In 2001, the American Food and Drug Administration (FDA) granted accelerated approval to imatinib, a drug for chronic myelocytic leukaemia. The FDA’s haste was unusual, although the reasoning behind it was not. It turned out that patients on imatinib achieved complete remission at rates that left oncologists largely speechless. The drug worked because it targeted a specific genetic mutation (the so-called Philadelphia chromosome) rather than battering all rapidly dividing cells indiscriminately, as chemotherapy does. Medicine had just glimpsed a different future.
Two and a half decades on, we tend to call that future personalised medicine (or, sometimes, precision medicine). In essence, it uses an individual’s genetic profile to guide diagnosis, drug selection, and dosing. Its recent expansion has been remarkable, brought about primarily by the reduction in cost of sequencing a human genome from 100 million US dollars in 2001 to under 100 US dollars today. The FDA now counts personalised treatments among more than a quarter of all new drug approvals since 2014. The global market for precision medicine, valued at roughly 152 billion US dollars last year, is projected to triple within a decade.
In many ways, personalised medicine is the greatest reinvention of medicine in a generation. But there are, as we shall see, some caveats.
The case for yes
Targeted therapies based on next-generation sequencing have demonstrated response rates of up to 85 per cent in certain cancers, against far lower rates for conventional chemotherapy. Identifying EGFR mutations in non-small cell lung cancer, or BRAF V600E in melanoma, now guides treatment selection as a matter of routine. Some patients once given a year are living four.
What’s more, the scope keeps widening. Pharmacogenomics, which matches dosing to an individual’s genetic metabolism profile, already shapes how many hospitals prescribe warfarin and some chemotherapy agents, cutting unnecessary toxicity.
Genomics England’s Generation Study aims to sequence 100,000 newborns, turning genetic risk screening into a population-level public health instrument. A 2025 paper in Nature Communications outlined a framework for integrating electronic medical records with real-time post-genomic data, something that represents a shift in medicine, at least in principle, from reactive to anticipatory.
The problem
The trouble is that the database underpinning much of this is small. As of 2023, 86.5 per cent of participants in genome-wide association studies were of European descent; participants of African ancestry accounted for less than 0.5 per cent. Genetic risk variants are population-specific. Polygenic risk scores calibrated on European cohorts perform materially worse in African, South Asian, and East Asian populations. The ‘personalised’ in personalised medicine, for a large portion of the world, is personalised for someone else.
Access compounds the problem, given that genomic testing remains inconsistently covered by insurers and public health systems alike, and the targeted therapies that follow often come with eye-watering costs. In the Philippines, one widely cited targeted cancer drug costs roughly the entire national drug budget. A 2024 review confirmed that lower rates of BRCA testing among Black women in the United States translate directly into lower uptake of risk-reducing interventions not because the science does not apply, but because the referrals are not made. The reinvention, on its current trajectory, risks deepening rather than dissolving the inequalities it might in theory address.
There is also a more fundamental problem, identified in that same Nature Communications paper is the fact that the integration of electronic medical records and genetic data has, in practice, “largely fallen short” of early expectations, partly because environmental factors (diet, pollution, housing) remain largely absent from clinical genomic models. Genes, it turns out, are not destiny. They interact with lives. A precision framework that ignores context is not really precise.
The honest verdict
Personalised medicine is real reinvention, in oncology at least. Its near-term potential in pharmacogenomics, rare diseases, and neonatal care is credible, grounded in data, and not going away. The collapse in sequencing costs has genuinely changed what is possible.
But reinvention of a corner of healthcare for a subset of patients is not reinvention of healthcare. The tools are powerful, the data infrastructure remains skewed, and the access architecture has not caught up with the science. If the AI models now being deployed to mine genomic datasets amplify the biases baked into those datasets (as researchers warn is likely) the gap could widen before it narrows.
Imatinib transformed outcomes for chronic myelocytic leukaemia. Twenty-five years later, it sits on the World Health Organisation’s list of essential medicines, available as a generic. That journey from breakthrough to broad access took a generation. For personalised medicine to earn its reinvention claims, the question is whether the same transition can be compressed, and just how many of us will benefit.
Photo: Dreamstime.
