Standardizing the Comparison of Systolic Blood Pressure vs. Pulse Pressure for Predicting Coronary Heart Disease

Abstract

Whether systolic blood pressure (SBP) or pulse pressure (PP) is more predictive of coronary heart disease remains controversial. The authors analyzed 6032 participants in the first National Health and Nutritional Examination Survey (NHANES I) followed up for an average of 16 years. Blood pressure was measured at baseline and coronary heart disease outcomes were determined from hospital or mortality records. Cox proportional hazard analyses were used to estimate the multivariate‐adjusted relative risk (RR) for increases of 10 mm Hg or 1 SD in SBP and PP, and the RR associated with PP was greater than the RR associated with SBP when using an increase of 10 mm Hg. However, when using an increase of 1 SD, the RR associated with SBP was larger than for PP. Although both are predictors, the authors conclude that SBP has a larger RR than PP for coronary heart disease events.

Whether systolic blood pressure (SBP), diastolic blood pressure (DBP), or pulse pressure (PP; defined as SBP minus DBP) is most predictive of coronary heart disease (CHD) is controversial, especially in older persons. Several epidemiologic studies have found that PP is superior to SBP for predicting CHD among the elderly, a finding that probably reflects increased arterial stiffening with age. ¹ , ² , ³ , ⁴ Other cohort studies including subjects older than 60 years have found that SBP is superior to PP for predicting CHD. ⁵ , ⁶ The largest published study to date (including more than 340,000 men) found that SBP is superior to PP for predicting CHD, but participants were no older than 57 at enrollment, ⁷ and the risk attributable to arterial stiffening is most significant in those older than 60 years. ⁸

In many of these studies, the analysis metric used to compare SBP with PP is the relative risk (RR) of a CHD event resulting from a 10‐mm Hg increase in SBP vs. a 10‐mm Hg increase in PP. We have cautioned against this analysis, since comparing the RRs of two different continuous variables (such as blood pressure [BP]) is equitable only if the amount of change is equivalent for each of the variables. ⁹ Achieving equivalent amounts of change is generally accomplished by using ±1 SD rather than absolute values, such as 10 mm Hg, to calculate RRs of continuous variables. While several studies ² , ³ , ⁴ , ⁵ , ⁶ have used either of these metrics, only one has examined both simultaneously. ¹ Furthermore, we have argued that comparing increased variance by adding each BP index to the same multivariate model more directly determines which index is more predictive of CHD. ⁹

We hypothesized that the metric used for risk estimation might affect risk relationships for CHD ascribed to the BP indices. Since the mean and SD for PP are smaller than those for SBP, using a change of 10 mm Hg compared with a change of ±1 SD for each variable would give an unfair advantage to PP. We tested whether the observed risks attributed to SBP, DBP, and PP were a function of the metric used to calculate incurred risk for CHD by calculating the respective risks using an increase of 1 SD compared with an increase of 10 mm Hg.

METHODS

The analysis presented herein included subjects in the longitudinal first National Health and Nutrition Examination Survey (NHANES I) ¹⁰ , ¹¹ , ¹² Epidemiologic Follow‐up Study (NHEFS) ¹³ , ¹⁴ , ¹⁵ , ¹⁶ who were 25–74 years old at baseline and underwent a detailed medical examination (n=6913). Four waves of follow‐up were conducted (1982–1984, 1986, 1987, and 1992). The mean follow‐up was 16 years (range, 7–22 years). Of the 6833 white and black individuals eligible for the study, 232 were unavailable for follow‐up at all four periods, 514 had a history of CHD at baseline, and 55 had unknown values for one or more of the variables assessed in the study. Thus, after all exclusions, 6032 subjects remained for analysis.

At baseline, physicians measured BP with a mercury or aneroid sphygmomanometer on the right arm of the seated participant, with the bell of the stethoscope used for auscultation. SBP was recorded at the level at which Korotkoff sounds first became audible. DBP was recorded at the level of complete cessation of Korotkoff sounds or, if there was no cessation, at the point of muffling. PP was calculated as SBP minus DBP. Measurements were recorded to the nearest 2 mm on the scale.

Hospital records and death certificates were used to identify a total of 1113 CHD cases (18%). We identified a case of CHD as a hospital discharge summary listing CHD as a discharge diagnosis (International Classification of Diseases, Ninth Revision [ICD‐9] ¹⁷ codes 410–414). The hospital admission date was used as the date of incidence. For persons with more than one record listing CHD, we used the earliest admission date. Possible bias due to loss to follow‐up or hospital record classification has been described elsewhere. ¹⁸ All death certificates were searched for any mention of these ICD‐9 codes (i.e., underlying cause or up to 20 contributing causes). For cases identified by both hospital record and death certificate, the date of incidence was taken from the hospital record. We identified 150 cases from death certificates only.

Risk‐adjusted Cox proportional hazards analyses were conducted adjusting for the following: alcohol use, exercise, age, race, sex, education, smoking status, history of diabetes mellitus, cholesterol level, body mass index, and symptoms of depression. The operational definitions of these variables have been described elsewhere. ¹⁸ The sample was divided into groups aged 25–59 and 60–74 at baseline. The RRs for the three BP measures (SBP, DBP, and PP) were first computed using 10 mm Hg and then using 1 SD. We computed χ² values by assessing increased variance explained by each BP index added to the aforementioned risk‐adjusted Cox proportional hazards model.

RESULTS

The risk‐adjusted RRs and 95% confidence intervals for each of the BP indices are presented by age group according to the metric used for risk calculation (Table). For the risk‐adjusted models, the RR of CHD associated with PP was larger than the RR associated with SBP using an increase of 10 mm Hg in both younger (1.17 vs. 1.14) and older (1.09 vs. 1.08) groups. However, when using an increase of 1 SD, the RR of CHD associated with SBP was larger than the RR associated with PP in both younger (1.32 vs. 1.24) and older (1.20 vs. 1.18) groups. Furthermore, the χ² values were somewhat larger for SBP than for PP. In general, the risk‐adjusted models did not substantially change the RRs for the BP indices compared with age‐adjusted models (data not shown).

Table.

Adjusted* Relative Risks for Coronary Heart Disease by Blood Pressure (BP) Indices and Comparison Method

	Relative Risk (95% Confidence Interval)**
	Mean (SD) by Age (yr)		Δ± 1 SD by Age (yr)		Δ± 10 mm Hg by Age (yr)		χ2 (1 df) by Age (yr)
BP Index	25–59	60–74	25–59	60–74	25–59	60–74	25–59	60–74
Systolic	127.3 (20.2)	146.3 (24.1)	1.32 (1.22–1.43)	1.20 (1.10–1.31)	1.14 (1.10–1.19)	1.08 (1.04–1.12)	49.59	16.31
Pulse pressure	44.1 (13.6)	60.3 (19.1)	1.24 (1.15–1.32)	1.18 (1.08–1.28)	1.17 (1.11–1.23)	1.09 (1.04–1.14)	36.23	13.68
Diastolic	83.2 (13.0)	86.0 (12.4)	1.20 (1.10–1.31)	1.10 (1.01–1.21)	1.14 (1.07–1.22)	1.08 (1.01–1.17)	17.28	4.25
Δ=change; df=degree of freedom; *adjusted for alcohol use, exercise, age, race, sex, education, smoking status, history of diabetes mellitus, cholesterol level, body mass index, and symptoms of depression; **p<0.05 for all values

This large, nationally representative, long‐term follow‐up study supports the finding that different results are obtained depending on the metric used to estimate risk for CHD associated with SBP or PP. Our findings are consistent with those of other published reports: studies using an increase of 10 mm Hg for comparison have found that the RR of CHD associated with PP is larger than the RR associated with SBP, ² , ³ , ⁴ while those using an increase of 1 SD have frequently found that the RR of CHD associated with SBP is larger than the RR associated with PP. ⁵ , ⁶ , ⁷ The different comparison metrics may lead to differing conclusions about which BP index is associated with a greater risk of CHD. For reasons we have described, ⁹ an increase of 1 SD is the more appropriate metric of comparison. The risk of CHD associated with increased SBP is greater than the risk associated with increased PP for predicting CHD; however, some of these RRs are only slightly larger. Based on our analysis, both SBP and PP are important independent predictors of future CHD events in both age groups.

CONCLUSIONS

The resolution of the controversy about which measure of BP best predicts CHD risk in older patients will determine future treatment. At the present time, national guidelines, such as the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7), ¹⁹ do not make any specific recommendations about PP. However, reports that tout the superiority of PP, particularly when results are based on a 10‐mm Hg increase, could lead to the premature adoption of PP as a primary target for treatment investigation. Such decisions should be based on appropriate analytic methods of comparison.

Acknowledgment and disclosure: We thank Monica Mungle for her editing assistance. Funding/support for this report was provided by grants HL04458, HL76857, HL076857, and HC25197 from the National Institutes of Health, Bethesda, MD.