Harmonizing limits of detection for FDA-authorized COVID-19 diagnostics
Comparing the limits of detection for COVID-19 molecular diagnostic tests offers a window into their relative sensitivity and potential to produce false negative results. New data published by FDA, based on a reference panel of patient samples, shakes up the standing of tests authorized early in the pandemic, whose initial metrics weren’t standardized and didn’t come from patient samples.
The new data, which compare the tests using the same samples and report results in the same units, highlight the benefits of harmonized protocols for decision-making, particularly in a pandemic (see “Having Touched the Third Rail of Data Sharing in the Pandemic, Drug Developers Should Hold on Tight”).
The biggest change in ranking was for the ID NOW COVID-19 test from Abbott Laboratories (NYSE:ABT), which dropped from being number three in sensitivity to being second to last (see Table).
The rapid point-of-care test, which uses a novel non-thermocycling probe technology and was widely deployed in the White House, came under scrutiny when an independent group found higher than expected rates of false negatives (see “New Scrutiny Prompts Re-Evaluation of Rapid Tests”).
More than 20 diagnostic tests received emergency use authorization (EUA) by FDA for COVID-19 testing between February and April. Because banked patient samples were scarce, manufacturers largely reported each test’s limit of detection (LoD) — the lowest amount of a target that the test can detect at least 95% of the time — based on samples from uninfected people spiked with viral RNA, in accordance with FDA guidelines.
A lower LoD means a more sensitive test, from an analytical sensitivity point of view. Determining a test’s clinical sensitivity requires measuring how frequently the test agrees with a reference test of known sensitivity.
As COVID-19 patient samples became more widely available, FDA developed a standardized reference panel to validate molecular diagnostic test performance, and began distributing it to test developers in May. To ensure accuracy of test results using the reference panel, FDA did not disclose the amount of viral material in the samples, which made the validation exercise a blind test for developers.
LoDs reported by FDA as of Sept. 17, which reflect the least sensitive extraction/instrument combination tested, are all reported in RNA Nucleic acid-based amplification test (NAAT) Detectable Units (NDUs)/mL.
As of April 2, test developers had reported LoDs in terms of viral genome copies per volume, copies per reaction, or using fractions of TCID50 (median tissue culture infectious dose) per mL, which represents the viral load at which 50% of cells are infected when a solution containing the virus is added to cell culture.
Most tests retained a similar relative standing, including Abbott’s RealTime SARS-CoV-2 assay, which was the second-most sensitive test based on manufacturer analyses as of April, and now ranks fourth based on FDA’s standardized tests in September.
The biggest improvement in standing among tests authorized by April 2 was for ScienCell Research Laboratories Inc., which was in the middle of the pack in April but jumped to the number two spot in September.
FDA’s analysis revealed a test from PerkinElmer Inc. (NYSE:PKI) to be the most sensitive as of Sept.17; previously, its LoD was reported in a unit not used by most other test developers, which made it difficult to gauge its standing.
Like the majority of COVID-19 diagnostics, PerkinElmer and ScienCell’s tests are both based on real-time polymerase chain reaction (RT-PCR).
The least sensitive test reported as of Sept. 17, after Abbott’s ID NOW, is a viral antigen test from Quidel Corp. (NASDAQ:QDEL), which measures viral proteins using a rapid immunoassay. Antigen tests are generally less sensitive than RT-PCR, but quicker, cheaper and more accessible (see “Test for the Utility of Antigen Tests”).
Both Abbott’s ID NOW and Quidel’s antigen test use dry nasopharyngeal swabs. The majority of tests were performed on nasopharyngeal swabs in viral transport media.
FDA data was available for one saliva test, from Fluidigm Corp. (NASDAQ:FLDM), which was authorized after April 2. The results suggest that test is less sensitive than most tests using RT-PCR on nasopharyngeal swabs, but about 5.5 times more sensitive than Abbott’s ID NOW, and 10 times more sensitive than Quidel’s antigen test.
LoDs reported by manufacturers as viral genomic sequence copies per volume are primarily reported in copies per mL; asterisks denote LoDs originally reported as copies per μL.
The CDC reported separate LoDs for two versions of its test involving different PCR primers.