Atazanavir: the HIV drug that turns some people's eyes yellow
The side effect this page is about looks alarming and is medically harmless, which is a difficult combination. A visible yellowing of the eyes and skin makes people stop a working HIV drug, and one common gene variant predicts almost exactly who it happens to.
Also sold as: Reyataz.
What atazanavir is
Atazanavir is an antiretroviral drug used to treat HIV, a protease inhibitor taken once a day as part of a combination and usually boosted by a second small-dose drug that keeps its level up. It is effective and long-established, though like efavirenz it has been overtaken by newer options in first-line use in many places.
Why your genes come into it
UGT1A1 is the enzyme that processes bilirubin, the yellow pigment the body makes when it recycles old red blood cells. Atazanavir blocks that enzyme as a side effect, so bilirubin can back up and produce visible jaundice, yellow eyes and skin, without any actual liver damage. How much it backs up depends on how much UGT1A1 capacity a person started with. People who carry two copies of the reduced-activity allele, which is the same genotype behind the benign condition called Gilbert syndrome, have less spare capacity, and in them atazanavir is far likelier to push bilirubin high enough to show. The published estimates are stark: at the reduced-function end the chance of jaundice bad enough to make someone stop the drug runs from roughly one in five up toward well over half.
What published guidance says
The rows below restate CPIC's published guidance for each possible UGT1A1 result. They describe what an HIV physician weighs, in particular whether visible jaundice would trouble a given person enough to favour a different drug from the start, not a change for anyone to make alone. The jaundice itself is not dangerous, but stopping a suppressive HIV drug is, so this is a conversation to have before a problem rather than a reason to act on your own.
UGT1A1
| If your UGT1A1 result is | What that means | Published CPIC guidance |
|---|---|---|
| Intermediate Metabolizer | Somewhat decreased UGT1A1 activity; low likelihood of bilirubin-related discontinuation of atazanavir. | Per CPIC (A): "There is no need to avoid prescribing of atazanavir based on UGT1A1 genetic test result. Inform the patient that some patients stop atazanavir because of jaundice (yellow eyes and skin), but that this patient's genotype makes this unlikely (less than about a 1 in 20 chance of stopping atazanavir because of jaundice)." |
| Normal Metabolizer | Reference UGT1A1 activity; very low likelihood of bilirubin-related discontinuation of atazanavir. | Per CPIC (A): "There is no need to avoid prescribing of atazanavir based on UGT1A1 genetic test result. Inform the patient that some patients stop atazanavir because of jaundice (yellow eyes and skin), but that this patient's genotype makes this unlikely (less than about a 1 in 20 chance of stopping atazanavir because of jaundice)." |
| Poor Metabolizer | Markedly decreased UGT1A1 activity; high likelihood of bilirubin-related discontinuation of atazanavir. | Per CPIC (A): "Consider an alternative agent particularly where jaundice would be of concern to the patient. If atazanavir is to be prescribed, there is a high likelihood of developing jaundice that will result in atazanavir discontinuation (at least 20% and as high as 60%)." |
Can your DNA file answer this?
UGT1A1 is not part of the standard chip or variant-file report, and the reason is the variant itself. The key change, the *28 allele, is a difference in the length of a repeated stretch of DNA rather than a single letter swap, which is the kind of variant genotyping arrays handle worst. Some consumer files carry a nearby marker that travels with *28, but that is a proxy rather than the variant. Aimosti reads UGT1A1 from aligned reads in a BAM or CRAM file as part of the Deep Read panel and returns Indeterminate where the repeat length is not resolved, rather than inferring it from a proxy.
A genotyping chip reads only a few hundred thousand pre-selected positions, about 0.02% of your genome, chosen for common variation. It does not sequence the rest, so it cannot find the rare or novel pathogenic variants the clinical and carrier modules look for: a “no finding” from chip data means “this chip never looked”, far more so than with whole-genome sequencing. Chip data suits common-variant traits, pharmacogenomic tag SNPs, and haplogroup ancestry, not clinical or carrier screening.
Common questions
I have Gilbert syndrome. Does that mean atazanavir will make me jaundiced?
It makes it much more likely, because Gilbert syndrome usually reflects the same reduced-function UGT1A1 genotype that leaves the least spare capacity to handle the bilirubin atazanavir backs up. It does not make it certain, and the jaundice, if it comes, is the harmless cosmetic kind rather than a sign of liver injury. It is a strong reason to raise the question before starting the drug so an alternative can be considered if the appearance would bother you.
Is the jaundice actually harmful?
No. It is unconjugated bilirubin rising because the enzyme that clears it is being blocked, the same mechanism as in Gilbert syndrome, and it does not indicate liver damage. The problem it causes is human rather than medical: people dislike looking visibly yellow and stop an otherwise effective HIV drug, which is the outcome the guidance is trying to head off.
If I am at high risk of jaundice, do I have to avoid atazanavir?
Not necessarily. The guidance suggests considering an alternative particularly where jaundice would concern the person, which leaves room for atazanavir if the appearance would not trouble you and it otherwise suits your regimen. That trade-off is a prescribing decision made with your preferences and your full treatment in view.