Article Text
Abstract
Background There is increasing evidence that azithromycin 1 g is driving the emergence of macrolide resistance in Mycoplasma genitalium worldwide. We undertook a meta-analysis of M. genitalium treatment studies using azithromycin 1 g single dose and azithromycin 500 mg on day 1 then 250 mg daily for 4 days (5-day regimen) to determine rates of treatment failure and resistance in both regimens.
Methods The online databases PubMed and Medline were searched using terms “Mycoplasma genitalium”, “macrolide” or “azithromycin” and “resistance” up to April 2016. Studies were eligible if they: used azithromycin 1 g or 5 days, assessed patients for macrolide resistant genetic mutations prior to treatment and patients who failed were again resistance genotyped. Random effects meta-analysis was used to estimate failure and resistance rates.
Results Eight studies were identified totalling 435 patients of whom 82 (18.9%) had received the 5-day regimen. The random effects pooled rate of treatment failure and development of macrolide antimicrobial resistance mutations with azithromycin 1 g was 13.9% (95% CI 7.7% to 20.1%) and 12.0% (7.1% to 16.9%), respectively. Of individuals treated with the 5-day regimen, with no prior doxycycline treatment, fewer (3.7%; 95% CI 0.8% to 10.3%, p=0.012) failed treatment, all of whom developed resistance (p=0.027).
Conclusion Azithromycin 1 g is associated with high rates of treatment failure and development of macrolide resistance in M. genitalium infection with no pre-existing macrolide mutations. There is moderate but conflicting evidence that the 5-day regimen may be more effective and less likely to cause resistance.
- Mycoplasma genitalium
- Azithromycin
- Meta-analysis
- Drug resistance
- Doxycycline
- Mutation
Statistics from Altmetric.com
Mycoplasma genitalium (MG) is an important, emerging STI which causes 15%–25% of acute non-gonococcal urethritis (NGU) in men and probably causes cervicitis, pelvic inflammatory disease, spontaneous abortion, preterm birth and tubal factor infertility in women.1–3
It is a relatively common anogenital infection, with the majority of those infected being asymptomatic.1 3 About 1%–3% of men and women in the general population are infected.4–7 MG is detected more often in those infected with chlamydia, with 3%–9% testing positive.4–9
Effective management strategies have been hampered not only by the lack of commercially available nucleic acid amplification test (NAAT) assays which have been evaluated to US Food and Drug Administration approval standard combined with only limited validation for those assays which have been CE marked,1 2 but also by the poor treatment efficacy in eradicating MG with doxycycline 100 mg twice a day for 7 days, or azithromycin 1 g single dose which are the current first-line treatments for NGU and cervicitis in the UK and USA.2 10 Failure to eradicate MG with 7 days doxycycline 100 mg twice a day occurs in >50% of cases and in 0%–60% of cases treated with azithromycin 1 g.1 11–13 Failure rates with azithromycin 1 g have increased with time, and a recent randomised controlled trial found no significant difference in treatment failure between azithromycin 1 g and doxycycline 100 mg twice a day for 7 days (40% vs 30%).13 This is probably due to the emergence of macrolide antimicrobial resistance in MG, worldwide.1 11 12 14 A number of experts believe this may be a consequence of extensive use of azithromycin 1 g for the treatment of STIs, as this regimen has been demonstrated to cause drug resistance in treatment failures.1 2 11 An extended 5-day regimen of azithromycin 500 mg on day 1 then 250 mg daily for 4 days was introduced in the 1990s for the treatment of MG and has been demonstrated by two groups to have high efficacy (>95%) and until recently had not been associated with macrolide resistance.15–17
There was no information on risk of developing macrolide drug resistance with either azithromycin 1 g or 1.5 g extended regimen in wild-type infections when compiling the 2015 UK NGU management guidelines.18 Expert opinion was divided about discontinuing azithromycin 1 g as first-line treatment in favour of doxycycline 100 mg twice daily for 7 days and/or changing to the azithromycin extended 5-day regimen for first-line treatment. There are no randomised trials comparing azithromycin 1 g to the 5-day regimen and the evidence supporting such a change was considered weak and no recommendation was made. A recent review by Manhart et al noted that although the evidence was suggestive of slower emergence of resistance with the extended dose, the evidence was weak and further data were needed.13 Undertaking a randomised controlled trial to demonstrate, this would take 5–7 years to publish its findings and would be complicated by the high rates of MG macrolide antimicrobial resistance observed worldwide (13%–100%).3 We therefore undertook a review of the literature to determine treatment failure and macrolide resistance rates for patients without pretreatment macrolide-resistant genotype infections, using both the single dose azithromycin 1 g regimen and a 5-day 1.5 g regimen. We also compared the treatment efficacies of both treatment regimens.
Methods
Search strategy
We undertook a review of the literature using the electronic online databases PubMed and Medline to identify published articles including the search terms ‘Mycoplasa genitalum’ AND (‘macrolide’ OR ‘azithromycin’) AND ‘resistance’ up to April 2016. Eligible studies were English Language prospective or retrospective treatment studies using azithromycin 1 g or 500 mg on day 1 then 250 mg daily for 4 days, in which patients who failed treatment were assessed for MG macrolide-resistant genetic mutations prior to and after treatment. Both men and women were included. Women and men were combined for the purposes of the analyses. Macrolide antimicrobial resistance was defined as MG with a known macrolide resistance genotype involving a mutation at the nucleotide position 2058 (2071) and/or 2059 (2072) in the 23S rRNA gene.12 19 Patients were excluded from the primary analysis if there was no information on the pretreatment specimen 23S rRNA macrolide genotype or if they had been pretreated with doxycycline. A secondary analysis was performed to include patients pretreated with doxycycline as many patients will have received this as first-line treatment for symptomatic NGU in Europe, prior to extended azithromycin.2
Data extraction and outcome
Identified studies were reviewed by KB, FG and PH and those meeting the inclusion criteria selected. Data were extracted by PH and FG and reviewed by SI for rates of failure and rates of macrolide resistance after treatment for MG with a course of azithromycin in patients with macrolide-susceptible pretreatment samples. These were defined as the percentage of individuals who returned for a test of cure who received a positive MG NAAT, and of those who were still positive, the percentage who were found to have a macrolide resistance mutation in the 23S rRNA gene, respectively.
Analysis
We combined the data to determine the absolute rates of treatment failure and development of resistance of the two regimens. Rates of failure and resistance to the 1 g regimen were quite variable across the different studies and therefore, we next took a meta-analytic approach. With this, approach we calculated the I2 statistics to assess the percentage of variability in treatment failure and development of antimicrobial resistance estimates that could be attributed to underlying study heterogeneity rather than chance alone. If I2 >25% random effects meta-analysis was used to estimate the pooled rate of macrolide resistance development and if I2 <25% fixed effects meta-analysis.11 20 Both fixed effects and random effects meta-analysis were undertaken.
We compared the absolute rates of treatment failure and development of resistance between the two regimens. We then assessed the differences in these rates and calculated the associated 95% CI and p values for the difference using Fisher’s exact test. As there were only two studies which had both data on the 1 g and 5-day regimens, we performed a sensitivity analysis where we only considered the difference in rates between the two regimens in these two studies. This sensitivity analysis helps to avoid the problem of heterogeneity across the studies and provides a more conservative estimate of the difference between the two regimens. We also undertook a separate sensitivity analysis including the additional patients treated with the 5-day regimen who had been pretreated with doxycycline as it is possible prior treatment could have an effect on azithromycin efficacy. There were insufficient numbers to separately examine whether gender, age, symptoms or study type affected treatment outcome.
Data were analysed using Stata V.13.1.
Results
Study selection and characteristics
The review process is shown in figure 1. Seventy-nine papers were identified and eight treatment studies met the inclusion criteria (table 1).15 17 19 21–26 Five were studies of patients tested for MG who were unselected or had NGU, cervicitis and/or pelvic inflammatory disease (PID) and/or were sexual contacts of infected partners and treated with azithromycin.15 19 21–23 One was a study of a random sample of men with NGU treated with azithromycin 1 g who were identified as MG positive.24 25 One was a female only study designed to test incidence, organism load and treatment failure after treatment with azithromycin 1 g.26 One was a prospective longitudinal cohort study comprising an observational study and a randomised treatment trial involving both men and women.17 In only three studies was there information on the use of the extended azithromycin regimen 500 mg then 250 mg daily 4 days.15 17 22
Azithromycin efficacy
Four hundred and thirty-five individuals were identified of whom eighty-two (18.9%) had been treated with the extended 5-day regimen and 353 (81.1%) were treated with azithromycin 1 g (table 1). Of individuals treated with azithromycin 1 g, 47 (13.3%, 95% CI 9.9% to 17.3%) remained MG positive and of those 43 (91.5%, 95% CI 80.0% to 97.6%) had a detectable mutation consistent with macrolide antimicrobial resistance. Of 82 individuals treated with the 5-day regimen, only 3 (3.7%, 95% CI 0.8% to 10.3%) failed and had a detectable mutation consistent with macrolide antimicrobial developed resistance, which gave a difference in failure and resistance rate compared with 1 g of 9.7% (95% CI 4.3% to 15.0%, p=0.012) and 8.5% (95% CI 3.2% to 13.8%, p=0.027).
The forest plot for treatment failure and antimicrobial resistance with azithromycin 1 g is shown in figure 2A,B, respectively, with details of the fixed effect and random effects meta-analyses. The I2 value was 60.5% and 41.5%, respectively, which indicates moderate heterogeneity between effect estimates from different studies. The random effects pooled rate of treatment failure and development of macrolide antimicrobial resistance mutation(s) with azithromycin 1 g was 13.9% (7.7% to 20.1%) and 12.0% (7.1% to 16.9%), respectively.
Of those with pretreatment macrolide resistance mutations 1 of 1 failed the 5-day regimen15 and 49 of 56,23 0 of 2,26 12 of 16,21 6 of 6,19 0 of 125 and 8 of 1017 failed azithromycin 1 g. Thus, 76 of 92 (82.6%, 95% CI 73.3% to 89.7%) individuals with a macrolide resistance mutation failed treatment with azithromycin.
Sensitivity analyses
We calculated the differences restricted to the Anagrius et al and Falk et al studies and including the three men who had previously been excluded due to missing or inconclusive pretreatment samples.15 17 This gave a difference in failure and resistance rate of 4.4% (95% CI −0.02% to 11.2%, p=0.38) for 1 g compared with the 5-day regimen.
We also undertook a separate analysis including the additional 56 patients treated with the 5-day regimen who had been pretreated with doxycycline. Of 138 individuals treated with the 5-day regimen, no additional failures were observed. Thus, 3 (2.2%, 95% CI 0.5% to 6.2%) failed and developed resistance, an 11.1% (95% CI 6.8% to 15.4%, p=0.0001) difference in failure rate and a 10.0% (95% CI 5.8% to 14.2%, p=0.0002) difference in macrolide resistance rate compared with the 1 g regimen.
Discussion
This meta-analysis provides good evidence that an azithromycin 1 g regimen is associated with rates of failure of 13.9% (7.7% to 20.1%), and of macrolide resistance of 12.0% (7.1% to 16.9%) in MG urogenital infection in which no pre-existing macrolide resistance mutations are present. Although fewer treatment failures were observed with the 5-day regimen in individuals with no prior doxycycline treatment, data from only three studies was available.15 17 22 When combining the data in a crude way, this equates to a difference in failure rate of 9.7% (95% CI 4.3% to 15.0%, p=0.012) and a difference in resistance rate of 8.5% (95% CI 3.2% to 13.8%, p=0.027), which provides moderate evidence of a difference. Individuals with a pre-existing macrolide 23S rRNA gene mutation had a treatment failure rate of 82.6% (95% CI 73.6% to 89%). These data are consistent with the hypothesis that the use of azithromycin 1 g is driving the increase in prevalence of macrolide-resistant MG genotypes.11 12
This is the first meta-analysis of the literature on azithromycin 1 g treatment studies of MG to include only individuals in whom there was no evidence of a macrolide resistance genotype prior to treatment, and hence determine the rate of resistance in fully sensitive isolates. Eight studies of whom only four were prospective17 22 23 26 were identified and none involved a direct comparison of the two regimens. The latest a person could be included as a reattendance ranged from 26 to 28 days for the four prospective studies17 22 23 26 but varied considerably in the four retrospective studies with a range of 28–364 days.15 19 21 25 In the four prospective studies, the loss to follow-up ranged from 3.1% to 21.9%17 22 23 26 which was in general lower than the four retrospective studies with a range of 12.7%–27.8% in two studies and unknown in the other two.15 19 21 25 As only those individuals who returned for a repeat test were included in this meta-analysis there are two potential biases. First, it is conceivable that a higher proportion of treatment failures were included in the outcomes, because it is likely, particularly in the retrospective studies, that patients successfully treated are less likely to return. Although Ito et al observed that three out of seven men who retested positive were asymptomatic.25 Second, with a longer duration of follow-up it is possible that some patients who retested positive were actually reinfections. Only patients with no previous doxycycline treatment were included in the primary analysis as doxycycline could reduce the MG load and thus the potential for pre-existing micro-organisms containing macrolide resistance mutations.12 14 27 Only three studies were identified with the 5-day regimen which precluded a rigorous meta-analysis of the efficacy of this regimen.15 17 22 Finally, some individuals who failed treatment and were macrolide resistant did not have suitable pretreatment specimens for resistance testing and were excluded from the analysis. As a sensitivity analysis, we calculated the differences restricted to the Anagrius et al and Falk et al studies and including the three men who had previously been excluded due to missing or inconclusive pretreatment samples.15 17 The study by Anagrius et al was undertaken in specimens collected from 1998 to 2005. In 2006, no macrolide resistance mutations were detected in any MG-positive patients.17 Including these individuals and comparing the treatment regimens in the only two studies which contained data on both the 1 g and 5-day regimens also demonstrated a reduction in failure and resistance rates of 4.4% (95% CI −0.02% to 11.2%, p=0.38).
Lau et al recently undertook a systematic review and meta-analysis of azithromycin 1 g treatment studies and demonstrated increasing failure rates which they attributed to emerging macrolide resistance over time.11 Notably prior to 2009, the pooled treatment efficacy was 85.3% (82.3%–88.3%), similar to 86.1% we observed in individuals with no prior macrolide resistance and 67% (57.0%–76.9%) after 2009. A previous study by Bjornelius et al demonstrated high efficacy (96%) of the extended 5-day azithromycin regimen and similar failure rate (14%) of azithromycin 1 g to the pooled efficacy from our meta-analysis and that observed in all studies prior 2009.11 16 However, all patients receiving the 1 g azithromycin regimen were Norwegian. Although no information on antimicrobial resistance is available, the study was undertaken in Norway and Sweden between 2002 and 2004. No macrolide resistance was identified in Sweden in 2006 but Norway had a 20%–30% macrolide failure rate in 2005–2006, which may explain the failure rate in the Norwegian patients.15 28 Two subsequent observational studies, which did not test for macrolide resistance prior to treatment, one from Norway undertaken 2005–200628 and one from Australia 2009–201329 did not observe any difference in treatment efficacy between the azithromycin 1 g and an extended 5-day regimen, 79% and 70%, 67% and 74%, respectively. However, these findings could be explained by the presence of macrolide-resistant micro-organisms in the population prior to treatment; the efficacy of both regimens is affected in macrolide-resistant isolates.1 28 Recently, Read et al in a retrospective observational study found that 4/34 (12%, 95% CI 3% to 27%) patients with wild-type infection had macrolide post-treatment mutations with the azithromycin-extended 5-day regimen.30 The duration of follow-up was median 36 days (14–100 days) with a loss to follow-up of 63 (37%) of 169 men. Although they remark the macrolide mutation rate was similar to a historical control treated with azithromycin 1 g and it is similar to the 12.0% from this meta-analysis, there are a number of possible explanations for the apparent conflict with our findings. The confidence limits are wide and within the 95% CI for the macrolide postmutation rate observed in this analysis 3.7% (95% CI 0.8% to 10.3%) and may just be a chance observation related to the small size of the study. In addition, given the high loss to follow-up, it is possible that the actual macrolide post-treatment mutation rate was lower. Third, macrolide resistance was associated with men who have sex with men (MSM). This group was highly sexually active with a median of 3 (2–6) sexual partners in the last 3 months prior to treatment. Seventy-six per cent of MG isolates were macrolide resistant in the local MSM population, and although review of the notes suggested that reinfection was not the cause this remains a possibility as it was not a prospective study. Finally, an alternative but not mutually exclusive explanation is that in populations with high level of resistance mixed infections, not detected with the current methodologies, would tend to contribute to apparent resistance development. Taken together, the conclusion of Read et al that the extended azithromycin 1.5 g was no more effective than a single 1 g dose should be viewed with caution and highlights the need for further high-quality prospective studies using this regimen.30 When the data from the Read study is added to this study and combined in a crude way, no difference is observed between azithromycin 1 g and the extended regimen in post-treatment macrolide mutations (p=0.09) but remains significantly different when patients with prior doxycycline are included (p=0.005).
Interestingly, even in the presence of macrolide antimicrobial resistance our findings suggest 17% of individuals will test negative following treatment with azithromycin. This might be explained by spontaneous clearance6 31–33 as a result of the adaptive immune response34 35 or other factors including either a temporary suppression of the infection and/or spontaneous fluctuations in the MG load which could lead to DNA shedding below the limit of detection, resulting in false negatives.17
This study provides only moderate evidence that fewer patients with MG urogenital infection will fail treatment and develop macrolide resistance with the extended 5-day regimen compared with azithromycin 1 g which conflicts with the recent findings of Read et al.30 Caution needs to be continued to be exercised when using the extended regimen as a replacement for azithromycin 1 g and all patients followed up to establish that they have been cured as some patients will still fail treatment either due to pre-existent macrolide resistance or to its emergence following treatment and may be asymptomatic.1–3 25 Interestingly, when the 56 patients with prior doxycycline treatment who were then treated with the extended regimen are considered, the superiority of the 5-day regimen is more pronounced with an 11.1% (95% CI 6.8% to 15.4%, p=0.0001) difference in failure rate and a 10.0% (95% CI 5.8% to 14.2%, p=0.0002) difference in macrolide resistance rate compared with the 1 g regimen. This is probably as a result of a reduced micro-organism load following doxycycline treatment which will bias the comparison of this group with azithromycin 1 g.12 13 27 Nevertheless, it provides strong support for the use of doxycycline 100 mg twice daily for 7 days as first-line treatment in men with NGU, in whom MG is a common pathogen, with the extended azithromycin reserved for those who fail treatment.2 There are no treatment studies on using other higher dose extended regimens such as azithromycin 1 g then 250 mg once a day for 4 days, and it is not possible to comment on whether other extended regimens are more efficacious than 500 mg then 250 mg once a day for 4 days.2 Further prospective studies comparing the extended azithromycin regimen with single dose and azithromycin treatment following doxycycline are urgently needed.
Key messages
There are no randomised controlled trials comparing the development of macrolide resistance in Mycoplasma genitalium infection between azithromycin 1 g with azithromycin 1.5 g over 5 days.
Azithromycin 1 g treatment is associated with 13.9% (7.7% to 20.1%) rate of failure and 12.0% (7.1% to 16.9%) risk of macrolide antimicrobial resistance.
There is moderate but conflicting evidence that the 5-day regimen may be more effective and less likely to cause resistance.
The difference in failure and resistance rates is 9.7% (95% CI 4.3% to 15.0%, p=0.012) and 8.5% (95% CI 3.2% to 13.8%, p=0.027), respectively.
Acknowledgments
We would like to thank Lars Falk,17 Dione Gesink22 and C Sarai Racey22 for providing additional data from their studies.
References
Footnotes
Contributors PH conceived the study idea. PH, FG and KB undertook the literature searches and FG KB and PH reviewed the records for eligibility. PH and FG extracted the data which was checked by SI. SI undertook the statistical analyses. All authors contributed to the data interpretation. PH wrote the first draft with input from FG, HM and SI and revised subsequent drafts following critical review by all authors.
Funding This work was supported by the National Institute of Health Research Health Protection Research Unit (NIHR HPRU) in Evaluation of Interventions at the University of Bristol in partnership with Public Health England (PHE).
Disclaimer The views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR HPRU, the Department of Health or PHE.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement There is no unpublished data from this meta-analysis, however, should someone wish to discuss the results in more detail please email the corresponding author.