Overdiagnosis

Overdiagnosis in epidemiology and public health is the detection of a histologically or biochemically confirmed "true-positive" disease, neoplasm, or physiological abnormality—typically in an asymptomatic person undergoing population screening, incidental cross-sectional imaging, or expanded disease-threshold testing—that would never have caused clinical symptoms, morbidity, or premature mortality during that individual's remaining lifetime. Systematically analyzed by H. Gilbert Welch, William C. Black, Alexandra Barratt, and Paul Glasziou, overdiagnosis is frequently confused with a false-positive screening result, yet the two phenomena are fundamentally distinct: a false-positive test occurs when an abnormal screening test (such as an elevated serum biomarker or suspicious mammogram) is refuted by a negative gold-standard biopsy, whereas an overdiagnosed lesion fulfills all gold-standard pathological criteria for real disease (such as a biopsy-proven papillary thyroid carcinoma or Gleason score 6 prostate adenocarcinoma) whose biological growth trajectory is so indolent relative to competing mortality risks that the patient would have died of an unrelated cause without ever knowing the lesion existed. Because a pathologist examining an excised tumor under a microscope cannot distinguish an overdiagnosed lesion from a progressive life-threatening lesion at the time of diagnosis, nearly all overdiagnosed patients undergo overtreatment (surgery, radiation, chemotherapy, or lifelong pharmacotherapy) that cannot improve survival and can only inflict physical, psychological, and financial harm.

As formalized by H. Gilbert Welch and William C. Black (2010) in the Journal of the National Cancer Institute, overdiagnosis requires the intersection of two prerequisites: (1) a vast subclinical reservoir of occult disease present in the general population, and (2) diagnostic screening or clinical workup activities (often amplified by diagnostic suspicion bias) that tap into that reservoir.

Even when a lesion is histologically malignant under a light microscope, it fails to cause clinical disease or death before the patient dies of competing causes (such as cardiovascular disease or stroke at age 82) under one of three biological trajectories:

1. Non-Progressive / Dormant Lesions: The cellular clone carries an oncogenic driver mutation (such as BRAF V600E in papillary thyroid microcarcinoma) but arrests proliferatively due to oncogene-induced cellular senescence, lack of angiogenic switch signaling, or intact host immune surveillance.
2. Spontaneously Regressing Lesions: The neoplasm undergoes spontaneous apoptosis, differentiation, or immunological clearance (well documented in infant stage 4S neuroblastoma, low-grade cervical intraepithelial neoplasia, and a subset of screen-detected ductal carcinomas in situ and small invasive breast tumors).
3. Very Slow-Growing (Indolent) Lesions Competing With Non-Disease Mortality: The tumor grows continuously at a nonzero rate, but its volume-doubling time is so long (e.g., >800 to 1,500 days) that competing mortality terminates the patient's life before the tumor crosses the clinical symptom horizon—representing the extreme limit of length-time bias and lead-time bias.

How do epidemiologists know how large the hidden reservoir of subclinical cancer is among healthy people? By examining systematic serial-section autopsy studies of adults who died of trauma, accidents, or acute myocardial infarction without any clinical history or suspicion of cancer during life:

• Prostate Cancer Autopsy Reservoir: Among men who died of unrelated causes without a clinical diagnosis of prostate cancer, systematic step-sectioning of the prostate gland discovers histologically confirmed prostate adenocarcinoma in 30% to 40% of men in their 60s and 50% to 70% of men in their 70s and 80s—compared with a lifetime risk of dying from prostate cancer of approximately 2.5% to 3.0%.
• Thyroid Cancer Autopsy Reservoir (Harach et al., 1985): When Finnish pathologists sliced the thyroid glands of consecutive autopsy patients at 2.5-millimeter intervals, they found histologically confirmed occult papillary thyroid carcinoma in 35.6% of adults—suggesting that nearly every adult human harbors microscopic papillary thyroid carcinomas if the gland is sliced thinly enough.
• Breast Cancer / DCIS Autopsy Reservoir: Systematic breast autopsy series detect occult invasive breast cancer in ~1.3% and ductal carcinoma in situ (DCIS) in 8.9% to 15.0% of women aged 40–70 who died of non-cancer causes.

Because individual biopsy specimens cannot be labeled "overdiagnosed" versus "progressive," overdiagnosis is proved at the population level by two epidemiological signatures:

1. Ecological Divergence Between Soaring Incidence and Flat Mortality

When a true biological epidemic of lethal cancer occurs (such as lung cancer in 20th-century cigarette smokers), age-standardized incidence and disease-specific mortality rise together in parallel. Conversely, when a screening program produces widespread overdiagnosis, age-standardized incidence of early-stage disease surges upward, while late-stage incidence fails to fall commensurately and disease-specific mortality remains completely flat:

• The South Korean Thyroid Cancer Screening Epidemic (Ahn, Kim, & Welch, 2014): Following the introduction of widespread high-frequency neck ultrasonography into routine adult health checkups in South Korea after 1999, the national incidence of thyroid cancer rose 15-fold between 1993 and 2011 (from 6.4 to >90 per 100,000 women), making thyroid cancer the most commonly diagnosed cancer in the country, while thyroid cancer mortality remained completely unchanged at ~0.6 per 100,000. Over 99% of the newly detected tumors were subclinical papillary microcarcinomas (<1–2 cm), resulting in hundreds of thousands of total thyroidectomies, lifelong levothyroxine dependence, recurrent laryngeal nerve palsies (1%–2%), and permanent hypoparathyroidism (2%–5%).
• Neuroblastoma Urine Catecholamine Screening in Infants (Woods et al., 2002; Schilling et al., 2002): Prospective population screening of infants in Quebec, Germany, and Japan for urinary vanillylmandelic acid (VMA) and homovanillic acid (HVA) doubled the diagnosed incidence of neuroblastoma without reducing late-stage disease or neuroblastoma mortality at all—proving that screening had simply uncovered an occult reservoir of neuroblastomas destined for spontaneous regression.

2. Persistent Excess Cumulative Incidence in Long-Term Randomized Screening Trials

The gold-standard method to quantify the exact percentage of overdiagnosis (*OD*%) is long-term post-trial follow-up of a randomized controlled screening trial (RCT) in which the control arm is never screened at the end of the trial (a "stop-screen" design). During the active screening phase, the screened arm always shows higher cumulative cancer incidence than the unscreened control arm because screening advances diagnosis in time (lead-time bias). Once screening stops, however, if every screen-detected cancer would eventually have presented clinically, the control arm must "catch up" over the duration of the longest preclinical sojourn time, equalizing cumulative incidence between the two arms. Any permanent excess of cumulative cases in the screened arm (I_screened − I_control) after sufficient post-screening follow-up is pure overdiagnosis:

• ERSPC Prostate-Specific Antigen (PSA) Trial & CAP Trial: In the European Randomized Study of Screening for Prostate Cancer (Schröder et al., 2009, 2014), preventing 1 prostate cancer death over 13 years required screening 781 men and diagnosing 27 additional men with prostate cancer—indicating that ~40% to 50% of PSA-detected prostate cancers represented overdiagnosis.
• Pulmonary CT Angiography and Subsegmental Pulmonary Embolism (Wiener, Schwartz, & Woloshin, 2011): Replacing ventilation–perfusion (V/Q) scans with multi-detector computed tomographic pulmonary angiography (CTPA) after 1998 caused an 80% surge in US pulmonary embolism diagnoses and anticoagulation hemorrhage complications alongside near-zero change in pulmonary embolism mortality and a 36% drop in apparent case fatality (an artifact of the Will Rogers phenomenon as tiny subsegmental clots were added to the case denominator).

3. Threshold Lowering in Chronic Disease Definitions

Overdiagnosis also arises outside oncology whenever consensus committees lower numerical diagnostic cutoffs for physiological continuous variables—such as lowering fasting plasma glucose thresholds for prediabetes, T-score thresholds for osteopenia, glomerular filtration rate cutoffs in elderly adults for stage 3a chronic kidney disease, or blood pressure cutoffs—expanding disease labels to tens of millions of low-risk individuals for whom the absolute risk reduction is minuscule and the number needed to treat exceeds the number needed to harm.

How does overdiagnosis differ from a false-positive screening test?

A false-positive occurs when an initial screening test is abnormal (e.g., a mammogram BI-RADS 4 shadow or a PSA of 5.2 ng/mL), but the subsequent diagnostic biopsy reveals no cancer; the patient is told within days or weeks that they are disease-free. Overdiagnosis, by contrast, occurs when the biopsy confirms a histologically genuine cancer (a true positive under the microscope), but that specific tumor is biologically non-progressive or grows so slowly that it would never have caused symptoms or death before the patient died of old age. Because overdiagnosed patients are told they have real cancer, they receive real surgery, radiation, or hormone therapy (overtreatment).

Why does overdiagnosis artificially inflate 5-year cancer survival statistics?

Five-year relative survival measures the proportion of diagnosed patients alive 5 years after diagnosis: Survival_5-yr = (True Progressive Survivors + Overdiagnosed Patients) / (All Progressive Patients + Overdiagnosed Patients). Because overdiagnosed patients carry non-lethal lesions with ~100% 5-year disease-specific survival, adding thousands of overdiagnosed cases to both the numerator and the denominator of the survival fraction drives 5-year survival toward 95%–99% even if exact same number of people die of progressive cancer every year! This creates a self-reinforcing cycle: the more a screening test overdiagnoses harmless lesions, the higher 5-year survival climbs, convincing the public and clinicians that screening is "saving lives" when age-adjusted population mortality has not moved at all.

How have clinical guidelines evolved to reduce overtreatment of overdiagnosed cancers?

To uncouple detection from overtreatment, oncology guidelines have implemented three reforms: (1) Active Surveillance instead of immediate radical prostatectomy or thyroidectomy for low-risk Gleason Grade Group 1 (3+3=6) prostate cancer and <1 cm intrathyroidal papillary microcarcinomas; (2) Nomenclature De-escalation, reclassifying indolent epithelial lesions to remove the word "carcinoma" (such as reclassifying encapsulated follicular variant of papillary thyroid carcinoma as NIFTP [noninvasive follicular thyroid neoplasm with papillary-like nuclear features]); and (3) Raising Screening Thresholds and Stopping Ages, discontinuing routine screening in frail adults with <10 years of life expectancy.

1. Welch, H. G., & Black, W. C. (2010). "Overdiagnosis in cancer." Journal of the National Cancer Institute, 102(9), 605–613. https://doi.org/10.1093/jnci/djq099
2. Ahn, H. S., Kim, H. J., & Welch, H. G. (2014). "Korea's thyroid-cancer 'epidemic'—screening and overdiagnosis." New England Journal of Medicine, 371(19), 1765–1767. https://doi.org/10.1056/NEJMp1409841
3. Wiener, R. S., Schwartz, L. M., & Woloshin, S. (2011). "Time trends in pulmonary embolism in the United States: evidence of overdiagnosis." Archives of Internal Medicine, 171(9), 831–837. https://doi.org/10.1001/archinternmed.2011.178
4. Woods, W. G., Gao, R. N., Shuster, J. J., Robison, L. L., Bernstein, M., Weitzman, S., Bunin, G., Levy, I., Brossard, J., Dougherty, G., Tuchman, M., & Lemieux, B. (2002). "Screening of infants and mortality due to neuroblastoma." New England Journal of Medicine, 346(14), 1041–1046. https://doi.org/10.1056/NEJMoa012387