ST 520 Statistical Principles of Clinical Trials - NCSU Statistics ...
ST 520 Statistical Principles of Clinical Trials - NCSU Statistics ...
ST 520 Statistical Principles of Clinical Trials - NCSU Statistics ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
CHAPTER 9 <strong>ST</strong> <strong>520</strong>, A. TSIATIS and D. Zhang<br />
Must be careful to choose θ, β, γ so that λ(t) ≥ 0 and if we also want a proper distribution, i.e.<br />
S(t) → 0 as t → ∞ then<br />
Λ(∞) = ∞.<br />
Other popular distributions include the log normal distribution where<br />
and the gamma distribution whose density<br />
log(T) ∼ N(µ, σ 2 ),<br />
f(t) is proportional to t ρ e −λt .<br />
Remark: In most clinical trials applications and research in survival analysis it has become<br />
common practice to use non-parametric and semi-parametric models where the shape <strong>of</strong> the<br />
distribution function is left unspecified.<br />
9.2 Censoring and Life-Table Methods<br />
Two important issues in clinical trials where the primary endpoint <strong>of</strong> interest is time to an event<br />
which are different than most studies:<br />
1. Some individuals are still alive (event <strong>of</strong> interest has not occurred) at the time <strong>of</strong> analysis.<br />
This results in right censored data.<br />
2. The length <strong>of</strong> follow-up varies due to staggered entry.<br />
This is illustrated in the schematic shown in the next page.<br />
The time to event <strong>of</strong> interest in most clinical trials is the time from entry into the study until<br />
death (right-hand panel).<br />
In addition to censoring occurring because <strong>of</strong> insufficient follow-up, it may also occur for other<br />
reasons such as<br />
• loss to follow-up (patient drops out <strong>of</strong> the study, stops coming to the clinic or moves away)<br />
PAGE 136