Hogg RS, Heath KV, Yip B, Craib KJP, O'Shaughnessy MV, Schechter MT, Montaner JSG. Improved Survival Among HIV-Infected Individuals Following Initiation of Antiretroviral Therapy. JAMA. 1998;279(6):450-454. doi:10.1001/jama.279.6.450
From the British Columbia Centre for Excellence in HIV/AIDS, St Paul's Hospital (Drs Hogg, O'Shaughnessy, Schechter, and Montaner; Mss Heath and Yip; and Mr Craib), Departments of Health Care and Epidemiology (Drs Hogg, Schechter, and Montaner), Pathology (Dr O'Shaughnessy), and Medicine (Dr Montaner), Faculty of Medicine, University of British Columbia, and Canadian HIV Trials Network (Drs O'Shaughnessy, Schechter, and Montaner), Vancouver, British Columbia.
Context.— Clinical trials have established the efficacy of antiretroviral therapy
with double- and triple-drug regimens for individuals infected with the human
immunodeficiency virus (HIV), but the effectiveness of these regimens in the
population of patients not enrolled in clinical trials is unknown.
Objective.— To characterize survival following the initiation of antiretroviral
therapy among HIV-infected individuals in the province of British Columbia.
Design.— Prospective, population-based cohort study of patients with antiretroviral
therapy available free of charge (median follow-up, 21 months).
Setting.— Province of British Columbia, Canada.
Patients.— All HIV-positive men and women 18 years of age or older in the province
who were first prescribed any antiretroviral therapy between October 1992
and June 1996 and whose CD4+ cell counts were less than 0.350×109/L.
Main Outcome Measures.— Rates of progression from initiation of antiretroviral therapy to death
or a primary acquired immunodeficiency syndrome (AIDS) diagnosis for subjects
who initially received zidovudine-, didanosine-, or zalcitabine-based therapy
(ERA-I) and for those who initially received therapy regimens including lamivudine
or stavudine (ERA-II).
Results.— A total of 1178 patients (951 ERA-I, 227 ERA-II) were eligible. A total
of 390 patients died (367 ERA-I, 23 ERA-II), yielding a crude mortality rate
of 33.1%. ERA-I group subjects were almost twice as likely to die as ERA-II
group subjects, with a mortality risk ratio of 1.86 (95% confidence interval
[CI], 1.21-2.86; P=.005). After adjusting for Pneumocystis carinii and Mycobacterium
avium prophylaxis use, AIDS diagnosis, CD4+ cell count,
sex, and age, ERA-I participants were 1.93 times (95% CI, 1.25-2.97; P=.003) more likely to die than ERA-II participants. Among
patients without AIDS when treatment was started, ERA-I participants were
2.50 times (95% CI, 1.59-3.93; P<.001) more likely
to progress to AIDS or death than ERA-II participants.
Conclusion.— The HIV-infected individuals who received initial therapy with regimens
including stavudine or lamivudine had significantly lower mortality and longer
AIDS-free survival than those who received initial therapy with regimens limited
to zidovudine, didanosine, and zalcitabine.
ANTIRETROVIRAL therapy has been shown to prolong survival in persons
with acquired immunodeficiency syndrome (AIDS) and those with intermediate-stage
human immunodeficiency virus (HIV) infection.1- 6
Between 1990 and 1995, didanosine and zalcitabine were increasingly used in
combination with zidovudine. The available options for combination antiretroviral
therapy expanded further when positive results were reported in clinical trials
involving lamivudine and stavudine7- 9
and, more recently, in trials involving protease inhibitors and nonnucleoside
reverse transcriptase inhibitors.10- 18
Recently, viral load–driven therapy has been adopted, as a growing body
of evidence has established plasma viral load as a meaningful predictor of
disease progression.19- 21
Overall, these developments have resulted in significant changes to the therapeutic
management of this disease22,23
and have led to a dramatic increase in the use of double- and, more recently,
triple-combination antiretroviral therapy regimens among HIV-infected persons.
Through clinical trials, much evidence has been generated in support
of newer treatment strategies. However, such evidence has not yet been obtained
from population-based studies. Therefore, we characterized the treatment regimens
and survival rates of HIV-infected individuals over time in the province of
British Columbia, building on previous research that showed a substantial
decline in AIDS-related mortality in British Columbia in recent years.24
Since the introduction of zidovudine monotherapy in British Columbia
in 1986, antiretroviral drugs have been centrally distributed at no cost to
eligible HIV-infected individuals. In October 1992, the Drug Treatment Program
became the responsibility of the British Columbia Centre for Excellence in
HIV/AIDS (the Centre). From 1986 to 1997, a total of 4775 HIV-positive British
Columbians received antiretroviral therapy in British Columbia. Of these,
3935 had ever been enrolled in the Drug Treatment Program, and 2192 were currently
receiving antiretroviral therapy at the time of this report. The Centre's
Drug Treatment Program remains the only free source of antiretroviral medications
in the province.
The Centre distributes antiretroviral drugs based on specific guidelines
generated by the Centre's Therapeutic Guidelines Committee. These guidelines
as well as treatment updates are distributed to all physicians participating
in the Drug Treatment Program. In 1992, the therapeutic guidelines recommended
double-combination therapy for individuals with CD4+ cell counts
lower than 0.350×109/L. In December 1995, this recommendation
was expanded to make double-combination therapy available to everyone with
CD4+ cell counts lower than 0.500×109/L.
A total of 5 antiretroviral agents were available in the province of
British Columbia during the study period. The nucleoside analogue zidovudine
has been available since 1986. The other 4 nucleoside analogues were made
available over a period of 3 years: didanosine and zalcitabine in October
1992, lamivudine in February 1994, and stavudine in April 1993. The dates
for lamivudine and stavudine reflect the availability of these therapies through
compassionate release25 ; lamivudine did not
become widely available as first-line therapy until December 1995 and stavudine
until July 1996. Protease inhibitors and nonnucleoside reverse transcriptase
inhibitors did not become available through the program until the summer of
Physicians enrolling an HIV-positive individual into the Centre's Drug
Treatment Program must complete a drug request enrollment form. The enrollment
request form acts as a legal prescription and compiles information on the
HIV-positive applicant's address and enrolling physician, HIV-specific drug
history, CD4+ cell counts, and current drug requests. Each request
is reviewed by a qualified practitioner to ensure that it meets the Centre's
established therapeutic guidelines.23 Approved
prescriptions are renewed every 2 months. At the time of the initial refill,
each participant is asked to complete an enrollment survey and a program consent
form, while the physician is asked to complete a clinical staging form. Patients
complete the survey and physicians complete the clinical staging forms annually.
The clinical staging form records participant-specific information on HIV-
and AIDS-related conditions according to the World Health Organization clinical
This analysis was restricted to all HIV-positive men and women who were
antiretroviral naive and were first prescribed any antiretroviral therapy
between October 1, 1992, and June 30, 1996. To limit the effect of changing
therapeutic guidelines, we restricted our analysis to persons who had CD4+ cell counts less than 0.350×109/L, thereby ensuring
that they were eligible for combination therapy throughout the study period.
Study subjects were divided into those who initially received therapy with
zidovudine, didanosine, and zalcitabine, denoted as ERA-I, and those who initially
received therapy regimens including lamivudine or stavudine, denoted as ERA-II.
The primary and secondary end points in this analysis were death and
a primary AIDS diagnosis, respectively. Deaths and AIDS diagnoses during the
follow-up period were identified on a continuous basis from physician reports
and through record linkages carried out with the British Columbia provincial
AIDS registry and Division of Vital Statistics. All-cause mortality was used
since more than 90% of deaths among participants were directly attributable
to HIV-related causes. Baseline clinical information, including primary AIDS
diagnosis and use of Pneumocystis carinii pneumonia
and Mycobacterium avium prophylaxis, was obtained
directly from the Drug Treatment Program records. Information on clinical
illnesses defined according to the 1993 Centers for Disease Control and Prevention
AIDS definition27 was collected from physician
reports and record linkages carried out in collaboration with the provincial
and national AIDS registries.
For the purposes of analysis, statistical methods followed the intent-to-treat
principle, with subjects retained in their initial treatment groups (ERA-I
or -II) irrespective of whether ERA-I participants subsequently switched to
regimens available in ERA-II. The outcomes examined were time from the start
of antiretroviral therapy to death or a primary AIDS diagnosis. Cumulative
mortality and AIDS-free survival rates were estimated using Kaplan-Meier methods.
Survival functions were compared using a log-rank test. Event-free subjects
were right-censored as of June 30, 1997. Persons unavailable for follow-up
were censored at the date of last known contact with the Drug Treatment Program.
Statistical comparisons were conducted using distribution-free methods.28 Categorical variables and ordinal and skewed continuous
variables were compared using the Mantel-Haenszel and Wilcoxon rank sum tests,
respectively. The Fisher exact test was used for 2×2 contingency tables
in which any of the expected cell frequencies was less than 5.
Cox proportional hazard models were used to estimate the hazard of death
and to estimate AIDS-free survival for the ERA-I group relative to the ERA-II
group, with associated 95% confidence intervals (CIs).29
In this analysis, we adjusted for a number of salient prognostic variables
at baseline, including use of P carinii pneumonia
and M avium prophylaxis, CD4+ cell count,
AIDS diagnosis, age, and sex. A diagnosis of AIDS, sex, and use of P carinii pneumonia and M avium prophylaxis
at entry were treated as fixed binary variables (yes vs no). Age (in years)
and CD4+ cell count (per 0.100×109/L) at baseline
were modeled as continuous variables. All reported P
values are 2-sided.
A total of 1687 individuals in British Columbia were first prescribed
antiretroviral therapy between October 1, 1992, and June 30, 1996. We excluded
446 subjects from this analysis because they did not meet the inclusion criteria,
ie, had baseline CD4+ cell counts of 0.350×109/L
or greater (408 participants) or because they were less than 18 years old
(38 participants). A further 63 subjects were excluded because CD4+
cell counts were not available within 1 year prior to the start of antiretroviral
treatment. The total study sample was based on the remaining 1178 subjects
(951 ERA-I, 227 ERA-II).
Compared with study participants, individuals excluded because CD4+ cell counts were not available within 1 year prior to the start of
antiretroviral treatment were more likely to be women (25% vs 9%; P=.001). However, these groups did not differ with respect to age (P=.63), P carinii pneumonia (P=.63) or M avium (P=.19) prophylaxis use, or a diagnosis of AIDS (P=.38) at baseline.
The overall median follow-up was 21 months (interquartile range, 14-34
months), with medians of 26 months (interquartile range, 17-38 months) and
15 months (interquartile range, 13-16 months), respectively, for subjects
in groups ERA-I and -II (P<.001). A total of 27
(26 ERA-I and 1 ERA-II) study subjects (2%) were unavailable for follow-up
in this analysis. In addition, 473 subjects (49.7%) in the ERA-I group switched
to ERA-II regimens prior to the end of the study period.
Of the 951 ERA-I subjects, 488 (51%) initially received monotherapy
and 463 (49%) received double therapy. Among the ERA-I subjects, 422 (44%)
were treated with zidovudine monotherapy, 266 (28%) with zidovudine and didanosine,
197 (21%) with zidovudine and zalcitabine, and 66 (7%) with other monotherapy
regimens. As shown in Figure 1,
the majority of ERA-I subjects started therapy before 1996: 124 (13%) in 1992,
325 (34%) in 1993, 204 (21%) in 1994, 231 (24%) in 1995, and 67 (7%) in 1996.
Of the 227 subjects in the ERA-II group, 205 (90%) were treated with
zidovudine and lamivudine, 10 (4%) with lamivudine, 2 (1%) with stavudine,
and 10 (4%) with other regimens, including lamivudine with didanosine or zalcitabine.
In the ERA-II group, 225 (99%) initially received antiretroviral therapy with
a regimen including lamivudine, 206 (91%) zidovudine, 2 (1%) stavudine, 5
(2%) didanosine, and 5 (2%) zalcitabine. A total of 27 ERA-II subjects (12%)
commenced therapy in 1995, and 200 (88%) in 1996.
Table 1 compares the baseline
demographic and clinical characteristics of the ERA-I and ERA-II groups. No
differences were observed at baseline between treatment groups with respect
to P carinii pneumonia (P=.92)
or M avium (P=.50) prophylaxis,
AIDS diagnosis (P=.24), or CD4+ cell count
(P=.48). However, ERA-I subjects were less likely
to be women (9% vs 13%; P=.05) and were younger (median,
36 vs 38 years; P=.02) than ERA-II subjects at baseline.
A total of 301 study subjects had AIDS at baseline (250 ERA-I, 51 ERA-II).
The numbers and percentages of the AIDS-defining illnesses in both groups
were P carinii pneumonia, 150 (50%); other opportunistic
infections, 74 (25%); Kaposi sarcoma, 37 (12%); wasting syndrome, 19 (6%);
neurologic disease, 14 (5%); and other malignant neoplasms, 7 (2%). There
was no statistical difference between the 2 groups in the proportion having
Kaposi sarcoma (P=.42) or other opportunistic infections
(P=.87). However, more subjects in the ERA-I group
than the ERA-II group had P carinii pneumonia (53%
vs 35%, P=.02).
As of June 30, 1997, a total of 390 deaths (367 ERA-I, 23 ERA-II) had
been identified, yielding a crude mortality rate of 33.1%. Figure 2 displays the Kaplan-Meier survival curves for the ERA-I
and ERA-II groups. Product-limit estimates (±SEs) of the cumulative
mortality rate at 15 months were 17.1% (±1.2%) and 10.0% (±2.0%)
for ERA-I and ERA-II subjects, respectively (P=.004).
ERA-I subjects were almost twice as likely to die as ERA-II subjects, with
a mortality risk ratio of 1.86 (95% CI, 1.21-2.86; P=.005).
The final multivariate model for the baseline factors associated with
mortality is presented in Table 2.
Use of ERA-II regimens (P=.003), a higher CD4+ cell count (P<.001), absence of AIDS (P<.001), younger age (P<.001),
and the use of M avium prophylaxis (P=.009) at baseline were independently associated with longer survival
after adjusting for sex and for use of P carinii
pneumonia prophylaxis. The adjusted mortality risk ratio from this multivariate
model indicates that study subjects in the ERA-I group were 1.93 times (95%
CI, 1.25-2.97) more likely to die following the initiation of antiretroviral
therapy than subjects in the ERA-II group.
We also restricted our analysis of survival to the 690 subjects who
were initially prescribed dual-nucleoside regimens. After adjusting for other
salient prognostic variables (use of P carinii pneumonia
and M avium prophylaxis, AIDS diagnosis, CD4+ cell count, sex, and age), ERA-I participants were 1.73 times (95%
CI, 1.09-2.74; P=.02) more likely to die than ERA-II
Finally, we repeated all relevant analyses with the time from the start
of antiretroviral therapy to diagnosis of AIDS or death as the outcomes of
interest. These analyses were restricted to the 877 subjects (701 ERA-I, 176
ERA-II) who were free of AIDS at baseline. As of June 30, 1997, a total of
300 events had been identified, 219 primary AIDS diagnoses (208 ERA-I, 11
ERA-II) and 81 deaths (71 ERA-I, 10 ERA-II). The numbers and percentages of
AIDS-defining illnesses in both groups were P carinii
pneumonia, 63 (29%); candida infections, 23 (11%); M avium complex infection, 21 (10%); cytomegalovirus infection, 13 (6%); other
opportunistic infections, 25 (11%); Kaposi sarcoma, 35 (16%); wasting syndrome,
23 (11%); neurologic disease, 8 (4%); and other malignant neoplasms, 8 (4%).Figure 3 displays the Kaplan-Meier curves
for cumulative mortality or progression to AIDS for the ERA-I and ERA-II groups.
Product-limit estimates (±SEs) of the rate of progression to AIDS or
death at 15 months were 23.8% (±1.6%) and 10.7% (±2.4%) for
ERA-I and ERA-II subjects, respectively (P<.001).
After adjusting for other prognostic variables (use of P carinii pneumonia and M avium prophylaxis,
CD4+ cell count, sex, and age), ERA-I participants were 2.50 times
(95% CI, 1.59-3.93; P<.001) more likely than ERA-II
participants to die or progress to AIDS. These results were not altered by
removing those AIDS-free participants (risk ratio, 2.45; 95% CI, 1.51-3.96; P<.001) who were receiving monotherapy at baseline or
by adjusting for an estimated AIDS reporting delay of 9 months (risk ratio,
2.23; 95% CI, 1.33-3.74; P=.002). When our analyses
were based on the time from the start of antiretroviral therapy to diagnosis
of AIDS, after adjusting for other prognostic variables, (use of P carinii pneumonia and M avium prophylaxis,
CD4+ cell count, sex, and age), ERA-I participants were 3.61 times
(95% CI, 1.95-6.67; P<.001) more likely than ERA-II
participants to progress to AIDS.
This analysis demonstrated a statistically significant improvement in
survival and AIDS-free survival among HIV-infected men and women who received
initial antiretroviral therapy using ERA-II regimens vs ERA-I regimens. Individuals
in the ERA-I group were nearly twice as likely to die as those in the ERA-II
group. This result remained statistically significant even after adjusting
for P carinii pneumonia and M avium prophylaxis use, CD4+ cell count, AIDS diagnosis, sex,
and age at baseline. Furthermore, ERA-I participants who were free of AIDS
at baseline were nearly 3 times more likely to progress to AIDS or die than
those in the ERA-II group.
Definitive evidence regarding the relative efficacy of various treatment
strategies can only be gathered in the context of randomized clinical trials.
However, it is also important to monitor the reproducibility of the benefits
obtained in controlled settings when therapies are applied to populations.
Our findings are likely to reflect, to a substantial extent, the effect of
the introduction of lamivudine, as recently established in clinical trials.30 The recently reported CAESAR trial showed that adding
lamivudine to regimens including zidovudine reduced the 1-year progression
to AIDS or death by 55%. This effect was maintained when subgroup analyses
by various CD4+ cell count ranges and prior exposure to antiretroviral
agents were performed within the CAESAR trial. In our study, the vast majority
of subjects in the ERA-II group received initial therapy with a regimen containing
lamivudine, and nearly all these subjects received both zidovudine and lamivudine.
In a population-based study, differences in outcomes between treatment
groups might result from nonrandom assignment to therapies or from differential
use of cointerventions between groups. In this context, it is reassuring to
note that no statistical differences were observed between the 2 treatment
arms at entry in P carinii pneumonia and M avium prophylaxis use, prior diagnosis of AIDS, and CD4+
Another possible concern is that subjects with more advanced disease
or symptoms may have chosen to initiate treatment sooner or to use more aggressive
therapy options.31,32 Those who
survive a short time from enrollment have less opportunity to select subsequent
treatment and are less likely to use any one treatment. In other words, rather
than treatment influencing progression to AIDS-free survival or death, survival
may lead to treatment use, thus overestimating the calculated treatment effect.
We have attempted to control these possible biases in several ways. As noted
above, comparisons of indicators of disease severity at baseline demonstrated
no significant differences between groups. Furthermore, our analysis avoids
any potential bias caused by individuals with more advanced disease discontinuing
treatment due to more severe illness. In this type of analysis, those who
switch treatment arms are retained in their original treatment group. Thus,
the fact that they have remained event-free long enough to initiate new treatment
becomes moot. Finally, as noted above, we used multivariate regression techniques
to adjust or control simultaneously for the effects of multiple baseline clinical
and demographic factors on mortality and AIDS-free survival. This approach,
commonly used in observational studies, 33
eliminates potential survival bias because the model compares only persons
in the different treatment groups who survive to the same time point.
Strengths of our study include the provision of antiretroviral treatment
at no cost, the intent-to-treat analysis, complete follow-up for 98% of the
cohort, and the extent of active reporting of deaths (n=175 [93%]; median
follow-up from date of death to date of reporting, 7 days [interquartile range,
In summary, our analysis demonstrated a significant improvement in mortality
and AIDS-free survival for men and women who received initial therapy with
regimens including stavudine or lamivudine compared with those who received
initial therapy with regimens restricted to zidovudine, didanosine, or zalcitabine.
Our results remained statistically significant even after adjusting for P carinii pneumonia and M avium
prophylaxis use, CD4+ cell count, AIDS diagnosis, sex, and age
at baseline. We therefore attribute this improvement to the introduction of
new antiretroviral therapy strategies. While the observational cohort study
is subject to a number of inherent constraints, adjustment for salient prognostic
covariates, such as age, sex, CD4+ cell count, and use of prophylaxis
for P carinii pneumonia and M avium, as well as restriction to antiretroviral-naive participants with
CD4+ cell counts lower than 0.350×109/L should
do much to attenuate these limitations. Furthermore, the treatment effect
estimated here is likely to be conservative and, for the most part, reflects
the efficacy of combinations including lamivudine and stavudine. Our results
confirm the importance of including data from large-scale observational studies
in the evaluation of the many available therapeutic strategies for persons
with HIV disease. The long-term impact of these and newer treatment strategies
remains to be explored.