I read a study on the accuracy of prostate MRI for predicting that no prostate cancer (PCa) is present. This is called the negative predictive value (NPV), and the authors make a great point in their introduction:
The reliability of a negative MRI is a key clinical metric – if the MRI is reported as positive for cancer, the patient will undergo a biopsy irrespective of whether this is a true positive or a false positive; however, a negative MRI may mean a patient is discharged without need for a biopsy. It is therefore important for an MRI to accurately exclude tumour, and a low false negative (FN) rate is a key determinate of whether this approach is clinically feasible.[i]
In other words, if a 3T multiparametric MRI (3T mpMRI) such as we offer shows no suspicion of cancer, can the radiologist be confident enough to send the patient home with the assurance that no cancer is there?
Well, we have to be careful in what we say and how we say it, and of course we have to take other factors into account (e.g. patient age, PSA, family history, DRE results, exposure to known carcinogens, etc.) And it’s important to go over the images with the patients, to explain that there are no suspicious areas. But we also have to clarify that microscopic PCa cells may be present, and not able to yet be detected by imaging alone.
The research team from England and Portugal noted that “mpMRI has the potential to accurately exclude prostate cancer and reduce biopsy burden.” They put mpMRI to the test by reviewing the MRI reports and actual biopsy results of 148 patients. The authors compared the images (scored by experienced radiologists using the PI-RADS 5-point scale (https://sperlingprostatecenter.com/pi-rads-score/) with fusion-guided targeted plus 24-36 systematic transperineal biopsy results (needle samples taken through the skin between the scrotum and the anus, rather than through the rectal wall). According to their findings, the imaging had a 19.6% false negative rate (there was PCa where the MRI showed nothing), and it was mostly in the anterior region. According to their criteria, 14 of the missed lesions were clinically significant (almost 10%). To be honest, this surprises me, because the majority of recent studies suggest that mpMRI may not detect small, clinically insignificant tumors, but excels at detecting significant PCa.
This leads to my critique, which does not concern the design, sample size, or statistical calculations (all of which I thought were excellent) but the use of fusion guidance (see a comparison of TRUS, real-time MRI and fusion-guided biopsies at https://sperlingprostatecenter.com/mri-guided-biopsy-vs-fusion-whats-difference/) I was disappointed that the authors did not specify which fusion device or technology they used. I gathered from reading the article that it was not cognitive fusion (basically an educated guess based on previous MRI images while placing needles under “blind” real-time ultrasound guidance). Therefore, I deduced that they were using one of the commercially available devices with special software to “fuse” the previously captured MRI with real-time ultrasound. I am familiar with the commercially available fusion devices, and I wish I knew which one they used. Admittedly, my disappointment stems from my unsatisfied curiosity.
Second, there is definitely more room for biopsy error with fusion than there is with real-time MRI. Things like patient motion, different position on the table, slightly inaccuracies when the physician marks point-by-point anatomic correspondences, and even small distortions in the software itself can mean that a targeted biopsy needled misses its target! So was it the original MRI scans that missed the tumor? The person who read and scored them? Inaccuracies of the fusion-guidance? I can’t be sure from reading the paper, so I have to reserve judgment regarding what I consider a high rate of false imaging negatives when compared with the biopsy results.
The authors point out that study limitations include a retrospective (not prospective) design, sampling error inherent in any biopsy approach, and the fact that they were testing the false negative rate of MRI in absence of other clinical factors such as PSA, family history, etc.
Personally, I have a high degree of confidence that when our 3T mpMRI protocol results are negative for significant prostate cancer, it’s a true negative, not a false one. Our team brings not only a high level of experience, but also years of networking with one of the world’s pre-eminent team of prostate MRI experts at Radboud University in Nijmegen (The Netherlands). I value all research into the accuracy of mpMRI, since it enlarges our collective knowledge about the imaging that is changing the landscape of PCa detection, diagnosis, and treatment – so I am grateful to Serrao and colleagues for their work with mpMRI compared with fusion-guided transperineal prostate biopsy.
[i] Serrao EM, Barrett T, Wadhwa K, Parashar D et al. Investigating the ability of multiparametric MRI to exclude significant prostate cancer prior to transperineal biopsy. Can Urol Assoc J. 2015 Nov-Dec;9(11-12):E853-8. doi: 10.5489/cuaj.2895. Epub 2015 Dec 14.