Temporal Trends in Early Clinical Manifestations of Perinatal HIV Infection in a Population-Based Cohort FREE

Worldwide, approximately 2.5 million children are infected with the human immunodeficiency virus (HIV),¹ and approximately 1700 new perinatal infections occur daily.² In the United States, more than 9300 HIV-infected children younger than 13 years have progressed to AIDS as of December 2003.² While the Pediatric AIDS Clinical Trials Group (PACTG) 076 regimen has reduced perinatal transmission by 67%,^3- 4 prenatal combination antiretroviral therapy (ART) has further reduced transmission to 2% or less.^4- 5 Moreover, earlier diagnosis and treatment has improved the prognosis of perinatal HIV infection.^6- 13 However, the impact of treatment on progression of perinatal HIV infection remains poorly characterized at the population level.

Perinatal HIV infection may progress in 2 patterns: early, with a median onset of age 4 months, or late, with a median onset of age 6 years.^14- 17 It is estimated that 20% to 30% of perinatal infections in the pre-ART era developed early progression,^14- 17 which was associated with a poor prognosis, and more often presented with Pneumocystis jiroveci pneumonia (PCP) or encephalopathy as the initial category C diagnoses.^{15- 16,18- 20} Early initiation of ART and PCP prophylaxis prior to progression to a category C diagnosis is advocated because clinical and laboratory values are unreliable predictors of early vs late progression.^13,21 Moreover, for some patients, early highly active antiretroviral therapy (HAART) results in prolonged control of viremia, seroreversion, and preserved immune function.^22- 26 However, these benefits must be balanced against potential drawbacks of early therapy, such as concerns regarding adherence with multiple unpalatable drugs that require frequent administration, risk of viral drug resistance, poorly characterized drug dosing guidelines in the first few months of life, and drug toxicity, especially when considering initiating ART in asymptomatic infants.^27- 28 There is also the issue of potential variation in immune response by age.²⁸

For infants, there are few data to guide when to initiate ART.²⁷ While many experts recommend initiation of HAART in any clinically or immunologically symptomatic infant, there is controversy regarding its use in clinically or immunologically asymptomatic infants.²⁷ Results from pediatric clinical trials^{6- 7,29- 38} and observational studies^8- 13,18 evaluating treatment are limited because of small sample size and limited time of follow-up. In addition, there are few population-based¹⁸ or longitudinal studies^8,10- 11,28 examining temporal trends in clinical outcomes, including disease progression and mortality, rather than surrogate laboratory markers. Population-based studies are particularly relevant for evaluating the full spectrum of clinical manifestations of a complex illness such as perinatal HIV infection.

Interventions for the prevention, detection, and treatment of perinatal HIV infection include the introduction of ART,³⁹ prenatal testing,⁴⁰ perinatal ART to reduce vertical transmission,^3,41 protease inhibitors (PIs),⁴² and HAART.⁴³ As a result, perinatal infection in the United States has decreased in incidence, and cases are being detected earlier.⁴⁴ Previously, we examined a population-based cohort of 105 of the HIV-infected children assessed in the analysis herein in Northern California born between January 1, 1988, through July 31, 1992, and demonstrated that PCP prophylaxis was associated with prolonged time to category C diagnosis, decreased rate of PCP, and increased prevalence of encephalopathy as the initial category C diagnosis.¹⁸

Herein, we extend analysis of that cohort to those born between January 1, 1988, and December 31, 2001. An earlier report regarding capture of data for infants by surveillance also involved some members of the cohort.⁴⁵ Our goal was to examine early progression of HIV, survival, and distribution of category C diagnoses in a perinatal population-based sample during different eras in prevention and management and in relation to early institution of any ART therapy.

Longitudinal surveillance data were extracted from the Northern California Pediatric Spectrum of Disease (PSD) project, a multicenter pediatric HIV surveillance project based at Stanford University since 1988 (which is when the process of reviewing medical records at each hospital site began).¹⁸ The surveillance area includes 12 counties in Northern California, with a total population of approximately 6 million.

The PSD surveillance system was used to identify all infants born to HIV-infected women. In the analyses herein, only data for HIV-infected children were used, but the data collection method was identical to that for children found to be HIV-exposed but uninfected. There were 289 HIV-infected children in the PSD database born through December 31, 2001 (n = 66 for the period through December 31, 1987; n = 111 for January 1, 1988, through July 31, 1992; and n = 112 for August 1, 1992, through December 31, 2001). January 1, 1988, is the start date to avoid selection bias because children in the cohort born before 1988 were rarely evaluated for HIV infection before age 3 years. A total of 223 children were potentially eligible for analysis (ie, HIV-infected and born between January 1, 1988, and December 31, 2001). Children were included in the analysis herein if they were identified at a surveillance hospital before their third birthday and had records through their third birthday or if they died of HIV-related causes before their third birthday and had records available before their death.

Study nurses went to each site having a pediatric HIV clinic (based at surveillance hospitals) where the children were followed up and reviewed the records for all children having mothers with HIV infection, whether HIV-infected, uninfected, or indeterminate, and used the records to determine if each child met the US Centers for Disease Control and Prevention (CDC) case definition for HIV infection (see below) based on physician notes and laboratory tests. The nurses went to each site on a regular basis to identify new clinic patients (an initial form for the baseline visit was filled out around the time of birth for those determined to be HIV-exposed at birth) and to review each child’s records at 6-month intervals and record cumulative data about each child for each 6-month interval on a standardized form. The information was entered into a standardized database. To maintain confidentiality and prevent duplication of records, children were assigned a unique but anonymous identifier (consisting of an alphanumeric code combined with each child’s birth date), which was entered into a Stanford database. A child could have had up to 7 chart reviews (1 at birth and 1 twice a year up to age 3 years) and a minimum of 1 chart review. All children had a chart review every 6 months from the time they were identified at the surveillance hospital, so that those who did not have 6 or 7 reviews and were alive at age 3 years were missing reviews from the time of birth until entry into clinical care at the surveillance hospital. The number of chart reviews per child was calculated assuming that the final review occurred after the child died or reflected information available at or after his or her third birthday. Of the 205 children included in the analysis, 50% of all children and 60% of children alive at age 3 years (n = 164) had 6 to 7 chart reviews (median, 5 [interquartile range {IQR}, 3-6]; mean, 4). Of children alive at age 3 years, median and mean number of chart reviews were 6 and 5, respectively.

In the PSD surveillance system, a case of perinatal HIV exposure is defined as a child younger than 13 years who had been born to a mother with HIV infection but having no history of blood or blood product transfusion prior to 1985. Data from the child’s medical record were used to determine cases of vertical transmission, using the CDC classification system for HIV infection in children younger than 13 years.^27,46 Food and Drug Administration–approved HIV tests were used at all institutions. Information on CD4 cell counts and viral load levels were collected if available from the medical record, but because they were not collected at identical time points for each child they were not used in this analysis. Surveillance included the examination of records from hospital-based clinics and from the California Children’s Services program, a state agency providing funding and case management for children with HIV infection. Hospital surveillance included all major pediatric referral hospitals in the Northern California area: Stanford University Hospital/Lucile Packard Salter Children’s Hospital at Stanford, University of California at San Francisco, San Francisco General Hospital, Children’s Hospital of Oakland, Santa Clara Valley Medical Center, University of California at Davis, and University Medical Center at Fresno. Records from all HIV-exposed patients cared for at these institutions in pediatric immunology and infectious diseases clinics were assessed. Records were examined from 2 California Children’s Services pediatric HIV programs: the HIV-Positive Children program, involved in funding, testing, and follow-up of children at risk for HIV infection or who have HIV infection and are asymptomatic; and the HIV Treatment program for children with HIV infection needing treatment. To ensure that the system captured data for as many HIV-infected children as possible, the California Children’s Services programs were reviewed on a regular basis (about every 2 years) to identify children who had been managed by the Stanford PSD system for perinatal HIV infection but not followed up by the surveillance hospital sites routinely visited for the 6-month chart reviews. One child was identified through these reviews; this child was born November 20, 1990, diagnosed with PCP at age 3 months, and died 2 days later, at which time identification of HIV seropositivity occurred (there was no treatment of relevance involved).

The PSD surveillance system included collection of data on sex and race/ethnicity, ascertained by physician documentation in the patients’ medical records. Data from the surveillance program were matched quarterly with data from the state AIDS and death registries. Approval was obtained on an annual basis from the institutional review boards of the institutions where children were enrolled and at Stanford University; the requirement for informed consent for this study was waived by all institutions because of the absence of personal identifiers in the database. Similarly, the requirement was waived for written informed consent for publication of data from the study in a medical journal.

In a separate study, we monitored the proportion of HIV-exposed infants identified by PSD surveillance⁴⁷ using a method similar to that previously reported.⁴⁵ The annual number of identified infants born to HIV-infected women was compared with the expected number of infants based on an anonymous HIV serosurvey, the Survey of Childbearing Women (SCBW).⁴⁸ This serosurvey was performed in 1988-1995 and 1998 by the state of California to follow the prevalence of HIV among women giving birth. Maternal seroprevalence data from the SCBW were matched to the 12 California counties from which the PSD database was derived to identify the number of observed vs expected HIV-exposed infants in the surveillance area.

The purpose of the matching between the PSD system and the SCBW serosurvey for HIV-exposed infants was to assess the extent to which the PSD system may have captured data for infants with HIV infection. Because the HIV infection status of infants may not be known until age 6 months or later, they are followed up in some cases for up to 1 year or longer until they meet the CDC definition for seroreversion. Most HIV-exposed infants will be uninfected.³ It is difficult to accurately estimate the number of HIV-infected infants born to these women because transmission rate data for each SCBW year are needed. The PSD database does not contain sufficient information to determine the proportion of women having perinatal prophylaxis. For each year in which SCBW data were available, the transmission rate could range from about 25% (with no perinatal prophylaxis³) to about 1% (rate with perinatal prophylaxis^4- 5). Thus, if it is assumed for the 1053 HIV-infected women identified in the SCBW serosurvey giving birth in 1988-1995 and 1998 (in the 12-county surveillance area) that the transmission rate was as high as 25% overall, 263 infected infants would be expected to be born in those years. Since 548 total HIV-exposed children born in those years were matched between the PSD system and the SCBW data (see “Results” section), this suggests that (since most exposed children are not infected) many of the HIV-infected infants were identified by the PSD system.

The study population included HIV-infected children, as defined by the CDC classification system,^27,46 born between January 1, 1988, and December 31, 2001, who had medical records available until at least age 3 years or who died prior to age 3 years from HIV-related complications. Treatment was defined to include therapies given continuously for at least 1 month and only in children who remained free of category C diagnoses⁴⁶ for 1 month after initiating therapy. For example, if a child progressed to a category C diagnosis within 1 month after starting ART, then he or she was considered to be in the “no treatment before category C diagnosis” category. This definition of “treatment” allows sufficient time to evaluate therapeutic benefits of ART in the prevention of category C diagnoses and is the same as that used in comparable studies.⁸ Evaluated therapies included ART and PCP prophylaxis. Specific ART received included zidovudine, didanosine, lamivudine, stavudine, zalcitabine, efavirenz, nevirapine, ritonavir, nelfinavir, amprenavir, and indinavir. Specific PCP prophylaxis included trimethoprim-sulfamethoxazole, pentamidine, dapsone, or atovaquone. Triple ART was defined as 3 antiretroviral agents of any class, with nucleoside reverse transcriptase inhibitors plus a PI or a nonnucleoside reverse transcriptase inhibitor. Of all the children who received triple ART, all received either a PI or a nonnucleoside reverse transcriptase inhibitor, so separate analyses for all triple nucleoside reverse transcriptase inhibitor regimens were not needed.

Of the 205 children in the study, 33 were treated with the PACTG 076 protocol.³ For consideration of PACTG prophylaxis, we assessed only children born in 1994 or later.³ There were a few infants for whom PACTG prophylaxis failed; it is not 100% effective in preventing perinatal transmission.³ Since the PACTG 076 protocol was not treatment per se, use of PACTG 076 was considered differently from drugs used for known HIV infection. No infants continued to receive PACTG 076 beyond 6 weeks of life, but they may have received additional ART after completing PACTG 076 if HIV infection was documented. Children not receiving any ART after PACTG 076 were considered “no treatment.” Children receiving ART immediately after or within 1 to 2 weeks after PACTG 076 were considered as having started treatment at birth and considered to have started mono ART at birth because of receiving zidovudine in the first 6 weeks, then categorized based on their subsequent therapy. Children completing PACTG 076 and starting ART for HIV infection at least 2 weeks after stopping PACTG 076 were considered to have started treatment when subsequent non-PACTG 076 treatment was initiated. The PACTG 076 protocol was only integrated into the analysis of therapy if virtually continuous with subsequent ART regimens (within 2 weeks after cessation of PACTG 076). Of the 33 children receiving the 076 regimen, 6 had received no ART beyond 6 weeks (2 received no treatment and 4 received PCP prophylaxis only [considered “no ART treatment”]; the 4 who received prophylaxis only were considered “PCP prophylaxis”). Six continued ART uninterruptedly within age 6 weeks (some infants were diagnosed with HIV infection before age 6 weeks and thus started on treatment ART before the 6-week PACTG regimen was completed); these children continued therapy until their end point (age 3 years, or death by age 3 years). These children were considered to have started mono ART at birth (subsequent therapy: 1 receiving triple ART at age 2 weeks, 1 triple ART at age 2 months, 2 dual ART at age 3 months, and 2 dual ART at ages 15 and 17 months). The 4 who received dual ART did not progress to triple ART. Twenty-one stopped ART for at least 14 days but restarted it later, considered as starting treatment at the time of their first non-PACTG 076 treatment (mono ART: n = 7; median age, 8 [IQR, 3-10] months; dual ART: n = 8; median age, 7 [IQR, 3-10] months; and triple ART: n = 6; median age, 2 [IQR, 2-3] months). All of the children except 1 had uninterrupted treatment until the end point. One child who completed PACTG 076 and then started mono ART at 8 months had multiple treatment interruptions until diagnosed with category C disease at age 20 months.

Adherence was assessed only from physician comments in patients’ medical records. Fifteen children of the 205 included in the analysis had evidence of ART noncompliance in their medical records. Three stopped all ART and did not restart (1 born February 1992 took ART from May 1993 to January 1994 with no evidence of category C disease at 3 years; 1 born July 1992 took ART from June 1994 to August 1994 with no evidence of category C disease at 3 years; a third born May 1991 took ART from August 1991 to October 1991, developed category C disease in December 1991, and died in April 1992 of encephalopathy). Eight children stopped ART for short periods followed by reinstitution of therapy. Of these, the mean and median time stopping therapy was 7.5 weeks and 3.5 weeks, respectively, with a range of 1 to 36 weeks. Five did not develop category C disease by age 3 years (mean, 10.8 weeks and median, 4 weeks, respectively, for those without category C [n = 5] and mean, 2 weeks and median, 2 weeks, respectively, for those with category C disease by age 3 years [n = 3]). One child started and stopped ART over the 3 years because of noncompliance and did not develop category C disease by age 3 years; total amount of time stopping therapy could not be calculated. Among children who received PCP prophylaxis, 3 children temporarily stopped prophylaxis once in the first year of life: 1 stopped therapy 20 days during the third month of life (no category C by age 3 years), 1 stopped therapy 2 weeks during the seventh month (category C at age 20 months), and a third stopped therapy 1 week during the seventh month (no category C). One of the 8 children who stopped ART for short periods also stopped PCP prophylaxis prematurely at age 9 months and did not develop category C disease by age 3 years.

Patients were classified into 3 cohorts by year of birth, with cutoff dates chosen to reflect major changes in management guidelines as follows: early cohort: January 1, 1988, through December 31, 1991; intermediate cohort: January 1, 1992, through December 31, 1995; and late cohort: January 1, 1996, through December 31, 2001. Children born between January 1, 1988, through December 31, 1991, were most commonly evaluated within the first year of life and children born after December 31, 1991, within the first month. December 31, 1991, was chosen to reflect the Food and Drug Administration approval of zidovudine for children in 1990³⁹ and the publication of guidelines for PCP prophylaxis in 1992.⁴⁹ By 1996, PACTG 076 had changed perinatal management,^3,41 PIs were introduced,⁴² and prenatal testing was widely recommended.⁴⁰

Evaluated variables included ages at initial HIV diagnosis, treatment, first category C diagnosis, and death; proportion of children treated with each type of ART (mono, dual, or triple) and/or PCP prophylaxis; the proportion of children who developed a category C diagnosis; and the proportion of children who died relative to age and type of therapy. Two children who died from HIV-related complications were considered as progressing to category C at the age of death for analysis of progression. Because all of the infants who ever received triple ART either initially or subsequently never progressed to category C, triple ART at any time was statistically highly protective; thus, triple ART was excluded from comparison of mono or dual therapy vs no therapy in analyses of outcome. Analyses of both initial and most advanced ART in describing temporal trends were also performed. Children with insufficient follow-up were excluded from analysis, as detailed in the “Results” section. There were no missing variables for children included in the analysis. The size of the study population has a power of 0.8 to detect a 20% difference in progression to category C between the treated and untreated groups and a 30% difference between triple ART and mono/dual ART, both with α = .05, 2-sided, and starting with the null hypothesis that 35% of untreated children will progress to category C in the first 3 years of life. Unless otherwise specified, discrete variables were compared using the 2-sided Fisher exact test. In specified places the Mantel-Haenzel test was used to test for trend in proportions. Continuous variables were always compared using the Kruskal-Wallis test. The analysis was performed using SAS release 8.2 (SAS Institute Inc, Cary, NC). P<.05 was considered statistically significant.

The Northern California PSD surveillance system identified 223 children with perinatal HIV infection born between January 1, 1988, through December 31, 2001. Of these children, 2 (1%) were excluded because they died prior to age 3 years from causes unrelated to HIV (Werdnig-Hoffman disease, child abuse/head injury). In addition, 16 (7%) were excluded because they were lost to follow-up before age 3 years (median age, 15 months; range, 2-28 months). Of those lost to follow-up, 3 (19%) had progressed to a category C diagnosis (at ages 2, 4, and 7 months) before their last recorded clinic visit. Of all 18 excluded children, 14 (78%) had received PCP prophylaxis, ART, or both. After exclusions, the study population comprised 205 children born between January 1, 1988, and December 31, 2001, who were followed up through age 3 years or who died of HIV-related causes before age 3 years (Table 1). Overall, 43 children died in the first 3 years of life (41 were included in the analysis). Two died of HIV-related causes but did not have category C disease at the time of death; the other 39 children died of category C disease.

The annual number of children identified through PSD with perinatal HIV infection was 16 to 25 between 1988-1996 but 6 or fewer after 1996. To determine whether this represented a true decline in perinatal transmission or decrease in ability to capture data on infants through surveillance, we compared the rates of identification of HIV-exposed infants in the 12-county surveillance area with the SCBW database. Based on a match of SCBW and PSD data, HIV-exposed infants born to 52.1% (548/1053) of HIV-infected women were identified by PSD surveillance in the 12 counties. The ratio of observed/expected infants of infected women increased from 29.6% in 1988 to 67.5% in 1998 (Table 2).⁴⁷ The 52.1% value is not a direct calculation of data capture for HIV-infected children, but assessment of data capture for HIV-exposed infants can provide a means of assessing whether data for most HIV-infected infants were captured by the surveillance system. For example, if it is assumed that the transmission rate was as high as 25% overall among the 1053 HIV-infected women giving birth in 1988-1995 and 1998 in the 12-county surveillance area, then 263 infected infants would be expected to be born during those years. Since 52.1% of potentially HIV-exposed infants identified in the SCBW matched with the number of HIV-exposed infants identified in the PSD, this would suggest that many of the HIV-infected infants were identified by the PSD. Most HIV-exposed infants do not become infected.^3- 5

Of the 205 children, 134 (65%) received ART, PCP prophylaxis, or both by age 3 years (Table 3). Of these, only 37 (28%) progressed to a category C diagnosis, compared with 44 of 71 (62%) untreated children (P<.001). Treatment was associated with delayed progression to the first category C diagnosis (median age at first category C diagnosis for treated vs untreated children, 14 months vs 4 months; IQR, 9-19 months vs 3-8 months; P<.001). Overall survival was 80% (164/205), and treatment was associated with improved survival (P = .02) as a result of decreased category C diagnoses. However, once children progressed to category C, survival and age at death were not associated with prior treatment. Similarly, age at progression was not associated with survival at age 3 years: the children who survived were diagnosed with a category C condition at median age 10 months (IQR, 4-16 months), while the children who died were diagnosed with a category C condition at median age 6 months (IQR, 4-11 months) (P = .23).

Among the 134 treated children, 99 (74%) received ART and PCP prophylaxis, 20 (15%) only PCP prophylaxis, and 15 (11%) only ART (Table 3). Of the 114 children treated with ART, 59 (52%) received mono, 32 (28%) dual, and 23 (20%) triple ART as their most advanced ART (Figure). All children who received triple ART received a regimen that contained a PI or a nonnucleoside reverse transcriptase inhibitor. Children not receiving ART were more likely to progress to category C, compared with those receiving mono and dual ART with or without PCP prophylaxis. None of the 23 children who ever received triple ART before age 3 years progressed to category C; this association was highly significant compared with mono or dual ART or PCP prophylaxis (P<.001). The difference in progression to category C between mono ART (37% [22/59]) and dual ART (28% [9/32]) was not statistically significant, nor was it statistically different from the proportion of children receiving PCP prophylaxis without ART (30% [6/20]) who progressed to category C. However, any use of ART was associated with a delayed onset of category C symptoms. The median age at initial category C diagnosis was 4 months without any treatment, 8 months with only PCP prophylaxis, and 16 months with mono or dual ART with or without PCP prophylaxis (P<.001).

To determine the association between very early initiation of ART with or without PCP prophylaxis and disease progression, we compared mono or dual ART with or without PCP prophylaxis started before age 2 months vs between 3 to 4 months. Children who received triple ART in the first 3 years of life were not included in this analysis because none progressed to category C. The age cutoffs were chosen because 4 months was the median age at which untreated children progressed to category C, whereas only 3 untreated children progressed prior to age 2 months. Earlier ART was associated with delayed and decreased progression by age 3 years (P = .02). Of the 10 children who started mono or dual ART by age 2 months, 1 progressed to category C by age 12 months and a total of 3 progressed to category C by age 3 years. In contrast, of 16 children who started mono or dual ART by age 3 to 4 months, 8 progressed to category C by age 12 months, and by age 3 years a total of 11 had progressed to category C. Administration of PCP prophylaxis did not differ significantly between these 2 early treatment groups, with only 1 of 26 children not receiving PCP prophylaxis. That child progressed to category C but did not develop PCP. Among all 26 children, only 2 developed PCP by age 3 years. Both started ART at age 3 to 4 months and PCP prophylaxis by 3 months and were diagnosed with PCP at approximately 1 year.

The rate of progression to a category C diagnosis declined over time (P = .048 for trend) (Table 4). Moreover, all 7 children who progressed to category C in the late cohort survived, compared with less than 50% in the earlier cohorts (P = .005). In addition, while 25% (22/89) and 23% (19/83) of the 2 earlier cohorts died by age 3 years, all 33 children in the late cohort survived. The decline in disease progression in later cohorts was associated with earlier median age at initial diagnosis of HIV infection: 6 months in the earliest cohort vs 14 days in the latest (P<.001) (Table 4). For all birth cohorts, treatment was associated with decreased progression to category C.

Later birth years were associated with increased proportion of treated children, treatment with dual or triple ART, and earlier institution of ART. First, the proportion treated with ART with or without PCP prophylaxis increased with time (P = .01 for trend) (Table 4). This trend was driven by the increasing proportion of children receiving ART. In contrast, administration of PCP prophylaxis did not change significantly over time. The type of ART administered changed with time, with all but 1 treated child in the early cohort receiving only mono ART and all treated children in the late cohort receiving dual or triple ART. Finally, the median age of starting either PCP prophylaxis or ART decreased with time. This trend was for initiating both ART and PCP prophylaxis.

While children in the late cohort (n = 33) had the best outcomes, 7 (21%) still progressed to category C. Of these, 6 had no treatment. Only 1 of 26 treated children progressed to category C. This includes 7 children who did not receive triple ART but who started ART before age 2 months. The child who progressed started treatment at age 6 months with dual ART and PCP prophylaxis.

Table 5 outlines the age at diagnosis and types of category C diagnoses in the first 3 years of life relative to treatment. We excluded the 2 children who died of HIV complications without a clear category C diagnosis in this analysis. Initial category C diagnoses included HIV encephalopathy (n = 33), PCP (n = 28), candidal infections (n = 8), cytomegalovirus infections (n = 7), Mycobacterium avium infection (n = 1) , and recurrent bacterial infections (n = 6). Among 35 treated children who developed category C diagnoses, 24 (69%) demonstrated encephalopathy and 3 (9%) PCP at the time of initial category C diagnosis, compared with 9 (20%) and 25 (57%), respectively, of 44 untreated children (P<.001). However, the proportion of all HIV-infected children who developed encephalopathy as their initial category C diagnosis or at any time by age 3 years did not differ significantly between untreated and treated children. Moreover, the age at encephalopathy diagnosis was not associated with treatment. This confirms that encephalopathy did not occur more often among treated compared with untreated children. Untreated children who progressed were more likely than treated children to demonstrate cytomegalovirus infections as their initial category C diagnosis (16% [7/44] vs 0%, P<.02).

There are few data to guide time to initiation and selection of therapy among young children with perinatal HIV infection.^21,27 Our novel finding of improved outcomes even with mono/dual ART begun by age 2 months vs 3 to 4 months, though limited by the small sample size, suggests the importance of very early diagnosis and treatment and is consistent with small clinical trials demonstrating a short-term protective effect of early vs delayed ART among perinatally infected infants.^24,31 Initiating ART within the first 2 months offers the potential to begin therapy during or near the time of primary infection.⁵⁰ Others have observed that ART begun by age 2 to 2.5 months is associated with reduction of viral replication and serum proviral DNA levels, seroreversion, and prevention of clinical and immunological symptoms.^24,31 Similarly, early combination ART begun prior to age 3 months has been associated with long-term suppression of viral replication, seroreversion, and preserved immune status,^22,51 independent of viral load at initiation of therapy.²⁶ However, because there are potential drawbacks of very early therapy (particularly in asymptomatic infants), including concerns regarding compliance, risk of viral drug resistance, adverse effects, and long-term toxicity, large, prospective clinical trials defining the differences between very early vs delayed institution of therapy are needed.^27- 28

Analysis of our population-based cohort confirms that for children with perinatal HIV infection, outcomes by age 3 years improved with time, as children were evaluated for HIV infection earlier and were treated earlier, more often, and with more advanced ART. Over time, there was a delay and a decrease in progression to a category C diagnosis, associated with greater survival. Among all birth cohorts, any treatment and particularly more advanced ART were associated with improved outcomes at 3 years: reduced disease progression, older age at category C diagnosis, and improved survival. Outcomes with triple ART were especially favorable; none of 23 children in this analysis who ever received triple ART in the first 3 years of life progressed to category C diagnosis.

Previously, we demonstrated that PCP prophylaxis and zidovudine monotherapy altered early disease progression in part by preventing PCP but not encephalopathy.¹⁸ Similarly, in the current study, we found that treatment with other regimens besides triple ART altered early disease progression to category C diagnoses but did not have an impact on the development of encephalopathy. However, triple ART protected against development of all category C diagnoses, including encephalopathy. Our analysis is consistent with other observational studies demonstrating that triple ART, administered both before and after a category C diagnosis, is more strongly associated with less disease progression and death than mono/dual ART.^8- 9,11 However, only a few other studies have analyzed the effect of any early ART on the progression to category C diagnosis.^8,13 In a study by Abrams et al,⁸ the investigators demonstrated a short-term effect of early mono/dual ART on progression and a long-term effect of triple ART among HIV-infected infants for up to 24 months of age. Our results demonstrate a protective effect of triple ART up to age 3 years.

The scope of our 16-year population-based observational study makes it unique in its ability to describe associations between advances in disease management and the natural history of perinatal HIV infection through at least age 3 years. Only a limited number of non-US studies^10- 11 have conducted longitudinal assessment of HIV-infected children identified from birth or through registries; a few US studies^8,52 have enrolled pregnant HIV-infected women and their infants in long-term follow-up.

There are limitations to this study. First, the observational nature of the study limits the ability to collect a broader array of variables and may provide variable information based on the documentation and practice customs of each treating physician. However, the collected variables were complete for all 205 infants included in the analysis and only 16 additional infants were excluded because they were lost to follow-up, reflecting the relatively consistent practices and follow-up among the physicians at each of the sites, which are all major pediatric specialty hospitals. Incomplete capture of data for all HIV-exposed infants in the surveillance area could have biased the study outcomes. However, the capture rate was based on identification of all HIV-exposed infants, including those not HIV-infected. Since most infants born to HIV-infected women will be uninfected, it is likely that most infants not identified through surveillance were uninfected. Also, hospital-based active surveillance such as that described herein may enhance the likelihood of identifying HIV-infected children. Temporal changes in the standard of care could also have accounted for some improvement in clinical outcome noted in this study, although the most significant of these temporal changes include increasing use of ART and PCP prophylaxis. Finally, the analysis of early vs delayed therapy was based on only a subset of the total study population. Clearly, these results are not sufficient to make recommendations about very early therapy, but this is one of the first observational studies to identify this correlation with results of clinical trials that have found short-term virological and immunological responses to early ART.^22- 26 A recent French study made a similar observation of positive effect of early vs delayed therapy but followed up a small number of infants only through age 2 years.¹³

In our study, 61% of children, and all of those receiving triple ART, remained free of category C diagnoses at age 3 years. Moreover, 50% of those who progressed were still alive at age 3 years. Analysis beyond age 3 years is needed to understand the long-term effect of treatment on progression and survival of HIV-infected children. Such studies should be feasible because PIs have been used for at least 8 years.

Despite advances in early identification and treatment, even in the most recent birth cohort, there were children who were not identified as HIV-infected until progression to category C occurred. In addition, only 1 of 26 treated children in the most recent cohort progressed to category C, and this child did not start treatment until age 6 months. This finding emphasizes that aggressive testing of pregnant women and early detection of HIV-infected infants must continue in order to maximize the benefits of early HAART and PCP prophylaxis. In addition, the effect of very early therapy among HIV-infected infants must be determined in efforts to improve the outcome of perinatal HIV infection.