Biomarkers

hs-CRP Levels: What the Number Means and Why One Reading Is Not Enough

On this page
  1. Why the statement asks for two measurements
  2. Above 10 mg/L is a different question
  3. The association is real, and it is not small
  4. And the protein itself is almost certainly not the cause
  5. Then why did JUPITER work?
  6. What to do with a number you already have

The statement that gave your lab its hs-CRP risk bands also told it to measure the marker twice, about two weeks apart, and average the two results. That second instruction is the most useful sentence attached to this test and the one almost nobody follows. A single hs-CRP value is a noisy reading of a real signal: the association between hs-CRP and cardiovascular events is large and has been replicated in hundreds of thousands of people, the within-person swing between two draws is roughly 30% to 75%, and the genetic evidence says the protein itself is almost certainly not causing any of it.

hs-CRPLabel in the 2003 AHA/CDC statementWhat it says, and does not say
Below 1.0 mg/LLow relative riskThe lowest band of a population distribution. Not a clean bill of health, and not a target to chase
1.0 to 3.0 mg/LAverage relative riskWhere most adults land. The band is wide partly because the measurement is noisy
Above 3.0 mg/LHigh relative riskHigh relative to the rest of the distribution. Not an absolute probability of anything
Above 10.0 mg/LNot a cardiovascular risk band at allThe statement says repeat the test and look for infection or another source of inflammation

Two labels before anything else, because both cause confusion on the report itself. CRP and hs-CRP are the same protein, measured by a more sensitive assay: standard CRP is used to detect obvious inflammation and often cannot resolve the low end at all, while hs-CRP is the version that can tell 0.6 from 2.4. And units differ by a factor of ten: hs-CRP is conventionally reported in mg/L, standard CRP is often reported in mg/dL, and 3.0 mg/L is 0.3 mg/dL. If two of your results appear to differ tenfold, check the unit column before anything else.

You could be asking this about yourself

is my hs-CRP settling, or spiking with something?

Connect BioTrakk to Claude, ChatGPT or Gemini. Log a reading in one sentence, and your own history becomes something you can ask about — in the chat you already have open.

Free. No BioTrakk account to make first. Claude takes one click.

Why the statement asks for two measurements

Because one is not a stable description of you. hs-CRP has one of the largest within-person variabilities of any routine blood marker, and the size of it is the single fact that should change how the number is read.

A 2024 systematic review and meta-analysis pooled 60 studies of repeated CRP and hs-CRP measurements and found a median within-subject coefficient of variation of 0.44 for hs-CRP, with individual studies ranging from 0.27 to 0.76. The pooled intraclass correlation was 0.62, meaning that a little under two fifths of the variation in readings is a person moving around their own average rather than one person differing from another. The same paper notes that the European Biological Variation Database puts the within-subject biological variation for hs-CRP at 58.9% (95% CI 53.2 to 66.0), and that several of the older published estimates may be unreliable.

Work that through with real numbers. Take someone whose true long-run average is 2.0 mg/L, sitting in the middle of the "average risk" band. At a within-person CV of ~44%, readings of 1.1 and 3.2 are both unremarkable for that person. The first lands in the low band, the second in the high band, and nothing about them has changed between the two draws. Your band is partly a property of the day you had blood taken.

This is why averaging two draws two weeks apart is not fussiness. Averaging two independent measurements cuts the noise by about 30%, which is the difference between a number you can track and a number that moves on its own.

Above 10 mg/L is a different question

The 2003 statement is explicit that a value above 10 mg/L is not read as the top of the cardiovascular scale. It says the test should be repeated and the person examined for sources of infection or inflammation, and the laboratory science group's companion report repeats the same instruction.

The reason is that acute inflammation moves this marker by orders of magnitude, not by percentages. A respiratory infection, a dental abscess, a flare of an inflammatory condition, recent surgery, an injury or a vaccination in the preceding days can all put hs-CRP far above anything a cardiovascular interpretation would survive. A reading taken four days into a cold is a measurement of the cold.

The statement's own phrase is metabolically stable: the two averaged measurements are meant to be taken when nothing acute is going on. That sounds obvious and is routinely lost, because the blood draw is scheduled weeks in advance and whatever happens that week happens anyway.

The association is real, and it is not small

None of the above makes hs-CRP a useless marker. The largest analysis of it is the Emerging Risk Factors Collaboration's 2010 individual-participant meta-analysis, which pooled records for 160,309 people with no history of vascular disease across 54 prospective studies, with roughly 27,769 fatal or non-fatal outcomes, and corrected for within-person variation in the risk factors it measured.

After adjustment for conventional risk factors, the hazard ratio for coronary heart disease was 1.37 (95% CI 1.27 to 1.48) per three-fold higher CRP concentration. The interval is comfortably clear of 1, so this is not a borderline finding.

The authors' own reading matters as much as the number: most of the crude association between CRP and heart disease was explained by risk factors already known to cause heart disease. CRP travels with smoking, adiposity, blood pressure, lipids and diabetes, which is a large part of why it predicts anything.

And the protein itself is almost certainly not the cause

This is where hs-CRP differs from, say, ApoB, and it is the part most self-tracking content leaves out.

Because the variants in the CRP gene that raise a person's lifelong CRP are allocated at conception and not by their habits, they can be used to ask whether CRP itself does anything. The CRP CHD Genetics Collaboration did exactly that across 194,418 participants in 47 studies, including 46,557 people with coronary heart disease, using four CRP-gene variants.

The result is one of the cleanest contrasts in cardiovascular epidemiology. Per 1 SD higher genetically raised CRP, the risk ratio for coronary heart disease was 1.00 (0.90 to 1.13). Per 1 SD higher measured circulating CRP in the same framework, it was 1.33 (1.23 to 1.43), and the difference between the two was significant (p = 0.001). The authors concluded that CRP concentration is unlikely to be even a modest causal factor in coronary heart disease.

The practical consequence is worth stating plainly, because a great deal of supplement marketing depends on it not being said: there is no good evidence that lowering your hs-CRP as an end in itself lowers your risk of anything. hs-CRP is a thermometer, not a thermostat. Reading it tells you something about the system. Taping an ice cube to it does not cool the room.

Then why did JUPITER work?

Because selecting people by a marker and treating the marker are different things, and the trial that gets cited for the second actually demonstrates the first.

JUPITER randomised 17,802 apparently healthy adults who had LDL cholesterol below 130 mg/dL and hs-CRP of 2.0 mg/L or higher to rosuvastatin 20 mg daily or placebo. It was stopped early at a median follow-up of 1.9 years, with 142 primary-endpoint events on the statin against 251 on placebo, roughly 44% fewer. On treatment, LDL cholesterol fell by about 50% and hs-CRP by 37%.

Read what that design can and cannot settle. It shows that an hs-CRP threshold found a group of people with ordinary LDL who nonetheless benefited from treatment, which is a real and useful result about whom to consider treating. It cannot separate the benefit of halving LDL from the benefit of lowering CRP, because the same drug did both in the same people. Treating it as proof that CRP was the thing being treated requires ignoring the 50% LDL reduction sitting next to it. The trial also recorded more physician-reported diabetes in the rosuvastatin group, which belongs in any honest one-paragraph summary of it.

The 2003 statement positioned hs-CRP in the same spirit: an optional test for people already judged to be at intermediate risk, to help a decision that was already being made, and explicitly not a screen for the whole adult population.

What to do with a number you already have

If a test has been run, or a clinician has a reason to run one, a few things make the result worth more than it would otherwise be:

  • Two draws, about two weeks apart, averaged, while nothing acute is going on. This is what the statement asks for and it is the single step that turns a noisy value into a usable one. One reading is a data point about a fortnight, not about you.
  • Write down the lab, the assay and the units. mg/L and mg/dL differ by a factor of ten, and hs-CRP and standard CRP are different assays of the same protein.
  • Record what was happening that fortnight. An infection, a dental problem, an injury, surgery, a vaccination, an unusually hard training block. Any of these can move hs-CRP more than a year of patient change, and without the note the spike is unreadable later.
  • Treat a value near a band edge as near a band edge. With a within-person CV around 44%, a 2.8 and a 3.2 are the same result wearing different labels.
  • Keep the trend, not the point. Several averaged pairs over a couple of years, same lab and same assay, say something a single result never can.
  • Take any decision to a clinician. This page contains no dose and no treatment. What an elevated value means for one person depends on their history, their other markers and their absolute risk, and that is not a question a reference band answers.

In practice the failure here is logistical rather than medical. The results arrive as PDFs months apart, the second confirmatory draw never gets booked, nobody writes down that the February test followed a week of flu, and two years later there is a column of numbers with no conditions attached to them. Logging each panel in BioTrakk (photograph the lab report or type the values in) keeps one trend line per marker with the context recorded beside it, so that a real move is distinguishable from a fortnight with a cold in it. The blood test tracker page describes how that storage works, homocysteine is the other marker on this site where one reading is routinely over-read, and if you want to test something against your own baseline rather than against a population average, that is what an n-of-1 experiment is for. On a marker this noisy, BioTrakk will often report "not distinguishable from chance", which on hs-CRP is usually the correct answer.

BioTrakk is not a medical device and this is not medical advice. hs-CRP is a clinical laboratory test: interpreting yours, and deciding anything on the basis of it, is a conversation with a clinician. Nothing here diagnoses, treats, monitors or prevents any condition, and this page recommends no treatment and no dose.

Frequently asked questions

What is a good hs-CRP level?

The bands most labs print come from the 2003 AHA/CDC scientific statement: below 1.0 mg/L is labelled low relative risk, 1.0 to 3.0 mg/L average, and above 3.0 mg/L high. Those are positions in a population distribution rather than targets, and the same statement asks for two measurements about two weeks apart, averaged, before a value is interpreted at all. A value above 10 mg/L is not the top of that scale: the statement says to repeat the test and look for infection or another source of inflammation.

Why does my hs-CRP change so much between tests?

Because large swings are normal for this marker. A 2024 meta-analysis of 60 studies found a median within-subject coefficient of variation of 0.44 for hs-CRP, with studies ranging from 0.27 to 0.76, and the European Biological Variation Database puts within-subject biological variation at 58.9%. At that level of variability, someone whose long-run average is 2.0 mg/L can read 1.1 on one draw and 3.2 on the next without anything having changed. This is exactly why two measurements are averaged.

Is an hs-CRP above 10 mg/L dangerous?

It is not read as a cardiovascular risk category at all. The 2003 AHA/CDC statement says a value above 10 mg/L should prompt a repeat test and an examination for sources of infection or inflammation, because acute inflammation moves this marker by orders of magnitude. An infection, a dental abscess, recent surgery, an injury, a vaccination or a flare of an inflammatory condition can all produce it. What it means for you is a question for the clinician who ordered the test, not for a reference range.

Does lowering hs-CRP lower my risk?

There is no good evidence that lowering CRP as an end in itself changes anything. The CRP CHD Genetics Collaboration studied 194,418 people across 47 studies and found that per 1 SD higher genetically raised CRP the risk ratio for coronary heart disease was 1.00 (0.90 to 1.13), against 1.33 (1.23 to 1.43) for measured circulating CRP. The authors concluded CRP is unlikely to be even a modest causal factor. It is a marker that travels with risk, not a lever that produces it.

Is hs-CRP the same as CRP?

It is the same protein measured with a more sensitive assay. Standard CRP is used to detect obvious inflammation and often cannot resolve the low end of the range, while hs-CRP can distinguish 0.6 from 2.4, which is what cardiovascular banding requires. Units are the other trap: hs-CRP is usually reported in mg/L and standard CRP often in mg/dL, so 3.0 mg/L and 0.3 mg/dL are the same result.

Should I get hs-CRP tested?

That is a clinical decision and this page does not make it. The 2003 AHA/CDC statement positioned hs-CRP as an optional test for people already judged to be at intermediate cardiovascular risk, to inform a decision that was already in play, and explicitly not as a screen for the whole adult population. If a test has already been run, the useful questions are whether a second confirmatory draw was taken two weeks later and whether anything acute was going on at the time.

Track it. Test it. Know for sure.

Log food, supplements, sleep and biomarkers by chatting with BioTrakk on Telegram, WhatsApp or the web — then let it find what actually moves your numbers.

Free on Telegram & Web. No app to install.