When it comes to rating the overall quality of mammography screening programs, sensitivity should take precedence over specificity, according to a commentary in the May American Journal of Roentgenology.
Dr. Lara Hardesty and colleagues cautioned against relying on positive predictive value (PPV1) as the sole indicator of whether a screening program is up to snuff. The group is from the University of Pittsburgh and Magee-Women's Hospital, both in Pittsburgh.
"A program that operates at the highest PPV1 may still miss a large number of potentially detectable cancers. If only the lesions with a high likelihood of being malignant are recalled, the high specificity could elevate PPV1 despite low sensitivity," they wrote (AJR, May 2005, Vol. 184:5, pp. 1505-1507).
In their article, Hardesty's group laid out three possible relationships between detection sensitivity and recall rates (see figure below). Linear applies to a linear progression between detection sensitivity and recall rates over a certain range of recall rates. In the convex relationship, the ratio between detected cancers and recall rates decreases with increasing recall rates.
 |
| Graph shows three possible models of the relationship between sensitivity and recall rates given an assumed technology-based upper limit of sensitivity for mammography at 90%: linear (diamond), convex (square), and sigmoidal (triangle). Hardesty LA, Klym AH, Shindel BE, Chough DM, Sumkin JH, Gur D, "Is Maximum Positive Predictive Value a Good Indicator of an Optimal Screening Mammography Practice?" (AJR 2005; 184:1505-1507). |
"If the relationship between sensitivity and recall rates is sigmoidal (S shape), this maximum PPV1 point will occur ... where the linear region between concave and convex begins to become convex," they stated.
Because this PPV1 point can happen at any level of sensitivity and recall rate, the authors cautioned that PPV1 should not be used as a primary measure for operational success. While PPV1 has been historically looked at as the "necessary balance" between sensitivity and specificity, it does not actually take into account the number of cancers that would be missed at a particular sensitivity level, they stated.
By Shalmali Pal
AuntMinnie.com staff writer
May 23, 2005
Related Reading
Breast imaging trends show longer waits for screening mammo, more biopsies, May 6, 2005
Breast biopsy costs big bucks, but so does cancer screening, January 19, 2005
Copyright © 2005 AuntMinnie.com
