Basic Neurosciences · Paper A
HIV disease markers
An HIV infection is described by two numbers that answer two different questions, and the one that moves fastest is not the one that tells you what the patient can catch. The plasma viral load counts how much virus is being made right now, so it falls within weeks of effective treatment and is the marker of whether the drugs are working. The CD4 count measures how much immune defence survives the damage already done, so it climbs back over months and years and is the marker that predicts which opportunistic infections can get in. A third test, the antibody test, answers a question neither of those does — has this person met the virus at all — and it answers late, which is the whole of the window period. For a psychiatrist the payoff is specific: those numbers decide whether a new cognitive problem needs an urgent scan, and in the treated era the brain no longer behaves like the blood.
Each number moves at the speed of the biology underneath it. HIV enters CD4-positive T-helper cells, uses them to make new virus and kills them in the process, and the virus particles it releases are cleared from plasma within hours. Because the pool in the blood turns over that quickly, the count of viral copies is a live read-out of production rather than a record of the past: block replication with antiretroviral therapy and the number collapses over weeks, which is exactly why it is the response marker. The CD4 count is the opposite kind of measurement. Every infected cell that dies is a cell the body must replace, and replacement depends on a thymus that has largely shut down by adulthood, so the count records accumulated loss and recovers slowly, typically by fifty to a hundred and fifty cells in the first year of treatment, most of it early, and more slowly after that. That is why the count, not the viral load, predicts what can invade: it measures the defence that is actually standing. The antibody test lags for a third reason. An adaptive immune response has to be built before it can be detected, so antibodies take around three to four weeks to appear, while viral genetic material is detectable from about ten days and the p24 antigen from roughly two weeks — which is why the modern laboratory test looks for antigen and antibody together and turns positive earlier than an antibody-only test. One compartment escapes all of this. The virus that establishes itself in the brain lives mainly in long-lived macrophages and microglia rather than in T cells, evolves separately there, and can keep a local inflammatory process running with the plasma number undetectable.
A man of thirty-eight is referred to you with four months of low mood and slowed thinking. He has HIV, he takes his tablets every day, and his last clinic letter says his viral load (the number of copies of the virus's genetic material in a millilitre of blood plasma) is undetectable. It is tempting to read that word and stop. Do not. Undetectable describes the virus, not the immune system, and the two recover on completely different timescales. Get the habit of asking which of the numbers in front of you moves in weeks, which moves in years, and which never moves again, and most of what the exam asks about HIV falls into place.
The plasma viral load counts virus. It is reported as copies per millilitre of plasma, and it tells you how much virus is being produced at this moment. The CD4 count (the number of CD4-positive T-helper cells in a cubic millimetre of blood) counts defence. Normal sits at roughly five hundred to fifteen hundred, and it tells you how much of the immune system is still standing after everything the infection has already destroyed.
So the two answer different questions. Is the treatment working? That is the viral load. What can this person catch? That is the CD4 count. Nothing in the topic matters more than keeping those apart.
Here is where careful readers still go wrong. An undetectable viral load means antiretroviral therapy (a combination of drugs that stops the virus copying itself) is doing its job, and it means something else genuinely important: with the level sustained undetectable, the infection is not passed on to a sexual partner. What it does not mean is that the immune system has been repaired.
A man diagnosed with a count of ninety may be undetectable within three months and still have a count of two hundred and fifty five years later. He is not infectious. He is also not immunologically normal, and the lowest count he ever reached — his nadir (the smallest value a measurement has ever taken in that person) — goes on predicting his risk long after the viral load has stopped being interesting.
Think of the infection as a bank account with a standing order draining it. The payment going out each month is the viral load; the balance left in the account is the CD4 count. Those are two genuinely different quantities, and the whole topic follows from that.
Cancel the standing order and next month's statement shows no payment. That is what treatment does, and it happens fast because virus particles are cleared from plasma within hours, so the number in the blood reflects production now rather than production last year. Cancelling the payment does nothing to the balance. Refilling an account is saving, and saving is slow: the count typically climbs by fifty to a hundred and fifty cells in the first year of treatment, most of that in the first few months, and more slowly after that, because replacing T-cells depends on a thymus that has largely shut down by adulthood.
And once the balance drops under certain lines, particular direct debits start to bounce. Below two hundred cells, pneumocystis pneumonia becomes the threat. Below one hundred, toxoplasmosis and cryptococcal disease. Below fifty, cytomegalovirus retinitis. Those lines are properties of the balance, never of the payment.
Suppose a man is diagnosed with a CD4 count of one hundred and twenty and a viral load of two hundred and forty thousand copies per millilitre. He is well below the two hundred line, so he starts co-trimoxazole as prophylaxis (a medicine given to prevent an infection rather than to treat one) alongside his antiretroviral therapy.
By twelve weeks his viral load is under fifty copies. The standing order is cancelled. His count, meanwhile, has moved from one hundred and twenty to about two hundred and ten over the first year — a rise of around ninety cells, which is squarely inside the fifty to a hundred and fifty typically seen in the first year, most of it arriving in the first few months, for people who start treatment with a low count. Two hundred and ten is over the line, so his prophylaxis can stop, though only once the count has held above two hundred for some months with the virus still suppressed.
Run it forward another year and he gains perhaps fifty more, to around two hundred and sixty. That is still far below the five hundred to fifteen hundred normal range quoted above. Twelve weeks of treatment fixed the virus; five years may not fix the count.
| The result in front of you | What it is actually reporting |
|---|---|
| Undetectable plasma viral load | Virus production in blood now — and non-infectiousness; nothing about immune recovery |
| CD4 count today | What can invade today; useless as a measure of whether the drugs are working |
| CD4 nadir | The worst damage ever sustained; predicts long-term recovery and cognitive risk |
| Cerebrospinal fluid viral load | The brain compartment, which the plasma result never covers |
The screening investigation in United Kingdom laboratories is a fourth-generation combined test, which looks for antibodies against the virus and for p24 antigen (a viral protein detectable in blood before antibodies appear) in the same assay. Antibodies take about three to four weeks to become detectable; antigen appears from around two weeks; viral genetic material from around ten days. The interval between infection and a positive test is the window period, and looking for antigen as well as antibody is what shortens it.
In practice that means a negative fourth-generation test taken around six weeks after a specific exposure is strong evidence against infection, with a repeat later where post-exposure prophylaxis has been taken, because the drugs delay the antibody response. A point-of-care test looking for antibodies alone has a longer window. Any reactive screening result is confirmed on a second sample before anyone is told they have HIV, and where very recent infection is suspected the test to ask for is an HIV RNA test rather than a repeat antibody test.
Antiretroviral therapy is offered to everyone at diagnosis whatever the count, and getting the viral load undetectable is the single intervention that changes both the individual's outcome and onward transmission. Co-trimoxazole is added as prophylaxis while the count sits below two hundred and stopped once it is reliably above. Those thresholds are British HIV Association standards; the World Health Organization's clinical staging system does the same job by symptoms where CD4 testing is unavailable.
Your own contribution is prescribing. Check every psychotropic against the antiretroviral regimen before you write it, using the BNF or a dedicated interactions resource: boosting agents such as ritonavir and cobicistat raise the levels of many psychiatric drugs substantially, and older regimens containing efavirenz carry vivid dreams and mood change in their own right, which is one reason it is now rarely first-line. Treating the depression in front of you is not a distraction from the HIV, because untreated depression is among the strongest predictors of people stopping their tablets.
The high-yield move is to name the question before reading the number. Treatment response is the viral load. Infection risk is the CD4 count. Exposure history is the antibody and antigen test, read against the calendar. Long-term cognitive risk is the nadir. And the brain is the compartment none of the blood numbers reports on, which is why a patient with perfect blood results can still be declining neurologically. Test yourself now: can you name which single number you would ask for in a suppressed patient whose thinking has worsened over six months, and say what a normal result would and would not rule out?
Hold the shape rather than the list: the viral load moves in weeks and reports on treatment, the CD4 count moves in years and reports on risk with its lines at two hundred, one hundred, and fifty, the antibody arrives weeks late and creates the window period, the nadir never moves at all, and the brain answers to none of them.