Article Text
Abstract
Objective To investigate the association between systolic and diastolic blood pressure (BP) and risk of hospitalisation among patients with type 2 diabetes.
Methods 4704 patients with type 2 diabetes from 18 general practices in Cambridgeshire were included. BP was assessed in 2008–2009. The primary outcome was cardiovascular hospital admissions in 2009–2011. Adjusted relative risks for each BP measurement group were estimated using Cox models. Further dose–response relationships between BP and risks were explored using restricted cubic spline models.
Results Over a median follow-up of 2 years, we recorded 5322 hospital admissions. There was a non-linear relationship between systolic (SBP) and diastolic (DBP) BP and the risk of cardiovascular hospitalisation (both p<0.001 for linearity test). The BP associated with the lowest risk of cardiovascular hospitalisation was 137 (95% CI 133 to 141)/78 (95% CI 76 to 80) mm Hg. The discrimination of the model could be significantly improved with either an SBP threshold or a DBP threshold (both p<0.0001).
Conclusions Among patients with type 2 diabetes, the risk of cardiovascular hospitalisation is lowest with a BP of 133–141/76–80 mm Hg. This concurs with the latest recommended randomised controlled trial based BP target of 140/80 from the American Diabetes Association.
Statistics from Altmetric.com
Background
The prevalence and cost of diabetes is growing rapidly worldwide.1 People with diabetes are twice as likely to be admitted to hospital, and at least 10% of those in hospital have diabetes at any one time.2 In some age groups, it is as many as one in five.3 The associated costs of excess admissions, as well as increased costs per admission, are significant contributors to the financial burden borne by healthcare systems from diabetes and often reflect preventable morbidity suffered by patients.4 Poor blood pressure (BP) control in patients with diabetes is an important risk factor for hospitalisation and might be useful in defining population risk of morbidity and increased health costs.5
Previous studies have reported an increased risk of mortality in patients with diabetes, with 80% dying from cardiovascular-related events.6 Some studies have described a non-linear relationship between BP measures and mortality,7 ,8 with a suggestion that aggressive BP lowering in high-risk patients could do harm rather than provide cardiovascular protection.9 However, few studies have investigated a possible relationship between BP measures and the risk of hospitalisation in people with type 2 diabetes.
We hypothesised that a non-linear relationship exists between BP measures and the risk of cardiovascular hospitalisation in people with type 2 diabetes. This study examines this relationship, and investigates whether a threshold exists for BP in predicting the risk of cardiovascular hospital admissions.
Methods
Study design and participants
Patient lists from 18 general practices across Cambridgeshire, England, in 2008/2009 were collated and linked with hospital admissions (Secondary Uses Service (SUS)) data as part of an evaluation of diabetes care across the county by the local health board, National Health Service (NHS) Cambridgeshire. This work was limited to volunteer practices using the Egton Medical Information Systems (EMIS) general practitioner (GP) software system, from which a predefined set of data could be extracted. There was no systematic selection process for these surgeries, and data extracted were for their entire diabetes population. All patients with diabetes (4704) had follow-up hospitalisation data to 2010–2011. Hospital admissions to NHS and private hospitals within and outside Cambridgeshire were followed-up. No personal identifiers were released to researchers, and all subsequent analyses were conducted on anonymised datasets. The work had approval from the Cambridgeshire research ethics committee as part of a wider service evaluation.
Briefly, 4704 patients were defined as having type 2 diabetes by GP diagnosis, International Classification of Diseases (ICD) description in the SUS dataset, the diagnosis time interval (if patients were defined as having both type 2 diabetes and type 1 diabetes in different datasets), and their age. All hospitalised patients had a diabetes ICD code (E10–E14) within the ICD coding.
The GP dataset included demographic characteristics, some treatment, body mass index (BMI) and metabolic variables including glycated haemoglobin (HbA1c) and lipid profiles. All patients with diabetes had their BP (systolic (SBP) and diastolic (DBP)) measured at least once a year after the diagnosis of diabetes. The BP in the dataset was the most recent and was taken before 1 April 2009. There was a minimum of 50 days before the first admission.
Diabetes duration was not universally recorded, and hence was not usefully available for analysis; however, the duration of diabetes registration in GP practices was recorded, which often reflected the diabetes duration. Diabetes therapy was not included in the dataset. Previous cardiovascular history was recorded. Lipid-lowering and aspirin treatment were recorded.
The primary outcome of the study was hospital admissions with cardiovascular disease (CVD) as the primary diagnosis (ICD-10: I20–I25, I60–I69 and I73 in the first ICD field) over the 2-year time period (1 April 2009–31 March 2011). Secondary analyses were for admissions (i) with all-cause and (ii) with diabetes as the primary diagnosis excluding CVD (ICD code (E10–E14) in the first ICD field).
Statistical analysis
Comparisons were conducted by logistic regression for categorical variables and analysis of covariance for continuous variables with adjustment for age.
For all outcomes, Cox proportional hazards models were fitted to estimate the HRs and the 95% CIs. The proportional hazards assumption was met by checking the proportional hazard test with p values and the Schoenfeld residual. Adjustments were made for age, gender, prevalent recorded history of cardiovascular/cerebrovascular disease, BMI, HbA1c, triacylglycerol, total cholesterol, low-density lipoprotein (LDL cholesterol), high-density lipoprotein (HDL cholesterol), duration of diabetes registration in GP practices, lipid-lowering treatment and aspirin treatment.
The percentage of missing values was 1.37%, 2.44%, 6.75%, 4.95%, 6.78% and 2.98% for BMI, HDL, LDL, triacylglycerol, total cholesterol and estimated glomerular filtration rate (eGFR), respectively. Although anorder effect was not revealed by the percentage of missing values, multiple imputation was still carried out by using all the other covariables plus the outcome.
The relationship between BP measures and risk of hospital admission was estimated using a linear model, a natural cubic spline model with three equally spaced knots determined from the levels of BP measures, and a quadratic spline model. The natural cubic spline model was chosen as the best fit model for the relationship curve by its minimum Akaike information criterion (AIC) compared with the linear model or quadratic spline model10 (the model fit statistics are listed in online supplementary table S1). For sensitivity analysis, the natural cubic spline models for the overall dataset were repeated using other potential knots, chosen to lie within the range for minimum to maximum measure of BP. The linear test was used in the natural cubic spline model to test the linearity of the relationship. The break-point test11 was carried out to target the potential thresholds (P5 to P95 of BP measures) by incorporating the piecewise term into the cubic spline model. The threshold with a significant break in the regression coefficients and achieving the minimum AIC was chosen as the final threshold. The 95% CI of the threshold was obtained from 1000 bootstrap samples.
The CI of the threshold was also used in a multivariate logistic regression model to observe the improvement in model discrimination by C-statistics.
Two sets of sensitivity analyses were implemented to observe the association between BP measures and risk of cardiovascular, all-cause and diabetes-coded hospitalisation. First, all analyses were limited to hospitalisation after the first 3 months after the start of follow-up. Second, all analyses were carried out in the BP measurement data-rich range (covering >95% patients): 100–180 mm Hg for SBP and 60–100 mm Hg for DBP, respectively.
All analyses were performed with Stata/SE V.11.0. All p values were calculated using two-tailed tests, and p<0.05 was taken as significant.
Results
In total, 4704 participants with type 2 diabetes were included in the analyses. The characteristics of patients are described by SBP or DBP group (table 1). Patients with SBP 130–139 mm Hg and DBP 80–84 mm Hg were defined as the reference groups. Other patients were classified into groups by every 10 mm Hg segment from the reference group in SBP, and every 5 mm Hg segment from the reference group in DBP.
Compared with patients with SBP of 130–139 mm Hg, patients with higher SBP (especially SBP ≥160 mm Hg) were more likely to be older and have higher total cholesterol, LDL and triacylglycerol concentrations, whereas patients with lower SBP (especially SBP <110 mm Hg) were more likely to be male, and have lower measures of BMI and less likely to have a history of coronary heart disease.
In contrast with patients with DBP of 80–84 mm Hg, patients with higher DBP were more likely to be younger and male and have higher levels of HbA1c, total cholesterol, LDL and triacylglycerol, and a lower percentage had lipid-lowering treatment; older age, lower BMI and lower prevalence of coronary heart disease history.
A BP of 137 (95% CI 133 to 141)/78 (95% CI 76 to 80) mm Hg was estimated to be associated with the lowest risk of cardiovascular hospitalisation, as tested by linear threshold models. Similar results were found in the secondary-outcome analyses. For both all-cause and diabetes admissions, the BP threshold was 137 (95% CI 133 to 141)/78 (95% CI 76 to 80) mm Hg. For cardiovascular admissions, the increased hospitalisation risk was more likely to be observed below the BP thresholds. For outcomes of all-cause and diabetes-coded admissions, increased hospitalisation risk was observed for both those with BP above and below the thresholds.
The level of discrimination of the main model measured by C-statistics in a multivariate logistic regression model was improved from 0.77 (95% CI 0.75 to 0.79) to 0.81 (95% CI 0.80 to 0.83) with inclusion of the SBP threshold (133–141 mm Hg) and to 0.79 (95% CI 0.77 to 0.81) with inclusion of the DBP threshold (76–80 mm Hg) (both p<0.0001).
Relationship curves were derived from the natural cubic spline models with adjustment of covariates in figure 1. There was a non-linear relationship between systolic/diastolic BP and risk of cardiovascular hospital admission (linearity test: p<0.001). Similar patterns of non-linear relationships were observed for the secondary outcomes of all-cause admissions and diabetes-coded admissions (linearity test: both p<0.001) (see online supplementary figure S1).
The non-linear association was also found in the sensitivity analysis to model the association in the data-rich range (100–180 mm Hg for SBP and 60–100 mm Hg for DBP) (figure 2; see online supplementary figure S2). A significantly higher risk of cardiovascular hospitalisation was found below/above the 137 mm Hg SBP threshold, whereas a significantly higher risk of cardiovascular hospitalisation was only found below the 78 mm Hg DBP threshold (figure 2). Higher risks of all-cause and diabetes-coded hospitalisation were found below/above the 137 mm Hg SBP and 78 mm Hg DBP thresholds (see online supplementary figure S2).
In another set of sensitivity analyses, after excluding hospital admissions in the first 3 months of follow-up, we obtained qualitatively similar findings for the association between BP measures and cardiovascular, all-cause and diabetes-coded hospitalisation (see online supplementary figure S3). Furthermore, the shapes of associations were still similar in the analysis within the data-rich range and with exclusion of hospitalisation in the first 3 months (see online supplementary figure S4).
Discussion
This investigation was undertaken to relate SBP and DBP to the risk of cardiovascular hospitalisation in a local cohort of patients with type 2 diabetes. We focused our investigation on the shape of the relationship assessing the evidence for non-linearity and in particular on the existence of a threshold. In all our analyses, we found evidence that the associations are non-linear. Threshold analysis provided evidence of a BP threshold: 137 (95% CI 133 to 141)/78 (95% CI 76 to 80) mm Hg. The significantly higher risk of cardiovascular admissions was found above and below 137 mm Hg SBP and below 78 mm Hg DBP. The risks of all-cause admissions and diabetes-coded admissions were both greater above and below 137/78 mm Hg.
Diabetes-related hospital admissions are an important component of the increased costs of diabetes and the preventable morbidity suffered by people with diabetes.4 The BP control target among those with diabetes is an important risk factor for hospitalisation and might be useful in defining population risk of morbidity and increased health costs.5 Few studies have set out to investigate the association between BP level and risk of cardiovascular hospital admission in people with type 2 diabetes.
Our investigation also identified a significant association between a higher risk of cardiovascular hospitalisation, and a lower BP, possibly reflecting BP management that is too tight among some of those with diabetes. Our dataset was not able to discriminate between those with such possible overtreatment and those with different forms of cardiomyopathy.
Clinical guidelines recommend maintaining BP levels at below 140/90 mm Hg in people with hypertension in the general population.12 ,13 Guidelines recommend further lowering the BP to a treatment goal of below 130/80 mm Hg in high-risk patients, including people with diabetes, cerebrovascular and coronary arterial disease, or kidney disease.14 ,15 For example, between 2002 and 2013, the American Diabetes Association (ADA) recommended that the BP treatment goal for patients with diabetes should be <130/80 mm Hg, concluding that ‘there is no threshold value for BP’, in keeping with early randomised trials such as the UK Prospective Diabetes Study (UKPDS)16 and Hypertension Optimal Treatment (HOT) trial.17 These trials showed major reductions in cardiovascular outcomes in the groups receiving tighter control of BP compared with those receiving conventional control. Furthermore, evidence from epidemiological studies suggests that cardiovascular risk starts to increase above a BP of 115/75 mm Hg.5 This evidence has led to the consensus that there is no lower threshold for BP lowering, supporting the idea that ‘the lower the better’.18 –20
Recently, such intensive control of BP in type 2 diabetes has been questioned. The Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial did not show further cardiovascular benefits of antihypertensive therapy by reducing SBP below 130 mm Hg in people with diabetes.21 Importantly, some research has suggested that BP lowering that is too aggressive in high-risk patients could be harmful rather than provide cardiovascular protection.9 ,22 Low BP, particular below 110/75 mm Hg, was associated with an increased risk of poor outcomes.23 A recent UK retrospective cohort study with 126 092 adult patients with a new diagnosis of type 2 diabetes revealed that BP below 130/80 mm Hg was not associated with reduced risk of all-cause mortality with or without CVD. Low BP, particularly below 110/75 mm Hg, was associated with an increased risk of poor outcomes.24 In 2013, the ADA amended its BP target to 140/80 mm Hg, although some patients (eg, younger, nephropathy) will have lower targets (similarly, some patients at risk of falls with postural hypotension may have a higher target).14
Our results extend previous findings, suggesting that lower levels of BP identify a subset of patients with significantly increased risk of cardiovascular hospital admission. This association was observed for both systolic and diastolic BP. The existence of a non-linear relationship between BP and cardiovascular outcomes has long been debated. The non-linear relationship was more marked in individuals with existing CVD, although it was also observed in patients without CVD.25 Data from HOT do not suggest the presence of a J-shaped relationship with regard to BP and cardiovascular outcomes,17 which has been the topic of some debate26; however, only 6% of the overall HOT participants had CVD at baseline, and lack of a J-shaped relationship might be a reflection of a population at low risk. Data from INVEST, which only enrolled patients with coronary artery disease, showed the relationship between BP and cardiovascular outcomes to be J-shaped in the whole population analysed20 and in the cohort with diabetes in particular.7 The Irbesartan Diabetic Nephropathy Trial also found a robust J-shaped relationship across increments of 10 mm Hg in DBP with an increased risk of myocardial infarction,9 confirming the association between BP and risk observed in the meta-analysis published previously.25 Our analysis especially focused on patients with diabetes, with adjustment for CVD history. We also observed the non-linear relationship between BP and risk of hospitalisation, with the threshold of 137/78 mm Hg. An increased risk of cardiovascular admissions was more likely to be found among those with BP below the threshold, which does not support the older recommendation of an SBP goal <130 mm Hg in patients with diabetes, consistent with the findings from ACCORD21 and INVEST.27 The significantly increased risk of hospitalisation was when SBP was <137 mm Hg for all-cause and diabetes-coded admissions. This threshold might be useful for differentiating patients with a higher risk of cardiovascular hospitalisation from those with a lower risk of cardiovascular hospitalisation.
Some limitations have to be considered in the interpretation of our findings. The associations were observed in a cohort with a short follow-up. The patients may have differed in other risk factors between BP groups. Although many factors had been adjusted in our analysis, these adjustments may not have been enough and might not have incorporated other unknown factors, especially antihypertensive therapy, as major residual confounders. A large proportion of patients accepted lipid-lowering therapy at the time of the diagnosis of diabetes, which might have reduced cardiovascular risk. While the BP provided reflects ‘real-life’ BP measurements, they were not measured with epidemiologically robust methods. This would have increased variance and reduced the chance of finding differences, but should not have influenced our findings unless there was a systematic bias. The relatively limited sample size, and without any breakdown of the nature of the cardiovascular hospitalisations, suggests these associations, and the thresholds should be tested in a larger cohort with a longer follow-up period in the future. There is a possibility that the index BP was measured after the episode of illness that led to the hospital admission (ie, reverse causality). As there was at least 50 days between the index BP and hospital admission, it is likely that the number of cases where this was material would be very small and unlikely to influence our findings. However, those with, for example, longstanding ischaemic or other forms of cardiomyopathy would be expected to be included in the low-BP group and to have an increased chance of hospitalisation.
In summary, there was a non-linear relationship between BP level and risk of cardiovascular hospital admission in this cohort of patients with type 2 diabetes. The BP threshold for the lowest hospital admission was 137 (95% CI 133 to 141)/78 (95% CI 76 to 80). SBP above 137 mm Hg was associated with a high risk of hospitalisation, and, in particular, SBP below 137 mm Hg or DBP below 78 mm Hg was significantly associated with a higher risk of hospital admission.
Key messages
-
What is known about this subject?
-
The association between blood pressure and all-cause mortality is well established. Few studies have addressed the association between blood pressure and hospitalisation rates, particularly cardiovascular hospitalisation rates among people with type 2 diabetes.
-
What might this study add?
-
A non-linear relationship between blood pressure measurements and cardiovascular hospital admission has been shown in this study, with an increased risk of hospitalisation not only among patients with high blood pressure but also among those with low blood pressure. The lowest risk of hospitalisation for patients with type 2 diabetes was identified as 137/78 mm Hg. This threshold might be useful for identifying people with a higher risk of cardiovascular hospitalisation.
-
How might this impact on clinical practice?
-
Our findings support trials of careful review of the management plans of those with lower as well as higher blood pressure to reduce the risk of hospitalisation. Our findings suggest that patients with low blood pressure are at high risk of hospitalisation.
Acknowledgments
DS thanks NIHR Cambridge Biomedical Research Centre for support. DS is guarantor. We thank NHS Cambridgeshire for the anonymised data.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Footnotes
-
Contributors DY analysed research data and wrote the manuscript. DS reviewed and edited the manuscript.
-
Funding This paper presents independent research funded by the National Institute for Health Research (NIHR) under its Research for Patient Benefit (RfPB) Programme (Grant Reference Number PB-PG-0808-17303). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
-
Competing interests None.
-
Ethics approval Cambridgeshire research ethics committee.
-
Provenance and peer review Not commissioned; externally peer reviewed.