Journal of the National Cancer Institute Advance Access originally published online on February 12, 2008
JNCI Journal of the National Cancer Institute 2008 100(4):284-285; doi:10.1093/jnci/djm310
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© The Author 2008. Published by Oxford University Press.
CORRESPONDENCE |
Response: Re: Low-Fat Dietary Pattern and Cancer Incidence in the Women's Health Initiative Dietary Modification Randomized Controlled Trial
Affiliations of authors: Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA (RLP); University of Arizona, Tucson/Phoenix, AZ (CAT); Kaiser Permanente Division of Research, Oakland, CA (BC); Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Torrance, CA (RTC)
Correspondence to: Ross L. Prentice, PhD, Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave North, PO Box 19024, Seattle, WA 98109-1024 (e-mail: rprentic{at}fhcrc.org).
We thank Dr Young for drawing additional attention to the important multiple testing issue, both in the context of our Women's Health Initiative (WHI) report and in epidemiologic research more generally. However, our report did explicitly address this issue, and we believe that our conclusion that "a low-fat dietary pattern may reduce the incidence of ovarian cancer among postmenopausal women" is both technically correct and an accurate summary of these valuable data.
The WHI randomized controlled Dietary Modification Trial examined the potential of a low-fat dietary pattern to reduce invasive cancer risk. The protocol prospectively designated breast and colorectal cancer as primary outcomes, whereas breast, colon, rectum, ovary, and endometrium were listed as "diet-related outcomes" that may benefit from the intervention tested. A weighted log-rank test was specified to compare intervention and comparison groups, with weights increasing linearly from zero at randomization to a plateau of 1.0 at 10 years after randomization.
The weighted log-rank statistical significance level was P = .03 for ovarian cancer. In interpreting this result, we noted that a statistical significance level as extreme as P = .03 could occur by chance alone with a probability as large as 15% when five comparisons are conducted, using a conservative Bonferroni correction. However, we also offered additional evidence in support of an ovarian cancer risk reduction in the intervention group. These results included a hazard ratio trend test with time from randomization for which the statistical significance level (P = .01) is not readily attributable to chance and lower ovarian cancer hazard ratios (P = .05) among women whose baseline dietary percentage energy from fat was relatively high (and who make comparatively larger reductions in the fat content of their diets). Statistical significance levels were described as "nominal" to alert the reader to the lack of multiple testing considerations in these supplementary analyses.
Tests were also presented for a list of other cancer sites because we anticipated that these would be of interest to the cancer and public health research communities. We included the proviso that these additional tests need to "be interpreted in the context of the entire set of approximately 25 site-specific comparisons." None of these additional analyses provided evidence for an intervention effect in the light of these multiple testing issues, and we described these results as "readily attributable to chance."
Given the trial's cancer focus, we also reported a nominal statistical significance level for total invasive cancer (P = .10). We described this P value as a "suggestion" and did not draw a related conclusion.
Dr Young's perspective that outcomes beyond the primary (breast and colorectal cancer) endpoints "would need to be verified in a new study" is not practical given the cost and logistical requirements of a dietary intervention trial on this scale. This, and the fact that virtually all other available data on the association of dietary fat to cancer risk are observational, argues for making the fullest defensible use of data from the present trial.
Related Correspondence
CiteULike 
Connotea 
Del.icio.us What's this?
J Natl Cancer Inst 2008 100: 284.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||