Inflammation Markers

High-Sensitivity CRP

hs-CRP

C-reactive protein is made by the liver in response to inflammation, driven by the signalling molecules interleukin-6 and interleukin-1 beta. That chain matters, because it is the reason C-reactive protein turns out to be a readout of something else and not a thing in itself. What inflammation is actually doing is in Inflammation, in the Body Guide.

Standard CRP answers whether there is significant inflammation right now. hs-CRP is used to assess the low-grade background inflammation associated with long-term cardiovascular and metabolic risk.

The difference between the 2 tests is smaller than it used to be

The distinction was real when it was made. A conventional CRP assay was built for the levels seen in acute illness and was blind to the low end. A high-sensitivity assay is defined by its ability to report reliably below 1 milligram per litre and up to at least 10.

On modern analysers the gap has largely closed. A routine wide-range CRP now quantifies down to about 0.3 milligrams per litre against about 0.15 for a high-sensitivity assay, and in a direct comparison the 2 correlated at 0.988, with a mean difference of 0.19 milligrams per litre and complete agreement above 3 milligrams per litre. What survives of the distinction is mostly regulatory and how the result is reported, and not a different measurement.

That is not an argument that the label is marketing. There is a genuine historical analytical difference and a live regulatory one. It is an argument against assuming the 2 tests are measuring different things.

What it does after an event, in hours and days

C-reactive protein has a half-life of about 19 hours. It starts rising 12 to 24 hours after an inflammatory stimulus, peaks at 2 to 3 days, and falls away over the following days once the stimulus stops. So after an isolated event it normalises over days, and the fortnight people quote comes from the retest convention and not from the biology.

Infection and injury move it by orders of magnitude.

Exercise does too, and the size of the response tracks the duration and the eccentric load, and not the effort. Studies of 50-kilometre races report 2 to 10 milligrams per litre at 24 hours and a return to baseline by 48. A 100-kilometre race exceeds 25 at 24 hours. A 24-hour treadmill ultramarathon has been reported peaking at 92, and the 246-kilometre Spartathlon produced a 152-fold rise. An ordinary gym session or a parkrun does not do this. The common claim that any hard training raises C-reactive protein is too broad, and what the data supports is that prolonged endurance work does, steeply.

A result taken in the days after an illness, an injury or a long race is measuring that event.

Where the numbers come from

The bands in general use come from a 2003 scientific statement issued jointly by the American Heart Association and the Centers for Disease Control and Prevention. Its 3 bands are below 1 milligram per litre, 1.0 to 3.0, and above 3.0.

Two details about them get dropped almost everywhere, and both change how the bands read.

The middle band is labelled average, and not moderate. The statement's own words. That is a different statement about the reader, because it says the middle band is where most of the population sits.

The cut points are population thirds. They correspond to approximate tertiles of hs-CRP in the adult population, with roughly twice the risk in the top third against the bottom. They were not derived from a point at which anything changes. They were derived by cutting the population into 3.

For scale on how many people that puts in the top band, in a US national survey 32.7% of adults had an hs-CRP of 3 or above. Bands built this way put a large fraction of an apparently healthy population into the high band.

Above 10 milligrams per litre the result is generally taken to reflect acute inflammation and not cardiovascular risk. The statement's instruction is to repeat it and look for a source of infection or inflammation. What has to be added is that a persistently raised value above 10 is not an absence of information, although the risk relationship has not been mapped above that point. The reason for repeating is to separate a transient cause from a persistent one.

The relationship is log-linear, which is not the same as linear

The bands are lines drawn across a continuum for practical purposes and not points where something changes, and that part is right. The shape of the continuum needs getting right, because it changes what the bands mean.

Pooled across the prospective studies, the risk climbs smoothly with the number and there is no step in it. A 3-fold higher C-reactive protein comes with about 60% more coronary heart disease once age and sex are allowed for. Allow for the conventional risk factors and that falls to under 40%, and allow for fibrinogen as well and it falls to around 20%. Vascular death behaves the same way. Stroke does not, because its risk shrinks once the conventional risk factors are allowed for and then edges back up when fibrinogen is added. Most of the association is other risk factors travelling with it.

Risk rises by multiplication and not by addition. Equal multiplicative steps carry equal risk, equal absolute steps do not, and it is the log scale that makes the tertile boundaries arbitrary lines and not meaningful ones.

Someone at 2.9 and someone at 3.1 are not in different states, and they may not be different numbers

That point deserves more weight than it usually gets.

External quality assessment between 2018 and 2023 found differences between manufacturers of up to 50% in 2018, around 20% from 2020 onward, and still above 35% for one manufacturer group by 2023.

The concrete version is this. Given the same sample, 75% of labs using one manufacturer's assay reported below 2 milligrams per litre, while more than 75% of labs using another reported above 2. That is the same blood, the same protein, and opposite sides of the threshold that current guidelines use.

How stable one measurement is, and how many you would actually need

Your own C-reactive protein varies by around 40% from one measurement to the next, with published figures running from 27% to 76% depending on the group studied. Only about 55% to 62% of all the variation in a population sits between people. The rest is the same people bouncing around. And the studies of within-person variation do not themselves publish a figure for how far apart 2 results have to be before the difference means anything.

It can be derived, and it should be labelled as derived. At a within-person variation of 44% and an analytical variation in the low single digits, 2 hs-CRP results have to differ by roughly 2.2-fold before the difference is more likely real than noise. The European federation of lab medicine publishes a lower within-person figure in its biological variation database, which puts it nearer 2-fold. Because C-reactive protein is strongly skewed, the true limits are asymmetric, so treat any of these as rough central figures and not precise ones.

The same arithmetic answers the question of how many measurements it takes to know your own average. To place it within 20% of the truth takes around 19 measurements. Within 30% takes around 8. Within 50% takes around 3.

Measured repeatedly in the same stable people over a year, the spread widens the further apart the samples are. It runs at about 0.07 milligrams per litre within a day, 0.19 within a week, 0.36 between weeks and 0.63 between months. Against the 2.0 threshold, 46% of those people crossed from one risk category into another at least once, and 21% of them landed 4 or more times in the low-risk category and 4 or more times in the high-risk category. Same people, no change in their health.

What that does and does not mean. It does not mean a single hs-CRP is useless. The associations quoted above were built on single baseline measurements, with statistical correction for within-person variability, and they work at the scale of a population. What a single hs-CRP cannot reliably do is place an individual person in a band. Those are different claims and the second one is the one that matters to someone reading their own report.

Getting a stable number

The convention set out in the 2003 statement is 2 measurements, optimally 2 weeks apart, with the results averaged, taken when metabolically stable, and fasting is not required.

Averaged. Taking the lower of the 2 builds in a downward bias and is not what the 2003 statement says to do. The purpose of the second measurement is to confirm, and not to minimise.

And 2 measurements averaged is better than one without being anything like a settled figure. At a within-person variation of 44%, 2 measurements do not establish a set point, and the numbers above say how far off that is.

What moves it, with sizes where sizes exist

Body weight is the strongest modifiable influence. Each additional unit of body mass index raises the odds of an hs-CRP at 3 or above by about 12%. Genetic analysis supports the direction running from body mass to C-reactive protein and not the reverse.

Sex. Women run higher, with roughly 1.6 times the odds of an hs-CRP at 3 or above, and an average difference of around 40%.

Oral oestrogen, where the route is the whole story. Oral conjugated oestrogens at 0.625 milligrams for 8 weeks more than doubled C-reactive protein. Transdermal oestradiol at 100 micrograms had no effect at all, in the same women, in a crossover design. A second study found oral oestradiol at 1 milligram raised it by 75% against placebo over 13 cycles, reversing by 49% on withdrawal, while transdermal oestradiol at 50 micrograms produced no change. The mechanism is that an oral dose passes through the liver first, and the liver is where C-reactive protein is made. The combined contraceptive pill works the same way, and trial data have C-reactive protein more than 3 times higher on it.

Smoking. Current smokers average 0.57 milligrams per litre against 0.41 in non-smokers, and there is no dose-response with the number of cigarettes. Quitting shows no short-term fall, with the odds of a high C-reactive protein no better a year after stopping. Beyond that they drop by about a fifth in ex-smokers overall, and by nearly half once more than 8 years have passed.

Statins. They lower C-reactive protein independently of their effect on cholesterol. Pooling the trials, only 2 doses separated from control at all, simvastatin at 40 milligrams a day and atorvastatin at 80. In the JUPITER trial described below, rosuvastatin at 20 milligrams lowered hs-CRP by around 37%.

Sleep, where the popular claim fails. Pooled across the observational work, sleep disturbance, short sleep and long sleep all carry small associations with C-reactive protein. Deliberately depriving people of sleep does not reproduce them. Updated in 2026, a single night of total or partial sleep deprivation produces no change in C-reactive protein at all. Three or more nights at around 4.5 hours does, and the effect is a substantial one. So one bad night does not move it. A sustained run of short nights does.

Gum disease. Treating periodontitis lowers C-reactive protein by about 0.69 milligrams per litre at 6 months, with the biggest effect in people starting above 3. That is the same size of reduction as conventional lifestyle or drug treatment achieves, which is a fair calibration for how much any lifestyle change moves this number.

Alcohol, which is U-shaped. C-reactive protein runs at about 2.60 milligrams per litre in people drinking less than monthly, 2.20 at 1 to 3 times a month, 1.70 at 1 to 4 times a week, 1.60 at 5 to 7 times a week, and back up to 1.80 at 2 or more drinks a day. The pattern holds after adjustment. That is a real finding about the marker and it is not a recommendation. It is a snapshot rather than a follow-up, and the lowest-intake group in any such comparison contains people who stopped drinking because they were unwell.

Ethnicity, and this one matters in the UK. Followed for nearly 9 years across European, African and South Asian groups, hs-CRP ran at 1.0 milligrams per litre in European participants, 1.3 in African and 1.6 in South Asian. It predicted cardiovascular events in the European and African groups. In the South Asian group it did not predict events at all, despite that group having both the highest C-reactive protein and the highest event rate. The people who found it called for different inflammatory markers for South Asian populations. A 2026 analysis of the same cohort, using different endpoints and models, found the risk gradient holding across all 3 ethnic groups, so this is not settled. For a UK reader that is a significant limitation, and one to state instead of bury.

Is there a published optimal hs-CRP?

No. Not from any professional body.

What exists is 2 numbers that get mistaken for one. Below 1 milligram per litre is the bottom population tertile from the 2003 statement. Below 2 is a threshold used in research on people already on treatment after a cardiovascular event, where it marks the absence of what the literature calls residual inflammatory risk. Neither has been published as a target for a healthy person.

The 2003 statement also carried a recommendation against using repeat hs-CRP measurements to monitor the effect of treatment, at its strongest grade, and no guideline has reversed that. The 2025 statement described below calls for measuring it more, not for treating to a target.

So any optimal hs-CRP you see quoted, whether that is 0.5 or 0.3 or as low as possible, has no professional-body source behind it. Given what follows, it does not have a mechanistic case either.

A marker of risk, and not a target to treat

This is the same structural point the HDL cholesterol entry makes about a different number, and it applies here for the same reason. The evidence on it has also moved, from not established to actively contradicted.

The evidence that a high C-reactive protein identifies people at higher risk is strong and consistent. The evidence that C-reactive protein itself causes that risk is not, on 3 separate lines.

The genetic line, which is the clearest

Inherited variants that raise a marker for life make it possible to ask whether the marker itself does the damage.

The test has been run on variants in the CRP gene itself, each shifting C-reactive protein by up to 30% per copy and affecting nothing else. Genetically raised C-reactive protein carries no increase in coronary heart disease whatever. In the same analysis, measured C-reactive protein carried about a third more. So C-reactive protein itself is unlikely to be even a modest cause of coronary heart disease, and a separate test on a different variant reached the same answer.

The part that makes this more than a null result is what the same method found one step upstream. There is a variant in the interleukin-6 receptor gene that raises interleukin-6 by about 9% per copy while lowering C-reactive protein by about 8%, which is the signature of the receptor being blocked. Carrying it comes with about 5% less coronary heart disease. So interleukin-6 receptor signalling does have a causal role, and C-reactive protein is the thing it leaves behind.

So variants that change C-reactive protein directly do nothing to coronary risk. Variants that change interleukin-6 signalling do change coronary risk, and they change C-reactive protein in the opposite direction to their effect on interleukin-6. C-reactive protein is a readout of the interleukin-6 axis, and moving the readout is not the same as moving the axis. That is why lowering C-reactive protein for its own sake is a category error and not merely an unproven strategy.

The trial line, which used to support the opposite conclusion

Three trials are usually cited here and they need separating.

JUPITER, 2008. A randomised trial of rosuvastatin at 20 milligrams against placebo, in people with no cardiovascular history, an LDL cholesterol below 3.4 millimoles per litre and an hs-CRP of 2.0 or above. Events fell by 44% and death from any cause by 20%. Over 5 years, 20 people had to be treated to prevent one event, and 29 to prevent one heart attack, stroke or death.

Three criticisms survive and all 3 matter. It was stopped at a median of 1.9 years against a planned maximum of 5, and stopping early at a statistical boundary systematically overstates the effect. Rosuvastatin lowered LDL cholesterol by around half and C-reactive protein by around 37%, so the trial cannot attribute the benefit to either. And, most importantly, every participant had an hs-CRP of 2 or above, with no comparison group who did not. JUPITER therefore shows that statins work in people with a raised C-reactive protein. It cannot show they work better in them, which is the claim it is routinely cited for, and that gap has never been filled. New-onset diabetes was also higher on treatment, at 3.0% against 2.4%.

CANTOS, 2017. A randomised trial of canakinumab, an antibody against interleukin-1 beta, against placebo, in people with a previous heart attack and an hs-CRP of 2 or above. It lowered hs-CRP by 26% to 41% depending on dose and did not touch cholesterol at all. Only the middle dose worked, cutting events by about 15%. Death from any cause did not improve. Fatal infection was higher on treatment.

So CANTOS remains the only positive trial of a pure anti-inflammatory drug for cardiovascular outcomes, and its effect was a 15% relative reduction at one dose of 3, with no mortality benefit and excess fatal infection.

CIRT, 2019. A randomised trial of low-dose methotrexate against placebo, and it found nothing. It is often described as a failure of the inflammatory hypothesis. It is better read as a failure of the drug to engage the pathway, because methotrexate did not lower interleukin-1 beta, interleukin-6 or C-reactive protein at all.

Taken together, those 2 trials were read as localising the benefit to the interleukin-1 beta to interleukin-6 to C-reactive protein axis specifically, and not to inflammation in general. That reading fitted the genetic evidence above, and it was the explicit reason the next trial was built.

And then the next trial tested it directly, in 2026

ZEUS reported in July 2026, so far only as a top-line company announcement and not in a peer-reviewed journal. A randomised trial of ziltivekimab, which blocks interleukin-6, against placebo, in people with established atherosclerotic cardiovascular disease, chronic kidney disease and an hs-CRP of 2 or above.

No benefit at all. The drug did what it was supposed to, with both free interleukin-6 and hs-CRP reduced as intended. There was no difference in death from any cause, and more serious infections on treatment.

That was the direct test of the mechanistic story, in exactly the population a raised hs-CRP identifies, with a drug that hit the target it was aimed at. It failed. Two further trials in the same programme are still running, one after a heart attack and one in heart failure.

The prognostic value of hs-CRP is untouched by any of this. The idea of it as something to treat is not.

What about colchicine

It comes up because it is the one anti-inflammatory drug that has reached cardiovascular guidelines, so it needs stating accurately for a UK reader.

The positive trials are real. In stable coronary disease, low-dose colchicine cut events by about 30%, with 35 people needing treatment over about 2 and a half years to prevent one. After a heart attack, a separate trial cut them by about 23%. It was approved in the US in June 2023 on that basis.

Four things have to go alongside that. A larger trial in acute coronary syndrome, published after the approval, found nothing, and it confirmed the drug was working by showing a C-reactive protein 1.28 milligrams per litre lower than placebo. The stable-disease trial showed a possible excess of non-cardiovascular death, at about 50% more, on figures too imprecise to settle. Guidelines give colchicine their weakest positive grade, and only for chronic stable coronary disease. And 2 pooled analyses published in July 2025 disagreed with each other by a factor of 2, one finding a 25% reduction in events and the other 12%, alongside a 35% increase in gastrointestinal side effects. A pros-and-cons debate piece published the same day argued the benefits are overestimated and the recommendations too strong.

For the UK specifically, colchicine is not licensed for any cardiovascular indication. Any UK cardiovascular use is off-label.

And no guideline anywhere selects people for colchicine on the basis of an hs-CRP.

What the guidelines actually say, which is not one thing

The positions diverge, they have reversed inside one organisation, and the UK is not in the conversation at all. Setting them out is more useful than picking one.

The 2003 joint statement recommended measuring it in people at intermediate risk, at physician discretion, and recommended against serial testing to monitor treatment.

The European Society of Cardiology's 2016 prevention guideline recommended against it, at its strongest negative grade, on the basis that circulating biomarkers did not add relevant prognostic information. Its 2021 prevention guideline listed biomarkers under risk modifiers and made no specific recommendation either way. Then its 2025 joint update on lipid disorders listed a persistently raised hs-CRP above 2 milligrams per litre as a risk modifier for reclassifying someone's risk. That is a reversal inside one society across 9 years, which is itself a fair way to convey how unsettled this is.

In the US, the 2019 primary prevention guideline made hs-CRP at 2.0 or above a risk-enhancing factor for people at intermediate risk, and the 2026 lipid guideline keeps it among the risk enhancers. Against that, the US Preventive Services Task Force concluded in 2018 that the evidence was insufficient to add it to risk assessment.

The strongest recent voice in favour is a 2025 scientific statement from the American College of Cardiology, which called for broad measurement of hs-CRP alongside LDL cholesterol in both primary and secondary prevention. Two things about it. It is a scientific statement and not a guideline, which is a real distinction in how much weight it carries. And it was published before ZEUS reported, though its own abstract already hedged that trials of anti-inflammatory therapy in secondary prevention have not all succeeded and that such evidence is needed before broad recommendations can be made.

In the UK, NICE does not use hs-CRP anywhere in cardiovascular risk assessment. The UK risk tool is QRISK3 and it does not include it. No NHS pathway uses it.

So it earns its place as something that refines a risk estimate, and not as a number to drive down on its own. The unusual thing about the current state of the evidence is that the bodies recommending measurement more often are doing so while the evidence for treating the number has got weaker, and both of those can be true at once, because a marker can be useful to know without being useful to target.

Erythrocyte Sedimentation Rate

ESR

How fast red cells settle in a tube over 1 hour, measured as the distance they fall in millimetres. They settle faster when inflammation-related proteins are present, so it is a general and non-specific inflammation marker.

The mechanism, because it explains everything else about this test

Red cells carry a negative surface charge and repel each other, which keeps them apart and slows their fall. Acute-phase proteins neutralise that repulsion, the cells stack into columns, and larger aggregates sink faster.

The dominant protein is fibrinogen. A regression analysis of routine lab data found independent contributions from the alpha-2, beta-2 and gamma globulin fractions, from IgG, IgA and IgM, from complement C3, from fibrinogen and from the platelet count, all pushing it up, and from serum albumin pushing it down, because albumin is itself negatively charged.

The finding from that analysis that matters most here is that C-reactive protein had the third-strongest raw correlation with ESR and no significant independent effect on it at all. The same paper found that people with a high ESR and a low C-reactive protein had a completely different profile of underlying conditions from people with a low ESR and a high C-reactive protein. Its conclusion was that the 2 tests have different roles and should be used according to context.

That is as close as the evidence gets to answering whether these 2 markers measure the same thing. They overlap heavily and they are not interchangeable.

Why it is slow at both ends

C-reactive protein turns over in about 19 hours, so its blood level tracks production almost in real time. ESR tracks fibrinogen and the immunoglobulins, whose turnover is measured in days to weeks.

So ESR rises slowly, over 24 to 48 hours, and then stays raised for weeks after the event that caused it has resolved. It lags going up and lags coming down. It is a slower-moving average of a slower-moving set of proteins, which is a real property and not a defect, and it is the reason the 2 tests can genuinely disagree.

What it is actually still used for

It is an older test, largely superseded by hs-CRP for everyday questions. Where it survives, it survives in named conditions and not as a general inflammation check, so naming them is more useful than a vague statement about patterns over weeks.

Giant cell arteritis. The 2022 classification criteria award 3 points for a maximum ESR of 50 or above or a maximum C-reactive protein of 10 or above, with 6 points needed to classify. Note the or. ESR is interchangeable with C-reactive protein there and does not add to it, and it is the weaker of the 2. Across a population-based series of confirmed cases, an ESR of 50 or above picked up 59.7% while a C-reactive protein of 10 or above picked up 87.3%. More than 40% of confirmed cases had an ESR below 50.

Polymyalgia rheumatica. A raised ESR or C-reactive protein is a required entry criterion in the 2012 provisional classification criteria.

Rheumatoid arthritis. It appears in the 2010 classification criteria and in the DAS28 disease activity score, which has an ESR-based version in routine use.

Myeloma, and this one is current UK primary care policy. NICE recommends that for adults aged 60 and over with persistent bone pain or an unexplained fracture, the tests to request are a full blood count, calcium, plasma viscosity or ESR, a paraprotein check by serum protein electrophoresis, and free light chains. ESR is not in the international diagnostic criteria for myeloma. It is a trigger into a pathway, which is a different job.

Bone and joint infection, where it is retained in scoring systems for infection around joint replacements, though weakly.

Early-stage Hodgkin lymphoma, where it carries prognostic weight in staging. One UK hospital's own lab handbook goes as far as saying ESR should only be requested for early-stage Hodgkin lymphoma and for clinical trials, and recommends plasma viscosity for everything else.

The thing about this test that nobody tells you, which is that it expires

ESR is not a measurement of a substance. It is a measurement of a physical process, performed on whole blood that has to still behave like whole blood.

The reference method requires prompt analysis. UK NHS lab handbooks commonly state that the sample must reach the lab within 4 hours of collection. Some newer analysers are validated for longer. One manufacturer-linked study of a non-sedimentation method reported stability up to 28 hours at room temperature and 48 hours refrigerated, and marketed that as 7 times the traditional limit. At room temperature the window is measured in hours, with the reference method holding up for about 10. Refrigeration stretches it, and current guidance allows 24 hours for the reference method and 48 for that alternative one.

The part that makes this worse than a simple limitation is the failure mode. Stored red cells change shape and surface properties and their settling behaviour drifts, so a delayed sample returns a number that looks like any other.

The methods also disagree with each other. Quality assessment data from over 6,000 labs found that only 28% still use the unmodified reference method, with 72% using modified or alternative methods, and that results from those can differ from the reference method by up to 142%, and from each other by up to 42%.

Where the numbers come from

No professional body publishes a healthy or optimal ESR, and there is no evidence base for one, because ESR has never been studied as something to change.

What exists is reference intervals, and they disagree. The most recent published derivation, in healthy people aged 18 to 100 using the reference method, gives 1 to 22 for women of any adult age, 1 to 14 for men aged 18 to 69, and 1 to 22 for men of 70 and over. The conventional table still in wide use gives 0 to 15 for men under 50, 0 to 20 for women under 50, 0 to 20 for men over 50 and 0 to 30 for women over 50. One UK hospital lab publishes 0 to 7 for men and 0 to 10 for women.

Those are not small differences. An upper limit of 7 and an upper limit of 14 for the same group, from 2 credible sources, is a doubling. This is method-specific and lab-specific to a degree that makes publishing any single band misleading.

The age rule of thumb, and what it rests on

There is a widely used formula for the upper limit of normal. It is age divided by 2 for men, and age plus 10 then divided by 2 for women.

Here is what it is. It comes from a single-page letter published in 1983, with no abstract and no formal validation cohort, and it has been in continuous textbook and calculator use for over 40 years on the strength of looking sensible. It also disagrees with the modern measured intervals above, and in the permissive direction. For a 70-year-old man it gives an upper limit of 35 against 22 in the 2025 derivation, and for a 70-year-old woman it gives 40 against 22.

So it is a rule of thumb of unvalidated provenance that is looser than the measured data. That needs saying plainly instead of presenting it as a reference range, which is how it is usually passed on.

What shifts it, beyond inflammation

Age, sex, pregnancy and anaemia all shift it, and the list is longer than that.

Things that raise it: advancing age, being female, pregnancy, anaemia or a low haematocrit, since fewer cells means less hindrance and more stacking, large red cells, a raised fibrinogen or immunoglobulin level, a paraprotein, which is one of the largest effects there is and the reason ESR sits in the myeloma pathway, a low albumin, obesity slightly, reduced kidney function, infection, inflammation and malignancy, some medications, and a tube that is not perfectly vertical.

Things that lower it, including into the normal range when something is genuinely wrong: a high red cell count, an extremely high white cell count, sickle cells and spherocytes, whose shapes resist stacking, a low fibrinogen, very thick plasma, inadequate mixing, and a sample that has partly clotted.

Above 100 the false-positive rate for serious underlying disease is low, and the usual causes are infection, connective tissue disease or metastatic cancer.

Is it obsolete?

Not obsolete, and not a general-purpose inflammation test either. Both of those get said and neither is right on its own.

The case for retiring it is substantial. Ordered together, ESR and C-reactive protein agree 67% to 81% of the time, and a 2025 pathology position states that routine ordering of both is unnecessary, with the 2 disagreeing in around a quarter of cases. Where both were measured, ESR was raised in 54% of people and hs-CRP in 76%. Nearly every raised ESR came with a raised hs-CRP, but only about two-thirds of raised hs-CRPs came with a raised ESR, so ESR mostly finds what C-reactive protein has already found and misses about a third of what it catches. The Choosing Wisely campaign has recommended against routine ESR since 2015, and programmes that discourage it have reduced how often it is ordered.

The sharpest version of the argument came in 2025, in a paper that called ESR and C-reactive protein zombie tests. Reviewing the evidence in bone and joint infection, endocarditis, appendicitis and childhood bone infection, it found both markers right between about half and four-fifths of the time, which is close enough to chance that the result does not change what anyone decides. It also cited a randomised trial in general practice where giving the clinician a point-of-care C-reactive protein result had no effect on antibiotic prescribing at all.

Two things bound that argument. It is specifically about diagnosing and predicting infection, and it does not address hs-CRP for cardiovascular risk or ESR in giant cell arteritis, polymyalgia rheumatica or rheumatoid arthritis. And it drew a published rebuttal and an author's reply, so it is a strong minority position inside infectious diseases and not a settled verdict.

Where that leaves it. ESR keeps defined roles in the named conditions above, including one that is current NHS primary care policy. Outside those, C-reactive protein is faster, more interpretable and less confounded, and the professional bodies advise against ordering both routinely. For a panel taken by someone who feels well, ESR has no guideline-supported role at all.

The other inflammation markers, and why they are not here

These come up often enough to need naming, with the honest status of each.

Ferritin is the one that matters most in practice, because it is already on the iron panel and it doubles as an inflammation marker. Inflammation raises ferritin independently of iron stores, which means a normal ferritin alongside a raised C-reactive protein may not mean iron repletion. The World Health Organization's 2020 guidance says to use higher iron-deficiency cut-offs when infection or inflammation is present, at below 70 micrograms per litre in adults against below 15 otherwise, and defines the relevant inflammation as a C-reactive protein above 5. Ferritin runs about 30% high where only C-reactive protein is raised and about 90% high where both acute-phase markers are, and ignoring that underestimates iron deficiency by roughly 14%. The iron entries have the detail, and how iron is stored and moved is in Iron, in the Body Guide.

Fibrinogen is the protein that actually drives ESR, so it is the mechanism behind this whole entry. Its story runs exactly parallel to C-reactive protein's. Higher fibrinogen tracks smoothly with coronary disease, stroke and vascular death, at roughly double the risk per gram per litre with age and sex allowed for, falling to about 1.8 times once the established vascular risk factors are added. The genetic test finds no causal effect at all. And adding either marker to a risk model barely moves it. In practice that comes out at roughly one extra cardiovascular event prevented for every 400 to 500 intermediate-risk people screened, which is a fair calibration for what any of these markers adds.

Interleukin-6 is the marker that is mechanistically more important than C-reactive protein and clinically less useful, which is an unusual combination. It sits upstream, the genetic evidence above implicates its receptor causally, and it predicts events strongly, at around two-thirds more risk in the top quarter of the population against the bottom. What rules it out in practice is that the assays vary between methods, it has a pronounced daily rhythm, there is no established cut-off, it costs a great deal more, and a single measurement captures only 35% to 41% of a person's true long-term average. The 2025 American College of Cardiology statement concluded it adds no predictive value on top of an hs-CRP of 2 or above, and a 2026 review from the same body stated that measuring it will not change management. And the drug that blocks this pathway is the one that failed in 2026.

Serum amyloid A behaves like C-reactive protein with a wider range. Its real role is monitoring a specific rare form of amyloidosis. It has no guideline role in cardiovascular risk, the assays are not standardised, and it is not widely available.

The neutrophil to lymphocyte ratio is free, because it is arithmetic on a full blood count you have already had. Pooled, it carries somewhere between about 1.6 and 2.4 times the risk of coronary disease, acute coronary syndromes and stroke. Its problem is fatal for reading an individual result. There is no standardised cut-off and every study picks its own, which is why the pooled figures are unstable and why the range around them is so wide. Publication bias is a stated concern on top of that. No guideline uses it. What neutrophils and lymphocytes each do is in Blood cells, in the Body Guide.

GlycA is worth naming because of what it shows about hs-CRP. It is a nuclear magnetic resonance signal from sugar groups on a set of circulating glycoproteins, so it is a composite index of the acute-phase response instead of a single protein. It predicts cardiovascular events comparably to hs-CRP across several large cohorts, and it correlates with hs-CRP only moderately, at 0.45 to 0.67, so the 2 are not the same signal. Its real advantage is stability. In the work comparing the 2 directly its within-person variation was around 5% against about 30% for hs-CRP, which is at the low end of the range quoted for hs-CRP above, and its sex difference was under 10% against hs-CRP's 40%. That contrast is a direct measure of how much of an hs-CRP result is noise. What blocks it is availability. It needs a nuclear magnetic resonance analyser, no standardised reference ranges exist and no guideline mentions it.

Plasma viscosity measures the same underlying protein changes as ESR, by measuring the plasma directly, and NICE treats it as an acceptable alternative to ESR in the myeloma pathway. Several NHS trusts now recommend it in preference. Because it is measured on plasma instead of on whole cells, it does not have ESR's stability problem.