Midlife Hypertensive Status and Cognitive Function 20 Years Later: The Southall and Brent Revisited Study

Methods

Results

Of 4,857 participants at baseline, 1,101 died, leaving 3,756. Of these, 306 did not respond, 241 had no UK address found, and 82 had emigrated. Of 3,433 traceable participants, 2,129 (62.0%) were directly followed up with 20 years later, and 1,484 of these had sufficient cognitive data to be included in the analysis (1,429 examined in clinic, 55 at home; 692 European, 551 South Asian, and 241 African Caribbean). Altogether, 172 (11.6%) participants with cognitive data were defined as having cognitive impairment. Compared to the baseline sample, those examined for cognitive function at follow-up were younger than the baseline sample (49.9 ± 6.2 and 52.4 ± 6.9, respectively) and had higher education (<10 years of education, 16.1% and 26.5%, respectively), but there was no substantial difference in sex distribution (female, 23.9% and 24.7%, respectively) or mean BP at baseline (121/78 and 125/79 mmHg, respectively). The range of intervals between baseline and follow-up examinations was 17 to 23 years (median 20 years).

As shown in Table 1, in unadjusted analyses, cognitive impairment was associated with the baseline characteristics of older age, female sex, lower education, diabetes mellitus, microalbuminuria, and higher BMI. Cognitive impairment was also associated with hypertension treatment at baseline, and significant heterogeneity was seen across quintiles of SBP and DBP, with a significant linear trend for SBP, PP, and MAP but not DBP. In terms of ambulatory BP in the subsample of 117 participants in whom these measurements were taken, there was a significant linear association between evening and nighttime ambulatory SBP and DBP and cognitive impairment. Figure 1 illustrates the relationship between BP quintiles at baseline and cognitive impairment at follow-up. Risk of cognitive impairment increased monotonically across the SBP and PP distributions, but the associations with DBP and MAP were U-shaped.

From analyses summarized in Table 2, the linear associations between SBP, PP, and cognitive impairment were attenuated after adjustment for demographic factors, although subsequent adjustment for other covariates had little further effect. An interaction term with ethnicity fell below statistical significance for SBP (likelihood ratio test chi-square (χ²) = 5.74, P = .06) and for PP (χ² = 3.39, P = .18). Logistic regression analyses of DBP and MAP showed significant nonlinear relationships with cognitive impairment, with linear and quadratic terms remaining significant in the fully adjusted models and little changed after further adjustment for AER. No interactions with ethnicity were found for either variable (DBP, χ² = .65, P = .72; MAP, χ² = 1.84, P = .40). With post hoc stratification according to baseline age, the associations between DBP and MAP and cognitive impairment were strongest in the older half of the sample (50–67; for DBP, fully adjusted odds ratio (OR) = 0.06, 95% confidence interval (CI) = 0.01–0.41 linear, OR = 1.20, 95% CI = 1.07–1.34 quadratic; for MAP: OR = 0.08, 95% CI = 0.01–0.57 linear, OR = 1.14, 95% CI = 1.04–1.26 quadratic) and were not significant in the younger half (40–49; for DBP: OR = 0.33, 95% CI = 0.05–2.25 linear, OR = 1.08, 95% CI = 0.96–1.21 quadratic; for MAP: OR = 0.22, 95% CI = 0.03–1.89 linear, OR = 1.09, 95% CI = 0.98–1.21 quadratic). The lower DBP component of the nonlinear association was stronger in those who were not receiving antihypertensive treatment at baseline (linear term OR = 0.21, 95% CI = 0.05–0.80; quadratic term OR = 1.11, 95% CI = 1.02–1.20) than in those receiving antihypertensive treatment (linear term OR = 0.01, 95% CI = 0.00–5.66; quadratic term OR = 1.43, 95% CI = 0.96–2.14).

Table 3 summarizes analyses of the association between hypertensive treatment at baseline and cognitive impairment at follow-up. In summary, although attenuated by adjusting for age and sex, this association persisted after adjusting for covariates and after further adjustment for AER. No significant interaction with ethnicity was found (χ² = .11, P = .94), and post hoc adjustment for DBP and MAP made little difference (data not shown).

Sensitivity analyses using inverse probability weighting (to take attrition between baseline and follow-up interviews into account) revealed similar findings. In fully adjusted models (Model 2 in Tables 2 and 3), ORs for cognitive impairment were as follows: SBP, OR = 1.08 (95% CI = 0.95–1.23), DBP linear, OR = 0.12 (95% CI = 0.03–0.54), DBP quadratic, OR = 1.15 (95% CI = 1.05–1.26), PP, OR = 0.93 (95% CI = 0.79–1.10), MAP linear, OR = 0.10 (95% CI = 0.02–0.57), MAP quadratic, OR = 1.13 (95% CI = 1.04–1.24), antihypertensive treatment, OR = 2.00 (95% CI = 1.16–3.42).

Table 4 displays logistic regression analyses from the subsample that underwent ambulatory BP monitoring. No associations were found with later cognitive function for average morning or nighttime BP, but there were linear associations between evening SBP and DBP and likelihood of cognitive impairment that were largely unaltered after individual adjustments for covariates, although evening SBP was no longer significant in the fully adjusted model. Further adjustments for resting DBP or MAP or for antihypertensive treatment had little effect on the association between evening DBP and cognitive impairment (data not shown).

Table S1 displays logistic regression analyses of cognitive impairment according to individual tests as dependent variables (with impairment on each defined on the basis of a score within the lowest 10% of the distribution within each ethnic group). In summary, most of the associations were strongest for verbal fluency as an outcome, particularly the U-shaped relationships with DBP and MAP. Antihypertensive treatment was most strongly associated with impaired immediate verbal and visual recall.

Discussion

Using data from a long-running multiethnic cohort study, associations between current cognitive impairment and markers of hypertensive disease 20 years earlier were investigated. In summary, an independent U-shaped association was found with resting DBP and MAP levels (cognitive impairment associated with high and low levels) and independent associations with midlife use of antihypertensive agents and with higher evening DBP in those who underwent ambulatory BP monitoring. The U-shaped associations with resting DBP and MAP were more prominent in older than younger participants. Considering potential underlying mechanisms, resting PP (as a marker of arterial stiffness) showed little evidence of association, and AER (as a measure of small vessel disease) did not have any substantial influence on the associations of interest. The associations did not vary substantially between ethnic groups.

Strengths of the study were the 20-year interval between exposure and outcome and comprehensive assessment of BP at baseline, allowing examination of a variety of different features of hypertension as exposures. Measures of ambulatory BP were only made in European and African-Caribbean participants and were therefore not representative of the sample at baseline, although they are unlikely to be biased with respect to their associations with cognitive function 20 years later. Cognitive assessment was comprehensive and used procedures that had been specifically designed for their cross-cultural applicability, in particular, the CSID, which has received substantial international evaluation.¹⁷ Weaknesses of this study include limited statistical power in some respects, particularly for stratified analyses. The sample examined should also be considered to be relatively healthy survivors constituting a small proportion of the original participants, although those with and without cognitive impairment were drawn from the same source (survivors from the original cohort), and a sensitivity analysis using inverse probability weighting to account for differential attrition did not meaningfully alter findings. Hypertension and cognitive impairment are predictors of mortality, so survival effects would more likely have obscured than exaggerated associations. A large number of covariates were considered, although residual confounding cannot be excluded absolutely. There was also no baseline information on cognitive function and therefore no means of measuring cognitive decline. Finally, no attempt was made to test mediating pathways, so measures of health status at follow-up were not considered as covariates.

The most consistent findings relating BP levels to cognitive impairment in later life have arisen from samples followed for at least 10 years. The findings of the current study that hypertension was associated with cognitive impairment 20 years later are in keeping with this literature, although other findings of worse late-life cognitive function associated with progressively higher midlife DBP ³ differ from those of the current study, in which a U-shaped relationship was found. Other studies of cognitive function with BP ascertainment at least 10 years earlier have reported associations with high midlife SBP (≥160 vs <110 mmHg) in men,²⁶ with progressively higher midlife SBP in people who had not received antihypertensive treatment,² with persistently high SBP (≥140 mmHg) over a 38-year period, and with high SBP (≥160 vs <140 mmHg).⁴ In the Framingham Offspring cohort study, midlife hypertension was specifically associated with worse late-life executive function in addition to progression of white matter hyperintensities on magnetic resonance imaging.²⁷ Findings from studies with dementia as an outcome include associations between high midlife SBP and DBP (≥160/95 mmHg) in men and dementia and AD, although only in those not previously taking antihypertensive agents,²⁸ and an association between midlife high SBP (≥160 vs <140 mmHg) but not high DBP and AD.²⁹ Therefore, although hypertension is implicated in late-life cognitive impairment, there is still a substantial level of controversy around what components of BP are most important and in what subgroups.

U-shaped associations between BP and cognitive impairment are not uncommonly found, and there is ample evidence of an exaggerated decline in BP (particularly SBP) preceding the clinical onset of dementia.¹ It is less usual to find a U-shaped relationship over a 20-year interval, as described here for DBP and MAP. The greater risk in those with lower pressures is unlikely to represent a secondary effect of neurodegeneration (often assumed to underlie more-contemporaneous associations between lower BP and cognitive impairment). Low BP might reflect atherosclerotic disease, or it might be a longstanding feature by which the individual is rendered vulnerable to episodes of hypoperfusion. It might also be a marker of a more-general frailty syndrome that predisposes to later cognitive impairment. The apparently stronger associations in older than younger participants at baseline supports this, although it requires replication. The association with lower DBP appeared stronger in people not taking antihypertensive agents and was therefore not likely to be iatrogenic in nature. Finally, the relationships were stronger for verbal fluency than memory. Although Type 1 statistical error is possible because this association was one of several exploratory analyses, this stronger association with verbal fluency may be consistent with subcortical damage secondary to cerebrovascular disease causing impairment in executive function rather than memory.

Fewer data have been published on midlife PP than on SDB and DBP as an exposure for late-life cognitive impairment. The Women’s Health Study found a five times greater likelihood of impairment on a measure of executive function in women with PP of 84 mmHg or greater than in women aged 76 to 80 with PP less than 68 mmHg and twice the incidence of impairment on immediate recall in participants aged 70 to 75 with PP of 71 mmHg than in those with PP less than 68 mmHg.¹⁰ The Kungsholmen study found a U-shaped relationship between PP and dementia in women aged 75 and older.⁸ These studies measured a potentially much later stage of vascular pathology in older age groups, but considerably less is known about the role of midlife PP in relation to later cognitive impairment. SBP is reported to increase with age, whereas DBP increases up to aged 50 to 59 and then begins to decline so that PP increases in later life.³⁰ This may explain why the current study found little evidence of midlife PP as an independent predictor of later cognitive impairment. Similarly, the Honolulu Asia Aging study found no association between PP in midlife and incidence of dementia in later life.³¹ One possible explanation is that PP is not important in the etiology of cognitive impairment, a second is that this only becomes important closer to the development of the outcome, and a third is that people with higher PP in midlife who survive to late life are healthy in other respects, masking associations with later cognition. Midlife AER did not appear to play a substantial role in the association between BP and cognition, even though baseline microalbuminuria was associated with cognitive impairment.

The findings on ambulatory BP are novel in terms of the length of follow-up between exposure and outcome. Evening BP was more strongly associated with cognitive impairment than morning and nocturnal measurements. Previous research³² has found high nocturnal SBP in older adults (mean age 71),³³ lack of a nocturnal dip,³⁴ and exaggerated variability in very elderly adults ³⁵ to be associated with cognitive impairment and dementia. The cohort in the current study was younger at the time of ambulatory measurement, and although some evidence of an association between nocturnal BP and later-life cognitive impairment was found, this disappeared after adjustment for age and sex. Mean resting BP might underestimate an additional influence of BP variability, although whether there are particular high-sensitivity periods during the day or night remains to be established. An important consideration is that the timing of bed rest influenced evening BP, and other factors might have confounded this association, which would require further investigation.

If midlife BP is associated with later-life cognitive impairment and subsequent dementia, management of BP in midlife could have important public health implications in terms of prevention. Several trials have been conducted of BP lowering and cognitive outcomes,^36–41 although only the Systolic Hypertension in Europe Study found a significant primary effect on risk of dementia, with a reduction in incidence from 7.7 to 3.8 per 1,000 person-years,³⁷ findings that persisted in open-label follow-up.⁴² In the current study sample, people receiving antihypertensive treatment at baseline were more likely to have cognitive impairment 20 years later, although antihypertensive use at baseline reflects chronic exposure to hypertension as well as its treatment. Overall, these findings may have important clinical implications in terms of the importance of detection and control of high BP in reducing the risk of cognitive impairment later in life.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article: