Skip Navigation

Long-Term Effectiveness of Highly Active Antiretroviral Therapy on the Survival of Children and Adolescents with HIV Infection: A 10-Year Follow-Up Study

  1. Kunjal Patel1,3,
  2. Miguel A. Herná n1,
  3. Paige L. Williams2,3,
  4. John D. Seeger5,
  5. Kenneth McIntosh4,
  6. Russell B. Van Dyke6,
  7. George R. Seage III1,3, and
  8. Pediatric AIDS Clinical Trials Group 219/219C Study Teama
  1. 1Department of Epidemiology, Harvard School of Public Health, Boston
  2. 2Department of Biostatistics, Harvard School of Public Health, Boston
  3. 3Department of Center for Biostatistics in AIDS Research, Boston
  4. 4Department of Harvard Medical School, Children' s Hospital, Boston
  5. 5Department of i3 Drug Safety, , Auburndale, Massachusetts
  6. 6Department of Pediatrics, Tulane University Health Sciences Center, New Orleans, Louisiana
  1. Reprints or correspondence: Dr. Kunjal Patel, 677 Huntington Ave., Dept. of Epidemiology, Harvard School of Public Health, Boston, MA 02115 (kpatel{at}hsph.harvard.edu).
 
Next Section

Abstract

Background. Previous observational studies found highly active antiretroviral therapy (HAART) to be associated with improved survival among human immunodeficiency virus (HIV)– infected children and adolescents. However, these studies had limited follow-up of HIV-infected children undergoing HAART. Given that HIV infection is chronic and that exposure to HAART is likely to be life-long, there is a need to evaluate the long-term effect of HAART on survival in this population.

Methods. The study included 1236 children and adolescents who were perinatally infected with HIV, who were on study or enrolled after January 1996 in a United States-based multicenter prospective cohort study (Pediatric AIDS Clinical Trials Group 219/219C), and who were not receiving HAART at baseline; subjects were observed for a maximum of 10 years through June 2006. A weighted Cox regression model was used to estimate the effect of HAART on survival, appropriately adjusted for time-varying confounding by severity.

Results. At the end of the 10-year follow-up period (median duration of follow-up, 6.3 years; interquartile range, 4.3– 9.8 years), 70% of participants had initiated HAART. Lower CD4 cell percentages, total lymphocyte counts, and albumin levels were associated with an increased probability of initiating HAART. Eighty-five deaths were observed, and the mortality hazard ratio associated with HAART, compared with non-HAART regimens, was 0.24 after adjusting for measured confounding by severity (95% confidence interval, 0.11– 0.51).

Conclusions. The use of HAART was highly effective in reducing mortality during the period 1996– 2006 among children and adolescents infected with HIV. With improved long-term survival, continued follow-up is necessary to evaluate the effects of prolonged use of HAART on potential adverse events, immune function, growth, sexual maturation, and quality of life in this population.

HAART reduces mortality among adults infected with HIV. A randomized trial among HIV-infected adults with CD4 cell counts of < 200 cells/mm3 found a 57% lower mortality associated with HAART, compared with a regimen including 2 nucleoside analogues [1]. This adult trial result may not be applicable to children, because the natural history of HIV infection in perinatally infected children differs from that in adults [2]. For example, in the first year of life, plasma HIV RNA levels in infected children are ∼ 10-fold higher than those in untreated adults and persist at high titers, reaching the steady-state values seen in adults only after ∼ 5 years of infection [3– 5]. In contrast, during primary HIV infection in adults, HIV RNA levels increase and then dramatically decrease over a period of 6– 12 months [3– 6].

Although randomized, controlled trials evaluating various antiretroviral therapies have been conducted in cohorts of children, the primary end points considered have been immunologic and virological factors, rather than morbidity and mortality [7, 8]. Quantification of the effectiveness of HAART in reducing mortality has, therefore, relied on observational studies. Several observational studies have focused on the association between HAART and mortality [9– 12], but only 2 studies have observed a large, population-based cohort of HIV-infected children and quantified the effect of HAART on mortality [9, 10]. De Martino et al. [9] observed 1142 children and found triple-combination therapy to be associated with a 71% reduction in the mortality rate (95% CI, 33%– 87%), compared with no antiretroviral therapy. Gortmaker et al. [10] observed 1028 children and found combination therapy including protease inhibitors to be associated with a 67% reduction in the mortality rate (95% CI, 42%– 81%), compared with other therapy. Both of these studies ended follow-up in 1999. Since then, the uptake of new antiretrovirals has increased, and changes in initial HAART regimens have been observed [13]. In addition, these previous studies had limited follow-up of children who initiated HAART, because HAART only became available in 1996. Because children are being exposed to HAART for extended lengths of time as a result of the chronicity of HIV infection, there is a need to evaluate the long-term effect of HAART on mortality.

These 2 observational studies were unable to appropriately adjust for measured confounding by severity over time because standard regression techniques to adjust for confounding may yield biased estimates when time-varying confounders are affected by previous treatment [14– 16]. In recent years, new analytic methods based on inverse probability weighting (i.e., marginal structural models) have been developed to appropriately adjust for this type of confounding [14– 16]. We therefore used marginal structural models to estimate the effectiveness of HAART in preventing mortality, using data for 1996– 2006 from a United States-based prospective study of HIV-infected children. This approach has been previously used to evaluate the effect of HAART on AIDS progression and mortality in adults [17], but it has never been used in the pediatric population.

Previous SectionNext Section

Patients and Methods

The study population included participants from Pediatric AIDS Clinical Trials Group (PACTG) protocols 219 and 219C, which are prospective studies designed to evaluate the long-term effects of HIV infection and in utero and postnatal exposure to antiretroviral therapy. Between April 1993 and September 2000, infected and uninfected children from >80 study sites in the United States were eligible for enrollment in PACTG 219 if they were born to HIV-infected mothers enrolled in PACTG perinatal trials or were themselves enrolled in PACTG perinatal or clinical trials and if they were < 21 years of age at study entry. In September 2000, all children in PACTG 219 were encouraged to enroll in PACTG 219C, which also expanded entry criteria to allow all HIV-infected children at the study sites to enroll in the cohort. These studies were approved by the human subjects review boards at each participating institution, and written informed consent was obtained from each child' s parent or legal guardian. The population eligible for this study included 1236 perinatally HIV-infected children enrolled in PACTG 219 and 219C from 1 January 1996 through 30 June 2006 with complete data on covariates of interest at some point prior to HAART initiation. The limited sample of behaviorally infected children in PACTG 219 and 219C prevented further analysis, thereby restricting our study population to perinatally infected children, and this time span includes the introduction of protease inhibitors into clinical practice and, therefore, captures data for those children who initiated HAART.

At each study visit, data on sociodemographic characteristics, clinical diagnoses, antiretroviral therapies, and numerous laboratory measurements were collected, including CD4 cell percentage, total lymphocyte count, WBC count, hematocrit, and albumin level. These laboratory measures have been previously identified as predictors for initiation of antiretroviral therapy [18, 19] and were therefore considered as potential confounders of the effect of HAART on mortality. HIV level is currently used by physicians to guide decisions about when to initiate HAART [19] and is also associated with mortality among HIV-infected children [20– 23]. However, HIV RNA measurements were available for only 36% of the population, because these measurements were not routinely collected or available before 2000. Therefore, we could adjust for HIV RNA level as a potential confounder only in a secondary analysis restricted to that subset. Clinical diagnoses were reviewed by a study physician and classified as category C or N/A/B on the basis of the Centers for Disease Control and Prevention (CDC) criteria for classification of HIV disease in children < 13 years of age [24] and ⩾ 13 years of age [25]. For statistical analyses, CD4 cell percentage was modeled in 4 categories (< 5%, 5%– 14%, 15%– 24%, and ⩾ 25%), CDC clinical category in 2 categories (N/A/B vs. C), total lymphocyte count in 2 categories (< 1500 cells/µ L vs. ⩾ 1500 cells/µ L), WBC count in 2 categories (< 3000 cells/µ L vs. ⩾ 3000 cells/µ L), hematocrit in 2 categories (< 35% vs. ⩾ 35%), albumin level in 2 categories (⩽ 4.0 g/dL vs. >4.0 g/dL), and HIV level in 4 categories (⩽ 400 copies/mL, 401– 5000 copies/mL, 5001– 50,000 copies/mL, and >50,000 copies/mL). Data were structured for analysis purposes into 13-week intervals, corresponding to the 3-month visit schedule in PACTG 219C, and covariate information was carried forward from the most recently observed value when missing.

HAART exposure was defined as the concomitant use of at least 3 drugs from at least 2 classes of HIV drugs. HIV drugs are classified into 3 main categories: nucleoside and/or nucleotide reverse-transcriptase inhibitors (NRTIs), nonnucleoside reverse-transcriptase inhibitors (NNRTIs), and protease inhibitors (PIs). Once a participant initiated HAART, he or she was considered to have continued HAART for the duration of follow-up. In the earlier PACTG 219 cohort, actual dates of initiation of medication or dates of changes in the use of medications were not available. After a previous evaluation of mortality using data from PACTG 219 [10], we assumed the midpoint between the visit date at which use of treatment was recorded and the date of the prior visit to be the date of initiation. We also conducted analyses in which the date of initiation was randomly assigned (unconditionally or conditionally on the basis of CD4 cell percentage) during the interval between the 2 visits. Analyses using the midpoint provided a more conservative estimate of the effect of HAART on mortality than did these alternative analyses (Appendix).

Each child was able to contribute a maximum of 40 person-visits of follow-up from their baseline visit to the last visit at which he or she was seen alive or the last visit before 30 June 2006 (i.e., completion of study), whichever came first. Children who were still alive as of their last visit were censored at that visit. For those children enrolled in PACTG 219 prior to 1 January 1996, their first study visit after this date was set as their baseline visit. For children enrolled after 1 January 1996 in either PACTG 219 or 219C, their first study visit was set as their baseline visit. For participants missing baseline data for any time-varying covariate, baseline was redefined as the first visit with complete data. Children with previous or current use of HAART at the baseline visit were excluded.

A weighted pooled logistic regression model was used to estimate the mortality hazard ratio for HAART versus no HAART, summarized over the entire follow-up period, of a marginal structural Cox proportional hazards model [15]. For comparison purposes, we also fit 2 unweighted pooled logistic models (one including only baseline covariates and the other including both baseline and time-varying covariates) that may not appropriately adjust for measured time-dependent con-founding.

To fit the weighted model, we estimated stabilized inverse probability weights for initiating HAART and for censoring [14, 15]. For each child and visit, the denominator of the weight can be viewed as the probability that they received their actual treatment history and remained uncensored up to that time, conditional on their past treatment and covariate history. These weights create a statistical population in which the probabilities of treatment with HAART and censoring are unrelated to the time-varying confounders included in the models used to estimate weights [14, 15]. Weighting, therefore, eliminates the problem in standard unweighted models of including time-varying confounders in a model that could also be affected by prior HAART or intermediates on the causal pathway from HAART to the mortality [14, 15]. We estimated the weights by fitting pooled logistic regression models for the probability of initiating HAART and of censoring at each visit given baseline and past time-varying covariates [15]. The covariates included in these models were age, sex, race/ethnicity, week of follow-up, and calendar year at baseline, together with baseline and time-varying values of CDC clinical category, CD4 cell percentage, total lymphocyte count, WBC count, hematocrit, and albumin level. Week of follow-up was modeled using flexible cubic splines with knots at 26, 65, and 91 weeks. Analyses were conducted using SAS, version 9 (SAS Institute). A sample SAS program to create weights and fit a marginal structural Cox proportional hazards model has been published [15].

Previous SectionNext Section

Results

Table 1 presents characteristics of the 1236 study participants at baseline. At baseline, 20% of subjects were < 4 years of age, 51% were female, 56% were black, 28% had CDC clinical stage C disease, and 20% had severe immunosuppression (CD4 cell percentage, < 15%). By the end of the follow-up period, 866 children (70%) had initiated HAART, and 47 deaths were observed among them, compared with 38 deaths observed among the 370 subjects who never initiated HAART. The total time of observation was 8138.7 person-years, including 5042.8 person-years of HAART, and the median length of follow-up for the cohort was 6.3 years (interquartile range, 4.3– 9.8 years). The median length of follow-up while receiving HAART was 5.8 years (interquartile range, 3.3– 8.3 years).

Figure 1
View larger version:
Figure 1

Number of subjects, deaths, and survival probability by treatment category and year of follow-up in a study of the influence of HAART on survival among HIV-infected children and adolescents.

Figure 2
View larger version:
Figure 2

Estimated effect of HAART on mortality among HIV-infected children and adolescents from unweighted (standard) and weighted models. All models include terms for the baseline hazard modeled as a cubic spline with knots at 26, 65, and 91 weeks.

Table 1
View larger version:
Table 1

Baseline demographic and clinical characteristics of the 1236 HIV– positive participants selected from the Pediatric AIDS Clinical Trials Group protocol 219/219C, January 1996– June 2006.

Figure 1 presents the number of subjects, deaths, and cumulative survival probability by treatment category and week of follow-up. At the start of follow-up, no subjects had initiated HAART. As follow-up continued, subjects began initiating HAART, with the proportion of initiators increasing over time. The deaths among the noninitiators of HAART were primarily observed within the first 2 years of follow-up, whereas the deaths among the initiators of HAART were observed throughout follow-up. This figure reveals the substantial effect of confounding by severity, particularly in the later weeks of follow-up, because HAART use appears to be associated with poorer long-term survival (i.e., higher mortality rates over time), compared with no HAART use.

The majority of the HAART regimens initiated by study participants included PIs (87%). Of these regimens, 29% also included an NNRTI. Among the regimens that included only a PI in combination with NRTIs, nelfinavir (48%) and ritonavir (44%) were the most common PIs, and lamivudine plus stavudine (48%) and lamivudine plus zidovudine (48%) were the most common NRTI components. Ritonavir with nevirapine and nelfinavir with nevirapine were the most common PI and NNRTI combinations (in 36% and 33% of the PI- and NNRTI-containing regimens, respectively). These PI and NNRTI regimens usually included 1 NRTI (stavudine in 67%) or 2 NRTIs (stavudine with lamivudine in 38% and with didanosine in 22%). Of the initial HAART regimens, 111 (13%) included only NRTIs and NNRTIs. These regimens included nevirapine (in 50% of the regimens) or efavirenz (50%), most commonly in combination with lamivudine and stavudine (31%) or didanosine and stavudine (22%).

Table 2 summarizes the time-varying predictors of HAART initiation. After adjustment for the other baseline and time-varying prognostic factors and for subject characteristics, HAART initiation was strongly associated with low CD4 cell percentages and weakly associated with low total lymphocyte counts, WBC counts, and albumin levels. CD4 cell percentages of < 5% were associated with a >5-fold increase in the rate of HAART initiation, compared with CD4 cell percentages of ⩾ 25% (hazard ratio, 5.63; 95% CI, 2.92– 10.85).

Table 2
View larger version:
Table 2

Association of time-varying prognostic characteristics with initiation of HAART by multivariable analyses.

The mortality hazard ratio of HAART versus non-HAART estimated from the weighted model was 0.24 (95% CI, 0.11– 0.51) (table 3). Baseline CD4 cell percentages of < 25% were strongly associated with >3-fold increases in mortality. Both baseline CDC clinical category C and low baseline total lymphocyte count were also independently associated with a >2-fold increase in mortality, with hazard ratios of 2.71 (95% CI, 1.47– 5.00) and 2.01 (95% CI, 1.04– 3.92), respectively.

Table 3
View larger version:
Table 3

Estimated effects of HAART and baseline prognostic characteristics on mortality.

Table A1
View larger version:
Table A1

Estimates of the effect of HAART on mortality obtained using alternative approaches for assigning dates of treatment initiation.

For comparison purposes, figure 2 presents the mortality hazard ratio estimate (0.24) for HAART versus no HAART from the weighted model and those from unweighted models that do not appropriately handle time-dependent confounding. The hazard ratio estimates from the unweighted models were as follows: unadjusted, 1.28 (95% CI, 0.81– 2.01); after adjustment for baseline covariates, 0.37 (95% CI, 0.20– 0.68); and after further adjustment for the time-varying covariates, 0.36 (95% CI, 0.19– 0.66).

In the secondary analysis restricted to the subset of the population with HIV RNA measurements (36% of participants), an HIV level of >50,000 copies/mL over time was associated with a 3-fold increase in the rate of HAART initiation, compared with an HIV level of ⩽ 400 copies/mL. The mortality hazard ratio for HAART versus no HAART estimated from the weighted model including adjustment for HIV RNA level was 0.30 (95% CI, 0.11– 0.82).

Previous SectionNext Section

Discussion

This study estimated a 76% lower mortality rate for HAART regimens, compared with non-HAART regimens, among HIV-infected children and adolescents observed from 1996 through 2006. This estimate is valid only under the assumption that all prognostic factors associated with HAART initiation were included in the analysis. We believe that this assumption was approximately true in our study, because we adjusted for the main indications for HAART initiation except HIV RNA level. A secondary analysis restricted to the subset of children with measured HIV RNA levels yielded a similar effect estimate, albeit with a wider confidence interval.

Our effect estimate (76% rate reduction) is slightly stronger than the effect observed by De Martino et al. [9] in Italian children (71%), even though these investigators compared triple combination therapy with no antiretroviral therapy, rather than with non-HAART therapies, as we did. We probably would have observed an even stronger protective effect of HAART on mortality if we had been able to perform the same comparison performed by De Martino et al. [9]. In our study cohort, however, only 20 children had not initiated any antiretroviral therapy during follow-up, thereby preventing a separate analysis. Although the ability to compare HAART regimens with no therapy has become limited in the United States and Europe as increasing numbers of children have already been exposed to antiretroviral drugs prior to enrolling in observational studies, this comparison is now less relevant, because clinical research questions have evolved to more interesting comparisons of HAART therapies with non-HAART therapies and are likely to evolve further in comparisons of novel HAART regimens with older HAART regimens.

The previous PACTG 219 study by Gortmaker et al. [10] involving children observed through 1999 found combination therapy with PIs to be associated with a 67% decrease in the mortality rate, compared with other regimens. When we mimicked their comparison in our marginal structural model analysis, we estimated a 77% reduction (95% CI, 53%– 88%). It is likely that a stronger effect would have been observed in the previous PACTG 219 study had a marginal structural model been used to further adjust for measured time-varying confounding by severity.

We also showed that standard methods to adjust for confounding tended to attenuate the association between HAART and mortality (figure 2). The unadjusted mortality hazard ratio was >1 (1.28), because sicker children were more likely than others to initiate HAART. Adjustment for baseline covariates had a dramatic effect on the estimate (it decreased to 0.37), but further standard adjustment for time-varying covariates had little effect in the estimate (0.36), which may be attributable to the potential bias introduced by including these time-varying variables in the regression model. In fact, adjustment for time-varying covariates using inverse probability weighting indicated that there was substantial time-varying confounding (the mortality hazard ratio from the weighted model was 0.24).

In our study, subjects who initiated HAART were assumed to continue receiving HAART for the duration of the study period, regardless of whether they actually continued this regimen. This assumption is consistent with clinical practice [13] and correctly classified ∼ 94% of the observed person-time. Our analysis is similar to an intent-to-treat analysis in a randomized controlled trial, which tends to provide a conservative estimate of treatment efficacy [26].

In conclusion, HAART use among children and adolescents infected with HIV was associated with significantly lower mortality than was seen with non-HAART regimens over a follow-up period of up to 10 years, supporting current pediatric guidelines, which strongly recommend the use of HAART as initial antiretroviral therapy [19]. Of course, initiation of such potent antiretroviral therapy should be weighed against any adverse events anticipated with its use and should be considered on an individual-child basis. As overall survival improves with the use of HAART, it will be increasingly important to observe infected children for long-term benefits and adverse effects of therapy. As this population ages and matures, the effects of prolonged use of HAART on immune function, growth, sexual maturation, and quality of life parameters will need to be evaluated.

Previous SectionNext Section

Pediatric AIDS Clinical Trials Group 219/219C Study Team

The following institutions and individuals participated in PACTG Protocol 219C, by order of enrollment. Baylor Texas Children' s Hospital: Mary E. Paul, Chivon D. Jackson, Faith Minglana, and Heidi Schwarzwald; University of Florida, Jacksonville: Mobeen H. Rathore, Ayesha Mirza, Kristy Champion, and Almer Mendoza; Chicago Children' s Memorial Hospital: R. Yogev and E. Chadwick; University of Puerto Rico, University Children' s Hospital AIDS Program: Irma L. Febo, Licette Lugo, Ruth Santos, and Ibet Heyer; Bronx Lebanon Hospital Center; M. Purswani, S. Baksi, E. Stuard, and M. Dummit; San Juan Hospital: M. Acevedo, M. Gonzalez, L. Fabregas, and M. E. Texidor; University of Miami: Gwendolyn B. Scott, Charles D. Mitchell, Claudia Florez, and Joan Gamber; University of Medicine and Dentistry of New Jersey: Arlene Bardeguez, Arry Dieudonne, Linda Bettica, and Juliette Johnson; Charity Hospital of New Orleans and Earl K. Long Early Intervention Clinic: M. Silio, T. Alchediak, C. Boe, M. Cowie, and R. Van Dyke; University of California, San Diego, Mother, Child, and Adolescent HIV Program: Stephen A. Spector, Rolando M. Viani, Mary Caffery, and Kimberly Norris; Howard University: Sohail Rana, Helga Finke, Patricia Yu, and Jhoanna Roa; Jacobi Medical Center: M. Donovan, R. Serrano, M. Burey, and R. Auguste; St. Christopher' s Hospital for Children, Philadelphia: J. Chen and J. Foster; Baystate Medical Center Children' s Hospital: B. W. Stechenberg, D. J. Fisher, A. M. Johnston and M. Toye; Los Angeles County Medical Center/University of Southern California: J. Homans, M. Neely, L. S. Spencer, and A. Kovacs; Children' s Hospital Boston: S. Burchett and N. Karthas; Children' s Hospital of Michigan: E. Moore and C. Cromer; St. Jude Children' s Research Hospital, Memphis: Aditya Gaur, Katherine Knapp, Nehali Patel, and Marion Donohoe; New York University School of Medicine/Bellevue Hospital: Maryam Minter, Thomas Hastings, Seham Akleh, and William Borkowsky; Children' s Hospital at Downstate: E. Handelsman, H. J. Moallem, D. M. Swindell, and J. M. Kaye; Columbia Presbyterian Medical Center and Cornell University New York Presbyterian Hospital: A. Higgins, M. Foca, P. LaRussa, and A. Gershon; Children' s Hospital of Philadelphia: Steven D. Douglas, Richard M. Rutstein, Carol A. Vincent, and Patricia C. Coburn; Children' s Hospital of Oakland: Ann Petru, Teresa Courville, Katherine Eng, and Karen Gold; University of California, San Francisco, Moffitt Hospital: Diane W. Wara, Nicole Tilton, and Mica Muscat; Children' s Hospital, University of Colorado, Denver: E. McFarland and C. Salbenblatt; Department of Pediatrics, Johns Hopkins University: N. Hutton, B. Griffith, M. Joyner, and C. Kiefner; Children' s Hospital and Regional Medical Center, University of Washington: Michele Acker, Ann J. Melvin, Kathleen M. Mohan, and Suzanne Phelps; Metropolitan Hospital Center: Mahrukh Bamji, Indu Pathak, Savita Manwani, and Ekta Patel; Children' s National Medical Center: Diana Dobbins, Deidre Wimbley, Tracy Perron, and Hans Spiegel; University of Massachusetts Medical School: K. Luzuriaga and R. Moriarty; University of Alabama at Birmingham: R. Pass and M. Crain; University of Maryland School of Medicine: D. Watson, J. Farley, K. Klipner, and C. Hilyard; Schneider Children' s Hospital: V. R. Bonagura, S. J. Schuval, C. Colter, and L. Campbell; Boston Medical Center: Stephen I. Pelton, E. R. Cooper, Lauren Kay, and Ann Marie Regan; University of Illinois: K. C. Rich, K. Hayani, M. Bicchinella, and J. Camacho; State University of New York at Stony Brook: Sharon Nachman, Denise Ferraro, Jane Perillo, and Michele Kelly; North Broward Hospital District: Ana M. Puga, Guillermo Talero, James Blood, and Stefanie Juliano; Duke University: Carole Mathison, Kareema Whitfield, Felicia Wiley, and Margaret Donnelly; Harlem Hospital: S. Champion, M. Frere, M. DiGrado, and E. J. Abrams; Cook County Hospital: James B. McAuley, Kenneth M. Boyer, Maureen Hak, and Jamie Martinez; University of South Alabama: Mary Mancao and Benjamin Estrada; Connecticut Children' s Medical Center: Juan C. Salazar and Gail Karas; University of North Carolina at Chapel Hill: Tom Belhorn, Jean Eddleman and Betsy Pitkin; Ruiz Arnau University Hospital: W. Figueroa and E. Reyes; State University of New York Upstate Medical University: L. B. Weiner, K. A. Contello, W. A. Holz, and M. J. Famiglietti; Children' s Medical Center of Dallas; University of Florida at Gainesville: R. Lawrence, J. Lew, C. Delany, and C. Duff; Children' s Hospital at Albany Medical Center: A. D. Fernandez, P. A. Hughes, N. Wade, and M. E. Adams; Lincoln Medical and Mental Health Center; Phoenix Children' s Hospital: J. P. Piatt, J. Foti, and L. Clarke-Steffen; Public Health Unit of Palm Beach County: J. Sleasman and C. Delaney; Medical College of Georgia: Stuart Foshee, Chitra S. Mani, Dennis L. Murray, and Christopher White; Yale University School of Medicine: Warren A. Andiman, Leslie Hurst, Janette de Jesus, and Donna Schroeder; Vanderbilt University Medical Center: G. Wilson; University of Rochester Medical Center: Geoffrey A. Weinberg, Francis Gigliotti, Barbra Murante, and Susan Laverty; St. Josephs Hospital and Medical Center, New Jersey: N. Hutchcon and A. Townley; Emory University Hospital: S. Nesheim and R. Dennis; University of South Florida: P. Emmanuel, J. Lujan-Zilberman, C. Graisberry, and S. Moore; Children' s Hospital of the King' s Daughters: R. G. Fisher, K. M. Cunnion, T. T. Rubio, and D. Sandifer; Medical University of South Carolina: G. M. Johnson; University of Mississippi Medical Center: H. Gay and S. Sadler; Harbor-UCLA Medical Center: Margaret A. Keller, Nasser Redjal, Spring Wettgen, and Sheryl Sullivan; Mount Sinai Medical Center: D. Johnson; Children' s Hospital of Los Angeles: J. Church, T. Dunaway, and C. Salata; Long Beach Memorial Hospital: Susan Marks, Karen Elkins, Jagmohan Batra, and Audra Deveikis; Robert Wood Johnson Medical School: S. Gaur, P. Whitley-Williams, A. Malhotra, and L. Cerracchio; Mt. Sinai Children' s Hospital: M. Dolan, J. D' Agostino, and R. Posada; The Medical Center, Pediatric Columbus, Georgia: C. Mani and S. Cobb; Medical College of Virginia: S. R. Lavoie and T. Y. Smith; Cooper Hospital-University Medical Center: A. Feingold and S. Burrows-Clark; University of Cincinnati: J. Mrus and R. Beiting; Columbus Children' s Hospital: M. Brady, J. Hunkler, and K. Koranyi; Sacred Heart Children' s Medical Center of Florida: W. Albritton; St. Luke' s/Roosevelt Hospital Center: R. Warford and S. Arpadi; Incarnation Children' s Center, New York: A. Gershon and P. Miller; Montefiore Medical-Albert Einstein College of Medicine: A. Rubinstein and G. Krienik; Children' s Hospital of Los Angeles: A. Kovacs and E. Operskalski; San Francisco General Hospital: D. Wara, A. Kamrin, S. Farrales, N. Tilton, and M. Muscat; Cornell University New York Presbyterian: R. Johan-Liang and K. O' Keefe; St. Louis Children' s Hospital: K. A. McGann, L. Pickering, and G. A. Storch; North Shore University Hospital: S. Pahwa and L. Rodriquez; and Oregon Health Sciences University: P. Lewis and R. Croteau.

Previous SectionNext Section

Acknowledgments

We thank the children and families for their participation in PACTG 219/219C and the individuals and institutions involved in the conduct of PACTG 219/219C. We also thank Dr. Roger Logan for technical assistance.

Financial support. Statistical and Data Analysis Center at Harvard School of Public Health (cooperative agreements 5 U01 AI41110 and 1 U01 AI068616) and the National Institutes of Health (AI07358 to K.P.).

Potential conflicts of interest. All authors: no conflicts.

Previous SectionNext Section

Footnotes

  • ↵Members of the study group are listed at the end of the text.

  • Received August 6, 2007.
  • Accepted September 28, 2007.
Previous SectionNext Section

Appendix

Previous SectionNext Section

Sensitivity Analyses of Time of HAART Initiation

In PACTG 219, start and stop dates of treatment were not routinely collected. Instead, at each study visit, antiretrovirals taken since the last visit were recorded. Typically, analyses of treatment within this cohort have assumed treatment to have started at the midpoint between the visit that treatment was recorded and the prior visit. We sought to assess whether our estimate of the effect of HAART on mortality was sensitive to this assumption by considering other reasonable scenarios for time of initiation between study visits.

First, analyses were conducted by randomly assigning a date of initiation between study visits using a uniform distribution. Next, more-informed approaches of assigning random dates based on CD4 cell percentages at the visit prior to treatment initiation were considered. These approaches reasoned that those with low CD4 cell percentages would be more likely to initiate treatment earlier in the interval between study visits, whereas those with higher CD4 cell percentages would be more likely to initiate treatment later in the interval. A piece-wise uniform, a truncated normal, and a continuously shifting distribution based on CD4 cell percentage were used to assign dates. For example, using the piece-wise uniform distribution, subjects with CD4 cell percentage of < 15% at the visit before treatment was recorded had a 70% probability of initiating their therapy earlier in the interval between study visits and a 30% probability of treatment initiation later in the interval. The converse was true for subjects with CD4 cell percentages of ⩾ 15%. Dates of treatment initiation were randomly assigned on the basis of these probabilities. All of these alternative approaches for assigning a date of initiation in PACTG 219 were run multiple times using different seeds for assigning random dates. Averaged estimates of the effect of HAART on mortality using these approaches are presented in table A1. Analyses using the midpoint between study visits for the date of treatment initiation provided more conservative estimates of the effect of HAART on mortality than did the other approaches considered. They were therefore presented as the primary analyses in our study.

Previous Section
 

References

  1. ↵
    1. Hammer SM,
    2. Squires KE,
    3. Hughes MD,
    4. et al
    . A controlled trial of two nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4 cells counts of 200 per cubic millimeter or less. AIDS Clinical Trials Group 320 Study Team. N Engl J Med 1997;337:725-33.
    CrossRefMedlineWeb of Science
  2. ↵
    1. Gavin PJ,
    2. Yogev R
    . The role of protease inhibitor therapy in children with HIV infection. Paediatr Drugs 2002;4:581-607.
    Medline
  3. ↵
    1. Palumbo PE,
    2. Kwok S,
    3. Waters S,
    4. et al
    . Viral measurement by polymerase chain reaction-based assays in human immunodeficiency virus-infected infants. J Pediatr 1995;126:592-5.
    CrossRefMedlineWeb of Science
    1. McIntosh K,
    2. Shevitz A,
    3. Zaknun D,
    4. et al
    . Age-and time-related changes in extracellular viral load in children vertically infected by human immunodeficiency virus. Pediatr Infect Dis J 1996;15:1087-91.
    CrossRefMedlineWeb of Science
  5. ↵
    1. Shearer WT,
    2. Quinn TC,
    3. LaRussa P,
    4. et al
    . Viral load and disease progression in infants infected with human immunodeficiency virus type 1: women and infants study group. N Engl J Med 1997;336:1337-42.
    CrossRefMedlineWeb of Science
  6. ↵
    1. Daar ES,
    2. Moudgil T,
    3. Meyer RD,
    4. Ho DD
    . Transient high levels of viremia in patients with primary human immunodeficiency virus type 1 infection. N Engl J Med 1991;324:961-4.
    MedlineWeb of Science
  7. ↵
    1. Nachman S,
    2. Stanley K,
    3. Yogev R,
    4. et al
    . Nucleoside analogs plus ritonavir in stable antiretroviral therapy-experienced HIV-infected children: a randomized controlled trial. Pediatric AIDS Clinical Trials Group 338 Study Team. JAMA 2000;283:492-8.
    Abstract/FREE Full Text
  8. ↵
    1. Wiznia A,
    2. Stanley K,
    3. Krogstad P,
    4. et al
    . Combination nucleoside analog reverse transcriptase inhibitor(s) plus nevirapine, nelfinavir, or ritonavir in stable antiretroviral therapy-experienced HIV-infected children: week 24 results of a randomized controlled trial. Pediatric AIDS Clinical Trials Group 377 Study Team. AIDS Res Hum Retroviruses 2000;16:1113-21.
    CrossRefMedlineWeb of Science
  9. ↵
    1. de Martino M,
    2. Tovo PA,
    3. Balducci M,
    4. et al
    . Reduction in mortality with availability of antiretroviral therapy for children with perinatal HIV-1 infection. JAMA 2000;284:190-7.
    Abstract/FREE Full Text
  10. ↵
    1. Gortmaker SL,
    2. Hughes M,
    3. Cervia J,
    4. et al
    . Effect of combination therapy including protease inhibitors on mortality among children and adolescents infected with HIV-1. N Engl J Med 2001;345:1522-8.
    CrossRefMedlineWeb of Science
    1. Sanchez JM,
    2. Ramos Amador JT,
    3. Fernandez de Miguel S,
    4. et al
    . Impact of highly active antiretroviral therapy on the morbidity and mortality in Spanish human immunodeficiency virus-infected children. Pediatr Infect Dis J 2003;22:863-7.
    MedlineWeb of Science
  12. ↵
    1. McConnell MS,
    2. Byers RH,
    3. Frederick T,
    4. et al
    . Trends in antiretroviral therapy use and survival rates for a large cohort of HIV-infected children and adolescents in the United States, 1989–2001. J Acquir Immune Defic Syndr 2005;38:488-94.
    CrossRefMedlineWeb of Science
  13. ↵
    1. Brogly S,
    2. Williams P,
    3. Seage GR III,
    4. Oleske J,
    5. Van Dyke R,
    6. McIntosh K
    . Antiretroviral treatment in pediatric HIV infection in the United States: from clinical trials to clinical practice. JAMA 2005;293:2213-20.
    Abstract/FREE Full Text
  14. ↵
    1. Robins JM,
    2. Herná n MA,
    3. Brumback B
    . Marginal structural models and causal inference in epidemiology. Epidemiology 2000;11:550-60.
    CrossRefMedlineWeb of Science
  15. ↵
    1. Herná n MA,
    2. Brumback B,
    3. Robins JM
    . Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology 2000;11:561-70.
    CrossRefMedlineWeb of Science
  16. ↵
    1. Cole SR,
    2. Herná n MA,
    3. Robins JM,
    4. et al
    . Effect of highly active antiretroviral therapy on time to acquired immunodeficiency syndrome or death using marginal structural models. Am J Epidemiol 2003;158:687-94.
    Abstract/FREE Full Text
  17. ↵
    1. Sterne JAC,
    2. Herná n MA,
    3. Ledergerber B,
    4. et al
    . Long-term effectiveness of potent antiretroviral therapy in preventing AIDS and death: a prospective cohort study. Lancet 2005;366:378-84.
    CrossRefMedlineWeb of Science
  18. ↵
    1. Mofenson LM,
    2. Harris DR,
    3. Moye J,
    4. et al
    . Alternatives to HIV-1 RNA concentration and CD4 count to predict mortality in HIV-1 infected children in resource-poor settings. Lancet 2003;362:1625-7.
    CrossRefMedlineWeb of Science
  19. ↵
    Working Group on Antiretroviral Therapy and Medical Management of HIV-Infected Children. Guidelines for the use of antiretroviral agents in pediatric HIV infection— October 26. 2006. Available at: http://aidsinfo.nih.gov/contentfiles/PediatricGuidelines.pdf. Accessed 1 August 2007.
  20. ↵
    1. Mofenson LM,
    2. Korelitz J,
    3. Meyer WA III,
    4. et al
    . The relationship between serum human immunodeficiency virus type 1 (HIV-1) RNA level, CD4 lymphocyte percent, and long-term mortality risk in HIV-1-infected children. J Infect Dis 1997;175:1029-38.
    Abstract/FREE Full Text
    1. Palumbo PE,
    2. Raskino C,
    3. Fiscus S,
    4. et al
    . Predictive value of quantitative plasma HIV RNA and CD4+ lymphocyte count in HIV-infected infants and children. JAMA 1998;279:756-61.
    Abstract/FREE Full Text
    1. Mofenson LM,
    2. Harris DR,
    3. Rich K,
    4. et al
    . Serum HIV-1 p24 antibody, HIV-1 RNA copy number and CD4 lymphocyte percentage are independently associated with risk of mortality in HIV-1-infected children. AIDS 1999;13:31-9.
    CrossRefMedlineWeb of Science
  23. ↵
    1. Lindsey JC,
    2. Hughes MD,
    3. McKinney RE,
    4. et al
    . Treatment-mediated changes in human immunodeficiency virus (HIV) type 1 RNA and CD4 cell counts as predictors of weight growth failure, cognitive decline, and survival in HIV-infected children. J Infect Dis 2000;182:1385-93.
    Abstract/FREE Full Text
  24. ↵
    Centers for Disease Control and Prevention. 1994 revised classification system for human immunodeficiency virus infection in children less than 13 years of age. MMWR Morb Mortal Wkly Rep 1994;43(RR-12):1-10.
  25. ↵
    Centers for Disease Control and Prevention. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Morb Mortal Wkly Rep 1992;41(RR-17):1-19.
    Medline
  26. ↵
    1. Aschengrau A,
    2. Seage GR
    . Essentials of epidemiology in public health. Sudbury, MA: Jones and Bartlett. 2003.
| Table of Contents

This Article

  1. Clin Infect Dis. (2008) 46 (4): 507-515. doi: 10.1086/526524
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 54 (5)
  1. Clinical Infectious Diseases
  1. Alert me to new issues

Most