The Systolic Blood Pressure Difference Between Arms and Cardiovascular Disease in the Framingham Heart Study
Abstract
Background
An increased inter-arm systolic blood pressure difference is an easily determined physical examination finding. The relationship between inter-arm systolic blood pressure difference and risk of future cardiovascular disease is uncertain. We described the prevalence and risk factor correlates of inter-arm systolic blood pressure difference in the Framingham Heart Study (FHS) original and offspring cohorts and examined the association between inter-arm systolic blood pressure difference and incident cardiovascular disease and all-cause mortality.
Methods
An increased inter-arm systolic blood pressure difference was defined as ≥10mmHg using the average of initial and repeat blood pressure measurements obtained in both arms. Participants were followed through 2010 for incident cardiovascular disease events. Multivariable Cox proportional hazards regression analyses were performed to investigate the effect of inter-arm systolic blood pressure difference on incident cardiovascular disease.
Results
We examined 3,390 (56.3% female) participants aged 40 years and older, free of cardiovascular disease at baseline, mean age of 61.1 years, who attended a FHS examination between 1991 and 1994 (original cohort) and from 1995 to 1998 (offspring cohort). The mean absolute inter-arm systolic blood pressure difference was 4.6 mmHg (range 0 to 78). Increased inter-arm systolic blood pressure difference was present in 317 (9.4%) participants. The median follow-up time was 13.3 years, during which time 598 participants (17.6%) experienced a first cardiovascular event including 83 (26.2%) participants with inter-arm systolic blood pressure difference ≥10 mmHg. Compared to those with normal inter-arm systolic blood pressure difference, participants with an elevated inter-arm systolic blood pressure difference were older (63.0 years vs. 60.9 years), had a greater prevalence of diabetes mellitus (13.3% vs. 7.5%,), higher systolic blood pressure (136.3 mmHg vs. 129.3 mmHg), and a higher total cholesterol level (212.1 mg/dL vs. 206.5 mg/dL). Inter-arm systolic blood pressure difference was associated with a significantly increased hazard of incident cardiovascular events in the multivariable adjusted model (hazard ratio 1.38, 95% CI, 1.09 to 1.75). For each 1-standard deviation unit increase in absolute interarm systolic blood pressure difference, the hazard ratio for incident cardiovascular events was 1.07 (CI, 1.00 to 1.14) in the fully-adjusted model. There was no such association with mortality (hazard ratio 1.02, 95% CI 0.76 to 1.38).
Conclusions
In this community-based cohort, an inter-arm systolic blood pressure difference is common and associated with a significant increased risk for future cardiovascular events, even when the absolute difference in arm systolic blood pressure is modest. These findings support research to expand clinical use of this simple measurement.
Background
An increased inter-arm systolic blood pressure difference is usually defined as 10 mmHg or greater1, 2 and can be found in up to 24% of healthy persons3–6. It has been suggested that identification of inter-arm systolic blood pressure difference is crucial for appropriate identification and treatment of hypertension7–9, for clinical decision making and surveillance1, 10, 11, 12, 13. Inter-arm systolic blood pressure differences have been studied in patients in primary care settings14–16, in patients with various manifestations of vascular disease17–19, and in population-based cohorts3, 6, 16. It has been found to correlate with classic and novel cardiovascular risk factors and may be predictive of cardiovascular events. Nevertheless, data on the epidemiology of inter-arm systolic blood pressure difference are relatively scarce3, 6, 16, 17. While an association between inter-arm systolic blood pressure difference and mortality was noted in three small prospective studies and a meta-analysis that pooled data from 20 older studies14, 15, 17, 18, 20, the evidence for an association with cardiovascular disease is less consistent20. The objectives of this study were to describe the distribution of inter-arm systolic blood pressure difference and risk factor correlates in the Framingham Heart Study (FHS) original and offspring cohorts and to determine the association between inter-arm systolic blood pressure difference and incident cardiovascular events and all-cause mortality.
Methods
Study participant sample
Methods of recruitment have been previously described for participants from the original cohort and the offspring study cohort of the FHS21, 22.
From the 5209 original FHS cohort that was recruited in 194823, 1026 men and women attended a FHS examination between 1991 and 1994. There were also 3532 men and women from the offspring study cohort who attended the 6th offspring cycle examination from 1995 to 1998. Of the 4558 total original and offspring cohort participants, 494 were excluded because they did not have a BP measurement in both arms. Of the remaining participants, 66 participants <40 years old and 608 with prevalent CVD were excluded. The Institutional Review Board at Boston University Medical Center approved the examination content and informed consent was obtained from study participants at the time of their examination.
Inter-arm systolic blood pressure measurement
For the purpose of calculating the inter-arm systolic blood pressure difference, brachial systolic blood pressure measurements were obtained as part of an ankle-brachial index (ABI) measurement by trained technicians according to a standard protocol after participants lay supine for at least 5 minutes. Measurements were taken in the sequence of right arm, left arm, right ankle, and left ankle and repeated twice. If the initial and repeat blood pressure differed by >10 mm Hg at any site, a third measurement was taken. When two blood pressure measurements were obtained in the same arm, the average blood pressure in that arm was used. The inter-arm systolic blood pressure difference was defined as the absolute difference between the single or average blood pressure measurement in each arm. An increased inter-arm systolic blood pressure difference was defined as inter-arm systolic blood pressure difference ≥10 mmHg14,18. The maximal difference between arms was used.
Definition of incident cardiovascular disease events and mortality
Participants were followed through 2010 for the first incident cardiovascular event, defined as coronary heart disease (i.e., a fatal coronary event, recognized myocardial infarction and unrecognized myocardial infarction, coronary insufficiency, or angina), a cerebrovascular event (i.e. ischemic stroke or transient ischemic attack), intermittent claudication24, or congestive heart failure25. An unrecognized myocardial infarction was adjudicated when serial changes on the electrocardiogram showed development of pathologic Q waves and neither the participant nor the physician considered the possibility of a myocardial infarction. Mortality data was obtained by review of all medical records including the last hospitalization, nursing home records, personal physical records and if needed next-of kin interviews along with death certificate evaluation. All cardiovascular events and deaths in both the original cohort and offspring participants were adjudicated by an endpoint committee of three senior investigators using the same criteria and all available medical records. Ischemic stroke and transient ischemic attack events were reviewed by a panel of study neurologists.
Definition of risk factor covariates
Covariates were obtained at the same time as the inter-arm systolic blood pressure difference measurement. Hypertension was defined as a blood pressure ≥140/90 mmHg or the use of anti-hypertension medications. Body mass index (BMI) was calculated as the participants’ weight in kilograms divided by the participants’ height in meters squared. Diabetes mellitus was considered present if the fasting glucose was ≥126 mg/100 mL (offspring), a non-fasting glucose of ≥200 mg/dL (original cohort), or if the participant was receiving treatment with insulin or oral hypoglycemic medications. Total cholesterol was analyzed as a continuous variable. Lipid-lowering therapy was entered into the model as a separate binary variable. Participants were queried at each examination about cigarette smoking and current smoking was defined as smoking at least one cigarette per day in the year preceding the examination. The 10-year Framingham risk score for general cardiovascular disease was calculated for each participant26.
Statistical Analysis
Descriptive statistics were generated. Continuous measures were summarized using mean and standard deviation. Categorical traits were summarized using percentages. We standardized Inter-arm Systolic Blood Pressure Difference (ISBPD) to mean=0 and SD=1 since this measure is skewed to the right. HR for continuous ISBPD were presented for every 1-SD unit increase in blood pressure. The main exposure, inter-arm systolic blood pressure difference, was first analyzed as a binary variable (i.e., <10 mmHg and ≥10mmHg), then as a 3-level ordinal variable <10 mmHg vs. 10 to <15 mmHg and ≥15 mmHg and then as a continuous trait. We tested for linear trend in the hazards across the 3-level ordinal variable. Two-sample t-test with pooled variance or unequal variance as indicated was used to compare means between participants who had inter-arm systolic blood pressure difference ≥10 mmHg versus those with inter-arm systolic blood pressure difference <10mmHg. The Chi-squared test was used to compare categorical traits. Age and sex adjusted time to first cardiovascular event was analyzed using Kaplan- Meier plots and evaluated using the log-rank Chi-squared test.
Next, after confirming the validity of the Cox proportional hazards regression assumptions, two Cox models were fitted with inter-arm systolic blood pressure difference < 10 mmHg as the referent category. The first model adjusted for the following variables: age, sex, hypertension, total cholesterol level, HDL cholesterol, cholesterol treatment, presence or absence of diabetes mellitus, and smoking status and the second model adjusted for the 10-year Framingham risk score for general cardiovascular disease26. We also assessed for linear trend using the 3-category ordinal inter-arm systolic blood pressure difference variable (i.e. inter-arm systolic blood pressure difference <10 mmHg, 10 to <15 mmHg, and ≥15 mmHg). Furthermore, similar models were fitted accounting for familial correlations between the original and offspring cohorts. We used generalized estimating equations (GEEs) to account for the familial correlations within our dataset. Similar analysis approaches were used restricting the sample to hypertensive participants and also using mortality as the outcome.
The predictive utility of the Cox model was examined by comparing the c-statistic. The incremental effect of adding inter-arm systolic blood pressure difference to the Framingham risk score for predicting cardiovascular outcomes was evaluated with the use of category based net reclassification improvement index (NRI)27. The NRI is used to assess how well an exposure “reclassifies” patients from one risk category to another. We calculated NRI using an extension to survival analysis that employs Kaplan-Meier estimates of event probabilities using categories: low risk (0 to <6%), intermediate risk (6 to 20%), or high risk (>20%). A large NRI indicates that the additional exposure causes a large improvement in reclassification.
Results
Our final study sample included 3,390 participants; mean age of 61.1±11.3 years, 56.3% women, followed for an average of 13.3 years. The median inter-arm systolic blood pressure difference was 3 (IQR 2,6) mmHg and 317 (9.4%) participants had an inter-arm systolic blood pressure difference ≥10 mmHg (Table 1). The distribution of systolic blood pressure difference between arms in the sample is shown in Figure 1. There were 3073, 246 and 71 participants with inter-arm systolic blood pressure difference of <10 mmHg, 10 to <15 mmHg and ≥15 mmHg, respectively.
Table 1
Age and Sex-Adjusted Comparison of Baseline Variables by Inter-arm Systolic Blood Pressure Difference
| Variable | ISBPD <10mmHg (n=3073*) | ISBPD ≥10mmHg (n=317*) | P Value |
|---|---|---|---|
| Age** | 60.9 (11.3) | 63.0 (11.5) | 0.002 |
| Male (%)** | 1343 (43.7) | 137 (43.2) | 0.87 |
| Body mass index kg/m2 | 27.4 (4.8) | 29.6 (6.2) | <0.001 |
| SBP mmHg | 129.3 (19.3) | 136.3 (20.3) | <0.001 |
| DBP mmHg | 74.9 (9.8) | 76.3 (9.9) | 0.02 |
| Hypertension (%) | 1249 (40.6) | 184 (58.0) | <0.001 |
| HTN treatment (%) | 789 (25.7) | 137 (43.6) | <0.001 |
| Total cholesterol mg/dL | 206.5 (39.3) | 212.1 (40.2) | 0.17 |
| HDL cholesterol mg/dL | 51.8 (16.2) | 51.2 (16.9) | 0.53 |
| Triglycerides mg/dL | 137.8 (135.2) | 151.2 (105.8) | 0.004 |
| Lipid lowering treatment (%) | 282 (9.2) | 38 (12.0) | 0.10 |
| Diabetes mellitus (%) | 229 (7.5) | 42 (13.3) | 0.003 |
| Current cigarette use (%) | 426 (13.9) | 39 (12.3) | 0.44 |
| Ever cigarette use (%) | 1384 (45.0) | 137 (43.2) | 0.54 |
| Framingham 10 year risk score | 12.3 (10.6) | 15.2 (12.1) | <0.001 |
Abbreviations: ISBPD – Inter-arm systolic blood pressure difference, SBP – Systolic blood pressure, DBP – Diastolic blood pressure, HTN – Hypertension
Overall, 598 participants (17.6%) experienced a first cardiovascular event including 83 (26.2%) participants with an increased inter-arm systolic blood pressure difference. The overall incidence rate of total cardiovascular disease was 17.64 per 1000 person-years (95% CI 16.25–19.11). Age and sex adjusted Kaplan-Meier plots stratified by binary inter-arm systolic blood pressure difference (<10 mmHg vs. ≥10mmHg) are shown in Figure 2 and show that the inter-arm systolic blood pressure difference ≥10mmHg category had a significantly higher probability for incident cardiovascular event than the interarm systolic blood pressure difference <10 mmHg category (log-rank test Chi-squared=18.2, p<0.0001).
Age and sex adjusted Kaplan-Meier plot for incident cardiovascular events in 3,390 participants with or without an inter-arm systolic blood pressure difference ≥10 mm Hg (Log-rank test Chi-squared p<0.0001).
An inter-arm systolic blood pressure difference ≥10 mmHg was associated with a significantly increased hazard of incident cardiovascular events in the multivariable adjusted Cox models (HR 1.38, 95% CI, 1.09 to 1.75) (Table 2). There was no change in the hazard with adjustment for Framingham risk score. In a multivariable model, an inter-arm systolic blood pressure difference ≥15 mmHg was associated with an increased hazard of incident cardiovascular events (HR 1.47, 95% CI 0.93–2.32). There was a statistically significant linear trend with increasing inter-arm systolic blood pressure difference category (p=0.009 in the multivariable adjusted model). An inter-arm systolic blood pressure difference ≥10 mmHg was also associated with a significantly increased hazard of incident cardiovascular events in a subgroup analysis of 1433 hypertensive participants (HR=1.50, 95% CI 1.14–1.97, in the multivariable model). In similar analyses for total mortality, there were no significant associations (Table 2). For each 1-SD increase in absolute inter-arm systolic blood pressure difference, the HR for incident cardiovascular events was 1.07 (CI, 1.00–1.14) in the multivariable-adjusted model (Table 3). There was no significant sex interaction with inter-arm systolic blood pressure difference. Adjustment for familial clustering did not materially alter the results.
Table 2
Hazard Ratios of an Incident Cardiovascular Event or Mortality According to ISBPD Category
| BP Category | Incident CVD HR (95% CI) | Mortality HR (95% CI) | |
|---|---|---|---|
| Primary Analysis for ISBPD≥10 mmHg: | |||
| MV adjusted* | <10 referent | 1.00 | 1.00 |
| ≥ 10 mmHg | 1.38 (1.09,1.75) | 1.02 (0.76,1.38) | |
| Framingham 10 year risk score | <10 referent | 1.00 | 1.00 |
| ≥ 10 mmHg | 1.38 (1.09,1.75) | 0.98 (0.72,1.33) | |
| Analysis for ISBPD Categories 10 to <15 mmHg and ≥15 mmHg: | |||
| MV adjusted* | <10 referent | 1.00 | 1.00 |
| 10 to <15 mmHg | 1.36 (1.04,1.77) | 0.99 (0.71,1.39) | |
| ≥15 mmHg | 1.47 (0.93,2.32) | 1.11 (0.62,1.99) | |
| Trend p-value | 0.009 | 0.82 | |
| Framingham 10 year risk score | <10 referent | 1.00 | 1.00 |
| 10 to <15 mmHg | 1.32 (1.01,1.72) | 0.95 (0.67,1.35) | |
| ≥15 mmHg | 1.62 (1.02,2.56) | 1.05 (0.57,1.95) | |
| Trend p-value | 0.005 | 0.97 | |
Abbreviations: ISBPD – Inter-arm systolic blood pressure difference, HR – Hazard ratio, CI – Confidence interval, MV – Multivariable
Table 3
Standardized ISBPD: Risk-Factor-Adjusted Relative Risk of a First Cardiovascular Event According to ISBPD
| Stratification | HR (95% CI)* | P-Value |
|---|---|---|
| MV adjusted** | 1.07 (1.00,1.14) | 0.04 |
| Framingham 10 year risk score | 1.08 (1.02,1.15) | 0.02 |
Abbreviations: ISBPD – Inter-arm systolic blood pressure difference, HR – Hazard ratio, CI – Confidence Interval, MV – Multivariable
We examined model performance and the utility of inter-arm systolic blood pressure difference in reclassification of overall risk for incident cardiovascular disease. Both the c-statistic for a multivariable model that included inter-arm systolic blood pressure difference and a multivariable model that did not include inter-arm systolic blood pressure difference were 0.74 (95% CI: 0.72–0.76). Category-based net reclassification for improvement changed significantly with inter-arm systolic blood pressure difference (NRI= 0.021; P= 0.02). The net reclassification is presented in table 4.
Table 4
Net Reclassification Stratified by CVD Events
| No CVD event (n=2783) | Incident CVD event (n=593) | |||||
|---|---|---|---|---|---|---|
| Risk category with ISBPD | <6% | 6–20% | >20% | <6% | 6–20% | >20% |
| Risk category without ISBPD | ||||||
| <6% | 543 | 16 | 0 | 177 | 6 | 0 |
| 6–20% | 23 | 1465 | 20 | 5 | 260 | 7 |
| >20% | 0 | 26 | 690 | 0 | 4 | 134 |
CVD – Cardiovascular; ISBPD – Inter-arm systolic blood pressure difference
Discussion
In this large prospective, community based cohort of middle-age men and women free of cardiovascular disease, an increased inter-arm systolic blood pressure difference was found to be present in nearly 10% of individuals and is associated with increased levels of traditional cardiovascular risk factors. Furthermore, an increased inter-arm systolic blood pressure difference is associated with an increased risk for incident cardiovascular events, independent of traditional cardiovascular risk factors. This association is consistent when inter-arm systolic blood pressure difference was examined as a binary and as a continuous variable and showed a positive linear trend across categories of increasing inter-arm systolic blood pressure difference. This is the first community based cohort study to show such an association.
While we did not examine the anatomical correlates of increased inter-arm systolic blood pressure difference, our findings are consistent with previous publications linking subclavian artery stenosis to increased cardiovascular risk3, 28 and are indirectly consistent with studies linking increased inter-arm systolic blood pressure difference with subclavian artery stenosis. Indeed, inter-arm systolic blood pressure difference >10 mmHg has been shown to have a high sensitivity29, 30, yet a low specificity31, for angiographically significant subclavian artery stenosis. Furthermore, in two population-based studies, subclavian artery stenosis has been found to correlate with conventional risk factors for atherosclerosis6, 16. While a recent meta-analysis did not find a correlation between an inter-arm systolic blood pressure difference ≥15 mmHg and a known pre-existing diagnosis of cardiovascular disease when pooling 6 studies20, subclavian artery stenosis has been shown to correlate with ischemic heart disease as detected by stress myocardial single-photon emission computed tomography32.
In 6743 participants free of clinical cardiovascular disease in the Multi-Ethnic Study of Atherosclerosis (MESA) an inter-arm systolic blood pressure difference ≥15 mmHg was found in 307 (4.5%) and correlated with subclinical measures of cardiovascular disease3. Consistent with our findings, the MESA participants with an increased inter-arm systolic blood pressure difference tended to be older, more likely to have diabetes mellitus and hypertension, and were more obese than participants with a normal interarm systolic blood pressure difference. An increased inter-arm systolic blood pressure difference was found to correlate with markers of atherosclerosis including peripheral artery disease (ABI≤0.9), an increase in carotid intima-media thickness and a high coronary calcium score. Mortality and cardiovascular events were not assessed.
Several small prospective studies in cohorts of selected participants and a single meta-analysis previously described an association between inter-arm systolic blood pressure difference and both all-cause and cardiovascular mortality14, 17, 18, 20, however an association with cardiovascular disease in those studies was inconsistent20. In a prospective study of 230 hypertensive patients (mean age 68.1 years) who were followed for a median of 9.8 years in a primary care setting, 55 (23%) had an inter-arm systolic blood pressure difference ≥10 mmHg14. An inter-arm systolic blood pressure difference ≥10 mmHg was associated with incident CV events and with all cause mortality and there was also an increase of 5–6% in the incidence of mortality for each 1 mmHg increase in inter-arm systolic blood pressure difference. In another prospective cohort recruited from two clinics, 421 consecutive patients (mean age 62.9 years), 95% of whom were men, were followed for a median of 5.6 years. Mortality was 31% and associated with an increase in inter-arm systolic blood pressure difference18. In contrast, overall mortality in our study occurred in 12.2%, over a longer period of time. Incident cardiovascular disease was not reported in this study. These two studies were small and selective, focusing on patients with hypertension or renal disease, making them less useful for drawing general conclusions. Adequately, in the secondary analysis in the subset of our study participants with hypertension, we observed a significant association between inter-arm systolic blood pressure difference and incident cardiovascular events. Finally, in another study combining three cohorts, 1,778 subjects were recruited from both non-invasive vascular diagnostic laboratories and the community and followed for a median of 9.4 years17. In a fully adjusted model accounting for cardiovascular risk factors and for baseline cardiovascular disease, an inter-arm systolic blood pressure difference ≥15 mmHg demonstrated only increasing but not significant hazards for cardiovascular mortality (HR=1.43, 95% CI 0.93–2.18). Incident cardiovascular disease was again not reported, the study population was enriched with high-risk patients being evaluated in vascular labs, and cause of death was adjudicated based on death certificate data. As expected our study participants were of lower risk and it is not surprising that our HR for cardiovascular events were lower. Our lower risk population may also have been underpowered to detect a mortality association with inter-arm systolic blood pressure difference. Post-hoc power calculations revealed that with 489/3390 deaths, 50 of which occurred in the inter-arm systolic blood pressure difference ≥10 mmHg, we only had 22% and 76% power to detect HR=1.2 and 1.5, respectively. For the association with mortality, our hazard ratio was low and not significant. Furthermore, of the deaths in our cohort, most were not cardiovascular disease related, further limiting our ability to assess an association with inter-arm systolic blood pressure difference.
Potential limitations of the present study include sequential as opposed to simultaneous blood pressure measurements2, 7, 33–35 and blood pressure measurements during one encounter as opposed to an average of separate encounters34, 36, 37. Nonetheless, in a cohort in which the prevalence of inter-arm systolic blood pressure difference >10 mmHg was 40%, the inter-arm systolic blood pressure difference was roughly stable, despite repeated measurements18 and in two studies comparing sequential and simultaneous measurements the inter-arm systolic blood pressure difference differences were well below the 10 mmHg threshold that we have used35, 38. Given that our results have shown significant and consistent prognostic value they are unlikely to represent random variation. Furthermore, sequential blood pressure measurement is the accepted method of measuring arm blood pressures in clinical practice39, making our results clinically applicable. Strengths of the present study include its large community-based sample, a comprehensive directly measured covariate assessment and validated cardiovascular events and deaths. Finally, the significant category-based net reclassification improvement index suggests statistically significant, albeit small, enhancement of incident cardiovascular event prediction beyond traditional risk factors by considering inter-arm systolic blood pressure difference.
In conclusion, in this community-based cohort, an inter-arm systolic blood pressure difference was common and was associated with a significant increase in risk for cardiovascular events even after adjustment for cardiovascular risk factors. Even modest differences in clinically-measured systolic blood pressures in the upper extremities reflect on cardiovascular risk. This study supports the potential value of identifying the inter-arm systolic blood pressure difference as a simple clinical indicator of increased cardiovascular risk. blood pressure is easily obtained in an office setting and our findings support recommendations for measurement of blood pressure in both arms both for accurate blood pressure detection and for detection of inter-arm systolic blood pressure difference. Our findings support research to expand clinical use of this simple measurement.
Acknowledgements
The Framingham Heart Study is supported by the NHLBI Framingham Heart Study (contract N01-HC-25195. Analysis for this project was supported by the NHLBI Division of Intramural Research.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of interest:
IW – None;
PG – None;
COD – None;
MRJ – Board member of VIVA physicians, a 501 c 3 non for profit education and research organization;
JMM – None
Author roles:
IW – Study design, literature search, data interpretation, writing
PG – Study design, data analysis, writing
CJD – Study design, data interpretation, editing
MRJ – Study design, data interpretation, editing
JMM – Study design, data interpretation, editing
All authors had access to the data.