Preoperative Nomograms Incorporating Magnetic Resonance Imaging and Spectroscopy for Prediction of Insignificant
A. Shukla-Dave1, H. Hricak1, O. Akin1, C. Yu2, K. L. Zakian1, V. Reuter1, P. T. Scardino1, J. Eastham1, and M. W. Kattan2
1Memorial Sloan-Kettering Cancer Center, New York, NY, United States, 2Cleveland Clinic, Cleveland, OH, United States
Prostate cancer is the most commonly diagnosed cancer in American men (1). With the widespread use
of prostate-specific antigen (PSA) testing, there has been a dramatic shift to early stage cancers (2).
Insignificant prostate cancers (PCa) pose little risk to life or health, but they are difficult to identify
clinically. Keeping this in mind, the aims of the present study were to 1) validate the previously
published preoperative MR-based nomograms for predicting the probability of insignificant prostate
cancer; and 2) design new nomograms incorporating MR Imaging (MRI), MR Spectroscopic Imaging
(MRSI) and clinical data without detailed biopsy data for predicting the probability of insignificant
prostate cancer (PCa).
Materials and Methods
Prospective study of 181 patients with low-risk PCa (Gleason score ≤6, PSA <10 ng/ml) who underwent
combined MRI/MRSI prior to radical prostatectomy. Written informed consent was obtained from each
patient prior to MR examination. Surgical pathology was used as the standard of reference (3). Data
were acquired on a 1.5 Tesla scanner (GE, Milwaukee, WI). The study consisted of MR imaging using
a pelvic phased array and expandable endorectal coil followed by standard MRSI protocol with PRESS
voxel excitation and water and lipid suppression (3, 4). MRSI data were obtained and processed using
GE software PROSE (prostate spectroscopy). The probability of insignificant cancer by MRI and MRSI
findings were recorded based on previously published (3) scoring system: 0-3 scale (0, definitely
insignificant PCa (no abnormality) - 3, significant PCa (definite abnormality >0.5cm3). The new
baseMRI model combines PSA, clinical stage, prostate volume on MR imaging and MRI score. The new
baseMRI/MRSI model has the same variables except that the MRI score is replaced with the overall
MRI/MRSI score (Figure1). The biopsy Gleason score is omitted from the new MR nomograms because
patients with only Gleason score 3+3 were included in the study, so the grade would not contribute to the
point scale in the nomograms. We used receiver operating characteristic (ROC) curves to assess the
incremental value of the 2 new MR models mentioned above to the existing Base, Medium, MRI and
MRI/MRSI models (3,5). The latter two now termed as mediumMRI and mediumMRI/MRSI models.
At pathology, twenty seven percent of the patients had insignificant PCa defined as organ confined
cancer ≤ 0.5cm3 in volume without poorly differentiated elements. The surgical Gleason score remained
the same as the biopsy Gleason score of 6 for 43.6% of patients but was higher for 56.4% of patients. All
MR models demonstrated good calibration. When the previously published Base and Medium models
were applied to the patient population in this prospective study, the resulting AUCs were 0.558 and 0.707,
respectively; the Medium model performed significantly better than the Base model (P = 0.001). These
results were consistent with the earlier findings. All four MR models were more accurate than the Base
model for discriminating insignificant from significant PCa (p <= 0.001 for all). However, none of them
was significantly more accurate than the Medium model (P ≥ 0.065 for all) (Figures 2 and 3). MR data was
helpful in predicting significant Pca in the low-risk patient population studied. 63/90 (70%) patients with
an MRI score of 3 (definitely significant PCa) had tumor volume > 0.5 cm3. As expected, the findings of a
non-nodular region with reduced T2-weighted signal > 0.5 cm3 (indeterminate MRI category) was non-
specific. 6/9 (67%) patients with an MRI score of 0 or 1 had tumor volume < 0.5 cm3 and 5/9 (56%) had
insignificant PCa. The misclassification of definitely or probably insignificant PCa was due to
underestimation of the Gleason grade at biopsy. With the addition of MRSI to MRI, the imaging
score was changed for 13 patients from indeterminate to significant cancer; the change was
correct in 11/13 patients; the two patients misclassified had total tumor volume <0.5 cm3 but
Gleason score 7. Overall the addition of MRSI improved predictive accuracy (Figure 3).
Discussion and Conclusion
In the present study, the biopsy Gleason score was upgraded in more than half of the patients (56.4%) at
surgical pathology. Physicians often repeat biopsy or opt for saturation biopsy to obtain a more
comprehensive evaluation, but they should exercise caution in making such decisions, as Gallina, et al.
have observed an association between prostate biopsy and an increased rate of mortality (6). The
certainty of this association, however, remains to be proven. We have successfully validated the
previously published MR nomogram models for predicting the probability of insignificant PCa in
patients with clinically low-risk disease. The new BaseMRI and BaseMRI/MRSI models performed
similarly to the Medium model incorporating detailed biopsy data and in future may obviate the need
for repeat biopsy to obtain for additional such data. Additionally, MRI and MRI/MRSI performed better
in identifying significant rather than insignificant disease. Hence, the nomogram models incorporating
MR findings may show that aggressive therapy is warranted in certain men whose disease would
otherwise appear to be low risk.
References 1. Cancer facts and figures 2010. pp. 1-68. Atlanta, GA: American Cancer Society, 2010; 2.
A.V.D’Amico, et. al., J Clin Oncol., 23(22), 4975, 2005; 3. A. Shukla-Dave et al, BJU Int 99:786,
2007; 4. Y. Mazaheri et al, Radiology 246:480-8, 2008; 5. M.W. Kattan, et. al., J Urol, 170, 1792 ,
2003. 6. A.Gallina et al. Int J Cancer 123:647, 2008
Figure1. BaseMRI/MRSI nomogram model. In the model
for locating the patient’s pretreatment PSA on the PSA
axis. Draw a line straight upwards to the Points axis to
determine how many points towards having an
insignificant cancer the patient receives for his PSA.
Repeat this process for each variable. Sum the points
achieved for each predictor and locate this sum on the
Total Points axis. Draw a line straight down to find the
patient’s probability of having insignificant PCa.
Figure 3. Comparison of ROC curves for the Base , Medium,
BaseMRI/MRSI and MediumMRI/MRSI models.
Figure 2. Comparison of receiver operating characteristic (ROC)
curves for the Base, Medium, BaseMRI and MediumMRI models.
Proc. Intl. Soc. Mag. Reson. Med. 19 (2011)46