Cancer screening is the systematic use of tests to identify possible cancer or precancerous abnormalities in people who do not have symptoms of the disease being sought. It is a form of medical screening, not a definitive diagnosis: an abnormal result usually requires further investigation, and a normal result does not exclude all cancer. Its principal purposes are to reduce deaths from particular cancers and, where precursor lesions can be detected and removed, to prevent invasive disease.
Screening differs from the prompt investigation of symptoms, from surveillance after cancer treatment, and from monitoring people with previously identified abnormalities. The same technology may serve several of these purposes, but its benefits, interpretation, and acceptable risks depend on the population and clinical setting. A test that helps diagnose symptomatic patients is not necessarily suitable for screening millions of apparently healthy people.
Cancer screening includes established programs for breast, cervical, colorectal, and lung cancer in defined populations, as well as more selective approaches for other malignancies. Its scientific evaluation must distinguish genuinely improved outcomes from earlier labeling of disease, detection of harmless abnormalities, and differences between people who do and do not participate. Effective screening therefore depends on an entire pathway—eligibility, testing, diagnostic assessment, treatment, and quality assurance—rather than on a test alone.
Biological Rationale and the Screening Pathway
Screening can improve outcomes when a cancer has a detectable phase before symptoms appear and intervention during that phase changes its subsequent course. The biological opportunity varies substantially between tumors. Some develop through recognizable precursor lesions over many years; others disseminate early or progress so rapidly that periodic testing rarely identifies a useful treatment window. Even within one organ, cancers differ in growth rate, molecular characteristics, and potential for metastasis.
Cervical and colorectal screening illustrate prevention through precursor detection. Persistent infection with certain human papillomavirus types can lead to cervical precancer, while many colorectal cancers arise through adenomatous or serrated lesions. Removing clinically significant precursors can reduce subsequent invasive cancer. Breast and lung screening more commonly aim to identify an existing malignancy at a stage when treatment is more effective. These distinctions influence the appropriate interval between tests and whether reductions in cancer incidence, mortality, or both are expected.
The detectable preclinical phase is not identical to the time since a tumor first developed. A lesion must reach a size, location, or biological state that a particular technology can identify. Mammography detects some tissue changes before they can be felt, whereas a blood-based biomarker test requires a sufficient and distinguishable biological signal. Test sensitivity can therefore vary with tumor size, subtype, anatomical location, and the characteristics of surrounding tissue.
A screening pathway begins with a defined eligible population and continues through invitation or access, informed participation, specimen collection or imaging, interpretation, communication of results, and diagnostic assessment. An abnormal stool test, for example, does not complete colorectal screening: it generally requires colonoscopy. Suspicious breast imaging may require additional views, ultrasonography, or biopsy. Diagnostic confirmation establishes whether cancer is present and provides information needed for treatment.
These dependencies explain why a screening test cannot be evaluated independently of the health system delivering it. Detecting an abnormality without timely access to diagnostic assessment and effective treatment may produce distress without the intended health benefit. Conversely, improvements in treatment can change the incremental value of screening over time. Screening policy consequently concerns both the biology of disease and the practical capacity to complete a reliable clinical pathway.
Measuring Test Performance and Clinical Benefit
The basic measures of diagnostic sensitivity and specificity describe how a test classifies people with and without the target condition. Sensitivity is the proportion of affected people with a positive result; specificity is the proportion of unaffected people with a negative result. Neither measure alone tells a participant how likely a positive result is to represent cancer. That probability also depends strongly on disease prevalence in the screened population.
The positive predictive value can be expressed as
where denotes sensitivity, specificity, and prevalence. In a hypothetical population of 10,000 people with a cancer prevalence of 1%, a test with 90% sensitivity and 95% specificity would identify approximately 90 true positives and 495 false positives. Its positive predictive value would be about 15%. This illustrates why screening low-prevalence populations demands particularly careful control of false-positive results.
Clinical benefit requires evidence beyond accuracy. A randomized controlled trial can compare an invitation to screening with usual care or another screening strategy. Cancer-specific mortality is a major outcome; invasive cancer incidence is also important when precursor treatment can prevent disease. All-cause mortality captures deaths from every cause but may require very large studies because deaths from the target cancer constitute only a fraction of all deaths. Advanced-stage cancer incidence can be informative, but it is not automatically a validated surrogate endpoint for mortality.
Survival after diagnosis is especially misleading as a stand-alone measure. Lead-time bias occurs when screening advances the date of diagnosis without postponing death, making measured survival appear longer. Length bias occurs because periodic screening preferentially detects slower-growing tumors with longer detectable preclinical phases. Participants may also differ from nonparticipants in health behavior and access to care, producing selection effects in observational comparisons.
Absolute effects are essential for interpretation. A 20% reduction in relative risk has different implications when the underlying risk is 1 in 100 than when it is 1 in 10,000. Useful reporting specifies the number invited or screened, duration of follow-up, screening rounds, cancer deaths prevented, and major harms. Intention-to-treat analysis preserves randomized comparisons by analyzing participants according to assignment, although nonattendance and screening in the control group can reduce the observed difference between trial arms.
False Results, Overdiagnosis, and Treatment-Related Harm
A false-positive screening result indicates an abnormality when the target cancer is absent. Consequences can include repeat testing, additional imaging, invasive procedures, expense, and psychological distress. Some abnormalities resolve with a short assessment; others require months of surveillance. False-positive rates depend on the positivity threshold, interpretation rules, previous test results, and the population being screened. Across repeated rounds, the cumulative probability of experiencing at least one false alarm can be substantially greater than the probability in a single round.
False-negative results occur when a test fails to identify an existing target condition. Causes include sampling error, lesions hidden by surrounding tissue, biological features that produce little detectable signal, and interpretation errors. An interval cancer is diagnosed after a negative screening examination and before the next scheduled screen. Such cancers may have been missed, been below the detection threshold, or developed rapidly afterward. Their frequency and characteristics are important indicators of program performance.
Overdiagnosis is distinct from a false positive. It is the diagnosis of a genuine cancer or precursor that would not have caused symptoms or death during the person’s remaining lifetime. It can result from very slow growth, biological nonprogression, or death from another cause before the lesion becomes clinically important. Because pathology may correctly identify malignant cells, greater diagnostic accuracy does not by itself eliminate overdiagnosis.
Overdiagnosed disease can lead to overtreatment, including surgery, radiation therapy, or drug treatment that cannot improve the affected person’s cancer outcome. These interventions may nevertheless cause infection, organ dysfunction, persistent symptoms, or other adverse effects. Overdiagnosis is difficult to quantify because the untreated lifetime course of an individual screen-detected cancer is generally unknown. Population estimates require long follow-up and careful allowance for lead time, underlying incidence trends, and competing mortality.
Screening procedures themselves also carry risks. Colonoscopy can cause bleeding or perforation, particularly when lesions are removed. Imaging may involve ionizing radiation, while incidental findings can trigger investigations unrelated to the target cancer. The balance between benefit and harm changes with age, underlying cancer risk, other illnesses, and life expectancy. Benefits may take years to emerge, whereas many harms occur immediately. Shared decision-making therefore concerns not simply whether a test can find cancer, but whether its expected consequences align with a person’s circumstances and preferences. Balanced information distinguishes a reduction in cancer mortality from a guarantee of avoiding cancer or death.
Breast Cancer Screening
Screening for breast cancer is based primarily on mammography, a low-dose X-ray examination that can reveal masses, architectural distortion, and microcalcifications. Modern systems include digital mammography and digital breast tomosynthesis, which reconstructs images from multiple projections. Mammography can detect invasive tumors and ductal carcinoma in situ, a noninvasive abnormality with a variable risk of progression. Not every detected lesion would become life-threatening.
Randomized mammography trials began in the 1960s and subsequent decades, including the Health Insurance Plan trial in New York and several Swedish trials. Their collective evidence supports a reduction in breast cancer mortality from screening in certain age groups, although estimates of the size of benefit and overdiagnosis vary with study design, follow-up, and analytic assumptions. Treatment has also improved substantially since these trials, complicating direct translation of older absolute benefits into contemporary settings.
Guidelines differ in starting age, interval, and upper age limit because organizations weigh evidence and trade-offs differently. In 2024, the US Preventive Services Task Force recommended biennial mammography for women aged 40–74 at average risk. Other organizations use different schedules, and organized national programs often invite narrower age ranges. Such recommendations are population policies, not claims that every person within an age band has identical risk.
Breast density affects interpretation because dense fibroglandular tissue can obscure tumors on mammograms and is itself associated with increased breast cancer risk. Supplemental ultrasonography or magnetic resonance imaging can detect additional cancers in selected settings, but may also increase false positives and biopsies. MRI has an established role in screening some people at substantially increased inherited or familial risk, commonly alongside mammography. Evidence that a supplemental test detects more lesions is not equivalent to evidence that it reduces mortality.
Breast screening also illustrates the interaction between diagnostic thresholds and treatment decisions. Ductal carcinoma in situ contributes to concerns about overdiagnosis because its natural history is heterogeneous and prediction of progression remains imperfect. Conversely, some aggressive tumors arise between screening rounds despite regular attendance. Clinical breast examination has been studied in particular settings, but its evidence base and implementation differ from those of mammography. Formal breast self-examination programs have not demonstrated a breast cancer mortality benefit in large randomized studies. Population screening and the diagnostic assessment of a newly noticed breast abnormality remain separate clinical activities.
Cervical Cancer Screening
Cervical cancer screening targets a disease with a well-characterized infectious cause and identifiable precancerous stages. Persistent human papillomavirus infection with oncogenic types is the central causal factor in nearly all cervical cancers. Most HPV infections clear without progressing, so the presence of HPV does not mean that cancer is present. Screening strategies distinguish transient infection from patterns associated with significant precancer or invasive disease.
Cervical cytology, commonly called the Pap test, examines cells collected from the cervix for morphological abnormalities. Its development is associated with George Papanicolaou and colleagues, whose work helped establish a practical method for detecting cervical disease before symptoms. Organized cytology programs introduced during the twentieth century were followed by major declines in cervical cancer incidence and mortality in several countries. Their effectiveness depended on repeated participation, laboratory quality, and reliable follow-up.
Primary high-risk HPV testing detects the causal infection rather than cellular changes alone. It is more sensitive than cytology for detecting important cervical precancer and can support longer intervals after a negative result in appropriate populations. Because positive results are relatively common, particularly at younger ages, programs use triage strategies such as HPV genotyping, cytology, repeat testing, or combinations of these. Diagnostic assessment may include colposcopy and directed biopsy.
Treatment of significant precancer interrupts progression toward invasive cancer. Excisional procedures remove affected tissue for examination, while ablative approaches destroy eligible lesions without providing an excision specimen. The choice depends on lesion characteristics, the ability to exclude invasive disease, and program resources. Excessive investigation or treatment of abnormalities likely to regress can cause harm; excisional treatment is associated with increased risk of some subsequent adverse pregnancy outcomes, particularly with larger or repeated excisions.
HPV vaccination changes the future screening population but does not make screening immediately unnecessary. Vaccinated cohorts have lower rates of vaccine-type infection and associated precancer, requiring eventual reassessment of starting ages, intervals, and triage thresholds. In 2020, the World Health Organization adopted a cervical cancer elimination strategy with 2030 targets of 90% HPV vaccination coverage among girls by age 15, 70% screening coverage with a high-performance test by ages 35 and 45, and appropriate treatment for 90% of those identified with cervical disease. HPV self-sampling can expand access, but a positive result still requires a functioning follow-up pathway.
Colorectal Cancer Screening
Colorectal cancer screening can both detect existing malignancy and prevent cancer by identifying lesions that can be removed. Many cancers develop through the adenoma–carcinoma sequence, while others arise through serrated pathways. These processes often unfold over years, creating opportunities for testing before symptoms develop. However, not all polyps progress, and some lesions are difficult to detect because of their flat shape, location, or subtle appearance.
Stool-based tests look for evidence associated with cancer or advanced precursors. Guaiac-based fecal occult blood tests detect the chemical activity of heme, whereas the fecal immunochemical test detects human hemoglobin. FIT is generally easier to use and permits adjustment of the positivity threshold, which changes both detection and the number of colonoscopies required. Multitarget stool DNA tests combine molecular markers with a blood-detection component. A positive stool test generally requires colonoscopy to determine its cause.
Endoscopic methods directly examine the bowel. Flexible sigmoidoscopy evaluates the rectum and distal colon, whereas colonoscopy examines the entire colon when complete and adequately prepared. Colonoscopy permits biopsy and polyp removal during the examination. Its performance depends on bowel preparation, completion, careful mucosal inspection, lesion recognition, and complete removal. CT colonography uses computed tomography to evaluate the colon but cannot remove lesions and may reveal findings outside the bowel.
Randomized trials of repeated guaiac-based stool testing and flexible sigmoidoscopy have demonstrated reductions in colorectal cancer mortality; sigmoidoscopy trials also demonstrated reduced incidence. Evidence for colonoscopy includes observational studies and randomized trials. The NordICC trial, reported in 2022, found lower colorectal cancer incidence after approximately ten years among people invited to a single screening colonoscopy, while the reduction in colorectal cancer mortality was not statistically significant at that follow-up. Only about 42% of those invited underwent the procedure, illustrating the importance of distinguishing invitation effects from outcomes among participants.
In 2021, the US Preventive Services Task Force recommended screening average-risk adults aged 45–75, with selective screening from 76–85 according to health, prior screening, and preferences. Other programs use different age ranges and test schedules. People with Lynch syndrome, certain family histories, prior advanced polyps, or longstanding colonic inflammatory bowel disease follow distinct surveillance pathways. At the program level, the effectiveness of stool testing depends on repeated participation and timely colonoscopy after positive results, not simply on distributing test kits.
Lung Cancer Screening
Screening for lung cancer uses low-dose computed tomography to identify pulmonary nodules and other suspicious abnormalities in populations with sufficiently high risk. Earlier approaches based on chest radiography and sputum cytology did not establish the mortality benefit subsequently demonstrated for low-dose CT. The principal screening evidence concerns people with substantial cigarette-smoking exposure, rather than the general population.
The US National Lung Screening Trial enrolled more than 53,000 participants aged 55–74 with heavy smoking histories. Its initial report in 2011 found a 20% relative reduction in lung cancer mortality with three annual rounds of low-dose CT compared with chest radiography. The Dutch–Belgian NELSON trial, reported in 2020, also demonstrated a lung cancer mortality reduction using a CT strategy incorporating nodule volume and growth. These trials established that CT screening can improve outcomes under defined conditions.
Most nodules found during screening are not cancer. Management systems therefore use size, morphology, growth, and other features to determine whether a finding requires routine screening, short-interval CT, additional imaging, or tissue sampling. Volumetric assessment can help distinguish stable benign nodules from lesions showing concerning growth. Standardized reporting systems reduce inconsistent management, although no classification can eliminate all false positives or missed cancers.
Harms include cumulative radiation exposure, incidental findings, overdiagnosis, and complications of invasive assessment. A biopsy or surgical procedure for a benign nodule can cause substantial harm, particularly in people with impaired respiratory function. Eligibility therefore depends not only on the probability of developing lung cancer but also on whether detection could lead to beneficial treatment. Severe competing illness can limit that benefit even when cancer risk is high.
In 2021, the US Preventive Services Task Force recommended annual low-dose CT for adults aged 50–80 with at least a 20 pack-year smoking history who currently smoke or quit within the previous 15 years. A pack-year represents smoking one pack per day for one year, or an equivalent cumulative amount. Other organizations use different eligibility criteria, including risk models or different treatment of time since quitting. Screening is distinct from evaluating respiratory symptoms, and it does not prevent the many noncancer consequences of tobacco exposure. Tobacco cessation remains a separate intervention with benefits extending well beyond lung cancer. Implementation studies consequently examine both imaging quality and the integration of screening with other relevant services.
Prostate Screening and Other Selective Approaches
Screening for prostate cancer commonly uses prostate-specific antigen, or PSA, measured in blood. PSA is produced by prostate tissue and is not cancer-specific: levels may increase with benign prostatic hyperplasia, inflammation, and other conditions. A higher level changes the probability of cancer but does not establish a diagnosis. Conversely, some prostate cancers occur at relatively low PSA concentrations.
Large randomized studies have produced findings shaped by screening frequency, background PSA testing, biopsy practices, and follow-up duration. The European Randomized Study of Screening for Prostate Cancer demonstrated a reduction in prostate cancer mortality with PSA-based screening, accompanied by additional diagnoses and treatment. The US PLCO trial did not show a mortality reduction, but substantial PSA testing in its comparison group complicated interpretation. These results illustrate why trial context matters as much as the label “screening versus no screening.”
Contemporary pathways may use repeated PSA measurement, risk calculators, additional biomarkers, and prostate MRI before deciding whether biopsy is warranted. MRI can improve the targeting of suspicious lesions and reduce some unnecessary biopsies, but it does not identify every clinically significant cancer. Active surveillance after diagnosis can reduce immediate treatment of selected low-risk cancers, yet it involves continued monitoring and does not remove the initial possibility of overdiagnosis. Population recommendations therefore commonly emphasize informed decisions rather than universal PSA testing.
For several other cancers, routine screening of average-risk populations has not demonstrated a favorable balance of benefits and harms. Large ovarian screening trials using CA-125 and ultrasonography did not show a significant mortality reduction; the final UKCTOCS mortality results were reported in 2021. Screening can lead to surgery for benign abnormalities. Routine thyroid ultrasound screening can detect many small papillary cancers without a demonstrated corresponding mortality benefit, a prominent example of overdiagnosis.
Risk-based surveillance is a different category. People with cirrhosis or selected chronic hepatitis B infections may undergo surveillance for liver cancer, commonly with ultrasound, sometimes combined with alpha-fetoprotein testing. Selected people with strong inherited or familial susceptibility to pancreatic cancer may enter specialized MRI and endoscopic ultrasound surveillance programs. Gastric cancer screening is established in some high-incidence countries, including Japan and South Korea, where background risk and available endoscopic infrastructure differ from those of lower-incidence settings. These examples show why evidence from a high-risk population cannot automatically justify general population screening.
Program Organization, Equity, and Informed Participation
Organized screening programs define an eligible population, issue invitations, specify testing and referral protocols, and monitor outcomes. Opportunistic screening occurs when testing is offered during individual healthcare encounters without the same population-wide structure. Both approaches can detect cancer, but organized programs make it easier to measure coverage, identify missed groups, and maintain consistent quality across the pathway.
Participation alone is an incomplete performance measure. A program must also track inadequate samples, recall rates, positive predictive value, diagnostic completion, time to treatment, complications, interval cancers, and stage distribution. Cancer registries and public health surveillance systems help connect screening activity with subsequent diagnoses and deaths. Evaluation requires long follow-up because an immediate increase in detected cancer can reflect earlier diagnosis or overdiagnosis rather than prevention.
Quality assurance is test-specific. Mammography programs assess image quality and interpretation; cervical programs monitor specimen adequacy and laboratory performance; colonoscopy services monitor bowel preparation, completion, lesion detection, and complications. Lung screening requires consistent CT protocols and nodule management. Information systems must distinguish routine recall from unresolved abnormal results so that an invitation for another screening round does not substitute for needed diagnostic investigation.
Access is shaped by social determinants of health, including income, geography, language, paid leave, transportation, and access to primary care. A nominally free test may still impose costs through travel, preparation, childcare, or lost earnings. Diagnostic follow-up can create larger barriers than the initial screen. Programs that expand testing without addressing these downstream obstacles may increase the number of identified abnormalities without achieving comparable improvements across populations. Health equity therefore concerns completion and outcomes, not only test uptake.
Informed participation requires communication of both benefit and harm in understandable terms. Absolute frequencies with a specified time horizon are generally more interpretable than relative reductions alone. Decision aids may explain false positives, overdiagnosis, treatment consequences, and the option not to participate. Consent is not adequately represented by achieving a participation target.
Economic assessment considers the full pathway, including invitations, tests, follow-up, treatment, complications, and capacity requirements. Outcomes may be expressed as costs per life-year or quality-adjusted life-year gained, but results depend on local cancer incidence, prices, participation, and treatment availability. Data linkage improves evaluation while creating obligations concerning data privacy, security, and appropriate use. Program governance must therefore connect clinical performance, public accountability, and protection of individual information.
Risk Stratification and Emerging Detection Technologies
Risk-stratified screening seeks to allocate testing intensity according to the likelihood of clinically important cancer and the expected benefit of early intervention. Existing programs already use age, smoking history, family history, and known inherited syndromes. More detailed approaches combine clinical factors with breast density, prior test results, environmental exposures, or genetic information. The objective is not merely to identify people likely to receive a cancer diagnosis, but to improve outcomes while limiting unnecessary procedures.
Genetic risk assessment includes rare pathogenic variants with large effects and polygenic risk scores that combine many common variants with individually small effects. These measures are not interchangeable. A score developed in one ancestry group may perform differently in another because of differences in genetic variation, background incidence, and data representation. Genetic counseling can be part of pathways involving inherited cancer susceptibility. Proposed screening schedules based on genetic scores require validation of clinical utility, not just evidence that the scores predict risk.
Blood-based detection is another major research area. Tumors can release circulating tumor DNA, proteins, and other signals into circulation. Investigational multi-cancer early-detection tests may analyze DNA methylation, mutations, fragment characteristics, or combinations of markers to identify a cancer-associated signal and sometimes estimate its tissue of origin. Very small or biologically low-shedding tumors can be difficult to detect, while noncancer processes can produce potentially confounding signals.
These technologies face challenges beyond laboratory discrimination. A positive signal without a localized lesion may lead to repeated imaging, endoscopy, or biopsy. Detecting more early-stage cancers does not itself demonstrate fewer late-stage cancers or deaths. A test covering many cancer types can also obscure major differences in sensitivity by organ, stage, and tumor biology. Establishing benefit requires prospective studies with clearly defined diagnostic pathways and clinically meaningful outcomes, while monitoring overdiagnosis and procedural harm.
Artificial intelligence is being studied and deployed to assist image interpretation, triage, and quality control. Algorithms can improve selected measures of detection or reduce reading workload, but performance may change across equipment, populations, and clinical workflows. Evaluation must consider missed cancers, recalls, downstream procedures, and whether human–algorithm interaction introduces new errors. For both computational tools and molecular assays, regulatory authorization, diagnostic accuracy, and commercial availability are separate questions from demonstrated population benefit. Ongoing research tests whether these approaches can improve the full screening pathway rather than simply increase the number of abnormalities detected.