NAT2 Genotype

If you're about to start isoniazid for tuberculosis, or you have taken it before and your liver enzymes climbed, this result changes the conversation. In one cohort of 120 tuberculosis patients, liver injury of any grade occurred in 65.5% of slow acetylators, 55.2% of intermediate acetylators, and 42.4% of rapid acetylators.

The gene is NAT2, short for N-acetyltransferase 2, and it does not change over your lifetime. A genotype result can stay useful each time a prescription involves a drug this enzyme handles.

What the Enzyme Actually Does

NAT2 encodes an enzyme made mostly in your liver cells and in the lining of your small intestine and colon. Its job is to attach a small chemical tag to certain compounds so your body can clear them. The targets are a specific family: aromatic amines, hydrazines, and the heterocyclic amines that form when meat is cooked at high heat.

You inherit two copies of the gene. Depending on which versions you carry, you land in one of three groups. Two higher-activity copies usually makes you a rapid acetylator. One higher-activity and one lower-activity copy makes you intermediate. Two lower-activity copies makes you slow.

With isoniazid, slower clearance can raise drug exposure and increase the chance that reactive byproducts injure liver cells. Rapid acetylators clear it faster, which sounds better until the drug level falls below what is needed to kill the bacteria.

Liver Injury from Tuberculosis Drugs

This is the association with the strongest evidence behind it, and one reason the test exists clinically. Isoniazid, the backbone of tuberculosis treatment, is a hydrazine. NAT2 is its main off-ramp.

Multiple meta-analyses pooling studies of tuberculosis patients found the slow genotype to be a risk factor for drug-induced liver injury during treatment. Recent pooled estimates put the odds at roughly three times those of faster acetylators, with individual analyses landing anywhere from about twice to nearly five times depending on population, injury definition, and which alleles are counted. In a Thai genome-wide study, NAT2 was the single strongest genetic risk factor for this kind of liver injury. In a separate Thai cohort of 894 people living with HIV, slow acetylators had roughly double the isoniazid exposure and about 2.4 times the odds of liver toxicity compared with rapid acetylators.

Timing matters as much as the odds. In one cohort, liver toxicity showed up in slow acetylators at a median of half a month into treatment, versus two months in the rapid and intermediate group, about three times the early hazard. If you are a slow acetylator, the danger window is front-loaded, which is when many monitoring schedules are still ramping up.

For a slow acetylator starting isoniazid, the case for frequent liver enzyme checks in the first weeks is much stronger than a standard schedule would suggest, and the case for discussing a lower dose with the physician managing your treatment is real. In a Japanese randomized trial of 172 patients, genotype-guided dosing eliminated early isoniazid liver injury in slow acetylators, none of whom were affected compared with 78% on standard dosing, and cut early treatment failure in rapid acetylators from 38% to 15%. That trial is the strongest direct evidence for acting on the result, and genotype-guided dosing is still not standard practice in most tuberculosis programs.

Why the Same Genotype Cuts Both Ways

Here is where it stops being a simple good-number, bad-number test. Rapid acetylators clear isoniazid so efficiently that standard weight-based dosing can leave drug levels too low, risking treatment failure and resistant bacteria. Slow is not bad and rapid is not good. Each carries a different failure mode with the same drug.

Cancer risk depends on the exposure, and on the exact chemical involved. Slow acetylators exposed to single-ring aromatic amines from tobacco smoke, such as 4-aminobiphenyl, have higher bladder cancer risk, because the compounds that need deactivating stay active longer. Rapid acetylators have been linked to higher colorectal cancer risk in some pooled analyses, likely because the enzyme can convert dietary compounds from charred meat into forms that damage DNA in the gut, but a 40-study analysis covering more than 40,000 people found no association at all, so that link remains contested rather than settled. One older case-control study of lung cancer found that people carrying two copies of the fully functional version had about three times the risk of people who did not, odds ratio 3.04. Later meta-analyses have mostly found no overall lung cancer association, and where one appears it often points the other way, toward the slow or intermediate phenotype, so treat that 1996 result as an outlier rather than a consensus.

The framework that makes all of this consistent: NAT2 is not a dial where more is better. It is a switch that routes chemicals down one path or another. Which path is safer depends entirely on what chemical you are handling.

Bladder Cancer and Smoking

The bladder cancer link is the best-studied cancer association, and it is a genuine interaction, not a standalone risk. The Spanish Bladder Cancer Study and its accompanying meta-analyses found the slow genotype raised bladder cancer risk by about 40% overall and particularly among cigarette smokers. A 54-study meta-analysis covering 13,343 cases reached the same conclusion. A later study covering the New England bladder cancer population showed the effect scales with how much you smoke: the genotype matters more in heavy smokers.

A UK Biobank analysis of over 390,000 people found current smokers who carried both a high genetic risk score and a slow phenotype had elevated bladder cancer risk. In a separate Swedish cohort of about 25,000 people, the combination of smoking and a bladder-cancer genetic variant produced more risk than either alone would predict.

The direction is not even fixed within bladder cancer. Among workers exposed to benzidine, a two-ring aromatic amine, slow acetylation turned out to be protective rather than harmful, because for that chemical the tagging step switches the compound on rather than clearing it. A nested analysis in the European EPIC cohort found no bladder cancer association with slow acetylation at all and argued against using the genotype to screen workers. So the occupational picture depends on which chemical is in the workplace, and the evidence there is genuinely mixed.

The smoking point is the solid one. If you carry the slow genotype and you smoke, quitting is doing more work for you than it would for someone with a rapid genotype. The genotype does not cause the cancer. The exposure does. The genotype helps decide how much damage the same exposure does, and for industrial chemicals the sign of that effect can flip depending on the compound.

Other Drugs This Affects

Isoniazid gets the attention, but the enzyme handles a short list of other drugs. A meta-analysis of nine cohort studies covering 1,077 patients taking sulfasalazine, used for inflammatory bowel disease and rheumatoid arthritis, found slow acetylators had roughly three times the odds of adverse reactions overall and about five times the odds of dose-related ones. The drug label states the same direction. Hydralazine for blood pressure and procainamide for heart rhythm are also cleared through this pathway, and slow acetylation has long been linked to drug-induced lupus with these agents.

Laboratory work in human liver cells confirms the pattern directly: cells from rapid, intermediate, and slow acetylators process hydralazine at measurably different rates.

A Signal in Metabolic Health

This part is newer and less settled. A large human study across multiple cohorts linked one specific coding change in NAT2 to reduced insulin sensitivity, independent of body weight, and named the gene as an insulin sensitivity gene. In that same work, the predicted fast-or-slow acetylator status itself was not associated with insulin sensitivity, so this is a separate signal from the result that drives drug dosing. Separate analyses connect non-coding variants near NAT2 to differences in plasma triglycerides and cholesterol.

A biobank analysis of over 330,000 people found that people carrying the NAT2*1 version got a slightly better LDL cholesterol response to simvastatin, on the order of one to five percent more reduction. That is a real finding and a small one. It is not a reason to pick a statin, and it does not replace checking your lipid panel.

One Test, Then Act on It for Life

Your genotype will not change. There is no trend to track and no reason to retest unless the lab result itself is in doubt, which happens more often than you would expect with older testing methods.

What does need ongoing tracking is the downstream picture. If you carry a slow genotype and you are on any of the drugs above, liver enzymes are the number to watch, and to watch early rather than on a leisurely schedule. If you carry a slow genotype and you smoke or have occupational chemical exposure, the relevant follow-up is exposure reduction, not more genetic testing.

What an Unexpected Result Should Make You Do

If you come back slow or intermediate and you are on or starting a drug from this list, the pathway is: order a liver panel now as a baseline, then repeat it more often than standard practice suggests during the first two months. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) are the two enzymes that move first when liver cells are being damaged.

Dose changes are a prescription decision. Bring the genotype to the physician managing the drug, whether that is an infectious disease specialist for tuberculosis or a rheumatologist for sulfasalazine, and ask whether genotype-guided dosing applies. If you also want direct confirmation that your drug levels are where they should be, therapeutic drug monitoring measures the actual concentration in your blood and settles the question that genotype only predicts.

If the genotype call is limited, ambiguous, or surprising for your ancestry, confirm it with full sequencing or a lab method that can resolve which variants travel together on the same copy of the gene. A pharmacogenetics specialist or genetic counselor is useful when the result changes a long course of treatment, or when relatives may act on it too.

If the result is rapid and you are on tuberculosis treatment that does not seem to be working, undertreatment is worth raising explicitly. Rapid clearance producing subtherapeutic levels is a known route to treatment failure, and it is easy to mistake for a resistant organism.

Where the Result Can Mislead You

The biggest problem with this test is not biology, it is the assay. NAT2 variants sit at different spots along the gene, and working out which versions you carry requires knowing which variants are on the same copy of the gene. Short-read sequencing cannot see that directly and has to guess statistically.

In a comparison across 1,828 people, long-read sequencing called the paired-copy result correctly over 99% of the time, while short-read sequencing matched only 64% of the time, because heterozygous positions could not be sorted onto the right copy. The practical damage is smaller than that gap sounds. In the same study, short-read data and even a four-variant panel still predicted the right acetylator phenotype in over 95% of people, and the paired-copy confusion changed the phenotype call in only about 4%. The phenotype is what gets acted on. Statistical guessing tools leave about 4.9% of white individuals with an unresolved call and get roughly 0.8% outright wrong.

Ancestry is where the gap actually bites. Most clinical panels test four or five common variants, and those perform well in European-ancestry populations. In non-white populations the discordant or indeterminate rate ran from 4.5% with a four-variant panel up to 25% with a five-variant one, because ancestry-specific versions of the gene are simply not on the panel. African-ancestry populations carry the slow NAT214 cluster at about 8.9% and NAT243 at about 3.6%, and both are essentially absent from European populations. Hispanic populations carry NAT2*7 at around 7%.

• Panel coverage: a limited-variant panel that comes back normal has only ruled out the variants it tests for, not every variant in the gene.
• Ancestry mismatch: if your ancestry is not primarily European, a four- or five-variant panel is more likely to give you an ambiguous or wrong phenotype call than a full sequencing result.
• Unclassified variants: full sequencing turns up versions of the gene whose effect on enzyme activity has not been characterized, and those get reported without a clear phenotype.
• Clinical-grade versus consumer data: a direct-to-consumer raw-data call may not include the variants, phasing, or reporting checks needed for prescribing decisions.
• Tumor-only sequencing: a result from cancer tissue is not the same as an inherited drug-metabolism result unless a matched normal sample confirms it.
• Genotype versus real-world enzyme activity: older studies comparing genotype with directly measured acetylation found they sometimes disagree, and factors like urine acidity and caffeine intake can shift measured activity without changing your genes.