Introduction

Cardiovascular disease (CVD) is the leading cause of premature morbidity and mortality [1]. High blood pressure (BP) constitutes a major cardiovascular risk factor [2, 3]. More than 7 million deaths worldwide and 1.27 million premature cardiovascular deaths in China were attributable to high BP [3, 4]. The number of hypertensive adults was predicted to increase by about 60 % to a total of 1.56 billion in 2025 all over the world [5], leading to an extremely huge global disease burden.

The relationship between BP and CVD risk is continuous. No definable threshold has been identified, down to a BP of at least 115/75 mmHg [6]. This might be one of the reasons that the concept of prehypertension was brought up in guidelines for the management of BP by the seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) [7]. The disease burden caused by prehypertension is impressive [4, 8]. Also considering the high prevalence and underdiagnosis rate of prehypertension [9, 10] and the high progression rate from prehypertension to hypertension [11, 12], effective strategies on prevention for this segment of the population would be of great value.

To identify persons at high risk of CVDs and to provide evidence for the prevention and treatment strategies of prehypertension, recognition of how risky prehypertension is for developing or dying of CVDs becomes the first step to take. There exist considerable studies showing that prehypertension is associated with stroke, myocardial infarction (MI), or coronary heart disease (CHD), but the results are inconsistent. For example, Qureshi et al. found that prehypertension was not associated with stroke [13], while a few other studies found the opposite relationship [14, 15]. It is difficult to confirm this issue in a single study, due to limited events. Therefore, we performed this meta-analysis to assess the associations between baseline prehypertension and incident stroke, CHD, MI and total CVD events at a prospective level.

Methods

Literature Search

We performed a systematic review of the published literature according to the recommendations of the Meta-analysis of Observational Studies in Epidemiology Group [16]. We conducted a comprehensive search of PubMed (1966 to June 2012) and the Cochrane Library (1988 to June 2012) without language restrictions. Search terms including MeSH words and text words were related to exposure (“prehypertensi*” or “high normal blood pressure”) and to outcomes (“cardiovascular disease”, “coronary disease”, “myocardial ischemia”, “myocardial infarction”, “coronary stenosis”, “acute coronary syndrome”, “atherosclerosis”, “ischemic heart disease”, “angina”, “stroke”, “cerebral infarction”, “intracranial hemorrhage”, “cerebrovascular disease”, “cerebrovascular attack”, “cardiovascular mortality”, “cardiovascular event” ). The literature search was undertaken by two authors (Guo X and Zhang XY) independently. Articles published in non-English language were reviewed and translated. We also manually searched the references of original and relevant reviews to ascertain additional studies. If the articles did not contain all of the necessary information, we contacted the authors for any possible additional published or unpublished data.

Inclusion and Exclusion Criteria

As described in our previous study [17], studies had to meet the following criteria for inclusion: (1) original article with prospective cohort design; (2) they assessed prehypertension or high normal BP as baseline exposure; (3) they assessed fatal or non-fatal incident stroke, CHD, MI, or total CVD events as outcome; (4) median follow-up of at least 3 years; and (5) they reported multivariate-adjusted risk ratio (RR) or hazard ratio (HR) and 95 % confidence interval (95 % CI) between exposure and outcomes with normal BP as reference. Multiple samples with different gender, age or ethnic groups from the same population were also included. If we identified multiple reports from the same study, the one with the most detailed information was adopted.

Studies were excluded if they met one of the following criteria: (1) no original data, such as reviews or comments; (2) only age-adjusted and gender-adjusted or unadjusted RR or HR was reported; (3) duplicated studies; (4) not conducted in human; and (5) data were derived from secondary analyses of clinical trials.

Data Extraction

Two investigators (Guo L and Li Z) extracted the data independently, with discrepancies resolved by an additional reviewer (Zheng) and through discussion. A standardized data extraction form was used. Information extracted included first author’s name, publication year, country, sample characteristics, prevalence of prehypertension, follow-up, definition of high BP, adjusted variables, outcome assessment, and multivariate-adjusted RRs or HRs and corresponding 95 % CIs. An electronic abstraction database was created in Microsoft Excel.

Assessment of Study Quality

As described previously [17], we assessed quality of all articles that met the selection criteria with the following eight items according to the guidelines developed by the US Preventive Task Force and the modified checklist [18–20]: (1) prospective study design; (2) maintenance of comparable groups; (3) adjustment of potential confounders; (4) documented loss of follow-up rate; (5) outcome assessed blind to exposure status; (6) clear and proper definition of exposures (prehypertension) and outcomes (stroke, CHD, MI and total CVD events); (7) temporality (BP measured at baseline, not at time of outcomes assessment) and (8) follow-up of at least 1 year. Studies were graded as good quality if they met seven to eight criteria; fair if they met four to six; and poor if they met less than four criteria.

Statistical Analysis

We obtained pooled estimates basing on the multivariate-adjusted RRs or HRs with 95 % CIs from included studies in order to estimate the quantitative association between prehypertension and the CVD outcomes. Between-study heterogeneity was tested by Q-statistic and quantified by the I2 statistic. I2 statistic of 0–40 % indicates unimportant heterogeneity, 30–60 % indicates moderate heterogeneity, 50–90 % indicates substantial heterogeneity, and 75–100 % indicates considerable heterogeneity [21]. If there existed statistically significant heterogeneity (P < 0.1 and I2 > 50 %), we chose a random-effects model, otherwise, a fixed-effects model was used.

Prehypertension was defined as systolic blood pressure (SBP) at 120–139 mmHg or diastolic blood pressure (DBP) at 80–89 mmHg. We further divided prehypertension into two BP ranges (i.e. low range: SBP of 120–129 mmHg or DBP of 80–84 mmHg; and high range: SBP of 130–139 mmHg or DBP of 85–89 mmHg), with normal BP (SBP < 120 mmHg and DBP < 80 mmHg) the reference category. The outcome assessment was the relative risk of fatal or non-fatal incident stroke, CHD, MI and total CVD in the whole prehypertensive range or in low-range and high-range prehypertension, respectively. Subgroup analyses were performed according to average age (<65 years vs. ≥65 years), gender (men vs. women), location (Asian vs. non-Asian), sample size (<10000 vs. ≥10000), follow-up (<10 years vs. ≥10 years) and study quality (good vs. fair).

Possible publication bias was evaluated visually by funnel plots and statistically by Begg’s and Egger’s tests. We also evaluated the influence of individual studies by sensitivity analysis to see the extent to which inferences depend on a particular study or group of studies. All analyses were performed using statistical package Stata version 11.0 and values of P < 0.05 were considered to be statistical significant.

Results

Literature Search and Study Characteristics

The initial database search generated 1,654 papers, of which 1,587 were excluded after review of title and abstract. Among the retrieved 67 articles, 29 articles met our inclusion criteria, with 1,010,858 participants [13–15, 22–47]. Figure 1 provides a diagram of the selection process and reasons for exclusion. The included studies varied in sample size from 566 [24] to 347,978 [29]. All but three of the studies [14, 29, 41] included both men and women. Eleven of the studies were conducted in the United States, four in China, eight in Japan, and one each in Germany, India, Iran, Korea, Singapore and Turkey. Follow-up ranged from 6.9 to 25 years. Most of included participants were free of CVDs and were not on use of antihypertensive medication. Table 1 summarizes the characteristics of the included studies.

Fig. 1
figure 1

Flow chart of the study selection process

Full size image
Table 1 Characteristics of prospective studies included in the systematic review and meta-analysis
Full size table

Prehypertension and Total Cardiovascular Disease

In the pooled analysis from twenty-one populations, both low-range and high-range prehypertension were associated with a greater risk of developing or dying of total CVD (low-range: RR: 1.24; 95 % CI: 1.10 to 1.39, P < 0.001; high range: RR: 1.56; 95 % CI: 1.36 to 1.78, P < 0.001) (Fig. 2a, b). The risk of total CVD was increased among the whole range of prehypertensive populations (RR: 1.44; 95 % CI: 1.35 to 1.53, P < 0.001) (Fig. 3a).

Fig. 2
figure 2

Association between two ranges of prehypertension, low range (a) and high range (b), and the risk of developing or dying of total cardiovascular disease. Low range prehypertension: 120–129/80–84 mmHg; high range prehypertension: 130–139/85–89 mmHg. CI confidence interval

Full size image
Fig. 3
figure 3

Association between prehypertension and the risk of developing or dying of total cardiovascular disease (a), stroke (b), coronary heart disease (c), and myocardial infarction (d). CI confidence interval

Full size image

Prehypertension and Stroke

Eight studies with eleven populations and nine studies with twelve populations evaluated the risk of low-range and high-range prehypertension for stroke, respectively. Both two ranges increased the risk of developing or dying of stroke (low-range: RR: 1.35; 95 % CI: 1.10 to 1.66, P = 0.004; high range: RR: 1.95; 95 % CI: 1.69 to 2.24, P < 0.001) (Fig. 4a, b). Seven studies investigated the association between the whole range of prehypertension and stroke, the pooled result of which showed that prehypertension was related to a greater risk of developing or dying of stroke (RR: 1.73; 95 % CI: 1.61 to 1.85, P < 0.001), with no evidence of heterogeneity (I2 = 0 %, P = 0.918) (Fig. 3b).

Fig. 4
figure 4

Association between two ranges of prehypertension, low range (a) and high range (b), and the risk of developing or dying of stroke. Low range prehypertension: 120–129/80–84 mmHg; high range prehypertension: 130–139/85–89 mmHg. CI confidence interval

Full size image

Prehypertension and Coronary Heart Disease

Seven studies with nine populations distinguished the two ranges of prehypertension. The pooled result showed that neither low-range nor high-range prehypertension was significantly associated with an increased risk of developing or dying of CHD (low-range: RR: 1.11; 95 % CI: 0.87 to 1.42, P = 0.39; high range: RR: 1.33; 95 % CI: 0.96 to 1.83, P = 0.085) (Fig. 5a, b). However, in the pooled analysis of four studies evaluating the whole range, prehypertension increased the risk of CHD (RR: 1.34; 95 % CI: 1.20 to 1.50, P < 0.001) (Fig. 3c).

Fig. 5
figure 5

Association between two ranges of prehypertension, low range (a and c) and high range (b and d), and the risk of developing or dying of coronary heart disease (a and b), and myocardial infarction (c and d). Low range prehypertension: 120–129/80–84 mmHg; high range prehypertension: 130–139/85–89 mmHg. CI confidence interval

Full size image

Prehypertension and Myocardial Infarction

In the pooled analysis from three populations, low-range prehypertension increased 1.43-fold risk of developing or dying of MI (P = 0.007), while high-range prehypertension was associated with a much higher risk (RR: 1.99; 95 % CI: 1.59 to 2.50, P < 0.001) (Fig. 5c, d). Among the whole range prehypertensive populations, risk of incident MI was also increased (RR: 1.79; 95 % CI: 1.45 to 2.22, P < 0.001) (Fig. 3d).

Sources of Heterogeneity

Tables 2 and 3 show further analysis stratified by different population groups in each range of prehypertension. Between-study heterogeneity was observed. The heterogeneity of effect was due to differences in gender, age, location, sample size, follow-up or study quality. No publication bias was observed (Begg’s test all P > 0.05; Egger’s test all P > 0.05, figures not shown). The sensitivity analysis showed that the omission of any of the studies from the analysis did not alter the overall finding.

Table 2 Subgroup analyses to explore source of heterogeneity in the low range prehypertension
Full size table
Table 3 Subgroup analyses to explore source of heterogeneity in the high range prehypertension
Full size table

Discussion

The present study provided a comprehensive review of the literature worldwide and quantitative estimates of prospective associations between prehypertension and CVDs among population-based studies. From multivariable adjusted studies, we found that prehypertension was associated with a clearly increased risk of incident stroke, MI and total CVD events, even within the lower range. The effects of prehypertension on CVD outcomes differed by many factors, such as gender, age group and study quality.

The overall prevalence of prehypertension in the US was 31 % according to the Third National Health and Nutrition Examination Survey (NHANES III) [48]. Over 32 million men and 21 million women aged 20 years or older in the US were estimated to be prehypertensive [49]. Given the large population and the robust impact of prehypertension on cardiovascular outcomes, the caused burden is considerable. It was estimated that in China alone, there were 0.22 million cardiovascular deaths in adults attributed to prehypertension in 2005 [4]. In the present study, we found that the baseline prehypertensive population had a 1.44-fold risk of developing or dying of total CVD. Even within the lower range group, the risk was increased, indicating that effective intervention in this early stage might reduce a substantial future burden. A recent meta-analysis including 18 studies reported a higher RR of 1.55 for CVD among prehypertensive participants [50]. This higher risk might be explained by the fact that some studies evaluating fatal events as outcomes were not included in that analysis. Although a secondary analysis of a clinical trial, which was not included in the present study, found that prehypertension was not associated with an increased risk of fatal CVD [51], the impact of drug use was not fully eliminated. In line with a former study in the Asia-Pacific region [52], we observed that the effect of prehypertension on total CVD vanished among participants aged 65 years or older.

Stroke is most highly correlated with BP among various vascular diseases [53]. The present study found that prehypertensive participants increased about 1.7-fold in risk of developing or dying of stroke, which was similar to the result from a previous study [54]. The RR for incident stroke was also quite similar in the analysis by Huang et al. [50], indicating a robust association between prehypertension and stroke. A recent meta-analysis of 16 randomized controlled trials indicated that antihypertensive therapy among prehypertensives could significantly reduce the risk of stroke [55], leading to a doubt that whether only lifestyle modification recommended by JNC 7 [7] is adequate. In our further subgroup analyses, we observed that among the low-range prehypertensive population, only men with prehypertension had a significantly increased risk for stroke, while the gender-specific result disappeared in the high-range population. The underlying mechanism is not understood. We also found that prehypertension was not associated with stroke among the elderly, the potential reasons of which were discussed by Lee et al. [54].

Also, we observed a 1.34-fold risk of CHD associated with prehypertension from four studies, which was a little higher than the result in the Asia-Pacific region [52] and lower than the result from Huang et al. [50]. While in the analyses of two separate range groups, the results were not statistically significant, even in the high-range prehypertension. This discrepancy is probably due to different distributions of the included populations. A novel finding is that high range prehypertensive population had a nearly two-fold risk of developing or dying of MI, meriting attentions in the clinical work. However, the number of included studies is relatively small, and more cohorts are expected. Of note, Butler et al. found that among the elderly, prehypertension was related to a 1.63-fold risk of incident heart failure over 10 years follow-up in the US [56]. The positive association was consistent with a previous study [14], indicating a roll of mildly raised BP in the mechanism of heart failure.

The strengths of our study include the comprehensive review of the literature worldwide and the large sample size we collected. There are limitations in the present study that merit discussion. First, the contributing studies varied in the degree of confounders. Although we only included multivariable adjusted studies to minimize the impacts, it remains a possibility that residual confounding and bias across the studies caused overestimation of the associations. Second, due to the limited number of studies, we pooled results from studies involving different definitions of CVD, which might compromise the results. In addition, although our literature search was extensive, there still was a possibility of omissions. Plus, a delay between search and publication was inevitable. Even though our meta-analysis has several limitations, it probably represents the most comprehensive review and the most accurate estimate to critically appraise the evidence surrounding the association between prehypertension and different types of CVD.

Conclusions

From the worldwide data, prehypertension was significantly associated with an increased risk of incident stroke, MI and total CVD. The impact was markedly different between the two BP ranges. Effective BP reduction treatment giving full consideration of the risk stratifications of different BP ranges should be initiated in this early stage to reduce a substantial future burden. Trials investigating the effects of BP reduction on different types of CVD outcomes among prehypertensive subjects are expected.