Population Scale Studies and Biomarkers of Disease Risk
One huge challenge is how best to extract maximal and highly valuable information from thousands of samples taken from epidemiology and biobank repositories and use it to better understand population-based factors in health and disease. This process is complicated by the fact that humans are highly variable with regard to lifestyles and diets and often are not open or truthful when answering questionnaires. Another problem encountered in these studies is the presence of drugs and their metabolites as well as other environmentally present materials. A huge increase in complexity is caused by the significant variability in gut microflora among individuals and the complex, two-way biochemical relationship between the microflora and the host. Nevertheless, metabolic phenotyping of epidemiology cohort biofluid samples has been successful with the derivation of population-wide risk factors and the comparison of such outcomes with information from GWASs.
In the first large-scale investigations of human urine, two studies used metabolic profiling by 1H NMR spectroscopy to study the metabolic variations within and among four distinct human populations—from China, Japan, the United States, and the United Kingdom; the information was collected as part of the INTERMAP project . The population differences could be related to documented differences in diet and diet-related risk factors and also to cardiovascular disease . The metabolites that discriminated between the populations were then also linked to data on blood pressure, and four metabolic biomarkers were determined . Some of these could be related to diet and to gut microbial function.
Following this approach, a number of large cohort investigations used both untargeted and targeted metabolic assay methods involving both NMR spectroscopy and MS. It seems useful to summarize these briefly here because they indicate the scale of what is now possible and the desirability of establishing dedicated phenome centers.
A preliminary study using a large number of samples examined plasma specimens by using NMR spectroscopy to discover subjects at high risk of death in the short term . To help identify biomarkers for all-cause mortality and to enhance risk prediction, over 100 candidate biomarkers were quantified by NMR spectroscopy of nonfasting plasma samples from a subset of the Estonian Biobank using nearly 10,000 samples. Significant biomarkers were validated and incremental predictive utility assessed in a separate 7500-strong population-based cohort from Finland. After adjusting for a range of conventional risk factors, four circulating biomarkers predicted the risk of all-cause mortality among the participants: (1) α-1-acid glycoprotein, (2) albumin, (3) very-low-density lipoprotein particle size, and (4) citrate. All four biomarkers were predictive of cardiovascular mortality, as well as death caused by cancer and other nonvascular diseases.
Similarly, GC-MS and LC-MS analyses of serum have been used to assess of risk of chronic kidney disease (CKD) in African Americans . Many metabolites could be associated with kidney function, including glomerular flow rate, and two possible biomarkers were candidate risk factors for CKD, namely, 5-oxoproline and 1,5-anhydroglucitol.
Quantitative NMR-based metabolic phenotyping has been employed to identify the biomarkers for incident cardiovascular disease during long-term follow-up . Biomarker discovery was conducted in the National Finnish FINRISK study on over 7000 samples. Replication of the discovered biomarkers and incremental risk prediction was assessed in approximately 6000 other samples from two other repositories. This study used a targeted analysis of 68 lipids and metabolites using MS, and 33 features were associated with incident cardiovascular events after adjusting for confounding parameters, that is, age, gender, blood pressure, smoking, diabetes mellitus, and medication. After further adjustment for routinely measured blood lipids, four metabolites were associated with future cardiovascular events, and higher quantities of serum phenylalanine and mono-unsaturated fatty acids were associated with increased cardiovascular risk, whereas higher ω-6 fatty acids and docosahexaenoic acid levels were associated with lower risk. The biomarker associations were further corroborated using MS in about 3000 samples from two further independent cohorts.
Another study was aimed at finding metabolic biomarkers of raised blood pressure, including the effects of lipid-lowering and antihypertensive drugs . This reported a metabolome-wide association study with 295 metabolites in human serum from 1762 participants of the KORA F4 (Cooperative Health Research in the Region of Augsburg) study population. The intention was to find variations of metabolite concentrations related to the intake of the various drug classes and to generate new hypotheses about the expected and unexpected effects of these drugs. For β-blockers, 11 metabolic associations were discovered, whereas for other drugs, lesser numbers were found as follows: angiotensin-converting enzyme (ACE) inhibitors (4), diuretics (7), statins (10), and fibrates (9). For β-blockers, significant associations were observed with metabolite concentrations that are indicative of drug side effects, such as increased serotonin and decreased free fatty acid concentrations. Intake of ACE inhibitors and statins were linked to metabolites, which provide insight into the action of the drug itself on its target, such as an association of ACE inhibitors with des-Arg(9)-bradykinin and aspartylphenylalanine, a substrate and a product of the drug-inhibited ACE. The intake of statins that reduce blood cholesterol resulted in changes in the concentration of metabolites of the biosynthesis as well as of the degradation of cholesterol.
Finally, a novel approach of linking variations in the human genome with variations in metabolic profiles can yield information on the genes and hence the pathways associated with NMR or MS peaks that have significant correlation with chosen biological and clinical endpoints. Thus it may be possible to identify those unassigned metabolites by using the known functions of the connected genes. This approach has been applied to human serum. To test this hypothesis, the first GWAS with metabolic phenotyping data was based on the quantitative measurement of 363 targeted metabolites in the serum of 284 male participants of the KORA study . Associations were found between common single nucleotide polymorphisms and the metabolite levels explaining up to 12% of the observed metabolic variance. Furthermore, using ratios of certain metabolite concentrations as a proxy for enzymatic activity, up to 28% of the variance could be explained. The study identified four genetic variants in genes coding for enzymes, in which the associated metabolites clearly matched the biochemical pathways where these enzymes are active. The results suggested that common genetic polymorphisms can induce considerable variation in the metabolic make-up of the human population. The authors proposed that this might lead to novel approaches to personalized health care based on a combination of genotyping and metabolic characterization.
The approach of studying large cohorts from epidemiologic collections or biobanks has led directly to the concept of dedicated centers for metabolic phenotyping analysis (phenome centers). This is discussed in more detail in chapter “Phenome Centers and Global Harmonization” of this book. The first such the center was set up at Imperial College London, following a £10 million grant from the MRC supported by the National Institute for Health Research (NIHR). To be successful, the approach has required strict operating procedures to ensure reproducibility  and access to state-of-the-art technologies.
In parallel to such innovations, the field of metabolic phenotyping has profited from major advances in analytical technologies. NMR spectroscopy is now routinely used for profiling at a frequency of 600 MHz, and access to higher frequency machines is available for metabolite identification. Although such instruments are not routinely used for profiling, Fig. 2.6 shows a 950 MHz 1H NMR spectrum of human urine; this should be compared with that shown in Fig. 2.2, taken at 400 MHz, to see the vastly increased amount of information present. Similarly GC-MS is now widely used in metabolic profiling studies, but more usually in a targeted fashion, for example, to assay a wide range of bile acids. The use of HPLC-MS has largely been superseded by UPLC-MS with its improved chromatographic resolution and the possibility of shortened separation times, enabling more samples to be measured in a given time.
Further details of studies in population screening and in the investigation of exposure effects are given in the chapter “Population Screening for Biological and Environmental Properties of the Human Metabolic Phenotype: Implications for Personalized Medicine” of this book.
Monitoring of glucose or glucose-related variables is of paramount importance in evaluating and adjusting glucose-lowering therapies. Self-monitoring of glood glucose (SMBG) and tests of glycated hemoglobin (A1C) levels allow measurement of fasting, timed or random glucose levels or estimates of mean glycemic exposure. There is however some evidence to suggest that control of glucose variability and post prandial glucose may be equally important in prevention of late complications, and there is therefore a need for additional tests to evaluate these aspects of glucose exposure.
1, 5 anhydroglucitol is a naturally occurring 1-deoxy form of glucose first identified in 1888 within the plant kingdom and in 1972 within humans. It circulates in the body mostly in its free form. When glucose levels surpass the renal threshold for glucosuria, 1,5 anhydroglucitol is excreted in the urine, thus lowering levels in serum or plasma. Poor glucose control is associated with low levels of 1, 5 anhydroglucitol (1, 5- AG). It can also reflect transient elevation of glucose over the previous 2-3 weeks.  This reflects changes in glycemic control on a shorter time scale than either A1C or fructosamine levels.
Image 1: Chemical structure of 1, 5 Anhydroglucitol
An assay is available and marketed as GlycoMark. The test is applicable to both serum and plasma samples. General normal ranges are from 11-24 mcg/ml. The assay has some limitations with individuals who have renal failure, Fanconi syndrome, liver disease (cirrhosis), those using a high soy diet or certain herbal medicines. However, it is unaffected by changes in hemoglobin, bilirubin or red blood cell life span which do affect the Hemoglobin A1C assay.
1,5 – AG has been considered potentially useful for monitoring glycemic excursions or glucose variability since it measures shorter-term changes in glycemic control than A1C or fructosamine. A1C measurement cannot differentiate between fasting and postprandial glucose levels.
Measurement of post-prandial glucose is considered important as a means of monitoring changes in diet and/or pharmacologic therapies. Since post-prandial glucose becomes an important contributor to glycemic control as A1C approaches 8%, and 1, 5-AG is a predictor of glycemic excursions, 1,5-AG could potentially offer an additional tool for assessing post-prandial glucose control. , 
Postprandial hyperglycemia is seen even in Type 1 patients with good glycemic control. A recent study among children with Type 1 DM indicated the utility of 1,5-AG in conjunction with A1C to evaluated therapy for DM and glucose control along with ability to target postprandial hyperglycemia.
According to various studies conducted, 1, 5-AG could serve to assist in evaluating glycemic control over shorter intervals than the 3 month response time of A1C. It could therefore provide earlier warning of worsening glycemic control and guide earlier therapeutic intervention. It is however unlikely that 1, 5-AG could replace A1C and there are few data or studies to indicate that the test reflects risk of microvascular or macrovascular complications.
A study undertaken in 2008 compared the relationship between changes in A1C and 1,5-AG excursions. The INITIATE Study comprised 233 patients with Type 2 DM who were randomized to BIAsp 30 or Insulin Glargine (IGlar). Patients were also taking Metformin between 1500-2550 mg per day. Assays were performed at baseline, week 12 and week 28. Irrespective of the treatment arm, serum levels of 1,5-AG were strongly related to A1C levels. 1,5-AG levels fell during hyperglycemia and increase when glycemic control is reestablished. The suggestion is that 1,5-AG may be useful as an independent marker of glycemic control. Since 1,5-AG may have utility as a measure of short term glycemic changes, this may enable physicians to select and adjust therapeutic treatments to minimize glycemic variability.
Disparity in 1,5-AG levels
An additional concern with markers of glycemic control is disparities among ethnic groups.
The DURABLE trial conducted from 2005-2007 enrolled 2094 patients in 243 sites throughout the US, North America, Europe and Asia. It evaluated A1C and 1,5-AG among the various racial and ethnic groups. It was noted that non-Causcasian individuals has generally higher baseline A1C and more variability of 1,5-AG was seen in these groups, particularly Asian and African patients. Asian patients were noted to have higher 1,5-AG levels despite higher PPG (postprandial glucose) levels. Several etiologies have been proposed but none have been delineated at present. Possible de novo synthesis of 1,5-AG; differences in renal metabolism of 1,5-AG along with differences in diet all are being investigated.
A recent study in the US investigated racial disparities in glycemic markers was conducted in participants of the ARIC (Atherosclerosis Risk in Communities). A1C and 1,5-AG levels were measured in both white and black participants. The results indicate that black individuals has significantly higher A1C levels and lower 1,5-AG levels than white individuals. It was noted that fasting glucose concentrations were similar in both groups. Thus, the implication is that the disparities noted might be driven by the differences in nonfasting glycemia which is reflected in levels of 1,5-AG. Additional studies are needed for further delineate this disparity and its possible mitigating factors.
Glycomark and SGLT-2 Inhibtors
There is a potential interference with SGLT-2 Inhibitor therapy. Glycomark levels can be falsely low and its use in individuals utilizing this therapy must be carefully considered. Further studies are continuing.
At present, 1, 5- AG is not widely utilized. However as more clinical data is forthcoming with regard to its utility in determining glucose control and variability, its use may broaden.