Whole-Body [¹⁸F]-Fluoride PET SUV Imaging to Monitor Response to Dasatinib Therapy in Castration-Resistant Prostate Cancer Bone Metastases: Secondary Results from ACRIN 6687

Round 1

Reviewer 1 Report

The authors present a well-structured introduction that includes the necessary background to understand the rationale for the study.

The study design is retrospective and presented as such. The described methods are appropriate to meet the goals of the study. Thresholds used to document lesion activity are supported by literature. Statistics are well described, and appropriate justifications are provided for the choice of statistical tests and for the lack of correction for multiple comparisons.

The discussion/conclusion section adequately describes the limitations of the study and positions the results adequately in regard of its goals.

Overall, this is a well written manuscript with great attention to statistical methodology.

Author Response

The authors would like to thank the reviewer for such nice complements on our work with the ACRIN clinical trial 6687 data.

Reviewer 2 Report

This manuscript presents a secondary analysis of data collected through the ACRIN 6678 trial whereby PET [F-18]NaF studies were used to monitor response to Dasatinib therapy in prostate Ca patients.

Strengths of the study: 1) A prospective trial, carefully designed through ACRIN. 2) This new analysis of the data looks at static, rather than dynamic imaging data analysis (as was preliminarily designed), which is much more practical in a clinical environment. As such, any significant positive results would be translatable clinically. 3) A novel whole body quantitative tool was investigated to analyze results. 4) Quantitative correlations with PFS or OS with quantitative imaging metrics has the potential to be clinically impactful.

Weaknesses of the Study:  1) The n was actually quite low, really only about 14 evaluable subjects, which makes conclusion-drawing somewhat problematic. 2) As this was a secondary analysis of the prospectively collected data, the analysis approaches did not appear to be hypothesis driven, and more a retrospective shot-in-the-dark attempt to find significant correlations between quantitative metrics and outcome. 3) As the static WB acquisitions were secondary to the dynamic data that was collected for the study's primary endpoints, the uniformity of uptake time was somewhat poor (Figure S2. Table S1) where 4/14 subjects had >12 minute differences in uptake time - this could be a fairly substantial variable, as the authors themselves acknowledge.

Specific Comments:

Repeatability of [F-18]NaF measures is critical for this analysis. Reference 13 is cited, and apparently results were used in the analysis, however the salient results are not mentioned in the manuscript, and probably should be (e.g. 14.1% for SUVmax).

I had to read the authors previous paper to understand the "flare" effect mentioned in the manuscript that was observed in the original analysis and previously published. In the original paper the flare was observed only in a couple of patients and was associated with better PFS.  However, the primary finding was that MOST TUMORS SHOWED A DECREASE IN SUV_MAX AND SUV_PEAK IN THE ORIGINAL ANALYSIS. This was not mentioned directly. This general lowering of SUV is concordant with the current study, but it is difficult to find this as a direct statement, anywhere. It seems important that the original analysis suggests a positive correlation whereby an INCREASE in [F-18]NaF uptake was associated with better PFS - this was potentially attributed to a "flare" when up to 5 lesions were chosen for analysis. However this is not followed up in the manuscript relative to the current analysis of the WB static data, no mention of this flare phenomena was in the submission. Either remove the reference to the flare from the first study, or include a similar analysis of Flare with the current analysis. This needs to be resolved somehow.

In section 2.3, it states lesions smaller than 1.0 cm³ as measured by PET were excluded. Reconstructed resolution of PET enters into that determination, and methodology should be have more detail.

qSUVmax, qSUVpeak, and qSUVtotal, and qVF particulars are generally not defined anywhere. I deduces that qSUVmax is the hottest SUV in the entire image set (not the average of 5 lesions, nor the average of all lesions), and this is also the case for SUVpeak. qSUVtotal - it is unclear whether the software is adding all SUVs, whether this is and SUVxVolume? and qVF, it is unclear what the denominator is.

I would suggest moving Table 1 up a couple of paragraphs. There are abbreviations that are used earlier in the document that are unclear until you get to the Table 1 (e.g. BAP, AR, uNTX...).

 

Table 2 results are a little confusing in that the Change on-disatinib column is not equal to the (baseline - On-desatinib). This suggests that you are calculating the "change" as the average of the 14 differences, rather than the difference between the average baseline and the average on desatinib. This should be clear from the legend or body of the manuscript, which it is not.

There is a lot of data presented in Tables, but the punch-lines seem somewhat unclearly stated.  If I understand correctly, the big-picture take-home messages seem to be:

With regard to "lesion-level" analysis, which is a bit of a misnomer, because not 1, but typically 5 of the hottest lesions are combined:

1) SUVmaxavg and SUVpeakavg appear to demonstrate a statistically significant DECREASE after dasatinib treatment. However no direct correlation with PFS or OS were observed. This result, and the plots (Figure 2) functionally show the same results as the original study and analysis. Nothing particularly new, except it was from WB data and uses static data. This is reportable, but not surprising.

2) Lots of patient-level analyses using QTBI analysis. But in this patient population with widespread metastatic disease, no meaningful results REGARDING CHANGE IN UPTAKE were found. Nothing close to statistical significance. This is perhaps reportable, but it needs to be explicitly stated. That is patient-level analysis in this patient population yielded no meaningful correlations, results, or relationships. This should have at least a sentence or two in the discussion. Intuition suggests that disease burden is just so high, that any response is buried either in the methodological analysis approach, or an ineffective treatment.

2) There appear to be Patient level correlations with OS and PFS with regard to total volume of disease, as measured at baseline.  That seems to almost be an obvious finding (more disease volume, worse prognosis).

3). The multi-variate analysis (Table 4) seems to demonstrate a correlation between SUV_maxavg1, SUV_peakavg1, and qSUV_peak1 WITH ln(BAP), age,  and PFS. But the nature of this functional relationship is not described. One wonders whether ln(BAP) and PFS are the drivers and SUV metrics are maybe just along for the ride.

The conclusion, as written, is not particularly helpful, as it doesn't summarize the results clearly.  The first sentence seems to say that quantitative results roughly mirror what was previously published by the group. If this is the case, they should simply state that.  The second sentence seems to state that the change analysis using the static data did not show any correlation with PFS or OS. The third sentence is clear - more tumor burden, worse prognosis. But this is not earth shattering news.  The last statement about ongoing studies elucidating the role of NaF... seems overly optimistic relative to the results presented here. So not sure it warrants stating this.

 

The supplemental data is actually quite good, and elucidating. I was better able to understand the manuscript through referencing this material.

Figure S5 is the one supplemental figure that requires some explanation.  How was "predicted PFS" calculated in the multivariate analysis?

Author Response

Please see the attachment named "ACRIN6687-Revision 2-Reviewer2 comments_MM.V4"

Author Response File: Author Response.docx

Reviewer 3 Report

In this interesting project the SUV, and the change in SUV, in bone metastases in fluoride PET scans is studied. The study originates from a multicenter study (4 centers if I remember correctly) investigating the influence of Dasatinib in mCRPC-patients with bone metastases. Even though it is a multicenter study only 17 patients participate, and even 3 patients leave the study due to clinical progression. Thus, it is debatable if the multicenter setup is an advantage.

 

General comments

The follow-up (FU) NaF is performed 12 +/- 4 weeks. Hence, the range between the FU-scans is 8 weeks. Why was the FU not performed exactly after 12 weeks? What are the potential implications of the large range? I consider the range as a study limit.

The authors do not discuss the possibility of flare in the follow-up scan. What implications could flare have on the results?

The five VOIs with highest uptake value are assessed on baseline scan. Please amplify if it is the same five lesions (same anatomical position) which are used on the follow-up scan, or is it five different VOIs (diffent anatomical position)? An image with baseline scan and FU scan and including VOIs could be illustrative instead of the images maybe designated for the original study.

Several abbreviations are not explained (e.g. what is delta qVF? What is delta IndexSUVpeak?). A list of abbreviations would help a lot.

Please specify how the VOI (I think it is a cube?) is placed when calculating SUVpeak. Is it done manually or automatically by the software? If manually paced the highest peak value is not measured (if it is the definition of peak from PERCIST) and furthermore it would induce an inter- and intra-observer variation.

Please write in short what are the inclusion criteria. Is it a heterogenous cohort?

Please notify if the same scanner was used for baseline and FU-scan.

 

Specific comments

Fit. S1: Where is the biopsy taken from? Is it from the prostate or a bone metastasis? Please also explain the abbreviation “AR”

Line 196: Please write which software package was used for statistical analyses.

Section 3.1 Patients: Most of the text seems redundant, e.g. “In our initial ACRIN 204 6687 publication on dynamic imaging results, two studies were omitted due to technical 205 issues with the dynamic scan, but their WB scan was useable for this analysis and there-206 fore were included.” What is the relevance of that information?

The main point of the section: There was 17 evaluable patients, but three of them did not undergo an on-treatment PET-imaging study due to clinical progression while on dasatinib; this resulted in early discontinuation from the trial prior to the second imaging time point.

Line 283-285: “Age and ln(BAP) were found to be strong predictors of disease progression in univariate analysis (data not shown).” Why are the data not shown?

Fig. S2: The figure is illustrative. I notice large variation of delay time in case 15 and 17 which is a limitation of the study.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Appreciate the changes. Looks OK now.  Only comment is that on line 115-116 you mention that reconstruction and image resolution values appear in supplemental materials table S1. I didn't see them in S1, or anywhere else.