Four years ago, we published an article in this journal reporting the results of an evaluation of more than 13,000 Gram-stained smears from men presenting with urethral discharge at the Denver Metro Health Clinic. Our findings provided a rationale to lower the traditional threshold at which treatment for nongonococcal urethritis (NGU) is given, from ≥5 to ≥2 polymorhonuclear cells per high-power field (PMNs/HPF) in the absence of intracellular diplococci. We arrived at this conclusion because the prevalence of Chlamydia trachomatis at this threshold (16.2%) was similar to or higher compared with prevalence in circumstances where treatment is currently recommended, for example, male partners of women diagnosed as having pelvic inflammatory disease (13.4%) or mucopurulent cervicitis (8.7%).1 Our findings added to the evidence from other studies that the traditional Gram stain threshold of at least 5 PMNs/HPF missed substantial proportions of both C. trachomatis and Mycoplasma genitalium infections among men presenting with urethritis symptoms.2,3 Subsequently, after review of the available literature on this topic,4 the Centers for Disease Control and Prevention included the threshold of at ≥2 PMNs/HPF in their 2015 sexually transmitted diseases (STD) treatment guidelines.5
In our article, we recognized the limitations of a single-site study and encouraged other STD clinics to validate our results, including clinics where clinicians both collect specimens and perform microscopy. One such study, very nicely conducted and described by Moi et al.6 is published in this issue of Sexually Transmitted Diseases and, at first blush, seems to be at odds with ours. Specifically, the study found low chlamydia prevalence rates at Loeffler's methylene blue (LMB) stain levels <10 PMNs/HPF and recommends against decreasing the treatment cutoff below this level.
We want to acknowledge that Moi et al. graciously sent us an earlier draft of their manuscript for our review. We provided feedback and suggested they submit the manuscript to Sexually Transmitted Diseases (disclosure: we recused ourselves from the review process after the manuscript was submitted). Many of our comments and suggestions have been addressed by the authors in the current version; however, we want to make a few points in this editorial in an attempt to reconcile the 2 articles.
First, there is an important point on which both studies agree. In contrast to gonococcal urethritis where most men are symptomatic and have Gram/LMB stain results >10 PMNs/HPF, the inflammatory expression of urethral chlamydial infection in men as measured by the presence of leucocytes in urethral discharge is much more gradual. Indeed, a large proportion of men with chlamydia infections in our 2 studies were asymptomatic, judged by the fact that approximately 50% of men with chlamydia infection in either study did not have a Gram/LMB stain performed. Among men with urethral discharge who did have a Gram/LMB stain performed, chlamydial prevalence gradually increased over the Gram/LMB stain strata as demonstrated in both studies. The prevalence at each stratum represents the positive predictive value of the stain to detect chlamydia at or above that level. This value increases as the overall prevalence in the population increases, and thus, the differential findings in the studies may simply be the result of differences in background chlamydia prevalence; more than twice as high in the Denver study (16.6%) compared with the Oslo study (7.5%). As a result, the predictive value for a positive chlamydia result of a positive Gram stain was 34.8% in our study at the threshold of ≥2 PMNs/HPF, compared with only 27.2% in the study by Moi et al. at the much more stringent cutoff of ≥10 PMNs/HPF.
There are several methodological differences between the 2 studies that may also explain part of the difference. First, having the clinician perform the microscopy may introduce a bias in favor of treatment and thus influence the overall judgment of the Gram/LMB stain result that may be avoided by having a less-biased laboratorian determine the outcome, as was done in our study. The authors included a small substudy with blinded laboratorians to examine such a bias, and this did not seem to be an issue. Second, the study by Moi et al. did not stratify their LMB stain results for each 1-PMN/HPF increment as was done in our study, but grouped the results in 3 strata: <5, 5–9, and ≥10 PMNs/HPF. Thus, they were not able to directly compare their results with ours and to determine a more exact cutoff at which they would have found similar chlamydia levels compared with our study. Based on the above considerations, one would expect this cutoff to be somewhere between 5 and 10 PMNs/HPF, but a more precise measurement would have been desirable and might have further minimized classification bias. Finally, there is the difference in staining technique, but we agree with Moi et al. that there is no evidence that Gram or LMB stains perform differently in the diagnosis of gonococcal urethritis or NGU. That said, as suggested by Moi et al., the method of obtaining urethral discharge for Gram/LMB stain was also different between studies, and the more invasive procedure described by them may indeed be more sensitive than the procedure used in our clinic. However, as Moi et al. concede, studies comparing the sensitivity of various collection methods for urethritis diagnosis are lacking. Regardless, the Gram/LMB stain is a very poor screening test for chlamydia because so many men are symptomatic and would not receive the test. When including men who did not receive a Gram/LMB stain assessment, the overall sensitivity in our study was only 43.3% and similarly low in the study by Moi et al. (both at the ≥5 PMN/HPF threshold). When limiting the results to men who received a Gram/LMB stain test, the sensitivity in the Oslo study rose to 95% and to 85% in ours. In a case-control study from Melbourne, sensitivity of the Gram stain at ≥5 PMNs/HPF to detect C. trachomatis infection was only 68% among symptomatic patients.7 Thus, when testing men for chlamydia, we do not rely on the Gram/LMB stain but rather on much more sensitive nucleic aid amplification tests.
The primary reason to conduct a Gram/LMB stain in men with urethral discharge is to identify gonococcal urethritis (where these tests perform much better), which determines therapeutic choice and reduces the unnecessary use of antimicrobials specifically used for gonococcal infections. Given the rising concerns of gonococcal antimicrobial resistance, this remains an important rationale for the continued use of Gram/LMB stain in men with urethral discharge. Secondarily, the Gram/LMB stain is used to diagnose NGU, only a small proportion of which is caused by C. trachomatis. Even when accounting for other known urethral pathogens, including M. genitalium and Ureaplasma urealyticum, 20% to 40% of NGU cases have no identified pathogen.8 However, treatment of NGU is generally recommended and may provide relief regardless of etiology.9 This brings us back to how NGU should be defined and, in the absence of another (better?) test, at what Gram/LMB stain cutoff the diagnosis should be made. We suggested a pragmatic proxy measure based on chlamydia prevalence, but the study by Moi et al. shows that this leads to threshold differences when the background chlamydia prevalence is lower and/or when potentially more sensitive smear tests are used.
Finally, these considerations are somewhat academic as the widespread use of gonorrhea and chlamydia NAAT testing in combination with reduced funding for public STD clinics has resulted in many clinics in the United States scaling back their stat laboratories. Many of them no longer perform Gram/LMB stains, and some even treat urethral discharge syndromically. Advocating for a more invasive approach to obtaining urethral smears as described by Moi et al. is not likely to change these trends. In fact, the emergence of rapid, sensitive, and specific testing technology providing a gonorrhea diagnosis, and perhaps also a more sensitive and specific assessment of NGU at the point of care, may ultimately make the Gram/LMB stain obsolete. Until that time, many of us will continue to use an imperfect test for which a universally acceptable interpretation may remain elusive.
1. Rietmeijer CA, Mettenbrink CJ. Recalibrating the Gram stain diagnosis of male urethritis in the era of nucleic acid amplification testing. Sex Transm Dis 2012; 39:18–20.
2. Couldwell DL, Gidding HF, Freedman EV, et al. Ureaplasma urealyticum
is significantly associated with non-gonococcal urethritis in heterosexual Sydney men. Int J STD AIDS 2010; 21:337–341.
3. Berntsson M, Löwhagen GB, Bergström T, et al. Viral and bacterial aetiologies of male urethritis: Findings of a high prevalence of Epstein-Barr virus. Int J STD AIDS 2010; 21:191–194.
4. Bachmann LH, Manhart LE, Martin DH, et al. Advances in the understanding and treatment of male urethritis. Clin Infect Dis 2015; 61(Suppl 8):S763–769.
5. Centers for Disease Control and Prevention. Sexually transmitted diseases treatment guidelines, 2015. MMWR Rcomm Rep 2015; 64.
6. Moi H, Hartgill U, Skullerud K, et al. Microscopy of stained urethral smear in male urethritis: Which cut-off should be used? Sex Transm Dis 2017; This issue.
7. Bradshaw CS, Tabrizi SN, Read TR, et al. Etiologies of nongonococcal urethritis: Bacteria, viruses, and the association with orogenital exposure. J Infect Dis 2006; 193:336–345.
8. Martin DH. Nongonococcal urethritis: New views through the prism of modern molecular microbiology. Curr Infect Dis Rep 2008; 10:128–132.
9. Manhart LE, Gillespie CW, Lowens MS, et al. Standard treatment regimens for nongonococcal urethritis have similar but declining cure rates: A randomized controlled trial. Clin Infect Dis 2013; 56:934–942.