Background Gonorrhoea can rapidly develop resistance to antimicrobials and treatment options are becoming depleted. Treatment guidelines require robust estimates of the prevalence of resistance but sentinel surveillance systems may not be representative nationally.
Objective To investigate the effectiveness of the sentinel Gonococcal Resistance to Antimicrobials Surveillance System (GRASP) at estimating resistance prevalence in England and Wales.
Methods Annual cross-sectional data on reported gonorrhoea diagnoses between 2000 and 2008 were compared between GRASP (26 clinics) and national mandatory (KC60) reporting (229 clinics). Resistance prevalence estimates in GRASP were weighted according to the national distribution of relevant patient characteristics: age group, gender, sexual orientation and geographical region. Trends in actual and weighted estimates were plotted.
Results Gonorrhoea cases reported through GRASP were more likely to be from London and to be men who have sex with men (MSM) and were less likely to be women and heterosexual men than those reported through KC60. Weighting for national distributions of demographic characteristics reduced estimates of resistance prevalence, particularly ciprofloxacin in 2006 (27% to 21%). Emerging resistance to cefixime in 2008 was reduced from 1.5% to 1.0%. Weighting did not adjust resistance prevalence above or below the 5% threshold for any antimicrobial.
Conclusions Although over-representing MSM and under-representing women and heterosexual men, GRASP has provided reliable estimates of resistance prevalence in England and Wales. However, weighting for the national distribution of patient characteristics should be considered in future. As resistance usually emerges in MSM, enhanced surveillance of high-risk populations could enable development of more tailored (and therefore optimal) treatment strategies.
- antimicrobial resistance
- epidemiology (general)
- chlamydia trachomatis
- vaginal infections
- STD surveillance
- laboratory diagnosis
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
- antimicrobial resistance
- epidemiology (general)
- chlamydia trachomatis
- vaginal infections
- STD surveillance
- laboratory diagnosis
The widespread emergence of antimicrobial resistance is of global concern.1 Resistance complicates and delays effective treatment, thus increasing the likelihood of poor health outcomes and onward transmission. Gonorrhoea, caused by the bacterium Neisseria gonorrhoeae, is the second most common bacterial sexually transmitted infection in the UK, with over 17 000 diagnoses made in genitourinary medicine (GUM) clinics in 2009.2 The ability of N gonorrhoeae to develop rapid resistance to antimicrobials is well described. Until recently, ciprofloxacin was the recommended first-line treatment for gonorrhoea in the UK but, as in many countries, the emergence of fluoroquinolone resistance precipitated a revision of treatment guidelines to recommend use of third-generation cephalosporins.3
According to WHO guidelines, use of an antimicrobial to treat an infection is usually no longer recommended once the overall level of resistance in the population exceeds 5%.4 Developing and updating treatment guidelines for gonorrhoea therefore require robust estimates of the overall prevalence of resistance and of how this may be changing among the main risk groups and over time. Surveillance programmes to inform treatment guidelines can be either comprehensive or sentinel, capturing data on all cases or from a limited sample, respectively, and can be either continuous or episodic. The choice is often made for practical reasons depending on resources available and numbers of cases (and hence isolates to be tested). Surveillance programmes for monitoring antimicrobial resistance in gonorrhoea include both those with comprehensive and continuous sampling (eg, Scotland,5 the Netherlands,6 Australia7) and those which are sentinel (eg, USA8). It is important that sample-based programmes are designed to permit reliable estimation of resistance prevalence in the population as a whole as well as in the main population subgroups at particular risk.
In England and Wales, the national surveillance of gonococcal resistance to antimicrobials was initiated in 2000 and is undertaken through the Gonococcal Resistance to Antimicrobials Surveillance Programme (GRASP). GRASP is a sentinel surveillance system, collecting consecutive gonococcal isolates for susceptibility testing over a 3-month period each year from selected laboratories across England and Wales.9 Although sites participating in GRASP cover all regions of England and Wales, the representativeness of any sentinel surveillance system needs to be understood to aid interpretation, especially for infections with highly localised distributions such as gonorrhoea. We used data from national mandatory reporting of gonorrhoea to investigate the representativeness of gonorrhoea sampled by GRASP by estimating the extent to which overall population-level estimates change when adjusted for the national distribution of relevant patient characteristics.
GRASP data sources and methodology
Detailed methodology for GRASP has been published elsewhere10 but, in brief, between June and August each year, GRASP collects consecutive gonococcal isolates and basic patient demographic data from 24 laboratories across England and Wales (figure 1). Susceptibility testing is performed centrally at the Health Protection Agency Sexually Transmitted Bacteria Reference Laboratory as described previously.10 Additional detailed clinical data and information on sexual behaviour are also collected on corresponding patients diagnosed with gonorrhoea in the associated 26 GUM clinics: 97% of men and 87% of women included in the 2008 collection were diagnosed in a GUM clinic.
Data source and methodology of national mandatory reporting of gonorrhoea
In England and Wales, all GUM clinics have a mandatory obligation to report diagnoses of sexually transmitted infections (STIs), including gonorrhoea. During the study the number of GUM clinics in England and Wales fluctuated between 227 and 229. Up to 2008 this information was collected on the KC60 statistical return.i The KC60 return provided aggregated data on all gonorrhoea diagnoses made each calendar quarter stratified by gender, age group and male sexual orientation. GUM clinics are open-access services providing specialist care for the diagnosis, treatment and management of STIs. Although other primary care providers such as general practitioners diagnose and treat STIs, national laboratory reporting data suggest that over 90% of men and almost 80% of women diagnosed with gonorrhoea in England and Wales receive their diagnosis in a GUM clinic.11
This study focused on gonococcal diagnoses made in GUM clinics only. Annual cross-sectional data from all patients diagnosed with gonorrhoea between 2000 and 2008 were extracted from GRASP and the KC60 reporting systems for analysis. Extraction of GRASP data was restricted to those patients diagnosed in a GUM clinic. Diagnoses of complicated gonorrhoea in the KC60 dataset do not include an age group breakdown and were excluded (however, these comprised fewer than 1% of cases in men and 5% of cases in women).
Demographic comparisons, 2000–2008
To examine the distribution of patient characteristics in the GRASP sample compared with national KC60 data, distributions of patient demographic characteristics as defined in the national surveillance systems were plotted and tabled. The distribution of reported diagnoses by gender, sexual orientation, age group and health region of report were compared between GRASP and KC60 surveillance data. The small number of patients with missing demographic data were excluded from stratified analyses but included in total population analyses.
Weighting for national distribution of patient characteristics
To estimate the impact of adjustment for the national distribution of relevant patient characteristics on population-level estimates of resistance prevalence from GRASP, weights to account for disproportionality between the achieved sample and the national distribution of patient characteristics reported through KC60 were calculated. A weight equal to the number of diagnoses reported in KC60 was assigned to every possible combination of year, health region, gender, sexual orientation category and age group. Weighted estimates of the percentage of diagnoses resistant to an antibiotic were calculated as a weighted average:
where wi is the stratum-specific weight and xi is the stratum-specific percentage of diagnoses resistant to an antibiotic (among those with a susceptibility test); and the summation is over all relevant strata.
Weighting for differential retrieval rates
A proportion of specimens submitted to the Sexually Transmitted Bacteria Reference Laboratory could not be retrieved for susceptibility testing because the organism was non-viable. Retrieval rates varied by submitting laboratory (and hence the associated GUM clinics). To estimate the impact of differential retrieval rates on resistance prevalence, weights to account for disproportionality between the achieved sample and the intended sample from laboratories were calculated. A weight equal to the number of diagnoses reported in GRASP was assigned to every combination of year and GUM clinic in the GRASP dataset. Weighted estimates of the percentage of diagnoses resistant to an antibiotic were calculated as above.
Graphs of trends in antimicrobial resistance prevalence comparing weighted and unweighted estimates were plotted. CIs were not presented as all analyses were based on the same antimicrobial resistance results and any differences between estimates were due to the weights, not sampling variation.
Comparisons of trends and patient demographics
Between 2000 and 2008, numbers of gonorrhoea diagnoses fell by 25% (20 899 to 15 631) for national KC60 reports, by 35% (9356 to 6088) for KC60 reports from participating GRASP clinics and by 49% (2478 to 1253) for GRASP reports (table 1; figure 2). In 2008, 37% (5852/15 631) of gonorrhoea diagnoses reported nationally were made in London, whereas for participating GRASP clinics, 46% (2790/6088) of diagnoses reported through KC60 surveillance and 47% (591/1253) of diagnoses reported through GRASP surveillance were from London (table 1).
Comparing KC60 reports in 2008 from all national clinics with those from GRASP clinics, 19% (2960/15 631) of gonorrhoea cases reported nationally were in men who have sex with men (MSM) compared with 25% (1515/6088) of those reported only from GRASP clinics, suggesting that the GRASP sample over-represented MSM (table 1). However, there was also evidence of some under-reporting of MSM in KC60 data, since the GRASP clinics reported a lower percentage of cases among MSM through KC60 surveillance (25%; 1515/6088) compared with GRASP surveillance (31%; 370/1184) (table 1). Also in 2008, 34% (5264/15 631) of cases reported nationally were women compared with 31% (1857/6088) of cases reported through KC60 surveillance from GRASP clinics and 30% (358/1184) of cases reported through GRASP surveillance, with a similar trend in heterosexual men with 47% (7407/15 631), 45% (2716/6088) and 39% (456/1184), respectively (table 1). There was no evidence of any seasonal variation in patient characteristics in the national KC60 reports: Combining all years' data, the percentage of cases aged <20 years in each calendar quarter ranged between 18.8% and 19.6%, while the percentage of cases in MSM ranged between 16.6% and 17.7%.
Weighting GRASP data according to the distribution of national patient characteristics reduced the proportion of male cases that were among MSM and slightly increased the proportion of total cases that were among women, but age distributions were similar (online supplementary figure 1).
Impact of weighting on resistance estimates
Weighting for national distributions of demographic characteristics reduced estimates of overall resistance prevalence for ciprofloxacin, penicillin (≥1 mg/l or β-lactamase+), Chromosomally-mediated resistant Neisseria gonorrhoeae (CMRNG), azithromycin and tetracycline (figure 3; online supplementary figure 2). The greatest effect was seen on the prevalence of ciprofloxacin resistance in 2006, reducing the estimate from 27% to 21% (figure 3). Emerging resistance to cefixime in 2008 was reduced from 1.5% to 1.0% as a result of weighting (figure 3). Weighting increased resistance prevalence estimates for PPNG (including or excluding Penicillinase-producing and tetracycline resistant Neisseria gonorrhoeae (PPNG-TRNG) very slightly in 2001 and 2005 (supplementary figure 2). In no case did weighting adjust resistance prevalence above or below the 5% threshold for any antimicrobial, including when the prevalence of ciprofloxacin resistance reached the 5% level in 2002.
Weighting for differential retrieval of specimens had little impact on estimates of resistance prevalence, for any antimicrobial in any year (figure 3; supplementary figure 2).
Understanding the coverage and representativeness of any disease surveillance system aids the interpretation of trends and delivery of robust public health advice. Surveillance based on the number of specimens or isolates submitted to laboratories for testing, although used in GRASP and by most gonococcal surveillance programmes, can be of limited value if submission patterns do not reflect underlying infection incidence.
In England and Wales the national mandatory (KC60) returns, which are not available in many countries, allowed us in this study to compare the characteristics of the GRASP patient sample with the national patient demographic. Compared with nationally reported cases of gonorrhoea in England and Wales, we showed that the GRASP sample over-represents cases diagnosed in London and those among MSM and slightly under-represents those in heterosexual men and women. Over-representation of MSM in GRASP was counterbalanced to some extent by some under-recording of men reporting sex with men in national surveillance, probably reflecting a tendency for some clinic software systems (which are used to generate national surveillance reports) to default to ‘heterosexual' orientation. However, the extent to which GRASP under-represents women may be greater than reported here: about 5% of cases in women, compared with only 1% in men, were excluded from the national (KC60) dataset before analysis as age group data are not collected for those with complicated infection. There was no evidence of any seasonal variation in patient characteristics in the national dataset, suggesting that restricting the GRASP collection to the summer months did not introduce further selection bias.
There was an observable if limited impact of the GRASP sampling approach on estimated resistance prevalence. Weighting for the national patient demographic tended to reduce estimates of resistance prevalence for most antimicrobials investigated. This was probably due, at least in part, to over-representation of diagnoses made in MSM in the GRASP sample relative to the national data, as the prevalence of resistance to most antimicrobials is higher among MSM. The greatest impact was for ciprofloxacin in 2006, where estimated prevalence was adjusted from 27% to 21% following weighting, although this would not have changed the recommendation by the GRASP steering group to change treatment guidelines. Although gonococcal infection is geographically clustered, the over-representation of diagnoses made in London probably had little impact on detecting emerging resistance in this study. However, the study highlights the potential for the GRASP sample to miss emerging resistance outside the capital and underestimate overall resistance prevalence.
Annual analyses of GRASP data have been weighted for variations in retrieval between laboratories since 2005.12 However in this study, surprisingly, we found this had negligible impact on resistance estimates even in years when retrieval rates were highly variable. This suggests that disproportionality in retrieval rates had a relatively minor influence on the case mix of the GRASP sample. As retrieval rates have improved in recent years, routine weighting for disproportionate specimen retrieval could probably be discontinued.
Improving the accuracy of prevalence estimates in emerging resistance, most often seen among MSM, might have greatest public health value. In this study for example, the overall prevalence of emerging resistance to cefixime was adjusted down from 1.5% to 1.0% following weighting for the national patient demographic, suggesting that cefixime could remain a useful drug for longer. However, while current guidelines advise switching recommended treatment once resistance levels have reached 5%,4 a single, generic policy may be neither desirable nor optimal economically,13 particularly when alternative treatments are lacking. Modelling work suggests that in high-risk populations such as MSM, resistance levels of 3% may warrant a change in treatment, while in populations where gonorrhoea prevalence is lower than 1%, continued use of an antimicrobial may continue to be viable even with resistance levels as high as 20%.13 There are many challenges facing surveillance programmes for antimicrobial resistance in gonorrhoea as treatment failure emerges to the current treatments,14 15 and therefore more aggressive, complex and tailored approaches to treatment will need to be developed and targeted at specific risk groups. Further empirical and mathematical modelling work is urgently needed to better understand the relationship between overall gonococcal prevalence, size and prevalence in population subgroups at risk and the impact of these parameters on optimal thresholds for switching treatments.
Given the data presented here, weighting of GRASP data for the national distribution of patients and patient characteristics should be adopted in future to ensure more accurate estimates of prevalence from GRASP, both overall and in high-risk groups such as MSM, and thereby better inform treatment strategies recommended in national guidelines. However, given the under-reporting of MSM in KC60 data and the importance of MSM in emerging resistance, improvements in the quality of data collected through national mandatory reporting of STIs from GUM clinics, especially sexual orientation, are needed. The recent replacement of KC60 surveillance in England with the GUM clinic activity dataset (GUMCAD)—an anonymised patient-level electronic surveillance system—should help facilitate improvements in the quality of national surveillance data as more robust checks are now imposed at data entry into clinic software.
The rapid decline in numbers of diagnoses reported through GRASP compared with national reporting is striking and probably reflects a reduction in specimens submitted to GRASP for culture. The collection of viable organisms for culture is necessary for the detection of emerging resistance. However, owing to their high sensitivity, reduced cost, use of non-invasive samples and simplicity, nucleic acid amplification tests (NAATs) are now routinely used to diagnose gonococcal infection. Increasingly, they are also being used to confirm the diagnosis in preference to culture.16 As the national data include all cases regardless of how confirmation was achieved, they will better reflect trends in underlying incidence and distribution of gonococcal infection in England and Wales. Clearly, as more clinics move towards confirmation with NAATs in the near future and as confirmation may not be required in high-prevalence populations, there is a risk that the GRASP sample could become increasingly biased towards symptomatic patients who are more likely to have samples taken for culture. If symptom presence varies by risk group (for example, MSM may be more likely to present with asymptomatic pharyngeal infections17), weighting the estimates of antimicrobial resistance prevalence in the GRASP sample for the national distribution of patient characteristics will therefore become increasingly important to help reduce bias. However, continuation of gonococcal culture methods is also vital for ensuring the early detection of emerging resistance.
This study indicates that the GRASP sample has provided sufficiently reliable estimates of the prevalence of resistance to antimicrobials across England and Wales to date, although the estimates may not be entirely representative of all gonorrhoea diagnoses. Neither of the weighted analyses reported here adjusted resistance estimates to a degree which would have resulted in a change to national recommendations, indicating that GRASP has informed the development of robust treatment guidelines. However, with emerging decreased susceptibility to cephalosporin and the inevitable challenges facing future treatment of gonorrhoea, the continuing need for robust, reliable measures of resistance, across the infected population and within risk groups, is clear.
The GRASP sample over-represents London and men who have sex with men (MSM) and under-represents women and heterosexual men but has to date provided reliable estimates of antimicrobial resistance prevalence to inform national treatment guidelines.
Weighting for the national distribution of patient characteristics could, in future, help reduce the bias in estimates of resistance prevalence both overall and in high-risk groups such as MSM.
Enhanced surveillance of resistance in specific risk groups such as MSM may also be needed to develop more complex and tailored approaches to treatment.
The authors would like to thank the remaining members of the GRASP Steering Group, Dr A Johnson, Prof G Kinghorn, Dr D Livermore, Dr R Mulla, Dr A Robinson, Prof J Ross, Dr J Wade, Dr C Bignell, Dr K Eastick and Dr J Paul and all members of the GRASP collaborative Group.
GRASP steering group
Professor Catherine Ison, Dr Catherine Lowndes, Ms Leah de Souza-Thomas, Dr Stephanie Chisholm, Mr John Anderson, Mrs Elisabeth Maclure-Webster, Dr Alan Johnson, Professor George Kinghorn, Dr David Livermore, Dr Rohinton Mulla, Mr Tom Nichols, Dr Angela Robinson, Professor Jonathan Ross, Dr Jim Wade, Dr Christopher Bignell, Dr Kirstine Eastick and Dr John Paul.
GRASP collaborative Group
EAST MIDLANDS: Dr M Minassian, Dr L Riddell, Dr V Weston, Dr C Bignell; EAST of ENGLAND: Dr M Farrington, Dr C Carne, Dr R Mulla, Dr T Balachandran. LONDON: Dr B Azadian, Dr K McLean, Dr A McOwan, Dr F Boag, John Hunter Clinic, Dr D Krahe, Dr M Nathan, Ms M Graver, Dr M Tenant-Flowers, Dr R Holliman, Dr P Hay, Dr A Jepson, Dr L Green, Dr B MacRae, Dr A Robinson, Dr S Bragman, Dr J Russell, Dr J Wade, Dr P Riley, Dr E Jungmann NORTH EAST: Prof. J Magee, Dr K N Sankar NORTH WEST: Dr T Neal, Dr P Carey, Dr A Qamruddin, Dr A Sukthankar, Dr J Anson SOUTH EAST: Dr J Paul, Dr G Dean, Dr A Stacey, Dr G Wildman, Dr M Cubbon, Dr A Tang SOUTH WEST: Dr R Spencer, Dr P Horner, Dr M Logan, Dr Z Sulaiman, Dr M Williams, Dr A Lees WALES: Dr R Howe, Dr H Birley. Dr E Kubriak, Dr R Das. WEST MIDLANDS Dr M Gill, Dr J Ross, H Jones, Dr J Gray, Dr D Dobie, Dr A Tariq YORKSHIRE & HUMBERSIDE: Dr M Denton, Dr J Clarke, Dr P Zadik, Dr G Kinghorn.
Funding GRASP has been funded totally (2000–2004) and partially (2005–2010) by the Department of Health (England). The views expressed in the publication are those of the authors and are not necessarily those of the Department of Health.
Competing interests None.
Ethics approval GRASP and KC60 are national public health surveillance systems run by the Health Protection Agency (HPA). The HPA has permission to handle these data under section 251 of the NHS Act 2006 (previously section 60 of the Health and Social Care Act 2001), renewed annually by the Ethics and Confidentiality Committee of the National information Governance Board. When GRASP was first established, ethics permission was obtained from local regional research committees and from the North West multicentre research ethics committee. Patients were informed of the study at the participating site through written notices.
Provenance and peer review Not commissioned; externally peer reviewed.
↵i Since 2009 in England this information has been collected in the genitourinary medicine clinic activity dataset (GUMCAD).