Prior Exposure to Azithromycin and Azithromycin Resistance Among Persons Diagnosed With Neisseria gonorrhoeae Infection at a Sexual Health Clinic: 2012–2019

Abstract

Background

There is conflicting evidence on whether prior azithromycin (AZM) exposure is associated with reduced susceptibility to AZM (AZM^RS) among persons infected with Neisseria gonorrhoeae (NG).

Methods

The study population included Public Health–Seattle and King County Sexual Health Clinic (SHC) patients with culture-positive NG infection at ≥1 anatomic site whose isolates were tested for AZM susceptibility in 2012–2019. We used multivariate logistic regression to examine the association of time since last AZM prescription from the SHC in ≤12 months with subsequent diagnosis with AZM^RS NG (minimum inhibitory concentration [MIC], ≥2.0 µg/mL) and used linear regression to assess the association between the number of AZM prescriptions in ≤12 months and AZM MIC level, controlling for demographic, behavioral, and clinical characteristics.

Results

A total of 2155 unique patients had 2828 incident NG infections, 156 (6%) of which were caused by AZM^RS NG. AZM^RS NG was strongly associated with receipt of AZM from the SHC in the prior 29 days (adjusted odds ratio, 6.76; 95% confidence interval [CI], 1.76 to 25.90) but not with receipt of AZM in the prior 30–365 days. Log AZM MIC level was not associated with the number of AZM prescriptions within ≤12 months (adjusted correlation, 0.0004; 95% CI, –.04 to .037) but was associated with number of prescriptions within <30 days (adjusted coefficient, 0.56; 95% CI, .13 to .98).

Conclusions

Recent individual-level AZM treatment is associated with subsequent AZM^RS gonococcal infections. The long half-life and persistence of subtherapeutic levels of AZM may result in selection of resistant NG strains in persons with recent AZM use.

Among sexual health clinic patients with culture-positive Neisseria gonorrhoeae at ≥1 anatomic site, azithromycin exposure in the prior 30 days was associated with azithromycin-resistant infection.

Neisseria gonorrhoeae (NG) has progressively developed resistance to all antimicrobials previously used for its treatment [1], leading the Centers for Disease Control and Prevention (CDC) to declare antimicrobial-resistant (AMR) NG an urgent health threat [2]. In December 2020 [3], the CDC changed its treatment recommendation for uncomplicated gonorrhea from a 2-drug regimen of ceftriaxone 250 mg and 1 g of azithromycin (AZM) to ceftriaxone 500 mg monotherapy. This change was prompted by surveillance data that demonstrated an increase in the percentage of Gonococcal Isolate Surveillance Project isolates with an AZM minimum inhibitory concentration (MIC) ≥2.0 µg/mL (indicating reduced susceptibility to AZM [AZM^RS]) from 0.6% in 2013 to 5.1% in 2019; among 2019 isolates from men who have sex with men (MSM), 8.8% [4] had an elevated MIC to AZM. Current British guidelines have similarly dropped the inclusion of AZM as a second agent in treating gonorrhea [5], though European [6] and Australian [7] guidelines continue to recommend AZM-inclusive regimens.

Currently, there is conflicting evidence regarding the relationship between individual-level AZM exposure and subsequent AZM resistance in persons with NG infection. Exposure to erythromycin in vitro induces high-level macrolide resistance in NG strains with low-level resistance [8], and there have been numerous case reports of de novo selection for AZM resistance during NG treatment [9, 10]. A study of sexually transmitted infection (STI) clinic patients in the Netherlands found that NG isolates from persons who had received AZM in the preceding 30 days had higher AZM MICs than isolates from patients who received AZM 31–60 days prior to diagnosis and persons with no exposure to AZM in the preceding 60 days [11]. These findings are at odds with 2 UK studies that evaluated the association of prior non-gonococcal urethritis (NGU), Chlamydia trachomatis (CT), or NG diagnoses [12] or NG alone [13]—proxies for AZM exposure—and AZM^RS in persons with culture-positive NG and found no such association.

We used data collected by the Public Health Seattle–King County (PHSKC) Sexual Health Clinic (SHC) to evaluate the association between individual-level receipt of AZM for STI treatment and subsequent NG infection with AZM^RS. We hypothesized that AZM treatment in ≤12 months preceding NG diagnosis would be associated with AZM^RS NG and that receipt of more courses of AZM would be associated with higher AZM MICs.

METHODS

Study Design, Study Population, Clinical Protocols, and Demographic and Behavioral Data

The study population included all patients from the PHSKC SHC with culture-positive NG at ≥1 anatomic site whose isolates were tested for AZM resistance between 1 January 2012 and 31 December 2019. Patients who presented to the clinic answered questions detailing demographic information, sexual history, and reason(s) for clinical visit using a computer-assisted self-interview. During the visit, clinicians verified sexual histories and documented exam findings, testing, and treatment. Throughout the study period, clinic policy recommended clinicians obtain specimens for culture from persons with urethral discharge and diplococci detected on Gram stain, persons with symptoms of NG (eg, cervicitis, pelvic inflammatory disease, urethritis, proctitis), persons reporting sexual contact to a partner with NG, and persons who previously tested positive for NG on molecular tests and are returning for treatment. Among men who have sex with men (MSM) evaluated as contacts to NG, clinicians collected culture specimens from all exposed anatomical sites. Among men who have sex with women exclusively (MSW) and women evaluated as contacts from 2012 through 2016, clinicians collected only genital culture specimens; after 2016, cultures were collected from all exposed anatomic sites; after 2016, clinicians added pharyngeal cultures for individuals who reported performing oral sex in the past 2 months. In 2019, all MSW and women were screened at the oropharynx regardless of their report of oral sex.

AZM Susceptibility Data

AZM susceptibility testing was performed at the University of Washington Neisseria Reference Laboratory. From 2012–2016, urethral isolates were tested using the agar dilution method [14], while isolates from extragenital and endocervical sites were tested using the Kirby-Bauer disk diffusion method with the CDC’s protocol and AZM disk diffusion zone diameter breakpoints [15]. Isolates tested using Kirby-Bauer disk diffusion with AZM zone diameters ≤30 mm (equivalent to agar dilution AZM MIC ≥0.5–1 µg/mL) were also tested using agar dilution. We imputed a low AZM MIC (0.25) for cultures with AZM zone diameters >30 mm. AZM susceptibility testing of isolates collected from 2017 through 2019 was performed using Etest [16]. Isolates were classified as having AZM^RS if the MIC was ≥2.0 µg/mL. If patients had a culture-positive infection at >1 anatomic site, we included the site with the highest AZM MIC. If the MICs at multiple anatomic sites were equal, we randomly selected 1 anatomic site to include. Isolates were linked to patients’ medical records by the PHSKC SHC data manager to create a deidentified dataset for analysis.

Statistical Analyses

We used a retrospective cohort study design comparing AZM use in the preceding 12 months between persons with NG infection with AZM^RS and without AZM^RS. Characteristics of patients with and without AZM^RS NG infections were compared using χ² tests for categorical variables. We categorized timing of last prescription for 1 g or 2 g AZM from the SHC in the preceding 12 months into 1 of 5 categories: <30 days, 30–89 days, 90–179 days, 180–364 days, and no AZM in the past 365 days. We then examined the association of timing of last prescription with having an NG infection with AZM^RS using univariate and multivariate logistic regression for clustered outcomes to account for persons with >1 visit in the study period. We then examined the association between the number of 1-g or 2-g AZM prescriptions in the prior 12 months and log AZM MIC (due to the fact that MICs are measured by dilution) using the Spearman rank correlation and multivariate partial Spearman rank correlation [17] with robust standard errors. We examined the association between the number of AZM prescriptions in the prior 30 days and log AZM MIC using univariate and multivariate linear regression due to the abundance of ties produced by the smaller range of possible prescriptions in this time period in Spearman rank correlation. Multivariate models adjusted for anatomic site of infection (penile urethral, cervicovaginal, rectal, or pharyngeal), number of sexual partners in ≤12 months (0, 1, 2–4, 5–9, ≥10, missing), self-reported human immunodeficiency virus (HIV) status, year (2012–2013, 2014–2015, 2015–2017, 2018–2019) and season of NG diagnosis (December–February, March–May, June–August, September–November) [18], race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic [any race], non-Hispanic Asian/Pacific Islander/Multiple Races, Declined/Missing/Unknown), patient’s gender and sex of sex partner(s) (cisgender MSM, cisgender MSW, cisgender female, genderqueer/trans/nonbinary), age in years, and number of clinic visits in the prior 12 months. We also repeated the univariate and multivariate models on restricted datasets of only pharyngeal and rectal isolates and of penile urethral isolates, given the higher proportion of AZM^RS infections detected at extragenital sites [19]. As a sensitivity analysis, we also imputed higher (1.0) and lower (0.125) MIC values for isolates that did not meet the threshold for Kirby-Bauer disk diffusion for each linear regression model. Analyses were conducted using Stata software v15.1 (StataCorp, College Station, TX). An α of 0.05 was considered significant.

Human Subjects Protection

The University of Washington’s Human Subjects Division determined that all analyses were performed as part of gonorrhea surveillance and control activities and were thus non-human subject research as defined by federal and state regulations.

RESULTS

Study Population and Univariate Associations of Demographic and Clinical Characteristics With AZM^RS NG

Between 2012 and 2019, 2155 unique patients were diagnosed with 2828 culture-positive NG infections, 156 (6%) of which had isolates with AZM^RS. Cisgender MSM, persons identifying as transgender, genderqueer and/or nonbinary, and persons of Hispanic ethnicity were more likely to have an AZM^RS NG infection (Table 1). Isolates from the rectum and pharynx were more likely to be AZM^RS compared with penile urethral and cervicovaginal isolates. There were no significant differences between persons with AZM^RS and susceptible isolates by self-reported HIV status or age. Persons with an AZM^RS NG infection were more likely to report having ≥10 partners in the previous 12 months (35% vs 24%, P = .02). The proportion of NG isolates with AZM^RS increased between 2012 and 2019, from 0% to 9% (P < .001). However, there was no significant association between AZM^RS and season of diagnosis or the number of clinic visits or AZM prescriptions in the 12 months prior. The geometric mean of AZM MIC in persons with susceptible isolates was 0.24 (range, 0.03–1.0), and the geometric mean in persons with AZM^RS isolates was 2.42 (range, 2–256; data not shown).

Table 1.

Study Population Demographics, Sexual History, and Clinical Characteristics by Azithromycin Minimum Inhibitory Concentration of Current Infection

Demographics, Sexual History, and Clinical Characteristics	AZM MIC <2.0 μg/mL	AZM MIC ≥2.0 μg/mL	P Value
	N = 2672 (94%)	N = 156 (6%)	…
	N (Column %)		…
Gender/Sex of sex partners
ȃCisgender male who has sex with males	2190 (82%)	140 (90%)	.005
Cisgender male who has sex with females	378 (14%)	9 (6%)
Cisgender female	67 (3%)	2 (1%)
Transgender, genderqueer, nonbinary	37 (1%)	5 (3%)
Age, y			.99
18–24	544 (20%)	32 (21%)
25–29	838 (31%)	49 (31%)
30–39	663 (25%)	37 (24%)
40+	627 (23%)	38 (24%)
Race/Ethnicity			.008
Hispanic	352 (13%)	31 (20%)
White	1413 (53%)	90 (58%)
Black	478 (18%)	13 (8%)
Native American, Pacific Islander, Asian, multiple races	345 (13%)	17 (11%)
Unknown/Missing	84 (3%)	5 (3%)
With human immunodeficiency virus	367 (14%)	20 (13%)	.75
Sexual partners in past 12 mo
0	698 (26%)	29 (19%)	.02
1	63 (2%)	2 (2%)
2–4	348 (13%)	20 (13%)
5–10	365 (24%)	24 (15%)
10+	633 (24%)	55 (35%)
Missing partner info	565 (21%)	25 (4%)	.02
Anatomic site of infection			.04
Urethra	1468 (55%)	72 (46%)
Pharynx	486 (18%)	31 (20%)
Rectum	675 (25%)	52 (33%)
Cervicovaginal	43 (2%)	1 (0.6%)
Year of Infection
2012–2013	478 (18%)	0 (0;0)	<.001
2014–2015	576 (22%)	25 (16%)
2016–2017	748 (28%)	45 (29%)
2018–2019	870 (33%)	86 (55%)
Season			.70
December–February	703 (26%)	46 (29%)
March–May	653 (24%)	39 (25%)
June–August	668 (25%)	39 (25%)
September-November	648 (24%)	32 (21%)
Number of prior clinic visits in past 12 mo
0	1212 (45%)	74 (47%)	.93
1	521 (20%)	30 (19%)
2–3	499 (19%)	26 (17%)
4+	440 (16%)	26 (17%)
Number of azithromycin prescriptions in past 12 mo
0	1821 (68%)	115 (74%)	.44
1	561 (21%)	28 (18%)
2	187 (7%)	7 (5%)
3+	103 (4%)	6 (4%)

Abbreviations: AZM, azithromycin; MIC, minimum inhibitory concentration.

Univariate and Multivariate Regression Models

Of the 156 patients with AZM^RS NG infections, 115 (74%) had not received a prescription for AZM from the SHC in ≤365 days, 4 (2.6%) had received a prescription in ≤29 days prior to a subsequent NG diagnosis, and 37 (24%) had received AZM in ≤30–364 days prior to NG diagnosis (Table 2). These proportions were similar when restricting to persons with extragenital vs penile urethral infections. In univariate analysis, diagnosis with AZM^RS NG was associated with prescription of AZM in <30 days (odds ratio [OR], 3.52; 95% confidence interval [CI], 1.32 to 9.38) compared with having no prescription for AZM in the previous 365 days (Table 2). This association was stronger after adjustment for clinical, demographic, and risk behavior characteristics (adjusted OR [aOR], 6.76; 95% CI, 1.76 to 25.90). Having a prescription for AZM in 30–89 days, 90–179 days, or 180–364 days prior was not associated with having AZM^RS NG in either univariate or multivariate logistic regression. In the multivariate model restricted to persons with a rectal or pharyngeal infection, having an AZM prescription in the prior 30 days was associated with having an NG infection with AZM^RS compared with having no AZM prescription in the prior 365 days (aOR, 5.40; 95% CI, 1.24 to 23.43; Table 2). Having a prescription for AZM in 30–89 days, 90–179 days, or 180–364 days prior was not associated with having AZM^RS NG rectal or pharyngeal infection in either univariate or multivariate logistic regression. Although there was no association of AZM^RS penile urethral infection and receipt of AZM in the 30 days prior to diagnosis, we observed only 7 such infections, and the aOR was similar to that observed for other anatomic sites (aOR, 8.17; 95% CI, .65 to 103.18).

Table 2.

Number and Proportion of Neisseria gonorrhoeae Infections With Reduced Azithromycin (AZM) Susceptibility and Association of Last AZM Dose With AZM Resistance^a in Logistic Regression

		Univariate			Multivariate
Timing of Last AZM Dose	Persons With AZMRS Infection/Persons in AZM Exposure Category (%)	OR	95% CI	P Value	OR	95% CI	P Value
Any anatomic site of infectionb
ȃNo AZM in past 365 d (ref)	115/1936 (6)	1.0	…	…	1.0	…	…
ȃ<30 d	4/22 (18)	3.52	1.32 to 9.38	.01	6.76	1.76 to 25.90	.005
ȃ30–89 d	9/216 (4)	0.69	.34 to 1.39	.30	0.75	.35 to 1.61	.46
ȃ90–179 d	10/301 (3)	0.54	.28 to 1.05	.07	0.51	−.24 to 1.08	.08
ȃ180–364 d	18/353 (5)	0.85	.51 to 1.41	.53	0.83	.47 to 1.45	.51
Pharyngeal or rectal infectionc
ȃNo AZM in past 365 d (ref)	60/860 (7)	1.0	…	…	1.0	…	…
ȃ<30 d	3/15 (20)	3.33	1.07 to 10.41	.04	5.40	1.24 to 23.43	.02
ȃ30–89 d	7/74 (8)	1.39	.61 to 3.16	.43	1.36	.55 to 3.40	.51
ȃ90–179 d	5/127 (4)	0.55	.21 to 1.39	.21	0.50	.17 to 1.51	.22
ȃ180–364 d	8/154 (5)	0.73	.34 to 1.56	.42	0.74	.32 to 1.68	.47
Urethral infectionc
ȃNo AZM in past 365 d (ref)	54/1030 (5)	1.0	…	…	1.0	…	…
ȃ<30 d	1/6 (17)	3.61	.41 to 31.51	.25	8.17	.65 to 103.18	.11
ȃ30–89 d	2/140 (1)	0.26	.06 to 1.10	.07	0.30	.06 to 1.49	.14
ȃ90–179 d	5/174 (3)	0.53	.21 to 1.36	.19	0.51	.18 to 1.47	.21
ȃ180–364 d	10/197 (5)	0.92	.49 to 1.92	.92	0.94	.45 to 1.96	.88

Abbreviations: AZM^RS, reduced susceptibility to azithromycin; CI, confidence interval; OR, odds ratio.

Minimum inhibitory concentration ≥ 2.0 μg/mL.

Multivariate regression adjusted for anatomic site of infection (reference = penile urethral infection), number of sex partners in past 12 months, year of diagnosis (reference = 2012–2013), self-reported human immunodeficiency virus (HIV) infection (reference = not HIV-positive), season (reference = summer [June–August]), race/ethnicity (reference = non-Hispanic White), gender/sex of sex partner(s) (reference = cisgender men who have sex with men), age, and number of clinic visits in past 12 months.

Multivariate regression adjusted for number of sex partners in past 12 months, self-reported HIV infection (reference = not HIV-positive), year of diagnosis (reference = 2012/2013), season (reference = summer [June–August]), race/ethnicity(reference = non-Hispanic White), gender/sex of sex partner(s) (reference = cisgender men who have sex with men), age, and number of clinic visits in past 12 months.

Log AZM MIC was not associated with the number of AZM prescriptions from the SHC in the 12 months prior to NG diagnosis (adjusted correlation [aCorr], 0.0004; 95% CI, –.04 to .037; Table 3). This positive, nonsignificant relationship was also observed in the model restricted to rectal and pharyngeal infections (aCorr, 0.049; 95% CI, –.011 to .11; Table 3). For each additional AZM prescription in the past 30 days, there was a 0.56 (95% CI, .13 to .98) increase in the log AZM MIC in multivariate analysis (Table 4). This relationship was also significant when restricted to extragenital infections (adjusted coefficient [aCoef], 0.55; 95% CI, .03 to 1.08) but not penile urethral infections alone (aCoef, 0.52; 95% CI, –.27 to 1.32). There were no substantial differences in the strength, direction, or significance of these associations when varying the imputed MIC for samples that did not meet the Kirby-Bauer disk diffusion threshold (data not shown).

Table 3.

Association of Azithromycin (AZM) Prescriptions in the Past 365 Days With Log AZM Minimum Inhibitory Concentration Level, μg/mL, in Univariate Spearman Rank Correlation and Multivariate Partial Spearman Correlation

	Univariate			Multivariatea
Anatomic Site of Infection	Coefficient	…	P Value	Coefficient	95% Confidence Interval	P Value
Any anatomic site of infectiona	0.006	–.03 to .04	.74	0.0004	–.04 to .037	.98
Rectal or pharyngeal infectionb	0.019	–.037 to .075	.51	0.049	–.011 to .11	.11
Urethral infectionb	0.001	−.049 to .051	.97	−0.029	−.078 to .020	.24

Multivariate correlation adjusted for anatomic site of infection (reference = penile urethral infection), number of sex partners in past 12 months, year of diagnosis (reference = 2012–2013), self-reported human immunodeficiency virus (HIV) infection (reference = not HIV-positive), season (reference = summer [June–August]), race/ethnicity (reference = non-Hispanic White), gender/sex of sex partner(s) (reference = cisgender men who have sex with men), age, and number of clinic visits in past 12 months.

Multivariate correlation adjusted for number of sex partners in past 12 months, year of diagnosis (reference = 2012–2013), self-reported HIV infection (reference = not HIV-positive), season (reference = summer [June–August]), race/ethnicity (reference = non-Hispanic White), gender/sex of sex partner(s) (reference = cisgender men who have sex with men), age, and number of clinic visits in past 12 months.

Table 4.

Association of Azithromycin (AZM) Prescriptions in the Prior 30 Days With Log AZM Minimum Inhibitory Concentration Level, μg/mL, in Univariate Linear Regression and Multivariate Linear Regression

	Univariate			Multivariatea
	Coefficient	95% CI	P Value	Coefficient	95% CI	P Value
Any anatomic site of infectiona	0.53	.11 to .96	.02	0.56	.13 to .98	.01
Rectal or pharyngeal infectionb	0.53	.004 to 1.06	.048	0.55	.03 to 1.08	.04
Urethral infectionb	0.32	−.38 to 1.02	.37	0.52	−.27 to 1.32	.20

Abbreviation: CI, confidence interval.

Multivariate regression adjusted for anatomic site of infection (reference = penile urethral infection), number of sex partners in past 12 months, year of diagnosis (reference = 2012–2013), self-reported human immunodeficiency virus (HIV) infection (reference = not HIV-positive), season (reference = summer [June–August]), race/ethnicity (reference = non-Hispanic White), gender/sex of sex partner(s) (reference = cisgender men who have sex with men), age, and number of clinic visits in past 12 months.

Multivariate regression adjusted for number of sex partners in past 12 months, year of diagnosis (reference = 2012–2013), self-reported HIV infection (reference = not HIV-positive), season (reference = summer [June–August]), race/ethnicity (reference = non-Hispanic White), gender/sex of sex partner(s) (reference = cisgender men who have sex with men), age, and number of clinic visits in past 12 months.

DISCUSSION

We found evidence for an association between recent (<30 days) receipt of AZM and subsequent AZM^RS NG infection. This relationship persisted when we restricted our analyses to extragenital infections. While the positive association between AZM receipt and NG infection with AZM^RS among persons with a penile urethral infection was not statistically significant, the similar effect size for this estimate along with wide confidence bounds may indicate that we did not have sufficient statistical power to detect such an association if it exists. We also found a significant relationship between the number of exposures to AZM and increasing levels of resistance in NG infections acquired within 30 days but not with the cumulative number of AZM exposures in the prior year. These findings support our hypothesis of a relationship between recent AZM use and the development of AZM resistance in NG infections at an individual level.

Prior work has identified 2 potential mechanisms by which AZM resistance emerges: direct selection through exposure to AZM for STIs (which this analysis was primarily designed to detect) and bystander selection through individual and community-level exposure to AZM for non-NG/STIs [20]. Our results suggest that direct selection related to STI treatment occurs. AZM therapy could lead to selective pressure on chromosomal mutations conferring AZM^RS given the long half-life of AZM and intracellular persistence of subtherapeutic levels up to 20 days [21]. This period closely corresponds to the 30-day exposure period in which we observed a strong association with AZM^RS. These results, which are consistent with results of the aforementioned Dutch study that found an association between AZM^RS NG infection and AZM receipt in the same exposure period, provide strong evidence of a link between individual-level AZM use and emergence of AZM^RS in NG infections. Our results contradict 2 UK studies that found no association between AZM exposure and AZM resistance [12, 13]. However, both studies used surveillance data to infer AZM exposure from STI [12] or NG [13] diagnoses. This may have led to misclassification of AZM exposure categories, particularly when using only NG diagnoses.

Our findings may also point to the role of commensal bacterial species in development of resistance. NG is highly adept at acquiring resistance determinants due to its propensity for genome modification via passive uptake of exogenous genetic material. Exposure to macrolides can induce resistance mutations in commensal Neisseria and other bacterial species that are plentiful in the human oropharynx [22–24]. These mutations may then be acquired by NG during a subsequent infection via conjugal transfer [23, 25]. Pharyngeal and rectal infections are often asymptomatic and can persist for longer periods of time, allowing a longer period of coexistence with commensals [26] and increasing the chances of AMR gene transfer [27]. Additionally, previous work has demonstrated low concentrations of antibiotics used for NG treatment in the oropharynx [28], leading to suboptimal concentrations that allow for emergence and proliferation of resistant strains [29].

Despite the strong association of AZM exposure in the previous 30 days with AZM^RS, the majority of persons with an AZM^RS NG infection were not exposed to AZM in this time period. This suggests that AZM resistance arises from direct selection, in at least some instances as a result of STI treatment, and then spreads through transmission of clonal resistant strains. Whole-genome sequencing studies detected the emergence and spread of NG with high-level AZM^RS in Scotland as early as 2004 [30]. In 2015, a cluster of NG with high-level AZM^RS was detected in Britain [31], and sustained transmission of this clonal strain was documented through at least 2018 [32]. Transmission of NG strains with high-level AZM^RS has been documented during the same period in Australia [33] and Canada [34] and later in the United States [35]. The majority of resistant isolates included in our analysis occurred after 2017, which may indicate rapid proliferation of newly emergent clonal strains. Additionally, the densely connected sexual networks of MSM, in combination with frequent testing and treating, have been proposed as a mechanism by which AMR NG strains can emerge and quickly propagate [36, 37].

The induction of resistant strains of NG may also result from an indirect bystander effect of population-level AZM exposure. High levels of AZM use for indications unrelated to STIs could induce NG resistance in people with unidentified gonorrhea or who acquire gonorrhea shortly after receiving AZM. Some of the strongest evidence for the role of population-level selective pressure comes from an ecological study linking population-level seasonal macrolide use with seasonal variation in AZM MICs in a nationally representative sample of NG isolates [18]. We did not observe the same association between AZM^RS and seasonality in our results, possibly due to an incomplete assessment of AZM use from other clinical sources. Other ecological studies have demonstrated similar associations between country-level macrolide use and prevalence of AZM^RS NG [38, 39]. We have previously demonstrated a marked increase in AZM use for epidemiologic and empiric treatment of STIs among PHSKC SHC attendees during the period examined, particularly among MSM [40], which may have contributed to high community-level selective pressure for AZM^RS determinants among this population. The documented increase in SHC AZM prescriptions closely correlates with the time periods in which AZM^RS infections were most commonly detected, and year of infection was one of the few covariates correlated with having an AZM^RS NG infection.

We acknowledge several limitations that may have impacted our analysis. Between 2013 and 2017, only extragenital isolates with AZM zone diameters ≤30 mm on Kirby-Bauer disk diffusion were selected for agar dilution, and it is possible that this method misclassified some extragenital isolates with lower-level (2 µg/mL) AZM^RS. We were limited to records of AZM episodes from a single sexual health clinic. Patients may have been prescribed AZM in other clinical settings, resulting in underestimates of total AZM use and misclassification of exposures, diminishing our statistical power. However, given that our clinic serves a population at high risk of bacterial STIs and provides regular screening and treatment, including quarterly screening for preexposure prophylaxis services, we believe our data capture a large proportion of total AZM use. Finally, we acknowledge that the sample size of patients exposed to AZM in the prior 30 days, the period of exposure for which we identified an association with AZM^RS NG, was relatively small (N = 22). However, the strength of the association identified was very strong in both univariate and multivariate analyses, and our results align with those from the only other study to capture AZM prescriptions explicitly [11] rather than using STI diagnoses as proxy [12, 13]. We also note that our study included more than 2800 anatomically diverse isolates with 156 AZM^RS isolates, making this one of the largest clinic-based analyses attempting to answer this question to date.

CONCLUSIONS

In conclusion, we found an association between recent receipt of AZM for STI treatment and AZM^RS NG infections among PHSKC SHC attendees. However, most persons with AZM^RS infections did not have AZM exposure in the previous year, suggesting transmission of clonal resistant strains and AZM use for non-STI indications also play an important role in the proliferation of AZM resistance. These findings underscore the timeliness of updated STI treatment guidelines that reduce the utilization of AZM in populations at high risk for NG and CT.

Notes

Acknowledgments. Gonococcal antimicrobial resistance surveillance activities at the University of Washington Neisseria Reference Laboratory are supported by the Centers for Disease Control and Prevention (CDC) Strengthening the US Response to Resistant Gonorrhea (SURRG) and the Antibiotic Resistance Laboratory Network (CK19-1904).

Financial support. This work was supported by the National Institutes of Health (NIH; grants 1R03AI156261-01 to L. A. B., TR002318 to E. R.). L. A. B. reports support from the CDC (Neisseria gonorrhoeae culture conducted through CDC-funded surveillance activities Gonococcal Isolate Surveillance Project and SURRG). C. T. received support from the CDC Epidemiology and Laboratory Capacity cooperative agreement CK19-1904 (grant funding to institution, Public Health–Seattle and King County).