| |
Virological Control during the First 6-18 Months after Initiating HAART as a Predictor for Outcome in HIV-Infected Patients: A Danish, Population-Based, 6-Year Follow-Up Study
|
| |
| |
".....Viral suppression during the first 6-18 months after HAART initiation predicts viral suppression, CD4+ cell count progression, and survival at 72 months...."
Clinical Infectious Diseases Jan 1 2006;42:136-144
Nicolai Lohse,1,2 Gitte Kronborg,3 Jan Gerstoft,4 Carsten Schade Larsen,5 Gitte Pedersen,7 Court Pedersen,1,2 Henrik Toft Sorensen,6,8 and Niels Obel1,2
1Department of Infectious Diseases, Odense University Hospital, and 2University of Southern Denmark, Odense, 3Department of Infectious Diseases, Hvidovre University Hospital, Hvidovre, 4Department of Infectious Diseases, Rigshospitalet, Copenhagen, 5Department of Infectious Diseases, Skejby University Hospital, and 6Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, 7Department of Infectious Diseases, Aalborg University Hospital, Aalborg, Denmark; and 8Department of Epidemiology, Boston University, Boston, Massachusetts
ABSTRACT
Background. Our objective was to examine whether virological control during the first 618 months after HAART initiation is a predictor for viral suppression, CD4+ cell count increase, and mortality in human immunodeficiency virus (HIV)infected patients 1890 months after initiation of highly active antiretroviral therapy (HAART).
Methods. We conducted a population-based observational cohort study in Denmark. Patients were divided into 3 groups, according to the proportion of time each patient had a detectable HIV RNA load (i.e., >400 copies/mL) during the 618 months after HAART initiation: 0% of the time interval (group 1), 1%-99% of the time interval (group 2), and 100% of the time interval (group 3). The proportion of patients with undetectable HIV RNA, CD4+ cell count changes, and mortality were examined by logistic, linear, and Cox regression analyses, respectively. We constructed cumulative mortality curves.
Results. We observed 2046 patients, for a total of 8898 person-years of follow-up that started at 18 months after HAART initiation. Mean CD4+ cell count increase rates during 72 months of follow-up were as follows: group 1, 3.3 X 106 cells/L per month (95% confidence interval [CI], 2.93.7 X 106 cells/L); group 2, 2.9 X 106 (95% CI, 2.5-3.3 X 106 cells/L); and group 3, 2.6 X 106 (95% CI, 2.0-3.3 X 106 cells/L). Survival at 72 months were as follows: group 1, 92.7% (95% CI, 90.5%94.4%); group 2, 85.6% (95% CI, 82.1%-88.5%); and group 3, 76.1% (95% CI, 70.6%-80.7%). At 72 months, 96% of group 1, 83% of group 2, and 57% of group 3 had an HIV RNA load of <400 copies/mL (P < .01). Treatment interruption before baseline was a predictor of mortality in group 2 (adjusted rate ratio, 2.94; 95% CI, 1.75-4.92]).
Conclusions. Viral suppression during the first 6-18 months after HAART initiation predicts viral suppression, CD4+ cell count progression, and survival at 72 months.
DISCUSSION
In this study, we found the degree of postprimary virological suppression (618 months after HAART initiation) positively associated with CD4+ cell count increases, survival, and the chance of having a viral load <400 copies/mL at 90 months (7.5 years) after HAART initiation.
The strength of our study was the use of a population-based, nationwide cohort and the large number of patients who were observed for at least 72 months after baseline. Few patients were lost to follow-up, and most of the viral load data were entered electronically into the database. The study recorded planned treatment interruptions as well as patient-reported noncompliance, both obtained from the medical chart. Although the algorithm for modelling viral load between measurements may be questioned, we obtained similar results when we reanalyzed data by carrying forward the previous viral load measurement. Therefore, we believe that measurement bias is likely to be small and that such bias could be in either direction.
Of 2404 patients who had initiated HAART, 181 patients were excluded because their first viral load measurement occurred >6 months after HAART initiation. These 181 patients had a cumulative 72-month survival from baseline of 91% (95% CI, 85%94%), which is similar to survival for patients included in the study. The potential selection bias introduced by the exclusion of this patient group is minor and could be in both directions. Not all patients were observed for the whole study period. Thus, the observed increase in CD4+ cell counts over time from baseline could be caused by patients with a decreasing CD4+ cell count ending follow-up early. However, only a few patients died, few were lost to follow-up, and there was an even distribution over time of patients who were censored because they reached the end of their observation period. Furthermore, we analyzed individual CD4+ cell count changes rather than computed means at different time points. Therefore, we believe that the observed CD4+ cell count increases reflect an actual improvement in immunological status.
A treatment interruption during the first 18 months of HAART was a strong predictor of death in patients with partial virological suppression. Treatment interruption could be associated with other risk factors for death, such as end-stage AIDS, drug toxicities, or non-HIV-associated conditions related to poor adherence. However, the excess mortality observed in patients who had an episode of treatment interruption before baseline remained constant throughout the 6-year period after baseline, making it unlikely that the excess could be explained by end-stage AIDS or drug toxicities at the time of treatment interruption. A smaller proportion of patients died of non-HIV-related causes of death in the subgroup with treatment interruption than in the subgroup without treatment interruption, suggesting that non-HIV-associated conditions were not responsible for the excess mortality. Treatment interruptions were not advocated in Denmark during the study period, and, therefore, comparison with the outcomes of trials examining the effect of physician-guided structured treatment interruption [28] should be performed with caution.
Episodes of low-level viremia after full virological suppression have been associated with higher levels of viral replication, drug resistance, impaired CD4+ cell count increases, and higher risk of subsequent sustained viral rebound [29-31], but other studies have used a number of different cutoff levels to define viremia, and the results have not been consistent. A recent study has shown that transient viremia in the range 50-200 copies/mL is due to random variation in the HIV RNA load around a steady state below this level [32], but, in another study, mathematical modelling has shown that these episodes are not evenly distributed among patients [33]. Our findings are in accordance with Easterbrook et al. [30], indicating strong clinical and paraclinical implications following even a short period of having an HIV RNA load of >400 copies/mL.
To our knowledge, this is the first study to take into account all available viral load measurements during a predetermined period to assess the prognostic value of intermittent viremia. Interestingly, we found that any degree of viremia involving a vial load of >400 copies/mL (groups 2A-D) was associated with higher mortality, compared with full virological suppression (group 1). The poor prognosis can be the result of a number of reasons. First, it could be associated with a drug-resistant virus emerging during the initial viremic period. Second, it could be associated with less strict virological targets in patients with competing comorbidity. Third, and what we believe is most likely, viremia and treatment interruption could be markers of poor adherence and even markers of nonHIV-related risk factors for death.
In conclusion, achieving an undetectable viral load is the primary treatment goal in daily clinical practice. Our study indicates that patients experiencing >1 viral load measurements >400 copies/mL are at considerably higher risk of clinical progression than are patients with fully suppressed virus. Additional studies are needed to explore the underlying causes of this treatment failure. Physicians should be aware of the even worse prognosis for these patients following a treatment interruption and should focus on improving adherence by means of intensive patient coaching, individualizing drug regimens, and treatment of underlying comorbidity to avoid these interruptions.
RESULTS
Study population. There were 2404 patients who had initiated HAART before 1 January 2002. The cumulative survival at 18 months after HAART initiation was 94.5% (95% CI, 93.5%-95.3%). There were 2046 patients who met the inclusion criteria at 18 months (baseline). Of the 358 patients who did not meet the inclusion criteria, 132 patients died before baseline, and 45 patients could not be observed for a full 18 months. Of these 45 patients, 28 left the country, and 17 had initiated HAART <18 months before their most recent clinic visit. An additional 181 patients who started HAART were excluded because their first viral load measurement occurred >6 months after HAART initiation. Of these 181 patients, 131 initiated HAART before mid-1997, when viral load measurements became routine for all patients. Of the 2046 patients, 1173 had undetectable viral loads throughout the whole period extending from 6 to 18 months after initiating HAART (group 1), 546 had detectable viral loads part of the time (group 2), and 327 had detectable viral loads throughout the whole period (group 3). The 2046 patients were observed for a total of 8898 person-years after baseline. Sixty-seven patients (3.2%) were lost to follow-up. Five hundred twenty-nine patients remained under observation at 72 months after baseline: 231, 182, and 116 patients from groups 1, 2, and 3, respectively (table 1). During the prediction period, 618 months after HAART initiation, viral loads were measured a median of 4 times for all groups, and the mean number of viral load measurements were as follows: 3.8 measurements, group 1; 4.3 measurements, group 2A; 4.2 measurements, group 2B; 4.2 measurements, group 2C; 4.0 measurements, group 2D; and 3.7 measurements, group 3. Patient characteristics are shown in table 2. The groups differed with respect to several characteristics, including CD4+ cell count at HAART initiation, injection drug use as the mode of infection, hepatitis C antibody status, having previous antiretroviral exposure at HAART initiation, having an AIDS diagnosis at HAART initiation, and date of HAART initiation.
Prevalence of patients with undetectable viral load. The final adjusted logistic regression model included no variables other than "patient group." At 72 months after baseline, 96% of patients in group 1 (95% CI, 93%-98%), 83% of group 2 (95% CI, 78%-89%), and 57% of group 3 (95% CI, 47%-66%) had an undetectable viral load (P < .01 for individual group comparisons). Figure 1 shows the prevalence of patients with an undetectable viral load (i.e., <400 copies/mL) commencing from baseline.
Progression of CD4+ cell counts. The final adjusted linear regression model for CD4+ cell count increase included CD4+ cell count at the beginning of each period. The mean CD4+ cell count increased significantly for 4 consecutive 18-month periods in all groups. The crude mean CD4+ cell count increase during the 72-month period after baseline was 3.3 X 106 cells/L per month (95% CI, 2.9-3.7 X 106 cells/L per month) for group 1, 2.9 X 106 cells/L per month (95% CI, 2.5-3.3 X 106 cells/L per month) for group 2, and 2.6 X 106 cells/L per month (95% CI, 2.0-3.3 X 106 cells/L per month) for group 3 (table 3). The difference between groups in this 72-month period was significant in the adjusted model (P < .01 for individual group comparisons). Figure 1 shows the progression of median CD4+ cell count commencing from HAART initiation.
Survival. The cumulative 72-month survival from baseline was 87.7% (95% CI, 85.9%-89.3%) for all patients and 92.7% (95% CI, 90.5%-94.4%), 85.6% (95% CI, 82.1%-88.5%), and 76.1% (95% CI, 70.6%-80.7%) for groups 1, 2, and 3, respectively.
The final adjusted Cox regression model included age at HAART initiation. With group 1 as the reference group, group 2 had a crude MRR of 2.38 (95% CI, 1.68-3.36), and group 3 had crude MRR of 3.96 (95% CI, 2.81-5.60). The adjusted MRRs were as follows: group 2, 2.63 (95% CI, 1.86-3.72); and group 3, 4.53 (95% CI, 3.20-6.42). When group 2 was divided into 4 subgroups (i.e., for a total of 6 groups), according to the proportion of time each patient had a detectable viral load, the mortality curves suggested similar mortality rates for subgroups A and B and for subgroups C and D (figure 2). Therefore, we compared subgroups C and D (51%-99% detectable) with A and B (1%-50% detectable) and found an age-adjusted MRR of 1.61 (95% CI, 1.00-2.59; P = .048).
Subgroup with treatment interruption. The group 2 subgroup that experienced treatment interruption for any reason during the first 18 months of HAART (n = 92) had a mortality rate of 6.87 deaths per 100 person-years at risk (95% CI, 4.57-10.34 deaths per 100 person-years). This mortality rate exceeded that for the group 2 subgroup without treatment interruption (adjusted MRR, 3.48; (95% CI, 2.10-5.79). It was also higher than that for group 3 patients without treatment interruption (adjusted MRR, 1.59; 95% CI, 0.972.60) but similar to that for group 3 patients with treatment interruption (adjusted MRR, 0.98; 95% CI, 0.52-1.87). The adjusted MRR of treatment interruption versus no treatment interruption within group 2 was at least 2.2 in all strata of the following variables: mode of infection (injection drug use vs. other), antiretroviral drug use preceding HAART initiation (naive vs. experienced), having an AIDS diagnosis before HAART initiation, having a CD4+ cell count at the time of HAART initiation (>100 vs. <100 cells/uL, and >200 vs. <200 cells u/L), sex, and date of HAART initiation (before vs. after 1 January 1999). Reasons for treatment interruption were noted for 85 of the 92 subjects. The most common reasons were compliance problems (30%), patient's wish (25%), and drug intolerance (27%). "Doctor's decision" was noted as a reason for only 6% of patients.
Causes of death. The causes of death were categorized as either HIV related (AIDS-defining conditions and bacterial infections), non-HIV related, or unknown. The prevalence of causes of death in group 1 were as follows: HIV related, 18%; non-HIV related, 58%; and unknown, 24%. The prevalence of causes of death in group 2 were as follows: HIV related, 35%; non-HIV related, 47%; and unknown, 18%. The prevalence of causes of death in group 3 were as follows: HIV related, 43%; non-HIV related, 40%; and unknown, 17%. The prevalence of causes of death among patients in group 2 without treatment interruption were as follows: HIV related, 31%; non-HIV related, 53%; and unknown, 16%. And prevalences of causes of death in group 2 with treatment interruption were as follows: HIV related, 43%; non-HIV related, 35%; and unknown, 22%.
Survival with undetectable viral load. In table 4, cumulative 72-month survival is combined with viral load measurements 72 months after baseline to predict the chance of being alive and virologically suppressed 90 months (7.5 years) after initiating HAART (conditioned on the patient being alive at baseline, 18 months [1.5 years] after initiating HAART). The probability of a successful outcome using these criteria ranged from 89% for group 1 (95% CI, 87%-90%) to 71% for group 2 (95% CI: 68%-74%), and 43% for group 3 (95% CI, 40%-46%).
SUBJECTS AND METHODS
The Danish HIV Cohort Study
The Danish HIV Cohort Study is a prospective, nation-wide, population-based cohort study of all HIV-infected individuals treated in Danish HIV clinics since 1 January 1995. Study methods have been described in detail elsewhere [25, 26]. Treatment for HIV infection in Denmark is restricted to 8 specialized treatment centers. The Danish health care system provides free, tax-supported medical care for all residents, including antiretroviral treatment for HIV-infected individuals. As of 1 March 2005, 4084 adult patients were enrolled in the study. The cohort study is ongoing, with continuous enrollment of both newly diagnosed HIV-infected individuals and individuals with HIV infection who move into the country. Use of the Danish 10-digit personal identification number enables treatment centers to avoid multiple registrations of the same patient and allows tracking of deaths and losses to follow-up due to emigration. Updates are performed annually.
Study Population
Eligible for the study were patients who initiated HAART before 1 January 2002, had at least 1 viral load measurement within 6 months after HAART initiation, and were alive at 18 months after HAART initiation. Patients not known to have died and not seen in the clinic since 1 January 2004 were considered to be lost to follow-up.
Antiretroviral Therapy
HAART was defined as combination antiretroviral treatment with at least 3 drugs, including 1 protease inhibitor (PI), or 1 nonnucleoside reverse transcriptase inhibitor (NNRTI), or abacavir. Patients classified as "naive" had been unexposed to antiretroviral treatment as of 2 weeks before HAART initiation.
Treatment Interruption
Treatment interruption was defined as a period of at least 2 weeks in which the patient was not taking antiretroviral drugs, after receiving HAART for the first time.
CD4+ Cell Count and Viral Load Measurements
To determine CD4+ cell counts at time points between measurements, the last measured value was carried forward. To model viral load values at time points between measurements, each measured value was carried forward for 30 days or until the next measurement, whichever came first. The measured viral load was extrapolated back to 30 days after the previous measurement. A viral load was considered to be detectable at a value of at least 400 copies/mL. We used this model because a (high) viral load measured at a visit to the health care clinic often leads to revision of the antiretroviral treatment regimens and, therefore, subsequent changes in viral load.
Prognostic Groups
We calculated the proportion of time that each patient had a detectable viral load during the 1-year period extending from 6 to 18 months after HAART initiation. For analysis purposes, on the basis of this proportion, patients were divided into 3 subgroups: group 1 had detectable HIV RNA for 0% of the time interval, group 2 had detectable HIV RNA for 1%-99% of the time interval, and group 3 had detectable HIV RNA for 100% of the time interval. For some analyses, patients were divided into 6 "detectable viral load" subgroups: 0% (group 1), 1%-25% (group 2A), 26%-50% (group 2B), 51%-75% (group 2C), 76%-99% (group 2D), and 100% (group 3).
Statistical Analyses
Viral load. The prevalence of patients with an undetectable viral load was calculated at baseline (18 months after HAART initiation) and every 18 months thereafter. Ratios between groups calculated at 72 months were analyzed using a logistic regression model.
CD4+ cell count. CD4+ cell count increases were computed for each of 4 consecutive 18-month periods after baseline and for the whole 72-month period. Individual increases were calculated for all patients observed throughout each period. Differences in mean CD4+ cell increases between groups were examined using a linear regression model.
Mortality. We computed the amount of time from baseline to death or end of follow-up and constructed cumulative mortality curves. Patients were censored at the time of their last clinic visit. We performed Cox proportional hazards regression analyses to estimate mortality rate ratios (MRRs) and to adjust for covariates.
Model building. Variables entered into the regression models included patient group, injection drug use as the mode of infection, hepatitis C antibody status, antiretroviral exposure preceding HAART initiation, having an AIDS diagnosis before HAART initiation, CD4+ cell count at HAART initiation (used in Cox and logistic models) or at the beginning of each period (used in the linear model), white race, sex, age at HAART initiation, initiation of HAART before 1 January 1999, and known treatment interruption before baseline. Using the patient group as the predictor variable, all other variables causing a 10% change in the risk estimates were entered into the regression models one by one [27]. Because of variation in the risk estimates, patients with a known treatment interruption were analyzed separately in the Cox model.
Prediction of survival with undetectable viral load. To predict the chance of being alive with an undetectable viral load at 72 months after baseline, cumulative survival and confidence intervals were multiplied by the proportion of patients with an undetectable viral load at that time.
|
|
| |
| |
|
|
|
