Statins and lower mortality in rheumatic diseases: An effect of immortal time bias?

Randomized controlled trials of the effectiveness of statins on rheumatic disease outcomes have shown limited or no benefit. On the other hand, observational studies have reported remarkable effectiveness of statins at reducing mortality in patients with rheumatic diseases. We evaluated these observational studies for the presence of immortal time bias, which tends to exaggerate the effectiveness of drugs by creating a survival advantage for exposed patients.

We searched PubMed for observational studies investigating the impact of statins in patients with common rheumatic diseases, including rheumatoid arthritis, osteoarthritis, lupus, ankylosing spondylitis, gout and systemic autoimmune diseases. Studies were included if estimates for all-cause mortality among statin users compared to non-users were reported. We evaluated each study for the presence of immortal time bias and estimated the impact of the bias on the published results.

We found eight observational studies investigating the impact of statins on mortality among patients with rheumatic diseases. Four studies were affected by the classical variant of immortal time bias, while the others introduced immortal time into the comparator via random-calendar date assignment. The studies with the classical form of immortal time bias, which tends to exaggerate drug effectiveness, reported protective effects of statins on mortality ranging from 13% to 57% reductions. In contrast, immortal time bias through random-calendar date assignment, which tends to play down the effectiveness by introducing immortal time in the comparator, reported 16% to 37% reductions in mortality.

Bias in observational studies may explain the discrepancy in findings with randomized controlled trials on the effectiveness of statins in patients with rheumatic diseases. Future observational studies will need to rely on incident and prevalent new-user designs that emulate randomized trials and avoid immortal time bias.