|
|
|
|
|
|
|
depends very much on the type of study. For a stable chronic disease, such as essential hypertension, large volumes of data should be available to provide good estimates of the number of patients expected. This will come from medical practitioners' records, but it is vital to modify any estimating database for the current study. What the investigator observes is not that all the hypertensives disappearthere are just as many as everbut that he had not applied the selection criteria when estimating recruitment. One of the author's client companies applies a standard 40 percent factor to all investigator recruitment estimates (i.e., they expect investigators to recruit 40 percent of what they predict). This is seldom completely accurate for an individual center (although much better than the original estimate), but averages quite well across all the centers. |
|
|
|
|
|
|
|
|
More sophisticated approaches to recruitment statistics are being evaluated. These data can be converted into valuable metrics for selecting sites (rather than for monitoring progress), especially if they are reduced to an ordinal scale [2], thus, making comparisons between sites simpler. The values can be adjusted for other factors, such as the proportion of patients evaluable, the monitor time spent on-site (a reflection of data quality), and the number of data queries, to give a global measure of site performance. |
|
|
|
|
|
|
|
|
For acute diseases, there is a higher risk that recruitment estimates are inaccurate because one is assuming that new cases arise with predictable frequency. For instance, some conditions are strongly seasonal, and some seasons will be better (or worse) than others. So it is vital to retrieve data far back enough in time to avoid being misled by an unusually high-prevalence season. Even if we are reassured by this, we should still ask the all-important project manager's question What happens if ? In this case, What happens if the next season is unusually benign? |
|
|
|
|
|
|
|
|
Earlier, we considered the challenge of achieving a protocol which does not need to be amended. Even if we meet this challenge, the next one is to ensure compliance. If the protocol is difficult to follow, and even if we have no problems in finding patients (a rare scenario), there is still the great danger that many of these patients are invalidated by protocol violations because the drug is now being used in the real world of clinical medicine. If it is critical that clinic assessments are carried out at particular times of day (for example, to coincide with trough drug levels or to plot the time course of postdose response), how confident can we be that this will be observed? Can we measure the impact on the study of exceptions to the rule? Can we estimate how many valid patients we might lose? |
|
|
|
|
|