Post-HAART Outcomes in Pediatric Populations: Comparison of Resource-Limited and Developed Countries

Abstract

CONTEXT:

No formal comparison has been made between the pediatric post–highly active antiretroviral therapy (HAART) outcomes of resource-limited and developed countries.

OBJECTIVE:

To systematically quantify and compare major baseline characteristics and clinical end points after HAART between resource-limited and developed settings.

METHODS:

Published articles and abstracts (International AIDS Society 2009, Conference on Retroviruses and Opportunistic Infections 2010) were examined from inception (first available publication for each search engine) to March 2010. Publications that contained data on post-HAART mortality, weight-for-age z score (WAZ), CD4 count, or viral load (VL) changes in pediatric populations were reviewed. Selected studies met the following criteria: (1) patients were younger than 21 years; (2) HAART was given (≥3 antiretroviral medications); and (3) there were >20 patients. Data were extracted for baseline age, CD4 count, VL, WAZ, and mortality, CD4 and virologic suppression over time. Studies were categorized as having been performed in a resource-limited country (RLC) or developed country (DC) on the basis of the United Nations designation. Mean percentage of deaths per cohort and deaths per 100 child-years, baseline CD4 count, VL, WAZ, and age were calculated for RLCs and DCs and compared by using independent samples t tests.

RESULTS:

Forty RLC and 28 DC publications were selected (N = 17 875 RLCs; N = 1835 DC). Mean percentage of deaths per cohort and mean deaths per 100 child-years after HAART were significantly higher in RLCs than DCs (7.6 vs 1.6, P < .001, and 8.0 vs 0.9, P < .001, respectively). Mean baseline CD4% was 12% in RLCs and 23% in DCs (P = .01). Mean baseline VLs were 5.5 vs 4.7 log₁₀ copies per mL in RLCs versus DCs (P < .001).

CONCLUSIONS:

Baseline CD4% and VL differ markedly between DCs and RLCs, as does mortality after pediatric HAART. Earlier diagnosis and treatment of pediatric HIV in RLCs would be expected to result in better HAART outcomes.

Highly active antiretroviral therapy (HAART) results in marked survival benefits for HIV-infected people.^1,2 In contrast to adults, who may defer HAART for several years, nearly half of HIV-1 infected children in Africa will die by the age of 2 if they are not treated.^3,4 By 2005 in Africa, where ∼90% of the world's HIV-infected children reside, children represented 13% of the population in need of antiretroviral treatment (ART) and only ∼5% of the population receiving ART.^5,6 The number of children receiving ART has since increased; however, children are still less likely than adults to receive therapy.⁵

Many factors impede the use of HAART in resource-limited settings, particularly in pediatric populations. Lack of infrastructure, health care professionals, and technology to diagnose HIV-1 and monitor treatment have initially delayed treatment of both adults and children.⁷ The threat of poor adherence and viral resistance continues to be a concern in resource-limited settings.^8,9 Children face additional barriers to treatment including dosing, formulations, higher costs for pediatric antiretroviral drugs, and high infant mortality rates.^3,10–13

Beginning in 2004, African countries began expanding access to antiretroviral medications as funding became available.^14,15 Better descriptions of clinical diagnosis, staging, and management of HIV-infected children facilitated scale-up of treatment.¹⁶ A number of publications in which treatment outcomes for pediatric populations in resource-limited settings were described have recently emerged. These studies, including 2 recent reviews by Sutcliffe et al¹⁴ and Ciaranello et al,¹⁷ reference outcomes from developed-countries (DCs) publications for informal comparison; however, no study has systematically compared outcomes and characteristics of pediatric ART between resource-limited countries (RLCs) and DCs.

The purpose of this study was to review the literature to quantify and compare major clinical end points and baseline characteristics for children receiving HAART in DCs versus RLCs.

METHODS

Search Strategy

A systematic literature search was performed through March 2010 for all studies for which outcomes (mortality, weight-for-age z score [WAZ], CD4%, and viral load [VL]) were reported after initiation of HAART in pediatric patients. The following databases were searched: PubMed, EBSCO, Global Health Host, AIDSLine, and the Cochrane Library. Conference abstracts from the International AIDS Society 2009 and Conference on Retroviruses and Opportunistic Infections 2010 were searched, because these data likely have not had time to be published. Search terms included “pediatric,” “children,” “HIV,” “HAART,” “Africa,” “resource-limited,” “developing country,” “outcomes,” “mortality,” “efficacy,” and “adherence” (or equivalents of these terms [ie, HIV-1, ART, antiretrovirals, ARV, therapy, treatment]). This search strategy was supplemented by searching references in the bibliographies of articles.

Study Selection

Observational cohorts and clinical trial studies were selected for review on the basis of predefined criteria. Full-length articles published in a language other than English were included if they had an abstract in English. Studies were selected on the basis of the following criteria: (1) patients were younger than 21 years and not limited to a narrow age range such as <24 months or >13 years), (2) patients had received HAART (≥3 antiretroviral medications), (3) the sample size was >20 patients, and (4) patients had at least 6 months' follow-up on HAART. Outcome measures included mortality, weight change, CD4 counts and percentages, and VLs. Two authors reviewed the reports and came to agreement on inclusion or exclusion of the publications.

Data Extraction

In addition to the outcome measures, information was extracted on the focus of the study, regimen, previous ART exposure including prevention of mother-to-child transmission, time from presentation to initiation of HAART, disease severity, predictors of mortality, orphan status, hospitalization, follow-up time, percentage of patients lost to follow-up, and intent-to-treat versus as-treated analysis. Articles were separated into 2 categories, RLCs or DCs, according to rankings by the United Nations Statistics Division.¹⁸ Articles were then subcategorized according to geographic location. Studies were also grouped on the basis of the cohorts' previous HAART exposure: HAART-naive (previous mono/dual therapy or antiretroviral naive) or HAART-exposed (3-drug regimen including a protease inhibitor or a nonnucleoside reverse transcriptase inhibitor).

Multiple reports were reviewed for the same study, and individual studies were compared for overlap. Overlap was evaluated by reviewing authors, location, date, duration, and specific interventions. When results overlapped, data from the largest cohort, most recent publication, or longest follow-up time were selected. Multiple reports for the same or overlapping cohort were included if they each provided unique outcome data (eg, 1 reported mortality, 1 reported CD4%, or each reported CD4 count at different time points). Unique outcome data were extracted and added so that no overlapping data points were used in calculations. Reports were excluded if data were not used (see Appendix).

Calculations and Statistical Analysis

The mortality percentage was collected directly or calculated from reported results. Deaths per 100 child-years (DPCY) was calculated by using the number of deaths and time of the mortality measurement, unless reported directly in the article. The total number of child-years was estimated as the sum of years contributed by living patients at the time of mortality measurement and one-half of this follow-up time for deceased patients. These estimations are likely to be less accurate with longer follow-up times.

For articles in which mortality rates at multiple follow-up time points were provided, the mortality measurement nearest 12 months' follow-up was used, because it was the most commonly used follow-up time point.

Mean baseline characteristics and mortality rates were calculated for comparison between RLCs and DCs. All mean calculations were weighted on the basis of cohort size. Hereafter, weighted means will be referred to simply as means. The means, ranges, SDs, and confidence intervals of HAART-naive studies were calculated for both mortality percentage and DPCY for each geographic subregion, RLCs, and DCs. In addition, the RLC and DC means, ranges, SDs, and confidence intervals for baseline characteristics including CD4 T-cell percentage, VL (log₁₀ copies per mL), age, and WAZ were calculated for HAART-naive studies and for all studies that included HAART-experienced cohorts. CD4% and the percentage of patients who achieved virologic suppression were graphed over time, and the mean levels 12 months after HAART were calculated. The RLC and DC baseline values and outcomes were compared by using independent-samples t tests.

RESULTS

Study Selection and Characteristics

The initial literature search produced 723 publications: 313 published articles and 410 conference abstracts. Abstracts, methods, and/or results were reviewed, and 199 reports were found to contain some relevant selection criteria. Of these reports, 131 were excluded for reasons listed in Fig 1 and the Appendix, and the remaining 68 were used for analysis (RLCs = 40, total N = 18 882 and 17 875 approximately correcting for overlap; DCs = 28, total N = 3150 and 1835 approximately correcting for overlap). Characteristics of included studies are summarized in Table 1.

FIGURE 1.

Study-selection flowchart. ARV indicates antiretroviral medication.

TABLE 1.

Study Characteristics of Pediatric Cohorts in RLCs and DCs

Study Authors (Year)	Location	N	Age, Mean (b) or Median (a)	Study Details	Follow-up Time, Total (b) or Median (a), mo
RLCs
Hien et al19 (2009)	Burkina Faso	52	6.8 ya	IAS abstract	24b
Fassinou et al20 (2004)	Cote d'Ivoire	78	7.2 ya	21% HAART-experienced	21a
Rouet et al21 (2006)	Cote d'Ivoire	78	6.5 ya	PI vs NNRTI comparison	36a
Nyandiko et al22 (2006)	Kenya	279	6.0 ya	Rural, orphan comparison	34a
Song et al23 (2007)	Kenya	29	8.5 yb	Adult comparison	15b
Van Winghem et al24 (2008)	Kenya	648	5.5 ya	Adherence: MSF	48b
Wamalwa et al25 (2007)	Kenya	67	4.4 ya	Early response	9a
Reddi et al26 (2007)	KwaZulu Natal	151	5.7 ya	16% HAART-experienced	8a
Leyenaar et al27 (2009)	Lesotho	284	2.2 ya	BIPAI center of excellence	14b
Cohen et al28 (2009)	Lesotho	283	NR	IAS abstract, rural: MSF	24b
Bong et al29 (2007)	Malawi	439	6.0 ya	FDC	24b
—16 (2006)	Malawi	A: 436	<15 y	FDC	6b
		B: 233		FDC	12b
Weigel et al30 (2010)	Malawi	497	8.0 ya	CROI abstract, growth	24b
Marazzi et al31 (2006)	Mozambique	297	4.4 yb	Integrated public health program	9a
Vaz et al32 (2009)	Mozambique	1007	3.0 ya	IAS abstract, growth	48b
van Griensven et al33 (2008)	Rwanda	315	7.2 ya	Nurse-based care: MSF	45b
Diack MBaye et al34 (2005)	Senegal	98	5.0 ya	Non-specific focus	36b
Barth et al35 (2008)	South Africa	66	8 mo to 11 y	Rural, ART-naive	12b
Eley36 (2006)	South Africa	409	1.9 ya	Severe clinical disease	12b
Jaspan et al37 (2008)	South Africa	391	2.2 ya	PI vs NNRTI comparison	48b
Jooste et al38 (2005)	South Africa	100	1–14 y	Non-specific focus	6b
Smit et al39 (2009)	South Africa	615	1.8 ya	IAS abstract: Cape Town	46b
Blè et al40 (2007)	Tanzania	59	3 mo to 11 y	Orphan study	12b
Kamya et al41 (2007)	Uganda	250	9.2 yb	Genotypic mutations	14b
Bolton-Moore et al42 (2007)	Zambia	2938	6.8 ya	Providers (nonphysicians)	12a
Gupta et al43 (2009)	Zambia	103	8.0 ya	IAS: Triomune FDC	36b
Walker et al44 (2007)	Zambia	93	8.8 ya	Non-specific focus	24b
Janssens et al45 (2007)	Cambodia	212	6.0 ya	Split FDC	36b
Myung et al13 (2007)	Cambodia	117	5.5 ya	DOT	26b
Zhang et al46 (2007)	China	A: 51	NR	HAART-naive	13b
		B: 32	NR	HAART-experienced	13b
Rajasekaran et al47 (2009)	India	295	7.6 yb	Non-specific focus	10a
Kline et al48 (2007)	Romania	414	13.0 yb	82% drug-experienced	51a
Aurpibul et al49 (2009)	Thailand	225	7.4 ya	IAS: growth	55b
Lapphra et al40 (2008)	Thailand	139	6.0 ya	Siriraj Hospital	36a
Puthanakit et al51 (2007)	Thailand	192	7.6 yb	Chiang Mai Hospitals	29a
Romanelli et al52 (2006)	Brazil	43	2.4 yb	Dual vs triple antiretroviral therapy	48b
Martins et al53 (2009)	Brazil	196	NR	IAS: growth	6b
Martins et al54 (2009)	Brazil	196	NR	IAS: immunosuppression	6b
George et al12 (2007)	Haiti	236	6.3 ya	Treatment-naive	20a
Severe et al55 (2005)	Haiti	94	<13 y	Adult and child study	12b
DCs
Ghaffari et al56 (2004)	US	40	7.1 ya	PI: University of Florida, Gainesville	22b
King et al57 (2005)	US	41	6.4 ya	PACTG 403, PI nelfinavir	11b
Krogstad et al58 (1999)	US	62	3 mo to 13 y	PI nelfinavir, age groups	10a
Krogstad et al11 (2002)	US	192	6.2 ya	PACTG 377	11b
McKinney et al59 (2007)	US	37	10.5 ya	PACTG 1021	22b
Melvin et al60 (2002)	US	36	6.0 ya	5 patients overlap PACTG	29a
Patel et al61 (2008)	US	1236	NR	PACTG 219 10-y follow-up	70a
Rosenblatt et al62 (2005)	US	192	6.2 ya	PACTG 377	11b
Soh et al63 (2003)	US	702	6.7 ya	PACTG 219 CD4 response	48b
Spector et al64 (2000)	US	57	8.0 ya	PACTG 382	11b
Starr et al65 (1999)	US	57	8.0 yb	PACTG 382	11b
Watson et al66 (1999)	US	72	NR	Adherence and efficacy	9a
Wiznia et al67 (2000)	US	192	6.2 ya	PACTG 377	11b
Yogev et al68 (2002)	US	245	7.4 ya	PACTG 338 subset	11b
Bracher et al69 (2007)	Denmark	49	6.7 ya	Long term follow-up	72b
Teglas et al70 (2001)	France	33	12.5 ya	Efavirenz study	9b
Thuret et al71 (1999)	France	22	6.5 ya	Non-specific focus	21b
Wintergerst et al72 (2008)	Germany	33	8.2 ya	Efavirenz study	50a
Fraaij et al73 (2005)	Netherlands	31	5.1 ya	Prospective PI study	48b
Scherpbier et al74 (2007)	Netherlands	36	6.6 ya	Efavirenz study	11a
van Rossum et al75 (2002)	Netherlands	32	5.4 ya	Non-specific focus	22b
van Rossum et al76 (2000)	Netherlands	28	6.0 ya	Non-specific focus	6b
Verweel et al77 (2002)	Netherlands	24	5.2 ya	HAART effect on growth	22b
Nadal et al78 (2000)	Switzerland	A: 37	6.3 ya	Ritonavir	28a
		B: 237	7.8 ya	Nelfinavir	28a
Rudin et al79 (2008)	Switzerland	133	6.3 ya	PI comparison	66b
Judd et al80 (2007)	UK, Ireland	156	NR	CHIPS 2003–2006 antiretroviral-naive	>9
Walker et al81 (2004)	UK, Ireland	265	4.2 ya	CHIPS, antiretroviral-naive	24b
PENTA82 (2002)	8 countries	103	5.3 ya	PENTA	11b

IAS indicates International AIDS Society; PI, protease inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; MSF, Medecins Sans Frontiere; BIPAI, Baylor International Pediatric AIDS Initiative; —, No author provided; FDC, fixed-dose combination treatment; CROI, Conference on Retroviruses and Opportunistic Infections; DOT, directly observed therapy; NR, not reported; PACTG, Pediatric AIDS Clinical Trial Group; CHIPS, Collaborative HIV Paediatric Study; PENTA, Paediatric European Network for Treatment of AIDS.

^aMean/total.

^bMedian.

Baseline Mean/Median Age

The mean baseline age in RLC studies was 5.4 and 5.7 years in HAART-naive and all studies, respectively (Table 2). In DCs, the mean age of patients in HAART-naive and all studies was 6.5 and 6.7 years, respectively. There was no significant difference in the mean/median age at baseline between RLCs and DCs for HAART-naive cohorts (P = .1) or all studies (P = .2).

TABLE 2.

Pooled Summary Statistics: Comparison of RLCs and DCs

	RLCs	DCs	P
Mean mortality, %
HAART-naive	7.6 (0 to 18.8)a ± <0.1b (8937)	1.6 (0 to 3.8)a ± <0.1b (1241)	<.001
Ref No.	12, 13, 16, 19, 22–25, 27, 28, 31, 33, 35, 42, 44, 45, 47, 51, and 52	59, 61, 69, 71, 73, 79, and 82
All studiesc	7.5 (0 to 18.8)a ± <0.1b (9663)	1.7 (0 to 3.8)a ± <0.1b (1277)	<.001
Ref No.	Additionally 21, 26, 46, and 48	Additionally 74
Mean DPCY
HAART-naive	8.0 (0 to 41.5)a ± 4.1b (8937)	0.9 (0 to 1.1)a ± 1.2b (1241)	<.001
Ref No.	12, 13, 16, 19, 22–25, 27, 28, 31, 33, 35–42, 44, 45, 47, 51, and 52	59, 61, 69, 71, 73, 79, and 82
All studiesc	7.7 (0 to 41.5)a ± 3.8b (9663)	1.0 (0 to 3.1)a ± 1.2b (1277)	<.001
Ref No.	Additionally 21, 26, 46, and 48	Additionally 74
Mean baseline CD4%
HAART-naive	12 (5 to 20)a ± 0.1b (8437)	23 (7 to 47)a ± 0.2b (647)	.01
Ref No.	12, 13, 19, 22, 25, 27, 29, 32, 33, 35–37, 39, 41, 42, 44, 45, 47, and 50–52	11, 56, 57, 59, 66, 69, 71, 73, 78, and 82
All studiesc	12 (5 to 20)a ± 0.1b (8666)	23 (7 to 47)a ± 0.1b (719)	.003
Ref No.	Additionally 21 and 26	Additionally 60 and 74
Mean baseline VL, log10 copies per mL
HAART-naive	5.5 (5.1 to 6.1)a ± 2.7b (1882)	4.7 (4.4 to 5.2)a ± 2.2b (647)	<.001
Ref No.	12, 19, 23, 25, 34–37, 41, 46A;, 51, and 52	56, 57, 59, 66, 67, 69, 71, 73, 78, and 82
All studiesc	5.5 (4.9 to 6.1)a ± 2.7b (1992)	4.7 (3.6 to 5.2)a ± 2.0b (752)	<.001
Ref No.	Additionally 21 and 46B	Additionally 60, 70, and 74
Mean baseline age, y
HAART-naive	5.4 (1.8 to 10.0)a ± 4.0b (9494)	6.5 (5.1 to 10.5)a ± 3.4b (575)	.1
Ref No.	12, 13, 19, 22–25, 27, 29, 31, 34, 36, 37, 39, 41, 42, 44, 45, 46, 47, and 50–52	56, 57, 59, 67, 69, 71, 73, 78, and 82
All studiesc	5.7 (1.8 to 13.0)a ± 3.8b (10169)	6.7 (5.1 to 12.5)a ± 2.9b (680)	.2
Ref No.	Additionally 21, 26, 46B, and 48	Additionally 60, 70, and 74
Mean baseline WAZ
HAART-naive	−2.2 (−3.8 to −1.6)a ± 2.1b (5748)	−0.4 (−0.8 to −0.3)a ± 0.1b (180)	<.001
Ref No.	13, 19, 23, 27, 30, 33, 36, 37, 40, 42, 44, 46, 49, 53, and 55	56, 59, 65, 71, and 77
All studiesc	−2.2 (−3.8 to −1.5)a ± 2.0b (6009)	−0.4 (−0.8 to −0.3)a ± 0.1b (180)	<.001
Ref No.	Additionally 20, 26, and 46B	—

N is the total number of patients involved in the statistic. Ref 60 overlaps 2 patients with the Pediatric AIDS Clinical Trial Group 377.⁶⁷

^aRange.

^b95% confidence interval for weighted mean.

^cIncludes HAART-experienced patients.

Baseline WAZ

The mean WAZ for children who were initiated on HAART in RLCs was −2.2 for both HAART-naive and all studies combined (Table 2). In DCs, the mean WAZ for both HAART-naive and all studies was −0.4. There was a large and statistically significant difference between baseline WAZ in RLCs and DCs (P < .001).

Post-HAART Mortality Outcomes

The mortality analysis included 38 cohorts: 30 cohorts from RLCs (N = 9663) and 8 cohorts from DCs (N = 1277). Only outcomes for patients on HAART (≥3 antiretroviral medications) were used in the analysis. Several calendar studies were excluded, because the authors reported mortality rates for birth cohorts without separating outcomes for patients on no ART, mono/dual therapy, or HAART; these studies revealed decreased mortality rates after the introduction of HAART.^2,80,83–85 In addition, 11 study reports provided mortality data but were not included in pooled analysis because of overlap with larger or more recent studies.^*

Post-HAART mortality data for HAART-naive studies are shown in Fig 2 and Table 3. Geographic subregions used for RLCs were Africa (20 studies), Asia/Eastern Europe (4 studies), and South America/Caribbean (2 studies). The mean mortality rates in Africa, Asia/Eastern Europe, and South America/Caribbean were 7.4%, 8.8%, and 7.6% per cohort and 7.5, 11.9, and 8.4 DPCY, respectively. In US and European HAART-naive studies, the mean mortality rates were 1.6% per cohort and 0.9 DPCY.

FIGURE 2.

Pediatric DPCY after HAART: HAART-naive cohorts. ^a Includes mono/dual ART-experienced patients. Vertical line (…), DC mean DPCY for HAART-naive studies; vertical line (—), RLC mean DPCY for HAART-naive studies; horizontal lines, 95% confidence intervals for weighted mean.

TABLE 3.

Pediatric Mortality Rates After HAART in RLCs and DCs

Country	N	Time	Lost to Follow-up, %	Mortality, %	DPCY
RLCs
Burkina Faso (Hien et al19)	52	12 mo	2	3.8	3.9
Kenya (Nyandiko et al22)	279	34 mo (median)	11	6.8	2.5
Kenya (Song et al23)	29	15 mo	3	0	0
Kenya (Van Winghem et al24)	648	12 mo	NR	1.1	1.1
Kenya (Wamalwa et al25)	67	6 mo	NR	9.0	18.8
Lesotho (Cohen et al28)	283	12 mo	2	5.0	5.1
Lesotho (Leyenaar et al27)	284	14 mo	1	6.3	5.5
Malawi A16	436	6 mo	11	5.0	10.3
Malawi B16	233	12 mo	15	13.0	13.9
Mozambique (Marazzi et al31)	297	6 mo	NR	8.4	17.5
Rwanda (van Griensven et al33)	315	24 mo	4	2.5	1.4a
South Africa (Barth et al35)	66	12 mo	17	9.0	9.5
South Africa (Eley36)	407	12 mo	NR	15.4	16.8
South Africa (Jaspan et al37)	391	12 mo	2	6.9	7.2
South Africa (Jooste et al38)	100	6 mo	3	9.0	18.8
South Africa (Smit et al39)	615	12 mo	NR	10.2	10.8
Tanzania (Bl+è et al40)	59	12 mo	0	0	0
Uganda (Kamya et al41)	250	12 mo	1	5.2	5.3
Zambia (Bolton-Moore et al42)	2938	36 mo	13	8.3	6.6a
Zambia (Walker et al44)	93	24 mo	NR	6.5	3.3
Africa mean	n/a	n/a	NR	7.4 ± 0.6b	7.5 ± 4.9b
Cambodia (Myung et al13)	117	6 mo	NR	18.8	41.5
Cambodia (Janssens et al45)	212	12 mo	2	6.0	6.2
India (Rajasekaran et al47)	295	10 mo (median)	2	8.1	10.2
Thailand (Puthanakit et al51)	192	48 wk	NR	6.7	2.8a
Asia/Eastern Europe mean	n/a	n/a	NR	8.8 ± <0.1b	11.9 ± 9.2b
Brazil (Romanelli et al52)	43	34 mo (median)	9.3	0	0
Haiti (George et al12)	236	12 mo	10	9.0	9.9
South America/Caribbean mean	n/a	n/a	NR	7.6 ± 0.1b	8.4 ± 16.4b
DCs
PENTA82	103	48 wk	NR	1.0	1.1
Denmark (Bracher et al69)c	49	108 mo	NR	2.0	0.4a
France (Thuret et al71)c	22	16 mo (median)	NR	0	0
Netherlands (Fraaij et al73)c	31	48 mo	16	3.1	0.4
Switzerland (Rudin et al79)c	133	66 mo	1	3.8	1.1a
US (McKinney et al59)	27	24 mo	NR	0	0
US (Patel et al61)c	866	70 mo (median)	NR	1.4	0.9a
Mean	n/a	n/a	NR	1.6 ± 0.2b	0.9 ± 1.2b

NR indicates not reported; n/a, not applicable.

^aDirectly reported number of child-years of follow-up.

^b95% confidence intervals for weighted mean.

^cIncluded mono/dual ART-experienced patients.

Comparisons between RLCs and DCs are listed in Table 2. Post-HAART mortality rates for HAART-naive cohorts in RLCs were ∼5 and 9 times greater than in DCs: 7.6% vs 1.6% and 8.0 vs 0.9 for mortality percentage and DPCY, respectively (P = .002 and P < .001, respectively). Mean mortality rates for all studies that included previously mono/dual protease inhibitor– and nonnucleoside reverse transcriptase inhibitor–treated children were 7.5% vs 1.7% mortality percentage and 7.7 vs 1.0 DPCY for RLCs and DCs, respectively.

Baseline and Post-HAART CD4 Levels

Forty-two study reports provided unique data for either baseline CD4% or CD4% over time: 25 from RLCs and 17 from DCs. Twenty-three RLC and 13 DC studies were pooled for baseline CD4%, restricting to 1 value from overlapping cohorts (Table 2). HAART-naive studies had mean baseline CD4% values of 12% (range: 5%–20%) and 23% (range: 7%–47%) for RLCs and DCs, respectively (P = .01). Twenty reports from RLCs and 9 reports from DCs described changes in CD4% after HAART initiation (Fig 3). In this graphical presentation, overlap exists between the Pediatric AIDS Clinical Trial Group publications (namely, refs ^{58, 60, 62, 63}, and ⁶⁵). Mean CD4% 12 months after HAART was significantly different between RLC and DC studies: 24% and 27%, respectively (P = .03).

FIGURE 3.

CD4 T-cell percentage change over time. A, RLCs; B, DCs.

Baseline VLs and Post-HAART Virologic Suppression

Forty-three study reports provided unique data for either baseline VL or percent virologic suppression: 21 from RLCs and 22 from DCs. Fourteen RLC and 14 DC studies reported baseline VL (Table 2). Baseline VLs in HAART-naive studies were 5.5 log₁₀ copies per mL in RLCs and 4.7 log₁₀ copies per mL in DCs (P < .001). Nineteen RLC and 20 DC reports described the percentage of patients who achieved virologic suppression (Fig 4). Viral suppression was defined as <400 copies per mL. Six study reports only defined viral suppression as <50 copies per mL.^{26,39,43,51,60,72} Overlap exists between refs ^{11, 58, 64, 65}, and ⁶⁸; the Paediatric European Network for Treatment of AIDS overlaps with ref 81 but not with ref 80, because data from this reference were extracted for the 2003–2006 birth cohort. Twelve months after HAART, the mean percentage of children who achieved viral suppression was 65% in RLC and 49% in DC studies, and there was no significant difference between the 2 groups (P = .4). Eleven of the DC and 7 of the RLC studies reported using an intention-to-treat approach when evaluating the rate of virologic suppression.^†

FIGURE 4.

Percentage of patients who achieved viral suppression over time. A, RLCs; B, DCs. ^a Intention-to-treat analysis.

Predictors of Mortality

Weighted least-squares regression was used to determine if differences in mortality between RLCs and DCs diminished after controlling for baseline WAZ, CD4%, or VL. Adjusting for baseline CD4 level, the mortality difference between RLCs and DCs persisted (6.7% mortality difference; P = .01), and there was negligible evidence of confounding. There were fewer studies for which VL and WAZ were reported; however, the mortality difference between RLCs and DCs seemed to be confounded by baseline WAZ and VL.

Studies from both RLCs and DCs revealed associations between mortality rate, baseline CD4%, and VL.^{11,12,13,36,42,61,63,68} Low WAZ was a risk factor for mortality.^{12,26,27,40,44,77} Several RLC studies revealed that younger age was associated with mortality, whereas DC studies revealed conflicting findings regarding age and mortality.^{12,22,36,51,73,80} Finally, 2 RLC studies revealed that orphans had a higher mortality rate, although programs with >50% orphans achieved relatively low mortality rates overall.^22,23,45

Additional Study Characteristics

Information was also collected surrounding initiation of HAART. RLC studies referenced various World Health Organization (WHO) pediatric ART guidelines that recommend initiation of HAART at WHO stage 3 or 4 or at a CD4% of <15%, <20%, <15%, or <20%, depending on age, or <200 cells per μL. The authors of ref 45 (Cambodia) noted that 37% of evaluated children did not meet initiation criteria, and 10% of eligible children died before initiation. DC studies often did not report specific initiation criteria. Between 61% and 99% of children in RLC studies initiated HAART at WHO stage 3 or 4 disease.^27,36 Between 10% and 62% of children in DC studies had Centers for Disease Control and Prevention class C disease.^78,82 Only 1 RLC study⁴² (Zambia) reported median age at diagnosis: 5.5 years. The authors of ref 69 (Denmark) reported median age of diagnosis at 1.5 years; median age of initiation was 6.7 years. Another study⁸⁰ found that between 2004 and 2006 foreign-born children presented later than UK-born children: 7.6 versus 0.8 years, respectively. Four RLC studies reported time between diagnosis and HAART initiation ranging from a median of 53 days to 26 months.^20,33,40,44 RLC studies reported that <10% of subjects received antiretroviral medications in attempt to prevent vertical transmission (<1%–12%); however, it is possible that perinatal nevirapine exposure was not systematically ascertained or was underreported.^{16,20,25,26,27,33,38,41} Prevention of mother-to-child transmission was reported as being widely available in DCs, but rates of utilization were not specified.^11,61,67,80

DISCUSSION

In this study, we determined and compared baseline status and outcomes of children who initiated HAART in RLCs and DCs. As anticipated, mortality rates were dramatically lower with HAART than in studies before HAART and <10% in both settings. Mortality rates were higher in RLCs than in DCs, but the mortality difference observed was less than would have been expected on the basis of general childhood mortality estimates from those regions, which suggests that the added contact with care providers enhances survival beyond baseline, likely by prevention of common infectious diseases. RLC cohorts involved children with significantly lower baseline CD4 counts and WAZ and higher VLs, all of which would be expected to also contribute to increased mortality rates. Efforts to initiate HAART earlier would be expected to identify children before substantial immunosuppression, which should translate into improved survival rates.

Comparisons between observational studies have inherent limitations; nonetheless, these comparisons are important for the evaluation of programs and to guide future treatment. Although authors of recent reviews have described post-HAART outcomes in Africa, none has systematically compared outcomes between regions.¹⁴ RLC outcomes have been compared informally to DC outcomes without careful consideration of the cohort selection criteria or treatment regimens. Informal comparisons have been made to DC study reports that provided outcomes according to calendar years, including patients not on ART and those on HAART for years, infant studies, overlapping studies, and cohorts of < 20 patients.^{2,14,17,80–85,89–91} Publications included in this study were systematically screened on the basis of age, cohort size, and regimens. Outcome data were extracted to avoid analysis of data from overlapping cohorts, and standardized outcomes were compared.

The post-HAART mortality rates for HAART-naive children were fivefold to ninefold greater in RLCs than in DCs. Although not directly comparable, this difference was less than the difference between overall mortality rates for children between the ages of 1 and 5 years in RLCs and DCs (exceptions included Brazil and Thailand).⁹² Six studies from Zambia, Kenya, Rwanda, and Tanzania found that post-HAART mortality rates fell below 3.5 DPCY despite higher regional child mortality rates (range: 37–56 deaths per 1000 live births between the ages of 1 and 5 years, as estimated by subtracting infant mortality rates from mortality rates of children <5 years old).^{22–24,33,40,44,92} Although HIV contributes to overall child mortality rates in high HIV-prevalence areas, HAART seems to provide some survival benefits to patients, perhaps by simply bringing children in contact with medical services.^93,94

The CD4 and VL data provide important contextual information for interpretation of the mortality results by demonstrating that HAART programs in RLCs have reported efficient and comparable increases in CD4% and declines in VLs as DC programs. Significantly lower CD4 levels observed in RLCs 12 months after HAART are likely a result of markedly lower baseline CD4 levels. In contrast, post-HAART VL-suppression rates were not significantly different between DCs and RLCs after HAART despite significantly higher baseline VL in RLCs.

Strengths of this analysis include the large number of studies evaluated, the study-selection process, and the standardized comparisons of mortality rates, baseline CD4 percentage, and baseline VLs. This comprehensive comparison spanned 11 years of publications, during which drug regimens, guidelines, and patient populations evolved, particularly in DCs. Although later publications from DCs would be expected to use more potent medications, earlier study reports described larger HAART-naive cohorts, which provides better comparison of baseline characteristics and perhaps slightly underestimates the difference between DCs and RLCs.⁷⁴ A limitation of the study is that outcomes were likely biased toward better outcomes, because children lost to follow-up could include unreported deaths or immunologic and virologic nonresponders. We excluded cohorts that had only included children with a minimum amount of follow-up for the same reason. The studies we included had survivor bias (mean baseline age: >5 years); in RLCs, untreated HIV-infected children have only 50% survival rates below 2 years. As-treated analysis also results in overestimation of virologic suppression, because the proportion of children who achieved virologic suppression has been reported within the denominator of patients with VL data rather than the total number of patients who initiated treatment.

Low baseline CD4, WAZ, and high VL levels were identified by individual studies as strong predictors of mortality in both RLCs and DCs. In multivariate analysis, mortality-rate differences between RLCs and DCs persisted even after adjusting for baseline CD4 count. However, we found that at least a portion of the mortality-rate difference was attributable to differences in WAZ or VL in these different settings. In RLCs, children were older at the time of diagnosis and had more advanced disease, and the majority of deaths occurred in the first 6 months of treatment.^‡ Earlier identification of children could improve post-HAART outcomes by identifying children with less advanced disease.^95,96 Revised 2010 WHO treatment guidelines recommend treatment for all children younger than 24 months and a new CD4 threshold of 25% or 750 cells per μL for children aged 2 to 5 years. Additional studies are needed to evaluate the effect of these new guidelines on post-HAART outcomes.⁹⁷

CONCLUSIONS

Pediatric HAART programs in RLCs are successfully achieving a reduction in HIV- related mortality; however, post-HAART mortality rates remain higher than the rates in DCs. Currently, children in RLCs begin HAART at higher baseline VLs and lower baseline CD4 levels. Continuing to improve child health with interventions including nutritional support and prevention and treatment of coinfections may additionally improve survival rates. With increased availability of treatment and earlier treatment, regions with high HIV prevalence should realize marked declines in HIV-related child mortality.

APPENDIX.

Excluded Studies

Country	Author	Reason for Exclusion
1. Cote d'Ivoire	Adje-Toure et al98	Overlap with Fassinou et al20, subset analysis excluded patients who died
2. Ethiopia	Biadgilign et al99	Adherence study, cross-sectional
3. Nigeria	Onankpa et al100	Epidemiology study, no post-HAART outcomes reported
4. South Africa	Cowburn et al101	Mortality not reported, hospitalization study
5. South Africa 2004	Eley102	Overlap with Eley,36 data not used in analysis
6. South Africa	Prendergast et al103	Limited to infants followed from birth
7. South Africa	van Kooten et al104	Cohort too small (17 patients)
8. South Africa	Violari et al95	Age limited to 6–12 wk, early vs delayed antiretroviral medication
9. Togo	Atakouma et al105	Cross-sectional study
10. Togo	Polisset et al106	Adherence study, no post-HAART outcomes reported
11. Argentina	Fallo et al107	Calendar-year comparisons
12. Brazil	Matida et al108	Calendar surveillance
13. Brazil	Candiani et al109	No. on HAART not specified
14. Guatemala	Samayoa et al110	Results combine HAART and non-HAART
15. Jamaica	Evans-Gilbert et al111	Mortality not reported, hospitalization
16. Cambodia	Madec et al112	Age limited to >13 y
17. India	Kumarasamy et al113	Excluded patients with follow-up at <18 mo
18. India	Lodha et al114	Excluded patients with follow-up at <3 mo
19. India	Natu et al115	Mortality not reported
20. India	Pensi et al116	Cohort too small (13 patients)
21. Romania	Ferris et al117	Overlap with Kline et al,48 focus on disclosure
22. Romania 2004	Kline et al88	Overlap with Kline et al,48 data not used in analysis
23. Thailand	Chearskul et al118	Overlap with Lapphra et al40, data not used in analysis
24. Thailand	Koekkoek et al119	Cohorts too small (16 patients)
25. Thailand	Plipat et al120	Cohort too small (19 patients)
26. Multiple	O'Brien et al10	Overlap with multiple studies
27. Multiple	Arrive et al121	Overlap with multiple studies
28. Multiple	KIDS ART-LINC122	Overlap with multiple studies
29. Multiple	Saez-Llorens et al123	Did not isolate data from RLCs and DCs
30. Africa/Romania	Weidle et al124	Mortality not reported, dosing study
31. Lat. America	Hazra et al125	Results combined HAART- and non–HAART-treated patients
Europe
32. Belgium	Hainaut et al126	Cohort too small (4 patients), age limited to <2 mo
33. France	Aboulker et al127	Age limited to <3 mo
34. France	Faye et al128	Age limited to <1 y
35. Germany	Funk et al90	Cohort too small (16 patients)
36. Germany 1998	Wintergerst et al129	Cohort too small (15 patients)
37. Italy	Canani et al130	Cohort too small (10 patients)
38. Italy	Chiappini et al131	Overlap with de Martino et al87, calendar study
39. Italy	de Martino et al87	Overlap with PENTA,82 data not used in analysis
40. Italy	Vigano et al132	Cohort too small (11 patients), heavily pretreated
41. Netherlands 1998	Cohen et al133	Cohort too small (13 patients)
42. Spain	Larru et al134	Limited to patients whose conditions failed to respond HAART
43. Spain 2006, 2004	Resino et al135,85	Calendar study
44. Spain 2003	Sanchez et al136	Overlap with Larru et al134, Kaplan-Meier survival
45. Spain	Guillen Martin et al137	Epidemiology/immigrant study, no post-HAART outcomes reported
46. Switzerland	Steiner et al138	Overlap with Nadal et al78/Rudin et al79 excludes patients with <72 wk follow-up, growth study,
47. UK, Ireland	Gibb et al91	Overlap with PENTA,82 calendar study
48. UK, Ireland	Doerholt et al89	Age limited to <12 mo
49. 9 countries	Newell et al139	Overlap with PENTA82/Scherpbier et al74, results combined HAART- and non–HAART-treated patients
50. Europe 2009	Goetghebuer et al140	Limited to infants followed from birth
51. US 2001	Abrams et al141	Calendar study
52. US 2003	Benjamin et al142	PACTG 300 mono/dual treatment, growth study
53. US 2004	Berrien et al143	Adherence study, post-HAART outcomes not reported
54. US 2001	Blazevic et al144	Cohort too small (11 patients)
55. US 2000	Borkowsky et al145	Overlap with PACTG 338 reports, data not used in analysis
56. US 2010	Brady et al146	Calendar/birth-cohort study
57. US 2005	Brogly et al147	Overlap with PACTG 219C reports, calendar study, <24 y of age
58. US 2004	Brundage et al148	Overlap with PACTG 382 reports, data not used in analysis
59. US 2001	Buchacz et al149	Overlap with PACTG 219 reports, growth measures not comparable
60. US 2003	Caudill et al83	Results combined HAART- and non–HAART-treated patients
61. US 2005	Chadwick et al150	Age limited between 4 wk and 24 mo
62. US 2008	Chadwick et al151	Age limited to <6 mo
63. US 2001	Chougnet et al152	Overlap with Mueller et al168,169, excluded patients with clinical/immune decline
64. US 2002, 2004	Church et al153,9	Cohort too small (14 patients), Enfuvirtide study
65. US 1999	Essajee et al154	Limited to severely immunocompromised patients
66. US 2004, 2007	Flynn et al155	Age limited to 8–22 y
67. US 2007	Glikman et al157	Cohort too small (9 patients), adherence study
68. US 2006	Gona et al158	Overlap with PACTG 219, calendar comparison, opportunistic-infection study
69. US 2001	Gortmaker et al2	Overlap with PACTG 219, calendar study
70. US 2001	Jankelevich et al159	Excluded patients with follow up at <96 wk
71. US 2001	Johnston et al160	Immune-reconstitution study
72. US 2009	King et al161	Age limited to 10–18 y, pharmacokinetic study
73. US 1998	Kline et al162	Cohort too small (12 patients)
74. US 2006	Lee et al163	Overlap with PACTG 219, quality-of-life study
75. US 2000	Lindsey et al84	Meta-analysis included mono/dual/HAART
76. US 1997, 2004	Luzuriaga et al164,165	Age limited between 2 wk and 24 mo
77. US 2005	McConnell et al166	Calendar study
78. US 1997	Melvin et al167	Cohort too small (9 patients)
79. US 1998, 1998	Mueller et al168,169	Follow-up included 16 wk of monotherapy and only 12 wk on HAART, overlap with Jankelevich et al159
80. US 2000	Nachman et al170	Overlap with PACTG 338, data not used in analysis
81. US 1999	Palumbo et al171	Mono and dual therapy study
82. US 2008	Patel et al172	Overlap with PACTG 219, CD4% comparison between patients with and without HAART initiation
83. US 1999	Pelton et al86	Results combined HAART- and non–HAART-treated patients
84. US 2005	Pelton et al173	Overlap with PACTG 338, data not used in analysis
85. US 2001	Polis et al174	Overlap with Mueller et al168,169, monotherapy
86. US 2000	Reddington et al175	Overlap with PACTG 219 is unclear, adherence study
87. US 2008	Robbins et al176	Limited to patients whose conditions failed to respond to HAART therapy, pharmacokinetic study
88. US 1997	Rutstein et al177	Results combined HAART- and non–HAART-treated patients
89. US 2005	Storm et al178	PACTG 219, cross-sectional quality-of-life study
90. US 2002	Van Dyke et al179	Overlap with PACTG 377, adherence subset, data not used in analysis
91. US 2004	Viani et al180	Calendar comparison
92. US 2007	Wiznia et al181	Required HAART for 4 mo before study initiation, Enfuvirtide study
IAS Abstracts 2009
93. Kenya	Ayaya et al182	Results combined HAART- and non–HAART-treated patients
94. Kenya	McGrath et al183	Growth study comparison of children <3 y/>3-y patient outcomes
95. Kenya	Owiso et al184	Calendar-study comparison
96. Kenya	Wamalwa et al185	Overlap with published study, data not used in analysis
97. KwaZulu Natal	Ndirangu et al186	Overlap with Reddi et al,26 growth study
98. Malawi	Braun et al187	Limited to “infant” cohort, age not specified
99. Malawi	Dow et al188	Age limited to <6 wk
100. Malawi	Kabue et al189	Limited to patients failing first line HAART
101. Malawi	Kabue et al190	Limited to “infant” cohort age not specified
102. Swaziland	Chouraya et al191	Age limited to <12 mo
103. South Africa	Colvin et al192	Mortality not reported
104. South Africa	Coovadia et al193	Post-PMTCT study
105. South Africa	Fatti et al194	Overlap with Eley,36 data not used in analysis
106. South Africa	Fenner et al195	Comparison of <5 y/>5-y patient outcomes
107. South Africa	Kaplan et al196	Overlap unclear, hospital not listed
108. Uganda	Kekitiinwa et al197	Limited to malnourished children
109. Cambodia	Augustinova et al198	Age limited to <18 mo
110. Cambodia	Isaakidis et al199	Cross-sectional survey
111. Cambodia	Sophan et al200	Limited to patients whose conditions failed to respond to first-line HAART
112. Indian	Pandian et al201	Overlap with published study, data not used in analysis
113. Thailand	McConnel et al202	Overlap with Puthanakit et al51
114. Brazil	Rezende et al203	Overlap with Romanelli et al52, limited to patients with follow-up at >48 wk
115. Multiple	Carter et al204	Comparison of children <12 mo/>12 mo
116. Multiple	Hansudewechakul et al205	Overlap multiple studies
117. Southern Africa	Davies et al206	IeDEA, virologic failure study, does not report virologic suppression
118. US	Palumbo et al207	IMPAACT trial, age limited to <36 mo
119. Unspecified	Bognon et al208	Results combined HAART- and non–HAART-treated patients
CROI 2010
120. South Africa	Venkatesh et al209	Limited to infants followed from birth
121. Uganda	Achan et al210	Outcomes reported virologic failure
122. Southern Africa	Becquet et al211	Survival study, not HAART-focused
123. India	Pandian et al212	Overlap with Rajasekaran et al47
124. Thailand	Sudjaritruk et al213	No mortality reported, hospitalization study
125. Spain	Palladino et al214	Fosamprenavir study, experimental
126. US	Nachman et al215	Cohort too small (10 patients)

PENTA indicates Paediatric European Network for Treatment of AIDS; PACTG, Pediatric AIDS Clinical Trial Group; IAS, International AIDS Society; PMTCT, prevention of mother-to-child transmission; IeDEA, International Epidemiologic Databases to Evaluate AIDS; IMPAACT, International Maternal Pediatric Adolescent AIDS Clinical Trials Group; CROI, Conference on Retroviruses and Opportunistic Infections.