| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
LETTER |
The authors are with the Meta-Analysis Research Group, Stevens Point, Wis. Michael Huncharek is also with the Department of Clinical Oncology, Marshfield Clinic Cancer Center, Marshfield, Wis.
Correspondence: Requests for reprints should be sent to Michael Huncharek, MD, MPH, Department of Clinical Oncology, Marshfield Clinic Cancer Center, Marshfield, WI 54481 (e-mail: huncharek.michael{at}marshfieldclinic.org; metaresearch{at}hotmail.com).
We agree with all of the points raised by Marshall et al. Our meta-analysis was designed to address the issue of whether the available epidemiological data support the contention that sunscreen use is associated with an increased risk of melanoma (as suggested by some in the existing literature). Marshall et al. are correct in suggesting that the lack of statistical heterogeneity indicated by the Q statistic for the studies using population controls does not completely rule out confounding. The tests for statistical heterogeneity suffer from limited statistical power, particularly in the context of meta-analyses pooling small numbers of studies.1 Although we cannot rule out residual confounding among those reports using population controls, the analysis for heterogeneity serves as a method for demonstrating and exploring sources of variation across the available studies, despite its recognized limitations. Our finding of statistical heterogeneity was followed by a number of sensitivity analyses designed to further characterize and explain its existence and implications. This is one of the most important features of any meta-analysis.
We also agree that "the potential for . . . upward confounding to obscure a protective sunscreen effect" is entirely possible, and this potential points out an important limitation of studies to date. (Our original manuscript briefly addressed this issue but was subsequently edited.) Since sunscreen users may in fact tend to stay in the sun longer than nonusers, our data indirectly suggest a protective effect.2 Further work is needed on the behavioral aspects of sunscreen use and their impact on disease risk.
Finally, the issue of SPF is an important one. With the limited literature cited in our analysis, it was not possible to formally address this topic. Again, additional studies are needed to better understand the interplay of host factors and behavioral issues related to type and method of sunscreen use, with SPF being one of many variables of interest.
References
1. Takkouche B, Cadarso-Suarez C, Spiegelman D. Evaluation of old and new tests of heterogeneity in epidemiologic meta-analysis. Am J Epidemiol.1999;150:206–215.
2. Autier R, Dore JF, Negrier S, et al. Sunscreen use and duration of sun exposure: a double-blind, randomized trial. J Natl Cancer Inst.1999;91:1304–1309.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
