Michael Tang, MD
Infectious Disease Fellow
Division of Infectious Diseases and Global Public Health
Department of Medicine
University of California, San Diego
1. HIV & Global Health Rounds
The UC San Diego AntiViral Research Center sponsors weekly
presentations by infectious disease and global public health clinicians,
physicians, and researchers. The goal of these presentations is to
provide the most current research, clinical practices, and trends in HIV,
HBV, HCV, TB, and other infectious diseases of global significance.
The slides from the HIV & Global Health Rounds presentation that you
are about to view are intended for the educational purposes of our
audience. They may not be used for other purposes without the
presenter’s express permission.
4. Overview
• Why is incidence important?
• Methods for incidence testing
• Repeated testing & Serial Prevalence
• CD4 Depletion Model
• Avidity Testing
• Viral Diversity
• Comparisons within the Primary Infection Resource Consortium
(PIRC) group
• Future directions
5. Definitions
• Acute HIV infection: earliest phase of HIV infection
• Prior to seroconversion generally depending on positive p24 antigen and
positive NAT test
• Duration depends on the diagnostic assay used
• Early HIV infection: ≤6 months from infection
• Recent HIV infection ≤12 months from infection
• Prevalent HIV infection: >12 months in duration
• Incidence: number of new infections (both diagnosed and
undiagnosed) in a given period of time
9. Ending the HIV Epidemic
• More than 50% of new HIV
diagnoses occurred in 48
counties, DC and San Juan,
Puerto Rico.
• These high burden jurisdictions
are charged with leading the
charge in decreasing HIV
incidence.
10. Why is Incidence important?
• Generally it is our outcome measure of choice for interventional
studies
• Both the 5 year and 10 year goal target new infections or incidence as the
outcome
• Identifying infections as incident or prevalent allows appropriate monitoring
HIV prevention efforts and allocation of resources.
• Clinical management includes rapid implementation of ART,
independent of CD4 or stage of HIV
• Diagnosis numbers can lag behind incidence and are influenced by
testing rates and diagnosis delays
11. Methods of measuring incidence
• Cohort studies
• Generally considered the gold standard
• Serial Prevalence
• CD4 Depletion Model
• Currently used by the CDC
• Avidity Testing
• Viral Diversity
12. Cohort study
• Longitudinal follow up of cohorts without HIV infection and frequent
retesting to determine when people acquire a new infection.
• Expensive, logistically difficult and time consuming
• Monitoring migration in and out of populations can be difficult
• These populations are not always representative of the overall
population we want to monitor
• Participants often receive risk reduction counseling, etc.
13. Example of Cohort study
• The Africa Health Research
Institute performs
surveillance 2x/year on 2
sites within the rural
KwaZulu-Natal region
• About 65-75% of eligible HIV
negative persons
participated in surveillance.
Vandormael et al. The state of the HIV epidemic in rural KwaZulu-Natal, South Africa: a novel application of disease metrics to assess
trajectories and highlight areas for intervention. International Journal of Epidemiology. January 2020:dyz269
14. Serial Prevalence
• Involves measuring prevalence of diagnosis at 2 time points, and
estimating incidence
• Uses the formula to calculate incidence
• Requires knowledge of mortality rates, migration. Also noted to be
expensive to perform, and unlikely to provide good information about
current trends in HIV incidence
• Used in the late 1980s when estimating incidence in the active duty
military and blood donors to set minimum estimates of HIV incidence
in general population
Karon et al. HIV prevalence estimates and AIDS Case Projections for the US. MMWR Nov 1990
Brookmeyer R. Measuring the HIV/AIDS Epidemic: Approaches and Challenges. Epidemiologic Reviews. April 2010
15. CD4 Depletion Model
• In 2011, Dr. Lodi and the CASCADE collaboration developed a CD4
depletion model to estimate duration from seroconversion to various
CD4 cut offs
• Utilized 25 cohorts of individuals with well-estimated dates of
seroconversion from Europe, Australia, Canada and sub-Saharan
Africa, with ~18,500 individuals included in the analysis
• Used a linear mixed model effect to estimate time to different CD4
cut offs from infection
• Rick Song, and the CDC used this depletion model to extrapolate
duration of infection from the initial CD4 count
16. Satcher Johnson, Anna & Ruiguang Song. (2017,
October 3). Estimating HIV incidence, prevalence,
and undiagnosed infection in the United States
[Webinar]. In Prevention Science and Methodology
Group Virtual Grand Rounds.
18. CD4 Depletion Model
• Using the calculated diagnosis date from above, able to determine a
diagnosis delay distribution and estimate % of infections from each
year that are still undiagnosed
• Diagnosis delay distribution is used to calculate population incidence
• Also able to calculate prevalence from cumulative incidence and
cumulative mortality rates
19. CD4 model – Advantages and Disadvantages
• Convenient way to estimate
incidence, prevalence, %
undiagnosed
• Inexpensive
• Relies on accuracy of CD4
depletion model (developed in
2011)
• Known issues with acute/early
infections, but felt that since these
are a small proportion
• Given changes in our treatment of
CD4, unlikely to be able to
improve accuracy of CD4
depletion model
20. Biomarker/Avidity testing
• Multiple types of assays, but the general principle is the same:
proportion of HIV specific antibody binding increases as time from
infection increases
• Values below a cutoff indicate a recent infection
• Important test characteristics include Mean Duration of Recent
Infection (MDRI) and False Recency Rate (FRR).
• Goal for MDRI is 1 year, although the longer that is possible, the more useful
the assay is.
• Goal for FRR is 0%, but generally shoot for <2%.
21. WHO Working Group on HIV Incidence Measurement Data Use Group. Boston, MA, USA: World Health Organization; 2018
22. Biomarker/Avidity testing - Limitations
• Misclassification: elite controllers, suppressed on ART, low CD4
(although less of an issue with newer assays)
• MDRI and FRR have some variability depending on the population
• Each of the recency tests were created with subtype B in mind (other
than BED), and characteristics of test changes with different subtypes
23. Effect of HIV-1 Subtype on Biomarker Assays
• Both MDRI and FRR will change as a result of different biomarker assays.
Kassanjee R, et al. Independent assessment of candidate HIV incidence assays on specimens in the CEPHIA repository: AIDS. 2014;28(16):2439-2449.
24. Viral diversity
• HIV infection occurs secondary to a single virus with viral diversity
increasing during infection
• Able to use genotypic resistance tests to evaluate viral diversity
• Generally, these are done via bulk sequencing. Nucleotides are
reported out at each position when >80% consensus is present. If not,
position is labeled ambiguous.
• The percentage of these ambiguous positions increases as HIV
infection progresses
• Based on Swiss cohort, cut off of 0.5% ambiguity was determined
with 87% sensitivity, and 70% specificity for <1 year of infection
Kouyos RD et al. Ambiguous Nucleotide Calls From Population-based Sequencing of HIV-1 are a Marker for Viral Diversity and the Age of Infection. Clinical
Infectious Diseases. 2011;52(4):532-539. doi:10.1093/cid/ciq164
25. Viral Diversity – Limitations
• Potential confounding with super infection
• Unable to use for people on or previously on ART, or those who a
genotype cannot be successfully performed.
• Costly to perform unless genotyping is needed
26. SD Primary Infection Resource Consortium
• 834 ART naïve, newly diagnosed HIV-1 individuals enrolled between
June 1996-July 2019
• Clinical data, including CD4+ counts were collected at baseline (Day
0), weeks 4, 12, 24, and every 24 weeks thereafter
• Persons with acute infection had additional CD4+ and VL measures at
weeks 2 and 8
28. PIRC Algorithm
• Acute – NAT positive, WB neg/indeterminate
• Early infection – stratified based on recency assay level
• Early infection – stratified based on last negative HIV test
• Prevalent infections – Positive WB, no previous test within 365 days
29. Goals
• Overall concerns – the CD4 depletion model may work well enough in
large population sizes (entire US population), concerned about its
ability to monitor changes at county level where there is only 200-
2000 new diagnoses each year
• Compare PIRC algorithm with the CD4 depletion model for
determining duration of infection
• Look at overall population as well as subpopulation of prevalent cases
• For prevalent cases, utilize patients who had longitudinal follow up >1 year
from their estimated date of infection prior to starting ART
• Determine if there is value in viral diversity as a marker for recency
30. Number of Participants 702
Gender Identity
Male sex - no. (%) 676 (96%)
Female sex - no. (%) 24 (3%)
Trans Male to Female - no. (%) 2 (0%)
Ethnicity - no. (%)
White 389 (57%)
Black 39 (6%)
Hispanic 205 (30%)
Other 23 (3%)
Age at presentation (years)
Median (IQR) 32.5 (26-40)
VL at entry
Median log10 copies/mL (IQR) 4.95 (4.11-5.69)
Pre-ART Peak VL
Median log10 copies/mL (IQR) 5.39 (4.74-5.94)
CD4 count at entry
Median cells/mm3 (IQR) 505 (378-652)
Duration of follow up without
treatment
Median days (IQR) 50 (12-232)
Transmission Risk Factor no. (%)
MSM 625 (89%)
MSM+IDU 27 (4%)
IDU 3 (0%)
Hetero Male 21 (3%)
Other/Unknown 26 (4%)
39. Summary of findings
• CD4 depletion model performs poorly in estimating duration of
infection for both recent and prevalent cases
• When simplified to a binary evaluation (incident vs prevalent), sensitivity and
specificity were both poor
• Viral diversity has some value in identifying incident infections, but
has a high FRR (>40%), and relatively low concordance with the PIRC
algorithm (74%)
40. Limitations
• Cohort of primarily recent infections
• No method to validate the PIRC method of incidence (no gold
standard)
• None of the recency assays are FDA approved – cannot be used
outside of research studies
• Recency assays are strain specific
• Migration of people in and out of the population
41. Next steps
• Feasibility of implementation
• Low hanging fruit: Collecting data on acute infections and negative tests
• Use of recency testing on remnant samples
• Further development of recency tests to reach benchmarks
• Evaluation of multi-assay algorithms to decrease the FRR
• Look at CD4 response to ART as a potential predictor of incidence
42. Acknowledgements
• Susan Little
• Sanjay Mehta
• Christy Anderson
• Felix Torres
• Participants in research studies
• AVRC staff
• ID program support
Say that we generally use these, but they were not established in any particular way – they are convention based.
Not only is it the outcome of interest, but each of the 48 targeted jurisdictions are charged with meeting these goals, and if we cannot objectively measure this outcome, particularly on the small scale (200-2000 new diagnoses each year)
For prevention studies such as PrEP, Test and Treat, etc
particularly important for a border city such as San Diego
We can see these changes that have occurred with scale up of prevention methods within South Africa
Scale up for Voluntary male circumcision and ART services between 2009-2011. Top is Women, bottom is men
R is relative mortality rate
Delta is difference in time.
Consider cutting as more of a population method
Purpose of Lodi’s study was to identify impact and cost effectiveness in changes in CD4 initiation thresholds.
SQRT of CD4 vs time yielded a linear model.
Please note that this is an example
Assume a linear decline in number/proportion of new diagnoses from a prior year’s infection over time.
None of the currently available assays are able to meet both these criteria
LAg, Bed, vitros, biorad
Talk about subtype dependence.
Examples of the different tests current available for Research/epidemiology use. None of these reach the benchmarks we previously discussed
Roger Kouyos and Swiss Cohort in 2011
Important to note that this is not just some random part of the sequence, but it’s from pol. Different parts of the sequence have variable rates of sequence variability
Feibig Staging
NAT – nucleic acid tests
WB – Western Blot
Paired T test
Also holds true with Wilcoxan ranked test
Should talk about aging process here
Interesting that there’s a normal distribution around 0
Note that patients were not double counted (so if used in the chronic group, not used in the incident group)
Added back in E3s here. May need to try to re-create combined table to double check my work
Note that still have fairly high prevalence. This leads to a pretty good PPV
Note that there are more patients here – this is because we could use E3 cases in this analysis. All analysis done from the date of the viral sequence
FRR would be 53/130=40%