"Quote:
Merck's sample size wasn't too small- it was too small to resolve the question about this particular difference.
According to who?"
I just meant, they didn't choose too few people. If the vaccine were to have the effect they needed, the size of the cohort would have been big enough for it to show up with statistical significance. However, it totally failed, and now we are talking about the cohort not being big enough to determine whether a tiny difference in the number infected was due to the vaccine or not. But that isn't their fault, because they couldn't have predicted that it would go the other way, with such a tiny difference between the two groups.
So I am curious to know...if it had gone differently and the groups had better evidence of efficiancy ~ it would not be too small a sample size? If perhaps there was 78% the other way? Then that would be evidence the vax works? But because it was 78% ineffective then it is too small a sample size to tell?
Wife to DH (06/09/01), Mother to DS (04/10/06) and rescuer of dachshunds ~ and that our Â arrived (06/10/11) safe and sound. Love Â our little one ~ and
If my groups consisted of two participants apiece, and the placebo group both got HIV while in the vaxed group neither person did, that would be a 100% difference. But no one would try to claim it was significant.
Well, we were referring to Mahtob's statements about it being a "tiny difference". No matter how you look at it, when there are 10 cases in the placebo group to give you a feel for the "natural occurance" and double that in the intervention group...something is quite possibly very wrong beyond coincidence and it's not exactly "tiny".
SM, what value did you enter for your t test analysis? Because it calls for the mean - which is the average of a series of numbers. What numbers are being averaged?
Anyone? Bueller?
Blessed, Bueller here. All I did was to assign numerical values to the vaccine and placebo groups and to the categorical yes/no. So in column 1 was 741 ones (1+ vaccine group); 762 twos (1+ placebo group); 672 threes (2+ vaccine group) and 691 fours (2+ placebo group). In column 2 I entered ones for HIV infected and zeros for non-HIV infected; selected the 2 groups I wished to compare and used an independent samples t-test. Which is all you can do since you are looking at the difference between the 2 groups without any other variables.
See what I mean? 25.8 might be X% higher than 25.3, but the difference between them is actually .5%, if you are looking at them as percentages.
So looking at the rate of protection based out of 100 (for Gardasil) I see that Merck claims that the vaccine is effective in the general population because it drops the number of cases from 1.5 in the placebo group to 1.2 in the vaccine group. That's an absolute difference of only 0.2%
The absolute difference in raw percentages is worthless as an indicator.
Quote:
Originally Posted by Science Mom
Which is all you can do since you are looking at the difference between the 2 groups without any other variables.
I am in complete agreement, which is why I've pointed out that the statistical significance determination can't be done here on this thread. But we can indeed call Merck's data significant by their own standard of reporting confidence intervals.
Sorry, I guess I'm no more clear about your choice to do it that way?
T test and ANOVA presume bell curve distribution of numerical data for valid interpretation of results. With a random choice of a meaningless number (1 = infected vax, 2 = noninfected placebo, or you could choose 10 million if you liked...) this is simply categorical data, not numerical. And the distribution is, of course, not bell curved, since all the numbers are the same.
Why not use methods specifically developed for categorical data, eg chi squared?
The trial is known as a "proof-of-concept" study because it enables researchers to test the concept that the vaccine candidate prevents HIV infection, or results in lower HIV levels in the blood of those who become infected with HIV. If the concept is proven â€” that is, if data generated by the study show that the vaccine candidate provides some protection against HIV, or delays or diminishes the course of HIV infection â€” this information will guide future research.
The recent announcement of the Phase II proof-of-concept efficacy trial of Merckâ€™s replication-defective adenovirus vector vaccine candidate, MRK-Ad5, is a very important step for the AIDS vaccine field. The collaboration between Merck and the HIV Vaccine Trials Network (HVTN) was announced in public discussions at the National Institutes of Healthâ€™s AIDS Vaccine Research Working Group (see Vaccine Briefs) and then at the recent AIDS Vaccine 2004 Conference in Lausanne, Switzerland. The trial, scheduled to begin at the end of this year, will test one of the most promising AIDS vaccine candidates in development.
In ongoing Phase I and II trials, MRK-Ad5 has elicited in up to 75% of vaccinees the most robust HIV-specific cellular immune responses yet seen in humans. The new trial will determine if cell-mediated immunity, at least as currently defined, can be effective in either preventing HIV infection or at least reducing post-infection viral load; the latter would hopefully lead to improved prognosis for individuals and lowered transmission rates for populations. Results of the trial are expected to be available in late 2007 or early 2008.
What is a test of concept trial?
As the name implies, a test of concept trial is about finding out if the vaccine concept or the type of vaccine being tested will be effective. A test of concept trial is not designed to establish the efficacy of a particular candidate but rather to help researchers decide if this candidate is worth testing in larger Phase III trials. These intermediate studies are also referred to as "proof of concept" or Phase IIb trials.
The number of volunteers required for such trials is smaller, only around 2-5,000 volunteers as compared to over 10,000 for Phase III trials. Phase IIb trials are therefore much easier to design and manage, and are less costly. Since fewer doses of vaccine are required, these trials are also much faster to implement because the manufacturing process is limited. Very importantly, they may also provide researchers with the immune correlates of protection, or the immune response generated by the vaccine that cause it to be effective. This can often be difficult to do in large Phase III trials.
However because Phase IIb trials are run in smaller populations, the precision of the trial is less. Therefore a vaccine can not be licensed based on the results of Phase IIb testing. If the results of a Phase IIb trial indicate that this approach is promising, a Phase III efficacy trial will be required before licensing and use of the vaccine. This means that the decision to run a Phase IIb trial will extend the total amount of time it takes to complete the clinical trials process. Phase IIb trials are an important screening step for different vaccine candidates and help organizations determine which ones to move forward into Phase III trials, without expending more time and money.
I wonder if it's possible to find out what endpoints would have been considered acceptable as "proof of concept" for this?
Do the people doing clinical trials publish exactly what they'll be looking for in advance? Do they agree beforehand on what exactly it is they're going for?
Is it published?
In speaking about this on online physician forums, there seems to general accpetance that the vaccine did cause an increase in HIV infections. So although the lay press didn't gain sufficient understanding to report this, it doesn't seem to be suppressed information from what I can tell.
The author has summarized the history of discovery, the mechanism and the clinical significance of antibody-dependent enhancement (ADE) of HIV infection. ADE has two major forms: (a) complement-mediated antibody-dependent enhancement (C-ADE) and (b) complement-independent Fc receptor-dependent ADE (FcR-ADE). The most important epitope responsible for the development of C-ADE-mediating antibodies is present in the immunodominant region of gp41 while antibodies mediating FcR-ADE react mainly with V3 loop of gp120. There are at least three fundamentally different hypotheses for the explanation of ADE in vitro: (a) increased adhesion of HIV-antibody-(complement) complexes to FcR or complement receptor carrying cells; (b) facilitation of HIV-target cell fusion by complement fragment deposited on the HIV-virions and (c) complement activation products may have a non-specific stimulatory effect on target cells resulting in enhanced virus production. FcR-ADE and C-ADE have been measured in vitro mostly by using FcR-carrying and complement receptor-carrying cell lines, respectively; no efforts have been made to standardize these methods. Several data support the possible clinical significance of FcR-ADE and C-ADE: (a) Cross-sectional and longitudinal studies indicate a correlation between the amounts of FcR-ADE and C-ADE-mediating antibodies and clinical, immunological and virological progression of the HIV-disease; (b) ADE may facilitate maternalâ€“infant HIV-1 transmission; (c) According to experiments in animal models, ADE are present and may modify the course of SIV (simian immunodeficiency) infection as well. The author raises a new hypothesis on the mechanism of the in vivo effect of C-ADE. According to the hypothesis, C-ADE-mediating antibodies exert their effect through enhancement of HIV propagation and consequent facilitation of the progression of HIV disease. Finally, according to observations from animal experiments and human clinical trials it cannot be excluded that ADE-mediating antibodies may develop, diminish the beneficial effect or may be harmful in volunteers vaccinated with HIV-1 candidate vaccines.
In speaking about this on online physician forums, there seems to general accpetance that the vaccine did cause an increase in HIV infections. So although the lay press didn't gain sufficient understanding to report this, it doesn't seem to be suppressed information from what I can tell.
You're right that this isn't rocket science. It's freshman level science that anyone with bio 101 should be able to interpret.
Honestly, if you can show this vaccine caused HIV, do it. You will be famous. I'm not kidding one tiny bit.
That would make worldwide headlines and would be a profoundly important human interest story.
You owe it to mankind.
Blessed. Do you remember when you mocked other people by calling them "local heroes" for claiming that Merck's trial data showed, as Merck observed, that the vaccine caused HIV infections.... Aren't you now claiming to be the local hero? Doesn't the above quote mean that you now owe it to mankind to publish your findings?
I'm not saying your chi square statistics are wrong. I'm more curious about the fact that you were so mocking and rude toward a position that you now support. For the record, I agree that your methodology accurately shows significance. Not that anyone here gives a poo what I think. I am just still bothered by your earlier venom. Was it misplaced, or do you think you need to publish your math and become famous as you claim will be the case.
By the way, if that 17.4% difference is significant, it does indeed mean that the vaccine caused those people to become infected with HIV.
Blessed, it seems to me that insider's statement was qualified. I bolded and underlined the relevant part. I can't find where he claimed that he had proven the statistical significance of the data before you launched those mocking attacks. His above quote is a mere statement of fact.
The supporting links and literature you're talking about were provided in the very first post on this thread. That's where Merck's numbers came from. Merck themselves observed and reported the negative impact of their vaccine. Your vitriol was clearly addressed against anyone who thought they could show the Merck data to be significant. And now you're claiming to have done just that.
The reason I seem to care is that I want to support your position now. But I can't really do that because you set the standard that anyone who claims the data is significant owes it to mankind to publish it and become famous. And I don't believe that type of mocking is accurate or fair. Merck and the mainstream are writing off this failed attempt as a noble enterprise. You seem to be admitting that now. Which makes your previous mocking a curiosity.
Sorry, I guess I'm no more clear about your choice to do it that way?
T test and ANOVA presume bell curve distribution of numerical data for valid interpretation of results. With a random choice of a meaningless number (1 = infected vax, 2 = noninfected placebo, or you could choose 10 million if you liked...) this is simply categorical data, not numerical. And the distribution is, of course, not bell curved, since all the numbers are the same.
Why not use methods specifically developed for categorical data, eg chi squared?
because in australia at least, none of the vaccine researchers use chi squared as a statistical method.
Location: back in the GREAT state of Minnesota! oh how i have missed you!
Posts: 4,980
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
SAD!!!!! :
Your life doesnĂ˘t change by the man whos elected. If your loved by someone you can't be rejected... decide what to be and go be it! If your a caged bird brake in and demand that somebody free it.
And we're done now. I'll be reviewing this thread later but we seem to have fully explored all the useful aspects of this discussion so I will not re-open it. PM me if you have any questions.
I have received a couple of compelling requests to continue discussion on this topic. Having addressed UAVs on three separate occasions, it is with considerable hesitation that I am re-opening this thread. This will be the last chance. If there is one more UAV not only will the thread be closed, I will have to consider removing it from the board completely. Please stay on topic regarding HIV vaccine and keep any personal comments to PM.
I'd really like to get to the bottom of the whole statistical significance thing.
Unfortunately, I'm probably at least borderline learning disabled at math (or maybe I just really dislike math so badly it pans out the same way...hard to say...lol) so I'm kinda at you guy's mercy here.
MK, Unfortunately the only statistic that could be applied I did. There are only the raw numbers to work with i.e. study participants that presumably were involved during the entire study period, vaccination status (albeit nebulous) and outcome HIV+/HIV-. Without more information the only conclusion that can be made (from where we are sitting) is that the vaccine failed.
I should add that the overall sample sizes were adequate to achieve a statistical difference should there be one BUT this was a multinational study and sample sizes in any one geographical region or country could have been inadequate since sample size selection (in studies such as this) is driven by disease prevalence which varies 50-fold (HIV) from country to country.
The "attack rate" was 78% higher, though, in the 2-3 dose group compared to placebo?
If it had gone the same, but opposite, wouldn't that have qualified as "proof of concept" for Merck to go into phase III, probably?
Is it just small numbers keeping significance from being reached according to whatever test you're using?
If it had been twice as many people with the same rates found, would that have been significant?
The "attack rate" was 78% higher, though, in the 2-3 dose group compared to placebo?
If it had gone the same, but opposite, wouldn't that have qualified as "proof of concept" for Merck to go into phase III, probably?
Perhaps.
Quote:
Is it just small numbers keeping significance from being reached according to whatever test you're using?
If it had been twice as many people with the same rates found, would that have been significant?
I think your grasp of stats is fine. If the rates stayed the same and the numerators were increased then you could potentially see a statistical difference.
Considering that you can change the categorical yes/no to 1/0 or whatever you like it is simple (and appropriate) to perform a t-test. I ran Chi-squares and the p values were even higher for the second group because chi-squares can be less precise particularly with low cell numbers.