Evidence-based laboratory medicine

Background

Evidence-based laboratory medicine (EBLM) is an approach to medical practice that integrates the best available research evidence about laboratory investigations with the clinical expertise of clinicians, to improve the health and health care outcomes of individual patients. Practicing EBLM enables laboratory professionals to translate test results to clinically relevant information that helps clinicians in delivering effective and cost-effective patient care.

Content

This chapter provides an overview on how evidence about laboratory tests is generated, how it is synthesized, and how it can be applied to questions about diagnosis, screening, prognosis, or monitoring. The topics covered here introduce the reader to the methodological and practical aspects of EBLM. They include (1) the process and methods of practicing EBLM, (2) the key components and types of evidence used in the evaluation of biomarkers, (3) tools for the assessment of the validity and applicability of the evidence, (4) key aspects of synthesizing the evidence in systematic reviews and meta-analyses, (5) basic principles of how EBLM is applied to other purposes of testing than diagnosis, (6) the challenges and tools of implementing the evidence for achieving best laboratory practice, and (7) the history and future challenges of EBLM.

Identify the clinical problem

The identification of a clinical problem is both the starting point and the foundation of the service provided by the health care professional.

In EBLM, clinical problems are dilemmas about the care of patients for a health professional that involves laboratory tests. A dilemma relates to a choice between actions; between doing one thing rather than another. In laboratory testing these are questions about whether or not to test, or issues about selecting one test rather than another.

An example of a test-related clinical problem is the following:

A general practitioner, Dr. Ann Brodie, is seeing a patient, Mr. Smith, a 69-year-old male, who presents with shortness of breath during exercise, increasing fatigue and weakness, and some swelling around the ankles. Mr. Smith has been a patient for many years. Dr. Brodie knows the history of Mr. Smith very well, including that he had no cardiovascular events so far. Dr. Brodie suspects congestive heart failure but hesitates about her next steps: sending Mr. Smith to the closest hospital to see a cardiologist, while she knows Mr. Smith does not like hospitals. Mr. Smith’s basic risk factors seem well controlled, and Dr. Brodie also considers the downsides associated with overtesting, should the cardiologist decide to run a large battery of tests. On the other hand, a missed diagnosis of congestive heart failure may also have clear consequences for Mr. Smith: if correctly identified as having heart failure, treatment is likely to improve his symptoms, and reduce his long-term risk of adverse events.

An alternative Dr. Brodie might consider is to order an N-terminal prohormone of brain natriuretic peptide (NT-proBNP) blood test. For this, Dr. Brodie has to consider how the test may guide her next steps. If she plans to send Mr. Smith to the cardiologist, regardless of the test result, then the test is obviously unnecessary. The NT-proBNP test can only be useful if it generates a result that will change her actions, compared to the default option; that is, sending Mr. Smith to the cardiologist. In this case, she may decide not to refer Mr. Smith if the NT-proBNP result is less than 300 ng/L (36 pmol/L).

Now we clearly have a clinical problem: a dilemma, well defined as a choice between two lines of action: the first option is to send the patient to a cardiologist in the nearby hospital for further evaluation, the second is to only do so if the NT-proBNP result is positive, and not to refer Mr. Smith if the result is negative.

Essential for EBM is that the solution for clinical problems like this one is driven by a consideration of the likely consequences of each line of action. What is likely to happen to Mr. Smith if he is sent to the cardiologist? How likely is this to affect his health, now and in the future? What is likely to happen to him if he is not sent to the cardiologist, now at least? How likely is this to affect his health, now and in the future?

In the interest of Mr. Smith, the line of action should be adopted that gives him the best options for a better health, now and in the future. Yet decisions should also take into account the consequences for others involved; for example, for his spouse or family, for society as a whole, which has to cover the costs of testing, further evaluation, treatment, and management of future events.

Ask the right question: The patients-intervention-comparator-outcomes format

As a laboratory professional practicing EBLM, you offer your help to Dr. Brodie. To help her solve the clinical problem, you both should be asking the right question.

Doctors and other health care professionals are constantly asking questions, and these can be divided into background questions and foreground questions.

Background questions, to date, are more commonly asked by newly qualified professionals by virtue of the ways they have been taught. Such questions typically deal with knowledge (or underlying science) of the pathophysiology or clinical presentation of a condition (e.g., Why is the circulating concentration of troponin I increased in acute coronary syndrome (ACS)? Why and how NT-proBNP is released into the circulation in congestive heart failure?).

Foreground questions are related specifically to the application of knowledge, of experience in treating the condition and using tests (e.g., Will a troponin I measurement help me determine whether this patient is suffering from an ACS? Does measurement of NT-proBNP help diagnose heart failure in addition to clinical signs and traditional diagnostic investigations in patients presenting with acute shortness of breath?).

Clinicians tend to ask more foreground (and fewer background) questions as their experience develops. This may change in coming years as a more evidence- and outcomes-based approach to teaching medical students, and training doctors, evolves.

Richardson and coworkers argued that all clinical problems could be expressed in the form of a question and went on to describe the PICO framework for formulating an answerable question :

The four letters P-I-C-O refer to the four elements of the question:

• P stands for “Patients”

• I stands for the “Index Test,” or the “Intervention” that is considered

• C stands for the “Comparator”

• O stands for the patient-relevant “Outcomes.”

In laboratory medicine we look at an Index test: the test that is being evaluated. This can be an in vitro medical assay to measure a biomarker (e.g., B-type natriuretic peptide, or high-sensitivity Troponin, or fecal hemoglobin), or a combination of such tests (see definitions and examples in Table 10.1 ). Alternatively, we can look at a strategy, in which the index test is used to guide further actions, as in a screening strategy.

When the intervention is considered in the context of laboratory medicine, it is worth considering the purpose of testing: for example, is the test for (1) screening, (2) diagnosis, (3) prognosis, or (4) monitoring of a condition (see definition and examples in Table 10.1 ).

An additional relevant question is about the role of the test in the clinical pathway, relative to other possible forms of testing (see definition and examples in Table 10.1 ). Is this test used for triage , for example, to prevent patients from undergoing other, more expensive, or invasive tests? In that case, only patients testing positive on the triage test will undergo these more expensive or invasive tests, and the ones who are triage test negative will not. An alternative role of a test considered in the clinical problem is replacement : should I do this test, rather than another? A third role of a test is an add-on one: should I add this test to one or more tests, already performed, or should I stop after this testing strategy?

In the example of Dr. Brodie, who considers NT-proBNP testing for Mr. Smith, the purpose of testing is clearly a diagnostic one. Mr. Smith decided to see his general practitioner because of the way his shortness of breath was affecting his life; he wanted relief of his symptoms (i.e., health outcomes). Considering the cause of these may be a steppingstone toward effective treatment.

The outcomes refer to patient-relevant health outcomes that the intervention is intended to influence. This could be removing symptoms and restoring health or preventing premature death or worsening of symptoms. However, in laboratory medicine we also often consider surrogates for these outcomes, such as a target condition that is being detected (e.g., heart failure), or a target event that could happen in the future (e.g., markers in patients with ACS, to predict the risk of cardiovascular events after discharge).

One crucial observation must be made regarding these well-phrased questions. In all examples testing itself will not directly affect the outcome considered. NT-proBNP testing will not directly reduce the severity of symptoms, and fecal hemoglobin testing will not directly remove any form of colorectal cancer. The effect between testing and outcomes is an indirect one: testing will be used to guide downstream actions, such as starting or stopping interventions, or communicating results to patients. Whether these downstream interventions are effective, and whether they are done in the right patients at the right time, will eventually determine the effect of testing on patient outcome.

In the example of fecal hemoglobin, a positive test result will be used to invite some screening participants for further investigation with colonoscopy. The colonoscopy will be used to look for cancer, advanced adenoma, or other precursor lesions, and to remove them, if possible.

This implies that it is often necessary, when phrasing the PICO type question, to define the clinical pathway : the typical chain of actions in the process of care that is guided by results testing, and able to affect the relevant outcomes.

The comparator is an alternative test, or an alternative intervention: what would we do if we do not apply the form of testing considered under the “intervention”? This could be another test (e.g., the GRACE score in the case of unstable angina patients), a clinical action (referral for patients with suspected heart failure, seen by the general practitioner), or no intervention right now (as in screening for colorectal cancer).

In EBLM, questions about testing are never considered in a void, but always for specific types of patients , in defined settings (or, perhaps more importantly, at specific points in a diagnostic pathway). In the example of Mr. Smith, again, we must ask ourselves what the defining features are in the clinical question. We know, for example, that he presents himself to a general practitioner with symptoms. That means that studies about NT-proBNP testing in symptomatic patients are relevant, while studies about screening with NT-proBNP in nonsymptomatic elderly people are not. This distinction is relevant because the clinical performance of NT-proBNP and its ability to improve health outcomes is not identical in symptomatic patients, compared to nonsymptomatic persons.

For a specific clinical problem, there may be just one PICO-style question, but it is also possible that the problem leads to two or more PICO style questions. This typically happens if we have multiple outcomes to consider, for example, short-term consequences versus longer-term effects, or positive outcomes (benefits) versus negative ones (harms).

An appropriate definition of the well-phrased question, designed to help solve the clinical problem, will guide the search for relevant evidence from sound research. Table 10.2 contains several structured questions in this PICO format for these various purposes of testing.

Acquire the evidence

Once we have phrased a well-defined question, we can look for the available evidence. Here a number of options are presented, in decreasing order of strength, according to the “5S” hierarchical structure, described by Brian Haynes. In this hierarchy, (1) original studies are at the base, followed by (2) syntheses (systematic reviews) of the evidence, (3) synopses of studies and syntheses, (4) evidence summaries (e.g., guidelines), and (5) the most evolved evidence-based information systems at the top ( Fig. 10.2 ).

Busy clinicians and laboratorians prefer having information and knowledge readily available right next to the point of health care delivery. With rapidly advancing information technology, electronic decision support systems can now be built and integrated into electronic medical records, but the lack of high-quality evidence and evidence-based guidelines related to laboratory testing still limits the availability of such smart solutions.

A few initiatives, such as a mobile application of a Partial Thromboplastin Time Advisor at the US Centers for Disease Control and Prevention, have already been successfully implemented and a free app has been made accessible by Apple via their iTunes App Store. For more information on clinical decision support systems and their application to diagnostic testing the reader is referred to a collection of useful resources ( Table 10.3 ).

Currently the busy health care professional more often turns to evidence-based practice (EBP) guidelines, where professional colleagues have used the principle of EBLM to develop recommendations for practice. We will discuss evidence-based guidelines below.

If such guidelines are not available, you may turn to critically appraised evidence synopses, which can be found in evidence-based journals, such as BMJ and EBM (see Table 10.3 ).

If you have no success still, move to the next level in the hierarchy as there may be others who have systematically searched and synthesized the available literature for you in form of a systematic review or meta-analysis that answers your PICO-type question. Unfortunately, few databases exist for systematic reviews that address diagnostic testing; a few resources are listed in Table 10.3 .

If you cannot find evidence-syntheses produced by others, you need to turn to the primary literature and search for reports of individual studies that can help answer the PICO type question. Various tools have been produced to help professionals search the available literature. We discuss a few of these below.

Appraise the evidence

If we have found forms of evidence synthesis, or single studies, we must carefully examine the validity and applicability of the available evidence. This means that we cannot just read the bottom lines of the discussion section of these study reports to find out the answer to the clinical question; we must study the research that has been performed.

Apply the knowledge gained from the evidence

Once we have found the evidence and have identified any limitations in applicability or validity, we can try to resolve the problem by combining the various pieces of evidence. There may be two or more PICOs addressing multiple outcomes.

Evidence in itself does not always tell us what we have to do. Whether the evidence applies, and whether evidence is sufficient to take or to recommend action, depends on values and preferences. Balancing harms and outcomes, or health gains and costs, can lead to different recommendations, based on differences in individual goals in life, or guided by the availability in scarce resources. What may be sensible for a young man may no longer be recommended for an elderly male, and what seems reasonable in a richer country may seem unattainable in resource-poor settings.