HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

Studies of Diagnostic Accuracy (STARD)

STARD Checklist (link to PDF)

STARD Checklist (link to MSWord document)

STARD Flowchart (link to PDF)

Explanatory document, with examples: http://www.clinchem.org/cgi/content/full/49/1/7

STARD guidelines: http://www.clinchem.org/cgi/content/full/49/1/1

For studies of diagnostic accuracy of tests, complete and return the STARD Checklist for Evaluations of Diagnostic Accuracy (1).

The STARD statement (1) and explanatory document (2) provide guidance helping authors to modify their manuscript as needed to provide the requested information.

Provide literature reference(s) describing the evaluated test(s) and criterion ‘gold’ standard test(s) or include detailed descriptions of them.

Follow accepted methodologic standards including the following:

1. Specify spectrum of evaluated patients (age and sex distributions, eligibility criteria, and summary of symptoms or disease stage).
2. Analyze pertinent subgroups of subjects (e.g., symptomatic and asymptomatic patients).
3. Avoid verification bias (usually by application of a ‘gold-standard’ test to all subjects rather than to a clinically selected subset).
4. Categorize test results and patients independently to avoid reviewer bias (usually by performance of tests with blinding to patient information and vice versa).
5. Provide confidence intervals (or SE) for indices of diagnostic accuracy such as sensitivity/specificity, likelihood ratios, and areas under receiver-operating characteristic (ROC) curves (3).
6. Indicate the number of indeterminate test results and their use (if any) in further data analysis.
7. Provide laboratory data on analytical imprecision of the test (usually day-to-day CV at two or more concentrations) or reproducibility of observer interpretation [e.g., for a visually read, dichotomous (e.g., positive/negative) test].

A flow diagram is strongly recommended (1,2).

Evaluation of diagnostic accuracy. In clinical studies, simple testing of the significance of differences between mean values of patient groups (e.g., by Student’s t-test) provides insufficient information to assess diagnostic accuracy.

Scatter plots of data, calculations of diagnostic sensitivities and specificities and their confidence intervals (3), and use of approaches such as ROC curves (4), cumulative distribution analyses (5), likelihood ratios (6), and discriminant analysis (7) provide information that is appropriate to specific situations.

Confidence intervals should be provided (1).

Discussions of predictive values in illustrative settings may be useful additions to assess the potential clinical utility of tests.

Analysis of serial measurements requires special attention (8).

The STARD checklist form is available in Microsoft Word format and can be opened by most word processing software.

Complete the checklist by filling in the proper fields within the attached table, saving the file with your changes, then returning it to us by uploading it as a Supplemental File with your submission.

Should you encounter any problems, feel free to send the file via e-mail to clinchemed{at}clinchem.aacc.org. Please include the manuscript number on all correspondence.

References

1. STARD Group. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD statement. Clin Chem 2003;49:1-6. [Abstract/Full Text]
2. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM, Moher D, Rennie D, de Vet HCW, Lijmer J. The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration. Clin Chem 2003;49:7-18. [Abstract/Full Text]
3. Harper R, Reeves B. Reporting of precision of estimates for diagnostic accuracy: a review. BMJ 1999;318:1322-3. [Full Text]
4. Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine [Review]. Note that in Figs. 4-12 in this paper, the labels for the x-axis at the top and bottom are reversed. The (correct) dual labeling of the x-axis solves the problem of whether to plot specificity or 1 - specificity on the x-axis. Clin Chem 1993;39:561-77. [Abstract/Full Text]
5. Krouwer JS. Cumulative distribution analysis graphs-an alternative to ROC curves [Technical Brief]. Clin Chem 1987;33:2305-6.
6. Albert A. On the use and computation of likelihood ratios in Clinical Chemistry. Clin Chem 1982;28:1113-9. . [Abstract/Full Text]
7. Solberg HE. Discriminant analysis. Crit Rev Clin Lab Sci 1978;9:209-42.
8. Matthews JNS, Altman DG, Campbell MJ, Royston P. Analysis of serial measurements in medical research. Br Med J 1990;300:230-5.