Other Markers

Creatine Kinase

CK, creatine phosphokinase

An enzyme held inside muscle fibres in large quantities, which leaks into the blood when those fibres are disrupted. It measures muscle breakdown, not muscle strength. What the enzyme is for, why creatinine comes out of the same chemistry, and what happens at the severe end are covered in Muscle, in the Body Guide. This entry is about reading the number, which for this marker is harder than for anything else in the guide.

Where the numbers come from

UK labs have largely settled on one interval. Pathology Harmony recommends 40 to 320 U/L in men and 25 to 200 U/L in women, and NHS lab directories publish those figures, but not all inform you of the caveat that applies. The caveat is that they were derived from a White Caucasian group and that other groups may have higher values.

The difficulty is not that UK labs disagree with each other. It is that the agreed figure sits a long way from everything else that has been measured. Applying the international reference procedure to hospital patients produced upper limits of 171 U/L in men and 145 U/L in women, roughly half the Pathology Harmony number. Population studies that excluded anyone who had trained in the previous 3 days put the male upper limit between 382 and 1,001 U/L, several times higher. The figure printed next to your result is a working convention adopted for good practical reasons, and it is not a measurement of where healthy people sit.

Ethnicity is the largest single reason those figures disagree

Work published in 2026 from primary care requests in the West Midlands, checked against a second dataset in south-west London, gives upper limits of 757 U/L for Black men, 357 U/L for Asian men and 314 U/L for White men. For women the same figures are 389, 188 and 170 U/L. Black upper limits run about 2.3 times White ones in women and 2.4 times in men, and Asian limits sit slightly above White ones.

Applying a single interval to everybody has a measurable cost. Doing so classifies between 22.6% and 30.3% of results from Black individuals as high, against 5.1% to 7.7% of results from White individuals. At 5 times the upper limit, the figure NICE attaches to its statin decisions, 1.82% of Black men sit above the line against 0.45% of White men.

It is not muscle mass. That explanation has been tested directly and rejected. Adjusting for body mass index, waist circumference and upper arm circumference left the difference essentially unchanged, which points at how much enzyme the tissue makes or how fast it is cleared instead of at how much tissue there is. The age effect goes the other way, and largely does disappear once body composition is accounted for. The 2 findings are easy to get backwards, and most secondary sources do.

Ethnicity-specific intervals are not available for routine use anywhere, and the people who produced the 2026 figures describe them as the first for a UK population. The Pathology Harmony caveat does exist but only a few NHS lab directories publish it alongside the interval, but no adjusted number comes with it. A requester is told the limit may be wrong for this person and given nothing to use instead.

The 2026 UK authors argue for separate intervals on 2 grounds. Over-flagging drives repeat testing and referrals that were never needed. And because statin decisions are keyed to multiples of the upper limit, a limit set too low for someone can lead to a statin being withheld or reduced when it should not be, which is a cardiovascular harm produced by a reporting convention.

The arguments against are real too. Ethnic categories are not consistent between countries, so a figure derived in one is not transferable to another. Recorded ethnicity in health data is incomplete and imperfect. And a higher threshold raises the chance of missing disease, although the same authors note that creatine kinase usually rises many-fold in genuine muscle disease, which makes that risk small. One question sits unresolved underneath all of it. The published multiples of 3 and 5 times were worked out against the conventional limit, and whether they can be carried over unchanged to a re-based one has been queried and not settled. Where someone has mixed ethnicity, the published advice is to read the result against the interval of the higher group.

Sex and age

Men run roughly twice as high as women in every dataset, within every ethnic group. Values reach adult levels at about age 13. In men they peak between 20 and 50 and decline from 60 onwards. In women they rise towards the end of the reproductive years and then fall. One claim to ignore is that oestrogen protects women from exercise-induced muscle damage. It does in animals. Human work shows either no difference between the sexes or women being somewhat more affected.

Exercise, and how long to leave it

Our instruction is 48 hours without heavy training. That is the general pre-test rule and it is the right instruction for most people most of the time. Two things make it an imperfect one for this marker specifically.

The first is that the only interval anyone has studied is 3 days. People who reported strenuous activity within the previous 3 days had close to twice the odds of a raised result. That establishes 3 days as a window in which exercise still moves the number. It does not establish that 3 days is enough, because no study has compared one interval against another.

The second is that the peak arrives in days, not hours. After maximal lengthening work the level was still more than 4 times baseline at day 10. Eccentric work, meaning the lowering half of a lift and downhill running, is what does it, and the size of the rise scales with how heavy the load was.

So 48 hours is a floor and not a guarantee, and after a first session of something genuinely unfamiliar no short interval is enough. If a result comes back high and you trained in the preceding week, a repeat is more informative than the number in front of you.

The same session does not produce the same number twice

Repeat an identical damaging session weeks later and the muscle damage markers, creatine kinase among them, come out 71% to 91% smaller. The protection develops after a single bout, lasts weeks to months, and a portion of it transfers to the opposite limb. Even trivial loads confer some of it.

This is the most important thing to know before comparing 2 of your own results. A reading taken after a novel session and a reading taken after a familiar one are not comparable, even in the same person doing the same exercise on paper. It is also why "I do this every week and my creatine kinase is fine" and "I did this once and it came back at 4,000" are both true statements about the same workout. Muscle, in the Body Guide sets out why the adaptation happens.

The spread between people is enormous, and it is not explained

Put a large group through an identical protocol and a quarter exceed 10,000 U/L while nearly half stay below 2,000 U/L. Same exercise, same muscle, same protocol. The terms high responder and low responder are a description of that, not an explanation of it. Candidate genetic signals exist, but no single variant accounts for a useful share of the variation and the studies behind them are small. Nobody can tell you from a blood test that you are a high responder by genotype.

Endurance events are a category of their own

A marathon commonly produces a few hundred U/L. A 160 km mountain race produces tens of thousands, and elite 24-hour runners have averaged above 50,000 U/L, with individuals spread very widely either side of that. The ultra figures should not be read as what running does. They are what extreme distance with a great deal of descent does.

Very high readings after prolonged hard exercise are common and frequently harmless, and there is no agreed level after exercise at which concern begins. The more useful finding is that among people admitted to hospital with exercise-related muscle breakdown, the creatine kinase level itself was not associated with who went on to develop kidney injury. Taking anti-inflammatory painkillers beforehand and being dehydrated were. Kidney injury was uncommon in that group, so the finding rests on a small number of events, so take the direction and not the size. Severe muscle pain, weakness, swelling or dark urine after training is a reason to be seen, and the number is not the thing to look up.

The thresholds that do exist were written for patients

NICE says to measure creatine kinase before starting a statin only in someone who has had persistent unexplained muscle symptoms, and during treatment only in someone who develops them. For people already on a statin and free of symptoms its instruction is explicit, and it is not to measure it at all. NICE's operative figure is 5 times the upper limit of normal, and even there it says to re-measure after 7 days and to act only if the level is still there.

European consensus guidance uses different figures. Above 4 times is where it says to act, above 10 times is where the drug should be stopped, and above 40 times is where kidney damage should be assessed. Some guidance elsewhere uses 3 times, a figure one published analysis calls arbitrary. The figures doing the same job, 3 times and 4 times and 5 times, span a factor of 1.67 between bodies. None of them was written for a well person who bought a test, and no body publishes a threshold for that situation.

Posting a sample introduces 2 errors that pull opposite ways

Haemolysis, meaning red cells breaking open, pushes creatine kinase up. A red cell enzyme released in the process behaves like creatine kinase in the assay. Reagents contain inhibitors for it and the inhibition is incomplete. The effect is proportionally largest when the true value is low, because a roughly fixed amount of interfering activity gets added whatever the starting point, which is the wrong direction for a healthy person. It is also strongly assay-dependent, with one kit crossing a 10% error at a haemoglobin concentration 4 times lower than another.

Time and warmth pull the other way, and less predictably. In separated serum held at room temperature creatine kinase falls. In whole blood at room temperature it changes substantially without a dependable direction, and it was named as one of the unstable markers in the work done specifically to establish whether blood could be posted at all.

The 2 errors do not cancel in any predictable way. Which one wins depends on the season, the transit and the individual sample. It is a genuine reason a repeat can differ from a first result with nothing having changed in you.

Is it standardised

There is an international reference method and a certified reference material, which is more than several markers in this guide have. That is not the same as results being portable. International quality assessment using material targeted by the reference method found that for enzymes as a class, standardisation had not been achieved in a fifth of labs and was questionable in the rest. In one Chinese city a recalibration exercise moved the share of labs within acceptable bias for creatine kinase specifically from 55.7% to 97.1%, which tells you where they had been sitting beforehand.

The reassuring part is local. The 2026 UK work compared Abbott and Roche analysers across the same population and found no detectable difference between them once age, sex and ethnicity were accounted for. The problem is the reference interval, not the measurement.

Macro-CK, and the pointless investigation it causes

A small number of people carry creatine kinase bound up in a larger complex that the body clears slowly, so it accumulates in the blood. There are 2 kinds. The more common one is the enzyme bound to an antibody, found mostly in older women. The other is an aggregate of the mitochondrial form.

What matters is how it presents. A raised total, isolated, unchanging over months or years, in somebody with no symptoms, no exercise explanation and normal everything else. It is a classic reason for a healthy person to be investigated repeatedly and to no purpose. A total creatine kinase cannot show it. Detecting it needs specific lab work that has to be asked for by name.

Among samples sent to a UK reference service over 10 years specifically to answer this question, about 12% were positive, and more than 80% of the antibody-bound cases were women. The second kind has been associated with malignancy in older hospital series, which makes it a reason for clinical assessment and not a conclusion.

How far a repeat has to move

On the figures in longest use, creatine kinase varies within a person by around 23% from one occasion to the next, and between people by around 40%. The ratio of those 2 figures is 0.57, which puts it in the band where a population range is a poor instrument for judging an individual. Your own previous result carries more information than the printed interval does.

On the same figures, a second result has to differ from the first by roughly two thirds before the change means anything at all. Treat that as an approximation and not a rule, because the distribution is heavily skewed and the real allowance is wider upwards than downwards.

What an isolated raised result usually turns out to mean

Among people referred to a specialist and fully investigated for an otherwise unexplained raised creatine kinase, a cause is found in about 1 in 4. The rest are left without one. People in that position who were followed for between 4 and 18 years did not deteriorate, in a group small enough that a quarter of them were lost along the way. That is the most reassuring finding attached to this marker and the least widely known.

Low results, and the thyroid

A low creatine kinase usually reflects carrying little muscle. An underactive thyroid can produce a low result, and it can also do the opposite. More than half of people with overt hypothyroidism have a raised creatine kinase. For subclinical hypothyroidism the one direct comparison found no significant difference from normal thyroid function, in numbers small enough that this is better read as nothing found than as nothing there. On thyroxine the level falls over a couple of weeks, often before TSH has corrected. European guidance names hypothyroidism first among the secondary causes to exclude in someone with muscle symptoms on a statin, and says that where a raised level persists, referral to a specialist in neuromuscular disease should be considered.

Biotin does not affect it

Creatine kinase is measured as an enzyme activity, by watching a reaction run, and not by an immunoassay. Biotin interference is a property of immunoassays built on a biotin capture step, so it does not arise here. The preamble lists which markers on the panel are affected and in which direction.

One thing to stop repeating

Routine drinking, and even heavy regular drinking, is not an established cause of a mildly raised creatine kinase. The largest population analysis found no association with heavy use, and the best clinical series of heavy drinkers found the enzyme raised in only 15% of them and called it an insensitive marker of alcoholic muscle disease. A binge can cause severe muscle breakdown with a very high reading, which is a different event with a different mechanism.

What it is really doing on this panel

Separating muscle from liver. Raised ALT and AST after heavy training, alongside a raised creatine kinase, is a muscle pattern on a report where 2 of those numbers are printed under a liver heading. The creatine kinase is what does the work there. A normal GGT alongside them makes an alcohol-related or cholestatic cause less likely without ruling liver disease out. Without creatine kinase on the panel there is no way to separate the muscle explanation from a liver one, and it is a common false alarm in people who train hard.

It is not a cardiac test. It was used as one before troponin existed, and it was displaced because the cardiac troponins are measured with assays aimed at heart-specific forms of the protein, which makes them far more specific to the heart than this is. They are not perfectly specific either, and muscle disease can raise one of them. A normal creatine kinase does not exclude a cardiac event, and a raised one does not suggest one.

Units

For this enzyme, units per litre and international units per litre mean the same thing, so a result reported either way needs no conversion.

Prostate-Specific Antigen

PSA

A protein made by the prostate. It belongs in semen, where it liquefies the fluid, and small amounts leak into the blood normally. More leaks when the gland is enlarged, inflamed or diseased. It is a marker of prostate activity and not a cancer test. What the protein does, why its name is wrong on 2 counts, and what a raised result opens up are covered in Prostate, in the Body Guide. This entry is about reading the number, and about the fact that the UK position on it changed twice in 2026.

What changed, and when

On 2 June 2026 the government accepted a recommendation from the UK National Screening Committee for England's first targeted prostate screening programme. It covers men aged 45 to 61 who have both a pathogenic BRCA2 variant and a family history of breast, ovarian, pancreatic or prostate cancer. A PSA test every 2 years, expected to begin rolling out in 2027.

The same recommendation goes the other way on everything else, and its reasons differ by group in a way that matters. For the whole population, and for men with a family history but no BRCA2 variant, the committee concluded that screening is more likely to cause more harm than good. For Black men it did not make that finding. It said there is ongoing uncertainty about whether screening would do more good than harm, and committed to continued work with UK researchers. The outcome is the same in both cases, which is no programme. The reasoning behind it is not.

The difference in risk that sits under those decisions is large. In England the lifetime chance of being diagnosed with prostate cancer runs at about 1 in 4 for Black men, about 1 in 8 for White men and about 1 in 13 for Asian men. For dying of it the same 3 groups come out at about 1 in 12, about 1 in 24 and about 1 in 44. Those are central estimates from English data now more than 15 years old, and the spread around each of them is wide, widest for Black men, where the diagnosis figure sits somewhere between 1 in 4 and 1 in 3.

The Asian figure gets almost no airtime, because the public argument is about the excess in Black men, which leaves a man of South Asian background to assume he sits where White men sit. He does not, on these numbers. The group counted as Asian here is the broad census category, so Indian, Pakistani, Bangladeshi and Chinese men are pooled into a single figure, and a lower recorded risk in a group that tests less is partly a lower chance of the disease being found rather than less of it being there. Men who do get a diagnosis are equally likely to die of it whatever their ethnicity.

BRCA1 appeared in the draft and was removed from the final version after further modelling put the risk it carries lower. The exclusions are contested. The TRANSFORM trial is designed to settle the question about Black men, who are built into its recruitment, and the family history question has been left with the model the committee used, which the government has funded to stay open for a further 3 years.

The Prostate Cancer Risk Management Programme has gone with it. That programme was the framework under which any man aged 50 or over could request a PSA test on the NHS and expect to be given it after a balanced discussion. Its 2 main government pages have been withdrawn and now redirect to the NHS test page, and in August 2026 the Department of Health and Social Care set out to GPs what replaces it.

Under the position that replaced it, a man can still ask his GP for the test whether or not he has symptoms, and the GP can still provide it, but the final decision rests with the GP. Age 50 on its own is no longer treated as a risk factor. GPs are pointed at the NICE summaries for the conversation.

The practical change is smaller than it sounds and the change in principle is larger. An entitlement became a discretion.

What that means for a test we sell

This is the one marker on our menu where a UK body has recently examined the evidence and drawn a line that excludes almost everybody who buys it. Most people buying a PSA test have no BRCA2 variant, no symptoms, and no family history, which puts them squarely in the group the screening committee concluded would be harmed more than helped by systematic testing.

We sell it because people ask for it, because the decision belongs to the person making it, and because a test bought with the argument in front of you is a different thing from a test bought without it. What we are not going to do is present it as a check everybody should have. The committee's reasoning is specific and it is not squeamishness. The test finds cancers that would have caused serious harm and cancers that would never have surfaced at all, and at the point of the blood draw there is no way to tell which has been found. The committee's own model put it that at age 60, about half the cancers screening finds would be ones that never needed finding. Its harm figures, measured 6 months after treatment, are 71% reporting bladder difficulty and 66% erectile difficulty after surgery, and 38% and 48% after radiotherapy. Set against that, a large European trial with a median follow-up of 23 years found prostate cancer deaths 13% lower in the screened group, with the balance of benefit against harm improving the longer the follow-up ran. Both of those things are true at once, and that is the whole difficulty.

Where the numbers come from, and the gap in the middle

NICE publishes age-specific thresholds in guideline NG12, in micrograms per litre. Below 40, clinical judgement. From 40 to 49, above 2.5. From 50 to 59, above 3.5. From 60 to 69, above 4.5. From 70 to 79, above 6.5. Above 79, clinical judgement.

Those figures are for people with symptoms. NG12 attaches them to people who already have lower urinary tract symptoms, erectile dysfunction or visible blood in the urine. They were never derived as a reference range for somebody with no symptoms who bought a test, and using them that way is a common error.

For someone with no symptoms, no UK body currently publishes a threshold. The 3.0 in wide circulation was the withdrawn programme's recommended referral value for asymptomatic men aged 50 to 69, and it went when the programme went. A UK clinical consensus exercise in 2024 set out to agree a replacement figure for asymptomatic men in general practice and failed to reach consensus on one, on the grounds that the evidence for any age-specific value was not there.

So publishing a number as the UK threshold for an asymptomatic man asserts something no UK body currently asserts. The honest construction is to show the figures that exist, say who each one was written for, and say plainly that the middle is empty.

The figures that do exist do not change with ethnicity. NICE looked at whether the referral thresholds should be lower for men at higher risk, including men of Black African family background, and found no evidence covering those populations. Its committee said that setting different thresholds without that evidence "could do more harm than good, because the risks of under and overtreatment are both important", and it issued a formal call for research into thresholds in exactly those groups. So the UK position is not that risk is equal across ethnic groups. It is that nobody has measured what a different cut-off would do.

Whether PSA itself sits at a different level by ethnicity is a separate question, and in the UK it is unanswered. No UK lab publishes a PSA reference interval broken down by ethnicity, and there is no UK dataset to build such a thing from. Work outside the UK has derived age-specific PSA distributions inside single national populations, Taiwan the clearest of them, and those come out different from the bands in general use here. That is a single country, East Asian rather than South Asian, and not a figure for a man tested in the UK.

What filling that gap locally has looked like

Where there is no national figure, regions write their own, and they have not written the same one. One East of England lab changed its reported ranges in 2021 so that any man aged 50 to 69 at 3.0 or above was flagged for urgent referral, on a cancer alliance instruction. For a period a man of 60 with a PSA of 4.0 was referred there and would not have been elsewhere. That lab has since moved back to the NICE bands, so the divergence closed without the question being answered.

A 2025 cancer alliance consensus in Wessex shows what the gap looks like now. Its threshold table is headed as being for men with symptoms, and for men without them it hands the question to the Prostate Cancer Risk Management Programme, which no longer exists. The hole is not hidden. It is written into the documents.

The calibration problem, which is larger than almost anyone assumes

Two calibration scales are in use for PSA, and results reported on the international standard read about 20% lower than results on the older one. Regulators have published the arithmetic behind that. The historic 4.0 cut-off on the older scale corresponds to 3.1 on the international scale.

That matters because the thresholds still in circulation were built in different eras against different scales, and are now applied to results from analysers traceable to whichever standard the manufacturer chose. A 4.0 and a 3.1 can be the same man on the same day. Across the main analyser platforms the spread in total PSA between methods sits around 11%, which clears the minimum agreed performance target without much room to spare. Measured a different way, as bias against a consensus value across international quality schemes including the UK one, PSA comes out inside the stricter target and among the better harmonised tumour markers. Both of those are true at once, and neither makes a threshold portable from one lab to another.

What moves the number, largest first

A urinary infection with a fever is among the largest non-procedural effects there is. In the acute phase the median rise in total PSA was about 14 micrograms per litre, several times a whole threshold. It does not settle quickly either. The free fraction drops back within a month, but total PSA declines slowly, and the ratio between them stays abnormal in between a half and two thirds of men at 1, 3 and 6 months, in a pattern the researchers described as comparable to what is seen in prostate cancer. One cancer alliance's consensus says to consider repeating after 6 weeks before referring anyone who has had a recent infection. A PSA taken in the months after a feverish one is not readable.

A prostate biopsy or a resection of the prostate raises it by several units and takes 2 to 3 weeks to settle, and the published instruction is to leave at least 6 weeks before measuring. Flexible and rigid cystoscopy have no meaningful effect on it. Acute urinary retention more than doubles it and it falls back once the retention is relieved. That looks like the retention doing the work more than the catheter, because it happens whichever way the bladder is drained, and the study behind that found the catheter made no clinically important difference.

Ejaculation raises it by about 41% at 1 hour, which in absolute terms averaged under 1 microgram per litre, so it belongs in the same size class as the 2 items after it and not above them. Most men are back to baseline by 24 hours and almost all by 48, although a second study found 40% still above their starting point at 24 hours. The absolute size of the rise is larger in older men and in men whose baseline is already higher, so the men most likely to be near a threshold are the men in whom it moves most. We ask for 48 hours, which is the interval the original work recommended and a cancer alliance consensus repeats.

Cycling is genuinely contested. Recreational riders aged 50 and over, sampled within 5 minutes of finishing a long ride, showed a rise of about 0.23. Elite cyclists, younger and over a longer distance, sampled after 10 minutes sitting still, showed nothing. The likely explanation is age, timing and sampling and not a disagreement about the underlying fact, since a PSA signal is hardest to detect in exactly the young athletes who made up the second group. Either way the effect where one is found is large enough to push a man sitting just under a line over it, and nowhere near large enough to produce an alarming figure out of nothing. We ask for 48 hours off it, which is the conservative reading of a split literature.

A digital rectal examination does far less than it is credited with. One study put the average rise at 0.26 and another the median at 0.4, and both sets of researchers called an effect that size clinically insignificant. The sensible rule is to take the blood before the examination. Waiting a week afterwards is advice that has outlived its evidence.

Carrying more weight lowers it, by dilution into a larger plasma volume. That was first shown in men already diagnosed, and it has since been found in general population samples too. PSA has been show to be roughly 12-22% lower in obese and morbidly obese men.

The medication section, and the one almost nobody declares

Finasteride and dutasteride roughly halve PSA. The 2 manufacturers give different instructions for handling that, and both instructions sit in the UK product information. Finasteride's says that in anyone treated for 6 months or more the PSA value should be doubled before it is compared against normal ranges for untreated men. Dutasteride's says to establish a new baseline after 6 months and to evaluate carefully any confirmed rise above the lowest recorded value, even where that figure is still inside the normal range for a man not taking the drug. That last clause is the whole point of the strategy. Both instructions are defensible and they are different strategies, so knowing which drug is involved matters.

Doubling is not quite enough for long-term users. The correction needed to preserve the same median rises from 2 at 2 years on the drug to 2.5 at 7 years, which means the standard advice systematically under-corrects the people who have been taking it longest. That drift was measured on the 5 mg dose.

And then the part that matters most for a customer base like ours. Finasteride at 1 mg a day, the dose sold for male pattern hair loss, does the same thing as the 5 mg prostate dose. It lowered median PSA by 40% in men in their 40s and by 50% in men aged 50 to 60, over a year of treatment, and what happens beyond that is not yet established. A man taking it for his hair does not think of it as a prostate drug, and very often does not list it as a medication at all. If you take it, at any dose, for any reason, that fact changes how the number should be read, and nobody will know unless you say so.

The proportion of free to total PSA is not meaningfully changed by either drug, and both manufacturers say so. Neither of them recommends using it. What they say is that if it is used, no adjustment is needed.

Testosterone replacement

PSA rises more on testosterone than on placebo during the first year of treatment. In men with low testosterone who were screened at the outset to exclude those at high prostate risk, trials running a year or longer found prostate cancer at any grade, urinary retention, biopsy and new urinary symptoms all uncommon and no different from placebo. Those are small numbers of events, so read it as nothing found in the groups studied.

The British Society for Sexual Medicine recommends assessing the prostate before starting, then reviewing at 3 months, at 6 to 12 months and every 12 months after that, with PSA checked annually. Two things in its guidance cut against how this is usually described. Because PSA rises early on treatment, it says the level at 6 months should be taken as the baseline for monitoring, and not the one from before the first dose. And it sets 2 triggers for urological assessment and closer surveillance, a rise of more than 1.4 within any single year and a rate of rise above 0.4 a year sustained across more than 2 years. The second of those is a rate-of-change criterion, and rate of change has not held up well as a way of reading a first result.

Why 2 results do not make a trend

PSA velocity and PSA doubling time both circulate as ways of reading a rising number. Adding velocity to a calculation that already contains age, PSA, examination findings, family history and any previous biopsy improved it by an amount too small to act on, and the analysis that showed this concluded it should not appear in practice guidelines. Doubling time belongs to monitoring disease that has already been diagnosed, whether under surveillance or after treatment, and it has no established role in reading a first result.

The figure that is useful is how far a repeat has to move, and the answer is further than most people assume. The tidiest estimate, a rise of 15% to 20%, comes from young healthy men with low values sampled under conditions designed to strip out everything except their own biology. In men of screening age sampled 2 weeks apart the measured variation came out around twice as wide, and among men retested a few months later before a biopsy decision around a third showed a rise too large for measurement noise to explain. So a move from 3.0 to 3.2 is certainly noise, and a move from 3.0 to 3.6 may well be. A single rise is a reason to repeat the test and not a reason to conclude anything from it.

Biotin pushes it the wrong way

PSA is measured on the kind of immunoassay in which biotin makes the result read falsely low, not falsely high. That is the dangerous direction, because a falsely reassuring result is the one nobody repeats. Copy asserting that biotin raises PSA has it backwards, and the error is common.

Two qualifications belong with that. Not every analyser uses the biotin-based chemistry, so on some of them the interference cannot happen at all. And the concentration a current reformulated assay tolerates is far above what an ordinary hair or nail supplement produces, with the very high doses used for other purposes being the ones that reach it. Even so, a falsely low PSA on a much smaller dose has been reported, and it delayed someone's treatment. The preamble covers the general rule, which is to stop it and repeat.

Free PSA, which we do not sell

Free PSA is harmonised markedly worse than total PSA, and it is the one measurement here that misses even the loosest agreed target, with 2 major platforms sitting more than 20% either side of the middle. The ratio is built on it and inherits the problem. The 25% cut-off in circulation was validated only in men whose total sits between 4 and 10 with a prostate that felt normal on examination, on one generation of one assay. It is not a general-purpose number. Prostate, in the Body Guide sets out what it and PSA density are used for and who arranges them.

What a raised result actually leads to

A scan before a needle. NICE directs that for suspected localised prostate cancer a multiparametric MRI comes first, reported on a 5-point scale. A score of 3 or more leads to a biopsy aimed at what the scan found. A score of 1 or 2 is grounds for considering no biopsy at all, as a shared decision after the risks have been talked through, and if someone chooses one anyway it is offered. In the trial behind that pathway, 28% of the men scanned had a reassuring result and had no biopsy. That is set out in Prostate, in the Body Guide, and it changes how a raised PSA should be read, because the overdiagnosis argument that shaped the screening debate was built in an era when a raised PSA led fairly directly to a needle.

Units

Micrograms per litre and nanograms per millilitre are the same unit, so UK and US figures need no conversion.