October 7th, 2019

Our HIV Testing Algorithm Has a Major Problem — Here’s How to Fix It

Mostly, HIV testing works great. It’s long been so accurate that we can strongly support HIV testing even in relatively low-risk people.

The 2014 revised lab testing guidelines made it even better, recommending a combined antigen/antibody screening test (called the 4th generation test), and replacing the Western blot with the HIV-1/2 differentiation immunoassay as the preferred confirmatory test.

However, if you’re a primary care, ID/HIV, or OB/GYN clinician who sends a lot of HIV tests, you’ve no doubt received the following perplexing result::

4th Generation Human Immunodeficiency Virus (HIV) Combination Antigen/Antibody:  Repeatedly Reactive

HIV-1/2 Antibody Differentiation Immunoassay:  Negative

Interpretation:  INDETERMINATE for HIV-1/2.

There are many problems with this outcome, which is common enough that it turns out to be the most common question about HIV testing by far. But three years have passed since I last addressed it, so let’s explore again and figure out how to fix it.

First, the reasons it’s so problematic:

  1. In patients at high risk for HIV, this is a critical result that should prompt rapid action. An indeterminate result with a positive HIV screen and negative HIV-1/2 differentiation assay could be acute HIV — patients may be in the “window” period before they have developed a sufficient antibody response. Why? It’s worth remembering that when we moved from the 3rd- to the 4th-generation screening test, we shortened the window period by adding p24 antigen to the screening test only. It’s now around 14 days after HIV acquisition. By contrast, the confirmatory differentiation test still only detects antibody. Window period for this test? A median of 33 days. And that difference may occur exactly when people with acute HIV are most symptomatic. Bad.
  2. In patients at low risk for HIV, this is a confusing result that almost never means anything and prompts only worry. Since we appropriately do a lot of HIV testing in low-risk people — especially pregnant women — the prior probability that they actually have acute HIV is extraordinarily low, especially if they are asymptomatic. The screening test’s 99.6% specificity is great (Lyme testing is jealous). But this means that around 4/1000 times, the HIV test will be positive even in a person without HIV — the prevalence of acute HIV in most screened populations is way lower than that. As a result, in many centers (including ours), most of the reactive HIV screens with negative confirmatory tests ultimately end up being false-positive screens.
  3. Sorting out what your patient has — acute HIV or a false-positive HIV screen — is straightforward medically, but far from easy from a logistical perspective. Sure, a subsequent negative quantitative HIV viral load means it’s a false positive, and a positive one (at pretty much any value) indicates acute HIV — simple to interpret medically. But try explaining that to a worried patient on the other end of your notification phone call, or telling them they have to come back in for another test to just be sure that your impression (whatever it may be) is correct. With so much testing done today as a matter of routine screening, we might end up having to do some of the risk assessment over the telephone, a distinctly uncomfortable scenario — especially tricky in an era when lots of our communication is via “patient portals” or some other non face-to-face method. Notably, some labs will “reflex” their indeterminate tests to a qualitative HIV nucleic acid test (NAT), but in discussion with my colleagues around the country, this is far from routine.
  4. Most medical centers can offer turnaround on a quantitative HIV viral load far faster than the HIV differentiation assay. This is particularly true if the screening test is done at your hospital’s lab, with reactive tests then sent to a reference laboratory. In our hospital, for example, when clinicians suspect acute HIV — patient may be at risk, or have symptoms — they often appropriately send an HIV viral load at the time of initial presentation. We then have this bizarre outcome where the HIV screen returns positive, the viral load comes back very high within a couple of days, and then — after a 3 or 4 day further delay — the differentiation assay return negative. In one such case, the HIV resistance genotype was also ordered at baseline, and it came back before the (negative) differentiation test!
  5. The tests are a nonsense word salad smothered with Thousand Island dressing. Read that result at the top in its entirety — what a bunch of gobbledygook. Look, we ID/HIV specialists know what these words mean, but ask some primary care clinicians or obstetricians, they will have no idea. Or ask them what it’s like to enter “HIV” in the order box of their electronic medical record. Chaos! (See image on the right — click on it for the full effect.) It’s completely understandable why they might be confused — this isn’t their specialty. So read the tests again, and pity the non-ID specialists: 4th Generation Human Immunodeficiency Virus (HIV) Combination Antigen/Antibody.  HIV-1/2 Antibody Differentiation Immunoassay. Good grief. How about just, “HIV screening test” and “HIV confirmatory test”?

How can we fix this? As usual when it comes to HIV testing, we turn to the top expert in this area, Dr. Bernie Branson. Fortunately for all of us, he just wrote a sensational review of HIV testing for Infectious Disease Clinics.

Here’s what Bernie says:

With current technology, it would be more efficient to reverse the current confirmatory testing sequence and, after a reactive Ag/Ab combination assay, perform a quantitative HIV-1 RNA viral load test to both confirm the diagnosis and contribute to immediate clinical management.

Of course! Confirm the reactive screen with an HIV-1 viral load! How sensible! This will give us the most critical result much more rapidly.

A detectable (at almost any level) HIV-1 viral load would prompt immediate HIV therapy, important for reducing HIV transmission, preserving immune function, and possibly providing our best candidates for HIV cure strategies when they become available.

With a negative HIV-1 viral load, the HIV-1/2 antibody assay (done next) would pick up the small proportion of those who are elite controllers, or are on ART and haven’t told you (HIV-1 antibody positive, viral load negative); have HIV-2 (HIV-2 antibody positive); or just had a false-positive HIV screen (HIV-1/2 antibody negative).

And those worried about a (slight) delay in diagnosis for HIV-2? This will occur rarely — with the prevalence of HIV-2 in this country exceedingly low, a CDC scientist recently described the routine use of this test as “High Burden, Low Yield”. 

So what are the barriers to making this change a reality?

  • There is no FDA-approved quantitative viral load test for HIV diagnosis. There is a qualitative HIV viral load test approved, but it is not widely available, requires plasma (see below), and a positive result would still warrant doing the quantitative test. According to Bernie’s paper, some are available internationally. Time to bring them here!
  • Many labs draw serum, not plasma, when doing HIV testing. Available quantitative viral load tests have been validated on plasma only. This makes the “reflex” strategy cited above in item #3 impractical. Apparently, serum-based quantitative viral load tests are also available in other countries.
  • Viral load testing is much more expensive. Tests considered “molecular” by payers cost a ton, even though the technology to do them has dramatically improved and become more efficient. This progress should theoretically make them more affordable, but some labs may be reluctant to give up the high price and revenues.

The above issues are not insurmountable barriers, but they will take time to overcome. But as an optimist by nature, I’m convinced this will happen — it just makes too much sense. And look, you apparently agree:

Until then, we’re stuck with this strange situation where the “confirmatory” test ironically may not confirm anything at all — either during acute HIV, or when the screening test is a false positive.

Fortunately, it doesn’t have to be this way — time to make the change!

12 Responses to “Our HIV Testing Algorithm Has a Major Problem — Here’s How to Fix It”

  1. Robert Cannon says:

    I believe you mean ‘prevalence’ rather than ‘incidence’ when you wrote “… the incidence of acute HIV in most screened populations …”.

  2. Malika Mohabeer Hart says:

    Great article, and full of sense. Thank you.
    The same situation is the the UK and with some developing countries still using the Western Blot for confirmation.
    Time to move forward with highly performing molecular tests.

  3. Dawd Siraj says:

    Great review as always Paul!!
    If changed, it will take away one of my most frequently asked questions! I always look smart answering this question!!

  4. Dr. Sax,

    A really great post! I want to highlight a point you have missed in favor of our current testing algorithm and that is the ability to diagnose HIV immediately through rapid/rapid testing. In Louisiana, our health department has endorsed a point of care 4th gen Ab/Ag followed by a point of care 3rd gen test for confirmation.

    This allows our testing programs to provide immediate results and, most importantly, start same-day ART or same-day PrEP. The POC testing is also essential for community-based testing and allows us to interact with folks who, otherwise, would not have engaged with the medical system.

    Of course, we still struggle with discordant results. But I vote for the traditional testing algorithm until we have an HIV PCR POC test.

    • Paul Sax says:

      Interesting approach, modeled on what’s done internationally when a second rapid test is performed to confirm the first. As you note, still not perfect — acute HIV will be missed by that 3rd generation test (which is why we went to 4th), and false-positive 4th gen screens (which are now more common given addition of p24 antigen) still don’t get resolved.

      Sigh. Let’s make this better today with a PCR!


  5. Kristy Morman, PA-C says:

    Excellent article! I am in primary care, and frequently screen for HIV. Our EMR also has tons of word salad test names. I appreciate the clarity offered in this article. Very helpful and, as always, with a touch of humor. Thanks Dr. Sax!

  6. Dr Srinivasan Ramesh says:

    Dear Dr Sax I thank you for a succint and illuminating article on HIV testing.In India a positive Western Blot test is mandatory before confirming a diagnosis of HIV infection.This is a legal requirement.It is not easy to convince other non ID/HIV specialists of the intricacies of the various tests .Dr Branson’s review in the Infectious diseases Clinics was very helpful too.One question remains .Why is Nucleic acid testing for HIV 1 and 2 not made a part of the diagnostic algorithm in high risk exposures to shorten the period to diagnosis and treatment.

  7. Aisha Khan says:

    Great article ! I agree, it would be nive if 4th gen test could be paired with reflex to PCR! I have had instance when I suspected acute HIV, I ordered both 4th gen and PCR…the PCR came back positive 3 days earlier than the differentiation test. Was able to start patient on ART before the confirmatory test even came back!

  8. Maryellen Cooley says:

    Hello Dr. Sax,
    Thank you for this update to your great article three years ago, which we shared broadly among local providers who were understandably perplexed by the new HIV algorithm.
    We do hear of a fair number of false 4G screenings at our LHJ, phone calls from providers asking if they need to report them, and from panicked patients.
    Very often these are ob patients, where a timely, reassuring RNA reflex would make all the difference.
    My question regards the window period for 4G testing. My understanding is that the POC 4G tests have a longer period than the lab-based 4G, by potentially a couple of weeks. I have heard that the rapids won’t pick up P24ag unless viraemia is around a million copies. Some of our local labs and many of our clinics use the POC test for screening, regardless of pt’s risk for HIV.
    Can you offer some clarity on POC vs Lab-based 4g window?
    Thank you!

  9. Jonathan Blum says:

    The ridiculous smorgasbord of tests on that EPIC screen shot almost guarantees a high rate of inappropriate ordering. This is a fixable problem. Your IT department cannot control every aspect of test display, but it can choose the names and what is shown. In my system, typing in “HIV” gets you three tests – Ab/Ag, quantitative RNA, and qualitative RNA. Genotype and rapid HIV test require one additional click to bring up. If your EPIC display looks like the one shown here, you should talk to IT about getting it streamlined.

    This issue comes up with a lot of tests. While it would be nice to have more decision support, I have seen that simply getting the correct tests on (and the wrong ones off) the menus and order sets of the various specialists goes a long way toward encouraging proper ordering. Unfortunately, it is a lot of work to clean up these menus, so it takes a dedicated group to get this done.

  10. Dana says:

    I just came across this forum in search of something to help me set my mind at ease. I am one of those pts that’s low risk US born with a – HIV1 and an indeterminate HIV2. I just retested today and will need to wait 48hrs for 2nd results. I am really believing the initial test to be false, but it’s torture not knowing for sure. I’m not seeking advise I’m just trying to release this energy.

HIV Information: Author Paul Sax, M.D.

Paul E. Sax, MD

Contributing Editor

NEJM Journal Watch
Infectious Diseases

Biography | Disclosures | Summaries

Learn more about HIV and ID Observations.