Elevated International Normalized Ratio is Associated with Ruptured Aneurysms

Abstract

Background and Purpose:

The effects of anticoagulation therapy and elevated international normalized ratio (INR) values on the risk of aneurysmal subarachnoid hemorrhage are unknown. We aimed to investigate the association between anticoagulation therapy, elevated INR values, and rupture of intracranial aneurysms.

Methods:

We conducted a case-control study of 4,696 patients with 6,403 intracranial aneurysms, including 1,198 prospective patients, diagnosed at the Massachusetts General Hospital and the Brigham and Women’s Hospital between 1990 and 2016 who were on no anticoagulant therapy or on warfarin for anticoagulation. Patients were divided into ruptured and non-ruptured groups. Univariable and multivariable logistic regression analyses were performed to evaluate the association of anticoagulation therapy, INR values, and presentation with a ruptured intracranial aneurysm, taking into account the interaction between anticoagulant use and INR. Inverse probability weighting using propensity scores was used to minimize differences in baseline demographics characteristics. The marginal effects of anticoagulant use on rupture risk stratified by INR values were calculated.

Results:

In unweighted and weighted multivariable analyses, elevated INR values were significantly associated with rupture status among patients who were not anticoagulated (unweighted OR 22.78, 95% CI 10.85–47.81 and weighted OR 28.16, 95% CI 12.44–63.77). In anticoagulated patients, warfarin use interacts significantly with INR when INR≥1.2 by decreasing the effects of INR on rupture risk.

Conclusions:

INR elevation is associated with intracranial aneurysm rupture but the effects may be moderated by warfarin. INR values should therefore be taken into consideration when counseling patients with intracranial aneurysms.

Introduction

Due to the increasing age of the general population and significant improvements in imaging techniques, the number of patients with unruptured and often incidental intracranial aneurysms is increasing.¹ Management of these patients is controversial and one of the major challenges facing neurosurgeons today, since the case fatality rate of aneurysmal subarachnoid hemorrhage (aSAH) remains high.^1–4 Although prevention of rupture with surgical or endovascular treatment is considered to be the most effective approach, treatment related risks and complications remain a significant problem.¹ In order to make a calculated assessment regarding the lifetime risk of rupture compared with the risk of treatment, it is of paramount importance to understand the natural history of unruptured intracranial aneurysms and its related risk factors. Although several aneurysm and patient related risk factors for aSAH have been identified, such as hypertension and smoking, the natural history of unruptured intracranial aneurysms has yet to be clearly defined.^5–8

As the use of anticoagulation therapy for atrial fibrillation, mechanical heart valves, and other indications is rapidly increasing with an aging population, neurosurgeons should be aware of the concomitant risks of this treatment. Recent population based studies have reported contradictory findings regarding the association between anticoagulant therapy and SAH risk.^9–12 However, these reports did not include or specify the number of aneurysmal SAH, and were limited due to a low number of patients on anticoagulant therapy¹², lack of inclusion of important confounding factors such as smoking and alcohol use⁹, the use of standardized diagnosis codes with possible misclassification and lower diagnostic validity as a result, and the use of control groups consisting of the general population without intracranial aneurysms. Moreover, only one study included INR values, albeit limited to patients on anticoagulation therapy.¹⁰ Given the widespread use of anticoagulant therapy and the potential fatal consequences of aSAH, we conducted a large case-control study investigating the magnitude and direction of the association between anticoagulation use, INR values, and the risk of aneurysmal subarachnoid hemorrhage.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request. The medical records of 4,696 patients who were diagnosed with an intracranial aneurysm between 1990 and 2016 at the Brigham and Women’s Hospital (BWH) and Massachusetts General Hospital (MGH) and who were either on no anticoagulant therapy or on warfarin at the time of diagnosis were reviewed. This study has been approved by our Institutional Review Board and considered minimal risk. Patient consent was therefore waived by the board. Patients were identified prospectively on clinical presentation (2007–2016) and retrospectively using natural language processing (NLP) in conjunction with the Partners Healthcare Research Patients Data Registry (RPDR) which includes 4.2 million patients who have received care from BWH and MGH (1990–2013).¹³ ICD-9 and CPT codes were used to obtain an initial set of potential aneurysm patients from the RPDR, and NLP was then used together with codified features to train a classification algorithm which classified 5,589 patients as having aneurysm with a positive predictive value of 0.91. Of the 5,589 patients, 727 were also seen on clinical presentation from 2007–2013 with prospectively collected data. 474 additional patients who were seen on clinical presentation from 2013–2016 were included with prospectively collected data. The medical records and imaging studies of the 6,063 patients were reviewed in detail (AC and RD) to identify 4,696 patients with definite saccular aneurysms and who were on no anticoagulant therapy or on warfarin. Patients on low-molecular-weight heparin were excluded. Of the 4,696 patients 1,198 were obtained prospectively. The results of the imaging studies, including intracranial aneurysm site and size, were recorded. Patients with possible infundibula or non-definitive diagnoses of aneurysms, feeding artery aneurysms associated with arteriovenous malformations, fusiform or dissecting aneurysms, and those lacking clinical notes or radiographic images were excluded from the present study. In addition, patients who received treatment of their aneurysm(s) prior to presentation were also excluded from the present study. Patients who presented with an aneurysmal subarachnoid hemorrhage were categorized as harboring a ruptured aneurysm.

Patient demographics (age, sex, race), comorbidities (hypertension, coronary artery disease (CAD), myocardial infarction (MI), atrial fibrillation (AF)), number and maximum size of intracranial aneurysms, family history of aneurysms or family history of SAH, and current tobacco and alcohol use were obtained. The diagnosis of aSAH was confirmed with a computed tomographic (CT) scan, cerebrospinal fluid analysis, or intraoperatively by a neurosurgeon. In addition, we collected detailed data regarding anticoagulant use, type of anticoagulant, indication, and INR values at the time of diagnosis of the intracranial aneurysms. A risk factor was assumed to be absent if we found no documentation of its presence. INR values were obtained at the time of diagnosis for ruptured aneurysms and within a year of diagnosis for unruptured aneurysms. For patients whose initial INR at the time of diagnosis were obtained while heparinized and falsely elevated, the next available INR after the partial thromboplastin time (PTT) had normalized was used. Patients were classified as anticoagulant users if they were on anticoagulants at the time of diagnosis. Four patients had been on anticoagulants prior to diagnosis and had discontinued their anticoagulation at least 2 months prior to diagnosis. These patients were classified as non-anticoagulant users. We obtained clinical notes with anticoagulation details by using the following search terms: anticoagulate, anticoagulated, anticoagulation, anti-coagulation, anti-coagulant, anticoagulant, anticoagulants, antithrombotic, anti-thrombotic, coumadin, coumarin, warfarin, heparin, molecular, vitamin K antagonist, unfractioned, argatroban, acova, dalteparin, fragmin, bivalirudin, angiomax, angiox, lepirudin, refludan, fondaparinux, arixtra, idraparinux, acenocoumarol, lovenox, enoxaparin, aggrastat, tirofiban, integrilin, and eptifibatide. These clinical notes were subsequently manually reviewed. Patients who were on low molecular weight heparin were excluded as that does not affect the INR values.

Differences in baseline characteristics between patients with anticoagulant therapy and without anticoagulant therapy were evaluated using t-tests for continuous variables and Pearson’s chi-square test for categorical variables. Univariable and multivariable logistic regression models were implemented to test for effects due to anticoagulant use and INR values, with a backward elimination procedure to identify significant confounders. Cut-off p-values of 0.1 were used to select the initial set of variables to be included in the initial multivariable model for backward elimination, with the exception of current anticoagulant use since this was one of the variables of interest. Adjusted odds ratios (OR) with 95% confidence intervals (CI’s) were calculated, and p< 0.05 was considered significant. Interaction terms between INR and anticoagulant use were included. Marginal effects of anticoagulant use on rupture risk stratified by INR values were calculated. In order to control for differences in baseline characteristics, inverse probability weighting using propensity scores was applied. Covariate balance was tested after inverse probability weighting using a chi-squared test. Missing values were accounted for by using multiple imputation with chained equations and inferential statistics were obtained from 40 imputed datasets. Sensitivity analyses using subgroups consisting of complete cases (i.e. without missing data) only, anticoagulation users only, and non-anticoagulation users only were also performed. All statistical analyses were performed using the Stata statistical software package (version 14, StataCorp. College Station,TX).

Results

Summary statistics of patient demographics and characteristics stratified by anticoagulant use are shown in Table 1. A total of 4,696 patients with 6,403 aneurysms were included, of which 1,300 (27.7%) patients presented with a ruptured aneurysm. In general, patients on anticoagulant therapy were significantly older, less frequently current smokers, and more frequently diagnosed with hypertension, coronary artery disease, myocardial infarction, and atrial fibrillation. In addition, INR values were significantly higher in anticoagulated patients.

Table 1:

Patient characteristics stratified by anticoagulant use.

Variables	All N=4,696	Missing	Anticoagulant group N=131*	Non-anticoagulant group N=4,565	P
Ruptured aneurysm (%)	1300 (27.7)	0	27 (20.6)	1273 (27.9)	0.07
Female (%)	3662 (78.0)	0	96 (73.3)	3566 (78.1)	0.19
White race (%)	3734 (79.5)	0	108 (82.4)	3628 (79.4)	0.40
Black race (%)	290 (6.2)	0	6 (4.6)	284 (6.2)	0.45
Hispanic race (%)	270 (5.7)	0	5 (3.8)	265 (5.8)	0.33
Asian race (%)	107 (2.3)	0	1 (0.8)	106 (2.3)	0.25
Other/unknown race (%)	295 (6.3)	0	11 (8.4)	284 (6.2)	0.31
Age at diagnosis (SD)	55.6 (13.7)	0	65.8 (12.5)	55.3 (13.6)	<0.01
Hypertension (%)	2150 (45.8)	0	74 (56.5)	2076 (45.5)	0.01
Coronary artery disease (%)	252 (5.4)	0	18 (13.7)	234 (5.1)	<0.01
Myocardial infarction (%)	193 (4.1)	0	13 (9.9)	180 (3.9)	<0.01
Atrial fibrillation (%)	140 (3.0)	0	47 (35.9)	93 (2.0)	<0.01
Size of largest aneurysm (SD)	6.9 (4.8)	92	6.3 (4.0)	6.9 (4.8)	0.17
Number of aneurysms (SD)	1.4 (0.8)	0	1.3 (0.7)	1.4 (0.8)	0.52
Family history aneurysms (%)	788 (16.8)	0	18 (13.7)	770 (16.9)	0.35
Family history SAH (%)	456 (9.7)	0	8 (6.1)	448 (9.8)	0.16
Current tobacco use (%)	1396 (30.4)	105	22 (17.1)	1374 (30.8)	<0.01
Current alcohol use (%)	2030 (46.7)	347	53 (42.4)	1977 (46.8)	0.33
International Normalized Ratio (INR) at time of diagnosis (SD)	1.09 (0.27)	1230	2.05 (0.9)	1.05 (0.1)	<0.01

^*131 patients on warfarin

Covariance balance was optimized with inverse probability weighting using propensity scores. The results of the unweighted and weighted multivariable analyses are shown in Table 2. Five out of 10 variables had a standardized mean difference (SMD) of 10% or higher, indicating possible confounding prior to inverse probability weighting. This number decreased to 2 after weighting (Table 3). In the weighted multivariable analysis, current alcohol use (OR 1.89, 95% CI 1.10–3.24) and higher INR values at the time of diagnosis among non-anticoagulant users (OR 28.16, 95% CI 12.44–63.77) were significantly associated with aneurysmal subarachnoid hemorrhage. The interaction term between anticoagulant use and INR indicates the additional effects of INR in patients who were anticoagulated compared to those who were not. The odds ratio of the interaction term was less than 1 (equivalent to a negative coefficient) and indicates a decreased risk of rupture in anticoagulated patients for a given INR. To examine the effects of this interaction for specific INRs, the marginal effect was calculated. We found that in anticoagulant users, the marginal effects of anticoagulant therapy on rupture risk were negative and significant for INR 1.2 and higher (Figure 1). Although the size of some coefficients changed slightly, the direction and significance of most coefficients remained the same in the sensitivity analysis using complete cases only, with the exception of tobacco use which became significant, and alcohol use which lost significance (Supplemental Table I). In the subgroup analysis of patients not on anticoagulation, INR values were also associated with a higher risk of rupture at presentation (Supplemental Table II). The etiology for elevated INR in non-anticoagulated patients included liver disease, bleeding disorder, and vitamin K deficiency (Supplemental Table III). In the subgroup analysis of anticoagulated patients, there was a similar trend for INR although it did not reach statistical significance (Supplemental Table II). Figure 2 shows the proportion of ruptured aneurysms stratified according to INR values among anticoagulant users and non-users. The odds ratios and significance levels for INR in the weighted models were similar to the unweighted models (Tables 2 and Supplemental Table I).

Table 2:

Unweighted and weighted multivariable logistic regression for rupture status in all patients (N= 4,696). Multiple imputation (40 imputations) with chained equations was used for missing data. Multivariable analysis was performed with an interaction term between anticoagulant use and INR values.

	UnivariableUnweighted		Multivariable Unweighted		Multivariable Weighted
Characteristics	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
Female	0.64 (0.55–0.74)	<0.01	0.72 (0.62–0.84)	<0.01	0.75 (0.46–1.24)	0.27
Black race (vs. white race)	1.92 (1.50–2.45)	<0.01	1.97 (1.52–2.56)	<0.01	1.64 (0.65–4.15)	0.29
Hispanic race (vs. white race)	0.32 (1.01–1.72)	0.04	1.36 (1.03–1.80)	0.03	0.71 (0.34–1.48)	0.36
Asian race (vs. white race)	1.86 (1.25–2.76)	<0.01	1.99 (1.31–3.01)	<0.01	1.52 (0.75–3.08)	0.25
Other/unknown race (vs. white race)	0.32 (1.02–1.71)	0.03	1.45 (1.11–1.89)	<0.01	2.53 (0.81–7.86)	0.11
Age at diagnosis	0.98 (0.97–0.98)	<0.01	0.99 (0.98–0.99)	<0.01	0.99 (0.98–1.01)	0.43
Hypertension	1.11 (0.98–1.26)	0.10	-	-	-	-
Coronary artery disease	0.70 (0.51–0.95)	0.02	0.75 (0.55–1.04)	0.09	1.20 (0.63–2.29)	0.57
Myocardial infarction	0.86 (0.62–1.20)	0.37	-	-	-	-
Atrial fibrillation	0.97 (0.67–1.42)	0.89	-	-	-	-
Size of largest aneurysm	1.01 (0.99–1.02)	0.38	-	-	-	-
Number of aneurysms	1.02 (0.94–1.11)	0.60	-	-	-	-
Family history aneurysms	0.60 (0.50–0.73)	<0.01	0.56 (0.46–0.68)	<0.01	0.74 (0.35–1.57)	0.43
Family history subarachnoid hemorrhage	0.60 (0.47–0.76)	<0.01	-	-	-	-
Current tobacco use (vs. not current)	2.01 (1.75–2.30)	<0.01	1.95 (1.69–2.25)	<0.01	1.65 (0.99–2.75)	0.05
Current alcohol use (vs. not current)	1.36 (1.19–1.56)	<0.01	1.34 (1.16–1.54)	<0.01	1.89 (1.10–3.24)	0.02
Current anticoagulant therapy (vs. not current)*	0.67 (0.44–1.03)	0.07	8.23 (2.05–33.07)	<0.01	19.59 (3.11–123.20)	<0.01
INR at time of diagnosis†,‡	1.60 (1.24–2.07)	<0.01	22.78 (10.85–47.81)	<0.01	28.16 (12.44–63.77)	<0.01
Interaction term between anticoagulant use and INR	0.92 (0.76–1.10)	0.36	0.07 (0.03–0.17)	<0.01	0.04 (0.01–0.11)	<0.01

^*From coefficient of anticoagulant use

^†INR = International Normalized Ratio

^‡From coefficient of INR values. In the interaction model, this corresponds to INR in non-anticoagulant users.

Table 3:

Standardized mean difference (SMD) of covariables before and after weighting.

Covariables	Unweighted	Weighted
Female	−0.11	0.05
Black race (vs. white race)	−0.07	0.02
Hispanic race (vs. white race)	−0.09	−0.006
Asian race (vs. white race)	−0.13	−0.11
Other/unknown race (vs. white race)	0.08	0.07
Age at diagnosis	0.81	0.03
Coronary artery disease	0.30	−0.03
Family history aneurysms	−0.09	−0.07
Current tobacco use (vs. not current)	−0.32	−0.09
Current alcohol use (vs. not current)	−0.08	−0.18

Figure 1:

The marginal effects of anticoagulant use on rupture risk stratified by INR levels. * = significant. Error bars represent 95% confidence intervals.

Figure 2:

Proportion of ruptured aneurysms stratified by INR-values among A) anticoagulant users and B) non-anticoagulant users.

Discussion

We evaluated whether anticoagulation therapy and elevated INR values in patients with intracranial aneurysms were associated with rupture. Our primary finding is that elevated INR is significantly associated with ruptured aneurysms, even when adjusted for a wide variety of confounders. Interestingly, the effects of INR appear to be ameliorated by anticoagulant use at INR levels of 1.2 and higher.

Whether the use of anticoagulants influences the risk of subarachnoid hemorrhage has been the focus of a limited number of population-based studies with inconsistent results.^{9–12, 14, 15} In a registry-based Danish study using prescription data, a significant association between anticoagulation therapy and SAH could not be found.¹² However, only 9 patients with SAH were current users of vitamin K antagonists, important confounders such as smoking and alcohol use were lacking, and inclusion of patients based on diagnosis-codes could have led to misclassification of SAH cases. In contrast, in a subsequent Dutch population-based study, vitamin K antagonists were significantly associated with SAH (OR 2.46, 95% CI 1.04–5.82).¹¹ The most recent population-based study on this topic showed no association between anticoagulant use and SAH (<1 month: OR 1.85, 95% CI 0.97–3.51, and >3 years: OR 1.24, 95% CI 0.86–1.77).⁹ However, in all three studies, INR values were not available, preventing the authors from drawing firm conclusions on coagulation status and the risk of aSAH, since non-compliance in anticoagulant users could bias the results. García-Rodríguez and colleagues recently demonstrated that warfarin use was associated with an increased risk of SAH compared with no therapy (OR 1.67, 95% CI 1.15–2.43), with a non-significant association with international normalized ratio (INR) values of >3 (OR 2.64, 95% CI 0.95–7.35).¹⁰ However, INR values of only 24 SAH patients on anticoagulation therapy were included. In contrast, we included the INR values of both anticoagulant users and non-users, since high INR values could also be due to liver disease, bleeding disorders, and dietary vitamin K deficiency. Moreover, their control group consisted of subjects with an unknown incidence of unruptured aneurysms, in contrast to our study. One of the major limitations of all previously mentioned population-based registries are misclassification bias and information bias, both on the level of outcome and exposure, due to the use of diagnosis codes and subsequent decreased accuracy in diagnosis.¹¹

Previous studies reported an increased risk of intracerebral hemorrhage (ICH) in anticoagulant users.¹⁶ In addition, ICH associated with oral anticoagulant use was highly associated with mortality risk due to early hematoma growth.^{17, 18} In contrast, we recently showed in a nationwide study that anticoagulant use was not associated with differential mortality or complication rates after aSAH.¹⁹ Our current results indicate that anticoagulants, by decreasing the effects of increased INR values, may be protective against intracranial aneurysm rupture at a given INR value. This protective effect could be due to additional pharmacodynamic characteristics of oral anticoagulants beyond the inhibition of vitamin K-dependent coagulation factors, such as possible immunomodulatory properties.²⁰ Indeed, it has been shown that anticoagulation with warfarin downregulates inflammation by inhibition of interleukin-6 production and tumor necrosis factor (TNF)-induced I-κβ phosphorylation, and consequent inflammatory signal transduction. ^{21, 22} Since vascular inflammation, specifically TNF-mediated inflammation, is thought to play a key role in intracranial aneurysm pathogenesis^{23, 24}, the anti-inflammatory effects of warfarin may provide the explanation for its ameliorating effects on INR elevation and subsequent aneurysm rupture, as shown in our study. On the other hand, it is also possible that the patients with elevated INR but who were not anticoagulated may have involvement of other parts of the coagulation pathway beyond what is detectable by INR and be more susceptible to aneurysm rupture as a result²⁵.

The main strengths of our study are the large sample size, the presence of a large control group with unruptured intracranial aneurysms, and the use of a high-quality, homogeneous database including INR values for anticoagulant users and non-users. The main limitation of our study includes the retrospective design for a portion of the patients. It is possible that the decreased association between the effects of INR in anticoagulated patients and aneurysm rupture is non-causal but due to the increased use of intracranial imaging in anticoagulated patients leading to the increased discovery of unruptured aneurysms. However, propensity score weighting was used to control for selection bias. The subgroup analysis of anticoagulated patients showed a trend in the association of elevated INR with aneurysm rupture. The lack of statistical significance may be due to insufficient power because of the small sample size of that subgroup.

With an increasing number of unruptured intracranial aneurysms being diagnosed every year, specifically in an older population with poly-pharmacy and co-morbidities such as atrial fibrillation and valvular diseases, the present study addresses a clinically relevant question for the clinician. In deciding on the optimal management of unruptured intracranial aneurysms in patients with elevated INR values, the treating physician must understand the effect of anticoagulation therapy and elevated INR values on the natural history of unruptured intracranial aneurysms. Our results advocate caution with anticoagulant prescriptions in this vulnerable but growing patient population.

Summary

In conclusion, we found that elevated INR values are significantly associated with an increased risk of aneurysmal subarachnoid hemorrhage in patients with unruptured intracranial aneurysms, and that the effects may be moderated by anticoagulation therapy. Future randomized control trials are needed to confirm these findings.