Association Between Antiseizure Drug Monotherapy and Mortality for Patients With Poststroke Epilepsy

Key Points

Question  Does survival vary with choice of antiseizure medication (ASM) for patients with poststroke epilepsy?

Findings  In this population-based cohort study of 2577 patients with poststroke epilepsy who received ASM monotherapy, those treated with lamotrigine had a significantly lower hazard of cardiovascular and all-cause death compared with those treated with carbamazepine. Patients prescribed valproic acid had a higher risk of cardiovascular and all-cause death compared with patients treated with carbamazepine and lamotrigine, and those treated with levetiracetam had a lower risk of death from cardiovascular disease than patients treated with carbamazepine.

Meaning  These findings suggest that there are differences in survival between specific ASMs and that lamotrigine and levetiracetam seem to be reasonable first-line treatment options for patients with poststroke epilepsy.

Importance  There is little evidence to guide the choice of antiseizure medication (ASM) for patients with poststroke epilepsy. Theoretical concerns about detrimental effects of ASMs on survival exist. Enzyme-inducing drugs could interfere with secondary stroke prevention. The US Food and Drug Administration recently issued a safety announcement about the potential proarrhythmic properties of lamotrigine.

Objective  To investigate whether mortality varies with specific ASMs among patients with poststroke epilepsy.

Design, Setting, and Participants  A cohort study was conducted using individual-level data from linked registers on all adults in Sweden with acute stroke from July 1, 2005, to December 31, 2010, and subsequent onset of epilepsy before December 31, 2014. A total of 2577 patients receiving continuous ASM monotherapy were eligible for the study. Data were analyzed between May 27, 2019, and April 8, 2021.

Exposures  The dispensed ASM (Anatomical Therapeutic Chemical code N03A) determined exposure status, and the first dispensation date marked the start of treatment.

Main Outcomes and Measures  The primary outcome, all-cause death, was analyzed using Cox proportional hazards regression with carbamazepine as the reference. Cardiovascular death (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes I0-I99 as the underlying cause) was assessed using Fine-Gray competing risk regression models.

Results  A total of 2577 patients (1400 men [54%]; median age, 78 years [IQR, 69-85 years]) were included. The adjusted hazard ratio of all-cause death compared with carbamazepine was 0.72 (95% CI, 0.60-0.86) for lamotrigine, 0.96 (95% CI, 0.80-1.15) for levetiracetam, 1.40 (95% CI, 1.23-1.59) for valproic acid, 1.16 (95% CI, 0.88-1.51) for phenytoin, and 1.16 (95% CI, 0.81-1.66) for oxcarbazepine. The adjusted hazard ratio of cardiovascular death compared with carbamazepine was 0.76 (95% CI, 0.61-0.95) for lamotrigine, 0.77 (95% CI, 0.60-0.99) for levetiracetam, 1.40 (95% CI, 1.19-1.64) for valproic acid, 1.02 (95% CI, 0.71-1.47) for phenytoin, and 0.71 (95% CI, 0.42-1.18) for oxcarbazepine.

Conclusions and Relevance  This cohort study’s findings suggest differences in survival between patients treated with different ASMs for poststroke epilepsy. Patients receiving lamotrigine monotherapy had significantly lower mortality compared with those receiving carbamazepine. The opposite applied to patients prescribed valproic acid, who had a higher risk of cardiovascular and all-cause death. Levetiracetam was associated with a reduced risk of cardiovascular death compared with carbamazepine, but there was no significant difference in overall mortality.

Quiz Ref IDStroke is the most commonly identified cause of new-onset epilepsy in adults.¹ The association of poststroke epilepsy (PSE) with mortality varies between studies,^2,3 but the hazard ratio (HR) of death seems increased on a population level.⁴ The reason for the increased mortality with PSE is unknown. Potential explanations include detrimental effects from antiseizure medications (ASMs), especially enzyme inducers, which could increase cardiovascular risk and interact with drugs used for secondary stroke prevention.^5-7 Recently, a US Food and Drug Administration (FDA) drug safety communication raised concerns about the proarrhythmic effects of lamotrigine, a commonly used drug in PSE.⁸

To our knowledge, there are no population-based real-world data on associations between ASMs and mortality in patients with PSE. Sweden offers unique conditions for examining this question owing to nationwide registers with nearly complete coverage of acute stroke events, ASM prescriptions, and mortality data. We investigated whether cardiovascular and all-cause death varied with different ASM monotherapies in PSE, taking into account a number of factors, including stroke severity.

We conducted a retrospective cohort study using individual-level data from Swedish population-based registers from July 1, 2005, to December 31, 2014. The exposure was ASM monotherapy, the primary outcome was all-cause death, and the secondary outcome was cardiovascular death. The unique personal identity number of each person with a Swedish citizenship or residence permit allowed linkage of register data. The Gothenburg Regional Ethics Review Board approved the study and waived the need for patient consent because the data were deidentified.

We merged data from 4 population-based registers. All adults admitted to a hospital owing to acute stroke are eligible for the Swedish Stroke Register (Riksstroke⁹), which collects information on stroke care, stroke severity, and cardiovascular comorbidities. A questionnaire follow-up is performed 3 months after the stroke and collects information on support and activities of daily living (ADL). Previous validations have shown a high consistency between the clinical data recorded in the register and medical records.⁹ In 2012, Riksstroke had an estimated coverage of 94% for patients with acute stroke in Sweden.¹⁰

The National Patient Register (NPR) contains information on all hospital admissions in Sweden since 1987,¹¹ and hospital-based physicians’ appointments and emergency department visits since 2001. It includes diagnostic coding for each admission (according to the International Classification of Diseases, Ninth Revision [ICD-9] and the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10]).

Every pharmacy in Sweden is required by law to provide data to the Swedish eHealth Agency on dispensations of prescribed drugs. The transfer of prescription data is automatic, and the information is obtainable through the Swedish Prescribed Drug Register (SPDR).¹²

The Swedish Cause of Death Register (CDR) contains mortality data for Swedish inhabitants since 1952 (eg, date of death, underlying cause of death, and contributing causes of death). The information originates from medical death certificates issued by physicians. In 2015, only 0.9% of all deaths were missing an underlying cause.¹³

All adults with an acute stroke event from July 1, 2005, to December 31, 2010, were identified in Riksstroke (n = 119 180). The NPR provided information on seizure-related diagnostic codes (ICD-9 codes 345 and 780D; and ICD-10 codes G40, G41, and R568) and the SPDR provided information on dispensed ASMs (Anatomical Therapeutic Chemical [ATC] code N03A). Patients with seizure-related diagnostic codes before the stroke were excluded. Poststroke epilepsy was defined as an ICD-10 code for epilepsy, seizures, or status epilepticus (codes G40, R568, and G41) more than 7 days after stroke. Quiz Ref IDAntiseizure medication treatment was tracked as previously described.¹⁴ Only those with persistent ASM treatment (ie, <1 year between dispensations throughout follow-up) were included. The standard ASM dispensation interval in Sweden is approximately 3 months. Patients with ASM dispensations before PSE or with dispensations of multiple ASMs were excluded. All included participants used a single ASM substance throughout follow-up (no switch or add-on). Figure 1 illustrates the eligibility process.

The dispensed ASM substance (ATC code N03A) determined exposure status, and the first dispensation date marked the treatment start. The CDR provided information on outcomes, including the date and the underlying cause of death. “Cardiovascular death” was defined as an ICD-10 code of I0-I99 (diseases of the circulatory system) as the underlying cause of death.

We extracted information on demographic characteristics, stroke characteristics, living arrangements, ADL, and smoking habits (before stroke) from Riksstroke. Our definition of ADL dependence (need assistance toileting, dressing, or moving around indoors) has been validated against the modified Rankin Scale (corresponds well to grade ≥4).¹⁵ The NPR provided information on status epilepticus (ICD-10 code G41). Hypertension, type 1 or 2 diabetes, and atrial fibrillation were based on either a diagnosis recorded during the stroke admission or a diagnostic code in the NPR before the start of ASM treatment (ICD-10 codes I10-I15 [hypertension], E10-E14 [diabetes], and I48 [atrial fibrillation]). The SPDR was the source of information on statins and antidepressants (dispensations of statins [ATC code C10AA] or antidepressants [ATC code N06A] between stroke and ASM treatment start).

Statistical analysis was conducted between between May 27, 2019, and April 8, 2021. We used IBM SPSS Statistics, version 26 (IBM Corp) to calculate descriptive statistics and Kaplan-Meier estimators. Survival rates were extracted from survival tables, and the SE was multiplied by 1.96 to obtain an estimate of the 95% CI. The survival time was calculated from treatment start to death (event), with censoring at study end (December 31, 2014). Group differences were assessed with the Fisher exact test, the χ² test, or the Kruskal-Wallis test, whichever was appropriate. All P values were from 2-sided tests, and results were deemed statistically significant at P < .05.

To estimate the risk of death, we used Cox proportional hazards regression. The proportional hazards assumption was checked by including time-dependent covariates in the model. Quiz Ref IDAntiseizure medication treatment was the only covariate in the univariable model. Carbamazepine was selected as a reference because it was frequently prescribed and has enzyme-inducing properties. The baseline characteristics listed in Table 1 were additional covariates in the multivariable model. When analyzing cardiovascular mortality, other underlying causes of death were considered competing risks. Consequently, we used the Fine-Gray competing risk regression model for these analyses.

Missing covariate data were handled using multiple imputation based on the standard fully conditional specification method in SAS, version 9.4 (SAS Institute Inc). Although some element of “missing not at random” is possible, we believe there are enough predictor variables available to justify the analysis. To increase the plausibility of the “missing at random” assumption, we included auxiliary variables (not used in the main analyses) that are associated with either missingness or the values of incomplete variables (eg, proxy data for stroke severity). The 27 variables included in the imputation procedure are provided in eMethods 1 in the Supplement. Twenty imputed data sets were created, each based on 20 iterations. Proportions of imputed variables before and after imputation are displayed in eTable 1 in the Supplement.

A total of 2577 patients (1400 men [54%]; median age, 78 years [IQR, 69-85 years]) with PSE receiving ASM monotherapy were included. Table 1 presents baseline characteristics, stratified for the 5 most commonly prescribed ASMs and oxcarbazepine. The median time between stroke and the first seizure-related diagnostic code was 347 days (IQR, 143-800 days). Median follow-up (from the start of treatment to death or censoring at study end) was 2.2 years (IQR, 0.7-4.3 years). Event-free patients had a median observation time of 4.3 years (IQR, 2.6-6.0 years).

Details on covariates with missing values are provided in eMethods 2 in the Supplement. When comparing survival between patients with complete data (n = 1966 [76%]) and patients with any missing covariate data (n = 611 [24%]), there was a moderate difference in the Kaplan-Meier survival estimators indicating a higher probability of survival for patients with complete data (3-year survival rate, 0.51 [95% CI, 0.49-0.54] vs 0.44 [95% CI, 0.40-0.48]). Characteristics of patients with complete and incomplete data are provided in eTable 2 in the Supplement. Missing values were imputed for the main analyses.

During the study period, 1550 deaths occurred. Figure 2 illustrates the crude survival functions, assessed with the Kaplan-Meier estimator. Lamotrigine had the highest 3-year survival rate (0.62 [95% CI, 0.56-0.67]), followed by levetiracetam (0.55 [95% CI, 0.49-0.61]), oxcarbazepine (0.54 [95% CI, 0.39-0.68]), carbamazepine (0.53 [95% CI, 0.50-0.56]), valproic acid (0.34 [95% CI, 0.30-0.39]), and phenytoin (0.32 [95% CI, 0.21-0.43]). When studying 5-year survival, the differences between lamotrigine, valproic acid, and carbamazepine remained statistically significant.

Table 2 displays the results of crude and adjusted analyses. The adjusted HR of all-cause death compared with carbamazepine was 0.72 (95% CI, 0.60-0.86) for lamotrigine, 0.96 (95% CI, 0.80-1.15) for levetiracetam, 1.40 (95% CI, 1.23-1.59) for valproic acid, 1.16 (95% CI, 0.88-1.51) for phenytoin, and 1.16 (95% CI, 0.81-1.66) for oxcarbazepine. The following control variables were significant in the multivariable model: age, sex, stroke type, living arrangements before and after stroke, ADL dependence before stroke, status epilepticus, diabetes, statin use, and smoking.

The corresponding analysis of patients with complete data (n = 1966) is provided in eTable 3 in the Supplement. The minor HR variations stem primarily from differences in estimates between the patients with complete data and the omitted patients with missing data (crude and age-adjusted HRs stratified for missingness are displayed in eTable 4 and eTable 5 in the Supplement). Owing to the limited number of patients in the cohort taking oxcarbazepine, the estimates of oxcarbazepine are more prone to chance variation.

Cardiovascular disease was considered the underlying cause of 969 deaths (63%). Table 3 displays crude and adjusted HRs. The results for lamotrigine and valproic acid were similar to the analyses for all-cause death. The adjusted HR of cardiovascular death compared with carbamazepine was 0.76 (95% CI, 0.61-0.95) for lamotrigine, 0.77 (95% CI, 0.60-0.99) for levetiracetam, 1.40 (95% CI, 1.19-1.64) for valproic acid, 1.02 (95% CI, 0.71-1.47) for phenytoin, and 0.71 (95% CI, 0.42-1.18) for oxcarbazepine.

Several sensitivity analyses were performed. First, we used a more specific definition of PSE, requiring a detected diagnostic code for epilepsy (ICD-10 code G40) later than 7 days after the stroke event (n = 2138). The combination of code G40 and ASM treatment is highly specific for epilepsy.¹⁶ The results for this analysis were similar to the main analyses. The adjusted HR of all-cause death was 0.70 (95% CI, 0.58-0.85) for lamotrigine, 1.05 (95% CI, 0.86-1.28) for levetiracetam, and 1.46 (95% CI, 1.27-1.69) for valproic acid. Second, because levetiracetam was not approved in Sweden at the beginning of the study period, we restricted the analyses to patients with a first seizure-related diagnostic code in 2009-2014 (n = 1589). The adjusted HR of cardiovascular death was 0.76 (95% CI, 0.57-1.00) for lamotrigine, 0.73 (95% CI, 0.56-0.97) for levetiracetam, and 1.36 (95% CI, 1.08-1.71) for valproic acid. Third, to test the association of shifting prescribing patterns and improved stroke care over time, the year of treatment start was added as a covariate in the multivariable model. The adjusted HR of death for lamotrigine was 0.75 (95% CI, 0.63-0.90) and for valproic acid was 1.44 (95% CI, 1.26-1.63), in agreement with the main analysis. Last, we limited our analyses to patients with less than 1 year between stroke and treatment start. The adjusted HR of death was 0.70 (95% CI, 0.55-0.91) for lamotrigine, 0.89 (95% CI, 0.64-1.23) for levetiracetam, and 1.41 (95% CI, 1.19-1.68) for valproic acid.

We also assessed differences in indicators of epilepsy severity (or drug effectiveness) between groups. The median number of hospital admissions with a main diagnosis of epilepsy, seizures, or status epilepticus per year of follow-up was 0 (IQR, 0-0) for carbamazepine, 0 (IQR, 0-0.234) for lamotrigine, 0 (IQR, 0-0) for levetiracetam, and 0 (IQR, 0-0.269) for valproic acid. Admissions were significantly more common among patients taking lamotrigine and valproic acid than those taking carbamazepine. We also compared epilepsy-related causes (ICD-10 code R568, G40, or G41) of death in the CDR. In the carbamazepine group, 14% of deaths (100 of 715) had a seizure-related diagnostic code listed as an associated cause, which was significantly lower than the other groups (valproic acid, 23% [83 of 366]; levetiracetam, 21% [31 of 145]; and lamotrigine, 22% [34 of 153]).

Quiz Ref IDPatients treated with lamotrigine monotherapy had significantly lower mortality compared with those treated with carbamazepine. The opposite applied to patients prescribed valproic acid, who had a higher risk of cardiovascular and all-cause death compared with those receiving carbamazepine or lamotrigine. Univariable analyses and models adjusted for multiple potential confounders, including stroke severity, yielded concordant results. Patients treated with levetiracetam had a lower risk of death from cardiovascular disease than patients treated with carbamazepine, but there was no statistically significant difference in overall mortality.

Our findings raise the possibility that specific ASMs influence the risk of cardiovascular and all-cause death, although our study design does not allow causal inference. Altered vascular risk is suspected to be the main reason behind our findings. Treatment with enzyme-inducing ASMs (carbamazepine and phenytoin) enhances metabolism of drugs commonly used in secondary prevention after stroke, including anticoagulants, calcium channel blockers, and statins.^7,17 In addition, ASMs can be directly associated with markers of vascular disease. For instance, carbamazepine and phenytoin have been associated with lipid abnormalities and increased levels of C-reactive protein, whereas valproic acid has been linked to weight gain, metabolic syndrome, and related endocrine abnormalities.^7,18,19Quiz Ref ID Carbamazepine, phenytoin, and valproic acid, but not lamotrigine, have also been associated with increased carotid intima-media thickness (a surrogate marker for stroke and myocardial infarction).²⁰ It seems less likely that the differences in mortality reflect epilepsy severity (see Limitations), and sensitivity analyses adjusted for the year of treatment start suggest that the differences are not explained by general improvements in stroke care over time.

To our knowledge, no previous studies have investigated associations between specific ASMs and survival in PSE. In the general population of patients with epilepsy, a Danish register-based cohort study reported a reduced risk of cardiovascular death with lamotrigine compared with carbamazepine monotherapy.²¹ In contrast to our findings, that study did not find significant differences in overall mortality between lamotrigine, valproic acid, and carbamazepine. One possible explanation for this inconsistency could be that patients with PSE may be particularly vulnerable to valproic acid’s unfavorable association with vascular risk. Furthermore, it suggests that our findings are not generalizable to a diverse cohort of patients with epilepsy.

Current ASM use is associated with an increased risk of sudden cardiac death in patients with and without epilepsy.²² The FDA recently issued a drug safety communication about the possibility of proarrhythmic effects of lamotrigine, and the International League Against Epilepsy has issued recommendations on management of cardiac risk with lamotrigine.⁸ Our real-world investigation does not suggest that lamotrigine should pose a particular risk for patients with PSE on a group level. Risk factors or biomarkers identifying those particularly at risk are needed.²³

Currently, treatment recommendations for PSE are based primarily on trials conducted in older adults with epilepsy of various causes. The evidence supporting the use of specific ASMs is limited, but experts have suggested lamotrigine and levetiracetam as valid treatment options.²⁴ Nonetheless, carbamazepine and valproic acid are commonly used among elderly patients with new-onset epilepsy.^14,25,26

Our findings suggest that ASM selection is associated with survival among patients with a history of cerebrovascular disease and support the use of lamotrigine as a first-line drug. Our findings regarding valproic acid indicate a need for more studies before which caution is advisable regarding use in this particular patient group. Further studies are needed to assess ASM effectiveness, drug-drug interactions, and the association of individual ASMs with cardiovascular outcomes, specifically in PSE. Irrespective of the selected ASM, the treating neurologist should be vigilant about risk factor intervention to modify general vascular risk.

This study has some limitations. We do not have information on all factors infuencing ASM selection, so some confounding is likely. In a setting of vascular disease and low survival rates, we believe age, functional status, and epilepsy severity to be the major determinants of ASM treatment. Hospitalizations for epilepsy during follow-up were less frequent in the carbamazepine group compared with the lamotrigine and valproic acid groups. Epilepsy was also a less common associated cause of death in the carbamazepine group. Taken together, the difference in mortality between lamotrigine and carbamazepine seems unlikely to be explained by epilepsy severity. Our eligibility criteria (ie, consistent use of a single ASM) also means the cohort probably consists of easy-to-treat patients. Our findings do not extend to patients requiring drug changes for seizure control; the impact of poorly controlled epilepsy may well outweigh other risks associated with ASMs. To reduce confounding by indication, multivariable analyses included adjustment for status epilepticus.

Another limitation is the lack of adjustment for comorbidities not available through Riksstroke. Nonvascular comorbidities are, however, unlikely to explain the lower hazard of death seen with lamotrigine or levetiracetam. These drugs have fewer drug-drug interactions and are more likely choices than carbamazepine in patients with pharmacologically treated comorbidities such as cancer and depression. We cannot dismiss the possibility that patients treated with valproic acid may have had a higher prevalence of comorbidities than patients treated with carbamazepine. That would, however, have less of an association with our secondary outcome, cardiovascular death.

Missing covariate data are another limitation. Removing patients with missing values from the analyses would result in a loss of around 24% of the sample. Our data set included a wide range of auxiliary variables from multiple sources, some of which were used to assess missing values and increase the plausibility of the “missing at random” assumption. Multiple imputation accounts for uncertainty in predicting missing values by adding variability. In retrospect, the HR estimates did not change notably when missing data had been imputed.

Because of the register-based approach, most variables have measurement errors to some degree. A discharge diagnosis of epilepsy (ICD-10 code G40) in the NPR has a positive predictive value of 90%.²⁷ Specificity is likely to be even higher in patients with a history of stroke and concomitant use of an ASM.¹⁶ Using the same definition of PSE, one study estimated the cumulative incidence rate to be 7.3%,⁴ which aligns with the results from other studies with long-term follow-up after stroke and suggests that we have identified most patients.^28,29 The SPDR covers all drugs issued via prescription in Sweden, but we have no information on the total daily dose or medication adherence after the filling of the prescription. We also used a conservative definition of treatment discontinuation, allowing up to 1 year between dispensations. Cardiovascular disease as the underlying cause of death (ICD chapter level) in the CDR has an accuracy of 87% compared with medical records.³⁰ These information errors should not differ between ASMs, but they may bias HRs slightly toward the null, making our estimates conservative.

Finally, our last data were captured in 2014, which means we could not investigate newer ASMs. The drugs studied are still the most commonly prescribed first ASMs among older adults with new-onset epilepsy.^26,31

In this nationwide register-based cohort study, we found differences in survival between patients with PSE treated with different ASMs. Patients treated with lamotrigine had significantly lower mortality than those treated with carbamazepine, whereas patients prescribed valproic acid had an increased risk of both cardiovascular and all-cause death. Levetiracetam was associated with a reduced risk of cardiovascular death compared with carbamazepine. Altogether, our findings suggest an association between ASM selection and mortality among patients with a history of cerebrovascular disease.

Accepted for Publication: October 18, 2021.

Published Online: December 13, 2021. doi:10.1001/jamaneurol.2021.4584

Corresponding Author: David Larsson, MD, Department of Neurology, Sahlgrenska University Hospital, Blå stråket 7, 413 45, Gothenburg, Sweden (david.gw.larsson@vgregion.se).

Author Contributions: Drs Larsson and Zelano had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Larsson, Zelano.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Larsson, Zelano.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Larsson, Zelano.

Obtained funding: Zelano.

Administrative, technical, or material support: Zelano.

Supervision: Johannessen Landmark, Zelano.

Conflict of Interest Disclosures: Dr Baftiu reported being employed as Medical Advisor, Neuroscience at Janssen Cilag AS. Dr Johannessen Landmark reported receiving honoraria from Eisai, GW, and UCB Pharma. Dr Åsberg reported receiving grants from Astra Zeneca outside the submitted work. Dr Zelano reported receiving grants from Swedish state (ALF), Swedish Society of Medicine, Swedish Society of Medical Research, Linnea and Josef Carlsson Foundation, Gothenburg Medical Society, Gothenburg University, and Magnus Bergvall Foundation during the conduct of the study; honoraria from Eisai, and UCB Pharma; being an investigator in a sponsored clinical trial for UCB Pharma, GW Pharma, SK Life Science, and Bial outside the submitted work. No other disclosures were reported.

Funding/Support: This study was funded by grants ALFGBG-715781 and ALFGBG-784921 from the Swedish state under the ALF agreement, grant SLS-881501 from Swedish Society of Medicine, grant S18-0040 from Swedish Society of Medical Research, grant 90_20180321_048 from Linnea and Josef Carlsson Foundation, grant GLS-780651 from Göteborg Medical Society, and grant 2017-01990 from Magnus Bergvall Foundation.

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.

Additional Contributions: Mattias Molin, BSc, Statistiska Konsultgruppen (a commercial statistical consultancy, Gothenburg, Sweden), performed regression analyses and the imputation procedure; he was compensated for his contribution.