Results of epidemiologic studies that investigated the significance of pulse pressure (PP) and mean arterial pressure (MAP) in terms of risk of cardiovascular disease (CVD) in patients with diabetes mellitus are inconsistent. We performed a quantitative meta-analysis to estimate CVD risk in relation to PP or MAP. Electronic literature search was conducted for prospective studies providing data on CVD risk for an increment in baseline MAP or PP in patients with diabetes mellitus. The pooled CVD risk for a 10-mm Hg increase in each blood pressure (BP) index was estimated with a random-effects model. There were 17 eligible studies consisting of 52,647 patients and 5,112 CVD cases. The pooled relative risk (95% confidence interval) of CVD for an increment of 10 mm Hg was 1.10 (1.04 to 1.16) for PP and 1.09 (0.98 to 1.21) for MAP. Significant between-study heterogeneity was observed (I 2 [p value]; 76.5% [p <0.001] for PP, 67.8% [p = 0.005] for MAP). In studies concurrently investigating CVD risk for the 4 indexes (i.e., PP, MAP, systolic BP, and diastolic BP), the pooled relative risk (95% confidence interval) was 1.17 (1.09 to 1.26) for PP, 1.11 (1.06 to 1.15) for MAP, 1.14 (1.06 to 1.22) for systolic BP, and 1.06 (0.94 to 1.19) for diastolic BP. In conclusion, the current meta-analysis suggested that PP was the strongest indicator among the 4 commonly used BP indexes. However, the large heterogeneity urged cautious interpretation of the study results.
There has been a well-established relation between blood pressure (BP), in particular systolic BP, and risk of cardiovascular disease (CVD) in patients with diabetes mellitus (DM) as well as in the general population. Recently, attention has been paid to 2 other indexes, mean arterial pressure (MAP) and pulse pressure (PP), whose components include both systolic and diastolic BP; MAP is calculated as1/3 × SBP + 2/3 × DBP, and PP is calculated as SBP − DBP, where SBP denotes systolic BP and DBP denotes diastolic BP. In the general population, compared with systolic and diastolic BP, PP has a lower predictive value for CVD whereas MAP has a comparable or greater predictive value. However, the relative magnitude among the BP indexes in terms of CVD risk is hypothesized to be specific for DM, considering that in subjects at high risk for CVD, systolic BP is higher and diastolic BP is lower (i.e., PP, but not necessarily MAP, is enlarged) in those with DM compared with those without DM. However, results of epidemiologic studies that investigated the significance of PP and MAP in terms of CVD risk in patients with DM are inconsistent. The aim of this meta-analysis is to comprehensively estimate CVD risk in relation to PP or MAP based on previously published prospective studies.
Methods
An electronic literature search using MEDLINE (from January 1, 1950 to April 2, 2013) and EMBASE (from January 1, 1974 to April 2, 2013) was conducted for studies providing data on future CVD risk in relation to baseline MAP or PP values in patients with DM. Study keywords were text words related to DM, MAP, PP, and CVD or thesaurus terms registered in MEDLINE (MeSH) or EMBASE (Emtree) related to DM (i.e., “diabetes mellitus, type 2” OR “diabetes mellitus” OR “diabetes mellitus, type 1” [in MeSH] and “insulin-dependent diabetes mellitus” OR “juvenile diabetes mellitus” OR “diabetes mellitus” OR “maturity-onset diabetes mellitus” OR “non–insulin-dependent diabetes mellitus” [in Emtree]) and CVD (i.e., “coronary disease” OR “coronary artery disease” OR “myocardial ischemia” OR “myocardial infarction” OR “cerebrovascular disorders” OR “brain ischemia” OR “stroke” OR “intracranial embolism and thrombosis” OR “intracranial hemorrhages” OR “brain infarction” OR “cerebral infarction” OR “subarachnoid hemorrhage” [in MeSH] and “acute coronary syndrome” OR “ischemic heart disease” OR “acute heart infarction” OR “heart infarction” OR “stroke” OR “brain ischemia” OR “subarachnoid hemorrhage” OR “brain hemorrhage” OR “transient ischemic attack” OR “brain ventricle hemorrhage” OR “cerebellum hemorrhage” OR “cerebrovascular disease” OR “cardiovascular disease” [in Emtree]). The 3 key concepts (i.e., DM, MAP or PP, or CVD) were combined using the Boolean operator “AND” after combining MAP and PP using the Boolean operator “OR.” Manual searches for relevant reports were added from examination of reference lists of the identified reports. No language restriction was imposed. Studies were included if (1) all participants had diabetes regardless of the type of diabetes; (2) incident CVDs were prospectively followed-up; (3) baseline values at cohort entry were presented for either PP or MAP or both PP and MAP; and (4) data on the relative risk (RR) of CVD for an increment in PP or MAP at cohort entry were provided.
The CVD end points included CVDs, coronary heart diseases (CHDs), and stroke that were symptomatic. Studies that investigated CHD or stroke apart from CVDs were included. Studies considering only fatal CVD as the study end point were also included. In studies that not only included data on risk of fatal CVD but also provided data on both fatal and nonfatal CVD, priority was given to data on risk of outcome that included both fatal and nonfatal events. If studies separately investigated fatal and nonfatal CVD, we chose the data on fatal CVD risk because it was the more serious event. Studies regarding peripheral vascular disease as a part of total CVD were also included. However, studies that mixed microvascular diseases (e.g., end-stage renal disease) and CVDs as study end points were excluded because these 2 end points involved entirely different concepts.
Two authors (SK and HS) independently extracted data, and discrepancies were solved by discussion. Data extracted from each study included the following: geographic region, type of DM (type 1, type 2, or nonspecified), definition of CVD outcomes, methods for ascertainment of CVD, mean age, gender, mean systolic BP, mean diastolic BP, proportion of patients taking antihypertensive drugs, duration of DM, follow-up periods, whether patients who already had CVD were excluded (yes or no), study covariates, and risk estimates for CVD. If the study reported multiple RRs for the same increment of BP, the most adjusted RR was used. For 1 study, in which the most adjusted RR could not be specified, we chose the RR adjusted for age, gender, and antihypertensive drugs. Study quality was assessed by modifying the Newcastle-Ottawa Quality Assessment Scale so that it was applicable to our theme. In summary, the Newcastle-Ottawa Quality Assessment Scale consists of 3 major items: S (selection, 3 questions), C (comparability, 2 questions), and O (outcome, 3 questions). For each “yes” answer, 1 point was awarded.
Data syntheses were separated by each combination of outcomes (i.e., CVD, CHD, or stroke) and by the BP indexes (i.e., PP, MAP, systolic BP, and diastolic BP). The RRs were transformed into natural logarithms—ln(RR)—and standardized into those for a 10-mm Hg increment. Each standardized ln(RR) was pooled with a random-effects model and the final RR was calculated by exponentiation of the pooled ln(RR). For PP and MAP, the data syntheses were conducted by each combination of outcomes (i.e., CVD, CHD, or stroke). Between-study heterogeneity was assessed by I 2 .
We compared the magnitude of CVD risk in relation to an increment in any 2 of the 4 BP indexes, limiting the analysis to studies that provided data on the RR for systolic BP or diastolic BP and for PP or MAP. For CVD risk in relation to PP or MAP, sensitivity analyses were added by (1) stratifying the included studies by mean age (<60 or ≥60 years) of participants and (2) meta-regression analysis, in which natural logarithms of RR for the 10-mm Hg increment were regressed on mean systolic and diastolic BP.
Publication bias was assessed by 2 formal tests: the Begg-adjusted rank correlation test and Egger’s regression asymmetry test as well as by visual inspection of a funnel plot. If publication bias was statistically suspected, we tried to adjust the risk estimates for publication bias using the trim-fill method. Data were analyzed using Stata software, version 11 (StataCorp, College Station, Texas). Two-sided p value <0.05 was considered statistically significant.
Results
Of 1,734 reports retrieved from MEDLINE and EMBASE searches, 17 eligible studies consisting of 52,647 patients with DM were included in this meta-analysis ( Figure 1 ). One study did not describe the number of cases. There were 5,112 CVD cases, 2,395 CHD cases, and 1,362 stroke cases in the remaining 16 studies. Although all but 2 studies analyzed PP, only 8 studies analyzed MAP. Two studies stratified their analyses into 2 subgroups according to age, but only 1 study also performed the analysis for the total study population.
Table 1 summarizes the characteristics of the 17 included studies. The covariates considered in each study are described in Table 2 . Only 1 study excluded patients with DM who took antihypertensive drugs, whereas, of the other 16 studies, 7 studies included taking medication for hypertension as a covariate. Excluding the covariate related to hypertension, 9 studies adjusted CVD risk for ≥6 of the 11 following potential cardiovascular risks factors that we a priori specified: age, gender, smoking, obesity indicators (body mass index, waist circumferences, or waist/hip ratio), duration of DM, systolic BP or MAP value, blood lipid values (total or low-density lipoprotein cholesterol, presence of hyperlipidemia or hyperlipidemia medication, high-density lipoprotein cholesterol, or triglycerides), nephropathy indicators (creatinine level, estimated glomerular filtration rate, urinary albumin excretion rate or presence of nephropathy), previous CVD or presence of chronic heart failure, and blood glucose level (fasting plasma glucose or hemoglobin A 1c ). Table 3 lists the results of assessment of study quality mainly based on data in Tables 1 and 2 . Subjects were considered to reflect typical patients with DM in 10 studies wherein outpatients who were receiving treatment for DM were exclusively recruited. Finally, the mean ± SD quality score was 5.1 ± 1.1 points. We judged the quality of 5 studies with a score<5 points as low.
| First Author (Year) | Country | Included Nonfatal Event | Type of DM | F/U Duration (Yrs) | Complete F/U | Mean Age (Yrs) | % of Men | DM Duration (Yrs) | No. of Patients | No. of Cases | Mean Systolic BP (mm Hg) | Mean Diastolic BP (mm Hg) | CVD (IHD) Prevalence (%) | Anti-HT Medication (%) ∗ | CVD Assessment | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CVD | CHD | Stroke | |||||||||||||||
| Nazimek-Siewniak (2002) | PL | No | 2 | 14 | No | 56 | 42 | 0 | 1,815 | 140 | 84 | 0 | 55 | M | |||
| Schram (2002) | NL | Yes | 2 | 8.6 | No | 66 | 43 | 0 | 208 | 34 | 146 | 84 | 26 | 41 | R | ||
| Schram (2003) | NL | No | 1 | 7.4 | Yes | 33 | 51 | 13 | 2,565 | 163 | 121 | 75 | 10 | 10 | M | ||
| Nakano (2004) | JP | No | 2 | 7.2 | Yes | 54 | 64 | 7.5 | 364 | 50 | 125 | 73 | 0 | 39 | M | ||
| Cockcroft (2005) | UK | No | 2 | 4.0 | Yes | 66 | 54 | 8.8 | 2,911 | 574 | 168 | 145 | 82 | ND | ND | R | |
| Nakano (2005) | JP | No | 2 | 8.3 | Yes | 46 | 69 | 6.7 | 228 | 19 | 20 | 122 | 74 | 0 | ND | M | |
| Ronnback (2006) | FI | Yes | 2 | 9.5 | Yes | 69 | 46 | 6.3 | 1,294 | 332 | 149 | 83 | 32 | 51 | R | ||
| Zoppini (2007) | IT | Yes | 2 | 10 | No | 66 | 45 | 12 | 1,128 | 146 | 68 | 42 | 155 | 87 | ND | ND | R |
| Tropeano (2008) | FR | Yes | 2 | 0.8 | Yes | 64 | 58 | 9.0 | 9,379 | 632 | 160 | 88 | 19 | 100 | M | ||
| Kengne (2009) | † | No | 2 | 4.3 | Yes | 66 | 58 | 7.0 | 11,140 | 1,000 | 559 | 433 | 145 | 81 | 32 | 69 | R |
| Nilsson (2009) | SWE | No | 2 | 5.6 | Yes | 63 | 55 | 8.0 | 11,128 | 1,728 | 1,175 | 699 | 148 | 82 | 0 | 60 | R |
| Hadaegh (2010) | IR | No | 1/2 | 8.4 | Yes | 54 | 45 | ND | 828 | 134 | 130 | 81 | 0 | 0 | S/M | ||
| van Hateren (2010) | NL | No | 2 | 9.8 | No | M | |||||||||||
| 60–75 yrs | 69 | 41 | 6.0 | 555 | 114 | 158 | 85 | 37 | 50 | ||||||||
| >75 yrs | 80 | 36 | 8.0 | 326 | 83 | 156 | 82 | 46 | 60 | ||||||||
| Gordin (2011) | FI | Yes | 1 | 5.3 | No | 39 | 52 | 22 | 4,509 | 269 | 134 | 79 | 9.0 | 38 | M | ||
| Hsieh (2012) | TW | No | 2 | 5.6 | Yes | 64 | 57 | ND | 2,161 | 25 | 135 | 78 | ND | 79 | R | ||
| Ruggenenti (2012) | IT | Yes | 2 | 9.1 | Yes | 62 | 53 | 6.2 | 1,208 | 189 | 151 | 87 | 4.3 | 55 | M/R | ||
| Theilade (2012) | DE | Yes | 1 | 8.0 | No | 44 | 57 | 28 | 900 | 213 | 139 | 79 | ND | ND | R | ||
∗ Proportion of patients taking antihypertensive drugs.
| First Author (Year) | Covariates |
|---|---|
| Nazimek-Siewniak (2002) | Gender, duration of DM, FPG, TC, TG, BMI, and (previous CVD) |
| Schram (2002) | Age, gender, MAP, and HT medication |
| Schram (2003) | Age, gender, HbA 1c , duration of DM, TC, HDL, LDL, TG, BMI, WHR, smoking, and HT medication |
| Nakano (2004) | Age, duration of DM, Cre, systolic BP, nocturnal fall in systolic BP, and (previous CVD) |
| Cockcroft (2005) | Age, gender, smoking, and TC/HDL |
| Nakano (2005) | Age, systolic BP, duration of DM, Cre, TC/HDL, nocturnal fall in systolic BP, and (previous CVD) |
| Ronnback (2006) | Age, gender, previous CVD, duration of DM, smoking, and HDL |
| Zoppini (2007) | Age, gender, duration of DM, BMI, FPG, smoking, DM medication, and PP variation |
| Tropeano (2008) | Age, gender, neuropathy, CHF, HbA 1c , smoking, WC, and HT medication |
| Kengne (2009) | Age and gender |
| Nilsson (2009) | Age, gender, duration of DM, HbA 1c , BMI, smoking, microalbuminuria, medication for HT, HL, and DM, MAP, and (previous CVD) |
| Hadaegh (2010) | Age, gender, FPG, WHR, FH of CVD, TC, smoking, aspirin, (previous CVD), and (HT medication) |
| van Hateren (2010) | Age, gender, BMI, duration of DM, Cre, TC/HDL, previous CVD, albuminuria, and medication for HL and HT |
| Gordin (2011) | Age, gender, HbA 1c , TC, eGFR, smoking, and HT medication |
| Hsieh (2012) | Age, gender, systolic BP, diastolic BP, HbA 1c , FPG, BMI, TC, TG, HDL, LDL, Cre, eGFR, and ACR |
| Ruggenenti (2012) | Age, gender, FH of CVD, smoking, BMI, HbA 1c , LDL/HDL, TG, Cre, UAE, uric acid, and medication for HL |
| Theilade (2012) | Age, gender, HbA 1c , TC, diastolic BP, smoking, previous CVD, Cre, and nephropathy |
| First Author (Year) | S1 | S2 | S3 | C1 | C2 | O1 | O2 | O3 | † Score |
|---|---|---|---|---|---|---|---|---|---|
| Nazimek-Siewniak (2002) | Yes | Yes | Yes | No | Yes | Yes | Yes | Yes | 7 |
| Schram (2002) | Yes | Yes | No | Yes | No | Yes | Yes | No | 5 |
| Schram (2003) | Yes | Yes | No | Yes | Yes | Yes | Yes | Yes | 7 |
| Nakano (2004) | No | Yes | Yes | No | No | Yes | Yes | Yes | 5 |
| Cockcroft (2005) | Yes | Yes | No | No | No | Yes | No | Yes | 4 |
| Nakano (2005) | No | Yes | Yes | No | No | Yes | Yes | Yes | 5 |
| Ronnback (2006) | No | Yes | No | No | No | Yes | Yes | Yes | 4 |
| Zoppini (2007) | Yes | Yes | No | No | Yes | Yes | Yes | No | 5 |
| Tropeano (2008) | No | Yes | No | Yes | No | Yes | No | Yes | 4 |
| Kengne (2009) | No | Yes | No | No | No | Yes | Yes | Yes | 4 |
| Nilsson (2009) | Yes | Yes | Yes | Yes | Yes | Yes | Yes | No | 7 |
| Hadaegh (2010) | Yes | Yes | Yes | Yes | Yes | No | Yes | Yes | 7 |
| van Hateren (2010) | Yes | Yes | No | Yes | Yes | Yes | Yes | No | 6 |
| Gordin (2011) | Yes | Yes | No | Yes | Yes | Yes | No | No | 5 |
| Hsieh (2012) | Yes | Yes | No | No | Yes | Yes | No | Yes | 5 |
| Ruggenenti (2012) | No | Yes | No | No | Yes | Yes | Yes | Yes | 5 |
| Theilade (2012) | No | Yes | No | No | Yes | Yes | Yes | No | 4 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
