Introduction
Chronic venous disease is a common condition with a diverse range of clinical presentations, including spider veins, varicose veins and venous ulceration. Despite its high prevalence, and varicose vein surgery being one of the most commonly performed procedures on the U.K. National Health Service, much is still unknown about this disease.
The prevalence of disease, although described by a number of epidemiological studies, is poorly characterized, particularly due to diverse methodology employed in the published literature. Reporting of primary, secondary care and questionnaire data yields diverse prevalence rates by the very nature of the data used for analysis, resulting in a varicose vein prevalence of 2 – 56% in men and < 1 – 60% in women. Poor characterization of disease epidemiology has important repercussions on both local and national disease management strategies. It further impacts on correct estimation of the future burden of the disease.
The pathophysiology of chronic venous disease is also poorly characterized. Increasingly, primary varicose vein weakness and dilatation are being recognized as important players in the development of venous reflux and hypertension, leading to the signs and symptoms observed in the different stages of chronic venous disease. Better characterization of the biological pathways involved in disease development and progression is key in improving the management of a disease that, in the new millennium, still employs the same principles used in Ancient Egypt to treat the condition, i.e. destruction of the refluxing vein, whether by removing, them, burning them or sclerosing them by injecting chemicals.

Aims
The aims of this PhD are as follows:
¥ To identify a differential metabolic signature according to CEAP classification in individuals with CVD in both serum and urine.
¥ To validate the results of previously performed departmental studies.
¥ To compare the metabolic signature of systemic serum versus serum from a refluxing lower limb vein.
¥ To identify systemic biomarkers for diagnostic, prognostic and therapeutic applications.


Methods
Following amendment of existing ethical approval for the study, both CVD patients and individuals with no symptoms of CVD and no history of the disease were invited to participate in the study. All participants underwent a comprehensive history and examination, duplex ultrasound assessment, quality of life measures and biofluid sampling in the form of blood and urine testing. Whole blood was centrifuged to obtain serum. All samples were collected following established departmental standard operating procedures (SOPs) (Appendix 9.2) and stored at -80°C. Nuclear Magnetic Resonance Spectroscopy experiments were performed on both serum and urine samples. Spectral data was extracted and processed in MATLAB® software. Multivariate statistical analysis employing principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) was performed in SIMCA software and enabled identification of discriminant spectral regions between the variables tested. Metabolite identification was performed with reference to the existing literature, using online databases and Statistical Total Correlation Spectroscopy (STOCSY).

Results
517 individuals with CVD and 105 healthy volunteers with no clinical symptoms of CVD were prospectively recruited to the study. In total, 622 individuals were recruited from a single centre from October 2014 to June 2016. Demographic, clinical and quality of life data were collected. The main differentiating criterion between CVD and control groups was symptomatology.
The serum experiments were performed on two separate machines, which resulted in a batch effect. After multiple attempts at analyzing the two batches together, it was decided to perform the analysis on the larger batch only to avoid confounding effects from having used a different machine. The urine experiments were all performed on the same machine. Following well-established pre-processing steps, the data was analysed via multivariate statistical tools.
Age and CEAP classification were the most statistically significant models upon multivariate analysis (p < 0.0001). A regression analysis was performed for the most significant metabolites across the CEAP spectrum.
Serum metabolites positively correlating with increasing CEAP score (and reaching statistical significance) included: 1-methylhistidine, phenylalanine, tyrosine, glycerol, lysine and succinate. Low-density lipoprotein was negatively correlated with increasing CEAP class. The discriminant metabolites for age included formate, phenylalanine, tyrosine, urea, glucose, creatinine, lysine, citrate and succinate, which were found in increased levels in the ≥ 60 group. Arm vs leg serum analysis revealed glycerol as the sole discriminant metabolite, with higher levels present in local serum compared to systemic serum.
Urinary metabolites generally correlated negatively with increasing CEAP class. Statistically significant trends included formate, creatinine, glycine, citrate, succinate, pyruvate and α-hydroxyisobutyrate. Age < 60 was associated with increased levels of creatinine, glycine, pyruvate and α –hydroxyisobutyrate.
The metabolites identified are involved in three main pathways: the tricarboxylic acid (TCA) cycle for energy metabolism, the hypoxia inducible factor (HIF) pathway activated in hypoxia and the one-carbon metabolism involved in amino acid synthesis and linked to the TCA cycle. These results suggest increased energy metabolism in higher CEAP classes, particularly in the C4 -6 group. This may be due to increased CVD severity, but the possibility that this is secondary to an underlying skin disorder (e.g. skin staining and venous ulceration), as opposed to a vein problem cannot be ruled out.
Age was an expected confounding factor; the existing surrounding literature on metabolic profiling and ageing presents varied metabolites. This may be due to the cross- sectional nature of many of these studies, which provide a ‘snapshot assessment’ of the metabolic environment.

Conclusions
Increasing CEAP class demonstrates correlation with specific metabolites, some of which have been previously reported in superficial venous disease. Possible pathways include energy metabolism, HIF pathway and the one-carbon metabolism. However, it was not possible to discern whether this metabolic change was due to the underlying venous disease or to skin damage caused by sustained venous hypertension. Local blood contains higher levels of glycerol compared to systemic blood. The urine profile revealed a negative correlation of the aforementioned metabolites with increasing CEAP class, which may be related to the fact that the metabolites are being consumed in upregulated intracellular pathways to generate energy.