REDUCTION IN VERTEBRATE ANIMAL USE IN RESEARCH

LIMITED AVAILABILITY OF SOME SPECIES OF RESEARCH ANIMALS, THE HIGH COST OF RESEARCH ANIMALS, AND CONCERNS ABOUT ANIMALWELFARE HAVE CREATED A NEED FOR MORE EFFICIENT EXPERIMENTAL DESIGNS THAN THOSE COMMONLY USED IN LARGE ANIMAL STUDIES. SEQUENTIAL DESIGNS UTILIZING STAIRCASE METHODS, AVAILABLE FOR OVER 40 YEARS, HAVE BEEN SHOWN TO REQUIRE AS FEW AS ONE-THIRD THE NUMBER OF SUBJECTS AS TRADITIONAL DESIGNS IN STUDIES INVOLVING ESTIMATION OF A THRESHOLD FOR AN ALL-OR-NONE RESPONSE. HOWEVER, THESE DESIGNS HAVE NOT BEEN WIDELY UTILIZED IN LARGE ANIMAL RESEARCH BECAUSE (1) THEY APPEAR TO USE LABORATORY RESOURCES INEFFICIENTLY, (2) THE EXISTING PROCEDURES ARE CUMBERSOME TO APPLY, (3) THEIR SENSITIVITY TOPROTOCOL DEVIATIONS HAS NOT BEEN ASSESSED, AND (4) GUIDELINES FOR FINDING USEFUL STANDARD ERRORS OF ESTIMATES FOR USE IN SIGNIFICANCE TESTS NEED TO BE DEVELOPED FOR SMALLSAMPLE PROBLEMS. THIS RESEARCH WILL (1) IDENTIFY RECENT PUBLISHED STUDIES FOR WHICH SEQUENTIAL DESIGNS WOULD HAVE RESULTED IN SUBSTANTIAL SAMPLE SIZE REDUCTION, (2) DEVELOP COMPUTER SOFTWARE TO REPLACE AND EXPAND THE LIMITED TABLES CURRENTLY USED FOR THE UP-AND-DOWN METHOD, (3) APPLY THIS SOFTWARE TO EVALUATE THE SENSITIVITY OF THESE METHODS TO PROTOCOL DEVIATIONS AND VIOLATIONS OF ASSUMPTIONS, AND (4) DEVELOP GUIDELINES FOR OBTAINING STANDARD ERRORS OF THE ESTIMATES FOR USE IN SIGNIFICANCE TESTING.