Will Rogers Phenomenon

Will Rogers phenomenon—known in clinical epidemiology, oncology, and biostatistics as stage migration or zero-time stage shift—is a paradoxical statistical artifact in which the introduction of a more sensitive diagnostic technology (such as positron emission tomography–computed tomography [PET-CT], brain magnetic resonance imaging [MRI], or sentinel lymph node biopsy) reclassifies a subset of patients carrying subclinical, microscopic disease from an early stage (Stage I) into an advanced stage (Stage II, III, or IV), causing the measured stage-specific survival rate to increase in every individual stage—even when no patient lives a single day longer and the overall survival rate of the entire cohort remains unchanged. Discovered and named in 1985 by Alvan R. Feinstein, Daniel M. Sosin, and Carolyn K. Wells at Yale University School of Medicine, the phenomenon takes its name from a humorous quip attributed to American humorist Will Rogers during the Great Depression: "When the Okies left Oklahoma and moved to California, they raised the average intelligence level in both states."

Why Moving the "Worst of the Good" to become the "Best of the Bad" Raises Both Averages

Let a total population of N patients with a newly diagnosed malignancy (such as non-small-cell lung carcinoma) be partitioned into two clinical stages—Early Stage I (localized tumor) and Advanced Stage II–IV (metastatic tumor)—under two different eras of diagnostic staging sensitivity:

• Era 1 (Low-Sensitivity Conventional Chest X-Ray Era): Patients can only be assigned to Stage II–IV if they have overt, macroscopic metastases visible on a plain chest radiograph. Patients who harbor silent, microscopic lymph-node or bone micrometastases invisible on plain X-ray are left inside Stage I.
• Era 2 (High-Sensitivity FDG PET-CT / Brain MRI Era): High-resolution whole-body imaging detects those silent micrometastases, migrating that intermediate subcohort out of Stage I and reclassifying them into Stage II–IV.

Notice the biological prognosis of the migrating subcohort (Group M, patients with microscopic occult metastases) relative to the other two subcohorts:

1. Compared to Truly Localized Stage I Patients (Group L): Group M has already seeded micrometastases, so Group M has a worse survival probability than Group L (S_M < S_L). Thus, when Group M is removed from Stage I, the worst-prognosis patients have been purged from Stage I, causing the average survival of the remaining Stage I cohort to rise!
2. Compared to Overtly Metastatic Stage II–IV Patients (Group A): Group M has only minimal, low-volume microscopic metastases rather than bulky, symptomatic organ failure, so Group M has a much better survival probability than Group A (S_M > S_A). Thus, when Group M is added into Stage II–IV, the best-prognosis new arrivals dilute the mortality of Stage II–IV, causing the average survival of the Stage II–IV cohort to also rise!

Exact Numerical Demonstration (N = 300 Cohort with Zero Change in Total Survival)

Consider a cohort of 300 lung cancer patients composed of three biological strata whose true biological survival is identical in 1980 (Era 1) and 2020 (Era 2):

• Stratum L (n = 100, Truly Localized Disease): 5-year survival = 80% (80/100 survive).
• Stratum M (n = 100, Occult Microscopic Metastases — The Migrating Group): 5-year survival = 40% (40/100 survive).
• Stratum A (n = 100, Overt Macroscopic Metastases): 5-year survival = 10% (10/100 survive).

Now compare the stage-specific 5-year survival rates reported before and after introducing PET-CT staging:

As this algebraic identity proves, a hospital or historical registry comparing Era 2 against Era 1 stratified by TNM stage would falsely conclude that a new chemotherapy protocol introduced in Era 2 had boosted Stage I survival from 60% to 80% and more than doubled Stage II–IV survival from 10% to 25%—even though 130 of 300 patients (43.3%) survived in both eras and not a single life was saved!

In their landmark 1985 New England Journal of Medicine study, Alvan R. Feinstein, Daniel M. Sosin, and Carolyn K. Wells investigated why TNM stage-specific 6-month survival rates for lung cancer patients treated at Yale–New Haven Hospital and the West Haven VA Medical Center in 1977 (n = 131) were substantially higher in every single TNM stage than for patients treated at the same hospitals in 1953–1964 (n = 1,266), despite the fact that no curative breakthroughs in lung cancer chemotherapy had occurred during that 15-year interval:

• Between 1953–1964 and 1977, three new non-invasive imaging modalities had entered routine clinical staging workups: radionuclide liver/spleen scans, radionuclide bone scans, and cranial/thoracic computed tomography (CT) scans.
• In the 1953–1964 cohort, asymptomatic metastases in the liver, bone, or brain went undetected unless the patient had clinical symptoms (such as bone pain, jaundice, or seizures).
• When Feinstein and colleagues reclassified the 1977 cohort according to "Old Era" symptomatic criteria (ignoring all metastases that had been discovered by silent radionuclide or CT imaging in asymptomatic patients), the apparent stage-by-stage survival improvements between 1953–1964 and 1977 vanished.

Since 1985, the Will Rogers phenomenon has been documented across virtually every domain of clinical medicine where diagnostic thresholds, histopathological grading, or imaging resolution have evolved (operating alongside lead-time bias, spectrum bias, and verification bias):

1. Sentinel Lymph Node Biopsy and Immunohistochemistry in Breast Cancer and Melanoma: Replacing standard hematoxylin-and-eosin (H&E) axillary lymph-node sampling with serial-section sentinel lymph node biopsy (SLNB) and anti-cytokeratin immunohistochemistry identifies isolated tumor cells and micrometastases (0.2–2.0 mm, staged as pN1mi), migrating patients out of node-negative Stage I into node-positive Stage II/III and improving apparent survival in both node-negative and node-positive breast cancer.
2. Gleason Grade Migration in Prostate Cancer (Albertsen et al., 2005): Between the 1990s and 2000s, surgical pathologists gradually abandoned assigning Gleason scores of 2–5 on needle biopsies and upgraded borderline architectural patterns into Gleason 6 or Gleason 7 (codified in the 2005 ISUP Modified Gleason System). Peter C. Albertsen and colleagues (2005) re-reviewed archival prostate biopsy slides from 1,858 men diagnosed in 1990–1992 and showed that contemporary pathologists upgraded 55% of the historic slides by a mean of +0.85 Gleason points—resulting in an apparent 28% reduction in Gleason-stratified prostate cancer mortality due purely to pathological grade migration.
3. The McDonald MRI Criteria in Multiple Sclerosis (Sormani et al., 2008): Revisions to the McDonald diagnostic criteria (2001–2017) allowed patients presenting with a first Clinically Isolated Syndrome (CIS) to be diagnosed immediately with Relapsing-Remitting Multiple Sclerosis (RRMS) if baseline brain/spinal cord MRI showed dissemination in space and time (or cerebrospinal fluid oligoclonal bands). Migrating milder, early CIS cases into the RRMS cohort simultaneously made the remaining CIS cohort appear to have an even lower conversion rate and made the RRMS cohort appear to have a much more benign disability progression course compared with historical Poser-criteria RRMS cohorts.

To prevent the Will Rogers phenomenon from corrupting clinical trials, cancer registries, and comparative effectiveness studies, public health and oncology methodology enforces four rules:

• Never Use Historical Controls for Stage-Stratified Survival Comparisons: Because imaging resolution (1.5T → 3T → 7T MRI, PSMA PET-CT, circulating tumor DNA [ctDNA] liquid biopsy) and TNM staging editions (AJCC 1st through 9th Editions) continuously drift upward in sensitivity over time, comparing a contemporary Stage III single-arm phase II trial against historical Stage III survival benchmarks is invalid; concurrent randomized controlled trials where both arms undergo identical baseline staging workups eliminate stage migration between treatment arms.
• Always Report Unstratified All-Stages Cohort Survival and Population Mortality Rates: True therapeutic progress must be verified by an improvement in the unstratified survival of the entire diagnosed cohort (Stages I–IV combined) and a decline in the age-standardized population disease-specific mortality rate (and years of potential life lost).
• Stage-by-Symptom and TNM Edition Cross-Walk Subclassification: As pioneered by Feinstein's "clinical-symptom staging" and modern AJCC sub-staging (such as separating pN0(i+) and pN1mi micrometastases from macroscopic pN1a nodal disease), registries must record the exact modality by which metastases were detected so historical trends can be standardized to a uniform diagnostic detection threshold.

1. Feinstein, A. R., Sosin, D. M., & Wells, C. K. (1985). "The Will Rogers phenomenon: stage migration and new diagnostic techniques as a source of misleading statistics for survival in cancer." New England Journal of Medicine, 312(25), 1604–1608. https://doi.org/10.1056/NEJM198506203122504
2. Albertsen, P. C., Hanley, J. A., Barrows, G. H., Penson, D. F., Kowalczyk, P. D. H., Sanders, M. M., & Fine, J. (2005). "Prostate cancer and the Will Rogers phenomenon." Journal of the National Cancer Institute, 97(17), 1248–1253. https://doi.org/10.1093/jnci/dji248
3. Sormani, M. P., Tintorè, M., Rovaris, M., Rovira, A., Vidal, X., Bruzzi, P., Filippi, M., & Montalban, X. (2008). "Will Rogers phenomenon in multiple sclerosis." Annals of Neurology, 64(4), 428–433. https://doi.org/10.1002/ana.21464
4. Chee, K. G., Nguyen, D. V., Brown, M., Gandara, D. R., Wun, T., & Lara, P. N., Jr. (2008). "Positron emission tomography and improved survival in patients with lung cancer: the Will Rogers phenomenon revisited." Archives of Internal Medicine, 168(14), 1541–1549. https://doi.org/10.1001/archinte.168.14.1541