In this meta-analysis, we summarized published evidence from 9 human intervention studies that investigated the effect of tea consumption on endothelial function as measured by FMD. It was found that moderate consumption of tea substantially enhances FMD. This is in line with findings from a previous meta-analysis on flavonoids/flavonoid-rich foods and cardiovascular risk factors 
, that included 2 studies on tea and FMD published at that time 
. The effect of tea on FMD seems robust because the estimated overall effect is large and the effect was observed in 8 out of 9 studies including different study populations and different tea types. Also, we could not detect indications of systematic publication bias.
During the last decade, FMD has been increasingly used as tool to assess effects of therapeutic interventions on endothelial function in humans. The relevance of FMD for predicting CVD risk independent of other well-established risk factors is crucial in this respect. Several prospective studies, but not all 
, indicate an independent inverse association between FMD and risk for cardiovascular events, not only in patients with varying stages of arterial disease 
, but also in subjects without diagnosis of CVD 
. FMD has been shown to add to the predictive value of ankle-brachial pressure index 
and hyperaemic flow velocity 
in patients with peripheral arterial disease. In addition, Chan et al. 
showed an interaction between carotid plaque burden and endothelial function for predicting future adverse vascular events in coronary artery disease patients. On the contrary, FMD does not seem to be independently associated with CVD outcomes when intima-media thickness is considered at the same time 
. Whether improvement in FMD in response to treatment can predict CVD risk was assessed in at least 2 studies. The study by Kitta et al. 
suggested that optimized therapy to reduce risk factors for coronary artery disease is effective in improving FMD in CVD patients, with persistent impairment in FMD being an independent predictor of events. Modena et al. 
showed that in postmenopausal women, antihypertensive therapy positively affects FMD, and that improvement in FMD identified patients with a subsequent more favourable prognosis for events when adjusted for changes in other risk factors such as SBP and DBP. Thus, taken together, an independent association is suggested between FMD and CVD risk, although causality remains to be proven.
Potential health properties of tea are likely due to certain chemical substances extracted from the tea leaves. The active tea substances responsible for the increase in FMD are still unknown, but a specific role for certain flavonoids, such as the catechins epicatechin, epigallocatechin, epigallocatechin gallate, and epicatechin gallate has been suggested 
. The caffeine in tea probably does not contribute to the effect; an oral dose of pure caffeine was found to not significantly affect FMD 
or even reduced FMD 
. Also, tea solids dose-dependently affect FMD when caffeine intake is kept constant 
. In one study 
, tea did not significantly improve FMD. An explanation may be found in the fact that FMD was measured only shortly (60–90 min) after tea consumption. Since catechin plasma concentrations peak 1–2 h after intake in fasted state and 2–3 h in postprandial state 
, it may be that the catechins were not yet fully absorbed at the time of the FMD assessment. However, whether the catechins are the true actives remains to be elucidated.
With the studies included in our analysis, we were not able to confirm a dose-response relation between amount of tea consumed and FMD response as was suggested by the study by Grassi et al. 
. The absence of a clear dose-response relation may be due to the inaccuracy of expressing tea intake as volume of beverage consumed per day rather than as amount of potentially active substances ingested in the different tea interventions. The included studies used different tea products and preparation methods with varying amounts of tea leaves, brewing time, and water temperatures. These differences presumably result in different concentrations of active substances per tea serving 
. It has been estimated that approximately 84% of total polyphenol content in tea is flavonoids 
. Black tea contains on average 992 mg/L of total polyphenols, whereas green tea contains 591 mg/L 
. A dose of 500 mL of tea (~2–3 cups) can thus be estimated to contain on average 415 mg of flavonoids for black tea and 248 mg for green tea. However, these are rough estimates and, in practice, the variation between tea brews will be large.
A proposed mechanism by which dietary flavonoids could affect FMD is that they improve the bioactivity of the endothelium-derived vasodilator NO 
by enhancing NO synthesis or by decreasing superoxide-mediated NO breakdown 
. Flavonoids may increase endothelial NO production 
by stimulating Akt-mediated endothelial-derived NO synthase activity 
, and additionally decrease levels of the vasoconstrictor endothelin-1 
. Another mechanistic explanation is that methylated flavonoids inhibit nicotinamide adenine dinucleotide phosphate oxidase activity, and thereby reduce the generation of reactive superoxide and hydroxyperoxide 
. However, the precise mechanism is not yet fully revealed, and different potentially active flavonoids and their metabolites may have different effects. No indication of endothelial-independent vasodilatory effects of tea was found in the included studies. Although not conclusive, this makes it unlikely that consumption of tea sensitizes arterial smooth muscle cells for NO.
Because heterogeneity between studies was significant, we analyzed the impact of potential covariates on the relationship between tea and FMD in order to identify factors that could explain differences in findings between studies. Only the study quality score was significantly associated with the size of the FMD response observed after tea consumption, with higher quality studies showing smaller improvements in FMD than the lower quality studies. However, this association was no longer significant when corrected for other covariates in the same model. It should be noted that scoring the quality of a study is intrinsically subjective. Thus, the apparent relation between study quality and observed effect on FMD should be interpreted with caution. Overall, the number of studies included in this meta-analysis was too limited to allow a thorough, reliable analysis of sources of heterogeneity.
Our study has other limitations. Firstly, reproducibility of a functional marker such as FMD is low as compared to most biochemical markers. Standardized protocols for imaging techniques, as for example recently described by Thijsen et al. 
, are required to reduce within-subject variability. These should include multiple measurements and duplicate readings, preferably using automated vessel wall boundary detection devices 
. Such an automated device was used in only 2 out of 9 studies included in the present meta-analysis. Also, the position of the occlusion cuff, either proximal or distal to the area of FMD measurement, is important to consider. A post-hoc analysis suggested that cuff position has an impact on the size of the FMD response after tea consumption, with larger effects on FMD when using the proximal occlusion method, which may explain part of the heterogeneity found between the studies. Doshi et al. 
have shown that FMD assessed by the distal occlusion method can be abolished by infusion of the NO synthase inhibitor NG-monomethyl-L-arginine, but is only partly inhibited (by ~35%) when using the proximal occlusion method. This indicates that dilation after proximal occlusion is not entirely NO-mediated. Because an enhanced tea effect on FMD was found with the proximal cuff placement, it may be speculated that tea affects the vasodilatory response beyond affecting NO bioavailability via mechanisms still to be revealed 
. A second limitation of our analysis is that the majority of included studies measured FMD acutely, i.e. about 2 hours after ingestion of a defined dose of tea. Although an improvement in FMD by tea was also seen after longer-term (4 weeks) regular tea consumption, the clinical relevance of acute improvements in FMD is unclear. Thirdly, this meta-analysis included 2 studies that compared more than 1 active treatment to the same control treatment 
. Although bias due to multiple inclusion of the same control group can not be excluded, we considered this as the best possible approach to not exclude valuable data from treatment arms (i.e. different doses). Excluding the study that contributed most to the overall estimate with 4 treatment arms vs. the same control group 
did not materially affect the results.
In conclusion, our findings indicate that tea consumption results in substantial effects on the vascular endothelium as indicated by an improved endothelial-dependent vasodilation in the first hours after intake, and also after longer-term regular consumption of tea. This effect may partly explain the relationship between tea consumption and reduced CVD risk in population studies when assuming that FMD offers independent predictive value for CVD endpoints. However, to what extent tea-mediated improvement in endothelial function is indeed causally related to a reduction in cardiovascular events can only be determined by large, long-term randomized trials on clinical endpoints.