We have the technology, but should we build the test?

KEYWORDS:

• Molecular diagnostics
• clinical utility
• health outcomes
• sexually transmitted infections
• microbiome

‘May you live in interesting times’ has long been considered to be a curse, and those of us in the field of molecular diagnostics are doing just that. With the exciting and exponential increase in molecular technologies, and during this era of microbiome exploration and next-generation sequencing, we are faced with incredible opportunities for new and improved molecular diagnostic assays. However, we must temper our enthusiasm lest we forget the meaning and importance of clinical utility. Here I will use examples from the field of sexually transmitted infections (STI) to illustrate the need to make rational decisions about the assays that we develop and that clinicians order. While I will focus on STI examples, these arguments apply to numerous potentially infectious or pathogenic organisms. I will argue that just because we can build an assay to identify a previously difficult to detect organism, does not necessarily imply that we should do so.

Genital microbiome studies, and particularly those looking at vaginal microbiota, have been numerous over the last two decades and provide increasingly greater depth of analysis regarding the largely hard to cultivate organisms that reside in this niche. It has been easy to identify organisms (e.g. Mycoplasma genitalia, Prevotella spp. and others) that are statistically associated with poor vaginal health. However, considering any one potential pathogen without taking into account the overall microbial community composition (the presence/absence of organisms and the relative abundance of each), the opportunity to develop an meaningless diagnostic assay looms large. One of the first difficulties in shifting our focus to community structure was to distinguish the community composition of healthy women from that of women with a disease such as bacterial vaginosis (BV) [1,2]. BV is a particularly appropriate example since this is a syndrome that is multifactorial, and in many cases, may result from a shift in the balance of organisms rather than a transmission event. During most cases of BV, the ratio of Lactobacillus spp. to Gardnerella vaginalis shifts so that the latter becomes predominant. As a result, identifying the presence of G. vaginalis with a highly sensitive qualitative (present/absent) molecular assay may not be clinically meaningful since this organism can also be found in vaginal samples from healthy women. Similarly, yeast ‘infections’ occur in women when an overgrowth of Candida spp. which are normal components of the microbial community in the vagina. These are examples of when it is necessary to consider the entire community composition (including presence/absence and organism load or relative abundance) rather than targeting amplification of a single target organism.

In the case of BV, it is more appropriate to utilize a panel of organisms that are implicated in BV and assess the ratios of those organisms to Lactobacillus spp. and to each other. This approach requires informatics programming support to analyze data and reach a clinical determination. Assays are becoming available that use this approach [3], however, achieving evidence of clinical utility has been quite difficult as a result of the poor comparison standards available for diagnosing BV [4]. When it is difficult to determine the true status of a patient (e.g. is asymptomatic BV clinically meaningful and does it require treatment?), it is difficult to understand the meaning of newer tests that may have exquisite sensitivity, but may not provide actionable results.

One of the organisms in many BV diagnostic panels is Mycoplasma hominis. This organism illustrates yet another problem facing development of molecular diagnostics: reliance on data that only generates statistical associations. Early data in the literature regarding M. hominis, when culture was used for detection, describe an association between this organism and BV [5]. Studies showed an increased prevalence of M. hominis among women with BV compared to those without BV. However, despite the fact that this organism is not sensitive to treatment regimens used for G. vaginalis, such treatment actually decreased the likelihood of recovering M. hominis. Several theories offer explanations for this finding, but the consensus is that that M. hominis is likely a biomarker of infection rather than the cause, or that the role of M. hominis is wholly dependent on organism load [6]. Further, with access to molecular detection methods, we now know that as many as 10–15% of women harbor M. hominis in the absence of signs or symptoms of clinical disease [7,8]. However, this has not stopped manufacturers from developing assays that report on the presence/absence of M. hominis or clinicians from ordering such tests [9]. At this time we do not really understand the clinical utility of treating for this organism and in the era of antimicrobial stewardship, we should approach prescriptions of antimicrobials in situations such as this with caution. Given the high prevalence of antimicrobial resistance in other mycoplasmas, treating M. hominis may be doing more harm than good.

In a similar vein, there has been a proliferation of mycoplasma diagnostics targeting other species and some of these assays are being offered commercially in many parts of the world. Ureaplasma urealyticum and Ureaplasma parvum molecular diagnostics are excellent examples of laboratory tests with unproven clinical utility that are widely used [10]. The history of these two ureaplasma organisms is quite murky and not often fully taken into account in epidemiologic studies. U. parvum and U. urealyticum were members of a single species (U. urealyticum biovar 1 and biovar 2, respectively) until about 2 decades ago and thus the literature about associations with genital diseases such as cervicitis and urethritis is not straightforward. At this point, it seems clear that U. parvum is nearly ubiquitous while U. urealyticum is statistically associated with cervicitis and non-gonococcal urethritis. Further, we do not have evidence of efficacy of currently used treatment regimens. Many cases are asymptomatic and the need for treatment in these cases is unclear [8]. Why then are diagnostic assays available, and widely used in southern and eastern Europe, that test for both of these organisms and why are clinicians using these tests for screening of asymptomatic patients?

Finally, the example of Mycoplasma genitalium (MG) provides yet one more example of the ability of technology to outpace our understanding of clinical utility. MG was first isolated from men with urethritis in the United Kingdom in the early 1980’s. We now know that this organism is likely transmitted between sexual partners and is strongly associated with discharge syndromes such as urethritis and cervicitis. In some populations, mycoplasma is more often identified in men with non-gonococcal urethritis than with chlamydia. However, this organism is extremely difficult to detect using culture and we now rely on molecular diagnostic tools [11]. Several assays are commercially available worldwide and are often used to screen for infections in asymptomatic patients. This is not consistent with the recommendations for diagnosis of mycoplasma, but the testing is often bundled with chlamydia/gonorrhea screening which is performed in asymptomatic populations. Thus, despite recommendations, the availability of molecular tools may be encouraging screening that could ultimately lead to selection pressure and drive continued increases in antimicrobial-resistant M. genitalium.

Not only do we now have assays that detect MG, newer tests can identify genetic markers of resistance to macrolides [12–14]. In longitudinal studies that track the presence or absence of MG as well as the resistance profile, data suggest that women may clear this infection without treatment. Further, even when the molecular assays demonstrate the presence of the macrolide resistance marker, patients may respond to treatment with Azithromicin [15]. How then do we interpret the resistance gene marker information? In the absence of information about the ratio of sensitive to resistant organisms, it is difficult to determine whether there is a threshold above which a positive result for a resistance marker has clinical relevance. Should we argue for changes to treatment guidelines based solely on molecular diagnostics? Will this argument then applies as well to Neisseria gonorrhoeae? When using a highly sensitive amplification technology, we run the risk of identifying a (potentially) irrelevant minority population that has a single resistance marker. Alternatively, we could identify a minority population with a susceptibility, or wild type, genetic marker and falsely conclude that no resistance will occur. We must keep in mind that with all molecular diagnostics that are currently commercially available, the result does not give us an indication of the viability of the organisms detected. Newer technologies have the capacity to assess viability and these will provide an improvement in our application of the results of these tests to clinical patient management.

The difficulty arising from not understanding the meaning of molecular test results, is that it is easy to imagine a situation when a clinician may feel that they are better off relying on clinical observations. For example, clinicians may focus only on patients presenting with signs of disease. Unfortunately, for most STI, and many other infectious diseases, patients may not exhibit symptoms and many STI present with very similar symptoms and co-infections are common. Therefore, molecular tests for screening or diagnostics are usually the most accurate method for managing patients at risk for these infections. Given our understanding of the important personal and public health consequences of untreated, asymptomatic infections (at least for chlamydia, gonorrhea, trichomonas and potentially mycoplasma), and the need to manage antibiotic use to avoid development of resistant organisms, identifying specific infections and targeting treatments accordingly is critical. Thus, we need to continue to develop new diagnostic methods, especially those that would provide either quantitative or viability results, but we must have data to support the clinical utility of these assays before they are widely adopted.

We have a phenomenal capacity to detect pathogens and genetic markers of interest and we learn of more potential organisms to target from microbiome and whole genome sequencing research every day. What is missing in our inexorable march to improved diagnostics is support for studies that would allow us to better understand clinical outcomes and thus the relevance of laboratory data. We need longitudinal data that demonstrates whether serious sequalae arise when asymptomatic people are not treated for infections. We need data regarding treatment efficacy based on clinical and microbiological outcomes and not just molecular detection of organisms. In this age of electronic medical records, some studies of clinical utility should be possible without the need for extensive, expensive clinical trial cohorts. Undertaking this type of work to provide information about clinical utility should fall on all of us in the health-care sector working collaboratively. A laboratory developing a new diagnostic assay may not have the resources to perform a clinical utility investigation, but in collaboration with public health agencies who should recognize the need for such studies, data could be obtained. Nor will diagnostic manufacturers take on the burden proving clinical utility, but they would likely be willing to work with public health agencies to achieve this common goal.

Ultimately, molecular diagnostic tests have revolutionized our ability to provide timely, sensitive, and specific identification of pathogens and causes of multifactorial syndromes. These examples from STI could be applied to current assays being developed to assess gut health using molecular tools and, in any of a number of other diagnostic areas. We should not try to hold back the tide of new assay development for the molecular laboratory, but we should not view new assays as a panacea for diagnostics in the absence of strong clinical utility and outcomes data. In the molecular diagnostics era, it is more critical than ever for laboratorians and clinicians to work together to unravel the complex meaning of laboratory results in order to, ultimately, improve patient outcomes.

Declaration of interest

The author received support from Abbott Molecular, Atlas Genetics; BD Diagnostics; Cepheid; Hologic; Luminex; Rheonix; Roche Molecular; SpeeDx. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript.

Reviewers disclosure

Peer reviewers on this manuscript have no relevant financial relationships or otherwise to disclose.

Additional information

Funding

This paper was not funded.