Sensitivity analyses for lobar intracerebral hemorrhage (ICH) in the primary prevention (left) and prior myocardial infarction (MI) (right) settings. Plots show quality-adjusted life-years (QALYs) off statin therapy (dashed line) and on statin therapy (solid line), as a function of the 10-year cerebrocardiovascular disease (CVD) event probability (CVD10) (A); the annual MI recurrence probability expressed as multiples of the base-case probability (Base-Case Probability MI) (E); the relative risk (RR) of ICH conferred by statin therapy (RR ICH on Statin) (B and F); and off-statin ICH annual recurrence probability [ICH Probability (Off Statin)] (C and G). Two-dimensional sensitivity analyses varying RR of ICH on statin vs CVD10 in primary prevention (D) and RR ICH vs base-case MI recurrence probability in secondary prevention (H) are also shown. For these plots, the third dimension (gray-scale gradient) depicts the QALYs on statin minus QALYs off statin, allowing one to see when the net difference is positive (favor statin continuation) vs negative (favor statin discontinuation). Clinical circumstances in which the QALY difference is negative are shown to the right of the thick decision boundary line, whereas net positive QALY differences are shown to the left of this decision boundary line. Base-case values are marked with white circles. Q indicates the quality-of-life adjustment factor.
Sensitivity analyses for deep intracerebral hemorrhage (ICH) in the primary prevention (left) and prior myocardial infarction (MI) (right) settings. Plots show quality-adjusted life-years (QALYs) off statin therapy (dashed line) and on statin therapy (solid line), as a function of the 10-year cerebrocardiovascular disease (CVD) event probability (CVD10) (A); the annual MI recurrence probability expressed as multiples of the base-case probability (Base-Case Probability MI) (E); the relative risk (RR) of ICH conferred by statin therapy (RR ICH on Statin) (B and F); and off-statin ICH annual recurrence probability [ICH Probability (Off Statin)] (C and G). Two-dimensional sensitivity analyses varying RR of ICH on statin vs CVD10 in primary prevention (D) and RR ICH vs base-case MI recurrence probability in secondary prevention (H) are also shown. For these plots, the third dimension (gray-scale gradient) depicts the QALYs on statin minus QALYs off statin, allowing one to see when the net difference is positive (favor statin continuation) vs negative (favor statin discontinuation). Clinical circumstances in which the QALY difference is negative are shown to the right of the thick decision boundary line, whereas net positive QALY differences are shown to the left of this decision boundary line. Base-case values are marked with white circles. Q indicates the quality-of-life adjustment factor.
Westover MB, Bianchi MT, Eckman MH, Greenberg SM. Statin Use Following Intracerebral HemorrhageA Decision Analysis. Arch Neurol. 2011;68(5):573-579. doi:10.1001/archneurol.2010.356
Copyright 2011 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2011
Although the benefits of 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) for reducing risk of cardiac and cerebrovascular disease are well established,1,2 more widespread use of statin therapy remains controversial. A particular subgroup of patients for whom the advisability of statin use is unclear are those at high risk for intracerebral hemorrhage (ICH).3 The reason for added concern is the increased incidence of ICH observed among subjects randomized to statin therapy in a clinical trial of secondary stroke prevention.2,4 This risk amplification might have greatest relevance to patients at high risk for hemorrhage by virtue of prior ICH, particularly hemorrhages in lobar brain regions characteristic of the degenerative vascular condition cerebral amyloid angiopathy.5,6 Because ICH survivors commonly have comorbid cardiovascular risk factors that would otherwise warrant cholesterol-lowering medication, it is important to weigh the risks and benefits of statin therapy in this population.
Given the uncertainty surrounding this clinical decision, we developed a decision analytic model.7 Decision analytic models have been applied to the clinical issue of anticoagulation in patients with a high risk of falling8 or a history of ICH,9 and to the cost-effectiveness of statin therapy in coronary and cerebrovascular disease.1,10,11 To provide guidance for the frequently encountered question of whether statin use is safe after ICH, we used a decision analytic model incorporating published data regarding the beneficial effects of statins,1 the risk of recurrent deep vs lobar ICH,9 and the reported impact of statin use on ICH risk.2,4
Simulated clinical trials were conducted with a Markov state transition model7 implemented in Matlab (The Mathworks, Natick, Massachusetts). The base case for these analyses is a 65-year-old male ICH survivor. The impact of statin therapy vs no statin therapy was considered under 3 basic scenarios involving differing risk for future cerebrocardiovascular events: (1) no prior cerebral ischemic event (transient ischemic attack or ischemic stroke) and no prior cardiac ischemic event (angina or myocardial infarction [MI]) (hereafter referred to as primary prevention); (2) prior ischemic stroke at least 1 year in the past (hereafter referred to as prior stroke); and (3) prior MI at least 1 year in the past (hereafter referred to as prior MI). For each scenario, we computed the expected total quality-adjusted life-years (QALYs) on vs off statin therapy. We separately considered hemorrhages occurring in the 2 brain locations that jointly account for more than 80% of hemorrhagic strokes12: “deep” ICH (thalamus or basal ganglia) and “lobar” ICH (frontal, parietal, temporal, or occipital), because of different recurrence risks reflecting distinct underlying pathophysiologies. Nontraumatic lobar ICH in the age range considered here is mainly due to cerebral amyloid angiopathy13 and carries a higher risk of recurrent ICH compared with deep ICH,9 which is primarily related to chronic hypertension.14,15
Details of the model structure and the associated assumptions can be found in the eAppendix, eTable, and eFigure. In brief, the model consists of states that correspond to disease risk, in which simulated patients can experience any combination of “events” (eg, ischemic stroke, MI, or ICH), which may lead to increased risk of future events, change in quality of life, or death. Values for risk, outcomes, and quality of life are adopted from a recent systematic review of statin therapy1 and a previously reported decision analysis.9 Quality-of-life adjustment factors were used to account for the relative decrease in quality of life following cardiac and cerebral ischemic events.1 Risks for recurrent ICH after prior deep or lobar ICH were taken from previously reported studies.5,6 The relative risk (RR) of ICH on statin therapy from the Stroke Prevention by Aggressive Reduction in Cholesterol Levels (SPARCL) trial analysis was 1.68 (95% confidence interval, 1.09-2.59); this risk was assumed to apply to both deep and lobar ICH. Event probabilities, RRs, and quality-of-life adjustment factors pertaining to the base case are summarized in Table 1. A model schematic, as well as further details, can be found in the eFigure and eAppendix.
We simulated the effect of statins on quality-adjusted life expectancy (ie, QALYs) for patients with a history of lobar ICH separately from its effect on QALYs for patients with a prior deep ICH. Table 1 lists the base-case values used for these analyses, and the eTable lists the transition probabilities and quality-of-life values for the individual states. The increased risk of ICH associated with statin therapy (RR, 1.68) was taken from the SPARCL study.16
Our base-case patient was a 65-year-old man with lobar ICH and a 20% risk of cerebrocardiovascular disease (CVD) in the following 10 years. In the primary prevention setting (ie, no prior history of ischemic cerebrovascular or cardiac events), avoiding statins was the preferred course. This option resulted in the accrual of 6.8 QALYs, whereas statin therapy resulted in the accrual of 4.6 QALYs, a net loss for statin use of 2.2 QALYs. Modeling of statin therapy under various secondary prevention scenarios also suggested a net benefit for avoiding statins. Predicted outcomes in the setting of a history of prior MI were 4.4 QALYs on statin therapy vs 6.2 QALYS off statin therapy; for history of prior ischemic stroke, the figures were 4.2 vs 6.0 QALYs, respectively.
In sensitivity analyses, avoiding statins remained the preferred option following lobar ICH over a wide range of values for the statin-associated RR of ICH (Table 2, Figure 1). Even at the lower limit of the 95% confidence interval of the RR for ICH reported in the SPARCL study (ie, 1.09),16 avoiding statins remained the preferred option by a small margin (net loss for statin use of 0.3 QALYs). For statin therapy to be favored, the RR of ICH would need to be less than or equal to 1.03 for primary prevention, 1.07 for secondary prevention after MI, and 1.06 for secondary prevention after ischemic stroke.
In complementary sensitivity analyses, we found the strategy of avoiding statins following lobar ICH to be robust to other values used in the base-case model as well. Avoiding statins remained the preferred strategy at (off-statin) annual ICH recurrence probabilities well below the base-case value of 14% per year (Table 2). Avoiding statins was also preferred over the entire range of 10-year CVD risk values considered (0%-80%) for primary prevention, and for risks of MI recurrence up to 6-fold greater than the base-case assumption for secondary prevention (Figure 1E). These results indicate that the risk of ICH on statin therapy is not offset by the secondary prevention benefits, even if the cardiovascular risks are artificially forced to be extremely high. Two-way sensitivity analyses varying both the RR of ICH and the 10-year CVD risk (for primary prevention; Figure 1D) or MI recurrence probability (for secondary prevention; Figure 1H) again demonstrate that avoiding statins is preferred through a wide range of values around the base-case assumptions.
We performed similar analyses for our other base-case patient, a 65-year-old man with a 10-year CVD risk of 20% at the time of a deep ICH. In this scenario, the risk of recurrent ICH is substantially lower.5,9 Assuming the 1.68 RR of ICH from the SPARCL data, in the primary prevention setting, statin therapy confers a net loss of 0.8 QALYs (13.0 vs 12.2 QALYs) (Table 3). In the secondary prevention settings, statin therapy also produced net QALY losses, although they were smaller (0.2 QALYs for the post-MI setting and 0.3 QALYs for the post–ischemic stroke setting).
Thresholds for the statin-associated RR of ICH below which statin therapy was preferred were 1.20 for primary prevention, 1.50 for secondary prevention after MI, and 1.41 for secondary prevention after ischemic stroke (Table 2 and Figure 2B and F). The off-drug probabilities for ICH recurrence at which statin therapy was preferred also came relatively close to the base-case value (2.1%), including the secondary prevention settings of prior MI (1.5%) or ischemic stroke (1.3%) (Table 2). In the post-MI scenario, statin therapy was preferred if the probability of MI increased to 1.4 times the base-case probability (Table 2). Decision boundaries from a 2-way sensitivity analysis of RR for ICH and 10-year CVD risk in the primary prevention setting and of RR for ICH and MI recurrence probability in the post-MI setting for deep ICH are shown in Figure 2D and 2H, respectively.
The differences in QALYs for the various scenarios discussed emerge from interactions among multiple competing factors during a patient's life span. It is also instructive to consider the effect of individual factors over shorter time spans. Using the base-case assumptions and measuring over a single year of follow-up, primary prevention with statin therapy is projected to prevent fewer than 2 deaths from either MI or ischemic stroke per 1000 patients per year, at the expense of causing 18 lobar ICHs (in patients with prior lobar ICH) or 3 deep ICHs (in patients with prior deep ICH) per 1000 patients per year. From the perspective of disability and resulting loss of quality of life, each year of primary prevention with statin treatment saves 2.6 QALYs from MI or 2.2 QALYs from ischemic stroke per 1000 patients per year, at the expense of 58.6 QALYs for lobar ICH or 9 QALYs for deep ICH.
We used a decision analytic model to evaluate a common dilemma facing physicians of patients with a history of prior ICH and indications for statin therapy: under what clinical circumstances should statin therapy be avoided because of risk of recurrent ICH?2,4,14,15 Our analysis indicates that in settings of high recurrent ICH risk, avoiding statin therapy may be preferred. For lobar ICH in particular, which has a substantially higher recurrence rate than does deep ICH, statin therapy is predicted to increase the baseline annual probability of recurrence from approximately 14% to approximately 22%, offsetting the cardiovascular benefits for both primary and secondary cardiovascular prevention.9 Our results were robust over a wide range of CVD event rates in both primary and secondary prevention settings and over a wide range of estimates for the statin-associated RR of ICH. In the case of deep ICH, the substantially lower baseline annual recurrence rate translates into a much closer balance between statins' risks and benefits, and consequently the optimal treatment option may vary with specific circumstances. For our base-case patient with deep ICH, we found that statin therapy was worse than no treatment, but this conclusion is sensitive to variations in the assumed baseline recurrence rates of ischemic events and ICH, and the statin-associated RR of ICH, such that, in some realistic secondary prevention settings, statin therapy may be preferred.
The different results for statin use after lobar vs deep ICH arise primarily from the different annual ICH recurrence probabilities in these 2 groups.5,6These 2 ICH locations appear to reflect different underlying pathophysiology13- 15: nontraumatic lobar ICH in the age range under consideration is mostly due to cerebral amyloid angiopathy,17 whereas hypertensive vascular disease is primarily responsible for deep ICH.16 The risk of recurrent deep ICH can be lowered by the use of appropriate antihypertensive medical therapy,18- 20 whereas cerebral amyloid angiopathy currently lacks an established preventive treatment. Patients with cerebral amyloid angiopathy are at risk for symptomatic ICH and for accumulation of clinically silent microhemorrhages, which may serve as the substrate for subsequent larger ICH events, explaining the observed continued risk of lobar ICH over the course of a lifetime.21
The mechanism by which statins might amplify the risk of hemorrhagic stroke remains unclear. Historically, concerns about increased ICH risk with lipid-lowering drug therapy have centered on epidemiological studies linking low cholesterol levels with an increased rate of hemorrhagic stroke. This apparent association may be weaker than originally thought,22 and the increased risk of ICH with statin therapy found in the SPARCL study was independent of low-density lipoprotein levels.2,4 Statins are known to have pleiotropic effects, independent of their effect on cholesterol levels,23- 25 and some of these have been proposed as possible mechanisms for increasing ICH risk.26 For example, there is evidence that statins may have antithrombotic27- 32 and fibrinolytic33- 38 effects, and may enhance the activity of other fibrinolytic agents.26,39 The dose and statin-type dependency of these effects is not yet well understood.
There are important limitations to this analysis. The data driving the statin-related RR of ICH derive from post hoc analysis of a single clinical trial, performed for ischemic and hemorrhagic stroke patients randomized to receive a single dose and statin agent (80 mg of atorvastatin calcium). It is therefore uncertain whether these results generalize across multiple populations, agents, and doses. A further potential limitation of our study is that we restricted our analysis to simple all-or-none strategies of either treatment or nontreatment. It is possible, however, that a “switch-over” strategy (eg, treating a patient with statins during the early high-risk period following MI or stroke followed by cessation of treatment) might be preferable. Despite these limitations, it is notable that the finding of net loss of QALYs in statin treatment of patients with lobar ICH persisted even at the lower 95% confidence limit observed in the SPARCL study, supporting its validity. A second major limitation that applies to all decision analyses is their reliance on parameters extracted from existing literature, which may contain uncertainties. Although only a randomized trial could definitively answer the questions raised in our study, we note that the sensitivity analyses performed to address some of this uncertainty support our fundamental findings. Another important caveat is the possibility (identified in some but not all observational studies40- 44) that patients on statin therapy at the time of ICH have better outcomes than those not on statin therapy. Although such statin-associated improvements in ICH outcome would partially mitigate the loss of QALYs associated with increased ICH incidence, they appear insufficient to significantly offset this effect in patients with lobar ICH in our model (data not shown).
In summary, mathematical decision analysis of the available data suggests that, because of the high risk of recurrent ICH in survivors of prior hemorrhagic stroke, even a small amplification of this risk by use of statins suffices to recommend that they should be avoided after ICH. In the absence of data from a randomized clinical trial (ideally comparing various agents and doses), the current model provides some guidance for clinicians facing this difficult decision.
Correspondence: Steven M. Greenberg, MD, PhD, Massachusetts General Hospital Stroke Research Center, 175 Cambridge St, Ste 300, Boston, MA 02114 (email@example.com).
Accepted for Publication: November 10, 2010.
Published Online: January 10, 2011. doi:10.1001/archneurol.2010.356
Author Contributions:Study concept and design: Westover, Bianchi, and Greenberg. Acquisition of data: Westover. Analysis and interpretation of data: Westover, Bianchi, Eckman, and Greenberg. Drafting of the manuscript: Westover and Bianchi. Critical revision of the manuscript for important intellectual content: Westover, Bianchi, Eckman, and Greenberg. Statistical analysis: Westover, Bianchi, and Eckman. Obtained funding: Greenberg. Study supervision: Eckman and Greenberg.
Financial Disclosure: None reported.
Funding/Support: This work was supported by grant R01 AG026484 from the National Institutes of Health.
Additional Contributions: We thank Sherry Chou, MD, for critical review and discussions of the manuscript, and Alessandro Biffi, MD, and Jonathon Rosand, MD, for critical review and for sharing unpublished data regarding the impact of statin use on ICH outcomes.