Blood Chemistry · Functional Ranges · Glucose Regulation
The NHS HbA1c threshold of 48 mmol/mol (6.5%) was designed to diagnose diabetes, not to protect you from getting there. Below that line, you're told everything is fine. But years of metabolic change can unfold before a conventional marker moves out of range, and quite different clinical pictures can look alike on a basic glucose panel.
In March 2024 I ran my own Randox Signature blood panel. My HbA1c came back at 30.53 mmol/mol (4.9%). By any measure — NHS, functional, or otherwise — that's optimal. No flags, no concern, no follow-up needed.
But alongside it: fasting glucose of 5.64 mmol/L (102 mg/dL).
The Randox reference range flags anything from 5.60 mmol/L (101 mg/dL) as pre-diabetic. On that single marker, in isolation, I was technically in the pre-diabetic zone. My HbA1c — which measures three-month average glucose — was at the low end of optimal. My fasting glucose — a snapshot from a single morning — was borderline by one conventional threshold.
Both numbers came from the same blood draw. One said fine. One said watch it. Without context, you'd have no way to know which one to believe — or whether either reading reflected anything clinically meaningful.
HbA1c measures the proportion of haemoglobin that has been glycated — sugar-coated — over the lifespan of a red blood cell, approximately 90–120 days. It is a three-month rolling average of blood glucose exposure. A single fasting glucose reading captures a moment. HbA1c captures a season.
That makes it a better marker than fasting glucose alone. But it has two important limitations that are rarely explained at the point of testing.
The first is that HbA1c is an average. Significant glucose variability — high post-meal spikes followed by reactive lows — can average out to a perfectly normal HbA1c. The person with extreme glucose swings and the person with stable, steady-state glucose can produce identical HbA1c values. The physiological experience underneath is completely different.
The second limitation is the threshold. The NHS uses 48 mmol/mol (6.5%) as the diagnostic threshold for diabetes, and 42–47 mmol/mol (6.0–6.4%) as pre-diabetes. Below 42 mmol/mol (6.0%), you're normal. But the relationship between HbA1c and health risk is continuous — it doesn't suddenly start at 42. Risk increases progressively from values well below the conventional threshold.
The EPIC-Norfolk study followed 4,662 men aged 45–79 for about four years. Deaths from all causes and from heart disease rose continuously across the whole HbA1c range, with the lowest rates below 5% (31 mmol/mol). Each 1% (11 mmol/mol) higher HbA1c was linked with 28% higher risk of death, and most of the excess deaths linked to an HbA1c of 5% (31 mmol/mol) or more were in men below the diabetes range (Khaw 2001). Grade: observational, men only; it shows a link, not that lowering HbA1c in people without diabetes prevents deaths. The diagnostic threshold is a convenience, not a cliff edge.
Functional blood chemistry interpretation uses tighter, health-optimisation ranges rather than disease-diagnosis thresholds. The distinction is important: conventional ranges are set at the outer boundaries of the population, calibrated to detect pathology. Functional ranges are set within the population, calibrated to reflect optimal physiology.
How sure is this? The NHS thresholds below come from NICE and the WHO. The “functional optimal” columns are the working ranges I use in functional practice. None of them has been validated against outcomes, and there is no standardised insulin assay, so an insulin range from one lab doesn’t transfer exactly to another. Read them as prompts to look closer, not as diagnoses. Both units are given; conversions are mine (glucose ×18, triglycerides ×88.6, insulin ÷6.945).
| Marker | NHS / Lab range | Functional optimal | Clinical note |
|---|---|---|---|
| HbA1c | Normal <42 mmol/mol (6.0%) | Optimal <35 mmol/mol (5.4%) Review 35–42 (5.4–6.0%) |
Values 35–42 warrant investigation of fasting insulin and HOMA-IR even if "normal" on standard criteria |
| Fasting Glucose | Normal 4.0–5.9 mmol/L (72–106 mg/dL) | Optimal 4.0–5.0 mmol/L (72–90 mg/dL) Review 5.0–5.6 (90–101 mg/dL) |
Above 5.0 mmol/L (90 mg/dL) with any elevation in fasting insulin warrants HOMA-IR calculation regardless of NHS threshold |
| Fasting Insulin | Lab range 17.8–173 pmol/L (3.0–29 µIU/mL) | Optimal <60 pmol/L (10 µIU/mL) Review 60–100 (10–17 µIU/mL) |
The lab range is wide enough, in my view, to include early insulin resistance. Fasting insulin above 60 pmol/L (10 µIU/mL) is worth investigating even with normal glucose |
| Triglycerides | Desirable <2.3 mmol/L (204 mg/dL) | Optimal <1.1 mmol/L (97 mg/dL) Review 1.1–1.7 (97–150 mg/dL) |
Fasting triglycerides above 1.1 mmol/L (97 mg/dL) are an early signal of impaired glucose metabolism and hepatic lipid handling, even within the "desirable" NHS range |
| TG:HDL Ratio | Not routinely calculated | Optimal <1.0 in mmol/L (about <2.3 in mg/dL) Concern >1.5 |
One of the most sensitive surrogate markers for insulin resistance. Calculated from standard lipid panel — requires no additional testing |
| HOMA-IR | Rarely calculated on NHS | Optimal <1.0 Concern >1.5 |
Requires fasting glucose and fasting insulin. Calculated, not directly measured. See the HOMA-IR post for the full explanation. |
Even with functional ranges applied, interpreting metabolic markers individually misses the clinical picture. The University of Oxford developed HOMA2 — an updated computational model that generates three values from a single fasting blood draw: HOMA2-IR (insulin resistance), HOMA2-%B (beta-cell secretory output), and HOMA2-%S (peripheral insulin sensitivity).
Individually, none of the three metrics provides meaningful data. Elevated HOMA2-%B, for example, could mean a healthy beta-cell reserve compensating for early insulin resistance — or it could mean diminished beta-cell capacity producing apparently high output against a low glucose input. Without the other two values, you cannot distinguish between these clinical realities.
Evaluated together, they differentiate five distinct glucose dysregulation patterns that look identical on a standard glucose panel.
Each pattern requires a different clinical approach. Treating them identically — as "blood sugar problem, reduce carbohydrates" — will help some, do nothing for others, and potentially worsen a third.
Type 2 diabetes starts with insulin resistance; Type 3c starts with damage to the pancreas, so insulin production is the main problem. Both raise fasting glucose and HbA1c, and a basic panel can’t tell them apart. The label matters: in the English primary-care cohort, 45.8% of people whose diabetes followed chronic pancreatic disease were on insulin within five years, against 4.1% with type 2 (Woodmansey 2017). Pancreatic disease can also affect digestion, which a type 2 label won’t prompt anyone to check. Fasting insulin or C-peptide alongside glucose is what shows whether output is low. Diagnosis and treatment belong with your GP or diabetes team. HOMA2 estimates beta-cell function and insulin sensitivity; it is not a diagnostic test for any type of diabetes, and %B means little read on its own (Wallace 2004).
The HOMA2 model can use either fasting insulin or C-peptide as the input. C-peptide — co-secreted with insulin at equimolar concentrations — is actually the more reliable marker for assessing pancreatic secretion. Unlike insulin, C-peptide is not affected by exogenous insulin administration, it has a longer half-life, and it is not subject to first-pass hepatic clearance.
Many clinicians who use the HOMA2 model recommend C-peptide as the preferred input — particularly in patients receiving insulin, or in any case where insulin dynamics are unpredictable. The Functional Platinum panel I use now includes both fasting insulin and C-peptide. Most NHS panels include neither routinely.
"Serial increases in fasting insulin levels in patients whose glucose levels remain within normal limits are a good indication that further testing with the full HOMA2 matrix would be beneficial rather than waiting until their glucose or A1c moves out of range."
Dr Brad Rachman — Optimal DX Podcast, Episode 5 (clinical opinion, not trial evidence)Back to my March 2024 results: fasting glucose 5.64 mmol/L (102 mg/dL), insulin 104 pmol/L (about 17 µIU/mL), C-peptide 2.06 ng/mL (0.68 nmol/L), triglycerides 0.78 mmol/L (69 mg/dL), HbA1c 30.5 mmol/mol (4.9%).
The triglycerides, the TG:HDL ratio of 0.58 and the HbA1c are all low. The insulin isn’t: 104 pmol/L (about 17 µIU/mL) is inside the lab range but well above the 60 pmol/L (10 µIU/mL) I use as a prompt to look closer. Put into the simple HOMA-IR formula (glucose × insulin ÷ 22.5), it gives about 4.3. There is no agreed cut-off for HOMA-IR, and cut-offs in studies vary widely between populations, but 4.3 is high by most of them. One fasting insulin is a noisy measure, and the long fast may have played a part, but read honestly, March 2024 was not the clean picture I first described here. It looked more like the early insulin resistance pattern than anything else.
Fifteen months later I retested. Insulin had halved to 49.9 pmol/L (about 8.3 µIU/mL), fasting glucose was 4.83 mmol/L (87 mg/dL), and HOMA-IR came out at about 1.8. HbA1c went the other way: from 30.5 to 38.4 mmol/mol (about 4.9% to 5.7%). That is still below the 42 mmol/mol (6.0%) NICE uses for prediabetes, and just below the American Diabetes Association’s 39 mmol/mol (5.7%), but it is inside my own “review” band. Triglycerides were 0.69 mmol/L (61 mg/dL).
In between, life happened. My mum died in June 2023. My dad’s health fell apart after that: grief, repeated hospital stays, emergency ambulance trips, and a dementia that sped up. I’m his eldest son, I hold his power of attorney, and I live the furthest away, on the other side of the country from my own family and work. From March 2025 I was also working late into the night, partly to catch up on time lost to caring, and partly to build Detective Health, which has become a new business as I wound the old one down. More cake, more alcohol, more broken nights, less time looking after me. Both blood tests sit inside that stretch.
I’m not telling you this as an excuse, and I’m not immune because I do this for a living. Clients need to see that I walk the walk, but walking it doesn’t mean life never knocks me off course. The numbers moved in two directions and they only tell part of the story; that’s exactly why I read a matrix and a history, not one number. My next step is the one I’d give a client: retest with fasting insulin and C-peptide, and deal with sleep, alcohol, food and stress before reaching for anything else.
If your HbA1c is below 42 mmol/mol (6.0%) and your GP has told you everything is normal, that may well be true. But it is worth knowing what "normal" actually means in this context — it means you have not yet crossed a diagnostic threshold. It does not mean your metabolic system is operating optimally, or that there is no process underway that will eventually move you across that threshold.
The questions worth asking after any metabolic blood test are not just "is this number in range?" but: what is fasting insulin doing relative to fasting glucose? What is the TG:HDL ratio? Is there a pattern across the matrix that tells a coherent story?
And crucially: is fasting insulin being measured at all? In most NHS metabolic screens, it is not. You can have significant early insulin resistance — detectable, addressable, reversible — with perfectly normal glucose and HbA1c, for years, without anyone running the test that would show it.
Early insulin resistance — Pattern 2 in the HOMA2 matrix — is highly reversible. Late insulin resistance is reversible with significant intervention. Once beta-cell exhaustion begins and glucose rises consistently, the window narrows but recovery is still possible with the right approach. The argument for early testing is not to create anxiety — it is to identify when the window is widest and intervention requires the least effort.
The Randox Signature panel I ran in March 2024 also measured adiponectin, leptin and resistin. I no longer use Randox. The Regenerus Functional Platinum panel I use now includes fasting glucose, fasting insulin, HbA1c, C-peptide, triglycerides and HDL cholesterol. That gives enough data to calculate HOMA-IR and TG:HDL, assess the full metabolic syndrome marker set, and begin constructing the HOMA2 matrix with C-peptide as the preferred input. Leptin and adiponectin are on the Doctor’s Data CardioMetabolic add-on, whose report marks them for research use only.
It also includes LDH (lactate dehydrogenase). LDH is not part of the HOMA2 model, which uses only fasting glucose with insulin or C-peptide (Wallace 2004). Some practitioner frameworks read LDH alongside the matrix; I haven’t found trial evidence that it separates these patterns, so I don’t use it that way.
The additional context from 83 markers across all systems means metabolic findings are never interpreted in isolation. An elevated fasting glucose in someone with elevated hsCRP and low albumin tells a different story from the same glucose in someone with optimal inflammatory markers and good protein status. The blood chemistry is read as a picture, not a list.
Fasting glucose and HbA1c alone are not enough. The Functional Platinum blood panel gives you the full marker set — including fasting insulin, C-peptide, and LDH — to build the metabolic matrix from a single fasting blood draw, alongside your history.
Book a conversation →