4.1 A clinical expert explained concerns about the global increase in bacteria developing resistance to antibiotics (antimicrobial resistance). Data from the UK's 5-year action plan for antimicrobial resistance 2019 to 2024 estimate that 700,000 people die every year globally because of infections caused by resistant strains of bacteria and this number will increase if no action is taken. The report notes that no new classes of antibiotics have been developed since the 1980s. Tackling antimicrobial resistance has been one of the key UK public health priorities for several years, and the use of antibiotics is gradually reducing. From 2014 to 2017, antibiotic use decreased by 7.3%, from 23.4 to 21.7 defined daily doses per 1,000 inhabitants per day. A key aim of the UK's 5‑year action plan for antimicrobial resistance is to implement diagnostic tests that can guide antimicrobial prescribing decisions. The committee noted that rapid tests for group A streptococcal infections (strep A) have been promoted for this purpose.
4.2 A clinical expert explained that sore throat is a common condition that is usually self-limiting, that is, it usually resolves without any antibiotic treatment. Usually a sore throat is caused by a virus and so treatment with antibiotics is not needed. The committee was aware of NICE's guideline on antimicrobial prescribing for acute sore throat. This highlights that treatment with antibiotics only reduces symptom duration by around 16 hours. However, the committee noted that this guideline covers all bacterial infections of the throat, and it was not clear what the treatment effect of antibiotics would be in people with a confirmed strep A infection. Antibiotics could reduce the risk of some complications of strep A, but these are usually either very rare or not serious. The committee heard that antibiotics are often prescribed because of perceived clinical need or patient and carers' expectations to have treatment. Clinical experts explained that NICE's guideline on antimicrobial prescribing for acute sore throat focuses on measures of self-care and advises that antibiotics should only be considered for people who are most likely to benefit from them. Delayed prescribing is an option for most people who need antibiotics; that is, when antibiotics are only dispensed if symptoms do not improve within a few days of the person visiting their GP. Full implementation of this guideline is anticipated to reduce the use of antibiotics in people with a sore throat. Patients may be reassured by a discussion that highlights the likelihood of a sore throat becoming more severe balanced with the risk associated with taking antibiotics. The committee concluded that, in people who are otherwise healthy, antibiotics are usually not needed for a sore throat. Therefore, the clinical need for rapid testing for strep A infections is unclear.
4.3 A patient expert explained about the needs of patients and carers when they are seeking medical advice for a sore throat, and making a decision about whether to have antibiotics or whether to self-care. Patients would value information on what the results of the rapid strep A tests mean, how reliable they are, what the test involves and whether this information influences a treatment decision. The patient expert noted that it could be more difficult to explain the test procedure or take a throat swab in younger children and in people with cognitive impairment or learning difficulties. Therefore, this could be challenging to do routinely in a standard appointment. They noted that some people with sore throat may appreciate point-of-care testing and almost immediate results, whereas others may prefer the samples being sent for laboratory processing because this may be seen as more reliable. The committee concluded that patients and carers seeking advice for sore throat may have different testing needs and preferences, and treatment expectations.
4.4 The committee discussed the available data on the diagnostic accuracy of the rapid tests for strep A in people with a sore throat. It noted that although 26 accuracy studies were identified by the external assessment group (EAG), most included a broad population and only 2 reported data separately for people with a high clinical score (Centor score of 3 or more). The rapid tests for strep A are most likely to be useful for people with a high clinical score. The committee noted that the prevalence of strep A is higher in people with high clinical scores than in people with low clinical scores or in an unselected population of people with a sore throat. Therefore, it concluded that studies in unselected populations or populations with lower clinical scores may not be applicable to NHS practice.
The accuracy of the rapid tests in routine clinical practice is uncertain and likely to be very variable
4.5 The committee discussed the accuracy data available for each of the tests. It noted that no data were available for 3 of the tests (Strep A Rapid Test Strip from Biopanda, Biosynex Strep A Cassette test, and Bionexia Strep A Plus Cassette test). The EAG highlighted the high level of uncertainty in the estimates of rapid test accuracy because of the limited evidence available and the high variability between the studies. The committee noted that some tests only had accuracy data from studies done under ideal conditions (such as in unpublished manufacturer studies), which are unlikely to be repeatable in routine clinical practice. This is because the rapid tests' performance is linked to the quality of sampling and processing the sample. A clinical expert commented that positive test results are usually correct, but negative results could be related to absence of strep A infection, poor test sensitivity, or poor sampling technique. The committee also noted the imperfect accuracy of the reference standard (microbiological culture of throat swabs), which is subject to similar limitations. Laboratory polymerase chain reaction (PCR) tests have higher sensitivity than microbiological culture of throat swabs, but the clinical significance of this is unclear. For example, laboratory PCR tests could detect strep A carriers rather than infection, resulting in false positive results. The committee noted that the imperfect reference standard could under- or overestimate the accuracy of the rapid tests but that the size of either bias was not known. Overall, the committee concluded that the performance of the rapid tests in routine clinical practice is uncertain and difficult to predict, and is likely to vary from practice to practice.
Rapid tests (used with clinical scoring tools) are unlikely to improve clinical outcomes compared with the use of clinical scoring tools alone
4.6 The committee reviewed the available evidence on the clinical effectiveness of using the rapid tests for people with suspected strep A throat infections. There was no evidence available on clinical outcomes such as morbidity, mortality, or onward transmission rate. The committee noted that severe complications of strep A are rare and there is no evidence to suggest that the rapid tests would reduce the risk of them happening. There were only 3 randomised controlled trials that reported antibiotic prescribing behaviours with or without rapid testing. The committee discussed the study by Little et al. (2013), done in UK primary care. The rate of delayed prescribing was lower in the rapid test group (23%) compared with the clinical score only group (43%). However, the reported use of antibiotics appeared similar in both groups (35% and 37%, respectively) and the level of immediate prescribing was also similar in both groups (18% and 16%, respectively). The EAG explained that data on reported antibiotic use were only available for 80% of enrolled patients so should be interpreted with caution. Also, a clinical expert commented that symptom severity and time to symptom resolution were comparable between these 2 groups (although the 2 groups were not formally tested for a difference). The committee concluded that the clinical benefit of the rapid tests was uncertain. The only study (Little et al. 2013) providing some evidence on this suggested there may be no benefit of rapid testing (used with the clinical scoring tool), compared with the use of clinical scoring tools alone.
4.7 The committee was aware that the rapid tests may be available in some community pharmacies. The EAG found no evidence on the diagnostic or clinical utility of rapid test accuracy when used in pharmacies, and therefore could not model this. Also, the committee noted that FeverPAIN had not been validated for use in pharmacies and that staff might need training to use clinical scoring tools. The committee concluded that it was not possible to assess the cost effectiveness of rapid tests for use in pharmacies.
4.8 The committee noted that test accuracy inputs for 9 of 14 rapid tests for which cost-effectiveness analysis was possible were from unpublished manufacturer data. This is likely to overestimate the accuracy of tests in routine clinical practice. Most accuracy estimates were from studies that were not applicable to NHS practice (see sections 4.4 and 4.5) because they did not use the tests with clinical scoring tools, or included unselected populations who did not necessarily have high scores from clinical scoring tools. The committee therefore concluded that the incremental cost-effectiveness ratios (ICERs) produced by the models were highly uncertain because of bias in the data used to model the accuracy of the rapid tests.
4.9 The committee discussed the estimated costs of the tests and of the staff time for running the tests in the models. It raised concerns about the Xpert Xpress Strep A test cost, which was much lower than the costs of the other 3 molecular tests. The EAG explained that the test costs also included analyser or test cassette reader costs (when this equipment is needed). For all tests except Xpert Xpress Strep A, it was assumed that 2 tests per day would be done in a medium-sized GP practice, based on expert opinion. For the Xpert Xpress Strep A test, it was assumed that 28 tests per day would be run, resulting in a lower cost per test and therefore more favourable ICERs. The committee noted that an updated price for the Cobas Strep A assay was submitted by the company during consultation. It understood this to be an average selling price, based on volume-based discounts; the range associated with the average selling price was not provided to NICE. The updated test cost did not change the conclusions of the analyses. Also, the committee noted that the updated cost did not include analyser costs. Therefore, the incremental costs associated with this test are likely to have been underestimated. Clinical experts commented that the time to process the rapid tests in routine clinical practice was likely underestimated in the models. They explained that the time included appeared to account only for the time for the test read-out, and not for the time needed to prepare the test and take the throat swab. The total time necessary to run the test would depend on the experience of the healthcare professional doing the test. This could vary considerably between practices depending on their set-up for point-of-care testing. The time needed to take the throat swab might also be longer for certain populations, for example younger children. The committee concluded that including a more realistic estimate of test processing time would further increase the costs and ICERs for the rapid tests.
4.10 The committee considered the adverse events in the models and noted that the rate of penicillin-induced anaphylaxis had a big effect on the ICERs in scenario analyses. Clinical experts advised that the rate of penicillin-related anaphylaxis, when penicillin is taken orally, is very low (about 1 in a million). Therefore, the rate assumed in the base-case scenario (0.01%) is more appropriate than the rate assumed in the scenario analyses (0.64%), but could even overestimate the rate of penicillin-related anaphylaxis in the UK. Clinical experts also noted that the costs of sepsis are not generalisable to the treatment costs of penicillin-induced anaphylaxis, as had been assumed by the EAG. The committee concluded that penicillin-induced anaphylaxis is rare, and that the results of the scenario analyses which included a higher rate were unrealistic.
4.11 The committee recalled that one of the suggested benefits of the tests was providing a more targeted approach to antibiotic prescribing (see section 4.1). It discussed the antibiotic use predicted by the models for both current practice and for the rapid tests (used with clinical scoring tools). The models predicted a 10% to 15% decrease in the absolute rate of antibiotic use with rapid tests, compared with current practice. This was based on predicted treatment decisions related to the Centor score or the rapid test results (see table 6). The committee questioned the external validity of this prediction because the study by Little et al. (2013) suggested similar antibiotic use between people who had the rapid test (with the clinical scoring tool), or the clinical scoring tool only (see section 4.6). It also recalled that the recent publication of NICE's guideline on antimicrobial prescribing for acute sore throat is expected to reduce antibiotic prescribing in sore throat further (see section 4.2). This could result in the tests having the potential to increase antibiotic prescribing. Clinical experts also advised that the FeverPAIN clinical scoring tool is more discriminative for strep A than the Centor tool. Therefore, the potential clinical benefits of the rapid tests compared with FeverPAIN could be even lower. The committee concluded that the predicted decrease in antibiotic use associated with the rapid tests might not be replicated in NHS practice.
The models do not account for all complications of strep A, but this is appropriate because they are very rare
4.12 The committee noted that the models do not capture all complications of strep A and discussed the effect of excluding those that are more serious. Clinical experts advised that the risk of serious complications, such as invasive strep A or sepsis, is very low and therefore unlikely to have a major effect on the modelling results. The rates of invasive strep A appear to have been increasing in the UK in recent years, but are still very low considering the high number of people presenting with a sore throat. The rate of serious complications is higher in people over 75 years, and the risk of associated mortality is higher for these people. Therefore, modelling serious complications could be important. However, modelling the use of rapid tests for people over 75 was not possible because of the lack of data for the models. The committee also discussed that the models did not capture the risk of scarlet fever, although they acknowledged that this was a possible complication in children presenting early with a sore throat. Scarlet fever is more likely in children than adults. The rates of scarlet fever appear to have been increasing in the UK in recent years but are still low considering the total number of children presenting with a sore throat. For the children and adult models, the committee concluded that excluding the more severe complications was unlikely to have had a big effect on the results.
4.13 The committee noted that the wider public health benefits of the rapid tests, such as contribution to antimicrobial stewardship or effect on onward transmission rate, were not captured in the models. The committee discussed the risk of onward transmission of untreated strep A infection to other household members, particularly the risk of onward transmission leading to invasive strep A infection. It noted that although this risk exists, it is very small. The risk of onward transmission could be higher during an outbreak, for example, in a care home. However, using the tests during an outbreak was outside the scope of this assessment. The committee also noted that currently the effect on public health (health effects and costs) of reduced antibiotic use has not been quantified. The modelling predicted a 10% to 15% reduction in the absolute rate of antibiotic use with the rapid tests (with clinical scoring tools), compared with using clinical scoring tools only. However, this had minimal effect on the total cost of the pathway because penicillin costs are very low. The only differences in costs and quality-adjusted life years (QALYs) related to antibiotic prescribing were those of managing less severe strep A complications and side effects of penicillin. The committee discussed that although bacterial resistance to penicillin is not thought to be as great a problem as resistance to other classes of antibiotics such as macrolides (for example, erythromycin) or cephalosporins, there is very limited evidence to quantify this. Therefore, further research on the contribution of different classes of antibiotics to antimicrobial resistance, and to quantify the wider effect of antimicrobial stewardship, is needed (see section 5.1). The committee concluded that this evidence will be important for developing tests to improve prescribing decisions, which have the greatest effect on reducing antimicrobial resistance.
4.14 The modelling predicted very small incremental costs and even smaller incremental QALYs. This resulted in ICERs between £1 million and £6 million per QALY gained, compared with current clinical practice, for most rapid tests (adult primary care model). These ICERs are far higher than those normally considered to be a cost-effective use of NHS resources and are based on the modelling predicting a decrease in antibiotic use that may not be replicated in NHS practice (see section 4.11). The committee noted that there is uncertainty about the model inputs and the most plausible ICERs. However, all sensitivity analyses suggested that the rapid tests are unlikely to be a cost-effective use of NHS resources. Also, the magnitude of any uncaptured benefit is not likely to be sufficient to change the interpretation of the results. The committee recalled the uncertain clinical role of the rapid tests in the context of NICE's recent guidance on antimicrobial prescribing for acute sore throat. The guidance advises that sore throat is self-limiting, and so in people who are otherwise healthy, antibiotics are usually not needed. The committee noted that the diagnostic accuracy of the tests in routine clinical practice is uncertain and likely to be highly variable. Also, there was no evidence to suggest that the rapid tests could reduce the rate of antibiotic prescribing or improve clinical outcomes in people with a sore throat. Therefore, the committee concluded that the most plausible ICERs for the rapid tests were too high, and their effect on wider public health benefits too uncertain (see section 4.13), to recommend routine adoption.