| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Commentary |
1 Washington Radiology Associates, PC, 2141 K St. NW, Ste. 900, Washington, DC 20037.
Received October 20, 2005; accepted after revision October 20, 2005.
This article is a commentary on the preceding article by Brenner et al.
Keywords: computer-aided detection • court cases
New technologies in medical imaging often herald great promise for clinical advancement. Yet their introduction may overturn the status quo of established and proven clinical practices. In the process, the implications of the new technology are likely to spill over into the legal arena, where accepted standards of care are ultimately determined. In radiology, computer-aided detection (CAD) for mammography, as well as for other screening techniques, follows this scenario of a new technology with the capability to significantly enhance our ability to detect potentially life-threatening diseases such as breast, lung, or colon cancer. However, because it changes our practice paradigm, this technology may result in an eventual change in the legally accepted standard of care.
Since its introduction in 1998, CAD software in mammography has established itself, for the most part, as a clinical blessing. The use of CAD in the community setting has been shown to increase breast cancer detection rates by almost 20% without unduly increasing recall rates for screening mammography [1]. Utilization of CAD in a regional screening program yielded similar overall increases in the cancer detection rate associated with a significant increase in the detection of small—1.0 cm or less—invasive cancers [2]. In addition, evidence shows that the use of CAD may increase radiologist sensitivity in mammographic interpretation and that CAD can potentially correctly identify missed cancers on false-negative mammograms [3, 4]. On the clinical front, this is all good news for the patient undergoing mammography and for the interpreting radiologist because the end result is that additional early, curable breast cancers will be detected, and more lives will be saved. But the use of CAD in mammography may have a darker side. Unresolved questions are looming regarding the legal implications of CAD, its introduction as evidence in a malpractice case, and establishment of a standard of care for its general use in mammographic interpretation.
Brenner and his colleagues [5] have provided a description and analysis of a medical malpractice case concerning delayed diagnosis of breast cancer and the subsequent appellate court decision wherein CAD data were introduced as evidence by the defense expert witness [6]. It is unlikely that the holding in this case will be used by plaintiffs against radiologists because the legal implications of this case are relatively restrictive. The appellate court affirmed the trial court's ruling that the CAD result was not hearsay because the CAD result was not an assertion made by a "person," as that word is used in evidence. The appellate court also affirmed that the standard-of-care objection made by the plaintiff was an irrelevant objection. Lastly, and most important, the appellate court made no determination about the scientific reliability of witness testimony concerning CAD results because the plaintiff had failed to raise this as an objection to the expert witness testimony at the time of the trial. Ohio evidence rule 702(c) permits "reliable scientific, technical, or other specialized information" to be admitted by expert testimony, over objection, if it meets certain criteria for reliability [7]. However, because the reliability of CAD was not challenged, the very heart of the issue of whether CAD data are considered scientifically sound in the courtroom setting has been left unresolved.
If CAD analysis could meet the scientific reliability criteria, it would be more likely to support the defense of a delayed diagnosis rather than further a plaintiff's claim of a missed cancer diagnosis because the reliability of a positive CAD mark is low as a result of the large number of false-positives. In contradistinction, the reliability of a negative CAD result is relatively high, as noted by Ulissey in his testimony at the trial. He stated that if no CAD marker was present, "there is greater than a 99 percent chance that no malignant calcium is on that film" [6]. In the case under discussion, the CAD results were used to bolster the defendant radiologist's position that there were no suspicious calcifications that warranted recall. However, if future court decisions favor a more global admissibility of CAD evidence, this may support a plaintiff's contention that nonrecall by the radiologist to further assess an area marked by CAD supports a breach in the standard of care.
Should this occur, it may well have far-reaching effects on how we practice mammography in the United States. One can imagine a future scenario in which a defendant radiologist did not recall a patient for what he or she considered a "subthreshold" finding and is later sued for delayed diagnosis of breast cancer. If the plaintiff can supply a CAD readout that marks the cancer that was purportedly "missed," this evidence, in literal black and white, may be difficult to refute. A positive CAD finding becomes a "smoking gun" against the defendant radiologist. A plethora of legal decisions favoring plaintiffs may produce an atmosphere of paranoia among mammographers whereby we feel compelled to recall most patients with positive CAD marks for fear of the legal repercussions. Mammographic interpretation then becomes CAD-driven. Once we descend this slippery slope, the result will be significantly increased recall rates and decreased positive predictive values. Capacity limits of mammography facilities will eventually be exceeded, patient anxiety concerning mammography will needlessly increase, and many more unnecessary biopsies will be performed. The legitimacy of mammographic screening as a whole will thereby be undermined and imperiled.
What can we do to mitigate potentially deleterious effects of CAD in the courtroom? First, we should promote restrictive and circumscribed evidentiary application of CAD data in court proceedings. Those radiologists who serve as expert witnesses should keep this in mind when giving their testimony, regardless of whether it is for the plaintiff or the defendant. The acceptance of CAD results as the final word in resolving contradictory claims, or the promotion by an "expert" of overblown claims as to the validity of current CAD systems, should be resisted because it will ultimately undermine the far greater significance and validity of the mammographic interpretive process that goes on in the gray matter of the radiologist. We should be careful not to imbue CAD with greater power in mammographic interpretation than it deserves. The most important elements of image interpretation remain within the complex and inimitable realm of human perception and cognition. This is supported by studies showing that CAD results probably have little influence on the interpretations of experienced or high-volume mammography interpreters, who generally do not show significant improvements in detection rates of cancer after the introduction of CAD in clinical practice [8].
Second, we should be realistic about the benefits and limitations of CAD in the clinical interpretation of mammograms. We know that the negative predictive value of CAD, as applied in mammography, is far greater than the positive predictive value because of the great number of false-positive CAD marks that are generated for every true-positive mark. We should emphasize to our health care colleagues and to our patients that CAD is a useful and important interpretive tool but does not constitute a double interpretation of the mammogram. We should not ourselves be lulled into complacency by a negative CAD result, nor should we be thrown into a frenzy over a readout that is peppered with CAD marks. We should be realistic and circumspect in our assessments of CAD marks for calcifications as opposed to masses and distortions because we know that the performance of CAD is better for the former than for the latter.
Finally, continued development of CAD software by both vendors and research scientists should be actively supported by the radiology community. The main objective should be to significantly reduce false-positive marks so as to increase the overall sensitivity of mammographic interpretation that is augmented by CAD. In addition, we should promote further investigations, particularly prospective studies, concerning the impact and efficacy of CAD in everyday clinical practice, especially in community-based radiology, where most women obtain their mammograms. Our active involvement in the ongoing refinement of CAD technology will help to ensure the survival of imaging-based screening techniques.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  J. F. Dick III, T. H. Gallagher, R. J. Brenner, J. P. Yi, L. M. Reisch, L. Abraham, D. L. Miglioretti, P. A. Carney, G. R. Cutter, and J. G. Elmore Predictors of Radiologists' Perceived Risk of Malpractice Lawsuits in Breast Imaging Am. J. Roentgenol., February 1, 2009; 192(2): 327 - 333. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
