Will My Medication Work for Me

Message us about a medicine that did not agree with you, or call 020 3951 3429.

When to get help today

Stop and get help now if a medicine has caused a rash with blistering, swelling of the face, lips or tongue, difficulty breathing, or a fever with a sore mouth. Call 999 or go to your nearest emergency department.

Do not stop a prescribed medicine because of anything on this page. Speak to whoever prescribed it first. A pharmacogenomic result changes a dose or a choice, and it does that through a prescriber rather than around one.

What a pharmacogenomic report actually is

Your liver breaks down most medicines using a small number of enzymes, and the genes that build those enzymes vary between people. Some people clear a drug quickly and get little effect at a standard dose. Some clear it slowly and get the side effects of a much larger one. A pharmacogenomic report reads those genes once and states which group you are in.

It is written against published guidelines rather than opinion. The Clinical Pharmacogenetics Implementation Consortium and the Dutch Pharmacogenetics Working Group both publish gene and drug recommendations that say what to do with a given genotype, and a report that does not name which guideline it followed is not a clinical document.

The result does not expire. Your genotype at 40 is your genotype at 70, so this is read once and used for every prescription after it.

The strongest result, and the half of it that is usually left out

PREPARE is the trial that moved this from interesting to actionable. Swen and colleagues randomised 6,944 patients across seven countries and tested 50 variants across 12 genes, published in The Lancet in 2023. Clinically relevant adverse drug reactions occurred in 21.5% of the genotype guided group against 28.6% of the control group, an odds ratio of 0.70 with a 95% confidence interval of 0.61 to 0.79 and p below 0.0001.

Write that as roughly a 30% reduction in the odds, because that is what an odds ratio of 0.70 means. It is not a 30% reduction in risk. In absolute terms the reduction on those 2023 Lancet figures was about 7 percentage points, which is real and is smaller than the headline sounds.

Three limitations travel with it. The trial was open label, so nobody was blinded. Randomisation was clustered at country and site level rather than at patient level. And a published correspondence in the same journal challenged whether a reduction of that absolute size justifies testing whole populations. We think it justifies testing the people in front of us who are already taking several medicines, which is a narrower claim.

Message us about the medicines you take

The result where the stakes are highest

Before certain chemotherapy, this stops being about tolerability and starts being about survival. One enzyme clears a widely used class of chemotherapy drug, and a person who makes too little of it is given a normal dose of something their body cannot handle.

Henricks and colleagues studied 1,103 evaluable patients across 17 hospitals, published in Lancet Oncology in 2018. Around 8% carried a heterozygous DPYD variant. Severe toxicity occurred in 39% of carriers against 23% of people without the variant, p equal to 0.0013. The MHRA Drug Safety Update of October 2020 recommends testing for DPD deficiency before starting 5-fluorouracil, capecitabine or tegafur.

The caveat most writing omits is the important one. In the same 2018 Lancet Oncology cohort, the 25% dose reduction used for two of the variants was not enough to bring risk back to normal. Only the 50% reductions brought carriers close to the risk of people without the variant. Genotype guided dosing lowers the danger here. It does not remove it, and any page that says otherwise is selling.

The trial that missed, and why we print it anyway

TAILOR-PCI tested whether choosing an antiplatelet drug by CYP2C19 genotype after a stent reduced major cardiovascular events. Pereira and colleagues randomised 5,302 patients, with 1,849 loss of function carriers in the primary analysis, published in JAMA in 2020.

At 12 months, on the 2020 JAMA figures, the primary endpoint occurred in 4.0% of the genotype guided arm against 5.9% of the conventional arm, a hazard ratio of 0.66 with a 95% confidence interval of 0.43 to 1.02 and p equal to 0.06. That does not reach statistical significance. It is a negative trial that trends in a favourable direction, and it should never be quoted as a positive one.

We print it because a clinic that only shows you the trials that worked is telling you something about how it reads the rest of the literature.

How many people this actually touches

Van Driest and colleagues genotyped 9,589 people at Vanderbilt and found 91% carried at least one actionable variant across five gene and drug pairs, published in Clinical Pharmacology and Therapeutics in 2014, rising to 96% in African American patients.

A much larger analysis by Chanfreau-Coffinier and colleagues, covering 7,769,359 people and published in JAMA Network Open in 2019, put the figure at 99%. That number is projected from reference allele frequencies rather than measured by genotyping everyone, and that caveat has to travel with it every time it is quoted. The same paper found 54.8% of people had received at least one medication carrying a CPIC Level A recommendation, and that 37.9% of new prescriptions were for one.

Its limitations are worth stating too. Projected variant prevalence, under representation of minority populations in the reference data, a single health system, and a medication landscape from 2011 to 2017.

What you can do before you come

Build a medicines history nobody has ever asked you for properly, and take it to any prescriber, here or anywhere else. It takes twenty minutes, it costs nothing, and it changes prescribing decisions on its own before any genetic result exists.

Every medicine that has ever disagreed with you. What it was, what happened, roughly when, and whether anyone stopped it or changed the dose. Include the ones that simply did nothing, because a medicine that had no effect at a normal dose is as informative as one that caused a reaction. Include family. A brother who could not tolerate a statin is a data point about you.

Then one thing that costs nothing and matters more than anything else on this page. If you are ever offered capecitabine, 5-fluorouracil or tegafur, ask whether your DPD status has been checked before the first dose. The MHRA recommended that in October 2020 and it is your right to ask.

And what not to buy. A standalone medicines panel bought online, read by nobody, and filed where your prescriber will never see it, is money spent on a document. The PREPARE result depended on the prescriber holding the report at the moment of prescribing. If the report will not be in front of that person, it will not do what the trial did.

What we would do

One. A conversation with a registered genetic counsellor, at no charge, and a medicines history taken properly, including the reactions nobody wrote down at the time and the medicines that simply did nothing.

Two. Saliva at home or a blood draw in Marylebone, sequenced at 30x in a laboratory accredited to ISO 15189, with the pharmacogenomic report written against the CPIC and DPWG guidelines.

Three. A doctor reads it with you, writes it into the record the clinic prescribes from, and it is applied to every prescription after that, for life.

Book Genome and Medicines

Message us about booking Genome and Medicines

Questions people ask

Will pharmacogenomic testing tell me if my medication will work? It tells a prescriber how you are likely to metabolise it, which changes the dose or the choice. In PREPARE, 6,944 patients across seven countries, clinically relevant adverse drug reactions fell from 28.6% to 21.5%, an odds ratio of 0.70, Swen and colleagues, The Lancet, 2023. That is roughly 30% of the odds and about 7 percentage points absolute.

Is pharmacogenomic testing proven? Partly. PREPARE was positive for adverse drug reactions. TAILOR-PCI, testing CYP2C19 guided antiplatelet choice after a stent, found 4.0% against 5.9% at 12 months with p equal to 0.06, Pereira and colleagues, JAMA, 2020, which does not reach significance. Both results are true at once.

Do I need this before chemotherapy? Ask about DPD deficiency before 5-fluorouracil, capecitabine or tegafur. The MHRA recommended testing in October 2020. Severe toxicity occurred in 39% of DPYD carriers against 23% of non carriers, Henricks and colleagues, Lancet Oncology, 2018, and reduced dosing lowers that risk without removing it.

How many people carry something worth acting on? Van Driest and colleagues found 91% of 9,589 people genotyped carried at least one actionable variant across five gene and drug pairs, Clinical Pharmacology and Therapeutics, 2014, and 96% in African American patients. A larger projected figure of 99% exists but is modelled from allele frequencies rather than measured.

Do I have to sequence my whole genome to get this? No, a targeted panel can produce a pharmacogenomic report. A whole genome gives you the same report plus everything else in the file, read again every year without a second sample. Genome and Medicines is £9,995 and includes both.

Does the result ever go out of date? Your genotype does not change, so the reading of your genes is done once. What moves is the guidance, because the Clinical Pharmacogenetics Implementation Consortium and the Dutch Pharmacogenetics Working Group both revise their recommendations. That is why the file is re read every year and why you are told only when something has changed for you.

Can I stop a medicine because of the report? No. The report goes to whoever prescribes for you and the change is made by them. Stopping a medicine on your own is the one way this causes harm rather than preventing it.

Will it tell me which antidepressant to take? It narrows the field rather than choosing for you. Metabolism is one input and diagnosis, previous response and preference are the others. Any service claiming to pick your antidepressant from a cheek swab alone is overstating what the evidence supports.

Message us with your medicines list, or call 020 3951 3429.

This page is information and not a diagnosis, and it is not a recommendation to start or stop any medicine. Prescribing decisions are made by the clinician responsible for your care. No result guarantees an outcome. If you are unwell today, see a doctor today. In an emergency call 999.

References

Swen JJ, van der Wouden CH, Manson LE, et al. A 12-gene pharmacogenetic panel to prevent adverse drug reactions, an open-label, multicentre, controlled, cluster-randomised crossover implementation study. The Lancet, 2023;401(10374):347-356. 10.1016/S0140-6736(22)01841-4

Henricks LM, Lunenburg CATC, de Man FM, et al. Lancet Oncology, 2018;19(11):1459-1467.

Pereira NL, Farkouh ME, So D, et al. JAMA, 2020;324(8):761-771. PMID 32840598.

Van Driest SL, et al. Clinical Pharmacology and Therapeutics, 2014;95(4):423-431.

Chanfreau-Coffinier C, et al. JAMA Network Open, 2019;2(6):e195345.

Medicines and Healthcare products Regulatory Agency. Drug Safety Update, October 2020, DPD deficiency and fluoropyrimidines.

Clinical Pharmacogenetics Implementation Consortium. Published gene and drug guidelines, cpicpgx.org, read 8 September 2026.

Dutch Pharmacogenetics Working Group. Pharmacogenetic recommendations, read 8 September 2026.

American College of Medical Genetics and Genomics. Recommendations for reporting of secondary findings in clinical exome and genome sequencing, ACMG SF v3.3. Genetics in Medicine, 2025.

ISO 15189:2022. Medical laboratories, requirements for quality and competence. International Organization for Standardization.