2023.02.10 | Questions 27-28
Biostats (2/3) - Pos and neg predictive values
Hello,
This is part 2 of 3 of the biostats series. In the explanations that follow, we review the 5 key performance characteristics common to all screening and diagnostic tests used in clinical medicine. This is an important set of concepts to understand, and I encourage you to read the explanations below and look through the additional resources at the end of this email.
As always, please feel free to share this post with anyone who might find it useful. You can reach me at studyraregenetics@gmail.com with any suggestions or feedback.
Have a great weekend!
-Daniel
Was this email forwarded to you? If so, subscribe for weekly genetics board prep questions!
Questions
Question 27
A 31-year-old woman at 9 weeks gestation presents for her first prenatal visit. She undergoes noninvasive prenatal testing (NIPT), which flags as “high risk” (i.e. positive) for Angelman syndrome. Natera’s test performance characteristics for Angelman syndrome are as follows: sensitivity 95%, specificity >99%, positive predictive value 10%, and negative predictive value >99.99%. Given the positive test result and knowledge of the test’s performance characteristics, what the probability of this fetus actually having Angelman syndrome?
Question 28
Which of the following would lower the positive predictive value of the NIPT test mentioned in question 27?
Explanation
Ans to Q27: 10%
Ans to Q28: Low prevalence of disease
Noninvasive prenatal testing (NIPT) is a screening method that uses maternal blood to detect chromosomal abnormalities in fetuses. Despite its widespread use, NIPT has been criticized in the media for the high number of false positive results that can cause unnecessary anxiety. For rare conditions such as Angelman syndrome, the Positive Predictive Value (PPV) can be low. This is because the PPV decreases as the prevalence of disease decreases (Question 28). To get a sense for how rare Angelman is, there are 15 babies with Down syndrome for every 1 baby with Angelman (1 in 1,000 births vs. 1 in 15,000 births, respectively)!
The PPV measures the probability that a “high risk” (i.e. positive) result accurately reflects a fetal abnormality. In the case of NIPT testing for Angelman syndrome, a “high risk” result has only a 10% chance of being associated with an actual case of Angelman syndrome (Question 27). This means that 90% of positive results will be false positives.
There are 5 key performance characteristics (including PPV) in all screening and diagnostic tests used in medicine that will be described below. The relationship between these 5 characteristics can be derived using the 4 values (TP, TN, FP, FN) listed in the cells of the confusion matrix shown below:
Sensitivity asks “Given that the patient actually has the target condition, what is the probability of a positive test result?” Put another way, it is the ability of the test to accurately identify individuals with the target condition. Sensitivity is expressed as the proportion of true positive cases divided by the total number of cases with the target condition (Sensitivity = TP / (TP + FN) ). Another term for sensitivity is recall.
Specificity asks “Given that the patient actually does not have the target condition, what is the probability of a negative test result?” Put another way, it is the ability of the test to accurately identify individuals without the target condition. Specificity is expressed as the proportion of true negative cases divided by the total number of cases without the target condition (Specificity = TN / (TN + FP). Note that, in contrast to PPV and NPV, both specificity and sensitivity are independent of disease prevalence.
Positive Predictive Value (PPV) asks “Given that the patient has a positive test result, what is the probability that they actually have the target condition?” In other words, how accurate is a positive test result? Another term for the PPV is precision. The PPV decreases as the prevalence of disease decreases (think “PPreValence” & “PPrecision”). Numerically, we can calculate PPV by dividing the total number of true positives (TP) by all positive predictions (TP plus false positives (FP)): PPV = TP / (TP + FP)
Negative Predictive Value (NPV) asks “Given that the patient has a negative test result, what is the probability that they actually do not have the target condition?” In other words, how accurate is a negative test result? The NPV increases as the prevalence of disease decreases. Numerically, we can calculate NPV by dividing the total number of true negatives (TN) by all negative predictions (TN plus false negatives (FN)): NPV = TN / (TN + FN)
Accuracy is the overall ability of the test to correctly identify individuals with or without the target condition. The accuracy is calculated as (TP + TN) / (total population), where the total population is TP + TN + FP + FN.
Learning objective
There are 5 key performance characteristics that reflect the utility of a screening or diagnostic test. Positive and negative predictive values are useful metrics that reflect the probability that a positive or negative test result, respectively, is accurate. In diseases with low population prevalence (e.g. Angelman), the PPV will be lower, and the NPV will be higher.
2023 ABMGG General Exam Blueprint | IX. Population screening → c) Other (page 4)
Resources
Confusion matrix (Wikipedia)
Sensitivity, Specificity, PPV, NPV (Youtube)