Letter to the Editor re: Perfusion MRI for the prediction of treatment response after preoperative chemoradiotherapy in locally advanced rectal cancer
- First Online:
- Cite this article as:
- Gollub, M.J., Gultekin, D.H. & Schwartz, L.H. Eur Radiol (2013) 23: 1297. doi:10.1007/s00330-012-2716-7
- 351 Downloads
• Rectal cancer treatment can be monitored by perfusion MRI.
• Ktrans values vary between responders and non-responders.
• Radiation and chemotherapy may affect Ktrans differently.
Two different and important clinical questions are addressed in these two articles: in the Lim study, the focus is on early response in order to allow a change in ineffective therapy to avoid morbidity; in the Gollub study, the issue of identifying pathological complete responders (pCR) at the end of treatment is emphasised in order to allow for a potential organ-sparing treatment approach. Both avenues of investigation zero in on important clinical targets. In the Lim study, there were seven pathological complete responders (TRG1) and nine “near-complete responders” (TRG2). Nonetheless, the post chemoradiotherapy (CRT) Ktrans was not helpful in distinguishing these from non-responders. In contradistinction, Gollub’s main study outcome was the ability to identify pCR, thereby potentially allowing for more individualised conservative treatment. We wonder if the differences in treatment in these two studies could explain the different findings. That is, radiation was used in the Lim study, and in Figure 1b, the post-treatment Ktrans in responders approached the same values as for the non-responders, but did not go lower, whereas in our study the mean Ktrans was significantly lower. Perhaps the use of radiation in the Lim study caused formation of granulation tissue, allowing continued perfusion and permeability such that the Ktrans could not decrease to a lower value, or perhaps the use of the anti-angiogenic agent Bevacizumab in our study may have more effectively de-vascularised the tumour, resulting in distinctively lower Ktrans values. These hypotheses would of course require validation with further studies of patients receiving CRT as we are currently investigating.
As the authors allude to, tumours are heterogeneous and in order to overcome this heterogeneity (e.g. necrosis, fibrosis, mucin) the Lim methodology included an average of regions of interest (ROIs) at multiple levels. However, another strategy is to select areas within one or more ROIs that are visually the most enhancing (e.g. quartile or decile values) to see if these more reliably show any trends during or at the completion of therapy. This might be thought of as akin to using SUVmax, or SUVmin to assess response in PET investigations. In our study, this approach sometimes showed significance whereas the entire ROI did not. We wonder whether a quartile or decile analysis of the data in the Lim study would allow for different conclusions.
The manual ROI method for tracing tumour is subjective and fraught with variability, which may significantly alter results as has been shown in DWI studies of rectal cancer . We applaud Lim et al. for using two observers, and wonder if they had much variability between the observers. This point is important since no matter how robust a measurement method is, if it is not reproducible among radiologists, it will not be able to be disseminated into the community and operationalised for use in a reliable way.
In any type of biomarker evaluation, such as performed in these three studies, confirmation of positive correlations must also be confirmed with independent data sets, so we share the enthusiasm for these studies and future analyses as well.
We look forward to Lim et al.’s response and for any more light that could be shed on their data. We endeavour to validate our own data in two ongoing studies involving standard CRT treatment in less selected patients than those in the Lim study (i.e. any CRM status), while still focusing on pCR at the end of treatment. In a prospective study, we will also be studying DCE-MRI early during radiation, and that study will more closely approximate the methodology of the Lim study.
Marc J. Gollub