Thomas M. is forty-five when he seeks advice for what he himself describes as “normal fatigue.” He heads the IT department of an international company, works primarily in front of several screens and travels regularly between different European sites. His schedule is meticulously organised. So is his medical follow-up.

Every year, Thomas undergoes routine laboratory testing. Every year, the results lead to continued monitoring that he finds reassuring. No acute event, no new diagnosis, no clear disruption capable of explaining what he is experiencing.

“Nothing concerning. We will simply continue to monitor it.”

This conclusion suits him all the more because it corresponds to his own interpretation: he works hard, moves less than he used to and is getting older. Yet, over the years, his daily functioning changes almost imperceptibly. He recovers less easily after a working week, gradually gains weight, his concentration becomes more variable and his sleep, although apparently adequate in duration, no longer leaves him feeling equally restored.

None of these elements is specific. Considered separately, each can be given a reasonable explanation: professional workload, reduced physical activity, stress, age, or the accumulation of responsibilities.

This interpretation is logical.

It becomes incomplete when each year is no longer examined in isolation.

When Thomas’s successive laboratory results are placed on a single timeline, the nature of the case changes.

In 2020, his fasting glucose is measured at 0.89 g/L. The following year, it reaches 0.93 g/L. In 2022, it is 0.98 g/L. In 2023, it reaches 1.03 g/L, followed by 1.08 g/L in 2024.

These figures require an important clarification. Under current American Diabetes Association criteria, fasting plasma glucose between 1.00 and 1.25 g/L falls within the range defining impaired fasting glucose, one of the possible criteria for prediabetes. The 2023 and 2024 results should therefore not be described as simply “normal” under ADA criteria. Their clinical interpretation nevertheless depends on the conditions under which the samples were obtained, reproducibility, other glycaemic parameters and the overall medical assessment.

This clarification does not weaken the reasoning of the case. It makes it more informative.

The information does not suddenly appear only when Thomas reaches 1.03 g/L. It is also contained in the way his profile had already been shifting over the preceding years.

The question is therefore no longer simply: “What is the value today?” It becomes: “What trajectory led to this value?”

Diagnostic thresholds are indispensable to medicine. They help define clinical states, standardise decision-making and prevent biological variation from being arbitrarily transformed into disease.

But a threshold and a trajectory do not answer exactly the same question.

A threshold helps determine whether, at the time of assessment, the criteria used to define a diagnostic category are present.

A longitudinal trajectory examines something different: direction, rate of change and, potentially, the concordance of several changes over time.

The ADA itself emphasises that fasting plasma glucose, HbA1c and two-hour glucose during an oral glucose tolerance test assess different aspects of glucose metabolism and do not identify exactly the same populations. The risk of type 2 diabetes is also continuous and does not biologically emerge at the precise moment a diagnostic boundary is crossed.

This prevents two symmetrical shortcuts.

A value below a threshold should not be artificially transformed into pathology. But the absence of threshold crossing does not mean that the trajectory leading towards that threshold contains no clinically relevant information.

In Thomas’s case, glucose is not changing in isolation.

His weight gradually increases. His waist circumference expands. His blood pressure shifts towards higher values. Recovery becomes less satisfactory. Fatigue occupies more of his daily life and concentration becomes less stable.

None of these elements, in isolation, is sufficient to establish a single mechanism.

Weight gain does not demonstrate insulin resistance. An increase in fasting glucose is not sufficient to explain fatigue. A change in blood pressure cannot, by itself, attribute the entire clinical picture to a single metabolic origin.

But when several variables shift during the same period, the clinical question moves to a different level.

It is no longer enough to ask each parameter separately:

“Are you pathological?”

It becomes necessary to ask the dataset as a whole:

“What evolution are you describing?”

One relevant hypothesis concerns the relationship between insulin sensitivity, insulin secretion and the maintenance of glucose homeostasis.

When tissue sensitivity to insulin decreases, pancreatic β-cells may increase their secretory response in order to compensate for that reduced sensitivity. As long as this adaptation remains sufficient, glucose tolerance may remain relatively preserved. The literature therefore describes a compensatory phase during which increased insulin secretion contributes to maintaining glucose homeostasis despite reduced insulin sensitivity.

This relationship matters because it demonstrates that a glucose concentration alone does not describe all of the mechanisms required to maintain it.

But it immediately requires a second precaution.

In Thomas’s case, the data presented here include neither an insulin measurement that can be meaningfully interpreted in this context, nor a direct measure of insulin sensitivity, nor a disposition index, nor another assessment capable of establishing that this compensatory mechanism is actually responsible for his trajectory.

Compensatory insulin secretion therefore represents a plausible physiological mechanism to investigate, not a diagnosis that these figures allow us to establish.

This distinction between an observable result and the regulatory mechanisms underlying it is central.

In the progression of type 2 diabetes, pathophysiological models describe a dynamic interaction between insulin sensitivity and the ability of β-cells to adapt insulin secretion. As long as compensation remains sufficient, reduced insulin sensitivity does not necessarily result in hyperglycaemia corresponding to overt diabetes.

When this adaptive capacity becomes insufficient relative to metabolic demand, glucose regulation progressively deteriorates.

This obviously does not mean that every person whose glucose concentration gradually increases is following this trajectory.

It means that rigorous clinical interpretation must distinguish three levels: the measured value, the mechanism that might explain it, and the actual demonstration of that mechanism in the individual patient.

This is where Thomas’s case becomes more informative than any one of his individual assessments.

A consultation provides a cross-sectional view. Five years of data provide longitudinal information. These two forms of information are not interchangeable.

A few additional hundredths of a gram per litre of glucose may appear minimally informative over one year. A few kilograms may be attributed to lifestyle. Slightly higher blood pressure may have many explanations. Less effective recovery remains an extremely nonspecific symptom.

But when these changes are placed back into sequence, their possible concordance becomes visible.

Time does not transform an association into causality.

It can, however, reveal what a succession of isolated measurements may conceal: the direction of the system.

Monitoring remains a legitimate medical strategy. In many situations, it is precisely the appropriate decision.

But “monitoring” and “assuming that nothing is changing” are not synonymous.

When an adverse cardiometabolic trajectory is confirmed, the passage of time may be accompanied by further weight gain, progression of dysglycaemia or the emergence of additional risk factors. This is precisely why current recommendations do not restrict prevention to established diabetes: they also address prediabetes and the assessment of associated risk factors.

The question, therefore, is not whether Thomas should be “treated as a person with diabetes” before he has diabetes.

The available data would not justify such a conclusion.

The question is whether his trajectory now warrants a more comprehensive assessment of metabolic and cardiovascular risk, rather than waiting for another threshold to be crossed.

There is no need to conclude that Thomas’s previous assessments were incorrect.

They may have been entirely consistent with the information available at the time.

The problem emerges when successive consultations remain independent episodes while the phenomenon being observed is longitudinal.

At each consultation, an implicit question may be asked: “What do the results show today?”

After several years, another question becomes possible: “What has changed since the first assessment, and are these changes evolving independently or in a common direction?”

The second question does not invalidate the first.

It uses information that the first cannot contain: the trajectory.

The report sent to the treating physician is therefore not intended to propose a parallel diagnosis.

It reconstructs the information already available and distinguishes what the data establish from what they merely suggest.

The progression in glucose is placed back into context. Changes in weight and waist circumference are considered alongside the blood-pressure trajectory. Fatigue and impaired recovery are integrated without artificially assigning them a metabolic origin.

Risk factors, possible treatments, physical activity, sleep, family history and other available biological parameters must then be reintroduced into the analysis.

Only after this reconstruction can a hypothesis be prioritised and a more precise question be addressed to the physician.

The record does not necessarily tell us which disease Thomas has. It may already show which evolution should no longer be regarded as a succession of independent results.

Crossing a diagnostic threshold is an important clinical event. It is not necessarily the first biological event in the trajectory leading towards it.

Conversely, observing an unfavourable trajectory can never predict with certainty that a person will develop the disease towards which certain risk factors may appear to point.

Longitudinal analysis lies precisely between these two errors: waiting until everything becomes pathological, or transforming every variation into pathology.

It does not seek to make abnormal what is not abnormal.

It seeks to determine whether the relationship between several pieces of information carries a meaning that none of them can carry alone.

A value describes a state.

A succession of values can describe a direction.

And, in some situations, that direction becomes the most important clinical information in the record.