VARIABILITY IN DENTAL CARIES EXPERIENCE AND ITS IMPLICATION UPON SAMPLE SIZE

¹ Harvard School of Dental Medicine, Boston, Mass.

1. Interpretative and other examining errors in DMF studies may be large, easily exceeding a 100 per cent difference between samples. In any research paper the investigator should present a clearly defined description of how the examination was conducted and what criteria of diagnosis were used. From these data the statistician must make common-sense evaluation of the accuracy of examinations prior to statistical testing of results.

2. Illustrations of actual data indicate that the standard deviations of observations about the means (averages) in DMF studies are large even where examining errors are reduced to a minimum. Standard deviations between 4 and 5 DMF teeth per person seem usual in currently available study groups in this country when the individuals are over the age of 12 years.

3. These two sources of variability imply that human DMF studies should be subjected to close scrutiny as to the validity of the data and statistical significance tests applied and reported wherever possible. Mere statements that "caries was reduced by x per cent" are not sufficient.

4. Where some estimate of the possible caries differences and standard deviations can be made in advance, experiments can be designed to have samples of sufficient size to meet the ordinary statistical tests. A formula and table are presented for the calculation of minimum sample size.

5. Presentation of data in terms of teeth rather than individuals obscures important findings relative to individual reactions, although it simplifies statistical analysis.