Cumulative incidence plots of absolute risk of dementia after injury. The x-axis shows time after injury, spanning 5 years. The y-axis shows cumulative incidence of dementia. Numerical counts indicate the number of patients alive without dementia at the corresponding time. Results show increasing incidence of dementia with time, particularly after a concussion, with reduction in risk associated with statin use. The patients with ankle sprain show minimal healthy-user bias.
Cumulative incidence plots of absolute risk of depression after injury. The x-axis shows time after injury, spanning 5 years. The y-axis shows cumulative incidence of depression. Numerical counts indicate the number of patients alive without depression at the corresponding time. Results show increasing incidence of depression with time, modest accentuation after a concussion, and no reduction in risk associated with statin use. Both patient groups show minimal healthy user bias.
eAppendix. Technical Appendix
Customize your JAMA Network experience by selecting one or more topics from the list below.
Redelmeier DA, Manzoor F, Thiruchelvam D. Association Between Statin Use and Risk of Dementia After a Concussion. JAMA Neurol. 2019;76(8):887–896. doi:10.1001/jamaneurol.2019.1148
Is statin use associated with an increased or decreased risk of subsequent dementia after a concussion?
In this large extended population-based double cohort study following 28 815 patients after a concussion, the 5-year incidence of dementia was substantial and statin use was associated with a significantly reduced risk of subsequent dementia.
Concussions are associated with an increased long-term risk of dementia, which is modestly reduced for patients receiving a statin.
Concussions are an acute injury that may lead to chronic disability, while statin use might improve neurologic recovery.
To test whether statin use is associated with an increased or decreased risk of subsequent dementia after a concussion.
Design, Setting, and Participants
Large extended population-based double cohort study in Ontario, Canada, from April 1, 1993, to April 1, 2013 (enrollment), and continued until March 31, 2016 (follow-up). Dates of analysis were April 28, 2014, through March 21, 2019. Participants were older adults diagnosed as having a concussion, excluding severe cases resulting in hospitalization, individuals with a prior diagnosis of dementia or delirium, and those who died within 90 days.
Statin prescription within 90 days after a concussion.
Main Outcome and Measure
Long-term incidence of dementia.
This study identified 28 815 patients diagnosed as having a concussion (median age, 76 years; 61.3% female), of whom 7058 (24.5%) received a statin, and 21 757 (75.5%) did not receive a statin. A total of 4727 patients subsequently developed dementia over a mean follow-up of 3.9 years, equal to an incidence of 1 case per 6 patients. Patients who received a statin had a 13% reduced risk of dementia compared with patients who did not receive a statin (relative risk, 0.87; 95% CI, 0.81-0.93; P < .001). The decreased risk of dementia associated with statin use applied to diverse patient groups, remained independent of other cardiovascular medication use, intensified over time, was distinct from the risk of subsequent depression, and was not observed in patients after an ankle sprain.
Conclusions and Relevance
In this study, older adults had a substantial long-term risk of dementia after a concussion, which was associated with a modest reduction among patients receiving a statin.
Concussions are a common cause of brain injury occurring in more than 1 million Americans each year and disproportionately involving older adults.1-4 The subacute consequences vary widely and include fatigue, headache, irritability, insomnia, inattention, photophobia, vertigo, and cognitive difficulties.5-9 Most patients recover from a concussion within weeks, although some can develop lingering mood disorders or chronic neuropsychiatric disorders.10-14 The extent of complications after a concussion is uncertain, and effective pharmacologic treatments remain elusive.15-17 Unfortunately, many medical treatments for traumatic brain injury that showed promise in animal models have subsequently failed in human clinical trials.18-22
Hydroxymethylglutaryl coenzyme A reductase inhibitors (statins) are a class of medications prescribed for the treatment of hyperlipidemia.23,24 Preclinical data suggest that statin use might mitigate injury-related brain edema, oxidative stress, amyloid protein aggregation, and neuroinflammation.25-29 The potential neuroprotective benefits from statins have also been speculated and include preserved cerebral blood flow, leading to decreased risks of Alzheimer disease, vascular dementia, and age-related cognitive decline.30-35 Together, these findings suggest that statin use could contribute to microvascular homeostasis and immune modulation independent of systemic lipid levels. However, statins do not improve cognition for patients already diagnosed as having dementia.36-38
The role of statins in recovery after a concussion has rarely been investigated.39,40 Furthermore, small sample sizes, selective enrollment, and brief follow-up have limited past studies41-43 assessing the long-term consequences of concussion. On the one hand, the potential neuroprotective associations of statin use may prevent subsequent dysfunction by neuron preservation and neural stem cell activation.44-47 On the other hand, the neurohazardous associations of statin use might contribute to memory difficulties from altered neurophysiology.48-54 We conducted a large extended population-based double cohort study using linked databases to test whether statin use is associated with an increased or decreased risk of dementia in older adults after a concussion.
This population-based multicenter double cohort study of older adults diagnosed as having a concussion throughout Ontario, Canada, was performed from April 1, 1993, to April 1, 2013 (enrollment period of 20 years), providing a minimum follow-up for survivors of 3 years and reflecting all data available. Ontario is Canada’s largest province, with a population of 12 407 300 in 2004 (the study midpoint), an annual incidence of dementia was 19 cases per 1000 patients for adults 65 years and older, and societal costs of dementia estimated at $18 440 per patient-year.55-58 During our study, universal health insurance covered outpatient medical care for all individuals, with no out-of-pocket costs to patients.59 In addition, the Ontario Drug Benefit Program covered prescription medications for all patients 65 years and older.60 The study protocol was approved by the Research Ethics Board of Sunnybrook Health Sciences Centre, including a waiver for direct patient consent. All data are available through the Institute for Clinical Evaluative Sciences in Ontario.
We identified patients 66 years and older diagnosed as having a concussion by assessing physician billing data using the concussion diagnostic criterion (International Classification of Diseases, Ninth Revision [ICD-9] code 850) from the Ontario Health Insurance Plan.61 This code for a concussion diagnosis has been validated with high specificity (99%) and moderate sensitivity (46%-76%).62,63 Patients who were admitted to a hospital within 2 days of a concussion or who survived less than 90 days were excluded to reduce confounding from severe brain injury.64 Patients with a history of dementia or delirium in the prior 5 years were also excluded to reduce confounding from past neuropsychiatric conditions.65 No patients were excluded otherwise. Patients with more than 1 concussion during the study were counted once based on the first incident.
The prescription of a statin was identified through the Ontario Drug Benefit Program database, which has an accuracy exceeding 99% in this setting.66 The specific statins included were atorvastatin, cerivastatin, fluvastatin, lovastatin, pitavastatin, pravastatin, rosuvastatin, and simvastatin. We classified patients based on the specific statin received immediately after their concussion. The primary analysis compared patients who had received a statin prescription in the 90 days immediately after a concussion with patients who had not. Secondary analyses explored statin prescriptions in the 90 days before the concussion and also considered statins as time-dependent exposures in analyses to account for fluctuating prescriptions over time.67
Additional baseline characteristics were defined on the day of the concussion and obtained by computerized linkage to health care records.68 A demographic registry was used to determine the patient’s age, sex, socioeconomic quintile, and home location.69,70 A physician services databases provided data on clinic visits, emergency department contacts, and hospitalizations in the prior year.60,71 The Ontario Drug Benefit Program database provided data on additional cardiovascular medications, neuropsychiatric medications, and miscellaneous medications.72-74 The available databases contained no information on smoking status, daily exercise, education level, family history, genetic factors, hearing loss, social isolation, or other factors that can alter dementia risk.75,76
The primary study outcome was a physician diagnosis of dementia (ICD-9 codes 290, 331, and 797) ascertained through the validated Ontario Health Insurance Plan database, as established in past research.77 These codes had a specificity of about 99% and a sensitivity of about 20% for dementia.78 To avoid false-positive results, we required a dementia diagnosis on 2 separate dates.79 Therefore, this outcome definition was highly specific and provided a conservative estimate of the incidence of dementia after a concussion. To corroborate this outcome, we also identified more extensive documentation of the diagnosis that could signify a worsening course over 3 years of follow-up. The available codes did not distinguish specific conditions underlying the dementia diagnosis.
To test the importance of a potential association between statin use and subsequent dementia, we replicated our entire selection strategy and analysis, focusing instead on older adults diagnosed as having an ankle sprain rather than a concussion. The objective of this secondary parallel analysis was to distinguish the long-term prognosis for patients who had an acute neurologic injury (concussion cohort) from patients who had a peripheral orthopedic injury (ankle sprain). We then assessed the long-term incidence of dementia for patients who received a statin after an ankle sprain compared with patients who did not receive a statin after an ankle sprain. If a patient experienced both a concussion and an ankle sprain, the individual was included in both cohorts (the exclusion of overlap patients yielded similar results).
Our primary analysis evaluated the incidence of dementia after a concussion, comparing patients who received a statin with patients who did not. Graphical displays were created using cumulative incidence curves. Statistical testing was based on proportional hazards analysis taking into account censoring for interval deaths and the study follow-up end date of March 31, 2016. Statistical testing examined associations before adjusting for measured baseline patient characteristics (basic analysis) and after adjusting for measured baseline patient characteristics (adjusted analysis) to check the robustness of relative risk estimates. All P values were 2 tailed, and .05 was the threshold for statistical significance.
Additional analyses explored the potential role of statins in preventing dementia after a concussion. Lipophilic statins (atorvastatin, cerivastatin, fluvastatin, lovastatin, pitavastatin, and simvastatin) were compared with hydrophilic statins (pravastatin and rosuvastatin).80 Higher dosages of statins (maximum accepted treatment dosage for a specific statin) were compared with lower dosages of statins (all smaller dosages for a specific statin).81 We also evaluated the incidence of dementia among patients receiving the most commonly used statins separately. Patients who received a statin before and after the concussion (continuous use) were also distinguished from patients who received their statin only after the concussion (initiation) and patients who stopped their statin after the concussion (discontinuation).
We conducted 3 further statistical analyses to explore the robustness of the primary analysis. The first analysis applied 1:1 propensity score matching of patients receiving statins with control patients to account for possible imbalances in baseline characteristics and unmeasured indications for statins. The second analysis introduced time-dependent covariates to account for subsequent changes in statin prescriptions over time and fluctuating adherence to statins. The third analysis considered the possible competing risks of other causes of death that might obscure a subsequent diagnosis of dementia.82 Each of the 3 statistical analyses was conducted before and after adjusting for measured baseline patient characteristics to check the robustness of relative risk estimates.
In addition, we used tracer analysis to check for confounding by replacing the dementia outcome with an alternative clinical end point. Specifically, we reasoned that subsequent depression, instead of dementia, is a different adverse neuropsychiatric outcome that is frequent, serious, important, and similar in some shared clinical features to a dementia diagnosis in older patients.83 However, the risk of depression is not substantially reduced by statins (due to distinct pathophysiology) but can be associated with shared determinants (eg, alcohol use and physical inactivity).84-86 Therefore, we repeated the analyses and examined the risk of depression rather than the risk of dementia after a concussion.
A total of 28 815 patients (median age, 76 years; 61.3% female) were diagnosed as having a concussion during the study. Overall, 7058 patients (24.5%) received a statin during the 90 days after a concussion, and 21 757 patients (75.5%) did not receive a statin. The distribution of baseline demographic characteristics was similar for both groups (Table 1). The typical patient receiving a statin was a 76-year-old woman who was also taking additional medications. On average, patients receiving statins had more cardiovascular medications and more prior physician visits than control patients. The frequency of neuropsychiatric medication use also tended to be higher (not lower) for patients receiving statins. About one-quarter (23.6%) of patients had a hospital admission in the prior year, with no imbalance between the 2 groups.
A total of 4727 patients developed dementia over a mean follow-up of 3.9 years after a concussion (Figure 1A). The absolute incidence was 1 case per 6 patients, and most (83.4%) had extensive documentation of dementia during follow-up. Patients receiving statins accounted for 1050 dementia cases over 28 129 patient-years (mean, 4.0 years), equal to an incidence of 37 cases per 1000 patients annually (twice the population norm).87 Control patients accounted for 3677 dementia cases over 85 339 patient-years (mean, 3.9 years), equal to an incidence of 43 cases per 1000 patients annually (more than twice the population norm). Together, statin use was associated with a 13% (95% CI, 7%-19%; P < .001) reduced risk of dementia compared with patients who did not receive a statin (relative risk, 0.87; 95% CI, 0.81-0.93; P < .001), equal to a number needed to treat of about 50 patients.
The decreased risk of dementia associated with statin use after a concussion was similar in the first half and second half of the cohort and persisted after adjusting for patient characteristics (Table 2). As expected, older age was associated with an increased risk of dementia, as was greater health care use (total prescriptions, physician visits, and prior hospitalizations). Patient sex was not associated with a consistent difference in dementia risk. Lower socioeconomic status and urban home location were associated with an increased risk of dementia. Adjustment for all measured baseline characteristics suggested that statin use was associated with a 16% (95% CI, 10%-22%; P < .001) reduction in the risk of subsequent dementia, equal to an E-value88 of 1.67.
Analysis of additional medication use suggested that the decreased risk of subsequent dementia was specific to statins. In particular, other lipid-lowering medications were not associated with a significant difference in the risk of dementia (Table 2). Other cardiovascular medications were not associated with a consistent decrease in the risk of dementia, with the possible exception of angiotensin II receptor blockers. Similarly, benzodiazepines, thyroid supplements, gastric acid suppressors, inhaled bronchodilators, and glaucoma eyedrops were not associated with a significantly decreased risk of subsequent dementia. As expected, major neuropsychiatric medications were associated with an increased risk of dementia, perhaps as a proxy for cognitive frailty in older patients.
Secondary analyses explored further nuances of statin use and the risk of subsequent dementia. Rosuvastatin use was associated with the largest risk reduction, and simvastatin use was associated with the smallest risk reduction (Table 3). Hydrophilic statins were marginally more beneficial than lipophilic statins. No greater benefit was found with higher dosages compared with lower dosages. Propensity score matching suggested that the risk reduction was not easily explained by baseline imbalances in measured patient characteristics. Patients who received a statin before and after the concussion explained most of the risk reduction. Those who initiated a statin after the concussion showed a significant risk reduction, and those who discontinued a statin after the concussion showed no significant risk reduction.
The parallel analysis identified a total of 307 890 patients diagnosed as having an ankle sprain, of whom 77 898 (25.3%) received a statin and 229 992 (74.7%) did not receive a statin. A total of 25 956 patients developed dementia over a mean follow-up of 4.3 years (Figure 1B). Patients receiving statins accounted for 6239 dementia cases over 336 251 patient-years (mean, 4.3 years), equal to an incidence of 19 cases per 1000 patients annually. Control patients accounted for 19 717 dementia cases over 1 001 606 patient-years (mean, 4.4 years), equal to an incidence of 20 cases per 1000 patients annually. Both groups were at the population norm, and statin use was associated with a 5% (95% CI, 3%-8%; P < .001) reduction in the risk of dementia, equal to a number needed to treat of about 220 patients.
The tracer analysis identified 1778 patients who were subsequently diagnosed as having depression after a concussion. The absolute incidence of depression was about 1 case per 16 patients (Figure 2). Patients receiving statins accounted for 440 cases of depression over 29 007 patient-years of follow-up, equal to an incidence of 15 cases per 1000 patients annually. Control patients accounted for 1338 cases of depression over 88 540 patient-years of follow-up, equal to an incidence of 15 cases per 1000 patients annually. Together, statin use was associated with an insignificant 4% (95% CI, −7% to 16%; P = .43) increased risk of depression before adjustment for measured baseline characteristics and an insignificant 4% (95% CI, −8% to 14%; P = .49) decreased risk of depression after adjustment for measured baseline characteristics.
We studied 28 815 older adults diagnosed as having a concussion to test whether statin use might influence a patient’s recovery after the concussion. In patients receiving statins, we found that the subsequent incidence of dementia was twice the population norm, and it was further accentuated in control patients who were not taking a statin. The relative reduction in dementia risk associated with statin use after a concussion was greatest for those taking rosuvastatin, was consistent for those receiving lower dosages, was accentuated after adjustments for measured patient characteristics, and was distinct from the risks for patients after an ankle sprain. No other cardiovascular or noncardiovascular medications were associated with a decreased risk of dementia after a concussion (with the possible exception of angiotensin II receptor blockers).
Our study adds to prior research on statin use after traumatic brain injury due to a larger sample size, longer follow-up, more detailed statistical analysis, and a priority on concussions (eAppendix in the Supplement). Four prior randomized trials yielded conflicting results, with 2 studies89,90 reporting a positive protective effect of statins on neurocognitive outcomes and the other 2 studies91,92 reporting no significant effect. Eleven prior nonrandomized studies also yielded conflicting results, with 5 studies93-97 reporting a positive protective association on clinical outcomes and the other 6 studies98-103 reporting a negligible association. No prior studies indicated a detrimental influence of statin use after traumatic brain injury, but almost all previous studies focused on patients with moderate to severe injuries rather than concussions.
Several limitations of our research merit attention. Our study was not a randomized trial, and the observed associations might reflect confounding due to earlier indications for statin use.104 Important missing covariates included smoking status, daily exercise, drug adherence, and other factors that influence the risk of developing dementia.105-108 These unknown differences in patients (healthy-user bias) might account for a reduction in the subsequent risk of dementia.109,110 Our study also lacks sufficient power to disentangle whether statins make a contribution before, during, or after a concussion. In addition, the diagnostic codes for concussion and dementia were not fully sensitive and may significantly underestimate the true incidence of dementia in patients after a concussion.111
The generalizability of our findings is also limited by a focus on older adults, the requirement for patients to survive at least 90 days after a concussion, and the exclusion of those already diagnosed as having dementia.112 We did not consider patients with severe brain injury that resulted in hospitalization; instead, we examined only concussion cases as the more common type of traumatic brain injury. We also lacked data on patients with injuries who did not seek medical care. Our study was based in Canada, did not account for geographic variances in health care delivery, and potentially underestimated risks due to random delays in diagnosing dementia. We also lacked nuanced data on functional status, and we found only a modest degree of healthy-user bias in our patients.113,114
Our study has further limitations for clinical care. Concussion and dementia are both varied disorders, so aggregate statistics do not necessarily apply to unique patients. The median follow-up duration was less than 5 years, whereas the course of dementia can span decades of subclinical changes. Our patients had years of unrecorded history and an unknown total number of concussions over a lifetime. The relative risks associated with concussions and with statin use were both modest because each addresses only one of many contributors to dementia. Our observational design also means that the unmeasured burden of cardiovascular and neuropsychiatric diseases may have been imbalanced against the patients receiving statins and may have led to analyses that underestimate the neuroprotection from statins.
The findings of our study suggest a potential long-term protective association between statin use and the risk of dementia after a concussion that justifies future research. The study also provides estimates of event rates, time profiles, effect sizes, and baseline frequencies needed for planning future trials. Of course, a randomized trial might face difficulties in patient enrollment because adults may not be willing to be randomized to receive a placebo rather than a statin.92 The 20-year patient enrollment interval of our study could also be prohibitive for a prospective trial. In addition, we know of no practical method to randomize patients to receive a statin immediately before a concussion. Therefore, analytic observational research may provide the best available data for the immediate future.
Concussion is often popularized as a problem in athletic youth and tends to be underdiagnosed in older individuals.115-117 The results of our study suggest that concussions are a common injury in older adults and indicate that dementia may be a frequent outcome years afterward. Therefore, more efforts to prevent concussions should be encouraged at all ages.118 Screening for past concussions might also offer new clinical insights for patients diagnosed as having dementia.119 A potential neuroprotective benefit may also encourage greater medication adherence for patients who are already prescribed a statin.120,121 In addition, a concussion should not be interpreted as a reason to stop statins, and a future randomized trial is justified.122 The long-term neurologic consequences of a concussion are substantial and merit attention.123
Accepted for Publication: January 10, 2019.
Corresponding Author: Donald A. Redelmeier, MD, MSHSR, Division of General Internal Medicine, Sunnybrook Health Sciences Centre, 2075 Bayview Ave, Ste G-151, Toronto, ON M4N 3M5, Canada (firstname.lastname@example.org).
Published Online: May 20, 2019. doi:10.1001/jamaneurol.2019.1148
Author Contributions: Dr Redelmeier had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Redelmeier, Manzoor.
Acquisition, analysis, or interpretation of data: Redelmeier, Thiruchelvam.
Drafting of the manuscript: Redelmeier, Manzoor.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Redelmeier, Thiruchelvam.
Obtained funding: Redelmeier.
Administrative, technical, or material support: Redelmeier, Manzoor.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by a Canada Research Chair in Medical Decision Sciences, the Canadian Institutes of Health Research, the BrightFocus Foundation, and the Comprehensive Research Experience for Medical Students at the University of Toronto.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The views are those of the authors and do not necessarily reflect the views of the Ontario Ministry of Health and Long-Term Care.
Additional Contributions: The following individuals at the University of Toronto provided helpful comments: Sandra Black, MD, Anthony Feinstein, MD, Michael Fralick, MD, Andrea Gershon, MD, Taaha Muhammad, BHSc, Sheharyar Raza, MD, Charles Tator, MD, Jason Woodfine, MD, Christopher Yarnell, MD, Jonathan Zipursky, MD, and Jeremy Zung, MD. None received compensation.