The “3 in 1” Study: Pooling Self-Taken Pharyngeal, Urethral, and Rectal Samples into a Single Sample for Analysis for Detection of Neisseria gonorrhoeae and Chlamydia trachomatis in Men Who Have Sex with Men

Abstract

Triple-site testing (using pharyngeal, rectal, and urethral/first-void urine samples) for Neisseria gonorrhoeae and Chlamydia trachomatis using nucleic acid amplification tests detects greater numbers of infections among men who have sex with men (MSM). However, triple-site testing represents a cost pressure for services. MSM over 18 years of age were eligible if they requested testing for sexually transmitted infections (STIs), reported recent sexual contact with either C. trachomatis or N. gonorrhoeae, or had symptoms of an STI. Each patient underwent standard-of-care (SOC) triple-site testing, and swabs were taken to form a pooled sample (PS) (pharyngeal, rectal, and urine specimens). The PS was created using two methods during different periods at one clinic, but we analyzed the data in combination because the sensitivity of the two methods did not differ significantly for C. trachomatis (P = 0.774) or N. gonorrhoeae (P = 0.163). The sensitivity of PS testing (92%) was slightly lower than that of SOC testing (96%) for detecting C. trachomatis (P = 0.167). For N. gonorrhoeae, the sensitivity of PS testing (90%) was significantly lower than that of SOC testing (99%) (P < 0.001). When pharynx-only infections were excluded, the sensitivity of PS testing to detect N. gonorrhoeae infections increased to 94%. Our findings show that pooling of self-taken samples could be an effective and cost-saving method, with high negative predictive values. (Interim results of this study were presented at the BASHH 2013 summer meeting.)

INTRODUCTION

In the United States and Europe, genital Chlamydia trachomatis and Neisseria gonorrhoeae are the most prevalent bacterial sexually transmitted infections (STIs). In the United Kingdom, rates of gonorrhea and chlamydia among men who have sex with men (MSM) continue to rise (1), while studies highlight increases in the numbers of HIV-positive and HIV-negative MSM reporting high-risk sexual behavior since the advent of combination antiretroviral therapy (cART) (2–4).

Coinfection with an STI and HIV has been shown to increase HIV transmission in epidemiological studies (5–7). Concomitant genital infections have been shown to increase shedding of HIV in the genital tract, in both vaginal secretions and seminal fluid (8, 9). Rectal C. trachomatis/N. gonorrhoeae infections have been associated with HIV acquisition in MSM (10), and it is plausible that they may also enhance HIV transmission.

Nucleic acid amplification tests (NAATs) are being used increasingly to screen MSM for pharyngeal and rectal C. trachomatis/N. gonorrhoeae infections, in addition to the use of urethral or first-void urine (FVU) samples, and the use of NAATs is now recommended in national guidelines (11). While there are no NAATs licensed for use with extragenital specimens, several studies have demonstrated that NAATs perform well for the detection of both pharyngeal and rectal C. trachomatis/N. gonorrhoeae infections in MSM (12–14). Studies evaluating the performance of these testing methodologies have demonstrated sensitivities and specificities exceeding 90% for both organisms from all three sites. Testing of pharyngeal and rectal samples in addition to urethral/FVU specimens has been shown to detect greater numbers of infections among MSM, compared to testing of FVU specimens using NAATs and culture of pharyngeal and rectal specimens for N. gonorrhoeae alone (12).

Recent data demonstrated that self-sampling from pharyngeal and rectal sites was feasible and acceptable among MSM and demonstrated similar performance for the detection of C. trachomatis/N. gonorrhoeae, compared to clinician-collected specimens (15, 16). This affords the possibility of testing MSM for STIs in nonclinical settings, including home sampling.

Pooling of samples between individuals has been shown to be sensitive and specific for the detection of C. trachomatis and can result in significant cost savings (17–19). Data suggest that pooling of samples from different sites in the same individual is effective for methicillin-resistant Staphylococcus aureus (MRSA) screening (20, 21), but this approach has not been assessed for STIs. If pooling of samples from the pharynx, urethra, and rectum within an individual is determined to be both sensitive and specific for C. trachomatis/N. gonorrhoeae detection, this could provide cost savings and could influence clinical practice.

The Aptima Combo 2 (AC2) assay (Hologic Inc., San Diego, CA, USA) has high clinical and analytical sensitivity and specificity for the detection of C. trachomatis/N. gonorrhoeae in FVU specimens and urethral swabs from asymptomatic and symptomatic men (14). The AC2 assay amplifies a specific region of the 23S rRNA of C. trachomatis and a specific region of the 16S rRNA of N. gonorrhoeae via DNA intermediates. Confirmation of positive results is possible using alternative single-analyte amplification assays for C. trachomatis and N. gonorrhoeae.

The primary aim of this study was to establish the sensitivity of AC2 testing for C. trachomatis/N. gonorrhoeae using pooled self-collected samples, compared to AC2 testing for C. trachomatis/N. gonorrhoeae from individual sites (current standard-of-care [SOC] testing). The secondary aims were as follows: (i) to determine the patient acceptability of the pooled-sample (PS) approach; (ii) to estimate the sensitivity of standard testing from individual sites for the detection of C. trachomatis/N. gonorrhoeae and the sensitivity of testing using pooled specimens, compared to a gold standard (the gold standard was considered to be any confirmed positive NAAT result, with either single-site or pooled specimens); (iii) to measure the number of missed/additional site-specific infections identified using either PS or SOC testing (from all three sites); (iv) to determine the effect of the volume of FVU specimens on the performance of PS testing; (v) to determine the association of symptoms with the performance of PS testing; and (vi) to estimate the budgetary impact of using pooled AC2 specimens versus separate AC2 specimens.

MATERIALS AND METHODS

Study groups.

The study was conducted in the Sexual Health and HIV clinics of the Mortimer Market Centre (site 1) and Guy's Hospital and St. Thomas' Hospital (site 2) between October 2012 and August 2013. MSM >18 years of age were eligible to participate if they (i) requested testing for STIs, (ii) reported recent sexual contact with either C. trachomatis or N. gonorrhoeae, or (iii) reported symptoms suggesting an STI. Patients were ineligible if they declined to participate or had received any antibiotics in the previous 4 weeks. Patients who agreed to participate provided written consent. The study was approved by the Hampstead Research Ethics Committee (United Kingdom).

Individuals were asked to complete a questionnaire regarding the acceptability of the self-collected pooled specimen, their recent sexual behavior, and presenting symptoms. The clinicians recorded physical signs. Patient demographic data were obtained from clinical records.

Specimen collection. (i) Collection methods.

Two different collection methods for pooled samples were evaluated, i.e., method A at site 1 and method B at site 2. All men underwent triple-site testing (pharyngeal, urine/urethral, and rectal specimens), in line with local clinic policy. Approximately halfway through the study, both sites switched to using method B, because early results suggested that method B was more effective and feedback suggested that it was easier for clinic and laboratory staff members. Samples collected using method A were analyzed in the study to determine whether there was a significant difference between methods A and B and, if not, to combine these data to give a single result on the performance of PS testing.

(ii) Urine/urethral specimens.

All men (with or without urethral symptoms) provided a FVU sample, and the weight of the urine specimen was recorded. Weight, rather than volume, was recorded because this was easier and quicker for the clinic staff to record (1 g = 1 ml, minus the weight of the urine container). Two milliliters of urine was transferred by pipette into an AC2 specimen pot for the individual specimen, and urine from the remaining FVU sample was used for the pooled specimen. Men who reported being in contact with gonorrhea or had urethral symptoms underwent collection of (i) first a urethral swab specimen for microscopy and culture for N. gonorrhoeae and (ii) then an AC2 urethral swab specimen to test for C. trachomatis/N. gonorrhoeae or a FVU sample (according to clinic policy and patient preference).

(iii) Pharyngeal and rectal specimens.

Samples from the pharynx and rectum (obtained at proctoscopy for symptomatic men and with a “blind” swab for asymptomatic men) were collected by clinicians in a standardized way (10, 12). Samples for culture for N. gonorrhoeae were also collected by clinicians, from the pharynx for men with gonorrhea contact and from the rectum for men with rectal symptoms or gonorrhea contact.

Allocation of the order of collection of specimens for each test was randomized and was determined using previously prepared sealed envelopes, with the exception of urethral swab specimens, which were always obtained prior to voiding of urine. Written information explaining the process of obtaining self-taken swabs was given to participants. The patients were asked to place both self-taken swabs into a universal container (method A) or directly into the AC2 urine tube (method B). For method A, after removal of 2 ml of urine, as described above for the SOC samples, both self-taken swabs were added to the FVU sample to produce the pooled specimen. For method B, the self-taken swabs were swirled and compressed against the inner wall of the tube, to release swab material into the AC2 urine tube, and then were removed and discarded. The 2 ml of urine was then added to this AC2 tube to form the method B pooled specimen.

Laboratory methods.

All specimens were stored and processed according to the manufacturer's instructions. Cultures for N. gonorrhoeae used Thayer-Martin selective medium (Oxoid Ltd., Basingstoke, Hampshire, United Kingdom) and were incubated in 10% CO₂ at 37°C for 48 h. Pharyngeal, urethral, FVU, and rectal samples were tested for C. trachomatis/N. gonorrhoeae with the AC2 assay. All specimens in which C. trachomatis or N. gonorrhoeae RNA was detected with the AC2 assay were tested for confirmation with the use of Aptima single-analyte C. trachomatis or N. gonorrhoeae assays (Hologic Inc., San Diego, CA, USA).

Interpretation of microbiological test results.

Negative results for C. trachomatis or N. gonorrhoeae from any individual site using the AC2 assay were considered negative. The following were required for a result to be considered a true-positive or resolved positive result: (i) positive culture results for N. gonorrhoeae from any site; (ii) positive results for C. trachomatis or N. gonorrhoeae from any anatomical site, confirmed using the respective Aptima single-analyte assay; or (iii) positive results for C. trachomatis or N. gonorrhoeae from the PS using the AC2 assay, confirmed using the respective Aptima single-analyte assay.

Sample size.

Assuming the prevalence of C. trachomatis and N. gonorrhoeae to be 10% among symptomatic MSM and MSM presenting as having contact with C. trachomatis or N. gonorrhoeae, a sample size of 1,400 men would be required to result in positive results for an estimate of 139 C. trachomatis and 139 N. gonorrhoeae cases. Assuming that, relative to testing from individual sites using the AC2 assays (i.e., based on cases identified as positive by this method), PS AC2 testing would demonstrate 95% sensitivity, a sample size of 139 positive results for each infection would allow us (i) to estimate the sensitivity of PS testing relative to SOC testing with high precision, with an expected 95% confidence interval (CI) of 90 to 98%, and (ii) to have 80% power to demonstrate that PS testing has acceptable sensitivity relative to SOC testing, with sensitivity greater than 88%. The sample size calculation was conducted using STATA v13.0 software (StataCorp, College Station, TX).

Statistical analyses.

Statistical analyses were performed using STATA v13.0. The performance of the PS and SOC tests were evaluated by calculating sensitivities for each test, using the definition of a resolved positive (see above) as the gold standard. This was done for PS testing overall, according to the method of PS collection (method A or B), and for symptomatic versus asymptomatic patients. The comparison of sensitivities between PS and SOC testing was based on McNemar's test for paired data. The comparisons of the sensitivities of PS testing between methods A and B and between symptomatic and asymptomatic patients were based on Fisher's exact test. Testing for an association between the volume of urine and the sensitivity of PS testing was based on a two-sample t test. Logistic regression was used to explore the association of each infection type with sociodemographic, symptom-related, and behavioral factors. To identify the key factors independently associated with C. trachomatis infections and N. gonorrhoeae infections, model selection based on a backward stepwise procedure was used, with a P value threshold for inclusion and exclusion of 0.05; only factors that were initially found to be significant in univariate analyses (P < 0.05) were included in the process. Subanalyses were performed by looking at sensitivities according to anatomical site and by excluding pharynx-only infections from analysis.

RESULTS

Participant characteristics.

In total, there were 1,064 participants, with a C. trachomatis prevalence of 16% (n = 168) and a N. gonorrhoeae prevalence of 27% (n = 292), which exceeded the numbers of each infection type (n = 139) required to power the study. A total of 586 participants were recruited using method A to produce the pooled specimen (PS) and 478 were recruited using method B to produce the PS. The median age of the participants was 37 years (interquartile range [IQR], 31 to 44 years), with 72% (n = 771) being asymptomatic and 42% (n = 446) being HIV positive; 47% (n = 475) reported an STI in the past year, with 11% (n = 112) acquiring HIV infection within the past year, and 49% (n = 481) reported more than two sexual partners in the past month (Table 1).

TABLE 1.

Characteristics of total sample and odds ratios for associations with C. trachomatis and N. gonorrhoeae infections

Variable	Total (% [no.])	C. trachomatisa			N. gonorrhoeae
		Positive (% [no.])	OR (95% CI)	Adjusted OR (95% CI)b	Positive (% [no.])	OR (95% CI)	Adjusted OR (95% CI)b
Overall	100 (1,069)	15 (164)			27 (288)
Age			P = 0.518			P < 0.001	P < 0.001
<35 yr	43 (460)	15 (66)	1		33 (152)	1	1
35–45 yr	37 (394)	17 (67)	1.22 (0.84–1.77)		26 (101)	0.70 (0.52–0.94)	0.69 (0.50–0.96)
>45 yr	20 (215)	15 (31)	1.01 (0.63–1.60)		16 (35)	0.39 (0.20–0.59)	0.43 (0.27–0.66)
Service provider factors
Method used for PS testing			P = 0. 934			P = 0.125
A	56 (590)	15 (91)	1		29 (170)	1
B	44 (479)	16 (73)	0.99 (0.71–1.38)		25 (118)	0.81 (0.61–1.06)
Clinic			P = 0.777			P = 0.242
1	30 (316)	17 (50)	1		28 (89)	1
2	70 (753)	15 (114)	0.95 (0.66–1.36)		26 (199)	0.83 (0.61–1.13)
Symptom-related and behavioral factors
Any symptoms			P < 0.001			P < 0.001	P < 0.001
No	28 (298)	10 (28)	1		16 (49)	1	1
Yes	72 (771)	18 (136)	2.07 (1.34–3.18)		31 (239)	2.28 (1.62–3.21)	2.01 (1.37–2.93)
C. trachomatis contact			P = 0.106
No	94 (1,008)	15 (150)	1
Yes	6 (61)	23 (14)	1.70 (0.92–3.17)
N. gonorrhoeae contact						P = 0.066
No	87 (932)				26 (242)	1
Yes	13 (137)				34 (46)	1.44 (0.98–2.12)
HIV status			P < 0.001	P = 0.005		P = 0.054
No	58 (623)	12 (73)	1	1	25 (154)	1
Yes	42 (446)	20 (90)	1.93 (1.38–2.70)	1.65 (1.16–2.36)	30 (134)	1.31 (0.97–1.72)
HIV in past year			P = 0.070			P = 0.284
No	89 (957)	15 (139)	1		26 (253)	1
Yes	11 (112)	22 (24)	1.59 (0.98–2.59)		32 (35)	1.26 (0.83–1.93)
STI ever			P = 0.115			P = 0.178
No	13 (92)	9 (8)	1		21 (19)	1
Unsure	3 (23)	13 (3)	1.57 (0.38–6.47)		39 (9)	2.47 (0.93–6.57)
Yes	84 (606)	16 (100)	2.08 (0.97–4.42)		27 (162)	1.40 (0.82–2.40)
STI in past year			P = 0.021			P < 0.001	P = 0.005
No	43 (441)	12 (53)	1		20 (87)	1	1
Unsure	10 (98)	17 (17)	1.54 (0.85–2.79)		25 (24)	1.32 (0.79–2.21)	1.22 (0.71–2.09)
Yes	47 (475)	19 (87)	1.66 (1.15–2.41)		34 (162)	2.11 (1.56–2.85)	1.69 (1.23–2.33)
No. of partners in past month			P = 0.002	P = 0.018		P = 0.001	P = 0.009
1	27 (271)	10 (27)	1	1	20 (53)	1	1
2	24 (242)	14 (34)	1.48 (0.86–2.53)	1.48 (0.85–2.57)	27 (65)	1.51 (1.00–2.28)	1.43 (0.93–2.20)
>2	49 (481)	20 (93)	2.17 (1.37–3.42)	1.95 (1.22–3.13)	32 (152)	1.90 (1.33–2.71)	1.79 (1.23–2.61)
Ever had sex with a woman			P = 0.231			P = 0.796
No	60 (596)	15 (86)	1		27 (158)	1
Yes	40 (400)	17 (69)	1.24 (0.87–1.75)		27 (109)	1.04 (0.78–1.38)

^aOR, odds ratio.

^bFactors were selected in a model selection process (see Materials and Methods).

Predictors of infection.

The C. trachomatis prevalence was 16% (n = 168). In the univariate analysis (Table 1), C. trachomatis infection was associated with having symptoms (P < 0.001); patients with STI symptoms had a prevalence 2-fold higher than that of patients without symptoms. C. trachomatis infection was associated with HIV-positive status (P < 0.001); HIV-positive patients had a prevalence almost 2-fold higher than that of patients who were HIV negative. Individuals who had experienced an STI in the past year had a higher C. trachomatis prevalence (P = 0.021). Individuals who had more than two sexual partners in the past month were more likely to be C. trachomatis positive than were those who had one sexual partner (P = 0.002). After model selection from among the factors in Table 1, STI symptoms (P = 0.006), HIV status (P = 0.002), and number of partners in the past month (P = 0.018) were also significantly associated with C. trachomatis infection.

The N. gonorrhoeae prevalence was 27% (n = 292). In the univariate analysis (Table 1), N. gonorrhoeae infection was associated with age (P < 0.001); older patients, i.e., 35 to 45 years and >45 years of age, were 30% and 60%, respectively, less likely to have N. gonorrhoeae infections than were those younger than 35 years. N. gonorrhoeae infection was associated with symptoms (P < 0.001); patients with STI symptoms had a prevalence >2-fold higher than did those without symptoms. Individuals who had experienced an STI in the past year had a higher N. gonorrhoeae prevalence (P < 0.001), and individuals who had more than two sexual partners in the past month were more likely to be N. gonorrhoeae positive than were those with one sexual partner (P = 0.001). After model selection from among the factors in Table 1, age (P < 0.001), STI symptoms (P < 0.001), having an STI in the past year (P = 0.005), and number of partners in the past month (P = 0.009) were found to be significantly associated with N. gonorrhoeae infection. Among the study subjects (who all self-identified as MSM), 60% had sexual intercourse with a woman at some point in their lives, but this was not a significant predictor of either type of infection.

Patient acceptability.

The study was well received by patients, who indicated a high level of acceptability of this method of sampling; 94% agreed or strongly agreed that they were happy to take their own samples, 93% agreed or strongly agreed that they found it easy to take their own rectal samples (the proportion was 89% for self-taken pharyngeal samples), and 91% agreed or strongly agreed that they were confident in taking their own samples and would be comfortable doing so in the future.

Pooling methods.

There was no significant difference in sensitivity between the two methods (P = 0.774 for C. trachomatis and P = 0.163 for N. gonorrhoeae); therefore, in further analyses, these data were analyzed separately as well as in combination, to give a single result on the performance of PS testing (Table 2).

TABLE 2.

Sensitivity of PS testing versus SOC testing

Infection and method	PS testing		SOC testing		P (PS testing vs SOC testing)
	No. of true-positive cases tested	Sensitivity (95% CI) (%)	No. of true-positive cases tested	Sensitivity (95% CI) (%)
C. trachomatis
Overall	160	91.9 (86.5–95.6)	164	96.3 (92.2–98.6)	0.167
Overall for method A	88	90.9 (82.9–96.0)	91	97.8 (92.3–99.7)	0.109
Overall for method B	72	93.1 (84.5–97.7)	73	94.5 (86.6–98.5)	1.000
Overall for asymptomatic patients	26	88.5 (69.8–97.6)	28	89.3 (71.8–97.7)	1.000
Overall for symptomatic patients	134	93.2 (87.5–96.9)	136	97.8 (93.6–99.5)	0.092
Excluding pharynxa	154	94.2 (89.2–97.3)	158	96.2 (91.9–98.6)	0.607
N. gonorrhoeae
Overall	286	89.9 (85.8–93.1)	288	98.6 (96.4–99.6)	<0.001
Overall for method A	168	87.5 (81.5–92.1)	170	98.8 (95.8–99.9)	<0.001
Overall for method B	118	93.2 (87.1–97.0)	118	98.3 (94.0–99.8)	0.109
Overall for asymptomatic patients	49	81.6 (68.0–91.2)	49	98.0 (89.1–99.9)	0.021
Overall for symptomatic patients	237	91.6 (87.3–94.8)	239	98.7 (96.4–99.7)	<0.001
Excluding pharynxa	231	94.4 (90.6–97.0)	233	98.3 (95.6–99.5)	0.049

^aInfections detected only at the pharyngeal site were excluded from the definition of true-positive results.

Effect of urine volume.

The sensitivities of PS testing to detect C. trachomatis infection were 89%, 91%, and 95% for urine volumes between 1 and 9 ml, between 10 and 19 ml, and between 20 and 90 ml, respectively; for N. gonorrhoeae detection, the sensitivities were 91%, 93%, and 91%, respectively. The associations between urine volume and PS testing sensitivity were not significant for either C. trachomatis (P = 0.816) or N. gonorrhoeae (P = 0.896).

Performance of PS testing. (i) C. trachomatis infections.

The sensitivity of PS testing to detect C. trachomatis infections was 92% (95% CI, 87 to 96%) (Table 2). This was not significantly different from that for SOC testing, i.e., 96% (95% CI, 92 to 99%). Using method A to produce the PS, the sensitivity was 91% (95% CI, 83 to 96%); using method B, the sensitivity was 93% (95% CI, 85 to 98%; P = 0.774). There was slightly higher sensitivity for detecting infections in symptomatic patients versus asymptomatic patients, i.e., 93% (95% CI, 88 to 97%) versus 89% (95% CI, 70 to 98%; P = 0.445).

(ii) N. gonorrhoeae infections.

The sensitivity of PS testing to detect N. gonorrhoeae infections (90% [95% CI, 86 to 93%]) was significantly lower than that of SOC testing (99% [95% CI, 96 to 100%]) (Table 2). Using method A to produce the PS, the sensitivity was 88% (95% CI, 82 to 92%); using method B, the sensitivity was 93% (95% CI, 87 to 97%; P = 0.167). The sensitivity of PS testing by method B was not significantly lower than that of SOC testing. There was somewhat higher sensitivity of PS testing for detecting infections in symptomatic patients versus asymptomatic patients, i.e., 92% (95% CI, 87 to 95%) versus 82% (95% CI, 68 to 91%; P = 0.064).

(iii) Missed infections.

The pooled-sample testing missed 44 infections (13 C. trachomatis and 31 N. gonorrhoeae infections), and SOC testing missed 10 infections (6 C. trachomatis and 4 N. gonorrhoeae infections). Of note, the proportion of infections that were missed using PS testing and that would have remained untreated was relatively small, as 37 of the 44 patients were either C. trachomatis or N. gonorrhoeae contacts and were treated epidemiologically as a contact or had symptoms of infection and were treated on clinical grounds on the day of attendance at the clinic.

(iv) Anatomical sites of infections.

The study was not formally powered to detect the performance of PS testing at different anatomical sites, but the findings are interesting. The sensitivities for detection of C. trachomatis in the pharynx, urethra, and rectum were 69% (95% CI, 39 to 91%; n = 13), 99% (95% CI, 93 to 100%; n = 73), and 92% (95% CI, 85 to 97%; n = 101), respectively (Table 3). Excluding pharynx-only infections increased the sensitivity of PS testing to detect C. trachomatis infections to 94% (95% CI, 89 to 97%) (Table 2).

TABLE 3.

Sensitivity of PS testing according to anatomical site of infection

Species and site of infection	No. of infections detected	No. of infections missed	Sensitivity (95% CI) (%)
C. trachomatis
Urethra	73	1	98.6 (92.6–100.0)
Rectum	101	8	92.1 (85.0–96.5)
Pharynx	13	4	69.2 (38.6–90.9)
N. gonorrhoeae
Urethra	140	3	97.9 (93.9–99.6)
Rectum	167	11	93.4 (88.5–96.7)
Pharynx	156	17	89.1 (83.1–93.5)

The sensitivities for detection of N. gonorrhoeae in the pharynx, urethra, and rectum were 89% (95% CI, 83 to 94%; n = 156), 98% (95% CI, 94 to 100%; n = 140), and 93% (95% CI, 89 to 97%; n = 167), respectively (Table 3). When pharynx-only infections were excluded, the sensitivity of PS testing to detect N. gonorrhoeae infections was 94% (95% CI, 91 to 97%) (Table 2).

(v) Numbers needed to test.

An analysis of the numbers needed to test demonstrated that, for C. trachomatis infections, there would be 1 missed C. trachomatis infection for every 153 people tested by PS testing rather than SOC testing and there would be 1 missed and untreated infection for every 267 people tested by PS testing rather than SOC testing. If pharyngeal samples were excluded from the PS, then there would be 1 missed C. trachomatis infection for every 353 people tested by PS testing rather than SOC testing. When pharyngeal samples were excluded from the PS, there were no missed and untreated C. trachomatis infections; therefore, the additional number needed to test could not be calculated. For N. gonorrhoeae infections, there would be 1 missed N. gonorrhoeae infection for every 43 people tested by PS testing rather than SOC testing and there would be 1 missed and untreated N. gonorrhoeae infection for every 63 people tested by PS testing rather than SOC testing. If pharyngeal samples were excluded from the PS, then there would be 1 missed N. gonorrhoeae infection for every 113 people tested by PS testing rather than SOC testing and there would be 1 missed and untreated N. gonorrhoeae infection for every 1,014 people tested by PS testing rather than SOC testing.

(vi) Negative predictive values.

The negative predictive values for C. trachomatis were 99% (95% CI, 98 to 99%) for PS testing and 99% (95% CI, 99 to 100%) for SOC testing. The negative predictive values for N. gonorrhoeae were 96% (95% CI, 95 to 98%) for PS testing and 99% (95% CI, 99 to 100%) for SOC testing.

DISCUSSION

Our findings show that pooling of self-taken samples could be an effective method of testing for gonorrhea and chlamydia infections in MSM, with high negative predictive values and an approach that is acceptable to patients. There was no statistical difference in the sensitivity of PS testing (92%) versus SOC testing (96%) for detecting C. trachomatis infections. For N. gonorrhoeae infections, PS testing (90%) did not perform as well as SOC testing (99%), and this was statistically significant. We also showed that, when pharynx-only infections were removed from analysis, the sensitivity of our PS testing increased to 94%, which reflects the likely performance of PS testing based on pooling of rectal and urine samples only to detect rectal or urethral infections.

The behavioral and symptom-related factors for the study group demonstrated that this was a relatively high-risk group; 49% had more than two partners in the past month, and 47% had been diagnosed with an STI in the past year. The factors that were significant for predicting infections in this group are not surprising, with being of younger age (less than 35 years) and having had an STI in the past year being predictors of N. gonorrhoeae infection, reporting of symptoms and being HIV positive being predictors of C. trachomatis infection, and having a greater number of partners being a significant predictor for both types of infection. These findings add to the evidence that targeting HIV-positive MSM for behavioral interventions to reduce their risks is something that we need to do more effectively in our clinics.

The costs of each AC2 assay vary according to the laboratory but currently range between approximately £10 and £20 within our services, equating to up to £60 per patient for triple-site testing and amounting to almost £7,000,000 annually for testing of MSM alone in the United Kingdom, based on figures for MSM tested in genitourinary medicine clinics in 2014 (1). Pooling of samples, as in this study, offers potential cost savings of up to two-thirds of the costs of the assays alone, as well as potential savings in consumables, processing time, and clinical pathway efficiency.

There are some limitations to this study. PS testing missed more infections (n = 44) than did SOC testing (n = 10), but it also detected 10 infections that SOC testing missed. Almost all of the missed infections (42/44 infections) were single-site infections (urethra, pharynx, or rectum). Therefore, lower organism loads might have contributed to those false-negative results. The dilutional effect of the volume of urine appeared not to be significant in the analysis. The subanalysis of PS testing according to anatomical site of infection showed that PS testing performed less well in detecting pharyngeal C. trachomatis (sensitivity, 69% [95% CI, 39 to 91%]) and N. gonorrhoeae (sensitivity, 89% [95% CI, 83 to 94%]). All of the samples for the PS were self-taken swabs, and this lower performance of PS testing in detecting pharyngeal infections may be due to sampling error or the lower organism loads seen in pharyngeal infections (22). There is evidence that the swabbing technique is important for optimal isolation of pharyngeal gonorrhea (23), and inadequate self-sampling could account for the lower detection rates for PS testing. The majority of N. gonorrhoeae infections were pharyngeal infections, which may explain the poorer performance of PS testing in detecting N. gonorrhoeae. The clinical significance of the false-negative PS testing results might be reduced; 37 of the 44 patients with such results were treated appropriately for the undetected infections, either as contacts of infection or because of their clinical presentation on the day of attendance at the clinic. SOC testing missed 10 infections, 6 of which would not have been treated had the PS testing not been done as well and detected the infections. Modification of the swabbing technique to obtain pharyngeal samples might improve the diagnostic yield; clinician-collected swabs might be preferable for a PS, and future studies are warranted to see if this achieves higher sensitivity than we achieved with a self-collected PS.

This study was conducted among MSM with high levels of sexual risk behavior in central London, where the STI prevalence is higher than in other populations in the United Kingdom. The performance of PS testing may vary depending on the risk group and the prevalence of infection; further studies are warranted to assess its performance in heterosexual men. There is a study currently assessing the performance of triple-site PS testing in women (24). Our study also assessed only a single platform, the Aptima Combo 2 assay, for the pooled samples. Further studies using other NAAT platforms and pooling techniques are warranted.

If PS testing became the standard of care for STI testing, it might affect how patients are treated when they return positive C. trachomatis or N. gonorrhoeae results. Among this group of MSM in London, C. trachomatis and N. gonorrhoeae infections at any anatomical site are treated in the same way, e.g., all C. trachomatis infections are treated with doxycycline for 1 to 3 weeks as first-line therapy and all N. gonorrhoeae infections are treated with ceftriaxone and azithromycin. However, guidelines may differ according to the setting and risk group, and treatment protocols may need review if PS testing alone is used for diagnosis. All individuals with lymphogranuloma venereum (LGV) who were identified at Guy's Hospital or St. Thomas' Hospital in the study had their pooled specimens tested for LGV DNA in addition to the single-site specimens, and the results were 100% concordant. As we have identified cases of urethral and pharyngeal LGV in our centers, the mucosal site of asymptomatic LGV is of less relevance, provided adequate treatment is given to infected patients. The majority of LGV cases seen in London are cases of symptomatic proctitis, and this is managed syndromically (25). It is recommended that any asymptomatic C. trachomatis detected in MSM be treated with doxycycline for 1 week to cover rectal infection. Reflex testing of all C. trachomatis-positive samples for LGV DNA could improve LGV detection, and an extended course of doxycycline or test of cure could be offered to individuals testing positive for LGV. Using PS testing also has implications for public health reporting of C. trachomatis and N. gonorrhoeae infections, as the site of infection would not be known, which may affect epidemiological data collection. Each center would have to reach a consensus regarding how this issue would be managed; one way might be to screen all patients using PS testing and then perform individual-site testing at the return visit for patients who test positive with the PS.

Pooling of self-taken swabs for C. trachomatis and N. gonorrhoeae detection is a new paradigm for STI testing and offers an acceptable and effective method of testing with a small sensitivity compromise for pharyngeal infections. There is a pragmatic compromise that is made between the major cost savings and the small decrease in the number of identified infections. In addition, the analytic sensitivity of the AC2 assay is higher than that of other commercial N. gonorrhoeae/C. trachomatis NAATs. Even with a slightly lower sensitivity with the pooled sample, the sensitivity is likely still as high or higher than that of some other commercial NAATs in current use for extragenital samples (12).

Use of the pooled-sample method provides the possibility of simplified three-site home testing, with significant cost savings over single-site testing. This could produce a public health benefit in reducing the burden of extragenital infections, with potential reductions in HIV acquisition in MSM (26). British Association for Sexual Health and HIV (BASHH) guidelines advocate the use of triple-site testing for detection of N. gonorrhoeae and C. trachomatis infections among MSM (11, 16). However, cost pressures have meant that many sexual health clinics across the country have not implemented this recommendation. At a service level, the use of pooled samples results in substantial cost savings, which is of great benefit both in the United Kingdom and internationally; it allows optimal testing strategies while containing costs. Cost savings of two-thirds would allow two additional patients to be tested with pooled samples for each patient currently being tested using standard-of-care triple-site testing. The cost of each AC2 assay is £10.80 at the Mortimer Market Centre, which equates to more than £30 per patient for each sexual health screen if separate specimens are used. Even pooling of just rectal and urine samples would allow savings of one-third and reduce the costs to approximately £20 per patient, with annual cost savings of approximately £100,000 (approximately 10,000 triple-site testing episodes per year). The cost of the AC2 assay varies across the United Kingdom, and total testing costs and potential savings might be considerably greater.

The use of pooled specimens from individuals may be even more cost-effective in settings with lower prevalence. The current clinic prevalence of C. trachomatis is 10% and that of N. gonorrhoeae is 16% (data from the Mortimer Market Centre from March to June 2015), lower than the study prevalence rates (16% and 27%, respectively). We can apply these lower prevalences to estimate the numbers of people we would need to test to detect missed infections if the pooled testing method was used instead of SOC testing. An additional 626 people would need to be tested to detect 1 missed C. trachomatis infection, and an additional 678 people would need to be tested to detect 1 missed and untreated C. trachomatis infection. An additional 94 people would need to be tested to detect 1 missed N. gonorrhoeae infection, and an additional 133 people would need to be tested to detect 1 missed and untreated N. gonorrhoeae infection. Using the cost of the AC2 assay at the Mortimer Market Centre, the cost to detect 1 missed C. trachomatis infection with PS testing would be £6,510, and the cost to detect 1 missed and untreated C. trachomatis infection would be £7,051. The cost to detect 1 missed N. gonorrhoeae infection with PS testing would be £978, and the cost to detect 1 missed and untreated N. gonorrhoeae infection would be £1,383. These are estimated numbers and should be interpreted cautiously, but they do demonstrate the potential effectiveness of using PS testing in lower-prevalence settings, with significant cost savings.

With further research, this approach could be used in other risk groups and with other NAAT platforms, leading to changes in current screening practices. No current method of testing optimizes sensitivity; even SOC testing missed 10 infections that PS testing detected. Decision-making regarding the choice and use of tests for STIs should be based on performance, costs, available resources, the priorities and objectives of the clinical service in which we operate, and the acceptability of the testing method to both patients and clinicians. This method of testing, at a time when we face not only significant financial constraints on our sexual health services but also expanding epidemics of STIs in MSM, might allow us to maintain the public health benefits of testing as many people as possible within our current budgets.

Funding Statement

The study was funded by local NHS bodies, Camden Provider Services, and the Guy’s and St. Thomas’ NHS Foundation Trust. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.