The possibility of non-invasive detection of “significant” cancer in the prostate remains an elusive goal. Partly, this is the result of lack of understanding of what makes prostate cancer cells significant. In terms of cause-specific death (the end point that matters) Gleason grade is generally acknowledged as predictive of distant failure and its consequences. The question of whether magnetic resonance spectroscopic imaging (MRSI) can be used to predict the Gleason score has been recently examined at our institution and positive correlations between voxels suspicious of cancer (i.e. with large R = (Choline + Creatine)/Citrate ratios) were reported. In this follow-up analysis we wish to quantify this correlation by calculating, as a function of R, the probability that a particular voxel has a pathologic Gleason score ≥4+4. We also include as cofactors in our model the sextant biopsy Gleason score, Bx (dichotomized to < or ≥4+4), and lesion volume, V.
Materials/Methods
The data consist of MRSI ratios R stratified by patient, lesion (a collection of contiguous voxels), voxels, biopsy and pathologic Gleason, and lesion volume (number of contiguous voxels). The data were analyzed using parametric Laplacianism, according to which the posterior probability distribution of interest is defined in terms of model parameters together with their prior distributions. Specifically, we use the logistic model logit(π) = α+βR+γBx+δV to describe the relationship between π (the probability of pathologic Gleason score ≥4+4) and (R,Bx,V). We evaluate Pr{α,β,γ,δ|data) and then the predictive density of y (a Bernoulli random variable distributed with intensity π) for different (R,Bx,V) combinations.
Results
Response surfaces were obtained which quantify the predictive power of MRSI ratio (R), sextant biopsy Gleason score (Bx), and lesion volume (V). The results reveal that MRSI R and V are highly predictive of pathologic score, while the maximum effect of Bx was <15%. An interesting result was the predictive power of V; for instance, for V = 1 (i.e., in the case of an isolated voxel with positive MRSI ratio) π < 20%, regardless of the values of R and Bx.
Conclusions
MRSI information may used, on a voxel-by-voxel basis, to initiate voxel-specific dose escalation. This is an example of biologically-motivated focal treatment for which IMRT and especially brachytherapy are ideally suited.
However, one would like to know: a) whether cells that display high-Gleason patterns are radioresistant or fast proliferating, and b) whether the elimination of such cells will decrease the probability of distant failure.
FULL TEXT