Just how accurate are lateral flow tests in the detection of COVID-19? That depends!

From the 11th January, COVID-19 testing rules are to be relaxed in England for people without symptoms. A positive lateral flow test (LFT) result, will no longer need to be confirmed with a follow-up positive polymerase chain reaction (PCR) test. Those with a positive LFT will still need to isolate, they just won’t need the extra PCR test. Firstly, what are the differences between PCR and LFTs?

PCR tests work by looking for a small genetic fragment of the COVID-19 virus and then using PCR technology in a lab to create multiple copies. The lower the number of cycles it takes to find the virus, the stronger the infection. There is a threshold which determines whether the signal is too weak to be an active virus. PCR testing has the ability to pick up tiny amounts of the virus, meaning that a person can test positive before they have a viral load high enough to be infectious and it’s also possible to test positive via a PCR test a while after infectiousness has ended.

LFTs work by placing a sample on a porous strip that has a line of antibodies designed to bind to the COVID-19 virus. When this binding occurs, the line changes colour, indicating a positive test. LFTs are not as good at detecting the virus as PCR tests, but their value lies in the speed at which they give results. A LFT gives a result in 30 minutes, compared with PCR tests which must be sent away to a lab for testing. LFTs do adequately pick up people while they are at their most infectious, however. LFTs have also been a crucial tool in picking up asymptomatic COVID-19 through routine regular testing.

Mina et al suggest the following relationship between a person's infection trajectory (blue line) and the performance of LFTs and PCR tests in detective COVID-19.

This article in the Guardian states that “The care minister, Gillian Keegan, told the Today programme that the change was intended to reflect the “accuracy and the amount of lateral flow tests” rather than expressly to curb staff shortages”. But LFTs haven’t changed, so how can they now be more accurate? Well it all depends of what you are using as your definition of accuracy.

You could be thinking about accuracy in terms of sensitivity and specificity, or you could be thinking of it in terms of positive predictive value (PPV) and negative predictive value (NPV). Crucially, PPV and NPV both depend on the underlying prevalence in the population, so how much COVID-19 there is circulating can affect the accuracy of your tests.

Consider the following contingency table which presents the four different possible test result scenarios:

 

 

Disease status

 

 

Positive

Negative

Test result

Positive

True positive

False positive

Negative

False negative

True negative

 
We can then consider the following definitions:

  • Sensitivity – probability of testing positive if you have the disease

 

True positive

 

 

Disease status positive

 

  • Specificity – probability of testing negative if you don’t have the disease

 

True negative

 

 

Disease status negative

 

  • PPV – probability of having the disease if you test positive

 

True positive

 

 

Test result positive

 

  • NPV – probability of not having the disease if you test negative

 

True negative

 

 

Test result negative

 

 PPV and NPV are arguably the key quantities of interest here. What we want to know is if you get a positive test, what is the probability that you actually have COVID-19 and if you get a negative test, what is the probability that you are truly negative. To see the effect of prevalence on PPV and NPV, we can look at some numbers.

According to the latest estimates from the Office for National Statistics Infection Survey (released 5th January 2022), on the 28th December 2021, it was estimated that 6% of the population had COVID-19. Let’s take 10,000 people and assume a 6% prevalence. This means that 600 people would have COVID-19 and 9,400 wouldn’t. The press release from the UKHSA states that LFTs “are over 80% effective at finding people with high viral loads who are most infectious and most likely to transmit the virus to others” and they “have an estimated specificity of at least 99.97% when used in the community”. Table 1 is a contingency table for the 10,000 people with 6% prevalence, 80% sensitivity, and 99.97% specificity:
 

 

 

Disease status

 

 

Positive = 600

Negative = 9,400

Test result

Positive

480

3

Negative

120

9,397

 

Table 1: Contingency table with 6% prevalence

We would expect a total of 483 positive LFTs and of these, 480 will be true positives giving a PPV of over 99%. We would also expect a total of 9,517 negative LFTs and of these, 9,397 of them will be true negatives giving a NPV of 98.7%. This means that if you test negative, there’s a 1.3% chance that you are really positive.

This BBC report states that guidelines will be reviewed when COVID-19 levels in the population drop below 1%. How would these numbers change if the underlying prevalence were 1%? We would have 100 people with COVID-19 and 9,900 without. Table 2 is a contingency table for the 10,000 people with 1% prevalence, 80% sensitivity, and 99.97% specificity:
 

 

 

Disease status

 

 

Positive = 100

Negative = 9,900

Test result

Positive

80

3

Negative

20

9,897

 

Table 2: Contingency table with 1% prevalence

We would expect a total of 83 positive LFTs and of these, 80 will be true positives giving a PPV of 96%. We would also expect a total of 9,917 negative LFTs and of these, 9,897 of them will be true negatives giving a NPV of 99.8%. This means that if you test negative, there’s a 0.2% chance that you are really positive.

So, we can see that as the underlying prevalence increases, the chance of a positive LFT being a false positive decreases, but the chance of a negative LFT being a false negative increases. This is due to the fact that as prevalence increases, there is more and more of the virus circulating, so positive test results are more likely to be true and negative test results are more likely to be false.

We can see that for the current estimated prevalence levels of 6%, if you get a positive LFT, it’s highly likely that you actually do have COVID-19 as the PPV is over 99%. This is why the decision has been made to scrap confirmatory PCR tests. Just 3 of our 10,000 test results in Table 1 were false positives.

What is also interesting to note, is that, at the 6% prevalence level, the chance of getting a false negative test result is 1.3%. This may not sound very big, but if you look at the most recent Test and Trace statistics (released 23rd December), for the period 9th – 15th December there were over 830,000 LFTs being conducted daily. And these are the LFTs that we know about, the true figure is likely to be much higher. We know that people take LFTs and don’t submit the results online. But even if the figure was 830,000, that would mean that there would still be almost 10,800 false negative LFTs every day. So if you get a negative LFT, with this much COVID-19 circulating, it doesn’t mean that you are definitely negative.