Do Nonsteroidal Anti-inflammatory Drugs Affect Blood Pressure? A Meta-Analysis

  1. Anthony G. Johnson, MBBS;
  2. Tuan V. Nguyen, MApplStat; and
  3. Richard O. Day, MD
  1. From St. Vincents Hospital and the University of New South Wales, Sydney, Australia. Requests for Reprints: Anthony G. Johnson, MBBS, FRACP, Department of Clinical Pharmacology, Princess Alexandra Hospital, Woolloongabba 4102 Qld Australia. Acknowledgments: The authors thank Dr. Robert Cummings for his comments and the Clinical Trials Coordinator, Suzanne Duffy, for her assistance with the blinding process. Grant Support: Dr. Johnson was a National Health & Medical Research Council of Australia scholar.
     
    Next Section

    Abstract

    Purpose: A meta-analysis of randomized trials studying the effect of nonsteroidal anti-inflammatory drugs (NSAIDs) on blood pressure.

    Data Sources and Study Selection: Eight databases were searched, yielding 38 randomized, placebo-controlled trials and 12 randomized but not placebo-controlled trials (comparing two or more NSAIDs).

    Data Extraction: Pooled mean treatment effects were computed in each trial for blood pressure, weight, creatinine clearance, plasma renin activity, and daily urinary excretion of sodium and prostaglandins. Meta-analyses of these variables were done for all randomized, controlled trials; for all randomized, uncontrolled trials; and for several subgroups.

    Data Synthesis: When pooled, NSAIDs elevated supine mean blood pressure by 5.0 mm Hg (95% CI, 1.2 to 8.7 mm Hg) but had no effect on variables other than blood pressure. Nonsteroidal anti-inflammatory drugs antagonized the antihypertensive effect of β-blockers (blood pressure elevation, 6.2 mm Hg; CI, 1.1 to 11.4 mm Hg) more than did vasodilators and diuretics. Among NSAIDs, piroxicam produced the most marked elevation in blood pressure (6.2 mm Hg; CI, 0.8 to 11.5 mm Hg), whereas sulindac and aspirin had the least hypertensive effect.

    Conclusions: Nonsteroidal anti-inflammatory drugs may elevate blood pressure and antagonize the blood pressure-lowering effect of antihypertensive medication to an extent that may potentially increase hypertension-related morbidity. Although certain NSAIDs and antihypertensive agents could be more likely to produce these effects, the underlying mechanisms require further study.

    Hypertension is prevalent [1, 2] and is a major determinant for stroke and coronary heart disease [3-5]. Further, the higher the blood pressure, the more marked is the reduction in life expectancy [6]. Nonsteroidal anti-inflammatory drugs (NSAIDs) are the most commonly prescribed drugs worldwide when grouped by generic categories [7], accounting for 4% to 9% of all prescriptions in developed countries [8-12]. Since 1967, their use has increased steadily, particularly among elderly persons [13].

    Certain NSAIDs have been reported in randomized, controlled trials to elevate blood pressure in some previously normotensive persons with [14] and without antihypertensive therapy [15-20], in patients with mild hypertension untreated [19, 21, 22] or treated with single doses of antihypertensive agents [23, 24], and in hypertensive persons whose blood pressure had been controlled by drug therapy [21, 25-39]. However, several well-designed studies have failed to show any effect of NSAIDs on blood pressure [40-51]. In addition, many randomized studies [52-63] have attempted to compare the effect of various NSAIDs on blood pressure control with widely differing results that have been thus far inconclusive. For most of the individual trials, the estimates may have been imprecise because the sample size was insufficient.

    In view of the anticipated increase in the prevalence of hypertension [64] and the substantial use of NSAIDs, any potential drug–drug (NSAID-antihypertensive agent) and drug-disease (NSAID-hypertension) interactions should be fully clarified. Consequently, we did a meta-analysis of randomized trials. Our primary goal was to produce a stable estimate of the overall effect of various NSAIDs on blood pressure; our secondary aims were to evaluate possible mechanisms by which NSAID therapy may alter blood pressure and to determine potential predisposing factors for this interaction.

    It has been proposed that NSAIDs may alter blood pressure through the effects of prostaglandin synthesis inhibition on body weight, cardiac output, or renal function. Consequently, where provided, we have included these data in the meta-analysis. Although NSAIDs have been associated with blood pressure elevation in normotensive persons and in both treated and untreated hypertensive persons, data pertaining to studies including persons in these subgroups were also analyzed separately to determine whether differential effects occurred in various population groups. Similarly, it was uncertain whether NSAIDs interacted with antihypertensive agents from different classes in the same manner and whether different NSAIDs altered blood pressure to the same degree. Consequently, we also studied these subgroups separately.

    Previous SectionNext Section

    Methods

    We did a MEDLINE search from 1966 to 1990. We also did the following additional searches: Embase (1974 to 1990), Biosis (1969 to 1990), Diogenes (1976 to 1990), Science Citation Abstracts (1972 to 1990), International Pharmaceutical Abstracts (1970 to 1990), IOWA Drug Information Service (1966 to 1990), and the Combined Health Information Database (1973 to 1990). For each search, key words relating to trial design (meta-analysis, research design, double-blind method, double-blind study, double, blind, random allocation, random, control, clinical trials) were crossed with the names of individual NSAIDs and the following terms: anti-inflammatory agents, nonsteroidal, hypertension, blood pressure.

    Articles reviewing the potential interaction between NSAIDs and blood pressure were selected from these searches, and their bibliographies were carefully checked for randomized trials not published elsewhere. We also reviewed textbooks on hypertension, clinical pharmacology, and NSAIDs, and checked reference lists of all randomized trials identified by any of the above means.

    We selected 194 articles as potentially fulfilling the entry criteria. Only 60 articles, however, described random allocation of one or more NSAIDs and measurement of the effect of NSAIDs on blood pressure. Of the 60 articles, 38 included randomized and controlled trials [14-51] (Appendix Tables 1 and 2). Twelve more articles [52-63] were reports of randomized but not placebo-controlled studies; in these studies, two or more NSAIDs were compared with each other rather than with placebo in terms of their effects on blood pressure control.

    The remaining 10 articles [65-74] could not be included: eight articles [65-72] because they did not provide measures of variance for the blood pressure data such as standard error or standard deviation, which was needed for our analysis; one article [73] because no blood pressure data were provided; and one article [74] because blood pressure measurement was not the primary goal of the study and because uncontrolled administration of morphine in that trial probably confounded blood pressure levels.

    The eight articles that did not provide variance data for the blood pressure measurements [65-72] and the article in which blood pressure data were not provided [73] were similar to the 50 articles included in the meta-analysis with respect to patient characteristics, relevant diagnoses, and treatment received. In addition, like the included trials, reporting of an effect of NSAIDs on blood pressure was inconsistent among these articles. However, many were completed several years ago, reducing the possibility of obtaining accurate, reliable data from the original authors; most were abstracts [66, 67, 69, 70, 73], and none had a large sample size with adequate follow-up. Consequently, we considered it appropriate to exclude these studies.

    Control and Measurement of Potential Bias

    To minimize selection bias, the Clinical Trials Coordinator coded the methods and results of articles that potentially fulfilled the entry criteria for the meta-analysis. The relevant methods and results sections were photocopied after any identifying information was blacked out. The decision to accept or reject articles was made by two authors, based on the methods alone without any knowledge of the source of the articles (authors or institution) and achieved by consensus.

    The same two authors subsequently developed an algorithm for assessing the methodologic rigor and scientific quality of the articles to be pooled and for calculating quality assessment scores (0 to 100 [the higher the score, the better the quality of the study assessed]), using only the methods and results sections. This algorithm was broadly based on a systematic evaluation of study quality published by Chalmers and colleagues [75]. Quality assessment scores by the two assessors highly correlated (r = 0.9), suggesting close interobserver agreement. In an attempt to control data-extraction bias, relevant data were drawn from papers by the assessors, who were blinded to all but the results sections, and in 89% of studies evaluated agreement in the data extracted was complete.

    One author provided a quality assessment score for all trials and extracted data from every trial included in the meta-analysis, whereas the other author reviewed a 20% random selection for quality assessment scoring and another 20% random selection for extraction of data.

    Appendix Tables 1 and 2 present the relevant features of all randomized, controlled trials with a placebo group. Table 1 provides details of patients at trial entry (age, sex, race) and key study features (sample size, percent dropouts, design, quality assessment score, geographic location, and source of support). Appendix Table 4 outlines the treatment received by patients in each trial (including diet, NSAID type and duration of use, and antihypertensive type and duration), and the average difference (in mm Hg) between mean blood pressure recorded on NSAID treatment and control treatment, respectively, adjusted for baseline mean blood pressure ([NSAID treatment mean blood pressure -baseline mean blood pressure] −[control treatment mean blood pressure − baseline mean blood pressure]). Appendix Tables 3 and 4 present similar information for randomized, controlled trials without a placebo group. Because only two trials [41, 51] used ambulatory blood pressure monitoring (both with same small sample sizes: n = 12) and all other trials assessed blood pressure using standardized mercury sphygmomanometers, the method of blood pressure measurement was not considered in the analysis.

    View this table:
    Table 1. Meta-analysis of Randomized, Placebo-Controlled Trials
    View this table:
    Appendix Table 1. Study Features of Randomized, Controlled Trials

    *Table OMITTED

    View this table:
    Appendix Table 2. Treatment Received in Randomized, Controlled Trials

    *Table OMITTED

    View this table:
    Appendix Table 3. Study Features of Randomized Trials without a Placebo Group

    *Table OMITTED

    View this table:
    Appendix Table 4. Treatment Received in Randomized Trials without a Placebo Group

    *Table OMITTED

    Combining the Trials

    The randomized, controlled trials assessing the effect of NSAIDs on blood pressure [14-51] may be readily subdivided into two broad categories (Appendix Table 3): those of predominantly white, middle-aged patients of both sexes with mild to moderate uncomplicated essential hypertension [29 trials] and those of predominantly white, healthy, young volunteers of both sexes with no history of hypertension and with normal blood pressure (15 trials). These broad categories may be further subdivided (Appendix Tables 1 and 2) into three groups of trials in hypertensive patients and two groups of trials involving normal volunteers based on whether or not antihypertensive therapy was coadministered and on its duration. Because all these trials were randomized and controlled, measured the effect of NSAIDs on blood pressure in patients with uncomplicated mild to moderate essential hypertension or healthy volunteers, and had no other apparent systematic differences between the trials in each of the five groups, we considered it clinically appropriate to pool their results.

    To assess whether these trials could be combined statistically, the randomized, controlled trials were subdivided into their five population groups (Appendix Tables 1 and 2): Analysis of variance showed no significant difference among any of the groups in terms of their pooled effect of NSAIDs on blood pressure (P > 0.2), suggesting that these groups were not statistically heterogenous and could be pooled.

    Similarly, analysis of variance showed no significant difference between trials grouped according to race (P = 0.17), duration of NSAID use (1 week or less compared with more than 1 week, P = 0.09; 4 weeks or less compared with more than 4 weeks, P = 0.09), antihypertensive type (P = 0.07) or duration (P > 0.2), diet (P > 0.2), study design (P > 0.2), quality assessment score (P = 0.19), location (P > 0.2), patients who dropped out after randomization (P > 0.2), and activity level (P > 0.2) in terms of their pooled NSAID effects on blood pressure.

    However, when trials were grouped according to NSAID type, analysis of variance showed marked differences in the pooled effects of different NSAIDs on blood pressure (P = 0.01). Overall, given that all the trials were randomized and controlled and that the pooled effect of NSAIDs as a group on blood pressure was consistent across a wide range of different analyses, we decided that these trials were sufficiently homogeneous to be combined in the meta-analysis.

    The 12 studies that were randomized but not placebo-controlled [52-63] could not be meaningfully combined with the placebo-controlled trials [14-51] because the compared effects differed. They were thus pooled separately.

    Statistical Analysis

    For each of the 50 articles included in the analysis, we recorded outcome data for all randomly assigned persons (intention-to-treat method). The mean difference between NSAID treatment and placebo (or control) treatment was calculated together with the standard error of the difference for each trial [76]. We then determined the pooled mean treatment effect using the standardized mean treatment effect weighted by the inverse of the variance of the difference for each trial [77, 78]. Using these data also allowed us to compute the pooled standard error and the corresponding 95% confidence intervals (CIs) [78].

    The standard error of the difference in crossover trials was calculated as if a parallel design had been used because individual participant data necessary to calculate the degree of covariance were not available for each trial [79]. Consequently, the true variance for crossover trials may have been overestimated, and the 95% CIs may therefore be expected to be spuriously wide, increasing the chance of a type II error.

    Each meta-analysis computed pooled mean treatment effects with accompanying 95% CIs for the following variables: blood pressure, weight, creatinine clearance, plasma renin activity and daily urinary sodium, and prostaglandin E2 and 6 keto-prostaglandin F1 α output. When mean blood pressure was not provided, it was calculated from the systolic (SBP) and diastolic blood pressures (DBP) using the following equation: mean blood pressure = 1/3(2 x DBP + SBP).

    Meta-analyses of the variables noted above were computed for all randomized, placebo-controlled trials, for all randomized trials without a placebo group, and according to the following subgroups: population type (five groups, Appendix Tables 1 and 2), NSAID type, and antihypertensive type (β-blockers, diuretics, or vasodilators). The analyses were done using SAS software (SAS Institute Inc; Cary, North Carolina).

    Because the results may vary depending on the overall quality of the primary trials and on whether, based on important criteria, particular trials have been included or excluded from the analysis [80], the subgroups outlined above were analyzed to determine if the effects of NSAIDs on blood pressure were consistent across a wide range of trial combinations.

    Previous SectionNext Section

    Results

    The meta-analysis included 771 volunteers or patients aged 47.6 ±2.5 years (mean ±SE). In the trials providing stratification by sex, overall slightly more men were studied, with a male:female ratio of 1.3:1; only 9% of the trials analyzed included blacks in their study cohort. Most persons or patients were studied primarily as outpatients (88%) and at rest (94%). Several trials [18, 30, 32, 37, 38, 43, 49] contained data on more than one NSAID, which were pooled for each trial to avoid using data from the same volunteers or patients more than once in the same analysis.

    Of the 38 articles (50 independent trials), which were randomized and placebo-controlled, 37 trials had a crossover design and 13 had a parallel design with the mean sample size per trial of 16 ±2 (Appendix Tables 1 and 2). The quality assessment score was 49.7 ±1.8, and most trials (95%) reported few (≤ 20%) dropouts. Most trials were done in English-speaking countries (72%), and 40% of the studies were funded at least partially by the pharmaceutical industry. For 62% of trials, the diet was either not specified or stated to be unrestricted. Although many NSAIDs were used (n = 9), indomethacin was administered in more than half of all trials and for most trials (66%) the duration of NSAID therapy was 1 week or longer. Many antihypertensive agents were administered as part of several trials, generally for 1 week or longer (76% of trials in which antihypertensive therapy was given).

    All 12 articles (16 independent trials) that were reports of randomized but not placebo-controlled trials used a crossover design with a mean sample size of 15 ±3 per study (Appendix Tables 3 and 4). The quality assessment score was 45.2 ±3.0 and as above, for most trials (94%), dropouts were few. In contrast to the randomized, controlled trials, however, most randomized, uncontrolled studies were done in non-English speaking countries (81%), and a single pharmaceutical company (Merck, Sharp and Dohme; Rahway, New Jersey) supported 44% of them. The diet was classified as unrestricted in 81% of trials and in all but one study (94%), sulindac and another NSAID, predominantly indomethacin (67%), were compared. The duration of NSAID use was 1 week or longer in 88% of studies. For these trials, different antihypertensive drugs were used, and the duration was 1 week or longer.

    When the outcome measurements described above were pooled for all the randomized, placebo-controlled trials, NSAIDs elevated supine mean blood pressure by 5.0 mm Hg (CI, 1.2 to 8.7 mm Hg) but had no significant effect on body weight, daily urinary sodium output, or creatinine clearance or urinary prostaglandin excretion per 24 hours (Table 1). Similarly, analyses according to the subgroups defined earlier also showed no significant effects of NSAIDs on the non-blood pressure variables assessed.

    When the randomized, placebo-controlled trials were classified based on population type (Figure 1), NSAIDs increased supine mean blood pressure in all subgroups, but only the trials involving controlled hypertensive persons achieved statistical significance with pooling (weighted mean increment, 5.4 mm Hg; CI, 1.2 to 9.6 mm Hg). The pooled increment in mean blood pressure with NSAID administration was slightly greater for those studies involving untreated hypertensive persons than those studies that included normotensive persons only (2.5 compared with 1.1 mm Hg, respectively) (Figure 1). Those studies in which antihypertensive therapy was administered were associated with a greater increase in blood pressure after NSAID treatment than those studies in which no antihypertensive agents were given (4.7 compared with 1.8 mm Hg, respectively) (Figure 1).

    Figure 1. N = total number of volunteers or patients; trials = total number of trials; HTN-Antihyp stat = patients with hypertension administered a single dose of antihypertensive medication; regular = patients with controlled hypertension; HTN-Uncontrolled = patients with uncontrolled hypertension; Normal-Antihyp = normotensive volunteers administered antihypertensive medication; Normal = normotensive volunteers with no history of hypertension. Dots represent mean difference in mm Hg between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure adjusted for baseline. The dashes identify the line of no difference between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure. Error bars show 95% CIs.
    View larger version:
    Figure 1. N = total number of volunteers or patients; trials = total number of trials; HTN-Antihyp stat = patients with hypertension administered a single dose of antihypertensive medication; regular = patients with controlled hypertension; HTN-Uncontrolled = patients with uncontrolled hypertension; Normal-Antihyp = normotensive volunteers administered antihypertensive medication; Normal = normotensive volunteers with no history of hypertension. Dots represent mean difference in mm Hg between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure adjusted for baseline. The dashes identify the line of no difference between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure. Error bars show 95% CIs. Meta-analysis of randomized, placebo-controlled trials according to population type.

    Among the randomized, placebo-controlled trials in which antihypertensive therapy was administered, NSAIDs antagonized the effect of all drug categories (Figure 2). However, when β-blockers and vasodilators were administered, NSAIDs produced a substantially greater increase in supine mean blood pressure than was produced in trials in which diuretics were administered. Despite these findings, only the pooled antagonistic effect of NSAIDs on β-blockers in trials in which both were administered achieved statistical significance with NSAID therapy, resulting in a weighted mean increase of 6.2 mm Hg (CI, 1.0 to 11.4 mm Hg).

    Figure 2. N = total number of volunteers or patients; trials = total number of trials. Dots represent mean difference in mm Hg between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure adjusted for baseline in patients taking different antihypertensive agents. The dashes identify the line of no difference between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure. Error bars show 95% CIs.
    View larger version:
    Figure 2. N = total number of volunteers or patients; trials = total number of trials. Dots represent mean difference in mm Hg between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure adjusted for baseline in patients taking different antihypertensive agents. The dashes identify the line of no difference between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure. Error bars show 95% CIs. Meta-analysis of randomized, placebo-controlled trials according to antihypertensive type.

    A meta-analysis of randomized, placebo-controlled trials according to NSAID type (Figure 3) showed that all NSAIDs increased supine mean blood pressure with piroxicam, indomethacin, and ibuprofen producing the most marked increases. However, only piroxicam had a statistically significant effect on supine mean blood pressure (6.2 mm Hg; CI, 0.8 to 11.5 mm Hg). Aspirin, sulindac, and flurbiprofen caused the smallest elevation in supine mean blood pressure, although the 95% CIs were wide for the latter two NSAIDs.

    Figure 3. N = total number of volunteers or patients; trials = total number of trials. Dots represent mean difference in mm Hg between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure adjusted for baseline. The dashes identify the line of no difference between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure. Error bars show 95% CIs.
    View larger version:
    Figure 3. N = total number of volunteers or patients; trials = total number of trials. Dots represent mean difference in mm Hg between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure adjusted for baseline. The dashes identify the line of no difference between the effects of the nonsteroidal anti-inflammatory drug and placebo on blood pressure. Error bars show 95% CIs. Meta-analysis of randomized, placebo-controlled trials according to nonsteroidal anti-inflammatory drug type.

    In the meta-analysis of randomized trials without a placebo group (Table 2), none of the pooled comparisons between NSAIDs attained statistical significance in terms of their effect on supine mean blood pressure. Combining trials that compared indomethacin with sulindac resulted in a weighted mean increase in supine mean blood pressure on indomethacin compared with sulindac treatment of 6.0 mm Hg (CI, −7.7 to 19.7 mm Hg). Indomethacin therapy similarly increased supine mean blood pressure by 5.7 mm Hg (weighted mean) more than imidazole salicylate treatment, but this was based on a single trial and the CI was wide (CI, −34.7 to 46.1 mm Hg). Although diclofenac, ibuprofen, and naproxen did not differ substantially in their effect on supine mean blood pressure from sulindac, these analyses were also based on few trials and the CIs were wide.

    View this table:
    Table 2. Meta-analysis of Randomized Trials without a Placebo Group
    Previous SectionNext Section

    Discussion

    Our results suggest that NSAIDs elevate mean blood pressure by approximately 5.0 mm Hg, at least over a period of several weeks. How clinically important is an increase in blood pressure of 5.0 mm Hg? An overview of randomized trials of antihypertensive treatment [81] found that a 5 to 6 mm Hg increase in diastolic blood pressure over a few years may be associated with a 67% increase in total stroke occurrence and a 15% increase in coronary heart disease. The latter may be higher if the difference in diastolic blood pressure is maintained longer (approximately 10 years) [82]. Also, these effects are apparent in both normotensive persons and hypertensive persons. Whether these results apply to patients using NSAIDs is not known because these studies included patients not taking NSAID medication.

    The hypertensive effect of NSAID administration was most marked in hypertensive patients taking antihypertensive medication and was less marked in normotensive volunteers administered antihypertensive drugs. The hypertensive effect of NSAID administration was reduced further in patients with uncontrolled hypertension but was least in normotensive volunteers not taking blood pressure-lowering medication.

    Increased plasma concentrations of prostaglandin E2 have been reported in patients with borderline and sustained essential hypertension [83]. Consequently, as detected in this meta-analysis, inhibition of prostaglandins by NSAIDs may be expected to produce a greater blood pressure increase in untreated hypertensive patients than in normotensive persons. However, the difference between the NSAID pressor effect on the blood pressure of untreated hypertensive and normotensive persons was small (1.4 mm Hg). The difference was not statistically significant, and its clinical importance requires clarification.

    Although certain antihypertensive agents may act through natriuretic and diuretic prostaglandins for their blood-pressure-lowering effect [84], it is more likely that vasodilatory prostaglandins have a counter-regulatory role, synthesized in response to reflex-mediated pressor mechanisms (such as sympathetic tone and angiotensin) that follow antihypertensive therapy [85]. Hence, hypertensive patients taking blood pressure-lowering medication would be expected to have the highest plasma levels of vasodilatory prostaglandins so that prostaglandin inhibition by NSAIDs in hypertensive patients taking blood pressure-lowering medication may be associated with the greatest increase in blood pressure. Although the hypertensive effect of NSAID administration was more marked in hypertensive patients taking antihypertensive medication than in normotensive persons not taking blood pressure-lowering therapy, the difference was not statistically significant and its clinical relevance is unclear.

    The duration of antihypertensive therapy may not be important (see Figure 1) given that the effect of NSAIDs on blood pressure was similar in the group where antihypertensive agents were administered to hypertensive patients for weeks to months and in the group where the antihypertensive agents were administered to normotensive volunteers for 1 day or less.

    A meta-analysis according to antihypertensive type showed that NSAIDs antagonized the blood pressure-lowering effect of β-blockers and vasodilators more than diuretics although the difference was not quite statistically significant. If it represents a true difference, this finding could be at least partially explained by varying effects of these antihypertensive agents on vasodilatory prostaglandin synthesis. Such a hypothesis remains speculative.

    Analysis of variance showed that NSAIDs varied significantly in their effects on blood pressure. Although piroxicam was the only NSAID that elevated supine mean blood pressure significantly in a statistical sense, indomethacin and ibuprofen had effects of a similar magnitude.

    However, the variance was relatively large, and a larger sample size may have been necessary for statistical significance to have been achieved. Sulindac and aspirin appeared to affect supine mean blood pressure minimally and probably not to a clinically important extent. For the remaining NSAIDs (tiaprofenic acid, diclofenac, naproxen, and flurbiprofen), the pooled, weighted mean increase in supine mean blood pressure was intermediate between the extremes described above or the CIs were very wide or both, suggesting that insufficient studies were available for pooling and that further studies are required to assess fully their effect on blood pressure control.

    Standard anti-inflammatory doses of each NSAID were used in the studies included in the meta-analysis. Hence, the degree of prostaglandin inhibition should have been approximately the same for individual NSAIDs. If prostaglandin inhibition is a part of the mechanism by which NSAIDs alter blood pressure, then similar hypertensive effects would be expected for different agents within this class of drug. The apparent variability in the effect of different NSAIDs on blood pressure raises the possibility that this effect is mediated through a mechanism other than prostaglandin inhibition. An alternative possibility, however, is that the variability in the pooled effect of different NSAIDs on blood pressure is caused primarily by confounding by the variable presence of other factors known to predispose to this effect in the studies pooled (for example, hypertension and the use of β-blocker drugs). Consequently, caution is required in the interpretation of the analysis by NSAID type.

    The meta-analysis of randomized but not placebo-controlled trials suggested that indomethacin probably increased supine mean blood pressure substantially more than did sulindac. The remaining comparisons between NSAIDs in this meta-analysis were inconclusive because of the relatively few trials included in each comparison and the resultant wide CIs.

    To assess possible mechanisms by which NSAIDs may influence blood pressure control, we analyzed six non-blood pressure variables. The pooled results showed that NSAIDs did not significantly alter body weight, daily urinary sodium output, or creatinine clearance. Importantly, NSAIDs did not appear to elevate blood pressure primarily by increasing salt and water retention because weight and urinary sodium were not altered by NSAID therapy, and the antagonism of blood pressure control was not more marked in patients taking diuretics compared with other antihypertensive medication. In addition, NSAIDs did not significantly alter plasma renin activity and 24-hour urinary prostaglandin E2 and 6-keto prostaglandin F1 α excretion. There was, however, a strong trend toward a decrease in plasma renin activity and the urinary prostaglandin output.

    The failure of the analysis to show any significant association between NSAIDs and the six non-blood pressure variables raises the possibility that other factors may contribute to the effect of NSAIDs in elevating blood pressure. In particular, the potential effect of NSAIDs on peripheral vascular resistance and cardiac function requires further evaluation. In a study in which indomethacin was administered on a short-term basis to normal men, Nowak and Wenmalm [86] reported a significant increase in arterial pressure and peripheral resistance with a variable decline in cardiac output.

    Meta-analysis is not intended to replace large randomized clinical trials. Further, the technique of meta-analysis has been carefully evaluated, and many potential problems have been identified [87-90]. However, the sample size for most trials assessing the effect of NSAIDs on blood pressure was inadequate. Consequently, meta-analysis produced a more stable estimate of the overall effect of NSAIDs on blood pressure and allowed consideration of related issues through subgroup analyses.

    Our meta-analysis provides evidence that, as a group, NSAIDs significantly elevate blood pressure. This effect is likely to have clinical relevance, particularly for the elderly, who have the highest prevalence of chronic illnesses, including musculoskeletal disorders and hypertension. If real, such an effect could be associated with a substantial increase in morbidity and mortality with a concomitant cost to the community. However, caution is required in extrapolating these findings from predominantly short-term NSAID-induced increases in blood pressure to complications of long-term blood pressure elevation, particularly in the elderly, because none of the studies in the meta-analysis included elderly volunteers or patients.

    The subgroup analyses (see Figures 1, 2, and 3) should be interpreted cautiously because the sample sizes were insufficient to facilitate stratification before analysis to account for potential confounders. Future studies with much larger sample sizes are required and should focus on the high-risk groups highlighted in this meta-analysis. Finally, NSAIDs have proven anti-inflammatory and analgesic properties, and the question of whether the risks of this class of drugs outweigh its benefits remains unanswered.

    Previous Section
     

    References

    1. 1.↵
      Hypertension prevalence and the status of awareness, treatment, and control in the United States. Final Report of the Subcommittee on Definition and Prevalence of the 1984 Joint National Committee. Hypertension. 1985; 7:457-68.
    2. 2.↵
      Davidson RA, Hale WE, Moore MT, May FE, Marks RG, Stewart RB. Incidence of hypertension in an ambulatory elderly population. J Am Geriatr Soc. 1989; 37:861-6.
    3. 3.↵
      Relationship of blood pressure, serum cholesterol, smoking habit, relative weight and ECG abnormalities to incidence of major coronary events: final report of the pooling project. The pooling project research group. Journal of Chronic Diseases. 1978; 31:201-306.
    4. 4.↵
      Kannel WB, Wolf PA, Verter J, McNamara PM. Epidemiologic assessment of the role of blood pressure in stroke. The Framingham study. JAMA. 1970; 214:301-10.
    5. 5.↵
      Shurtleff D. Some characteristics related to the incidence of cardiovascular disease and death: Framingham Study, 18-year follow-up. In: Kannel WB, Gordon T, eds. The Framingham Study: An Epidemiological Investigation of Cardiovascular Disease. Washington, DC: U.S. Department of Health, Education and Welfare, publication no. (NIH) 74-599; 1974, section 30.
    6. 6.↵
      Hypertension Detection and Follow-up Program Cooperative Group. Five-year findings of the hypertension detection and follow-up program. I. Reduction in mortality of persons with high blood pressure, including mild hypertension. JAMA. 1979; 242:2562-71.
    7. 7.↵
      National Center of Drugs and Biologics. Drug utilization in the United States by categories, 1981. 3rd annual review. Washington, DC: 1982:17.
    8. 8.↵
      Baum C, Kennedy DL, Forbes MB. Utilization of nonsteroidal antiinflammatory drugs. Arthritis Rheum. 1985; 28:686-92.
    9. 9.↵
      Committee on Safety of Medicines. CSM update. Non-steroidal anti-inflammatory drugs and serious gastrointestinal reaction −1.Br Med J. 1986; 292:614.
    10. 10.↵
      Annual Report of the Department of Community Services & Health 1988/89, Statistical Supplement;1989:48.
    11. 11.↵
      Griffin JP. Survey of the spontaneous adverse drug reaction reporting schemes in fifteen countries. Br J Clin Pharmacol. 1986; 22:83S-100S.
    12. 12.↵
      O'Brien B. Patterns of European Diagnoses and Prescribing. London: Office of Health Economics; 1984.
    13. 13.↵
      Walt R, Katschinski B, Logan R, Ashley J, Langman M. Rising frequency of ulcer perforation in elderly people in the United Kingdom. Lancet. 1986; 1:489-92.
    14. 14.↵
      Cinquegrani MP, Liang CS. Indomethacin attenuates the hypotensive action of hydralazine. Clin Pharmacol Ther. 1986; 39:564-70.
    15. 15.↵
      Staessen J, Fagard R, Lijnen P, Moerman E, De Schaepdryver A, Amery A. Effects of prostaglandin synthesis inhibition on blood pressure and humoral factors in exercising, sodium-deplete normal man. J Hypertens. 1983; 1:123-30.
    16. 16.↵
      Hardy BG, Bartle WR, Myers M, Bailey DG, Edgar B. Effect of indomethacin on the pharmacokinetics and pharmacodynamics of felodipine. Br J Clin Pharmacol. 1988; 26:557-62.
    17. 17.↵
      Staessen J, Cattaert A, Fagard R, Lijnen P, Moerman E, De Schaepdryver A, et al. Hemodynamic and humoral effects of prostaglandin inhibition in exercising humans. J Appl Physiol. 1984; 56:39-45.
    18. 18.↵
      Cowley AJ, Stainer K, Rowley JM, Wilcox RG. Effect of aspirin and indomethacin on exercise-induced changes in blood pressure and limb blood flow in normal volunteers. Cardiovasc Res. 1985; 19:177-80.
    19. 19.↵
      Mookerjee BK, Patak RV, Bentzel CJ, Lee JB. Indomethacin antagonism of the effects of furosemide in normal and hypertensive man (Abstract). Kidney Int. 1975; 8:441.
    20. 20.↵
      Walter E, Kaufmann W, Oster P. Does chronic aspirin treatment increase blood pressure in man? Klin Wochenschr. 1981; 59:297-9.
    21. 21.↵
      Beckmann ML, Gerber JG, Byyny RL, LoVerde M, Nies AS. Propranolol increases prostacyclin synthesis in patients with essential hypertension. Hypertension. 1988; 12:582-8.
    22. 22.↵
      Gerber JG, LoVerde M, Byyny RL, Nies AS. The antihypertensive efficacy of hydrochlorothiazide is not prostacyclin dependent. Clin Pharmacol Ther. 1990; 48:424-30.
    23. 23.↵
      Webster J, Petrie JC, McLean I, Hawksworth GM. Flurbiprofen interaction with single doses of atenolol and propranolol. Br J Clin Pharmacol. 1984; 18:861-6.
    24. 24.↵
      Reimann IW, Ratge D, Wisser H, Frolich JC. Are prostaglandins involved in the antihypertensive effect of dihydralazine? Clin Sci. 1981; 61:319s-21s.
    25. 25.↵
      Quilley J, Duchin KL, Hudes EM, McGiff JC. The antihypertensive effect of captopril in essential hypertension: relationship to prostaglandins and the kallikrein-kinin system. J Hypertens. 1987; 5:121-8.
    26. 26.↵
      Pugliese F, Simonetri BM, Cinotti CA, Ciabattoni C, Catelle F, Vaocano G, et al. Differential interaction of piroxicam and sulindac with the antihypertensive effect of propranolol (Abstract). Eur J Clin Invest. 1984; 14:54.
    27. 27.↵
      Salvetti A, Abdel-Haq B, Magagna A, Pedrinelli R. Indomethacin reduces the antihypertensive action of enalapril. Clin Exp Hypertens. 1987; 9:559-67.
    28. 28.↵
      Kirch W, Stroemer K, Hoogkamer JF, Kleinbloesem CH. The influence of prostaglandin inhibition by indomethacin on blood pressure and renal function in hypertensive patients treated with cilazapril. Br J Clin Pharmacol. 1989; 27:297S-301S.
    29. 29.↵
      Salvetti A, Arzilli F, Pedrinelli R, Beggi P, Motolese M. Interaction between oxprenolol and indomethacin on blood pressure in essential hypertensive patients. Eur J Clin Pharmacol. 1982; 22:197-201.
    30. 30.↵
      Wong DG, Spence JD, Lamki L, Freeman D, McDonald JW. Effect of non-steroidal anti-inflammatory drugs on control of hypertension by β-blockers and diuretics. Lancet. 1986; 1:997-1001.
    31. 31.↵
      Smith MD, Kupa A, Weatherall M, Henstridge JD, Brooks PM. The effect of tiaprofenic acid on blood pressure control in treated hypertensive patients. Curr Med Res Opin. 1985; 9:388-93.
    32. 32.↵
      Puddey IB, Beilin LJ, Vandongen R, Banks R, Rouse I. Differential effects of sulindac and indomethacin on blood pressure in treated essential hypertensive subjects. Clin Sci. 1985; 69:327-36.
    33. 33.↵
      Radack KL, Deck CC, Bloomfield SS. Ibuprofen interferes with the efficacy of antihypertensive drugs. A randomized, double-blind, placebo-controlled trial of ibuprofen compared with acetaminophen. Ann Intern Med. 1987; 107:628-35.
    34. 34.↵
      Mills EH, Whitworth JA, Andrews J, Kincaid-Smith P. Non-steroidal anti-inflammatory drugs and blood pressure. Aust N Z J Med. 1982; 12:478-82.
    35. 35.↵
      Salvetti A, Pedrinelli R, Magagna A, Abdel-Haq B. The influence of indomethacin on some pharmacological actions of atenolol. In: Dunn MJ, Patrono C, Cinotto GA, eds. Prostaglandins and the kidney: biochemistry, physiology, pharmacology, and clinical applications. New York: Plenum Medical Book Company; 1983:287-95.
    36. 36.↵
      Wing LM, Bune AJ, Chalmers JP, Graham JR, West MJ. The effects of indomethacin in treated hypertensive patients. Clin Exp Pharmacol Physiol. 1981; 8:537-41.
    37. 37.↵
      Chalmers JP, West MJ, Wing LM, Bune AJ, Graham JR. Effects of indomethacin, sulindac, naproxen, aspirin, and paracetamol in treated hypertensive patients. Clin Exp Hypertens. 1984; 6:1077-93.
    38. 38.↵
      Baez MA, Alvarez CR, Weidler DJ. Effects of the non-steroidal anti-inflammatory drugs, piroxicam or sulindac, on the antihypertensive actions of propranolol and verapamil. J Hypertens. 1987; 5:S563-6.
    39. 39.↵
      Watkins J, Abbott EC, Hensby CN, Webster J, Dollery CT. Attenuation of hypertensive effect of propranolol and thiazide diuretics by indomethacin. Br Med J. 1980; 281:702-5.
    40. 40.↵
      Shaw W, Shapiro D, Antonello J, Cressman M, Vlasses P, Oparil S. Indomethacin does not blunt the antihypertensive effect of lisinopril (Abstract). Clin Pharmacol Ther. 1987; 41:219.
    41. 41.↵
      Wright JT, McKenney JM, Lehany A, Bryan DL, Cooper LW, Lambert CM. The effect of high-dose short-term ibuprofen on antihypertensive control with hydrochlorothiazide. Clin Pharmacol Ther. 1989; 46:440-4.
    42. 42.↵
      Hollifield JW. Failure of aspirin to antagonize the antihypertensive effect of spironolactone in low-renin hypertension. South Med J. 1976; 69:1034-6.
    43. 43.↵
      Oparil S, Horton R, Wilkins LH, Irvin J, Hammett DK. Antihypertensive effect of enalapril in essential hypertension: role of prostacyclin. Am J Med Sci. 1987; 294:395-402.
    44. 44.↵
      Cinquegrani MP, Liang CS. Antihypertensive effects of pinacidil in patients with and without indomethacin pretreatment. Clin Exp Hypertens. 1988; 10:411-31.
    45. 45.↵
      Davies JG, Rawlins DC, Busson M. Effect of ibuprofen on blood pressure control by propranolol and bendrofluazide. J Int Med Res. 1988; 16:173-81.
    46. 46.↵
      Reuse C, Leeman M, Degaute JP, Abramowicz M, Prost JF, Naeije R. Preserved renal perfusion during β blockade by tertatolol with and without cyclooxygenase inhibition in normal humans. J Clin Pharmacol. 1988; 28:312-6.
    47. 47.↵
      Pedrinelli R, Magagna A, Salvetti A. The effect of oxprenolol and indomethacin on renin and aldosterone of normal subjects during low sodium diet. Eur J Clin Invest. 1982; 12:107-11.
    48. 48.↵
      Sugimoto K, Fujimura A, Kumagai Y, Kotegawa T, Ebihara A. Influence of indomethacin on a reduction in forearm blood flow induced by propranolol in healthy subjects. J Clin Pharmacol. 1989; 29:307-10.
    49. 49.↵
      Roberts DG, Gerber JG, Barnes JS, Zerbe GO, Nies AS. Sulindac is not renal sparing in man. Clin Pharmacol Ther. 1985; 38:258-65.
    50. 50.↵
      Gullner HG, Gill JR Jr, Bartter FC, Dusing R. The role of the prostaglandin system in the regulation of renal function in normal women. Am J Med. 1980; 69:718-24.
    51. 51.↵
      McKenney JM, Wright JT, Goodman RP, Cooper L, Yunker N, Lambert C. Effect of high-dose ibuprofen on 24-hour blood pressure in healthy women. DICP. Ann Pharmacother. 1987; 21:517-20.
    52. 52.↵
      Abate MA, Layne RD, Neely JL, D'Alessandri R. Effect of naproxen and sulindac on blood pressure response to atenolol. DICP Ann Pharmacother. 1990; 24:810-3.
    53. 53.↵
      Lewis RV, Toner JM, Jackson PR, Ramsay LE. Effects of indomethacin and sulindac on blood pressure of hypertensive patients. Br Med J. 1986; 292:934-5.
    54. 54.↵
      Koopmans PP, Thien T, Thomas CM, Van den Berg RJ, Gribnau FW. The effects of sulindac and indomethacin on the anti-hypertensive and diuretic action of hydrochlorothiazide in patients with mild to moderate essential hypertension. Br J Clin Pharmacol. 1986; 21:417-23.
    55. 55.↵
      Steiness E, Waldorff S. Different interactions of indomethacin and sulindac with thiazides in hypertension. Br Med J. 1982; 285:1702-3.
    56. 56.↵
      Trimarco B, De Simone A, Cuocolo A, Ricciardelli B, Volpe M, Patrignani P, et al. Role of prostaglandins in the renal handling of a salt load in essential hypertension. Am J Cardiol. 1985; 55:116-21.
    57. 57.↵
      Koopmans PP, Kateman WG, Tan Y, van Ginneken CA, Gribnau FW. Effects of indomethacin and sulindac on hydrochlorothiazide kinetics. Clin Pharmacol Ther. 1985; 37:625-8.
    58. 58.↵
      Abdel-Haq B, Magagna A, Favilla S, Salvetti A. The interference of indomethacin and of imidazole salicylate on blood pressure control of essential hypertensive patients treated with atenolol. Preliminary report. Int J Clin Pharmacol Ther Toxicol. 1987; 25:598-600.
    59. 59.↵
      McMahon FG, Jain AK, Vargas R, Ryan M, Ryan JR. Does sulindac attenuate pressure control among hypertensives? Curr Ther Res. 1989; 45:143-51.
    60. 60.↵
      Salvetti A, Pedrinelli R, Magagna A, Ugenti P. Differential effects of selective and non-selective prostaglandin-synthesis inhibition on the pharmacological responses to captopril in patients with essential hypertension. Clin Sci. 1982; 63:261S-3S.
    61. 61.↵
      Salvetti A, Pedrinelli R, Alberici P, Magagna A, Abdel-Haq B. The influence of indomethacin and sulindac on some pharmacological actions of atenolol in hypertensive patients. Br J Clin Pharmacol. 1984; 17:108S-11S.
    62. 62.↵
      Koopmans PP, Thien T, Gribnau FW. The influence of ibuprofen, diclofenac and sulindac on the blood pressure lowering effect of hydrochlorothiazide. Eur J Clin Pharmacol. 1987; 31:553-7.
    63. 63.↵
      Koopmans PP, Thien T, Gribnau FW. Influence of non-steroidal anti-inflammatory drugs on diuretic treatment of mild to moderate essential hypertension. Br Med J. 1984; 289:1492-4.
    64. 64.↵
      Curb JD, Borhani NO, Schnaper H, Kass E, Entwisle G, Williams W, et al. Detection and treatment of hypertension in older individuals. Am J Epidemiol. 1985; 121:371-6.
    65. 65.↵
      Ebel DL, Rhymer AR, Stahl E. Effect of sulindac, piroxicam and placebo on the hypotensive effect of propranolol in patients with mild to moderate essential hypertension. Adv Ther. 1985; 2:131-42.
    66. 66.↵
      Koopmans PP, Thien T, Thomas CM, Gribnau FW. Do indomethacin and sulindac antagonise the antihypertensive effect of hydrochlorothiazide (Abstract)? Pharm Weekbl Sci. 1985; 7:87.
    67. 67.↵
      Abbott EC, Daniel J, Watkins J, Dollery CT. The effects of indomethacin on blood pressure in treated primary hypertension (Abstract). Clin Res. 1978; 26:834A.
    68. 68.↵
      Jackson SH, Pickles H. Indomethacin does not attenuate the effects of hydralazine in normal subjects. Eur J Clin Pharmacol. 1983; 25:303-5.
    69. 69.↵
      Motolese M, Beggi P, Salvetti A, Arzilli F. Interaction between oxprenolol and indomethacin on blood pressure of essential hypertensive patients (Abstract). Eur J Clin Invest. 1981; 11:22.
    70. 70.↵
      Alvarez CR, Baez MA, Weidler DJ. Effect of sulindac and piroxicam administration on the antihypertensive effect of propranolol (Abstract). J Clin Pharmacol. 1986; 26:544.
    71. 71.↵
      Lewis RV, McLay J, Maclean D, Tregaskis B. The effects of indomethacin and sulindac upon the blood pressures of individuals with untreated labile or mild hypertension. J Hum Hypertens. 1989; 3:233-7.
    72. 72.↵
      Halabi A, Linde M, Zeidler H, Konig J, Kirch W. Double-blind study on the interaction of oxaprozin with metoprolol in hypertensives. Cardiovasc Drugs Ther. 1989; 3:441-3.
    73. 73.↵
      Hurwitz GA, Halushka PV, Privitera PJ, Webb JG. Indomethacin does not enhance sympathetic responses to exercise in man (Abstract). Clin Res. 1979; 27:234A.
    74. 74.↵
      Maunuksela EL, Olkkola KT, Korpela R. Intravenous indomethacin as postoperative analgesic in children: acute effects on blood pressure, heart rate, body temperature and bleeding. Ann Clin Res. 1987; 19:359-63.
    75. 75.↵
      Chalmers TC, Smith H Jr, Blackburn B, Silverman B, Schroeder B, Reitman D, et al. A method for assessing the quality of a randomized control trial. Controlled Clin Trials. 1981; 2:31-49.
    76. 76.↵
      Gardner MJ, Altman DG. Confidence intervals rather than P values: estimation rather than hypothesis testing. Br Med J. 1986; 292:746-50.
    77. 77.↵
      Rosenthal R, Rubin DB. Comparing effect sizes of independent studies. Psychol Bull. 1982; 92:500-4.
    78. 78.↵
      Cappuccio FP, MacGregor GA. Does potassium supplementation lower blood pressure? A meta-analysis of published trials. J Hypertens. 1991; 9:465-73.
    79. 79.↵
      Fleiss JL. The Design and Analysis of Clinical Experiments. New York: John Wiley & Sons; 1986.
    80. 80.↵
      Sacks HS, Berrier J, Reitman D, Ancona-Berk VA, Chalmers TC. Meta-analyses of randomized, controlled trials. N Engl J Med. 1987; 316:450-5.
    81. 81.↵
      Collins R, Peto R, MacMahon S, Hebert P, Fiebach NH, Eberlein KA, et al. Blood pressure, stroke, and coronary heart disease. Part 2. Short-term reductions in blood pressure: overview of randomised drug trials in their epidemiological context. Lancet. 1990; 335:827-38.
    82. 82.↵
      MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, et al. Blood pressure, stroke, and coronary heart disease. Part 1. Prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet. 1990; 335:765-74.
    83. 83.↵
      London GM, Hornych A, Safar ME, Levenson JA, Simon AC. Plasma prostaglandins PGE2 and PGF2 α, total effective vascular compliance and renal plasma flow in essential hypertension. Nephron. 1982; 32:118-24.
    84. 84.↵
      Davis A, Day RO, Begg EJ. Interactions between non-steroidal anti-inflammatory drugs and antihypertensives and diuretics. Aust N Z J Med. 1986; 16:537-46.
    85. 85.↵
      Wing LM. NSAIDs: Interactions with antihypertensives and diuretics. In: Brooks PM, Day RO, eds. Non-steroidal Anti-inflammatory Drugs, Basis for Variability in Response. Basel: Birkhauser Verlag; 1985:91-7.
    86. 86.↵
      Nowak J, Wennmalm A. Influence of indomethacin and of prostaglandin E1 on total and regional blood flow in man. Acta Physiol Scand. 1978; 102:484-91.
    87. 87.↵
      Bulpitt CJ. Meta-analysis. Lancet. 1988; 2:93-4.
    88. 88.↵
      Wachter KW. Disturbed by meta-analysis? Science. 1988; 241:1407-8.
    89. 89.↵
      Thacker SB. Meta-analysis. A quantitative approach to research integration. JAMA. 1988; 259:1685-9.
    90. 90.↵
      Grossberg GT. The pitfalls of meta-analysis (Editorial). J Am Geriatr Soc. 1990; 38:607.
    « Previous | Next Article »Table of Contents

    This Article

    1. Ann Intern Med August 15, 1994 vol. 121 no. 4 289-300
    1. AbstractFREE
    2. Figures
    3. » Full Text
    4. Full Text (PDF)

    Rapid Responses

    1. Submit a response
    2. No responses published

    Navigate This Article

    Current Issue

    1. November 17, 2009, 151 (10)
    1. Alert me to current issues