1. To control the bias caused by poor medication compliance in the Helsinki Heart Study three methods were used to measure medication compliance during the total 5 years follow up time: continuous capsule counting, semi-annual urine gemfibrozil analysis and a new method, the digoxin marker at the end of the third and fifth study years. 2. The serum lipid responses to gemfibrozil treatment varied linearly with the level of medication compliance, e.g. the mean change in serum total cholesterol was -11.4% among those whose apparent capsule consumption was greater than or equal to 90% of the scheduled dosage, -11.2% among those who had greater than or equal to 90% positive gemfibrozil analyses and -11.4% among those with good compliance according to both digoxin marker measurements. In contrast the mean serum cholesterol change was only -0.02% if the mean daily capsule count was less than 50%, -1.7% with fewer than 50% positive gemfibrozil analyses and -1.1% if the result was poor in both digoxin marker measurements. 3. Combining the different method findings revealed that the cholesterol changes tended to be small in those groups who had poor compliance classification measured by any of the methods, even if the other results showed good compliance.