(Unlearned) lessons from John Graunt and Kenneth Rothman: A "CLASSic" example

Abstract

This commentary reviews the work of John Graunt and Kenneth J. Rothman on statistical significance and the need for prespecification of study end points. The authors argue that it is dangerous to substitute oversimplifications based exclusively on whether a result has reached statistical significance for a rational process of causal inference. An example is given based on the Celecoxib Lonngterm Arthritis Safety Study.

References

Rothman KJ: Lessons from John Graunt . Lancet 347 37-39, 1996.

Silverstein FE, Faich G, Goldstein JL, Celecoxib Long-term Arthritis Safety Study: Gastrointestinal toxicity with colecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: The CLASS study: A randomized controlled trial . JAMA 284 1247-1255, 2000.

Hommes DW, Bura A, Mazzolai L: Subcutaneous heparin compared with continuous intravenous heparin administration in the initial treatment of deep vein thrombosis. A meta-analysis . Ann Intern Med. 116 279-284, 1992.

Messori A, Scroccaro G, Martini N: Calculation errors in meta-analysis . Ann Intern Med. 118 77-78, 1993.

Rothman KJ, Lanes S, Robins SJ: Causal inference . Epidemiology 4 555-556, 1993.

Rothman KJ, Greenland S: Approaches to statistical analysis . Rothman KJ, Greenland S: Modern Epidemiology Philadelphia Lippincott Williams & Wilkins. 183-199, 1998.

Ahlbom A: The p-value, the p-value function and the confidence interval . Ahlbom A: Biostatistics for Epidemiologists Boca Raton, Fla Lewis Publishers. 35-53, 1993.

Oakes MA: Critique of significance tests . Oakes MA: Statistical Inference Chestnut Hill, Mass Epidemiology Resources Inc. 22-74, 1990.

Jüni P, Rutjes AW, Dieppe PA: Are selective COX 2 inhibitors superior to traditional non steroidal anti-inflammatory drugs . BMJ 324 1287-1288, 2002.

Mukherjee D, Nissen SE, Topol EJ: Risk of cardiovascular events associated with selective COX-2 inhibitors . JAMA 286 954-959, 2001.