Examination of the FDA Warning for Statins and Cognitive Dysfunction

Background: The ACC/AHA released new guidelines in December of 2013 for treatment of high blood cholesterol to simplify identification and treatment of patients most likely to benefit from statins. These guidelines may result in more patients receiving statin therapy, and at younger ages. In 2012, the U.S. Food and Drug Administration (FDA) mandated warnings for all statin drugs for possible adverse effects on cognitive performance. Statins can be classified as having greater lipophilic or hydrophilic solubility properties with lipophilic statins more readily crossing the blood brain barrier, and possibly differentially inducing detrimental cognitive effects. Objective: We sought to analyze generalizability of the FDA statin class warning. Methods: De-identified publicly-available data were analyzed from the FDA Adverse Event Reporting System (AERS) in relation to reports of cognitive dysfunction (primary outcome), and by type of statin (lipophilic, hydrophilic) versus “control” drugs used in the general population. Results: Significantly higher proportional reporting ratios (PRRs) were observed for lipophilic statins, which more readily cross the blood-brain barrier, (range: 1.47-3.51) compared to hydrophilic statins (range: 0.69-1.64). However, fluvastatin, lovastatin, and pitavastatin (lipophilic) had relatively few adverse reports. The signal of higher risk of cognitive dysfunction was observed for the lipophilic statin atorvastatin (PRR = 2.59, 95% confidence interval: 2.44-2.75) followed by simvastatin (PRR = 2.22, 95% confidence interval: 2.04-2.31). Hydrophilic statins (rosuvastatin, pravastatin) showed essentially no evidence suggestive of heightened risk of cognitive dysfunction. Fluvastatin, lovastatin, and pitavastatin had relatively few adverse reports, and no evidence of a higher proportion of cognitive dysfunction reports compared to the control drugs in aggregate (PRR range: 0.22 to 1.48). Conclusions: Inconsistent with the FDA class warning, highly lipophilic statins with specific pharmacokinetic properties (atorvastatin, simvastatin) appear to confer a significantly greater risk of adverse cognitive effects compared to other lipophilic statins and those with hydrophilic solubility properties.


Introduction
Statins are widely prescribed in the United States for primary and secondary cardiovascular risk reduction and constitute an estimated $26 billion dollar per year market for pharmaceutical manufacturers [1]. The numbers of patients receiving statin therapy is likely to increase in the coming years based on the dissemination in 2013 of the American College of Cardiology/American Heart Association's (ACC/ AHA) updated clinical guidelines for the assessment and management of blood cholesterol for primary and secondary prevention [2]. Although the new guidelines simplify clinical treatment strategies, they promote a broader risk assessment tool which may increase the numbers of patients between the ages of 20 -79 eligible for statin use for primary prevention to an estimated 46 million Americans [3]. Statins are believed to reduce cardiovascular risk by lowering serum total cholesterol and low density lipoprotein (LDL) cholesterol, although, a number of pleiotropic effects including reductions in inflammatory cytokines and plasminogen activator levels, improved endothelial function, anti-thrombotic effects, and plaque stabilization have been proposed as equally important vascular benefits [4][5][6]. Although statins are among the most widely prescribed medications, they are also among the most controversial. Clinical trials dating back to the 1990s have generated compelling evidence as to the effectiveness of statins for the reduction of cardiovascular morbidity and mortality in patients with established coronary artery disease (i.e. secondary prevention) [7][8][9]. While these trials provide evidence for secondary prevention of coronary heart disease, statins have been plagued by conflicting evidence of potential associations with higher all-cause mortality including cancer, trauma and suicide, as well as adverse physical, psychological, and cognitive effects [10][11][12][13][14][15][16]. More recently, controversies surrounding statins stem from a lack of consensus in the scientific community as to the benefits of statins for primary prevention in the absence of established coronary artery disease [17][18][19][20]. Despite these unresolved controversies, statins are increasingly endorsed and prescribed for daily, long-term use for primary prevention across a broader spectrum of low-to moderaterisk patients including among children through older adults.
In the practice setting, the risk/benefit considerations of advising statin use in low-to moderate-risk patients has been checked, at least in part, by considerations of the economic costs associated with longterm statin use in healthy, asymptomatic patients, as well as by concerns over the potential for adverse effects on liver function and muscle tissue. Commonly recognized adverse effects associated with statins have mainly focused on their well-known potential for induced myalgia and rarely, rhabdomyolysis. Although reports of adverse neurocognitive effects associated with statin use have been previously published in the scientific literature [12,21] and in patient-based reporting systems [22], adverse neurocognitive effects have been less well recognized and possibly missed by clinicians [22] particularly in light of research suggesting that statins may confer neuroprotective benefits on neurovascular and neurodegenerative diseases including vascular and Alzheimer's dementia [23][24][25][26][27]. In contrast, a post marketing surveillance study unmasked a number of adverse effects of statins on cognitive performance and psychological symptoms including memory loss, amnesia, altered mood and irritability, sleep disruption and confusion such that in 2012, the Food and Drug Administration (FDA) mandated warnings on all statin labels [28]. Although this warning extends to all statins, previous research to examine statin-induced neurocognitive effects has provided clinically important insights against a class effect depending on the lipophilicity and resultant tissue selectivity of the statin [29][30][31][32]. Specifically, statins can be classified as having greater lipophilic or hydrophilic solubility properties, with lipophilic statins tending to achieve higher levels of exposure in non-hepatic tissues as lipophilic statins passively and nonselectively diffuse into both hepatocyte and non-hepatocyte tissue; hydrophilic statins rely largely on active transport into hepatocyte to exert their effects [33][34][35][36][37][38]. High hepatoselectivity is thought to translate into a reduced risk of adverse effects [38].
Accordingly, we sought to investigate the effect of type of statin (lipophilic versus hydrophilic) on reports of cognitive dysfunction using the FDA Adverse Event Reporting System (AERS), and to gauge the generalizability of the FDA mandated statin-at-large warning. To our knowledge, no previous studies have used the AERS database for this purpose and in comparison to a range of "control" drugs commonly used in the general population (i.e. to parallel statin use).

Materials and Methods
The FDA AERS is a computerized information database designed to support the FDA's post-marketing safety surveillance program for all approved drug and therapeutic biologic products. The FDA uses AERS to monitor for new adverse events and medication errors that might occur with these marketed products, in a post-marketing setting [39]. Of note, adverse events reported for particular medications do not list the primary indication for which the medication was used (e.g. specific disease condition and on-label versus off-label use, dosages). Adverse events in AERS are coded to Medical Dictionary for Regulatory Activities (MedDRA) terminology [40]. Importantly, reporting of adverse events from the point of care into AERS is voluntary (passive surveillance) in the U.S., with estimates of only 1% to 10% of all true adverse reactions being chronicled [41,42]. With this known underreporting and other limitations-including the fact that FDA does not require proof of a causal relationship between a product and event and that the incidence of adverse events cannot be calculated from AERS-this system is most appropriately used for "signal" identification for potential safety concerns which may provide the rationale for further evaluation, such as through the conduct of epidemiological studies [43,44] or prospective research.

Database utilization and search terms
All quarterly data files from the FDA AERS website [39] from January 1, 2004 through September 30, 2012 were downloaded in ASCII text format and converted to SAS® databases using the SAS System®. These data included the 7 separate quarterly data files publicly available (demographics, drug information, reaction information, outcome information, source of reporting, therapy dates, and manner of use) which were merged together using a common deidentified unique record number. Although the data are publicly available and de-identified, formal Institutional Review Board approval for analyses of these data was obtained from the University of South Florida, which verified compliance with Federal Exemption criteria outlined in 45CFR46.101(b).

AERS definitions
The AERS reports contain limited clinical information, with more detailed information on timing and nature of each reported event. Occupation: The occupation of the reporter of the event (physician, pharmacist, other health professional, lawyer, or consumer).

Control drugs
We selected comparator drugs with similar consumption properties as statins, yet no known or suspected relationship with cognitive dysfunction. This included drugs with a high prevalence, potential daily use, different indications (e.g. pain, hypertension, gastroesophageal reflux disease), and frequent use among adults and older adults. The control drugs and their trade names were as follows: aspirin, ibuprofen, acetaminophen, diovan (Valsartan), enalapril (Renitec, Vasotec), hydrochlorothiazide, esomeprazole (Nexium), omeprazole (Losec, Omesec, Prilosec), clopidogrel (Plavix). For both statins and control drugs, adverse reports were only selected when the reported role in the adverse event was classified as "primary."

Outcome reactions
For determination of the primary outcome "cognitive dysfunction," two of the authors (F.S., K.K) reviewed a frequency listing of all names of adverse reports listed in AERS. By consensus, the following adverse reports were selected to comprise the broad category of cognitive dysfunction: amnesia (both unspecified type and global amnesia), cognitive disorder, confusional state, dementia, delusion, and memory impairment.

Statistical analysis
Adverse event reports for each individual statin and type of statin were considered the "signal" group and compared individually against the control drugs selected. Initially, chi-square tests were used to compare reporting characteristics of the statin and control drugs for reports in which cognitive dysfunction (as defined above) was listed as an adverse reaction. For all adverse event drug reports (statin and control), the proportion in which cognitive dysfunction was listed as an adverse reaction was calculated. These proportions formed the basis for calculating proportional reporting ratios (PRRs) with the statin drug (or type) in the numerator and control drug in the denominator. Thus, PRRs greater than one are indicative that the statin drug has a higher proportion of reported adverse events listed as cognitive dysfunction in comparison to the control drugs. When interpreting the PRR, prior reports (and in this analysis) of "signal" detection of heightened risk have used a minimum of three or more cases, PRR ≥ 2, and chi-square value of ≥ 4.0 [48]. P-values of the PRRs were calculated using chi-square tests with Yates correction and defining statistical significance as P<0.05.

Results
For the period January 1, 2004 through September 30, 2012, the "Drug" file within AERS contained a total of 16,471,747 records for evaluation. From these records, 1,270,721 adverse reports were identified for the collective set of statin and control drugs. Of these reports, 190,727 (15.0%) were classified as being the primary suspected drug in the reaction and formed the basis of the analysis. The remaining adverse reports were classified as secondary suspect drug (11.2%), concomitant drug (73.4%), and interacting drug (0.3%).
Of the 190,727 drug reports selected, 3,287 were for statins reported as the primary suspected drug related to cognitive dysfunction, and 1,580 were for the control drugs. The control drug reports for cognitive dysfunction were more likely than the statin reports to be among females (58.1% versus 46.4%) and submitted by consumers (52.2% versus 29.0%). Of note, among lipophilic statin reports of cognitive dysfunction, 11.1% were submitted by a lawyer compared to just 0.5% among hydrophilic statin reports. For lipophilic statin reports of cognitive dysfunction, the percentages submitted by a lawyer over time were

Proportional reporting of cognitive dysfunction
Comparing the proportion of all adverse reports listed as cognitive dysfunction for statins versus the 8 individual control drugs, much higher PRRs were observed for lipophilic statins (range: 1.47 to 3.51) compared to hydrophilic statins (range: 0.69 to 1.64) [ Table 2]. Six of the 8 PRR estimates for lipophilic statins versus individual control drugs exceeded the value of >2 specified as a signal of heightened risk

Proportional reporting of individual lipophilic statins
Given that the signal of cognitive dysfunction risk was restricted to the class of lipophilic statins, PRRs for individual lipophilic statins were calculated [ Figure 1]. Fluvastatin, lovastatin, and pitavastatin had relatively few adverse reports in the AERS database [ Figure 1, Panel A], and there was little to no evidence of a higher proportion of cognitive dysfunction reports compared to the control drugs in aggregate [PRR range: 0.22 to 1.48, Figure 1, Panel B]. In contrast, the signal suggestive of higher risk of cognitive dysfunction was observed for

Proportional reporting of individual types of cognitive dysfunction
In examining individual types of cognitive dysfunction reported, and restricting the analyses to atorvastatin and simvastatin (i.e. primary signals identified), the highest PRRs compared to all control drugs were for cognitive disorder (atorvastatin: 10.05; simvastatin: 6.31) followed by amnesia (atorvastatin: 8.12; simvastatin: 5.98) and then memory impairment (atorvastatin: 2.92; simvastatin: 2.32) [ Table  3]. For all adverse events reported for atorvastatin, 3.2% reported amnesia as a primary adverse reaction. Comparing atorvastatin versus

Subgroup analyses
For the two statins with the highest signal (atorvastatin and simvastatin) for adverse cognitive outcomes (amnesia and cognitive disorder), subgroup analyses were conducted. For atorvastatin, the PRR versus all control drugs for amnesia was markedly higher for reports that were submitted by health professionals (PRR=11.84, 95% confidence interval: 9.60, 14.62) compared to reports submitted by consumers (PRR=3.77, 95% confidence interval: 3.17, 4.49). For reports of atorvastatin and the occurrence of cognitive disorder, the PRRs were exceptionally high (>12) among males and during the time period 2006-2009. For simvastatin adverse reports, PRRs were generally similar among all subgroups except an apparent higher proportion of reports for cognitive disorder during the period 2010-2011.

Discussion
In this analysis of the FDA AERS database, we identified a significantly higher proportion of adverse reports of cognitive dysfunction associated with commonly prescribed potent, lipophilic statins, specifically, atorvastatin and simvastatin, compared to less frequently prescribed lipophilic statins (fluvastatin, lovastatin, pitavastatin) and to selected drugs commonly used in the general population. In addition, hydrophilic statins (rosuvastatin and pravastatin) showed essentially no evidence suggestive of heightened risk of cognitive dysfunction and were shown to have a statistically lower proportion of reported cognitive dysfunction compared to hydrochlorothiazide, a diuretic commonly prescribed for blood pressure management. Lipophilic statins metabolized through the CYP2C9 pathway (pitavastatin and fluvastatin; Table 1) or with combined low absorption and bioavailability (lovastatin) were associated with relatively few adverse reports in the AERS database with similar reports (proportionally) of cognitive dysfunction as the control drugs in aggregate.

Potential adverse mechanism
Statins can be classified as having greater lipophilic or hydrophilic solubility properties, with lipophilic statins tending to achieve higher levels of exposure in non-hepatic tissues because lipophilic statins passively and non-selectively diffuse into both hepatocyte and nonhepatocyte tissue; hydrophilic statins rely largely on active transport into hepatocyte to exert their effects [6,33,35,37,49,50]. High hepatoselectivity is thought to translate into a reduced risk of adverse effects [31,33]. Atorvastatin and simvastatin are potent lipid lowering statins, both metabolized through the CYP3A4 pathway, and possessing either long half-life (atorvastatin) or high absorption (simvastatin) [ Table 1]. In contrast, fluvastatin undergoes extensive first-past metabolism rendering lower bioavailibilty and extensive protein binding reducing systemic exposure to non-hepatocyte tissue [31,33]. To maintain cell integrity and function, neurons require a sufficient supply of unesterified cholesterol. While all statins may potentially affect brain cholesterol metabolism by lowering available plasma cholesterol levels, lipophilic statins are thought to directly influence brain cholesterol metabolism by crossing the blood brain barrier and inhibiting the synthesis of cholesterol within nerve cells or by limiting available extracellular cholesterol. This mechanism is consistent with reports of lipophilic statins having a greater distribution and severity of cognitive dysfunction, including memory loss, compared with hydrophilic statins [16,29,31,32,34,51].
Although the mechanism for transient or reversible cognitive dysfunction associated with statins is not known, Engelberg [52] proposed that adverse psychological and cognitive effects result when low cholesterol in brain cell membranes lead to lower lipid microviscosity, which in turn, may affect neurotransmitter exposure resulting in decreased synaptic binding and uptake. This effect may become intensified when lower cholesterol levels are achieved through the use of lipophilic statins, which may also adversely affect neuronal function though immunomodulatory or inflammatory mechanisms [53]. Statins are also known to lower circulating fat-soluble micronutrients including vitamin E which may indirectly affect the integrity of neuronal membranes resulting in neurocognitive deficits [21]. Our results lend support for the proposed mechanism whereby lipophilic statins with certain pharmacokinetic properties may differentially impact brain cholesterol synthesis and contribute to neuronal dysfunction in susceptible individuals. As statins are increasingly prescribed to a broader spectrum of at risk patients, insights as to the effects of highly lipophilic statins on brain metabolism are clinically relevant to assessing overall risk/benefit ratio and monitoring for potential adverse effects.
We did not observe a signal of adverse cognitive effects for the lipophilic statins pitavastatin, lovastatin, and fluvastatin. In addition to these statins having much lower historical market penetration compared to atorvastatin and simvastatin, they differ in pharmacokinetic properties including being metabolized primarily through the CYP2C9 pathway with relatively high absorption (pitavastatin and fluvastatin), or having a relatively short half-life (lovastatin). At this time, it is unclear as to the specific pharmacokinetics of atorvastatin and simvastatin in addition to being highly lipophilic that may predispose to adverse cognitive effects.

Passive surveillance of adverse events
The AERS is a voluntary reporting system (passive surveillance). In our analysis, statin-induced adverse cognitive effects were significantly more likely to be reported to the FDA AERS database by consumers, both men and women, than by physicians or other health care providers. Warnings issued by the FDA are relevant to the practice community and consumers, yet the propensity to report adverse events may differ. From a survey of 650 community-based statin users studied to explore patient-physician interactions related to patient reported adverse effects of statins, Golomb and colleagues [22] found that of 113 respondents who reported adverse cognitive symptoms while on statins, only 65% reported their symptoms to their physician. Of these, nearly 50% of respondents reported that their physician denied any relationship between their symptoms and statin use. An additional 34% of respondents received no confirmation from their physician that their symptoms were, or were not, related to statin use. Golomb and coauthors [22] concluded that physicians may be missing less well-known adverse events such that post-marketing surveillance supported by patient-targeted reporting of adverse events may represent an important supplement to clinician reported adverse events, particularly for new or under-recognized symptoms.
In addition, it is noteworthy that voluntary reporting to AERS captures only a fraction of all true adverse events, estimated at between 1% to 10% [41,54]. In our analysis, a total of 2,566 reports of cognitive dysfunction were present for atorvastatin and simvastatin in AERS [ Table 3] over an approximate 8-year period. By extrapolation, and assuming a true causal relationship between these two statins and cognitive dysfunction, this translates to roughly 3,000 to 30,000 occurrences each year. The magnitude of these crude estimates provides a rationale for greater awareness and surveillance of potential statin-induced adverse cognitive effects. While clinicians recognize that all drugs have the potential to exert adverse effects, the occurrence of cognitive dysfunction, including memory impairment, cognitive disorder, and dementia represent particularly worrisome adverse effects that may significantly impact quality of life and thus impact treatment options and medication adherence.
To date, clinician awareness of the potential of statins to induce memory impairment and symptoms of dementia may be underrecognized as compared to the more widely recognized effects of statins on muscle tissue. Complicating the recognition of the potential detrimental cognitive effects of statins is paradoxical experimental [23,25,55] and clinical research [56,58,59] indicating significant benefits for statins in reducing the neuropathological features and severity of cognitive symptoms in Alzheimer's and vascular dementia. In many of these studies, investigators chose to test their hypotheses using atorvastatin, simvastatin and lovastatin based on the suspected pleiotropic and lipophilic properties. Thus, clinicians may be preemptively prescribing highly lipophilic statins for cognitive protection in high-risk patients.

Treatment selection
In 2009, 211 million prescriptions for statins were recorded which exceeds the estimated prevalence of cardiovascular, coronary artery and cerebrovascular disease combined [60,61] and reflects the extent to which statins are widely prescribed for primary prevention. In the clinical setting, the selection of a particular statin has been traditionally based on formulary constraints and the potency of a given statin to achieve targeted lipid lowering goals for individual patients. As there are no current recommendations for the selection of statins based on solubility properties or adverse event potential, our analysis provides emerging support for the clinical utility of matching individualized patient risk assessments with the relative solubility properties of the statin to reduce the likelihood of more severe cognitive adverse effects. A greater awareness of the effects of lipophilic versus hydrophilic statins may allow informed substitutions of better tolerated statins, thereby reducing adverse events and improving patient safety, adherence, and outcomes. This is consistent with a recent literature review that suggested that less lipophilic statins may be a logical alternative in cases where cognitive impairment that is secondary to a more lipophilic statin is suspected [32]. However, this review tended to minimize the risk of cognitive impairment with statin use by concluding it to be a rare occurrence, a conclusion at odds with the sheer number of adverse reports in the AERS database and known underestimation of true cases with this database. Going forward, epidemiological and prospective clinical research is needed to guide clinical decision-making in statin selection.
In addition, clinical decision making for statin use for the purpose of primary versus secondary prevention of cardiovascular disease has blurred over time. A Cochrane review [19] reported only limited evidence that primary prevention with statins may be cost effective and improve patient quality of life, and urged caution in prescribing statins for primary prevention among persons at low cardiovascular risk. Taking into account conventionally-recognized potential side effects of statins, absence of compelling primary prevention data in both children and adults (i.e. randomized controlled trials), and our results suggesting a greater occurrence of statin-induced cognitive dysfunction in adults (atorvastatin and simvastatin), we echo the need for great caution in the selection and use of statin therapy for children overall and adults at low cardiovascular risk.

Limitations
There are limitations to the FDA AERS database. AERS provides a tool for signal detection through the use of pharmacovigilance, which intends to search for unidentified patterns and automatically detect important signals [44]. A signal warrants further investigation but is not an "alert" which mandates wider communication or regulatory change. Statistical data mining tools, such as the PRR, allow the detection of signals that is not always possible using case series or reviews. The information from these types of studies identifies areas which are appropriate for additional study as well as continued surveillance. Causality assessments cannot be performed using the AERS database due to underreporting of events, variable reporting quality, reporting biases, and substantial missing data [62]. Moreover, numerator and denominator data are not available to provide incidence or risk estimates, and differences in magnitude of signal scores do not directly imply differences in risk [62].

Conclusions
To our knowledge, this is the first analysis of the AERS database to distinguish differential associations of lipophilic and hydrophilic statins on reported cognitive dysfunction. Our analysis identified potent, lipophilic statins (atorvastatin and simvastatin) as receiving a significantly greater proportion of reported adverse neurocognitive effects compared to other less potent or less lipophilic statins or hydrophilic solubility properties. These findings provide direction for future epidemiological and clinical research, as well as insights for informed treatment options, substitutions of more tolerable statins, and improved patient safety and adherence.