Dr. Atkins notes that the main effect of prednisone on quality-of-life end points is statistically significant when compared with that of acyclovir. Our recommendation for combination therapy is predicated on treatment effects on both quality of life and cutaneous healing. The main effect of prednisone for cutaneous crusting and total healing was not significant compared with that of acyclovir. Because acceleration of cutaneous healing may reduce the complications of herpes zoster (that is, secondary bacterial infection), we believe this to be of clinical value. This recommendation is supported by the Cox regression analyses shown in Table 1.

The combined therapy group achieved significantly accelerated time to total crusting and healing and a faster return to usual activity when compared with the group receiving prednisone. The differences for cessation of acute neuritis, uninterrupted sleep, and total cessation of analgesic use were not statistically significant. Thus, our recommendation that corticosteroids should not be used without concomitant antiviral therapy incorporates disease resolution and quality-of-life end points.

Dr. Herbert suggests that risk ratios are of no clinical value. With all due respect, we disagree. Dr. Carson raises a similar concern but from a slightly different perspective. Contemporary analyses of clinical trial data have moved beyond a comparison of means. The variation in Kaplan-Meier survival outcomes among groups for the follow-up period is such that the most accurate interpretation of the data uses risk ratios. Use of different statistical measures introduces unacceptable bias. Nevertheless, the median times to achieve the three key end points were as follows: 5 days in the combined therapy group compared with 26 days in the placebo group for uninterrupted sleep, 3 days in the combined therapy group compared with 21 days in the placebo group for return to normal activity, and 14 days in the combined therapy group compared with 28 days in the placebo group for cessation of analgesic use. Our concern, however, is the potential for bias. Using the example of return to normal activity, some readers might conclude that there is a sevenfold acceleration for this end point. Indeed, the risk ratio is 3.22, indicating only a threefold acceleration in activity resumption. Thus, we will continue to use analytic tools that yield the most honest and unbiased interpretation.