Do You Need to Fast Before Blood Work?
For almost every blood test your doctor orders, you can eat breakfast first. A new analysis of 9.7 million outpatient results in JAMA Internal Medicine found that only a handful of common lab values shift enough with fasting to change your care. The list is short and predictable: glucose, triglycerides, lipase, GGT, and a few others. Long water-only fasts also have real costs, especially if you take insulin or a sulfonylurea.
Comparing 8 to 12 hours of fasting with more than 12 hours, no analyte crossed the total allowable error threshold that labs use to flag a clinically meaningful change. Ten analytes moved enough with shorter or longer fasts to matter. The rest, from cholesterol to kidney function to most electrolytes, sat still.
The short list that still needs a fast
After a shorter fast, triglycerides ran about 20% higher and GGT 15%. Two moved after an unusually long fast: total bilirubin up 17%, and 25-hydroxyvitamin D up 10%.
Only some of those shifts change a clinical decision. If your doctor is diagnosing diabetes off a fasting plasma glucose, the American Diabetes Association threshold is 126 mg/dL and assumes at least eight hours without calories, so eating before that specific test can push a borderline result into the diabetes range on paper. If a lipase is being drawn to work up pancreatitis, food matters. And if triglycerides have previously been very high, say 400 mg/dL or above, the 2026 ACC/AHA dyslipidemia guideline prefers a fasting confirmation.
What breakfast doesn't touch
The routine panels your annual gets built on are largely immune. Total and LDL cholesterol drop slightly after a meal, HDL barely moves, and none of it changes risk classification. In the ASCOT-LLA trial, fasting and non-fasting panels drawn four weeks apart in the same people agreed on risk category about 95% of the time. HbA1c doesn't budge with a single meal; it reflects glycation over the last two to three months. Kidney function, most electrolytes, liver enzymes other than GGT, and CBC all stayed within normal biological noise in the JAMA data.
Calculated LDL is the exception worth knowing about. The standard Friedewald equation assumes a fixed ratio between triglycerides and VLDL cholesterol, and that assumption wobbles when triglycerides run high after a meal. In a small study of patients with type 2 diabetes, Friedewald-calculated LDL after a meal underestimated the fasting value enough to move a substantial share of them into a lower risk category, and an even larger share of the statin users. Newer equations (Martin-Hopkins, Sampson) hold up better in the fed state. If you're near a decision threshold for starting or intensifying a statin, ask for LDL to be measured directly or calculated with Martin-Hopkins, or confirm with a fasting draw.
The hidden cost of a fast you didn't need
Skipping breakfast is not neutral. In an outpatient survey, about a quarter of patients with diabetes had hypoglycemic symptoms while preparing for fasting labs, and nearly half of self-fasters restricted water too. More than half of insulin users and nearly three-quarters of people on oral diabetes drugs kept taking their medication on the morning of a fast, which is exactly the setup for a low. A meta-analysis of older adults links hypoglycemia to markedly higher rates of falls, fractures, and dementia.
Beyond diabetes, an early fasted appointment often means a missed meal, missed water, a headache, and sometimes a missed appointment. When most of what shows up on a comprehensive panel doesn't change with breakfast, that trade doesn't pay off.
What to do with your next lab slip
Read the order. If it includes a fasting plasma glucose to diagnose diabetes, a lipase for possible pancreatitis, or a repeat triglyceride after a prior very high reading, fast the way you've been told. For a routine annual with a lipid panel, HbA1c, comprehensive metabolic panel, and CBC, eat, drink water, and take your medications on schedule. If a result comes back near a treatment threshold, especially LDL, a fasting draw afterward is a reasonable way to confirm it. These are all markers you can order and track yourself, and comparing your own readings taken under the same conditions is the cleanest way to follow them over time.
| Test | Fast needed? | Why |
|---|---|---|
| Standard lipid panel (total, LDL, HDL, non-HDL) | No | Post-meal shifts are small and don't change risk category |
| Triglycerides, screening | No | Non-fasting values predict cardiovascular risk as well or better |
| Triglycerides, if prior ≥400 mg/dL | Yes | Fasting confirms the reading and guides pancreatitis-risk management |
| HbA1c | No | Reflects months of glycation, so a single meal is irrelevant |
| Fasting plasma glucose (diabetes diagnosis) | Yes | The diagnostic threshold assumes a fasted draw |
| Comprehensive metabolic panel | No | Kidney function, most electrolytes, and routine liver enzymes stay inside biological noise |
| Lipase | Yes | Food raises it enough to muddy a pancreatitis workup |
| GGT | Usually no | Shift is modest and matters only in specific liver evaluations |
| CBC | No | Not meaningfully affected by meals |
The 12-hour water-only fast before routine blood work is a habit the evidence has outgrown. A few things could change that: a Quest or LabCorp update rescinding blanket fasting instructions, a formal US screening statement adopting non-fasting lipids as default, or head-to-head data showing that eating skews iron, ferritin, cortisol, or insulin enough to matter. Until then, ask which specific tests on your order need fasting, and assume the rest do not.