Cyclo-oxygenase-2 inhibition and endothelium-dependent vasodilation in younger vs. older healthy adults

Abstract

Aim

A major feature of endothelial dysfunction is reduced endothelium-dependent vasodilation, which in ageing may be due to decreased production of endothelial prostacyclin, or nitric oxide (NO), or both.

Method

We tested this hypothesis in 12 younger (age 18–38 years, six women) and 12 older healthy adults (age 55–73 years, six post-menopausal women). Endothelium-dependent vasodilation was assessed by the forearm vascular conductance (FVC) response to intra-arterial acetylcholine (ACh) (0.5, 1.0, 2.0, 4.0 μg dl⁻¹ forearm tissue min⁻¹) before and 90 min after inhibition of the enzyme cyclo-oxygenase-2 (COX-2) with oral celecoxib (400 mg), followed by the addition of endothelial NO synthase inhibition with intra-arterial N^G-monomethyl-l arginine acetate (L-NMMA).

Results

Ageing was associated with a significantly reduced FVC response to ACh (P = 0.009, age-by-dose interaction; highest dose FVC ± SEM in ageing: 11.2 ± 1.4 vs. younger: 17.7 ± 2.4 units, P = 0.02). Celecoxib did not reduce resting FVC or the responses to ACh in any group. L-NMMA significantly reduced resting FVC and the responses to ACh in all groups, and absolute FVC values following L-NMMA were similar between groups.

Conclusion

In healthy normotensive younger and older adults, there is minimal contribution of prostacyclin to ACh-mediated vasodilation, yet the NO component of vasodilation is reduced with ageing. In the clinical context, these findings suggest that acute administration of medications that inhibit prostacyclin (i.e. COX-2 inhibitors) evoke modest vascular consequences in healthy persons. Additional studies are necessary to test whether chronic use of COX-2 medications reduces endothelium dependent vasodilation in older persons with or without cardiovascular risk factors.

What is already known about this subject

• A cardiovascular hazard associated with cyclo-oxygenase-2 (COX-2) inhibitors is a reduction of endothelial prostacyclin. This may exacerbate vascular dysfunction in the setting of reduced NO, a condition that occurs even in healthy ageing. The contribution of prostacyclin to vasodilation is reduced in patients with hypertension, but it is not known if prostacyclin inhibition with a clinical dose of the COX-2 inhibitor celecoxib blunts endothelium-dependent vasodilation in older adults.

What this study adds

• After a clinical dose of celecoxib in both younger and older adults, there is minimal contribution of prostacyclin to acetylcholine (ACh)-mediated vasodilation, yet the NO component of vasodilation is reduced with ageing. For clinical situations where acute non-opioid analgesia is desirable, it is reassuring that short term use of celecoxib evokes minimal vascular effects. Additional studies are necessary to test whether chronic use of COX-2 inhibitors reduces endothelium dependent vasodilation in healthy ageing.

Introduction

Non-steroidal anti-inflammatory drugs (NSAIDS) are among the most widely prescribed drugs in the world. The selective cyclo-oxygenase-2 (COX-2) inhibitors or ‘coxibs’ have anti-inflammatory and analgesic therapeutic effects while sparing gastrointestinal and bleeding side effects associated with inhibition of COX-1 [1,2]. Short term use of coxibs is particularly desirable for non-sedating and opioid-sparing effects in acute pain syndromes. For example, our institution has shown that in surgical patients, addition of coxibs to a pre-emptive, multi-modal analgesic regimen improves patient outcome and significantly reduces healthcare costs when compared with patients receiving traditional opioids, particularly among patients with significant comorbidities such as heart disease [3,4]. Unfortunately, chronic inhibition of COX-2-dependent formation of prostacyclin is a cardiovascular hazard, especially among patients with cardiovascular risk factors [5,6]. Taken together, the effect of short term prostacyclin inhibition on vascular function is a key clinical question.

Endothelial dysfunction, abnormalities in the regulatory functions of the vascular endothelium, is one of the earliest detectable stages in the pathophysiology of vascular disease [7]. A hallmark of endothelial dysfunction, even in healthy ageing, is reduced endothelium-dependent vasodilation. Formative work demonstrated this concept by a reduced forearm blood flow (FBF) response to brachial artery infusion of acetylcholine (ACh) in healthy older humans, which in all subjects was inversely proportional to age [8].

ACh stimulates endothelial cholinergic muscarinic receptors to activate endothelial nitric oxide synthase (eNOS) to increase production of nitric oxide (NO). NO diffuses to the smooth muscle to evoke vasodilation via cyclic guanosine monophosphate (cGMP) [9]. ACh also stimulates phospholipase A₂ to convert membrane lipids into arachidonic acid for conversion by cyclo-oxygenase enzyme (COX) to form prostanoids, including vasoconstricting substance thromboxane A₂ and vasodilating substance prostacyclin [9]. While COX-1 is expressed in most cells and mediates production of thromboxane A₂, COX-2 is largely responsible for prostacyclin production [10–12]. Prostacyclin release onto smooth muscle cell IP receptors activates cyclic adenosine monophosphate (cAMP) to evoke smooth muscle vasodilation [13,14].

The NO pathway can be partially blocked by inhibition of eNOS with the competitive inhibitor N^G-monomethyl-l arginine acetate (L-NMMA) [15,16]. L-NMMA competes with L-arginine for the active site on eNOS, blocking the ability of L-arginine to be converted to NO, leading to blunted vasodilation. Inhibition of COX-2 by intravenous parecoxib (pro-drug of valdecoxib) decreases ACh-induced forearm vasodilation in patients with hypertension [17] but oral rofecoxib does not impair ACh-induced forerm vasodilation in healthy young adults [18]. Taken together, COX-2 inhibition may have no consequence in healthy young individuals who are presumably free of vascular dysfunction [18]. In contrast, COX-2 inhibition may be critically important for individuals with vascular dysfunction who already may have a reduced production of vasodilators, such as NO.

Endothelium-dependent vasodilation is blunted in healthy ageing, but whether this is due to reduced prostacyclin, reduced NO, or both is unclear. Therefore, the purpose of this study was to characterize the effect of celecoxib and L-NMMA on endothelium-dependent vasodilation in ageing. We hypothesized that compared with younger men and women, older men and women would demonstrate reduced endothelium dependent vasodilation and the mechanisms would include both prostacyclin and NO pathways.

Methods

This study was approved by the Mayo Institutional Review Board. After obtaining written informed consent, we enrolled 12 healthy older adults (six men and six post-menopausal women not on hormone replacement therapy) between the ages of 55 and 73 years. Inclusion age for the older subjects (≥55 years) was based on evidence that endothelial dysfunction accelerates after age 41 years for men and 53 years for women [7], and cardiovascular disease precipitously increases at age 55 years, even in previously healthy people [19]. Simultaneously we enrolled 12 healthy younger adults (six men and six premenopausal women) aged between 18 and 38 years. All subjects were screened by medical history, physical examination, and blood chemistry. Exclusion criteria consisted of a history of smoking, hypertension, hyperlipidaemia, heart disease, diabetes, autonomic disorders, hypersensitivity to sulfonamides (i.e. celecoxib), and other conditions or medications that might alter cardiovascular regulation. On screen visit, subjects were excluded for screening blood pressure >145/90 mmHg on more than two measurements, hyperlipidaemia (triglycerides >200 mg dl⁻¹, total cholesterol >240 mg dl⁻¹, LDL >160 mg dl⁻¹), and a body mass index (BMI) >30 kg m⁻². Physical activity was assessed by the Minnesota Leisure Time Activity questionnaire. Individuals who took nutritional supplements and NSAIDS held these medications for 4 weeks prior to the study day. Premenopausal women were studied during the early follicular phase of the menstrual cycle or placebo phase of those taking oral contraceptive pills.

Subjects checked into the Mayo CTSA Clinical Research Unit after an overnight fast. A 20-gauge brachial artery catheter was placed in the non-dominant arm using an aseptic technique, subcutaneous lidocaine 1%, and ultrasound guidance. Limb blood flow was measured by forearm venous occlusion plethysmography as described in detail elsewhere [14]. During recording, a wrist cuff was continuously inflated to suprasystolic pressure (250 mmHg) to occlude arterial blood flow to the hand while a venous occlusion cuff around the upper arm was inflated to 50 mmHg for 7.5 of every 15 s, providing one blood flow measurement every 15 s and four flows min⁻¹. Drugs were administered based on forearm volume in dl of forearm volume (FAV) as determined by water displacement. Normal saline (0.9% sodium chloride) was infused for baseline measurements. The forearm infusion protocol is depicted in Figure 1. To test endothelium-dependent vasodilation, intra-arterial ACh (Michol-E, Novartis) was infused for 2 min at 0.5, 1.0, 2.0, 4.0 μg dl⁻¹ FAV min⁻¹ with each dose infused at the same rate. After the first trial of ACh infusions, celecoxib 400 mg (Pfizer) was immediately administered orally to block the production of COX-2 mediated prostanoids including prostacyclin. After 90 min an arterial blood sample was collected and sent to a commercial laboratory for determination of serum celecoxib concentrations (NMS Labs, Willow Grove, PA, USA). Intra-arterial L-NMMA (Clinalfa) was administered as a bolus dose of 5 mg min⁻¹ for 10 min, followed by a maintenance infusion of 1 mg min⁻¹. Arterial blood sampling was repeated at 150 min for celecoxib concentrations prior to the subsequent infusion of ACh. Intra-arterial sodium nitroprusside (NTP, Nitropress, Hospira) was infused for 2 min periods at 0.25, 0.5, 1.0, and 2.0 μg dl⁻¹ FAV min⁻¹ to test endothelium-independent vasodilation by direct delivery of NO to vascular smooth muscle.

Figure 1.

Study timeline. After brachial artery catheter placement, ACh was infused at four dose levels. Celecoxib was given orally and ACh was repeated 90 min later. Then N^G-monomethyl-l arginine acetate (L-NMMA) was loaded and infused for the remainder of the protocol. ACh was infused a third time under ‘double blockade’ conditions. Nitroprusside (NTP) was the final vasodilator. *Indicates arterial blood draw to determine plasma celecoxib concentration

Sample-size/statistical power considerations

In a previous study by DeSouza et al., the change in FBF from baseline to the highest dose of ACh was (mean ± SEM) 4.0 ± 0.2 to 12.3 ± 0.7 ml dl⁻¹ FAV min⁻¹ for older men (n = 41) compared with a change from 3.9 ± 0.2 to 17.1 ± 1.1 for young men (n = 22) [20]. Given the sample sizes for that study, this was consistent with a difference in FBF response to ACh of approximately 1.1 SD units. In general, a sample size of 12 older and 12 younger subjects in the present study provided statistical power (one-tailed, α = 0.05) of 82% to detect a difference between groups of 1.1 SD units. Men and women were represented equally in each age group for an exploratory analysis to determine whether gender would influence the findings.

Statistical analysis

Data are presented as FBF in ml dl⁻¹ FAV min⁻¹. To account for age-related variability in mean arterial pressure (MAP), forearm vascular conductance (FVC) was included and calculated as FVC = (FBF/MAP × 100) expressed as arbitrary units. While the sample sizes were powered to detect an age effect on endothelial function, this protocol also generated preliminary data on possible gender differences. Subject characteristics were analyzed using two-way anova with the two factors being age (young vs. old) and gender (male vs. female) (JMP 9.0.1, SAS Institute, Cary, NC, USA). Forearm data were analyzed using mixed linear models (PROC MIXED, SAS version 9.3, SAS Institute, Cary, NC, USA). For these models, the dependent variable was the change in FBF and FVC from baseline, ACh dose was modelled as a within subject effect while age group and gender were modelled as between-subject effects. In addition to the main effects, all two-way interaction terms were included in the model. Separate models were fitted for each treatment condition. To assess the treatment condition comparisons of interest (celecoxib vs. no treatment, celecoxib vs. celecoxib + LNMMA) additional models were fitted which included treatment conditions as within subject effects. To account for the repeated measures study design an unstructured covariance matrix was used. Data are expressed as mean ± SEM. Statistical significance was set at P < 0.05.

Results

Table 1 displays the subject demographics by age. The mean age of the younger and older subjects was 28 ± 2 and 62 ± 2 years, respectively. Homocysteine was the only variable that was greater in the older vs. younger subjects (P = 0.02). For a sub-analysis based on gender, men had a greater height, weight, and BMI (P < 0.001 for all). Systolic and mean arterial pressure were greater in men (P < 0.01) but not diastolic blood pressure. Heart rate was greater in women (P < 0.01). Similarly, women had lower mean haemoglobin (P < 0.001) and creatinine concentrations (P < 0.01), and greater mean high density cholesterol (P < 0.001). Importantly, none of the gender-dependent characteristics above were different based on age. For leisure time physical activity, the estimated yearly metabolic equivalent expenditure was highly variable in the older men, and the comparison between younger and older subjects was not significant (P = 0.07). Finally, mean plasma celecoxib concentrations just prior to the second and third ACh trials were greater than the clinical therapeutic reference range (590–780 ng ml⁻¹, NMS labs). These plasma concentrations were decreasing from 90 min to 120 min, suggesting that peak plasma concentrations had occurred. While celecoxib concentrations tended to be greater in the older subjects, these values were not significant.

Table 1.

Subject characteristics

Characteristic	Younger subjects	Older subjects
Height (cm)	175.2 ± 3.6	171.7 ± 2.1
Weight (kg)	75.7 ± 4.9	75.8 ± 3.1
Body mass index (kg m−2)	24.6 ± 0.7	25.6 ± 0.7
Waist-to-hip ratio	0.82 ± 0.03	0.87 ± 0.03
Systolic blood pressure (mmHg)	115.3 ± 3.9	124.0 ± 4.7
Diastolic blood pressure (mmHg)	68.7 ± 3.2	74.7 ± 2.5
Mean arterial pressure (mmHg)	84.2 ± 3.1	91.1 ± 2.9
Heart rate (beats min-1)	67.2 ± 3.4	61.4 ± 3.1
Haemoglobin (g dl−1)	13.9 ± 0.5	13.9 ± 0.3
Sodium (mEq l−1)	140.2 ± 0.5	141.1 ± 0.4
Potassium (mEq l−1)	4.5 ± 0.08	4.3 ± 0.09
Creatinine (mg dl−1)	0.9 ± 0.05	0.9 ± 0.05
Cholesterol (mg dl−1)	173.5 ± 7.6	196.4 ± 9.2
Triglycerides (mg dl−1)	95.6 ± 11.5	96.5 ± 8.7
HDL (mg dl−1)	58.4 ± 3.1	66.7 ± 5.0
LDL (mg dl−1)	96.0 ± 7.9	110.5 ± 7.6
C-reactive protein (mg l−1)	2.1 ± 0.7	1.3 ± 0.2
Homocysteine (mcmol l−1)	7.4 ± 0.5	9.3 ± 0.6*
Yearly METs	1639 ± 385	4034 ± 1211
Plasma celecoxib 90 min (ng ml−1)	911 ± 174	1200 ± 224
Plasma celecoxib 120 min (ng ml−1)	871 ± 170	1125 ± 202

Variables are displayed as mean ± SEM. HDL, high density lipoprotein; LDL, low density lipoprotein; METs, metabolic equivalents of leisure time physical activity. Also included are the plasma celecoxib concentrations drawn at 90 min and 120 min in the protocol.

^*indicates P = 0.02.

Table 2 displays the mean FBF values in the young and older groups in three experimental conditions: no treatment (i.e. ACh dose–response before celecoxib), ACh after celecoxib and ACh after celecoxib + L-NMMA To account for age-related variability in blood pressure, Figure 2 displays the ACh dose–responses in FVC. For endothelium-dependent vasodilation, young adults had a greater FVC response to ACh than older adults (P < 0.01, main effect of age; P < 0.01, age-by-dose interaction). Celecoxib did not blunt the vasodilator response to ACh in young or older subjects (P = 0.5, main effect of celecoxib), and the age-dependent differences in the response to ACh were preserved (P = 0.02, main effect of age; P = 0.02, age-by-dose interaction). L-NMMA significantly blunted the vasodilator response to ACh in young and older subjects (P < 0.001, main effect of L-NMMA), and there were no longer differences in the vasodilator response to ACh in young vs. older subjects (P = 0.2, main effect of age; P = 0.8, age-by-dose interaction). A linear regression of the ACh dose–responses showed no linear effect of age on vasodilation within age groups, before or after administration of celecoxib.

Table 2.

FBF responses to ACh before and after celecoxib, and celecoxib + L-NMMA

Drug	FBF, ml dl−1 FAV min−1		P values
	Younger	Older	Dose	Age	Interaction
No treatment:
ACh (μg dl−1 FAV min−1)
Baseline	2.65 ± 0.34	2.39 ± 0.20	<0.0001	0.007	0.006
0.5	11.07 ± 1.70	8.78 ± 0.86
1.0	12.33 ± 1.58	7.82 ± 1.03
2.0	15.61 ± 2.15	8.28 ± 0.79
4.0	17.69 ± 2.44	11.23 ± 1.43
Celecoxib:
ACh (μg dl−1 FAV min−1)
Baseline	3.08 ± 0.47	2.28 ± 0.29	<0.0001	0.02	0.02
0.5	11.78 ± 1.69	7.83 ± 1.25
1.0	13.72 ± 2.83	7.06 ± 1.33
2.0	16.13 ± 3.21	9.78 ± 1.59
4.0	19.27 ± 3.10	10.71 ± 1.47
Celecoxib + L-NMMA:
ACh (μg dl−1 FAV min−1)
Baseline	1.78 ± 0.18*	1.61 ± 0.16*	<0.0001	ns	ns
0.5	6.31 ± 0.82	5.60 ± 1.09
1.0	4.88 ± 0.63	4.23 ± 1.13
2.0	6.46 ± 1.42	5.11 ± 1.07
4.0	8.68 ± 1.80	6.79 ± 1.38
NTP (μg dl−1 FAV min−1)
Baseline	1.78 ± 0.13	1.68 ± 0.15	<0.0001	0.04	0.05
0.25	8.73 ± 0.90	7.58 ± 0.73
0.5	11.58 ± 1.31	9.37 ± 0.98
1.0	16.04 ± 1.75	11.78 ± 1.26
2.0	20.83 ± 2.37	13.31 ± 1.47

Variables are displayed as mean ± SEM. ACh, acetylcholine; L-NMMA, N^G-monomethyl-l arginine acetate; NTP, sodium nitroprusside. FAV, forearm volume.

^*indicates a significant reduction in baseline blood flow when compared with the previous celecoxib only condition.

Figure 2.

Forearm vascular conductance (FVC) response to ACh in healthy younger vs. older humans. ACh was administered in the baseline or ‘no treatment’ condition (circles), followed by the celecoxib condition (triangles), followed by the addition of N^G-monomethyl-l arginine acetate (L-NMMA, squares). Younger subjects (left panel, n = 12) had a greater vasodilator response to ACh than older subjects (right panel, n = 12). Celecoxib did not affect the vasodilator response in either group. L-NMMA significantly decreased baseline FVC. The vasodilator response to ACh was significantly inhibited by L-NMMA to a greater extent in younger subjects, such that the FVC response to ACh was no longer different between groups. *P value <0.05 indicates main effect of celecoxib + L-NMMA condition on ACh dose responses compared with the previous condition (celecoxib only). , no treatment; , celecoxib; , celecoxib + L-NMMA

For a sub-analysis of the forearm vasodilator response within genders, young men had a greater vasodilator response to ACh than older men (P = 0.03, main effect of age; P < 0.01, age-by-ACh dose interaction). After celecoxib, these responses tended to be greater but were no longer significant (P = 0.1, main effect of age; P = 0.3, age-by-ACh dose interaction). After L-NMMA there were no differences in the vasodilator response to ACh in young vs. older men (P = 0.3, main effect of age; P = 0.4, age-by-dose interaction). Women did not display an age-dependent difference in the vasodilator response to ACh (P = 0.1, main effect of age; P = 0.2, age-by-ACh dose interaction). After celecoxib, the main effect of age on the FVC response remained non-significant (P = 0.1), while there was a small but significant age-by-dose interaction (P = 0.03). After L-NMMA there were no differences in the vasodilator response to ACh in young vs. older women (P = 0.4, main effect of age; P = 0.9, age-by-dose interaction).

When analyzing the forearm vasodilator response between genders, in all three conditions (no treatment, celecoxib, celecoxib + L-NMMA) there were no differences in the forearm vasodilator responses to ACh when comparing young men vs. young women, and no differences when comparing older men vs. older women. Finally, the forearm vasodilator response to NTP reached small but statistical significance based on age, as younger subjects had a greater FVC response to NPT than older subjects (P = 0.04, main effect of age; P = 0.05, age-by-ACh dose interaction).

Discussion

This study confirmed previous evidence that endothelium dependent vasodilation, as measured by the forearm vasodilator response to ACh, is reduced in healthy older adults. The new finding is, that contrary to our hypothesis, prostacyclin inhibition with a single dose of celecoxib did not reduce resting FBF or endothelium-dependent forearm vasodilation in healthy older adults. When L-NMMA was added to inhibit production of endothelial NO, resting FBF and endothelium-dependent forearm vasodilation were significantly decreased to similar end points in the older vs. younger subjects. Taken together, these findings suggest that the major mechanism by which endothelium-dependent vasodilation is blunted in ageing is due to NO production, while the contribution of prostacyclin to basal vascular tone and endothelium-dependent vasodilation is negligible in healthy normotensive older adults.

The reduction of endothelial prostacyclin in the setting of reduced NO may cause an increased risk of major cardiovascular and cerebrovascular events [21] such as increased blood pressure [22]. The endothelium is a key contributor to the balance between vasodilation and vasoconstriction [23]. Mechanistic changes in age-related endothelial dysfunction include decreased bioavailability of NO, reduced activity of eNOS, and increased oxidative stress [14,24,25]. Additionally, age-associated endothelial dysfunction results in a decrease of vasodilating prostacyclin and an increase in vasoconstricting thromboxane [9,17]. Therefore, in the setting of reduced endothelial NO, long term inhibition of prostacyclin by use of COX-2 inhibitors may accentuate the balance of vascular tone towards vasoconstriction and increased risk of major adverse cardiovascular and neurologic events [26,27]. Indeed, rofecoxib (Vioxx) was shown to increase blood pressure [22] and cardiovascular event rates which led to the discontinuation of this once popular medication [21]. A recent meta-analysis reported that chronic use of COX-2 inhibitors increased major vascular events by a third when compared with placebo [28]. An interesting aspect of that study was that the NSAID ibuprofen was implicated in the risk of vascular events similar to the coxibs, but naproxen did not increase major vascular events, which was postulated to be associated with COX-1 inhibition to a large enough degree that its anti-platelet effect offset the vascular effect of COX-2 inhibition [28].

Our hypothesis was formulated on several lines of investigation including endothelial function in healthy older adults as measured by ACh and exogenous prostacyclin infusion, COX-2 inhibition in patients with risk factors for vascular disease and COX-2 inhibition in healthy young adults. While reduced forearm vasodilator responses to ACh in ageing is well-established [8,20], our laboratory demonstrated that in healthy older adults, the forearm vasodilator response to intra-arterial infusion of prostacyclin was diminished compared with younger subjects. Further, when NO production was decreased with L-NMMA and prostacyclin infusions were repeated, the dose–response curves were blunted in the younger subjects but unchanged in the older subjects, suggesting that the reduced vasodilator effect of prostacyclin may be attributable to reduced NO in the older subjects [14]. The present study extended these findings by confirming that the NO component of endothelium-dependent vasodilation is blunted, and provided new evidence that there is no compensatory increase in prostacyclin production in healthy older subjects. This idea is also consistent with forearm exercise models in healthy ageing subjects. Our laboratory demonstrated that L-NMMA evoked a 40% reduction in the NO component of exercise hyperaemia vasodilation but ketorolac (non-selective COX-1 and COX-2 inhibitor) evoked no appreciable change in the prostaglandin component [29]. Crecelius et al. reported that the augmentation of forearm exercise hyperaemia by acute ascorbic acid administration is mediated primarily by an increase in bioavailability of NO with minimal effect following non-selective COX inhibition with ketorolac [30].

Along the lines of non-selective COX inhibitors, Singh et al. administered aspirin (3, 9, and 30 μmol min⁻¹) to 18 young and 15 older healthy subjects and found that aspirin caused a greater dose-related reduction in FBF in the younger vs. the older subjects [24]. This reduction was interpreted to suggest that the role played by prostaglandins in the regulation of vascular tone diminishes with age. Formative work by Bulut and colleagues showed that in hypertensives, ACh-mediated forearm vasodilation was reduced with systemic intravenous infusion of a COX-2 inhibitor (parecoxib) suggesting the importance of endothelial production of prostanoids in patients with vascular dysfunction [17]. Conversely, Verma and colleagues reported that in healthy young adult volunteers, a 7 day course of the oral COX-2 inhibitor rofecoxib did not blunt the forearm vasodilator response to ACh [18]. Thus, we reasoned that in healthy older adults the mechanism of endothelial dysfunction might include a component of reduced prostacyclin. However, our results were consistent with exercise hyperaemia studies that used non-selective COX blockade in healthy older adults.

The findings in this study can be explained by both patient and procedural factors. First, the older subjects were not recruited based on pre-existing evidence of endothelial dysfunction, such as reduced flow-mediated dilation [31]. However, demographic and screening characteristics of our older subject cohort appeared ‘physiologically’ similar to the younger subjects in every aspect except serum homocysteine concentrations. In this context, a landmark study by Solomon and colleagues determined the cardiovascular risk associated with celecoxib in three dose regimens and assessed the relationship between baseline cardiovascular risk and effect of celecoxib on cardiovascular events in 7950 patients in six placebo-controlled trials [6]. The authors found evidence of differential cardiovascular risk as a function of celecoxib dose regimen and baseline cardiovascular risk. In other words, the patients who received 400 mg of celecoxib day⁻¹ (similar to the dose used in our study), did not suffer increased risk of cardiovascular events if their baseline cardiovascular risk was low, similar to our older population. Taken together, our findings suggest that in healthy older people, compensatory prostacyclin production may not be as critical to vascular health as it may be for patients with pre-existing cardiovascular risk factors.

The procedural factors that deserve consideration regarding our findings centre on limited methods to block endothelial prostacyclin. Currently there is no direct acting injectable selective COX-2 inhibitor available for human use. Although the injectable COX-2 inhibitor parecoxib is available in Europe, it is not approved for use in the United States. In the current study, parecoxib may have provided a method for intravenous drug delivery vs. oral administration [17]. However, intra-arterial administration would not have proved useful because paracoxib is a prodrug of valdecoxib that requires hepatic activation. Therefore, it is not immediately efficacious for inhibition following intra-arterial administration. While we acknowledge that the plasma concentration of paracoxib or celecoxib required to fully inhibit endothelial COX-2 has not been established, we elected to administer the widely used, clinically recommended dose of oral celecoxib for this study, and we confirmed plasma celecoxib concentrations with blood sampling prior to the sercond and third administration of ACh. Moreover, to our knowledge this is the first study to evaluate endothelium-dependent vasodilation in healthy older subjects without use of traditional NSAIDs (i.e. indomethacin, ketorolac) that are limited by non-specific prostaglandin inhibition including thromboxane A₂.

There are also study-related limitations that require consideration. First, the forearm model used is not representative of all human vasculature. Second, we performed subgroup analyses within and between genders to provide exploratory comparative data, but the subgroup sample sizes were not powered to detect differences in these populations (type II error). Thus, these findings should be interpreted as preliminary. However, because celecoxib did not affect the forearm vasodilator response to ACh, it is unlikely that gender-dependent differences in the prostacyclin component of vasodilation would be found in larger sample sizes of healthy older men and women. Finally, we were puzzled by the reduced vasodilator response to NTP in the older vs. younger subjects that reached statistical significance. By study design NTP was not administered until after the ACh dose–responses and one cannot rule out an unforeseen effect of the duel presence of L-NMMA and celecoxib on age-dependent responsiveness to NTP. A more likely explanation is that NTP may evoke a small but appreciable decreased vasodilation in older subjects, suggested by the original data showing an inverse correlation between age and vasodilator responsiveness to ACh and also a smaller but significant relationship for NTP [8]. A final caveat is that endothelium-derived hyperpolarizing factors have been shown to be enhanced in conditions of NO deficiency [32] but exploration of these mechanisms was limited based on protocol duration during arterial cannulation. Future mechanistic studies are needed to test whether chronic use of COX-2 medications reduces endothelium dependent vasodilation in older persons with or without cardiovascular risk factors.

Author contributions

Conception and design of the work: JHE, LRG, ARA, SLK, WTN.

Acquisition of data: JHE, LRG, ARA, SLK, WTN.

Analysis and interpretation of data: JHE, LRG, ARA, SLK, WTN.

Drafting or revising the article critically for intellectual content: JHE, ARA, WTN.

Final approval of the manuscript: JHE, LRG, ARA, SLK, WTN.

This work was supported by US National Institutes of Health grants UL1 TR000135 and HL-89331.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare JHE had support from United States National Institute of Health for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.

The authors are grateful to the study volunteers for their participation. We also thank Jean N. Knutson, Sarah C. Wolhart, and Pamela A. Engrav for their technical assistance.