Genetic Variation Impacts Drug Efficacy. Could Testing Help?

In April 2019, Chloe Meadows was diagnosed with ADHD and began working with her doctor to find a drug cocktail to relieve her symptoms. Among the medicines she took was Wellbutrin, in late 2020. She recalls that about a month into taking it, however, she sat down to eat pizza, suffered a seizure, and fell, dislocating her shoulder. Family members later told her she hit her head so hard that her earring flew out. She was unconscious, she told Undark, and only woke up during the subsequent ambulance ride to the emergency room.

Afterward, she stopped taking Wellbutrin and later added a different drug to her regime, a generic version of the ADHD drug Concerta, but she said that she soon began to experience thoughts of self-harm every night: “I just mentally referred to it as Hell Hour.”

One day, Meadows missed a dose, and Hell Hour didn’t happen. Wondering if there was a connection between her prescription and the awful evenings, she changed to a generic version of Adderall. The brand-name drug and its generics are used by more than 41 million Americans with ADHD.

She had heard brand-name Adderall worked better, but the Food and Drug Administration announced a national shortage in October 2022. Since then, many ADHD patients have had difficulties getting the medication — and even its generics — at certain times. Meadows said the generic version often wore off by day’s end.

The difficulties Meadows experienced are not uncommon for people managing ADHD and other illnesses. Medications can affect people in unexpected and highly variable ways: A drug that works wonders for one person may be ineffective, or even harmful, for another. In part, that’s because the biological process that influences how a medicine works — how it is absorbed by tissues, digested, and metabolized in the body — can vary depending on a person’s age, sex, and myriad other personal factors.

Among those factors are genetics. In recent decades, researchers have uncovered numerous genetic variants that seem to play a role in people’s responses to painkillers, cancer drugs, heart disease pills, and other medicines, spawning a field known as pharmacogenetics or pharmacogenomics. The FDA lists a few hundred medicines linked with poor performance or concerning side effects in people who carry certain genetic variants.

The biological process that influences how a medicine works can vary depending on a person’s age, sex, and myriad other personal factors — including genetics.

For those carriers, the consequences can range from mild to severe. For instance, people with certain versions of the gene CYP2C9, which encodes a protein that breaks down common drug compounds, tend to digest ibuprofen slowly, allowing the drug to persist longer in their bodies and raising the risks of stomach bleeding and other side effects. More rarely, genetic variants can render common antibiotics such as vancomycin lethal.

The discovery of these gene-drug links has fueled a push by many researchers and health care providers to incorporate pharmacogenetic tests into standard health care — tests designed to identify a person’s troublesome variants ahead of time and help guide them to the medicines and doses that are best for them.

At a recent appointment, a physician’s assistant suggested to Meadows that a pharmacogenetic test might shed light on why certain drugs seemed to work poorly on her, and she jumped at the opportunity for answers. “Any scrap of information that we get is more than we know currently,” she said.

Chloe Meadows has experienced difficulties in finding an ADHD medication that works well for her. One clinician suggested pharmacogenetic testing.

Visual: Courtesy of Chloe Meadows

There is a growing body of evidence that this information can improve health outcomes. For about 15 years, international groups of experts have worked to develop evidence-backed pharmacogenetic recommendations. Last year, researchers conducted the first large clinical trial in Europe to assess whether a pharmacogenetic test administered before prescribing at least one of 39 drugs could help reduce rates of side effects. The study of more than 6,000 people, 93.5 percent of whom had variants associated with gene-drug interactions, reported that participants who received tests were 30 percent less likely to suffer severe adverse events.

But data also suggests that pharmacogenetic tests have been slow to gain traction in health care. A high number of people likely carry one or more variants that are thought to either raise the risk of side effects or diminish the performance of a prescription medicine, said Elizabeth Phillips, a professor of medicine at Vanderbilt University Medical Center in Tennessee. Only a fraction of people — some studies suggest less than 10 percent — ever get the tests.

Experts say that’s partly because clinicians, patients, and organizations that seek to establish testing standards often disagree on when, and to what extent, the tests are useful. In studies, it can be difficult to control for placebo effects and other factors that might confound results.

Moreover, the case for pharmacogenetic interventions — which hinges not only on the evidence linking variants to side effects, but also on individual perceptions of how harmful those side effects are — is stronger for some drugs and variants than for others, complicating efforts to standardize their use.

Evaluating such evidence is subjective, said Jasmine Luzum, a pharmacogenetics researcher at the University of Michigan, who added: “I’m really concerned that we’re not using pharmacogenetics more.”


One of the earliest, clearest pharmacogenetic success stories was a genetic test developed in the 2000s for an antiviral drug called abacavir, used to treat people living with HIV. At the time, people who took abacavir would sometimes develop rashes, fevers, or severe gastrointestinal symptoms, typically within days or weeks of starting the medication. Although symptoms usually subsided when the person stopped taking the pill, the real danger still lurked: If they tried taking abacavir again, the drug would trigger a massive, potentially fatal immune reaction.

“Nothing like it had ever been described before,” said Phillips, who recalls caring for many HIV patients at that time. In 2002, researchers discovered that a version of the gene HLA-B, which encodes proteins that play a critical part in the immune system, was associated with the response to abacavir, and Phillips and her colleagues found that a skin patch test could identify those at risk. A 2008 randomized controlled trial involving about 1,600 people found that about half the people who carried the variant were at risk of a severe reaction, whereas people who didn’t carry it never suffered that side effect. In another study, published later that year, the pattern held true across people who had otherwise varying genetic profiles. Around that time, the U.S. Department of Health and Human Services published guidelines recommending that clinicians test for the variant before prescribing abacavir.

The results were a boon to the young field of pharmacogenetics. They demonstrated, for the first time, that doctors could use genetic tests to preempt a drug’s problematic side effects.

Professor of medicine Elizabeth Phillips, along with her colleagues, found that a skin patch test could be used to identify patients at risk of reacting to the HIV drug abacavir. She has since identified potential genetic triggers for other drug reactions. Visual: Courtesy of Elizabeth Phillips

But, in many ways, abacavir was an easy target. When people living with HIV go to a clinician’s office, both doctor and patient know they will need medication, and so it’s typically not a problem to order a genetic test as part of a treatment plan and wait a few days for results. As Phillips recalls, it also helped that the clinicians who prescribed abacavir were often the same ones who witnessed and treated patients suffering from the drug’s side effects. It made it easier for them to put two and two together, and made them more invested in finding a solution to the problem of abacavir hypersensitivity, Phillips said.

Since the immune reaction to abacavir could be traced to a single variation, in a single gene, with a clear link to patient outcomes, the genetic test could easily help avoid the drug’s side effects, Phillips said. “Something like abacavir is really the low hanging fruit,” she added.

For other pharmacogenetic targets that have come along since, however, the links between genetic variants and drug performance have often been less straightforward — and so too have the choices for doctors.


As Tasha Tolliver remembers it, her 15-year-old daughter Izzy developed a fever in September 2015 after she began taking Bactrim, an antibiotic prescribed for acne. Then a puffy rash sprang up on Izzy’s torso, and the family consulted doctors. Izzy’s symptoms quickly worsened. Her body swelled so badly her face became tough to recognize. Skin peeled off her hands and feet. She suffered nausea and other symptoms that doctors would later recognize as signs of a failing liver. She spent months bouncing between the hospital and her home, continuing to get sicker even months after she stopped taking the drug.

At each hospital visit, Tolliver recalls, clinicians puzzled over her daughter’s symptoms, to no avail. After a week in the hospital, Izzy was discharged with a six-week prescription for steroids. The treatment helped her feel well enough to go to school. But as the drugs wore off, her symptoms flared up again. She fainted at school one Friday in November, and felt so much worse the next day that her family rushed her to the pediatrician’s office, and then to the hospital. The next day, she suffered fatal heart failure. “She was in school the day before she died,” Tolliver recalled.

Tasha Tolliver’s 15-year-old daughter Izzy experienced a severe drug reaction called DRESS in 2015, which led to Izzy’s death. Visual: Courtesy of Tasha Tolliver
Izzy and her younger sister, pictured before Izzy became ill. In the wake of her death, the family has searched for answers in pharmacogenetic testing. Visual: Tasha Tolliver

Only after an autopsy did doctors and the Tolliver family learn that Izzy had suffered a rare condition known as Drug Reaction with Eosinophilia and Systemic Symptoms, or DRESS, in which immune cells known as eosinophils multiply and infiltrate the liver, heart, and other organs.

The family’s subsequent grief and desperate search for answers was coupled with terror that Izzy’s younger sister, who was 13 at the time, might also be at risk of DRESS. (The condition has been linked to at least 44 drugs, including common antibiotics such as vancomycin, the gout drug allopurinol, and carbamazepine, a medicine to treat seizures.) The question drove the Tollivers to their pediatrician, who redirected them to a researcher, who suggested they speak with Phillips.

Phillips and other researchers have identified potential genetic triggers for DRESS. But as is often the case with gene-drug interactions, the connections are not straightforward. DRESS is a rare condition, and because it can be triggered by several different drugs — and because patients’ symptoms can range from a severe skin rash to organ failure — it has proven difficult to gauge just how many people are at risk of the condition, and how much of that risk can be attributed to genetic variations alone.

Tasha Tolliver said that she herself carries at least two genetic variants that have been linked to the syndrome — yet she has taken Bactrim in the past with no side effects. Tolliver’s younger daughter carries just one of the problematic variants, yet that’s no guarantee that she won’t have a severe reaction to Bactrim or one of the dozens of other drugs linked to DRESS. (To be safe, Tasha and her younger daughter avoid the antibiotic and certain other drugs.)

Izzy’s first symptom of DRESS was a fever, followed by a puffy rash that sprang up on her torso. At each hospital visit over the next months, clinicians puzzled the symptoms, to no avail. Visual: Courtesy of Tasha Tolliver
Izzy’s symptoms quickly worsened to swelling and skin peeling. Treatment with steroids helped her feel better, but as the drugs wore off, her symptoms flared and she died of heart failure. Visual: Courtesy of Tasha Tolliver

Researchers suspect that DRESS is caused by a combination of genetic and non-genetic factors, such as high levels of a virus named HHV-6, which is typically latent and harmless in most people but can flare up to cause symptoms in people who are severely immunocompromised. In people susceptible to DRESS, drugs can also trigger those viral flareups. Tolliver recalls that HHV-6 levels in her daughter Izzy’s blood were about 200 times higher than normal when she died.

The many nuances that influence physiological responses to medication can make it difficult to conclusively pin bad drug reactions to specific genetic variants. While some genetic tests can provide yes-no answers and a clear course of clinical action, that’s not always the case in pharmacogenetics, said Minoli Perera, a pharmacogenomics researcher at Northwestern University Feinberg School of Medicine. “It can be clear cut. It may not be,” she said. “It’s more complicated than just a straight ‘you carry a variant, and the answer is X’.”


In recent years, organizations like the FDA, the Clinical Pharmacogenetics Implementation Consortium, or CPIC, and the Dutch Pharmacogenetics Working Group, or DPWG, have sought to cut through the uncertainty surrounding pharmacogenetic tests, aiming to bring standardization to the field. CPIC and DPWG convene experts to assess scientific evidence on gene-drug interactions and translate that knowledge into practical guidelines that clinicians can use: lists of known gene-drug associations, for instance, and recommendations for modifying prescriptions or doses for patients with troublesome variants.

But, as with most scientific fields, the accumulating sea of pharmacogenetic results is marked by gray areas, and the guidelines these organizations offer can conflict. A 2022 analysis led by research geneticist Daryl Pritchard, the senior vice president of science policy at the Personalized Medicine Coalition, found that, although the FDA’s list included 106 medicines with known gene-drug interactions and CPIC’s list covered 59, only 39 were common to the two. The two groups often differed on which genetic variants were linked to which drugs — and on the steps clinicians should take in light of those gene-drug interactions. Only for five of the drugs did the two lists fully agree.

“It’s more complicated than just a straight ‘you carry a variant, and the answer is X’.”

Moreover, companies that provide pharmacogenetic testing services are typically not beholden to the recommendations of CPIC, or other bodies. Although they take cues from organizations such as CPIC, they often also use their own proprietary algorithms to analyze and interpret data.

In April 2019, the FDA warned Virginia-based Inova Genomics Laboratory about illegally marketing pharmacogenetic tests without agency review; the FDA previously sent a more general statement directed at consumers about pharmacogenetic tests. “A lot of the variants had very low levels of evidence or had been published in one study and weren’t replicated,” said Michael Eadon, a nephrologist and pharmacogenetics researcher at Indiana University School of Medicine, speaking about companies that offer these tests.

In a 2020 case report, a group of researchers described the conundrum that testing inconsistencies can create for patients. The study recounted the experience of a 65-year-old woman with anxiety and depression who took pharmacogenetic tests from two different companies. One company’s report informed her that she carried a variant of a gene called CYP2D6 that made her a poor metabolizer of antidepressant drugs known as selective serotonin reuptake inhibitors, or SSRIs. The other company identified the same variant but concluded that it made her an intermediate metabolizer of SSRIs. The conflicting reports “worsened the patient’s anxiety surrounding medication response,” the study authors wrote.

Both companies are “considered reputable in the industry and stand by their interpretations of the tests,” the authors wrote. However, they added, “the evidence supporting their claims is not easily accessible and the thought process is not transparent.”


Although groups like CPIC have worked to standardize the interpretation of pharmacogenetic tests, they typically stop short of recommending that a patient should get a genetic test before being prescribed a drug.

“We don’t say whether or not we think the evidence is strong that you should order a genetic test for this patient,” said Luzum, who works with CPIC occasionally. But if a patient happens to have their test results because they or a clinician ordered them, an interested clinician could potentially turn to CPIC’s database to understand the implications.

Even the FDA rarely makes recommendations on whether a pharmacogenetic test should be given before a prescription. As a result, the clinical decision to use — or not use — the testing in patient care depends largely on the clinician, the health care facility, and a patient’s condition.

Jasmine Luzum, a pharmacogenetics researcher at the University of Michigan, has studied variability in how clinicians use the results of testing, and how that impacts patient care. Visual: Courtesy of Jasmine Luzum
Michael Eadon, a nephrologist and pharmacogenetics researcher at Indiana University School of Medicine, has worked to implement pharmacogenetic testing for about 20 years. Visual: Courtesy of Michael Eadon

In a 2021 study, Luzum and her colleagues described how this variability can impact care. They chronicled the disparate experiences of two patients. One, a 55-year-old clinical geneticist who was diagnosed with breast cancer, discussed with her oncologist the scientific literature on genetic variants that could influence her response to tamoxifen, a common breast cancer drug. CPIC guidance suggested that variants of the gene CYP2D6 were linked to slower drug metabolism and to poorer survival rates among people who take tamoxifen. But guidelines from the American Society of Clinical Oncology and the National Comprehensive Cancer Network didn’t recommend pharmacogenetic testing for tamoxifen responses. The patient advocated for the test anyway and, based on the results, was prescribed an alternative to tamoxifen.

The 2021 study also recounted the experience of a 45-year-old woman who requested a pharmacogenetic test through her primary care provider, in hopes of understanding her often-troublesome responses to drugs for depression, and acid reflux. Although her primary care doctor ordered the tests, one specialist put the results in her chart and never discussed them with her, while another dismissed the results entirely.

Luzum said that clinicians may overlook the results of a genetic test in this way if a patient’s medical history suggests that the benefits of a drug will outweigh the genetics-related risks. She notes that genetics are not the only factors that hint at how helpful or harmful a medicine will be — nor are they the most firmly established. Often, patients and clinicians must weigh genetic results against other health markers, such as a person’s family history of heart disease, that have long been used to make clinical decisions, Luzum said. For clinicians, she added, “it’s a very personalized decision how strong that evidence needs to be for you to change what you do.”

“Patients are willing to accept pharmacogenetics at way lower levels of evidence because they’re the ones taking the medication.”

Health insurers have also been slow to embrace pharmacogenetics. Medical geneticist Patricia Dickson, who is in charge of the pharmacogenetic clinic at Washington University School of Medicine in St. Louis, said their team has “had zero luck in getting pharmacogenomics testing covered by insurance.” Emily Cicali, a University of Florida pharmacist specializing in pharmacogenetics, said that in her experience the tests are sometimes covered, but it depends on the health insurance company and the reason for the test.

“We’re kind of in a variable landscape right now,” Cicali said. Even when insurance companies pay for the test itself, she added, the interpretation service needed to translate dozens of pages of genetic details into something a clinician can use is never covered at their clinic.

Luzum suggested that tension can arise when patients, clinicians, and insurers differ on the question of how much evidence is enough to justify making decisions based on a pharmacogenetic test. “Patients are willing to accept pharmacogenetics at way lower levels of evidence because they’re the ones taking the medication,” Luzum said. “They’re like, we want to know if this drug is going to work on me.”


Researchers and clinicians have long argued that clinical and logistical hurdles are complicating access to pharmacogenetic tests and preventing tests from reaching many people who might benefit from them. (An exception, said Vanderbilt University’s Phillips, is the field of cancer, where genetic tests are widely used for a variety of decisions.)

Across the U.S., programs have sprung up at several medical centers and other health care facilities to try to streamline the use of pharmacogenetic information in clinical decision-making. In these settings, clinicians, pharmacists, and, occasionally, genetic counselors work together to figure out how to use test results, often using CPIC guidance and FDA information.

“At the end of the day, the doctor who is prescribing the medication has to be invested, has to understand the results, has to be able to be comfortable using those results,” said Cicali, who works with a program at the University of Florida that aims to help providers throughout the state get pharmacogenetic testing and learn how to use it.

Cicali and her colleagues have sought, among other things, to simplify the steps doctors must take to order pharmacogenetic tests for their patients.

“They want to go order the test and then are kind of hit with, oh, it’s not so easy to order,” Cicali said. To smooth that process, the program works with an on-site laboratory that it is equipped to do in-house genetic testing — an approach other centers have also taken.

Although pharmacogenetic tests offered in the University of Florida program may be covered by insurance, the services offered by Cicali’s team — which include interpreting the results and placing them in the electronic health record in a way that clinicians can use — costs a few hundred dollars and are not typically reimbursed.

But experts say these services are essential to incorporating pharmacogenetics into health care. “Anyone can order pharmacogenomic testing, it’s the interpretation that’s really hard,” said Washington University’s Dickson. In her experience, having pharmacogenetic results in hand — without adequate support from pharmacists and other experts — is “almost more of a headache for most doctors,” simply because “they don’t know what to do about these results.”

Emily Cicali (left) provides a consultation with a patient seeking pharmacogenetic testing. Through a program at the University of Florida, Cicali and her team help doctors order the tests, and they also interpret and place the tests in the electronic health record in a way that clinicians can use. Visual: UF Health

At Indiana University, nephrologist Michael Eadon and his colleagues have worked to implement pharmacogenetic testing for a variety of drugs for about 20 years. In a 2022 study, they tested whether clinicians who were provided genetic test results for chronic kidney disease patients would consider that data when prescribing blood pressure medications. At the conclusion of the year-long study, clinicians said they discussed the results with about 85 percent of their patients, and roughly one in three said they adjusted medications based on the genetic testing results.

“Some of the clinicians were very excited to participate,” Eadon said, and “other clinicians said, ‘thank you very much, you can enroll my patients, but there’s no way I will ever use any of this information.’”

Eadon, who is working with CPIC to develop pharmacogenetic standards for a blood pressure drug named hydralazine said doctors’ views on the value of pharmacogenetic results are partly subjective. Research suggests, for instance, that the effectiveness of the blood pressure medicines thiazides depends in part on which variant of YEATS4 — a gene thought to be involved in making RNA — a person carries. In one study, patients with one version of the gene saw their blood pressure drop on average by 13 mm of mercury, whereas people who carried a different version of the gene saw a drop of only 9 mm of mercury. (Systolic and diastolic blood pressures at or below 120 mm Hg and 80 mm Hg, respectively, are considered healthy.) “Some providers would look at that information and say, well, it looks like the thiazide still works at nine millimeters of mercury, it’s only a 4 millimeter of mercury difference, I’m going to ignore this information,” Eadon said. “And other providers might say, well, that might mean the difference between having to be on one drug or two drugs to control your blood pressure.”

“As a society, we have kind of accepted this idea that we’re just going to try this drug and some people will have this negative effect. But why have a bad effect if you don’t have to?”

But some health care providers neglect pharmacogenetic information because of long-standing habits, according to Daryl Pritchard, at the Personalized Medicine Coalition, an advocacy group that, among other things, seeks to expand the field of available pharmacogenetic tests and therapies. They reason that most patients with conditions where there are drugs known to be effective might bounce between medicines, but will eventually land on one that controls their symptoms, Pritchard said. A genetic test might speed up that journey, but simply trying out different drugs can seem simpler and less onerous. Pritchard described this trial-and-error culture in medicine as perhaps pharmacogenetics’ greatest challenge — a sentiment that others echoed.

“As a society, we have kind of accepted this idea that we’re just going to try this drug and some people will have this negative effect,” Dickson said. “But why have a bad effect if you don’t have to?”


This May, Chloe Meadows received the results of the pharmacogenetic tests she hoped would help end the cycle of bad drug reactions she’s endured in recent years and steer her to more effective treatments for her ADHD. They came in the form of a document about 17-pages long that her clinician gave her at an appointment. Hardly surprising for this still-maturing subfield of medicine, she said that the report answered some questions, and left many others open.

The report did suggest that Meadows had genetic variants that may have affected her response to some of the drugs she had tried in the past, she told Undark, including the drug that she believes triggered her seizures.

But Meadows’ genetic makeup did not explain why the generic version of Adderall seemed to work poorly for her; the variants on the test indicated she is a normal metabolizer of the drug, she said. Overall, the results provided “a lot of information and not a lot of conclusive answers.”

Still, Meadows said she’s grateful for those scraps of new knowledge, and relieved to know that she does not have genetic susceptibilities to many other ADHD medications. She expects to have to try other ADHD medicines in the future, especially given the ongoing Adderall shortage: “It is nice to know that even if I have issues with them for other reasons, that I’ve ruled out issues on some front.”


Reporting for this story was supported, in part, by a fellowship from the Good Science Project and the Johns Hopkins University MA in Science Writing program.

Jyoti Madhusoodanan is a senior contributor to Undark and a science writer based in Portland, Oregon.