CD4+ T cell recovery during suppression of HIV replication: an international comparison of the immunological efficacy of antiretroviral therapy in North America, Asia and Africa

Abstract

Background: Even among HIV-infected patients who fully suppress plasma HIV RNA replication on antiretroviral therapy, genetic (e.g. CCL3L1 copy number), viral (e.g. tropism) and environmental (e.g. chronic exposure to microbial antigens) factors influence CD4 recovery. These factors differ markedly around the world and therefore the expected CD4 recovery during HIV RNA suppression may differ globally.

Methods: We evaluated HIV-infected adults from North America, West Africa, East Africa, Southern Africa and Asia starting non-nucleoside reverse transcriptase inhibitor-based regimens containing efavirenz or nevirapine, who achieved at least one HIV RNA level <500/µl in the first year of therapy and observed CD4 changes during HIV RNA suppression. We used a piecewise linear regression to estimate the influence of region of residence on CD4 recovery, adjusting for socio-demographic and clinical characteristics. We observed 28 217 patients from 105 cohorts over 37 825 person-years.

Results: After adjustment, patients from East Africa showed diminished CD4 recovery as compared with other regions. Three years after antiretroviral therapy initiation, the mean CD4 count for a prototypical patient with a pre-therapy CD4 count of 150/µl was 529/µl [95% confidence interval (CI): 517–541] in North America, 494/µl (95% CI: 429–559) in West Africa, 515/µl (95% CI: 508–522) in Southern Africa, 503/µl (95% CI: 478–528) in Asia and 437/µl (95% CI: 425–449) in East Africa.

Conclusions: CD4 recovery during HIV RNA suppression is diminished in East Africa as compared with other regions of the world, and observed differences are large enough to potentially influence clinical outcomes. Epidemiological analyses on a global scale can identify macroscopic effects unobservable at the clinical, national or individual regional level.

Introduction

Although the first-line non-nucleoside reverse transcriptase inhibitor (NNRTI)-based regimens widely used in resource-limited settings reliably suppress HIV RNA replication in adherent patients, less certainty exists about the equivalence of CD4 + T cell recovery during HIV RNA suppression—an outcome which also has a direct effect on clinical outcomes.¹ First, host genetic factors such as CCL3L1 copy number,² CCR5 and cytokine polymorphisms,^3,4 and mitochondrial DNA⁵ can influence CD4 recovery during HIV RNA suppression. The prevalence of these genetic factors—as well as others not yet identified likely differ in human populations around the world. Second, characteristics of HIV itself, such as tropism, also influence CD4 recovery after HIV RNA suppression.⁶ For example, the proclivity of subtype D for X4 tropism and the preponderance of subtype D in certain regions may influence CD4 recovery in those regions.^7,8 Finally, immunological activation due to environmental exposure to microbial antigens has emerged as a unifying theory that explains both CD4 loss before antiretroviral therapy (ART) initiation and attenuated CD4 rise on suppressive ART.⁹ Implicated organisms include commensal gut flora,¹⁰ hepatitis C virus,¹¹ Mycobacterium tuberculosis,¹² Cryptococcus neoformans,¹³ helminths¹⁴ and herpesviruses such as cytomegalovirus.¹⁵ Populations in resource-limited settings are exposed to a higher burden of many infections starting at a younger age.^16,17

To date, CD4 recovery during ART-mediated HIV RNA suppression—which we call ‘immunological efficacy’—has not been directly compared across geographically disparate populations. Most existing multi-regional studies have included all patients regardless of their virological response.¹⁸ Although such designs provide a picture of population-level effectiveness of ART, they do not distinguish whether differences in CD4 recovery are due to socio-structural and behavioural determinants of medication adherence or to biological factors that act on the immune system. Most analyses which do restrict observation to CD4 recovery during HIV RNA suppression have been carried out within Europe and therefore did not assess wider geographical differences.^1,19 One cross-regional study carried out in the setting of a randomized trial found diminished CD4 recovery among South African patients as compared with Europeans, even though the South Africans had higher rates of HIV RNA suppression.²⁰ However, these intriguing results only included data from one country in Africa (South Africa), in which the socioeconomic setting differs markedly from the rest of sub-Saharan Africa.

In the present analysis, we sought to understand the effect of differences in host, virus and environment on CD4 recovery during ART-mediated HIV RNA suppression across geographically large regions of the world. We analysed patients from five regions participating in the International Epidemiological Database to Evaluate AIDS (IeDEA) Consortium: North America, West Africa, East Africa, Southern Africa and Asia. The size and reach of IeDEA allowed us to identify, and restrict the analysis to, patient populations followed with serial plasma HIV RNA testing in resource-limited settings. Regional differences in immunological efficacy—should they exist—may inform existing and spur additional hypotheses for biological researchers seeking to understand the mechanisms of immunological destruction by HIV and immunological restoration after ART initiation. Differences could also influence assessments about the risks and benefits of ART initiation at different CD4 thresholds. Finally, differences in immunological efficacy may inform public health scientists who seek to assess or model population-level health benefits of the global roll-out of ART.

Methods

Design

We conducted a multi-site cohort analysis using data collected from five regions in the IeDEA Consortium. Our objective was to assess the effect of region (with attendant but unmeasured differences in human genetics, HIV subtype and environmental exposures) on CD4 count recovery during ART-mediated suppression of plasma HIV RNA. We adjusted the effect of region on CD4 count recovery for factors already known to influence CD4 count recovery, such as sex, age, pre-therapy CD4 count, pre-therapy HIV RNA level, and ART regimen composition (e.g. the use of zidovudine). We used a directed acyclic graph (Figure 1) to formally express the research question in a causal framework²¹ and to guide decisions about adjustment.²²

Figure 1.

Directed acyclic graph depicting research question. We seek to estimate the direct effect of region on CD4 + T cell recovery during antiretroviral-mediated HIV RNA suppression, apart from other previously established mediators. Restriction, in the study design, to time during HIV RNA suppression closes pathways mediated by medication access and adherence. Statistical adjustment closes pathways mediated by factors known to influence CD4 + T cell recovery such as age and sex. We interpret any remaining association between region and CD4 + T cell recovery to represent the effect of additional regional host, viral and environmental factors that act through biological mechanisms.

Patients

IeDEA is an NIH-funded research consortium which pools data from geographically dispersed cohorts to address macroscopic epidemiological questions regarding the HIV epidemic. Cohorts from five regions participated in this analysis: North America (USA and Canada), West Africa (Ivory Coast, Senegal and The Gambia), East Africa (Uganda), Southern Africa (South Africa and Malawi) and Asia (China, Thailand, Vietnam, Indonesia,and Japan). Institutional review boards in the respective IeDEA regions approved the study. Within each region, we included all ART-naive adults (adults defined as age >17 years) who initiated NNRTI-based ART, had a pre-therapy CD4 count of ≤350 cells/µl, were monitored with routine HIV RNA testing, and achieved at least one plasma HIV RNA level <500 copies/ml in the first 48 weeks after initiation of ART. Observation began at ART initiation and continued as long as patients remained on an NNRTI-based regimen with a plasma HIV RNA of <1000 copies/ml. Observations were censored at the last HIV RNA level <1000 copies/ml that occurred before the first of any of the following: (i) HIV RNA rebound to ≥1000 copies/ml; (ii) switch to a protease inhibitor or other non-NNRTI-based regimen; (iii) an interval of 9 months without an HIV RNA determination (to minimize the risk of incorporating CD4 values obtained during unascertained HIV viraemia); (iv) death; (v) loss to follow-up (defined as 6 months without a clinic visit); or (vi) database closure.

Measurements

Socio-demographic and clinical variables such as sex, age, ART regimen and the dates of clinical events (e.g. ART initiation, follow-up visits etc.) were collected in the respective IeDEA-associated cohorts during the course of routine care and in research-based cohorts through protocols. CD4 counts and plasma HIV RNA levels were obtained from the respective clinical or research laboratories associated with the various sites.

Analysis

First, we conducted a non-parametric comparison of region-specific median CD4 counts (and interquartile ranges) before ART and at 4, 12, 24 and 36 months after ART initiation. For patients with multiple CD4 count determinations within a 90-day window of the target date, we used the median value.

Second, we used a mixed-effects regression model to estimate the effect of region on mean CD4 count values after initiation of ART, adjusting for age, sex, pre-therapy CD4 count, pre-therapy plasma HIV RNA level and initial ART regimen composition. Random intercepts and slopes, with unstructured covariance, were used to accommodate patient-specific departures from the trajectories determined by the fixed effects. To best conform to the model assumptions of normality and homoscedasticity, we used the square root transformation of the on-therapy CD4 values. We subsequently back-transformed estimates into the native CD4 scale.^23,24 The pre-therapy CD4 value was treated as a predictor and not included as an outcome.²⁵ This analysis was performed using full information maximum likelihood estimation in M-plus version 6 (Muthén & Muthén, Los Angeles, CA), which provides consistent estimates under the assumption that missing predictor data were missing at random.²⁶

Time was specified as a linear spline with knots at 4 months and 1 year to account for the different rates of CD4 count change known to occur in these periods.^27–30 We explored the significance of an additional knot at month 24. Interaction terms between region and time as well as pre-therapy CD4 and time were used to accommodate the influence of these factors on changes in CD4 slope, resulting in a distinct piecewise linear trajectory for each region and pre-therapy CD4 count. Pre-therapy CD4 count was specified as a restricted cubic spline with three knots, to accommodate non-linear associations between pre-therapy CD4 counts and CD4 counts over time. Two-way interactions between time and all other pre-therapy patient characteristics (e.g. age, sex) were also included to allow adjustment for these factors over time. We also explored a three-way interaction term between pre-therapy CD4 cell count, region and time, motivated by the hypothesis that regional differences in CD4 slope may differ by pre-therapy CD4 cell count. Our final model was chosen based on examination of Akaike and Bayesian information criteria.

Age was treated as a continuous variable and pre-therapy HIV RNA level was categorized according to convention at log₁₀ values of <4.0, 4.0–4.5, 4.51–5.0, 5.1–5.5 and >5.5 copies/ml.³¹ The presence or absence of a particular antiretroviral medication in the initial regimen was classified using an indicator variable for the medication in question. For illustrative purposes, we estimated mean CD4 counts at 4 months, 1 year and 3 years after ART initiation for patients with pre-therapy CD4 counts of 25/µl, 50/µl, 100/µl, 150/µl, 200/µl and 300/µl in each region. CD4 recovery was estimated for a ‘prototypical’ patient, defined as having the mean value in the entire study population for age, sex, pre-therapy HIV RNA and composition of ART regimen. At each of the three follow-up time points and six pre-therapy CD4 count combinations, we generated between-region differences in mean CD4 count for a total of 180 contrasts.

Results

In total, 28 217 patients from 105 cohorts contributed a total of 37 825 patient-years of observation (Table 1). In the entire cohort, the median age at ART initiation was 36 years (IQR: 30–42), 59% were women and the median pre-therapy CD4 value was 116 cells/µl (IQR: 50–181). Patient characteristics differed across regions. Women, for example, comprised only 20% of the patients from North America but approximately 70% in the African regions. Each patient contributed a median of 12.7 months (IQR: 6.94 –24.0) of follow-up during HIV RNA suppression. During follow-up, the median number of CD4 determinations made was 3 (IQR: 2–5) and median time between CD4 determinations was 114 days (IQR: 84–168). The most common reason for censoring was database closure (i.e. administrative) in 41% of patients. Loss to follow-up occurred in 15% of patients. The reasons for censoring differed across regions: for example, observation end due to 9 months without an HIV RNA determination was present in 46% of persons in Asia compared with 13% in North America.

Table 1.

Patient characteristics

Characteristics	North America N = 4450	West Africa N = 649	East Africa N = 2,202	Southern Africa N = 20,323	Asia N = 593	Total N = 28,217
Pre-therapy
Clinic sites, n	15	4	19	52	15	105
Age in years, median (IQR)	40 (34–46)	37 (31–44)	35 (30–41)	35 (30–41)	36 (31–42)	36 (30–42)
Female, n (%)	881 (20)	456 (70)	1451 (66)	13597 (67)	187 (32)	16572 (59)
CD4 + T cells/µl, median (IQR)	168 (62–250)	151 (75–223)	149 (71–210)	105 (48–165)	124 (34–200)	116 (50–181)
Plasma HIV RNA log10 copies/ml, median (IQR)a	4.9 (4.5–5.3)	5.4 (4.9–5.8)	5.12 (4.6–5.6)	4.79 (4.2–5.3)	5.08 (4.6–5.6)	4.9 (4.3–5.4)
Zidovudine in initial regimen, n (%)b	2415 (54)	198 (31)	1458 (66)	2972 (15)	115 (19)	7158 (26)
Nevirapine in initial regimen, n (%)c	841 (19)	415 (64)	646 (80)	6158 (30)	420 (71)	8480 (32)
ART initiation date, median (IQR and range)	2-Feb-03 (9-Mar-01 to 14-Apr-05; 15-Sep-96 to 5-May-09)	21-Jun-06 (10-Feb-06 to 3-Nov-06; 11-Jan-00 to 12-Apr-07)	13-Mar-05 (20-Feb-2005 to 8-May-08; 19-Apr-04 to 20-Nov-09	16-Mar-05 (2-Feb-05 to 27-Mar-07; 08-Apr-08 to 31-Aug-09)	5-May-04 (17-Feb-03 to 29-Jun-06; 12-Aug-99 to 29-Sep-08)	9-Jan-06 (19-Aug-04 to 2-Mar-07; 15-Sep-96 to 20-Nov-09)
Database closure date	03-Sep-09	27-Nov-07	05-Oct-09	15-Dec-09	23-Jun-09
Follow-up
Duration months, median (IQR)	31.7 (18.8–33.9)	12.1 (8.3–18.0)	33.2 (22.2–33.2)	20.9 (11.7–30.9)	26.0 (17.7–31.5)	12.7 (6.9–24.0)
Time between CD4 counts days, median (IQR)	98 (73–121)	175 (96–185)	112 (84–168)	134 (102–181)	89.5 (56–147)	114 (84–168)
CD4 determinations median (IQR)	5 (2–9)	2 (1–3)	6 (3–8)	2 (1–4)	4 (2–7)	3 (2–5)
Time between HIV RNA determinations days, median (IQR)	98 (71–119)	176 (128–184)	168 (85–168)	139 (110–182)	82 (56–160)	117 (87–168)
HIV RNA determinations, median (IQR)	4 (2–8)	1 (1–2)	4 (2–6)	2 (1–4)	3 (1–6)	2 (1–5)
Reasons for end of observation
Administrative censor, n (%)	2048 (46)	440 (68)	1764 (80)	7198 (35)	219 (37)	11669 (41)
Nine-month gap without HIV RNA determination, n (%)	572 (13)	70 (11)	102 (5)	6802 (34)	271 (46)	7817 (28)
Loss to follow-up, n (%)	544 (12)	65 (10)	66 (3)	3561 (18)	25 (4)	4261 (15)
Switch or stop of ART, n (%)	333 (8)	3 (1)	9 (0.4)	443 (2)	5 (1)	793 (3)
HIV RNA rebound, n (%)	889 (20)	59 (9)	221 (10)	1947 (10)	71 (12)	3187 (11)
Death, n (%)	64 (1)	12 (2)	40 (2)	372 (2)	2 (0.3)	527 (2)

The unadjusted distribution of CD4 counts across regions demonstrated lower values in East Africa at most time points, and this difference was apparent in each of the pre-therapy CD4 categories (Figure 2). For example, after ART initiation for patients from East Africa with a pre-therapy CD4 value of ≤50/µl, the median CD4 count during HIV RNA suppression was 111/µl (IQR: 72–158) at 4 months, 191/µl (IQR: 139–265) at 1 year and 240/µl (IQR: 199–350) at 2 years. For a patient from Southern Africa with the same pre-therapy CD4 count of ≤50/µl, the median CD4 counts were 120/µl (IQR: 75–180) at 4 months, 219/µl (IQR: 158–301) at 1 year and 344/µl (IQR: 251–462) at 2 years. Similar trends were observed for patients who started ART at higher CD4 levels.

Figure 2.

CD4 + T cell counts before and after antiretroviral therapy-mediated HIV RNA suppression. Distribution of CD4 + T cell values for given pre-therapy CD4 + T cell count strata at 4, 12 and 24 months after antiretroviral therapy initiation, stratified by region. These values are not adjusted for other patient characteristics.

Results of the mixed-effects, piecewise linear regression of expected CD4 counts during ART-mediated suppression of HIV RNA showed that after adjustment for age, sex, composition of initial regimen, pre-therapy CD4 count and pre-therapy HIV RNA level, CD4 recovery was lower in East Africa as compared with the other regions across all pre-therapy CD4 levels (Figure 3). For example, for prototypical patients with 150 CD4 cells/µl at ART initiation, the mean CD4 value 3 years after ART initiation was 437/µl (95% CI: 425–449) in East Africa whereas it was 529/µl (95% CI: 517–541) in North America, 494/µl (95% CI: 429–559) in Western Africa, 515/µl (95% CI: 508–522) in Southern Africa and 503/µl (95% CI: 478–528) in Asia (Table 2).

Figure 3.

Model estimated CD4 + T cell counts during antiretroviral therapy-mediated HIV RNA suppression. Mean CD4 + T cell counts for prototypical patients by region and pre-therapy CD4 + T cell count, at 4, 12 and 36 months after antiretroviral therapy initiation. Estimates were derived from piecewise mixed effects linear regression with knots at 4 and 12 months. Estimates are shown for a patient with mean values in the entire population for age, sex, medication regimen composition and pre-therapy HIV RNA level.

Table 2.

Average CD4 + T cell counts for patients for given pre-therapy CD4 levels. Values are shown in each region at 4, 12 and 36 months after antiretroviral therapy initiation and during HIV RNA suppression. Estimates were obtained from a piecewise mixed-effects linear regression with knots at 4 and 12 months. Values are adjusted for sex, age, composition of initial regimen, pre-therapy CD4 level and pre-therapy HIV RNA level

We also expressed CD4 change as differences between mean values in each region for each given pre-therapy CD4 count at 4, 12 and 36 months after ART initiation (Table 3). Significant differences in CD4 recovery between East Africa and the other regions tended to occur after 4 months of therapy, and the observed differences tended to grow over time. For example, for patients with pre-therapy CD4 level of 150/µl, 4 months after ART initiation the mean CD4 level in North America was 13 cells/µl higher (95% CI: 4–22) than in East Africa. The difference rose to 51 cells/µl (95% CI: 41–61) at 12 months and 92 cells/µl (95% CI: 77–108) 36 months after ART initiation.

Table 3.

Cross-regional differences in expected CD4 levels between patients. Values are shown for patients with given pre-therapy CD4 + T cell counts at 4, 12 and 36 months after ART initiation. Patients are assumed to have population-averaged characteristics of sex, age, composition of initial regimen and pre-therapy plasma HIV RNA levels. The value in each cell represents the estimated difference in CD4 recovery between an individual in the region indicated in the column and the region indicated in the row. For example the first value of 6 in the top left cell indicates that the average CD4 rise in North America is 6 cells higher than in West Africa 4 months after ART initiation for patients who initiate antiretroviral therapy with a CD4 count of 25 cells/µl

Diff., difference; NA, North America; WA, West Africa, EA, East Africa; SA, South Africa; AS, Asia.

Addition of a knot in the piecewise regression at 24 months after ART initiation did not change results substantially. A comparison of each of the three sites from East Africa showed that all sites demonstrated diminished CD4 recovery as compared with other regions and therefore that observed differences were not driven by one site (data not shown).

Discussion

This study describes across geographical regions, for the first time, large and—especially in the case of East Africa as compared with other regions—potentially clinically meaningful differences in CD4 + T cell recovery among HIV-infected patients during ART-mediated suppression of plasma HIV RNA. These estimates may also inform modelling exercises to understand potential effects of ART treatment such as the UNAIDS investment framework and others. Patients from East Africa demonstrated the lowest CD4 recovery after adjustment for factors known to influence CD4 rise, including age, sex and pre-therapy CD4 count and plasma HIV RNA. The attenuated CD4 recovery in East Africa was consistent across a range of pre-therapy CD4 counts. Differences were not clearly apparent during the first 4 months of therapy, a period in which changes mainly result from the redistribution of CD4 cells from lymphoid tissues,- but rather became increasingly pronounced during subsequent periods when de novo T cell production drives CD4 count increases in the peripheral blood.

Our findings extend existing analyses comparing CD4 recovery among different patient populations globally. Although a previous cross-regional comparison found that patients in high- and low-income countries experienced similar rates of CD4 recovery on ART, the comparison was limited to 6 months of observation after ART initiation¹⁸ whereas we observed patients for up to 3 years. Previous work in The Netherlands that did follow patients for longer periods—for up to 5 years of HIV RNA suppression—found CD4 increases among patients who originated from sub-Saharan Africa to be on average 40 cells/µl lower than among patients originating from Europe or North America.¹⁹ By including populations currently living in Africa, however, our analysis was able to capture the effects of ongoing environmental factors as well as genetic or past environmental exposures, which may explain the greater differences we observed.

Chronic exposure to microbial antigens may explain, at least in part, the observed differences in ‘immunological efficacy’ of ART. Population-based surveys have found that the distribution of CD4 levels in HIV-uninfected people in East Africa does not differ from other regions.³² This implies that the diminished CD4 recovery in East Africa cannot be explained by different normal CD4 ‘set points’, but rather represents poor recovery per se. Although this blunted recovery is consistent with a number of different hypotheses, the emerging paradigm—that immunological activation drives CD4 cell depletion during untreated HIV disease¹⁰ as well as attenuating CD4 recovery after HIV replication is controlled by ART⁹—offers an intriguing potential explanation. Commensal microbial antigens from the gut as well as a variety of infections such as tuberculosis,³³ malaria, schistosomiasis¹⁴ and herpesviruses such as cytomegalovirus¹⁵ have all been implicated as causes of immunological activation. East Africans experience a higher occurrence of most of these infections from a younger age than North Americans.

Exposure to microbial antigens, however, does not fully explain diminished CD4 recovery in East Africa because residents of Southern and especially West Africa experience a similar prevalence of infections but exhibited more robust CD4 recovery. Viral factors, therefore, might play a role. HIV subtype D, which is rare in other regions of the world—including in other areas of Africa—accounts for 30–50% of HIV infections in some areas of East Africa.^34,37 Subtype D has a greater predilection for X4 tropism⁷ and X4 tropism, in turn, has been associated with suboptimal CD4 recovery after HIV RNA suppression.⁶ Nutrition may also play a role. A randomized trial in the USA suggested that protein intake was associated with better CD4 recovery, and populations in rural East Africa may have particularly low protein intake.³⁸ The effect of trimethoprim-sulfamethoxazole—a medication that can impair haematopoiesis—may be magnified in populations with low protein intake. Host genetic characteristics unique to East Africa may also contribute to diminished capacity for CD4 + T cell restoration.^4,5 Further studies to elucidate the mechanisms that determine the differences we observed between regions—which may be multifactorial—are needed.

The diminished CD4 recovery in East Africa is large enough to potentially influence health outcomes. A large cohort analysis from the Collaboration of Observational HIV Epidemiological ResearchEurope (COHERE) found reductions in AIDS progression and mortality: by 65% per rise of 100 CD4 cells/µl among patients with a most recent CD4 level <200/µl; by approximately 20% for patients with CD4 levels from 201/µl to 500/µl; and by 4% for those with a CD4 level >500/µl.¹ Given our observation of an approximately 100-cell/µl difference between East Africa and North America at 3 years following ART initiation, the results of COHERE suggest that the poorer CD4 recovery in East Africa is likely clinically meaningful across a range of CD4 counts. Inpatient health systems in Africa, furthermore, often lack sophisticated and expensive diagnostic and treatment options available in Europe, thus magnifying potential consequences of attenuated CD4 recovery in East Africa. Finally, diminished CD4 recovery after ART initiation in East Africa—and therefore a longer period of immunosuppression—implies that the benefits of initiating ART at higher CD4 levels may be greater in East Africa than in other regions of the world.

Several limitations were present in this study. First, we did not conduct the CD4 measurements ourselves in a centralized laboratory, and therefore regional differences in assay performance cannot be addressed. In East Africa, however, CD4 testing was carried out in three administratively and geographically distinct laboratories, thus making laboratory artefact unlikely. Second, losses to follow-up occurred in 15% of patients and raise concerns about selection bias. Our analysis, however, restricted observation to time during HIV RNA suppression, and this strategy provides stronger protection against bias due to informative censoring. Also, conditions such as tuberculosis that could cause both losses to follow-up as well as diminished CD4 recovery during HIV RNA suppression—and therefore informative censoring—are more likely to occur in East Africa than in the USA and would lead to underestimates of the differences observed. In addition, the longer interval between HIV RNA determinations in East Africa as compared with North America may have allowed longer periods of undetected rebound HIV RNA viraemia and therefore biased estimates of CD4 recovery during HIV RNA suppression downward in East Africa. The interval between HIV RNA measurements, however, was also relatively long in West and Southern Africa, yet those regions exhibited more robust CD4 responses as compared with East Africa. Therefore, we believe that artefact due to interval of CD4 measurements is an unlikely explanation for our findings. In North America, restriction of the cohort to patients starting NNRTI-based regimens implies that many patients (e.g. those starting protease inhibitor-based regimens) were excluded, thus introducing the possibility they were unlike other patients who were included in unmeasured ways. A final limitation is that all our sites in East Africa were located in Uganda and we lacked data from Kenya or Tanzania, thus potentially compromising the generalizability of the findings.

In summary, we found notable regional differences in CD4 count recovery on suppressive ART, with patients in East Africa experiencing a clinically significant attenuation as compared with other regions. We speculate that the differences in the ‘immunological efficacy’ of ART globally may be explained in part by the notion that HIV pathogenesis both before ART and after ART is driven by microbial exposure and immunological activation. Epidemiological analyses carried out on a global level, made possible for the first time by large research consortia such as IeDEA, can identify macroscopic effects otherwise unobservable at the clinical, national or individual regional level.

Funding

Funding has been provided by the National Institutes of Health [K23 AI084544, U01 AI069918, U01 AI069919, U01 AI069924, U01 AI069911, U01 AI069907, R01 MH054907 and P30 AI027763] and the Division of Intramural Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health.

Conflict of interest: None declared.

Key Messages.

• Globally, CD4 + T-cell recovery during suppressive antiretroviral therapy is not uniform.

• Patients in East Africa exhibit the most blunted CD4 recovery and differences were large enough to potentially influence clinical outcomes.

• We speculate that differences in the ‘immunological efficacy’ of antiretroviral therapy may be explained in part by differences in microbial exposure and immunological activation, as well as other viral and host factors.