How important is peri‐operative hypertension?

Pre-operative hypertension (Table 1) 1 is associated with an increased risk of peri-operative cardiac events and/or mortality (for example 2-5); however, its omission from various risk stratification algorithms, such as the Revised Cardiac Risk Index 6 amongst others, has precipitated a fall in profile of this condition. This fall in profile is at odds with its frequency. In our recent work using Hospitals Episode Statistics data from England and Wales, 64%, 55% and 40% of patients undergoing cardiac surgery 7, abdominal aortic aneurysm repair 2 and major joint arthroplasty 2, respectively, had a diagnosis of hypertension. In several studies, pre-operative hypertension predicts mortality, or mortality and major morbidity, in a wide spectrum of surgery, from cardiac 3, 8, to orthopaedic 2, to day-case surgery 4. This should be unsurprising to anaesthetists for two primary reasons: (i) the pathophysiology of hypertension reflects vascular disease and stiffening of the arterial tree with subsequent myocardial changes including left ventricular hypertrophy and diastolic dysfunction; and (ii) great attention is paid to measuring blood pressure (BP) in the peri-operative period. However, definitive information on how to interpret BP numbers in the peri-operative period is lacking. In this editorial, the difficulties in interpreting pre-operative BP and targeting intra-operative BP are covered. The intention is both to highlight the importance of pre-operative hypertension and peri-operative haemodynamic control in peri-operative risk, and to stimulate a research agenda to improve peri-operative care. Numerous studies demonstrate that hypertension is an important pre-operative risk factor, though several studies have not (described in 9, 10). This may be due to varying definitions of pre-operative hypertension used in different studies, e.g. community-based diagnosis, measurement on the day of surgery, or clinical coding. Alternatively, some studies lack statistical power to identify an independent effect of hypertension when adjusting for other associated risk factors such as renal failure or previous stroke (that typically impose a higher peri-operative risk). In this regard, we have found that in-hospital coding of pre-operative hypertension was a risk factor for peri-operative mortality, prolonged length of stay and readmission, in a study of 414 895 major joint arthroplasties (with 164 380 hypertensive patients) 1. In contrast, in a much smaller dataset of 14 524 abdominal aortic aneurysm repairs, hypertension (affecting 7970 patients) affected the length of stay and the risk of postoperative readmission but not peri-operative mortality 1. The difference in mortality between the two datasets may be due to loss of statistical power in the mortality analysis (due to the comparatively low event rate of mortality and smaller sample size) in the latter dataset. This sort of ambiguity has clouded this area of research, leaving several fundamental questions unanswered, for example those below. It is unclear whether BP on admission to hospital is the correct measure to use for judging peri-operative risk. Indeed, white coat hypertension may either reveal an underlying tendency to moderate or severe hypertension 11, or provide an inaccurate estimate of the baseline BP (presuming it is the baseline BP that is associated with increased risk) 12. Other reasonable BP estimates, such as those from the community, may be superior as they may reflect the chronic BP state more accurately 12. We are currently attempting to address whether clinic BP from general practice can predict peri-operative risk, but ambulatory BP may yet again be superior to this 12, 13. The studies addressing the numerical values of BP and peri-operative risk have been underpowered to date 5, 14. While it appears intuitive that the higher the BP, the larger the risk, this is yet to be definitively demonstrated. The most convincing peri-operative evidence found a linear trend for each 10-mmHg increase in systolic BP on admission (OR 1.20 (95% CI 1.01–1.42); n = 183) and increased odds of myocardial ischaemia 15 similar to community data 16. However, given Howell et al.'s limited sample size, a large, contemporary study is warranted. Other approaches have included comparing patients who died postoperatively with single ‘matched’ controls, but such studies have not found important differences in pre-operative BP between the two groups 14, 17, 18. While current guidance indicates that surgery should not be cancelled for severe hypertension (Table 1) 9, 19, there is a lack of information about the risks of operating on patients with BP above this level. One study randomly assigned 989 hypertensive patients who presented for surgery with a diastolic BP of > 110 mmHg, just before surgery, to intranasal nifedipine-mediated BP control followed by immediate surgery, or delayed surgery with conventional BP control over several days 20. It found no benefit to delaying surgery 20. A shift to consideration of pulse pressure (systolic minus diastolic BP) may aid the interpretation of pre-operative BP, as small increases in community-measured pulse pressure vastly increase the risk of stroke and other cardiovascular outcomes 21. Increased pulse pressure predisposes to an increased risk of cerebral events and mortality, based on a large cohort of patients undergoing cardiac surgery 22, 23. In contrast, pulse pressure was not associated with an increased incidence of major adverse cardiac events or mortality in two, potentially flawed 10, smaller cohort studies of peripheral vascular surgery 24, 25. Clearly, more data are required on the utility of pulse pressure to predict peri-operative risk. In summary, most studies conducted so far to identify associations between different numerical thresholds and peri-operative outcomes appear underpowered. Given our improved access to ‘Big Data’ 26, a contemporary analysis of the impact of pre-operative numerical BP values and postoperative outcomes is warranted. The proportionality between BP values and change in peri-operative risk needs to be addressed. The relationship may be linear (as observed in the small sample of Howell et al. 15) or alternatively there may be an inflection point in the curve (for example, when reaching severe hypertension) after which risk escalates rapidly. Of course, it is also possible that there is no relationship between numerical values and meaningful postoperative outcomes, and that any peri-operative risk imposed by hypertension can be explained by secondary end-organ (largely cardiac, renal and cerebral) effects. Such a conclusion requires rigorous analysis of large datasets. The interaction of different drug treatments, numerical BP values and peri-operative outcomes remains unknown. For example, controversy surrounds the peri-operative use of beta-blockers in hypertensive patients 5, 27 (see 28 for a recent commentary); it may be that we need to modify patient selection 29 or drug administration 30 to realise the benefit from peri-operative beta-blockade. Furthermore, with National Institute of Health and Care Excellence (NICE) guidelines recommending angiotensin-converting enzyme inhibitors or angiotensin receptor blockers as first-line treatment in patients under 55 years old 13, there will be more patients on these drugs presenting for surgery. These drugs have been associated with increased intra-operative hypotension 31, and studies in cardiac surgery have suggested an association with peri-operative acute kidney injury and death 32, 33, though the data are less clear from non-cardiac surgery 34, 35. More evidence about how angiotensin-converting enzyme inhibitors should be managed in the peri-operative period is required. For example, in one small study of 40 patients, omission of angiotensin-converting enzyme inhibitors in patients taking them chronically did not provide tangible benefit 36. Evidence remains lacking about how to target intra-operative BP. It is worth highlighting that this is another area of peri-operative research that is plagued by non-standardised definitions, creating inconsistencies in the message. A systematic review identified 140 definitions of intra-operative hypotension in 130 different papers on the subject 37. For example, a systolic or mean arterial pressure ± 20% of a pre-operative measurement is often recommended, with a limited evidence base 37. However, it is likely that we are poor at controlling intra-operative haemodynamics: 93% of patients in one recent dataset had a systolic BP that dropped by more than 20% from baseline 37. Evidence-based, intra-operative haemodynamic targets are urgently required. Recent evidence has shown that both systolic hypotension and hypertension are potentially problematic 3. This is reasonable, given that both hypoperfusion and hyperperfusion, as well as effects on cardiac work, can be deleterious. In a seminal study of cardiac surgery, Aronson et al. showed that a function of time spent with an intra-operative systolic BP > 135 mmHg or 145 mmHg or < 75 mmHg postoperatively), and magnitude of change in BP above or below these levels, was associated with increased peri-operative mortality 8. It is important to note that in this analysis, pre-operative hypertension (defined as 160/105 mmHg – similar to stage-2 hypertension) was retained as a risk factor, so both the variability in control of intra-operative haemodynamics and a diagnosis of pre-operative stage-2 hypertension were important. In this cohort study, the data of patients with stage-1 hypertension were not explicitly reported, so it is unclear if patients in this category had elevated risk after adjusting for intra-operative haemodynmaic instability. Other attempts to define the lower limits of mean arterial pressure have been successful to a small degree. An empirical threshold for mean arterial pressure of 55 mmHg was defined by Walsh and colleagues, using a robust statistical approach to define the risk of acute kidney injury in non-cardiac surgery 38. The authors also showed that a mean arterial pressure < 55 mmHg for more than 20 minutes was associated with increased 30-day mortality 38. However, it seems unlikely that a single threshold could adequately capture the risk in all patients; hence proportional changes from an accurate baseline BP (wherever that is measured) is likely to be a more appropriate strategy. Along these lines, a 30% drop from baseline mean arterial pressure was associated with postoperative stroke in a case-control study 39. However, how extreme BP changes must be (and for how long and in which patients) to increase peri-operative risk still remains ill-defined. Cerebral autoregulation should not be taken for granted 40, particularly in hypertensive patients where impaired autoregulation is likely to contribute to cognitive decline 41. Up to 20% of patients undergoing cardiac surgery with cardiopulmonary bypass have impaired autoregulation, indicating that cerebral blood flow is dependent on pressure 42. Of the remaining 80%, the lower limit of autoregulation is variable, ranging from 40 to 90 mmHg 42. Hence, definitions of hypotension based on a supposed lower limit of autoregulation, or on empiric definitions based on populations in which the capacity for autoregulation varies, may not be helpful for an individual patient. What is necessary is to define the lower limit of autoregulation for individual patients in real time. Various approaches to this problem have been attempted, largely driven by the technology of near infra-red spectroscopy (NIRS) 42-45. This technique can provide data on cerebral oxygen saturation, giving an end-organ approximation of the oxygen content of the blood. In this context, reduced cerebral oxygen saturation (with a potential lower limit of 75% 45), a surrogate of end-organ oxygen extraction, has been associated with postoperative morbidity after cardiac surgery 45. Other studies have suggested similar benefits, but perhaps of most interest is the use of NIRS to define the lower limit of autoregulation. Ono and colleagues recently showed that a correlation coefficient associating mean arterial pressure with low-frequency oscillations in the NIRS signal could define this lower limit. They then showed that the time under this lower limit of autoregulation was associated with increased peri-operative morbidity 44 (including acute kidney injury 43) and mortality 44 from cardiac surgery. While it may only be possible to define the lower limit of autoregulation in 80% of patients, this is a potentially important advance, especially as identifying that there is no lower limit of autoregulation in the remaining 20% may itself define care: knowing that blood flow is pressure-dependent at all measured BPs could clearly be important. Even if we can identify that the patient's BP is inappropriate, the question of how to treat that BP remains unanswered. This depends both on why the pressure is inappropriate (for example, changes in cardiac output versus total peripheral resistance) and on the available pharmacological interventions. It is worth recognising that administration of a vasopressor may not improve end-organ oxygen delivery; indeed, cerebral oxygen saturations may drop with phenylephrine administration 46. However, there are limited data on functional outcomes such as morbidity and mortality to guide the use of a specific pharmacological agent. I have identified some areas of peri-operative care with insufficient data to inform clinical practice. Patients with hypertension diagnosed in the community frequently present for a wide range of surgery and, at present, we inadequately understand their peri-operative risk. Clearly a holistic approach is required, where hypertension, pre-operative pharmacology and intra-operative haemodynamic management are all regarded as part of a package of care that encompasses other variables impacting on patient outcomes. The next step is to identify if, and how, patient care may be optimised in the peri-operative period; this may involve real-time measurements of autoregulation, end-organ perfusion or a change in peri-operative pharmacological management (including anti-hypertensive and anaesthetic drugs) evaluated in randomised controlled trials. Significant progress has been made in understanding the impact of BP on peri-operative outcomes, but the unanswered questions are too fundamental to clinical care not to receive our full attention. I thank Professor Pierre Foex for his helpful comments before and during preparation of this manuscript. RDS has received funding from the AAGBI/Anaesthesia to study the stratification of pre-operative BP thresholds.