Laboratory Assay Evaluation Demystified: A Review of Key Factors Influencing Interpretation of Test Results Using Different Assays for SARS-CoV-2 Infection Diagnosis

Abstract

Laboratory tests are an integral part of the diagnosis and management of patients; however, these tests are far from perfect. Their imperfections can be due to patient health condition, specimen collection, and/or technological difficulty with performing the assay and/or interpretation. To be useful clinically, testing requires calculation of positive predictive values (PPVs) and negative predictive values (NPVs). During the current global pandemic of COVID-19 (coronavirus disease 2019), multiple assays with unknown clinical sensitivity and specificity have been rapidly developed to aid in the diagnosis of the disease. Due to a lack of surveillance testing, the prevalence of COVID-19 remains unknown. Hence, using this situation as an clinical example, the goal of this article is to clarify the key factors that influence the PPV and NPV yielded by diagnostic testing, By doing so, we hope to offer health-care providers information that will help them better understand the potential implications of utilizing these test results in clinical patient management.

Keywords: clinical interpretation; negative predictive value; positive predictive value; sensitivity; specificity; test parameters.

Figure 1

Positive predictive value (PPV) and negative predictive value (NPV) as a function of sensitivity and specificity when the prevalence of the disease is 5%. A, As seen in this graph, specificity is the most important factor in the PPV when the disease prevalence is low. The PPV almost does not change significantly across the range of sensitivities of the assay. Nevertheless, the specificity must be close to 100% to obtain good PPV, and its effect on PPV is exponential (ie, small decrease in specificity leads to very large decrease in PPV). B, When the disease prevalence is low (5% in this example), sensitivity and specificity (as long as specificity is approximately >0.4) do not have a major effect on NPV.

Figure 2

Positive predictive value (PPV) and negative predictive value (NPV) as a function of sensitivity and specificity when the prevalence of the disease is 95%. A, When the disease prevalence is high, sensitivity (as long sensitivity is approximately >0.4) and specificity do not have a major effect on PPV. B, As seen in this graph, sensitivity is the most important factor in the NPV when the disease prevalence is high (95% in this example). The NPV almost does not change significantly across the range of specificities of the assay. Nevertheless, the sensitivity must be close to 100% to obtain good NPV, and its effect on NPV is exponential (ie, small decrease in sensitivity leads to large decrease in NPV).

Figure 3

Positive predictive value (PPV; part A) and negative predictive value (NPV; part B) as a function of sensitivity and specificity when the prevalence of the disease is 40%.