Transcript
Good morning. As many of you already noticed, I'm not a Dr. Boskos. I'm Alessandra Gorgulho, and although he had planned to attend to this meeting, it was unfortunate he could not come. So I'll do my best to give his talk in his behalf. What I am going to present to you is our volumetric analysis in 117 brain metastases from melanoma, treated with radiosurgery.
So, the purpose of our study was to analyze the volumetric changes, aiming to correlate those with the clinical outcome of these patients. Melanoma is well established as the third more common primary tumor associated with brain metastases following lung and breast cancer. The cerebral mets occur in 10% up to 40% of the patients with stage IV melanoma, which is, unfortunately, the most common stage diagnosed in patients with melanoma. Survival is very grim, ranging from three to four months, and majority of the patients die from complications of CNS disease spread. So we performed radiosurgery, in some cases we used iPlan, in the other cases we used a Brainscan for targeting and the planning was also always performed with Brainscan. We use the axial T1 with gadolinium images for contouring the lesions. And the slice thickness of these MRIs were 1.5 millimeters.
So this is the inclusion criteria that carries some problems, but it was what we could come with to perform this study. So, the inclusion criteria was patients who had at least five months after SRS for their melanoma brain metastases. What is already a biased population, since I just showed to you in the slide before, that the mean survival ranges from three to four months. So our inclusion criteria already got the best population. The other criteria was that we only include patients who had image follow up at UCLA. For logistic reasons, it would be very difficult to try to get follow-up MRI that is important to the Brainscan from outside facilities without being Dicom compatible, so that induced selection bias in our sample. That's the methodology. So we used...the MRI's were important in Brainscan. They made...all the metastases were manually delineated, what achieved a very good precision in the volume.
So taking in account our inclusion criteria, there were 39 patients with 117 mets that we're analyzing. This study is light male predominance as expected from the literature. Median age 53.1 years. The median initial volume of these tumors was 0.47 cc, mean was 2.11 cc, the median interval between diagnosis and radiosurgery was 0.9 months, the median treatment, number of lesions treated per radiosurgery session was 2.5 lesions, and the median follow up was to 9.2 months.
So going towards results and analysis. Response defined only as any reduction in comparison to the initial volume was observed in 95.3%. That doesn't mean success rate. And lack of response was observed in 5% of the case. So that means it's straight failure
For the local control, this is what defines success. It was observed in 82.6% of the case. And I'm going just to describe to you the Kaplan-Meier curve that's shown in the next slide. So according to our Kaplan-Meier, one should expect 75% of local tumor control at 17 months after stereotactic radiosurgery. So the local control rate at 6 months was quite high, 21.6, and I'm going to show to you that actually the mean average time for failure, that's not immediate failure is 8 months. So these lesions are included there, that's why we have such a high rate at six months. At 1 year 82%, and 2 years 77%. And here's the Kaplan-Meier and it's actually in days, that's why I made that quick translation for you before.
Regarding volumetric analysis, as I mentioned to you before, any kind of reduction in comparison to the initial volume, you need 95% of the case, but actually, initial response followed by relapsing is failure as well. So we ended up with 82% of local control. Sixty-two percent of the lesions were either stable, or they only decreased in size. In a minority of the lesions, 19% presented and narrowly increasing size followed by reduction.
And this is just an overall regression of the graphic showing the patterns of response identified in this sample. And that's as I mentioned to you, eight months, that's when we started to notice more, that we were really losing this patient. Regarding the results, the average delay to observe tumor reduction was 3.25 months, better for those who presented the straight responses in comparison to those who presented failure after all. The average delay for response was eight-month, and this three months for the nonresponse is just reflecting our protocol. We follow these patients every three months so that's why this number is here.
So then, what he did, he used Cox proportional model using backward stepwise regression for model selection and using local outcome as the outcome, local control as the outcome. Supratentorial location of the lesion, median is speed for reduction after radio search, and initial volume were all predictors of local control. When focusing the analysis into the local control group, and this is just the raw data presented to you, and this is the regression line based on the raw data, all these volumetric points are measured. And this is for the group showing late response, also success, that's the regression line.
So what it was observed is that two new concepts were introduced by looking at this data. The first one is percentage of shrinkage and the speed of shrinkage. For the group that had a straight response, the lesions shrunk faster in comparison to, and more, in comparison to those that had late response. The median tumor volume for this group was 0.41 cc, and that was used as a cut-off to look for a complete response. So only those that were successfully treated were looked to check whether we could find something to predict a complete response, and volume was the significant predictor as expected. So there was a correlation between initial tumor volume and complete response using the volume as a dichotomous variable.
Here is initial response followed by relapse, so these are the failures. And the volume of these tumors was larger in comparison to the mean, due to the median tumor size in the group that has local control, and the median percentage of regrowth was 205%. So then, I present to you how we calculated these parameters and then we analyze them regarding the speed of tumor volume changes in the group. Dr. Boskos performed this analysis and it's basically showing to you that the volumetric reduction follows pretty much a linear pattern, while the speed to achieve reduction follows a sinusoidal curve. And he brought some examples here to show that that was consistent, and I'm showing the group with a straight response. So basically that's the summary of what I just told you. Volumetric reduction follows a linear shape, while the speed follows a sinusoidal shape. And for the group with a delayed response, so that's also success, you can notice the same sinusoidal regression line.
Now, when he looked at the group that failed, this sinusoidal aspect curve wasn't noticed. So in conclusion, volumetric follow-up is a useful tool to precisely determine the tumor volume changes. It allows the definition of new, potentially useful parameters for the future to evaluate. When you are seeing a lesion increasing inside, it's always hard. Everybody has been under this situation, three months, six months, it's increasing in size so you consider failure or not. So if we can validate this in future studies, free of selection bias, then it might be an interesting parameter to be used. Small size melanomas have a higher possibility of complete response, what is pretty much intuitive, but it was proven here. The median interval for relapse and response may be used as parameters for a more accurate follow-up if, again, this difference is confirmed in future studies, including 100% of your sample. And the speed of metastasis reduction is not linear, but it increases and decreases until it reaches a particular value.
And also there might be a technical problem on this because we know that not every follow-up MRI has exactly the same slice. So you may account for these little variations in the speed that were noticed due to just technical limitations, the images are not exactly the same. And the small size of the relapse group, maybe, it was a disadvantage for analyzing our group. And that's his final slide. Maybe since he couldn't come here he's inviting us to go to Paris. Thank you.
So, the purpose of our study was to analyze the volumetric changes, aiming to correlate those with the clinical outcome of these patients. Melanoma is well established as the third more common primary tumor associated with brain metastases following lung and breast cancer. The cerebral mets occur in 10% up to 40% of the patients with stage IV melanoma, which is, unfortunately, the most common stage diagnosed in patients with melanoma. Survival is very grim, ranging from three to four months, and majority of the patients die from complications of CNS disease spread. So we performed radiosurgery, in some cases we used iPlan, in the other cases we used a Brainscan for targeting and the planning was also always performed with Brainscan. We use the axial T1 with gadolinium images for contouring the lesions. And the slice thickness of these MRIs were 1.5 millimeters.
So this is the inclusion criteria that carries some problems, but it was what we could come with to perform this study. So, the inclusion criteria was patients who had at least five months after SRS for their melanoma brain metastases. What is already a biased population, since I just showed to you in the slide before, that the mean survival ranges from three to four months. So our inclusion criteria already got the best population. The other criteria was that we only include patients who had image follow up at UCLA. For logistic reasons, it would be very difficult to try to get follow-up MRI that is important to the Brainscan from outside facilities without being Dicom compatible, so that induced selection bias in our sample. That's the methodology. So we used...the MRI's were important in Brainscan. They made...all the metastases were manually delineated, what achieved a very good precision in the volume.
So taking in account our inclusion criteria, there were 39 patients with 117 mets that we're analyzing. This study is light male predominance as expected from the literature. Median age 53.1 years. The median initial volume of these tumors was 0.47 cc, mean was 2.11 cc, the median interval between diagnosis and radiosurgery was 0.9 months, the median treatment, number of lesions treated per radiosurgery session was 2.5 lesions, and the median follow up was to 9.2 months.
So going towards results and analysis. Response defined only as any reduction in comparison to the initial volume was observed in 95.3%. That doesn't mean success rate. And lack of response was observed in 5% of the case. So that means it's straight failure
For the local control, this is what defines success. It was observed in 82.6% of the case. And I'm going just to describe to you the Kaplan-Meier curve that's shown in the next slide. So according to our Kaplan-Meier, one should expect 75% of local tumor control at 17 months after stereotactic radiosurgery. So the local control rate at 6 months was quite high, 21.6, and I'm going to show to you that actually the mean average time for failure, that's not immediate failure is 8 months. So these lesions are included there, that's why we have such a high rate at six months. At 1 year 82%, and 2 years 77%. And here's the Kaplan-Meier and it's actually in days, that's why I made that quick translation for you before.
Regarding volumetric analysis, as I mentioned to you before, any kind of reduction in comparison to the initial volume, you need 95% of the case, but actually, initial response followed by relapsing is failure as well. So we ended up with 82% of local control. Sixty-two percent of the lesions were either stable, or they only decreased in size. In a minority of the lesions, 19% presented and narrowly increasing size followed by reduction.
And this is just an overall regression of the graphic showing the patterns of response identified in this sample. And that's as I mentioned to you, eight months, that's when we started to notice more, that we were really losing this patient. Regarding the results, the average delay to observe tumor reduction was 3.25 months, better for those who presented the straight responses in comparison to those who presented failure after all. The average delay for response was eight-month, and this three months for the nonresponse is just reflecting our protocol. We follow these patients every three months so that's why this number is here.
So then, what he did, he used Cox proportional model using backward stepwise regression for model selection and using local outcome as the outcome, local control as the outcome. Supratentorial location of the lesion, median is speed for reduction after radio search, and initial volume were all predictors of local control. When focusing the analysis into the local control group, and this is just the raw data presented to you, and this is the regression line based on the raw data, all these volumetric points are measured. And this is for the group showing late response, also success, that's the regression line.
So what it was observed is that two new concepts were introduced by looking at this data. The first one is percentage of shrinkage and the speed of shrinkage. For the group that had a straight response, the lesions shrunk faster in comparison to, and more, in comparison to those that had late response. The median tumor volume for this group was 0.41 cc, and that was used as a cut-off to look for a complete response. So only those that were successfully treated were looked to check whether we could find something to predict a complete response, and volume was the significant predictor as expected. So there was a correlation between initial tumor volume and complete response using the volume as a dichotomous variable.
Here is initial response followed by relapse, so these are the failures. And the volume of these tumors was larger in comparison to the mean, due to the median tumor size in the group that has local control, and the median percentage of regrowth was 205%. So then, I present to you how we calculated these parameters and then we analyze them regarding the speed of tumor volume changes in the group. Dr. Boskos performed this analysis and it's basically showing to you that the volumetric reduction follows pretty much a linear pattern, while the speed to achieve reduction follows a sinusoidal curve. And he brought some examples here to show that that was consistent, and I'm showing the group with a straight response. So basically that's the summary of what I just told you. Volumetric reduction follows a linear shape, while the speed follows a sinusoidal shape. And for the group with a delayed response, so that's also success, you can notice the same sinusoidal regression line.
Now, when he looked at the group that failed, this sinusoidal aspect curve wasn't noticed. So in conclusion, volumetric follow-up is a useful tool to precisely determine the tumor volume changes. It allows the definition of new, potentially useful parameters for the future to evaluate. When you are seeing a lesion increasing inside, it's always hard. Everybody has been under this situation, three months, six months, it's increasing in size so you consider failure or not. So if we can validate this in future studies, free of selection bias, then it might be an interesting parameter to be used. Small size melanomas have a higher possibility of complete response, what is pretty much intuitive, but it was proven here. The median interval for relapse and response may be used as parameters for a more accurate follow-up if, again, this difference is confirmed in future studies, including 100% of your sample. And the speed of metastasis reduction is not linear, but it increases and decreases until it reaches a particular value.
And also there might be a technical problem on this because we know that not every follow-up MRI has exactly the same slice. So you may account for these little variations in the speed that were noticed due to just technical limitations, the images are not exactly the same. And the small size of the relapse group, maybe, it was a disadvantage for analyzing our group. And that's his final slide. Maybe since he couldn't come here he's inviting us to go to Paris. Thank you.