Age Modulates the Association of Caffeine Intake With Cognition and With Gray Matter in Elderly Diabetics

Abstract

Background

The association between caffeine and cognitive performance has not been tested in older individuals with type 2 diabetes (T2D). Its association with brain volume in T2D has been tested only in animals.

Methods

We examined the association of caffeine with cognitive function and brain volume in a sample of elderly diabetics participating in the Israel Diabetes and Cognitive Decline Study (n = 638) and the moderating effect of age on this association. In a subsample (n = 185) with magnetic resonance imaging, we also examined these associations with gray and white matter volumes (GM/WM).

Results

Using linear regression adjusting for cognition-related covariates, we found that higher caffeine intake was associated with better function in overall cognition (p = .018), attention/working memory (p = .002), executive functioning (p = .047), and semantic categorization (p = .026). Interaction analyses of caffeine intake with age were significant for semantic categorization (p = .025), and approached significance for overall cognition (p = .066). This association was driven by the older group (above-median) for whom the association of caffeine intake with semantic categorization (p = .001), attention/working memory (p = .007), executive functioning (p = .005), and overall cognition (p = .002) were significant. In the magnetic resonance imaging subsample, there was an interaction (p = .034) of caffeine intake with age for GM volume; in the older group, higher caffeine intake was associated with greater GM volume (β = .198, p = .033).

Conclusions

Caffeine intake may have a beneficial role in cognitive functioning of elderly adults with T2D, which may be moderated by age. Greater GM volume may be a mechanism underlying the association of higher caffeine intake with better cognitive function.

Caffeine is among the most commonly ingested stimulants worldwide (1), naturally occurring in coffee, cocoa, and tea, among other foodstuffs. It is generally considered safe, with numerous potential benefits such as improved alertness, but also potential dangers, including anxiety and impaired sleep (2). Acute caffeine consumption has been linked to temporary improvements in cognitive functioning, especially in reaction time and focused attention (2). Habitual or longer-term caffeine use has also been linked to better global and specific cognitive performance and cognitive maintenance (3,4). However, caffeine may not be beneficial for all cognitive domains, may be more beneficial for women (5), and may not be dose-dependent (6,7). Similarly, caffeine has been associated with lower risk of dementia and Alzheimer’s disease (AD) (5,8,9) but see (6,10) and in AD mouse models, caffeine has been shown to reverse memory impairment (11), and with lower risk of other neurodegenerative disorders, including Parkinson’s disease (12). Finally, there is evidence indicating its association with hippocampal volume (13), more white matter (WM) lesions in the elderly adults (14), changes in cerebral blood flow (1), functional connectivity (1), and with fewer AD and vascular neuropathologic lesions (15).

Age is the strongest risk factor for cognitive decline and dementia. Age has been shown to modify the relationship of caffeine intake with cognition (3,4), so that the potential advantage provided by caffeine to the brain is stronger with advancing age. The causes for this moderating effect of age are unknown but longer accumulating effects of caffeine (16), a stronger physiological reaction to caffeine in older individuals (17) are potential explanations. Age is also associated with decreased gray matter (GM) volume (18), as is T2D (19), however very little is known about the relationship between caffeine and GM.

Elderly adults with type 2 diabetes (T2D) are at particularly high risk of developing cognitive impairment and dementia (20). However, despite the potential neuroprotective effects of caffeine, its relationship with cognition has not been investigated in this high-risk population. Caffeine may be especially beneficial for T2D; it has been found to reverse T2D-related neurological damage and prevent memory loss in T2D model mice. Thus, we examined the relationships of caffeine intake with cognition and the impact of age on this relationship in a large cohort of T2D elderly adults participating in the Israel Diabetes and Cognitive Decline (IDCD) study. The IDCD study is an investigation of the relationships of long-term T2D-related characteristics with cognitive decline. In a subsample of the cohort that had a magnetic resonance imaging (MRI), we examined the relationship of caffeine with gray and WM volume. The study is a collaboration between the Icahn School of Medicine at Mount Sinai (ISMMS), NY, the Sheba Medical Center, Israel, and the Maccabi Healthcare Services, Israel.

Method

Study Population

The IDCD study enrolls subjects with T2D, aged 65+ years, and living in the area of Tel-Aviv. They are randomly selected from the approximately 11,000 T2D individuals that are in the Maccabi Health Services diabetes registry. The IDCD study is approved by the ISMMS, Sheba, and Maccabi IRB committees. Inclusion criteria requires that participants are nondemented at the time of enrollment, have at least three hemoglobin A1c (HbA1c) assessments, have an informant, and are fluent in Hebrew. Exclusion criteria prohibit an ICD code for dementia or dementia subtypes, prescribed cholinesterase inhibitors, or a major psychiatric or neurological condition that could affect cognitive performance. The IDCD has recruited 1,288 subjects of whom 897 were eligible (8.5% refused participation and 21.9% were excluded by eligibility criteria) and 744 had caffeine intake information; 638 had complete data on all relevant variables (caffeine, full cognitive testing, and covariates).

Maccabi Health Services Diabetes Registry

The Maccabi Diabetes Registry is an electronic patient record system that includes diagnoses, medications, and blood exam values for any Maccabi client with diabetes. In order to be listed in the Maccabi registry, patients must have one of the following: (i) HbA1c >7.25%, (ii) Glucose >200 mg/dL on two exams more than 3 months apart, (iii) purchase of diabetic medication twice within 3 months, and (iv) diagnosis of diabetes (ICD9 code) by a general practitioner, internist, endocrinologist, ophthalmologist, or diabetes advisor, supported by a HbA1c>6.5% or Glucose>125 mg/dL within half a year.

Dementia Evaluation and Diagnostic Consensus Conference

Each participant in the IDCD completes a demographic, medical and psychiatric history, neuropsychological testing, and physical testing. Laboratory results are received from the Maccabi registry. For each participant, a diagnostic consensus conference is jointly conducted by neurologists, psychiatrists, geriatricians, and neuropsychologists (at least two types of expertise). Only subjects deemed as cognitively normal are included in the IDCD study.

Neuropsychological battery

The outcome measures of this study are derived from a neuropsychological test battery and have been adapted to Hebrew and the Israeli culture (21). It is administered by a neuropsychologist. All test scores were transformed into Z scores, reversed if necessary, so positive values represented good cognition. Factor analysis summarized the neuropsychological measures into four factors: episodic memory (immediate recall, delayed recall, and recognition), executive functioning (Trails A and Trails B, constructional praxis, and digit symbol), semantic categorization (similarities, letter fluency, category fluency), and attention/working memory (diamond cancellation, digit span forward, and digit span backwards). The score for each domain was the sum of its z-scores; an overall cognition score was the sum of all z-scores.

Caffeine Intake

Daily caffeine intake was assessed using the Food Frequency Questionnaire (FFQ), a questionnaire specifically developed and validated for the elderly population in Israel (22,23). Instant coffee, coffee, tea, and soft drinks were considered sources of caffeine (with coffee accounting for 84% of the caffeine consumed). Beverage portion sizes were based on the standard portion sizes, as published by the Israeli Ministry of Health. Daily caffeine intake was calculated by multiplying portion by the values of caffeine per 100 mL and by frequency of intake per month, and dividing by 30.5 (an estimate of days/month) in order to get an estimation of caffeine intake per day.

Estimation of Gray and White Matter Volume

Randomly selected participants from the IDCD cohort (n = 897, n = 744 with caffeine intake information) underwent a MRI scan. MRI scans were performed in the diagnostic imaging department, at Sheba Medical Center with a 3 Tesla scanner (GE, Signa HDxt, v16VO2). High-resolution (1 mm³) images were acquired using a 3D inversion recovery prepared spoiled gradient-echo (FSPGR) T1-weighted sequence (TR/TE = 7.3/2.7s, 20° flip angle, TI 450 ms). T1-weighted anatomical images for each subject were processed using the Voxel Based Morphometry (VBM) toolbox, developed by Gaser (http://www.fil.ion.ucl.ac.uk/spm/ext/#VBMtools) and implemented in Statistical Parametric Mapping (SPM8) software. This procedure included automated iterative skull stripping, segmentation of the images into GM, WM, and cerebrospinal fluid probability images, and spatial normalization of the GM images to a customized GM template in standard MNI (Montreal Neurological Institute) space. In order to optimize signal to noise, the GM maps were smoothed using an 8 mm Gaussian kernel. GM probability maps were held at the threshold of 0.1 to minimize inclusion of incorrect tissue types. Total intracranial volume (ICV) was calculated using the segmented and thresholded images (ICV = GM + WM + CSF).

Measurement of Covariates

Sociodemographic covariates were age, sex, and years of education. Cardiovascular covariates were low-density lipoprotein cholesterol (LDL) and high-density lipoprotein cholesterol (HDL) cholesterol, and systolic and diastolic blood pressure. Inflammation covariates were IL-6 and C-reactive protein (CRP). T2D-related covariates were duration of T2D (calculated as the time from the first date the subject was identified by the Diabetes Registry as having T2D, and the date of the baseline cognitive assessment), and hemoglobin A1c [HbA1c] the gold standard measure of glycemic control. APOE genotype (dichotomized as having or not having an APOE4 allele), serum albumin, and serum thyroid-stimulating hormone, each associated with cognitive functioning (24–26), were also covariates, as was smoking (never, previously, current) Cardiovascular covariates were each calculated by averaging all measurements available for each subject in the Maccabi Diabetes Registry since its inception in 1998 or since the participant entered the Registry. The first two HbA1c measurements in the Diabetes Registry were not included in the analyses to avoid residual effects from the time before the diagnosis.

Statistical Analysis

Regression analyses, controlling for sociodemographic, caloric intake, cardiovascular, albumin, thyroid-stimulating hormone, and T2D characteristics, were performed to examine the associations of caffeine intake with the four cognitive domains and overall cognition.

To examine modulation by age of the relationship of caffeine with cognitive outcomes, regression analyses examined the relationship of cognitive outcomes with an interaction variable comprised of the product of age and caffeine. For a concrete description of the change in the association of caffeine with cognition as age increases, regression analyses examined the relationships of caffeine with cognitive outcomes split by the median age (which was similar for the whole sample [71.50] and the MRI subsample [71.08]). Median age was used because it was similar for both the whole sample and the MRI subsample, allowing for comparisons of two groups of similar size.

Finally, to identify the relationship between age, caffeine, and brain volume, regression analyses were used to examine (i) the association of caffeine with GM and WM in N = 183 participants with MRI, (ii) the interaction of age and caffeine on MRI outcomes, and (iii) the correlations between caffeine and MRI outcomes within younger and older participants.

Results

Participant Characteristics

Participant characteristics, for the entire sample, and above and below median of age are presented in Table 1. The younger age group had higher caffeine intake, systolic blood pressure and albumin levels, grater GM volume, shorter duration of diabetes, and lower thyroid-stimulating hormone. There were no significant differences among the groups in years of education, caloric intake, total intracranial volume, WM volume, HDL and LDL cholesterols, diastolic blood pressure, hemoglobin A1C, smoking, ApoE4, CRP, or IL-6.

Table 1.

Characteristics of the Study Population

	Entire Sample (N = 638)	Entire Sample by Age Split
Mean (SD)	All Participants	Younger Group (age 64–71.5) N = 317	Older Group Age (age 71.5–86) N = 321	p-value Difference
Overall cognition	0.68 (7.3)	2.4 (6.9)	−1.02 (7.3)
Semantic categorization	0.21 (2.4)	0.67 (2.3)	−0.25 (2.3)
Attention/working memory	0.08 (2.2)	0.30 (2.2)	−0.14 (2.1)
Executive functioning	0.23 (2.9)	0.91 (2.8)	−0.43 (3.0)
Episodic memory	0.15 (2.2)	0.52 (2.1)	−0.21 (2.2)
Female (%)	38.2%	36.6%	39.9%	.220
Years of education	13.3 (3.5)	13.5 (3.3)	13.0 (3.6)	.306
Caloric intake (Kcal)	2,066.7 (703.37)	2,091.18 (745.48)	2,042.56 (659.41)	.383
Caffeine (mg)	204.2 (114.5)	214.94 (125.4)	193.64 (101.7)	.025
HDL cholesterol (mg/dL)	47.5 (10.7)	46.9 (10.4)	48.0 (10.9)	.195
LDL cholesterol (mg/dL)	101.7 (19.0)	103.1 (19.7)	100.3 (18.3)	.070
Diastolic blood pressure	84.0 (14.9)	84.9 (14.5)	83.2 (15.1)	.160
Systolic blood pressure	165.9 (9.7)	166.8 (9.6)	165.1 (9.8)	.027
Albumin	4.31 (.2)	4.33 (.2)	4.28 (.2)	<.001
TSH	2.5 (1.5)	2.3 (1.4)	2.6 (1.6)	.015
Duration of diabetes	8.7 (2.6)	8.5 (2.7)	8.9 (2.5)	.042
HbA1c	6.8 (.80)	6.8 (.86)	6.7 (.73)	.224
Apoe4	13.5%	12.3%	14.6%	.227
T2D medication	75.9% hypoglycemic only	76.3% hypoglycemic only	75.4% hypoglycemic only	.892
	0.8% insulin only	0.6% insulin only	0.9% insulin only
	9.9% insulin and hypoglycemic	9.1% insulin and hypoglycemic	10.6% insulin and hypoglycemic
	13.5% none	13.9% none	13.1% none
Smoker	Never 39.7%	Never 37.9%	Never 41.4%	.470
	Previously 48.7%	Previously 49.2%	Previously 48.3%
	Current 11.6%	Current 12.9%	Current 10.3%
Il-6 (pg/mL)	2.81 (2.4)	2.77 (2.7)	2.85 (2.2)	.205
CRP (mg/mL)	1.14 (1.9)	1.19 (2.2)	1.09 (1.5)	.464
Gray matter (N = 185)	514.21 (50.8)	532.04 (50.6) (N = 101)	492.78 (43.34) (N = 84)	<.001
White matter (N = 185)	515.72 (66.8)	521.5 (71.9) (N = 101)	508.8 (59.7) (N = 84)	.196
Total intracranial volume (N = 185)	1,335.82 (135.91)	1,350.04 (146.4) (N = 101)	1,318.72 (120.7) (N = 84)	.119

Note: CRP = C-reactive protein; HDL = high-density lipoprotein cholesterol; LDL = low-density lipoprotein cholesterol; TSH = Thyroid-stimulating hormone.

Caffeine and Cognition

Adjusting for the covariates, regression analyses showed that higher caffeine intake was significantly associated with better attention/working memory (R² = .211, β = .117, t(619) = 3.108, p = .002), overall cognition (R² = .372, β = .080, t(619) = 2.372, p = .018), semantic categorization (R² = .286, β = .080, t(619) = 2.229, p = .026), and executive functioning (R² = .342, β = .069, t(618) = 1.992, p = .047), but not episodic memory (R² = .129, β = −.030, t(619)= −.753, p = .452).

Age modified the relationship of caffeine with semantic categorization (R² = .292, β = 1.272, t(618) = 2.245, p for interaction = .025), and at trend level for overall cognition (R² = .375, β = .980, t(618) = 1.841, p = .066), but not executive functioning (R² = .344, β = .759, t(617) = 1.391, p = .165), attention/working memory (R² = .212, β = .305, t(618) = .511, p = .610), or episodic memory (R² = .130, β = .573, t(618) = .914, p = .316).

To describe the direction of the interaction, linear regression analysis of caffeine with the cognitive outcomes was performed on subgroups above and below median, and showed that the relationship was strongest in the older age group (see Table 2), where caffeine was significantly associated with overall cognition (R² = .399, β = .151, t(302) = 3.176, p = .002), semantic organization (R² = .326, β = .163 t(302) = 3.250, p = .001), attention/working memory (R² = .220, β = .148, t(302) = 2.733, p = .007), and executive functioning (R² = .402, β = .135, t(301) = 2.860, p = .005), but not with episodic memory (R² = .142, β = .004, t(302) = .078, p = .938. There were no significant associations between caffeine and any cognitive factor in the younger age group.

Table 2.

Interactions of Caffeine and Age on Cognitive Functions, for All Participants, and Linear Associations of Caffeine With Cognition Median Age, Controlling for Demographic, Cardiovascular, Diabetes, Smoking, and ApoE Variables (linear contrast p values)

	Interaction of Caffeine With Age on Cognition	Entire Sample by Age Group
	All Participants N = 638	Youngest Group (age 64–71.5) N = 101	Oldest Group (age 71.5–84) N = 84
Overall cognition	R 2 = .375, β = .980, p = .066*	R 2 = .312, β = .033, p = .517	R 2 = .399, β = .151, p = .002**
Semantic categorization	R 2 = .292, β = 1.272, p = .025**	R 2 = .270, β = .015, p = .783	R 2 = .326, β = .163, p .001**
Attention/working memory	R 2 = .212, β = .305, p = .610	R 2 = .224, β = .101, p = .065	R 2 = .220, β = .148, p = .007**
Executive functioning	R 2 = .344, β = .759, p = .165	R 2 = .234, β = .034, p = .520	R 2 = .402, β = .135, p = .005**
Episodic memory	R 2 = .130, β = .573, p = .361	R 2 = .103, β = −.059, p = .311	R 2 = .142, β = .004, p = .938

Note: * trend level at >.10. **significant at .05 level.

We also assessed the association of caffeine with GM and WM in the participants with MRI (see Table 3). In this subsample, caffeine was neither associated with GM nor WM volumes. However, there was an interaction between age and caffeine for GM (R² = .716, β = 1.694, t(164) = 2.135, p = .034) such that in those below median age, higher caffeine intake was not associated with GM (R² = .743, β = −.064, t(66) = −1.024, p = .309) but in those above median age, higher caffeine intake was associated with greater GM (R² = .580, β = .198, t(65) = 2.177, p = .033). There was no interaction of caffeine intake with age on WM (R² = .861, β = −.422, t(164) = −.760, p = .448).

Table 3.

Interactions of Caffeine and Age on Brain Volume, for MRI Sample, and Linear Associations of Caffeine With Brain Volume by Median Age, Controlling for Demographic, Cardiovascular, Diabetes, Smoking, and ApoE Variables

	Interaction of Caffeine With Age on Brain Volume	MRI Sample by Median Age
	MRI Sample N = 185	Youngest Group (age 64–71.5) N = 101	Oldest Group (age 71.5–84) N = 84
Gray matter	R 2 = .716, β = 1.694, p = .034*	R 2 = .550, β = −.064, p = .309	R 2 = .580, β = .198, p = .033*
White matter	R 2 = .861, β = −.422, p = .448	R 2 = .865, β = .014, p = .745	R 2 = .866, β = −.032, p = .536

Note: MRI = Magnetic resonance imaging.

*Significant at .05 level.

To investigate whether T2D-related characteristics affected the results, we repeated the analyses controlling only for age, sex, and education and all results were very similar.

Discussion

In this sample of elderly individuals with T2D, age moderated the relationships of caffeine both with cognition and with GM, such that higher caffeine intake levels were significantly associated with better cognitive performance and greater GM volume in older T2D individuals but not in younger, after controlling for numerous relevant covariates. Caffeine intake had a small but significant association with overall cognition, working memory/attention, semantic categorization, and approached significance for executive functioning. No associations were found with WM.

Older age is associated with worsening cognitive functioning and greater risk of all dementia types (27). Our results are consistent with those of others showing an age-related association between caffeine and cognitive performance, such that the relationship was strongest in older individuals (3,4). Longer exposure to and accumulating benefits of caffeine (16), and higher sensitivity to caffeine in older individuals (17) have been proposed as potential underlying mechanisms behind the moderating effect of age on the association of caffeine with cognition. Finally, hypotension, associated with cognitive decline, is common among older individuals (28) and individuals with T2D (29). Caffeine may provide a slight rise in blood pressure without the risk of hypertension (30), thereby mitigating the impact of hypotension on cognition; this study included the mean of blood pressure measurements over several years as a covariate, and its inclusion did not alter the results. However, we do not have information on daily fluctuations of blood pressure. Finally, though all participants are nondemented, those in the older group do have lower cognitive functioning compared to the younger group. Within the older group, higher caffeine is associated with better cognitive functioning, however, this relationship does not exist in the younger group. It is possible that, due to stronger cognitive functioning in the younger group, a difference between those who use larger amounts of caffeine and those who do not cannot yet be seen.

The result of an association between caffeine and cognitive outcomes in this sample of individuals with T2D is strengthened by our finding of an interaction between age and caffeine on GM (such that higher levels of caffeine are associated with more GM volume in the older group of participants). Caffeine consumption has been found to reverse T2D-related neurological damage and prevent T2D-related memory impairment in mice (31) suggesting that long-term use of caffeine is potentially beneficial in reducing T2D- related neuropathology and cognitive loss. Our results of greater volume of GM in older subjects who have higher intake of caffeine are consistent with the reduction of neuronal loss with age by chlorogenic acids and polyphenols (8,32). As longer duration of T2D is associated with brain volume loss, particularly in the GM (19), the moderation of age on the relationship of caffeine with GM may suggest that as individuals with T2D age, caffeine may protect against T2D-related GM loss and therefore also against cognitive decline. Finally, it is important to recognize that caffeine is associated with cerebral blood flow (33). In this study, caffeine intake was not associated with WM volume, in line with findings of no relationship between WM loss and T2D (34).

Several mechanisms by which caffeine and coffee (one of the most popular caffeine-providing beverages) may be neuroprotective have been suggested. Caffeine appears to reduce beta-amyloid deposits in mice (35), potentially due to caffeine’s role as an adenosine antagonist (36). Caffeine also protects the blood brain barrier, improving efflux of beta-amyloid and thereby reducing the risk of AD (as well as Parkinson’s disease) (12,37). Finally, coffee contains polyphenols, specifically chlorogenic acid, an antioxidant, which may reduce inflammation and protect against dementia (32).

Caffeine has been shown to be associated with lower risk of dementia, AD, and cognitive decline (8,16,17) and with better cognitive functioning (32). Here, we show that the association between caffeine and cognition is relevant to T2D elderly adults, specifically older elderly adults. This study was limited by its cross-sectional outcomes. Longitudinal outcomes of the IDCD study are ongoing. We do not have information on whether caffeine intake was a constant habit for the participants. This is especially important in light of a recent longitudinal analysis of caffeine consumption, in which a constant, moderate caffeine consumption was associated with reduced risk of MCI, but an increase in caffeine consumption was associated with an increased risk (7). We cannot rule out that persons with incipient, preclinical dementia, decrease their caffeine intake (38), leading to an apparent association of higher caffeine intake with better cognition. Although there is often a decline in caloric intake prior to the onset of dementia (39), which might include caffeine intake, caloric intake was controlled for in this study. This study is strengthened by a large sample, strong validity for T2D diagnosis and cardiovascular variables, and a broad cognitive assessment, permitting examination of global and specific cognitive domains as well as by a subsample who had MRI so brain markers related to the potential neuroprotective role of caffeine could be examined. The study innovates by focusing on a sample of T2D patients, who are at high dementia risk. Further, we could not find studies examining the association of caffeine with GM volume. Our study did not include a non-T2D sample impeding conclusions regarding specific effects of T2D. However, results were similar with and without inclusion of T2D-related covariates, possibly suggesting that elderly adults without T2D might benefit from caffeine. The range of caffeine intake in the IDCD sample was from 0 to 626.5 mg, averaging 205.2 mg, with very few participants’ intake above recommended values. These levels are consistent with those of U.S. samples of comparable ages (40). There were few participants with very high caffeine intake levels; this impeded testing of whether very high levels may deleteriously (or beneficially) affect cognition, however, a study investigating different levels of caffeine found a “lack of a distinct dose-response association (6)”. Based on Fulgoni, Keasit, and Lieberman (2015) (40), a high caffeine intake dose was defined in our paper as 439 mg/d (90th percentile, (40)). Twenty-seven participants in this study reported a daily intake of caffeine above 439 mg/d. Caffeine is an inexpensive, readily available drug, so understanding its potential role in cognition and GM volume, and how age moderates it, may suggest preventive strategies against dementia.

Funding

This study was supported by the National Institutes of Health (Grant/Award Number: R01 AG034087/AG/NIA NIH HHS), Alzheimer’s Association (Grant/Award Number: NIRG-11-205083 2012), the American Federation for Aging Research, The Bader Philanthropies, and the LeRoy Schecter Foundation.

Author contributions

R.K.W. conducted statistical analysis and wrote manuscript. R.R-S. researched data, reviewed/edited manuscript. A.L. reviewed/edited manuscript. A.H. researched data, reviewed/edited manuscript. D.S. extracted caffeine intake data, reviewed the manuscript. D.L. reviewed/edited manuscript and contributed to discussion. R.P. collected data and reviewed the manuscript. R.Z. collected data and reviewed the manuscript. J.M.S. reviewed/edited manuscript. M.S-B. designed the study, performed the experiments, analyzed and researched data, and reviewed/edited the manuscript.

Conflict of Interest

None reported.