How to Avoid Bumping into the Translational Roadblock
Translating neuroprotective efficacy from animal studies to clinical trials in humans has been fraught with difficulty. This failure might be because animal studies were falsely positive or clinical trials were falsely negative. Here, I focus on the measu
- PDF / 104,070 Bytes
- 9 Pages / 504 x 720 pts Page_size
- 100 Downloads / 197 Views
1. Introduction “…you will meet with several observations and experiments which, though communicated for true by candid authors or undistrusted eye-witnesses, or perhaps recommended by your own experience, may, upon further trial, disappoint your expectation, either not at all succeeding, or at least varying much from what you expected” (Robert Boyle 1693, Concerning the Unsuccessfulness of Experiments)
Valid experiments are those which give appropriate descriptions of some biological truth in a system being studied. Internal validity relates to the extent to which an experiment accurately describes what has happened in that model system. External validity relates to the extent to which the results from that model system can be generalized to predict what might happen, for instance, in a group of patients with the disease being modeled in response to the drug being tested. Ulrich Dirnagl (ed.), Rodent Models of Stroke, Neuromethods, vol. 47, DOI 10.1007/978-1-60761-750-1_2, © Springer Science+Business Media, LLC 2010
7
8
Macleod
Stroke is one condition where, despite substantial efforts in the neuroscience community, translation of efficacy to humans has proved exceptionally difficult (1). The modeling of stroke in rodents usually involves measuring some outcome – be it a change in gene expression, an increase in protein phosphorylation, a volume of cerebral infarction or post stroke behavior – and it may also involve determining a change caused by an experimental intervention in that outcome. For our purposes, here we concentrate on experiments testing the efficacy of candidate stroke drugs, but the same considerations apply to all experiments where any outcome is measured. 1.1. The Validity of Individual Experiments
Following treatment, infarct volume and neurobehavioral score may improve, worsen, or be unchanged. We hope that results from our experiments will reflect “biological truth”, but of course this is not always the case. Measurement error and biological variability mean that our sample of animals can never describe completely what the outcome would be across all animals. These are random errors, and while they reduce the precision of the estimate of biological effect, they are as likely to underestimate effects as they are to overestimate effects. The likely scale of this random error can be estimated in preliminary experiments, and we can then predict how closely the results of an experiment of given size reflect the population response. This allows us to estimate how large an experiment should be to have a reasonable (specified) probability of detecting a treatment effect of a given size. There are also, however, sources of nonrandom error (bias), which cause the estimate of effect to be consistently understated or, more usually in this context, overstated. These sources of bias include selection bias, performance bias, ascertainment bias, and attrition bias (see also Chap. 19).
1.1.1. Selection Bias
The only difference between experimental groups should be the different treatments they recei
Data Loading...
