Relationship between Arterial Hypertension with Cognitive Performance in Elderly. Systematic Review and Meta-Analysis

Background: Previous systematic reviews report that arterial hypertension (AHT) is associated with lower performance in cognition in the elderly. However, some studies show that with higher blood pressure, a better cognitive performance is obtained. Objective: The aim of this study was to determine the relationship between AHT with cognitive performance in the elderly. Methods: the review involved a search on PubMed, Scopus and PsycINFO databases from January 1990 to March, 2020 to identify the relationship among AHT and cognitive performance in older people. Results: 1170 articles were identified, 136 complete papers were reviewed, a qualitative analysis of 26 studies and a quantitative analysis of eight studies were carried out. It was found that people with AHT have a lower performance in processing speed SMD = 0.40 (95% CI: 0.25, 0.54), working memory SMD = 0.28 (95% CI: 0.15, 0.41) in short-term memory and learning SMD = −0.27 (95% CI: −0.37, −0.17) and delayed recall SMD = −0.20 (95% CI: −0.35, −0.05). Only one study found that higher blood pressure was associated with better memory performance. Conclusion: Our results suggest that high blood pressure primarily affects processing speed, working memory, short-term memory and learning and delayed recall.


Introduction
Regular blood pressure is 120/80 mmHg. When it gets to 140/90 or more in a chronic manner, it is considered as high blood pressure or arterial hypertension (AHT) [1]. The American Heart Association and the American Heart College have proposed to reduce this diagnostic criterion to 130/80 mmHg [2]; however, the relevance for Latin America is still pending [3].
It is estimated that there are 1130 million people worldwide with AHT [1]. It is associated with other diseases like diabetes [4], cancer [5], psychosocial stress [6], dementia or cognitive impairment [7]. It has such relevance that AHT has been considered by a group of experts as the main modifiable factor from middle age on to prevent dementia or cognitive impairment [8].
Hypertension is associated with anatomical and physiological changes which harm the brain. AHT increases by atherosclerosis and stiffness in blood vessels. Atherosclerosis is the thickening of arteries produced by the accumulation of fat, cholesterol and other substances, thereby, decreasing the blood vessels' internal diameter. Additionally, it favors clot formation or thrombogenesis. Such stiffness is triggered by the blood vessels endothelium's hypertrophy, increasing collagen and fibronectin deposition. All the above explains why hypertension is one of the main factors associated to stroke and dementia [9,10]. Blood vessels' rigidity is a predictor of vascular events [11] and it generates mini strokes in basal ganglia and white matter arterioles, therefore, causing the so-called small vessel disease [12,13]. In addition, AHT is a risk factor for atrial fibrillation, a condition that increases the probability of developing cognitive impairment [14]. 2020 through Pubmed, Scopus and PsycINFO data bases with the search key words: "hypertension" OR "blood pressure" AND "cognition" AND "older".

Eligibility Criteria
For the studies to be included, they needed to fulfill the predefined requirements of PEO: Population, Exposition and Outcome. For this systematic revision we included the studies with the following designs: cross-sectional study, cohort study, case-control study; with a population between 50 and 80 years old; with one or more groups with hypertension and with a group or population without hypertension to assess the effect of AHT. Furthermore, the studies had to use at least two valid neuropsychological instruments to measure cognitive functions. We excluded systematic revision articles, case studies and basic research studies with animals. We also excluded studies with a population with a neurological or psychiatric disease (cerebrovascular event, dementia, cognitive impairment, depression, etc.) or some other disease affecting cognition in a significant manner like cancer, AIDS and renal insufficiency. Finally, we also excluded articles which only used a screening test to evaluate cognitive functions, for example, the Folstein Mental State mini test or the Montreal Cognitive Assessment.

Article Selection
An independent article revision was performed by the authors (JM S-N and UD R-S) at every stage, considering inclusion/exclusion criteria. We used Excel software to keep track of the revision process. When disagreement occurred among both authors, a third author (VM M-N) was involved to participate in the discussion. First, we eliminated repeated articles. Next was the selection of articles based on the title and abstract. Subsequently, the full articles were reviewed, and the articles were selected for qualitative synthesis and meta-analysis.

Analysis and Data Synthesis
The authors independently registered study type, sample characteristics (size, male/female ratio, age, scholarship and blood pressure, their hypertension definition and main results). Subsequently, we compared the compiled data and, in case of disagreement, the article was reviewed again.
The author JM S-N classified the tests and implemented tasks in each study, using the following categories: executive functions, work memory, processing speed, cognitive inhibition, short-term memory learning, delayed memory, and reasoning. We selected these processes because they are the ones most affected during aging [22,24].
The Newcastle-Ottawa Scale (NOS) was used to assess the quality case-control studies in the meta-analysis [30]. When the interventions and associated outcomes were assessed as sufficiently homogeneous and when sufficient information was available from the studies, quantitative data were pooled in the Review Manager (Version 5.3, The Cochrane Collaboration 2014) for meta-analysis. In this sense, in the event that the group with hypertension was divided by some characteristic, the scores were calculated to obtain only a mean score, standard deviation and sample size. The meta-analysis of these values was performed using the random effects model. The I 2 statistic was used to assess inconsistencies between studies and describe the percentage of variability in effect. Heterogeneity was considered substantial if the I 2 statistic was ≥50%. All effect sizes were calculated using standardized mean differences (SME), as all studies used a wide variety of scale measures.

Studies Selection
In the initial search strategy, we found 1169 articles, mainly on PubMed (After removing the duplicated publications), 1033 articles were rejected according in their titles and abstracts. 136 studies were reviewed, and out of these we excluded 37 due to differences regarding their methodology. Additionally, we excluded another 67 because those only used one screening instrument, and six were disqualified for other reasons. Upon completion, we ended up with 26 articles which met the eligibility criteria, of which 8 were included in the meta-analysis ( Figure 1).

Studies Characteristics
We selected 26 studies, of which six were cross-sectional, nine were cohort and 11 were case-control ( Table 1). The sample size of patients with hypertension ranged from 12 [31] to approximately 3200 [32]. The age of the participants was mostly around 70 years ( Table 1). The percentage of men varied from 7% [33] to 69.4% [34]. Some studies perform a division by gender [35,36], and in five we did not find information on the composition by gender [37][38][39][40][41]. Most of the studies included participants with eight or more years of education (Table 1), in only two studies did most of the participants have less than eight years of education [42,43].  The definition of hypertension was by medical diagnosis or by being prescribed medications for high blood pressure (Table 1). In three of them we did not find the criteria to define arterial hypertension [41,44,45]. The cut-off points of systolic pressure to define hypertension varied in the studies, one of them was greater than 130 mmHg 3200 [46], five were greater than 140 mmHg [35,36,38,39,42], and seven studies were greater than 160 mmHg [32,40,43,[47][48][49][50].
Eleven case-control studies were found without risk of bias and with adequate quality to perform a quantitative analysis (Appendix B). In some studies, they divided the sample with hypertension into two groups: controlled blood pressure (systolic pressure < 140 mmHg) or out of control [38,51], or into three groups: controlled blood pressure, out of control, and without treatment [39,42]

Relationship of Cognitive Performance with Arterial Hypertension
In most studies it is found that the higher the blood pressure or hypertension, the lower the cognitive performance. Only in one study was it found that higher the diastolic pressure contributed to a better cognitive performance [47]. In three studies it was found that a low or high blood pressure is related to a lower cognitive performance [32,45,51] and in two studies no relationship was found between hypertension and cognitive performance [52,53] ( Table 1).

Processing Speed
The processing speed was tested in 17 studies (Table 2). Seven studies found that higher blood pressure is associated with lower performance on processing speed tests [31,[36][37][38]41,48,54]. Two studies found a relationship between lower performance in cognitive tests with lower diastolic pressure in people who do not receive treatment [51], and systolic in participants who suffered myocardial infarction [32].  Four studies with a total of 759 participants with hypertension and 771 controls were included in the processing speed meta-analysis ( Figure 2). The results were analyzed from the digit and symbol substitution tests [37,46], trail making test A [31] and a measure composed of several tests [48]. In this indicator, a higher score indicates a lower cognitive performance. Two studies had subdivisions of the group with hypertension: one with several treatments [37], and another in untreated, uncontrolled and controlled [31], for which they were united, forming three groups; controlled hypertensive, uncontrolled hypertensive, and the control group.

Working Memory
The working memory was evaluated in 13 studies, mainly with the Trail Making Test B and Digit Span Backwards (Table 2). Six studies found that higher blood pressure is associated with lower performance on working memory tests [31,[36][37][38]40,48,51]. One study found that low diastolic pressure is associated with lower performance on working memory tests, particularly in the low-educated population [51].
Three studies with a total of 481 participants with hypertension and 499 without hypertension were included in the meta-analysis of working memory (Figure 3). The results used were from the Trail Making Test B [31,37,52]. In this indicator, a higher score indicates a lower performance. The studies divided the groups into ingesting different drugs [37] and untreated, uncontrolled and controlled [31], for which they were united forming three groups: uncontrolled hypertension, controlled hypertension, and the control group. The group with uncontrolled hypertension compared to the control group had a lower performance in working memory SMD = 0.28 (95% CI: 0.15, 0.41; I2 = 0%; p = 0.65; n = 3 studies; n hypertension = 469; n control = 499). The group with controlled hypertension compared to the control group had a better performance SMD = −0.39 (95% CI: −1.02, 0.23; p = 0.1 n = 1 studies; n hypertension = 12; n control = 57), but it is not statistically significant.
Six studies with a total of 798 participants with hypertension and 819 without hypertension were included in the meta-analysis of short-term memory and learning (Figure 4). The results used were the California Verbal Learning Test (CVLT) [52], the Selective Reminding Test (SRT) [55], the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) [39], the Hopkins Verbal Learning Test (HVLT) [37], and the Logical Memory test [31,36]. Several studies had subdivisions in the hypertension group: with and without subjective memory problems [55], ingesting different medications [37], in men and women [36], untreated, uncontrolled and controlled [31,39]; reason why they were united forming three groups; controlled hypertensive, uncontrolled hypertensive and control group.

Delayed Memory
Delayed memory was included in 15 studies, and the time variation was from 2 to 30 min (Table 2). In two, higher pressure was found to be related to lower performance in delayed memory [31,32]. In one study it was found that lower diastolic pressure is related to lower performance in delayed memory, particularly in people who have suffered a stroke [32]. A study found that higher pressure is related to better performance in delayed memory in men [47].
Five studies with a total of 355 participants with hypertension and 439 without hypertension were included in the delayed memory meta-analysis ( Figure 5). The indicators used were CVLT delayed memory [52], SRT [55] CERAD [39], WISC, and logical memory [31,36]. Several studies had subdivisions in participants with hypertension: with and without subjective memory problems [55], in men and women [36], untreated, uncon-trolled and controlled [31,39], for which they were united forming three groups: controlled hypertensive, uncontrolled hypertensive and control group.

Other Cognitive Processes
Executive functions were assessed in 12 of the 26 studies (Table 2). In seven studies, higher blood pressure was found to be related to lower cognitive performance [34,35,42,48,51,52,56]. In one study, low and high diastolic pressure were found to be related to lower performance in executive functions [46].
Cognitive inhibition is only assessed in two studies using the Stroop interference task ( Table 2). No significant difference was found between hypertensive and normotensive participants [31,52].
Reasoning was evaluated in seven studies (Table 2). In four studies, it was found that higher blood pressure is related to lower performance in reasoning tests [50,52,54,55].

Discussion
The present review was carried out with the objective of determining the relationship between high blood pressure and cognitive performance in older adults. The tests used were grouped into cognitive processes that decline during aging [22,24]. It was found through a quantitative analysis that higher blood pressure affects processing speed, working memory, short-term memory, and delayed memory.
The observed results in a qualitative analysis suggest that elderly people with hypertension have a lower level of performance than old people without this condition. This applied to executive function (seven out of 12 studies), work memory (six out of 13 studies), processing speed (seven out of 17 studies), short-term memory (nine out of 23 studies) or delayed memory (two out of 15 studies) and reasoning (four out of 7 studies). Nonetheless, in cognitive inhibition there is a need to perform more studies to be able to suggest or rule out an effect.
Our results contrast with another systematic revision made by Van den Berg, et al. [27], who concluded that the most affected process in hypertension patients is memory. In their revision, they performed a quantitative analysis where two articles included mild cognitive impairment patients [57,58], a condition which mainly affects memory [59].
In this sense, our revision has two main differences with the one made by Van de Berg et al. (2009) [27]. We excluded studies which integrated patients with mild cognitive impairment and we divided memory in two categories: short-term and delayed memory. The first one is related with processes like attention, while delayed memory is related to storage, which is an indicator of the hippocampus's integrity [60]. This division was relevant, because it allowed us to show that in most studies there is a difference among AHT patients and people without this disease.
Most of the studies found reported that AHT is related to lower cognitive performance; however, in two studies a paradoxical effect is found. One study reported that higher diastolic pressure in men is related to higher cognitive performance [47], the other that the group with controlled AHT had higher performance compared to the control group [31], which was corroborated in the meta-analysis carried out in our study. One explanation for the paradoxical effect may be due to increased perfusion in the brain, which benefits from thickening of the arteries [61]. Another possible explanation may be an effect of the medications. In a systematic review, it was found that angiotensin II receptor blockers can prevent cognitive deterioration in people with AHT [62], and that these could have a positive effect on cognition. However, to corroborate the above it is necessary to carry out more research in this regard.
In four studies it was found that people with lower diastolic blood pressure had a lower performance, mainly in working memory and processing speed [32,45,51], particularly in people with less education [51]. Also, a lower working memory performance was found in people with low systolic pressure, especially if they had a myocardial infarction [32].
This is consistent with studies that report that both high and low blood flow can cause damage to the brain. On the one hand, hypertension can cause cerebrovascular accidents and lesions in the white matter [63]. On the other hand, low blood pressure may be associated with ischemic injuries [64,65]. Furthermore, very low blood pressure has been associated with the presence of Alzheimer's disease 60. Considering the above, it would be necessary to maintain homeostatic blood pressure to preserve cognitive functioning.
A limit point is used to diagnose AHT, which causes significant variation. The value most frequently used is a blood pressure higher than 140/90 mmHg [66]. Another higher value set is 160/95 mmHG [67] or lower at 130/80 mmHg [2]. The latter can also explain the AHT effect on cognitive functions and inconsistencies in the different studies. For example, in this present study we did not find a significant difference among people with AHT and patients without this condition. Regarding work memory and processing speed, in one article authors used systolic pressure > 160 mmHg as diagnostic criteria [42]. Consequently, people with lower blood pressure might have already been affected, which would nullify the differences. On the other hand, in a different study, AHT was diagnosed with a blood pressure of 130 mmHg [46]. Thus, there could be people with these blood pressure levels who are not experiencing any impact on their cognitive functions.
Another relevant criterion is the previous diagnosis. For instance, people with AHT who maintain a controlled blood pressure by taking medication like angiotensin II receptor blockers can experience a lower impact on cognitive functions [62]. However, it is necessary to include the blood pressure criteria because if only the previous diagnosis is considered, people with a systolic pressure > 140 mmHg could be included in the control group [57]. That being the case, it would be better to divide AHT patients in at least three groups: controlled, uncontrolled with treatment and uncontrolled without treatment, besides the control group. [68,69] We found that systolic pressure is related with a lower cognitive performance. In this sense, similar findings have been pointed out and reported in reviews narrative and systematic [64,65]. Another relevant indicator could be the pulse pressure, which is the difference between systolic and diastolic pressure, as whichever is higher might negatively affect cognitive functions [38,42,52].
The instruments used to measure cognitive functions are quite relevant to measure the effect of AHT as well. The most used ones in which there was a significant difference were the Trail Making Test A and B [36][37][38] and the digit-symbol substitution test [37,46,52]. Meanwhile, other instruments showed no difference, the work memory digits task being one example [31,36,39].
The National Institute of Neurologic Disorders and the Canadian Stroke Network have proposed three different evaluation protocols with AHT patients; the differences lie within the time (60 min, 30 min and 5 min) in which the tests mentioned above are included [70]. It is likely that these tests are sensitive enough to detect changes caused by AHT without structural damage; however, more research needs to be done to verify this.
One of the most relevant characteristics in the sample is the education level. In our revision, only two studies had a population with less than eight years of education [42,43]. Education is the main factor associated with cognitive reserve; this term is defined as a person's capacity that if brain damage occurs, they can live with an adequate cognitive performance [71]. Therefore, the effects caused by an alteration in blood pressure may be greater; [51] however, more research is needed on this.
Our revision presents certain limitations: we failed to include studies with people older than 80 years old, and we did not revise grey literature nor the previously registered protocols. It would be advisable to carry out more studies analyzing other variables such as age, sex, follow-up time and academic level, comorbidities (atrial fibrillation and other arrhythmias, diabetes mellitus), medications and lifestyle (exercise, nutrition, sleep, smoking, and alcohol intake).

Conclusions
Our findings suggest that older people with AHT present a lower performance than old people without this condition, mainly in terms of processing speed, work memory, and memory. However, more studies with AHT patients need to be done, in both controlled and uncontrolled treatments. Furthermore, instruments sensitive to cognitive functions caused by AHT changes are essential in these kinds of studies, as is the capacity to identify the right AHT limit point (systolic or heart rate) where a change in cognitive functions is presented. Funding: This research was supported by Dirección General de Asuntos del Personal Académico, Universidad Nacional Autónoma de México (DGAPA-UNAM) (PAPIME PE203421); Programa para la Investigación Bibliográfica Científica sobre Salud (PIBCIS) de la FES Zaragoza, UNAM; Red Académica Asesora de Revisiones Sistemáticas (RAARS).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.