C, Lange JM, et al. Virologic and
immuno-homosexual men. AIDS1993; 7:1325 – 9. logic benefits of initial combination therapy with zidovudine and
zalci-30. Kahn JO, Lagakos SW, Richman DD, et al. A controlled trial comparing tabine or didanosine compared with zidovudine monotherapy. J Infect
continued zidovudine with didanosine in human immunodeficiency vi-Dis1996; 173:1354 – 66.
rus infection. N Engl J Med1992; 327:581 – 7 (comment in: N Engl J 11. Larder BA, Kohli A, Bloor S, et al. Human immunodeficiency virus type
Med1992; 327:639 – 41, N Engl J Med 1992; 327:1598 – 9). 1 drug susceptibility during zidovudine (AZT) monotherapy compared
31. Spruance SL, Pavia AT, Peterson D, et al. Didanosine compared with with AZT plus 2*,3*-dideoxyinosine or AZT plus 2*,3*-dideoxycytidine
continuation of zidovudine in HIV-infected patients with signs of clini-combination therapy. J Virol1996; 70:5922 – 9.
cal deterioration while receiving zidovudine. A randomized, double-12. van Everdingen JJ, Lange JM. Consencus vroegtijdige interventie met
blind clinical trial. Ann Intern Med1994; 120:360 – 8. zidovudine. Ned Tijdschr Geneeskd1991; 135:1547 – 50.
32. Volberding PA, Lagakos SW, Grimes JM, et al. A comparison of immedi-13. Centers for Disease Control. Revision of the CDC surveillance case
defini-ate with deferred zidovudine therapy for asymptomatic HIV-infected tion for acquired immunodeficiency syndrome. MMWR1987; 36(suppl
adults with CD4 cell counts of 500 or more per cubic millimeter. N S1):1S – 5S.
Engl J Med1995; 333:401 – 7 (comment in: N Engl J Med 1995; 333: 14. Schneider MM, Hoepelman AI, Eeftinck Schattenkerk JK, et al. A
con-450 – 1). trolled trial of aerosolized pentamidine or
trimethoprim-sulfamethoxa-33. Schuurman R, Nijhuis M, van Leeuwen R, et al. Rapid changes in human zole as primary prophylaxis against Pneumocystis carinii pneumonia in
immunodeficiency virus type 1 RNA load and appearance of drug-patients with human immunodeficiency virus infection. N Engl J Med
resistant virus populations in persons treated with lamivudine. J Infect 1992; 327:1836 – 41 (comment in: N Engl J Med 1993; 328:1499).
Dis1995; 171:1411 – 9. 15. Collet D. Modelling survival data in medical research. London:
Chap-34. de Jong MD, Veenstra J, Stilianakis NJ, et al. Host-parasite dynamics and man & Hall,1994.
outgrowth of virus containing a single K70R amino acid change in 16. Stein DS, Graham NM, Park LP, et al. The effect of the interaction of
immunodefi-ciency virus type 1 RNA load suppression by zidovudine. Proc Natl 39. Bartlett JA, Benoit SL, Johnson VA, et al. Lamivudine plus zidovudine Acad Sci USA1996; 93:5501 – 6. compared with zalcitabine plus zidovudine in patients with HIV infec-35. Loveday C, Kay S, Tenant-Flowers M, et al. HIV-1 RNA serum-load and tion. A randomized, double-blind, placebo-controlled trial. Ann Intern
resistant viral genotypes during early zidovudine therapy. Lancet1995; Med1996; 125:161 – 72.
345:820 – 4. 40. Lange JMA. Combination antiretroviral therapy: back to the future? Drugs 36. Katlama C, Ingrand D, Loveday C, et al. Safety and efficacy of lamivudine- 1995; 49(suppl 1):32 – 7.
zidovudine combination therapy in antiretroviral-naive patients. A
ran-41. Ho DD. Time to hit HIV, early and hard [editorial]. N Engl J Med1995; domized controlled comparison with zidovudine monotherapy. JAMA
333:450 – 1. 1996; 276:118 – 25.
42. Gulick RM, Mellors J, Havlir D, et al. Potent and sustained antiretroviral 37. Eron JJ, Benoit SL, Jemsek J, et al. Treatment with lamivudine,
zidovu-activity of indinavir (IDV), zidovudine (ZDV) and lamivudine (3TC) dine, or both in HIV-positive patients with 200 to 500 CD4/cells per
[abstract Th.B.931]. In: Program and abstracts: XI International Confer-cubic millimeter. N Engl J Med1995; 333:1662 – 9 (comment in: N Engl
ence on AIDS (Vancouver, Canada), Vancouver: WordPower Business J Med1995; 333:1704 – 5).
Services,1996. 38. Staszewski S, Loveday C, Picazo JJ, et al. Safety and efficacy of
lamivu-43. Carpenter CC, Fischl MA, Hammer SM, et al. Antiretroviral therapy for dine-zidovudine combination therapy in zidovudine-experienced
pa-HIV infection in 1996. Recommendations of an international panel. tients. A randomized controlled comparison with zidovudine
