Patients attending MSHC who have nongonococcal urethritis, pelvic inflammatory disease, cervicitis, or proctitis are routinely tested for M. genitalium, as are their sex partners. From August 2012 through November 2014, patients for whom azithromycin and moxifloxacin failed and for whom no other treatment options were available were treated with pristinamycin at a dose of 1 g 4 times/day for 10 days. Because of promising preliminary results and side effects data, in December 2014, pristinamycin became a second-line treatment after azithromycin failure. Two other pristinamycin regimens were evaluated during the study period: 1 g 2 times/day for 10 days and 1 g 3 times/day in combination with doxycycline. The pristinamycin/doxycycline combination was used based on evidence indicating this combination was effective in treating methicillin-resistant Staphylococcus aureus and theoretical and empirical evidence supporting antimicrobial drug combinations in several infections (24,25).

Patients receiving pristinamycin from MSHC’s pharmacy were prospectively followed by a research nurse. The nurse extracted the following information and recorded it in a database: results of tests of cure, adverse effects, and posttreatment sex with an untreated or inadequately treated partner.

Patients were routinely advised to return for a test of cure 2–4 weeks after treatment. Microbiological cure was defined by a negative test for M. genitalium 14–90 days after start of treatment. Patients reporting posttreatment sex with an untreated partner (where the relationship preceded treatment) were excluded regardless of the result of their test of cure because of their high risk for reinfection. Patients who reported sex with treated partners or new partners were retained in the analysis and coded as being at risk for reinfection so that this information could be analyzed as a risk factor. The Alfred Hospital Ethics Committee approved the study (project 490/16).

During August 2, 2012–February 9, 2016, a total of 133 patients were treated with pristinamycin for M. genitalium infection. Patients were excluded from further study for 4 reasons: no test of cure (n = 6), test of cure <14 days from start of treatment (n = 4), test of cure >90 days from start of treatment (n = 5), and sex with untreated partners (n = 4) (Figure 1). The analysis comprised the remaining 114 patients. These patients had been unsuccessfully treated with >1 of the following antimicrobial drugs before pristinamycin: single-dose azithromycin (1 g; 34 [30%] patients); azithromycin 1.5 g (500 mg, then 250 mg/d for 4 d; 76 [67%]); moxifloxacin (400 mg/d for 10 d; 19 [17%]); or doxycycline (100 mg 2×/d for 7 d; 19 [17%]). Twenty-five (22%) patients had been treated with >2 courses of antimicrobial therapy. Four (4%) received pristinamycin as initial treatment because their partners had a resistant infection.

Of the 114 patients in the analysis, data were available for 99 for analysis of bacterial load (before and after pristinamycin) and 74 for genetic sequencing of >1 region of interest. Median time to test of cure was 30 days (interquartile range 23–41).

Our investigation found that pristinamycin cured M. genitalium infections in 75% of patients who were unsuccessfully treated with azithromycin; in 17% of this population, moxifloxacin had failed to cure infection. The effectiveness of pristinamycin was the same when given at a dose of 1 g 4 times/day as when given as 1 g 3 times/day in combination with doxycycline at 100 mg 2 times/day. Higher pretreatment bacterial load was associated with treatment failure; the odds of failure increased almost 2-fold with each 1 log₁₀ increase in load. Failed treatment reduced bacterial load by a mean 1.3 log₁₀. Almost all of this population had macrolide-resistant infections, precluding assessment of the effect of any 23S ribosomal gene mutation on treatment response, but the site of MRM (2058 vs. 2059) and ribosomal protein gene mutations were not associated with treatment outcomes.

Pristinamycin was well-tolerated with a low rate (7%) of reported side effects, but in this large case series, it did not perform better than moxifloxacin (89% in a recent meta-analysis) for macrolide-resistant M. genitalium infection (28). Given the lack of alternatives, pristinamycin remains an option during pregnancy and in other situations where fluoroquinolones have failed or are contraindicated.

Our study has some limitations. Because it was an evaluation in a clinical service and not a trial with controls, the documentation of reinfection risk and treatment adherence is not as complete. Nevertheless, because previous treatments had failed and pristinamycin is not approved to treat M. genitalium infections, sexual health physicians at MSHC were asked to be vigilant for these factors, and when documented (60%–80% of records), neither reinfection risk nor treatment adherence proved significant. Further limitations are that the combination of doxycycline with the 3 g/day dose and the few patients who received 2 g/day limit our power to compare the efficacies of different doses.

The only other report of the efficacy of pristinamycin in patients with M. genitalium infection was a series of 6 patients at MSHC in whom azithromycin and moxifloxacin treatment had failed (12). These patients were treated with pristinamycin at a dose of 1 g 4 times/day for 10 days at MSHC in 2013, and all 6 infections were cured (100% [97.5% 1-sided CI 54%–100%]). In contrast, infections were cured in 75% of the 51 patients in our study who received this regimen. The initial 6 cures are consistent with 25% therapy failure in our study when the CIs are considered.

Pristinamycin is listed as a third choice in the 2016 European guideline on M. genitalium, and our findings do not encourage any stronger recommendation (5). No oral alternatives to pristinamycin are available, but 1 case of M. genitalium was cured by 1 week of daily injected spectinomycin, which, if available, is a less convenient option (29).

Given initial treatment successes with pristinamycin and promising in vitro data, we did not expect treatment to fail in 25% of macrolide-resistant M. genitalium infections. The reasons for this percentage of failures are not clear and did not appear related to adherence or reinfection, but higher bacterial load was strongly associated with pristinamycin failure. This finding has been observed in studies of the closely related macrolides, which show that higher load infections are more likely to fail treatment and lead to detection of resistance mutations (11,30). Higher load infections might be more likely to harbor heterotypic resistance within a subpopulation of organisms.

Antimicrobial susceptibility testing at the Statens Serum Institute (Copenhagen, Denmark; J.S. Jensen, unpub. data) indicated for both macrolide-resistant (n = 17) and susceptible (n = 23) M. genitalium strains, MICs were all susceptible (defined as <1 mg/L) (31). However, these MICs were close to the expected breakpoint, so we used the maximum dose of 1 g 4 times/day for 10 days. Macrolide-resistant strains had significantly higher MIC₉₀ (0.50 mg/L) for pristinamycin than susceptible strains (MIC₉₀ 0.125 mg/L; p = 0.003) but remained within the susceptible range (32). MRM are known to confer resistance to the streptogramin B component of pristinamycin (18), and mutations in position 2058 in M. genitalium have been associated with a higher MIC against macrolides, such as solithromycin (33), and in M. pneumoniae with erythromycin (34). Collectively, our data showing all 6 patients with no 23S MRM were cured, the MIC data, and the finding that infections were cured in 58% of patients with 2058 mutations, compared with 73% with 2059 mutations (p = 0.26), may indicate that 23S mutations influence the efficacy of pristinamycin and that there may be a differential effect of 2058 versus 2059 mutations in M. genitalium; however, this possibility requires further study.

Of the L4 and L22 ribosomal gene mutations present, only 1 of each resulted in an amino acid change. The L4 mutations detected were not close to the L4 loop near position 69 (M. genitalium numbering, position 66 E. coli numbering) known in other bacteria to cause macrolide resistance. By in vitro selection of resistance with subinhibitory concentrations of pristinamycin in the closely related M. pneumoniae, variable numbers of G insertions in position G60 (G59 in M. genitalium) was the only mechanism identified for pristinamycin resistance. These insertions occurred after a high number of passages suggesting a low potential for selection of resistance with pristinamycin (20). We identified no similar mutations in our dataset. The combination of doxycycline and pristinamycin was also no more effective than pristinamycin monotherapy, despite in vitro data suggesting a small additional effect of the combination, although not true synergy (J.S. Jensen, unpub. data).

We do not have an effective, safe, inexpensive oral drug for treating M. genitalium now that macrolide resistance is so prevalent. Fluoroquinolones are potentially toxic and vulnerable to increasing resistance. As our results indicate, pristinamycin cures 75% of macrolide-resistant infections with a course of up to 80 tablets; its price and availability vary around the world, but it may play a role where quinolones are contraindicated or ineffective. Two findings are consistent across several studies: 1) infections with lower bacterial loads are more likely to be cured (even in the presence of resistance) and 2) failed treatment with pristinamycin and azithromycin reduces bacterial load (11). These data highlight the urgent need for further work to determine the activity of new and existing antimicrobial drugs, and combinations of antimicrobial drugs, against this neglected infection.