The two-period crossover design with baseline measurements

We consider a two-period crossover study in which each patients measured on the response variable at the start as well as at the end of both periods. We examine models in which the carryover effect at the start of the second period may be different from the carryover effect at the end, and in which the correlations between observations decrease as a function of the time between them.

In trials with a relatively short washout period, we recommend that the second baseline measurement not be incorporated into the analysis and that the data be evaluated by analysis of covariance, with the difference between the post-treatment values as the response variable and the first period's baseline value as the covariate. The absence of carryover effects must be assumed.

When the washout period is moderately long (comparable in length to either treatment period), the preferred analysis for a difference between direct treatment effects will again generally be based on the differences between post-treatment values. An analysis based on changes from baseline would, under certain assumptions about the form of the variance-covariance matrix, be preferred only for quite long washout periods and large correlations between observations. Even then, the efficiency of the test for equality of direct effects is improved if the difference between the baseline values is used as the covariate.     

A Capture-Recapture Problem When Information is Obtained From Two Qualitatively Different Sources

We first consider a wildlife-related capture recapture problem with a known total number of fish. Our objective is to estimate the number of healthy fish, H. We first use a rod that only attracts healthy fish, which are tagged and returned to the water. Later, we use a net that scoops both sick and healthy fish. Three assumptions regarding the probability of being caught by the net, conditional on health status and being caught by the rod, lead to three different estimates of H. We give approximations to expected values of the three estimates and give a condition under which they bracket H.

A potential application of these methods is to the follow-up of World Trade Center responders. Responders are disease-free when they arrive at the clean-up site and are asked to report for a visit after a fixed period of time, but some fail to do so. Some responders, whether they come to the scheduled return visit or not, spontaneously report a disease before the scheduled visit, but absence of disease is never reported in this manner. We use the methods developed to estimate the total number of subjects with disease by the time of the scheduled return visit.     

Confidence intervals for dose or treatment effect in several 2 × j tables

Consider dichotomous observations taken from T strata or tables, where within each table, the effect of J>2 doses or treatments are valuated. 'Ihe dose or treatment effect may be measured by various functions of the probability of outcomes, but it is assumed that the effect is the same in each table. Previous work on finding confidence intervals is specific to a particular function of the probabilities, based on only two doses, and limited to ML estimation of the nuisance parameters. In this paper, confidence intervals are developed based on the C, test, allowing for a unification and generalization of previous work. A computational procedure is given that minimizes the number of iterations required. An extension of the procedure to the regression framework suitable when there are large numbers of sparse tables is outlined.     

Results of National Alcohol Screening Day: college demographics, clinical characteristics, and comparison with online screening

OBJECTIVE: The authors evaluated the efficacy of the 2002 college-based National Alcohol Screening Day (NASD) by determining: (1) the demographic and clinical characteristics of the participants that were screened and (2) the degree to which those scoring at hazardous drinking levels received clinical intervention or were referred for further assessment or treatment. PARTICIPANTS: Of 45,368 participants, 23,334 were screened and 14,598 received some form of clinical intervention, defined as advice or referral. METHODS: The authors conducted nonparametric and univariate analyses to test for statistical differences in demographics, clinical-characteristics, and interventions-as a function of-screening mode. RESULTS: Nearly 34% of those screened in person had an AUDIT score of 8 or higher, indicating harmful or hazardous drinking. Of these, only 10% had ever undergone alcohol treatment. More than 58% of those screened online scored 8 or higher, and of this group, fewer than 6% had ever undergone alcohol treatment. CONCLUSIONS: These data suggest that the in-person event and the online interactive program associated with NASD are serving markedly different populations, particularly with regard to clinical indicators.