Our specific aim in this study of 6 +/- 2 serial lipoprotein(a) (Lp(a)) measurements over 1.7 +/- 0.5 years in 145 patients (64 men, 81 women, 135 whites, 10 blacks, mean age 63 years) with high Lp(a) (> or = 25 mg/dl) was to assess whether, to what degree, and why Lp(a) levels change over time. A second specific aim was to study the biological coefficient of variation of Lp(a) (C.V.b). No Lp(a) lowering drugs were given. In the 145 patients, mean (S.D.) and median Lp(a) on entry (70 +/- 34 and 59 mg/dl) did not differ from levels on follow-up (72 +/- 34 and 64 mg/dl) (P > 0.4). Mean (S.D.) C.V.b was 18 +/- 9%, median 17%, minimum 1.7% and maximum 58%. Three or four samples for Lp(a) measurement should be very adequate to establish a 'true' Lp(a) value, since C.V.bs based on three to four samples did not differ (P > 0.1) from those based on five to six, seven, eight, or nine samples. For patients with mean +/- S.D. baseline Lp(a) levels of 39 +/- 3, 49 +/- 3 and 59 +/- 2, 95% of the follow-up values fell in the following ranges: 25-78, 24-93 and 34-104 mg/dl. Lp(a) on study entry was highly correlated with Lp(a) on follow-up (r = 0.84, P = 0.0001) and with Lp(a)'s C.V.b on follow-up (r = 0.24, P = 0.003). Entry Lp(a) was inversely correlated with absolute and percent change in Lp(a) on follow-up (r = -0.38, P = 0.0001 for both).(ABSTRACT TRUNCATED AT 250 WORDS)