Both Elisaf and colleagues and Dr. De Lima report data that confirm elevation of serum Lp(a) levels in patients with early renal failure. Other groups have also reported similar data on patients with renal failure treated conservatively (1, 2). Elisaf and colleagues noted similar frequency distribution of apolipoprotein(a) isoforms in patients with normal and moderately impaired renal function. This finding is consistent with our observations. They also showed that serum Lp(a) levels are significantly increased only in the patients expressing the high-molecular-weight apolipoprotein(a) isoforms, whereas the patients expressing low-molecular-weight isoforms have Lp(a) levels similar to those in persons with normal renal function. We have further analyzed apolipoprotein(a) isoforms in 97 patients with renal failure and 153 age- and sex-matched controls by combining the apolipoprotein(a) phenotypes in five subgroups according to the molecular weight of the smallest isoforms, as performed in earlier studies (3) (Table). This analysis indicates the presence of an inverse relation between isoform protein size and serum Lp(a) levels in patients with normal and in those with moderately impaired renal function. However, in contrast to Elisaf and colleagues' findings, it shows that Lp(a) levels differ significantly only in the subgroup with the apolipoprotein(a) isoforms of the lowest molecular weight.

De Lima and his coworkers found similar serum Lp(a) levels in two groups of patients with renal failure (serum creatinine levels of 2.0 to 4.0 mg/dL and >4.0 mg/dL) and did not find an association between Lp(a) levels and serum creatinine levels. He concludes that maximum Lp(a) levels are reached early in the course of renal disease. We disagree with these conclusions: In addition to the limitations caused by the poor definition of the degree of renal failure by measurement of serum creatinine, the groups were too small (16 and 18 patients, respectively) to have sufficient statistical power. The wide range of serum Lp(a) levels and the skewed distribution of these levels require cautious statistical evaluation, and statistical simulations indicate that approximately 100 patients per group are required for the detection of a 30% difference at a probability less than 5% (4).