Table of contents
Part 3 - Practice: Improving Health
Chapter 9 Screening
Issues in deciding whether or not to screen
The difference between screening and diagnostic testing can be defined merely in terms of differences in the risk of the disease for which a patient is being screened or tested. In screening, the patient being screening is asymptomatic and at apparently low risk of the disease being screened for. In diagnosis, the patient has symptoms or signs indicating a high likelihood of the disease being tested for. Due to the similarity of the two situations, some of the issues mentioned below for screening are also relevant to diagnostic testing.
Identifying who will develop clinical disease
Few indicators of pre-clinical disease are dichotomousthat is, few differentiate sharply between who will develop the condition and who will not. Blood sugar, cholesterol, and blood pressure are all continuous scales. Based on the test sensitivity and specificity (sometimes combined using the Receiver Operating Characteristic (ROC) curve) experts decide on a point on the scale that distinguishes those who are at high risk of developing the clinical condition from those who are at low risk. Even in cancer screening, although some findings on imaging or cytology can be categorized definitely as neoplastic or as benign, other images or specimens show intermediate changes for which even experienced readers may have differing interpretations. Furthermore, some people develop disease even though the screening results put them in the unlikely group (false negatives), and some people, apparently at high risk, never develop disease (false positives). If the cut-off point is set too far towards the high risk end of the scale, too many patients with early stage disease will be missed, so the sensitivity and the negative predictive value will be low. If the point is too far towards the low risk end of the scale, the costs of investigating and reassuring large numbers of low-risk patients will outweigh the benefits of treating patients early who otherwise would have developed a clinically significant condition, and the specificity and positive predictive value will be low.
Figure 9.1: Screening test performance
No test is 100% specific and 100% sensitive, so there will always be false negatives and false positives. Furthermore, the prevalence of the illness under study affects the positive predictive value of the test. Apparently healthy populations tend to have a low prevalence of the disease being screened for; therefore, the positive predictive value of screening tests tends to be less than that of diagnostic tests.
Because it is often hard to detect disease in its early stages, only tests with excellent performance characteristics (sensitivity and specificity) can be used for screening. In order for mammography to be acceptable as a screening test, technicians and radiologists, many of whom were experienced in diagnostic mammography, needed further training to ensure that the performance characteristics of their screening matched those in breast cancer screening trials.
Because it detects disease at a pre-symptomatic stage, that is before it causes perceivable damage or disability, screening may identify disease that would never harm the patient. While Chapter 1 presented the idea of natural historynatural historyapplied to a disease, the characteristic evolution of the symptoms and signs as a disease runs its course and is not treated. as though disease progresses steadily, this is not always the case. Disease may regress because of the patients natural defences. For example, about two thirds of cases of mild cervical dysplasia regress to normal, whereas progression from mild to severe dysplasia or worse occurs at a rate of about 1% per year.2 Slow progression may mean that the patient dies from other causes before succumbing to the cervical cancer detected by the screening. There is, as yet, no method to predict what will happen in individual cases. The old adage that "more men die with prostate cancer than die of prostate cancer" is true. The prostate specific antigen (PSA) test is only about 70% sensitive but is still useful in detecting early cases of prostate cancer.3 However, autopsy series find that approximately 30%40% of men over the age of fifty have prostate cancera proportion that rises with ageyet only about 3% of men die from it. It is not yet certain what should be done once early prostate cancer has been detected. Similarly, studies of breast cancer screening have shown that up to 30% more breast cancers are diagnosed in the screened group than in the unscreened group, probably due to the detection of cancers that would never otherwise have become apparent.4
What the PSA developer thinks
To read the opinion of Prof Richard J. Ablin, who developed the PSA test, on how the test is used, go to: http://www.nytimes.com/2010/03/10/opinion/10Ablin.html
Because of progress in imaging and screening techniques, cancers can now be diagnosed at an extremely early stage. The natural history of these very early lesions is poorly understood. In particular, it can be difficult to distinguish between inflammation and early neoplasia. Those responsible for diagnosispathologists and radiologistsprefer to err on the side of caution and so may overdiagnose borderline cases.