Immunodeficiency at the start of combination antiretroviral therapy in low-, middle- and high-income countries

Abstract

Objectives

To describe the CD4 cell count at the start of combination antiretroviral therapy (cART) in low-income (LIC), lower middle-income (LMIC), upper middle-income (UMIC) and high-income (HIC) countries.

Methods

Patients aged ≥16 years starting cART in a clinic participating in a multi-cohort collaboration spanning six continents (International epidemiological Databases to Evaluate AIDS and ART Cohort Collaboration) were eligible. Multi-level linear regression models were adjusted for age, gender and calendar year; missing CD4 counts were imputed.

Findings

379,865 patients from nine LIC, four LMIC, four UMIC and six HIC were included. In LIC the median CD4 cell count at cART initiation increased by 83% from 80 to 145 cells/μl between 2002 and 2009. Corresponding increases in LMIC, UMIC and HIC were from 87 to 155 cells/μl (76% increase), 88 to 135 cells/μl (53%) and 209 to 274 cells/μl (31%). In 2009, compared to LIC, median counts were 13 cells/μl (95% CI -56 to +30) lower in LMIC, 22 cells/μl (-62 to +18) lower in UMIC and 112 /μl (+75 to +149) higher in HIC. They were 23 cells/μl (95% CI +18 to +28) higher in women than men. Median counts were 88 cells/μl (95% CI +35 to +141) higher in countries with an estimated national cART coverage >80%, compared to countries with <40% coverage.

Conclusions

Median CD4 cell counts at start of cART increased 2000-2009 but remained below 200 cells/μl in LIC and MIC and below 300 cells/μl in HIC. Earlier start of cART will require substantial efforts and resources globally.

Introduction

The prognosis of HIV-positive patients has dramatically improved with the advent, in 1996, of combination antiretroviral therapy (cART) [1, 2]. Suppressed viral replication allows reconstitution of the immune system: peripheral CD4 cell counts increase rapidly first from redistribution from lymphoid tissues, and then gradual by de novo synthesis [3, 4]. Since 2002, the Global Fund for Tuberculosis, AIDS and Malaria (GFTAM), US President’s Emergency Plan for AIDS Relief (PEPFAR) and other funders have sharply increased global cART availability. The World Health Organization (WHO) estimated, by 2010, that 6.6 million of the 15 million who needed cART in low and middle income countries had access [5].

When to initiate cART to maximize the benefit of therapy has been debated for years [6]. Benefits of early initiation, at high CD4 cell counts, must be balanced against drug toxicities and the potential for drug resistance. Conversely, starting therapy late, as measured clinically or by CD4 count, is associated with poorer prognosis and increased mortality [7]. A sub-study of the Strategies for Management of Antiretroviral Therapy (SMART) trial showed that delaying cART until the count fell below 250 cells/μl more than tripled the rate of AIDS or death compared to starting above 350 cells/μl [8]. Analyses that combined data from cohort studies also indicated that starting cART above 350 CD4 cells/μl is beneficial, and some, but not all, showed benefit with a threshold of 500 cells/μl [9-11]. The START (NCT00821171) and TEMPRANO (NCT00495651) trials will provide further data on the efficacy of early versus late initiation of cART.

However, many patients enter care at late. An analysis of treatment programs in 12 countries in sub-Saharan Africa, South America and Asia showed that while CD4 cell counts at initiation increased from 2001 to 2005/2006, most started well below recommended thresholds [12]. Similarly, a United States of America (USA) and Canada cohort showed that median CD4 cell count at first presentation for HIV care was 317 cells/μl in 2007: more than half of patients initiated therapy below 350 cells/μl [13]. A recent Latin American study reported that the percentage of patients initiating cART late ranged from 56% in Argentina to 91% in Honduras [14].

Early initiation of cART is recognized as having a broader role in HIV prevention [15]. Already established as a means to prevent mother-to-child transmission [5], the HIV Prevention Trials Network (HPTN) 052 trial found cART reduced heterosexual HIV transmission by 96% between discordant couples [16]. Combined with other proven prevention tools, immediate or early cART might contribute to achieving the goal of an AIDS-free generation [17].

We examined trends and determinants of the CD4 cell count at cART initiation in patients starting therapy between 2002 and 2010 in low, middle and high income countries by combining data from two HIV cohort consortia, which together span six continents.

Methods

Data sources

The International epidemiological Databases to Evaluate AIDS (IeDEA) is a global consortium structured through regional centres to pool clinical and epidemiological data on HIV-positive individuals, particularly patients on cART. The seven regions included in IeDEA are North America, Caribbean/Central and South America, Asia/Pacific, East Africa, West Africa, Central Africa and Southern Africa. Regional cohorts of IeDEA have been described in detail elsewhere [18-21]. The European cohorts of the ART Cohort Collaboration, a network of cohort studies of patients on cART in high-income countries, were also included [7]. Pooling of data and their use in collaborative analyses were approved by local ethics committees and institutional review boards. For the present study, regional centers sent de-identified data to the University of Bern, Switzerland for cleaning and analysis.

Inclusion criteria and definitions

Patients aged ≥16 years at cART initiation were eligible. cART was defined as a regimen of at least three antiretroviral drugs, typically from two drug classes. Baseline CD4 cell count was defined as the count nearest to the date of cART start with a window of 6 months before to 1 month after starting. CD4 cell counts above 5,000 cells/μl (i.e. >3 times above the upper reference range in Caucasians [22]) were considered invalid and set to missing. Countries were grouped according to the World Bank classification of Gross National Income per Capita 2010 as low-income (LIC, US$1,005 or less per year); lower middle-income (LMIC, US$1,006 to 3,975 per year); upper middle-income (UMIC, US$3,976 to 12,275 per year) and high-income (HIC, US$12,276 or more per year) [23]. Data on national cART coverage for 2009 (based on the WHO 2006 treatment guidelines [24]) were obtained from the 2010 progress report on the Global HIV/AIDS response published by WHO [25] for LIC and MIC.

Statistical analysis

Descriptive analyses were stratified by country, gender and World Bank country income group. CD4 cell counts at cART start and other baseline characteristics were summarized as medians with interquartile ranges (IQR) or numbers with percentages. To address the problem of generalizing data from a small sample to an entire country, further analyses were restricted to countries that contributed at least 500 patients with CD4 cell counts from two or more sites and had observations after 2007.

Missing CD4 counts were multiply imputed using predictive mean matching and chained equations stratified by gender, age and country income level. Technical details on the multiple imputation are provided in the supplementary materials. All analyses were performed both for the imputed data and for the complete case dataset. To assess trends and examine individual-level and country-level predictors of median baseline CD4 cell counts, we aggregated the data by sex, age, year and country. We then fitted three types of weighted linear regression models. First, a simple linear regression estimated gender-specific annual changes in median baseline CD4 cell counts by country (model 1). In sensitivity analyses, we weighted individual country data to create a dataset that was representative, within each income group, of the number of patients on cART in each country in 2009, as estimated by WHO [5]. The derivation of the weighting factors is shown in supplementary Table S1.

Second, a weighted mixed-effects linear regression was used to estimate the median CD4 cell counts at start of cART in 2009 and to examine the influence of age, sex, country income level and, in an analysis restricted to LIC and MIC, of national cART coverage (model 2). Age, calendar year, gender and country income level were entered as fixed effects and country as random effect. The third model included calendar year, gender and country income level and was used to model median CD4 cell count trends between 2002 and 2010 (model 3). Finally, the proportions of patients starting cART with CD4 cell counts below 50, 100 and 200 cells/μl were analyzed using generalized linear mixed effects models. Age, calendar year, gender and country income level were entered as fixed effects and intercept and slope as random effects, by country. Data were analysed using Stata 12.0 (Stata Corporation, College Station, Texas, USA) and R 2.12 (The R Development Core Team, Vienna, Austria).

Results

Descriptive analyses

Data from 492,915 patients ≥16 years of age who started cART in 48 countries were submitted to the data center (Figure 1). Among 437,230 ART-naïve patients with known age, gender and start date, 309,564 (63%) had a CD4 cell count at start of cART. Compared to the 127,666 patients without a CD4 cell count, patients with counts at cART initiation were younger, more likely to be male, and more likely to be from a high income country. Importantly, those with CD4 cell counts were less likely to have advanced disease (WHO stage III/IV) than patients without CD4 cell counts (Table S2). Only 4.6% (20,217 patients) had an earlier CD4 cell count, 6 months or more before starting ART.

Figure 1.

Flow chart of patients included and excluded from analyses.

The number of included patients from each country varied from 60 from Japan to 147,029 from Zambia (Table 1). There were 12 countries with up to 500 patients, 23 with 501 to 5,000 patients, four with 5,001 to 10,000 patients, seven with 10,001 to 50,000 and two with more than 50,000 patients (Figure S1). The percentage of female patients ranged from 4% in Taiwan to 73% in Burundi, and median age ranged from 30.9 years in Indonesia to 41.5 years in Nigeria (Table 1). The median year of cART initiation ranged from 2000 in Australia and Italy to 2009 in the Philippines. The median CD4 cell count at cART-initiation for the entire study period ranged from 56 cells/μl in Indonesia to 290 cells/μl in Australia.

Table 1.

Characteristics of patients starting cART by country and World Bank income groups.

Country	No. of patients		Median age in years (IQR)		Median calendar year of starting cART (IQR)		Median CD4 cell count at start of cART in cells/μl (IQR)
	Women	Men	Women	Men	Women	Men	Women	Men
Low income
Benin	413	298	33.8	40.1	2005	2005	119	90
Burkina Faso	822	348	35.7	41.3	2007	2006	182	137.5
Burundi	332	139	36.4	43.0	2009	2009	209.5	157
Cambodia	107	110	33.8	36.6	2005	2005	103	70.5
DR Congo	1828	800	38.5	43.3	2008	2008	145	139.5
Gambia	140	77	37.4	45.0	2006	2006	160	130
Haiti	780	668	38.0	40.0	2004	2004	115.5	92.5
Kenya	19454	10571	35.2	39.6	2007	2006	135	109
Malawi	4337	3365	33.3	37.2	2008	2008	150	122
Mali	1020	650	32.7	40.2	2006	2006	134	101
Mozambique	284	177	32.1	37.5	2008	2008	242	204
Rwanda	1500	521	33.8	38.0	2006	2007	218	183
Tanzania	2168	1050	36.5	41.4	2007	2007	107	106
Uganda	4229	2552	34.2	38.3	2007	2007	130	114
Zimbabwe	3112	1388	37.1	39.7	2009	2009	133	96
Overall	40526	22714	35.1 (29 - 41)	39.4 (33 - 46)	2007 (2006 - 2008)	2007 (2006 - 2008)	139 (65 - 213)	113 (40 - 186)
Lower middle income
Cameroon	1821	829	33.9	39.3	2009	2009	158	139
Côte d’Ivoire	8531	4857	34.5	41.1	2006	2005	156	125
Honduras	101	159	35.0	38.0	2005	2005	120	98
India	91	364	31.9	34.1	2004	2003	177	129.5
Indonesia	92	227	29.1	31.4	2008	2008	89.5	45
Nigeria	4728	2678	41.2	41.3	2006	2006	162	127
Philippines	27	234	39.3	30.5	2008	2009	186	180
Senegal	183	126	35.4	42.0	2004	2004.5	122	97.5
Zambia	50096	32054	33.5	37.3	2008	2008	150	129
Overall	65670	41528	34.5 (28 - 41)	38.4 (33 - 44)	2007 (2006 - 2008)	2007 (2006 - 2008)	152 (76 - 228)	128 (55 - 201)
Upper middle income
Argentina	208	479	34.0	36.0	2004	2004	181	148
Botswana	549	367	34.3	38.4	2003	2003	140	116
Brazil	371	740	37.0	36.1	2007	2007	208	186
Chile	56	383	34.1	36.6	2003	2003	138	112
China	124	409	38.9	39.6	2005	2007	169.5	91
Malaysia	111	380	34.1	36.2	2007	2008	159	113
Mexico	49	349	35.0	35.0	2004	2004	131	84
Peru	208	495	32.0	35.0	2005	2005	85.5	74
South Africa	32920	21673	33.6	39.0	2007	2007	124	111
Thailand	731	907	35.7	36.3	2007	2008	95	89
Overall	35327	26182	33.7 (28 - 39)	38.5 (32 - 45)	2007 (2006 - 2008)	2006 (2004 - 2008)	124 (60 - 188)	111 (42 - 180)
High income
Australia	74	1261	33.7	39.1	2001	1999	280	290
Canada	850	3651	36.0	41.0	2002	2002	210	220
France	10023	20739	34.3	38.8	2003	2001	237	233
Germany	269	955	33.9	38.5	2004	2002	177	173
Italy	915	2312	35.4	37.6	2000	2000	265	244
Japan	4	56	46.1	38.1	2006	2007	236	253.5
Netherlands	1617	5125	32.4	39.4	2003	2002	210	189
Singapore	19	83	33.9	41.4	2004	2003	134	57
South Korea	10	182	41.2	38.0	2003.5	2006	123	222.5
Spain	2596	8406	34.0	36.8	2002	2003	224	201
Switzerland	1414	3124	33.9	38.4	2001	2001	201	205
Taiwan	8	186	32.2	32.9	2001.5	2003	132	225
UK	369	1016	34.3	36.9	2003	2002	158	206
USA	3362	8991	39.0	40.0	2003	2003	241.5	215
Overall	21530	56087	34.9 (28 - 41)	38.9 (33 - 45)	2002 (2000 - 2004)	2002 (1999 - 2005)	230 (110 - 350)	217 (89 - 345)

Results from descriptive analyses based on 309,564 patients.

Regression models

A total of 379,865 patients from 23 countries were included in the regression models, including 86,390 patients from nine LIC (23%), 176,858 from four LMIC (47%), 82,152 from four UMIC (22%) and 34,465 from six HIC (9%). The Democratic Republic of the Congo, Kenya, Malawi and Mali were overrepresented when compared to the WHO estimates of the number of patients on cART in the LIC included in our study. In contrast, Tanzania and Zimbabwe were underrepresented. The group of LMIC countries was dominated by Zambia, and that of UMIC countries by South Africa. Among HIC, France contributed almost half of all patients whereas the United States and Italy were underrepresented (Table S1).

The annual increase in median CD4 cell counts at start of cART initiation from 2002 to 2009, estimated through linear regression (model 1), varied within and across World Bank country groups and tended to be greater among women than men (Table 2). Among LIC, the annual change in median CD4 cell counts ranged from -14 cells/μl (95% CI -37 to +9 cells/μl) among women in the Democratic Republic of the Congo to +32 cells/μl (95% CI +22 to +42 cells/μl) among Rwandan women. For LMIC the corresponding range was from -2 cells/μl (95% CI -14 to +9 cells/μl) among Nigerian men to +13 cells/μl (95% CI +8 to +19 cells/μl) among women in Côte d’Ivoire. In UMIC the range was from -1 cells/μl (-17 to +15 cells/μl) in Botswana men to +15 cells/μl (95% CI 0 to +29 cells/μl) among Brazilian men. Finally, in HIC the range was from -3 cells/μl (95% CI -12 to +7 cells/μl) in Australian men to +16 cells/μl (95% CI +12 to +20 cells/μl) in Canadian men. Crude (unweighted) and weighted pooled estimates of annual increases in CD4 cell count at ART-initiation by country group were generally similar (Table 2). Also, results were similar in the complete case analysis (supplementary Tables S2 and S3).

Table 2.

Annual change between 2002 and 2009 in median CD4 cell count at the start of cART in low income, lower middle income, upper middle income and high income countries, by gender.

Country		Women		Men
		CD4 cells/μl	(95% CI)	CD4 cells/μl	(95% CI)
Low income
Benin		+8	(-15 to +31)	+4	(-11 to +19)
DR Congo		-14	(-37 to +9)	+6	(-11 to +22)
Kenya		+15	(+13 to +18)	+11	(+8 to +14)
Malawi		+9	(+2 to +15)	+12	(+4 to +21)
Mali		+12	(-2 to +27)	+4	(-1 to +10)
Rwanda		+32	(+22 to +42)	+31	(+12 to +51)
Tanzania		+7	(-1 to +14)	+2	(-4 to +8)
Uganda		+23	(+11 to +35)	+16	(+1 to +31)
Zimbabwe		+8	(+1 to +15)	+5	(-1 to +11)
Pooled	crude	+12	(+6 to +17)	+11	(+7 to +15)
	weighted*	+11	(+4 to +17)	+9	(+4 to +14)
Lower middle income
Cameroon		+10	(-4 to +24)	+13	(-4 to +31)
Côte d’Ivoire		+13	(+8 to +19)	+10	(-0 to +20)
Nigeria		+4	(-13 to +21)	-2	(-14 to +9)
Zambia		+11	(+9 to +13)	+8	(+4 to +11)
Pooled	crude	+9	(+6 to +12)	+7	(+4 to +10)
	weighted*	+6	(+1 to +11)	+5	(+1 to +9)
Upper middle income
Botswana		+9	(-4 to +23)	-1	(-17 to +15)
Brazil		+8	(-8 to +24)	+15	(+0 to +29)
South Africa		+9	(+8 to +11)	+4	(+1 to +7)
Thailand		+10	(+2 to +17)	+6	(-6 to +18)
Pooled	crude	+9	(+6 to +11)	+4	(+1 to +7)
	weighted*	+7	(+1 to +13)	+6	(-1 to +13)
High income
Australia		+10	(-18 to +38)	-3	(-12 to +7)
Canada		+7	(-1 to +15)	+16	(+12 to +20)
France		+11	(+8 to +13)	+9	(+3 to +15)
Italy		+4	(-3 to +11)	+6	(-3 to +14)
Spain		+3	(-1 to +7)	+6	(+2 to +10)
USA		+9	(+2 to +16)	+13	(+6 to +20)
Pooled	crude	+9	(+6 to +11)	+9	(+6 to +12)
	weighted*	+7	(+4 to +11)	+9	(+6 to +13)

Results from linear regression (model 1) based on 347,919 patients, with missing values imputed using multiple imputation.

^*Weighted by the number of patients on cART in the respective country, as estimated by WHO [5].

Table 3 shows estimated median CD4 cell counts at start of cART in 2009 by age, gender and country income level, as estimated from the mixed effects linear regression (model 2). Median CD4 cell counts were higher in women compared to men (difference +23 cells/μl, 95% CI +18 to +28 cells/μl) and lower for patients aged from 30 to <40 (difference -16 cells/μl, 95% CI -21 to -10) and 40 to <50 years (difference -17 cells/μl, 95% CI -24 to -10) compared to those below 30 years of age. Median counts were similar in LIC, LMIC and UMIC, but were higher in HIC (difference compared to low income countries +112 cells/μl, 95% CI +75 to +149 cells/μl). In the analysis restricted to LIC and MIC, those with cART coverage of 80% or greater (Botswana, Brazil, Cambodia, Rwanda and Zambia) had substantially higher CD4 cell counts at cART initiation (difference +88 cells/μl; 95% CI +35 to +141 cells/μl) than countries with a coverage below 40% (Côte d’Ivoire, Democratic Republic of the Congo, Indonesia, Nigeria and Malaysia). Results were again similar in the complete case analysis (Table S4).

Table 3.

Individual-level and country-level predictors of the median CD4 cell count at the start of cART in 2009.

Variable	Median CD4 cell count (cells/μl)
Sex
Male	164 (intercept, 140 to 189)
Female	23 (18 to 28)
Income group
Low	164 (intercept, 140 to 189)
Lower middle	-13 (-56 to 30)
Upper middle	-22 (-62 to 18)
High	112 (75 to 149)
Age group (years)
< 30	164 (intercept, 140 to 189)
30 to <40	-16 (-21 to -10)
40 to <50	-17 (-24 to -10)
≥50	-3 (-12 to 6)
National cART coverage (%)*
< 40	144 (intercept, 103 to 185)
40 to < 60	10 (-38 to 57)
60 to < 80	-6 (-58 to 46)
≥80	88 (35 to 141)

Results from mixed effects linear regression (model 2) based on 55,007 patients starting cART in 2009, missing values imputed using multiple imputation. Intercepts and coefficients (95% confidence intervals) are shown. All models include calendar year, age, gender and income group. The intercept of 165 cells/μl corresponds to men in low income countries.

^*Separate analysis based on 52,482 patients starting cART in 2009 in low income and middle income countries (model 2). The intercept of 146 cells/μl corresponds to men in low income countries. Estimates of national cART coverage in 2009, based on WHO 2006 guidelines were as follows [25]: Benin 72%, Democratic Republic of the Congo 26%, Kenya 65%, Malawi 63%, Mali 65%, Rwanda >95%, Tanzania 44%, Uganda 53%, Zimbabwe 49%, Cameroon 41%, Côte d’Ivoire 39%, Nigeria 31%, Zambia 85%, Botswana >95%, Brazil 80%, South Africa 56%, Thailand 76%.

Figure S2 shows trajectories of median CD4 cell counts at cART initiation and the percentage of patients starting cART below 200 cells/μl and below 100 cells/μl, by country income group and gender, estimated by model 3. In LIC median CD4 cell count at start of cART in women increased by 91% from 2002-2009, from 82 cells/μl to 157 cells/μl, and by 62% from 79 cells/μl to 127 in men. Corresponding increases for LMIC were from 83 cells/μl to 166 cells/μl in women (a 102% increase) and from 97 cells/μl to 138 cells/μl in men (a 42% increase). The increases in UMIC were from 86 cells/μl to 141 cells/μl in women (a 63% increase) and from 91 cells/μl to 125 cells/μl in men (a 36% increase). Finally, in HIC median CD4 cell counts at cART initiation increased by 26% from 211 cells/μl to 266 cells/μl in women and by 34% from 208 cells/μl to 280 cells/μl in men.

In LIC the percentage of patients starting cART below 200 cells/μl declined from 85% in 2002 to 81% in 2009 in women and from 88% to 70% in men. Corresponding figures for LMIC were from 78% to 59% in women and from 79% to 67% in men and for UMIC from 86% to 65% in women and from 78% to 71% in men. In HIC the decline was from 46% to 31% in women and from 49% to 35% in men. For threshold below 100 cells/μl the declines were from 50% to 28% in women and from 57% to 36% in men in LIC, from 47% to 27% in women and from 51% to 36% in men in LMIC, from 54% to 31% in women and from 49% to 40% in men in UMIC and from 22% to 15% in women and from 27% to 17% in men in HIC. Trends were similar in the complete case analysis (Figure S2).

Discussion

This global analysis of CD4 cell counts at cART initiation between 2002 and 2010 was conducted in two HIV cohort collaborations, which together span six continents [7, 18-21]. We found that median CD4 cell counts at cART initiation were substantially higher in HIC, with only small differences between LIC and MIC. Median CD4 cell counts at start of cART increased over the study period in most countries. These increases were greater in LIC and MIC than in HIC, and greater in women than in men. Among LIC and MIC, median CD4 cell counts at start of cART were substantially higher in the few countries with national cART coverage of 80% or above.

In LIC and MIC the median CD4 cell count remained well below 200 cells/μl between 2002 and 2009, despite the 2001WHO recommendation to start when the CD4 gets near or falls below 200 cells/μl or, in persons with advanced clinical disease (WHO stage IV), irrespective of the CD4 cell count [26]. Following a 2006 recommendation to consider treatment at CD4 cell counts below 350 cells/μl for patients in WHO stage III, WHO indicated in 2009 that cART should be initiated at 350 cells/μl irrespective of clinical symptoms [27]. National guidelines in resource-limited settings generally echoed WHO guidelines [28], while HIC have more rapidly increased the CD4 cell count threshold for initiation. Of note, North American guidelines recently converged in their recommendation that cART should be offered to all HIV-positive individuals, irrespective of the CD4 cell count [29, 30].

A substantial rise HIV testing in many countries, supported by GFTAM, PEPFAR, national or state/provincial level governments and other donors may have contributed to increasing CD4 cell counts at the start of ART. A monitoring and evaluation analysis of PEPFAR-supported HIV care clinics in eight sub-Saharan African countries found that CD4 cell counts at cART initiation increased with HIV testing coverage in the region [31]. In sub-Saharan Africa, PEPFAR supported over 140 million testing and counseling sessions between 2004 to 2011, with the number of sessions increasing from 1.9 million in 2004 to over 40 million in 2011 [32]. Two Demographic and Health Surveys (DHS) from seven PEPFAR countries conducted between 2003 and 2010 showed a dramatic increase in population level coverage of HIV testing and counseling [32]. In Kenya, for example, the percentage of men reporting testing in the last 12 months increased from 7.5% to 22.7%. The corresponding increase was even greater in women, from 6.7% to 29.3%. Similarly, in Lesotho, testing increased from 6.3% to 42.0% in women and from 4.8% to 24.0% in men. Provider-initiated testing and counseling may not, however, be sufficient to prevent late HIV diagnosis. Uganda adopted provider-initiated HIV testing in the health care setting in 2005, but in a recent randomized controlled trial half of HIV-positive patients screened had CD4 cell counts below or equal to 250 cells/μl [33].

The steeper increase of HIV testing coverage among women compared to men may be explained by scale-up of programs to prevent mother to child transmission (PMTCT). More frequent testing leads to earlier diagnosis and then earlier initiation of cART in eligible women. The scale-up of PMTCT could thus also account for the greater increases of CD4 cell counts at the start of cART in women than men observed in the present study. A review of national program data for 2004 to 2005 showed that the scale-up of PMTCT had gained momentum in many countries, and that provider-initiated (opt-out) HIV testing had become nearly universal in some regions (for example in East and Southern Africa) but not in others (for example in West Africa) [34]. A systematic review and meta-analysis of 44 studies of pregnant women who attended PMTCT programs in sub-Saharan Africa showed that uptake at antenatal care services was 94% for opt-out testing, compared to 58% for opt-in testing [35]. Coverage with any type of antiretroviral prophylaxis was 70%, and 62% of pregnant women eligible for cART received treatment [35]. Similarly, a meta-analysis of six studies that reported on numbers of adults followed between HIV diagnosis and start of cART showed that of every 100 patients with a positive HIV test, 72 had a CD4 cell count measured, 40 were eligible for cART and 25 started cART. Of note, men were more likely to be lost to program and less likely to start cART than women [36].

Our study has several limitations. We could only examine information up to 2010, because more recent data were not yet available from many sites participating in the IeDEA and ART cohort collaborations. CD4 cell counts at start of cART were missing in approximately one quarter of patients, who were more likely to be from LIC and MIC and in a more advanced stage of disease than patients with measured CD4 cell counts. It is thus likely that our estimates of median CD4 cell counts at cART initiation are biased upwards for these countries. Data from some countries were limited to a small number of patients from a single clinic. We decided to include these in descriptive analyses, but because the data are probably not representative of all patients on cART in those countries, we excluded them from analyses of time trends and predictors of CD4 cell count at start of cART. In sensitivity analyses we weighted individual country data, with the aim of creating a dataset that was representative, within each income group, of the number of patients on cART in each country, as estimated by WHO [5]. Some of the data included in these sensitivity analyses may, however, not be representative of all patients on cART in the country. In particular, the clinics from LIC and MIC participating in IeDEA are mainly urban and capture data in electronic databases, indicating a higher level of resources. They may more closely reflect best practice in urban settings than in the country as a whole. Nevertheless, our study is a unique source of information on trends and determinants of the CD4 cell count in adult patients starting cART across the globe.

In conclusion, our results illustrate the enormous challenges that lie ahead. Despite the massive scale up of cART in LIC, from 300,000 patients on cART in 2002 to 6.6 million by the end of 2010, the increases in median CD4 cell count at the start of cART during this time period have been modest. In 2009, most patients in LIC and MIC countries and many patients in HIC started cART with CD4 cell counts below 200 cells/μl, which means that they were already at high risk of complications and had spent years at high risk of HIV transmission. Substantial efforts and resources are needed to achieve earlier implementation of cART globally. Finally, continued monitoring of the CD4 cell count at which HIV-positive patients start cART in LIC, middle and high income countries is needed to evaluate the results of these efforts.

Figure 2.

Trends of median CD4 cell counts at the start of cART (upper panel) and in proportions of men and women starting cART below 200 cells/μl (middle panel) or below 100 cells/μl (lower panel) in low, middle and high income countries, 2002 to 2010.

The shaded areas represent the 95% confidence intervals for each year. Results from mixed effects linear regression (model 3) based on 379,865 patients, missing values imputed using multiple imputation.

Appendix

Writing committee

Dorita Avila¹, Keri N Althoff², Catrina Mugglin¹, Kara Wools-Kaloustian³, Manuel Koller¹, François Dabis⁴, Denis Nash⁵, Thomas Gsponer¹Somnuek Sungkanuparph⁶, Catherine McGowan⁷, Margaret May⁸, David Cooper⁹, Cleophas Chimbetete¹⁰, Marcelo Wolff¹¹, Ann Collier¹², Hamish McManus⁹, Mary-Ann Davies¹³, Dominique Costagliola¹⁴, Brenda Crabtree-Ramirez¹⁵, Romanee Chaiwarith¹⁶, Angela Cescon¹⁷, Morna Cornell¹³, Lameck Diero¹⁸, Praphan Phanuphak¹⁹, Adrien Sawadogo²⁰, Jochen Ehmer²¹, Serge P. Eholie²², Patrick CK Li²³, Matthew P Fox^24,25, Neel E Ghandi²⁶, Elsa González²⁷, Christopher KC Lee²⁸, Christopher J Hoffmann^29,30, Andrew Kambugum³¹, Olivia Keiser¹, Rossana Ditangco³², Hans Prozesky³³, Fiona Lampe³⁴, Nagalingeswaran Kumarasamy³⁵, Mari Kitahata¹², Emmanuel Lugina³⁶, Rita Lyamuya³⁷, Saphonn Vonthanak³⁸, Valeria Fink³⁹, Antonella d’Arminio Monforte⁴⁰, Paula M Luz⁴¹, Yi-Ming A Chen⁴², Albert Minga⁴³, Jordi Casabona^44,45, Albert Mwango⁴⁶, Jun Y Choi⁴⁷, Marie-Louise Newell⁴⁸, Elizabeth A. Bukusi⁴⁹, Kapella Ngonyani⁵⁰, Tuti P Merati⁵¹, Juliana Otieno⁵², Mwebesa B Bosco⁵³, Sam Phiri⁵⁴, Oon T. Ng⁵⁵, Kathryn Anastos⁵⁶, Jürgen Rockstroh⁵⁷, Ignacio Santos⁵⁸, Shinichi Oka⁵⁹, Geoffrey Somi⁶⁰, Christoph Stephan⁶¹, Ramon Teira⁶², Deo Wabwire⁶³, Gilles Wandeler⁶⁴, Andrew Boulle¹³, Peter Reiss⁶⁵, Robin Wood⁶⁶, Benjamin H Chi⁶⁷, Carolyn Williams⁶⁸, Jonathan A Sterne⁹, Matthias Egger^1,13

1) Institute of Social and Preventive Medicine (ISPM), University of Bern, Switzerland; 2) Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, USA; 3) Indiana University School of Medicine, Indianapolis, IN, USA; 4) Institut de Santé Publique, Epidémiologie et Développement (ISPED), Université Bordeaux Segalen, Bordeaux, France; 5) CUNY School of Public Health, New York, USA; 6) Faculty of Medicine Ramathibodi Hospital, Mahidol University, Bangkok, Thailand; 7) Vanderbilt University School of Medicine, Nashville, TN, USA; 8) School of Social and Community Medicine, University of Bristol, UK; 9) The Kirby Institute, The University of New South Wales, Sydney, Australia; 10) Newlands Clinic, Harare, Zimbabwe; 11) Fundación Arriarán, Universidad de Chile, Santiago, Chile; 12) University of Washington School of Medicine Center, Seattle, USA; 13) Centre for Infectious Disease Epidemiology and Research, School of Public Health and Family Medicine, University of Cape Town, South Africa; 14) 1-UPMC Univ Paris 06, UMR_S 943, F-75013, Paris, France; 15) Instituto Nacional de Ciencias Médicas y Nutrición, Salvador Zubirán, México D.F., México; 16) Research Institute for Health Sciences, Chiang Mai, Thailand; 17) British Columbia Centre for Excellence in HIV/AIDS, Vancouver, Canada; 18) Academic Model Providing Access to Health Care (AMPATH), School of Medicine, Faculty of Health Sciences, Moi University, Eldoret, Kenya; 19) HIV-NAT/Thai Red Cross AIDS Research Centre, Bangkok, Thailand; 20) CHU Sourô Sanou, Bobo-Dioulasso, Burkina Faso; 21) SolidarMed African HIV prevention and treatment program, Lucerne, Switzerland; 22) Service de Maladies Infectieuses et Tropicales (SMIT), CHU de Treichville, Abidjan, Côte d’Ivoire; 23) Queen Elizabeth Hospital, Hong Kong SAR, China; 24) Center for Global Health and Development and Department of Epidemiology, Boston University and Boston University School of Public Health, Boston, USA; 25) Health Economics and Epidemiology Research Office, University of the Witwatersrand, Johannesburg, South Africa; 26) Departments of Epidemiology, Global Health & Infectious Diseases, Emory University Rollins School of Public Health, Atlanta, GA USA; 27) Instituto de Medicina Tropical Alexander von Humboldt, Universidad Peruana Cayetano Heredia, Lima, Peru; 28) Hospital Sungai Buloh, Sungai Buloh, Malaysia; 29) Aurum Institute, Johannesburg, South Africa; 30) John Hopkins University School of Medicine, Baltimore, USA; 31) Infectious Diseases Institute (IDI), Kampala, Uganda; 32)Research Institute for Tropical Medicine, Manila, Philippines; 33) Division of Infectious Diseases, University of Stellenbosch and Tygerberg Hospital, Cape Town, South Africa; 34) Research Department of Infection and Population Health, University College London, UK; 35) YRG Centre for AIDS Research and Education, Chennai, India; 36) Ocean Road Cancer Institute (ORCI), Dar-es-Salaam, Tanzania; 37) Morogoro Regional Hospital, Morogoro, Tanzania; 38) National Center for HIV/AIDS, Dermatology & STDs, Phnom Penh, Cambodia; 39) Fundación Huésped, Buenos Aires, Argentina; 40) Clinic of Infectious Diseases & Tropical Medicine, San Paolo Hospital, University of Milan, Italy; 41) Fundação Oswaldo Cruz, Rio de Janeiro, Brazil; 42) Taipei Veterans General Hospital and AIDS Prevention and Research Centre, National Yang-Ming University, Taipei, Taiwan; 43) Centre Médical de Suivi de Donneurs de Sang/CNTS/PRIMO-CI, Abidjan, Cote d’Ivoire; 44) Centre d’Estudis Epidemiològics sobre ITS/VIH/SIDA de Catalunya, Badalona, Spain; 45) CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain; 46) Zambian Ministry of Health, Lusaka, Zambia; 47) Division of Infectious Diseases, Department of Internal Medicine, Yonsei University College of Medicine, Seoul, South Korea; 48) Africa Centre for Health and Population Studies, University of KwaZulu-Natal, Mtubatuba, South Africa; 49) Family AIDS Care and Education Services (FACES), Kisumu, Kenya; 50) Tumbi Regional Hospital, Kibaha, Tanzania; 51) Faculty of Medicine Udayana University & Sanglah Hospital, Bali, Indonesia; 52) Nyanza Provincial Hospital, MTCT-plus, Kisumu, Kenya; 53) Mbarara University of Science and Technology (MUST), ISS Clinic, Mbarara, Uganda; 54) Lighthouse Trust, Lilongwe, Malawi; 55) Tan Tock Seng Hospital, Singapore; 56) Departments of Medicine and Epidemiology and Population Health, Albert Einstein College of Medicine, New York, USA; 57) Department of Internal Medicine I, University Hospital Bonn, 53105 Bonn, Germany; 58) Hospital La Princesa, Madrid, Spain; 59) National Center for Global Health and Medicine, Tokyo, Japan; 60) National AIDS Control Program, Dar es Salaam, Tanzania; 61) HIVCENTER am Klinikum der Johann Wolfgang Goethe-Universität, Frankfurt, Germany; 62) Hospital Sierrallana, Torrelavega, Spain; 63) Mulago University JHU Collaboration (MUJHU MTCT-plus), Kampala, Uganda; 64) Department of Infectious Diseases, Bern University Hospital and University of Bern, Switzerland; 65) Academisch Medisch Centrum bij de Universiteit van Amsterdam, Amsterdam, The Netherlands; 66) Desmond Tutu HIV Centre, University of Cape Town, South Africa; 67) Center for Infectious Disease Research in Zambia, Lusaka, Zambia; 68) Epidemiology Branch, Division of AIDS, National Institute of Allergy and Infectious Diseases, Bethesda, MD, USA.