Tests are done to screen for disease, diagnose disease, classify and measure the severity or stage of disease, and monitor the course of a disease, especially its response to treatment.

Screening Tests: Screening tests are used to try to detect a disease when there is no apparent evidence that a person has the disease. For example, most doctors recommend that women over 40 have a mammogram every year or every 2 years to look for breast cancer even if they have no breast lumps. Screening is based on the natural idea that outcome will be better if a disease is recognized and treated in its early stages. Although logical enough, this idea is not always correct. For some diseases, such as testicular cancer and ovarian cancer, there does not seem to be any difference in outcome between people whose disease is detected by screening and those whose disease is diagnosed after the first symptoms appear.

A further potential problem with screening tests is that the results usually require confirmation by a more definitive test. For example, women who have a mammogram with abnormal results need to have a breast biopsy. Such definitive tests are often invasive, uncomfortable, and sometimes a bit dangerous. For example, a lung biopsy can cause a collapsed lung. Because results of screening tests are sometimes abnormal in people without disease (which is common because no test is 100% accurate), some people undergo an unnecessary test that might harm them.

Clinical trials are necessary to tell which screening tests are effective and which people should undergo them. Despite these concerns, it is clear that for some diseases such as high blood pressure and cervical cancer, screening saves lives. To be useful, tests used for screening must be accurate, be relatively inexpensive, pose little risk, cause little or no discomfort, and improve outcomes.

Diagnostic, Classification, and Monitoring Tests: Diagnostic tests confirm or rule out a disease when a doctor suspects that a person has the disease. For example, a doctor who suspects serious heart disease might recommend cardiac catheterization. This test is a good diagnostic test but not a good screening test because it is expensive, can have serious side effects, and is uncomfortable. However, these drawbacks are outweighed by the need for this test when the presence or absence of disease must be confirmed.

Some tests are used to classify and measure the severity of a disease that has already been diagnosed. Results may lead to more specific and effective choices for treatment. For example, after a diagnosis of breast cancer is confirmed, additional tests are done to determine if and where the cancer has spread.

Tests are also used to monitor the course of a disease over time, often to determine the response to treatment. For example, blood tests are done periodically in people who take thyroid hormone to treat hypothyroidism to determine whether the hormone dose best meets their needs. A decision about how often such testing is needed is made based on the person's situation.

Doing Tests and Interpreting Results: When deciding whether to recommend testing for a disease, especially to make a diagnosis, doctors estimate how likely it is that a person has the disease (the pre-test probability of disease). In coming up with an estimate for a specific person, doctors may consider the following:

• Information about the disease in their practice area, including how common the disease is (prevalence) and how many new cases of the disease occur during a specific period of time (incidence)
• A person's particular characteristics (risk factors, such as family history of the disease) that increase or decrease the likelihood of being affected

With this information, doctors can then select the best test to screen for or confirm the presence of the disease.

Doctors must also be prepared to determine what test results may mean. Unfortunately, tests are not perfect. Results are sometimes normal in people who have the disease being tested for. That is, tests have false-negative results. Results are sometimes abnormal in people who do not have the disease being tested for. That is, tests have false-positive results. Therefore, important characteristics of a test are its sensitivity (the likelihood that results will be abnormal in people with the disease being tested for) and its specificity (the likelihood that results will be normal in people without the disease). Doctors can mathematically combine the pre-test probability of disease with the results of the test and with information about the test's sensitivity and specificity to more accurately estimate the likelihood that the person has the disease (post-test probability).

Does Everyone Need a Test?
In short, no. Although many people find medical tests reassuring, test results are not always right. Doctors must consider what they know about the disease they are testing for, what they know about the test itself, and what they know about the person. For example, parents are concerned that their 4-year-old daughter might have a urinary tract infection (UTI) because she is walking holding her thighs together. In the office, the doctor discovers that the girl is not urinating more frequently or complaining of pain with urination. Her physical examination is normal. Based on these findings, the doctor thinks the likelihood of a UTI is very low, about 5%, and reassures the parents that nothing needs to be done unless other symptoms develop. The parents say they would feel better if the doctor did a test. Would a test help the doctor? In this case, even a fairly accurate test may give confusing results. Suppose the doctor did a test that had a sensitivity of 90% and specificity of 90% for UTI (typical figures for many medical tests). Sensitivity of 90% means that done in 100 people with a UTI, the test will correctly be positive in 90. Specificity of 90% means that done in 100 people without a UTI, the test will correctly be negative in 90 but give a false-positive result in 10. This test seems very accurate (and is). But would it help this little girl? If the test result is positive, does that mean the child actually has a UTI? Suppose the doctor tests 1,000 children who each have a 5% chance of having a UTI. The test's sensitivity and specificity (plus some simple arithmetic) tell the doctor that there will be 45 children who have true-positive test results, but 95 children who have false-positive results. That means that a positive result in this particular child is twice as likely to be false as it is to be true. Thus, even a positive test result should not change the doctor's decision not to treat. Because the doctor would not do anything different, it makes no sense to do the test in the first place. It would be a different story if the doctor thought the likelihood of a UTI was 50-50. Because only 10% of test results (positive or negative) would be false, testing would be helpful.

Another characteristic of a test is its reliability. A highly reliable test gives the same result when a person undergoes the test more than once, unless the disease being tested for has actually lessened or worsened. Results from a less reliable test may change randomly.

Before doing a test, doctors weigh the potential harm of the test against the potential benefit of the information the test may provide. Doctors must also consider how the results will be used. It may not be useful to do a test if the results will not change the recommended treatment. For example, if a test is being considered to determine whether a particular treatment is an option for a person, but the person has already decided not to undergo that treatment, the test need not be done.

Last full review/revision June 2007 by Thomas V. Jones, MD, MPH