Having paps?
Gardasil has, thus far, shown a low to nothing effectiveness against cervical cancer. At best, it's 17-20% effective for mildish neoplasisa. And it goes to hell after that point.
The reason effectiveness against cervical cancer has not been established for Gardasil is because a study has not been performed that would be able to demonstrate a difference. This is quite different from actually performing a study looking at cervical cancer and finding no difference.
The progression from HPV infection to CIN grades 1-3 to carcinoma in situ to cervical cancer is well-established. Also well-established is that treatment at an earlier stage prevents progression to any of the later stages. This makes it reasonable to start by looking at whether or not any differences can be demonstrated in the earlier part of the process, especially when your final outcome is uncommon or rare.
If you wish to demonstrate whether a difference exists, a rough rule of thumb is that you need 50 events in your control group in order to have an 80% chance of detecting a significant difference (for p < 0.05 as the test of significance). This is demonstrated in the FUTURE I study, where there were an impressive-sounding six times as many cases of carcinoma in situ in the non-vaccinated group, but the real numbers were 1 and 6 leading to a p-value right on the line for statistical significance.
A study to demonstrate reduction in cervical cancer would probably require about 50,000 participants, and the evidence of benefit is too overwhelming for it to pass ethical consideration.
That reductions in the CIN grade I group account for the efficacy of the vaccine is to be expected. It is prevention of the primary disease, rather than modification of pre-existing disease, that leads to the benefit. The higher grade lesions that occurred in the vaccinated group would be expected to represent pre-existing disease in the early months or first year of the study. And this is confirmed by the graphs which demonstrate that the cumulative incidence of higher grade lesions plateaued in the vaccinated group after 18 to 24 months, whereas they kept rising in the non-vaccinated group (when looking at lesions due to vaccine serotypes).
Just a technical note about drawing causal inferences (in respect to you comments about the vaccine "causing" disease worsening or increased serotype prevalence). Causal inferences can reasonably be drawn from experimental data
for those variables on which the groups were randomized or for overwhelming non-experimental data. It is not reasonable to draw causal inferences from tiny, non-significant patterns found by the a posteriori sifting through a large pile of data not subject to randomization on the relevant characteristic, simply because they are likely to be spurious. However, while under most circumstances, the patterns you referred to would be ignored (like the finding that Libras and Geminis did not benefit from aspirin in the ISIS-2 study), because in this case they would be important if present, it is arguably reasonable to include the ability to study this question in further research on this vaccine (at least for the shifting serotypes question).
Linda
