Statistical Analyses

The goal of the study was to estimate the incidence of bacteremia and BSIs by oral microorganisms following an invasive dental cleaning in cancer patients with CVCs and meeting the eligibility criteria as listed. Using serum from the baseline blood draw, we made an initial determination of the presence of bacteremia in blood drawn through the CVC. The prevalence of bacteremia at baseline was estimated by calculating the proportion of patients who test positive for oral microorganisms in blood drawn through the CVC at that time. The proportion was reported with an exact binomial 95% confidence interval. The exact confidence interval carries the benefit of maintaining 95% coverage of the “true” prevalence of bacteremia in this clinical population even if the observed number of infected individuals at baseline is zero.

Individuals who tested positive for bacteremia at baseline were excluded from calculation of the incidence of bacteremia at 20 minutes after initiation of cleaning, at end of cleaning + 30 minutes, and 24 hours after initiation of the dental cleaning per protocol. The reason for this exclusion is that such individuals are no longer “at risk” of developing incident bacteremia once the dental cleaning begins.

Incidence rates for bacteremia were calculated for 20 minutes after starting the cleaning, 30 minutes after finishing the cleaning, and 24 hours after the time of dental cleaning initiation by calculating the proportion of individuals who are found to have become bacteremic at each of those time points. Incidence rates are reported with exact binomial 95% confidence intervals.

The incidence of BSI with oral microorganisms during the 30 days following a dental cleaning was estimated by calculation of an incidence density. As in estimating the incidence of bacteremia, any individual who was bacteremic with oral microorganisms at baseline was excluded from the incidence density calculation. The rationale for this exclusion is that it allowed us to capture incidence of BSI that was specifically related to bacteremia that developed through a dental cleaning and not through any pre-existing bacteremia. An incidence density was calculated instead of a cumulative incidence rate because it accounts for variation in the times at which BSIs are detected and for variation in the lengths of the disease-free follow-up periods for such subjects. The incidence density is reported in conjunction with an exact Poisson 95% confidence interval.