We've already seen that there's a need for more evidence to help guide clinicians through the diagnostic process. It's clear that diagnostic problems are generally more complex than determining whether or not someone has a disease. Research must therefore reflect this complexity, taking into account several aspects of the clinical problem it aims to address. A traditional approach to research in the field of diagnostics has been to assess the performance of a single diagnostic test, in order to determine whether it should be used in clinical practice. A new test, such as a point of care skin prick test or a new blood marker, may be trialed within a group of patients with and a group of patients without the disease. And the diagnostic accuracy of the single test is then assessed. Often this is done in controlled research conditions outside of the usual clinical setting. This kind of research, which we'll refer to as test research, still makes up the majority of published research in the field of diagnostics. But what's the intrinsic clinical value of this kind of research? Take, for example, the results of some test research that claims that a new immunological bio marker can be used to diagnose active tuberculosis. In such a study, often a group of patients with known active tuberculosis and a group of individuals without active tuberculosis, for example healthy volunteers, are included. Then the accuracy of this single test in diagnosing active tuberculosis is assessed and possibly compared with the diagnostic accuracy of an already established test. How useful are these findings for clinical practice? Well, how often is a diagnosis based on only a single test result using no other information? Not very often. And is it realistic that the tests were performed in patients who already had a confirmed diagnosis, and in healthy volunteers? No, not really. In almost all situations, a diagnosis in daily practice is not based on a single test. It's based on a diagnostic workup which consists of several pieces of information. This will include findings from the patient's medical history and from symptoms and signs and often will include results from laboratory tests, imaging or functional tests. Furthermore, in order to reflect clinical practice, it makes much more sense in a research setting to assess the performance of diagnostic tests in those patients in whom you're utmost likely to use these tests in clinical practice. That means patients you suspect of a specific disease, in this case, active tuberculosis. And not in patients with a confirmed diagnosis of tuberculosis, let alone in healthy volunteers you don't even suspect of having the disease. An alternative approach to test research, referred to by clinical epidemiologists as diagnostic research, focuses on conducting research that reflects the clinical setting and as a result doesn't suffer from the same limitations as test research. In contrast to test research, in diagnostic research it's recognized that a diagnostic workup is a multi-variable process. And so multiple tests are performed. And their diagnostic value is assessed in a sequential manner, starting with the more simple, readily available tests and then expanding to more advanced tests. Furthermore, the patients recruited in diagnostic research studies do not have a confirmed diagnosis but instead are suspected of having the disease of interest. The distinction between these two kinds of research is an important one and it can have a huge impact on the validity and applicability of the results that you see. So, it's now clear that if we want to address a diagnostic problem that we see in clinical practice, we'll need to take a diagnostic research approach. Now let's think back to our patient from the previous lecture. A baby girl who may or may not have an allergy to cow milk. We already decided that because there's a lack of existing evidence we need to conduct a clinical epidemiological study. So how should we go about designing our study so that it falls in line with the diagnostic research approach? First let's consider our research domain. Our clinical problem relates solely to patients who are suspected of having a cow milk allergy. We may want to restrict our research to young children if we're concerned that there may be a difference between age groups. From this it becomes clear that our study should only include young patients who are suspected of having cow milk allergy. For example, those presenting with symptoms and signs suggestive of cow milk allergy. In diagnostic research, it's also vitally important to define the health setting that our clinical problem relates to. Diagnosis of cow milk allergy starting from a primary care setting is quite different to diagnosis within a specialist setting, such as an allergy clinic. We're originally interested in diagnosing or excluding the disease in primary care. And so, this could also be part of the definition of our domain. Now, we need to think about the different tests, the determinants in epidemiology terms, that we want to measure in our study. The patient's medical history and findings from physical examination are essential pieces of information. The signs and symptoms of the patients that led to the suspicion of allergy will also be key. We also want to know whether blood markers can be used to make accurate predictions. So we'll need to measure these too. Importantly these tests will need to be measured in all of the patients in our study. And we'll need to think very carefully about how to do this so that our testing best reflects the way these tests are performed in clinical practice. As is always the case in clinical research, we need to think very carefully about how we'll measure the outcome in our study. In this diagnostic study, this means that we have to do our very best to assess whether the children actually have an allergy to cow milk or not. Ideally, we should know this with absolute certainty so that we can assess the accuracy of the diagnostic tests we're interested in. The gold standard test for assessing cow milk allergy is the oral food challenge test, and it's seen as a means of reaching a definitive conclusion. Therefore we can use this test to determine the outcome. But is measuring the outcome really that straightforward? Depending on how invasive and risky such a gold standard really is, it could be considered unethical to perform it in every individual in our study. And what about research problems where there's no such perfect test? It may be the case in diagnostic research that we're interested in finding an accurate combination of tests, because there's no clear way to reach a conclusive diagnosis. For this reason in diagnostic research it's common to use the term reference standard instead of gold standard when discussing the outcome measure. Because for most diagnoses there simply is no gold standard. How to measure the outcome on diagnostic research warrants a great deal of thought. And this highlights the need for careful planning when designing a diagnostic study. Obviously, the reference standard should be measured in all individuals included in our study. We've now carefully decided which information we'll need to gather from our patients. And we've begun thinking about how this should be done in order to reflect clinical practice and in tum providers with useful and realistic results. The next step will be to consider how we'll conduct this information. And then how we'll translate the information we collect into results that are useful for clinicians.