Pharmacological treatment of hypertension in people without prior cerebrovascular disease for the prevention of cognitive impairment and dementia

Abstract

Background

This is an update of a Cochrane Review first published in 2006 (McGuinness 2006), and previously updated in 2009 (McGuinness 2009). Hypertension is a risk factor for dementia. Observational studies suggest antihypertensive treatment is associated with lower incidences of cognitive impairment and dementia. There is already clear evidence to support the treatment of hypertension after stroke.

Objectives

To assess whether pharmacological treatment of hypertension can prevent cognitive impairment or dementia in people who have no history of cerebrovascular disease.

Search methods

We searched the Specialised Register of the Cochrane Dementia and Cognitive Improvement Group, CENTRAL, MEDLINE, Embase, three other databases, as well as many trials registries and grey literature sources, most recently on 7 July 2020.

Selection criteria

We included randomised controlled trials (RCTs) in which pharmacological interventions to treat hypertension were given for at least 12 months. We excluded trials of pharmacological interventions to lower blood pressure in non‐hypertensive participants. We also excluded trials conducted solely in people with stroke.

Data collection and analysis

Two authors independently assessed trial quality and extracted data. We contacted study authors for additional information. We collected information regarding incidence of dementia, cognitive decline, change in blood pressure, adverse effects and quality of life. We assessed the certainty of evidence using GRADE.

Main results

We included 12 studies, totaling 30,412 participants, in this review. Eight studies compared active treatment with placebo. Of the four non‐placebo‐controlled studies, two compared intensive versus standard blood pressure reduction. The two final included studies compared different classes of antihypertensive drug. Study durations varied from one to five years.

The combined result of four placebo‐controlled trials that reported incident dementia indicated no evidence of a difference in the risk of dementia between the antihypertensive treatment group and the placebo group (236/7767 versus 259/7660, odds ratio (OR) 0.89, 95% confidence interval (CI) 0.72 to 1.09; very low certainty evidence, downgraded due to study limitations and indirectness).

The combined results from five placebo‐controlled trials that reported change in Mini‐Mental State Examination (MMSE) may indicate a modest benefit from antihypertensive treatment (mean difference (MD) 0.20, 95% CI 0.10 to 0.29; very low certainty evidence, downgraded due to study limitations, indirectness and imprecision).

The certainty of evidence for both cognitive outcomes was downgraded on the basis of study limitations and indirectness. Study durations were too short, overall, to expect a significant difference in dementia rates between groups. Dementia and cognitive decline were secondary outcomes for most studies. Additional sources of bias include: the use of antihypertensive medication by the placebo group in the placebo‐controlled trials; failure to reach recruitment targets; and early termination of studies on safety grounds.

Meta‐analysis of the placebo‐controlled trials reporting results found a mean change in systolic blood pressure of ‐9.25 mmHg (95% CI ‐9.73, ‐8.78) between treatment (n = 8973) and placebo (n = 8820) groups, and a mean change in diastolic blood pressure of ‐2.47 mmHg (95% CI ‐2.70, ‐2.24) between treatment (n = 7700) and placebo (n = 7509) groups (both low certainty evidence downgraded on the basis of study limitations and inconsistency).

Three trials ‐ SHEP 1991, LOMIR MCT IL 1996 and MRC 1996 ‐ reported more withdrawals due to adverse events in active treatment groups than placebo groups. Participants on active treatment in Syst Eur 1998 were less likely to discontinue treatment due to side effects, and participants on active treatment in HYVET 2008 reported fewer 'serious adverse events' than in the placebo group. There was no evidence of a difference in withdrawals rates between groups in SCOPE 2003, and results were unclear for Perez Stable 2000 and Zhang 2018. Heterogeneity precluded meta‐analysis.

Five of the placebo‐controlled trials provided quality of life (QOL) data. Heterogeneity again precluded meta‐analysis. SHEP 1991, Syst Eur 1998 and HYVET 2008 reported no evidence of a difference in QOL measures between active treatment and placebo groups over time. The SCOPE 2003 sub‐study (Degl'Innocenti 2004) showed a smaller drop in QOL measures in the active treatment compared to the placebo group. LOMIR MCT IL 1996 reported an improvement in a QOL measure at twelve months in one active treatment group and deterioration in another.

Authors' conclusions

High certainty randomised controlled trial evidence regarding the effect of hypertension treatment on dementia and cognitive decline does not yet exist.

The studies included in this review provide low certainty evidence (downgraded primarily due to study limitations and indirectness) that pharmacological treatment of hypertension, in people without prior cerebrovascular disease, leads to less cognitive decline compared to controls. This difference is below the level considered clinically significant. The studies included in this review also provide very low certainty evidence that pharmacological treatment of hypertension, in people without prior cerebrovascular disease, prevents dementia.

Plain language summary

Treating high blood pressure with medicines to prevent future dementia or problems with memory and thinking

Why this review is important

Observational studies following participants over many years have suggested that high blood pressure is associated with subsequent dementia or new problems with memory and thinking (cognitive impairment). It is important to establish if treating high blood pressure can reduce the risk of dementia and problems with memory and thinking. There is already clear evidence to support the treatment of high blood pressure after stroke. The evidence for treating high blood pressure in the absence of stroke was not as well established when this review protocol was written.

Main findings of the review

We included 12 trials, with a total of 30,412 participants, in this review. Treating high blood pressure with medicines may reduce the risk of dementia and problems with memory and thinking, but we were unable to confirm this using the data we have available. This may be due to shortcomings in the evidence available.

Some participants had side effects from medications, such as falls, that led to them stopping the medications. This is also an important consideration.

Limitations of the review

We were unable to find evidence to confirm that treating high blood pressure with medicines will prevent future dementia or problems with memory and thinking. Based on the evidence available, all we can say is that the methods used in the studies we looked at were not sufficient to answer our study question. In practice, we want to know if treating high blood pressure in mid‐life reduces dementia and problems with memory and thinking in later life. These studies were too short in duration to answer this question. They tended to be designed to investigate heart attack and stroke as outcome measures, with memory and thinking tests added on, which means they were less likely to include the numbers of people with problems with memory and thinking necessary to reliably answer the question. In the studies comparing active medicines and placebo (inactive medicine), many people in the placebo groups ultimately did receive active medicine which further complicated the results.

Antihypertensive treatment may prevent new problems with memory and thinking, or dementia, or both, despite the findings of this review. To be more confident in our conclusions, we would need studies that have more distinct treatment and placebo groups, and that treat participants starting at a younger age and with a longer follow‐up.

Summary of findings

Summary of findings 1. Antihypertensive therapy compared to placebo for patients without prior cerebrovascular disease for prevention of cognitive impairment and dementia.

Antihypertensive therapy compared to placebo for people without prior cerebrovascular disease for prevention of cognitive impairment and dementia
Patient or population: participants without prior cerebrovascular disease for prevention of cognitive impairment and dementia Setting: community Intervention: antihypertensive therapy Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with antihypertensive therapy
Number of cases of dementia Range of follow‐up 2 to 5 years	Study population		OR 0.89 (0.72 to 1.07)	15,427 (4 RCTs)	⊕⊝⊝⊝ VERY LOWa,b,c	‐
	34 per 1000	30 per 1000 (25 to 36)
Change in cognitive score (MMSE) Range of follow‐up 2 to 5 years	The mean change in MMSE ranged from ‐0.15 to ‐0.08	The mean change in MMSE was 0.20 points higher (0.10 points higher to 0.29 points higher)		9435 (4 RCTs)	⊕⊕⊝⊝ VERY LOWa‐d	‐
Change in systolic blood pressure level (mmHg) Range of follow‐up 2 to 5 years	The mean change in systolic blood pressure level (mmHg) ranged from ‐33.5 to ‐15.77	The mean change in systolic blood pressure level (mmHg) was 9.25 points lower (9.73 points lower to 8.78 points lower)		17,793 (6 RCTs)	⊕⊝⊝⊝ LOWc,e	‐
Change in diastolic blood pressure level (mmHg) Range of follow‐up 2 to 5 years	The mean change in diastolic blood pressure level (mmHg) ranged from ‐13.1 to ‐4.21	The mean change in diastolic blood pressure level (mmHg) was 2.47 points lower (2.70 points lower to 2.24 points lower)		15,209 (5 RCTs)	⊕⊝⊝⊝ LOWc,e	‐
Adverse effects Range of follow‐up 2 to 5 years	Results varied across studies, with some reporting evidence of more withdrawals due to adverse effects in the antihypertensive therapy group (SHEP 1991, LOMIR MCT IL 1996, MRC 1996), some reporting evidence of fewer withdrawals in the antihypertensive therapy group (Syst Eur 1998, HYVET 2008), and one reporting no evidence of a difference between groups (SCOPE 2003)			16,830 (6 RCTs)	⊕⊝⊝⊝ VERY LOWb,c,e	Heterogeneity precluded meta‐analysis
Quality of life measured via Sickness Impact Profile, PGWB Positive well‐being, and EuroQoL Current Health Range of follow‐up 2 to 5 years	Two studies reported evidence of a difference between the active treatment and placebo groups (LOMIR MCT IL 1996: P = 0.03; SCOPE 2003: PGWB Positive well‐being = ‐0.8 vs ‐1.1, P = 0.04, and EuroQoL Current Health = ‐3.1 vs ‐5.3, P = 0.008). Three studies reported no evidence of a difference between groups (SHEP 1991, Syst Eur 1998, HYVET 2008).			15,795 (5 RCTs)	⊕⊝⊝⊝ VERY LOWb,c,e	Heterogeneity precluded meta‐analysis
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; OR: odds ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect

^aDowngraded 1 level due to indirectness – studies not powered to detect significant differences in dementia rates or rate of cognitive decline as measured using MMSE between groups. Duration of studies likely too short to detect enough cognitive impairment and dementia in participants of the age studied. Conversely, younger participants may be more likely to benefit long‐term cognitively from prolonged antihypertensive treatment.
^bDowngraded 1 level due to study limitations – likely attrition bias as cognitively impaired and frailer participants more likely to be lost to follow‐up.
^cDowngraded 1 level due to study limitations – proportions of participants in placebo group receiving active treatment, particularly in SCOPE 2003 and Zhang 2018, means there is a high risk of contamination bias.
^dDowngraded 1 level due to imprecision – this is below the statistically significant threshold suggested by Prince 2014 and the clinically significant threshold suggested by Lopez 2005 and Jones 2009.
^eDowngraded 1 level due to inconsistency ‐ significant heterogeneity. Number of possible reasons, baseline BP varied between studies, different antihypertensive drugs were used, large proportions of participants in placebo groups, especially in SCOPE 2003 and Zhang 2018, received active treatment.

Background

This review updates and replaces a previous Cochrane Review, 'Blood pressure (BP) lowering in patients without prior cerebrovascular disease for prevention of cognitive impairment and dementia', which was first published in 2006 (McGuinness 2006), and updated in 2009 (McGuinness 2009). These iterations found no convincing evidence that BP lowering prevents the development of dementia or cognitive impairment in hypertensive people with no apparent prior cerebrovascular disease.

Description of the condition

The relationship between BP, cognition and dementia is complex and not yet fully established. Analysis of the epidemiological data must take into account the following methodological issues. Firstly, longitudinal studies can suggest whether a causal relationship exists between BP and cognition or dementia, whereas cross‐sectional studies can not determine causality. Secondly, the definitions of high or low BP and cognitive impairment are heterogeneous across the studies and may limit comparability. The methods used to screen for and to assess cognitive impairment are also heterogeneous. However, the diagnosis of dementia, and its aetiology, do follow established, standardised criteria. Thirdly, studies may vary in controlling for the effect of other vascular risk factors or antihypertensive treatment. Nevertheless, analysis of the available epidemiological data does allow some conclusions to be drawn: (1) elevated BP in mid‐life is associated with cognitive impairment and dementia in later life; (2) treatment of elevated BP in mid‐life would appear to mitigate this risk; and (3) the relationship between BP and cognitive impairment and dementia in later life is more complex.

Several longitudinal studies (summarised by Iadecola 2016 and Walker 2017) have reported a consistent relationship between elevated BP in mid‐life (age 40 to 64 years) and cognitive impairment in later life (age 65 years or older) (Elias 1993; Swan 1998; Kilander 2000; Launer 2000; Elias 2004; Gottesman 2014). This effect was more marked if the high BP was not treated (Elias 1993; Kilander 2000; Gottesman 2014).

Several of the same data sets also indicated a relationship between mid‐life hypertension and the incidence of dementia and Alzheimer's disease in later life (Launer 2000; Kivipelto 2001; Wu 2003; Whitmer 2005). In the Honolulu‐Asia Aging Study, this association was only present in those participants not treated with antihypertensive agents (Launer 2000). One Japanese study reported a relationship of mid‐life high BP to later life vascular dementia but not to Alzheimer's disease (Yamada 2003). A systematic review and meta‐analysis of longitudinal studies reported an association between hypertension and both incident and prevalent vascular dementia (Sharp 2011).

The available data regarding later life BP and cognition or dementia are inconsistent (Birns 2009, Rouch 2015). Some studies showed no relationship between high BP and cognitive impairment (Hebert 2004; Tervo 2004; Gottesman 2014), or dementia (Yoshitake 1995; Brayne 1998). Other studies did find a relationship between later life high BP and cognitive decline (Elias 1993), or dementia (Skoog 1996). Low BP in later life has been more consistently associated with cognitive impairment and dementia (Guo 1996; Morris 2001; Ruitenberg 2001; Bohannon 2002; Qiu 2003; Verghese 2003; Hebert 2004; Qiu 2004; Waldstein 2005).

Description of the intervention

Randomised controlled clinical trial (RCT) evidence has demonstrated a reduction in cardiovascular and cerebrovascular morbidity and mortality associated with treatment of hypertension across the life course (EWPHE 1985, SHEP 1991, STOP 1991, MRC 1996, Syst Eur 1998, HYVET 2008, SPRINT MIND 2019). Both non‐pharmacological and pharmacological approaches have been shown to be effective. Current guidance in the United Kingdom (UK) recommends encouragement of a healthy diet, regular exercise and reduced alcohol, caffeine and sodium consumption as the first line intervention in people with high BP. UK guidelines recommend pharmacological antihypertensive treatment for people under 80 years old with BP above 140/90 mmHg (or above 150/90 mmHg for those over 80 years) (NICE 2019).

The 2017 American College of Cardiology (ACC) and American Heart Association (AHA) guidelines and the 2018 European Society of Cardiology (ESC) guidelines recommend a pharmacological antihypertensive treatment threshold of BP over 130/80 mmHg where cardiovascular risk is high, and BP over 140/90 mmHg where cardiovascular risk is low (Whelton 2018; Williams 2018). The ESC have further age‐specific advice recommending pharmacological treatment of systolic blood pressure (SBP) 140 to 159 mmHg in fit people over 65 years but younger than 80 years where treatment is tolerated. The ACC/AHA recommend a target SBP of below 130 mmHg in fit people over 65 years, and advocate practitioner discretion for people over 65 years with higher levels of comorbidity. The ESC recommends a target SBP of 120 to 129 mmHg for people younger than 65 years, and a target SBP of 130 to 139 mmHg for those older than 65 years and, where tolerated, for those older than 80 years. They recommend a target diastolic blood pressure (DBP) of below 80 mmHg across the board.

Recommended first line antihypertensive agents are calcium channel blockers (CCBs) for people aged over 55 years or of African‐Caribbean origin without type 2 diabetes, and angiotensin‐converting enzyme (ACE) inhibitors or angiotensin receptor blockers (ARBs) for everyone else. Combination therapy with the alternative group and the addition of thiazide‐like diuretics is advocated thereafter (NICE 2019).

As stated, treatment of high blood pressure has been reported in several observational studies to be protective against cognitive impairment and dementia (Guo 1999; Qiu 2003a; Yasar 2005; Khachaturian 2006; Gottesman 2014). It is unclear whether individual drug classes confer a benefit beyond that of blood pressure lowering, where that association has been found. A systematic review of CCB use and cognitive decline in the elderly did not find evidence to support either a protective or detrimental association between their use in the elderly and subsequent cognitive function (Peters 2014). A meta‐analysis by Levi Marpillat 2013 suggested ARBs may carry an additional benefit but was not able to draw conclusions.

How the intervention might work

The pathological processes through which hypertension might influence cognitive function are numerous. A position paper on the subject by the American Society of Hypertension suggested that these mechanisms may be divided into five classifications (Gorelick 2012): functional (relating to regulation of cerebral blood flow and the blood‐brain‐barrier); structural (relating to increased white matter lesions, cerebral atrophy and the amyloid plaques and neurofibrillary tangles composed of tau that are neuropathological hallmarks of Alzheimer's disease); pharmacological (relating to the role of angiotensin II); stroke‐related (relating to stroke in strategic areas or the neurone loss associated with stroke, or both); and other factors (including the metabolic syndrome and inflammation, age, genetic predisposition and smoking). In vitro, animal and observational studies are ongoing in an effort to better understand these mechanisms. These processes are interrupted by pharmacological interventions proven to reduce blood pressure. Interested readers are also directed towards the more recent scientific statement from the American Heart Association on the impact of hypertension on cognitive function (Iadecola 2016).

Why it is important to do this review

This is an update of a Cochrane Review first published in 2006 (McGuinness 2006), and previously updated in 2009 (McGuinness 2009), which found no convincing evidence that BP lowering prevents the development of dementia or cognitive impairment in hypertensive people with no apparent prior cerebrovascular disease.

The aim of the original systematic review was to determine the effects of BP lowering on development of cognitive decline and dementia in people with hypertension but no history of stroke or transient ischaemic attack (TIA). Given the burden of both hypertension and dementia, identification of treatments that reduce the rate of cognitive decline or incidence of dementia would be of considerable clinical, public health and societal benefit.

This update is important to ensure that relevant data published since the first review and subsequent update are included in feasible meta‐analyses, allowing, where possible, the best possible estimate of the effect of BP lowering on cognitive decline to be obtained. The emergence of landmark trials such as ACCORD MIND 2014 and SPRINT MIND 2019 have strengthened the case for an update.

The previous review could have included both pharmacological and non‐pharmacological interventions, but in practice included only double‐blind, placebo‐controlled trials of antihypertensive medications. We have excluded non‐pharmacological interventions which are often complex in nature. These will be dealt with in a separate review.

Given the evidence to support treatment of hypertension and the change in guidelines since the initial review was planned, further placebo‐controlled trials would be unethical. For this update, we have therefore broadened the search criteria to include any RCTs in which pharmacological treatment for hypertension was administered for at least 12 months and cognitive outcomes were measured. It now also includes RCTs comparing different antihypertensive drugs or different BP targets. We dropped the requirement that trials had to be double‐blind because we did not think this was essential for head‐to‐head comparisons.

Objectives

To assess whether pharmacological treatment of hypertension can prevent cognitive impairment or dementia in people who have no history of cerebrovascular disease.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs) in which pharmacological interventions to lower blood pressure (BP) were administered for more than 12 months. We chose 12 months as the minimum trial duration because we considered it infeasible that cognitive effects could be detected in shorter trials.

Types of participants

We included people with a diagnosis of hypertension made according to the established criteria for the time. For studies included in the previous reviews, BP readings were at least 160/90 mmHg for entry into essential hypertension studies. Systolic blood pressure (SBP) was 160 to 219 mmHg, and diastolic blood pressure (DBP) was below 90 mmHg for entry into an isolated systolic hypertension study. The most recent National Institute for Health and Care Excellence (NICE) and Joint National Committee on the Prevention, Detection, Evaluation and Treatment of High Blood Pressure (JNC) 8 guidelines recommend thresholds of 140/90 mmHg and 150/90 mmHg, respectively (NICE 2019; JNC 8). Contemporary guidelines at the time many recent studies were conducted were more stringent, however, with the JNC 7 recommending thresholds of 140/90 mmHg and 130/80 mmHg for diabetics (JNC 7). Therefore, more recent included studies enrolled participants who had lower blood pressures at entry than did the older studies.

We excluded studies conducted solely in people with previous cerebrovascular disease as the evidence base for secondary prevention is addressed elsewhere. We did not exclude participants on the basis of prior use of antihypertensive therapy.

We had previously excluded studies on the basis that their populations were too young to expect a significant incidence of dementia. Given the emerging evidence of a relationship between vascular risk factors and mid‐life cognition, we have removed this criterion. We reviewed studies previously excluded on the basis of participant age.

We excluded studies with a sole focus on people living with dementia, as incidence of dementia is an outcome measure.

Types of interventions

We included these pharmaceutical agents, or classes of therapy, in this review:

• angiotensin‐converting enzyme inhibitors (ACE inhibitors);

• angiotensin receptor blockers (ARBs);

• beta adrenergic blockers;

• combined alpha and beta blockers;

• calcium channel blockers (CCBs);

• diuretics;

• alpha adrenergic blockers;

• central sympatholytics;

• direct vasodilators;

• peripheral adrenergic antagonists; and

• sympathomimetics.

We included trials giving these medications at any dose for at least 12 months. With the exception of Perez Stable 2000, all included studies employed a stepped approach to pharmacological antihypertensive treatment, with dosages increased or drugs added, or both, if BP was above given thresholds at predefined points in the studies. Further details of individual study approaches can be found in the Characteristics of included studies and Table 2.

1. Study properties.

Study	Number of participants	BP Entry Criteria (mmHg)	Baseline BP	Comparison	Study Duration
Placebo‐controlled RCTs
HYVET 2008	n=3336, n=1687 Rx group; n=1649 placebo n=1469 completed 2 yr MMSE	SBP 160‐200, DBP < 110	Mean SBP 173.1 Rx group; 172.9 placebo	stepped (diuretic then ACE inhibitor) vs placebo	5 years, 2 year results reported for HYVET‐COG
Perez Stable 2000	n=312, n=156 each Rx and placebo groups dementia incidence reported for n=102 and n=101 respectively	DBP 90‐104	Mean BP 140/96 Rx group; 141/96 placebo	placebo vs BB also lifestyle intervention factorial design (not included here)	12 months
SCOPE 2003	n=4937, n=2477 Rx group; n=2460 placebo	BP 160‐179/90‐99	Mean BP 166.0/90.3 Rx group; 166.5/90.4 placebo	stepped (ARB then diuretic) vs placebo	44.6 months
SHEP 1991	n=4736, n=2365 Rx group; n=2371 placebo	SBP > 160, DBP < 90	Mean BP 170.5/76.7 Rx group; 170.1/76.4 placebo	stepped (diuretic then BB then centrally acting agent) vs placebo	Mean follow‐up 4.5 years
Syst Eur 1998	n=2418, n=1238 Rx group; n=1180 placebo	SBP 160‐219, DBP < 95	Mean BP 173.4/86.0 Rx group; 173.5/86.1 placebo	stepped (CCB then ACE inhibitor then diuretic) vs placebo	Median follow‐up 2.0 years
LOMIR MCT IL 1996	n=368 (n=124; n=120; n=124)	DBP 95‐105	Mean BP 154.5/99.7; 152.0/99.3; 150.7/99.8	stepped, isradipine vs methyldopa vs placebo, dose doubled then ACE inhibitor added	12 months
MRC 1996	n=2651 with n=2567 and n=2584 completing PALT and TMT assessments, n=633 diuretic n=640 BB n=1311 placebo	SBP 160‐209, DBP < 115	Mean BP diuretic 184.9/90.3; BB 184.2/90.7; placebo 183.5/90.5	stepped, BB vs diuretic vs placebo, then BB dose doubled, other study drug added then CCB added; additional drugs permitted	54 months
Zhang 2018	n=732 (n=366 both groups)	SBP ≥140 and/or DBP ≥ 90 or self‐reported use of BP lowering meds in previous 2 weeks	Mean 156.1/71.3 ARB; 156.5/71.2 placebo	telmisartan vs placebo but all participants also Rx hydrochlorothiazide	Mean follow‐up 59.8 months
Intensive versus standard BP reduction
ACCORD MIND 2014	n=1439, n=745 intensive group, n=694 standard group	SBP ≥ 130 on ≥ 2 occasions on no meds, SBP 130‐160 on 0‐3 meds, 161‐170 on 0‐2 meds, 171‐180 on 0‐1 meds	Mean 138.8/76.0 intensive group; 139.2/76.3 standard	< 120 vs < 140 mmHg	40 months
SPRINT MIND 2019	n=9,361, n=4678 intensive group, n=4683 standard group	SBP 130 ‐ 180 on ≥ 2 occasions	Mean 139.7/78.2 intensive group; 139.7/78.0 standard	< 120 vs < 140 mmHg	Median intervention period 3.34 years
Other included studies
AVEC 2012	n=47 (n=17; n=17; n=13)	> 140/90 or on antihypertensives	Mean 153/85 ACE inhibitor; 149/81 ARB; 155/83 diuretic	stepped, ARB vs diuretic vs ACE inhibitor, dose increased then CCB added then BB added	12 months
CAMUI 2013	n=142 (n=74 and n=68)	> 140/90 or > 130/80 if diabetic or renal disease and already on an ARB	Mean 156/83 ARB+C; 155/83 ARB+D	stepped, ARB + CCB vs ARB + diuretic, further ARB added then alpha blocker then BB then antisympathetic agent added	12 months

RCTs: randomised controlled trials; MMSE: mini mental state examination; SBP: systolic blood pressure; DBP: diastolic blood pressure; Rx: hypertensive treatment group; ACE inhibitor: angiotensin converting enzyme inhibitor; BB: beta‐blocker; ARB: angiotensin receptor blocker; CCB: calcium channel blocker; PALT: paired associate learning test; TMT: trail making test.

Previous versions of this review included only placebo‐controlled studies (McGuinness 2006; McGuinness 2009). For this update, Cochrane requested that all RCTs of pharmacological treatment of hypertension be considered.

Types of outcome measures

Primary outcomes

• Incidence of dementia. For the diagnosis of dementia, we accepted standard criteria, such as the American Psychiatric Association Diagnostic and Statistical Manual of Mental Disorders (DSM) (APA 1987), International Classification of Diseases‐10 (ICD‐10) (WHO 1992); National Institute of Neurological and Communicative Disorders and Stroke ‐ Alzheimer's Disease and Related Disorders Association (NINCDS‐ADRDA) (McKhann 1984), or acceptable equivalents (Dubois 2007, McKhann 2011).

• Cognitive change from baseline. We required cognitive function to have been formally assessed using, for example, the Mini‐Mental State Examination (MMSE) or acceptable alternative.

Mild cognitive impairment was not considered as an outcome measure in either of the previous versions of this review and was not intended to be included in this update.

Secondary outcomes

• Blood pressure level

• Incidence and severity of adverse effects

• Quality of life

After termination of the Syst Eur 1998 trial, any participants exiting double‐blind therapy before the end of the study were followed as open‐label follow‐up until the end of the study and analysed using intention‐to‐treat analyses. In the previous versions of this review (McGuinness 2006; McGuinness 2009), we had included data on the changes in SBP and DBP for the entire Syst Eur 1998 cohort. For this update, we felt it more appropriate to include only the participants from the cognitive outcome analyses. The updated data are shown in Analysis 3.1 and Analysis 3.2.

3.1. Analysis.

Comparison 3: Change in blood pressure from baseline, Outcome 1: Change in systolic blood pressure level (mmHg) placebo studies

3.2. Analysis.

Comparison 3: Change in blood pressure from baseline, Outcome 2: Change in diastolic blood pressure level (mmHg) placebo studies

Search methods for identification of studies

Electronic searches

We searched the Cochrane Dementia and Cognitive Improvement Group’s Specialised Register, ALOIS (alois.medsci.ox.ac.uk/), on 7 July 2020. ALOIS is maintained by the Information Specialists of the Cochrane Dementia and Cognitive Improvement Group, and contains dementia and cognitive enhancement studies identified from:

• quarterly searches of the Cochrane Library’s Central Register of Controlled trials (CENTRAL);

• monthly searches of a number of major healthcare databases: MEDLINE Ovid, Embase Ovid, CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature), PsycINFO Ovid and LILACS (Latin American and Caribbean Health Science Information database);

• monthly searches of a number of trials registers: metaRegister of Controlled Trials (mRCT); Umin Japan Trial Register; World Heath Organization (WHO) portal (which covers ClinicalTrials.gov, ISRCTN registry (www.isrctn.com/), Chinese Clinical Trial Registry, German Clinical Trials Register, Iranian Registry of Clinical Trials, the Netherlands National Trials Register, plus others); and

• monthly searches of a number of grey literature sources, including ISI Web of Knowledge Conference Proceedings.

Additional separate searches were run in each of the above sources to ensure that the most up‐to‐date results were retrieved. The search strategies used for the retrieval of reports of trials can be seen in Appendix 1.

Searching other resources

We searched conference proceedings in the above searches. We contacted authors for further information as described in the dealing with missing data section below.

Data collection and analysis

Selection of studies

For this update, two review authors, ELC and BMcG, independently screened the search results. Both review authors agreed upon and tested the MeSH terms and search strategy used. The other review authors (PP, ST and RB) acted as adjudicators and reviewed the process.

ELC and BMcG independently selected trials for relevance against the defined inclusion criteria. We excluded those trials that did not fulfil the criteria from further analysis. No language restrictions were imposed. We resolved any disagreements through discussion and adjudication.

In the case of multiple publications relating to the same study, data are included only once.

Data extraction and management

We extracted data on trial characteristics and results from the published reports. The summary statistics required for each trial and each outcome for continuous data were the mean change from baseline, the standard error of the mean change, and the number of participants for each treatment group at each assessment. Where changes from baseline were not reported, the mean, standard deviation, and the number of participants for each treatment group at each time point were extracted. The outcomes measured in clinical trials of dementia and cognitive impairment often arise from ordinal rating scales. Where the rating scales used in the trials had a reasonably large number of categories (more than 10), we treated the data as continuous outcomes arising from a normal distribution.

For binary data, we sought the numbers in each treatment group and the numbers experiencing the outcome of interest. We defined the baseline assessment as the latest available assessment prior to randomisation, but no earlier than two months prior.

We have incorporated feedback regarding our erroneous interpretation of change in MMSE data from HYVET 2008 in the previous update.

Assessment of risk of bias in included studies

We assessed risk of bias in the included studies using Cochrane's tool for assessing risk of bias, applied to the published methods and reports for each study. We assessed selection bias, performance bias, detection bias, attrition bias, reporting bias, and any other sources of bias. Regarding randomisation, we considered studies to be at low risk of bias if robust methods were described and unclear risk if no methods were described. Regarding blinding, we considered studies to be at low risk (of performance and detection bias) if participants and assessors or adjudicators, respectively, were blinded, and high risk if they were not. Regarding attrition bias, we considered reporting of overall follow‐up rates, differential follow‐up rates between study arms, and whether associations were found between baseline characteristics and loss to follow‐up.

We used GRADE methods to rate the certainty of evidence (high, moderate, low or very low) behind each effect estimate in the review (Guyatt 2011). This rating refers to our level of confidence that the estimate reflects the true effect, taking account of risk of bias in the included studies, inconsistency between studies, imprecision in the effect estimate, indirectness in addressing our review question and the risk of publication bias. We produced a 'Summary of findings' table to show the effect estimate and the quantity and quality of supporting evidence for the following outcomes: incidence of dementia; change in MMSE; change in systolic blood pressure; change in diastolic blood pressure; adverse effects requiring discontinuation of treatment; quality of life measures.

Measures of treatment effect

For continuous outcomes, the measure of the treatment effect was the mean difference (MD) where the pooled trials used the same rating scale or test, and the standardised mean difference (SMD), which was the absolute mean difference divided by the standard deviation, where they used different rating scales or tests.

For binary outcomes, such as development of dementia or no dementia, we used the odds ratio (OR) to measure treatment effect.

Unit of analysis issues

This review includes randomised controlled trials only. As a multi‐arm study, we divided up the control group for MRC 1996. For dichotomous outcomes, we planned to divide up both the number of events and the total number of participants. For continuous outcomes, we planned to divide up only the total number of participants and leave the means and standard deviations unchanged.

Dealing with missing data

We sought data on every participant assessed for each outcome measure. To allow an intention‐to‐treat analysis, we sought the data irrespective of compliance, whether or not the participant was subsequently deemed ineligible, or otherwise excluded from treatment or follow‐up. If intention‐to‐treat data were not available in the publications, we sought 'on‐treatment' data or the data of those who completed the trial and indicated it as such. We contacted authors to request additional information where required. We corresponded with: Peters and colleagues regarding the error in reporting of their HYVET 2008 data; Saito and colleagues to clarify the CEREBRAL study is still ongoing; and Zhang and colleagues to seek additional data (Zhang 2018). We are grateful to these authors for their co‐operation. We attempted unsuccessfully to correspond with: Williamson and colleagues regarding SPRINT MIND 2019 results, though these were subsequently published; Wei and colleagues regarding their outcome measures (Wei 2013); and Ma and colleagues regarding cognitive outcomes reported in a conference abstract from 2018 (CHIEF 2012).

Assessment of heterogeneity

We decided whether or not to synthesise data using a meta‐analysis based on the clinical heterogeneity between trials (including consideration of participants, interventions and outcomes). We performed tests for heterogeneity using a standard Chi² statistic and a significance level of P = 0.10, as suggested by Cochrane. We also assessed the impact of heterogeneity on the meta‐analyses using the I² statistic. This informed our decision to used random‐effects versus fixed‐effect model.

Assessment of reporting biases

We considered whether results for prespecified outcomes published in protocols and methods sections were reported. Where there was a difference between outcome measures as planned in protocols and as reported in published results, we incorporated this into our GRADE assessments.

Data synthesis

We planned to undertake an analysis assessing the combined effect of all antihypertensive medications. We conducted meta‐analyses only when we considered the studies to be sufficiently similar clinically. For example, studies comparing placebo and active treatment were amenable to meta‐analysis. It was not possible to pool studies comparing active treatment and placebo, different active treatments and different BP targets in a single meta‐analysis. Where meta‐analyses were feasible, we presented the overall estimate from a fixed‐effect model. If there was significant statistical heterogeneity, we also conducted sensitivity analyses assessing the impact of using a random‐effects model.

Subgroup analysis and investigation of heterogeneity

We did not plan to undertake any subgroup analyses. We investigated heterogeneity as outlined above.

Sensitivity analysis

We undertook sensitivity analyses to assess the robustness of the results to fixed‐effect versus random‐effects models, and on the inclusion or exclusion of studies of poor quality.

Summary of findings and assessment of the certainty of the evidence

We summarised our findings in the Summary of Findings table (Table 5). We planned to present both primary and secondary outcomes measures where available. We focused on placebo controlled RCTs in keeping with original aim of the review. Certainty of evidence was assessed using the GRADE system and GRADEpro software.

Results

Description of studies

Results of the search

We have undertaken a total of seven searches for this review, including the latest search in July 2020. These are outlined in Figure 1. In addition to the previously included four studies, describing 15,427 participants, we identified eight new eligible studies, including 14,985 participants, for this update.

1.

Study flow diagram for searches December 2014 (212), November 2015 (309), October 2016 (145), August 2017 (220), June 2018 (279), June 2019 (177), July 2020 (259)

In total, we included 12 randomised controlled trials (RCTs), reporting outcomes for 30,412 participants, in this review.

Included studies

Of the eight placebo‐controlled RCTs, two used a diuretic‐based regimen (SHEP 1991; HYVET 2008), one used an ARB (SCOPE 2003), one used a CCB‐based (calcium channel blocker‐based) regimen (Syst Eur 1998), and one used a beta‐blocker (BB) regimen (Perez Stable 2000). LOMIR MCT IL 1996 and MRC 1996 had three arms each: two drug arms and one placebo arm. LOMIR MCT IL 1996 ran parallel isradipine, methyldopa and placebo groups with an ACE inhibitor added if necessary, while MRC 1996 ran parallel diuretic and beta‐blocker‐based groups with matching placebo groups. Zhang 2018 randomised participants using a 2 × 2 factorial design with ARB vs. placebo and statin vs. placebo arms. In addition, all participants were given hydrochlorothiazide.

Of the non‐placebo‐controlled studies, two compared intensive versus standard BP reduction (ACCORD MIND 2014; SPRINT MIND 2019). Of the two final included studies, AVEC 2012 compared ACE inhibitor, ARB, and diuretic groups, and CAMUI 2013 compared an ARB plus CCB group with an ARB plus diuretic group.

We outline the algorithms for pharmacological management in each trial in Table 2 with further details available in the Characteristics of included studies table.

SHEP 1991, LOMIR MCT IL 1996, Syst Eur 1998, Perez Stable 2000, SCOPE 2003 and HYVET 2008 were double‐blind, placebo‐controlled RCTs. Following the removal of the double‐blind, placebo‐controlled requirement for this update, we have also included MRC 1996 which had a placebo arm but was single‐blind.

The remaining five studies have been completed since the last review update. Zhang 2018 was a placebo‐controlled RCT comparing telmisartan to placebo in an elderly hypertensive population with white matter hyperintensity burden as the primary outcome measure.

The Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial was an open‐label RCT comparing target standard and intensive SBP targets below 140 mmHg and 120 mmHg, respectively, including only participants with type 2 diabetes mellitus. The cognitive sub‐study is ACCORD MIND 2014. The Systolic Blood Pressure Intervention Trial, SPRINT MIND 2019, is an open‐label RCT mirroring ACCORD in a non‐diabetic population, comparing target standard and intensive SBP targets below 140 mmHg and 120 mmHg, respectively.

The Antihypertensives and Vascular, Endothelial and Cognitive Function trial, AVEC 2012, is a double‐blind RCT comparing lisinopril, candesartan and hydrochlorothiazide with respect to cognitive outcomes in an elderly, hypertensive population with early cognitive impairment. CAMUI 2013 was a multicentre, open‐label, non‐placebo‐controlled RCT that administered a combination antihypertensive therapy in the elderly.

Sample size

We outline study sample sizes in Table 2. MRC 1996, Syst Eur 1998, HYVET 2008, ACCORD MIND 2014 and SPRINT MIND 2019 were all sub‐studies of larger RCTs.

Participants

Demographics and baseline cognition

We outline the gender, age, geographical location and, where reported, ethnicity, of participants in Table 3. There were more female than male participants overall. Most participants were middle‐aged or older. Study populations were global but predominantly recruited in North America and Europe. Study populations consisted mostly of ambulatory participants recruited from the community or primary care facilities.

2. Study demographics.

Study	Gender	Age criteria	Age	Geographical area	Race and ethnicity	Educational background	Baseline cognition
Placebo‐controlled RCTs
HYVET 2008	60% female	≥ 80 years	Mean age 83.5 years	Eastern Europe (56%), China (40%), Western Europe, Australasia and Tunisia	Not reported	27% no education 28% primary 29% secondary 12% higher 3% further	Median MMSE 26
Perez Stable 2000	34% female	18‐59 years	Mean age 46/45 years	California, USA	76%/71% white	Mean 16 years	Mean (SD) DSST 61 (10) See Characteristics of included studies for other tests
SCOPE 2003	65% female	70‐89 years	Mean age 76.4 years	Primarily Europe	Not reported	10% <primary 44% primary 40% >primary 6% university	Mean MMSE 28.5
SHEP 1991	57% female	≥ 60 years	Mean age 71.6 years	Exclusively USA	86.1% white, 13.9% black, 5% oriental, 2% Hispanic, 1% other	Mean 11.7 years	0.3% active Rx 0.5% placebo groups 'cognitive impairment scale score of 4 or greater', based on the short‐CARE assessment tool
Syst Eur 1998	66% female	≥ 60 years	Mean age 69.9 years	Exclusively Europe	Not reported	Mean 12.3 years	Median MMSE 29
LOMIR MCT IL 1996	Solely male	40‐65 years	Mean age 52 years	Israel	Not reported	Not reported	Mean semantic memory score 21.4, 20.8 and 22.3 across groups
MRC 1996	58% female	65‐74 years	Mean age 70.3/70.4 years	UK	Not reported	Not reported	Mean PALT score 17.0 all 3 groups; mean TMT score 59.9, 59.9 and 61.0 across 3 groups
Zhang 2018	48% female	≥ 60 years	Mean age 70.7 years	China	100% Han Chinese	Mean 7 years	Mean (IQR) MMSE 26 (25,28); mean (IQR) DRS 134 (129, 139)
Intensive versus standard BP reduction
ACCORD MIND 2014	55% female	≥ 55 years	Mean age 62.3/62.5 years	North America	66% white, 22% black, 6% Hispanic, 6% other	14/12% < high school 24/28% high school 37/35% some college 25% ≥ college	Mean MMSE 28
SPRINT MIND 2019	36% female	≥ 50 years	Mean age 67.9 years	USA	58% white,30% black, 10% Hispanic, 2% other	Not reported	Median MoCA 23, range 20‐26; Median logical memory 8; Median digit symbol coding test 51
Other included studies
AVEC 2012	57% female	≥ 60 years	Mean age 72/72/71 years	Boston, USA	70% White, 23% African American	19% up to high school, 81% college or more	Mean (SD) MMSE 26(2), 26(2), 25(2) across groups
CAMUI 2013	52% female	≥ 65 years	Mean age 74.1/74.5 years	Japan	Not reported	Not reported	Mean (SD) MMSE 27(3) and 26.5(3.2) across groups

RCTs: randomised controlled trials; MMSE: mini mental state examination; Rx: hypertensive treatment group; DSST: digit symbol substitution test; short‐CARE: short‐Comprehensive Assessment and Referral Evaluation; PALT: Paired Associate Learning Test; TMT: Trail Making Test; MoCA: Montreal Cognitive Assessment

We also outline educational background and baseline cognition, where available, in Table 3.

Blood pressure (BP) entry criteria

We outline the BP entry criteria for the included studies in Table 2, with more complete descriptions in the Characteristics of included studies tables. More recent studies have lower BP eligibility thresholds in line with the evolving guidelines and clinical practice.

Interventions

With the exception of Perez Stable 2000, all studies employed a stepped approach to pharmacological antihypertensive treatment with dosages increased or drugs added, or both, if BP was above given thresholds at predefined points in the studies. Further details of each study's drug treatment protocol can be found in the Characteristics of included studies tables and Table 2.

Duration of study

Planned mean length of study follow‐up varied between the studies (one year to five years) and is described in Table 2. Three studies ‐ Syst Eur 1998, HYVET 2008 and SPRINT MIND 2019 ‐ were stopped early on safety grounds. After termination of the Syst Eur 1998 trial, any participants exiting double‐blind therapy before the end of the study were followed as open‐label follow‐up until the end of the study and analysed using intention‐to‐treat analyses, the results of which have been incorporated into this review. SPRINT MIND 2019 had planned for 24‐month and 48‐month cognitive follow‐up with a further assessment at study closeout if that was more than one year from the four‐year follow‐up visit. Participants were recruited between November 2010 and March 2013. The decision was taken to halt the study on safety grounds in August 2015. SPRINT MIND 2019 reported dementia and mild cognitive impairment (MCI) incidence as ascertained from final extended study visits conducted between October 2017 and July 2018 (median total follow‐up 5.11 years). Limited results were also reported for outcomes ascertained during the study period (median intervention period 3.34 years) and it is those results we report here where possible. ACCORD MIND 2014 reported 20‐month and 40‐month follow‐up. The glycaemic control arm of the ACCORD study was terminated early due to concerns regarding increased mortality in the intensive glycaemic control group.

Outcomes

Dementia

Five studies measured incidence of dementia (SHEP 1991, Syst Eur 1998, SCOPE 2003, HYVET 2008, SPRINT MIND 2019). Triggers for (a) further diagnostic evaluation, and (b) the criteria for diagnosis of dementia were:

• Syst Eur 1998 ‐ (a) MMSE score below 23 points or symptoms or signs reported by participant or carer or found by doctor; (b) Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM‐III‐R), with all cases validated by a treatment allocation‐blinded review board;

• SCOPE 2003 ‐ (a) significant cognitive decline was a reduction by 4 or more points in the MMSE in two consecutive visits in comparison to baseline; (b) International Classification of Diseases‐10 (ICD‐10) criteria, adjudicated by an independent clinical event committee based on information supplied by the local investigator;

• HYVET 2008 ‐ (a) decline in MMSE score to below 24 points or fall of more than 3 points in one year; (b) Diagnostic and Statistical Manual of Mental Disorders, Fourth edition (DSM‐IV) (APA 1994), with consensus from a treatment allocation‐blinded central committee;

• SPRINT MIND 2019 ‐ (a) Montreal Cognitive Assessment (MoCA) score below 21 points for white participants with 12 or more years of education, below 19 points for white participants with fewer than 12 years of education or nonwhite participants with 12 or more years of education, or below 17 points for nonwhite participants with fewer than 12 years of education, OR a decrease in MoCA score of 5 or more points led to administration of the Functional Activities Questionnaire to a pre‐identified proxy. A score of more than 0 on the Functional Activities Questionnaire OR 1 or less on the 5‐point delayed recall subtest of the MoCA underwent further testing using an extended cognitive battery (see Characteristics of included studies). A telephone assessment was undertaken where in‐person follow‐up was not possible. In these cases a score of 31 points or less on the Modified Telephone interview for Cognitve Status (instead of a MoCA cut‐off) prompted administration of the Functional Activities Questionnaire. Where a participant had died or could not be contacted, the Dementia Questionnaire was administered to a pre‐specified contact; (b) McKhann 2011 for dementia and Albert 2011 for mild cognitive impairment (MCI) as adjudicated by an expert multidisciplinary panel blinded to treatment allocation.

In SHEP 1991, dementia was diagnosed by an expert and confirmed by the central coding panel according to the DSM‐III‐R criteria (APA 1987) after referral triggered by score on the short‐Comprehensive Assessment and Referral Evaluation (short‐CARE) at follow‐up. In addition, as reported by Applegate 1994 (see references for SHEP 1991), a subgroup of six centres undertook additional cognitive assessments comprising the Digit Symbol Subsitution Test, Addition Test, Finding As Test, Boston Naming Test, Delayed Recognition Span Test and the Letter Sets Test.

Cognitive decline

Mini‐Mental State Examination

The most common cognitive test used was the Mini‐Mental State Examination (MMSE) with six studies reporting change in MMSE: SCOPE 2003 (adjusted change at 44.6 months); Syst Eur 1998 (mean change at 2 years); HYVET 2008 (mean change at 2 years); Zhang 2018 (baseline and 1, 2, 3, 4 and 5 years' mean scores); ACCORD MIND 2014 (baseline, 20‐month and 40‐month adjusted mean scores); and CAMUI 2013 (baseline and 1 year scores). In addition, HYVET 2008 and SCOPE 2003 reported cognitive decline as defined by a fall in MMSE: to less than 24 points, or of more than 3 points in one year (HYVET 2008); or a fall of 4 or more points (SCOPE 2003). Of note, HYVET 2008 reported MMSE at two years for 1469 of the total 3336 participants. The numbers in each group are not reported in the HYVET COG paper of 2008 but the original HYVET paper in the New England Journal of Medicine notes 2‐year follow‐up blood pressures for 754 participants in the active treatment group and 701 participants in the placebo group, so we have used these numbers here. We are grateful to Zhang 2018 for providing unpublished data regarding change in MMSE and BP over the study period. We have incorporated the 60‐month values in our analyses. AVEC 2012 used MMSE as a screening tool at baseline but did not measure change.

Other cognitive measures

LOMIR MCT IL 1996 measured semantic memory using referenced but non‐specified tools (Weingartner 1983). MRC 1996 reported baseline scores and rate of change over time, adjusted for baseline score, of the Paired Associate Learning Test (PALT) and Trail Making Test A (TMT‐A). A sub‐study of Syst Eur 1998 used TMT‐A and TMT‐B. Perez Stable 2000 reported 3‐month and 12‐month change of the Digit Symbol Substitution Test, Stimulus Evaluation/Response Selection Continuous Performance Task and California Verbal Learning Test. In addition to MMSE, Zhang 2018 measured the Mattis Dementia Rating Scale (DRS) at baseline and annual review. In addition to reporting MMSE and DRS scores as continuous variables, Zhang and colleagues considered cognitive impairment as a binary yes/no outcome comprised of 'MMSE score of <=23 points at any annual follow‐up visit or decline by <=3 points between any two annual follow‐up visits and/or DRS score <=123 points at any annual follow‐up visit' (Zhang 2018).

As well as MMSE, ACCORD MIND 2014 reported baseline, 20‐month and 40‐month adjusted mean Digit Symbol Substitution Test, Stroop and the Rey Auditory Verbal Learning Test (RAVLT). SPRINT MIND 2019 reported baseline only MoCA, Logical Memory form II of the Wechsler Memory Scale and Digit Symbol Coding Test of the Wechsler Adult Intelligence Scale tests. In addition to dementia, incidence of mild cognitive impairment, at follow‐up but not as baseline, was reported.

AVEC 2012 reported change over 12 months as adjusted least square means for the TMT, Digit Span Test (DST) and Hopkins Verbal Learning Test (HVLT).

Blood pressure

All studies measured sitting BP on study visits. None of the included studies used ambulatory BP monitoring to generate outcome measurements.

Incidence and severity of adverse effects

There was considerable variability in the reporting of adverse effects.

Of the placebo‐controlled RCTs, SHEP 1991, SCOPE 2003, LOMIR MCT IL 1996, MRC 1996 and Syst Eur 1998 reported withdrawal for (SHEP 1991, SCOPE 2003, MRC 1996) or including (LOMIR MCT IL 1996 and Syst Eur 1998) adverse effects. Perez Stable 2000 reported the numbers in the intervention group not taking the study drug at twelve months but not the reasons for this nor the corresponding number in the placebo group. Zhang 2018 reported study withdrawals over five years but not the reasons for these. Data regarding adverse effects and discontinuation of therapy due to adverse effects were not available for the pre‐specified cognitive sub‐study groups for MRC 1996 and Syst Eur 1998, and so we used the results from the overall cohorts. In addition, SHEP 1991 reported 'prevalence of symptoms ever characterised as troublesome or intolerable' and serum clinical chemistry changes. SCOPE 2003 reported most common adverse effects, adverse effects indicating hypotension and routine clinical chemistry changes. LOMIR MCT IL 1996 reported most frequently reported adverse symptoms. HYVET 2008 reported serious adverse events but not withdrawals due to adverse effects.

ACCORD MIND 2014 reported adverse events for the overall BP trial categorised as serious adverse events (including those attributed to antihypertensive medications), symptoms affecting quality of life, adverse laboratory measures and clinical measures. SPRINT MIND 2019 reported serious adverse events classified as possibly or definitely related to the intervention. Neither study reported data for adverse events requiring discontinuation of treatment.

AVEC 2012 reported incidence of the 'most common and serious' adverse events for the three treatment groups and treatment withdrawal due to adverse events. CAMUI 2013 reported a single withdrawal due to adverse drug effects.

Quality of life

Seven trials provided quality of life (QOL) data. SHEP 1991 used the short‐CARE and Social Network Questionnaires and, at the six centres undertaking additional testing, three global quality of life questions. LOMIR MCT IL 1996 measured QOL using a tailor‐made tool (Bar‐On 1993), in addition to measures of sleep and physical and sexual function (Croog 1986), workplace stress (Frankenhaeuser 1976), and an adapted life‐events scale (LOMIR MCT IL 1996). A sub‐study of Syst Eur 1998 used three QOL assessments: Sickness Impact Profile, Brief Assessment Index and a checklist of 32 symptoms associated with hypertension and the side effects of antihypertensive treatment. A SCOPE 2003 sub‐study, Degl'Innocenti 2004, reported QOL data using three validated assessments: Psychological General Well‐Being Index (PGWB), Subjective Symptoms Assessment Profile (SSA‐P), and EuroQoL Health Utility Index (EuroQoL). HYVET 2008 used 'a quality of life questionnaire including the short‐form (SF) 36 and a list of symptoms associated with hypertension and hypertensive drugs' in a sub‐study of participants.

ACCORD MIND 2014: the ACCORD BP health‐related quality of life (HRQL) sub‐study reported on 1028 participants who completed baseline and one or more follow‐up QOL evaluations, including the Diabetes Symptoms Distress Checklist; the 36‐Item Short Form Health Survey, Version 2 (SF‐36) (RAND Corporation, Santa Monica, CA); the Patient Health Questionnaire (PHQ) depression measure (Pfizer Inc, New York, NY); the World Health Organization (WHO) Diabetes Treatment Satisfaction Questionnaire (DTSQ); and the EuroQol Feeling Thermometer (EuroQol Group, Rotterdam, Netherlands) (O'Connor 2012). Participants in SPRINT MIND 2019 completed the Veterans RAND 12‐Item Health Survey (VR‐12), the Patient Health Questionnaire (PHQ) depression measure (Pfizer Inc, New York, NY) and a five‐point Likert scale rating patient satisfaction with blood pressure medications and care from very satisfied to very dissatisfied.

Five studies did not report QOL measures (MRC 1996; Perez Stable 2000; AVEC 2012; CAMUI 2013; Zhang 2018).

Excluded studies

Three previously excluded studies were included in this update: LOMIR MCT IL 1996 (participants aged 40 to 65 years), MRC 1996 (single‐blinded), and Perez Stable 2000 (participants aged 18 to 59 years).

Of the excluded studies, we excluded most because they did not report cognitive outcomes; for example, the placebo‐controlled European Working Party on High Blood Pressure in the Elderly (EWPHE 1985) study, STOP 1991, Syst China 1998, Black 2001, ALLHAT 2003, Hsieh 2003, VALUE 2004 and ASCOT 2005.

As the evidence for treating hypertension post‐stroke is now well‐established, we considered a history of cerebrovascular disease an exclusion criterion for this review. Therefore, we excluded the 'Perindopril pROtection aGainst REcurrent Stroke Study' (PROGRESS 2003), and the 'Prevention Regimen For Effectively Avoiding Second Strokes' study (PRoFESS 2008), which recruited only participants with a history of transient ischaemic attack (TIA) or stroke. INFINITY 2019 required participants to have a significant white matter hyperintensity burden on MRI at baseline and so was excluded.

We also excluded studies that did not require participants to be hypertensive. ONTARGET 2008, a non‐placebo‐controlled trial, compared an ACE inhibitor, ARB, or both, and TRANSCEND 2008, a placebo‐controlled RCT, compared an ARB with placebo, with regard to a composite cardiovascular event primary outcome measure, with secondary dementia and cognitive decline outcome measures in a population at high cardiovascular risk. The 'Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation' (ADVANCE 2010) study compared additional antihypertensive agents with placebo in an exclusively diabetic population and included dementia as a secondary outcome but participants were not required to be hypertensive. HOPE 3 2016 recruited people in middle age or older with additional cardiovascular risk factors but did not require participants to be hypertensive. The cognitive sub‐study, HOPE 3 2019, included participants aged 70 years and over.

The 'Discontinuation of Hypertensive Treatment in Elderly people' (DANTE 2015) study was open‐label and compared the cognitive effects of antihypertensive treatment discontinuation versus continuation in those aged over 75 years with mild cognitive impairment. Discontinuation studies have been considered by another Cochrane Review and are not included in this review (Jongstra 2016).

Risk of bias in included studies

Allocation

Nine of the studies reported a robust method of randomisation (ACCORD MIND 2014; AVEC 2012; CAMUI 2013; HYVET 2008; Perez Stable 2000; SCOPE 2003; SHEP 1991; Syst Eur 1998; Zhang 2018). LOMIR MCT IL 1996 did not describe a method of randomisation. MRC 1996 states randomisation was within stratified blocks but the method of randomisation is unclear. SPRINT MIND 2019 reported use of 'an internet‐based, web browser randomisation procedure'.

Blinding

In contrast to the previous update (McGuinness 2009), this review included studies that were single‐blind or non‐blinded. In the cases of ACCORD MIND 2014 and SPRINT MIND 2019, the intensive treatment groups had more visits than the standard treatment groups, which increases the risk of performance bias. The risk of detection bias is theoretically increased in those studies where cognitive tests were administered by unblinded assessors (i.e. MRC 1996, ACCORD MIND 2014 and CAMUI 2013). The expert adjudication panel in SPRINT MIND 2019 was blinded to treatment allocation.

Incomplete outcome data

We judged most studies to be at high risk of attrition bias on the basis of lack of reporting or dropout rates associated with older age or worse cognitive function at baseline. Overall loss to follow‐up was similar between study group arms. Less than half of participants in SHEP 1991 and Syst Eur 1998 completed 5‐year and 4‐year follow‐up, respectively, though numbers were similar across the groups. Perez Stable 2000 and AVEC 2012 reported 12‐month cognitive outcomes for only two‐thirds of participants without explanation. LOMIR MCT IL 1996 and CAMUI 2013 did not explicitly report numbers completing 12‐month cognitive assessments. In MRC 1996, ACCORD MIND 2014 and SPRINT MIND 2019, those with lower baseline scores were less likely to be followed up. SPRINT MIND 2019 showed that more frail participants were also less likely to be followed up.

Due to early termination of the trial, only 44% of participants in HYVET 2008 completed a 2‐year follow‐up visit. The study was probably too short to exhibit attrition bias. SCOPE 2003 reported dementia and cognitive decline outcomes for most participants and was therefore also judged to be at low risk of attrition bias. Zhang 2018 reported annual follow‐up rates of over 90%, up to an average follow‐up of 59.8 months.

Selective reporting

Several studies reported data for positive findings only, which increases the risk of selective reporting bias. SPRINT MIND 2019 reported mild cognitive impairment (MCI) as a secondary outcome; however, in the original study protocol, it was described as 'not a primary or secondary outcome'. CHIEF 2012, a randomised, open‐label, blinded endpoint study comparing CCB plus diuretic and CCB plus ARB in 13,542 hypertensive Chinese participants, measured MMSE at baseline and during follow‐up. A conference abstract from 2018 reports headline results; however, our attempts to contact the authors for further information were unsuccessful.

Other potential sources of bias

Dropout rates due to adverse events or lack of treatment effect, or both, did differ across groups and raise the possibility of contamination bias. In the eight placebo‐controlled trials, a proportion of the participants assigned to the placebo group received antihypertensive treatment because their BP exceeded pre‐set 'escape criteria'. Also, a proportion of participants assigned to antihypertensive treatment groups stopped taking their medications due to side effects or because they achieved normal blood pressure off medication. The degree to which participants cross over from one group to another reduces the strength of the results of a study. The percentage of participants assigned to placebo groups who were receiving antihypertensive medication by the end of trials were as follows: SHEP 1991 ‐ 44%; Syst Eur 1998 ‐ 27%; and SCOPE 2003 ‐ 84%. In LOMIR MCT IL 1996, 27% of participants in the placebo group withdrew. In MRC 1996, 51% (1309/2584) of participants completed the study in their randomised group, 47% of the placebo group completed the study in their randomised group and without additional antihypertensive therapy (622/1311). In Zhang 2018, all participants in both groups were commenced on hydrochlorothiazide at the start of the study with the dose increased if required.

The percentage of participants assigned to antihypertensive treatment groups who had ceased taking antihypertensive treatment by the end of the trials were as follows: SHEP 1991 ‐ 10%; LOMIR MCT IL 1996 ‐ 25% (estimated 36/144 as 21 withdrawals in the run‐in phase not described by group and final numbers not reported); Syst Eur 1998 ‐ 18%; Perez Stable 2000 ‐ 13%; and SCOPE 2003 ‐ 0%.The percentage of participants taking initially assigned medication alone, where reported, were as follows: SHEP 1991 ‐ 30%; Syst Eur 1998 ‐ 30%; and SCOPE 2003 ‐ 25%. An estimated 20% to 50% of participants in LOMIR MCT IL 1996 were prescribed captopril in addition to the original study drug but explicit figures were not reported. In HYVET 2008, at the 2‐year follow‐up, 0.8% of the antihypertensive treatment group were not taking one of the three treatment steps specified in the protocol, and 0.6% of the placebo group were not taking one of the matching placebo steps specified in the protocol. We have therefore judged SHEP 1991, LOMIR MCT IL 1996, MRC 1996, Syst Eur 1998, Perez Stable 2000, SCOPE 2003 and Zhang 2018 to be at high risk of contamination bias. This is likely to have led to a systematic underestimate of the effect size.

SPRINT MIND 2019 reported headline outcomes from an extended follow‐up visit, two years after the study had been terminated. We have included the results from the in‐study visits here, where possible.

The process of identifying cases of dementia in Syst Eur 1998, SCOPE 2003, HYVET 2008 and SPRINT MIND 2019 may have led to a systematic underestimation of the effect by relying on screening tests such as MMSE, which are known to lack sensitivity (Creavin 2016), and in the case of SPRINT MIND 2019, a MoCA score below that shown to be most sensitive (Davis 2015).

Effects of interventions

See: Table 1

Comparison 1: Antihypertensive therapy compared to placebo

Incidence of dementia

Incidence of dementia was a secondary outcome in five of the included trials (SHEP 1991; Syst Eur 1998; SCOPE 2003; HYVET 2008; SPRINT MIND 2019). The combined results of the four placebo‐controlled trials suggest that antihypertensive treatment results in little to no difference in dementia incidence (OR = 0.89, 95% CI 0.72 to 1.09; 4 trials, 15,427 participants; Analysis 1.1). We considered this to be very low certainty evidence, downgraded due to study limitations and indirectness. The incidence of dementia was lower in the antihypertensive group but there was uncertainty regarding the size and direction of effect.

1.1. Analysis.

Comparison 1: Incidence of dementia, Outcome 1: Number of cases of dementia placebo studies

Of the individual studies, only Syst Eur 1998 reported a significant benefit of the intervention. There were 21 cases of incident dementia in the placebo group and 11 in the antihypertensive group. The numbers of patient years were 2737 and 2885, respectively. Active treatment reduced the rate of dementia by 50% (95% CI 0 to 76) from 7.7 to 3.8 cases per 1000 patient‐years (P = 0.05). However, when the numbers of participants in the antihypertensive (1238) and placebo groups (1180) are considered, consistent with analysis of the other included studies, the benefit with antihypertensive treatment is no longer significant (OR 0.49, 95% CI 0.24 to 1.03; Analysis 1.1).

We conducted a sensitivity analysis excluding SCOPE 2003 (due to the very high rate of antihypertensive use in the placebo group), but evidence still suggested that antihypertensive treatment results in little to no difference in dementia incidence.

Cognitive change from baseline

Syst Eur 1998, SCOPE 2003, HYVET 2008 and Zhang 2018 compared antihypertensive treatment to placebo and provided data on cognitive change from baseline. The combined results from these four trials, which reported change in MMSE, showed that MMSE was 0.20 points higher (indicating better cognitive function) in the active treatment group (MD 0.20, 95% CI 0.10 to 0.29; 4 studies, 9335 participants; very low certainty of evidence downgraded due to study limitations, indirectness and imprecision; Analysis 2.1). However, the magnitude of the effect is below that considered statistically significant by convention, as described in the 2014 World Alzheimer's Report (Prince 2014), as well as that considered minimally clinically significant in the literature (Lopez 2005; Jones 2009; Howard 2012; Andrews 2019).

2.1. Analysis.

Comparison 2: Cognitive change from baseline, Outcome 1: Change in MMSE placebo studies

We conducted a sensitivity analysis excluding SCOPE 2003 and Zhang 2018 (due to the very high rate of antihypertensive use in the placebo group), which in fact further reduced the statistical significance of the difference in MMSE decline between the intervention and placebo groups (MD ‐0.01, 95% CI ‐0.16 to 0.14; 2 studies, 3873 participants; very low certainty of evidence downgraded due to study limitations, indirectness and imprecision).

A subgroup of SHEP 1991 (n = 798 hypertensive treatment and n = 766 placebo) completed baseline and follow‐up cognitive testing using the digit symbol substitution test (DSST), addition test, finding A's test, Boston naming test, letter sets test and delayed recognition span test. No differences were found in mean changes in cognitive tests between the groups.

MRC 1996, comparing beta‐blocker (n = 640), diuretic (n = 633) and placebo (n = 1311) groups, reported adjusted (for age, sex and baseline intelligence, cognitive and depression scores) PALT (Paired Associate Learning Test) and TMT‐A (Trail Making Test‐A) coefficients where these were calculated as 'the slope of the regression lines derived by regressing the cognitive test scores on time for each subject.... assigning time values of 1,2,3,4, and 5 to cognitive tests scores at entry and 1,9,21 and 54 months respectively' (MRC 1996). No difference was found between treatment and placebo groups on intention‐to‐treat or per protocol analyses. Change in overall mean TMT‐A and PALT scores were reported for the whole cohort only.

LOMIR MCT IL 1996 reported that after twelve months 'patients treated with the combination of isradipine and captopril showed a statistically significant improvement from baseline in semantic memory (p<0.001) ... the methyldopa group treated as monotherapy or with captopril showed a negative change from baseline in semantic memory'. Results for the placebo group were only represented graphically and appeared to show a deterioration in semantic memory over the twelve months.

Perez Stable 2000 reported cognitive performance at twelve months for propranolol and placebo groups for all four cognitive tests used. No statistically significant differences were found between propranolol and placebo groups at twelve months.

Change in systolic blood pressure (SBP)

The combined results from six trials reporting change in SBP indicated evidence suggesting antihypertensive treatment reduces SBP (MD ‐9.25 mmHg, 95% CI ‐9.73 to ‐8.78; 6 studies, 17,793 participants; low certainty of evidence downgraded on the basis of study limitations and inconsistency; Analysis 3.1) (SHEP 1991; MRC 1996, Syst Eur 1998; SCOPE 2003; HYVET 2008; Zhang 2018). Intention‐to‐treat analysis figures were used. The low numbers who ended the study in the randomised group should be noted. All included studies indicated a significant benefit of treatment. Given the presence of heterogeneity, we undertook a sensitivity analysis comparing fixed‐effect versus random‐effects models, which gave the same evidence.

Reporting of results varied. MRC 1996, Syst Eur 1998 and HYVET 2008 reported change in BP only, not final values. SHEP 1991 provided yearly systolic and diastolic blood pressure measures for active treatment and placebo groups for five years along with difference between active treatment group and placebo. LOMIR MCT IL 1996 reported baseline SBP and DBP across groups; twelve month BP and change in BP were not reported. Perez Stable 2000 reported baseline and twelve month SBP and DBP only and did not report change. SCOPE 2003 reported baseline, final and adjusted (for country and baseline BP) change in BP for the two groups. Zhang 2018 reported baseline and annual mean SBP for each group and provided us with unpublished data describing change when requested.

Change in diastolic blood pressure (DBP)

The combined results from five trials reporting change in DBP level indicated evidence suggesting antihypertensive treatment reduces DBP (MD ‐2.47 mmHg, 95% CI ‐2.70 to ‐2.24; 5 studies, 15,209 participants; low certainty of evidence downgraded on the basis of study limitations and inconsistency; Analysis 3.2) (SHEP 1991; Syst Eur 1998; SCOPE 2003; HYVET 2008, Zhang 2018). All included studies indicated a significant benefit of treatment although Syst Eur 1998 included participants with isolated systolic hypertension only. Given the presence of heterogeneity, we undertook a sensitivity analysis comparing fixed‐effect versus random‐effects models, which gave the same evidence. As above, LOMIR MCT IL 1996 did not report change in DBP. MRC 1996, comparing beta‐blocker (n = 640), diuretic (n = 633) and placebo (n = 1311) groups, did not report change in DBP. A previous report from the authors of Perez Stable 2000 reported a mean decrease in DBP of 7.7 mmHg in the propranolol only group and 5.9 mmHg in the placebo only group.

Incidence and severity of adverse effects

In SHEP 1991, significantly more participants on active treatment withdrew due to adverse events (OR 1.98, 95% CI 1.63 to 2.42). LOMIR MCT IL 1996 reported a statistically significant increase in sleep and sexual disorders and withdrawals due to side effects in the methyldopa compared to the isradipine and placebo groups. The overall MRC 1996 study reported a significantly higher rate of withdrawals due to major side effects in the beta‐blocker (333/1102) compared to the diuretic (160/1081) and placebo groups (82/2213). Syst Eur 1998 reached significance, with participants on active treatment less likely to discontinue treatment due to side effects (OR 0.42, 95% CI 0.33 to 0.54). Perez Stable 2000 reported that 20/156 participants in the propranolol group were not taking the study drug at 12‐month follow‐up but did not specify reasons for this. Discontinuation of treatment in the placebo group was not reported. No significant difference between the active treatment and placebo groups in adverse events requiring discontinuation of treatment was found in SCOPE 2003, possibly because of the large proportion of participants in the placebo group who commenced active treatment during the study. HYVET 2008 reported fewer 'serious adverse events' in the active treatment group than in the placebo group (358 versus 448, P = 0.001). Zhang 2018 reported 31 deaths, 16 withdrawals and 5 losses to follow‐up for their study population of 732 participants over five years. Reasons for study withdrawal were not reported. Hetergeneity precluded meta‐analysis.

Quality of life (QOL)

SHEP 1991 reported no significant differences between active treatment and placebo groups in any of the three global quality of life questions at baseline or final visit for those in that subgroup. Similarly, the sub‐study of Syst Eur 1998 showed no significant difference in change in Sickness Impact Profile at two years between active treatment and placebo groups. The SCOPE 2003 sub‐study, Degl'Innocenti 2004, showed significant differences in change in QOL measures between the active treatment and placebo groups including Psychological General Well‐Being (PGWB) Index Positive well‐being (‐0.8 versus ‐1.1, P = 0.04) and EuroQoL Current Health (‐3.1 versus ‐5.3, P = 0.008). LOMIR MCT IL 1996 reported a statistically significant (P = 0.03) improvement in a subjective measure of QOL at twelve months in the isradipine and captopril combined therapy group. The methyldopa monotherapy group showed a deterioration in subjective QOL at twelve months but statistical significance was not reported. HYVET 2008 reported no difference in SF‐36 scores between active treatment and placebo groups over time. MRC 1996, Perez Stable 2000 and Zhang 2018 did not report QOL measures.

Comparison 2: intensive versus standard blood pressure reduction

Incidence of dementia

SPRINT MIND 2019 did not show a significant difference in dementia incidence between standard and intensive treatment groups (HR 0.93 (95% CI 0.73‐1.18), P = 0.54).

Cognitive change from baseline

ACCORD MIND 2014, comparing standard versus intensive BP control, reported baseline, 20‐ and 40‐month performance in MMSE, digit symbol substitution test (DSST), Rey Auditory Verbal Learning Test (RAVLT) and Stroop, and 40‐month change, adjusted for baseline score of the relevant measure. No significant difference was noted between groups for any of the tests. Change in adjusted mean MMSE over 40 months in the intensive (n = 690) and standard (n = 659) groups were ‐0.25 (CI ‐0.42 to ‐0.08) and ‐0.30 (CI ‐0.48 to ‐0.13), respectively. SPRINT MIND 2019 reported baseline but not follow‐up cognitive assessment data.

Change in systolic blood pressure (SBP)

ACCORD MIND 2014, comparing standard versus intensive BP control, reported baseline and 40‐month SBP as mean (SD) as outlined in the data analyses tables. At 40 months, mean SBP was 133.2 (14.8) versus 119.0 (14.7) mmHg respectively. SPRINT MIND 2019, when it was halted in August 2015, reported mean SBP in the standard and intensive treatment groups of 134.8 mmHg (95% CI 134.1 to 135.6 mmHg) and 121.6 mmHg (95% CI 120.8 to 122.3 mmHg), respectively.

Change in diastolic blood pressure (DBP)

ACCORD MIND 2014, comparing standard versus intensive BP control, reported baseline and 40‐month DBP as mean (SD), as outlined in the data analyses tables. At 40 months, mean DBP was 64.0 (10.1) versus 30.2 (9.9) mmHg. SPRINT MIND 2019 did not report change in DBP.

Incidence and severity of adverse effects

ACCORD MIND 2014 reported significantly more (P < 0.001) serious adverse events attributed to antihypertensive medications in the intensive treatment arm (77/2362, 3.3%) than those in the standard treatment arm (30/2371,1.3%). SPRINT MIND 2019 reported serious adverse events classified as possibly or definitely related to the intervention for 118 participants (2.5%) in the standard treatment group and 220 participants (4.7%) in the intensive treatment group (hazard ratio (HR) 1.88, P < 0.001).

Quality of life (QOL)

The ACCORD BP HRQL sub‐study of ACCORD MIND 2014 reported no significant differences in five out of six QOL measures. Those assigned to intensive BP control had statistically, but not clinically, significant worsening of physical component scores of SF‐36 (‐0.8 versus ‐0.2, P = 0.02). The SPRINT MIND 2019 group reported, in 2017, no significant difference in QOL measures between treatment groups.

Comparison 3: non‐placebo‐controlled RCTs

Incidence of dementia

Neither AVEC 2012 nor CAMUI 2013 reported incidence of dementia.

Cognitive change from baseline

AVEC 2012, comparing ARB, diuretic and ACE inhibitor groups, reported change in Trail Making Test (TMT), Hopkins Verbal Learning Test (HVLT) and Digit Span Test (DST) over 12 months as least square means adjusted for baseline age and MMSE. Data were incompletely reported in the 2013 paper, as follows: 12‐month least square mean decrease in time taken to complete TMT‐B of 17.1 seconds in ARB group (n = 17), 4.2 seconds in diuretic group (n = 13) and an increase of 14.4 seconds in the ACE inhibitor group (n = 17) (between group P = 0.008); improvement in recognition portion of HVLT in ARB group (P = 0.03); no group difference in immediate and delayed recall portions of HVLT and DST. Data were shown in figures but not reported in a 2012 report. Data were not extrapolated from the figures for these analyses.

CAMUI 2013, comparing ARB plus CCB and ARB plus diuretic combinations, reported no significant difference in MMSE values at baseline and one year in either group, or between groups. Mean MMSE at baseline and one year for the ARB plus CCB group (n = 68) were 27.0 and 26.6 points, respectively, and for the ARB plus diuretic group (n = 74), 26.5 and 26.5 points, respectively. Group means only were reported for final MMSE.

Change in systolic blood pressure (SBP)

AVEC 2012, comparing ARB (n = 17), diuretic (n = 13) and ACE inhibitor (n = 17) groups, reported baseline, 6‐ and 12‐month age‐adjusted least square mean (SE) SBP for the three groups, and also change in SBP over 12 months as mean reduction with standard errors for each of the three groups: ARB 26 (5) mmHg, diuretic 25 (6) mmHg and ACE inhibitor 27 (5) mmHg, as shown in the data analyses tables. These were reported as not significantly different between groups (P = 0.93).

CAMUI 2013, comparing ARB plus CCB and ARB plus diuretic combinations, reported as a figure baseline, 3‐, 6‐, and 12‐month SBP for both groups. Mean and SD data were reported explicitly only for baseline, 3‐ and 6‐month BP data, with 6‐month SBP of 132 (13) in the ARB plus CCB group (n = 68) and 133 (16) in the ARB plus diuretic group (n = 74) (P = 0.575).

Change in diastolic blood pressure (DBP)

AVEC 2012, comparing ARB (n = 17), diuretic (n = 13) and ACE inhibitor (n = 17) groups, reported baseline, 6‐ and 12‐month age‐adjusted least square mean (SE) DBP for the three groups. CAMUI 2013, comparing ARB plus CCB and ARB plus diuretic combinations, reported as a figure baseline, 3‐, 6‐ and 12‐month DBP for both groups. Mean and standard deviation data were reported explicitly only for baseline, 3‐ and 6‐month BP data, with 6‐month DBP of 72 (9) in the ARB plus CCB group (n = 68) and 73 (11) in the ARB plus diuretic group (n = 74) (P = 0.791).

Incidence and severity of adverse effects

AVEC 2012, comparing ARB (n = 17), diuretic (n = 13) and ACE inhibitor (n = 17) groups, reported incidence of the 'most common and serious' adverse events for the three treatment groups with no significant difference between the groups. One participant withdrew due to adverse events from each of the ACE inhibitor and ARB groups. CAMUI 2013 reported no participants withdrawn from the ARB plus CCB group (n = 68) due to adverse events, and one participant withdrawn from ARB plus diuretic group (n = 74) (due to uric acid elevation).

Quality of life (QOL)

Neither AVEC 2012 nor CAMUI 2013 reported QOL measures.

Discussion

Summary of main results

The studies included in this systematic review do not provide evidence that pharmacological treatment of hypertension reduces cognitive impairment or dementia. Firm conclusions cannot be drawn, however, due to the certainty of the evidence available, as summarised in the 'Summary of findings' table.

Regarding our secondary outcomes, we found evidence, as expected, that pharmacological treatment of hypertension reduces blood pressure. We found evidence that pharmacological treatment of blood pressure causes adverse events but not convincing evidence that this leads to discontinuation of treatment. We were unable to draw conclusions regarding quality of life measures.

Overall completeness and applicability of evidence

This review identified 12 studies which address the updated review question of whether pharmacological treatment of hypertension prevents cognitive decline or dementia. However, these studies were not sufficient to answer the review question.

The shift towards lower BP targets is due to the accumulating evidence supporting stricter control of BP to prevent cardiovascular morbidity and mortality (Whelton 2018; Williams 2018; NICE 2019). The context in which we continue to test the hypothesis that treatment of BP ‐ in addition to reducing cardiovascular events ‐ will reduce dementia and cognitive decline, is therefore further constrained.

In order to answer our research question of whether pharmacological treatment of hypertension reduces dementia and cognitive decline, we need to consider whether to treat, which agent to treat with and at which BP target to aim. The inclusion of ACCORD MIND 2014 and SPRINT MIND 2019 in this review increases the relevance of the evidence to the research question but does not provide evidence that intensive BP management reduces dementia or cognitive decline.

Included studies with the longest durations are unlikely to have sufficient follow‐up periods to provide evidence of cognitive benefit from pharmacological antihypertensive treatment, if it does indeed exist. The shorter, more recent trials certainly do not afford such an opportunity. The testing of pathophysiological hypotheses in the shorter studies of AVEC 2012 explain why these brief durations of follow‐up were undertaken.

The utilisation of drug combinations within and across studies means conclusions regarding individual drug classes cannot be drawn, especially with the small numbers of participants in the newly included studies. In Syst Eur 1998, it was speculated that dementia prevention was facilitated by the neuroprotective role of the CCB nitrendipine. One meta‐analysis suggests differential effects between dihydropyridine calcium channel blockers (CCBs) and diuretics, on the one hand, and inhibitors of the renin‐angiotensin system, on the other (Wang 2003). Potential different class effects have been considered elsewhere (Levi Marpillat 2013; Peters 2014; Peters 2018), and further work is ongoing.

It is important to note that cardiovascular, including cerebrovascular, outcomes were the primary endpoints in the majority of the included studies in this review, with dementia and cognitive function as secondary outcomes. The earlier trials were terminated once benefits for the primary endpoints were shown. It is therefore possible that beneficial effects on cognition were not observed prior to the cardiovascular benefits becoming apparent.

For these reasons, the studies identified are not sufficient to answer the research question.

RCTs have provided evidence to support the active treatment of hypertension, even in the very elderly. Interestingly, a retrospective analysis has shown no interaction effect between frailty and the effect of antihypertensive treatment in HYVET (Warwick 2015). Given the shift towards age‐specific BP thresholds and targets, as evidenced by the ESC guidelines, it is imperative that studies include participants across age ranges. The included studies did cover a relatively broad age range.

Placebo‐controlled trials, at least at the BP thresholds used in the older RCTs, are no longer directly relevant and the expansion of our inclusion criteria reflects this. We excluded studies without specific BP entry criteria, which may have excluded potentially relevant studies such as ONTARGET 2008, TRANSCEND 2008, ADVANCE 2010 and the recently published cognitive sub‐study of HOPE 3 2016.

ACCORD MIND 2014 recruited only diabetics. LOMIR MCT IL 1996 and Perez Stable 2000 recruited participants with diastolic dysfunction. Included studies all used office BP measurements; continuous BP monitoring is assumed to be similar but may affect validity.

Open‐label extension studies

Only those cognitive outcomes measured during the study period were analysed and reported in HYVET 2008. Syst Eur 1998 reported both the per protocol analysis, including only data collected while participants were on double‐blind treatment, and intention‐to‐treat analysis, including data collected during the study period but while some participants were on open‐label treatment (median follow‐up two years). It is the latter results which are included in this review. A further open‐label, active treatment follow‐up study in the same population based on the original active study medication was later completed (Forette 2002). The median follow‐up was 3.9 years. From the placebo group, 80.5% of participants ended up on antihypertensive medication due to blood pressure level. Throughout follow‐up, systolic/diastolic blood pressure was 7.0/3.2 mmHg higher in the 1417 participants on placebo than in the 1485 participants originally randomised to antihypertensive treatment. Compared with the controls, long‐term antihypertensive therapy reduced the risk of dementia by 55% from 7.4 to 3.3 cases per 1000 patient‐years (43 versus 21 cases, P < 0.001). There was no significant change in MMSE between placebo and antihypertensive treatment groups.

Subsequent to publication of ACCORD MIND 2014, results of an open‐label observational extension study, ACCORDION MIND (Murray 2017), were published showing no difference in cognitive outcomes between the intensive and standard groups at 80 months. There was significant loss to follow‐up, and the differences in BP seen between the intensive and standard BP groups during the study were not maintained.

At the end of the SPRINT MIND 2019 study phase in August 2015, the mean SBP in the intensive and standard groups were 121.6 mmHg (95% CI 120.8 to 122.3 mmHg) and 134.8 mmHg (95% CI 134.1 to 135.6 mmHg), respectively, with a mean difference of 13.3 mmHg (95% CI 12.3 to 14.3 mmHg) between groups. During the extended study visits in 2017 to 2018, the between‐group mean difference had reduced to 6.4 mmHg (95% CI 4.2 mmHg to 8.5 mmHg) which the investigators attributed largely to the increase in SBP in the previously intensively treated group to 129.2 mmHg (95% CI 127.7 mmHg to 130.7 mmHg).

Outcome measures

As discussed by Iadecola 2016 and Elias 2018, cognitive outcome measures used in studies to date, especially those with younger participants and shorter study durations, are likely inadequate, and have contributed to under powering of studies. More thorough cognitive assessments and outcomes, focused on the most relevant domains ‐ for example, executive function ‐ will in time address these problems.

Non‐pharmacological interventions

In keeping with the original review (McGuinness 2006) and update (McGuinness 2009), non‐pharmacological studies were not included in this update. Interest is now moving to more complex interventions, targeting multiple vascular risk factors and combining pharmacological and behavioural elements. These require different review methods and will be covered in a new Cochrane Review.

Quality of the evidence

We identified 12 studies including 30,412 participants. The body of evidence identified in this review does not allow robust conclusions to be drawn regarding whether pharmacological treatment of hypertension prevents dementia and cognitive impairment.

Our concerns regarding indirectness for the dementia and cognitive impairment outcomes were due to relatively short studies ‐ several of which did not report dementia incidence ‐ being included in a review asking a question regarding pharmacological antihypertensive treatment, which tends, in practice, to be a long‐term undertaking. The participants in the included studies were, on average, older than those in whom treatment of hypertension has been associated with later cognitive benefit in observational studies (Walker 2017). The window for successfully treating hypertension with a view to preventing cognitive decline and dementia may have already passed by the time these participants were enrolled in the studies. In addition, we have combined studies employing different antihypertensive agents, although as outlined above emerging evidence would seem to suggest that certain agents may have disease‐modifying benefits over others. For nine of the included studies, cognition was a secondary endpoint. Only ACCORD MIND 2014, Syst Eur 1998 and HYVET 2008 were powered at the outset to detect a significant difference in cognitive decline between the groups. Cardiovascular endpoints continue to be the main outcome measure for hypertension studies and continue to be the endpoints on which treatment guidelines are based (Reboussin 2018, Williams 2018). This increases the risk of indirectness in the studies where cognitive outcomes measures are add‐ons.

The heterogeneity of the study methods and the results give rise to concern regarding inconsistency for the dementia and cognitive impairment outcomes. In particular, the varying durations of the wider range of studies included in this updated review are of concern.

We have deemed the included studies to be at a serious risk of bias overall for all outcomes. This is primarily due to attrition and contamination bias.

The loss to follow‐up of participants who did not complete follow‐up cognitive assessments increases the risk of attrition bias. As MRC 1996, ACCORD MIND 2014 and SPRINT MIND 2019 show, these participants may not be randomly lost to follow‐up. A follow‐up study to SHEP 1991, carried out by Di Bari 2001, aimed to evaluate whether assessment of cognitive and functional outcomes was biased by differential dropout. Characteristics of participants who did or did not participate in follow‐up cognitive assessments were compared. Assignment to placebo group and the occurrence of cardiovascular events independently predicted missed assessments. When 20% to 30% of the participants who missed the assessment were assumed to be cognitively impaired, assignment to antihypertensive treatment reduced the risk of cognitive impairment. The authors concluded that, in SHEP 1991, the cognitive evaluations were biased toward the null effect by differential dropout. This might have obscured the appraisal of a protective effect of treatment on the incidence of cognitive decline in older hypertensive adults. It is also important to note that HYVET 2008 and Syst Eur 1998 were halted early on safety grounds.

The treatment of participants in placebo groups with antihypertensive agents was necessitated by the changing ethical requirements as a result of emerging evidence in the earlier studies. In addition, those randomised to a specific treatment group often went on to require additional antihypertensive agents from a different class. A large proportion of participants included in this review ended studies on a different treatment to which they had been initially randomised (see Assessment of risk of bias in included studies section above). The high proportions of the placebo group in SCOPE 2003 and Zhang 2018 are a particular cause for concern. Our sensitivity analysis removing SCOPE from the dementia outcome analysis made no difference. Removing both studies from the change in MMSE analysis served only to slightly reduce the statistical significance of the difference in MMSE detected between the groups.

Potential biases in the review process

We used well‐described, stringent search criteria and methods for this review. Two authors (ELC and BMcG) independently reviewed search results and any disagreements were settled by consensus (ELC, BMcG, ST and PP). We reviewed the high number of systematic reviews addressing similar questions and no additional studies were identified, increasing the likelihood that we identified all relevant studies.

It is possible that our study selection has introduced bias. We chose to include only studies of participants with hypertension, as diagnosed using published criteria. We excluded studies which did not have BP entry criteria (ADVANCE 2010, ONTARGET 2008, TRANSCEND 2008 and HOPE 3 2016). We feel these are appropriate criteria for this review. They do introduce the risk that we have excluded studies that are relevant to our study question. The reasons for exclusion of these studies from the current meta‐analysis, according to the prespecified criteria in the Methods, may be found in the Characteristics of excluded studies section.

We attempted to contact the authors of the CHIEF 2012 study but were unsuccessful.

Agreements and disagreements with other studies or reviews

Several systematic reviews and meta‐analyses have considered related questions.

Parsons 2016 systematically reviewed placebo‐controlled RCTs of antihypertensive treatment with a mean participant age of 65 years or older. Their meta‐analysis contained the same four RCTs as our 2009 review.

Weiss 2017 systematically reviewed studies considering intensive blood pressure treatment with a mean participant age of at least 60 years. They described results from SCOPE 2003, SHEP 1991, Syst Eur 1998, HYVET 2008, MRC 1996, ACCORD MIND 2014 and PROGRESS 2003. Their only meta‐analysis incorporated ACCORD MIND 2014 and three observational studies and found that, in those studies combined, using antihypertensives to achieve moderate BP control had no effect on rates of incident dementia (OR 0.89, 95% CI 0.74 to 1.07).

Fink 2018 conducted a wide‐ranging systematic review of the literature and considered studies in three groups – comparing the effects of pharmacological antihypertensive treatments versus placebo, intensive versus standard pharmacological antihypertensive approaches, and different pharmacological antihypertensive treatments on the outcomes of dementia, mild cognitive impairment (MCI) and cognitive domain‐specific tests. Fink 2018 included shorter studies than we included, with a minimum required duration of six months (Starr 1996; Fogari 2003; Fogari 2006; Goldstein 1990), studies without BP entry criteria (ADVANCE 2010; ONTARGET 2008; TRANSCEND 2008), and a study in which the majority of participants were likely to have dementia at the outset (Tedesco 1999). They concluded that 'low‐strength evidence in adults with normal cognition showed that antihypertensive treatment versus placebo does not reduce risk for dementia, and moderate strength evidence showed no difference between these treatments on global cognitive screening tests' and otherwise presented narrative conclusions.

Larsson 2018 conducted a systematic review and meta‐analysis of RCTs and prospective studies of antihypertensive drug treatment with dementia as an outcome measure. They combined SCOPE 2003, SHEP 1991, Syst Eur 1998, HYVET 2008, and PROGRESS 2003 in an analysis and reported a non‐significant relative risk of dementia of 0.84 (95% CI 0.69 to 1.02) for active treatment compared to placebo.

Van Middelaar 2018 systematically reviewed studies of BP lowering by both pharmacological and non‐pharmacological means, measuring dementia incidence. They included SCOPE 2003, SHEP 1991, Syst Eur 1998, HYVET 2008, studies of secondary stroke prevention (PROGRESS 2003 and PRoFESS 2008), studies without a BP entry criteria (ADVANCE 2010) and multicomponent interventions (PreDIVA 2016 and Look AHEAD 2017) and reported a pooled risk ratio of 0.93 (95% CI 0.84 to 1.02) for intervention compared to control.

Elias 2018 considered placebo‐controlled RCTs of antihypertensive medication use that included cognitive outcomes in their quest to examine the quality of cognitive outcomes assessment in these trials.

Peters 2019 combined data from eight RCTs of antihypertensive therapy that measured dementia incidence (SHEP 1991, SCOPE 2003, HYVET 2008, PROGRESS 2003, PRoFESS 2008, ADVANCE 2010 and SPRINT MIND 2019), and reported a relative risk reduction of dementia of 0.93 (95% CI 0.86 to 1.00) associated with antihypertensive treatment.

Most recently, Hughes and colleagues have systematically reviewed RCTs of antihypertensive agents of 1000 or more participants and lasting at least one year, that measured dementia incidence or change in cognitive scores, or both (Hughes 2020). They included SHEP 1991, MRC 1996, Syst Eur 1998, PROGRESS 2003, SCOPE 2003, HYVET 2008, PRoFESS 2008, ADVANCE 2010, ONTARGET 2008, TRANSCEND 2008, ACCORD MIND 2014, SPS3 2014, SPRINT MIND 2019, and HOPE 3 2016. Meta analysis of 12 studies (excluding MRC 1996 and ACCORD MIND 2014) found a reduced risk of dementia or cognitive impairment (7.0% versus 7.5%; OR 0.93, 95% CI 0.88 to 0.98; absolute risk reduction 0.39%, 95% CI 0.09% to 0.68%). Meta‐analysis of eight studies (PROGRESS 2003, SCOPE 2003, HYVET 2008, PRoFESS 2008, TRANSCEND 2008, ONTARGET 2008 (2 arms), SPRINT MIND 2019 and HOPE 3 2016) showed a reduced risk of cognitive decline (20.2% versus 21.1%; OR 0.93, 95% CI 0.88 to 0.99; absolute risk reduction 0.71%, 95% CI 0.19% to 1.2%).

Gupta and colleagues systematically reviewed RCTs of antihypertensive treatment to reduce cognitive decline but did not require participants to be hypertensive or free of dementia (Gupta 2020). They did require mean or median participant age to be 60 years or older, and the comparison group to be no intervention, placebo standard of care or a higher BP goal and at least one year follow‐up. They included nine RCTs with 34,994 participants: SHEP 1991, MRC 1996, Syst Eur 1998, SCOPE 2003, Pantoni 2005, HYVET 2008, TRANSCEND 2008, ACCORD MIND 2014 and SPRINT MIND 2019. Meta‐analysis yielded a standardised mean difference change in cognition of ‐0.049 (95% CI ‐0.078 to ‐0.019) associated with antihypertensive treatment.

Several authors have conducted systematic reviews comparing different classes of antihypertensive agents with no firm conclusions drawn (Levi Marpillat 2013; Peters 2014; Zhuang 2016).

Authors' conclusions

Implications for practice.

The inclusion of additional studies does not alter the conclusions from the initial review and subsequent update. There is insufficient evidence to recommend pharmacological treatment of hypertension BP to prevent dementia or cognitive decline.

There is evidence to support pharmacological treatment of BP to prevent cardiovascular and cerebrovascular morbidity and mortality. People with hypertension are therefore likely to receive pharmacological treatment. It remains unclear what the appropriate BP thresholds are for pharmacological treatment of hypertension in the prevention of cognitive decline or dementia, which agents ‐ if any ‐ confer the greatest cognitive benefit, what the appropriate BP targets should be, and at which points in the life course.

Implications for research.

Despite the wealth of evidence from observational studies associating mid‐life hypertension and late‐life cognitive impairment and dementia, and reduction of risk associated with treatment of hypertension, there remains a stubborn lack of RCT evidence to support the hypothesis that treatment of hypertension prevents dementia and cognitive decline. The pervading wisdom is that this reflects the shortcomings of studies to date rather than a rebuttal of the hypothesis.

Studies need to include mid‐life participants with decades‐long follow‐up. Studies need to be designed to consider cognition as a primary outcome measure. Appropriate cognitive outcome measures need to be employed. Studies comparing BP targets have been undertaken (ACCORD MIND 2014, SPRINT MIND 2019). Future studies will need to compare different classes of antihypertensive drugs. Future studies will also need to include biomarkers of the different dementia‐causing neurodegenerative processes so we can understand the underlying pathophysiology.

Trials of BP lowering treatment regimens should therefore be designed to include participants with hypertension in middle age with prolonged follow‐up and assessment of cognition into older age. These trials should be adequately powered to determine, as a primary outcome, whether cognitive impairment and dementia can be prevented.

Feedback

Peer review and editorial comment, July 2009

Summary

Peer reviewers and editor suggested rewriting parts of the background section and discussion from the initial review to facilitate ease of reading as parts of these sections were also repeated in other sections of the review; for example, characteristics of included studies.

A few inconsistencies in presentation through the various sections of the review were highlighted.

Comments throughout initial review and this update have returned to the included and excluded studies.

Reply

Rewriting of the background and discussion sections in the updated review were done as suggested.

Consistent use of terms and phrasing now used throughout the updated review.

The brief given to the reviewers by the Review Group was to analyse the effect of blood pressure lowering on cognitive impairment in participants with no evidence of prior cerebrovascular disease. The reasons for exclusion of certain studies and the problems with analysis of the included studies have been dealt with in detail in the initial review and in this update.

Contributors

ST and BMcG; reviewed and commented on by PP.

What's new

Date	Event	Description
7 July 2020	New search has been performed	The most recent search for this review was performed on 07 July 2020
7 July 2020	New citation required but conclusions have not changed	New studies added, methods updated, conclusions unchanged

History

Protocol first published: Issue 1, 2003
Review first published: Issue 2, 2006

Date	Event	Description
29 June 2019	New search has been performed	Top‐up searches were performed for this review in December 2014, November 2015, October 2016 and August 2017. An update search was performed for this review on 28 June 2018 and 29 June 2019. Authors were contacted for further information as described. . Non‐pharmacological studies were not included in this update and will be the focus of a separate review. Feedback (from 2009) regarding error in interpretation of HYVET‐COG results incorporated into updated review. New studies included. New author added.
27 July 2009	Feedback has been incorporated	Further feedback from editor incorporated into updated review.
26 November 2008	New citation required and conclusions have changed	The authors have recommended that no further trials are conducted assessing the effect of antihypertensive interventions for cognition.
26 November 2008	New search has been performed	Update of search of February 2008 retrieved several studies for consideration by the authors: upon assessment, one new study has been included (HYVET 2008); there are now four studies included in the review, with a total of 15,936 participants. In addition, two new studies have been excluded.
8 February 2006	New citation required and conclusions have changed	Substantive amendment
30 September 2005	New search has been performed	Minor update

Appendices

Appendix 1. Sources searched and search strategies

Source	Search strategy	Hits retrieved
1. ALOIS (www.medicine.ox.ac.uk/alois) [last searched 07 July 2020]	hypertension OR hypertensive OR hypertensives OR HYP OR "blood pressure"	Dec 2014: 10 Nov 2015: 13 Oct 2016: 4 Aug 2017: 0 Jun 2018: 0 Jun 2019: 1 Aug 2020: 72
2. MEDLINE In‐process and other non‐indexed citations and MEDLINE 1946‐present (Ovid SP) [last searched 07 July 2020]	1. (cognit* or memor* or "quality of life").ti,ab. 2. Cognition/ 3. Memory/ 4. dement*.ti,ab. 5. Dementia/ 6. or/1‐5 7. (hypertens* or antihypertens* or anti‐hypertens* or "blood pressure").ti,ab. 8. Blood Pressure/ 9. Hypertension/ 10. Antihypertensive Agents/ 11. (diuretic* or "calcium channel blocker*" or "ACE inhibitor*").ti,ab. 12. or/7‐11 13. 6 and 12 14. randomized controlled trial.pt. 15. controlled clinical trial.pt. 16. randomly.ab. 17. drug therapy.fs. 18. groups.ab. 19. placebo.ab. 20. or/14‐19 21. 13 and 20 22. (2008* or 2009* or 2010* or 2011* or 2012* or 2013* or 2014*).ed. 23. 21 and 22	Dec 2014: 2840 Nov 2015: 412 Oct 2016: 963 Aug 2017: 958 Jun 2018: 743 Jun 2019: 726 Aug 2020: 806
3. Embase 1980‐27 June 2019 (Ovid SP) [last searched 07 July 2020]	1. (cognit* or memor* or "quality of life").ti,ab. 2. Cognition/ 3. Memory/ 4. dement*.ti,ab. 5. Dementia/ 6. or/1‐5 7. (hypertens* or antihypertens* or anti‐hypertens* or "blood pressure").ti,ab. 8. Blood Pressure/ 9. Hypertension/ 10. Antihypertensive Agents/ 11. or/7‐10 12. 6 and 11 13. randomized controlled trial/ 14. controlled clinical trial/ 15. "randomly allocated".ti,ab. 16. randomi?ed.ti,ab. 17. "randomly divided".ab. 18. placebo.ab. 19. or/13‐18 20. 12 and 19 21. (2008* or 2009* or 2010* or 2011* or 2012* or 2013* or 2014*).em. 22. 20 and 21	Dec 2014: 2535 Nov 2015: 617 Oct 2016: 1117 Aug 2017: 1069 Jun 2018: 1433 Jun 2019: 648 Aug 2020: 911
4. PSYCINFO 1806‐June week 3 2019 (Ovid SP) [last searched 07 July 2020]	1. (cognit* or memor* or "quality of life").ti,ab. 2. Cognition/ 3. Memory/ 4. dement*.ti,ab. 5. Dementia/ 6. or/1‐5 7. (hypertens* or antihypertens* or anti‐hypertens* or "blood pressure").ti,ab. 8. Blood Pressure/ 9. Hypertension/ 10. (diuretic* or "calcium channel blocker*" or "ACE inhibitor*").ti,ab. 11. or/7‐10 12. 6 and 11 13. randomi?ed.ti,ab. 14. randomly.ab. 15. "double‐blind*".ti,ab. 16. "single‐blind*".ti,ab. 17. groups.ab. 18. placebo.ab. 19. or/13‐18 20. 12 and 19 21. (2008* or 2009* or 2010* or 2011* or 2012* or 2013* or 2014*).up. 22. 20 and 21	Dec 2014: 461 Nov 2015: 87 Oct 2016: 138 Aug 2017: 185 Jun 2018: 143 Jun 2019: 107 Aug 2020: 110
5. CINAHL (EBSCOhost) (1980 to June 2018) [last searched 07 July 2020]	S1 TX cognit* OR memor* OR "quality of life" S2 (MH "Cognition") S3 (MH "Memory") S4 TX dement* S5 (MH "Dementia") S6 TX hypertens* S7 TX antihypertens* S8 TX anti‐hypertens* S9 TX "blood pressure" S10 (MH "Blood Pressure") S11 (MH "Hypertension") S12 (MH "Antihypertensive Agents") S13 TX diuretic* S14 TX "calcium channel blocker*" S15 TX "ACE inhibitor*" S16 (MH "Angiotensin‐Converting Enzyme Inhibitors") S17 (MH "Calcium Channel Blockers") S18 (MH "Candesartan") OR (MH "Candesartan Cilexetil Hydrochlorothiazide") S19 (MH "Chlorthalidone") S20 TX Chlorthalidone S21 TX Candesartan S22 TX Enalapril S23 TX Ramipril S24 TX Quinapril S25 TX Perindopril S26 TX Lisinopril S27 TX Benazepril S28 TX Imidapril S29 TX Trandolapril S30 TX Cilazapril S31 TX Fosinopril S32 (MH "Angiotensin II Type I Receptor Blockers") S33 S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 OR S20 OR S21 OR S22 OR S23 OR S24 OR S25 OR S26 OR S27 OR S28 OR S29 OR S30 OR S31 OR S32 S34 S1 OR S2 OR S3 OR S4 OR S5 S35 S33 AND S34 S36 TX randomly S37 TX placebo S38 TX randomised S39 TX randomized S40 TX groups S41 TX "double‐blind*" S42 TX "single‐blind*" S43 TX RCT S44 (MH "Randomized Controlled Trials") S45 S36 OR S37 OR S38 OR S39 OR S40 OR S41 OR S42 OR S43 OR S44 S46 S35 AND S45	Dec 2014: 243 Nov 2015: 182 Oct 2016: 266 Aug 2017: 274 Jun 2018: 358 Jun 2019: 907 Aug 2020: 744
6. Web of Science – all databases [includes: Web of Science (1945‐present); BIOSIS Previews (1926‐present); MEDLINE (1950‐present); Journal Citation Reports] [last searched 07 July 2020]	(dement* OR alzheimer* OR "lew* bod*" OR frontotemporal OR FTD OR FTLD OR "cognit*impair*") AND TOPIC: (hypertension OR hypertensive OR hypertensives OR HYP OR "blood pressure") ANDTOPIC: (randomly OR randomised OR randomized OR placebo OR "double‐blind*" OR trial OR RCT OR CCT)	Nov 2015:818 (all dates) Oct 2016: 272 Aug 2017: 213 Jun 2018: 164 Jun 2019: 50 Aug 2020: 174
7. LILACS (BIREME) [last searched 07 July 2020]	"blood pressure" OR "da pressão sanguínea" OR hipertensos OR hypertension [Words] and dementia OR demencia OR cognition OR cognitive [Words]	Nov 2015: 25 (all dates) Oct 2016: 3 Aug 2017: 19 Jun 2018: 0 Jun 2019: 0 Aug 2020: 33
8. CENTRAL (The Cochrane Library) (November 2015) [last searched 07 July 2020]	#1 cognit* or memor* #2 dement* #3 MeSH descriptor: [Cognition] explode all trees #4 MeSH descriptor: [Memory] explode all trees #5 MeSH descriptor: [Dementia] explode all trees #6 #1 or #2 or #3 or #4 or #5 #7 hypertens* or antihypertens* or anti‐hypertens* or "blood pressure" #8 MeSH descriptor: [Blood Pressure] explode all trees #9 MeSH descriptor: [Hypertension] explode all trees #10 MeSH descriptor: [Antihypertensive Agents] explode all trees #11 diuretic* or "calcium channel blocker*" or "ACE inhibitor*" #12 #8 or #9 or #10 or #11 #13 #6 and #12 Publication Year from 2013 to 2015, in Trials	Dec 2014: 225 Nov 2015: 79 Oct 2016: 53 Aug 2017: 66 Jun 2018: 181 Jun 2019: 62 Aug 2020: 147
9. Clinicaltrials.gov (www.clinicaltrials.gov) [last searched 07 July 2020]	(antihypertensive OR antihypertensives OR "anti‐hypertensive" OR diuretic OR "calcium channel blocker" OR "ACE inhibitor") AND (dementia OR cognition OR cognitive)	Dec 2014: 12 Nov 2015: 6 Oct 2016: 15 Aug 2017: 17 Jun 2018: 16 Jun 2019: 11 Aug 2020: 16
10. ICTRP Search Portal (http://apps.who.int/trialsearch) [includes: Australian New Zealand Clinical Trials Registry; ClinicalTrilas.gov; ISRCTN; Chinese Clinical Trial Registry; Clinical Trials Registry – India; Clinical Research Information Service – Republic of Korea; German Clinical Trials Register; Iranian Registry of Clinical Trials; Japan Primary Registries Network; Pan African Clinical Trial Registry; Sri Lanka Clinical Trials Registry; The Netherlands National Trial Register] [last searched 29 June 2019. Database not available 07 July 2020]	(antihypertensive OR antihypertensives OR "anti‐hypertensive" OR diuretic OR "calcium channel blocker" OR "ACE inhibitor") AND (dementia OR cognition OR cognitive)	Dec 2014: 4 Nov 2015: 4 Oct 2016: 4 Aug 2017: 2 Jun 2018: 24 Jun 2019: 1
TOTAL before de‐duplication		Dec 2014: 6330 Nov 2015: 2239 Oct 2016: 2835 Aug 2017: 2803 Jun 2018: 3062 Jun 2019: 2513 Aug 2020: 3013 TOTAL: 22795
TOTAL after de‐duplication and first‐assessment by the CDCIG information specialist		Dec 2014: 212 Nov 2015: 309 Oct 2016: 145 Aug 2017: 220 Jun 2018: 280 Jun 2019: 178 Aug 2020: TOTAL: 1344

Data and analyses

Comparison 1. Incidence of dementia.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Number of cases of dementia placebo studies	4	15427	Odds Ratio (M‐H, Random, 95% CI)	0.89 [0.72, 1.09]

Comparison 2. Cognitive change from baseline.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Change in MMSE placebo studies	4	9435	Mean Difference (IV, Fixed, 95% CI)	0.20 [0.10, 0.29]

Comparison 3. Change in blood pressure from baseline.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Change in systolic blood pressure level (mmHg) placebo studies	6	17793	Mean Difference (IV, Fixed, 95% CI)	‐9.25 [‐9.73, ‐8.78]
3.2 Change in diastolic blood pressure level (mmHg) placebo studies	5	15209	Mean Difference (IV, Fixed, 95% CI)	‐2.47 [‐2.70, ‐2.24]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

ACCORD MIND 2014.

Study characteristics
Methods	Multisite study. Randomised, double 2x2 factorial trial. All participants entered into intensive versus standard hypoglycaemic treatment trial plus either BP or lipid trial. Randomisation via coordinating centre using unique, computer‐generated sequences using permuted blocks of 4, 8 or 12 participants. Participants in BP trial randomised to open‐label, intensive versus standard BP lowering treatment. Prespecified subgroup considered cognitive outcomes. Study duration: 40 months.
Participants	Geographical area: North America. Of the 6 American and 1 Canadian clinical centre networks in the overall ACCORD study, 6, including 54 clinics, participated in ACCORD MIND. ACCORD study criteria (n = 10,251): Type II diabetic patients with HbA1c ≥ 7.5%, aged 40 ‐ 79 years, at high risk for cardiovascular disease events. BP study criteria (n = 4,733): SBP 130‐180 mmHg with minimal proteinuria. ACCORD MIND study criteria (n = 2,977): age ≥ 55. All participants recruited into ACCORD MIND within 45 days of randomisation into ACCORD study. ACCORD MIND participants recruited between August 2003 and December 2005. Participants randomised to BP or lipid arm. N = 1,439 participants were entered into the ACCORD MIND BP trial, n = 745 randomised to the intensive BP lowering arm, n = 694 to the standard BP lowering arm.
Interventions	Visit schedule differed between the intensive and standard BP measurement groups. Participants in the intensive BP group were seen at least monthly until SBP < 120 mmHg achieved at, minimum, months 1, 2, 3 and 4, and two‐monthly thereafter. Milepost visits scheduled at 4‐month intervals for first 2 years with a different class of antihypertensive added if systolic BP ≥ 120 mmHg. Participants in the standard BP lowering group had therapy up‐titrated if SBP was ≥ 160 mmHg on one visit or ≥ 140mmHg on two successive visits. Therapy was down‐titrated if SBP was < 130 mmHg on one visit or < 135mmHg on two successive visits.
Outcomes	Analysed in review: digit symbol substitution test (DSST); Mini‐Mental State Examination; Rey Auditory Verbal Learning Test; Stroop test; Health Related Quality of Life as measured using the Medical Outcomes Study 36‐item short form health survey, Diabetes Treatment Satisfaction Questionnaire and Patient Health Questionnaire‐9. Not analysed in review: total deaths; deaths due to CVD; non fatal myocardial infarction (MI), non fatal stroke; revascularisation; heart failure hospitalisations; microvascular composite outcomes; biochemical analyses; Patient Health Questionnaire; self‐reported ability to manage diabetes therapy questionnaire.
Notes	Exclusion criteria overall ACCORD study included: recent significant hypoglycaemic event; BMI > 45; serum creatinine > 1.5 mg/dL; active liver disease; cardiovascular event or procedure within previous 3 months; weight loss; malignancy; 'any condition that, in the judgement of clinical study staff members, would preclude full participation in the study e.g. preexisting clinical evidence of dementia, substance abuse'. Of the n = 745 randomised to the intensive BP lowering arm and n = 694 to the standard BP lowering arm, n = 618 (of n = 659) and n = 649 (of n = 690), respectively, had 40 month data reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'unique randomisation sequences for ACCORD were computer generated centrally at the coordinating centre using permuted blocks of 4, 8 or 12 participants' 'baseline characteristics were compared between intervention groups...characteristics that differed between groups were adjusted for in post‐hoc analyses'
Allocation concealment (selection bias)	Low risk	Quote: ' unique randomisation sequences for ACCORD were computer generated centrally at the coordinating centre using permuted blocks of 4, 8 or 12 participants'
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: '(the lipid trial) is the only masked intervention in ACCORD'.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: '(the lipid trial) is the only masked intervention in ACCORD'. Quality control measures are described for cognitive testing but no mention made of blinding the assessors.
Incomplete outcome data (attrition bias) All outcomes	High risk	Imputation methods for missing data reported. Numbers deceased, lost to follow‐up and with missing assessments reported (12% missing or deceased intensive group, 11% standard group). Of the 745 and 694 in the intensive and standard BP intervention groups, 690/740 ad 659/690 with baseline DSST measurements were included in the final analyses. Those missing follow‐up data were significantly older, had higher SBP and lower DSST scores at baseline.
Selective reporting (reporting bias)	Low risk	40‐month BP measurements reported for 86% of the randomised participants (83% intensive group, 88% standard group). 40‐month results reported for DSST, the prespecified outcome, but for a selection only of the secondary cognitive outcome measures, stating that 'mean 40‐month cognitive function did not differ between intervention groups for any of the other 3 cognitive tests'.
Other bias	Unclear risk	Only included participants with type 2 diabetes mellitus. Quote: 'All participants receive nutrition and physical activity counselling, as well as a recommendation to use aspirin daily...current smokers receive smoking cessation counselling...information on current guidelines for lipids and blood pressure treatment is provided to participants' personal physicians' Quote: 'regular visit schedules differ by treatment group assignment' schedule for intensive group described more thoroughly than that for standard group, so magnitude of difference unclear. Visit effect accounted for in statistical analyses. Participants had already been randomised to intensive or standard BP management when they were invited to participate in the MIND substudy. To account for this, differences in baseline characteristics between the two groups were analysed and corrected for in post‐hoc analyses. Add‐on antihypertensive therapy was required for one participant each in the ARB and diuretic groups and three in the ACE inhibitor group.

AVEC 2012.

Study characteristics
Methods	Single centre study. Three‐arm double‐blind randomised controlled trial. Randomisation: computer‐generated random allocation sequence. Study personnel and participants blinded to allocation. Study duration: 1 year.
Participants	Geographical region: USA. Setting: community. Aged ≥ 60 years. BP entry criteria: SBP > 140 mmHg or DBP > 90 mmHg or receiving antihypertensive therapy. Early cognitive impairment without dementia defined as executive impairment, as measured using the CLOX test or early memory impairment as measured using the Repeatable Battery for the Assessment of Neuropsychological Status.
Interventions	Three treatment arms with starting doses: lisinopril 10 mg once daily; candesartan 8 mg once daily; hydrochlorothiazide 12.5 mg daily. Participants seen fortnightly, average of two seated BP measurements taken, medications up‐titrated until BP < 140/90. Lisinopril, candesartan and hydrochlorothiazide titrated up through 20 mg and 40 mg, 16 mg then 32 mg and 25 mg respectively. Next step long acting nifedipine added at 30 mg titrated through 60 mg to 90 mg then long acting metoprolol added at 12.5 mg titrated through 25 mg to 50 mg. Once target BP reached, participants followed for 12 months.
Outcomes	Analysed in this review: Trail Making Test A and B; Hopkins Verbal Learning Test ‐ Revised; Digit Span Test Not analysed in this review: health questionnaire; instrumental activities of daily living; cerebral blood flow haemodynamics
Notes	Exclusion criteria: untreated SBP > 200 mmHg or DBP > 110 mmHg or treated SBP > 180 mmHg or DBP > 100 mmHg; MMSE < 20 or clinical diagnosis of dementia or Alzheimer's disease; raised serum creatinine or potassium; treatment with > 2 antihypertensive agents; congestive heart failure; diabetes; stroke. Of the n = 18 (ACE inhibitor), n = 20 (ARB) and n = 15 (diuretic) participants recruited, n = 11, n = 9 and n = 11, respectively, completed 12‐month follow‐up.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'this random allocation is performed at the central pharmacy...randomisation is performed using computer generated random numbers leading to a random allocation sequence'
Allocation concealment (selection bias)	Low risk	Quote: 'this random allocation is performed at the central pharmacy...randomisation is performed using computer generated random numbers leading to a random allocation sequence'
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: 'both study personnel and participants are blinded to the group assignment'
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: 'both study personnel and participants are blinded to the group assignment'
Incomplete outcome data (attrition bias) All outcomes	High risk	Baseline data reported for the 47 participants of the 53 participants with successful middle cerebral artery isolation at baseline. Only 47/53 and 31/53 completed 6‐month and 12‐month follow‐up, respectively (40/47 and 29/47 with reported outcomes). No analysis of influences on loss to follow‐up reported.
Selective reporting (reporting bias)	Unclear risk	Intention to treat analyses used. Only significant differences in cognitive outcomes between the groups reported.
Other bias	Unclear risk	Study protocol published, stated aiming for 100 participants, but only 63 eligible individuals screened. 53 randomised but results reported for n = 47. Described as a pilot study. Study aimed to recruit those with early cognitive impairment but only MMSE used to exclude participants with more advanced cognitive impairment, diagnosis of dementia given as exclusion criteria but no mention of criteria for making such a diagnosis. Mean (SD) MMSE are relatively high however. Add‐on antihypertensive therapy utilies for n = 1 each in ARB group and diuretic group and n = 3 in ACE inhibitor group.

CAMUI 2013.

Study characteristics
Methods	Multisite study. Randomised, open‐label, parallel comparison study. Randomisation conducted by trial data centre using dynamic allocation method: stratified by age, sex and SBP. Study duration 1 year.
Participants	Geographical setting: Japan. Hospital outpatients. Epoch: recruitment November 2008 until November 2010, follow‐up until November 2011. Hypertensive participants aged ≥ 65 years who had not attained target BP with ≥ one month of ARB monotherapy. BP entry criteria: SBP > 140 mmHg and/or DBP > 90 mmHg or, in the case of diabetes or renal disease, SBP > 130 mmHg and/or DBP > 80 mmHg.
Interventions	Four‐week screening and run‐in phase. Participants switched from existing ARB therapy to either losartan 50 mg plus hydrochlorothiazide 12.5 mg or existing ARB therapy plus amlodipine 5 mg. Follow‐up every 3 months. If target BP (i.e. as per entry criteria) was not reached, a further ARB was added as a first step with an alpha‐blocker, beta‐blocker or antisympathetic agent added as the next step.
Outcomes	Analysed in review: MMSE Not analysed in review: laboratory tests
Notes	Exclusion criteria: secondary hypertension; heart failure greater than NYHA Class III; history of severe hepatic or renal disease; critical liver damage
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'the randomisation was conducted at the CAMUI trial data centre by the dynamic allocation method'
Allocation concealment (selection bias)	High risk	This was an open‐label study.
Blinding of participants and personnel (performance bias) All outcomes	High risk	This was an open‐label study.
Blinding of outcome assessment (detection bias) All outcomes	High risk	This was an open‐label study.
Incomplete outcome data (attrition bias) All outcomes	High risk	Total of n = 128 completed study follow‐up, numbers completing MMSE at 12 months not reported.
Selective reporting (reporting bias)	High risk	Intention‐to‐treat and per protocol analyses undertaken and reported for change in BP. Final group means only provided for MMSE data.
Other bias	Unclear risk	A sample size of n = 92 for each group was required; however, only n = 68 and n = 74 in each group were recruited. Seven from each group were withdrawn as they discontinued the study medication or because laboratory tests could not be checked. One adverse event (AE) in the ARB+D group reported. Additional drugs permitted, number of participants requiring additional drugs reported: ARB+D ‐ 7 at 3 months, 3 at 6 months, 3 at 9 months; ARB+C ‐ 4 at 3 months, 3 at 6 months, 3 at 9 months.

HYVET 2008.

Study characteristics
Methods	Multisite study. Randomisation: stratified by age, sex, and centre. Participants randomly allocated through interactive voice response system setup by coordinating centre. Participants blinded; investigators blinded; outcome assessors blinded. Planned duration: 5 years.
Participants	Geographic region: Europe, China, Australasia, Tunisia. 195 centres in 13 countries. Study setting: community. N = 3845 (60.5% female). Age: ≥ 80 years, mean 83.6 years. Ethnicity: not reported; 86 western Europe, 2144 eastern Europe, 1526 China, 19 Australasia, 70 Tunisia. Mean sitting blood pressure (BP) at entry: 173.0/90.8 mmHg. BP entry criteria: sustained systolic BP 160 to 199 mmHg (1 month apart); standing systolic BP > 140 mmHg; sitting diastolic BP 90 to 109 mmHg (protocol amendment 2003 < 110 mmHg). Of the total n = 3845 HYVET population (n = 1933 hypertensive treatment group, n = 1912 placebo), n = 3336 were recruited into HYVET COG (n = 1687 hypertensive treatment group, n = 1649 placebo).
Interventions	Minimum 2 month single daily placebo tablet run‐in phase with all other antihypertensive treatment stopped. Treatment: step 1 ‐ indapamide SR 1.5 mg/day; step 2 ‐ indapamide SR 1.5 mg/day and perindopril 2 mg/day; step 3 ‐ indapamide SR 1.5 mg/day and perindopril 4 mg/day. Control: matching placebos. Target sitting systolic BP < 150 mmHg and diastolic < 80 mmHg (sitting systolic BP ≥ 150 mmHg accepted if standing systolic BP < 120 mmHg). Average follow‐up: median 1.8 years (mean, 2.1; range, 0 to 6.5). Difference in BP at end of study (treatment ‐ control) systolic/diastolic: ‐15.0/‐6.1 mmHg (sitting).
Outcomes	Analysed in the review: blood pressure level; incidence of dementia; change in cognitive function (measured by MMSE); incidence of side effects. Not analysed in the review. Primary: fatal and non‐fatal stroke. Secondary: total mortality; cardiovascular mortality; cardiac mortality; stroke mortality; skeletal fracture rate. Other outcomes: incidence of retinal lesions; overt heart failure; renal failure; dissecting aortic aneurysm; acute myocardial infarction (MI). Cognitive function outcomes not specified in original trial protocol.
Notes	Exclusion criteria: known accelerated hypertension; overt clinical congestive heart failure, requiring treatment with diuretic or ACE inhibitor; renal failure (serum creatinine > 150 μmol/L); documented cerebral or subarachnoid haemorrhage in previous 6 months; condition expected to severely limit survival; known secondary hypertension; gout; clinical diagnosis of dementia; nursing home residence; contraindication to trial medication (serum potassium < 3.5 µmol/L or > 5.5 µmol/L); inability to stand or walk. Overall HYVET trial: % not on assigned treatment at 2 years: 0.8% active treatment; 0.6% control group. Of the n = 1687 hypertensive treatment group, n = 1649 placebo recruited to HYVET COG 96 died and 141 declined to continue and 132 died and 146 declined to continue in the two groups respectively, with 5 lost to follow‐up in each group. Due to early termination of the trial, not all participants reached at least one year of follow‐up. Of the n = 3336, only n = 1469 (i.e. 44%) attended their 2‐year follow‐up visit.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'Randomisation was stratified according to age and sex; permuted blocks of 4 and 6 of any 10 patients used to ensure roughly equal assignment to each of the two groups within large centres'
Allocation concealment (selection bias)	Low risk	Quote: 'when the coordinating centre has received the entry form on a patient and checked they are eligible, they inform the interactive voice response system (IVRS) to permit randomisation according to schedule. The IVRS then automatically, in real‐time, faxes the centre to inform them that their patient is eligible to enter the trial and that the local investigator can make a call to the system to receive number of the treatment pack that their patient is to receive'
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Active treatment and matching placebo. Allocation concealed from investigator and participant.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Active treatment and matching placebo. Allocation concealed from investigator and participant. End‐Points Committee blinded to treatment allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: 'Data from patients were analysed for the groups to which the patients were assigned, regardless of which study drugs (or which doses) the patients actually received and regardless of other protocol irregularities. Patients from closed centres were included in intention‐to‐treat population and contributed (data) up to the date of closure of the centre'. In the hypertensive treatment group, 303/1687 participants did not complete the study, of which five were lost to follow‐up. In the placebo group, 343/1649 did not complete the study of which, again, five were lost to follow‐up. Comment: Similar numbers for each event for which data were censored between treatment and placebo groups. Only 44% of participants completed a 2‐year follow‐up visit. Quote: 'repeating the analyses with all patients, regardless of whether they had more than one cognitive function assessment, did not affect the results'. Shorter duration of study more of an issue than attrition bias.
Selective reporting (reporting bias)	Low risk	Outcomes specified in the protocols reported in the main results and cognitive sub‐study.
Other bias	Low risk

LOMIR MCT IL 1996.

Study characteristics
Methods	Multisite study. Randomised, double‐blind, case‐controlled study. Age, weight, height, duration of hypertension, blood pressure and heart rate reported for groups at baseline. Study duration 1 year.
Participants	Geographical setting: Israel. Healthcare setting not specified but visits mentioned therefore likely outpatients. Epoch: unclear. Male participants aged 40 to 65 years with newly‐diagnosed or inadequately treated hypertension. BP entry criteria: DBP 95 to 105 mmHg. Total n = 368 participants: n = 124 (isradipine) n = 120 (methyldopa) n = 124 (placebo).
Interventions	Washout phase for those on pre‐existing hypertensive treatment followed by 2 to 4 weeks placebo period. Participants randomly assigned, at week 0, to 1.25 mg isradipine, 250 mg methyldopa or placebo twice daily. Dose doubled after 4 weeks if DBP still ≥ 95 mmHg. At week 6, captopril 25 mg daily (single‐blinded to participant only) added if DBP ≥ 95 mmHg. Dose of captopril added every 8 weeks if required. Participants examined every 4 weeks after the titration period.
Outcomes	Analysed in review: semantic memory tests at six and 12 months; QOL measures at six and 12 months
Notes	Exclusion criteria: secondary hypertension; malignant hypertension; unstable angina; recent myocardial infarction. Final numbers followed up unclear. Numbers completing cognitive assessments at 1 year not explicitly given, though dropouts described. Differential withdrawal from study groups: 15/124 isradipine group, 21/120 methyldopa group, 34/124 placebo group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: 'all patients who entered the study were randomly assigned to one of the three parallel treatment groups'
Allocation concealment (selection bias)	Unclear risk	Study described as double blind but details not provided.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Study described as double‐blind. Lead‐in placebo phase blind only to participants. Additional captopril described as single‐blinded.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Study described as double‐blind but blinding of assessors not described.
Incomplete outcome data (attrition bias) All outcomes	High risk	N = 21 withdrew prior to the double blind phase; group breakdown not stated. N = 70 withdrew (n = 15 isradipine group, n = 21 methyldopa group, n = 34 placebo group; for under‐treatment and side effects); of these, n = 60 were followed until the end of the study. Numbers completing cognitive assessments at 12 months not given.
Selective reporting (reporting bias)	Unclear risk	Intention‐to‐treat analyses appear to be reported.
Other bias	Unclear risk	No power calculation provided. Absolute numbers receiving captopril not reported ‐ graph of proportions provided, estimated approximately 20% (isradipine group) to 50% (placebo group). DBP < 95 mmHg achieved by 64.4% of those on isradipine monotherapy, 49.6% on methyldopa and 36.3% on placebo, increasing to 90.4%, 84% and 75.2%, respectively, when captopril added. Results therefore presented for three intervention groups when could have represented different captopril doses or final BPs.

MRC 1996.

Study characteristics
Methods	Multisite study. Participants 'randomly allocated ... in stratified blocks of eight within each sex and clinic'. Single‐blinded study: participants blinded; investigators and outcome assessors not blinded. Planned duration: 5 years. Lost to follow‐up: 25%.
Participants	Geographic region: United Kingdom. Study setting: participants recruited from general practice. Epoch: whole study recruited 1982‐7, study completed 1989, cognitive sub‐study recruited 1983‐5. Age criteria: 65 to 74 years. Whole study n = 4396, cognitive sub‐study n = 2651. Blood pressure entry criteria: mean of second and third seated, untreated, measurements taken, SBP 160 to 209 and DBP < 115 mmHg.
Interventions	Participants allocated in equal proportion to four treatment groups: i) hydrochlorothiazide plus amiloride; ii) matching placebo; iii) atenolol 50 mg once daily; iv) matching placebo. Two different diuretic regimens were initially used, hydrochlorothiazide 50 mg and amiloride 5 mg once daily or hydrochlorothiazide 25 mg and amiloride 2.5 mg once daily with all participants transferred to the lower dosing in 1985 on the basis of a sub‐study. Target SBP was 150 or 160 mmHg depending on whether run‐in SBP had been < or ≥ 180 mmHg respectively. If BP had not responded (definition not given) after 12 weeks or if target BP had not been achieved after 6 months, either atenolol dose was doubled, the other trial drug was added or, next, nifedipine was added in doses of up to 20 mg daily. Other supplementary drugs were also permitted. Participants were followed fortnightly for one month, then monthly for up to three months and three monthly thereafter.
Outcomes	Analysed in this review: Paired Associate Learning Test (PALT); Trail Making Test‐A (TMT‐A) Not analysed in this review: stroke; coronary events; other cardiovascular events; death from any other cause; depression rating scale
Notes	Exclusion criteria: known or suspected secondary hypertension; taking antihypertensive medications; cardiac failure or another accepted indication for antihypertensive treatment; on treatment for angina pectoris; myocardial infarction or stroke in the previous three months; impaired renal function; diabetes; asthma; serious intercurrent illness; serum potassium concentration ≤ 3.4 or > 5.0 µmol/L. Of the n = 633 (diuretic) n = 640 (BB) and n = 1311 (placebo), n = 236, n = 134 and n = 622, respectively, completed the study on the randomised allocation with no additional antihypertensive agents. Of the surviving participants, n = 265, n = 589, n = 988 and n = 550 had 2, 3, 4 and 5 TMT data points, respectively.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: 'all trial entrants were randomly allocated in equal portions to one of four treatment categories...in stratified blocks of eight within each clinic'. Method of randomisation unclear.
Allocation concealment (selection bias)	Unclear risk	Quote: 'all trial entrants were randomly allocated in equal portions to one of four treatment categories...in stratified blocks of eight within each clinic'. Method of randomisation unclear.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: 'the trial was single blind: patients did not know which treatment group they were in but the doctors and nurses did'
Blinding of outcome assessment (detection bias) All outcomes	High risk	Quote: 'the trial was single blind: patients did not know which treatment group they were in but the doctors and nurses did'
Incomplete outcome data (attrition bias) All outcomes	High risk	Both intention‐to‐treat and per protocol analyses used and reported. Numbers completing 2, 3, 4 and 5 data points all reported. Older age and worse cognitive function at baseline were associated with less follow‐up. All participants with two or more TMT values (96%) were included in analyses in an attempt to ameliorate bias. Quote: 'restriction of the analysis to those subjects who had completed the trial did not alter the result'.
Selective reporting (reporting bias)	Unclear risk	Results reported for those completing study on and off randomised Rx for all participants.
Other bias	High risk	Aim was to recruit 5000 participants to overall study, 4,396 recruited. Alternative drugs permitted. Dosage of diuretic regimen changed during study. GPs BP measurements were lower than the study measurements (by mean 10/3 mmHg). Only 37% of the diuretic group, 21% of the beta blocker group and 47% of the placebo group completed the study on the randomised treatment allocation and with no additional antihypertensive medications. More participants randomised to the beta‐blocker group required supplementary drugs than those randomised to the diuretic group (52% vs 38%). 69% of the study time was spent on placebo for those in the placebo group.

Perez Stable 2000.

Study characteristics
Methods	Single site study. Participants 'randomly assigned ... using a factorial design and a computer generated list of random numbers'. Propranolol and placebo groups matched for sex, race, annual household income, smoking status, alcohol intake, age, educations, body weight, BP and heart rate. Double blinded study. Duration: 1 year.
Participants	Geographic region: California, USA. Study setting: 'Potential participants were recruited from the community by mailings and by public service announcements and from patients seen at UCSF Medical Center'. Epoch: unclear. Age criteria: 18 to 59 years. Blood pressure entry criteria: average DBP 90 to 104 mmHg of three readings at two visits two weeks apart.
Interventions	Participants assigned to 'Lifestyle Focus Group' + placebo, 'Lifestyle Focus Group' + propranolol, placebo alone or propranolol alone. Initial dose 40 mg daily for three days then increased to 80 mg daily. 'Subsequent dosages were modified according to DBP at each of nine follow‐up visits'. Maximum study dose 400 mg propranolol daily in divided doses. Cognitive outcomes measured at baseline, three months and 12 months.
Outcomes	Analysed in this review: Stimulus Evaluation/Response Selection; Continuous Performance Task; Digit Symbol Substitution Test; California Verbal Learning Test Not analysed in this review: Center for Epidemiological Studies Depression Scale; Beck Depression Inventory; sexual function questionnaire
Notes	Exclusion criteria: insulin, bronchodilator or antidepressant treatment; antihypertensive treatment within the previous month; coronary artery disease; valvular heart disease; renal insufficiency; cerebrovascular disease
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'patients were randomly assigned ... using a factorial design and computer‐generated list of random numbers'
Allocation concealment (selection bias)	Low risk	Quote: 'propranolol and matching placebo were purchased from generic suppliers. Study medications were prepackaged...the maximum dose of study medications was five 80mg tablets of propranolol or five placebo tablets...'
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: 'propranolol and matching placebo were purchased from generic suppliers. Study medications were prepackaged...the maximum dose of study medications was five 80mg tablets of propranolol or five placebo tablets...study personnel in contact with participants were blinded to drug assignment.'
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: 'study personnel in contact with participants were blinded to drug assignment.'
Incomplete outcome data (attrition bias) All outcomes	High risk	Cognitive outcomes reported for 102/156 (propranolol group) and 101/156 (placebo group) at 12 months.
Other bias	High risk	Neither group had systolic hypertension at the outset, mean SBP 140 mmHg and 141 mmHg in the two groups. Quote: 'At 3 months of follow‐up, 72 participants were taking 80 mg of propranolol, 41 were taking 160 mg, and 21 were taking 240 to 400 mg per day. At 12 months, 63 participants were taking 80 mg of propranolol, 19 were taking 160 mg, and 35 were taking 240 to 400 mg per day.Twenty participants were not taking any study medications at the end of follow‐up, and 19 others were lost to follow‐up.' Dropouts not described for placebo group. Intention‐to‐treat analyses used.

SCOPE 2003.

Study characteristics
Methods	Multisite study. Randomisation: participants allocated by a central, computer‐generated randomisation schedule in a 1:1 ratio. Participants blinded; providers blinded; outcome assessors blinded. Eight participants lost to follow‐up. Percentage not on assigned therapy at study end: 84% control group, 75% treatment group. Duration of trial: 5 years.
Participants	Geographic region: Europe, Canada, USA and Israel. Study setting: community. N = 4964. 64% female. Age range: 70 to 89 years, average 76.4 years. Mean blood pressure at entry 166/90 mmHg. Blood pressure entry criteria: SBP 160 to 179 mmHg or DBP 90 to 99 mmHg, or both. Mini‐Mental State Examination score of 24 or above.
Interventions	Control: matching placebo. Treatment: step 1 ‐ candesartan 8 mg daily; step 2 ‐ candesartan 16 mg daily; step 3 ‐ hydrochlorothiazide 12.5 mg daily. Other drugs, except angiotensin‐converting enzyme inhibitors and AT1‐receptor blockers, could be added later. Mean duration of trial: 44.6 months. Difference in blood pressure at study end (Treatment‐Control) systolic/diastolic: ‐3.2/‐1.6 mmHg
Outcomes	Analysed in the review: blood pressure level, incidence of dementia, cognitive change measured by MMSE, incidence of side effects. Not analysed in the review: major cardiovascular events (cardiovascular deaths, non‐fatal myocardial infarction (MI), non‐fatal stroke), total mortality, fatal and non‐fatal stroke, new onset diabetes mellitus. Dropouts due to side effects (no significant difference between groups): placebo: 17%; treatment: 15%. Quality of life or functional status outcomes: both treatment regimens were well tolerated.
Notes	Exclusions: Related to hypertension: secondary hypertension, SBP > 180 mmHg, orthostatic hypotension, need for antihypertensive treatment other than hydrochlorothiazide during run‐in; stroke or myocardial infarction (MI) within 6 months; decompensated heart failure; serum AST or ALT > 3 times upper limit of normal (ULN); serum creatinine >180 µmol/L (men) or > 140 µmol/L (women); contraindication to study drug or hydrochlorothiazide; serious concomitant diseases affecting survival; alcoholism or drug abuse. Related to dementia: dementia; treatment with drugs for dementia; conditions which preclude MMSE; vitamin B12 deficiency treated < 12 months; hypothyroidism treated < 12 months; neurosyphilis or AIDS; severe brain disorder which may interfere with cognitive function; certain mental disorders; psychopharmacological treatment started within 6 months. 'Executive and Steering committees had full access to all data and were free to suggest analyses, interpret results and write ... independently of study sponsor'. Of the n = 2477 and n = 2460 in the hypertensive treatment group and placebo groups, dementia incidence was available for n = 2416 and n = 2409, respectively.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'Patients allocated by central, computer‐generated randomisation schedule, in a 1:1 ratio to candesartan or control treatment'.
Allocation concealment (selection bias)	Low risk	Quote: 'Patients allocated by central, computer‐generated randomisation schedule, in a 1:1 ratio to candesartan or control treatment'. 'The investigator sent fax with patient data for central randomisation and received treatment allocation (i.e. patient study number) by return fax'.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: 'All patients [received] the study drug ... or matching placebo similar in appearance and taste'. 'All suspected clinical events [were reported to] the coordinating centre ... dementia events were adjudicated by the Independent Clinical Event Committee.' 'All clinical events were strictly and prospectively defined.' 'Every person involved in the adjudication process was blinded with respect to the patient's treatment group allocation'
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: 'All patients [received] the study drug ... or matching placebo similar in appearance and taste'. 'All suspected clinical events [were reported to] the coordinating centre ... dementia events were adjudicated by the Independent Clinical Event Committee.' 'All clinical events were strictly and prospectively defined.' 'Every person involved in the adjudication process was blinded with respect to the patient's treatment group allocation'
Incomplete outcome data (attrition bias) All outcomes	Low risk	6 in treatment group and 2 in control group lost to follow‐up at study end; included in intention‐to‐treat analysis. 27 excluded of 4964 randomised (all 13 at 1 centre due to concerns on individual participant data; 14 ‐ no study drug dispensed).
Selective reporting (reporting bias)	Low risk	Outcomes specified in protocol reported.
Other bias	High risk	84% of participants in the placebo group ended the study on an antihypertensive agent (not ACE or ARB) to ensure the ethical treatment of hypertension. Comment: Fewer than anticipated events and lower between group differences in BP reduced power of study to detect outcome differences.

SHEP 1991.

Study characteristics
Methods	Multisite study. Double‐blind, placebo‐controlled stepped‐care treatment programme. Study duration: 5 years. Randomisation: stratified by site and by antihypertensive medication status at initial contact; participants randomly allocated by coordinating centre. Participants blinded; providers blinded; morbidity and mortality assessors blinded. Type of trial: randomised, placebo‐controlled. Percentage lost to follow ‐up: < 1%. Percentage not on assigned therapy at study end: placebo group: 44%; treatment group: 10%
Participants	Geographic region: United States of America. Study setting: community. N = 4736 (55.8% female). Age range 60 to > 80 years, mean 71.6 (SD 6.7). Race: White non‐Hispanic (79.2%), Black (13.8%), Hispanic (1.8%), Asian (4.3%), other (0.9%). Education level: mean 11.7 years (SD 3.5). Diagnosis: systolic hypertension. Mean blood pressure at entry: 170/77 mmHg. Blood pressure (BP) entry criteria: systolic BP 160 to 219 mmHg and diastolic BP < 90 mmHg.
Interventions	Control: matching placebo. Treatment: step 1 ‐ chlorthalidone 12.5 or 25 mg daily; step 2 ‐ atenolol 25 or 50 mg or reserpine 0.05 or 0.10 mg daily. Average length of follow up: 4.5 years. Difference in blood pressure at study end (Treatment ‐ Control) systolic/diastolic: ‐11.1/‐3.4 mmHg.
Outcomes	Analysed in the review: incidence of dementia; quality of life; incidence of side effects including depression; blood pressure level. Not analysed in the review: total mortality; total stroke; sudden cardiac death; rapid cardiac death; non‐fatal myocardial infarction (MI); fatal MI; left ventricular failure; other cardiovascular death; transient ischaemic attack (TIA); coronary artery bypass grafting or angioplasty; renal dysfunction. Dropouts due to side effects: control group: 7%; treatment group: 13%. 90% of hypertensive treatment group group and 44% of placebo group on active hypertensive treatment at end of 5 years. Number of participants in groups at end of 5 years: 773/2365 hypertensive treatment group, 738/2371 placebo group. Quality of life or functional outcomes: no perceptible negative effect of treatment compared to control on measures of cognitive, physical and emotional function.
Notes	Exclusions: history and/or signs of major cardiovascular diseases likely to require pharmacologic and other treatment (e.g. previous myocardial infarction, coronary artery surgery, major arrhythmias, conduction defect, recent stroke, carotid artery disease, history of transient ischaemic attack (TIA) with bruit matched with TIA localisation, two or more TIAs and signs or symptoms in a single neurological distribution); other major diseases (e.g. cancer, alcoholic liver disease, established renal dysfunction) with competing risk factors for the primary endpoint ‐ stroke; presence of medical management problems (e.g. insulin dependent diabetes, history of dementia, evidence of alcohol abuse); bradycardia; people maintained on beta‐blockers, diuretics, other antihypertensive drugs, anticoagulants, or experimental drugs on recommendation of their physicians.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'After verification of eligibility ... randomly allocated by coordinating centre to one of two treatment groups. Randomisation stratified by centre and antihypertensive medication use at initial contact.' Comment: Probably adequate
Allocation concealment (selection bias)	Low risk	Quote: 'Random assignment ... made by coordinating centre and transmitted to local centre by telephone after verification of eligibility.' Comment: probably adequately concealed.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: 'Participants randomised in a double‐blind manner to once‐daily dose of active drug treatment or matching placebo.' Data on study endpoints collected by investigators and confirmed by a panel of three physicians blind to allocation.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: 'Participants randomised in a double‐blind manner to once‐daily dose of active drug treatment or matching placebo.' Data on study endpoints collected by investigators and confirmed by a panel of three physicians blind to allocation.
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: 'All analyses were by treatment assignment at randomisation.' Comment: Subsequent publication (Di Bari 2001) addressed effect of differential dropout between treatment and control groups and concluded that the cognitive evaluations were biased toward the null effect by differential dropout.
Selective reporting (reporting bias)	Low risk	Prespecified outcomes reported.
Other bias	High risk	About 35% of participants in the placebo group took antihypertensive medication during the study for pre‐defined BP levels.

SPRINT MIND 2019.

Study characteristics
Methods	Multisite, randomised, controlled, open‐label study. Randomised using 'an internet‐based, web browser randomization procedure' to open‐label, intensive versus standard BP lowering treatment. All participants had an adjudicated dementia diagnosis. Prespecified subgroup considered cognitive outcomes. Recruitment between November 2010 and March 2013 with 24‐ and 48‐month follow‐up planned. Study halted on safety grounds in August 2015.
Participants	Geographical area: 102 centres in United States including Puerto Rico. Study participants (n = 9,361) were all aged 50 years or over with an SBP of 130 to 180 mmHg AND deemed to be at increased risk of cardiovascular events by virtue of one or more of: clinical or subclinical cardiovascular disease other than stroke; chronic kidney disease; a 10‐year risk of cardiovascular disease of 15% or greater on the basis of the Framingham risk score; an age of 75 years or older. Exclusion criteria included diagnoses of diabetes mellitus and previous stroke. It was intended that rate of cognitive decline would be measured in a subgroup of 2,800 participants, 640 of whom would also undergo MRI brain. Of the n = 4683 and n = 4678 in the standard and intensive BP reduction groups, n = 4285 and n = 4278, respectively, took part in the SPRINT MIND sub‐study.
Interventions	Treatment algorithms similar to that of ACCORD used (see above). N = 4683 randomised to standard and n = 4678 to intensive treatments.
Outcomes	Analysed in review: dementia and mild cognitive impairment. Quality of life measures. Not analysed in review: deaths due to CVD, myocardial infarction, stroke, heart failure, non‐myocardial infarction acute coronary syndrome measured over 6 years; extended cognitive assessment battery results not yet reported.
Notes	Mixture of cognitive assessment methods, including telephone assessments, used.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: 'an internet‐based, web browser randomization procedure'
Allocation concealment (selection bias)	Low risk	Quote: 'stratified according to clinical site'
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: 'participants and study personnel were aware of the study‐group assignments'
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: 'the adjudicators were masked to treatment assignment'
Incomplete outcome data (attrition bias) All outcomes	High risk	Imputation methods for missing data reported. Numbers without any follow‐up cognitive assessment reported and similar in both groups: n = 398 standard treatment group and n = 400 intensive treatment group. Those missing follow‐up data were less likely to be white and had lower cognitive and higher frailty scores at baseline. Extended follow‐up (median 5.11 years) available for 59.2% and 61.1% of participants in standard and intensive treatment groups respectively.
Selective reporting (reporting bias)	High risk	Mild cognitive impairment (MCI) reported as a secondary outcome but the study protocol states 'while not a primary or secondary outcome MCI is important because of its relevance to dementia'.
Other bias	Low risk	As with ACCORD, visit schedules differed by treatment group assignment. Between group differences in BP reduced during open‐label follow‐up but only results from active study period included in this review.

Syst Eur 1998.

Study characteristics
Methods	Multisite study. Randomisation: stratified by centre, sex and previous cardiovascular complications. Participants randomly allocated to placebo or active treatment by a centralised system or local computer system. Participants blinded; providers blinded; outcome assessors blinded. Study duration: 5 years. Lost to follow‐up: 2% at 2 years. Percentage not on assigned therapy at study end (2 years) including open follow‐up and lost to follow‐up: placebo group: 27%; treatment group: 18%
Participants	Geographic region: 23 countries across western and eastern Europe, mainly from Finland, Bulgaria, the Russian Federation, Belgium, Italy, Israel, UK, France, Estonia, Lithuania, Spain, Poland and Romania. Study setting: community based and referral clinic. N = 4695 (66.8% female). Age range: ≥ 60 years; mean 70.3 years. Race: not reported. Mean BP at entry: 174/86 mmHg. BP entry criteria: sitting systolic BP 160 to 219 mmHg and sitting diastolic BP < 95 mmHg. Standing systolic BP at least 140 mmHg. Of the n = 2398 and n = 2297 in the hypertensive treatment and placebo groups, n = 1238 and n = 1180, respectively, were recruited into the cognitive sub‐study.
Interventions	Treatment: step 1 ‐ nitrendipine 10 to 40 mg/day; step 2 ‐ enalapril 5 to 20 mg/day and/or hydrochlorothiazide 12.5 to 25 mg/day. Control: matching placebos with stepped therapy schedule similar to treatment groups. Mean daily doses: nitrendipine 28.2 mg; enalapril 13.5 mg; hydrochlorothiazide 21.2 mg. Average follow‐up: 2 years (median). Difference in blood pressure at end of study (Treatment ‐ Control) systolic/diastolic: ‐10.1/‐4.5 mmHg at 2 years.
Outcomes	Analysed in the review: blood pressure level. Not analysed in the review: total mortality; coronary heart disease (CHD) mortality; CHD morbidity and mortality (M&M); cerebrovascular mortality; cerebrovascular M&M; cardiovascular mortality; cardiovascular M&M. Dropouts due to side effects (no significant difference between groups): placebo: < 7.3%; treatment: < 7.8%. Quality of life or functional status outcomes: across 4 years of follow‐up, participants receiving active treatment were more likely to report problems on the Social Interaction scale than placebo treated participants (OR 1.32, 95% CI 1.02 to 1.69). There were no significant differences between active and placebo treatment in the other Sickness Impact Profile dimensions or in the measure of depression.
Notes	Exclusions: hypertension secondary to a disorder that needed specific medical or surgical treatment; retinal haemorrhage or papilloedema; congestive heart failure; dissecting aortic aneurysm; serum creatinine concentration at presentation of 180 µmols/L or more; history of severe nose bleeds, stroke or myocardial infarction in the year before the study; dementia; substance abuse; any disorder prohibiting a sitting or standing position; any severe concomitant or non‐cardiovascular disease.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'After stratification by centre, sex, and previous cardiovascular complications, the patients were randomly assigned double‐blind treatment with active medication or placebo by means of a computerised random function.'
Allocation concealment (selection bias)	Low risk	Quote: 'Randomisation by coordinating centre of patients satisfying all entry criteria and in whom none of the exclusion criteria are present' 'patients were randomly assigned ... by means of a computerised random function.'
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: 'Placebo tablets were identical to study drugs, with similar schedule'.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: 'Endpoint committee was unaware of patients treatment status'
Incomplete outcome data (attrition bias) All outcomes	High risk	Open follow‐up of participants who withdrew from treatment; if this was not possible, data collected annually on vital status, endpoints and antihypertensive medication use. Less than half of participants in each group followed up at 4 years, associations between baseline age and cognition and loss to follow‐up not reported. Comment: Similar numbers of non‐supervised open and lost to follow‐ups in each group; higher number of supervised open follow‐ups in placebo group.
Selective reporting (reporting bias)	Low risk	Prespecified outcomes reported
Other bias	High risk	144 participants in placebo group and 72 participants in active treatment group taking open‐label antihypertensives at final visit.

Zhang 2018.

Study characteristics
Methods	Appears to be single site study. 2x2 factorial design to telmisartan vs placebo and rosuvastatin vs placebo. All participants on hydrochlorothiazide. Randomisation: participants allocated using computer‐generated randomisation schedule in a 1:1:1:1 ratio. Participants blinded; providers blinded; unclear whether cognitive outcome assessors blinded. High rate of follow‐up. Duration of trial: 5 years.
Participants	Geographic region: Shangdong area, China. Study setting: community dwelling participants. N = 732 (47.8% female). Age range: ≥ 60 years; mean (SD) 70.7 (6.2) years. Ethnicity: all participants Han Chinese. BP entry criteria: 'SBP ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg, or self reported use of blood pressure‐lowering medications in the last 2 weeks'. Mean BP at entry: 156/71 mmHg. Mean education: 7 years.
Interventions	Participants randomised to 'telmisartan vs. placebo, telmisartan 40 mg increased to 80 mg given once daily if needed...open‐label medication, hydrochlorothiazide (12.5 mg increased to 25 mg daily if needed), was used as a baseline medication in all treatment arms...Clinical follow‐up visits were conducted weekly during the washout period, and then at trial months 1, 3, and 6, and every 6 months thereafter, until the conclusion of the study...The targeted blood pressure was defined as <140/90 mm Hg. Medication compliance was assessed by counting the number of tablets taken.' Precise up‐titration steps unclear.
Outcomes	Analysed in the review: 'Cognitive function was assessed annually using the Chinese versions of the MMSE and the Mattis Dementia Rating Scale (DRS) by experienced neuropsychology research assistants, who are experts in the measurement of cognitive function...For the progression of cognitive impairment, the changes in scores of the MMSE and DRS and possible cognitive impairment were used. Possible cognitive impairment was identified using the following: an MMSE score of ≤23 points at any annual follow‐up visit or a decline by ≤3 points between any two annual follow‐up visits, and/or a DRS score ≤123 points at any annual follow‐up visit.' Not analysed in the review: WMH.
Notes	All participants received diuretic treatment 'open‐label medication, hydrochlorothiazide (12.5 mg increased to 25 mg daily if needed), was used as a baseline medication in all treatment arms'
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: 'We used a two‐by‐two factorial design to randomly assign eligible patients on a 1:1:1:1 ratio into antihypertensive intervention (telmisartan vs. placebo, telmisartan 40 mg increased to 80 mg given once daily if needed) and lipidmodulating intervention (rosuvastatin vs. placebo, rosuvastatin 10 mg given once daily) arms separately.'
Allocation concealment (selection bias)	Low risk	Quote: 'We used computer‐generated randomization according to the order of recruitment with a block size of eight, without stratification.'
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: 'Members of our institution who were not directly working on the study executed the randomization and supplied study medications. During the double‐blind phase, all patients and investigators were masked to treatment assignment.' Comment made that neuroradiologists interpreting the MRI images were blinded to clinical and cognitive patient data.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: 'Treatment allocations were not unmasked until the study was completed and after final clinical database lock down'. No mention made of blinding of the neuropsychology research assistants assessing cognition.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: 'Among the 732 patients...725 (99.0%), 714 (97.5%), 703 (96.0%), 686 (93.7%), and 676 (92.4%) completed the first, second, third, fourth, and fifth annual cognitive functional assessments, respectively, over a mean follow‐up time of 59.8 (range 12–65) months.The most common reasons for non‐complete of the trial were death (31, 55.4%), withdrawal (16, 28.6%), and loss to follow‐up (5, 5.4%).'
Selective reporting (reporting bias)	Low risk	Quote: 'Analyses followed the intention‐to‐treat principle.'
Other bias	High risk	HIgh risk contamination bias given that all participants in placebo group received diuretic therapy. Study powered to detect differences in WMH on MRI, not differences in cognitive decline between the groups.

CVD: cerebrovascular disease; MI: myocardial infarction; AST: aspartate aminotransferase; ALT: alanine aminotransferase; WMH: white matter hyperintensity.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
ADVANCE 2010	No BP inclusion criteria
ALLHAT 2003	No cognitive outcomes reported
ASCOT 2005	No cognitive outcomes reported
Black 2001	No cognitive outcomes reported
Croog 1986	Six months study duration only
DANTE 2015	Discontinuation study
Efimova 2014	Six months study duration only
Efimova 2015	Not randomised
ESH CHL SHOT	Participants had prior stroke or transient ischaemic attack
EWPHE 1985	No cognitive outcomes reported
FINGER 2015	Non‐pharmacological study
Fogari 2003	Six months study duration only
Fogari 2006	Six months study duration only
Goldstein 1990	Cognitive outcomes reported for six month maintenance periods
HEART 2018	Ongoing study. Eight month study duration only.
HOPE 3 2016	No BP inclusion criteria
HOT 1998	No cognitive outcomes reported
Hsieh 2003	No cognitive outcomes reported
INFINITY 2019	Established cerebrovascular disease (WMH on MRI) an inclusion criterion
Lasser 1989	Seven month study duration only
LOW BP VIENNA 2018	No cognitive outcomes reported
ONTARGET 2008	No BP inclusion criteria
PreDIVA 2016	Non‐pharmacological study
PRESERVE 2018	Established cerebrovascular disease an inclusion criterion
PRoFESS 2008	Participants had prior stroke
PROGRESS 2003	Participants had prior stroke or transient ischaemic attack
rrAD 2019	multicomponent interventions (ongoing)
SPS3 2014	Participants had prior stroke
Starr 1996	Six months study duration only
STOP 1991	No cognitive outcomes reported
Syst China 1998	No cognitive outcomes reported
Tedesco 1999	Majority of participants likely to have had dementia at outset
TRANSCEND 2008	No BP inclusion criteria
VALUE 2004	No cognitive outcomes reported
Wei 2013	No reported dementia ascertainment methods
Zheleznyh 2016	Observational study

Characteristics of studies awaiting classification [ordered by study ID]

CHIEF 2012.

Methods	Multicentre study. Randomised open‐label controlled trial with blinded end‐point assessment. Method of randomisation: 'Allocation to study treatment was carried out by internet using a central, computer‐based randomization service in 2 × 2 factorial design.' Study duration: median follow‐up 41 months
Participants	Geographic region: 180 centres in China. Age range: 50 to 79 years, mean age 61.5 years. 51% female. BP entry criteria: SBP > 140 mmHg or DBP > 90mmHg, mean BP at screening 149/89 mmHg and at randomisation 157/93 mmHg. In addition to hypertension, all participants were required to have had at least one additional cardiovascular risk factor. At baseline 10.8% and 10.6% of participants in the two groups had a history of stroke, and 10.4% and 11.1% had a history of transient ischaemic attack (TIA).
Interventions	Participants were randomised to either amlodipine 2.5 to 5 mg plus amiloride/hydrochlorothiazide 1.25 to 2.5 mg/12.5 to 25 mg or amlodipine 2.5 to 5 mg plus telmisartan 40 to 80 mg. 'Patients with serum TC levels between 4.0–6.1 mmol l−1were further randomized to statin therapy or a standard management group. All eligible subjects were randomly assigned to an intensive lifestyle intervention group or a standard intervention group, according to the community area where the patients were resident.'
Outcomes	The study was powered for a composite primary endpoint of non‐fatal stroke, non‐fatal myocardial infarction and cardiovascular death. MMSE and quality of life are listed among secondary outcomes in the 2011 protocol paper by Wang et al which states 'Mini‐mental state examination (MMSE) assessment is also conducted in the first visit, 2‐year visit and last visit. The life quality and lifestyle questionnaire, and hypertension knowledge questionnaire are completed every year.' In 2018 a conference abstract describes 'no significant between‐group difference for MMSE scores and HRQoL. However, a significant correlation was observed between the magnitude of BP decrement AND the increase of MMSE scores and the improvement of HRQoL. Compared with the baseline, the difference of mean MMSE scores, the EQ‐5D index and the EQ VAS was 0.36, 0.03, 1.62 respectively (p < 0.001; p < 0.001; p = 0.030) between patients with systolic blood pressure of 130‐140 mmHg and >=140mmHg.' We attempted to contact the authors but did not receive a reply.
Notes

Characteristics of ongoing studies [ordered by study ID]

CEREBRAL.

Study name	Comparison between angiotensin‐converting enzyme inhibitors and angiotensin receptor blockers on the risk of stroke recurrence and longitudinal progression of white matter lesions and silent brain infarcts on MRI (CEREBRAL study)
Methods	Multicentre study. Randomised open‐label controlled trial with blinded end‐point assessment. Method of randomisation: 'list of random numbers ... [with] minimization method to adjust for baseline characteristics'. Study duration: 24 months
Participants	Geographic region: Japan. Study setting: Healthcare setting not specified but visits mentioned therefore likely outpatients. Age range: 65 years and over. BP entry criteria: SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg during clinic blood pressure measurements, or participants on antihypertensive treatment regardless of blood pressure control Other inclusion criteria: previous cerebrovascular events and/or asymptomatic cerebrovascular lesions as evaluated by MRI examined within six‐months prior to enrolment
Interventions	Any existing antihypertensive treatment stopped then participants randomised to ACE inhibitor or ARB therapy. Target BP 140/90 mmHg with dose increase in the first instance and then additional agents permitted but not simultaneous ACE inhibitor and ARB therapy.
Outcomes	Relevant to review: MMSE; 'cognitive impairment defined by investigator‐reported diagnosis of dementia or significant cognitive dysfunction, or a score of 23 on the MMSE during follow‐up in patients without dementia or cognitive impairment at baseline. Cognitive decline was defined as a decrease of 3 points on the MMSE from baseline during follow‐up' Not relevant to review: cerebrovascular events and progression of cerebrovascular lesions on MRI; acute coronary syndromes; hospitalisation for congestive cardiac failure; death
Starting date	'Patient enrollment began in October 2004 and concluded in January 2011'. Follow‐up was due to complete in 2013.
Contact information	See reference
Notes	Exclusion criteria include: atrial fibrillation; cancer; intracerebral haemorrhage; renal dysfunction; congestive heart failure

ESPRIT.

Study name	Effects of Intensive Systolic Blood Pressure Lowering Treatment in Reducing RIsk of Vascular evenTs (ESPRIT) Study (ESPRIT)
Methods	Multicentre study. Randomised open‐label controlled trial with blinded end‐point assessment. Study duration: 4 years.
Participants	Geographic region: China. Age range: 50 years of age or older. BP entry criteria: clinic SBP 130 to 180 mmHg. Participants also required to have 'vascular disease or [be] at high vascular risk'.
Interventions	Intensive control of SBP aiming for SBP < 120mmHg. Control group described as standard control of SBP aiming for SBP < 140mmHg using the same medications.
Outcomes	Primary outcome: 'Composite outcome of myocardial infarction, coronary or non‐coronary revascularization, chronic or acute decompensated heart failure hospitalization or emergency department visit, stroke, or death from cardiovascular disease'. Multiple secondary outcomes including: all‐cause dementia or mild cognitive impairment.
Starting date
Contact information	Information taken from: https://clinicaltrials.gov/ct2/show/NCT04030234
Notes	Information taken from: https://clinicaltrials.gov/ct2/show/NCT04030234

OPTIMAL DIABETES.

Study name	Optimal Blood Pressure for the prevenTIon of Major vAscuLar Events in Patients With DIABETES Mellitus (OPTIMAL‐DIABETES)
Methods	Multicentre study. Randomized open‐label controlled trial with blinded end‐point assessment. Study duration: anticipated median follow up 3.5 years.
Participants	Geographic region: Brazil. Age range: 50 years of age or older. BP entry criteria: clinic SBP 130 to 180 mmHg. Participants also required to have type 2 diabetes mellitus and 'be considered as having a high cardiovascular risk'.
Interventions	Intensive control of SBP aiming for SBP < 120 mmHg. ACE inhibitor, ARBs, thiazide type diuretics and CCBs to be encouraged. Control group described as standard control of SBP aiming for SBP < 140 mmHg using the same medications.
Outcomes	Primary outcome: 'Time to first event of cardiovascular death, non‐fatal myocardial infarction (MI), non‐fatal stroke, hospitalization for unstable angina or hospitalization for heart failure'. Multiple secondary outcomes including: time to mild cognitive impairment and/or all‐cause dementia; total brain volume; white matter lesion volume.
Starting date
Contact information	Information taken from: https://clinicaltrials.gov/ct2/show/NCT04040634
Notes	Information taken from: https://clinicaltrials.gov/ct2/show/NCT04040634

Differences between protocol and review

This review updates and replaces a previous Cochrane Review, 'Blood pressure lowering in patients without prior cerebrovascular disease for prevention of cognitive impairment and dementia', which was first published in 2006 (McGuinness 2006), and updated in 2009 (McGuinness 2009). The previous review could have included both pharmacological and non‐pharmacological interventions, but in practice included only double‐blind, placebo‐controlled trials of antihypertensive medications. It found no convincing evidence that BP lowering prevents the development of dementia or cognitive impairment in hypertensive participants with no apparent prior cerebrovascular disease. Because randomising participants with hypertension to placebo treatment would no longer be considered ethical, and because we were interested in whether there might be differential effects of different antihypertensive treatment regimes, the inclusion criteria for this review were widened. It now also includes RCTs comparing different antihypertensive drugs or different BP targets. We excluded non‐pharmacological interventions, which are often complex in nature, and which will be dealt with in a separate review. We dropped the requirement that trials had to be double‐blind because we did not think this was essential for head‐to‐head comparisons.

Contributions of authors

For this update:

ELC: drafting of versions, search for trials, obtaining hard copy of trials, selection of trials for inclusion/exclusion, extraction of data, entry of data, interpretation of data analysis

BMcG: drafting of versions, search for trials, obtaining hard copy of trials, selection of trials for inclusion/exclusion, extraction of data, interpretation of data analysis

ST: drafting of versions, adjudicator and reviewer of process

PP: adjudicator and reviewer of process

RB: adjudicator and reviewer of process

For the previous update:

BMcG: drafting of versions, search for trials, obtaining hard copy of trials, selection of trials for inclusion/exclusion, extraction of data, entry of data, interpretation of data‐analysis

ST: drafting of versions, search for trials, obtaining hard copy of trials, selection of trials for inclusion/exclusion, extraction of data, entry of data, interpretation of data analysis

PP: adjudicator and reviewer of process

RB: adjudicator and reviewer of process

Contact editors: Richard Harvey of CDCIG, Jim Wright and Francois Gueyffier of Cochrane Hypertension Group
Consumer editor: Gerard McVeigh

Contact editor: Leon Flicker

Sources of support

Internal sources

• No sources of support provided

External sources

• NIHR, UK

  This update was supported by the National Institute for Health Research (NIHR), via a Cochrane Programme Grant to the Cochrane Dementia and Cognitive Improvement group. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Systematic Reviews Programme, NIHR, National Health Service or the Department of Health

Declarations of interest

None known.

New search for studies and content updated (no change to conclusions)