Of late, there have been statistical models surfacing that boldly assert Celsion's (NASDAQ:CLSN) phase III HEAT clinical success. My own thesis favors clinical success largely because of the history of ineffectiveness for radiofrequency ablation [RFA] targeting tumors in the so-called "larger" plus 3 to 7 cm range. This however is a general observation as I have argued that the study targets the sickest hepatocellular carcinoma subjects in which RFA has not been effective. Nevertheless, whether RFA plus Thermodox [RFAT] improves progression-free survival is the primary focus of Celsion's phase III HEAT study followed long-term survival (the secondary measure). However, I have been rethinking the statistical and mathematical arguments being touted as a predictability measure. I conclude that these models are not full-proof; in fact, I am very hesitant to accept recent assertions. Therefore, I would like to evaluate this approach as the announcement of results is imminent.

First, we know that 380 events were recorded by November 2012. However, the percentage of progression-free survival between RFA versus RFAT is yet unknown. There are a wide range of unknown variables that could skew these probability equations.

For example, the biggest variable of them all is the human subject and each human subject represents a wide range of variables such as: age, number of tumors, size of tumors, location of tumors, stage of cancer, genetic factors, general health of the subject, alcoholism, etc. The list is actually quite long and every factor represents a new variable. Plus, the technique of using RFA will vary from surgeon to surgeon and with multiple test centers, there may be other variables from location to location. Furthermore, undertake a serious study of the scientific articles reporting on RFA and you will be challenged by the variability between studies. It is not as 'cut and dry' as some imply once you read the literature. Therefore, I strongly caution against relying upon statistical models to predict the current phase III HEAT study outcome.

Second, while I don't favor betting your cash on a statistical model, I decided to pursue this approach and with acknowledgement to SA's Alex Heisenberg article ("Celsion's Phase III Blockbuster Data Revealed And It's Been Right Underneath Your Nose"), I have borrowed the enrollment dates and the numbers enrolled (See below). Note: I did add 1 additional subject to the last enrollment: 101 versus Heisenburg's 100--this is minor point to make the final tally 701.

Third, I divided the enrollment: 50% RFA versus 50% RFAT. This is an assumption as the actual percentage in the treatment arms may not be exact. As double-blinded study, I do wonder if the clinical study arms will be 'even Steven.' This too could skew a predictability model.

Fourth, there is serious debate surrounding the time to the primary triggering events (death, local tumor recurrence in the liver, distant tumor recurrence in the liver, and distant tumor recurrence outside the liver in another body organ) defined by the HEAT study. Unfortunately, there is not definitive agreement on the time to a primary event reflected in multiple studies in the literature and especially for tumors in the greater than 3 to 7 cm range.

Thus, we are left make a predictive assumption about the percentage of recurrence at 1, 2, and 3 years. I have chosen to use Poon's referenced study of overall recurrence free survival (1 year = 54.6%; 2 years = 27.3%; 3 years = 20%), but in the inverse; thus, at 1 year 45.4% had a recurrence, at 2 years 72.7% had a recurrence, and at 3 years 80% had a recurrence. I use the inverse since the goal is to predict how many RFA events can be predicted between 9/30/2008 - 11/9/2012. The one challenge and this is a huge variable is whether or not at one year, it is 45.4% or 40%, 50%, 60% etc. but once the results of the current HEAT study are published that answer will come. My hunch and it's really nothing more than a guess that the larger size tumors in the subjects will push the rate of recurrence at one year to more higher than 45.4%, but that too is yet to clarified.

Now let's look at the table.

Date of Enrollment | Incremental RFA Enrollment |
RFA Primary Triggering Event (45.4%) 12 Months |
RFA Primary Triggering Event (72.7%) 24 Months | RFA Primary Triggering Event (80%) 36 Months |
Rounded Off |

9/30/2008 | 20 | 8 | 8 | ||

8/5/2009 | 131 | 52.4 | 52 | ||

11/10/2009 | 45 | 18 | 18 | ||

1/20/2010 | 54 | 19.629 | 20 | ||

5/4/2010 | 81 | 29.4435 | 29 | ||

8/3/2010 | 74 | 26.899 | 27 | ||

9/30/2010 | 45 | 16.3575 | 16 | ||

11/11/2010 | 24 | 8.724 | 9 | ||

2/11/2011 | 42 | 9.534 | 10 | ||

3/25/2011 | 18 | 4.086 | 4 | ||

5/12/2011 | 24 | 5.448 | 5 | ||

7/11/2011 |
30 | 6.81 | 7 | ||

8/3/2011 | 12 | 2.724 | 3 | ||

5/30/2012 | 101 | *11.4635 | 11 | ||

Total | 701 | 219 |

Calculation notes:

- For simplicity, I did not divide a year into (1/12) parts otherwise if I did, the number of RFA events would be slightly higher and therefore the number of RFAT events would be slightly lower.
- *For 5/30/2012 I did use a 1/2 year (22.7%) since a portion of 101 have most likely had a recurrence.

Fifth, as of November 2012, Celsion reported 380 primary events, we can calculate using the above chart that:

- 219 RFA primary events (57.63% out of 100% of 380)
- 161 RFAT primary events (42.37% out of 100% of 380)
- Total: 380 primary events (15.26% difference out of 100% of 380)

Furthermore, as of November 2012, we can extrapolate:

- 131.5 RFA no events (40.97% out of 100% of 321)
- 189.5 RFAT no events (59.03% out of 100% of 321)
- Total: 321 no events (18.06% difference out of 100% of 321)

Now adding the two sets of differences together, we arrive at: 15.26% + 18.06% = 33.32% as the predictive measure.

Very curiously, 33% is the primary target for the phase III HEAT study and that would explain why the trial was completed because it reached its effective measure.

Conclusion

As I've stated, my view of the phase III HEAT study is not based on the math and I look forward to what I expect will be a flurry of comments about how I arrived at 33.32%. However, I do think other models are not only flawed, but way off the study's target of 33%. As I also argued, there are far too many unknown variables to predict the trial's outcome. Therefore, I encourage caution and that you should never invest more than you are willing to lose. Nevertheless, do not assume because my predictability model is much lower than others that I am any less bullish. I've been bullish of RFAT for more rationalized reasons based upon an observation of the logic and the science. With my nerves on edge for imminent news, I remain very bullish.

**Disclosure: **I am long CLSN, CVM, ATRS, QCOR. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

**Additional disclosure:** Investors buy and/or sell at their own risk. You are warned not to use this article for individual investment advice. You are warned that you could lose your entire investment by trading in the stock market. This article is for entertainment purposes only. For me 'long' is until I sell. I do not 'short' stocks. I declare that I may trade (buy and/or sell) any stock at any time mentioned in this article.