% | $
Quotes you view appear here for quick access.

Celsion Corp. Message Board

  • daviscupper daviscupper Nov 11, 2012 8:26 PM Flag

    Model Results at 380 PFS Events.

    This analysis will take four posts. Part I.
    Now that we have arrived at the pivotal 380 PFS events and data lock down it is time to #$%$ what the outcome of the Heat Trial is likely to be. If we assume management's 12 month median PFS figure for the control group is correct, which is based on historical facts and the collective wisdom of the world’s KOLs (knowledgeable opinion leaders), the results are likely to be truly remarkable.
    Let’s begin by looking at the known data:

    11/1/2012, 700,
    10/1/2012, 700,
    9/1/2012, 700,
    8/1/2012, 700,
    7/1/2012, 700,
    6/1/2012, 700, 10, 5.5, 1.147, 0.246
    5/1/2012, 690, 10, 6.5, 1.355, 0.290
    4/1/2012, 680, 10, 7.5, 1.564, 0.335
    3/1/2012, 670, 10, 8.5, 1.772, 0.379
    2/1/2012, 660, 10, 9.5, 1.981, 0.424
    1/1/2012, 650, 10, 10.5, 2.189, 0.469
    12/1/2011, 640, 10, 11.5, 2.398, 0.513
    11/1/2011, 630, 10, 12.5, 2.577, 0.558
    10/1/2011, 620, 10, 13.5, 2.727, 0.603
    9/1/2011, 610, 10, 14.5, 2.877, 0.647
    8/5/2011, 600, 42, 16.247, 13.297, 3.046
    5/12/2011, 558, 24, 18.434, 8.321, 1.975
    3/25/2011 , 534, 18, 19.96, 6.653, 1.604
    2/11/2011, 516, 42, 22.186, 16.467, 4.160
    11/11/2010, 474, 24, 24.3695, 9.933, 2.611
    09/30/2010, 450, 45, 26.003, 18.460, 5.224
    8/3/2010, 405, 74, 28.4205, 33.509, 9.389
    5/4/2010, 331, 81, 31.638, 38.308, 11.441
    1/20/2010, 250, 54, 34.5495, 26.521, 8.329
    11/10/2009, 196, 45, 37.2945, 22.500, 7.493
    8/5/2009, 151, 131, 43.975, 65.500, 25.719
    9/30/2008, 20, 19, 50.572, 9.500, 4.290
    6/30/2008, 1, 1, 52.072, 0.500, 0.232

    The enrollment data, columns one and two are from press releases and the assumption that the last 100 patients added to the trial were distributed evenly over a 10 month period, beginning on 9/1/2011 and ending on 6/1/2012.
    Continued on part II.

    Sentiment: Strong Buy

    SortNewest  |  Oldest  |  Most Replied Expand all replies
    • Sweet....and it will get sweeter!

      Sentiment: Strong Buy

    • Model looks better and better with each and every passing day. Cheers.

      Sentiment: Strong Buy

    • Daviscupper

      Assuming null difference between control & treatment groups and a say 18 mnth PFS median when would we have reached 380 PFU's? Jan 2012? or before?

    • Never a bad thing to review the known data. If you put a little time to read and understand the model you will see how well the results are likely to be. Cheers.

      Sentiment: Strong Buy

    • I would love to hear CEO Michael Tardugno reconfirm today his belief that the control arm has a 12 month median PFS rate. With results scheduled to ge released very shortly that kind of reconfirmation could be explosive!

      Sentiment: Strong Buy

    • davis,

      Your model didn't prove any thing. Every long knows if control arm is at 12-month, the trial will be a blockbuster success. The question is will control arm at 12-mon?

      What does your model show if control arm median is 18-month with 54-month distro? Can you list the treatment group median and PFSs? Now, that will be telling.

      Have fun!

      • 2 Replies to carrix_fool
      • carrrix_fool,

        Since I got a little board I ran the model assuming an 18 month median PFS rate for the control group. That is, first 18 months 2.78% of the patients event each month. The next 18 months 2% event each month. And the final 18 months .78% event each month. That is an 18 month PFS rate with a 54 month distribution. This run generated 239 PFS events for the control group.

        I then used a 32 month median PFS rate for the treatment group with a 72 month distribution. That is the first 32 months 1.5625% of patients evented each month. The next 40 months 1.25% of patients evented each month. This generated 153 PFS events for the treatment group.

        Total PFS events = 239+153 = 392 events. A little greater than 380 events. At any rate this yields a PFS ratio = 32/18 = 1.78, which is beyond the 1.33 requirement.

        So even if we exceed the PFS rate of the control group by 50 percent we will still far surpass the endpoint requirement. Cheers.

        Sentiment: Strong Buy

      • What I have shown is that if management's insistence that the median PFS rate for control is 12 months is correct then the ThermoDox will be a wild success. Beyond what many on this board have imagined.

        If you want to obtain the results of an 18 month median PFS rate for the control arm with a right-skewed 54 month distribution then I have shown you how to do it. Get crackin'. Cheers. I look forward to your report and results.

        Hint: It will still meet endpoints.

        Sentiment: Strong Buy

    • Part II.
      As stated before the first column is the date, the second column shows the patients enrolled on that date, the third column reveals the change in patients on the date from the previously reported enrollment date, the fourth column shows how long a given sample of patients have been enrolled from the average date of enrollment to November 1st, 2012, the fifth column shows how many “control” PFS events occurred from that sample of patients and the sixth column shows how many “treatment PFS events occurred from the same sample of patients.

      Note the fourth column on 6/30/2008, one person was enrolled. From 6/30/2012 to 8/1/2012 that person was enrolled for 4 years, four month and one day, or 52.072 months. On 9/30/2008, 19 new patients were enrolled since 6/30/2008, that is 3 months later. If we assume these new patients were added evenly over those three months then they were on average in the trial 1-1/2 months less than the first person in the trial, 50.572 months (52.072-1.5=50.572).

      How do we calculate the number of PFS events from any given sample of patients, the data in the fifth and final column? It is easy. We assume a 12 month PFS median event rate, and a right-skewed distribution over 36 months. What does this distribution look like? I will simplify it and clearly show you how you can do the calculations yourself. I will assume that for the first 12 months, 4.167 percent of the patients experience a PFS event each month, the next 8 months 3 percent of the patients had an event, the next 8 months 2 percent experience an event and the final 8 months 1.25 percent experience an event (12x4.167%+8x3%+8x2%+8x1.25%=100%). This accounts for all the patients and is a 12 month median PFS rate, right-skewed distribution over 36 months.

      So let’s do a sample “control” PFS event calculation: Take the 5/4/2010 entry date. There are 331 patients, 81 new patients since the last entry and these new patients have been in the trial for an average of 31.638 months. The number of estimated “control” PFS events from this new sample of patients is:
      81(.04167x12+.03x8+.02x8+3x.0125)/2 = 38.308 events.

      Why did I divide by 2? Because only half of all patients are in the control group.

      The .04167x12, .03x8 and .02x8 account for the first 28 months of the period. The .0125x3.638 accounts for the last 3.638 months of the total 31.638 months.

      We will show you how to derive the treatment patient events, the sixth column, later.

      Now let’s look at the 8/5/2009 entry, there are 65.5 PFS events from the control group. That is half of the 131 new patients. How did we get that so quickly? Simple. These patients have been in the trial for 43.975 months. We are assuming a 36 month distribution so all of the control patients have experienced an event.

      If you add up all the control PFS events in the fifth column we get 290 PFS control group events. We have just achieved the required 380 PFS event number in the Heat Trial and it should be abundantly clear the treatment group is doing extraordinarily well. If there have been 290 control events then there are only 90 PFS events from the treatment group.
      Continued on Part III.

      Sentiment: Strong Buy

      • 3 Replies to daviscupper
      • I just discovered another typo. In Part II, paragraph four, line five the equation reads: "81(.04167x12+.03x8+.02x8+3x.0125)/2 = 38.308 events."

        It should read as: 81(.04167x12+.03x8+.02x8+.0125x3.638)/2 = 38.308 events. The last term in the parenthesis was incorrectly copied. However, there was no error in calculations. Just a typo of the last term.

        Sentiment: Strong Buy

      • Noticed a typo in the second sentence of the second paragraph of Part II. The second sentence reads: "From 6/30/2012 to 8/1/2012..." It should read: "''From 6/30/2012 to 11/1/2012..."

        No errors in data or calculations have been made. Just a simple typo. Cheers.

        Sentiment: Strong Buy

      • Part III.
        Now how do we determine the values in the sixth column, the number of treatment patients that have evented for each sample of patients. And most importantly how do we calculate the median PFS rate of the treatment group. This is where it gets interesting. From the above calculations we determined there have been 90 PFS treatment group events. We also know the dates of treatment enrollment and how long these patients have been enrolled.

        So, I ran several Monte Carlo simulations using different median PFS rates and distributions until the sum of PFS events in column six totaled 90 PFS events. I tried an 18 month median PFS rate and got an alarming number of PFS events. I tried 24 months and still achieved way too many. I bumped it up to 36 months and got 135 PFS events, still too many. I tried 48 months and still too many. It wasn’t until I used a 56 month median PFS rate for the treatment patients that I finally arrived at 90 PFS events.

        A 56 month median PFS rate assumes that for the first 56 months, .8929% (.008928, less than one-percent) of patients achieve an event each month.

        So let’s do a sample calculation for the date 8/5/2009 and find the number of treatment patients that had an event in that sample, column six. The calculation is: 131 patients x 43.975 months x .008929 patient events/month x .5 (only half of patients are in treatment group) = 25.719 events in treatment group.

        Let’s summarize what we have determined through this model:
        The hazard ratio = 290/90 = 3.22, that is the control group is more than 3 times more likely to experience a PFS event than the treatment group during the trial period just prior to data lock down.

        The PFS rate ratio = 56/12 = 4.66, that is the median PFS rate for the treatment group is 4.66 times the control group. That blows away the 1.33 requirement.
        These are truly awesome results!
        Summary continued in Part IV.

        Sentiment: Strong Buy

1.21-0.02(-1.63%)Jul 22 3:59 PMEDT