Recurrent Symptomatic Hemorrhage in Cerebral Cavernous Malformations After Discontinuation of Atorvastatin or Placebo

Abstract

Background

A recent randomized prospective controlled trial demonstrated that atorvastatin for up to 2 years was safe but did not significantly alter rebleeding in cerebral cavernous malformations. However, any consequences of discontinuing atorvastatin remain unknown. We hypothesized that symptomatic hemorrhage (SH) recurs more frequently in cerebral cavernous malformations after discontinuation of atorvastatin than placebo.

Methods

We conducted a 12‐month posttreatment follow‐up of patients enrolled in the Atorvastatin Therapy in Cavernous Angiomas with Symptomatic Hemorrhage Exploratory Proof of Concept (AT CASH EPOC) trial (41 randomized to atorvastatin, 39 to placebo) to identify potential recurrent SH after trial drug discontinuation. Every SH was adjudicated by review of imaging and corresponding symptoms. Patients were excluded from follow‐up for <90% compliance with study drug, for its discontinuation <3 months after trial enrollment, for statin reinitiation <3 months after discontinuation, or for lack of follow‐up. Cases were censored during follow‐up upon cerebral cavernous malformation resection/radiation or later statin reinitiation.

Results

Follow‐up included 33 patients who had been randomized to placebo and 32 who had taken atorvastatin. Four SH events occurred at 3, 49, 84, and 225 days after atorvastatin discontinuation, and 1 SH at 395 days after discontinuing placebo. There was significantly lower symptomatic hemorrhage‐free survival in the atorvastatin‐discontinuation group (log‐rank χ²=4.136, P=0.042). The hazard ratio was 0.162 (95% CI, 0.027–0.977) for placebo versus atorvastatin discontinuation.

Conclusions

Discontinuation of atorvastatin was associated with a higher risk of recurrent SH compared with placebo discontinuation. Additional studies are warranted to confirm this hypothesis‐generating observation, examine potential mechanisms, and how best to mitigate this risk.

REGISTRATION

URL: https://www.clinicaltrials.gov; Unique identifier: NCT02603328.

Nonstandard Abbreviations and Acronyms

CCM

cerebral cavernous malformations

SH

symptomatic hemorrhage

Clinical Perspective.

What Is New?

• Given that the risk of brain bleeding upon statin discontinuation had never been documented, particularly in the presence of hemorrhagic pathology such as cerebral cavernous malformations, we here observe a high risk of bleeding upon discontinuation of atorvastatin 80 mg per day, as compared with placebo, in patients with cerebral cavernous malformations and recent symptomatic hemorrhage.

What Are the Clinical Implications?

• Although atorvastatin has been shown to be safe in patients with cerebral cavernous malformations with recent symptomatic hemorrhage, clinicians should herein consider the risk of discontinuing high‐dose atorvastatin in this clinical setting.

• It remains unclear if this applies to all statins and other doses, and whether gradual tapering of the drug might mitigate the risk, and the risk of atorvastatin discontinuation should be examined heretofore in the setting of other hemorrhagic brain pathologies.

There is much interest in developing therapies to prevent bleeding in cerebral cavernous malformations (CCMs). A leading proposed mechanistic therapeutic target is ρ‐kinase activation caused by loss of CCM gene function, resulting in vascular permeability, and CCM lesion proliferation and hemorrhage. ¹ Statins at high doses affect pleotropic ρ‐kinase inhibition and have been shown to inhibit CCM lesion development and bleeding in preclinical models. ² , ³ A recent prospective randomized clinical trial, Atorvastatin Therapy in Cavernous Angiomas with Symptomatic Hemorrhage Exploratory Proof of Concept (AT CASH EPOC, clinicaltrials.gov NCT02603328) demonstrated that atorvastatin administration for up to 2 years in patients with CCMs with symptomatic hemorrhage (SH) in the prior year was safe, but had only a weak nonsignificant benefit on CCM iron deposition and recurrent SH rates. ⁴

Atorvastatin cessation has long been reported to cause harmful rebound effects including serious cardiovascular events and ischemic strokes. ⁵ , ⁶ This has been attributed in part to exaggerated pleiotropic effects including proinflammatory factors, ρ‐kinase activation, or oxidative stress. ⁷ , ⁸ , ⁹ Risk of statin cessation in hemorrhagic stroke is more controversial, ¹⁰ and there is to date no information on any potential risk after discontinuation of statins in patients with CCMs. We examined this question in the cohort of AT CASH EPOC subjects who stopped atorvastatin or placebo during or upon conclusion of the trial. We hypothesized that SH recurred more frequently in hemorrhagic CCMs in the year after discontinuation of atorvastatin than placebo and examined potential demographic or clinical differences in patients who bled or did not bleed following atorvastatin or placebo cessation.

Methods

The data dictionary and the complete deidentified data set from the parent shall be made available as required by the US National Institute of Neurological Disorders and Stroke within a year of publication of the primary study on https://www.ninds.nih.gov/currenty‐research/research‐funded‐ninds/clinical‐research/archived‐clinical‐rersearch‐datasets. Data from this observational cohort study shall be made available to other researchers upon request from the corresponding author.

This is a cohort follow‐up study of patients previously enrolled in the AT CASH EPOC prospective randomized clinical trial. Details of the trial and its primary results have been published. ⁴ Briefly, the trial included patients ≥18 years old who had suffered an SH from a CCM lesion the prior year, and had not received surgical resection of the hemorrhagic CCM, nor prior brain irradiation, nor taken any statins in the prior year. It enrolled 80 patients randomized 1:1 to receive atorvastatin 80 mg/d for 2 years versus placebo, with double blinding to the assigned therapy until completion of the trial and reporting of trial results. Clinical and demographic features of the cohorts randomized to atorvastatin and placebo were well balanced, with no significant differences. The primary outcome in the trial was the percent change in lesional iron content as a surrogate of lesional hemorrhage, assessed by quantitative susceptibility mapping on magnetic resonance imaging in a modified intention‐to‐treat cohort of cases who received at least 1 pair of annual quantitative susceptibility mapping assessments. The full trial cohort was assessed by intention‐to‐treat for recurrent SH rates and any harmful effects of the drug versus placebo. The trial was conducted with US Food and Drug Administration investigative new drug exemption and was approved by the institutional review board at the University of Chicago, including full consent for participation in the trial and clinical follow‐up.

We conducted a 12‐month (±30 days) posttreatment follow‐up of patients randomized in the AT CASH EPOC trial by review of clinical visits at the trial site during the year after trial drug discontinuation (27 patients), or by structured electronic mail and telephone follow‐up (53 patients). We excluded patients with <90% compliance with the study drug during the trial (days used versus days in the trial), those who discontinued thr trial drug <3 months after enrollment in the trial, restarted any statin medication <3 months after trial drug discontinuation, or for lack of follow‐up. The remaining patients constituted the follow‐up cohort and were queried for 2 primary events: (1) any statin use during the year after discontinuation of trial drug and (2) any potential recurrent SH from the date of trial drug discontinuation. An SH was defined in the trial as the occurrence of new neurologic symptoms attributable to a CCM lesion and evidence of new bleeding on imaging studies. ¹¹ Patients were queried for any medication use, new clinical symptoms, and any imaging during the year after trial drug discontinuation. Patients had discontinued the drug during the trial for various reasons or upon completion of 2 years of therapy. They were assessed during follow‐up of 365 (±30) days from the date of trial drug discontinuation.

Cases in the follow‐up cohort were censored upon CCM resection or radiation, or upon initiation/reinitiation of any statin therapy during the follow‐up period (any SHs were counted until the date of censoring including any SH events during the follow‐up leading to lesion excision). Enrolled subjects were informed about their trial drug assignment upon publication of study results, 9 months after discontinuation of the study drug in the final trial subject; hence, all had remained blinded to treatment assignment during the follow‐up except for 5 patients (3 placebo and 2 atorvastatin) who knew about their drug assignment during <3 of the 12 months of follow‐up after trial drug discontinuation.

Statistical Analysis

The number of SH events was counted in the patients previously receiving placebo and atorvastatin, and the time to recurrent SH from the date of trial drug discontinuation was noted. Kaplan‐Meier survival tables for freedom from recurrent SH were constructed for cases previously receiving placebo versus atorvastatin during follow‐up. Differences were analyzed using the log‐rank (Mantel‐Cox) and Gehan‐Breslow‐Wilcoxon tests, the latter giving extra weight to early events. Hazard ratios with 95% CIs were calculated. We queried the trial database to compare clinical and demographic features of patients in the follow‐up cohort who had rebled and did not rebleed after atorvastatin or placebo discontinuation. We also assessed the respective duration of atorvastatin use in cases that bled and did not rebleed after discontinuation, dates from last prior SH until atorvastatin discontinuation, and peripheral blood leukocyte ρ‐kinase activity during the trial. These were compared using the Fisher exact test.

All statistical analyses were conducted using GraphPad Prism Version 10.5.0 (673), and significance was assumed at P<0.05.

Results

Of 39 patients initially randomized to receive placebo in the AT CASH EPOC trial, 6 were excluded from the follow‐up study (1 for noncompliance, 3 lost to follow‐up, and 2 for treatment duration of <3 months in the trial). Of 41 patients initially randomized to receive atorvastatin, 9 were excluded (3 for noncompliance, 1 for treatment duration in the trial <3 months, and 3 who resumed a statin drug <3 months posttrial). Thus, 33 patients initially assigned to placebo and 32 patients initially assigned to atorvastatin were included in posttreatment follow‐up analyses (Figure 1).

Figure 1. Profile of cases analyzed in follow‐up after trial drug discontinuation.

Eighty patients randomized in the AT CASH EPOC or EPOC (Atorvastatin Therapy in Cavernous Angiomas with Symptomatic Hemorrhage Exploratory Proof of Concept) trial were assessed for posttreatment follow‐up after discontinuation of the trial drug. Thirty‐nine patients originally assigned to placebo and 41 assigned to atorvastatin were evaluated for 1 year after stopping the study medication. In the placebo group, 33 patients were included in the follow‐up analysis and 6 were excluded (1 noncompliance, 3 lost to follow‐up, 2 treatment duration <3 months). Of those included, 27 contributed a full year of follow‐up, and 6 were censored during follow‐up (5 underwent CCM resection, 1 initiated a statin). In the atorvastatin group, 32 patients were included and 9 were excluded (3 noncompliance, 2 lost to follow‐up, 1 treatment duration <3 months, 3 resumed a statin <3 months posttrial). Of those included, 25 contributed a full year of follow‐up and 7 were censored (2 underwent CCM resection, 5 initiated a statin). Censoring occurred at the date of CCM resection, initiation/reinitiation of statin therapy during follow‐up, or at last contact. All confirmed bleeds during the follow‐up were counted until the date of censoring. The follow‐up window was planned for 12 months (±30 days) from the date of trial drug discontinuation. CCM indicates cerebral cavernous malformation.

During the 365 days (±30 days) of posttrial drug cessation follow‐up, 6 patients in the placebo group were censored at various time points (5 for resection of hemorrhagic CCM lesion and 1 who started a statin drug during the follow‐up period); 7 patients in the atorvastatin group were censored at various time points (2 for resection of hemorrhagic CCM lesion and 5 who started a statin drug during the follow‐up period) (Figure 1). Indications for CCM resection after atorvastatin discontinuation were subclinical bleeding without new symptoms in 1 case and worsening subjective symptoms without new bleeding in a second case. Indications for CCM resection after placebo discontinuation were SH events during the trial before study drug discontinuation in 4 cases (these were counted as events in the parent trial outcomes) and subclinical bleeding without new symptoms in 1 case.

During follow‐up, 4 SH events occurred at 3, 49, 84, and 225 days after atorvastatin discontinuation, and only 1 SH 395 days after discontinuing placebo. All rebleeds occurred in the index hemorrhagic CCM lesion that had triggered trial enrollment. The Kaplan‐Meier survival analyses demonstrated significantly lower SH‐free survival in the atorvastatin‐discontinuation group (log‐rank χ²=4.136, P=0.042) (Figure 2). The hazard ratio was 0.162 (95% CI, 0.027–0.977) for placebo versus atorvastatin discontinuation. The Gehan‐Breslow‐Wilcoxon test (χ²=4.080, P=0.043) highlighted early divergence in recurrent SH risk.

Figure 2. Kaplan‐Meier survival estimates showing time to symptomatic hemorrhage after trial drug discontinuation (follow‐up cohort).

Four symptomatic hemorrhage events occurred in 4 patients originally assigned to the atorvastatin group vs 1 symptomatic hemorrhage event in 1 patient originally assigned to the placebo group during the 1‐year follow‐up after discontinuation of the trial drug. Symptomatic hemorrhage times were measured from the date of discontinuation; events in the atorvastatin discontinuation group occurred at 3, 49, 84, and 225 days, and the event in the placebo‐discontinuation group occurred at 395 days. Seven patients in the prior atorvastatin group were censored (2 resections, 5 atorvastatin reinitiations) and 6 in the prior placebo group (5 resections, 1 atorvastatin initiation), noted as vertical dashes, with the censored date as the date of CCM resection or initiation/reinitiation of atorvastatin therapy, or at last follow‐up. The follow‐up window was planned for 365 days (±30 days) from discontinuation. Numbers at risk are displayed below the x axis at 0, 49, 84, 225, and 365 (±30) days. Shaded regions denote 95% CIs for the survival estimates. The log‐rank test indicated a significant difference in symptomatic hemorrhage‐free survival between groups (hazard ratio 0.162 [95% CI, 0.027–0.977]; P=0.042), consistent with fewer and later postdiscontinuation symptomatic hemorrhage events after prior placebo than after prior atorvastatin. CCM indicates cerebral cavernous malformation.

There were no differences in the demographic and clinical features among follow‐up cases receiving atorvastatin or placebo who had or did not have a recurrent SH (Tables S1 and S2). There were no differences in time from last prior SH until drug discontinuation, duration of drug use, or peripheral leukocyte ρ‐kinase activity during the trial among atorvastatin follow‐up cases who had or did not have a recurrent SH (Figures S1‐S3). Notably, 3 of 4 rebleeds occurred in patients who discontinued atorvastatin >2 years after the last prior SH.

Discussion

Current evidence‐based guidelines endorse the use of statins for cardiovascular and cholesterol indications in patients with CCMs. ¹² Despite the lack of significant benefit of atorvastatin to prevent rebleeding in patients with CCMs with recent SH, it was safe in that setting. ⁴ Yet there has never been any study of the risk of atorvastatin cessation in patients with CCMs.

We report significantly greater and earlier risk of SH in patients who stopped atorvastatin than placebo among patients with CCM with SH previously randomized to the respective treatment. Cases that bled after atorvastatin or placebo cessation had no significantly different demographic or clinical features than those who did not. Their atorvastatin treatment duration and their peripheral leukocyte ρ‐kinase activity during the trial were not notably different. Interestingly, 3 of 4 rebleeds after atorvastatin discontinuation occurred >2 years after the initial SH; hence, rebleeding was not limited to cases with recent prior bleed. All rebleeds occurred in the same lesion that had bled previously; hence, there did not seem to be a risk of new bleeding in CCMs that had not recently bled.

It is unclear why this rebound effect occurred after drug discontinuation when atorvastatin appeared to have only weak and nonsignificant benefits in the trial proper. Speculation about potential mechanisms have been suggested in other settings of exaggerated cardiovascular events and strokes after atorvastatin discontinuation. ⁷ , ⁸ , ⁹ Mechanisms in CCM lesions may be different and require further investigation.

There are several limitations to this study. Our cohort was small, and blinding was partially unmasked during follow‐up; thus, ascertainment bias cannot be excluded. It is possible that SH events may have been missed during follow‐up. More patients underwent lesion resection after placebo than atorvastatin discontinuation. Although we counted all SHs before lesion resection, this may have eliminated bleeding risk unevenly during follow‐up. We only examined differences in SH during follow‐up, with rigorous adjudication of evidence of new bleed and attributable symptoms in every case. We had not designed the study to systematically screen for subclinical bleeds or lesional iron content after drug or placebo discontinuation. With too few cases of recurrent SH, demographic or clinical differences among cases with rebleed may not have been evident in our analyses and could not be meaningfully assessed in multivariate control models.

Our results are merely hypothesis generating and should be verified in future studies and queried in large registries of patients with CCMs. We do not know if there would be such risk with atorvastatin doses <80 mg/day used in our study, other statin drugs, or in patients with CCMs without prior SH. Furthermore, it is unclear if gradual drug discontinuation would mitigate the risk. Yet our results call for caution on discontinuing atorvastatin in patients with hemorrhagic CCMs. A rebound effect of drug discontinuation should hereafter be examined as part of the future assessments of the safety of statins and other pharmacotherapies in patients with CCMs. ¹³

Conclusions

Discontinuation of atorvastatin was associated with a significantly higher and earlier risk of recurrent SH compared with placebo discontinuation. This raises concern for a possible rebound effect during the year after statin withdrawal in patients with hemorrhagic CCMs. Such observations during posttrial cohort analyses are merely hypothesis generating. Additional studies are warranted to confirm these observations, clarify mechanisms, and examine how best to mitigate the potential risk.

Sources of Funding

Funded by the National Institute of Neurological Disorders and Stroke, National Institutes of Health (NIH) R01NS107887.

Disclosures

Dr Awad is a consultant for Neurelis and Ovid, Rx. Dr Flemming was a consultant for Recursion Pharmaceuticals and Ovid Therapeutics. The other authors report no conflicts.

Supporting information

Table S1–S2

Figures S1–S3