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

This Article Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Obuchowski, N. A. Search for Related Content PubMed PubMed Citation Articles by Obuchowski, N. A.

One Less Bias to Worry About [letter]

Departments of Biostatistics and Epidemiology and Radiology, The Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195-5196. e-mail: nobuchow@bio.ri.ccf.org

Editor:

I recently reported (1) on a dozen biases that are common in radiology studies and discussed strategies to avoid them. They included imperfect gold standard bias, convenience-sample bias, verification bias, reading-order bias, and context or prevalence-effect bias. The latter describes the situation where observers’ inherent diagnostic performance changes as the prevalence of disease changes in the sample of images they are interpreting. The bias was reported by Egglin and Feinstein (2), where six readers who interpreted pulmonary arteriograms for pulmonary emboli were significantly more accurate when the prevalence of pulmonary emboli in the sample was 60% versus when it was only 20%.

In the July 2003 issue of Radiology, Dr Gur and colleagues (3) performed a comprehensive evaluation of the prevalence effect in a laboratory setting. They incorporated in their study five diseases (lung nodules, pneumothorax, interstitial disease, alveolar disease, and rib fractures), 14 observers (eight board-certified radiologists, two fellows, and four residents), and 1,632 patients (including a nested test set of 179), who were separated into those with conditions that were subtle and conditions that were difficult to diagnose. They found no change in observer performance as the prevalence of disease varied from 2% to 28%. The evidence from this study is quite strong: The absence of a prevalence effect was consistent across the relatively large number and different experience levels of observers, for all five diseases, over the range of tested prevalence rates, and for conditions that were subtle and difficult to diagnose.

Thus, on the basis of this most recent study, I believe that investigators can ignore the prevalence effect in laboratory studies for the range of prevalence rates considered by Dr Gur and colleagues. This means that studies of observer performance do not need to be performed in an environment where the sample prevalence of disease matches the population prevalence for the study results to be generalizeable to the population. Rather, the sample and population prevalence rates can be different, at least to the extent considered by Dr Gur and colleagues, and still provide meaningful findings that can be transferred to the population. More research is needed to determine the effect of prevalence, if any, on clinical studies involving human observers.

REFERENCES

1. Obuchowski NA. Special topics III: bias. Radiology 2003; 229:617-621.[Abstract/Free Full Text]
2. Egglin TK, Feinstein AR. Context bias: a problem in diagnostic radiology. JAMA 1996; 276:1752-1755.[Abstract/Free Full Text]
3. Gur D, Rockette HE, Armfield DR, et al. Prevalence effect in a laboratory environment. Radiology 2003; 228:10-14.[Abstract/Free Full Text]

This article has been cited by other articles:

				M. van Straten, H.W. Venema, C.B.L.M. Majoie, N.J.M. Freling, C.A. Grimbergen, and G.J. den Heeten Image Quality of Multisection CT of the Brain: Thickly Collimated Sequential Scanning versus Thinly Collimated Spiral Scanning with Image Combining AJNR Am. J. Neuroradiol., March 1, 2007; 28(3): 421 - 427. [Abstract] [Full Text] [PDF]

This Article Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Obuchowski, N. A. Search for Related Content PubMed PubMed Citation Articles by Obuchowski, N. A.