Validating marker-based incidence estimates in repeatedly screened populations

Assessment of BED HIV-1 incidence assay in seroconverter cohorts: effect of individuals with long-term infection and importance of stable incidence

Background

Performance of the BED assay in estimating HIV-1 incidence has previously been evaluated by using longitudinal specimens from persons with incident HIV infections, but questions remain about its accuracy. We sought to assess its performance in three longitudinal cohorts from Thailand where HIV-1 CRF01_AE and subtype B′ dominate the epidemic.

Design

BED testing was conducted in two longitudinal cohorts with only incident infections (a military conscript cohort and an injection drug user cohort) and in one longitudinal cohort (an HIV-1 vaccine efficacy trial cohort) that also included long-term infections.

Methods

Incidence estimates were generated conventionally (based on the number of annual serocoversions) and by using BED test results in the three cohorts. Adjusted incidence was calculated where appropriate.

Results

For each longitudinal cohort the BED incidence estimates and the conventional incidence estimates were similar when only newly infected persons were tested, whether infected with CRF01_AE or subtype B′. When the analysis included persons with long-term infections (to mimic a true cross-sectional cohort), BED incidence estimates were higher, although not significantly, than the conventional incidence estimates. After adjustment, the BED incidence estimates were closer to the conventional incidence estimates. When the conventional incidence varied over time, as in the early phase of the injection drug user cohort, the difference between the two estimates increased, but not significantly.

Conclusions

Evaluation of the performance of incidence assays requires the inclusion of a substantial number of cohort-derived specimens from individuals with long-term HIV infection and, ideally, the use of cohorts in which incidence remained stable. Appropriate adjustments of the BED incidence estimates generate estimates similar to those generated conventionally.

Interpopulation variation in HIV testing promptness may introduce bias in HIV incidence estimates using the serologic testing algorithm for recent HIV seroconversion

OBJECTIVES

The serologic testing algorithm for recent HIV seroconversion (STARHS) calculates incidence using the proportion of testers who produce a level of HIV antibody high enough to be detected by ELISA but low enough to suggest recent infection. The validity of STARHS relies on independence between dates of HIV infection and dates of antibody testing. When subjects choose the time of their own test, testing may be motivated by risky behaviour or symptoms of infection and the criterion may not be met. This analysis was conducted to ascertain whether estimates of incidence derived using STARHS were consistent with estimates derived using a method more robust against motivated testing.

METHODS

A cohort-based incidence estimator and two STARHS methods were applied to identical populations (n=3821) tested for HIV antibody at publicly funded sites in Seattle. Overall seroincidence estimates, demographically stratified estimates and incidence rate ratios were compared across methods. The proportion of low-antibody testers among HIV-infected individuals was compared with the proportion expected given their testing histories.

RESULTS

STARHS estimates generally exceeded cohort-based estimates. Incidence ratios derived using STARHS between demographic strata were not consistent across methods. The proportion of HIV-infected individuals with lower antibody levels exceeded that which would be expected under independence between infection and testing.

CONCLUSIONS

Incidence estimates and incidence rate ratios derived using methods that rely on the changing antibody level over the course of HIV infection may be vulnerable to bias when applied to populations who choose the time of their own testing.

Accuracy of serological assays for detection of recent infection with HIV and estimation of population incidence: a systematic review

We systematically reviewed the accuracy of serological tests for recent infections with HIV that have become widely used for measuring population patterns incidence of HIV. Across 13 different assays, sensitivity to detect recent infections ranged from 42-100% (median 89%). Specificity for detecting established infections was between 49.5% and 100% (median 86.8%) and was higher for infections of durations longer than 1 year (median 98%, range 31.5-100.0). For four different assays, comparisons were made between assay-derived population incidence estimates and a reference incidence estimate. The median percentage difference between the assay-derived incidence and reference incidence was 26.0%. Serological assays have reasonable sensitivity for the detection of recent infection with HIV, but are vulnerable to misclassifying established infections as recent-potentially leading to biases in incidence estimates. This conclusion is highly qualified by the apparent absence of a standardised approach to assay evaluation. There is an urgent need for an internationally agreed framework for evaluating and comparing these tests.

Prevalence and incidence of hepatitis C in injecting drug users attending genitourinary medicine clinics

Surveillance reports and prevalence studies have indicated that injecting drug users (IDUs) contribute more to the hepatitis C epidemic in the United Kingdom than any other risk group. Information on both the prevalence and incidence of hepatitis C in IDUs is therefore essential to understanding the epidemiology of this infection. The prevalence of hepatitis C in specimens from the Unlinked Anonymous Prevalence Monitoring Programme collected in 1995, 1996, 1998, 1999, 2000, and 2001 was determined using residual syphilis serology specimens from IDUs attending 15 genitourinary medicine (GUM) clinics in and outside London. These specimens were tested for antibodies to hepatitis C virus (anti-HCV). Using this cross-sectional design, anti-HCV-negative specimens were tested for HCV RNA to identify incident infections during the 'window' period of infection, and thus to estimate HCV incidence. Results of the multivariable analysis showed that there was marked variation in prevalence by clinic (P<0.0001) and age (P<0.0001). Overall the majority of infections were in males and the overall prevalence in injectors declined over the study period from 36.9% to 28.7%. The annual incidence in these injectors was estimated as being 3.01% (95% CI 1.25-6.73). Over the study period HCV incidence decreased by 1.2% per year. Genotyping of the incident infections identified the most common genotype as type 1 with type 3 being more frequently seen after 1998. Of the prevalent infections, genotype 1 was the most common. The study has confirmed a higher prevalence of anti-HCV in IDUs in the London area compared to those outside London. How representative of the current injecting drug user population are IDUs attending GUM clinics is unclear. Even so, such studies allow prevalence and incidence to be estimated in individuals who have ever injected drugs and inform ongoing public health surveillance.

Diagnosis of acute hepatitis C virus infection and estimated incidence in low- and high-risk English populations

The diagnosis of acute hepatitis C virus (HCV) infection is not straightforward; few people exhibit clinical symptoms and genome/antigen detection techniques do not indicate when infection had occurred. Here, a strategy to detect HCV RNA in the absence of antibody ('window-period') for diagnosis of acute infection is assessed. The sentinel surveillance of hepatitis testing study was used to retrospectively identify anti-HCV negative samples from high-risk individuals (2002-2003), for testing singly for HCV RNA. Additional samples were identified prospectively (2005) and tested in pools for HCV RNA. Positive samples were genotyped. Incidence and costs of adopting the pooling strategy were estimated. In the retrospective study, 8/390 (2.1%) samples were confirmed HCV RNA positive, anti-HCV negative. Prospectively, 3237 samples were tested in 325 pools. Five positive pools identified four confirmed HCV RNA positive patients (one false positive). Estimated incidence was 12.9 per 100 person-years in injecting drug users (IDUs) (retrospective study) and 3.7 per 100 person-years among drug/alcohol services and prison attendees (prospective study). Estimated costs were pound 850 per positive sample, in areas of higher risk. The yield from a window-period strategy depends upon the population tested. Pooled HCV RNA testing of anti-HCV negative samples from the current IDUs is realistic and relatively inexpensive to identify recently infected individuals.

Uma metodologia para estimar a incidência a partir de prevalências obtidas em Pesquisas Nacionais por Amostra de Domicílios: o exemplo da alfabetização

Conhecer a incidência de algum evento por idade e tempo é sem dúvida um dos maiores interesses da epidemiologia. Devido à relação matemática que existe entre as medidas de incidência e prevalência, alguns métodos foram desenvolvidos com o objetivo de obter a incidência a partir da prevalência. Este estudo descreve, de forma tutorial, um método que fornece estimativas de incidências por idade e tempo desde que existam prevalências repetidas de um evento, que a ocorrência do mesmo esteja relacionada à idade do indivíduo e que este possua características de uma doença crônica. O estudo foi realizado para a alfabetização da população brasileira, utilizando as prevalências por idade obtidas na Pesquisa Nacional por Amostra de Domicílios (PNAD) 1981, 1985, 1989, 1992, 1995 e 2001, e permitiu recuperar as estimativas da incidência de alfabetização nas idades entre 5 e 20 anos entre os anos de 1960 a 2000. O método ajusta os parâmetros de incidência para idade e tempo com o auxílio de um modelo linear generalizado para dados categóricos e utiliza o algoritmo EM para obter os dados não observados.

Human immunodeficiency virus type 1 incidence among blood donors in France, 1992 through 2006: use of an immunoassay to identify recent infections

BACKGROUND

In France, blood donations found to be positive for the presence of human immunodeficiency virus type 1 (HIV-1) are further tested to detect recent infections (< or =180 days) using an enzyme immunoassay (EIA-RI) developed in 2002. The characteristics of recently infected donors, estimates of HIV-1 incidence, and the residual risk of transfusion-transmitted HIV-1 are presented, in both first-time and repeat donors.

STUDY DESIGN AND METHODS

Of the 1027 donations found to be HIV-1-positive between 1992 and 2006, a total of 459 could be retrospectively tested with the EIA-RI. Multivariate analysis was performed to determine the donor characteristics associated with recent infection. Incidence rates and residual risk obtained with the EIA-RI were compared to classical cohort estimates derived from repeat donor histories.

RESULTS

Of the 459 HIV-1-positive donors studied, 105 (22.9%; 95% confidence interval [CI], 19.2-27.0) were identified as recently infected. Factors independently associated with recent infection were repeat donor status (adjusted odds ratio [AOR], 4.0; 95% CI, 2.4-6.9) and non-B subtypes (AOR, 2.0; 95% CI, 1.2-3.6). Incidence decreased from 4.3 (95% CI, 1.9-9.4) in 1992 through 1994 to 1.3 (95% CI, 0.6-2.8) per 10(5) in 2004 through 2006 in first-time donors and from 3.2 (95% CI, 2.0-5.0) to 0.8 (95% CI, 0.4-1.4) per 10(5) in repeat donors. Incidence and residual risk estimates were similar to those obtained with the classical cohort method.

CONCLUSION

This study suggests that the EIA-RI can be used to estimate HIV-1 incidence in a population with low HIV incidence. The estimated HIV-1 incidence in the blood donor population confirms the extremely low risk (1 in 3,350,000 donations) of HIV-infected blood donations entering the blood supply in France.

Comparison of three methods to measure HIV incidence among persons seeking voluntary, anonymous counseling and testing

OBJECTIVES

The authors compared 3 practical methods to estimate human HIV incidence rates using existing data from persons seeking anonymous testing in San Francisco between 1996-2002. Each method was assessed for strengths and limitations.

METHODS

Three different approaches were used to determine HIV incidence: one based on self-reported dates of prior tests, one based on linking records of prior tests using an anonymous unique testing code, and one based on the Serologic Testing Algorithm for Recent HIV Seroconversion (STARHS).

RESULTS

The 3 methods found comparable rates of seroconversion overall (1.0, 1.2, and 1.3 per 100 person-years) and among men who have sex with men (1.4, 1.6, and 2.0 per 100 person-years). Incidence for all 3 methods saw a peak during 1999 followed by a decline. Greatest variability of incidence was observed among lower-risk populations, in whom few infections were expected.

CONCLUSIONS

The 3 methods had complementary strengths and limitations, which may prevent proper interpretation of HIV incidence if any one method is analyzed alone. HIV incidence rates among persons seeking HIV testing should be interpreted cautiously using corroborative data on risk behavior and sexually transmitted diseases and other contextual information.

A new strategy for estimating risks of transfusion-transmitted viral infections based on rates of detection of recently infected donors

BACKGROUND

Estimates for human immunodeficiency virus (HIV)-1 and hepatitis C virus (HCV) transfusion-transmitted risks have relied on incidence derived from repeat donor histories and imprecise estimates for infectious, preseroconversion window periods (WPs).

STUDY DESIGN AND METHODS

By use of novel approaches, WPs were estimated by back-extrapolation of acute viral replication dynamics. Incidence was derived from the yield of viremic, antibody-negative donations detected by routine minipool nucleic acid testing (MP-NAT) of 37 million US donations (1999-2002) or from sensitive/less-sensitive HIV-1 enzyme immunoassay (S/LS-EIA) results for seropositive samples from 6.5 million donations (1999). Incidences and WPs were combined to calculate risks and project yield of individual donation (ID)-NAT.

RESULTS

The HIV-1 WP from presumed infectivity (1 copy/20 mL) to ID-NAT detection was estimated at 5.6 days, and the periods from ID to MP-NAT detection and from MP-NAT to p24 detection at 3.4 and 6.0 days, respectively; corresponding estimates for HCV were 4.9, 2.5, and 50.9 days (the latter represents period from MP-NAT to HCV antibody detection). The HIV-1 incidence projected from MP-NAT yield or from S/LS-EIA data was 1.8 per 100,000 person-years, resulting in a corresponding HIV-1 transfusion-transmitted risk of 1 in 2.3 million. The HCV incidence from MP-NAT yield was 2.70 per 100,000 person-years with a corresponding risk of 1 in 1.8 million donations. Conversion from MP-NAT to ID-NAT was projected to detect two to three additional HIV-1 and HCV infectious units annually.

CONCLUSIONS

MP-NAT yield and S/LS-EIA rates can accurately project transfusion risks. HCV and HIV-1 risks, currently estimated at 1 per 2 million units, could be reduced to 1 in 3 to 4 million units by ID-NAT screening.

What a test for recent infection might reveal about HIV incidence in England and Wales

BACKGROUND

A laboratory method has been developed that detects recent HIV infection and allows incidence to be estimated by testing single stored antibody-positive specimens. A theoretical exploration of the method's surveillance utility was carried out.

METHODS

Using various data sources, HIV incidence rates were postulated. The confidence intervals (CI) for these postulated incidences were calculated using the expected number of recent infections for each postulated incidence, the actual number tested for HIV, and the known number of HIV-1 positives. A test for trend was used to determine when an important change in incidence could be recognized.

RESULTS

If the incidence was 5% per annum (p.a.) in homosexual/bisexual men attending sexually transmitted diseases (STD) clinics in London, 64 recent infections would be expected in the 392 HIV-seropositive specimens and, if observed, would result in a 95% CI of 3.1-7.9% p.a. for the incidence rate. An incidence of 1% p.a. in pregnant women would be most unlikely as this would require detection of 193 recent infections, 26 more than the total 167 HIV-seropositive specimens found in 1997. In African women attending STD clinics in London, 30% of prevalent infections would be classified as recent if the incidence was 5% p.a. Further, if the incidence in homosexual/bisexual men were to fall by 50% over 3 years, a decrease of this magnitude would be recognized as significant within 2 years.

CONCLUSIONS

The detuned assay will increase the information from HIV serosurveys even where prevalence and incidence are relatively low. Existing surveillance systems should be redesigned to take full advantage of the method.