Studying the Relationship of Patient Positioning and Intravesical Bacillus Calmette Guerin Dwell Time to Improve the Treatment of Non-muscle Invasive Bladder Cancer

Will be measured on the same participant under three test positions (ad lib, supine, and seated). A generalize linear mixed model for correlated data will be used to compare the three test conditions. Dwell time will be modeled as a function of test condition and any other covariates (i.e. patient characteristics and/or environmental aspects) thought to be related to dwell time. Interaction effects with test condition will be considered. If a majority of participants (> 75%; n = 10) have dwell times of 2 hours, the statistical method used to compare the 3 test conditions will be changed. To determine factors that impact the ability to maintain the BCG for two hour, a generalized estimating equation (GEE) model will be used. The outcome of interest will be a binomial indicator of dwell time equal to (coded as 1) or less than (coded as 0) 2 hours. Covariates will include all patient characteristics, and any environmental aspects suspected of influencing dwell time.

The probability of maintaining the BCG for at least 1 hour for each test position will be compared using a GEE model. Any patient or environmental characteristics felt to impact on dwell time will be included as covariates in the GEE model.

Compliance with study instructions (supine and seated positions only) will be measured using the Fitbit's count of number of steps taken and nursing staff observation. Compliance will be defined as zero or > zero steps (i.e. a binomial variable) during the test session. A GEE model will be constructed to compare each positions probability of remaining compliant.

Compliance to Fitbit tracking will be characterized as the time worn from appointment arrival until first void. Distribution of this variable is likely to be correlated with dwell time thus a GEE model will be used. Any patient or environmental characteristics felt to impact Fitbit tracking compliance will be included as covariates in the GEE model.

The difference between Fitbit and clinician (i.e. the bias) will be computed and compared to zero (i.e. no difference between methods). Additionally, the difference (y-axis) will be plotted as a function of clinician determined vital sign (x-axis). Bland and Altmann suggest using the gold standard (clinician determined values of vital signs) on the horizontal axis when comparing new methods to a commonly used technique. In the absence of any pattern in the plot, the 95% limits of agreement will be calculated. This interval describes the range of expected differences in 95% of patients, thus, describing the potential accuracy of the Fitbit. The mean absolute percent error for both step count and heart rate will be calculated so that the results may be compared to those reported in other medical journals.