Diagnostic accuracy: 60% of the time, it works every time

Both doctors and patients take the reliability of diagnostic tests for granted, but are they really as accurate as they appear to be? Hannah Lippman explains more here.

Diagnostic tests play a crucial role in the management of health and disease. These tests are vital for identifying a specific condition, disease stage, therapeutic response or the chance that a person will develop a condition in the future.

In many cases, fast and accurate diagnosis could mean the difference between life and death, and so the field of diagnostics must rapidly evolve to improve accuracy and reliability of test results.

What it’s all about

A diagnostic accuracy study is designed to investigate how well a particular diagnostic test is able to identify a target condition, in comparison to a reference test or ‘gold standard’.

Diagnostic accuracy studies are used to evaluate the ability of one or more diagnostic tests to correctly identify a target condition. These studies should not be confused with diagnostic studies, which simply generate information about the prevalence, nature or severity of a particular condition in a population.

A diagnostic accuracy study is designed to investigate how well a particular diagnostic test is able to identify a target condition, in comparison to a reference test or ‘gold standard’.

In order to do this, participants taking part in these studies are tested using both an index test (the test which is being evaluated in the study) and a clinical reference standard (the method currently considered best at identifying that condition).

By comparing the results of these two tests, the accuracy can be determined through the evaluation of the specificity (the amount of people who have not got the condition who have a negative result) and sensitivity (the amount of people who have got the condition who have a positive result).

For instance, a very specific test would provide negative results for all (or almost all) patients who are disease free, and a very sensitive test would identify all (or almost all) patients who have the disease.

In general, diagnostic accuracy studies are relatively simple to conduct. On the whole, they are observational studies, following a case-control (where a group of people who are suffering from a condition are compared to a control group who are known not to have the disease) or cross-sectional (where data is collected from a specific group of people at a single point in time) study design.

So what’s the problem?

As with all clinical research, diagnostic accuracy studies are subject to various sources of bias and variation.

Although the methodology of this type of study seems simple enough, it is not without its limitations. As with all clinical research, diagnostic accuracy studies are subject to various sources of bias and variation.

Reference standard tests are rarely 100% specific or sensitive, and it has been found that in some studies the overall estimates of sensitivity or specificity significantly differs from the true value (such as over or underestimating the accuracy of the reference standard).

The accuracy of the reference standard can be affected by a number of variables, ranging from the experience of the person performing the test to the type of patients that are included in the study.

Another major problem that is encountered is incomplete reporting. It is vital to ensure that a detailed methodological description of the index tests and reference standards are included, so that any possible bias can be identified.

In addition to the problems within the study methodology, it has been found that in many cases the studies themselves were not being reported correctly. An article published in Trials discusses the drawbacks of the primarily observational study designs used, proposing that conducting randomized controlled trials investigating diagnostic accuracy could help eliminate this bias and arguably have a greater impact in the implementation of changes in standard practice.

Improving visibility and reporting

By all accounts, the STARD checklist dramatically has improved the quality of reporting diagnostic accuracy studies.

In 2003, a group of scientists and editors compiled the STARD (Standards for the Reporting of Diagnostic Accuracy) statement, including a checklist of 25 mandatory items with the intention of improving the accuracy and completeness of reporting diagnostic accuracy studies.

By all accounts, the STARD checklist dramatically has improved the quality of reporting diagnostic accuracy studies, however a paper published in 2006 in the journal BMC Medical Research Methodology showed that that there were still teething problems, calling for further clarification of the guidelines.

Following review by international experts, in 2015 this checklist underwent a substantial update, to include 30 items, ensuring the transparent and complete reporting of diagnostic accuracy trials.

In collaboration with Cochrane (the world’s largest organization producing systematic reviews in health and social care), the Cochrane Register of Diagnostic Test Accuracy Studies has been founded. This register aims to encourage the submission of systematic reviews, discussing completed diagnostic accuracy studies. It also makes the process of searching for reviews diagnostic accuracy trials far simpler, with a comprehensive search function.

Systematic reviews of diagnostic accuracy studies allow for the results of diagnostic accuracy trials to be pooled to create a precise estimate of the accuracy of a specific diagnostic test, whilst providing a transparent overview of different studies designs and conduct, to allow for the elimination of bias.

What we can do

By making the registration of diagnostic accuracy trials mandatory, it is hoped that this will increase transparency within the field.

Substantial evidence has been found to suggest that many diagnostic accuracy studies remain unpublished, that only a small proportion of published diagnostic studies have been registered.

The registration of diagnostic accuracy trials has therefore now been added to the updated STARD checklist as a compulsory requirement. By making the registration of diagnostic accuracy trials mandatory, it is hoped that this will increase transparency within the field.

In the ISRCTN registry, the registration of diagnostic accuracy studies is steadily increasing. For example, a recently registered study out of Imperial College London describes a diagnostic accuracy study into the assessment and classification of ovarian cancer. This type of study could have far-reaching implications in the diagnosis and treatment of life-threatening illnesses, and so their registration is vital for visibility within the world of research.

A one-stop shop for registering diagnostic accuracy trials

The ISRCTN registry is dedicated to supporting transparency in the reporting of clinical trials. Diagnostic accuracy trials submitted to ISRCTN receive expert curation in line with the STARD criteria, to ensure that all relevant information is included and visible in the trial record.

The team are able to offer advice and assistance in the registration of a range of study types, including diagnostic accuracy, ensuring that guidelines are explicitly followed to ensure maximum visibility.

For further information about registering diagnostic accuracy trials, please contact ISRCTN.

Apply online to register your trials with the ISRCTN registry: http://www.isrctn.com/trial/draft

View the latest posts on the On Medicine homepage

Comments

By commenting, you’re agreeing to follow our community guidelines.

Your email address will not be published. Required fields are marked *