Estimation of Antimicrobial-Resistant Mycoplasma genitalium Infections in People Aged 14–59 Years: Results From the National Health and Nutrition Examination Survey, United States, 2017–2018

Abstract

Mycoplasma genitalium antimicrobial resistance is common in US sexual health clinics. Population-level prevalence is unknown. We tested National Health and Nutrition Examination Survey (2017–2018) specimens for M genitalium resistance mutations (n = 58). Macrolide resistance prevalence was 42.7% (95% CI, 24.4%–62.7%). Quinolone resistance was not detected. Macrolide resistance testing should guide azithromycin treatment.

Mycoplasma genitalium (MG) is on the Centers for Disease Control and Prevention's antimicrobial resistance (AMR) threats report watch list, indicating potentially significant treatment challenges [1]. MG causes male urethritis [2] and is associated with pelvic inflammatory disease, female infertility, and preterm delivery [3]. Antimicrobials with efficacy against MG are limited and AMR occurs frequently.

MG lacks a cell wall, making antibiotics that target peptidoglycan inherently ineffective. Doxycycline has low efficacy (30%–45%) [4]. Azithromycin was previously recommended as first-line therapy, but macrolide resistance has rapidly spread [5]. In 2020, 59% of MG infections among sexual health clinic (SHC) attendees in the United States were macrolide resistant [6]. Given this high prevalence of macrolide resistance, resistance-guided therapy was recommended for MG infections in the 2021 update to the Centers for Disease Control and Prevention’s sexually transmitted infection (STI) treatment guidelines [7]. Under resistance-guided therapy, high-dose azithromycin is used for macrolide-susceptible infections, and moxifloxacin is used for macrolide-resistant infections or when macrolide susceptibility is unknown. However, the prevalence of quinolone resistance–associated mutations (QRAMs) in the parC gene linked with moxifloxacin treatment failure [8] in US SHCs was 6.2% in 2020 (L.E.M., unpublished data). Additional mutations in the gyrA gene may enhance the likelihood of moxifloxacin treatment failure [8]; these have been rarely reported in US populations.

Most US data on MG AMR come from SHCs, where the prevalence of STIs and AMR is higher than in other settings. The extent of MG AMR in the general population is unknown. We tested MG-positive urine specimens from the National Health and Nutrition Examination Survey (NHANES; 2017–2018) to estimate population-level prevalence and correlates of MG AMR.

METHODS

NHANES uses a clustered sampling strategy [9], collecting sociodemographic and behavioral data during household interviews in 2-year cycles. During a mobile examination center (MEC) health examination, sensitive questions are asked via a computer-assisted personal interview system, and urine and blood samples are collected. In NHANES 2017–2018, nucleic acid amplification testing in urine specimens from participants aged 14 to 59 years was performed using the Aptima® Mycoplasma genitalium Assay (Hologic, Inc).

Of 63 pristine MG-positive urine specimens (never thawed/refrozen), 58 (92%) were available for AMR testing. MG nucleic acid was extracted from 1 mL of thawed Aptima transport medium and eluted in 0.1 mL on a Roche MagNaPure 96 instrument. Macrolide resistance mutations (MRMs) in the 23S rRNA gene were detected via a modification of the published method employing a reverse transcriptase polymerase chain reaction, followed by PyroMark24 sequencing [10]. To detect QRAMs, conventional polymerase chain reaction assays amplified the parC and gyrA genes, followed by Sanger sequencing [10]. Specimens not amplified were reanalyzed using double the amount of extracted nucleic acid template. MRMs included A2058C/G/T and A2059C/G/T; ParC QRAMs included D87H/N/Y and S83I/R; and GyrA QRAMs included M95I, D99N, and D99G.

We accessed limited access NHANES data through the National Center for Health Statistics Research Data Center and estimated weighted prevalence of MRM/QRAMs overall and for sociodemographic, behavioral, and clinical subgroups. We calculated exact binomial 95% CIs using Korn-Graubard and Wald methods, suppressing estimates based on cell sizes <5. We used modified Poisson regression with a log link to estimate univariable weighted prevalence ratios (PRs). Small numbers precluded adjusted analyses. The flow diagram in the supplementary figure outlines the relationship between the underlying NHANES 2017–2018 sample and the specimens for which we have AMR results. To assess the representativeness of the specimens that we tested, we compared the weighted distribution (using MEC weights) by age group, sex, and race/Hispanic origin between the entire MG-positive sample (n = 67) and the sample with macrolide resistance results (n = 49). Age group and sex distributions differed by ≤2%. Differences by race/Hispanic origin were larger (eg, 53% of the 49 were non-Hispanic White vs 47% of the total 67 in the MG-positive sample), but the demographic composition was similar enough to proceed with analysis using MEC weights [9]. After estimating subgroup AMR prevalence, we multiplied the 2017–2018 sex- and age-specific totals from the American Community Survey conducted by the US Census Bureau [11] by the weighted prevalence of MG AMR in each category and summed these totals to estimate the absolute number of resistant MG infections in the civilian, noninstitutionalized population aged 14 to 59 years.

We used R version 4.4.3 (R Foundation for Statistical Computing) and the “survey” package for statistical analyses. NHANES procedures were approved by the National Center for Health Statistics Ethics Review Board (protocols 2011-17 and 2018-01). These analyses were exempt from University of Washington Institutional Review Board review.

RESULTS

In original analyses, MG prevalence was 1.7% (95% CI, 1.1%–2.7%) [12]. Of 58 MG-positive specimens, 49 (84.5%), 41 (70.7%), and 42 (72.4%) yielded MRM, ParC QRAM, and GyrA QRAM results, respectively. Weighted MRM prevalence was 42.7% (95% CI, 24.4%–62.7%), equating to ∼1.4 million (95% CI, .2–4.7 million) macrolide-resistant MG infections in 2017–2018. The most common mutation was A2058G (26.9%; 95% CI, 12.4%–46.2%; Table 1). A2058C/T was detected in <5 specimens each. A2059C/T was not detected.

Table 1.

Prevalence of Specific MRMs and QRAMs Among Persons Aged 14–59 Years With Macrolide and Quinolone Resistance Testing Results, NHANES, United States, 2017–2018

Mutation	No. With Mutation (Unweighted)a	Weighted % (95% CI)b
Macrolide resistance (n = 49)
MRM
Present	24	42.7 (24.4–62.7)
Absent	25	57.3 (37.3–75.6)
Specific MRM detected
Wild type	25	57.3 (37.3–75.6)
A2058C	…c	…c
A2058G	14	26.9 (12.4–46.2)
A2058T	…c	…c
A2059C	0	0
A2059G	6	9.1 (1.3–28.2)
A2059T	0	0
Quinolone resistance (n = 41)
QRAMe
Present	0	...
Absent	41	100.0 (100.0–100.0)d
Specific QRAM detected
ParC mutations
Wild type	35	87.6 (61.3–98.6)
ParC QRAMe	0	…
Other mutationsf	6	12.4 (1.4–38.7)g
GyrA mutations
Wild type	39	96.8 (82.2–99.9)
GyrA QRAMh	0	…
Other mutationsi	…c	…c

Abbreviations: MRM, macrolide resistance mutation; QRAM, quinolone resistance–associated mutation; NHANES, National Health and Nutrition Examination Survey.

^aNumbers may not sum to the total due to missing data.

^bKorn-Graubard 95% CIs, unless otherwise specified.

^cCell size <5; data suppressed.

^dWald 95% CI calculated because Korn-Graubard 95% CI could not be calculated.

^eParC QRAM mutations included D87H, D87N, D87Y, S83I, and S83R.

^fOther -ParC mutations detected included c234t (silent) and P62S; n ≤ 5 for eac.

^gRelative standard error ≥50%; estimate is unstable. There were no estimates with relative standard error >30% and <50% (interpret with caution).

^hyGrA QRAM mutations included M95I, D99N, and D99G.

ⁱOther GyrA mutations detected included t345c (silent) and c217t (R73C); n ≤ 5 for each.

Weighted MRM prevalence was higher in people aged <25 years (64.3%) vs ≥25 years (30.7%), declining 5% per year as age increased (PR, 0.95 per year; 95% CI, .92–.98; Table 2). MRM prevalence was highest among people who reported mental health care (92.9%; PR, 2.7; 95% CI, 1.53–5.12) or a new sex partner (76.1%; PR, 2.7; 95% CI, 1.34–5.34) in the past year and was associated with not being married or cohabiting (PR, 2.8; 95% CI, 1.21–6.46). Too few people reported nonheterosexual orientation, chlamydia/gonorrhea diagnoses, or chlamydial coinfection to estimate PRs. None of those tested reported HIV infection.

Table 2.

Estimated Prevalence of MRMs Among 49 Persons Infected With Mycoplasma genitalium Aged 14–59 Years With Macrolide Resistance Testing Results, NHANES, United States, 2017–2018

	No. With MRMa (Unweighted)	Weighted Prevalence, % (95% CI)	Weighted PR, % (95% CI)	P Value
Overall	24	42.7 (24.4–62.7)	…	…
Sex
Male	7	33.0 (3.3–80.1)b	1 [Reference]
Female	17	52.5 (25.6–78.4)	1.6 (.53–4.76)	.36
Age
Per year	24	…	0.95 (.92–.98)	.004
<25 y	13	64.3 (14.5–97.2)	2.1 (1.00–4.38)
≥25 y	11	30.7 (14.8–50.9)	1 [Reference]	.05
Race and ethnicityc
Non-Hispanic Black	7	29.3 (7.1–62.7)b	0.8 (.18–3.07)b	.64b
Non-Hispanic White	8	39.1 (1.1–94.5)b	1 [Reference]
Hispanic	8	70.1 (…) d	1.8 (.52–6.19)	.30
Another	…e	…e	…e	…e
Education level
Less than high school	8	37.0 (2.3–88.6)b	0.6 (.18–1.85)b	.31b
High school graduate/GED	6	26.9 (1.6–76.4)b	0.4 (.17–1.02)b	.05b
Some college/college graduate or above	10	64.6 (38.8–85.4)	1 [Reference]
Income/poverty ratio
≤185% federal poverty line	9	31.9 (7.6–67.0)b	0.7 (.22–1.97)b	.40b
>185% federal poverty line	11	49.0 (6.8–92.4)b	1 [Reference]
Marital status
Currently married/living with partner	7	19.6 (5.1–44.5)b	1 [Reference]
Not currently married/living with partner	11	57.4 (31.3–80.8)	2.9 (1.33–6.42)	.02
Health insurance coverage (current)
Yes	19	46.7 (19.9–75.0)	1 [Reference]
No	5	32.9 (…)d	0.7 (.32–1.55)	.34
Seen mental health provider past year
Yes	6	92.9 (.3–100.0)	2.7 (1.53–5.12)	.006
No	8	34.3 (15.7–57.1)	1 [Reference]
Age at sexual debut (per year)	24	…	1.1 (.91–1.24)	.43
Sex with new partner past yearf
Yes	9	76.1 (.1–100.0)	2.7 (1.34–5.34)	.011
No	12	28.4 (10.4–53.6)	1 [Reference]
Ever had anal sex
Yes	8	40.0 (11.6–74.6)	0.8 (.38–1.73)	.56
No	15	49.1 (24.7–73.8)	1 [Reference]
Sex without condom past yearf
Yes	15	43.0 (18.0–71.0)	1.1 (.37–3.08)	.88b
No	5	40.1 (.0–100.0)b	1 [Reference]b
Lifetime number sex partners (per partner)	24	…	0.98 (.95–1.01)	.16

Abbreviations: MRM, macrolide resistance mutation; NHANES, National Health and Nutrition Examination Survey; PR, prevalence ratio; GED, general educational development.

^aNumbers may not sum to the total due to missing responses to the question.

^bRelative standard error >30% and <50%; interpret with caution.

^cHispanic ethnicity includes Puerto Rican, Cuban/Cuban American, Dominican Republic, Mexican American, Central/South American, other Latin American, other Hispanic or Latino. Other race and ethnicity include American Indian/Alaska Native, Asian, Native Hawaiian/Pacific Islander, other, multiple.

^dKorn-Graubard confidence interval could not be calculated

^eCell size <5; data suppressed.

^fAmong those reporting ever having had sex.

No QRAMs in the 41 samples with parC results were detected, nor were any QRAMs detected in the 42 samples with gyrA results. Mutations not associated with resistance were found in both genes (Table 1).

DISCUSSION

MRMs were present in slightly less than half of US respondents infected with MG aged 14 to 59 years in NHANES 2017–2018, equating to ∼1.4 million macrolide-resistant MG infections. MRMs were nearly universal in people infected with MG reporting recent mental health care and were present in ∼75% of those with a new sex partner in the past year. Macrolide resistance was more common in younger than older people and in unmarried/noncohabiting people than stably partnered people. QRAMs were not detected.

One other nationally representative survey has estimated MG AMR prevalence. The British National Survey of Sexual Attitudes and Lifestyles tested specimens collected during 2010 to 2012 (Natsal-3). MG prevalence was similar (Natsal-3 vs NHANES: 1.2% vs 1.8% in men and 1.3% vs 1.7% in women) [12, 13], but MRM prevalence in NHANES was nearly 3 times that in Natsal-3 (42.7% vs 16.1%). Natsal-3 was conducted 6 to 8 years before NHANES 2017–2018; had both surveys been conducted at the same time, the MRM prevalence may have been more similar. Specimen degradation over time likely does not explain the difference in prevalence between surveys. Specimens from Natsal-3 and NHANES had been stored for the same amount of time prior to MRM testing (up to 6 years). MRMs may be more widely disseminated in the United States. In the United Kingdom, MRM prevalence in SHCs during the same period was higher than in Natsal-3 (41% vs 16%) [14], whereas MRM prevalence in US SHCs and NHANES differed less (59% vs 43%) [6].

ParC QRAMs were detected in 3.3% in Natsal-3 but not in NHANES. However, the number with ParC results in Natsal-3 (n = 61) and NHANES (n = 41) was small. Had we detected ParC QRAMs in a single NHANES participant, unweighted prevalences would have been similar.

Few characteristics were associated with MRMs and 95% CIs were wide for all subgroup estimates. Associations with MRMs should be interpreted cautiously. Although race and ethnicity have been consistently associated with MG [12], they were not associated with MRMs, consistent with US SHC data [6]. Associations with younger age, new partners, and marital status may reflect more frequent chlamydial infection and treatment with azithromycin (first-line treatment in the United States until 2021), which selects for MRMs in 10% to 12% of MG infections for each round of treatment. The relationship with mental health care may be driven by depression or attention-deficit/hyperactivity disorder. Depressive symptoms and sexual risk behavior were correlated in SHC attendees (L.E.M., unpublished data), and attention-deficit/hyperactivity disorder is characterized by impulsivity. Both could result in exposure to resistant MG or treatment for chlamydia.

The absence of QRAMs in NHANES 2017–2018 contrasts with meta-analysis estimates of 9.1% prevalence in the United States (2018–2021) [5], reflecting the predominance of SHC data in the meta-analysis. Unlike MRMs, QRAMs in the United States may be more concentrated in populations with high STI risk. As moxifloxacin was not officially recommended for MG treatment in the United States until 2021 [7], the minimal antibiotic pressure in the general population prior to that time may explain the absence of quinolone resistance in NHANES 2017–2018.

The primary strength of this analysis is the nationally representative nature of NHANES. All previous US estimates of MG AMR are derived from SHCs, yet most STIs in the United States are diagnosed in non-SHCs [15]. These data are more generalizable. Leveraging total population size data to estimate the absolute number of macrolide-resistant infections provided new insight into the total burden of resistant MG infections.

Our analysis has limitations. First, despite the large overall sample size in NHANES, the number of MG-positive specimens was small, resulting in limited statistical power and high variances. This is particularly true for population subgroup estimates, which may be unstable and should be cautiously interpreted. Second, we did not correct for multiple comparisons; some statistically significant associations may have occurred by chance. Third, resistance mutations do not always translate to treatment outcomes. Although MRMs correlate well with azithromycin treatment failure, cure can be obtained in the presence of ParC QRAMs [8]. The wide range in minimum inhibitory concentrations for the same ParC mutation, probably further modified by the simultaneous presence of gyrA mutations, contributes to variation in cure rates. Nevertheless, the absence of QRAMs in parC or gyrA suggests that quinolone resistance was rare in 2017–2018. Fourth, specimens were not available from all participants who were MG positive, and not all tested specimens yielded results. Despite our weighted analysis, residual bias in our estimates may remain. Fifth, these data are from 2017–2018 and resistance varies over time. Contemporary prevalence may differ.

The high MRM prevalence likely resulted from widespread use of macrolides. Although QRAMs were not detected, quinolone resistance has increased in other countries [5] and will increase in the United States as MG testing and moxifloxacin treatment expand. MRM testing is increasingly available. It should ideally accompany diagnostic testing for MG, and moxifloxacin should be prescribed only when azithromycin cannot be used. Ongoing AMR surveillance is needed to inform treatment and prevention strategies.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.