Table 2.
Findings from prespecified and post hoc secondary analyses.
| Analysis* | Ratio of adjusted IRRs (95% CIs) for antihyperlipidemic of interest vs. referent | |||||||
|---|---|---|---|---|---|---|---|---|
| fibrates | statins | |||||||
| feno | gem | rosuva | atorva | prava | lova | simva | ||
| Further elucidating the association between antihyperlipidemic discontinuation and outcome | ||||||||
| Stratify risk segment | Days 1–30 | 0.25 (0.03–1.76) | 5.11 (0.10–265.8) | 0.21 (0.02–2.64) | 0.44 (0.15–1.30) | referent | 0.29 (0.06–1.47) | 0.42 (0.14–1.32) |
| Days 31–60 | ND | ND | ND | ND | referent | ND | ND | |
| Days 61–90 | ND | ND | ND | 0.41 (0.03–5.63) | referent | ND | 0.21 (0.01–5.33) | |
| Deconstruct composite outcome | VTE | 0.15 (0.01–1.77) | 1.45 (0.05–41.1) | 0.35 (0.02–7.56) | 0.43 (0.12–1.54) | referent | 0.63 (0.11–3.73) | 0.65 (0.18–2.40) |
| IS | 0.45 (0.02–13.1) | ND | ND | 0.88 (0.11–6.70) | referent | ND | 0.47 (0.06–3.55) | |
| Lump antihyperlipidemics of interest by likelihood of CYP2C9 inhibition** | 0.71 (0.37–1.35) | referent | ||||||
| Lump antihyperlipidemics of interest by likelihood of interacting with warfarin, per Truven Micromedex and Facts & Comparisons DDI module ratings** | 0.86 (0.45–1.64) | referent | † | |||||
| Assessing SCCS underlying assumptions, minimizing the role of bias and/or confounding | ||||||||
| Increase maximum length of risk segment from 90 to 120 days†† | 0.19 (0.03–1.32) | 1.09 (0.02–65.0) | 0.21 (0.02–2.04) | 0.46 (0.16–1.29) | referent | 0.39 (0.08–1.81) | 0.53 (0.18–1.56) | |
| Exclude segments occurring before the first risk segment (i.e., conduct left-censored unidirectional SCCS) | ND | ND | ND | 0.71 (0.06–8.36) | referent | ND | 0.36 (0.03–4.20) | |
| Exclude segments occurring after the first risk and indeterminate risk segments (i.e., conduct right-censored unidirectional SCCS) | 0.27 (0.03–2.49) | 2.54 (0.06–104.6) | 0.51 (0.03–7.94) | 0.82 (0.26–2.59) | referent | 0.85 (0.12–5.78) | 1.01 (0.31–3.27) | |
| Reclassify second or later risk and indeterminate risk segments as non-risk segments | 0.22 (0.03–1.59) | 1.51 (0.04–59.3) | 0.29 (0.02–4.41) | 0.54 (0.18–1.58) | referent | 0.43 (0.08–2.14) | 0.66 (0.22–1.98) | |
| Include average daily dose of warfarin as covariate in outcome model | 0.24 (0.03–2.09) | 8.55 (0.28–259.0) | 0.35 (0.02–5.79) | 0.90 (0.27–3.03) | referent | 0.37 (0.06–2.29) | 0.97 (0.28–3.32) | |
| Exclude subjects with >1 outcome during the observation period | 0.37 (0.05–2.77) | 2.42 (0.06–98.7) | 0.34 (0.02–5.27) | 0.86 (0.26–2.89) | referent | 0.74 (0.12–4.42) | 1.19 (0.35–4.02) | |
| Exclude subjects that die during the observation period | 0.21 (0.03–1.52) | 1.19 (0.04–36.7) | 0.15 (0.01–2.30) | 0.43 (0.14–1.27) | referent | 0.34 (0.06–1.86) | 0.62 (0.20–1.90) | |
Atorva = atorvastatin; CI = confidence interval; CYP = cytochrome P450; DDI = drug-drug interaction; feno = fenofibrate; gem = gemfibrozil; IRR = incidence rate ratio; IS = ischemic stroke; lova = lovastatin; ND = not detectable/model produced unstable estimates; prava = pravastatin; rosuva = rosuvastatin; SCCS = self-controlled case series; simva = simvastatin; VTE = venous thromboembolism.
*Examining VTE/IS as composite outcome, unless otherwise noted.
**Post hoc analysis.
†IRR for combined fenofibrate/gemfibrozil/rosuvastatin/lovastatin/simvastatin vs. atorvastatin/pravastatin listed in merged fenofibrate-gemfibrozil-rosuvastatin cell.
††Thereby increases maximum length of indeterminate risk period from 90 to 120 days.