Diagnosis is an act of naming: the point at which we are told what kind of health problem we’re dealing with. With a successful diagnosis, uncertainty ends, explanation is offered and appropriate treatment can start. From a biological point of view, however, diagnosis arrives late.
By the time a condition like type 2 diabetes is named, the biological processes that produced it have been running for years. Genetic susceptibility interacts with ]environmental exposures, social context, and a series of compensatory physiological responses. Before blood glucose crosses the diagnostic threshold for diabetes, the body has already adapted, struggled and partially failed.
Diagnosis is therefore not the start of disease. It’s the moment when accumulated biological change becomes visible.
The moment of diagnosis is important because the diagnostic label signals the arrival of a “system state” that meets agreed criteria and qualifies for care. Naming triggers action: patients get an explanation, clinicians get a framework for treatment, and healthcare systems unlock care pathways, medications and resources.
Because of this, we strive to diagnose earlier. The hope is that earlier intervention will reverse what is reversible, slow what cannot be reversed, and prevent downstream damage.
Precision medicine complicates this logic.
As measurement becomes more precise, diagnosis can come earlier. It also becomes narrower and more individualised. Three problems follow.
- First, diagnosis that’s too early can label people who are at the edge of pathology rather than firmly inside it. Treatment may be unnecessary. Harms may outweigh benefits. Risk becomes disease by definition rather than through its physical consequences.
- Second, as we measure more molecules, image tissues in greater detail, and stratify more finely, people who share a diagnosis begin to share less biology. The common core of disease thins out. Ultimately, we arrive at an “n of 1”.
- Third, as the group dissolves, so does the evidence base. Randomised clinical trials depend on defining reasonably homogeneous populations united by a shared pathology. When the commonality fragments, it becomes harder to test treatments, compare outcomes or generalise results.
At this point a deeper question emerges. If disease can no longer be easily defined, what exactly are we diagnosing? And if “disease” becomes unstable, what does “health” mean?
These are not philosophical curiosities. They have practical consequences for patients, clinicians and healthcare systems.
Diagnosis, it turns out, is both a beginning and an endpoint.
Type 2 diabetes is often treated as a single disease. In reality, it is one of the clearest examples of how a diagnosis persists even as the biology it names disassembles.
Clinically, type 2 diabetes is defined by hyperglycaemia. Biologically, people arrive at their hyperglycaemia by different routes. In some, insulin resistance dominates: insulin is produced but tissues fail to respond. In others, insulin secretion fails early: the pancreas cannot meet even modest demands. Many fall somewhere in between.
When patients with newly diagnosed diabetes are grouped using routine clinical features, several reproducible subtypes emerge, each with different trajectories, risks of complications, and responses to treatment. Some develop kidney disease early, others eye disease. Some require insulin soon after diagnosis; others respond well to lifestyle change and oral medication.
None of this is a surprise to clinicians. Many have long suspected that “type 2 diabetes” is a convenient umbrella rather than a single entity. Precision medicine has given us the tools to show this systematically.
Diabetes in Africa: precision reveals difference
Much of what we think we know about type 2 diabetes is based on studies conducted in European or North American populations.
In multiple studies, people of African ancestry develop type 2 diabetes at younger ages and at lower body mass index than their European counterparts. Many present with more severe hyperglycaemia despite being relatively lean. In some populations, markers of insulin secretion are lower, suggesting earlier pancreatic beta-cell failure rather than predominant insulin resistance.
A lean, middle-aged patient of African ancestry with diabetes may not fit the mental model clinicians associate with “type 2 diabetes”, leading to delayed diagnosis, misclassification, or suboptimal treatment. In some cases, patients present with episodes of ketoacidosis, traditionally associated with type 1 diabetes, yet later manage without insulin, blurring conventional categories.
Even our diagnostic tools are not neutral. HbA1c, used to diagnose and monitor diabetes, is influenced by red blood cell biology. Genetic traits such as alpha-thalassaemia, common in certain African populations, can lower HbA1c independently of glucose levels. Thresholds developed in European populations may therefore under-diagnose or delay diagnosis when applied uncritically elsewhere.
Precision medicine exposes the limits of universal categories and reveals how disease definitions are often population-specific artefacts rather than biological constants.
If the same diagnostic word is applied to systematically different biological states, what exactly is it capturing?
Not just diabetes
These issues are not unique to diabetes.
Asthma varies widely in how it presents and responds to treatment. Breast cancer is no longer one disease but several molecular subtypes with different outcomes. Heart failure has distinct subgroups based on how the heart functions. Depression also varies greatly between individuals, with recognised clinical subtypes. Neurodegenerative disorders like Alzheimers show biological and clinical variation that suggests distinct subgroups.
As measurement improves, diseases break up.
This does not mean disease labels are wrong, just that they are imprecise tools, useful starting points rather than complete explanations.
What happens when the group disappears?
Medicine is organised around diseases as stable entities. Clinical trials, guidelines, reimbursement, training programmes and quality metrics all assume that diseases define meaningful groups.
Randomised clinical trials, still the gold standard for evidence, compare interventions in populations unified by a shared diagnosis. As shared biology thins out, trial results become averages over increasingly heterogeneous groups.
At the same time, healthcare systems struggle to make use of finer, more precise stratification. If “type 2 diabetes” becomes five subtypes, each behaving differently, how should guidelines be written and formularies be designed? How should clinicians be trained? And do these changes actually improve population health, or individual care?
…..
Heterogeneity of diabetes figure in Leslie, Richard David et al. The Lancet Diabetes & Endocrinology, 11(1): 848 – 860 (Nov, 2023). Each dot represents a person with diabetes and each colour shows the relative contribution of different biological pathways to their disease. Most people with type 2 diabetes have a mix of causes (grey). A minority have diabetes driven mainly by one dominant mechanism. Type 1 and monogenic diabetes have distinct biological drivers, but some individuals show overlapping features.
…..
There is risk at both extremes. Ignore heterogeneity and care becomes blunt and inefficient. Embrace it without restraint and we lose the ability to generate evidence, scale interventions, or organise care.
This is not a technical inconvenience so much as a structural challenge to how modern medicine produces knowledge.
Does disease disappear in the age of precision medicine?
Diagnosis is still essential. It is how suffering becomes actionable in healthcare systems, patients gain access to care and clinicians coordinate action. Without diagnosis, medicine is incoherent.
But diagnosis is no longer a final explanation. Biologically, it marks the point at which accumulated dysregulation crosses a threshold that we can name. Operationally, it is a beginning: the trigger for care, treatment, and support.
In the era of precision medicine, diagnosis must be understood as provisional – a useful fiction that enables action but requires follow-up questions. Which subtype? Which mechanisms? Which risks? Which trajectory?
The danger is not that disease disappears. The danger is that we cling too tightly to disease labels when biology has already moved on.
Diagnosis is still medicine’s most powerful tool. In the age of precision medicine it may not be the final destination but where better understanding begins.
Readings
Ahlqvist E, Storm P, Käräjämäki A, et al. Novel subgroups of adult-onset diabetes and their association with outcomes: A data-driven cluster analysis of six variables. The Lancet Diabetes & Endocrinology. 2018;6(5):361–369. DOI: 10.1016/S2213-8587(18)30051-2.
https://www.thelancet.com/journals/landia/article/PIIS2213-8587(18)30051-2/fulltext