APOE genotype, hypertension severity and outcomes after intracerebral haemorrhage

Abstract Intracerebral haemorrhage in the elderly is a severe manifestation of common forms of cerebral small vessel disease. Nearly 60% of intracerebral haemorrhage survivors will develop clinical manifestations of small vessel disease progression including recurrent haemorrhage, ischaemic stroke, dementia, late-life depression and gait impairment within 5 years. Blood pressure measurements following intracerebral haemorrhage are strongly associated with this risk. However, aggressive blood pressure lowering in the elderly carries substantial risks. In order to determine whether there might be an opportunity to select individuals at the highest risk for small vessel disease progression for aggressive blood pressure reduction, we investigated whether APOE gene variants ɛ2/ɛ4 modify the association between blood pressure and small vessel disease clinical progression after intracerebral haemorrhage. We conducted a single-centre longitudinal study at a tertiary care referral centre (Massachusetts General Hospital in Boston, MA, USA), analysing 716 consecutive survivors of acute intracerebral haemorrhage, enrolled from January 2006 to December 2016. We conducted research interviews at the time of enrolment and obtained APOE genotypes from peripheral venous blood samples. We followed patients longitudinally by means of validated phone-based research encounters, aimed at gathering measurements of systolic and diastolic blood pressure, as well as information on small vessel disease clinical outcomes (including recurrent haemorrhage, incident ischaemic stroke, incident dementia, incident depression and incident gait impairment). APOE ε4 and systolic blood pressure were associated with the risk of recurrent haemorrhage, ischaemic stroke and post-haemorrhage dementia, depression and gait impairment (all P < 0.05). APOE ε4 and systolic blood pressure interacted to increase the risk of recurrent haemorrhage, ischaemic stroke, dementia and gait impairment (all interaction P < 0.05). Among patients with elevated blood pressure following intracerebral haemorrhage (average systolic blood pressure 120–129 mmHg and diastolic blood pressure <80 mmHg) only those with one or more APOE ε4 copies were at increased risk for one or more small vessel disease outcomes (hazard ratio = 1.97, 95% confidence interval 1.17–3.31). Among haemorrhage survivors with hypertension (stage 1 and beyond) APOE genotype also stratified risk for all small vessel disease outcomes. In conclusion, APOE genotype modifies the already strong association of hypertension with multiple small vessel disease clinical outcomes among intracerebral haemorrhage survivors. These data raise the possibility that genetic screening could inform blood pressure treatment goals in this patient population.


Introduction
Intracerebral haemorrhage (ICH) is the most severe form of stroke, accounting for 10-15% of all acute cerebrovascular events, and for 50% of stroke-related mortality and disability (Qureshi et al., 2001;Poon et al., 2014). Most spontaneous ICH cases are the acute manifestation of age-related cerebral small vessel disease (SVD; Pantoni, 2010;Biffi and Greenberg, 2011). ICH survivors are, therefore, at high risk for all manifestations of progressive SVD: recurrent ICH, ischaemic stroke (especially small vessel infarcts), cognitive impairment, late-life depression and gait impairment (Pantoni, 2010;Benedictus et al., 2015;Biffi et al., 2016;Moulin et al., 2016). The APOE gene has been robustly associated with SVD and with ICH risk; indeed, APOE variants e2 and e4 represent by far the most potent genetic risk factors for ICH (Greenberg et al., 1995;Tzourio et al., 2008;Biffi et al., 2010b).
While blood pressure (BP) control is widely advocated as effective for reducing ICH risk, the optimal degree of BP reduction remains controversial. Published ICH management guidelines recommend achieving goals of Systolic BP (SBP) <130 mmHg and Diastolic BP (DBP) <80 mmHg for secondary prevention (Hemphill et al., 2015). However, findings from randomized trials and a large meta-analysis suggest that individuals at high risk for cardiovascular diseases, as ICH survivors often are, benefit from achieving normal BP (i.e. SBP < 120 and DBP < 80 mmHg; Ettehad et al., 2016;Group et al., 2015). Indeed, the revised ACC/ AHA guidelines recently proposed more stringent BP control goals for the general population (Whelton et al., 2018). We ourselves reported increased ICH recurrence risk among individuals with average SBP 120-129 mmHg (Biffi et al., 2015). On the other hand, pharmacological BP reduction in the elderly (the population most at risk for ICH) has been associated with increased risk of ischaemic stroke, cognitive impairment and gait impairment/falls (Eigenbrodt et al., 2000;Hyman and Taffet, 2009;Rose et al., 2010;Butt and Harvey, 2015).
Because of substantial variation in ICH risk based on APOE genotype, this genetic information may be of assistance in guiding BP management among ICH survivors. We, therefore, sought to test whether APOE genetic variation influences the association between BP and ICH recurrence risk, as well as the risk of other clinical manifestations of progressive SVD (ischaemic stroke, dementia, late-life depression and gait impairment) in a cohort of consecutive ICH survivors.

Materials and methods
Patient recruitment and baseline data collection All participants were enrolled in an ongoing single-centre longitudinal cohort study of ICH as previously described (Biffi et al., 2010a;Biffi et al., 2012;Biffi et al., 2015), and selected based on the following inclusion criteria: (i) age 18 years; (ii) admitted to Massachusetts General Hospital from January 2006 to December 2016; (iii) diagnosed with spontaneous ICH confirmed by CT scan (Fig. 1); and (iv) survived at least 90 days after index ICH (Biffi et al., 2015). APOE genotypes for variants e2 and e4 were determined from DNA samples derived from peripheral venous blood, drawn at time of enrolment (Biffi et al., 2010b;. The study protocol was approved by the Massachusetts General Hospital Institutional Review Board. Written informed consent was obtained from all study participants or their surrogates. Additional information on recruitment and data collection can be found in the Supplementary Appendix. Longitudinal follow-up ICH survivors and/or their caregivers were contacted and interviewed by dedicated study staff at 3, 6, 12 months after index ICH, and every 6 months thereafter, based on established protocols (Biffi et al., 2015). Investigators obtained information about ICH recurrence, ischaemic stroke (and subtype based on the TOAST method; Adams et al., 1993), cognitive impairment, gait impairment, mood impairment, death and medication use/dosing. Cognitive performance was evaluated using the Modified Telephone Interview for Cognitive Status (TICSm) test and the 16-item (short) version of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE-16;Brandt et al., 1988;Jorm, 1994;de Jager et al., 2003;Barber and Stott, 2004;Knopman et al., 2010;Seo et al., 2011;Pendlebury et al., 2013). Incident depression was identified using the four-item version of the Geriatric Depression Scale (GDS-4; Almeida and Almeida, 1999;Pocklington et al., 2016). Gait impairment was defined as a newly developed requirement for assistance (caregiver or device) for everyday ambulation, as determined by the patient and/or caregiver report. Additional information on follow-up methodology can be found in the Supplementary Appendix. At each followup, time-point study staff also captured information on BP measurements as previously described (Biffi et al., 2015). In brief, study staff inquired about the most recent BP measurements obtained in a medical setting by medical personnel. If participants could not provide reliable BP measurements, medical records were obtained for review. We pre-specified data capture targets of 1 BP measurement per 6-month period.

Statistical analysis
Separate statistical models were created for each outcome of interest, as well as for a composite outcome including all SVD-related clinical diagnoses (recurrent ICH, ischaemic stroke, dementia, depression and gait impairment). We determined factors associated with each outcome using log-rank tests (univariable analyses) and Cox regression (multivariable analyses). Additional details on multivariable modelling are provided in the Supplementary Appendix. We conducted interaction analyses for SBP/DBP with APOE e2/e4 if primary terms were found to be significant in multivariable analyses. We then separately performed analyses stratifying subjects by both hypertension severity (based on the 2017 ACC/ AHA guidelines) and APOE genotype. We estimated yearly risk for SVD outcomes of interest for graphical plotting purposes, by combining the Nelson-Aalen cumulative hazard function with the Cox model determined statistical risk effects using the predictSurvProb function in the pec R package. Estimated risks were graphically subdivided based on: (i) number of APOE e4 copies and (ii) hypertension severity (per 2017 ACC/AHA guidelines) during follow-up. We found that SVD outcomes showed significant correlation with each other (Supplementary Table 1), and therefore, did not meet the criteria for Bonferroni adjustment. We, therefore, addressed multiple testing burden by adopting the false discovery rate method (Keselman et al., 2002.). All significance tests were two-tailed, and the significance threshold was set at P < 0.05 (after false discovery rate adjustment). All analyses were performed with R software v 3.5.2 (The R Foundation for Statistical Computing). Additional information on the statistical methodology can be found in the Supplementary Appendix.

Literature review and attempted replication of results
To attempt external replication of our findings, we conducted a search of published literature and publicly available data, to identify suitable datasets for analysis. We searched PubMed, Embase, Ovid, Google Scholar, Dryad, Figshare, Zenodo and OSF for articles and data published prior to August 2018, using a dedicated electronic search strategy (see details in Supplementary Appendix). We selected for further manual review studies that (i) included only patients diagnosed with primary (i.e. spontaneous) ICH; and (ii) studies that had either APOE genotype or BP data available in the original dataset.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.

Study participants
A total of 750 patients met the initial criteria for inclusion ( Fig. 1). Patients who declined consent (n ¼ 14) were missing an index CT scan (n ¼ 2) or had no available APOE genotype (n ¼ 3) were excluded from all analyses. A total of 11 participants were missing BP measurements for one or more 6-month periods (Fig. 1), and were removed from all analyses; their forced re-introduction did not alter results substantially (Supplementary Table 2). Discrepancies between telephone-collected and EMR-collected follow-up data resulted in the removal of four patients from the present study (Fig. 1), and their removal did not alter results substantially (Supplementary Table 3). A total of 716 patients (Table 1) were, therefore, included in our analyses.

Follow-up information and post-ICH outcome rates
During a median follow-up time of 52.8 months [interquartile range (IQR) 29.8-69.5), we observed an average rate of loss to follow-up of 1.4% per year. We observed a total of 89 recurrent ICH events among 716 study participants, corresponding to a recurrence rate of 3.4%/year (95% CI 2.1-5.4%), and 59 ischaemic stroke events, corresponding to an incidence rate of 2.1%/year (95% CI 1.4-3.0%). Among ischaemic stroke events, 21 (36%) were categorized as small vessel stroke by TOAST criteria, corresponding to an incidence rate of 0.9%/year (95% CI 0.3-1.4%). A total of 122/716 (17%) ICH survivors developed new-onset dementia, corresponding to an incidence rate of 5.2%/year (95% CI 4.6-5.7%). We observed that 182/716 (25%) ICH survivors developed new-onset depression, corresponding to an incidence rate of 6.4%/year (95% CI 5.7-6.9%). Finally, 95/716 (13%) participants developed new-onset gait impairment, corresponding to an estimated incidence rate of 3.6%/year (95% CI 3.1-4.0%). We present detailed information on study sample size, mortality, loss to follow-up and post-ICH outcomes during the first 5 years of follow-up in Supplementary Table 4. APOE genotype, hypertension and outcome after intracerebral haemorrhage In univariable analyses (Supplementary Table 5), SBP and APOE e4 were associated with risk of recurrent ICH, small vessel ischaemic stroke, incident dementia, incident depression and incident gait impairment. These findings were confirmed in multivariable analyses (Table 2), after adjustment for relevant covariates (see Supplementary Appendix for additional details). We subsequently tested for interaction between SBP and APOE e4 in multivariable analyses, and found association with increased risk of recurrent ICH, small vessel ischaemic stroke, incident dementia and incident gait impairment (Supplementary Table 6, interaction P < 0.05 for ICH recurrence, small vessel ischaemic stroke, dementia and gait impairment). We repeated multivariable analyses for dementia, depression and gait impairment for subjects who did not experience ICH or ischaemic stroke during follow-up (n ¼ 566). In this subset, we identified consistent interaction between SBP and APOE e4 in determining the risk of dementia and gait impairment (Supplementary Table 7).
To quantify APOE-dependent effects on the relationship between BP and post-ICH outcomes, we first explored associations between APOE e4 and the composite post-ICH poor outcome endpoint (including recurrent ICH, small vessel ischaemic stroke, dementia, depression and gait impairment) within each hypertension severity category. We found that APOE e4 was associated with increased risk for poor outcome among patients with elevated BP, hypertension stages 1 and 2 (Supplementary Table 8). We then repeated all multivariable analyses after stratification for hypertension severity (according to 2017 ACC/AHA guidelines) and number of e4 copies (Table 3). These analyses demonstrated that non-hypertensive ICH survivors with elevated BP (i.e. SBP of 120-129 mmHg and DBP <80 mmHg) were at increased risk for a composite endpoint of recurrent ICH, small vessel ischaemic stroke, dementia, depression and gait impairment only if they possessed one or more APOE e4 copies: hazard ratio (HR) ¼ 1.97, 95% confidence interval (CI) 1.17-3.31, P ¼ 0.011 for comparison between participants with elevated BP with versus without APOE e4 ( Fig. 2A). Specifically, APOE e4 carriers with elevated BP showed significant differences in risks for recurrent ICH (HR ¼ 2.11, 95% CI 1.06-4.21, P ¼ 0.036), dementia (HR ¼ 1.89, 95% CI 1.05-3.41, P ¼ 0.037) and depression (HR ¼ 1.66, 95% CI 1.02-2.70, P ¼ 0.044) as detailed in Fig. 2B. ICH survivors with hypertension Stage 1 and beyond were also at increased risk for poor outcomes after ICH, with APOE genotype further increasing risk among those with one or more e4 copies ( Fig. 2A and B). We provide detailed results of association analyses for the composite post-ICH poor outcome endpoint (including recurrent ICH, small vessel ischaemic stroke, dementia, depression and gait impairment), stratified by APOE genotype and hypertension severity, in Supplementary Table 9. Finally, we estimated whether modelling the SBP/APOE e4 genotype interaction improved predictive ability for post-ICH outcomes, and found significant differences (compared with models including both variables but no interaction term) for prediction of future risk of ICH recurrence, ischaemic stroke, dementia, depression and gait impairment (Supplementary Table 10).

Systematic review and attempted replication of results
After reviewing published literature and publicly available data, we identified 50 original reports of ICH survivors (see Supplementary Appendix). Of these, 16 were conducted at Massachusetts General Hospital and enrolled participants also included in the present study. Of the remaining 34 studies, 16 included APOE genotype for enrolled ICH survivors and only 2 had available BP data during follow-up. Of the latter, one included APOE genotype but only evaluated ICH recurrence as the outcome of interest. We, therefore, concluded that independent replication of our findings was not feasible at the time of our analyses.

Discussion
We demonstrate that APOE e4 interacts with BP following primary ICH to increase the risk for recurrent ICH, small vessel ischaemic stroke, incident dementia and incident gait impairment. This effect extended to non-hypertensive ICH survivors with SBP of 120-129 mmHg and DBP <80 mmHg) who were at increased risk for a composite endpoint of recurrent ICH, small vessel ischaemic stroke, dementia, depression and gait impairment only if they possessed APOE e4. Thus, APOE genotype identified high-risk individuals who would otherwise be deemed to be relatively low risk. These results, which represent a unique example of interaction between a common genetic variant and a modifiable risk factor (BP) to influence multiple outcomes for a prevalent, highly relevant neurological condition. The finding that APOE e4 modifies the association of BP with multiple common clinical manifestations of SVD among ICH survivors has clinical implications. First and foremost, counselling of ICH patients and their caregivers may benefit from the inclusion of APOE genotype. Carriers of e4 might be selected for closer BP monitoring and/or more aggressive management. From a broader perspective, published guidelines for BP reduction following ICH may merit reconsideration and inclusion of genetic information. We also demonstrated that modelling the interaction between APOE e4 and BP improved the predictive capability for most outcomes of interest. This is relevant for future research studies in the field of ICH and SVD; more accurate modelling of outcome risk would allow investigators to design randomized controlled trials focused on highest-risk individuals, thus maximizing success rate for identification of truly beneficial interventions (Stanley, 2007). Furthermore, clarification of the biological basis for the described interaction between APOE genotype and BP control may well highlight novel aspects of SVD pathophysiology, thus offering additional targets for potential intervention. Finally, the opportunity to act on a modifiable risk factor based on easily ascertained genetic data would represent an ideal 'sandbox' to explore psychological, economic and societal implications of potential upcoming advances in precision medicine (Gray et al., 2014;Collins and Varmus, 2015;Lander, 2015).
APOE genotype plays a crucial role in determining the risk and severity of cerebral amyloid angiopathy (CAA), a common form of cerebral SVD characterized by accumulation of amyloid-b (Ab) in the CNS leptomeningeal medium and small arteries (Rannikmae et al., 2013). Common clinical manifestations of CAA include ICH, lacunar ischaemic stroke, cognitive and gait impairment (Biffi and Greenberg, 2011). As previously described for parenchymal Ab accumulation, hypertension severity likely leads to worsening damage to CAA-prone arterial vessels among APOE e4 carriers, increasing the risk for a variety of associated clinical outcomes (de Leeuw et al., 2004; Ref. Ref.
Ref. de Frias et al., 2014;Kester et al., 2014). Due to the frequent co-existence of vascular Ab pathology (in the form of CAA) and parenchymal Ab pathology (in the form of Alzheimer's disease), a proportion of the interaction effects we identified likely reflects known relationships between hypertension severity and Alzheimer's disease progression (Smith and Greenberg, 2009). Taken together, these considerations reflect the established role of (iii) incident dementia; (iv) incident depression; and (v) incident gait impairment. Risk distributions are separated based on APOE genotype (e4: 0 copies versus 1-2 copies) and hypertension severity during follow-up. P-values are calculated for each group in reference to normotensive subjects (regardless of APOE genotype) using the Log-rank test. (B) Estimates of yearly risk for individual post-ICH outcomes of interest, stratified by APOE genotype (e4: 0 copies versus 1-2 copies) and hypertension severity during follow-up. Vertical error bars indicate one standard deviation in risk estimate. Single asterisk indicates P-value < 0.05 for comparison with normotensive subjects (regardless of APOE genotype) using the Log-rank test. Double asterisks indicate P-value < 0.01 for comparison with normotensive subjects (regardless of APOE genotype) using the Log-rank test. ICH ¼ intracerebral haemorrhage; SV ¼ small vessel.

Elevated
APOE in risk for Alzheimer's disease/CAA by dint of their role in Ab aggregation, deposition and clearance (Kanekiyo et al., 2014). However, APOE gene products also play a critical role in non-amyloid biological pathways, including inflammatory response (Tai et al., 2015), CNS lipid homeostasis (Mahley, 2016), neurogenesis and synaptic plasticity (Liu et al., 2013;Kim et al., 2014) and mitochondrial resistance to oxidative stress (Jofre-Monseny et al., 2008). As a result, APOE e4 acts directly or in concert with age, head injury, oxidative stress, ischaemia and inflammation to alter disease onset, progression and prognosis in a variety of neurological disorders (Maiti et al., 2015). Finally, APOE e4 has also been directly linked with cerebrovascular dysfunction via a variety of mechanisms, including pericyte migration/activation, astrocyte activation, smooth muscle cell damage, basement membrane degradation and alterations in brain endothelial cells (Zlokovic, 2013;Tai et al., 2016). We, therefore, hypothesize our findings to reflect more broadly the biological interaction between the damaging effects of hypertension and the pathological substrate represented by APOE e4 across a multitude of mechanistic pathways. The robustness of our results is supported by the strengths of our design, which include consistent capture and incorporation of relevant neuroimaging, genetic and BP information and follow-up based on standardized procedures, capturing multiple relevant outcome endpoints of immediate clinical relevance. The study's limitations derive most substantially from its small sample size. Furthermore, subjects were recruited and followed in an observational manner at a single academic tertiary care centre. These findings will, therefore, require replication in different centres and healthcare delivery settings. In our systematic review, we were unable to identify currently available resources that would allow for a ready replication of our findings. In addition, we were limited in our characterization of cognitive, mood and gait disorders. Specifically, our ability to better delineate cognitive impairment by sub-domains, capture severity of mood symptoms or describe patterns of gait impairment is minimal, due to the nature of the screening tools employed. Finally, BP data capture was non-standardized and likely resulted in imprecision in capturing hypertension severity. However, this is unlikely to have systematically affected individuals based on APOE genotype, and more likely eroded statistical power instead of generating false-positive associations.
In summary, we demonstrate that APOE genotype interacts with average SBP to influence long-term clinical outcomes following ICH. Among non-hypertensive ICH survivors with elevated BP, APOE genotype identified those at higher risk for poor outcome. Although these findings require replication, their incorporation in clinical practice and future clinical trials may guide precision approaches for BP control in this very high-risk population.

Supplementary material
Supplementary material is available at Brain Communications online.