Contributed Commentary by Timothy E. Sweeney
January 10, 2020 | Antimicrobial resistance (AMR) and sepsis are among the most pressing public health challenges of our time, accounting for growing numbers of deaths and healthcare costs worldwide. A recent report warns that AMR could cause over 10 million deaths globally per year by 2050, while another review estimates that sepsis is responsible for over 5 million deaths annually.
Unfortunately, simultaneously addressing both AMR and sepsis in the clinic can be challenging. Good antibiotic stewardship promotes prescribing fewer antibiotics and only after careful diagnostic testing, while sepsis guidelines recommend prescribing broad-spectrum antibiotics earlier for patients with suspected bacterial infection. Underlying this tension is a fundamental diagnostic problem: When a patient presents with signs and symptoms of infection, there are few tools available to accurately guide antibiotic prescribing that are fast enough for rapid workflows. For example, emergency room physicians struggle with decision-making in a large fraction of the approximately 15 million annual patients with suspected acute infection and sepsis.
Diagnostics for acute infections typically either seek to identify pathogens directly or measure specific host biomarkers to try to establish a diagnosis indirectly. Both methods have major shortcomings.
Pathogen testing is typically slow, and most locations in the body are not easily sampled (although some new respiratory pathogen tests run relatively quickly). For example, there is no easy way to sample a belly infection at the bedside. Further, while an enormous effort has been made to improve bloodstream infection testing, most patients with infections do not have pathogens that can be detected in the blood, so a negative result does not rule out the need for antibiotics. Take, for instance, a patient in the emergency department with a radiographically confirmed pneumonia. Today, such a patient might get a nasal respiratory pathogen panel, sputum culture, blood culture, and urine antigen tests, and yet still find no pathogen 62% of the time (https://doi.org/10.1056/NEJMoa1500245).
Biomarker testing is typically faster, but common biomarkers such as C-reactive protein (CRP) and procalcitonin, are nonspecific for infection and not accurate enough to be routinely actionable for antibiotic prescribing. Thus, in acute settings, such as clinics, urgent care centers and emergency departments, physicians choose whether to initiate antibiotics without confirmatory testing, and typically choose which antibiotics to use based on guidelines and local antibiograms, not on microbiological data. Data suggest that physicians frequently guess wrong, both in choosing whether to prescribe, and in selecting which antibiotics are likely to be optimal.
The need for more accurate diagnostics for acute infections is clear, and several novel and promising diagnostic technologies have come to market recently, with more in the pipeline. More importantly, new advances in acute infection diagnosis are creating opportunities to ensure that the right patients get the right treatments in a cost-efficient manner. This more effective and efficient workflow will be critical to tackling AMR and sepsis.
Broadly, new approaches for acute infection diagnosis can be split into three categories: (1) rapid pathogen detection, (2) rapid pathogen antibiotic susceptibility testing (AST), or (3) host-response measurements. Some technologies are combining rapid pathogen detection and rapid AST, typically genotypic (detection of resistance markers) not phenotypic (direct measurement of bacterial growth inhibition by antibiotics). Elsewhere there are excellent reviews of rapid ID and AST and host response technologies.
To understand the role that these approaches may play in antibiotic prescribing, it is important to consider the decisions that physicians must make for patients with suspected infection. These are, namely: Whether to initiate antibiotics, which antibiotics to choose, and when to stop treatment. Physicians may use pathogen-detection and host-response technologies at all phases. That said, the primary utility of host-response is in choosing whether to start antibiotics and when to stop them, while the primary utility of pathogen-detection (and especially AST) is in choosing the right antibiotics for the patient.
(Novel host-response technologies that can rapidly and more accurately identify bacterial and viral infections may show improved utility for both starting and stopping antibiotics but have not yet entered the market. Inflammatix’s HostDx tests fall into this category.)
Rapid pathogen detection may sometimes aid in initial prescribing (e.g. Gram status), but my personal bias is that genotypic AST is likely to be viewed as a rule-in test (e.g., prescribing vancomycin when MecA is detected) rather than a rule-out test (e.g., choosing to prescribe methicillin because MecA was not detected). Many physicians may be reluctant to ‘downgrade’ to a narrow antibiotic without phenotypic AST. However, due to the need to capture or grow enough bacteria to assay them phenotypically, such technologies will always take longer than the genotypic approach. Thus, there will likely always be a tradeoff of speed for actionability, and different AST technologies will fill the space in different ways.
Taking into account turnaround time and likelihood of clinical impact, future diagnostic workflows are likely to differ dramatically from today’s practice. In the future, perhaps testing would start with a rapid (<30 minute) host-response test that identified the patient as having only a viral infection or only a bacterial infection (aiding in whether to prescribe). Host-response indicators that suggest a bacterial infection would be followed up with immediate initial antibiotics and then rapid (1-6 hour) direct pathogen detection and AST, such that the patient may only get one broad-spectrum dose before being narrowed to appropriate therapy. Host-response monitoring could then help with the decision on when to stop (or switch to orals and discharge). In occasional severe or chronic infections, follow-up testing with metagenomics for pathogen nucleic acids (hours to days) may aid in identifying rare infections.
It is clear that patients with bacterial infections will benefit from early, accurate antibiotic administration, and that this will aid in reducing morbidity and mortality from sepsis. It is also clear that avoiding antibiotics in patients who do not need them will decrease antibiotic-associated adverse drug events, reduce costs, and reduce AMR. A combination of rapid host-response profiling and AST should be just what the doctor ordered.
Timothy E. Sweeney, M.D., Ph.D., is cofounder and chief executive officer of Inflammatix, Inc. Inflammatix is a molecular diagnostics company building rapid, novel host-response tests for acute infectious and inflammatory conditions. Dr. Sweeney trained in general surgery at Stanford before cofounding Inflammatix. He can be reached at firstname.lastname@example.org.