We have identified over 20 different medical factors that might potentially confound our study on aging, such as race, blood pressure, diabetes, type or size of tumor adjacent to the normal section etc (supplemental Table 1). Fourteen medical factors affected less than ten patients (such as diabetes or proteinuria), making it unlikely that they could account for age-related change in gene expression in the 74 patients analyzed. Six medical factors occurred in ten or more patients (race, two types of tumors, size of tumor and hypertension), but it is unlikely that these affect our aging study for the following reasons.First, with the exception of transitional cell carcinoma, none of the other medical factors were skewed with respect to age, and would not be expected to bias gene expression in an age-related fashion (Supplemental Figure 1).
Second, the nephrectomies contained renal cell carcinoma and transitional cell carcinoma that were localized to an isolated region of the kidney. Our normal samples were obtained from the region of the kidney furthest from the carcinoma, and our samples are not directly contaminated with cancer cells. This procedure for obtaining kidney samples has been used in previous publications to profile gene expression in normal kidney (Higgins et al., 2004) and as a normal control in a kidney cancer study (Higgins et al., 2003).
Third, we used regression models to directly test whether these medical factors affect age-related expression, and found that they do not (Supplemental Figures 2-7).
Finally, five of the samples were from kidneys that did not have a tumor, and
two of these were for transplantation that had no associated pathology. These
five samples behaved similarly to the other samples in Figure 3. Although five
samples are not statistically significant, the observation that they do not
stand out from the other samples taken from patients with pathology supports
our case that our results are not confounded by disease.