American Heart Association

epidemiology and genetics

Early Appearance of Spot Sign on CT Perfusion Associated with Hematoma Expansion and Poor Outcome in Small Retrospective Study


Intracerebral hemorrhage (ICH) causes a significant amount of stroke-related morbidity and mortality. Of the various prognostic factors in ICH, hematoma expansion (HE) is one of the few potentially modifiable ones and as such has been a topic of increasing research.
Unfortunately, large-scale randomized controlled trials aimed at preventing hematoma expansion have not shown robust results, possibly owing to the limited ability of clinicians to predict which patients are at greatest risk. The “spot sign,” a radiographic sign representing the leakage of contrast with a hematoma on CT scan has recently become a topic of extensive study with respect to its ability to predict hematoma expansion. As described previously, a recently published meta-analysis suggested that the sensitivity and positive predictive value of the spot sign was related to the time from ictus to scan acquisition and may not adequately predict HE when it is detected. Additionally, other studies have shown that using CT perfusion (CTP) improves the detection rate of the spot sign. Wang et al. sought to explore the relationship between spot sign characteristics on CTP (including number, timing, and maximum density) to evaluate the relationship between these characteristics and the risk of HE as well as clinical outcome.
The authors’ retrospectively reviewed a total of 83 patients from a prospectively collected database of consecutive patients with supratentorial SICH. Patients receiving surgical evacuation of their hematoma were excluded from the outcome analysis, and additionally were excluded from HE analysis if the intervention took place prior to the 24 h follow-up CT scan. Patients who died prior to the 24 h follow-up CT scan were excluded from the HE analysis.  Patients with secondary causes of ICH were also excluded. A total of eleven patients (6 with positive spot sign) were excluded from the HE analysis, and twelve patients (7 with positive spot sign) were excluded from outcome analysis due to receiving surgical evaluation. Excluded patients had higher SBP and DBP, over a three-fold greater median hematoma volume (52.30 mL vs 15.80 mL, P=0.029) and a nearly two-fold higher median NIHSS (21 vs 11, P=0.004). 

Baseline clinical variables included patient demographics, medical history, medications, onset to imaging time (OIT), baseline National Institute of Health Stroke Scale (NIHSS), and laboratory results. The clinical outcomes assessed were NIHSS at 24 h, in-hospital mortality, and modified Rankin Scale (mRS) including death at 3 months follow-up. Spot sign was assessed with CTP and the timing of occurrence (time from the start of scan to first detection of spot sign), total number of spots, maximum spot attenuation, and axial dimensions were recorded. The median time of onset to CP was 180 (120 to 240) minutes. Hematoma expansion was defined as an absolute ICH growth ≥ 6 mL or relative growth ≥ 33% as recorded on 24 h follow-up CT scan. 

The rate of spot sign was higher in patients with HE than those without (62.5% vs 12.5%, P<0.001). The presence of spot sign was independently associated with HE after correcting for APTT, glucose, NIHSS, baseline hematoma volume, and antiplatelet use. There was a trend for an association between the presence of spot sign and 3-month mortality (32% vs 8%, OR=5.649; 95% CI 0.913-34.954; P=0.063) after multivariate analysis. Spot sign was detected much earlier in patients with HE than those without (18.75 s vs 26.87 s, P=0.007). The timing of spot sign was significantly correlated with both absolute and relative ICH volume growth, and there was no association between the other spot sign characteristics and HE. The authors defined a subset of spot sign patients as having an “early occurring spot sign (EOSS),” defined as detection of the spot sign before 23.13 seconds. EOSS was found to have 0.67 sensitivity and 0.90 specificity for HE. On multivariate analysis, EOSS was an independent predictor of HE (OR=28.835; 95% CI, 6.960-119.458; P<0.001) and 3-month mortality (OR=22.377, 95% CI, 1.773-282.334; P=0.016). EOSS maintained a higher specificity for HE compared to spot sign (91% vs 74%).

In this single-center cohort of SICH patients, spot sign as assessed by CTP was associated with HE. The authors also found that early detection of the spot sign was the most important characteristic with respect to predicting HE and significantly correlated with ICH growth. The primary limitations of the study were related to the retrospective nature of the study, the relatively small sample size, and the use of a single center for recruitment and determining imaging parameters. External validation with a larger sample size and standardized imaging parameters will be necessary. In addition, the high rate of excluded surgical evacuation patients with positive spot sign may have led to an underestimation of the spot sign’s PPV with respect to mortality. It is also important to note that the predictive value and associations with spot sign on CTP in this study should not be applied patients evaluated with single-phase CT angiography. However, replicating this study with multi-phase CTA could be feasible and represents a potential future avenue of research.

Recent and Regular Use of Cocaine Significantly Increases Odds of IS in a Case-Control Study of Young Adults

Benjamin R. Kummer, MD

Cheng Y-C, Ryan KA, Qadwai SA, Shah J, Sparks MJ, Wozniak MA, et al. Cocaine Use and Risk of Ischemic Stroke in Young Adults. Stroke. 2016


Case series have reported biologically plausible associations between cocaine use and stroke in young patients, but few studies have rigorously investigated this relationship or the effects of drug ingestion timing or route of ingestion in large populations. Using a case-control design, Cheng and colleagues attempted to confirm the existence of the link between cocaine use and stroke in young patients, and delineate the relationship between timing and route of cocaine on the risk of ischemic stroke (IS). In this study, the authors drew their population from an existing, prospectively collected, case-control registry of 1,100 cases between the ages of 15 and 49 with first-ever IS, and approximately 1,100 controls. Cocaine use and timing/route were recorded on the basis of individual patient recollections (a small subset of patients had urine drug screening), as was confounding clinical and demographic data.
After adjusting for age, gender, ethnicity, smoking, hypertension, and alcohol use, patients that used cocaine in the 24 hours prior (OR 5.7, 95%CI 1.7-19.7) had significantly increased odds of IS compared to controls. However, the odds ratio for IS lost its statistically significance (OR 3.5, 95%CI 0.9-12.6) after removing higher-odds study periods. Patients that had smoked cocaine in the previous 24 hours had even higher adjusted odds of IS (OR 7.9, 95% 1.8-35.0), although no adjustment was made for smoking, hypertension, or alcohol use. Patients that reported smoking cocaine at any point also had slightly increased but not statistically significant adjusted odds of IS (OR 1.2, 95%CI 0.9-1.6). Patients that reported use of at least once in the past 1-30 days had a slightly increased, non-significant adjusted odds of IS (OR 1.1; 95% CI 0.7-1.9).

Overall, the number of patients exposed to cocaine use was very small relative to the sample size. Aside from the small number of observations, which caused effect estimates to be very wide, cases also had a greater proportion of patients with stroke risk factors (HTN, DM2, smoking, black race) than controls. The determination of drug use history could have been influenced by recall bias, and there may have also been a bias in under-reporting drug use by the study population (although this would have generated a conservative bias for the effect estimates). Also, while the logistic regression model was adjusted for some confounders, it was not adjusted for income level, education, employment, and insurance status, as well as important risk factors in young patients, such as hypercoagulable state, hyperlipidemia, family history of stroke, or obesity.
Despite its small number of observations, this study by Cheng and colleagues argues for a biologically plausible relationship between acute cocaine smoking and IS in young patients. It would have been interesting to see whether the association between acute cocaine smoking and IS disappeared with further adjustment for markers of health care access. These findings support urine toxicology screening for young patients with acute IS, although the implications for acute stroke management in such cases are unclear.

Intracranial Aneurysm Growth and Rupture Have Risk Factors in Common in a Large Meta-Analysis

Benjamin R. Kummer, MD

Backes D, Rinkel GJE,  Laban KG, Algra A, and Vergouwen MDI. Patient- and Aneurysm-Specific Risk Factors for Intracranial Aneurysm Growth: A Systematic Review and Meta-Analysis. Stroke. 2016

Three percent of the general population harbors an unruptured intracranial aneurysm (UIA). Due to their unfavorable interventional risk-benefit profile, small, asymptomatic UIAs are often left untreated and followed with serial imaging for growth—a well-established risk factor for aneurysm rupture. Because many factors associated with aneurysm growth are also associated with rupture, patients needing increased frequency of monitoring or even intervention might be identifiable by characteristics associated with aneurysm growth. Building on their previous work on aneurysm growth published in 2015 in Stroke, Baackes and colleagues further the aim of identifying patients at increased risk for rupture in this systemic review and meta-analysis of the clinical and aneurysm-specific factors associated with the growth of UIAs.

Approximately 4,000 patients and 5,000 unruptured intracranial aneurysms were studied, drawn from 18 separate publications and 15 unique patient cohorts (5 separate studies overlapped on 2 independent cohorts), followed over a mean of 2.8 years. The analysis only pooled studies that had clear definitions and effect estimates of aneurysm growth; however, a significant amount of epidemiologic and methodologic heterogeneity was present. Prospective and retrospective designs, studies with different population sizes, growth definitions, imaging modalities (CTA, DSA, MRA), Newcastle-Ottawa publication bias scores (mean 6, SD 1.9), as well as the numbers of growing UIAs were all analyzed together. Sensitivity analyses were performed to account for some, but not all of this heterogeneity. 

As previously shown, many factors associated with aneurysm rupture, like female sex, smoking, hypertension, larger aneurysm size, posterior circulation location and aneurysm shape, were also associated with aneurysm growth. However, the degree of heterogeneity in the effect estimates was moderate to substantial. The presence of multiple UIAs, which has not been readily identified as a factor associated with aneurysmal rupture, was shown to be significantly associated with aneurysm growth, raising the possibility that patients with multiple aneurysms might have a possibly connective-tissue based predilection towards more rapid growth than patients with single aneurysms, although it is important to note that this signal was also confounded by quite heterogeneous effect estimates. In contrast to their well-established association with aneurysmal rupture, family history of SAH and previous SAH were not significant factors for aneurysmal growth. Interestingly, UIA cohorts from Japan (where rates of aneurysmal rupture are higher than in European and North American cohorts) were had a lower risk of aneurysmal growth than North American cohorts. After stratifying effect estimates by study design and quality, the overall relative risk of aneurysm growth was lower in higher quality and prospective study designs than in lower-quality and retrospective designs, again suggesting that bias may have also contributed to the overall results.  

Despite the degree of heterogeneity, which the authors attempted to incompletely adjust for with sensitivity analyses, this meta-analysis suggests that aneurysmal growth and rupture may share a common pathophysiology, and yields some interesting hypothesis-generating findings on UIA growth and rupture.

Silent Brain Infarctions Are Associated With an Increased Risk of Future Stroke

Neal S. Parikh, MD



Citing the potential utility of silent brain infarcts (SBI) as both a possible stroke risk factor and a surrogate outcome in stroke trials, Dr. Gupta and colleagues performed a systematic review and meta-analysis to evaluate the precise association between SBI and subsequent stroke.

The authors selected studies of adults who had baseline MRI-ascertained SBI and at least 12 months of follow-up for clinical stroke. Studies were vetted and abstracted in a rigorous, pre-specified manner. Meta-analyses were performed with random effects modelling, which controls for occult heterogeneity between individual study samples. The risks of selection bias and publication bias were assessed and found to be low.

Thirteen studies were ultimately included. All studies defined SBI as T2 hyperintense lesions at least three millimeters in size, with various methods employed to distinguish SBI from leukoaraiosis and dilated perivascular spaces. Clinical ischemic stroke was typically defined as a neurological deficit lasting greater than 24 hours in the absence of hemorrhage. Not all studies distinguished between ischemic and hemorrhagic stroke.

A total of 14,764 subjects were included, and the mean follow-up for clinical stroke was 76 months. Most subjects were middle-aged or elderly. 3,007 (20%) had SBI on MRI. The crude relative risk of clinical stroke in patients with SBI was 2.94 (95% confidence interval (CI), 2.24-3.86). Eight of the 13 studies provided covariate-adjusted hazard ratios (HR); the aggregate adjusted HR was 2.08 (95% CI, 1.69-2.56).

Subgroup analyses found that the association between SBI and subsequent stroke was present in both stroke-free community-dwelling patients and in patients with prevalent stroke. In both groups, patients with SBI experienced a two-fold increased risk of subsequent stroke.

The major limitations of their study are: variable definitions of SBI as diagnosed on MRIs of variable magnetic field strength, variable inclusion of relevant covariates, inconsistent reporting of ischemic versus hemorrhagic stroke, non-tissue based definition of ischemic stroke, and outdated cohorts (e.g. not subject to contemporary maximum medical therapy of vascular risk factors). These limitations do not, on the whole, negate the study’s findings.

There appears to be a clear, positive association between SBI and subsequent stroke. In current clinical practice, incidental SBIs do not prompt thorough stroke mechanism evaluation or personalized secondary prevention. Our evolving understanding of this entity will surely impact our practice patterns.

MHTFR C677T Mutation Is Associated With an Increased Risk of Lacunar Stroke

Alexander E. Merkler, MD

Rutten-Jacobs LCA, Traylor M, Adib-Samii P, Thijs V, Sudlow C, Rothwell PM, et al. Association of MTHFR C677T Genotype With Ischemic Stroke Is Confined to Cerebral Small Vessel Disease Subtype. Stroke. 2016

Discovering modifiable risk factors of ischemic stroke is the future of stroke prevention. Hyperhomocysteinemia is associated with an increased risk of stroke and the most common MHTFR mutation, the MHTFR C677T mutation, leads to increased levels of homocysteine. However, previous studies evaluating the effect of lowering homocysteine levels with B vitamins have been negative in terms of reducing overall stroke risk. 

In the current manuscript, Dr. Rutten-Jacobs et al evaluate the association between the MHTFR C677T variant and small vessel disease. The authors evaluated the association between the MHTFR C677T mutation and both stroke subtype and white matter hyperintensity volume under the hypothesis that hyperhomocysteinemia only increases the risk of small vessel disease. In addition, as previous literature has shown an association between hypertension and homocysteine, the authors stratified patients by the presence of hypertension.

The authors identified 1359 cases of lacunar strokes, almost 4000 cases of large vessel and cardioembolic strokes and over 14 000 controls. The MHTFR C677T mutation was significantly associated with lacunar stroke and white matter hyperintensity volume, but not with other stroke subtypes. Furthermore, stratifying the lacunar strokes by hypertension status, confirmed the association the MHTFR C677T mutation with lacunar strokes in hypertensive, but not in normotensive patients.

The MHTFR C677T variant is associated with an elevated risk of specifically lacunar stroke and in patients who are hypertensive. The study adds further support that hyperhomocysteinemia is a risk factor for small vessel disease but not other subtypes of ischemic stroke. The manuscript should serve as a prime example of the importance of correctly choosing an outcome – perhaps B vitamins may successfully reduce stroke in patients with hyperhomocysteinemia when specifically lacunar stroke is chosen as the outcome of interest.

Validating the Montreal Cognitive Assessment for Stroke and TIA Patients

Neal S. Parikh, MD

Citing the high prevalence of cognitive impairment in patients with stroke, the authors call for better methods to easily detect cognitive impairment in this population. The authors aim to validate the Montreal Cognitive Assessment (MoCA) for classifying patients into three groups: low-, intermediate- and high-likelihood of moderate-severe cognitive impairment.


Three hundred and ninety patients referred to a stroke prevention clinic after stroke or TIA or other potentially cerebrovascular disease completed the MoCA. The mean age was 62 (range 17-94), and 53% were female. Patients were non-aphasic and English-speaking. Of the 390 patients, 34% had ischemic or hemorrhagic stroke, 34% had possible or probable TIA, and 32% had other vascular or non-vascular diagnoses (e.g. migraine). The gold standard for receiver operator characteristic/area under the curve analyses was an extensive neuropsychological battery.

The median MoCA score was 25 and did not differ significantly between diagnosis groups. By neuropsychological testing, 13% had moderate-severe cognitive impairment, and 30% had mild impairment.

Using a single cut-point for classification as moderate-severe impairment, test characteristics were optimal for MoCA <23 (sensitivity 60%, specificity 90%).


When two cut-points were assessed, the intermediate likelihood range was 23-27. In other words, a score of >27 reliably excluded patients with moderate-severe impairment (sensitivity 96%, negative predictive value 98%). A score of <23 reliably identified patients with moderate-severe impairment (specificity 90%, positive predictive value 49%). Of patients scoring in the intermediate likelihood range (23-27), 8% had moderate-severe impairment. The authors do not report the breakdown of the remaining 92% in terms of mild versus no cognitive impairment.


The MoCA may therefore help for identify patients unlikely to have cognitive impairment (e.g. “rule out”), but, as expected for a screening test, is not specific (e.g. not diagnostic). In their sample, 62% of patients fell into the intermediate group, and only 8% of these patients had moderate-severe cognitive impairment; this suggests that the intermediate range may not be a helpful category. However, for targeting assessments and interventions in clinical practice and research, the high specificity cut-point (MoCA<23) will certainly be of use.






Preexisting Heart Disease Underlies Newly Diagnosed Atrial Fibrillation After Acute Ischemic Stroke

Allison E. Arch, MD

Rizos T, Horstmann S, Dittgen F, Täger T, Jenetzky E, Heuschmann P, et al. Preexisting Heart Disease Underlies Newly Diagnosed Atrial Fibrillation After Acute Ischemic Stroke. Stroke. 2016

Atrial fibrillation (afib) is a risk factor for ischemic stroke. In patients who just had a stroke, a proportion of them will be diagnosed with afib in the post-stroke period. Does afib diagnosed immediately after a stroke carry the same risk for future stroke as afib that was already a known diagnosis? Is it associated with cardiovascular disease, or does the stroke itself cause a neurogenic arrhythmia that may not portend the same future stroke risk? These are the questions that Rizos and colleagues addressed in their single center observational study.

The authors included 1,397 stroke patients who were entered into a prospective database. In the 320 patients who had afib recorded in the hospital, 64% had a known diagnosis of afib. The remaining 36% had a new diagnosis of afib. Comparing those with known afib to those newly diagnosed, both groups had the same amounts of coronary artery disease, left atrial dilation, and low ejection fractions on echocardiogram. In other words, the newly diagnosed afib group had the same cardiovascular profile as the known afib group. Therefore, patients with new afib and stroke may just have afib due to underlying cardiovascular disease, and not from neurogenic strain on the heart.

This has implications for secondary stroke prevention. Whether or not atrial fibrillation was a known or a new diagnosis at the time of ischemic stroke, Rizos and colleagues show that both groups have the same underlying cardiovascular risk profile. Therefore, one may conclude that both groups should undergo the same stroke secondary prevention strategies.  

Impaired Higher-Level Functional Capacity Predicts Stroke Occurrence

Ilana Spokoyny, MD

Murakami K, Tsubota-Utsugi M, Satoh M, Asayama K, Inoue R, Ishiguro A, et al. Impaired Higher-Level Functional Capacity as a Predictor of Stroke in Community-Dwelling Older Adults: The Ohasama Study. Stroke. 2016

Does impairment of higher level functional capacity (intellectual, social, and instrumental activities of daily living such as cooking, cleaning, shopping) predict stroke in patients who are independent in basic ADLs? Should we probe as deeply into our patients’ higher level functional capacity as we do into their past medical history? The authors of this study address these important questions in a study of 1500 Japanese patients over age 60. These patients were independent for basic ADLs, and had not had a prior stroke.

After following the patients for about 10 years, the authors found that 191 of 1493 developed a first-time stroke. As expected, age played a role: those with impaired higher level functional capacity were older than those with normal capacity, and those who had strokes were older than those who did not. Overall TMIG-IC score (measurement of higher level functional capacity) was associated with significantly higher probability of developing stroke overall in multivariate analysis. The intellectual activity subscore was the only subscore which remained significant for the overall group. When the group was stratified by age, sex, and hypertension, the intellectual activity subscore was significantly associated with stroke in women only. Social role was significantly associated with stroke in those patients 75 years or older.



It is important to ask if the association is causative – or whether there are common risk factors for stroke as well as impairment in higher level functional capacity. In this study, the association persisted after adjusting for age and risk factors. The authors suggest that higher level functional impairment may represent early cerebrovascular disease, such as silent strokes. Determining the degree of white matter disease and correlating with higher level functional status would be the logical next step to follow this study. Determining baseline function, including higher level functional capacity, is useful and may help predict stroke. It would be interesting to compare the post-stroke level of independence of stroke patients with and without baseline higher level functional impairment. I suspect that those with higher level functional impairment, even if they were independent for basic ADLs, would have more significant post-stroke disability after adjusting for location and size of stroke. Lower cognitive reserve may portend a more protracted course and more difficulty rehabilitating from a stroke. Having data on a patient’s baseline status would be helpful in predicting stroke occurrence (shown in this study) and possibly predicting outcomes post-stroke.

Breakfast Intake Inversely Associated With Incident Stroke

Neal S. Parikh, MD

Kubota Y, Iso H, Sawada N, Tsugane S, The JPHC Study Group. Association of Breakfast Intake With Incident Stroke and Coronary Heart Disease: The Japan Public Health Center–Based Study. Stroke. 2016

The association between breakfast intake and good health is often touted in popular culture.

The authors cite studies demonstrating associations between skipping breakfast and multiple metabolic derangements and vascular risk factors including obesity, hypertension, dyslipidemia and glucose intolerance. Given these findings and the findings of a study of male US health professionals suggesting an inverse relationship between breakfast intake and coronary heart disease, the authors studied the association between breakfast intake and incident stroke and coronary heart disease.

The study was a population-based, prospective study using the Japan Public Health Center-Based (JPHC) study. Of 140,420 Japanese adults who were eligible, the study included 82,772 participants who had answered a breakfast intake questionnaire in the late 1990s and were free of prevalent stroke or coronary heart disease. Breakfast intake was defined as frequency of breakfast intake per week with daily breakfast as the reference. Covariates included age, sex, BMI, hypertension, hyperlipidemia, diabetes, blood pressure and cholesterol medications, smoking, exercise, sleep, stress, cohabitation, nature of employment, alcohol intake, caloric intake, and intake of vegetables, fruit, fish, soy, dairy, nuts, fat, fiber, and sodium. The outcomes were incident stroke (ischemic, subarachnoid hemorrhage, cerebral hemorrhage), MI, and sudden cardiac death.

The 82,772 participants provided 1,050,030 patient-years of follow-up. Participants who ate breakfast less frequently were less healthy and less likely to be on appropriate medication (e.g. for hypertension and hyperlipidemia). There were 4,642 cases of coronary heart disease and 3,772 strokes (including 1050 cerebral hemorrhages, 417 subarachnoid hemorrhages, 2,286 ischemic strokes).

After adjusting for potential confounders including dietary and lifestyle factors, total cardiovascular disease was associated with complete breakfast omission (HR, 1.14; CI 1.01-1.27) as was total stroke (HR 1.18, CI 1.04-1.34). An association was seen between breakfast omission and cerebral hemorrhage (HR, 1.36, CI 1.10-1.70) but not with coronary heart disease, subarachnoid hemorrhage, or cerebral infarction. The association between breakfast omission and stroke was only seen in non-users of anti-hypertensive medications.

The main limitation is that breakfast intake frequency was a time-fixed variable, which means that the methods did not account for the possibility of breakfast habits changing over time. Second, those omitting breakfast were less healthy, which raises the real possibility of residual confounding, especially as the confidence interval nearly crossed 1.0 for total cardiovascular disease.

Nonetheless, as the authors explain, there is a plausible causal pathway from breakfast omission to morning hypertension and cerebral hemorrhage. Additionally, to its merit, the study exhaustively controlled for vascular risk factors and dietary and lifestyle factors. This study suggests that failure to regularly eat breakfast is associated with cardiovascular disease, particularly hemorrhagic stroke. It may therefore indeed be healthful to eat breakfast daily.

Hemorrhagic stroke risk increased with certain NSAIDs: A meta-analysis of the data

Peggy Nguyen, MD

Ungprasert P, Matteson EL, and Thongprayoon C. Nonaspirin Nonsteroidal Anti-Inflammatory Drugs and Risk of Hemorrhagic Stroke: A Systematic Review and Meta-Analysis of Observational Studies. Stroke. 2016

The risk of cardiovascular events with NSAIDs has recently been publicized in the general media, with the FDA strengthening its warning on the association of NSAIDs with myocardial infarction and strokes in the past year. The relationship between NSAIDs and stroke risk, however, particularly hemorrhagic stroke, is ill-defined. The authors here performed a comprehensive meta-analysis reviewing the risk of hemorrhagic stroke in patients taking NSAIDs.

This meta-analysis identified 10 studies (7 case-control, 3 cohort studies) encompassing 1,489,120 patients. There was some heterogeneity across the studies, but most studies identified NSAID exposure as use of any NSAIDs until the index date or within 30 days prior to index date. Pertinently, there was a small, but not statistically significant, increased risk of hemorrhagic stroke (RR 1.09, 95% CI 0.98 – 1.22). Statistical analysis was also done for individual NSAIDs when data was available from a minimum of 3 studies, with a statistically significant increase for hemorrhagic stroke with diclofenac (RR 1.27, 95% CI 1.02 – 1.59) and meloxicam (RR 1.27, 95% CI 1.08 – 1.50). The highest risk estimate was seen in rofecoxib users, although it was not statistically significant (RR 1.35, 95% CI 0.88 – 2.06). The statistical significance was not significantly changed with sensitivity analyses, although the relative risk was increased for diclofenac and meloxicam in one analysis to 1.46 and 1.48, respectively.


As the authors point out, the results of the meta-analysis should be interpreted with caution. The lack of a significant association with NSAIDs as a whole may be a function of the reversibility of NSAID inhibition on COX-1. Selection bias may exist; due to the cardiovascular warning associated with NSAIDs, the exposed group may in fact be healthier than the non-exposed group, in which case, the risk of hemorrhagic stroke may actually be underestimated. Nevertheless, the results are at least suggestive that in patients who are at increased risk for hemorrhagic stroke, for example, a patient with amyloid angiopathy, it is not unreasonable that these medications be avoided.