Event rates sample size

In tong term trials there will usually be an opportunity to check the assumptions which underlay the original design and sample size calculations. This may be particularly important if the trial specifications have been made on preliminary and/or uncertain information. An interim check conducted on the blinded data may reveal that overall response variances, event rates or survival experience are not as anticipated. A revised sample size may then be calculated using suitably modified assumptions... ... 

Fig. 1. An example of power calculation to detect the indicated odds ratio for a range of risk factor prevalence and event rate with a sample size of 300 patients.

This instrument was designed to yield information intermediate between the ARC and the DSC. A sample of 0.2-0.5 g is loaded into a tube-like container and placed into the device (larger sample sizes may be used at slower scan rates). A thermocouple is connected to the outside of the tube and the cell is fitted with a pressure transducer. A similar, empty cell in the same oven with thermocouple serves as a thermal reference. The oven is heated at a slow, linear rate (0.5 to 1 °C/min), and the pressure and differential thermal data are collected. The data are presented in a fashion similar to DSC - Heat Rate (mW) vs. Temperature (°C). The thermal data are enthalpically calibrated by means of a series of standards (cahbration at high heat rates may be non-linear). Detection of thermal events approaches the sensitivity of the ARC. 

Adverse events are sometimes termed type A (usually pharmacologically predictable, relatively frequent, seldom fatal and usually identified during clinical trials) or type B (unpredictable idiosyncratic reactions which are usually infrequent but can be very serious or fatal) (Rawlins and Thompson, 1977 Venning, 1983). Postmarketing ADR monitoring usually identifies the more serious, type B reactions. The sample size needed in clinical trials to detect differences between an incidence rate of 1/10 000 and 2/10 000 is about 306 000 patients (e.g. for a placebo comparison of chloramphenicol-induced aplastic anemia, which occurs in 1/30 000 Lasagna, 1983). Clinical trials at this scale are simply impractical. 

There are numerous possible explanations for why initial phar-macogenetic associations have failed to be replicated in subsequent studies. Outcome studies are particularly problematic in this field, because they almost uniformly lack statistical power owing to insufficient event rates (due to small sample size and/or short follow-up time). The choice of end point is also important, because studies more often than not note differences in clinical outcome end points without detecting differences in surrogates such as LVEF or heart rate [29, 41]. 

Here, we discuss two alternatives if primary assurance of safety must come from a boimded hazard ratio, such as in the FDA guidance for evaluating CV risk for new antidiabetic drugs, we illustrate an adaptive trial for choosing the optimal sample size when the true event rate is not well known, while also requiring type 1 error control necessary in a confirmatory study. Type 1 error control will be illustrated via simulation. 

Multiple factors contribute to the study s power and average sample size the maximum allowable sample size, accrual rate, CV event rate, stopping boundaries, and maximum follow-up length after the last patient is enrolled are key. The maximum sample size, follow-up times, and stopping boundaries are a priori defined design features. The accrual rate can be controlled to some extent by the number of sites and inclusion criteria. The hardest to predict or control is the CV event rate. One can attempt to project this based on the study s inclusion criteria and literature however, it can be particularly difficult to predict in noncardiac populations (e.g., a trial for rofecoxib for arthritis) for which there is little literature on CV outcomes. 

Table 7.2 shows how the design enrolls more patients when the event rate is lower to ensure a sufficient number of observed CV events in equivalent situations, the adaptive design enrolls more patients when the event rate is lower, 4909 patients for 3.5 evenfs per 100 patient-years versus 4561 for 5.0 events per 100 patient-years. Furthermore, the sample size decreases when accrual decreases. This is because each enrolled patient produces greater exposure. 

The adaptive sample size can protect power over a range of event rates, but trials still need to be quite large when event rates are low. Another possibility is to introduce the risk difference in combination with the risk ratio [6]. 

Often misused rules-of-thumb in the evaluation of thermal hazards are the 100 Degree Rule and similar rules which state that, if the operating temperature of a process is 1(X)°C or some other temperature difference lower than the nearest detectable exotherm observed in a small-scale test, then the process operation will not experience this thermal event and it is not necessary to obtain more detailed information from other, more sensitive tests. Several factors govern the temperature dependent rates of heat generation as detected in small-scale tests. These include the physical aspects of the test procedure such as sample size, the Phi factor, sensitivity and agitation, and the thermokinetic aspects of the reaction being studied, in particular the activation energy. ... 

Table 13.2 Impact of sample size and cardiovascular event rate on trial durations at 90 and 80 %...

In addition, it has become accepted (and even expected) that a cardiovascular safety outcome trial be conducted to accrue the number of cardiovascular events to discharge the 1.3 threshold. The approach utilized to discharge the 1.8 and 1.3 thresholds will determine the timing of when this trial will be initiated (during Phase III or post-submission), a topic discussed shortly. Typically, these cardiovascular safety outcome trials are event driven, i.e., they are designed to accrue a prespecified number of outcomes. The number of outcomes will determine the statistical power of the study. The sample size and cardiovascular event rate, contingent on the underlying risk of the population, will impact the duration of the trial, as shown in Table 13.2. 

A large sample size has an effect on the thermal curve comparable to that of a faster scanning rate. A large sample yields a thermal curve that has a large broad peak with poor baseline resolution. Larger sample sizes are used for subtle events that need magnification, such as glass transition temperatures. 

As with many thermoanalytical measurements, there must be a compromise established between such experimental parameters as heating rate and sample size and the quality of the results in terms of sensitivity and accuracy. Faster heating or cooling rates and larger sample sizes will enhance the intensity of the signal (AT ) and, therefore, the sensitivity. On the other hand, they lead to greater thermal lag between the temperature observed for the event and its actual equilibrium value. 

---