Letters to the Editor
Sensitivity and specificity - is your test reliable?
- Peter E. Hickman, Julia M. Potter
- Aust Prescr 2002;25:131-2
- 1 December 2002
- DOI: 10.18773/austprescr.2002.130
The Editorial Executive Committee welcomes letters, which should be less than 250 words. Before a decision to publish is made, letters which refer to a published article may be sent to the author for a response. Any letter may be sent to an expert for comment. When letters are published, they are usually accompanied in the same issue by their responses or comments. The Committee screens out discourteous, inaccurate or libellous statements. The letters are sub-edited before publication. Authors are required to declare any conflicts of interest. The Committee's decision on publication is final.
Editor, – The recent article on sensitivity and specificity (Aust Prescr 2002;25:107) is of concern in that it implies that sensitivity and specificity are invariant when applied to a particular disease state. This is not so. We give examples below.
Following occlusion of a coronary artery during myocardial infarction, cardiac troponin will be released. However, troponin is a protein and will not get into the circulation until some hours after the coronary occlusion has occurred. Thus samples collected early, say at two hours post-event, will have a poor diagnostic sensitivity for identifying myocardial infarction, while samples collected later, say at 12 hours post-infarction, will have a very high diagnostic sensitivity. These two clinical settings with very different sensitivities are not covered by the usual statement that 'cardiac troponin has a sensitivity for myocardial infarction approaching 100%'.
Consider the use of ferritin measurement to establish or exclude a diagnosis of iron-deficiency anaemia. A low ferritin concentration is considered to support the diagnosis of iron-deficiency anaemia. If samples are collected only in the general practice setting there will be very few 'false normal' results. If however, samples are collected in the acute hospital setting, where there is a relatively higher prevalence of liver disease with release of tissue ferritin, then there will be proportionately more people falsely identified as having 'normal' iron homeostasis. The apparent diagnostic sensitivity in these two populations, if compared to the best test available - bone marrow biopsy and quantitation of stored iron - would be quite different, because of the characteristics of the two populations.
Both of the examples above demonstrate that diagnostic sensitivity can vary for a particular disease state, and are one of the reasons why tests appear to perform differently in the reports in the literature. It is important to define very precisely the population that is being studied, when diagnostic sensitivity and specificity is being discussed.
Peter E. Hickman
Director of Chemical Pathology
Princess Alexandra Hospital
Julia M. Potter
Director of Chemical Pathology
Royal Brisbane Hospital
Director of Chemical Pathology, Princess Alexandra Hospital Woolloongabba, Qld
Director of Chemical Pathology, Royal Brisbane Hospital Herston, Qld