Since the mid-1990s, there has been a significant reduction in perinatal transmission of HIV in the United States1 as well as increased survival into adolescence and adulthood of the population of US children who were infected with HIV at birth or during childhood. Most of these children and adolescents have received long-term antiretroviral (ARV) therapy.2 Since the introduction of ARV therapy in the early 1990s, drug therapy guidelines have changed from suboptimal single- to dual- to potent triple-drug therapy, including protease inhibitors (PIs) or nonnucleoside reverse transcriptase inhibitors (NNRTIs). The introduction of genotypic and phenotypic resistance testing has also allowed for modification of regimens to improve clinical outcomes.3 Reasons why clinicians change ARV regimens often depend on a variety of factors, including patient adherence, response to therapy, previous ARV use, and the presence of viral resistance.
Although the efficacy of combination ARV therapy in adults and children enrolled in clinical trials has been demonstrated, the prescribing patterns for ARVs and survival trends in observational cohort populations have been less well described. Furthermore, despite the association of highly active combination therapy and PIs with decreased mortality,4-8 studies of these regimens in adults have begun to address the question of whether the effects of sequential ARV regimens are less durable than the effects of the initial regimens,9 and the extent to which children and adolescents are receiving sequential combination ARV regimens has not yet been determined.
Herein, we describe patterns in ARV use and the association of these patterns with survival among HIV-infected children and adolescents enrolled in the Centers for Disease Control and Prevention's (CDC's) Pediatric Spectrum of HIV Disease Study (PSD). The PSD cohort includes children and adolescents who were all infected with HIV at a young age and follows them for much longer than do most clinical trials. The cohort also includes follow-up in both university tertiary care and community settings, which allows for greater generalization of results than does follow-up conducted exclusively at tertiary care settings.
Pediatric Spectrum of Disease
The PSD is a longitudinal epidemiologic chart review study of HIV-exposed and HIV-infected children who were younger than 13 years of age at the time of enrollment. The PSD has collected data since 1989 on children and adolescents in 6 US geographic areas with some of the highest HIV prevalence rates: Puerto Rico, Texas, Massachusetts, Los Angeles County, New York City, and Washington, DC.10,11 The study has been approved by the institutional review boards at all participating sites.
Study Population and Data Collection
Participants in this analysis were all HIV-infected children or adolescents in the PSD and were included in the analysis until death or last follow-up. They were considered to be active in the cohort in a given year if they had data available for either of 2 6-month abstraction periods each year. Complete drug prescription data were available for participants who were active in the cohort from 1994 to 2001, and drug prescription data, excluding start and stop dates, were available for those participants active in the cohort before 1994. Data for the survival analysis were included for participants active in the cohort between 1989 and 2001.
Data were collected from HIV clinic charts and hospital records and included information on birth history, risk for HIV infection, date of initial HIV evaluation, laboratory confirmation of HIV diagnosis, presence of HIV-related signs and symptoms, history of AIDS-defining illnesses, hospitalizations, immunizations, age at death, type of medical reimbursement, and clinical trial participation. The CDC pediatric HIV classification system was used to define HIV infection status as well as clinical and immunologic categories.12 For the purpose of this report, we presume that ARV drugs prescribed were taken; ARV drugs were included in the analysis unless there was documentation that they were not taken. For drug prescriptions that had no start dates indicated, the start date was considered to be the date of the next chart abstraction during which a continuation of the drug was documented.
We began ARV analyses in the PSD cohort as of 1994 because of the increased availability and use of ARVs beginning in that year, and trends in ARV prescription and sequential triple-therapy regimens were reported beginning in 1994. Trends in annual therapy use (monotherapy, dual-drug therapy, and triple-drug therapy) were assessed for children and adolescents who continued receiving these therapies for 6 months or more in a calendar year. The 6-month period was chosen to classify patients based on the predominant type of therapy in a given year. Patients with unclassified ARV use were those who changed between these 3 categories of drug regimens more frequently than every 6 months and thosewho died or were lost to follow-up during the 6-month period.
Trends in Sequential Triple Therapy
A change in a triple-therapy regimen was defined as a change in 2 or more drugs while continuing on at least a 3-drug regimen. To define a regimen change in accordance with the guidelines for use of antiretroviral agents in both children and adolescents, 2 drugs were chosen as necessary.3,13 Reasons for stopping drugs were not included in this definition because of limitations in chart documentation of this information and the high number of unknown reasons for drug changes.
Trends in Survival Rates
Survival rates were compared for children and adolescents in the following birth cohort groups: 1989 to 1993, 1994 to 1996, and 1997 to 2001. These years were chosen because they reflect sentinel time points in the announcement of pediatric clinical trial results or the availability of new ARV drugs licensed for pediatric use. In 1989, most children were not receiving ARV drugs; in 1994, most were receiving 1 or 2 ARV drugs; and by 1997, many children were receiving triple therapy with or without PIs. Survival analyses using Cox proportional hazard models were done to compare the survival rates of children based on birth cohort and use of triple-ARV therapy. Log rank tests were used to assess their differences.
For the survival analysis of those receiving triple-ARV therapy, all eligible children were included and categorized according to whether they had ever received triple-ARV therapy, and the analysis was adjusted for birth cohort. To address the potential bias of patients starting triple-ARV therapy because they had survived long enough for it to become available, an additional analysis was done for those patients who had ever received triple therapy, using time-varying covariates to account for time receiving therapy. This model was also adjusted for sex, race, site, and birth cohort. The adjusted hazard of death in this analysis was relative to that before starting triple therapy. The outcome measure for all survival analyses was age at time of death, and the null hypothesis was that survival times were equal for all 3 birth cohorts and for those ever or never receiving triple-ARV therapy.
Odds ratios (ORs), confidence intervals (CIs), and χ2 tests for trend were calculated using Epi-Info, version 6.0 (Centers for Disease Control and Prevention, Atlanta, GA). Kaplan Meier survival curves were calculated using SPSS (SPSS, Chicago, IL). Other data analyses were done with SAS, version 8.0 (SAS Institute, Cary, NC).
The number of children and adolescents included in the ARV analyses varied between 463 in 1989, 2196 in 1994, and 2040 in 2001. A minimum of 4% of participants in 2001 to a maximum of 10% of participants in 1989 were lost to follow-up. The birth cohort survival analysis had 1790 participants in the 1989 to 1993 birth cohorts, 643 in the 1994 to 1996 birth cohorts, and 331 in the 1997 to 2001 birth cohorts. The triple-therapy survival analysis had 1576 study participants who had received triple-ARV therapy and 1183 who had not.
Because of the decline in HIV-infected newborns in the United States,1 the PSD cohort represented, in large part, the same population of children and adolescents each year after 1994. The primary change in demographics from 1994 to 2001 was the median age of cohort participants, which increased from 4 years in 1994 to 9 years in 2001. Other changes in the cohort included an increasing proportion of African Ameri-cans and a greater proportion of children covered by Medicaid (Table 1). The mode of HIV infection remained primarily perinatal.
Assessment by birth cohort found earlier cohorts had a higher median age at initial HIV evaluation, with the median being 4.4 (interquartile range [IQR] = 19.3) months for the 1989 to 1993 birth cohort and 0.5 (IQR = 7.9) and 0.2 (IQR = 2.9) months for the 1994 to 1996 and 1997 to 2001 birth cohorts, respectively. There were also significant differences by race (P = 0.003) and site (P < 0.001). The overall groupings for race remained the same (6%-9% white, 53%-61% African American, and 33%-38% Hispanic), whereas the proportion in Massachusetts and New York decreased as the proportion in Texas increased. There were no differences by sex, and the proportion in later birth cohorts with non-perinatal transmission was too small for analysis.
Antiretroviral Therapy Use
In 1996, only 6% of children and adolescents in the PSD were taking 3 or more drugs, but in 2000 and 2001, 68% and 66% of patients, respectively, were prescribed triple therapy for at least 6 months per year (P < 0.0001, χ2 for trend; Fig. 1). In 2000 and 2001, 14% and 12% of patients, respectively, continued to receive dual therapy for more than 6 months of the year.
The characteristics of patients receiving different types of therapy for at least 6 months in 2001 were assessed. Differences by race/ethnicity included a significantly higher proportion of white and Hispanic patients (73% and 70%, respectively) receiving triple therapy compared with African-American patients (63%) (OR = 1.62 and OR = 1.36, 95% CI: 1.11-2.35 and 95 CI: 1.11-1.66, respectively). The racial differences, although trending in the same direction, were not significant in earlier years. For all patients in 2001, regardless of whether they were taking monotherapy, dual therapy, or triple therapy, the median viral load was less than 5000 copies/mL and all had a median CD4 count greater than 500 cells/mm3.
From 1994 through 2001, ARV therapy use also increased overall, with the greatest increase in the use of PIs. By 2001, 94% of patients were taking nucleoside reverse transcriptase inhibitors (NRTIs), 70% were taking PIs, and 35% were taking NNRTIs.
Trends in Sequential Triple-Therapy Regimens
Assessment of trends in sequential ARV regimens indicated that since 1997, increasing numbers of patients in the PSD cohort had received 3 or more triple-therapy regimens (P < 0.0001, χ2 for trend; Fig. 2). In 1997, 81% of patients receiving triple therapy were receiving their first regimen, 15% were receiving their second regimen, and 4% were receiving their third or greater regimen. In 2001, 56% of patients receiving triple therapy were receiving their first regimen; 27% were receiving their second regimen, and 17% were receiving their third or greater regimen. Furthermore, in 2001, 41 patients (3%) were receiving their fifth or greater triple-therapy regimen.
The durability of sequential triple-therapy regimens also decreased. The median duration of the first regimen was 13 months (IQR = 4-26 months), that of the second was 9 months (IQR = 3-21 months), and that of the third or greater was 7 months or less. Patients with a short duration of their first regimen (<1 year) were not significantly more likely to have a short duration of their second regimen.
In 2001, there were no significant differences in gender, mode of infection, clinical trial participation, or type of reimbursement between those who were receiving their third or greater triple-therapy regimen and those receiving their first or second regimen. Adolescents aged 13 years or older were significantly more likely to be receiving a third or greater triple-therapy regimen than were children aged 1 to 12 years (OR = 2.44, 95% CI: 1.84-3.25), however, and white patients were significantly more likely to be receiving their third or greater triple-therapy regimen than were African-American or Hispanic patients (OR = 1.64, 95% CI: 1.08-2.5).
The pattern of ARV drug use also differed for those patients receiving their third or greater regimen compared with those receiving any ARV regimen. Patients receiving their third or greater regimen in 2001 were at least twice as likely to use abacavir and lopinavir/ritonavir than were patients in the cohort as a whole (OR = 2.43 and OR = 3.07, 95% CI: 1.86-3.18 and 95% CI: 2.38-3.96, respectively). Patients receiving their third or greater triple-therapy regimen were also significantly more likely to receive amprenavir and efavirenz (OR = 1.64 and OR = 1.70, 95% CI: 1.19-2.26 and 95% CI: 1.29-2.24, respectively).
Coincident with the increased prescription of triple therapy and PIs has been a significant improvement in survival rates in the 1997 to 2001 birth cohorts compared with earlier birth cohorts (P < 0.0001). Survival rates also improved between 1989 to 1993 and 1994 to 1996, but this difference was not statistically significant (Fig. 3).
Survival rates improved among those who received triple-ARV therapy for any length of time in contrast to those who never received triple therapy (P < 0.0001); this difference continued to be significant when birth cohort was included in the analysis (P = 0.006; see Fig. 3). Comparison of survival rates for those patients who had ever received triple-ARV therapy, adjusted for time taking the therapy, indicated that the hazard of death decreased with time receiving triple-ARV therapy. When adjusted for birth cohort, race, sex, and site, the 1994 to 1996 birth cohorts had significantly greater survival rates than the 1989 to 1993 birth cohorts (P < 0.05).
Triple-ARV therapy use and survival rates have increased among HIV-infected children and adolescents in the PSD cohort. The proportion of patients using newer ARV drugs and the proportion receiving their third or greater triple-ARV regimen have also increased between 1994 and 2001, however.
In accordance with guidelines to prescribe 3 ARV drugs initially, preferably including a PI or an NNRTI, most children in this observational cohort have been receiving triple therapy since 1999 and PIs since 1998. The finding that 12% of patients continued to receive dual-ARV therapy in 2001 can likely be attributed to the fact that these patients continue to have low viral loads and adequate immune function.
Although the proportion of children receiving triple therapy and PIs has remained relatively stable since 2000 and 1998, respectively, demand for new ARV drugs has continued. In particular, since 1998, there have been significant changes in the use of certain PIs and NNRTIs, such as lopinavir/ritonavir and efavirenz. Drugs such as T20 and tenofovir have also either recently been approved by the US Food and Drug Administration (T20) or are being studied in children (tenofovir) after being approved for use in adults.14 These and other new ARV drugs need to be tested further in pediatric and adolescent populations to determine the optimal ARV drug regimens for those who have already received a number of different ARV drugs.15-18
Coincident with the increased use of new ARV drugs are increasing numbers of patients who have received 3 or more triple-therapy regimens. Furthermore, the finding of decreasing duration of sequential triple-ARV regimens corroborates findings in the adult population.9 These trends pose new challenges for the treatment and care of HIV-infected children and adolescents and highlight the importance of determining the optimal treatment modifications for patients who do not respond to initial and sequential combination ARV therapy regimens.15
Nevertheless, survival rates of children and adolescents in the birth cohorts from 1997 to 2001 have improved since the introduction of triple-ARV therapy, and an improved survival rate is significantly associated with triple-ARV therapy use, even after adjustment for birth cohort. Furthermore, a Kaplan-Meier estimate of birth cohort survival found improved survival in both the 1989 to 1993 and 1994 to 1996 birth cohorts around 1998, when PIs and triple-ARV therapy became more common. Our findings support those of previous studies in the United States and Europe. An improved survival rate associated with triple-ARV therapy use was demonstrated in an observational cohort study of children perinatally infected with HIV-1, which was followed up through 1999 by de Martino and colleagues in Italy.5 The same finding was noted in a clinical trial population of pediatric patients followed up from 1996 through 1999 at tertiary care centers by the Pediatric AIDS Clinical Trials Group (PACTG).4 Given that our study had even longer follow-up times, we show that gains in survival rates have been sustained so far. Furthermore, the results of our analysis demonstrate overall longer survival in all cohort groups than did the study by de Martino et al.5,19 The trend toward improved survival rates and improved immune status of HIV-infected children and adolescents20 is likely influenced by increasing numbers of patients who were prescribed triple therapy as well as by improved HIV diagnostic technologies allowing for earlier diagnosis and the increasing prescription of prophylaxis for opportunistic infections. Early gains in survival in the 1994 to 1996 birth cohorts may be a result of these factors, which preceded the widespread introduction of triple-ARV therapy. Despite more patients receiving sequential triple-ARV regimens, the continued availability of new drugs may also be contributing to the sustainability of higher survival rates.
There are certain limitations to the PSD data and this analysis. In a prospective chart review, reasons for changes in ARV therapy (eg, potential ARV resistance, nonadherence, toxicity) are not always available; therefore, reasons for drug changes were not included in the analysis of changes in triple therapy. Clinician decisions about particular ARV regimens are based on laboratory values, history of adherence to ARV drugs, past ARV drug history, drug side effects, presence or absence of viral resistance, and clinical judgment. Although these factors do affect the trends in prescription of ARV drugs and survival rates reported here, the extent of each one's contribution could not be assessed. An additional limitation of the analysis is that it is often difficult to assess whether patients actually adhere to ARV therapy; although adherence can be inferred from medical chart data, charts more precisely convey prescribing patterns than actual ARV use. The findings presented here may also overestimate children receiving highly active antiretroviral therapy with either a PI or an NNRTI, but restriction of the definition of triple-ARV therapy would only strengthen the survival analysis findings. Lastly, the time period that patients were receiving a particular regimen may be underestimated because of missing start dates, which were noted in 10% of the 6-month chart abstractions.
In summary, these findings from the PSD cohort demonstrate that substantially more patients are receiving triple therapy and PIs than in earlier years and that survival rates have significantly improved for HIV-infected children since 1994. As the cohort of HIV-infected children and adolescents in the United States ages and continues to change regimens, however, demand for new ARV drugs continues. Further research is needed in the area of effective modifications of therapy and the role that ARV resistance and resistance monitoring plays in choosing optimal ARV regimens for those patients who have been taking them for a long time. As children and adolescents on ARV therapy go through hormonal changes and growth spurts associated with puberty, the late complications and toxicities of chronic exposure to highly active antiretroviral therapy also need to be carefully monitored and evaluated.
The authors thank the following individuals: CDC: Kenneth Dominguez, Mary Glenn Fowler, Alan Greenberg, Beverly Bohannon, and Thom Sukalac; Massachusetts: Ho-Wen Hsu, Joyce Cohen, and Catherine Reddington (University of Massachusetts Medical School, State Laboratory Institute); Barbara Stechenberg, Eileen Theroux, and Maripat Toye (Baystate Medical Center-Springfield MA); Stephen Pelton, Anne Marie Regan, and Sam Theodore (Boston Medical Center); Kenneth McIntosh and Catherine Kneut (Boston Children's Hospital); Katherine Luzuriaga, Dorothy Smith, and Donna Picard (University of Massachusetts Medical School); Gerard Coste and Margaret Lynch (Cambridge Hospital); New York City: Vicki Peters, Kai-Li Liu, Polly Thomas, Chere Mapson, Annette Brooks, Myrna Beckles, Patricia Diggs, Stephanie Manning, Carol McFarlane, and Karla McFarlane (New York City Department of Health and Mental Hygiene); Arye Rubinstein (Albert Einstein Hospital); Saroj Bakshi (Bronx Lebanon Hospital); Edward Handelsman (Downstate University Hospital); Elaine Abrams (Harlem Hospital); Cathy Painter (Incarnation Children's Center); Andrew Wiznia (Jacobi Hospital); Ninad Desai (Kings County Medical Center); Nathan Litman (Montefiore Hospital); Joseph Stavola (New York Hospital); Jacob Abadi (North Central Bronx Hospital); Washington, DC: Tamara Rakusan, Hans Spiegel, Andrew Bonwit, Vonterris Hagan-Temple, and Waldo Perez (Children's National Medical Center); Sohail Rana, Renee Jenkins, Davene McCarthy-White, and Linda Hart (Howard University Hospital); Puerto Rico: Idith Ortiz, Juan Carlos Orengo, Aida Melendez, Myribel Santiago-Torres, Evelyn Rivera, Ruth Santos, and Emily Maldonado (Puerto Rico Departmento de Salud); Eleanor Jimenez and Luis A. Martinez (San Juan City Hospital); Irma Febo and Lissette Lugo (University Pediatric Hospital); Wanda Figueroa and Odette Garcia (Bayamon Regional Hospital); Jose Vazquez Julia (Caguas Regional Hospital); Rosa Delgado, Ortando Quincoce, and Judith Gautier (Ponce Regional Hospital); Texas: Sharon Melville, Richard Yeager, and Mary James (Texas Department of Health); Octavio Ramilo and Janeen Graper (Children's Hospital-Dallas); Terence Doran, Rachel Davis, and Mary Jane Varela (University of Texas-San Antonio); Gilberto Handel and Tony Millon (Texas Tech Medical Center-El Paso); Gloria Heresi and Kathleen Paul (University of Texas-Houston); Mary Paul and Amy Leonard (Baylor College of Medicine-Houston); Janak Patel and Andrea Smith (University of Texas Medical Branch-Galveston); Saramistha Hauger, MariFran Shannon, and Nelda Garcia (Children's Hospital of Austin); Los Angeles County: Toni Frederick, Laurene Mascola, Yon Silvia Walker, Janielle Jackson-Alvarez, and Priya Mukhopadhyay (Los Angeles County Department of Health); Yvonne Bryson (UCLA School of Medicine); Joseph Church (Children's Hospital of Los Angeles); Audra Deveikis (Memorial Miller Children's Hospital); Margaret Keller (Harbor-UCLA Medical Center); Deborah Lehman (Cedars-Sinai Medical Center); Andrea Kovacs (LAC and USC Maternal Child Clinic); Steve Taylor (Martin Luther King, Jr/Drew Medical Center); Victor K. Wong (Southern California Kaiser Permanente).