Determination and Selection of the Optimum Number of Sites and Patients for Clinical Studies

Abstract

Calculating required sample sizes is a critical step in the design of any study. For dental studies, the sample size needs to be specified at two levels: (1) the number of patients (n) enrolled in the study, and (2) the average number of sites (m) examined per patient. In general, m and n should be selected in such a way that the precision of the research findings is maximized, while the cost of the study is minimized. This objective can be realized by taking stock of the components of variation and the costs involved with enrolling patients and examining sites. The research cost for n patients ($C1/patient), at an average of m sites per patient ($C2/site), can usually be approximated by nC1 + nmC2. The precision varies as a function of the variance components, m, and n. To optimize precision for a fixed cost, the average number of sites examined per patient (m(opt)) should be approximately equal to [formula: see text] where rho is the within-patient correlation coefficient of the site-specific variable measured. When m(opt) is approximately equal to or in excess of the average number of sites available per patient, whole-mouth examinations are indicated. When m(opt) is considerably smaller than the average number of sites available, the sample of optimum size should be obtained by some random mechanism. Examination of a number of sites considerably different from m(opt) results in a waste of resources, regardless of the number of patients studied. Standard statistical analyses for determination of the patient sample size required to obtain a pre-specified precision or power are discussed.

Response of Pulp Sensibility Test Is Strongly Influenced by Periodontal Attachment Loss and Gingival Recession

“…Interestingly, in our multivariate analysis, patient age did not seem to have influenced the relationship between predictor variables and the outcome of interest. In addition, our methodology considered the subject as the unit of analysis to avoid limitations previously described for the use of the tooth as such unit (23,24).…”

Power and money: Designing statistically powerful studies while minimizing financial costs.

“…Several authors have considered the minimization of costs in this and closely related situations (e.g. Bloch, 1986; Cochran, 1977; Hedayat & Sinha, 1991; Heitjan, Derr, & Satyaswaroop, 1992; Hujoel & DeRouen, 1992; Lui, 1991; Lui & Cumberland, 1992; Sokal, 1981; Sukhatme & Sukhatme, 1970; Williams et al, 1990). Hujoel and DeRouen (1992, p. 1518) showed that the optimal number of replicate measurements is equal to where ρ is the average pairwise correlation among replicates.…”

“…Bloch, 1986; Cochran, 1977; Hedayat & Sinha, 1991; Heitjan, Derr, & Satyaswaroop, 1992; Hujoel & DeRouen, 1992; Lui, 1991; Lui & Cumberland, 1992; Sokal, 1981; Sukhatme & Sukhatme, 1970; Williams et al, 1990). Hujoel and DeRouen (1992, p. 1518) showed that the optimal number of replicate measurements is equal to where ρ is the average pairwise correlation among replicates. An estimate of ρ can often be derived from past research, pilot data, or general knowledge of the field.…”

Methods for the statistical analysis of binary data in split‐mouth designs with baseline measurements

“…It is well-known that the valid application of split-mouth designs requires the assumption of negligible carry-over effects of treatment across the different experimental units in a mouth [23,24]. In the present context, this implies that the treatment applied to sites in a given segment has a negligible effect on both the baseline and follow-up measurements observed in differently treated segments.…”