Older Age at Infection and Nulliparity are Associated with Long-term Non-progression in Female Sex Workers Infected with Non-Subtype B HIV-1

Abstract

Background:

Studies have reported on HIV-infected, antiretroviral therapy (ART)-naïve individuals who show minimal disease progression despite prolonged infection. The characteristics of these long-term non-progressors (LTNPs) are not well-characterized in populations predominantly infected with non-subtype-B HIV-1.

Methods:

Female sex workers in Mombasa, Kenya who acquired HIV-1 were studied to ascertain immunological disease progression. Long-term non-progression was defined as an ART-naïve duration of infection ≥7 years and a majority of CD4+ counts ≥600 cells/µL with a non-declining CD4+ trend. Correlates of long-term non-progression were determined using multivariable logistic regression.

Results:

Between February 1993 and March 2014, 332 women acquired HIV-1. Of these, 77 (23%) had ≥7 years of follow-up and 13 (17%) were categorised as LTNPs. Factors associated with long-term non-progression included age >30 years at infection (aOR=9.41, 95% CI: 1.48–59.86, P=0.005) and nulliparity (aOR=20.19, 95% CI: 1.36–299.90, P=0.03). Each log10 increase in viral set point was associated with a lower likelihood of being a LTNP (aOR=0.31, 95% CI: 0.12–0.79, P=0.01).

Conclusion:

These findings suggest that age and parity may influence the likelihood of long-term non-progression through mechanisms that are not mediated by the effects of these variables on viral load. Future studies should seek to determine whether the associations presented are reproducible.

Introduction

In the early 1990’s, a small proportion of antiretroviral therapy (ART)-naïve individuals were described who showed little or no evidence of immunologic decline despite being infected with HIV-1 for ≥7 years^1–4. These individuals were referred to as long-term non-progressors (LTNPs). Studies of LTNPs have contributed to understanding the pathogenesis of HIV-1 and are relevant in advancing the HIV cure research agenda^{5, 6}.

There is however, paucity of data describing the phenomenon of long-term non-progression among individuals infected with non-B HIV-1 subtypes^{7, 8}. The present study sought to describe the presence and characteristics of long-term non-progression in a population of female sex workers (FSWs) who remained ART-naïve for ≥7 years after HIV-1 infection in Mombasa, Kenya. The predominant infecting HIV-1 subtypes in this setting are A, C, and D^9–11.

Methods

Study procedures for the Mombasa Cohort of FSWs have been published¹². Briefly, at enrollment and monthly follow-up visits, participants completed face-to-face interviews to ascertain demographic information as well as sexual and reproductive history. A pelvic examination was performed with collection of specimens for genital tract infection screening. In HIV-1-seronegative women, blood was collected monthly for HIV-1 screening, and an aliquot of plasma was stored at −80°C. In those who seroconverted for antibodies to HIV-1, blood was collected quarterly to determine CD4⁺ counts and plasma viral load (PVL).

Women were eligible for this analysis if they were ART-naïve, had accrued ≥7 years of follow-up after HIV-1 infection, and had ≥2 results from CD4⁺ counts available, excluding those in the first year after seroconversion (period of acute infection). To estimate the time of HIV-1 infection, stored plasma samples obtained prior to seroconversion were tested to identify any that were HIV-1 RNA positive. If ≥1 pre-seroconversion sample was HIV-1 RNA positive, the estimated time of infection was considered to be 17 days prior to the first RNA-positive sample¹³. If no pre-seroconversion samples were HIV-1-RNA-positive, the estimated time of infection was considered to be the midpoint between the last seronegative visit and the first seropositive visit.

HIV-1 RNA in stored plasma samples was measured using the Gen-Probe (Gen-Probe Inc., San Diego, CA, USA) first-generation viral load assay with a lower limit of quantitation of 100 copies/mL. Set point viral load was defined as the first PVL measurement between 120 – 730 days (2 years) after the estimated time of HIV-1 infection. Infecting HIV-1 subtype was determined as previously described¹⁴.

Based on previous studies, LTNPs were defined as having a majority (at least two-thirds) of available CD4⁺ counts ≥600cells/µL^15–18. Follow-up was measured from the estimated time of infection to the last available visit before loss to follow-up, initiation of ART, death, or administrative censoring on 31^st March 2014. For each potential LTNP, a manual review of the linear change in CD4⁺ counts over time (trend) was required to show a pattern consistent with non-progression (i.e. rising or constant/non-declining).

Screening for HIV-1 was performed using an enzyme linked immunosorbent assay (ELISA): Detect-HIV 1–2 (BioChem ImmunoSystems, Montreal, Canada) was used between February 1993 and January 2010. A second ELISA assay, Pishtaz HIV 1–2 (Pishtaz Teb Diagnostics, Tehran, Iran) was used between February 2010 and March 2014. Positive screening results were confirmed with a second ELISA. Recombigen HIV 1–2 (Cambridge Biotech, Worcester, MA, USA) was used from February 1993 through August 2004, Biorad HIV 1–2 (Biorad Laboratories, Hercules, CA, USA) from September 2004 through May 2006, and Vironostika HIV-1 Uni-Form II Ag/Ab (Biomeriux, Marcy-l’Etoile, France) after June 2006.

Enumeration of CD4⁺ counts from April 1993 until January 1998 was performed using Cytosphere (Coulter, Haileah, FL, USA). Between February 1998 and October 2004, the Zymmune (Zynaxis Inc., Malvern, PA, USA) system was used. Thereafter, CD4⁺ counts were performed using FACSCount (Becton Dickinson, Franklin Lakes, NJ, USA).

For this analysis, baseline characteristics were summarized as median (interquartile range [IQR]) for continuous data and counts (N [%]) for categorical data. Univariate logistic regression was used to compare variables in LTNPs versus non-LTNPs. Comparison categories were formed by dichotomizing continuous data at the median. The outcome of interest was LTNP status, as defined above. While multiple CD4 cell counts were evaluated to determine outcome status, each participant had only one outcome assigned (i.e. LTNP or non-LTNP).

To identify independent predictors of LTNP status, exposures associated with long-term non-progression in univariate analyses with P≤0.10 were included in a multivariable model. The analysis evaluated correlates associated with a single outcome (i.e. LTNP or non-LTNP) without violating the assumption of independent observations. Data were analyzed using STATA Version 12 (StataCorp LP, College Station, TX, USA) and IBM SPSS Statistics Version 25 (IBM Corp, Armonk, NY, USA).

Beginning in March 2004, all participants who were eligible were offered ART according to Kenyan guidelines. Under these guidelines, ART was initially recommended for patients with CD4 counts <200 cells/µL. This threshold was raised to <350 cells/µL between October 2010 and June 2014 then <500 cells/µL until July 2016 when the universal test-and-start guideline was adopted. Until this time, women being followed as LTNPs were not eligible to initiate ART since, by definition, they continued to have CD4 counts well above the thresholds used to define ART eligibility.

Ethical approval for this study was obtained from the review committees of Kenyatta National Hospital, the University of Washington, and the Fred Hutchinson Cancer Research Center. All participants provided written informed consent.

Results

Between February 1993 and March 2014, there were 332 new HIV-1 infections in the Mombasa Cohort. Seventy-seven (23%) newly infected women were eligible for this analysis based on having ≥7 years of ART-naïve follow-up. Of these, 13 (17%) were categorized as LTNPs (Figure 1).

Figure 1.

Identification of LTNPs among HIV-1 seroconverters in the Mombasa Cohort of FSWs. LTNP: long-term non-progressor.

Among the 77 eligible women, the median age at sexual debut was 17 (interquartile range, IQR 15–18) years. The median age at time of HIV-1 infection was 29 (IQR 26–34) years (Table 1). The women reported a median of 3 (IQR 2–6) years of sex work and 2 (IQR 1–3) prior live births. The median viral load set point was 4.43 log₁₀ copies/mL (IQR 3.86–5.17). Among 52 (68%) seroconverters with infecting HIV-1 subtype data available, 39 (75%) were infected with subtype A, 7 (13%) with subtype D, and 4 (7%) with subtype C.

Table 1:

Characteristics of female sex workers who seroconverted to HIV-1 and were evaluated for long-term non-progression in the Mombasa Cohort

	All women evaluable as LTNPs	Women identified as LTNPs	Women evaluable but not identified as LTNPs
	N (%) or Median (IQR)	N (%) or Median (IQR)	N (%) or Median (IQR)
	(N=77)	(N=13)	(N=64)
Age at infection	29 (26–34)	33 (31–36)	28 (25–32)
Age at sexual debut	17 (15–18)	18 (17–19)	17 (15–18)
Number of prior births	2 (1–3)	1 (1–2)	2 (1–3)
Years of education	8 (7–10)	8 (6–11)	8 (7–10)
Years of sex work	3 (2–6)	7 (2–11)	3 (2–6)
Alcohol use	66 (86)	12 (92)	54 (84)
Genital tract infections
HSV-2	70 (91)	10 (77)	60 (94)
TV	4 (5)	0 (0)	4 (6)
VVC	6 (8)	0 (0)	6 (9)
BV	15 (19)	2 (15)	13 (20)
GC	4 (5)	0 (0)	4 (6)
Contraceptive use
None	20 (26)	4 (31)	16 (25)
DMPA	28 (36)	2 (15)	26 (41)
OCP	15 (19)	2 (15)	13 (20)
Infecting HIV subtype*
A	39 (75)	6 (86)	33 (73)
A-recombinant	2 (3)	1 (14)	1 (2)
C	4 (8)	0 (0)	4 (9)
D	7 (14)	0 (0)	7 (16)
Viral load set point (log10 copies/mL)	4.43 (3.86–5.17)	3.91 (3.27–4.68)	4.59 (3.94–5.23)

Among 52, 7 and 45 women respectively, with infecting HIV-1 subtype data available; TV; Trichomonas vaginalis, VVC; vulvovaginal candidiasis; BV; bacterial vaginosis, GC; gonococcus, DMPA; depot medroxyprogesterone acetate, OCP; oral contraceptive pill

In univariate analyses, older age (≥16 versus <16 years) at sexual debut (OR 3.53, 95% confidence interval, CI: 0.72–17.26, P=0.1), older age (>30 versus ≤30 years) at HIV-1 infection (OR 9.17, 95% CI: 1.87–44.92, P=0.006), more years (>3 versus ≤3 years) of sex work (OR 2.67, 95% CI: 0.73–9.78, P=0.1), and nulliparity (OR 11.46, 95% CI: 0.96–137.39, P=0.05) were associated with significantly increased odds of being a LTNP. Each log₁₀ copies/mL increase in viral set point was associated with 54% lower odds of being a LTNP (OR 0.46, 95% CI: 0.22–0.96, P=0.04) (Table 2).

Table 2:

Correlates of long-term non-progression among female sex workers with ≥7 years of ART-naïve follow-up in the Mombasa Cohort

Exposure	LTNPs N (%)/median (IQR) (N=13)	Non-LTNPs N (%)/median (IQR) (N=64)	Univariate OR (95% CI)	P-value	Multivariate aOR* (95% CI)	P-value
Age at sexual debut
≥16 years	11 (85)	39 (61)	3.53 (0.72 – 17.26)	0.1	5.94 (0.91 – 38.88)	0.06
Age at HIV-1 infection
>30 years	11 (85)	24 (38)	9.17 (1.87 – 44.92)	0.006	9.41 (1.48 – 59.86)	0.005
Years of sex work
>3 years	8 (67)	27 (42)	2.67 (0.73 – 9.78)	0.1	1.84 (0.44 – 7.66)	0.4
Parity at enrollment
No previous birth	2 (15)	1 (2)	11.46 (0.96 – 137.39)	0.05	20.19 (1.36 – 299.9)	0.03
Viral load set-point (log10 copies/mL)	3.91 (3.27 – 4.68)	4.59 (3.94 – 5.23)	0.46 (0.22 – 0.96)	0.04	0.31 (0.12 – 0.79)	0.01
Infecting HIV subtype
Subtypes C/D/recombinant compared to subtype A	1 (14)	12 (27)	0.44 (0.05 – 4.09)	0.5
Contraceptive use at infection
None/Condoms/IUCD	5 (56)	17 (29)	1.0
Depo-Provera	2 (22)	26 (45)	0.28 (0.05 – 1.59)	0.2
Norplant/OCPs	2 (11)	15 (26)	0.51 (0.09 – 3.06)	0.5
Any genital infection
Yes	0 (0)	8 (18)	N/A

^*Adjusted for age at HIV-1 infection, years of sex work, parity and PVL set point. Due to collinearity, age at HIV-1 infection and parity were excluded from the multivariate models

In multivariate analyses, older age at HIV-1 infection (aOR 9.41, 95% CI: 1.48–59.86, P=0.005) and nulliparity (aOR 20.19, 95% CI: 1.36–299.90, P=0.03) remained significantly associated with increased odds of being a LTNP. Conversely, each log₁₀ copies/mL increase in viral set point was associated with 69% lower odds of being a LTNP (OR 0.31, 95% CI: 0.12–0.79, P=0.01).

Table 3 presents results from a sensitivity analysis that retains the following exposures (age at sexual debut, age at HIV-1 infection, years of sex work and number of prior live births) as continuous variables in the logistic regression model. In multivariate analyses, each 1-year increment in age at HIV-1 infection was significantly associated with increased odds of being a LTNP (aOR 1.28, 95% CI: 1.09–1.51, P=0.003). Conversely, each additional live birth (aOR 0.32, 95% CI: 0.13–0.80, P=0.02) as well as each log₁₀ copies/mL increase in viral set point (aOR 0.46, 95% CI: 0.19–1.08, P=0.07) was associated with lower odds of being a LTNP.

Table 3:

Correlates of long-term non-progression among female sex workers with ≥7 years of ART-naïve follow-up in the Mombasa Cohort (sensitivity analysis without dichotomizing continuous exposures)

Exposure	LTNPs N (%)/median (IQR) (N=13)	Non-LTNPs N (%)/median (IQR) (N=64)	Univariate OR (95% CI)	P-value	Multivariate aOR* (95% CI)	P-value
Age at sexual debut
1-year increment	18 (17 – 19)	17 (15 – 18)	0.99 (0.93 – 1.05)	0.7
Age at HIV-1 infection
1-year increment	33 (31 – 36)	29 (26 – 33)	1.15 (1.01 – 1.30)	0.03	1.28 (1.09 – 1.51)	0.003
Years of sex work
1-year increment	6.5 (2.3 – 11.0)	3.0 (2.0 – 6.0)	1.00 (0.97, 1.04)	0.9
Parity at enrollment
Per live birth	1 (1 – 2)	2 (1 – 3)	0.54 (0.28 – 1.03)	0.06	0.32 (0.13, 0.80)	0.02
Viral load set-point (log10 copies/mL)	3.91 (3.27 – 4.68)	4.59 (3.94 – 5.23)	0.46 (0.22 – 0.96)	0.04	0.46 (0.19, 1.08)	0.07
Infecting HIV subtype
Subtypes C/D/recombinant compared to subtype A	1 (14)	12 (27)	0.44 (0.05 – 4.09)	0.5
Contraceptive use at infection
None/Condoms/IUCD	5 (56)	17 (29)	1.0
Depo-Provera	2 (22)	26 (45)	0.28 (0.05 – 1.59)	0.2
Norplant/OCPs	2 (11)	15 (26)	0.51 (0.09 – 3.06)	0.5
Any genital infection
Yes	0 (0)	8 (18)	N/A

^*Adjusted for age at HIV-1 infection, parity and PVL set point. Due to collinearity, age at HIV-1 infection and parity were excluded from the multivariate models

Discussion

In this prospective study of FSWs predominantly infected with HIV-1 subtype A and with ≥7 years of ART-naïve follow-up, 17% were identified as LTNPs. Women were significantly more likely to be LTNPs if they were older at the time of HIV-1 infection and if they were nulliparous at enrolment into the cohort. These findings were independent of the effect of PVL on HIV-1 progression, suggesting that age and parity may influence the likelihood of being a LTNP through mechanisms that are not mediated entirely by the effects of these variables on PVL.

Previous studies on age and HIV-1 progression have had mixed results^19–21. Generally, older age has been associated with more rapid progression, which is hypothesized to result from weaker immunologic responses due to age-related immunosenescence^22–24. In contrast, this study found that women who were older at the time of HIV-1 infection were more likely to be LTNPs. In this FSW population, an interesting hypothesis to explain this observation is that women may have acquired immunologic responses to the virus from multiple prior mucosal exposures before becoming HIV-infected, resulting in delayed disease progression^{25, 26}.

Studies of HIV-1-seropositive women who become pregnant have suggested that pregnancy-mediated immune alterations can lead to rapid disease progression²⁷. It is possible that similar immunologic mechanisms among women who acquire HIV-1 following one or more prior pregnancies might explain the lower likelihood of LTNP status associated with prior pregnancies in this cohort.

A key strength of this study was that long-term follow-up in the Mombasa Cohort provided an ideal opportunity to answer this unique research question. Follow-up procedures included longitudinal collection of data and biological specimens before and after HIV-1 infection, facilitating the study of various potential correlates. Due to the global shift toward immediate ART initiation for all HIV-infected individuals, it is unlikely that such a group would be studied in the future. Nonetheless, it may be possible to examine questions related to LTNPs in existing datasets with extended follow-up of populations prior to initiation of ART.

This study analyzed correlates of long-term non-progression in the population of women with ≥7 years of ART-naïve follow up. Excluding women who died or were lost to follow-up <7 years following infection would predictably generate an estimate of the proportion of LTNPs that is higher than the proportion that would have been identified among all women who initially acquired HIV-1. Based on the high rate of long-term attrition due a variety of factors including migration and death²⁸, estimates in the full dataset could have been biased in ways that are difficult to predict. The present analysis, while it does not estimate the proportion of LTNPs among all FSWs infected with HIV, had the advantage of including a clearly identifiable group that could be used for comparison in the analysis for correlates.

These findings need to be interpreted in the context of several limitations. First, the small number of LTNPs identified, including only 3 who were nulliparous (two LTNP and one non-LTNP), provided limited statistical power, leading to wide CIs. Given that LTNPs are a rare population, the structure of this analysis meant that the risk of limited statistical power was inevitable. Second, due to the nature of data collection in the cohort, parity was evaluated at the time of cohort enrollment and not when infection was acquired. Third, known immunogenetic correlates of long-term non-progression such as human leukocyte antigen type were not assessed. Despite these limitations, it is notable that the present analysis represents one of the largest cohorts of LTNPs infected with non-B subtypes described to date.

In conclusion, this analysis found that older age (>30 years) at time of HIV-1 infection and nulliparity were associated with increased likelihood of being a LTNP. These effects were independent of the effect of PVL on HIV-1 progression. Future research using historical datasets should determine whether these associations are reproducible. If so, it will be important to explore potential mechanisms linking older age and nulliparity with long-term non-progression.