There are three kinds of risk used in medicine for assessing patients.  Each is looking at the data in a slightly different way.  Lifetime risk and Absolute Risk are much more useful than Relative Risk.

Cancer comes in different varieties and screening has variable benefits.  The non-progressive or very slow group has little benefit for the patient being diagnosed early.  The fast group we cannot diagnose early.  Only the slow group is a useful group to screen for early diagnosis.

Are we really diagnosising the cancers that are dangerous and count?  The left graph would say yes and the right graph is no.  Refer  to these graphs when comparing screening for the various cancers listed below.

Changing what we define as normal vs disease has significant consequence.  Many more people who previously classified healthy now have "mild" disease.  Intervention in "mild" disease has much less benefit.

High blood pressure is a good example of different benefits for disease severity.  The top table shows severe hypertension benefits a large number of patients.  The lower table emphasizes little difference in benefit for treatment of very mild hypertension vs no treatment.  Bottom line - severe hypertension - you need treatment, mild to very mild hypertension - treatment may be worse than the disease.

Near normal osteoporosis treatment benefits only a small percent while potentially damaging many more people.  Drug therapy has potential severe side effects.

Treatment of near normal cholesterol benefits only a few while overtreating many.  The financial aspects of overtreatment are staggering.

Should men be screened for abdominal aortic aneurysm?  A lot of overdiagnosis for a benefit to a very few.  

More sophisticated screening methods detect more milder disease with less likelihood of benefit from the screening.  Screening should detect  those at high risk without many false positives.

Prostate cancer was aggressively screened for using PSA in the early '90s.  The death rate stayed the same.  We are finding and treating much less aggressive cancers which are not killing people = overdiagnosis. 

The diagnoses of breast cancer are going up over time due in part to increased screening ie mammography.  There is a slight dip in cancer deaths which implies screening predominantly finds non-killing cancers.

The table on the left demonstrates most women do not benefit from mammography though a few do and it varies with age of the patient.

The decision of an individual patient to be screened or not screened by mammography is not the same for all women.  It should be an informed decision based on all the evidence.  The summary on the left is a good one.

Thyroid cancer diagnoses have been going up over time.  Deaths are staying the same.  Are we doing people a favor by diagnosing non-killing cancers?

Same problem with kidney cancer.

Cervical cancer itself does not seem to be overdiagosed but it's "precursor" lesions are greatly overdiagosed and overtreated.  It is a shotgun approach of overtreating many to benefit a very few.

There appears to be no overdiagnosis or overtreatment of colon cancer.  Colon polyps are another matter entirely.

Screening for lung cancer has not been successful.  No methodology has really worked.

Classic graph of overdiagnosis.  Melanoma deaths are stable despite screening and increased diagnoses.

This is an extremely important concept.  Eg Bulging discs seen by MRI are used to diagnose disc disease and probably an operation.  But 50% of people with no back pain have bulging discs.  Go figure.

Incidentalomas are things found unrelated to why you went to the doctor.  It is a significant driver behind rising healthcare costs and most physician just hate them.  You can't ignore them but most turn out not to be anything.