What prostate cancer is. The prostate is a gland the size of a walnut that sits below the bladder and surrounds the top of the urethra, and it makes the fluid that carries sperm. More than 95 in every 100 prostate cancers arise in its acini, the small glands that make that fluid, which is why the ordinary type is called acinar adenocarcinoma. It usually starts on the outer part of the gland, which is why it usually causes nothing at first, and why so much of it is found by a blood test rather than by a symptom.
Why the same cancer gets different risk labels on either side of the Atlantic. Britain and America stratify localised prostate cancer with different systems. NICE NG131 recommendation 1.2.15 uses the five Cambridge Prognostic Groups; the NCCN uses six bands, from very low to very high, and splits intermediate risk into favourable and unfavourable using criteria the Cambridge system does not have, including PSA density and the proportion of positive cores. NICE replaced its own three-tier table, which came from D'Amico, in the 2021 amendment because the three-tier model could not tell Gleason 3+4 from 4+3, and the Cambridge system can. A man reading an American source about his own diagnosis is reading a different map of the same ground.
Is the Gleason score being replaced by the grade group? No. They are used together, and a UK report in 2026 gives both. The Royal College of Pathologists dataset sets out the grade groups to be used 'in tangent with the Gleason score', and its proforma asks for each separately. The grade groups exist to split Gleason 7, which is really two diseases, 3+4 and 4+3, into grade groups 2 and 3. NICE's Cambridge table is written with both scales in every row for the same reason: the Gleason score is what appears in older notes and in most of the literature.
Why does my report mention cribriform or intraductalwhen my grade is only 7? Because since October 2024 the UK reporting dataset asks for it, and because within grade group 2 and 3 those patterns are where the risk sits. Invasive cribriform carcinoma is the worst-behaving form of Gleason pattern 4, and intraductal carcinoma is cancer filling ducts that still have their outer basal layer. Either one argues against active surveillance, and either one is a reason to be offered germlinegenetic testing under both the NCCN and the Philadelphia consensus recommendations.
Grade group 1 prostate cancer and the word cancer. Grade group 1, Gleason 6, is the grade that has prompted a long argument about whether it should be called cancer at all, because left alone it very rarely spreads. NICE's answer is practical rather than semantic: Cambridge Prognostic Group 1, which is grade group 1 with a PSA under 10 and stage T1 or T2, is the one band where active surveillance is offered first and radical treatment is a fallback if surveillance is unsuitable or unacceptable to the person (NG131 1.3.8).
Am I being watched, or watched and left? Active surveillance and watchful waiting are different plans with different intentions. Active surveillance monitors with PSA, repeat MRI and repeat biopsy so that a cancer that changes can still be cured. Watchful waiting, in NICE's own definition, is a strategy for controlling rather than curing, for people who do not want radical treatment or for whom it is not suitable, and relies on deferred hormone therapy. Both involve not treating today; only one keeps cure on the table.
How common is it, and what is the outlook? Worldwide, 1,546,112 new cases and 419,849 deaths a year, fourth for cases among all cancers and eighth for deaths (GLOBOCAN 2024). In the UK, 57,898 new cases and 12,300 deaths a year, and 78.9 percent of men survive ten years or more; the lifetime risk of being diagnosed is 17 percent (Cancer Research UK). In the United States 333,830 cases and 36,320 deaths are projected for 2026, and 69.5 percent are found while still confined to the prostate, where five-year relative survival is 100 percent (SEER). These are averages over everybody diagnosed and they are not a personal prognosis; for most men the question the numbers raise is not whether treatment will work but whether treatment is needed at all.
Why the histology tier here is short. The WHO fifth edition names ductal adenocarcinoma, treatment-related neuroendocrine prostatic carcinoma, adenoid cystic (basal cell) carcinoma, and squamous and adenosquamous carcinoma as types separate from acinar adenocarcinoma, and lists PIN-like carcinoma as a subtype of acinar adenocarcinoma with the atrophic, pseudohyperplastic, microcystic and foamy-gland appearances as patterns rather than entities. The first two have pages here because they change management; the rest are rare enough that the honest treatment is a sentence on this page and a glossary entry, not a thin page of their own. Urothelial carcinoma of the prostatic urethra is covered in the urinary tract chapter of the same classification, so it is not a prostate cancer record here.
Living with prostate cancer, care and decisions: This is orientation from public patient pages and guidelines, not advice for your case: your own team's instructions and 24-hour number come first. The eight decision rows on this page, the question sets for the urology, surgical, oncology, hormone therapy, sexual function and continence and palliative care appointments and for partners and carers, the first 60 days checklist and the red cards were written from NICE NG131 and NG234 and from the NHS, Prostate Cancer UK, Macmillan, Cancer Research UK, Marie Curie and Maggie's, with the patient-reported figures taken from the ProtecT trial's own publications rather than from summaries of them.
On the numbers in these rows. They come from ProtecT, a UK trial that randomly assigned 1,643 men aged 50 to 69 with PSA-detected, mostly Cambridge Prognostic Group 1 to 3 disease to active monitoring, prostatectomy or radiotherapy, and followed them for a median of 15 years with questionnaire response rates above 80 percent. They describe what happened to that group. They are not a forecast for one man, and no sentence here should be read as one. Where NICEhas put the same figures into its own box 2, the guideline's wording is used, because that is the table a clinician is meant to work through with you.
On plain words. These pages call incontinence leaking and calls impotence not being able to get an erection, and gives both their numbers. Men do not raise either subject, surveys consistently find they are not asked about them either, and the result is that the commonest long-term consequences of treating this cancer arrive unannounced. NICE NG131 (1.1.12) requires that sexual function, physical appearance, continence and other aspects of masculinity are discussed before treatment. Nothing here assumes the reader is old, uninterested in sex, or has a partner.
On hormone therapy. It is the longest-running part of this illness for most men who need it and the least covered anywhere. The row on living with hormone therapy collects what NICE offers for each effect, because almost every one of them has an answer: medroxyprogesterone for hot flushes, a fracture risk assessment and bone drugs where osteoporosis is present, six-monthly metabolic checks, supervised resistance and aerobic exercise twice a week for 12 weeks for fatigue, prophylactic breast bud radiotherapy for men on long-term bicalutamide, psychosexual counselling for couples, and intermittent therapy where the setting allows it.
The words that carry the screening argument, in one place. Lead time is how much earlier a screening test found the cancer, and lead-time bias is the mistake of counting that extra knowing as extra life; overdiagnosis is a lead time longer than the rest of the man's life; overtreatment is the harm that follows when an overdiagnosed cancer is treated. The number needed to screen turns all of it into one figure a health service can weigh: at 16 years in the European randomised trial, 570 men invited and 18 extra cancers diagnosed to prevent one prostate cancer death.
Two words for the same scan, and they are not the same score. The international system is PI-RADS, now version 2.1, which builds a 1 to 5 assessment category from fixed per-sequence rules. NICE NG131 1.2.2 asks United Kingdom reports to use a 5-point Likert scale instead, which is the radiologist's overall probability judgement using the clinical picture as well as the images. Both run 1 to 5 and both trigger a biopsy at 3 or more, and a man reading an American source about his own scan is reading a different scale.
The science in detail, 1 of 7, in full; the one-sentence version is on the overview. The androgen receptor, five routes to the same place. Amplification of the gene at Xq12: 48.9% of 444 samples in prad_su2c_2019, 52.0% of 150 in prad_su2c_2015 and 57.0% of 149 in prad_fhcrc, against 1.0% of 489 TCGA primaries and 4.0% of 424 metastaticcastration-sensitive samples, first shown as a treatment effect in 7 of 23 recurrent tumours with none in the matched pre-treatment specimens. Amplification of an intergenic enhancer 624 kb upstream, in 81% of 101 deeply sequenced metastases, invisible to coding-exon panels. Ligand-binding-domain mutation in 13.7 to 18.0% of castration-resistant cohorts against 0.4% of primaries, clustered on four residues: L702H, which makes the receptor responsive to glucocorticoid; T878A and H875Y, which open it to progestogens; and F877L, which turns enzalutamide and apalutamide into agonists. The AR-V7 splice variant, which deletes the drug-binding domain altogether, detectable in circulating tumour cells from 39% of men starting enzalutamide and 19% starting abiraterone. And simply raising receptor expression, which was shown in xenografts to be necessary and sufficient for resistance on its own. In plasma, some potential androgen receptor resistance alteration is present in 42% of 2,213 evaluable men, often polyclonal or compound.
The science in detail, 2 of 7, in full; the one-sentence version is on the overview. The landscape by stage. Localised disease, from the TCGA primary cohorts: ERG fusion 45.6% of 333, ETV1 8.7%, ETV4 4.8%, FLI1 1.2%; SPOPmutation 11.1%; FOXA1 mutation 5.7%; PTEN deep deletion 17.4%; TP53 mutation 11.5%, falling to 3.4% in the CPC-GENE localised whole genomes; NKX3-1 deletion 8.0 to 15.9%; CHD1 deletion 7.0%; MYC amplification 8.2%; RB1 deletion 9.4%; BRCA2 mutation 1.6% with deletion 3.5%; ATM mutation 4.0%; CDK12 mutation 1.8%. Metastatic castration-resistant disease, from prad_su2c_2019 and its companions: TP53 36.7%, PTEN deletion 25.7%, AR amplification 48.9%, MYC amplification 23.6%, RB1 deletion 9.7%, BRCA2 mutation 8.3%, ATM 6.1%, CDK12 5.9%, SPOP falling to 5.6%. Metastatic castration-sensitive disease sits between, from 424 men: TP53 30.4%, PTEN deletion 14.4%, AR amplification 4.0%, SPOP 12.5%. Ancestry adds a third axis: in 65 Chinese tumours, CHD1 deletion reaches 20.0% while ERG rearrangement is 6.2%, and in 2,069 men at one centre, tumours from Black men carried more AR alterations, fewer PTEN mutations and more 8q gain, and tumours from Asian men more FOXA1 and ZFHX3 alterations.
The science in detail, 3 of 7, in full; the one-sentence version is on the overview. DNA repair, germline against somatic. Inherited mutations in 20 repair genes were present in 82 of 692 men with metastatic prostate cancer unselected for family history, 11.8%: BRCA2 5.3%, ATM 1.6%, CHEK2 1.9%, BRCA1 0.9%, RAD51D 0.4% and PALB2 0.4%, against 4.6% in localised disease and 2.7% in a 53,105-person reference population, with no difference by family history or age at diagnosis. On a broader panel in 3,607 men, 17.2% had a positive result and only 30.7% of those were BRCA, with mismatch repair variants in 1.74% of all men tested and 37% of positives ineligible for testing under the guideline then in force. Somatically, BRCA2, BRCA1 and ATM aberrations together were 19.3% of 150 metastatic castration-resistant tumours. A tumour report cannot answer the germline question and a separate consented test is needed; a plasma report cannot answer it either, and in 10% of men an ATM, BRCA2 or CHEK2 variant in plasma has come from the bone marrow rather than the prostate, which a paired whole-blood control distinguishes. Pathology gives a hint: intraductal or ductal histology was present in 48% of germline carriers against 12% of non-carriers in one 150-man series, and localised tumours from germline BRCA2 carriers already look metastatic in their genomes.
The science in detail, 4 of 7, in full; the one-sentence version is on the overview. The four PARPinhibitor eligibility lists, read from the labels. Olaparib alone after enzalutamide or abiraterone: 14 genes, ATM, BRCA1, BRCA2, BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, RAD51B, RAD51C, RAD51D and RAD54L, germline or somatic. Olaparib with abiraterone: BRCA1 and BRCA2 only. Rucaparib after androgen receptor-directed therapy: BRCA1 and BRCA2 only, germline and/or somatic. Niraparib with abiraterone: BRCA2 alone in castration-sensitive disease and BRCA1 or BRCA2 in castration-resistant disease. Talazoparib with enzalutamide: 12 genes, ATM, ATR, BRCA1, BRCA2, CDK12, CHEK2, FANCA, MLH1, MRE11A, NBN, PALB2 and RAD51C, of which ATR, FANCA, MLH1, MRE11A and NBN are not on olaparib's list and seven of olaparib's genes are not on this one, with the label stating that no approved test for them exists. The response data behind the widest list: among men with a single-gene alteration and measurable disease in PROfound, confirmed objective response was 24 of 43 with BRCA2, 56%, 3 of 30 with ATM, 10%, and 2 of 34 with CDK12, 6%, with none among the eleven rarer genes, one or two men each. PPP2R2A was enrolled and then left out of the indication for unfavourable benefit and risk.
The science in detail, 5 of 7, in full; the one-sentence version is on the overview. PSMA, an imaging target and a therapeutic one. The tissue survey behind it found PSMA detectable in prostatic epithelium, duodenal mucosa, a subset of proximal renal tubules and colonic neuroendocrine cells and in no other normal tissue, staining 33 of 35 primary prostate adenocarcinomas and 7 of 8 nodalmetastases but only 8 of 18 bone metastases, with expression falling as tumours dedifferentiate. At staging, PSMA PET-CT was 27% more accurate than CT with bone scan in 302 randomised men, 92% against 65%, with sensitivity 85% against 38% and specificity 98% against 91%, management change in 28% against 15%, equivocal findings in 7% against 23%, and radiation of 8.4 against 19.2 mSv. At biochemical recurrence it localised disease in 475 of 635 men, 75%, with detection rates of 38% below PSA 0.5 ng/mL, 57% at 0.5 to 1.0, 84% at 1.0 to 2.0, 86% at 2.0 to 5.0 and 97% above 5.0, and positive predictive value 0.84 by histopathology. For radioligand therapy the criterion is lesion by lesion: at least one lesion with uptake greater than normal liver, and no lesion above a size threshold with uptake at or below liver. Of 291 men screened for TheraP, 200, 69%, were eligible. What the scan cannot do is give a histology or a genotype, and a PSMA-low, FDG-avid site in a man on hormonal treatment is one of the ways lineage change announces itself.
The science in detail, 6 of 7, in full; the one-sentence version is on the overview. Leaving the androgen receptor behind. The transition is epigenetic before it is genetic: whole-exome sequencing showed substantial genomic overlap betweencastration-resistantadenocarcinoma and neuroendocrine disease from the same men, fitting divergent clonal evolution, while methylation separated the two sharply and reassigned some tumours that looked like adenocarcinoma. The genotype that permits it is loss of both RB1 and TP53, shown in mouse models where Rb1 loss enabled lineage plasticity and metastasis and additional Trp53 loss caused antiandrogen resistance, with Ezh2 and Sox2 raised and EZH2 inhibition restoring receptor expression, and in human cells where the shift from luminal to basal-like identity required TP53 and RB1 loss and was mediated by SOX2. Combined RB1, TP53 and PTEN alterations were enriched in clinically defined aggressive variant disease, with RB1 copy loss the strongest single discriminator at 54% of 44 samples. Neuroendocrine is not the only exit: an androgen receptor-null, neuroendocrine-null phenotype sustained by FGF and MAPK signalling has emerged over two decades, and deep phenotyping resolves five states rather than two. Recognition is morphological first, using the six named categories, with AURKA and MYCN co-amplification in 40% of neuroendocrine tumours against 5% of adenocarcinomas, and cell-free DNA methylation discriminated the two states with an area under the curve of 0.96 and 1.0 in two cohorts totalling 101 men.
The science in detail, 7 of 7, in full; the one-sentence version is on the overview. Liquid biopsy, and why it matters more here. Prostate cancer metastasises to bone, and bone is the hardest tissue to biopsy well, so plasma is not a convenience here but often the only practical sample. Above a 2% tumour fraction, which 75.6% of paired samples reached, every somaticmutation found in a matched metastatic biopsy was also present in plasma, with copy-number calls in actionablegenes 88.9% concordant. At scale, 94% of 3,334 men had detectable circulating tumour DNA at a median fraction of 7.5%, and 93% of tissue-detected BRCA mutations were found in plasma including 100% of predicted germline variants. Plasma sees more than tissue where it counts: potential androgen receptor resistance alterations in 42% of 2,213 men, including polyclonal and compound mutations and exon 8 deletions, and multiclonal BRCA2 reversions at PARP inhibitor resistance. It cannot separate germline from somatic without a consented germline test, cannot give a histology, cannot measure PSMA, and cannot by itself distinguish a repair-gene variant from the tumour from one from the bone marrow: at a 2% threshold, 19% of men had clonal haematopoiesis variants in plasma and 10% had one in a gene used to decide PARP inhibitor candidacy, most often ATM, accounting for almost half of all somatic repair-gene variants detected. A paired whole-blood control fixes it. Before treatment starts, the useful plasma signal is in BRCA2, ATM and TP53 rather than in the AR amplification that dominates the report.
Frequencies marked cBioPortal were computed from the public API on 25 September 2026 on the deposited sample lists of prad_tcga_pan_can_atlas_2018 (494 sequenced, 489 with copy number, 494 with structural variants), prad_tcga_pub (333), prad_broad (112), prad_cpcg_2017 (477, mutations only), prostate_dkfz_2018 (313, mutations only), prad_eururol_2017 (65), prad_p1000 (1,013), prad_mskcc_2017 (504), prostate_msk_2024 (2,260), prad_msk_stopsack_2021 (2,069), prad_cdk12_mskcc_2020 (1,465), prad_mcspc_mskcc_2020 (424), prad_su2c_2019 (444), prad_su2c_2015 (150), prad_mich (61), prad_fhcrc (141 sequenced, 149 with copy number), nepc_wcm_2016 (114, mutations only) and mpcproject_broad_2021 (82), as sample-level counts of non-synonymous mutation, high-level amplification, deep deletion or structural variant. They are not the papers' own percentages, which are quoted alongside where the abstract gives them.
What the deposits cannot see, stated once so the rows do not have to repeat it. Fusion counts come from the structural-variant profile, so a cohort with no such profile reads zero rather than negative for TMPRSS2-ERG, and a targeted DNA panel that does not bait the breakpoint introns reads about 25% where whole-exome and RNA cohorts read 41 to 46%. The metastatic whole-exome deposits call high-level amplification on chromosome arms gained as a block, so genes near a real amplicon read high: NBN at 8q21 reads 20.5% in prad_su2c_2019 next to MYC at 8q24, and BRIP1, RAD51C, ATR and PMS2 read similarly inflated there, which is why only AR, MYC and the ETS loci are quoted as focal from those deposits. The CPC-GENE and DKFZ localised deposits carry mutation calls only, so their low TP53 and PTEN figures are partly a property of the deposit. Panel cohorts mix primary and metastatic samples, so their figures sit between the two stages rather than describing either.
Machine-readable versions
The same record for scripts and assistants; checked 2026-09-25. Data CC BY-NC 4.0, attribute “Data from OnCo (onco.cc)”.