Immunodeficiency at Antiretroviral Therapy Start: Five-Year Adult Data (2012–2017) Based on Evolving National Policies in Rural Mozambique

Abstract

Before the 2015 implementation of “Test and Start,” the initiation of combination antiretroviral therapy (ART) was guided by specific CD4 cell count thresholds. As scale-up efforts progress, the prevalence of advanced HIV disease at ART initiation is expected to decline. We analyzed the temporal trends in the median CD4 cell counts among adults initiating ART and described factors associated with initiating ART with severe immunodeficiency in Zambézia Province, Mozambique. We included all HIV-positive, treatment-naive adults (age ≥ 15 years) who initiated ART at a Friends in Global Health (FGH)-supported health facility between September 2012 and September 2017. Quantile regression and multivariable logistic regression models were applied to ascertain the median change in CD4 cell count and odds of initiating ART with severe immunodeficiency, respectively. A total of 68,332 patients were included in the analyses. The median change in CD4 cell count under “Test and Start” was higher at +68 cells/mm³ (95% CI: 57.5–78.4) compared with older policies. Younger age and female sex (particularly those pregnant/lactating) were associated with higher median CD4 cell counts at ART initiation. Male sex, advanced age, WHO Stage 4 disease, and referrals to the health facility through inpatient provider-initiated testing and counseling (PITC) were associated with higher odds of initiating ART with severe immunodeficiency. Although there were reassuring trends in increasing median CD4 cell counts with ART initiation, ongoing efforts are needed that target universal HIV testing to ensure the early initiation of ART in men and older patients.

Introduction

Viral suppression through the initiation of combination antiretroviral therapy (ART) is an integral component of the management of persons living with human immunodeficiency virus (HIV). Historically, the benefits of ART initiation in asymptomatic persons with HIV and CD4 cell counts greater than 500 cells per cubic millimeter (cells/mm³) were not as clear, because there was a paucity of data to guide treatment recommendations in such individuals.^1–3 However, in 2015, results from two randomized controlled trials demonstrated clinical benefits of ART initiation regardless of CD4 cell count.^1,4

The “Strategic Timing of Antiretroviral Treatment (START)” trial, in which ∼50% of included sites were in sub-Saharan Africa, 4,685 patients with HIV and CD4 cell counts greater than 500 cells/mm³ were randomized to two groups: immediate ART initiation versus deferred ART initiation until their CD4 cell count had decreased to less than 350 cells/mm³. Serious morbidity and mortality (AIDS and non-AIDS related) were 57% less likely to occur in persons randomized to the immediate ART initiation group compared with persons in the deferred ART group. In addition, there was no significant difference in the occurrence of serious ART medication-related toxicities between the two groups.¹ Similar findings were revealed in the “Early Antiretroviral Treatment and/or Early Isoniazid Prophylaxis Against Tuberculosis in HIV-infected Adults” (ANRS 12136 TEMPRANO) trial conducted in West Africa whereby the immediate initiation of ART decreased the risks of severe HIV-related illness and death.⁴

Subsequently, the World Health Organization (WHO) updated its ART guidelines in 2015 with recommendations to treat all persons with HIV as soon as possible after diagnosis, regardless of immune status.⁵ This strategy, commonly referred to as “Test and Treat” or “Test and Start,” has the potential for significant public health benefits, because the early initiation of ART has been shown to decrease the rates of HIV transmission, as encapsulated by the mantra that “treatment is prevention.”⁶ In addition, this “Treat All” approach was anticipated to significantly accelerate progress toward achieving the 90-90-90 goals set by the Joint United Nations Program on HIV/AIDS (UNAIDS) by 2020.^7,8 Several sub-Saharan African countries, including Mozambique, began implementing the “Test and Start” strategy in 2016.

Promising trends have emerged as the CD4 cell count thresholds for ART initiation have evolved. In Mozambique, the prevalence of advanced disease at ART initiation declined from 73% to 37% between 2004 and 2014.^9,10 Median CD4 cell counts at the start of ART initiation increased from 2002 to 2015 across various country income designations.^11,12 As scale-up efforts progress, the prevalence of advanced disease at ART initiation is expected to continue to decline.

Vanderbilt University Medical Center (VUMC), through its wholly owned subsidiary Friends in Global Health (FGH), has provided technical assistance in Zambézia Province, Mozambique since 2006 through support from the U.S. Government Centers for Disease Control and Prevention (CDC)/President's Emergency Plan for AIDS Relief (PEPFAR). Zambézia is the country's second most populous province with ∼5.1 million persons as of 2017 and has an HIV prevalence (2015) of 15.1%.^34,35 During the time of this evaluation, VUMC/FGH provided technical assistance to 15 of Zambézia's 22 districts and supported more than 140 health facilities that provided comprehensive HIV services, including the provision of ART to more than 100,000 persons. The key programmatic areas of VUMC/FGH's technical assistance initiative entitled “Avante: Towards Epidemic Control” cover: (i) HIV prevention (including elimination of mother to child transmission [EMTCT]), (ii) adult care and treatment, (iii) HIV/tuberculosis (TB) coinfection, and (iv) pediatric care and treatment. Of note, the “Test and Start” strategy commenced in Zambézia Province in a phased approach, starting first in the VUMC/FGH-supported provincial capital of Quelimane in August 2016, followed by select larger urban districts, which included the VUMC/FGH-supported district of Namacurra in April 2017. Of note, “Test and Start” has now been implemented in all VUMC/FGH-supported districts as of November 2017.

The objectives of this study were to analyze temporal trends in the median CD4 cell counts of adults (≥ 15 years of age) at the start of ART and to identify risk factors associated with initiating ART with severe immunodeficiency.

Methods

Data sources

An electronic database, Open Medical Record System (OpenMRS)™, is utilized at VUMC/FGH-supported health facilities for routine patient monitoring. Aggregated, deidentified data on patient demographics, clinical information, and ART pick-up were extracted for secondary analysis following approval by local Ethics Committees and Institutional Review Boards. This study was approved by the Vanderbilt University Institutional Review Board (160549); the Institutional Research Ethics Committee for Health of Zambézia Province (Comité Institucional de Bioética para Saúde-Zambézia) (02-CIBS-Z-16); the Provincial Directorate of Health of Zambézia Province (Direcção Provincial de Saúde Zambézia), and the Centers for Disease Control and Prevention of Mozambique (2016-163/163a).

This analysis includes a review of programmatic data through September 20, 2017; therefore, it only includes preliminary “Test and Start” data for two VUMC/FGH-supported districts, namely Quelimane and Namacurra, as “Test and Start” was not fully implemented in all supported health facilities until after this September 20, 2017 data analysis cutoff date.

Study settings

VUMC/FGH-supported health facilities in 15 districts in Zambézia Province were included in this evaluation. Each district-level health system consists of one large central health facility in the district capital and smaller peripheral health facilities. Quelimane is a substantially larger district; its health facilities were characterized as ‘urban large’ (>2,000 active patients on ART) and ‘urban other’ (<2,000 active patients on ART). Each VUMC/FGH-supported health facility offers varying levels of comprehensive services, including clinical care, laboratory testing, and pharmacy services. Referrals to the HIV treatment facilities originate from several sources, including voluntary counseling and testing (VCT), provider-initiated testing and counseling (PITC), antenatal care (ANC) clinic, youth clinics known as Serviço Amigo do Adolescente e Jovem (SAAJ), and the National Tuberculosis Control Program (PNCT).

Inclusion criteria and definitions

We included all HIV-positive adults (≥15 years of age) who initiated ART at a VUMC/FGH-supported health facility between September 30, 2012 and September 20, 2017. Patients were treatment naive and had at least one CD4 cell count value documented within 6 months before ART initiation or up to 2 months following ART initiation. ART initiation date was defined as the first ART pick-up at a health facility.

ART initiation policies were divided into three categories to reflect the changes to national guidelines during the study period. Policy I was defined as ART initiation among persons having a CD4 cell count <350 cells/mm³ and/or WHO clinical stage 3 or 4 disease and was implemented between September 30, 2012 and February 29, 2016 in all the districts. Policy II was defined as ART initiation among persons having a CD4 cell count <500 cells/mm³ and/or WHO clinical stage 3 or 4 disease and was first implemented in all the districts on March 1, 2016. The proportions of persons with WHO clinical stage 3 or 4 disease that had not qualified by CD4 criteria per Policy period were 35.5% (4,152/11,698) in Policy I and 26.1% (1,744/6,690) in Policy II. Policy III was defined as “Test and Start” and was implemented in the districts of Quelimane and Namacurra on August 29, 2016 and April 11, 2017, respectively. The remaining 13 supported districts had not implemented policy III and remained on Policy II by the end of the study period and were therefore combined into one district group.

Descriptive analyses

Data were aggregated and summarized based on the following characteristics: policy change, sex (including a distinct sex group for pregnant/lactating women), age, program year, type of health facility, referral type, district group, civil status, level of education, sexual orientation, WHO clinical stage, and presence of coinfection with TB. The median CD4 cell counts and interquartile ranges for each of the baseline characteristics were summarized. The Kruskal–Wallis test was used to detect distribution differences between comparison groups.¹³ Missing data from the variables WHO clinical stage, educational level, civil status, and sexual orientation were categorized as “not available” and included in the analyses.

Quantile regression of the median change in CD4 cell count

We applied a multivariate quantile regression to model the median change on the CD4 cell count. In comparison to conventional linear regression, which models the conditional mean, the quantile regression models the conditional quantile. In this study, the median of the CD4 cell count was conditioned on risk factors such as policy change and patient characteristics. The coefficients of the model were reported as the adjusted measurements of the risk effect on the median CD4 cell count and the associated confidence intervals were estimated as asymptotic confidence intervals with independent and identically distributed errors.¹⁴

There were 11 covariates included in the model, which were specified by the investigators before the analysis. The covariates were variables obtained from the OpenMRS database and have been widely used in previous studies.¹⁵ They included: policy change, the year patient-initiated ART, age group, sex group, type of health facility, civil status, level of education, WHO clinical stage, referral type, sexual orientation, and the presence of TB coinfection. The variable, year of ART initiation, was a confounder for the policy change and was purposefully included in the model as a continuous variable, adjusting for programmatic quality improvements implemented as the program evolved to derive the most conservative conclusions. Being an explanatory model, all available covariates were included in the model to identify potential risk factors while also controlling for confounding factors. No variable selection process was implemented in the analysis.

There were some data fields that were not available and were considered missing in our analysis, including five covariates with percentages of missing values: civil status (24.9%), educational level (18.0%), WHO clinical stage (0.7%), referral type (2.5%), and sexual orientation (39.1%). Multiple imputation was implemented in the analysis on all records, including those with missing values on some variables, and the combined results were reported. Compared with a complete case analysis in which subjects with incomplete data are excluded, imputation allows us to include all subjects. In addition, multiple imputation accounts for the variability due to unknown values by averaging over multiple parameter estimates on imputed data. For favorable reproducible results, it is suggested that 100f imputations should be chosen for the imputation analysis, where f is the fraction of the cases having incomplete data.¹⁶ There were 36 imputed datasets generated for the analysis accounting for the maximum of 39.1% missing values on the covariates, also accounting for the efficient parallel computation using six central processing unit (CPU) cores simultaneously. After imputation, the model parameters and associated variances were combined using Rubin's rule.¹⁷ As a sensitivity analysis, a complete case analysis was done on 28,614 subjects with all covariates available. Another sensitivity analysis was done on all subjects with missing values categorized as a separate group.

Multivariable logistic regression for ART initiation with severe immunodeficiency

Severe immunodeficiency defined as having a CD4 cell count <200 cells/mm³ was converted into a binary variable.¹⁸ Similar to the quantile regression, the 11 covariates were included in the logistic regression model to determine their effects on initiating ART with severe immunodeficiency. The results were combined from the same 36 imputed datasets. Similar sensitivity analyses were performed on the complete case dataset and another on the full case dataset with missing values categorized as a separate group.

The data were analyzed using Stata 14.0 (Stata Corporation, College Station, TX, USA) and R (R Core Team 2018).¹⁹ The multiple imputation was implemented using R “mi” package and quantile regression was implemented using R package “quantreg.”^20,21

Results

Descriptive analyses

Data were obtained on 275,608 patients. We excluded 207,276 patients who did not meet the inclusion criteria of being treatment naive and having at least one CD4 cell count value documented within 6 months before ART initiation or up to 2 months following ART initiation, and included 68,332 patients in our analyses (Fig. 1). The median age of patients was 29.1 years [interquartile range (IQR) 23.2–36.2], and the majority of patients (61.9%) were in the 25–49-year age group. Of those included in the study, 47,323 (69.3%) were female, of which 29,108 (42.6% of included females) were neither pregnant nor lactating at ART initiation. TB coinfection was observed in 2,674 (3.9%) of patients. Less than 1% of the patients had attained a university level of education compared with 65.5% having either no formal education or only primary school education. Rapid program expansion accounted for 21,769 (31.9%) patients being included in the 2017 programmatic year. Most patients received care at peripheral health facilities compared with other facility types and were referred through VCT (44.4%) (Table 1).

FIG. 1.

Patient selection for inclusion in the analyses.

Table 1.

Baseline Characteristics and Median CD4 Cell Counts of Patients at Antiretroviral Therapy Initiation

Characteristics	Number of patients (%) N = 68,332	CD4 count median in cells/mm3 [IQR]	p-value
Age (year)	Median [IQR]: 29.1 [23.2–36.2]
Age Group			<0.0001
15–24	21,996 (32.2)	407 [254–598]
25–49	42,317 (61.9)	313 [177–479]
50+	4,019 (5.9)	281 [161–422]
Gender group			<0.0001
Male	21,009 (30.7)	271 [142–421]
Female (nonpregnant/nonlactating)	29,108 (42.6)	320 [185–489]
Female (pregnant/lactating)	18,215 (26.7)	456 [310–632]
TB Coinfection			<0.0001
No	65,658 (96.1)	339 [198–517]
Yes	2,674 (3.9)	295 [144–471]
Policy Definition			<0.0001
I	34,822 (51.0)	307 [175–490]
II	25,683 (37.6)	376 [227–527]
III	7,827 (11.4)	377 [208–568]
WHO Clinical Stage			<0.0001
Stage 1	34,162 (50.0)	387 [252–567]
Stage 2	13,714 (20.1)	298 [174–446]
Stage 3	17,802 (26.1)	282 [140–469]
Stage 4	2,195 (3.2)	233 [100–418]
Not available	459 (0.7)	356 [186–541]
Education level			<0.0001
No formal education	13,434 (19.7)	348 [205–532]
Primary school	31,294 (45.8)	342 [199–520]
Secondary school	11,082 (16.2)	326 [184–492]
University	226 (0.3)	261 [141–449]
Not available	12,296 (18.0)	328 [192–508]
Avante Programmatic Year			<0.0001
2013	7,020 (10.3)	277 [155–427]
2014	10,796 (15.8)	319 [183–509]
2015	11,162 (16.3)	311 [181–491]
2016	17,585 (25.7)	354 [208–513]
2017	21,769 (31.9)	377 [221–547]
Facility type			<0.0001
District capital	26,179 (38.3)	333 [193–506]
Peripheral health facility	29,294 (42.9)	349 [210–527]
Urban large	7,226 (10.6)	296 [153–467]
Urban other	5,633 (8.2)	356 [199–555]
Referral type			<0.0001
VCT	30,361 (44.4)	293 [161–451]
VCT–Community	429 (0.6)	360 [236–507]
PITC–Inpatient	1,105 (1.6)	244 [114–414]
PITC–Outpatient	9,485 (13.9)	296 [165–456]
Antenatal Care (ANC) clinic	19,420 (28.4)	452 [307–628]
Not available	1,708 (2.5)	320 [186–500]
Other noncoded	4,452 (6.5)	323 [182–492]
SAAJ	344 (0.5)	354 [218–520]
TB Clinic–PNCT	1,028 (1.5)	287 [138–448]
Civil status			<0.0001
Living with partner	23,789 (34.8)	344 [203–527]
Married	11,176 (16.4)	361 [217–535]
Never married	13,413 (19.6)	315 [174–484]
Spouse died/separated/divorced	2,907 (4.3)	299 [168–464]
Not available	17,047 (24.9)	337 [198–520]
District			<0.0001
Namacurra	10,390 (15.2)	334 [182–532]
Quelimane	12,859 (18.8)	322 [173–508]
Other	45,083 (66.0)	343 [206–513]

VCT, voluntary counseling and testing; PITC, provider initiated testing and counseling; SAAJ, Serviço Amigo do Adolescente e Jovem (SAAJ); PNCT, Programa Nacional de Controlo da Tuberculose.

Trends in median CD4 cell counts and quantile regression of the median changes in CD4 cell counts

The youngest age group (15–24 years) had the highest median CD4 cell count, 407 cells/mm³ (IQR: 254, 598) at ART initiation. The median CD4 cell count was significantly higher in women [370 cells/mm³ (IQR: 228, 557)] compared with men [271 cells/mm³ (IQR: 142, 421)] at ART initiation (p = 0.0001). Among women, pregnant/lactating women and those referred from the ANC clinic had considerably higher CD4 cell counts (Fig. 2). After adjusting for other risk factors, the quantile regression showed that compared with the reference age group of 15–24, the median changes in CD4 cell count at ART initiation for the 25–49 and 50+ age groups were −44 and −45 cells/mm³, respectively. The median change in CD4 cell count was +100 cells/mm³ for pregnant/lactating women and +53 cells/mm³ for nonpregnant/nonlactating females compared with men (Table 2).

FIG. 2.

Comparison of median CD4 counts by age and sex groups over time.

Table 2.

Quantile Regression with Covariates Included and Missing Value Imputations

Variable	Change in median CD4 cell count (95% CI)
Policy
I (CD4 cell count <350)	Ref
II (CD4 cell count <500)	41 (33.8–47.8)
III (Test-and-Start)	68 (57.5–78.4)
Year ART Initiated	7 (4.2–9.5)
Age Group
15–24	Ref
25–49	−44 (−48.7 to −39.8)
50+	−45 (−53.6 to −35.6)
Sex group
Male	Ref
Female nonpregnant/nonlactating	53 (48.4–58.0)
Female pregnant/lactating	100 (92.8–106.6)
Facility type
District periphery	Ref
District capital	−5 (−9.5 to −0.8)
Urban other	−23 (−32.4 to −14.3)
Urban large	−34 (−41.5 to −26.4)
Civil status
Married	Ref
Living with partner	−7 (−12.4 to −1.6)
Never married	−22 (−28.5 to −15.3)
Spouse died/separated/divorced	−24 (−34.2 to −14.2)
Education level
No formal education	Ref
Primary school	2 (−3.5 to 6.9)
Secondary school	−13 (−20.2 to −6.8)
University	−30 (−64.6 to 3.8)
WHO clinical stage
Stage 1	Ref
Stage 2	−37 (−42.5 to −32.0)
Stage 3	−33 (−38.2 to −27.9)
Stage 4	−72 (−83.2 to −60.9)
Referral type
VCT	Ref
VCT–community	44 (19.8–68.6)
PITC—inpatient	−35 (−50.6 to −20.3)
PITC—outpatient	−2 (−4.3 to 7.4)
Antenatal care (ANC) clinic	74 (67.8–80.7)
Other noncoded	24 (15.7–31.9)
SAAJ	13 (−13.3 to 40.3)
TB clinic–PNCT	4 (−15.7 to 23.7)
TB Coinfection
No	Ref
Yes	5 (−7.1 to 17.9)

Ref, reference group; VCT, voluntary counseling and testing; PITC, provider initiated testing and counseling; SAAJ, Serviço Amigo do Adolescente e Jovem (SAAJ); PNCT, Programa Nacional de Controlo da Tuberculose.

Patients that were coinfected with TB had lower median CD4 cell counts at ART initiation compared with those without TB. The lowest median CD4 cell counts (261 cells/mm³) were noted in patients with university-level education. Median CD4 cell counts increased as the programmatic years evolved, with the highest value observed at 377 cells/mm³ in 2017 (Table 1).

In the district of Quelimane, the median CD4 cell count increased from 271 cells/mm³ (IQR: 137, 415) to 376 cells/mm³ (IQR: 206, 566) as the guidelines transitioned from Policy I to III. Namacurra, the other district that implemented “Test-and-Start,” had a decrease in median CD4 cell count from 405 cells/mm³ (IQR: 236, 595) to 378 cells/mm³ (IQR: 221–583) between Policy II and III. The median CD4 cell counts in the other district groups increased from 316 cells/mm³ (IQR: 189–504) to 376 cells/mm³ (IQR: 230–521) as they transitioned from Policy I to II (Supplementary Table S1). After including the year in the quantile regression model which accounted for an average increase of 7 cells/mm³ per year, the median changes in CD4 cell count were +41 and +68 cells/mm³ for Policies II and III, respectively, compared with policy I (Table 2).

Multivariable logistic regression

The adjusted odds of initiating ART with severe immunodeficiency under Policy III (“Test and Start”) were 32% lower when compared with Policy I (95% CI: 0.61–0.75). Patients in the age groups 25–49 and 50+ years had 1.36 (95% CI: 1.30–1.42) and 1.25 (95% CI: 1.16–1.36) higher odds, respectively, of initiating ART with severe immunodeficiency compared with the youngest age group. Women had lower odds of initiating ART with severe immunodeficiency compared with men, with pregnant/lactating women having the lowest adjusted odds ratio of 0.42 (95% CI: 0.39–0.45) and those referred from an ANC clinic having an odds ratio of 0.50 (95% CI: 0.47–0.54). Attainment of a university level of education was associated with 1.29 times higher odds (95% CI: 0.97–1.73) of initiating ART with severe immunodeficiency compared with those without any formal education. The odds of initiating ART with severe immunodeficiency was 2.19 times (95% CI: 2.00–2.41) higher in patients with WHO stage 4 clinical disease compared with those with stage 1 (Table 3). The sensitivity analyses showed similar trends as a result from the 36 imputations.

Table 3.

Multivariable Logistic Regression for Determinants of Initiating Antiretroviral Therapy with Severe Immunodeficiency, Defined as Having a CD4 Cell Count Less Than 200 Cells/mm³

Variable	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Policy
I (CD4 cell count <350)	Ref	Ref
II (CD4 cell count <500)	0.65 (0.62–0.67)	0.73 (0.68–0.78)
III (Test-and-start)	0.74 (0.70–0.79)	0.68 (0.61–0.75)
Year of ART initiation	0.86 (0.85–0.87)	0.96 (0.93–0.98)
Age group
15–24	Ref	Ref
25–49	1.91 (1.84–1.99)	1.36 (1.30–1.42)
50+	2.32 (2.15–2.50)	1.25 (1.16–1.36)
Sex group
Male	Ref	Ref
Female nonpregnant/nonlactating	0.67 (0.65–0.70)	0.65 (0.62–0.67)
Female pregnant/lactating	0.20 (0.19–0.21)	0.42 (0.39–0.45)
Facility type
District periphery	Ref	Ref
District capital	1.15 (1.10–1.19)	1.06 (1.02–1.11)
Urban other	1.09 (1.02–1.16)	1.28 (1.18–1.40)
Urban large	1.56 (1.48–1.65)	1.34 (1.25–1.43)
Civil status
Married	Ref	Ref
Living with partner	1.13 (1.07–1.19)	1.07 (1.01–1.13)
Never married	1.46 (1.38–1.55)	1.21 (1.14–1.29)
Spouse died/separated/divorced	1.54 (1.41–1.69)	1.23 (1.11–1.35)
Level of education
No formal education	Ref	Ref
Primary school	1.06 (1.01–1.11)	0.97 (0.93–1.02)
Secondary school	1.19 (1.12–1.26)	1.07 (1.01–1.14)
University	1.79 (1.36–2.36)	1.29 (0.97–1.73)
WHO clinical stage
1	Ref	Ref
2	2.15 (2.05–2.25)	1.48 (1.41–1.55)
3	2.81 (2.69–2.93)	1.66 (1.59–1.74)
4	4.01 (3.67–4.39)	2.19 (2.00–2.41)
Referral type
VCT	Ref	Ref
VCT—community	0.52 (0.41–0.66)	0.63 (0.49–0.80)
PITC—inpatient	1.44 (1.28–1.63)	1.33 (1.17–1.51)
PITC—outpatient	0.98 (0.93–1.02)	0.97 (0.92–1.02)
Antenatal care (ANC) clinic	0.24 (0.23–0.26)	0.50 (0.47–0.54)
Other (noncoded)	0.81 (0.75–0.86)	0.85 (0.79–0.91)
SAAJ	0.59 (0.45–0.76)	0.80 (0.61–1.05)
TB clinic–PNCT	1.15 (1.01–1.31)	0.93 (0.78–1.11)
TB Coinfection
No	Ref	Ref
Yes	1.59 (1.47–1.72)	1.00 (0.90–1.12)

Ref, reference group; VCT, voluntary counseling and testing; PITC, provider-initiated testing and counseling; SAAJ, Serviço Amigo do Adolescente e Jovem (SAAJ); PNCT, Programa Nacional de Controlo da Tuberculose; ART, antiretroviral therapy.

Discussion

In our analyses of immunodeficiency at ART initiation over a 5-year period in Zambézia Province, we found a reassuring trend of increasing median CD4 cell counts, which correlated with evolving WHO treatment guidelines and national policies. These trends have been reported in prior studies that examined aggregated data by country income levels.^11,12 To our knowledge, our study is one of the first to demonstrate the positive effects of the policy changes, including “Test and Start” on the median change in CD4 cell count at the time of ART initiation within a sub-Saharan African country. This was particularly evident in the district of Quelimane, which was the first to implement the “Test and Start” strategy within Zambézia Province. The results were not as dramatic in the district of Namacurra, likely due to insufficient data as a result of a shorter implementation duration within the study period.

A multivariable analysis revealed that male sex, older age, WHO stage 4 clinical disease, and referrals from inpatient PITC services were associated with higher odds of initiating ART with severe immunodeficiency. The gender disparities in immunodeficiency at ART initiation have been well documented across sub-Saharan Africa.^22–27 These studies demonstrated that men were more likely to engage in care with more advanced disease compared with women and as a result had slower immune recovery and higher mortality. In a large urban clinic in South Africa, men were found to have a 20% greater risk of death at 24- and 36 months of follow-up compared with women.²³ Although these gender discrepancies are likely as a result of the emphasis placed on EMTCT and implementation of Option B+ programs (ART initiation in pregnant/lactating women regardless of immune status/CD4 cell count) across the region, the negative economic impact of stigma on the livelihood of men is also considered to play a major role.^28,29 Strategies to improve male engagement in care have been recommended and include community-based HIV testing and ART initiation, flexible clinic hours, private access to care, and integration of HIV testing with chronic disease monitoring.³⁰

Similar to our study, older individuals in Malawi were shown to initiate ART with more advanced disease and have less robust immune responses compared with younger HIV-positive individuals.³¹ The authors postulated that limited access to health care and low index of suspicion of HIV infection in older individuals were contributing factors. Although there is a paucity of data regarding HIV treatment outcomes in the elderly in sub-Saharan Africa, prudent approaches to the co-management of HIV and chronic diseases are warranted in this population.

Since patients classified as WHO stage 4 have more advanced disease; they are more likely to be hospitalized with AIDS-defining illnesses, which would prompt testing for the diagnosis of HIV. Hence, this could explain the higher odds of severe immunodeficiency at ART initiation observed in patients referred from inpatient PITC services. Similar observations were noted in studies across HIV clinics in Ethiopia and Mozambique.^26,32 Since WHO clinical stage 4 is a condition of severe immunodeficiency, our model was intended to be descriptive rather than causal.

Although a small proportion of the patients (<1%) had received university-level education, we observed nonsignificant trend toward severe immunodeficiency at ART initiation in this subgroup. This finding is in contrast to prior studies where no associations were observed.^26,33 We postulate several reasons for this finding in Zambézia Province. University-educated individuals are more likely to be employed as civil servants and have clerical jobs, which makes it difficult to seek care during regular clinic hours. Additionally, HIV stigma may play a significant role in their hesitation to present to care due to their higher socioeconomic status. Further studies are needed to better understand this phenomenon since the small sample size of university-educated patients in our analyses limited statistical significance and the few patients who might have sought care in the private sector would need to be accounted for. Some of the male engagement strategies described above may also prove to be useful in this population.

Our study has several limitations. The relatively short period of the implementation of “Test and Start” throughout Zambézia Province meant that there was insufficient data to analyze the long-term effects of this policy. As more data are collected, further analyses utilizing a stepped-wedge design are anticipated. There was a significant amount of missing data for the following baseline variables: WHO clinical stage, sexual orientation, civil status, and level of education. These missing variables were included in our analyses after multiple imputation was employed but could have reduced the representativeness of our sample. However, the large sample size, a major strength of our analysis would curb this effect. Moreover, our sensitivity analyses showed similar trends as the multiple imputation analysis.

We report an increasing trend in median CD4 cell count with evolving policies in Zambézia Province and risk factors associated with initiating ART with severe immunodeficiency. Also, as “Test and Start” becomes ubiquitous in Mozambique, considerable public health efforts should be employed toward universal HIV testing, as demonstrated in several sub-Saharan countries, to ensure that men, elderly patients, university-educated individuals, and patients diagnosed in inpatient settings are rapidly linked to care and treatment.^36,37

Conclusions

Although there were reassuring trends in increasing median CD4 cell counts with ART initiation, ongoing efforts are needed that target universal HIV testing to ensure the early initiation of ART in men and older patients.

Author Contributions

F.A. and C.W.W. designed and implemented the study. WG performed the statistical analyses. F.A. wrote the initial draft of the article. F.A., W.G., A.F.G., C.D., J.C., W.S., G.C., J.T., M.S., E.G., S.V.R., M.F.A.S., S.T., T.M., and C.W.W. contributed to the interpretation and presentation of the findings. All authors were involved in article revisions and approved the final version of the article for submission.

Author Disclosure Statement

The authors have no competing interests to declare. The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of the CDC.

Funding Information

This evaluation has been supported by the President's Emergency Plan for AIDS Relief (PEPFAR) through the Centers for Disease Control and Prevention (CDC) under the terms of Cooperative Agreements U2GGH002071 and U2GGH001943.

Supplementary Material

Supplementary Table S1