Understanding Positive and Negative Predictive Values
And why this is so important in clinical practice!
Case: You are seeing a 50-year-old male patient who recently immigrated from Brazil presenting with a 3-month history of post-prandial abdominal pain, nausea, and loss of appetite. H. Pylori infection is on your differential – what test should you order?
As clinicians we order tests, and when ordering tests we need to understand diagnostic accuracy. These are tough concepts to comprehend and integrate into practice, but it’s important for us to have basic understanding of the theory.
Earlier we reviewed sensitivity and specificity. What’s the difference between those test characteristics compared with negative predictive value (NPV) and positive predictive value (PPV)?
Simply put, sensitivity and specificity look solely at the test characteristics (ability to classify patient has having or not having the disease), whereas NPV and PPV apply those characteristics to the patient in front of you to answer the following:
“The test is positive, what is the probability the patient has the disease?” (PPV)
or “The test is negative, what is the probability the patient is disease free? (NPV)
Hang in there, I promise this will make sense.
Most disease states have a “gold standard” or “reference standard” test – i.e. a test that correctly identifies diseased patients as positive and non-diseased patients as negative (1). We can’t always order gold standard tests due to cost, availability, invasiveness, resource-consumption, etc. (1). In practice, the tests we order can produce false negative and false positive results.
The NPV states: “If the test returns negative, what is the probability that the patient doesn’t have a disease? In other words, what is the probability this is a true negative test (opposed to a false negative)?
The PPV is the opposite: “If the test returns positive, what is the probability that the patient actually has the disease? In other words, what is the probability this is a true positive test (opposed to a false positive)?
So why do we care about this in practice? We do not know whether the patient has the disease or not when we order a diagnostic test (that is why we order tests!) Instead, we interpret the result of our test – ether positive or negative. NPV and PPV consider the prevalence of the disease in our population, which helps us determine if our result is a true positive or true negative based on the patient in front of us.
What is prevalence? the proportion of people in a population with a disease/condition at a specific point in time(2).
Back to the case: One of the gold standard tests for identifying active H. pylori infection is the urea breath test (3). Alternatives are the stool antigen test (high specificity, poor sensitivity), and H. Pylori serology (it cannot discriminate between acute infection or old/treated infection) (3). Due to cost and availability limitations, you order H. Pylori serology as you have no prior documented infection and your pre-test probability for active H. Pylori infection is medium to high. The test comes back negative.
Negative predictive value: Remember, NPV is the probability the patient is disease free if they test negative. When you review his negative test result, you think to yourself: how likely is it that he really does NOT have H. pylori infection? How reassured should I be? This is the negative predictive value.
Positive predictive value: Remember, PPV is the probability a patient has the disease if they test positive. In other words, if the test result was positive, you may think of the following: how likely is it that he has an active H. pylori infection? This is the positive predictive value.
One important point is that the sensitivity and specificity of tests are applicable in every population, whereas the PPV and NPV are dependent on the prevalence of the disease in a given population. Thus, the NPV and PPV are derived from research studies using a specific population, therefore are not always directly applicable to your setting, unless you think (or know) the prevalence of that disease in your practice area is similar.
Key Take Home Point
PPV and NPV use the sensitivity and specificity of a test combined with disease prevalence to tell you the following:
NPV: if a test is negative, what is the probability this is a true negative?
PPV: if a test is positive, what is the probability this is a true positive?
*Scroll down below if you want to review an illustrative example.
References/Readings
Monaghan TF, Rahman SN, Agudelo CW, Wein AJ, Lazar JM, Everaert K, et al. Foundational Statistical Principles in Medical Research: Sensitivity, Specificity, Positive Predictive Value, and Negative Predictive Value. Medicina (Kaunas, Lithuania). 2021;57(5):503.
Tenny S, Hoffman MR. Prevalence. [Updated 2023 May 22]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2023 Jan-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK430867/
Patel SK, Pratap CB, Jain AK, Gulati AK, Nath G. Diagnosis of Helicobacter pylori : What should be the gold standard? World journal of gastroenterology : WJG. 2014;20(36):12847-59
A helpful 6-minute video: https://www.youtube.com/watch?v=psELBu7muNY
An Illustrative Example of PPV and NPV
The statistics used in this example are illustrative only (not derived from a specific study).
