Distraction From Randomization in Diagnostic Research
Introduction
In almost every system to grade epidemiologic studies according to their level of evidence, randomized studies or meta-analyses of randomized studies receive the highest classification. In epidemiologic research, randomization comes into perspective when two (or more) etiologic determinants are compared and evaluated on their ability to affect patient outcome. Hence, randomization is particularly required in etiologic studies of preventive or therapeutic strategies, i.e., intervention studies of intended treatment effects. Randomization prevents so-called confounding by indication. Provided a large enough number of subjects and properly executed, it ensures similar distribution of all (other) risk factors across the studied intervention groups. Any observed difference in patient outcome between both groups then can be attributed to the intervention studied. This intervention can be a single intervention, combination of interventions, or test–treatment intervention.
The ultimate goal of diagnostic testing is, like all medical care, to improve patient outcome. Hence, it has been advocated widely that when establishing a test's diagnostic accuracy, the impact of the test on patient outcome also must be quantified. Consequently, the paradigm of a randomized design often is assigned to diagnostic research questions, as well 1, 2, 3, 4, 5, 6. Diagnostic questions are not etiologic questions, but randomization transforms a diagnostic study into an etiologic or intervention study, which may not be necessary in many instances. Type of diagnostic question and existing knowledge determine the appropriate study design 7, 8. In this report, we describe the instances for which a randomized comparison is no prerequisite to quantify the beneficial effect of a diagnostic procedure or test on patient outcome and when it is indicated.
A diagnostic test, whether a patient history or physical examination finding, radiologic or electrophysiology examination, or laboratory measure of a biologic component, can be used for several reasons. It can be used to screen for disease presence in asymptomatic persons (screening), detect disease presence in patients suspected of that disease (diagnosis), monitor treatment success, and confirm previous test results. In the present report, we focus on the use and evaluation of tests for diagnostic purposes, i.e., on their ability to determine the presence or absence of disease in suspected patients.
Section snippets
Diagnostic Practice
Setting a diagnosis in itself is not a therapeutic process, but rather a vehicle to guide therapies. A diagnosis in itself is a label that, despite being highly valued by medical professionals, is of no consequence to a patient other than to set the course of therapy. A diagnostic test commonly has no direct therapeutic effects and therefore does not directly influence patient outcome. After a diagnosis, or rather the probability of the most likely diagnosis, is established and assessment of
Diagnostic Research Without Randomization
Viewing diagnostic practice as defined, most diagnostic studies or test evaluations include cross-sectional studies quantifying the accuracy of a new or existing diagnostic test to predict the presence or absence of a disease as determined by the existing reference test in suspected patients. The reference is the best available method to determine disease presence or absence 11, 12, 14. Commonly, there is no patient follow-up (only as ultimate reference test), and patient outcome is not
Randomization In Diagnostic Research
Situations in which randomized follow-up studies of patient outcome are needed to properly quantify the value of a diagnostic test on patient outcome include the following examples:
- 1.
When there is no established (or imperfect) reference test for the disease under study, such as depression, irritable bowel syndrome, and congestive heart failure;
- 2.
The diagnostic technology under study might be better to the extent that it provides new information potentially leading to other treatment choices than
Concluding Remarks
Various research methods have their particular advantages and disadvantages. The popular belief that only randomized studies produce results applicable to clinical practice with confidence and observational studies may always be misleading does a disservice to patient care, clinical investigation, and the education of health care professionals 7, 8, 32. In many instances, randomized studies in diagnostic research are not necessary to validly estimate the effect of the diagnostic test on patient
References (32)
Evidence-based diagnostic radiology
Lancet
(1997)- et al.
Measuring the effects of imaging: An evaluative framework
Clin Radiol
(1995) - et al.
Sensitivity and specificity should be deemphasized in diagnostic accuracy studies
Acad Radiol
(2003) - et al.
New immunoassay for the detection of Helicobacter pylori infection compared with urease test, 13C breath test and histology: Validation in the primary care setting
J Microbiol Methods
(2001) - et al.
Diagnostic and therapeutic value of selective salpingography and tubal catheterization in an unselected infertile population
Fertil Steril
(2003) - et al.
When should we remain blind and when should our eyes remain open in diagnostic research
J Clin Epidemiol
(2002) - et al.
Assessment of the accuracy of diagnostic tests: The cross-sectional study
J Clin Epidemiol
(2003) - et al.
Randomised comparisons of medical tests: Sometimes invalid, not always efficient
Lancet
(2000) - et al.
Managing the small pulmonary nodule discovered by CT
Chest
(2004) - et al.
Baseline findings of a randomized feasibility trial of lung cancer screening with spiral CT scan vs chest radiograph: The Lung Screening Study of the National Cancer Institute
Chest
(2004)