Background
Antibiotic resistance is a major challenge for the management of gonorrhoea globally: extended-spectrum cephalosporins are the last antibiotic class remaining for empirical treatment of gonorrhoea [
1,
2], and 42 countries have already reported
Neisseria gonorrhoeae strains with decreased susceptibility against them [
2]. The first strain with high-level resistance to the recommended combination therapy with ceftriaxone and azithromycin was recently described [
3]. With an estimated 78 million new gonorrhoea cases each year [
4], new control strategies are urgently needed before gonorrhoea becomes untreatable.
Conventional diagnostic tests for gonorrhoea, such as nucleic acid amplification tests (NAATs) and culture, are not sufficient to control antibiotic resistance. Commercially available NAATs, the most commonly used diagnostic gonorrhoea tests in high-income countries, cannot detect antibiotic resistance [
5,
6]. Cultures of
N. gonorrhoeae can be used to determine antibiotic-resistance profiles, but reliable results depend on stringent collection and transport of specimens [
7]. Culture and susceptibility testing need several days to deliver results [
8]. NAATs can deliver results in a few hours [
9], but in routine use specimens might be tested in batches [
10] and NAAT results might be delivered only after several days [
11]. While symptomatic gonorrhoea patients usually receive empirical treatment at their first visit, asymptomatic patients might have to return for treatment. Loss to follow-up and the further spread of resistant infections can result.
Point-of-care (POC) tests promise to help control antibiotic resistance [
12]. POC tests provide results rapidly and allow informed clinical decisions about treatment at the first visit of a patient. POC tests, therefore, reduce the time to treatment and avoid loss to follow-up. A modelling study suggested that POC tests can reduce gonorrhoea prevalence if no antibiotic resistance is present in the population [
13]. Though not yet commercially available [
12], POC tests that detect resistance promise to reduce the use of antibiotics [
14] and to spare last-line antibiotics through individually tailored treatment [
15,
16]. One modelling study illustrated that individualised treatment could slow down the spread of resistance as much as combination therapy [
17]. However, reduced time to treatment and increased follow-up with POC tests might increase the rate of gonorrhoea treatment. Since higher treatment rates can lead to the faster spread of resistance [
18,
19], POC tests might increase resistance levels. We extended a previously developed mathematical model of gonorrhoea transmission [
19] to compare the effects of current conventional tests (culture and NAATs) with POC tests that reduce time to treatment and loss to follow-up. We investigated the potential impact of POC tests on resistance and on the number of gonorrhoea cases for a population at high risk of infection [
20], men who have sex with men (MSM), and a population at lower risk of infection, heterosexual men and women (HMW).
Discussion
Using a mathematical transmission model, we compared the expected impact of POC tests on gonorrhoea cases and antibiotic resistance with conventional tests, culture and NAAT. We found that POC tests that detect antibiotic resistance with a sensitivity of 99% avert more gonorrhoea cases than any other test across all simulated settings. Additionally, we found that POC tests can slow down the spread of resistance if their test sensitivity to detect resistance is sufficiently high. If the test sensitivity of a POC test to detect resistance is higher than 0–40%, resistance spreads more slowly than with NAAT, and if POC sensitivity to detect resistance is higher than 80–95%, resistance spreads more slowly than with culture.
We captured the basic principles of the gonorrhoea testing and treatment process for culture, NAAT, and POC in a single model structure. The parameters describing the sexual behaviour and the natural history of gonorrhoea were estimated and calibrated in a previous study [
19]. The default parameters that describe testing and treatment of gonorrhoea were based on literature values and are measurable. The model results are robust in sensitivity analyses (Figs.
3,
4 and
5, Additional file
1: Figures S5–S14). The model can be used to help design trials comparing different test strategies and guide the introduction of POC tests in the future.
Mathematical models generally depend on assumptions that should be taken into consideration when interpreting model results. We managed the complexity of our model with the following assumptions.
First, we did not consider test specificity. A low test specificity to detect resistance against the first-line antibiotic would result in increased use of the second-line antibiotic, and thus, simultaneously decrease the level of resistance against the first-line antibiotic and increase the level of resistance against the second-line antibiotic. Since we focused on resistance against the first-line antibiotic, we could not capture the impact of test specificity appropriately.
Second, our model does not include a change in antibiotic recommendations: undetected resistant infections are always treated with the first-line antibiotic, even if all infections in the population are resistant. This clinical pathway increases the average duration of resistant infections and possibly the observed cases. Whilst this is unlikely in high-income countries with good antibiotic resistance surveillance, it is not an unrealistic scenario in resource-poor settings without surveillance where 71–100% of gonococcal strains are resistant to fluoroquinolones [
28]. In our model, MSM have a substantial level of resistant gonorrhoea infections after 5 years using NAAT. We expect that our model overestimates the observed cases using NAAT and the observed cases averted using culture and POC+R compared with a model including a change in antibiotic recommendations.
Third, we considered treatment with a single antibiotic although current treatment guidelines recommend a combination therapy with two antibiotics simultaneously [
1,
7]. The model results are fully applicable to treatment with combination therapy if antibiotic-resistant gonorrhoea is interpreted as resistance against both antibiotics used for combination therapy.
Fourth, we investigated the effects of one test at a time and did not consider the effects of mixed testing. Our results, therefore, show only what the ideal effects of each test could be.
Fifth, we simplified the testing and treatment process. To compare the testing scenarios better, we did not model care-seeking and returning for treatment as separate processes, but approximated the overall treatment rates. In accordance with WHO [
20] and CDC recommendations [
22], we assumed that re-treatment of resistant infections occurs with the second-line antibiotic because a resistance profile has been determined after the second visit.
Sixth, for better comparability we assumed that culture, NAAT, and POC tests have the same sensitivity to detect gonorrhoea, even though culture has a lower sensitivity to detect rectal or pharyngeal gonorrhoea than molecular tests [
29]. Our sensitivity analysis showed that a lower culture test sensitivity to detect gonorrhoea of 90% had a small effect on the model results.
Finally, it should be mentioned that the minimal POC test sensitivity to detect resistance that is necessary to slow down resistance spread will depend on the tests currently used, the setting, and population, and should be subject to validation.
Currently, there are no commercial POC tests that can detect antibiotic-resistant
N. gonorrhoeae [
12] and there remain challenges for their development. First, molecular POC test that detect resistance need molecular markers that reliably predict phenotypic resistance. So far only markers that predict resistance against some antibiotics are known [
12,
30,
31]. Second, diagnostic tests need to deliver results fast to be considered POC. The fastest molecular diagnostic test for gonorrhoea that is commercially available takes 90 minutes [
9,
32], which might be too long to wait for some patients. Finally, costs and training requirements for molecular tests have hindered their availability in low-income countries so far [
33].
This study addresses two key questions for gonorrhoea control and resistance [
34]. First, we investigated the potential impact of a POC test that detects antibiotic resistance (POC+R). We found that POC+R can slow resistance spread and reduce the number of gonorrhoea cases compared with culture or NAAT. The impact of POC+R is particularly strong when the fraction of asymptomatic individuals who return for treatment (
λ
A,baseline) and the fraction of successfully treated individuals who were symptomatic (
ψ) were low before POC+R is introduced. However, when the POC test cannot detect resistance (POC−R), the benefits of POC are outweighed by accelerated resistance evolution. Because fewer patients are lost to follow-up, more patients are treated and more antibiotic treatment selects more strongly for antibiotic resistance. Since resistance cannot be detected, resistance levels increase and fewer cases are averted.
Second, we investigated the impact of POC tests in two populations with different levels of risk of gonorrhoea, MSM and HMW. We found that in both populations, POC tests with reliable resistance detection (POC+R) slow down the spread of resistance and avert the highest number of cases. POC tests without resistance detection (POC−R) avert about as many cases as POC+R in HMW, but clearly fewer cases than POC+R in MSM. Since resistance usually spreads faster in MSM [
19], the faster spread of resistance caused by POC−R impacts the cases averted after 5 years in MSM, but not in HMW. POC tests that detect resistance reliably are crucial for both populations and both populations need culture-based surveillance of resistance to keep molecular markers for POC resistance detection updated.
Acknowledgements
Not applicable.