Beta Cell Function and Insulin Resistance

1. Two Numbers, One Feedback Loop

Fasting glucose and fasting insulin are set together. A resistant liver and muscle need more insulin to hold the same glucose; a beta cell that can supply it keeps the glucose normal at the cost of hyperinsulinaemia, and one that cannot lets the glucose rise. The homeostasis model of Matthews and colleagues, 1985, reads the pair against a computer model of that loop and reports two things.

The reference point is a fasting glucose of 81 mg/dL with an insulin of 5 µIU/mL, which the two formulae turn into exactly 1.0 and 100 per cent. Both indices are approximations of the model, and Wallace, Levy and Matthews wrote in 2004 that HOMA is reported twenty times more often for resistance than for beta cell function, and that reading beta cell function in isolation is a misuse.

2. What the Hyperbola Shows, and What It Cannot

Kahn and colleagues, 1993, measured insulin sensitivity and first-phase insulin response in 93 healthy people, lean to obese, and found the two related as a hyperbola: sensitivity multiplied by secretion is constant. The product is the disposition index, and a person whose beta cell keeps pace with falling sensitivity slides along the curve; one whose beta cell does not falls beneath it.

• The figure in this tool redraws that shape with HOMA-derived axes, sensitivity as 100 divided by HOMA-IR and function as HOMA-%B, and the reference curve through the normal young adult's point. Kahn used the minimal model and an intravenous glucose load, not HOMA; the transposition is this tool's.
• The distance from the curve is the fasting glucose in disguise. Divide the %B formula by the IR formula and the insulin cancels: %B / IR = 145,800 / [glucose × (glucose − 63)] in mg/dL. It is 100 at 81, 60 at 90 and 18 at 126 mg/dL. So a point below the curve says only that the glucose is above 81; it is not a second finding.
• What the point does add is where along the curve the patient sits: the high-insulin end, where the beta cell is working hard against resistance, or the low-insulin end, where a raised glucose with little insulin behind it is deficiency rather than resistance.

3. There Is No Agreed Cut-Off, and Here Is the Evidence

The 2010 insulin standardisation statement of Staten and colleagues says it plainly: there are no criteria by which an individual can be classified as insulin sensitive or resistant, or as having mild, moderate or severe impairment of secretion. Population studies draw their own line, usually at a centile.

A value of 2.5 is the figure most often quoted for adults, and it is also the Indian adolescent figure from Delhi, which is why the tool names it. Nobody has published an Indian adult threshold derived by the same method. In young lean normoglycaemic South Indian men studied against the clamp at Vellore, HOMA-IR and QUICKI discriminated poorly, with a receiver operating characteristic area of about 0.3 for QUICKI, which is the strongest reason to treat any single threshold with caution in this population.

4. Reasons the Indices Lie

• Exogenous insulin. The insulin assay measures what was injected along with what was secreted. HOMA-IR and HOMA-%B are meaningless on insulin; the C-peptide is the measure of the patient's own secretion.
• A sulfonylurea or a meglitinide drives the fasting insulin up. Both indices read high, and a high %B on glibenclamide is the drug, not the gland.
• Glucocorticoids raise glucose and insulin together; an acute illness or a hyperglycaemic emergency suppresses the beta cell, and the ADA and EASD ask that a C-peptide not be read within two weeks of one.
• A non-fasting sample, or a fast shorter than 8 hours, is outside the model.
• Hypoglycaemia on the sample. Below 72 mg/dL a low insulin or C-peptide is the appropriate response, not deficiency; the %B formula itself is undefined at 63 mg/dL and below.
• Renal impairment raises the C-peptide, because the kidney clears about half of it, and dialysis clears it to a variable extent.
• The assay. Insulin assays are not standardised and the same sample gives different numbers on different platforms; C-peptide assays disagree most at high values. A figure from one laboratory should not be read against a threshold derived on another.

5. The C-Peptide in Practice

Jones and Hattersley, 2013, reviewed the evidence and set thresholds for insulin-treated diabetes. They are approximate, and values close to them "should be treated with great caution".

• Its clinical job is classification, above all in the insulin-treated patient 3 to 5 years from diagnosis where a preserved C-peptide argues against type 1. Its value as a measure of resistance is limited, and the same review says a C-peptide HOMA is not advised in practice and is not valid on insulin.
• A regression the other way exists. Li and colleagues, 2004, fitted C-peptide-substituted HOMA formulae in 21 normal subjects and proposed them for patients on exogenous insulin. This tool prints that approximation when a C-peptide is entered, labelled as research, and prints the Exeter caution beside it. The two positions are not reconciled here.
• Non-fasting is often better. A random C-peptide with a glucose above 144 mg/dL (8 mmol/L) classified diabetes better than a fasting or glucagon-stimulated sample in the largest study; EDTA whole blood is stable for 24 hours at room temperature, serum gel for 6.

6. The Indian Phenotype

• Thin-fat. Rural Pune newborns were small in every measurement except the subscapular skinfold, which was relatively preserved (Yajnik 2003). At six years, Indian children had a lower BMI than white British children but more body fat and a higher HOMA-IR, by about 0.25 units, after adjusting for that fat (Lakshmi 2012).
• The cut-offs move with it. The 2009 Indian consensus sets overweight at a BMI of 23 and obesity at 25, and abdominal obesity at a waist above 90 cm in men and 80 cm in women. This tool raises a thin-fat flag when the BMI is below 23 and either the waist is above the cut-off or HOMA-IR is 2.5 or above; that pairing is the tool's, not a published criterion.
• The decline is the disease. In UKPDS the fasting glucose rose on every therapy over six years and the rise tracked a progressive loss of beta cell function. A %B that looks acceptable today is a snapshot of a moving picture.

7. What to Do With the Result

• Insulin resistance, glucose normal: the beta cell is compensating. Weight, waist, activity and the lipid profile are the levers, and the atherogenic profiler and the cardiovascular risk tool take the same patient further. Nothing in this tool changes a drug on its own.
• Raised glucose with low insulin and C-peptide: think deficiency before resistance. In an adult who is not obese, that is the moment for a GAD65 antibody and a classification, not a third oral agent.
• Raised glucose with high insulin: the beta cell is working hard and losing. The outpatient pathway chooses the agent; the HbA1c target tool sets what it is aiming at.
• Repeat rather than trust a single value near a threshold. The coefficient of variation of both HOMA indices is about a third.