Principles of cancer staging, screening and tumour markers

• Staging describes how far a cancer has spread in the body. The commonest language is the TNM system: T for the primary tumour, N for nearby lymph nodes, and M for distant metastasis, often grouped into stages 0 to IV. ^[1][2]
• Staging is anatomical; it says where the disease is, not how abnormal the cells look (that is grade) or what the cells secrete (that is a marker).
• Before any treatment decision, performance status is recorded: how well the patient can carry on ordinary life. The two common scales are ECOG (0 to 5) and Karnofsky (0 to 100). ^[4]
• Screening is testing people who have no symptoms in order to find early disease or a removable precursor. It is judged by criteria such as those set out by Wilson and Jungner, and by modern additions on bias, harm and informed choice.
• Tumour markers help in known cancer: to support diagnosis, to estimate prognosis, to choose treatment, and to watch response or recurrence. They have generally failed as screening tests because they are neither sensitive nor specific enough.
• A single abnormal marker, scan or symptom never proves cancer; staging, screening and marker results are always combined with tissue diagnosis and the clinical picture.

By the end of this page you should be able to:

1. Define staging and distinguish it from grading, and give the purpose of cancer staging.
2. Read a TNM designation and say what T, N and M describe, and how the categories are grouped into stages 0 to IV.
3. Explain the difference between clinical and pathological staging, and the staging prefixes c, p and y.
4. Describe when a cancer does not use TNM, and why.
5. Record performance status with the ECOG and Karnofsky scales and explain its clinical uses and limits.
6. State the Wilson and Jungner screening criteria and the modern additions, and name the biases that distort screening results.
7. Explain screening, surveillance and re-screening, and the difference between them.
8. State what tumour markers are for, and why they are unreliable as tests for new cancer.

Staging, screening and tumour markers are the shared grammar of oncology. They appear in every organ chapter, in every tumour board, and in every clinical trial. A trainee who can stage a cancer, read a performance score, appraise a screening test and interpret a marker in context is already most of the way to competent cancer practice. None of these tools is, by itself, a diagnosis: each is a piece of evidence that has to be fitted together with the history, examination, imaging and biopsy.

This page is a principles page. The organ chapters (breast, lung, colorectal, prostate, cervix, thyroid, blood cancers and others) carry the site-specific rules, the treatment implications and the exact staging tables. When a statement here disagrees with an organ chapter on treatment, the organ chapter wins on treatment; this page wins on the shared biology and definitions.

The guideline editions used to write this page are the UICC TNM Classification of Malignant Tumours, 9th Edition, published in 2025 ^[2], with the structural definitions that teaching material still anchors to the 8th Edition, published in 2016 ^[2]; the AJCC Cancer Staging Manual editions summarised by the SEER training programme (8th edition, 2018 onward) ^[3]; the National Cancer Institute cancer staging summary, reviewed 14 October 2022 ^[1]; the NCI summary of screening tests, updated 27 September 2024 ^[7]; and the NCI tumour marker fact sheet, reviewed 7 December 2023 ^[8]. Performance status is taken from the ECOG scale published by Oken and colleagues in 1982 and maintained in the public domain by ECOG-ACRIN ^[4].

A few working definitions settle the terms before anything harder is asked of them.

These four ideas are often blurred. Three confusions matter most for a student:

• Staging is not grading. Stage measures place and spread; grade measures how abnormal the cells look under the microscope and how fast they are likely to grow. A tiny high-grade tumour can be stage I, and a large low-grade tumour can be stage IV if it has spread. Both are recorded, and both matter: from breast cancer onward, the stage group is determined by TNM, grade and biomarker status together ^[9].
• Staging is not diagnosis. Staging tools such as imaging, endoscopy and physical examination tell you where disease is; the diagnosis rests on tissue ^[1].
• A marker is not evidence of cancer. An elevated marker can come from benign conditions, and a normal marker does not exclude cancer ^[8].

Staging is the part of the record that does the most work. It determines treatment, gives a first approximation of prognosis, allows results to be compared between hospitals and over time, lets cancer registries and screening programmes measure stage shift, and defines which clinical trials a patient can enter. Because stage at diagnosis never changes, even a good response to treatment is recorded as “downstaged on treatment” while the original stage remains the label used for statistics ^[1].

The practical aim is to describe the disease accurately enough to make decisions and no more. In most cancers this is achieved by answering four questions: where is the primary tumour, how far has it grown, are the nearby lymph nodes involved, and is there distant spread. TNM does exactly this.

The TNM system is the most widely used cancer staging system ^[1]. It was developed by the UICC and AJCC and is the common language in which cancer extent is recorded worldwide ^[2]. It is an anatomically based system: it records the primary and regional nodal extent of the tumour and the absence or presence of metastasis ^[2]. It is applied at every hospital because its categories are what imaging, surgery and pathology are all translated into.

• T describes the size and extent of the primary tumour, that is, the original tumour where the cancer began ^[1][2]. It usually combines a size threshold with the invasion of neighbouring structures. Higher numbers mean a larger or more locally extensive tumour. T is often subdivided for precision (for example T3a and T3b) ^[1].
• N describes the number and location of regional lymph nodes that contain cancer ^[1][2]. N0 means no involved nodes, and N1, N2, N3 describe increasing numbers or involvement of more distant regional nodes.
• M describes whether the cancer has metastasised: whether it has spread from the primary tumour to other parts of the body ^[1][2]. M0 means no distant spread; M1 means distant spread has occurred.

The first category that matters clinically is often M. If M is 1, the patient has metastatic disease, and the local size of the primary tumour becomes secondary in deciding whether the intent of treatment is curative or palliative.

The individual values seen in pathology reports carry specific meanings ^[1]:

• TX, NX, MX: the corresponding feature cannot be assessed.
• T0: no primary tumour can be found.
• N0: no cancer in nearby lymph nodes.
• M0: cancer has not spread to other parts of the body.
• M1: cancer has spread to other parts of the body.

Stage group is assigned using a table in the site chapter ^[3], but the letters and numbers are always written together in a short code (for example cT2N1M0 or pT3N0M0), so that the detail is preserved even after grouping.

The TNM combinations for a given cancer are grouped into fewer stages for clinical use. The grouped stage is what “stage I” and “stage II” on a report mean ^[1]:

Substages such as 2A, 2B and 2C are common and usually depend on the exact TNM combination and on tumour-specific factors such as grade. Stage IV in particular is not a single state: it ranges from a solitary metastasis to widespread disease, and the difference is governed by the M category and the site-specific table.

A second, simpler anatomical language exists and is used more by registries than by clinicians ^[1]:

• In situ: abnormal cells present but not spread.
• Localized: cancer confined to the place it started, with no sign of spread.
• Regional: spread to nearby lymph nodes, tissues or organs.
• Distant: spread to distant parts of the body.
• Unknown: not enough information to assign a stage.

Timing is the key to this distinction ^[3]. Recording at what point information was obtained is at least as important as the tumour size, because the same tumour has a different treatment before and after surgery.

Clinical staging is based on all the information gathered before definitive treatment begins: the physical examination, imaging, endoscopy and biopsy ^[3]. It is the stage used to decide whether a patient is a candidate for surgery, for neoadjuvant treatment, or for palliative care, and it is assigned only when the necessary fields can be completed ^[3].

Pathological staging is assigned after resection of the primary cancer and analysis of the surgical specimen, usually including the regional lymph nodes ^[3]. Pathology sees what imaging could only infer, so the stage can move. In some sites a tumour can be pathologically staged without a resection of the primary if a distant site is pathologically evaluated ^[3].

Where both exist, both are recorded and are marked with a prefix:

• c (clinical) before treatment begins, such as cT2N1M0.
• p (pathological) after surgery, such as pT3N0M0.
• y (post-neoadjuvant) when systemic or radiotherapy was given before surgery, indicating that the pathological stage reflects treatment effect rather than the original tumour.

The TNM classification thus has a dual system: a pre-treatment clinical classification and a post-surgical pathological one ^[3]. The pathological stage is the more accurate local statement, but it cannot be assigned to patients who are not operated on, which is why the clinical stage remains the one used for them.

A few rules follow from this. The stage does not chase the disease: a cancer is always referred to by the stage given at diagnosis ^[1], with new information about spread or response added on rather than replacing it. Staging is not restaging by default. It is also an error to quote a pathological stage for a patient who received neoadjuvant therapy without the y prefix, because the specimen then reflects treated disease.

TNM is not the staging system for every cancer ^[1]. Brain and spinal cord tumours and blood cancers are common examples with different systems ^[1], and the UICC/AJCC approach is only one of many named classification schemes that exist for specific sites ^[2]. Other cancers carry their own named systems (for gynaecological and haematological cancers, for instance) which must be used where the site chapter requires them. The reason is pragmatic rather than ideological: a tumour of the brain, a leukaemia or a lymphoma does not spread in the same anatomical sequence as a solid tumour growing into adjacent tissue and draining regionally, so forcing it into T, N and M would misdescribe it.

Even within the solid tumour world, staging systems are revised as evidence accumulates. Breast cancer is a good illustration: before surgery, stage is assigned as clinical prognostic stage, and after surgery as pathological prognostic stage. Both are numbered 0 to IV, and the number is determined by TNM, tumour grade and biomarker status together ^[9]. This means a modern breast stage is not a purely anatomical label, and comparing breast stage with, for example, lung stage on the same 0-to-IV scale can mislead. Lung cancer retains the more traditional anatomical TNM grouping.

Once the extent of the tumour is known, the second statement on the chart is how the patient feels and functions. Performance status is this statement. It is quick, costs nothing, and is among the most powerful bedside prognostic tools in oncology.

The ECOG (Eastern Cooperative Oncology Group) Performance Status Scale runs from 0 to 5 ^[4]:

The scale is in the public domain. It should be quoted with its name, and the original reference is Oken and colleagues, 1982 ^[4]. Key elements of it first appeared in the literature in 1960, in the earlier Zubrod report ^[4].

The other widely used scale is the Karnofsky Performance Status, a percentage scale from 100 (normal, no complaints, no evidence of disease) down to 0 (dead), in steps of 10 and first introduced in 1949 ^[4]. The two scales are not interchangeable in detail, but they track each other, and one commonly used mapping is ^[4]:

Performance status is used in three ways, and a bedside clinician should record it for all three:

• Treatment selection. Whether a patient can tolerate a proposed therapy is the first question in most solid tumour guidelines, and it is answered by function as well as by organ function. Fit patients with good performance status are the ones in whom systemic anticancer therapy is usually attempted; where performance status is poor, the risks of treatment may outweigh the benefits. The final decision still rests on the individual, the tumour biology and the available regimens.
• Prognosis and communication. A good performance status is one of the strongest favourable factors in most advanced cancers; a deteriorating score over weeks is a more reliable signal of decline than a single measurement.
• Trial eligibility and data. Performance status is a standard entry criterion for cancer trials and a required variable in cancer registries, so that results can be stratified and compared fairly.

Performance status has two well-known limits that a student should know rather than discover later. The first is subjectivity: clinician assessment of performance status can vary, and some tools are being studied to make it more objective ^[4]. The second is that performance status, by itself, is not a measure of quality of life. It measures the ability to perform self-care and usual roles; it says nothing about mood, symptoms, family support or preferences. A patient with ECOG 2 and severe pain is not the same as a patient with ECOG 2 and no symptoms, and the score alone should not drive a treatment decision.

In practice the two are recorded as a pair before the first treatment decision: “T3N1M0, ECOG 1”. Clinical stage decides what is technically and oncologically possible; performance status decides whether the patient can survive the treatment that the stage would imply. Neither stage nor score is a diagnosis, and neither extends survival on its own. Both are needed, and both are reversible only in the sense that a better assessment can replace an earlier, less informed one.

Screening is, in principle, simple: find disease early and treat it. In practice it is one of the most contested areas of medicine, because it is offered to healthy people, most of whom will never develop the disease. It therefore has to meet a standard of evidence that symptomatic treatment does not.

A screening test cannot be judged by the same tests as a diagnostic test. For a diagnostic test in a symptomatic patient, sensitivity and specificity describe performance in people who already have a reason to be tested. For a screening test applied to thousands of well people, the absolute numbers of false positives, false negatives, overdiagnoses and missed cases dominate the harm–benefit balance.

The starting point is the report commissioned by WHO in 1968, Principles and practice of screening for disease, by Wilson and Jungner. The ten classic criteria are still quoted as the gold standard of screening assessment ^[5]:

1. The condition sought should be an important health problem.
2. There should be an accepted treatment for patients with recognised disease.
3. Facilities for diagnosis and treatment should be available.
4. There should be a recognisable latent or early symptomatic stage.
5. There should be a suitable test or examination.
6. The test should be acceptable to the population.
7. The natural history of the condition, including development from latent to declared disease, should be adequately understood.
8. There should be an agreed policy on whom to treat as patients.
9. The cost of case-finding (including diagnosis and treatment of patients diagnosed) should be economically balanced in relation to possible expenditure on medical care as a whole.
10. Case-finding should be a continuing process and not a “once and for all” project ^[5].

Wilson and Jungner knew the exercise would not stay static; they wrote that any appearance of dogmatism should stimulate discussion, because real development depends on an exchange of views ^[5].

A modern, practical condensation comes from population screening work. Three ideas capture most of what matters ^[6]:

1. There is a screening test acceptable to those it is intended for, reasonably accurate, and offerable to large numbers of appropriate people.
2. The disease is treatable, and treating it at an early (screen-detected or preclinical) stage gives better outcomes than treating it when it presents with symptoms.
3. The harm and cost of detecting and treating early disease by screening are less than the harm and cost of not screening for it. This is cost-effectiveness, not cost-saving: any gain in quality or length of life must not come at a cost that society cannot sustain, including the physical and psychological harm of offering the test ^[6].

A systematic review of screening criteria proposed over the decades since 1968 found that most adaptations still overlap with Wilson and Jungner, and drew out a further set of “emerging” criteria ^[5]:

• The screening programme should respond to a recognised need.
• Its objectives should be defined at the outset.
• There should be a defined target population.
• There should be scientific evidence of effectiveness of the screening programme.
• The programme should integrate education, testing, clinical services and programme management.
• There should be quality assurance, with mechanisms to minimise the potential risks of screening.
• The programme should ensure informed choice, confidentiality and respect for autonomy.
• The programme should promote equity and access for the entire target population.
• Programme evaluation should be planned from the outset.
• The overall benefits of screening should outweigh the harm ^[5].

Three other modern lessons frame the debate ^[5]:

• The biology of invasive cancer ranges from indolent to aggressive; the test should target progressive, potentially lethal disease rather than detect anything it can find.
• Not every precursor is an obligate precursor; treating all of them means overdiagnosis.
• Effective screening reduces advanced disease, so registry data should be followed for stage shift rather than only for case counts; and benefit is not shared equally across risk groups, so the target population should be defined by risk and expected lifespan, not by age alone.

Being clear about these labels prevents both over-screening and missed cases ^[6]:

• Screening actively approaches large numbers of asymptomatic people, most of whom are free of disease, and offers a test that is either diagnostic or that identifies a high-risk group for further investigation ^[6].
• Surveillance is the testing of people at very high risk because of a known condition (an inherited mutation, Lynch syndrome, BRCA carriage, chronic inflammatory bowel disease). It does not involve pro-actively identifying high-risk individuals from within an average-risk population ^[6].
• Re-screening is the continuing part of screening. A negative test is never a lifelong exemption: most programmes re-invite people at fixed intervals, because both the person and the disease biology change over time ^[5].

In practice the line between screening and surveillance is grey, and the pragmatic solution is that some body (a screening committee or the relevant specialist service) takes ownership and applies an evidence-based method, with dialogue between organisations so that patients do not fall between two services ^[6].

Screening looks good so easily because of bias, and a student should be able to name the four classic ones ^[6]:

• Lead-time bias. Survival is measured from diagnosis to death. Screening lengthens the interval between diagnosis and death by finding the disease earlier, without necessarily postponing death. Apparent survival improves while mortality does not.
• Volunteer bias. People who attend screening are, on average, healthier and less deprived than those who do not, so screening appears to benefit a healthier group.
• Length (length-time) bias. Screening tends to find relatively slow-growing disease, because slow-growing tumours spend longer in the detectable preclinical phase. Survival figures are then better than for symptomatically detected fast-growing disease, even with identical treatment.
• Overdiagnosis. Screening inevitably detects disease that was never destined to cause symptoms or death, because the person would have died of something else first. This leads to treatment that only causes harm.

These biases mean that a screening programme cannot be validated by simple before-and-after or case-series data. Well-conducted population-based randomised trials, with outcomes analysed on an intention-to-screen basis (including non-attendees and interval cancers), are the only way to be sure ^[6]. Even a mortality reduction in a trial is not enough on its own if the harms and costs are unacceptable ^[6].

An effective screening test finds cancer early, reduces the number of people in the screened population who die of that cancer, and has more benefit than harm—accepting that harm includes bleeding and other physical injury, false positives, false negatives, overdiagnosis and overtreatment ^[7]. With that benchmark, current programmes fall into three groups ^[7]:

Tests that are generally recommended for average-risk adults:

• Breast cancer. Screening mammography reduces breast cancer deaths in women aged 40 to 74, most clearly in women aged 50 to 69. Expert groups generally recommend that women begin breast cancer screening at age 40 ^[7].
• Cervical cancer. HPV tests and Pap tests, used alone or together, find and treat abnormal cells before cancer develops. Testing generally begins at age 21 and ends at age 65 for women with adequate prior screening and no high risk ^[7].
• Colorectal cancer. Colonoscopy, sigmoidoscopy and stool tests reduce deaths. People at average risk are generally recommended to begin regular screening between ages 45 and 75 ^[7].
• Lung cancer. Low-dose helical CT reduces lung cancer deaths in heavy smokers. Expert groups generally recommend screening some current or former heavy smokers from ages 50 to 80 ^[7].

Tests that are generally restricted, shared decisions or high-risk only ^[7]:

• Prostate cancer. PSA may slightly reduce prostate cancer mortality, but expert groups generally advise against routine PSA testing in average-risk men because of overdiagnosis and false positives; it may be offered to men at higher risk, such as those with a family history or of Black African origin, after discussion.
• Ovarian cancer. CA-125 (often with transvaginal ultrasound) has not been shown to be an effective screening test in asymptomatic women.
• Skin cancer. Routine clinical skin examination has not been shown to reduce deaths from skin cancer.
• Liver cancer. Alpha-fetoprotein with liver ultrasound is used in high-risk groups (for example cirrhosis), not as a general-population test.

Tests that are still being evaluated:

• Multi-cancer early detection (MCD) tests. These measure multiple biomarkers in blood and may signal several cancer types at once. Whether they reduce cancer deaths, and how to use them safely, is not yet established, and none has regulatory approval for screening at the time of writing ^[7].

Two reminders of the screening paradox are worth holding onto. In prostate cancer screening, randomised evidence of mortality reduction came with the cost of treating 27 men needlessly for every prostate cancer death prevented ^[6]. In breast cancer, mammographic overdiagnosis of ductal carcinoma in situ has repeatedly called that programme into question ^[6]. Screening always does some harm; the policy question is whether it does more good than harm in a well-defined population.

Screening is not a scanner: a screening programme needs adequate diagnostic and treatment services behind it, or finding a cancer cannot be followed by anything useful. Large parts of the population bear the greatest burden of cancers that are curable when found early but reach services late. In those settings the same principles apply, but the practical actions are different: visual inspection and biopsy for persistent lesions, treatment of chronic infection (for example H. pylori), HPV vaccination and vaccination-linked screening where cold chains allow, and clinical follow-up of symptoms that do not fit, bring patients to diagnosis earlier than buying a new technology. Where treatment is not available locally, build the referral pathway first.

A tumour marker is anything present in or produced by cancer cells, or made by the body in response to them, that gives information about the cancer ^[8]. Markers have traditionally been proteins or other substances (in blood, urine, stool, tumour tissue or other fluids) found at higher levels in people with cancer. Increasingly, genomic markers—tumour gene mutations, patterns of gene expression and epigenetic changes in tumour DNA—are used as markers, both in the tumour itself and in fragments shed into blood ^[8].

Some markers belong to one cancer (prostate-specific antigen to prostate cancer, thyroglobulin to differentiated thyroid cancer); others are raised in several (carcinoembryonic antigen, CA 19-9, CA 125) ^[8].

Tumour markers carry information about a cancer in seven ways ^[8]:

• Diagnosis support. A marker may help when cancer is suspected, but an elevated level does not prove cancer, because benign conditions can raise it. Diagnosis rests on tissue and imaging, with the marker in support.
• Type of cancer. A marker can suggest an origin when the primary site is unknown, or subclassify a tumour.
• Stage. Marker levels often correlate with tumour burden, which is one reason they appear in staging discussions.
• Prognosis. Some markers give an independent estimate of outcome.
• Treatment selection. Tissue markers such as HER2, oestrogen receptor and progesterone receptor identify candidates for targeted or hormonal therapy.
• Response to treatment. Serial measurements during therapy show whether the tumour is responding.
• Recurrence. Periodic measurement after treatment can detect relapse before imaging does.

Because a raised level can come from benign disease and a normal level can occur in cancer, marker results are combined with biopsy and imaging rather than used alone ^[8]. Serial measurements over time are usually more informative than a single value, because a trend reduces the noise of biology and of assay variation.

The table is a memory aid, not a substitute for the site chapter; volumes and cut-offs are method-dependent and are given in the site chapters and the NCI list of markers in common use ^[8].

The same thing that makes markers useful in known cancer makes them poor in screening. The NCI summary is explicit: studies of circulating tumour markers for screening have generally found that they do not work well, because they often fail to identify everyone with the disease (they are not sensitive enough) or flag people who do not have it (they are not specific enough) ^[8]. Early cancers release tiny amounts; many markers are also raised by infection, inflammation, pregnancy, liver disease, smoking and ageing. Measuring a panel rather than one marker does not yet solve the problem. Researchers are still working on multi-cancer detection tests that analyse multiple biomarkers together, but much remains to be learned about their harms and benefits, and none has regulatory approval for screening ^{[7] [8]}.

Marketing a marker as “an early warning of cancer” is therefore an ethical as well as a technical error. A patient with cancer and a negative result has been falsely reassured. Sensible practice keeps markers inside the indications above and insists on tissue and imaging to confirm or exclude disease.

Markers are not part of TNM staging, but they travel alongside it. A staging system still groups patients by anatomy; the added prognostic information from the pathology report, receptor status and marker trend is applied once the stage is fixed. For example, breast cancer staging integrates grade and biomarker status into the prognostic stage group ^[9], but the TNM element itself remains anatomical. In biliary and pancreatic cancer, marker trends are often used as endpoints in treatment trials, which means the stage and the marker each tell part of the story; neither replaces the other.

Where the level of a marker conflicts with the expected stage, the marker is not always wrong, but the disagreement should trigger review of the imaging, the pathology and the assay before any change in management. A modest marker level that climbs during treatment is a more trustworthy signal than a single high value.

Staging is not supposed to happen before diagnosis, but in practice a clinician works forward under uncertainty. A common pathway looks like this:

The flowchart is the shape of most oncology practice: tissue first, stage next, performance status throughout, and team-based decisions at the end. Note where this page deliberately stops. The exact T and N cut-offs, the allowed imaging, and the precise timing of restaging belong to the site chapter and to the current site-specific guideline; they change as evidence accumulates, and they should not be memorised from general notes.

Staging rules only become useful when read alongside the chapter for the organ involved. These are the points where this page meets those chapters; where they make a different management point on treatment, the organ chapter wins:

• Breast cancer. Clinical prognostic stage before surgery, pathological prognostic stage after; both 0 to IV, determined by TNM, grade and biomarker status (HER2, oestrogen and progesterone receptor) ^[9]. Sentinel node results and receptor status change management even when the anatomical stage does not ^[9].
• Lung cancer. TNM remains the backbone; performance status (ECOG) drives whether systemic therapy is offered. Smoking history plus age range determine low-dose CT screening eligibility ^[7].
• Colorectal cancer. TNM plus pathological nodal status after resection; CEA is used for follow-up rather than for diagnosis; screening age and modality follow the regional programme ^{[7] [8]}.
• Prostate cancer. PSA is used for detection and follow-up, grade and T status for treatment selection; routine PSA screening in average-risk men is not recommended because of overdiagnosis ^[7].
• Cervical cancer. HPV and Pap cytology screen; precursors (CIN, HSIL) by grade; cancers are stage-clinical because of imaging and examination, with pathological assessment after surgery where undertaken ^[7].
• Hepatocellular carcinoma. Alpha-fetoprotein with ultrasound is used in high-risk groups (such as cirrhosis) rather than in the general population; the marker path is paired with liver function assessment ^{[7] [8]}.
• Ovarian, endometrial and skin cancers. CA-125, transvaginal ultrasound, skin examination and PSA are examples of tests that are useful in selected high-risk groups but have not been shown to reduce mortality in average-risk populations ^{[7] [8]}.
• Brain and spinal cord tumours and haematological malignancies. These do not use TNM and carry their own staging systems; treat their chapters as the reference ^[1].
• Rare tumours, sarcomas and germ cell tumours. Marker- and stage-specific rules dominate (AFP and beta-hCG in germ cell disease, for example); see the relevant chapter rather than this page ^[8].

• Clinical staging needs only an examination, baseline blood tests, imaging and biopsy. These are achievable in many settings and should be recorded fully before treatment; their absence is what turns a curable cancer into an exploratory one.
• Performance status costs nothing and should be recorded for every patient.
• Where pathology services are limited, the staging and marker discussion should still happen on clinical grounds, with the pathway documented so that later biopsy results can refine it.
• Screening should only be offered where there is a diagnostic and treatment pathway behind it. Where these do not exist, clinical care of symptomatic patients does more good than screening.
• Markers are not a substitute for biopsy. In settings where biopsy is difficult, imaging and marker trends may be the only available information, but they should be interpreted as suggestive rather than diagnostic.

Breast cancer and breast cancer screening; lung cancer and low-dose CT; colorectal cancer and polyps; prostate cancer; cervical cancer and HPV; hepatocellular carcinoma and hepatitis B or C; ovarian and fallopian tube disease; thyroid cancer; gastric peptic disease and gastric cancer; haematological malignancies (leukaemia, lymphoma, multiple myeloma); brain and spinal cord tumours; metastatic disease overview; hypercalcaemia and SIADH; thromboembolism in cancer; survivorship and follow-up; health systems and cancer control."