Can false positives/negatives be reliably detected by redoing the test?

Question

As the title states, I'm wondering whether simply redoing a test is a valid means of eliminating the risk of a given result being false.

The question is specifically about repeating the same exact test, not using a different testing methodology. If relevant, I'm mostly interested in blood tests, more specifically for the diagnosis of brucellosis.

Or, put another way, what are the most common reasons for false results? There's obviously the risk of lab error and cross-contamination, but I assume there are also reasons intrinsic to the patient themselves.

I've read that some tests can generate false positives if the patient is infected with some other virus or bacteria (which may or may not be harmful) [1, p.2]. Is this the only cause or are there other problems which may interfere with the test (possibly including cases where the test doesn't work on this person for reasons unknown)?

External factors would likely be eliminated by repeating the test (if the lab is usually reliable). However, I assume the reasons intrinsic to the patient would not be affected, such that the second test would give the same false result if the "problem" is the patient.

Which raises the further question of whether there are any general statistics on the most likely causes of false results: are they usually due to user error (and therefore identifiable by redoing the test) or due to some underlying condition in the patient (and likely doomed to give the same mistaken result every time)?

Answer 1

Although somewhat buried in the question, it appears you are interested in Brucellosis.

Brucellosis is an infectious disease caused by bacteria of the Brucella genus including B. melitensis, B. abortus, B. suis, and B. canis.

Diagnostic testing for Brucellosis can generally be placed into 4 categories¹:

1. Bacterial culture: the causative bacteria are grown in a specialized medium outside of a host
2. Agglutination test: antibodies from the patient are mixed with an antigen (such as Brucella bacteria) and cause visible clumping
3. ELISA: antibodies themselves are identified (read more)
4. Genetic studies: genetic material from the bacteria are identified in a patient sample

Bacterial Culture

Obviously, getting a bacterium to grow in the lab requires you actually collect the bacterium. Many pathogenic bacteria are surprisingly hard to grow outside a human. The sensitivity of culture for Brucellosis is often less than 50%². If the clinician has not actually collected the bacteria, repeating the test can potentially work.

Bacteremia (bacteria in the blood) can be low level as 1 to 5 bacteria per mL and occur intermittently³. Repeat culture can be particularly useful if testing a different body site, for example cerebral spinal fluid if there are neurologic symptoms.

Bacterial cultures can often grow many different types of organisms and laboratories have standard methods to identify what has grown. In resource rich areas, 16S ribosome sequencing has made this task relatively straightforward.

Agglutination test and ELISA

Both the agglutination type tests and ELISAs rely on detecting patient's antibodies in a sample. Antibody production is complex but generally follows a predictable pattern.

Hypothetical antibody curves adapted from this WHO report.

Depending on when a sample is taken in this curve, antibodies may not have been produced in detectable quantities yet. Repeating a sample might be helpful.

However, the reality is that Brucella are endemic in many parts of the world, and so many people have pre-existing antibodies. It is for this reason that the US Centers for Disease Control recommends routinely taking two samples to observe the increase in antibody levels.

The agglutination tests are also susceptible to the prozone phenomenon⁴ where antibody levels are so high that they coat the bacteria in such a way that the do not cross link and clump. This can usually be overcome with diluting the sample rather than a repeat sample.

Sometimes, a patient's antibodies to another bacterium might also bind to the Brucella in the agglutination test. This would be a false positive and a second sample would be unlikely to change the result. Using an ELISA with a more specific antigen would be one possible alternative approach.

Genetic Testing

Molecular genetic testing for Brucella is not widely used in clinical practice⁵, but detection of genetic material requires it be present in the sample at a reasonable level. The same principles as described in the bacterial culture section apply.

Of note, the 16S ribosome sequence of all Brucella species sequenced so far have been identical⁵, so determination of the particular species is not possible.

Laboratory Error

In addition to all of the above considerations, there are obviously process problems that can occur. For example, a sample could be miss-labeled or inadvertently moved. A critical reagent in the process might also fail. In some of these cases, repeating the test, potentially at a different laboratory, might address the issue.

Conclusions

Thus, overall, for some types of false negative results, repeating a sample could be useful. However, for most false positive results, repeating a test may not be helpful, and so an orthogonal approach should be employed.

An experienced clinician with training in infectious disease practicing in a research rich setting will be familiar with the issues in their particular community and know how to address them.