Bad blood?

Blood donation – Image by ANKAWÜ via Wikimedia Commons – CC-BY-SA-3.0

As regular as clockwork, at this time of year, articles appear about the Australian Red Cross Blood Service and their desperate need for blood donations across the holiday period.  If you’re eligible, please consider giving.  If you move in the circles I do, then shortly after this comes the call to open donations up for men who have sex with men.

 

I saw this excellent article by Nic Holas (twitter) in the Sydney Morning Herald just before Christmas, talking about the issue of whether or not MSM should be able to give blood.  It’s a controversial topic – and I think Nic’s article was excellent.

It’s something I’ve discussed a few times at work, and have often meant to blog about, so thanks to Nic for giving me the kick up the bum to actually do it.

 

Warning:  Post contains medical-school style evidence-based medicine, 2×2 tables, generous use of Fermi estimations, and all my personal opinions which are nothing to do with anyone I work with or for and are not medical advice etc etc.

So.  I’m going to leave aside the ethical question of whether or not gay men should be able to donate, and just focus on the risk-management and the maths.  But first, for those of you that haven’t done med school EBM, some definitions:

Sensitivity (of a test):  The likelihood a test will correctly detect the thing that you’re testing for (ie: in this context, how likely is a HIV test to be positive, assuming the donor has HIV. Another way of expressing it is 1-(false negative rate) )

Specificity (of a test): The likelihood a positive result from a test truly represents a positive result (ie:  if the test is positive, does the donor really have HIV – or another way, it’s 1-(false positive rate of the test) )

Sensitivity and specificity are functions of the test being used.  Obviously, you’d like a test where these are very close to 100%, particularly in the setting of HIV testing, but in reality, no test is perfect, and you often end up with a trade-off – highly sensitive tests (ie: few false negatives) may come at the expense of a slightly lower specificity (ie: a few false positives).

Positive predictive value (PPV):  What is the chance that a positive test result is actually a truly HIV+ person rather than a false positive?

Negative predictive value (NPV):  What is the chance that a negative test result is a truly HIV- person rather than a false negative.

Unlike the sensitivity and specificity, PPV and NPV depend on both the test, but also the disease prevalence – which is a key issue for this discussion, and I’ll illustrate by way of example.

 

The key to understanding the blood donor screening questionnaire is to think of it as a risk-mitigation tool.  The Red Cross Blood service want to manipulate the population of donors such that the risk of HIV-positivity is as low as possible;  this in turn makes the likelihood of a false negative test even smaller.  As Nic points out in his article, HIV disproportionately affects men who have sex with men, therefore by excluding them, the donor pool is less likely to have positive individuals within it.

From the ASHM PEP guideline (pdf), HIV seroprevalence various populations are as follows:

  • MSM in major Australian centres:  4-12%
  • MSM who inject drugs: 29%
  • Non-MSM PwID: 1%
  • Whole Australian population: 0.1%
  • People who “pass” the blood donor questions:  0.0004%

Lets call the prevalence in MSM a nice round 10% (because it makes my numbers easier), and round the blood donor prevalence up to 0.001% for the same reason (this will reduce the difference in risk somewhat).

Modern HIV tests are pretty good – quoted performance is a sensitivity (picking up people with HIV) of 99.8% and a specificity (positive test is correct) of 98%.

So, you can construct a cross-table of the prevalence of the virus and the performance of the test, and I’ve done this under three different scenarios: 1) people who are eligible to donate based on the ARCBS questionnaire, the unscreened general population, a population with HIV seroprevalence of 10%.

 

Lets see what it shows:

 

Eligible blood donors

HIV Status
PositiveNegativeTotals
HIV Test ResultPositive998999,9901,000998
Negative298,999,01098,999,012
Totals100099,999,000100,000,000
HIV Seroprevalence: 0.001%
All negative tests: 98,999,012
True negative tests: 98,999,010
Likelihood that a negative test means really no HIV (ie: Negative predictive value) = 99.999998%

= 1 false negative test / 49,499,506 tests done

 

Unscreened Population

HIV Status
PositiveNegativeTotals
HIV Test ResultPositive998990010,898
Negative2989,010989,012
Totals1,000999,0001,000,000
Seroprevalence: 0.1%
All negative tests: 989,012
Truly negative tests: 989,010

NPV: 99.9998% = 1 false negative test / 494,506 tests

 

Population Seroprevalence 10%

HIV Status#colspan
PositiveNegativeTotals
HIV Test ResultPositive99,8009,000108,800
Negative200891,000891,200
Totals100,000900,0001,000,000
Seroprevalence: 10% (ballpark for Sydney / Cairns MSM)
All negative tests: 891,200
Truly negative tests: 891,000

NPV: 99.98% = 1 false negative test / 4456 tests done

 

So, you can see from these tables, that the screening questions improve the performance of the test by reducing the pre-test probability of HIV infection in the population;  it is essentially a way of minimising the small risk of a false negative test to becoming an even more rare event.  Hopefully this example has shown that there is benefit in donor screening.

The question is then, is it reasonable for any sex with a man being reason for deferral?  Seeing a sex worker in the previous 12 months is also indication for deferral, despite the fact that it is a requirement that Australian sex workers are tested every three months and also that sex workers do not offer unprotected services.

As well as this, just as HIV is unevenly distributed in the general population, it’s also unevenly distributed amongst gay men – older men who may have been diagnosed in the 80s, sexually adventurous men (known in the sexual health business as SAMs).  At what point do you decide to jump off the curve of diminishing returns in making the donor questions even more complex?

It’s particularly worth thinking about this in light of the final estimation I’ll do in this post.

Around 3.5% of the eligible donor population give blood.

Given there are 25,000 people living with HIV in Australia, and around 75% acquired it through MSM, lets say that there are 20,000 MSM with HIV.  Based on a seroprevalence of 10%, that means there are around 200,000 MSM in Australia.  Say 5% of them would give blood if they were eligible, that would be an extra 10,000 donors – or an increase in the donor pool by 2%.

Sure, a 2% increase in the number of donors would give some extra reserve in the system, but would need either a major overhaul of the donor screening questions and the increased potential for false negative tests.

ARCBS have reviewed the guidelines relating to sexual behaviours and deferrals in 2012, and based on the principle that blood safety was paramount, elected to leave restrictions as they are.  The full report (pdf) acknowledges that this policy may exclude some low-risk MSM, but on the balance felt that removing the restriction was an unacceptable risk.

 

There’s no easy answer to this question, and to be honest, I’m not entirely decided either way.  Please feel free to leave your comments.

(thanks to Nic Holas for his article and to Daniel Reeders, -his blog name I have shamelessly ripped off for the title of this post)

 

 

 

 

You may also like...

4 Responses

  1. I thought a major motivation for the screening rules was to cut down on the chance of donation from someone with a recent infection who isn’t yet making antibodies, rather than to handle the assay false-negative rate.

    • Trent says:

      Yes, that’s also part of it – but they screen using a 4th generation combo assay (p24 antigen + antibody) and a RNA PCR as well, so in theory the RNA (or the p24) should pick up people even in that window.

      I haven’t included the stats on double testing here – it’s substantially more complex as they pool samples for testing.

  2. badblood says:

    ARCBS also screen pooled samples with NAAT — nucleic acid amplification testing — in pools of 1,000. I’m actually having a hard time finding good estimates of its sensitivity, but it was described to me by a pathologist friend as ‘fucking sensitive’. So ARCBS’ position isn’t really about safety, it’s about cost: including higher risk categories requires smaller pools and further NAAT testing when a pool comes back reactive.

Leave a Reply

%d bloggers like this: