It starts with a fever, a cough, red eyes, the kind of symptoms that might accompany a bad cold. Then, everything gets much worse. A rash appears on the skin, with target-like dots that blister and burst. Lips crust and bleed. Eyes become inflamed. The throat develops such painful ulcers that swallowing becomes impossible. Skin starts peeling off in sheets. The damage is so extensive that patients typically end up in burn wards, in excruciating pain, unable to eat, urinate, or open their eyes. Many beg to die; some do.

In its milder forms, this condition is called Stevens-Johnson syndrome (SJS). If more than a third of a person's skin peels off, it becomes known as toxic epidermal necrolysis (TEN). Either way, “it's the worst thing I’ve seen in 30 years in clinical medicine,” says Teri Manolio at the National Human Genome Research Institute.

SJS/TEN is a disease of devastating irony. Most cases happen when people take drugs that are meant to improve their health and their bodies revolt in catastrophic fashion. These hypersensitivity reactions are rare. They are only triggered by certain drugs, and only in people with specific genetic variants in a cluster of immunity genes. And over the past decade, scientists have identified many of these ruinous drug-gene combinations.

Which means that SJS/TEN should be almost entirely preventable.

As Wasun Chantratita from Mahidol University puts it, it's “the low-hanging fruit of genomic medicine.” To deal with it, you don't need to edit genes, or turn to stem cells, or prescribe drugs that target mutations in a patient's DNA. You just need to screen people for the risky variants and withhold the triggering drugs.

But, as is often the case, things aren't that straightforward. The case of SJS/TEN shows how the promise of genomic medicine, even in simple cases, can be stymied by mundane obstacles.

Back in 2002, SJS/TEN was a regular feature in the dermatology ward of Taiwan's Chang Gung Memorial Hospital. “Two patients were admitted every week and this disease was supposedly ‘rare,’” recalls Wen-Hung Chung, then a newly minted attending physician. Chung noticed that everyone who got the condition had taken either carbamazepine, used to treat epilepsy and seizures, or allopurinol, used to treat gout. But he didn't know why some people absorbed these medicines calmly, while others reacted disastrously.

To Chung's wife, the immunologist Shuen-Iu Hung, the condition looked like some kind of berserk immune response, and the duo wondered if some patients had genes that predisposed them to such violent reactions. The HLA genes were the most likely candidates. These make proteins that grab suspicious molecules and brandish them for the immune system to inspect. They come in a vast smorgasbord of variants, each one shaped to recognize a different molecule. Perhaps some people have HLA variants that grab onto drugs, triggering misplaced immune assaults. And since the drugs typically diffuse throughout a patient's body, these assaults lead to systemic inflammation.

By sequencing the HLA genes in their patients, Chung and Hung found one variant—HLA-B*15:02—that shows up in every single person who developed SJS/TEN after taking carbamazepine, compared to just 3 percent of patients who tolerated the drug. That result, published in 2004, became a classic example of pharmacogenomics—the study of how our genes affect our responses to drugs.

More examples followed. Chung and other scientists soon identified many other drugs that trigger SJS/TEN, including anti-HIV medications like abacavir and nevirapine, and anti-convulsants like phenytoina and lamotrigine. Each drug is associated with its own particular risky HLA variants.

The obvious solution would be to phase out these drugs and, to an extent, that's happened. In Taiwan, carbamazepine prescriptions have fallen by around 70 percent in the last decade, as doctors have swapped to a safer second generation of anti-epileptics. But since these newer drugs are also more expensive, some countries like Thailand have found it harder to make the switch.  “In developed countries, you’d use the new drug first,” says Chantratita. “But in our country, the majority of medications are old drugs.”

The solution is a genetic test. Chung predicted this in his 2004 paper, concluding: “It should be possible to exploit this association in a highly reliable test to predict severe adverse reaction.” Indeed, it was. In Taiwan, Thailand, Singapore, and other Asian countries, doctors now carry out routine genetic tests for the HLA-B*15:02 variant before prescribing carbamazepine. It takes just 24 hours and if anyone tests positive, they get the newer, more expensive drugs.

These tests might seem expensive for cash-starved medical systems, but they pale in comparison to the costs of caring for seriously ill SJS/TEN patients. So in 2010, Taiwan's national health insurance started covering the cost of the screening test. Thailand did the same after an SJS/TEN patient who was blinded by her condition successfully sued the government.

These steps have worked resoundingly well. A decade ago, dozens of people would check into Taiwanese hospitals every year with SJS/TEN; today, there's just a handful of annual cases. This is, as one review put it, a success story in genomic medicine from resource-limited countries.

But SJS/TEN stubbornly refuses to die out completely, which Chantratita ascribes to human error. “Our hospitals are very congested,” he says. “A doctor might ask you to have a blood test but when the information comes back, it's not put in your file. A second doctor comes along and sees no notification, so prescribes a lethal drug.” An electronic-health system that automatically notified doctors and pharmacists of a patient's test results would solve these problems, but no such system exists. In the era of genomic medicine, lives are still being lost because of shortfalls in IT.

Chantratita's colleagues at Bangkok's Ramathibodi Hospital opted for a low-tech approach. After every patient gets an HLA test, their results are entered into a “pharmacogenomic wallet card,” a vibrant fuchsia rectangle that they carry around and show to future doctors. It's simple, cost-effective, and although it's still being formally evaluated, Chantratita says that all signs suggest that it's working.

“I just presented the data at a U.K. meeting, and I almost didn’t want to,” he says. “These guys were talking about whole-genome sequencing for millions of people, and we just have a simple plastic card. But many people said: Wow, we should do that here.”

The genetic variants that are most commonly associated with SJS/TEN are more common in parts of Asia. For example, HLA-B*15:02 is found in up to 8 percent of people in Taiwan, Thailand and other parts of South-East Asia, but is vanishingly rare in Japan, Korea, or Europe. This explains not only why SJS/TEN is more common in South-East Asia, but also why scientists there have taken the lead in understanding and managing the condition.

But SJS/TEN is a global problem. Western countries have their own deadly combinations of drugs and HLA variants. “I'm thrilled about what Taiwan and Thailand have done, but look how poorly it's played out in resource-intense countries,” says Neil Shear from the University of Toronto.

There are a few exceptions. In 2002, Elizabeth Phillips, now at Vanderbilt University Medical Center, showed that people with the HLA-B*57:01 variant can develop SJS/TEN after taking abacavir. That variant is common in Europe and has become the subject of routine testing, largely because abacavir is used specifically to treat HIV. “HIV physicians are an organized group who practice guideline-based medicine,” says Phillips. “It was their drug that had this solvable problem.” By contrast, other SJS/TEN-inducing medications are used more broadly; carbamazepine, for example, is sometimes prescribed by dentists for jaw pain.

“Reaching so many different kinds of physicians with this knowledge is a huge challenge,” Phillips says. “So this has to be codified. A box has to pop up in an electronic medical record that says: you're prescribing carbamazepine, have you done this test? It's too hard for physicians to remember without these kinds of prompts and support measures.”

The low awareness and prevalence of SJS/TEN doesn't help. Nor does the fact that drug-hypersensitivity reactions are often ignored in the West, drowned out by more (in)famous problems like cancer, heart disease, or diabetes.  But here, too, things are changing.

In January 2014, Chantratita's simple wallet-card system floored the delegates at a genomic-medicine meeting in Washington, D.C. “Here was a real opportunity, not only a genetic finding but a way of implementing it,” says Teri Manolio. “I went back to the NIH and asked what we were funding on SJS/TEN and we had nothing.” Manolio organized a two-day meeting in March 2015 to discuss research priorities for the disease, and how to monitor or manage drug-hypersensitivity reactions more broadly. Chung and Chantratita were there, as were scientists from Indonesia, Singapore, Europe, and the USA.

“With NIH's involvement, we can actually get the ball rolling,” says Shear. “Every dean of medicine says: Blah blah blah, personalized medicine, blah blah blah. Well here’s an opportunity to do something. It's personalized medicine, and it would direct people to safer therapies. I'd like to see this disease disappear.”