How far to lower blood pressure in the long term, after a stroke?

There have been numerous studies in the past few years questioning the wisdom of setting lower thresholds and targets for blood pressure control in high-risk patients, such as those with coronary disease, stroke or diabetes. The majority of these studies have been observational, some planned as such ab initio, but most have been post-hoc observational studies, performing non-randomized analyses of data from large clinical trials [1–7]. These studies have exerted considerable influence, and many guidelines have reviewed their target blood pressures for high-risk patients from 130/80 mmHg, upwards to 140/90 mmHg. In doing so, they have removed the differential between patients with uncomplicated hypertension and high-risk patients with diabetes or with established vascular disease [8–10]. Recommendations for blood pressure thresholds and targets are clearly of paramount importance, given that around 40% of the global adult population has hypertension, as defined by a blood pressure over 140/90 mmHg, and that the burden of blood pressure-related disease is also considerable in high-risk patients with a blood pressure below 140/90 mmHg [11,12]. It is also clear that there must be a J-curve relating blood pressure to cardiovascular risk at some point, below which autoregulation and tissue perfusion of vital organs will fail. But there is much uncertainty whether there is a J-curve within the usual range of blood pressure to which patients with hypertension or those with vascular disease might be exposed by treatment with blood pressure-lowering drugs. It is also uncertain whether any such J-curve reflects the patients’ essential risk profile and pre-existing disease, or the effects of blood pressure-lowering medications. Finally, it is not clear whether there are differences between different outcomes, such as myocardial infarction and stroke, or between patient groups with different risk profiles, such as those with previous coronary disease, cerebrovascular disease or diabetes [8,13,14]. In this issue of the Journal of Hypertension, Kim et al.[15] report on the association between blood pressures recorded after 5 years of follow-up in 483 survivors, out of 1589 patients with previous stroke, who entered the North East Melbourne Stroke Incidence Study (NEMESIS) between 1997 and 1999, with outcomes recorded in a further 5 years of follow-up, to a total of 10 years. The two outcomes studied were all-cause mortality and a composite vascular outcome comprising non-fatal stroke, non-fatal myocardial infarction and all-cause mortality. The study was meticulously conducted with five annual follow-up visits between the 5 and 10-year mark, and with rigorous checking and verification of outcome data against a variety of local sources in Melbourne and national databases in Australia. The authors split the patients into four quarters according to the SBP at 5 years with SBP of 92–120, 121–130, 131–141, and 142–210 mmHg, respectively [15]. They report the presence of a J-curve with 61% greater risk of the composite vascular outcome at 10 years in the group with SBP between 92 and 120 mmHg compared to the reference group, with SBP between 131 and 144 mmHg, but no significant difference between those in this reference group and those with SBP between 121 and 130 mmHg or 142 and 210 mmHg. There were similar findings for all-cause death alone. Age, male sex and living in an institution were associated with poorer outcomes at 10 years, whereas, rather surprisingly, a high cholesterol at 5 years was associated with a better outcome at 10 years, even after adjustment for statin therapy. Exclusion of 41 patients with heart failure at baseline weakened the association between SBP and outcomes. The major strengths of this study are its novelty, in being a community-based study with very long follow-up (10 years), and the rigour with which it was conducted and analysed. It also has clear limitations. The first is the small sample size, with only 483 patients analysed in the final model, exposing the study to considerable random error, particularly in relation to multiple subgroup analyses. The second is shared with a great number of other studies reporting J-curves and that is its observational nature, which clearly limits its interpretation to that of generating a hypothesis. As the authors acknowledge, randomized evidence from the Perindopril Protection against Recurrent Stroke Study (PROGRESS) strongly supports lowering blood pressure after a stroke, regardless of baseline blood pressure, without any suggestion of a J-curve [16–18]. More recently, the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial in patients with type 2 diabetes demonstrated a clear benefit with a strong reduction in the risk of stroke in the group targeting (and achieving) an SBP of less than 120 mmHg, compared to that targeting an SBP of less than 140 mmHg [19]. This was accompanied by a tendency of reduction in the primary composite outcome and in myocardial infarction, both of which were reduced by more than 10%, though this was not significant in either case [19]. There is also evidence from a very large meta-analysis of randomized trials that active reduction of SBP (and DBP) leads to substantial reduction in the incidence of stroke, with no heterogeneity between the reductions seen in patients with previous stroke, previous myocardial infarction or without any history of previous vascular disease [20]. The evidence, to the contrary, supporting a J-curve for cardiovascular events in high-risk patients, comes mainly from post-hoc, non-randomized, observational analyses of data from large clinical trials, but even those have suggested that the J-curves occurred mainly in relation to coronary disease and not to stroke [1,3,9]. The main problem with observational studies of this sort lies in the difficulty of excluding indication bias and reverse causality, which could well be responsible for the greater risk observed in those with the lowest SBP in the present study [15]. This is particularly cogent when it is considered that the blood pressure in the group with SBP below 120, ranged as low as 92 mmHg, is surely not a pressure that any responsible clinician or authoritative guidelines would recommend or target. Indeed, few would contemplate lowering blood pressure in patients with an SBP below 115 or 110 mmHg. Another example of reverse causality was seen with reports some years ago from observational community studies suggesting that lower blood pressures in elderly patients were associated with greater risks of stroke or of mortality, findings that have been clearly refuted by the Hypertension in the Very Elderly (HYVET) trial which demonstrated considerable reduction in stroke and in heart failure in those receiving active blood pressure-lowering treatment [21], as acknowledged by Kim et al.[15]. One of the analyses that will cause some concern is the one reporting that there was no significant difference in stroke risk between the reference group with SBP between 131 and 141 mmHg and the group with the highest SBP, between 142 and 210 mmHg. This may well be interpreted by some readers to mean that there is little point in lowering blood pressures in the majority of patients who have had a stroke, even if quite clearly hypertensive. This comes at a time when it is clear from studies such as the Prospective Urban Rural Epidemiology (PURE) study that on an average, only 15% of the hypertensive population has blood pressure controlled to the recommended less than 140 mmHg, across 17 high, middle and low-income countries [12]. The interpretation that there is little to be gained from treating patients with previous stroke who are clearly hypertensive, drawn by some readers, coupled with the suggestion of harm in those with SBP less than 120 mmHg, could be sufficient to dissuade many practitioners from making the greater efforts that are clearly needed to control and reduce the growing burden of blood pressure-related disease [11]. There is a clear need for new randomized clinical trials to address these issues and help us to determine what blood pressure thresholds and targets are appropriate for a number of patient groups with increased cardiovascular risk, including those with stroke. Meanwhile, it would seem prudent to continue to lower blood pressure assiduously in the majority of patients who have suffered a previous stroke. ACKNOWLEDGEMENTS Conflicts of interest J.C. is the Co-Principal Investigator for the PROGRESS trial of blood pressure lowering after a stroke and the INTERACT2 trial of blood pressure lowering in the acute phase of intra-cerebral haemorrhage. He has received research grants from Servier International and from the National Health of Medical Research Council of Australia, administered through the University of Sydney.