Utility of the Xpert MTB/RIF Assay for Diagnosis of Tuberculous Meningitis

Summary

article has not abstract

Tuberculous meningitis (TBM) is characterized by copious cerebrospinal fluid (CSF) inflammation and yet few Mycobacterium tuberculosis. This combination creates a disease that is notoriously difficult to definitively diagnose. In this week's issue of PLOS Medicine, Patel and colleagues report the diagnostic performance of the GeneXpert system's Xpert MTB/RIF assay for the diagnosis of TBM in a cohort of 204 South African, predominantly HIV-infected, adults presenting with suspected meningitis of whom 59 had definitive TBM [1]. The Xpert MTB/RIF assay's overall sensitivity was 62%, and specificity was 95%. The performance was better using larger volumes of centrifuged CSF among HIV-infected persons with sensitivity of approximately 80% and excellent specificity for microbiologically confirmed TBM. Xpert MTB/RIF performance was less impressive using uncentrifuged CSF with a sensitivity of ≤50%, and Xpert MTB/RIF performance was negligible in HIV-uninfected persons.

What Is GeneXpert?

The GeneXpert System (Cepheid) is a single use cartridge-based real-time PCR fully automated system that performs sample decontamination, sonication, automated nucleic acid amplification, and fluorescence-based quantitative PCR [2]–[4]. The Xpert MTB/RIF assay, developed by David Alland, detects M. tuberculosis DNA in approximately 2 hours with minimal hands-on time [3]. This new technology was endorsed by the World Health Organization in December 2010, and as of June 30, 2013, a total of 1,402 GeneXpert instruments and over 3 million Xpert MTB/RIF cartridges have been procured in the public sector in 88 countries [5]. The concessional pricing is US$9.98 per cartridge for 145 low- and middle-income countries [6]. The same GeneXpert platform also can be used for a variety of US Food and Drug Administration (FDA)-approved testing (e.g., influenza, Clostridium difficile, methicillin-resistant Staphylococcus aureus).

What Is the Performance of Xpert MTB/RIF?

There is a rapidly emerging literature regarding the performance of the Xpert MTB/RIF assay. Fundamentally, the sensitivity depends on the burden of organisms and thereby the target DNA present in the specimen. The published Xpert MTB/RIF detection threshold is approximately 100–130 colony forming units (cfu)/ml of sample [2],[3]. Patel and colleagues observed a similar threshold of 80–100 cfu/ml of CSF in this study [1]. In comparison, the detection threshold is <10 cfu/ml for mycobacterial liquid culture and is >5,000 cfu/ml for Ziehl-Neelsen staining for acid fast bacilli (AFB) via standard microscopy in sputum [7]–[9]. In real world clinical terms, this means 98%–99% detection by Xpert MTB/RIF of AFB smear-positive pulmonary TB, and approximately 75% detection of smear-negative, culture-positive pulmonary TB [3],[10],[11].

The threshold of detection is a key principle. The Xpert MTB/RIF test performs better when there is a larger burden of infectious organisms present in the specimen being tested. Yet, TB meningitis is a paucibacillary condition with few organisms. More organisms are likely present when the host is immunocompromised, or when a larger input volume is used for the test. Thus specimen centrifugation can compensate and should improve diagnostic yield, as demonstrated by a 35% improvement in sensitivity in this study [1].

The prior data on Xpert MTB/RIF testing of CSF are limited. In India, the Xpert MTB/RIF assay detected two of seven culture-positive specimens using an input volume of ∼1 ml [12]. In an Italian study, 11 of 13 TBM patients were detected by Xpert MTB/RIF using an input volume of 2 ml into the cartridge without the standard N-acetyl-L-cysteine-sodium hydroxide (1% NALC-NaOH) decontamination and mucolytic step (i.e., GeneXpert Sample Reagent) [13]. This sample reagent was designed for sputum samples. Numerous commercially available PCR assays exist for other pathogens (e.g., herpes simplex PCR) without such a decontamination step [13], and the necessity of using the sample reagent for non-bloody CSF is unclear.

Public Health Significance

Although two commercial TB PCR tests previously existed [14], the innovation is that the GeneXpert platform is fully automated and is being rolled out in low- and middle-income countries. Thus, GeneXpert is an actual technology that can be—and is being—widely used globally. However, immediate implementation of a US$10 Xpert MTB/RIF assay for all cases of meningitis is unwise and unsustainable. Further research is needed on how best to incorporate the Xpert MTB/RIF test into diagnostic testing for meningitis, to ensure that it is a cost-effective intervention that improves health and does not waste resources.

Patel and colleagues modeled a clinical score to predict who had such high pretest probability of TBM that Xpert MTB/RIF was unnecessary to perform [1]. Yet health systems also need the opposite, a clinical score or algorithm to identify who has such low pretest probability that they do not require testing. Several investigators have developed meningitis diagnostic algorithms, yet broader validation is needed [15]–[17]. Ordering comprehensive testing of all the available diagnostic tests for every patient with suspected meningitis, including Xpert MTB/RIF testing, is 3–4-fold more expensive, without any better diagnostic yield than a targeted stepwise approach [15].

Key Principles of TB Meningitis Diagnosis

Despite molecular diagnostics, there remain a number of key pieces of information that inform clinicians as to the likelihood of a TBM diagnosis, so as to target testing in a cost-effective manner. The first is history. TBM is a subacute illness. Symptoms <6 days are atypical for TBM, yet near universal for bacterial meningitis [16]. Second is the immunology of the patient. Immunosuppression due to HIV/AIDS or age (e.g., infants, elderly) are key drivers of TBM, and immunosuppression increases the bacillary burden of M. tuberculosis organisms, likely increasing the diagnostic yield of molecular testing. In the current study, the GeneXpert performed poorly among HIV-uninfected persons, and the performance in children is unknown. Third is the CSF profile. TBM is classically a lymphocytic meningitis (i.e., >30%–50% lymphocytes in >90% of persons [16],[18]); with a low CSF glucose of <60% of serum glucose or an absolute CSF glucose concentration <2·2 mmol/l (<40 mg/dl) in >92%–95% [18]–[20].

In HIV-infected adults, the clinical history and CSF profile overlap extensively with meningitis due to Cryptococcus neoformans, and cryptococcal meningitis is overall the most common meningitis etiology in adults in sub-Saharan Africa [15]. Thus before a US$10 Xpert MTB/RIF test is performed for a less common condition, a US$2 cryptococcal antigen lateral flow assay should likely be performed for a more frequent condition [15].

If there is insufficient CSF volume available for testing (i.e., <3 ml), in a clinically stable patient treated presumptively for bacterial meningitis a repeated lumbar puncture in 48 hours is likely a better strategy than sub-optimal Xpert MTB/RIF testing using a limited volume. A repeat lumbar puncture can collect a sufficiently large volume as well as reassess CSF glucose. At 48 hours, the CSF glucose should have risen by >100% of the initial level in treated bacterial meningitis [16]. Persistently low CSF glucose levels at 48 hours coupled with excluding cryptococcal meningitis should prompt Xpert MTB/RIF testing and/or empiric anti-TB therapy [16].

Xpert MTB/RIF appears to be a highly useful test to “rule in” the diagnosis of TBM, yet the clinical acumen of physicians remains a necessity for the wise use of any new diagnostic test. Careful application of these new diagnostic tools should improve clinicians' ability to deliver timely, cost-effective care to patients with suspected TBM throughout the world, an approach that future studies should systematically evaluate.