Everyone is clamoring to get COVID-19 at-home tests this holiday season. The tests come in two flavors. One are tests from the folks at Detect and Cue Health that provide PCR like results in your home. The others are the more common rapid antigen tests. With so many on the market you might ask yourself, is one better than the others? While we’re still waiting for more data and third party studies to validate the accuracy of these tests, it is hard to say for sure. However I have collected the self-reported and some third-party studies for each of the tests so you can see the difference in performance of them.
In general most rapid antigen tests are good for telling you whether the sniffles you have are COVID or a cold. The data on detecting symptomatic infection is good for a majority of them. We have less data on how effective they are at detecting asymptomatic infection and the ones that we do have data for tend not to be as good at it. If you care about asymptomatic COVID detection my recommendation is to use one of the PCR like tests from Detect or Cue health, or find a local PCR test. If a rapid-test is your only option, prefer Flowflex.
If you cannot get a PCR test or do not want to pay the higher price for the at-home PCR-like kits, there are a few things you can do to have more confidence in your rapid-tests. The first is to take them as close in time as possible to when your gathering is. The theory is that even if you are infected, if the amount of virus in your nose is below the limit of detection of the test you’re less likely to be contagious. Second, you can take a second test 24-48 hours after your first one. This series procedure is how the FDA has authorized most tests (other than Flowflex) for asymptomatic negative testing.
The data below contains tables I compiled from various data sheets submitted to the FDA from the test companies. The provided tables sometimes break down the data in different ways so I standardized it to provide a row for test sensitivity and test specificity. Sensitivity is how well a test is able to detect a positive COVID case. Specificity is how accurately it reports a negative result. Many tests break out symptomatic v. asymptomatic data for the test, so other than for the at-home PCR like tests which did not provide this breakdown, I do the same. The raw counts are in the column with the test name and the Lab PCR comparison counts. Higher percentages are better on both attributes. The larger the raw number of study participants is, the more confident you can be that the study represents how well the test will perform in the real world.
Detect
Subjects that were symptomatic or had a recent exposure.
Symptomatic or Exposed | Detect | Lab PCR | |
---|---|---|---|
Sensitivity | 90.9% | 30 | 33 |
Specificity | 100% | 79 | 79 |
Cue
Cue Health conducted prospective studies at 4 urgent care locations and at 2 Cue Health locations to evaluate use of the Cue COVID-19 Test by lay users in a simulated home use environment. All subjects successfully followed the instructions in the Cue Health App to run the Cue COVID-19 Test, start to finish without any assistance.
Symptomatic and Close Contacts | Cue | Lab PCR | |
---|---|---|---|
Sensitivity | 91% | 22 | 24 |
Specificity | 98% | 239 | 243 |
iHealth
Symptomatic | iHealth | Lab PCR | |
---|---|---|---|
Sensitivity | 94.3% | 33 | 35 |
Specificity | 98.1% | 102 | 104 |
Ellume
Symptomatic | Ellume | Lab PCR | |
---|---|---|---|
Sensitivity | 96% | 25 | 26 |
Specificity | 100% | 38 | 38 |
Asymptomatic | Ellume | Lab PCR | |
---|---|---|---|
Sensitivity | 91% | 10 | 11 |
Specificity | 96% | 118 | 123 |
Binax Now
Study 1
Symptomatic | BinaxNow | Lab PCR | |
---|---|---|---|
Sensitivity | 91.7% | 22 | 24 |
Specificity | 100% | 28 | 28 |
Study 2
Symptomatic | BinaxNow | Lab PCR | |
---|---|---|---|
Sensitivity | 84.6% | 99 | 117 |
Specificity | 98.5% | 338 | 343 |
Massachusetts Third Party Study
1,380 adults and 928 children 974/1,380 (71%) adults and 829/928 (89%) children were asymptomatic.
Symptomatic | BinaxNow | Lab PCR | |
---|---|---|---|
Sensitivity | 96.5% | 82 | 85 |
Specificity | 100% | 270 | 270 |
Asymptomatic | BinaxNow | Lab PCR | |
---|---|---|---|
Sensitivity | 70.2% | 40 | 57 |
Specificity | 99.6% | 913 | 917 |
Flowflex
108 symptomatic and 64 asymptomatic patients
Symptomatic | Flowflex | Lab PCR | |
---|---|---|---|
Sensitivity | 93% | 28 | 30 |
Specificity | 100% | 78 | 78 |
Asymptomatic | Flowflex | Lab PCR | |
---|---|---|---|
Sensitivity | 92% | 11 | 12 |
Specificity | 100% | 52 | 52 |
AccessBio
Symptomatic | AccessBio | Lab PCR | |
---|---|---|---|
Sensitivity | 95% | 19 | 20 |
Specificity | 98% | 120 | 123 |
Asymptomatic | AccessBio | Lab PCR | |
---|---|---|---|
Sensitivity | 70% | 7 | 10 |
Specificity | 97.6% | 123 | 126 |
Massachusetts Study
1245 adults and 253 children, 83% asymptomatic
Symptomatic < 7 days | AccessBio | PCR | |
---|---|---|---|
Sensitivity | 84.8% | 43 | 51 |
Specificity | 97.2% | 115 | 118 |
Asymptomatic | AccessBio | PCR | |
---|---|---|---|
Sensitivity | 50.0% | 62 | 124 |
Specificity | 99.1% | 904 | 912 |
BD Veritor
597 Symptomatic Subjects
Symptomatic | BD Veritor | Lab PCR | |
---|---|---|---|
Sensitivity | 84.6% | 33 | 39 |
Specificity | 99.8% | 557 | 558 |
Intrivo On/Go
A total of 153 subjects were evaluated in this study
Symptomatic | On/Go | Lab PCR | |
---|---|---|---|
Sensitivity | 84.6% | 26 | 30 |
Specificity | 99.8% | 120 | 123 |
Asymptomatic | On/Go | Lab PCR | |
---|---|---|---|
Sensitivity | 70% | 7 | 10 |
Specificity | 97.6% | 123 | 126 |
Limits of Detection
Besides comparing real world study performance of the tests, another way to infer whether it’s likely to detect COVID is to look at its limit of detection. The limit of detection is a lab test that ascertains the minimum amount of virus that needs to be present to trigger a positive test result 95% of the time. Tests that have a lower limit of infection should see COVID in samples that tests with higher limits cannot see. A lower limit of detection is better if you also have confidence in the test from the study data above. For example while BD Veritor self-reports the best limit of detection here, the lack of study data on asymptomatic detection means I would not use it for that yet.
What is the impact of a limit of detection? One article frames it this way:
These results suggest that each 10-fold increase in LoD is expected to increase the false negative rate by 13%, missing an additional one in eight infected patients. The highest LoDs on the market will miss a majority of infected patients, with false negative rates as high as 70%.
Brand | Limit of Detection (TCID50 / ml) |
---|---|
BD Veritor | 1871 |
Detect | 8002 |
Cue | 1,300 |
Flowflex | 2,500 |
Intrivo On/Go | 2,800 |
Ellume | 6,310 |
QuickVue | 19,100 |
iHealth | 20,000 |
Abbott BinaxNow | 40,4003 |
AccessBio | 540,0004 |
Conclusions
The data provided by manufacturer studies is helpful to better understand test performance. However this data is not as robust as real world third party study data. We have some of this for tests like BinaxNow and see that for symptomatic patients this test does fine, but the accuracy does not hold up as well for asymptomatic patients. You can mitigate the asymptomatic accuracy of the tests by taking them closer in time to when you’re going to gather and taking them in a two test series.
The at-home PCR like tests have the lowest limits of detection other than the BD Veritor rapid test which has a figure so low it is a bit suspect. I would not trust it until they release more data about asymptomatic detection. The at-home PCR like tests or Flowflex are currently going to be your best bet for detecting asymptomatic COVID before a gathering.
-
This seems suspiciously low, especially given their 84% sensitivity. I do not trust this number at this time. ↩
-
Over two lots of tests, the LoD of the DetectTM Covid-19 Test was determined to be 313 copies per swab, which is equivalent to 800 copies per mL if all virus is transferred from the swab to the buffer. ↩
-
Analytical Sensitivity of the Abbott BinaxNOW COVID-19 Ag Card is a third party study showing it has a limit of detection of 4.04 × 10^4 to 8.06 × 10^4 copies/swab. Abbott claims a sensitivity of 140.6 TCID50/mL but given the other figures and third party performance characteristics, I am skeptical. ↩
-
AccessBio claims their limit of detection is 2,800 TCID50/mL but given the poor results in this third party study I do not think the first-party study is reliable. ↩