Dear Sir:

We write to respond to the letter of Dr. Wu et al. referring to our article “BMI and Cervical Cancer Screening among White, African-American, and Hispanic Women in the United States,” which was published in the July 2005 issue of Obesity Research. In their letter, Wu et al. presented their results from analyses similar to those in our published study examining the association between BMI and Pap smear testing stratified by race. While their findings for white women were largely similar to ours, they report that moderately obese African-American women (BMI = 35 to 40 kg/m2) were actually almost one-half as likely to receive Pap testing (or 1.93 times as likely to under-use Pap tests) as their normal weight counterparts. In contrast, we found that moderately obese African-American women were only slightly less likely to undergo Pap testing (6% less likely), and this difference was not statistically significant.

There were several methodological differences between our study and the analyses by Wu et al. While Wu et al. excluded women who received Pap testing for reasons other than routine screening, we chose to retain these women in our sample. Interestingly, despite their additional exclusion, Wu et al. had greater numbers of white and African-American women in their sample; nonetheless, the unadjusted relationships between BMI and Pap testing among African-American women that they report are similar to our published results. The unadjusted relative risk for screening in moderately obese compared with normal-weight African-American women was 0.94 (83% divided by 88%) in our study and 0.99 (79% divided by 80%) in the analyses by Wu et al. The primary discrepancy in our findings relates to the adjusted findings. Whereas our adjusted findings of a relative risk of 0.94 were not different from our unadjusted relative risks, the adjusted relative risk of Wu et al. of 0.52 for Pap testing (or 1.93 for under-use) deviates substantially from their unadjusted relative risk of 0.99. We agree with Wu et al. that this inconsistency is likely a result of the additional factors that Wu et al. adjusted for in their model. In our study, we carefully selected the factors that we included in our adjusted models. To evaluate the true effect of obesity as a potential barrier to screening that was independent of potential confounders, we adjusted for previously known or suspected barriers to care, such as sociodemographic factors, and access to care in our initial model. In subsequent models, we serially adjusted for factors that we thought might be in the causal pathway in the relationship between obesity and screening to determine the explanatory power of these factors (e.g., illness burden, physical functioning, self-esteem and psychological functioning, and health habits). Results from these subsequent models were similar to results from our initial model. In addition to some of these factors, Wu et al. adjusted for variables that are likely not true confounders but rather may be collinear with or proxies for the outcome of interest (Pap testing). Some examples include their adjustment for other cancer screening, immunizations, and respondent perception of cancer risk. Such overadjustment may lead to misleading results, which we suspect may have occurred in analyses by Wu et al.