Friday, October 29, 2010

Evaluating magnetic resonance mammography

Breast cancer detection can be a contentious area of debate. A recent article published in the British Medical Journal is presented here. The article acts as a reply to an article written by Dr. Kell who argued that MRI was providing no benefits to patients who've already been diagnosed with breast cancer. My reply points out that cancer detection rates are not necessarily a good way to evaluate a new screening method after years of monitoring a population (ie. in longitudinal studies). It sounds counter-intuitive but the reasoning is simple: If you're monitoring a population and you add a new screening method that is more sensitive, then it picks up more tumours as it is initially applied but by catching those tumours it has created a population that has fewer tumours remaining which lowers the cancer detection rate. Thus comparing cancer detection rates after years in a longitudinal study might be a very bad way to compare the efficacy of a new screening method! Surprised?

You can check out the article here

BMJ didn't take out a copyright on the letter so we are reproducing it here:
Evaluating magnetic resonance mammography
The article titled "Magnetic resonance mammography" by Malcolm Kell [1] presents a simple yet important point: women who've had breast cancer detected (for example by x-ray mammography) who then get an MRI examination may receive an unnecessary mastectomy.

Aside from simply warning about unnecessary mastectomies, Dr. Kell refers to the literature on detecting cancer in the contralateral breast with MRI after an initial diagnosis of cancer. Dr. Kell states that "the notion that detecting occult disease with magnetic resonance mammography would benefit patients was not borne out after longer patient follow-up." The reason that Dr. Kell has concluded that no benefit is available to the patients is because the "rates of contralateral disease are the same (6%) at eight years regardless of whether or not magnetic resonance mammography was performed at diagnosis."[2] Dr. Kell has jumped to a potentially incorrect conclusion. If MRI is more sensitive than mammography, then after many years of monitoring a population we might expect MRI to produce detection rates similar to those in mammography. This sounds counter intuitive, but the reasoning is quite simple. If MRI is more sensitive to detecting malignancies in the contralateral breast, then near the beginning of MRI based contralateral breast screening, the technique will catch more cancers than would have been caught by a less sensitive screening method. After initially catching many cancers, the MRI monitored population has fewer malignancies left to be caught in future rounds of screening. Thus MRI's task of screening the patient population becomes more challenging after years of screening by virtue of creating a more challenging pool of patients to screen. This effect has been previously reported, where Dr. Nishikawa showed that long term rates of detected disease are not necessarily a good method for evaluating a newer more sensitive screening method [3]. Dr. Nishikawa's paper was written in the context of screening patients for cancer with and without computer-aided detection technologies [3], however, it appears as though his arguments apply much more generally, including in MRI based screening of the contralateral breast after a diagnosis of cancer.

To determine if detecting occult contralateral disease with MRI benefits the patient, it may be inappropriate to compare rates of contralateral disease after many years of surveillance. The fact that MRI detects occult contralateral cancer in 3-4% of patients with a recently diagnosed malignancy [4-6] indicates that MRI is doing its job of detecting small occult tumours. Additionally, to determine if detecting occult contralateral disease with MRI benefits the patient, we can look at the size, grade and nodal status of the tumours being caught by MRI and whether or not catching these tumours earlier on in their development means that the patients will receive easier treatments for their disease.

Dr. Kell states that "Magnetic resonance mammography identifies occult disease in the breast that may not be visible on other imaging modalities, and this may lead to inappropriate treatment decisions."[1] While it is possible that MRI exams can lead to inappropriate treatment decisions, an inappropriate treatment is a very strong possibility when we don't bother detecting an existing occult cancer in the contralateral breast. The author's quote implies that a patient's most appropriate treatment may involve leaving a detectable yet undetected tumour to grow inside the patient and to not specifically treat it until years later after its detection by a less sensitive method. Waiting until the tumour grows before detecting it and providing the most appropriate therapy may lead to treatments which are harder on the patient.

Jacob Levman, PhD, Imaging Research, Sunnybrook Research Institute, University of Toronto

[1] Kell MR. Magnetic resonance mammography. British Medical Journal, 2010;341:c5513.

[2] Solin LJ, Orel SG, Hwang WT, Harris EE, Schnall MD. Relationship of breast magnetic resonance imaging to outcome after breast-conservation treatment with radiation for women with early-stage invasive breast carcinoma or ductal carcinoma in situ. J Clin Oncol 2008;26:386-91.

[3] Nishikawa RM, and Pesce LL. Computer-aided detection evaluation methods are not created equal. Radiology 251: 634-636, 2009.

[4] Lehman CD, Gatsonis C, Kuhl CK. MRI evaluation of the contralateral breast in women with recently diagnosed breast cancer. N Engl J Med 2007;356:1295-303.

[5] Pediconi F, et al. Contrast-enhanced MR Mammography for Evaluation of the Contralateral Breast in Patients with Diagnosed Unilateral Breast Cancer or High-Risk Lesions. Radiology 2007;243(3):670- 680.

[6] Lee SG, et al. MR Imaging Screening of the Contralateral Breast in Patients with Newly Diagnosed Breast Cancer: Preliminary Results. Radiology 2003;226(3):773-778.

P.S. As cited in the article, Dr. Nishikawa is to be thanked for noticing this effect when evaluating x-ray mammographic screening methods with and without the use of computer-aided detection. His article is written accessibly - you don't need to be a super computer-aided diagnosis nerd to read it, you can check it out here.