Estimating the fitness cost and benefit of cefixime resistance in Neisseria gonorrhoeae to inform prescription policy: A modelling study

Abstract

Background: Gonorrhoea is one of the most common bacterial sexually transmitted infections in England. Over 41,000 cases were recorded in 2015, more than half of which occurred in men who have sex with men (MSM). As the bacterium has developed resistance to each first-line antibiotic in turn, we need an improved understanding of fitness benefits and costs of antibiotic resistance to inform control policy and planning. Cefixime was recommended as a single-dose treatment for gonorrhoea from 2005 to 2010, during which time resistance increased, and subsequently declined.

Methods and findings: We developed a stochastic compartmental model representing the natural history and transmission of cefixime-sensitive and cefixime-resistant strains of Neisseria gonorrhoeae in MSM in England, which was applied to data on diagnoses and prescriptions between 2008 and 2015. We estimated that asymptomatic carriers play a crucial role in overall transmission dynamics, with 37% (95% credible interval CrI 24%-52%) of infections remaining asymptomatic and untreated, accounting for 89% (95% CrI 82%-93%) of onward transmission. The fitness cost of cefixime resistance in the absence of cefixime usage was estimated to be such that the number of secondary infections caused by resistant strains is only about half as much as for the susceptible strains, which is insufficient to maintain persistence. However, we estimated that treatment of cefixime-resistant strains with cefixime was unsuccessful in 83% (95% CrI 53%-99%) of cases, representing a fitness benefit of resistance. This benefit was large enough to counterbalance the fitness cost when 31% (95% CrI 26%-36%) of cases were treated with cefixime, and when more than 55% (95% CrI 44%-66%) of cases were treated with cefixime, the resistant strain had a net fitness advantage over the susceptible strain. Limitations include sparse data leading to large intervals on key model parameters and necessary assumptions in the modelling of a complex epidemiological process.

Conclusions: Our study provides, to our knowledge, the first estimates of the fitness cost and benefit associated with resistance of the gonococcus to a clinically relevant antibiotic. Our findings have important implications for antibiotic stewardship and public health policies and, in particular, suggest that a previously abandoned antibiotic could be used again to treat a minority of gonorrhoea cases without raising resistance levels.

Fig 1. Usage and resistance of cefixime in England and Wales.

The proportion of gonococcal isolates in Gonococcal Resistance to Antimicrobials Surveillance Programme (GRASP) that are resistant to cefixime over time is compared with the proportion of gonorrhoea diagnoses treated with cefixime. Dashed lines show the dates of treatment guideline changes. MSM, men who have sex with men.

Fig 2. Flow diagram of model compartments with rates of transition between infection states.

Susceptible individuals (S) become infected with either cefixime-susceptible (s = sus) or cefixime-resistant strains (s = res). Infections initially pass through an incubation period (U_s) before the individuals with the infection either develop symptoms (E_s) or remain asymptomatic carriers (A_s). Symptomatic individuals seek treatment (T_s;p) and are prescribed either cefixime (p = cef) or another antibiotic (p = oth). The 2 sides are symmetric with the exception of the 2 arrows highlighted in red, which correspond to the cost and the benefit of resistance.

Fig 3. Posterior distributions of parameters.

The diagonal plots show histograms of the posterior distributions for all sampled parameters. The blue lines show prior distributions, and the red lines indicate posterior mean and 95% credible intervals. The plots below the diagonal show scatter plots based on 1,000 samples from the posterior, illustrating the relationships between pairs of estimated parameters. An orange background indicates a correlation higher than 0.8, a yellow background indicates a correlation between 0.5 and 0.8, a green background indicates a correlation between 0.3 and 0.5, and a white background indicates a correlation less than 0.3. The plots above the diagonal show the corresponding correlation coefficients with the 95% credible intervals in parentheses.

Fig 4. Comparison of simulated and observed cases of gonorrhoea.

Panel A shows the total number of cases, and panel B shows only the cefixime-resistant cases. Observed data are shown in orange, with the shaded area showing the 95% posterior predictive interval (based on 1,000 simulations using samples from posterior distribution). Note different scales. GUMCAD, Genitourinary Medicine Clinic Activity Dataset.

Fig 5

(A) Histogram of the posterior estimate of $R_0^{\mathrm{sus}}$. (B) The 95% credible interval of $R_0^{\mathrm{res}}(\pi )$ against π with dashed lines showing the 95% credible interval for ${\pi }^{\{R_0^{\mathrm{res}}=1\}}$. (C) Histogram of the posterior estimate of ${\pi }^{\{R_0^{\mathrm{sus}}=R_0^{\mathrm{res}}\}}$: the threshold of cefixime prescriptions above which $R_0^{\mathrm{res}}>R_0^{\mathrm{sus}}$.

Fig 6. Incidence of gonorrhoea in 2015 based on simulations from 2004 to 2015 with varying levels of cefixime prescribing.

(A) Incidence of the cefixime-resistant strain. The red lines show the 95% credible interval for ${\pi }^{\{R_0^{\mathrm{res}}=1\}}$. (B) Incidence of the cefixime-susceptible strain. The red lines show the 95% credible interval for ${\pi }^{\{R_0^{\mathrm{sus}}=R_0^{\mathrm{res}}\}}$. (C) The overall incidence from both cefixime-resistant and cefixime-susceptible strains. The shaded areas show the 95% posterior predictive intervals (based on 1,000 simulations using samples from posterior distribution).