HIV incidence in South African blood donors from 2012 to 2016: a comparison of estimation methods

Abstract

Background:

Measuring incidence is important for monitoring and maintaining the safety of the blood supply. Blood collected from repeat-donors has provided the opportunity to follow blood donors over time and has been used to estimate the incidence of viral infections. These incidence estimates have been extrapolated to first-time donors using the ratio of NAT yield cases in first time versus repeat-donors. We describe a model to estimate incidence in first-time donors using the Limiting Antigen (LAg) avidity assay and compare its results with those from established models.

Methods:

HIV positive first-time donations were tested for recency using the LAg assay. Three models were compared; incidence estimated for 1) first-time donors using LAg avidity 2) first-time and repeat-donors separately using the NAT yield window period (WP) model and 3) repeat-donors using the incidence/WP model.

Results:

HIV incidence in first-time donors was estimated at 3.32 (CI 3.11, 3.55) and 3.81 (CI 3.07, 4.73) per 1000 PY using the LAg assay and NAT yield WP models, respectively. Incidence in repeat-donors was between 2.0 and 2.5-fold lower than in first-time donors estimated at 1.56 (CI 1.37, 1.77) and 1.94 (CI 1.86–2.01) per 1000 PY using the NAT yield/WP and incidence/WP models, respectively.

Conclusion:

Testing HIV positive donations using the LAg assay provides a reliable method to estimate incidence in first-time donors for countries that collect the majority of blood from first-time donors and do not screen with NAT.

Introduction

Measuring the incidence of HIV is important for public health and maintaining the safety of the blood supply. Historically, incidence has been studied in expensive longitudinal cohort studies. Blood transfusion services that collect blood from mostly repeat-donors provide the opportunity to follow donors over time and have been used to estimate the incidence of viral infections in the blood supply[1]. Using the estimated incidence and an estimate of the mean duration of pre-detectable infectiousness (based on viral load growth during the ramp-up phase of infection and the HIV-1 minimum infectious dose), we can estimate the proportion of screen-negative blood donations which are potentially infectious ie. residual risk (RR)[2]. These repeat-donor based incidence and consequent RR estimates have been extrapolated to first-time donors by multiplying the repeat-donor estimates with a factor such as the first-time to repeat-donor prevalence ratio or the first-time to repeat-donor NAT yield (HIV RNA positive, HIV antibody negative) ratio. The prevalence ratio is expected to overestimate the incidence and subsequent RR in first-time donors, especially in high endemic countries, as repeat-donors are prescreened for infection.[3] The NAT yield ratio can only be used in contexts that perform NAT testing in parallel with serology, and in most countries, the numbers of NAT yield cases are relatively small resulting in poor precision of the NAT yield ratio. Many blood services in developing countries collect blood mostly from first-time donors.

With the use of more sensitive NAT in parallel with serological assays a “NAT yield window period (WP) model” can estimate first-time donor incidence and RR[4]. Many blood services that use NAT to screen blood, use this NAT yield WP model to estimate RR. However, the small number of NAT yield donations, typically detected, limits the precision of risk estimates and the cost of NAT limits its application. Newer “incidence assays” with lower sensitivity to HIV antibody allow the definition of a tunable (but longer) “recent infection” WP compared to NAT. Unlike NAT testing on serological negative specimens, incidence assays need only be performed on the HIV antibody-positive donations and can yield a much larger sample of recent infections providing more precision for incidence estimations in first-time donors[5]. However, these more recent serologic approaches for incidence estimation have not been extensively validated in the blood bank setting.

As part of a five-year study, all HIV seropositive donations were tested with an incidence assay to determine recent HIV infections.[6–8] We compared incidence estimates calculated from three approaches: 1) the cross-sectional LAg incidence assay method using first-time donors[7]; 2) the NAT yield WP model[4] in both repeat-donors and first-time donors, and 3) the incidence/WP (I/WP) classic method using repeat-donors[2]. In addition we compared the incidence in first-time donors with that reported in the general population[9] and examined differences in incidence over the study period in different geographical regions and in donors of different ethnicities, gender and ages.

Methods

Setting

SANBS collects approximately 900,000 blood donations per annum in a high HIV endemic context and screens blood in parallel for NAT and anti-HIV leading to approximately 1600 confirmed HIV seropositive and 60 HIV NAT yield donations per year.

Laboratory testing algorithms

All blood donations were screened by the Prism (Abbott, Delkenheim, Germany) 3^rd generation anti-HIV, HBsAg and anti-HCV chemiluminescent immunoassays (ChLIA) in parallel with the Procleix Ultrio (Plus) NAT assay (Grifols, Barcelona, Spain) for HIV RNA, HCV RNA, and HBV DNA. An extensive confirmatory and follow-up algorithm (previously described)[10] was in place for HIV, which classifies donations into four categories: 1) HIV negative, 2) HIV concordant positive (HIV RNA by NAT and anti-HIV by ChIA), 3) HIV “NAT yield” and 4) HIV “serology yield” (RNA negative, anti-HIV ChIA repeat reactive and Western blot positive). The NAT yield cases are confirmed through seroconversion to anti-HIV positive at recall (approximately 70% return) or, for non-returning donors, replicate testing by Procleix Ultrio (Plus) discriminatory HIV NAT and quantitative PCR (Cobas Ampliprep/Cobas Taqman HIV-1/2, Roche, Basel, Switzerland) or Real time HIV-1 m2000rt (Abbott, Delkenheim, Germany), using stored fresh frozen plasma from the index donation.

HIV concordant positive donations were tested for recency of infection using the Limiting Antigen (LAg) avidity assay (Sedia Biosciences Corporation, Portland, Oregon). The single-well LAg avidity EIA is responsive to the avidity of HIV-1 specific IgG, as it quantifies antibody binding to a multi-subtype recombinant HIV-1 antigen coated onto assay plates at low density (hence the term “limiting antigen”, typically offering just a single binding site to multivalent IgG antibodies, and hence demonstrating a progressive increase in signal intensity over 3–6 months following seroconversion that can be used to infer duration of infection.[11] A normalized optical density (ODn) of <1.5 for the recent/longstanding threshold has been demonstrated to define a state of “recent infection” with a mean duration of approximately 180 days for Clade C HIV infections.[12] Five recency or incidence assays, including the LAg avidity assay, have been robustly evaluated and validated by the Consortium for the Evaluation and Performance of HIV Incidence Assays. A false recency rate (FRR) of 1.3% (95% CI 0.3–3.2) was estimated for the LAg assay. [5]

Incidence Models

Only allogeneic whole blood donations were included in the analysis. The definition of a repeat-donor is different for model 2 and 3 (see appendix). Age categories were collapsed for reporting based on incidence into 3 groups; younger than 20 years, 20 to 30 years and older than 30 years. Confidence limits for incidence estimates were derived from Poisson regression.

Model 1: The LAg first-time donor model

First-time donors were categorized into three groups: HIV negative, recent HIV infection, and longstanding HIV infection. Recent HIV infection was defined as 1) HIV concordant positive and LAg avidity recent (ODn<1.5) or 2) NAT RNA positive, anti-HIV negative. There were 123 (2%) confirmed HIV positive donors with missing LAg results and were imputed as recent or longstanding HIV using a fully conditional specification multiple imputation logistic regression method (see appendix). Donors with longstanding infection (ODn ≥1.5) were classified as prevalent and excluded from the analysis. Incidence was calculated as incident cases/1000 person-years (PY) using only cases classified as recent infections.

The denominator was defined as the total time for at-risk first-time donors with each uninfected first-time donor contributing the full mean duration of recent infection (MDRI) estimate and each recently infected donor contributing half the MDRI estimate. The MDRI of 195 (96% CI: 168–222) days estimated by Grebe and colleagues[12] was used which includes the length of the period that anti-HIV is positive and LAg with an ODn<1.5 (179 days) and the length of the period that NAT is positive and anti-HIV is negative (16 days)[12].

Model 2: NAT yield WP model

The NAT yield WP model previously described[4] classifies incident infections as NAT positive, anti-HIV negative. Incidence was calculated as NAT yield cases/1000 PY. Each uninfected donor contributed the full NAT yield detection period and each NAT yield donor contributed half the NAT yield period to the denominator. The length of the NAT yield detection period of 15.4 days was used[13, 14]. As this model assesses new infections cross-sectionally based on a brief NAT yield detection period, we used it for estimation in first-time and repeat-donors. For this model a repeat-donor is classified as a donor who donated previously within or outside of the study period.

Model 3: the classic incidence/WP model.

This model originally described by Schreiber et al[2] only estimates incidence in repeat-donors. Donors are classified as repeating donors if they contributed 2 or more donations during the study period, the term “repeating” is used to differentiate these donors from “repeat” donors in model 2. (see appendix) A donor is classified as having an incident HIV infection if they have two or more donations during the study period of which the first donation tested HIV negative and any subsequent donation tested HIV positive. Incidence was calculated as cases/1000 PY. The denominator was defined as the total follow-up time for the at-risk repeating donors. For those who remain free of infection throughout the study period, time at risk is the time from first to last donation in the study period. For those found to be infected at the second or subsequent donations, time at risk is time from the first donation in the study period to halfway between the last donation at which the donor was uninfected and the donation at which infection was detected.

Incidence in the general population

We compared the incidence derived in first time donors with the incidence reported in the general population from the UNAIDS data 2019 report. [9]

Results

SANBS collected 4,019,985 whole blood donations from January 2012 until December 2016. Of these 3,504,481 (87%) were collected from 723,166 repeat-donors and 515,504 (13%) were collected from first-time donors (used in models 1 and 2). There were 3,661,669 (91%) donations classified as coming from repeating donors for model 3. Table 1 provides the number of donations used to calculate person-years (PY) for each model. Table 2 provides the HIV positives used as incident cases in each model by year, province, ethnicity, gender and age. HIV prevalence in all donations, in first-time donations and in repeat donations were 0.23%, 1.11% and 0.1%, respectively, giving a first-time to repeat-donor HIV prevalence ratio of 11:1. The NAT yield rate in all donations, first-time donations and repeat donations were 0.0078%, 0.016% and 0.0066%, respectively, giving a first-time to repeat-donor NAT yield ratio of 2.42:1.

Table 1:

Donations for each model

	Model 1	Model 2			Model 3
	LAg first time donors*	All	First time	Repeat	Repeating
Total	513334
Total used in denominator	508796	4 019 985	515 504	3 504 481	3 661 669
Year
2012	98690	780 826	100 229	680 597	697581
2013	108055	796 638	109 755	686 883	728479
2014	103000	803 641	104 323	699 318	742204
2015	102033	828 449	103 291	725 158	767212
2016	97018	810 431	97 906	712 525	726223
Province
Gauteng	226558	1 805 636	229 145	1 576 491	1645924
KwaZulu Natal	103638	686 877	105 339	581 538	612057
Mpumalanga	46014	395 308	46 962	348 346	361922
Eastern Cape	47967	373 306	48 543	324 763	342237
Free State	26895	269 849	27 249	242 600	250974
North West	23940	227 523	24 140	203 383	210101
Limpopo	20842	138 395	21 074	117 321	124689
Northern Cape	12936	123 090	13 052	110 038	113748
Ethnicity
White	157523	2 296 050	158 340	2 137 710	2160998
Black	272117	1 156 029	277 350	878 679	991200
Asian	36174	301 480	36 525	264 955	272142
Coloured	27927	215 940	28 120	187 820	197415
Unknown	15055	50 486	15 169	35 317	39944
Gender
Female	283185	1 741 727	287 345	1 454 382	1544856
Male	225603	2 278 249	228 151	2 050 098	2116843
Age
16–19	235252	720 898	238 229	482 669	605 952
20–30	142808	991680	144 647	847 033	872 909
>30	130736	2 309 771	132 450	2 177 321	2 180826

^*Note: the first time donations classified as long standing are subtracted from the total.

Longstanding donations are excluded from the denominator

Table 2:

classification of HIV positives for each model

	Model 1		Model 2									Model 3
	LAg recent		RNA+/Ab+			RNA−/Ab+/WB+			RNA+/Ab−			New infections in Repeating donors
	First time	Missing
			All	First time	Repeat	All	First time	Repeat	All	First time	Repeat
Longstanding	4538
recent	879
Total	879	123	8607	5462	3 145	234	202	32	312	82	230	2425
Year
2012	176	29	1548	1076	472	18	15	3	60	17	43	178
2013	228	33	1858	1239	619	36	34	2	82	26	56	448
2014	162	34	1795	1115	680	39	35	4	57	15	42	545
2015	165	7	1811	1082	729	63	54	9	54	14	40	644
2016	148	20	1595	950	645	78	64	14	59	10	49	610
Province
Gauteng	312	22	3310	2154	1 156	93	82	11	110	28	82	886
KwaZulu Natal	235	44	1943	1165	778	57	50	7	84	32	52	573
Mpumalanga	142	23	1361	948	413	35	30	5	46	16	30	326
Eastern Cape	70	26	798	459	339	15	13	2	29	2	27	276
Free State	61	2	517	302	215	12	9	3	22	2	20	169
North West	18	5	211	141	70	6	5	1	8		8	58
Limpopo	29	1	311	203	108	12	10	2	8	1	7	93
Northern Cape	12	0	156	90	66	4	3	1	5	1	4	43
Ethnicity
White	16	3	308	85	223	7	5	2	13	1	12	139
Black	831	114	7790	5126	2 664	217	187	30	281	77	204	2113
Asian	2	1	85	31	54	3	3	0	5	1	4	35
Coloured	20	2	295	125	170	2	2	0	10	1	9	116
Unknown	10	3	129	95	34	5	5	0	3	2	1	22
Gender
Female	651	83	5732	3763	1 969	187	162	25	213	66	147	1563
Male	228	40	2875	1699	1 176	47	40	7	99	16	83	862
Age
16–19	268	34	1840	1371	469	28	26	2	59	22	37	426
20–30	418	42	3855	2263	1 592	62	46	16	162	45	117	1211
>30	193	47	2913	1828	1 085	144	130	14	91	15	76	788

Incidence

Model 1:

Among the 515,504 donations by first-time donors, 5,540 (1.07%) confirmed anti-HIV positive. Of these, 879 (16%) were classified as recent infections. The denominator consisted of PY contributed by 510,862 at-risk donors [excluding 4,642 HIV positive donors with longstanding infections]. The overall incidence using model 1 and the LAg assay to determine recency in first-time donors was 3.32 (CI 3.11, 3.55) per 1000 PY.

Table 3 shows incidence by year, province, gender, ethnicity and age using model 1. Incidence varied slightly by year, declining from 4.08 in 2013 to 2.90 per 1000 PY in 2016. The highest incidence was in the Mpumalanga and KwaZulu-Natal provinces and the lowest incidence was in the North West and Northern Cape provinces (Figure 1)

Table 3:

Incidence determined by three models

	Incidence (Lag FT)/1000 person years	Incidence (NAT yields)/1000 person years			Incidence (Repeat, classic)/1000 person years
	Model 1	Model 2			Model 3
		First time	All	Repeat
Overall			1.84 (1.65, 2.06)
Donor type
First time	3.32 (3.11, 3.55)	3.81 (3.07, 4.73)
Repeat				1.56 (1.37, 1.77)	1.94 (1.86–2.01)
Year
2012	3.48 (3.01, 4.02)	4.06 (2.53, 6.54)	1.82 (1.42, 2.35)	1.50 (1.11, 2.02)	1.74 (1.50–2.01)
2013	4.08 (3.59, 4.64)	5.68 (3.87, 8.34)	2.45 (1.97, 3.04)	1.93 (1.49, 2.51)	1.79 (1.55–2.07)
2014	3.05 (2.62, 3.55)	3.45 (2.08, 5.72)	1.68 (1.30, 2.18)	1.42 (1.05, 1.93)	1.75 (1.52–2.02)
2015	3.05 (2.62, 3.55)	3.25 (1.92, 5.48)	1.55 (1.19, 2.02)	1.31 (0.96, 1.78)	2.00 (1.75–2.29)
2016	2.90 (2.47, 3.41)	2.45 (1.32, 4.55)	1.73 (1.34, 2.23)	1.63 (1.23, 2.16)	1.63 (1.40–1.89)
Gender
Female	4.43 (4.11, 4.78)	5.52 (4.34, 7.03)	2.91 (2.54, 3.33)	2.40 (2.04, 2.82)	2.68 (2.55–2.82)
Male	1.94 (1.70, 2.20)	1.68 (1.03, 2.73)	0.86 (0.70, 1.07)	0.96 (0.77, 1.19)	1.29 (1.20–1.37)
Age
16–19	2.21 (1.96, 2.48)	2.20 (1.45, 3.35)	1.95 (1.51, 2.51)	1.82 (1.32, 2.51)	2.27 (2.07–2.50)
20–30	5.58 (5.08, 6.14)	7.49 (5.59, 10.04)	3.89 (3.33, 4.53)	3.27 (2.73, 3.92)	3.57 (3.37–3.79)
>30	2.87 (2.49, 3.30)	2.72 (1.64, 4.52)	0.93 (0.76, 1.15)	0.83 (0.66, 1.04)	1.07 (0.99, 1.15)
Female
16–19	3.51 (3.09, 3.99)	3.50 (2.23, 5.49)	3.29 (2.48, 4.36)	3.16 (2.20, 4.55)	3.51 (3.15–3.91)
20–30	6.94 (6.20, 7.77)	10.18 (7.31, 14.18)	5.68 (4.71, 6.84)	4.69 (3.74, 5.88)	4.61 (4.28–4.97)
>30	3.23 (2.71, 3.85)	3.88 (2.21, 6.84)	1.40 (1.07, 1.82)	1.19 (0.88, 1.60)	1.40 (1.28–1.55)
Male
16–19	0.65 (0.47, 0.90)	0.66 (0.21, 2.04)	0.70 (0.39, 1.26)	0.72 (0.36, 1.43)	1.05 (0.86–1.28)
20–30	3.77 (3.15, 4.50)	3.89 (2.10, 7.24)	2.33 (1.78, 3.06)	2.13 (1.57, 2.88)	2.58 (2.34–2.85)
>30	2.40 (1.90, 3.02)	1.24 (0.40, 3.85)	0.62 (0.45, 0.86)	0.59 (0.42, 0.83)	0.80 (0.72–0.90)
Race
White	0.19 (0.12, 0.32)	0.15 (0.02, 1.06)	0.13 (0.08, 0.23)	0.13 (0.08, 0.23)	0.20 (0.17–0.23)
Black	5.88 (5.50, 6.29)	6.71 (5.37, 8.39)	5.80 (5.16, 6.52)	5.52 (4.81, 6.33)	5.81 (5.57–6.06)
Asian	0.12 (0.03, 0.48)	0.65 (0.09, 4.61)	0.39 (0.16, 0.94)	0.36 (0.13, 0.95)	0.33 (0.24–0.46)
Coloured	1.38 (0.89, 2.13)	0.85 (0.12, 6.01)	1.10 (0.59, 2.04)	1.14 (0.59, 2.18)	1.67 (1.39–2.00)
Unknown	1.27 (0.68, 2.36)	3.15 (0.79, 12.58)	1.41 (0.46, 4.38)	0.67 (0.09, 4.77)	1.47 (0.96–2.23)
Province
Gauteng	2.61 (2.34, 2.92)	2.92 (2.02, 4.24)	1.45 (1.20, 1.74)	1.23 (0.99, 1.53)	1.55 (1.45–1.66)
KwaZulu Natal	4.42 (3.90, 5.02)	7.28 (5.15, 10.30)	2.91 (2.35, 3.60)	2.12 (1.62, 2.78)	2.47 (2.27–2.68)
Mpumalanga	5.92 (5.02, 6.97)	8.25 (5.05, 13.46)	2.77 (2.07, 3.70)	2.04 (1.43, 2.92)	2.81 (2.52–3.14)
Eastern Cape	2.98 (2.36, 3.75)	0.99 (0.25, 3.94)	1.85 (1.28, 2.66)	1.97 (1.26, 3.03)	2.34 (2.08–2.64)
Free State	4.29 (3.34, 5.51)	1.76 (0.44, 7.04)	1.94 (1.27, 2.94)	1.95 (1.26, 3.03)	2.00 (1.71–2.34)
North West	1.42 (0.90, 2.26)	0	0.83 (0.42, 1.67)	0.93 (0.47, 1.86)	0.84 (0.64–1.11)
Limpopo	2.61 (1.81, 3.75)	1.14 (0.16, 8.07)	1.37 (0.69, 2.75)	1.42 (0.67, 2.97)	2.28 (1.86–2.81)
Northern Cape	1.74 (0.99, 3.06)	1.83 (0.26, 12.98)	0.96 (0.40, 2.32)	0.86 (0.32, 2.30)	1.12 (0.83–1.52)

Figure 1:

Incidence in first-time donors determined using model 1

Model 2:

As the NAT yield method can be used to estimate incidence in all donations, first-time donations and repeat donations, we used it as a bridge to allow comparison to both model 1 and 3. In addition, we used this model to compare incidence in first-time and repeat-donors. Incidence, based on 312 NAT yields in all donations, was estimated at 1.84 (CI 1.65, 2.06) per 1000 PY. Using the 82 and 230 NAT yields from first-time and repeat-donors, incidence was estimated to be 2.44-fold higher in first-time donors at 3.81 (CI 3.07, 4.73) compared to 1.56 (CI 1.37, 1.77) per 1000 PY in repeat-donors.(Table 3).

Model 3:

The classic model to determine incidence included only donations from donors with at least two donations during the study period.

Among 3,661,699 whole blood donations by 602,329 repeating donors, 2425 tested confirmed HIV positive. Number of donations varied between 2 to 33 per donor. The median follow-up times, the amount of time during the study periods (used in the denominator) for HIV positive and negative donors were 383.5 and 636 days, respectively. Incidence estimated using repeating donors was 1.94 (CI 1.86–2.01) per 1000 PY (Table 3).

Comparisons between Models

When comparing the models to determine incidence in first-time donors, model 2 yielded an estimate that was 1.15-fold higher than model 1 but this difference was not statistically significant. When we extrapolated the incidence derived from repeat-donors using model 3 to first-time donors using the NAT yield ratio of 1:2.42, the estimated incidence in first-time donors was 4.69 (CI 4.51, 4.89) which was significantly higher (1.41-fold) than in model 1 but not statistically different from the results from Model 2.

Model 2 and Model 3 were used to estimate incidence in repeat-donors. Model 3 estimated incidence as 1.24--fold and significantly higher than in model 2. To investigate the reason for the higher estimates in model 3, we assessed the sensitivity of the incidence estimate on the inclusion of donors who donated twice only compared to donors who donated three or more times and found a 2-fold higher incidence in the former group. Thus, model 3 repeating donors included some donors, who were observed from their first to the second donation, when incidence is higher, whereas model 2 repeat donors only included second and subsequent donations, when the incidence is lower.

Being female, aged 20 to 30 years old, of black ethnicity and living in KwaZulu-Natal and Mpumalanga were associated with a higher incidence for all models.

Comparison to HIV Incidence in the General Population

Incidence was approximately 3.5 times higher in the general population compared to first-time donors (Figure 2). In the general population, incidence significantly decreased from 13.30 (CI 12.6, 13.9) in 2012 to 10.6 (CI 9.89, 11.3) per 1000 PY. A similar but not significant decrease, albeit lower, was seen in the first-time blood donors which declined from 3.48 (CI 3.01, 4.02) in 2012 to 2.90 (CI 2.47, 3.41) in 2016 (Table 3).

Figure 2:

Incidence in first-time donors using model 1 and 2 and incidence in repeat-donors using model 2 and 3 compared to the incidence in the general population[9]

Discussion

The incidence of a viral infection in blood donors is used to estimate the RR of a transfusion transmitted infection occurring as it establishes the number of new infections which may go undetected by the blood services screening strategy. South Africa has the largest burden of HIV in the world and although the pre-screening donor interview and testing of first-time donors removes the majority of the prevalent infections, a substantial number of incident HIV infections are detected in our donor population and hence we project an elevated RR of transfusion-transmitted HIV infection relative to most other countries even with application of ID-NAT.[3] Accurate measurement of HIV incidence and consequent RR is essential to monitor risk reduction strategies. In this study, we were able to test HIV positive samples from first-time donors to determine recent infection status using the LAg avidity assay and use the incident infections to compare the incidence derived by LAg with that derived by parallel testing using ID-NAT and the NAT yield WP model.[4] In addition we compared incidence in repeat-donors derived by the I/WP model[2] with the NAT yield WP model[4]. Our high-level findings demonstrate that incidence rates determined by several methods were broadly consistent which is reassuring, although small discrepancies should spur re-examination of accuracy and implications of differences between the models.

Among first-time donors, we found similar overall HIV incidence using the LAg avidity and NAT yield models, adding confidence that these two cross-sectional incidence approaches are comparable and that the LAg avidity approach has promise in that it yields significantly larger numbers of recent infection cases and hence more precise incidence estimates than the NAT yield method. However, in repeat-donors we estimated a significantly higher incidence using the classic model compared to the NAT yield model. Furthermore, we found an extrapolated incidence from repeat-donors to first-time donors using the classic model that was one and a half-fold higher than the LAg avidity and NAT yield models, suggesting that previous corrections using NAT yield ratios for derivation of first-time incidence may be overestimated.

We speculate additional reasons for the higher incidence derived by the classic method: 1) in our study the median follow-up time (the time during the study period that contributes to the denominator) of HIV positive repeating donors was nearly half the length of HIV negative donors, as negative donors were able to give multiple donations that were counted for the entire study period, compared to HIV positive donors who would have had truncated follow-up following detection of infection and deferral; 2) the inter-donation intervals for HIV-positive and HIV-negative donors were not random (data not shown), in part due to the pre-set periodic times of mobile clinics where 60% of the blood is collected; and 3) the classic method using repeating donors, included some first-time donors who donated a second donation during the study period. These newer donors may be at higher risk of having an incident infection compared to established repeat-donors. When we estimated incidence in donors who donated their second donation in the study and compared this to the incidence if they donated their 3^rd or later donation, the former had a 2-fold higher incidence giving credibility to this assumption (data not shown).

The findings using the classic method may have relevance to the issue of self-deferral during the earlier stages of infection by repeat-donors, which consequently results in over estimation of RR. This is because a repeating donor may appropriately self-defer from donating blood for the mandatory 3 months following high-risk behaviour; therefore although classified as an incident HIV infection using the classic model, such donors would not be contributing donations relevant to the RR as they self-deferred during the window period and would subsequently be detected using the current laboratory screening strategies. The lower incidence in repeat-donors based on the NAT yield WP approach is likely more accurate since detection of a NAT yield case is a direct measurement of a donor presenting to donate during the RNA positive pre-seroconversion stage of infection that immediately follows the undetectable pre-NAT infectious WP.

Incidence in first-time donors was more than twice that of repeat-donors. This difference may be due to confounding or selection bias. First-time donors are enriched with young, female and black individuals, subgroups with higher HIV incidence. In addition they have had no selection and education compared to repeat-donors who may self-defer from donating blood if their risk behaviour changed since their last negative donation. In contrast, first-time donors have not been educated on being a safe blood donor when they present for the first time and could donate blood for test seeking without the stigma of going to a testing clinic.

Like South Africa, many developed countries such as Canada, Australia, the USA and countries in Europe collect only a minority of their blood donations from first-time donors[1, 15–17]. For example, Canada collects 13% of their blood from first-time donors[15]. They calculated incidence using the I/WP model and repeat-donors and subsequently multiplied the proportion of first-time donors by 2 based on the hypothesis that incidence is double that in repeat-donors.[16]. Australia collect 6% of their blood from first-time donors and include only NAT yield first time donors in their incidence and RR estimates.[18]

Poorer countries however, tend to have more first-time donors. In Tehran 40% of the blood comes from first-time donors and these donations were excluded from their incidence calculations. [19] Zimbabwe similarly collects 44% of the blood from first-time donors[20], however they estimated RR in first-time donations by modifying the I/WP model; instead of using the inter-donation interval of a seroconverting repeat donor they substituted the prevalence and duration of asymptomatic WHO stage 1 and 2 of five years. [21] Recently a Brazilian study reported the prevalence, incidence and RR calculated using the I/WP model. They reported that 91% of donations came from first time donors and were therefore excluded from the incidence calculations. It would be interesting to know whether the HIV positives in the first-time donors were long standing infections (and therefore would have little impact on the RR) or recent infections (which would have an impact on the RR) [22].

There are some limitations to this study. The LAg avidity assay is known to falsely classify as “recent” approximately 60% of donors on anti-retroviral therapy (ART) and 13% of Elite controllers as recent infections[5, 7] due to a muted immune response especially if treatment is started early after acquisition[23, 24]. To mitigate this we excluded all donations that tested ID-NAT negative, anti-HIV positive - which is the test result pattern most commonly seen among HIV positive donors taking surreptitious ART [25]. However it is possible that some of our ID-NAT positive donors could be on ART with incomplete viral suppression and hence falsely classified as recent infections.[25] A second limitation is that we did not account for differences in inter-donation intervals of the seroconverting donors compared to the negative donors, non-random inter-donation intervals, and particularly longer inter-donation intervals in the HIV-positive donors in the immediate pre-seroconversion time-frame relative to their prior intervals which could bias the results of the I/WP model for repeat-donors and extrapolation of the results from that model to first-time donors using the NAT yield ratio[2]. Finally, although we performed LAg avidity testing on the HIV-concordant positive donations from repeat-donors, an analysis to compare these data with the I/WP model results was beyond the scope of this analysis. Such an analysis would require adjustment to the contribution of person-time to the denominators for repeat-donors with inter-donation intervals that are shorter than the MDRI for LAg assay, which is common at SANBS.

Conclusion:

The incidence model based upon LAg avidity testing of first-time donors provided similar incidence estimates when compared to the NAT yield WP model and substantially lower incidence compared to extrapolation of incidence in first-time donors from the classic method, supporting broader use of the LAg plus NAT yield based approach in measuring HIV incidence. We believe testing HIV positive donations from first-time donors using the LAg assay could provide blood establishments that do not perform NAT screening with a good tool to determine incidence in their first-time donor populations. By extrapolation, these findings also support use of new incidence assays in cross-sectional HIV surveys of the general population.