Statistical analyses were performed with commercial software (SAS ver. 8.2 for windows; SAS Institute, Cary, NC). Statistical power was estimated with another program (Power and Precision software; Biostatistics, Inc., Morristown, NJ). Power calculations a priori were based on our previous report on association between exposure to
C. pneumoniae and AMD progression.
21 The proportion of the high/medium (upper and middle tertiles) titers of antibodies to the
C. pneumoniae AR39 elementary bodies in the progressive cases of early AMD was 79%, compared to 61% in nonprogressors. To detect a similar 18% difference in proportions with the high/medium (upper and middle) titers of the antibodies would require a sample size of 101 subjects per group with 80% power and 132 per group with 90% power (two-tailed α = 0.05). The available numbers of cases with prevalent early AMD from our sample (
n = 159) should enable our study to detect a smallest difference of 15% with 80% power, or a difference of 18% with 92% power, between cases and controls. We should be able to detect odds ratios (ORs) of 1.7 at a significance level of 0.05 for any (early or late) AMD, comparing the upper to the lower tertiles of the antibody titer. In regard to late AMD, the available number of cases with prevalent or incident late AMD in this study were too small to be able to detect such a small difference, with a power of only 53% for prevalent and a power of 27% for incident late AMD. Logistic regression analyses of associations between the
C. pneumoniae antibody level and AMD were adjusted for age and smoking status, the two major risk factors known to be independently associated with AMD. Since family history of AMD may reflect heritability of the disease, whereas history of cardiovascular events may be associated with both the outcome and covariates of interest, additional adjustment for AMD family history and histories of heart attack or stroke was performed to minimize potential confounding effects. We used analysis of covariance (PROC GLM, SAS) to compare the means of the log-transformed antibody titers adjusted for age, gender, and current smoking. Tests with
P < 0.05 were considered significant.