Insulin Tolerance Test Interpreter
Insulin Tolerance Test Interpreter
Nothing in an insulin tolerance test means anything until the glucose has actually fallen below 2.2 mmol/L with symptoms. After that, the cortisol peak is read against a threshold that depends on which assay measured it — modern specific assays read several hundred nanomoles lower than the historical 500 to 550 — and the growth hormone peak against two competing guideline figures.
Insulin tolerance test
Glucose gate + cortisol + GH → interpretationGlucose nadir 1.9 mmol/L with symptoms; peak cortisol 395 nmol/L on an Abbott platform; peak GH 2.1 µg/L
The gate, then the two axes
Peak cortisol ≥ the threshold validated for YOUR assay (320–550 nmol/L)
Peak GH < 3 µg/L = severe GH deficiency (GRS 2007, Endocrine Society 2011)
Peak GH ≤ 5 µg/L = deficient by the AACE criterion
- < 2.2 mmol/L with symptoms
- the gate. Below 40 mg/dL on a laboratory glucose, with sweating, tremor, tachycardia or drowsiness. Without it the test has not asked its question and neither peak means anything
- insulin 0.10–0.30 units/kg
- 0.15 units/kg is typical; 0.10 where hypopituitarism is likely, and 0.2 to 0.3 in acromegaly, Cushing syndrome or diabetes, where insulin resistance blunts the fall. Samples at baseline and then every 30 minutes for two hours
- 320–550 nmol/L cortisol
- not a range but a set of assay-specific thresholds. 550 nmol/L is historical; 500 is the legacy figure; 416 nmol/L was validated on the Abbott platform across 300 tests, keeping sensitivity at 100.0% and raising specificity to 86.7%; the Roche Elecsys Cort I and Cort II 2.5th percentiles in healthy participants were 404 and 320 nmol/L
- 3 µg/L versus 5 µg/L
- two guideline thresholds for the growth hormone peak, both current. The GH Research Society 2007 and Endocrine Society 2011 use below 3.0 µg/L for severe adult deficiency; AACE 2009 and 2019 use 5.0 µg/L or below
- what makes this test dangerous
- deliberate hypoglycaemia. Contraindicated in ischaemic heart disease, epilepsy or unexplained blackouts, untreated hypothyroidism, untreated adrenal insufficiency with a 09:00 cortisol below 100 nmol/L, pregnancy, and at the extremes of age. A normal ECG, an indwelling cannula, medical supervision throughout and rescue glucose to hand are not optional
Worked example
Glucose nadir 1.9 mmol/L with symptoms; peak cortisol 395 nmol/L on an Abbott platform; peak GH 2.1 µg/L
1.9 mmol/L is below 2.2 with symptoms, so the hypoglycaemic stimulus was adequate and the peaks can be read
395 nmol/L is below the 416 nmol/L threshold validated for this platform → subnormal cortisol response
2.1 µg/L is below 3 µg/L → severe growth hormone deficiency by the GH Research Society and Endocrine Society criterion
→ both axes subnormal
Now change only the assay threshold. On the Roche Elecsys Cort II figure of 320 nmol/L, the same 395 nmol/L is a normal cortisol response and the result becomes growth hormone deficiency alone. On the historical 550 nmol/L it would have been subnormal by a wide margin. One patient, one sample, three verdicts — which is why the threshold has to come from the laboratory that ran the assay
Cortisol thresholds by assay generation
| Threshold | Where it comes from | What it is validated against |
|---|---|---|
| 550 nmol/L | Historical; polyclonal assays with substantial cross-reactivity | Convention rather than a derivation study |
| 500 nmol/L | The legacy figure in wide use | Assumes anyone not reaching it is deficient and needs full replacement |
| 416 nmol/L | Lazarus 2024, 300 insulin tolerance tests over 12 years, Abbott platform | Sensitivity 100.0%, specificity improved to 86.7% |
| 404 nmol/L | Okutan 2026, 90 healthy participants, Roche Elecsys Cort I | 2.5th percentile of the normative response; 5th percentile 451 nmol/L |
| 320 nmol/L | Okutan 2026, same participants, Roche Elecsys Cort II | 2.5th percentile; 5th percentile 348 nmol/L |
Contraindications and precautions — the reason this test is supervised
| Detail | |
|---|---|
| Ischaemic heart disease | Absolute. Hypoglycaemia provokes catecholamine release and ischaemia |
| Epilepsy or unexplained blackouts | Absolute. Hypoglycaemia lowers the seizure threshold |
| Untreated hypothyroidism | Absolute. Replace thyroxine first; the GH and cortisol responses are blunted and recovery from hypoglycaemia is slowed |
| Untreated adrenal insufficiency | Do not proceed if the 09:00 cortisol is below 100 nmol/L — the test is unnecessary and dangerous |
| Age | Avoid above 60 to 65 years and below 16; risk rises and alternatives exist |
| Pregnancy | Contraindicated |
| Before starting | A normal ECG, an indwelling cannula, and 10% dextrose or glucagon 1 mg immediately available |
| During | Continuous medical supervision. Rescue with intravenous glucose does not invalidate the test once the nadir and the symptoms have been documented |
One gate, two axes, and a threshold that belongs to the assay
The insulin tolerance test provokes hypoglycaemia deliberately and watches what the pituitary does about it. Insulin is given intravenously, usually at about 0.15 units/kg, and cortisol, growth hormone and glucose are sampled at baseline and then every 30 minutes for two hours. An intact hypothalamic-pituitary-adrenal axis answers a falling glucose with a surge of ACTH and therefore cortisol; an intact somatotroph axis answers it with growth hormone. It remains the reference test for both, because it stresses the whole axis from the hypothalamus down rather than testing the adrenal gland's response to exogenous ACTH.
Everything depends on a single precondition. The stimulus is the hypoglycaemia, so unless the laboratory glucose actually fell below 2.2 mmol/L — 40 mg/dL — with symptoms, a flat response is evidence of nothing at all. This is not a technicality. A test that failed to produce hypoglycaemia and was read anyway will label a patient with secondary adrenal insufficiency and commit them to lifelong glucocorticoid replacement, sick-day rules and the small but real mortality that goes with unnecessary steroid. Insulin resistance is the usual reason the glucose does not fall far enough, which is why the dose is raised to 0.2 to 0.3 units/kg in acromegaly, Cushing syndrome and diabetes.
The cortisol threshold has the same problem as the short Synacthen test, and it is not a small one. The familiar 500 to 550 nmol/L figures were derived on polyclonal immunoassays that cross-reacted appreciably with cortisone, 11-deoxycortisol and synthetic glucocorticoids, and therefore over-read. Modern monoclonal assays, and mass spectrometry, report lower numbers for the same physiology. An audit of 300 insulin tolerance tests on the Abbott platform found that lowering the threshold from 500 to 416 nmol/L kept sensitivity at 100.0% and lifted specificity to 86.7% — which means the old threshold was misclassifying people. A normative study in 90 healthy participants put the 2.5th percentile at 404 nmol/L on the Roche Elecsys first-generation cortisol assay and 320 nmol/L on the second-generation one. The practical rule is that a peak cortisol without the name of the assay beside it cannot be interpreted.
The growth hormone side carries a smaller but similar ambiguity: below 3.0 µg/L is the GH Research Society and Endocrine Society criterion for severe adult deficiency, while AACE uses 5.0 µg/L or below, so a peak of 4 µg/L is deficient under one guideline and not the other. Obesity blunts the peak, and assays differ in which growth hormone isoforms they detect, so a cut-off does not travel between platforms. And because the test works by making someone hypoglycaemic on purpose, the contraindications are absolute rather than relative — ischaemic heart disease, epilepsy or unexplained blackouts, untreated hypothyroidism, untreated adrenal insufficiency, pregnancy — with a normal ECG, a cannula, supervision and rescue glucose required before anyone starts.
Frequently asked questions
What counts as adequate hypoglycaemia in an insulin tolerance test?
A laboratory glucose below 2.2 mmol/L (40 mg/dL) accompanied by symptoms such as sweating, tremor, tachycardia or drowsiness. A meter reading is not sufficient to define the endpoint. If the glucose did not fall that far, neither the cortisol nor the growth hormone response can be interpreted and the test must be repeated with a higher insulin dose.
What peak cortisol is normal on an insulin tolerance test?
It depends on the assay. The historical figures were 500 to 550 nmol/L, derived on less specific assays. An audit of 300 tests on the Abbott platform supported 416 nmol/L, keeping sensitivity at 100.0% while improving specificity to 86.7%, and a normative study put the 2.5th percentile at 404 nmol/L on Roche Elecsys Cort I and 320 nmol/L on Cort II. Use the threshold your own laboratory has validated.
What peak growth hormone indicates deficiency?
Below 3.0 µg/L according to the GH Research Society 2007 and the Endocrine Society 2011, and 5.0 µg/L or below according to AACE 2009 and 2019. Both are current, so a peak between 3 and 5 is deficient or not depending on which guideline the centre follows. Obesity lowers the peak independently of any deficiency.
When is the insulin tolerance test contraindicated?
Ischaemic heart disease, epilepsy or unexplained blackouts, untreated hypothyroidism, untreated adrenal insufficiency with a 09:00 cortisol below 100 nmol/L, pregnancy, and at the extremes of age — generally avoided above 60 to 65 and below 16. A normal ECG is required before starting, along with an indwelling cannula, continuous medical supervision and immediate access to 10% dextrose or glucagon.
Does giving rescue glucose invalidate the test?
No, provided the glucose nadir and the symptoms were documented first. The stimulus has already been delivered at that point, and the cortisol and growth hormone responses continue to develop afterwards. Patient safety takes precedence; sampling continues on schedule.
Why not use a short Synacthen test instead?
For the cortisol axis alone, often you should — it carries none of the hypoglycaemic risk. The insulin tolerance test earns its place where growth hormone and cortisol are both in question, or soon after pituitary surgery, when the adrenal glands have not yet atrophied and a Synacthen test can be falsely normal.
Related calculators
References
- Yuen KCJ, Biller BMK, Radovick S, et al. Growth hormone stimulation tests in assessing adult growth hormone deficiency. In: Endotext. MDText.com; NCBI Bookshelf NBK395585. — glucose below 40 mg/dL (2.2 mmol/L) with symptoms; peak GH <3.0 µg/L (GRS 2007, Endocrine Society 2011) against ≤5.0 µg/L (AACE 2009, 2019); calibration to IS 98/574.
- Lazarus K, Hayes A, Narula K, Papadopolou D, Tan TM-M, Choudhury S. Redefining ITT cortisol thresholds on Abbott platforms to prevent misdiagnosis of adrenal insufficiency. Clin Endocrinol (Oxf). 2024;101(3). — 300 tests; 416 nmol/L gave sensitivity 100.0% and specificity 86.7%.
- Okutan S, Borresen SW, Hilsted L, Feldt-Rasmussen U, Klose M. Defining normative cortisol responses to the insulin tolerance test: a comparison of Elecsys Cort I and II assays. Pituitary. 2026;29:3. — 90 healthy participants; 2.5th percentiles 404 and 320 nmol/L.
- Molitch ME, Clemmons DR, Malozowski S, Merriam GR, Vance ML. Evaluation and treatment of adult growth hormone deficiency: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2011;96(6):1587–1609.
- Manchester University NHS Foundation Trust. Insulin Tolerance Test — Adults. 2023. — insulin 0.10–0.30 units/kg by indication; glucose <2.2 mmol/L measured by the laboratory; contraindications including ischaemic heart disease, epilepsy, untreated hypothyroidism and a 09:00 cortisol below 100 nmol/L; normal ECG and rescue glucose required.
Medical Disclaimer: The tools and content provided here are for educational and reference purposes only. They are not intended to substitute for professional medical advice, diagnosis, or treatment. Clinical decisions should always be based on the comprehensive assessment of a qualified healthcare professional.