Click this link
to close this window and return to the course once you have completed the case study.

CASE STUDY: Expedience, Misrepresentation, or Falsification?*

Dr. Leyos is a senior scientist in an internationally respected cancer research institute. His research group has recently been studying the regulation of a novel gene that may be a primary factor in allowing tumors to metastasize. Three pilot (preliminary) experiments have shown positive results consistent with the hypothesized role, but none of these studies is publishable. In the first case, investigators were not blinded to the origins of the samples for data collection and analyses. In a second case, adequate controls were not included and in the third case, the freezer thawed resulting in some sample degradation. However, despite these limitations, all three pilot experiments were consistent with a clear confirmation of the hypothesis. Based on these results, a definitive experiment was designed and carried out. Because of the long incubation periods and assay times, the experiment required six months to complete. Because demonstration of the effect required pooling of 20 mice for each data point, the experiment was extremely costly both in lives of animals and dollars spent. On completion of the final assays, Dr. Leyos learned that labels fell off of two samples, one for a control group and the other for an experimental group. If the two samples are omitted from the analysis, the results just miss reaching the accepted level of statistical significance (P<0.05). If the samples are assigned to control and experimental groups one way, the final analysis is also not statistically significant and is not consistent with the previous three experiments. However, if the unlabeled samples are applied in the opposite way, the final analysis is statistically significant and consistent with the pilot studies. Dr. Leyos is trying to decide among the following courses of action.

Click on one of the following choices:
Repeat the experiment (at a cost of approximately 6 months, 300 animals, and $40,000).

Attempt to publish the findings omitting the questionable samples.

Assign the two samples to their likely groups and publish the statistically significant and convincing results.