Field of the Invention
The present invention relates generally to an NMR analysis of in vitro
blood plasma or serum biosamples from a subject.
Background of the Invention
Type 2 diabetes mellitus (T2DM or "diabetes") is one of the most costly and burdensome chronic diseases in the U.S. and other countries. The defining feature of T2DM is hyperglycemia, a reflection of impaired carbohydrate (glucose) utilization resulting from a defective or deficient insulin secretory response. T2DM is a late manifestation of metabolic derangements that begin many years earlier. Its cause is believed to be a progressive increase in insulin resistance coupled with deteriorating β-cell function. So long as the pancreatic β-cells are able to secrete enough insulin to compensate for the progressive resistance of target tissues to insulin's hypoglycemic effects, the patient is able to maintain normal fasting glucose levels. Hyperglycemia and the transition to T2DM occur as a consequence of progressive β-cell dysfunction which leads to failure to maintain hypersecretion of insulin in the face of increasing insulin resistance.
Type 2 diabetes has been traditionally diagnosed by the detection of elevated levels of glucose (sugar) in the blood (hyperglycemia). While hyperglycemia defines diabetes, it is a very late stage development in the chain of events that lead from insulin resistance to full-blown diabetes. Accordingly, it would be desirable to have a way of identifying whether or not a subject is at risk for developing Type 2 diabetes (i.e.,
is predisposed to the condition) prior to the development of the classic symptoms, such as hyperglycemia. Earlier detection of indicators of the disease (e.g., detection before glucose levels are elevated enough to be considered hyperglycemia) may lead to more effective treatment of the disease, if not actual prevention of the onset of the disease.
The most direct and accurate methods for assessing insulin resistance are laborious and time-consuming, and thus impractical for clinical application. The "gold standard" among these research methods is the hyperinsulinemic euglycemic clamp, which quantifies the maximal glucose disposal rate (GDR, inversely proportional to insulin resistance) during the clamp. Another arduous research method which is somewhat less reproducible (CV 14-30%) is the frequently sampled intravenous glucose tolerance test (IVGTT) with minimal model analysis, which measures insulin sensitivity (Si
), the inverse of insulin resistance.
Risk of progression to Type 2 diabetes is currently assessed primarily by fasting glucose, with concentrations 100-125 mg/dL defining a high-risk "pre-diabetes" condition and for which T2DM is currently defined in patients having fasting plasma glucose levels at 126 mg/dL and above. However, the actual risk of individual patients with pre-diabetes (those at greatest risk of developing T2DM in the near future) varies widely.
NMR spectroscopy has been used to concurrently measure low density lipoproteins (LDL), high density lipoproteins (HDL), and very low density lipoproteins (VLDL), as LDL, HDL and VLDL particle subclasses from in vitro
blood plasma or serum samples. See, U.S. Patent Nos. 4,933,844
. U.S. Patent No. 6,518,069 to Otvos et al.
describes NMR derived measurements of glucose and/or certain lipoprotein values to assess a patient's risk of developing T2DM.
Generally stated, to evaluate the lipoproteins in a blood plasma and/or serum sample, the amplitudes of a plurality of NMR spectroscopy derived signals within a chemical shift region of NMR spectra are derived by deconvolution of the composite methyl signal envelope to yield subclass concentrations. The subclasses are represented by many (typically over 60) discrete contributing subclass signals associated with NMR frequency and lipoprotein diameter. The NMR evaluations can interrogate the NMR signals to produce concentrations of different subpopulations, typically seventy-three discrete subpopulations, 27 for VLDL, 20 for LDL and 26 for HDL. These sub-populations can be further characterized as associated with a particular size range within the VLDL, LDL or HDL subclasses.
An advanced lipoprotein test panel, such as the LIPOPROFILE® lipoprotein test, available from LipoScience, Raleigh, N.C., has typically included a total high density lipoprotein particle (HDL-P) measurement (e.g., HDL-P number) that sums the concentration of all the HDL subclasses and a total low density lipoprotein particle (LDL-P) measurement that sums the concentration of all the LDL subclasses (e.g., LDL-P number). The LDL-P and HDL-P numbers represent the concentration of those respective particles in concentration units such as nmol/L. LipoScience has also developed a lipoprotein-based insulin resistance and sensitivity index (the "LP-IR™" index) as described in U.S. Patent No. 8,386,187
Despite the foregoing, there remains a need for evaluations that can predict or assess a person's risk of developing type 2 diabetes before the onset of the disease.
The invention concerns a method of evaluating a subject's risk of developing type 2 diabetes, comprising: performing Nuclear Magnetic Resonance measurements of at least one lipoprotein component, at least one branched chain amino acid, and GlycA from an in vitro
blood plasma or serum biosample from the subject, wherein GlycA is a composite Nuclear Magnetic Resonance signal from carbohydrate portions of acute phase reactant glycoproteins containing N-acetylglucosamine and/or N-acetylgalactosamine moieties that is detectable at about 2.00 ppm when Nuclear Magnetic Resonance measurements are performed at about 47 degrees C; and performing a computer-implemented calculation of a diabetes risk index for the subject using a mathematical model of risk of developing type 2 diabetes that includes the Nuclear Magnetic Resonance measurements of the at least one lipoprotein component, the at least one branched chain amino acid, and GlycA. The invention also concerns a system as defined in claim 12. The other claims specify preferred embodiments.
Further features, advantages and details of the present invention will be appreciated by those of ordinary skill in the art from a reading of the figures and the detailed description of the preferred embodiments that follow, such description being merely illustrative of the present invention. Features described with respect with one embodiment can be incorporated with other embodiments although not specifically discussed therewith. That is, it is noted that aspects of the invention described with respect to one embodiment, may be incorporated in a different embodiment although not specifically described relative thereto. That is, all embodiments and/or features of any embodiment can be combined in any way and/or combination, as far as such a combination is covered by the appended claims.
Brief Description of the Figures
Figure 1A is a graph showing a 5 year conversion risk of developing T2DM based on fasting glucose levels (mg/dl) and first and fourth quartiles (Q1, Q4, respectively) of a diabetes risk index according to embodiments of the present invention. The data in the table represent the percent of subjects in the 1st and 4th quartiles of six glucose subgroups that convert to type 2 diabetes over a 5 year period.
Figure 1B is a graph showing a diabetes conversion rate (%) to diabetes based on MESA with upper and lower DRI scores at different glucose ranges according to embodiments of the present invention.
Figure 2A is a schematic illustration of different lipoprotein subclass populations and exemplary size groupings according to embodiments of the present invention.
Figure 2B is a schematic illustration of different lipoprotein subclass components that provide positive and negative risk associations such as those used to assess insulin resistance and CHD risk according to embodiments of the present invention.
Figure 2C is a table of HDL subclasses H1-H26 with subpopulations grouped to optimize risk association with T2DM for intermediate risk patients (patients having glucose between high and low levels) according to embodiments of the present invention.
Figure 3A shows the diabetes conversion rates (%) for MESA study participants grouped into 4 fasting glucose categories. 411 subjects out of the 4985 individuals in the MESA study population converted to diabetes during the 6-year follow-up period. The dotted line divides the study population into those with prediabetes, defined by a fasting glucose level >100 mg/dL, and those with normal glucose (≤ 100 mg/dL).
Figure 3B shows the diabetes conversion rates (%) for MESA study participants grouped into 3 fasting glucose categories. 411 subjects out of the 4985 individuals in the MESA study population converted to diabetes during the 6-year follow-up period. The dotted lines divide the study population into those with low-risk glucose (<90 mg/dL), intermediate-risk glucose (90-110 mg/dL), and high-risk glucose (>110 mg/dL) according to embodiments of the present invention.
Figure 4A is a graph showing the diabetes conversion rates (%) for MESA subjects with high (5th quintile) and low (1st quintile) DRI scores within each of 4 glucose subgroups, from a study in which 411 out of 4985 total MESA participants converted to diabetes during a 6-year follow-up period, according to embodiments of the present invention.
Figure 4B is a graph showing the diabetes conversion rates (%) for MESA subjects with high (top decile) and low (bottom decile) DRI scores within each of 4 glucose subgroups, from a study in which 411 out of 4985 total MESA participants converted to diabetes during a 6-year follow-up period, according to embodiments of the present invention.
Figure 5 is a graph illustrating diabetes risk associations for 9 different size groupings or sub-populations of the 26 HDL subpopulations with three boxes of further groupings of selected HDL subclasses according to embodiments of the present invention. The χ2 values from the logistic regression model indicate the strengths and signs of the risk associations as determined in the MESA study population during 6 years of follow-up among 4968 MESA participants with 411 incident cases of diabetes diagnosed (all 9 subpopulations were included in the same logistic regression model, adjusted for age, gender, race, and glucose) according to embodiments of the present invention.
Figure 6 is a table of DRI prediction model parameters with statistical measures of relevance for an intermediate risk glucose subgroup (e.g., FPG between 90-110 mg/dL) as contemplated by embodiments of the present invention.
Figure 7 is a graph showing the incremental prediction of incident diabetes in MESA (411 out of 4968 participants converting to diabetes during 6 years) beyond that given by age, gender, race, and glucose level, as quantified by the LR χ2 statistic, for 4 different logistic regression models that include, in addition to age, gender, race, and glucose, the variables listed below each of the data bars.
Figure 8 is a flow chart of exemplary operations that can be used to assess risk of developing T2DM according to embodiments of the present invention.
Figure 9 is an NMR spectrum showing inflammation markers in the plasma NMR spectrum (N-acetyl methyl signals from glycosylated acute phase proteins) associated with defined NMR markers, GlycA and GlycB, respectively, according to embodiments of the present invention.
Figures 10A is an example of a fitting function/deconvolution model that uses four Valine (quartet) signals to calculate NMR measures of Valine according to embodiments of the present invention.
Figure 10B is an expansion of the plasma NMR spectrum containing methyl signals from lipoproteins and branched-chain amino acids according to embodiments of the present invention.
Figure 10C shows the full NMR spectrum containing methyl signals from lipoproteins and branched-chain amino acids with an expansion showing the location of signals from the noted metabolites according to embodiments of the present invention.
Figure 11A is an NMR spectrum showing glucose signal as multiplets at several locations according to embodiments of the present invention.
Figure 11B is a region of the blood plasma proton NMR spectrum containing glucose peaks according to embodiments of the present invention.
Figures 12A and 12B are schematic illustrations of the chemical structures of the carbohydrate portion of N-acetylglycosylated proteins showing the CH3 group that gives rise to the GlycA NMR signal.
Figures 13A and 13B are schematic illustrations of the chemical structures of the carbohydrate portion of N-acetylneuraminic acid modified glycoproteins showing the CH3 group that gives rise to the GlycB NMR signal.
Figure 14A is a graph showing an expanded section of the plasma NMR spectrum containing the signal envelope from the plasma lipoproteins and the underlying GlycA and GlycB signals,
Figures 14B and 14C are graphs of the NMR spectral region shown in Figure 14A illustrating deconvolultion models to yield NMR signal for measurement of GlycA and GlycB.
Figure 14D is a table of different components in a GlycA/B deconvolution model.
Figure 14E is an NMR spectrum showing metabolite A present in a sample at typical normal (low) concentration according to embodiments of the present invention.
Figure 14F is an NMR spectrum showing metabolite A present in a sample at an elevated (high) concentration according to embodiments of the present invention.
Figures 15A-15D are graphs of the GlycA NMR spectral region illustrating spectral overlap from lipoprotein signals (particularly from VLDL/Chylos) for samples with high TG (triglycerides).
Figure 16A is a table of different measures of GlycA concentration, depending on a protein component used in the deconvolution (e.g., "fitting") model.
Figures 16B-16D illustrate the GlycA and GlycB "fits" (deconvolution) of the same plasma sample using deconvolution models with different protein components (#1-#3 in the table in Figure 16A).
Figure 17 is a schematic screen shot of the deconvolution of a 10 mmol/L reference sample of N-acetylglucosamine, used to generate a conversion factor relating GlycA and GlycB signal areas to glycoprotein N-acetyl methyl group concentrations.
Figure 18A is a flow diagram of an NMR Valine test protocol according to embodiments of the present invention.
Figure 18B is a flow chart of exemplary pre-analytical processing that can be used prior to obtaining NMR signal of biosamples according to embodiments of the present invention.
Figure 18C is a flow diagram of operations that can be used to evaluate Valine using NMR according to embodiments of the present invention.
Figure 19 is a chart of prospective associations of hs-CRP and NMR-measured GlycA and NMR-measured Valine levels with various disease outcomes in MESA (n=5680) according to embodiments of the present invention.
Figure 20 is a chart of characteristics of MESA subjects by NMR measured GlycA quartile (in "NMR signal area units") according to embodiments of the present invention.
Figure 21 is a schematic illustration of a system for analyzing a patient's predictable risk using a DRI risk index module and/or circuit using according to embodiments of the present invention.
Figure 22 is a schematic illustration of a NMR spectroscopy apparatus according to embodiments of the present invention.
Figure 23 is a schematic diagram of a data processing system.
Figure 24 is a flow chart of exemplary operations that can be used to assess a risk of developing T2DM in the future and/or having prediabetes, according to embodiments of the present invention.
Figure 25A is an example of a patient report that includes a GlycA measurement and/or a diabetes risk index.
Figure 25B is another example of a patient report with a visual (typically color-coded) graphic summary of a continuum of risk from low to high according to embodiments of the present invention.
Figure 26 is a prophetic example of a graph of DRI versus time that can be used to monitor change to evaluate a patient's risk status, change in status, and/or clinical efficacy of a therapy or even used for clinical trials or to contradict planned therapies and the like according to embodiments of the present invention.
Figures 27A and 27B are graphical patient/clinical reports of % risk of diabetes versus FPG level and DRI score and risk pathway. Figure 27A shows patient #1's score while Figure 27B shows patient #1's score in comparison with a lesser risk patient (patient number 2) having the same FPG. While each patient has the same FPG, they have different metabolic issues identified by the DRI scores stratifying risk according to embodiments of the present invention.
Figures 28A- 28C are graphical patient/clinical reports of diabetes conversion rate (%) versus FPG level and DRI score (high DRI, Q4 and low DRI, Q1) according to embodiments of the present invention. Figure 28A is for a 4-year risk of conversion to diabetes. Figure 28B is a 5-year risk of conversion and Figure 28C is a 6 year risk of conversion.
Figure 29 is a graphical patient/clinical report of a Q4/Q1 relative risk of diabetes conversion (1-8) versus FPG level and DRI score for both 6-year (upper line) and 2-year conversions periods.
Figure 30 is a graphical patient/clinical report of log scale 5 year conversion with a diabetes conversion rate (%) versus FPG level and DRI score (high DRI, Q4 and low DRI, Q1) color coded from green, yellow, pink/orange to red and with a legend textually correlating the risk as very high, high, moderate and low.
Figure 31 is a graphical patient/clinical report of 5 year conversion to diabetes with a diabetes conversion rate (%) versus FPG level and DRI score (high DRI, Q4 and low DRI, Q1) color coded from green, yellow, pink/orange to red and with a legend textually correlating the risk as very high, high, moderate and low.
Figure 32 is a table of data that show the performance of DRI in the IRAS dataset, the MESA dataset, and IRAS dataset (using the glucose subgroups from MESA).
Figure 33 shows the performance of DRI (w/o glucose) in the same dataset criteria as Figure 32.
The foregoing and other objects and aspects of the present invention are explained in detail in the specification set forth below.
Detailed Description of Embodiments of the Invention
The present invention now is described more fully hereinafter with reference to the accompanying drawings, in which embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
Like numbers refer to like elements throughout. In the figures, the thickness of certain lines, layers, components, elements or features may be exaggerated for clarity. Broken lines illustrate optional features or operations unless specified otherwise.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items. As used herein, phrases such as "between X and Y" and "between about X and Y" should be interpreted to include X and Y. As used herein, phrases such as "between about X and Y" mean "between about X and about Y." As used herein, phrases such as "from about X to Y" mean "from about X to about Y."
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the specification and relevant art and should not be interpreted in an idealized or overly formal sense unless expressly so defined herein. Well-known functions or constructions may not be described in detail for brevity and/or clarity.
It will be understood that, although the terms first, second, etc. may be used herein to describe various elements, components, regions, layers and/or sections, these elements, components, regions, layers and/or sections should not be limited by these terms. These terms are only used to distinguish one element, component, region, layer or section from another region, layer or section. Thus, a first element, component, region, layer or section discussed below could be termed a second element, component, region, layer or section without departing from the teachings of the present invention. The sequence of operations (or steps) is not limited to the order presented in the claims or figures unless specifically indicated otherwise.
The term "programmatically" means carried out using computer program and/or software, processor or ASIC directed operations. The term "electronic" and derivatives thereof refer to automated or semi-automated operations carried out using devices with electrical circuits and/or modules rather than via mental steps and typically refers to operations that are carried out programmatically. The terms "automated" and "automatic" means that the operations can be carried out with minimal or no manual labor or input. The term "semi-automated" refers to allowing operators some input or activation, but the calculations and signal acquisition as well as the calculation of the concentrations of the ionized constituent(s) are done electronically, typically programmatically, without requiring manual input.
The term "about" refers to +/- 10% (mean or average) of a specified value or number.
The term "prediabetes" refers to a risk state for a patient or subject rather than a disease state. Thus, the term "prediabetes" refers to someone that has not been diagnosed with type 2 diabetes and, as currently defined by the American Diabetes Association, is associated with individuals that have a fasting plasma glucose level that is between 100 and 125 mg/dL, an oral glucose tolerance test level that is between 140-199 (mg/dL) or an A1C percent that is between 5.7 to 6.4 as represented in Table 1
below (the greater the level, the higher the risk of type 2 diabetes for each type of test).
Table 1: Blood Test Levels for Diabetes and Prediabetes
| ||A1C (percent)||Fasting Plasma Glucose (mg/dL)||Oral Glucose Tolerance Test (mg/dL)|
||6.5 or above
||126 or above
||200 or above|
||5.7 to 6.4
||100 to 125
||140 to 199|
||99 or below
||139 or below|
|Definitions: mg = milligram, dL = deciliter|
For all three tests, within the prediabetes range, the higher the test result, the greater the risk of diabetes
Embodiments of the invention may be particularly suitable to stratify risk for patients having the same or similar fasting glucose levels. See, e.g., Figures 1A
. Generally stated, it is contemplated that a diabetes risk index score can be used to stratify risk for developing type 2 diabetes in the future alone or with FPG or other measure of glucose such as A1C (a non-fasting sample using hemoglobin A1C) or oral glucose tolerance measurements. The diabetes risk score can stratify type 2 diabetes risk for patients having the same glucose level, but different underlying metabolic situations.
Embodiments of the invention can evaluate a patient's risk of having or developing type 2 diabetes in the future. The risk may be generated with respect to any suitable timeline, typically stated as within a 5-25 year time frame, more typically within about a 5 year or a 6 year time frame using one or more defined models of conversion to type 2 diabetes in the future using a plurality of risk model parameters.
Embodiments of the invention provide new biomarkers that can stratify risk of developing type 2 diabetes in the future for patients in an intermediate risk category.
The term "patient" is used broadly and refers to an individual that provides a biosample for testing or analysis.
The term "GlycA" refers to a new biomarker that is derived from a measure of composite NMR signal from carbohydrate portions of acute phase reactant glycoproteins containing N-acetylglucosamine and/or N-acetylgalactosamine moieties, more particularly from the protons of the 2-NAcGlc and 2-NAcGal methyl groups. The GlycA signal is centered at about 2.00 ppm in a plasma NMR spectrum at about 47 degrees C (+/-0.5 degrees C). The peak location is independent of spectrometer field but may vary depending on analysis temperature of the biosample and is not found in urine biosamples. Thus, the GlycA peak region may vary if the temperature of the test sample varies. The GlycA NMR signal may include a subset of NMR signal at the defined peak region so as to include only clinically relevant signal contributions and may exclude a protein contribution to the signal in this region as will be discussed further below.
The term "GlycB" refers to a new biomarker that is derived from a measure of composite NMR signal from the carbohydrate portions of acute phase reactant glycoproteins containing N-acetylneuraminic acid (sialic acid) moieties, more particularly from the protons of the 5-Nacetyl methyl groups. The GlycB signal is centered at about 2.04 ppm in the plasma NMR spectrum at about 47 degrees C. The peak location is independent of spectrometer field but may vary depending on analysis temperature of the biosample. Thus, the GlycB peak region may vary if the temperature of the test sample varies (and is not found in urine samples).
As used herein, the chemical shift locations (ppm) refer to NMR spectra referenced internally to CaEDTA signal at 2.519 ppm. Thus, the noted peak locations discussed and/or claimed herein may vary depending on how the chemical shift is generated or referenced as is well known to those of skill in the art. Thus, to be clear, certain of the described and/or claimed peak locations have equivalent different peak locations in other corresponding chemical shifts as is well known to those of skill in the art.
The term "biosample" refers to in vitro
blood, plasma, serum, CSF, saliva, lavage, sputum, or tissue samples of humans or animals. The invention is based on an evaluation of human blood plasma or serum biosamples. The blood plasma or serum samples may be fasting or non-fasting. Where glucose is measured by NMR, the biosample is typically fasting blood plasma or serum samples. However, glucose may be measured by any suitable means.
The terms "population norm" and "standard" refer to values defined by a large study or studies such as the Framingham Offspring Study or the Multi-Ethnic Study of Atherosclerosis (MESA) or other study having a large enough sample to be representative of the general population. However, the instant invention is not limited to the population values in MESA as the presently defined normal and at-risk population values or levels may change over time. Thus, a reference range associated with values from a defined population in risk segments (e.g., quartiles or quintiles) can be provided and used to assess elevated or reduced levels and/or risk of having a clinical disease state.
The term "clinical disease state" is used broadly and includes an at-risk medical condition that may indicate medical intervention, therapy, therapy adjustment or exclusion of a certain therapy (e.g., pharmaceutical drug) and/or monitoring is appropriate. Identification of a likelihood of a clinical disease can allow a clinician to treat, delay or inhibit onset of the condition accordingly.
As used herein, the term "NMR spectral analysis" means using proton (1
H) nuclear magnetic resonance spectroscopy techniques to obtain data that can measure the respective parameters present in the biosample, e.g., blood plasma or blood serum. "Measuring" and derivatives thereof refers to determining a level or concentration and/or for certain lipoprotein subclasses, measuring the average particle size thereof. The term "NMR derived" means that the associated measurement is calculated using NMR signal/spectra from one or more scans of an in vitro
biosample in an NMR spectrometer.
The term "downfield" refers to a region/location on the NMR spectrum that pertains to the left of a certain peak/location/point (higher ppm scale relative to a reference). Conversely, the term "upfield" refers to a region/location on the NMR spectrum that pertains to the right of a certain peak/location/point.
The terms "mathematical model" and "model" are used interchangeably and when used with "DRI", "diabetes risk index", or "risk", refer to a statistical model of risk used to evaluate a subject's risk of developing type 2 diabetes in the future, typically within 2-7 years. The risk model can be or include any suitable model including, but not limited to, one or more of a logistic model, a mixed model or a hierarchical linear model. The risk models can provide a measure of risk based on the probability of conversion to type 2 diabetes within a defined time frame, typically within 5-7 years. The risk models may be particularly suitable for providing risk stratification for patients having "intermediate risk" associated with slightly to moderatley elevated glucose values associated with prediabetes ranges. The DRI risk model can stratify a risk of developing T2DM as measured by standard χ2 and/or p values (the latter with a sufficiently representative study population).
Examples can include biomarkers that link to diabetic pathophysiology, including two or more of: insulin resistance, impaired β-cell function, inflammation and defective non-insulin (NI) dependent glucose uptake as shown in Table 2.
TABLE 2: BIOMARKERS/PATHOPHYSIOLOGY
|Defective NI Glucose Uptake
The role of HDL is complex and HDL-C is considered to be a relatively crude biomarker. Recently, researchers have suggested that HDL is an active player in diabetic pathophysiology rather than a bystander. See, Drew et al., The Emerging Roles of HDL in Glucose Metabolism, Nat. Rev., Endocrinol., 8, 237-245 (2012) published online 24 Jan. 2012
. The proposed HDL biomarkers indicated in the table above refer to defined subpopulations of HDL which can include negative and positive T2DM risk association as will be discussed further below. In some particular embodiments, DRI scores can be generated using DRI risk models that have a plurality of HDL components that represent each of the four different pathophysiologies recognizing that HDL subclasses play different roles in a person's risk of developing diabetes.
Inflammation can be associated with many different disease states including, but not limited to T2DM and CHD. It is also believed that inflammation may modulate HDL functionality. See, e.g., Fogelman, When Good Cholesterol Goes Bad, Nature Medicine, 2004
. Carbohydrate components of glycoproteins can perform biological functions in protein sorting, immune and receptor recognition, inflammation and other cellular processes.
The DRI model(s) can include at least one inflammatory marker, at least one lipoprotein component and at least one other defined metabolite or biomarker. In some embodiments, the DRI models include at least one interaction parameter.
The term "interaction parameter" refers to at least two different defined parameters combined (multiplied) as a mathematical product and/or ratio. Examples of interaction parameters include, but are not limited to, medium HDL-P/total HDL-P, (HMDM
)( GlycA), (HMDM
)(HZ), and a ratio of GlycA to medium HDL-P or HMDM
The term "LP-IR™" refers to an insulin resistance score that rates a subject's insulin sensitivity from insulin sensitive to insulin resistant using a summation of risk scores associated with different defined lipoprotein components. See, e.g., U.S. Patent No. 8,386,187
for a detailed discussion of the LP-IR score. Generally stated, large VLDL, VLDL size, and small LDL have a positive risk association while large HDL, LDL size and HDL size have a negative association (Figure 2B).
The term "lipoprotein component" refers to a component in the mathematical risk model associated with lipoprotein particles including size and/or concentration of one or more subclasses (subtypes) of lipoproteins. Lipoprotein components can include any of the lipoprotein particle subclasses, concentrations, sizes, ratios and/or mathematical products (multiplied) of lipoprotein parameters and/or lipoprotein subclass measurements of defined lipoprotein parameters or combined with other parameters such as GlycA.
The DRI mathematical models can use other clinical parameters such as gender, age, BMI, whether on hypertension medicine and the like.
 Figure 1A
is a graph that presents diabetes conversion rates of subjects within 6 glucose subgroups. The dotted line connects the data points giving the diabetes conversion rates for all subjects in each of the 6 glucose subgroups. The other data points within each glucose subgroup give the conversion rates for those subjects in the upper and lower quartiles of the DRI score. As shown, the risk of developing diabetes at any given glucose level is substantially greater for DRI index values in Q4 vs Q1. Notably, the diabetes risk index is a better predictor than glucose alone because it effectively stratifies risk at any given glucose level, without requiring any additional clinical information about the patient.
 Figure 1B
is a graph that illustrates the diabetes risk stratification of subjects within each of 4 fasting glucose ranges who have DRI scores ranging from high (DRI=10; top decile) to low (DRI=1; bottom decile). In this embodiment, the DRI scores were calculated using a multivariable risk equation that used the beta-coefficients derived from the logistic regression model shown in Figure 6
applied to the MESA subpopulation having glucose levels from 90-110 mg/dL. The 5 parameters included in the DRI risk equation were LP-IR, HM x HZ (where HZ refers to HDL size), GlycA, GlycA x HM, and Valine. In this embodiment, glucose was not included as one of the terms in the risk equation that generated the DRI score, but was included in the regression model from which the beta-coefficients were calculated for the 5 DRI parameters.
A graph such as shown in Figure 1B
can be given to a clinician or produced by an APP (e.g., an application program on a smartphone or electronic notebook) or other electronic program to help determine the actual degree of risk of a patient by taking into account both the patient's DRI score and a previously known or provided glucose level. The glucose level can be provided using glucose measurements corresponding to FPG, A1C, or glucose from an oral glucose tolerance test (Table 1).
A person's risk of developing T2DM can be presented as a DRI index score with respect to a defined range of risk, from low, intermediate and high risk. The "index" can be a simple guide or predictor of a person's risk status. The diabetes risk index is generated from a statistically validated mathematical model of risk that can characterize a subject's risk of developing T2DM in a future timeframe, in a range of from low (e.g., less likely) to high (more likely) relative to a population norm. The "low" value can be associated with DRI values that are in the lower half of a population norm, typically with a 1st
quartile or 1st
quintile. High risk DRI values can be associated with DRI values in a fourth quartile or fifth quintile or third tertile or even a 9th
decile of a population norm and indicates a high likelihood of converting to type 2 diabetes in the future. Intermediate risk DRI values can be associated with values between the low and high ranges.
While it is contemplated that the DRI index will be particularly useful when provided as a numerical score, the risk index can be presented on a patient report in different manners. The DRI index can be provided as a result expressed numerically or alphanumerically, typically comprising a numerical score on a defined scale or within a defined range of values. For example, in particular embodiments, the DRI risk index can be provided as or include a score within a defined range, such as, for example, between 0-.1, 0-1, 0-5, 0-10, 0-24, 0-100, or 0-1000 and the like. Typically, the lowest number is associated with the least risk and the higher numbers are associated with increased risk of developing T2DM in the future, typically within 5-7 years, although over time frames may be used for some embodiments. The lower value in the range may be above "0" such as 1, 2, 3, 4 or 5 and the like, or may even be a negative number (e.g., -1, -2, -3, 4, -5 and the like). Other index examples, include, for example, alphanumeric indexes or even icons noting degrees of risk, including but not limited to, "LR1" (low risk), IR5 (intermediate risk) and "HR9" (high risk), terms such as "DRI positive", "DRI high", "DRI neutral", "DRI low", "DRI good", "DRI bad", "DRI watch" and the like.
As noted above, the diabetes risk index can be decoupled from glucose measurements. Thus, for example, a DRI can be calculated for patients as a screening test. If the DRI index (e.g., score) is in an intermediate or high range, a clinician can request a glucose measurement. The testing can also be carried out in the reverse. If a patient presents with an intermediate or high glucose measurement, a DRI test can be ordered to stratify risk and/or understand a risk trajectory based on the DRI score and the glucose level. That is, if a glucose level is low, but a DRI score is elevated, the disease progression can be relatively early and a near term risk in the next 1-6 years can be low, but a lifetime risk is still of concern. This information may warrant increased monitoring of DRI and/or glucose and/or influence therapy or lifestyle choices.
To help understand the information provided by the two different measurements, instructional guidelines and/or an electronic program can be provided to a clinician that generates a test result when both data points are supplied. The combined data evaluation can be provided as a download from a laboratory or from an offering company, such as, for example, LipoScience (Raleigh, NC). Instructional guidelines can be provided to a clinician so that the clinician can understand the risk stratification provided by the DRI score and can inform a clinician whether to order a glucose test which may be more time consuming, expensive or inconvenient for a patient. Thus, the glucose test may be ordered less often or only when a patient presents with an intermediate or high DRI risk score. An electronic risk analysis circuit can also be provided (e.g., a portal accessible via the Internet) that can generate risk information based on glucose and/or DRI scores (see, e.g.,
discussion below with respect to Figure 21).
The DRI can be generated independent of and/or without requiring concurrent glucose measurements and may be used to allow a clinician to consider what risk category a respective patient may belong to.
In other embodiments, the diabetes risk index models may include glucose as a parameter or may use a glucose measurement as a separate parameter used with a DRI score to characterize risk and/or a risk trajectory and the glucose and DRI scores can be provided to a clinician as a single test summary not requiring separate test orders on the part of the clinician.
In some embodiments, for example, where the a patient has a diabetes risk index that is in the fourth quartile or fourth or fifth quintile of a population norm, the patient can be identified as at increased risk for diabetes relative to a population norm.
In some embodiments, the diabetes risk index can be provided without a glucose measurement and/or without including such a measurement in the model. Thus a DRI score in the fourth quartile or fourth or fifth quintile, alone, can identify those at risk of developing diabetes, typically in the next 5-7 years or lesser or longer timelines as discussed above.
In some embodiments, such as where the diabetes risk index is in the third or fourth quartile or in the fourth or fifth quintile and has a glucose measurement (e.g., for FPG measurement) that is above 95 mg/dl, typically between 100-125 mg/dl, the patient can be identified as at increased risk or associated with a likelihood of developing T2DM in the future, typically within about 5-7 years but other time frames may be utilized. Where glucose measurements are used, FPG glucose measurements may be used or A1C or other glucose measurements can be used (see,
for example, Table 1).
As shown with respect to Figures 1A
, the risk of developing diabetes in the future at any given glucose level is substantially greater where the DRI value is in Q4 or Q% versus Q1 or at a score of 10 versus 1 (or other defined score ranges, depending on the model and score values employed). Thus, the DRI can be a simple NMR-based risk score that can effectively stratify risk without requiring additional clinical information.
Lipoproteins include a wide variety of particles found in plasma, serum, whole blood, and lymph, comprising various types and quantities of triglycerides, cholesterol, phospholipids, sphyngolipids, and proteins. These various particles permit the solublization of otherwise hydrophobic lipid molecules in blood and serve a variety of functions related to lipolysis, lipogenesis, and lipid transport between the gut, liver, muscle tissue and adipose tissue. In blood and/or plasma, lipoproteins have been classified in many ways, generally based on physical properties such as density or electrophoretic mobility or measures of apolipoprotein A-1 (Apo A-1), the main protein in HDL.
Classification based on nuclear magnetic resonance-determined particle size distinguishes distinct lipoprotein particles based on size or size ranges. For example, the NMR measurements can identify at least 15 distinct lipoprotein particle subtypes, including at least 5 subtypes of high density lipoproteins (HDL), at least 4 subtypes of low density lipoproteins (LDL), and at least 6 subtypes of very low density lipoproteins (VLDL), which can be designated TRL (triglyceride rich lipoprotein) V1 through V6.
As shown in Figure 2A
, current analysis methodology allows NMR measurements that can provide concentrations of 73 subpopulations with 27 VLDL, 20 LDL and 26 HDL subpopulations to produce measurements of groups of small and large subpopulations of respective groups. Figure 2A
shows one example of groupings of the lipoprotein components (such as size groupings for LP-IR™ measurements) but other size groupings may be employed for other models or model components. For example, to optimize risk association with type 2 diabetes, different size groupings of HDL subpopulations can be used as will be discussed further below.
The NMR derived estimated lipoprotein sizes, e.g., HDL-P particle sizes for H1-H26 (Figure 2C),
noted herein typically refer to average measurements, but other size demarcations may be used.
In preferred embodiments, the DRI risk progression model parameters can include NMR derived measurements of deconvolved signal associated with a common NMR spectrum of lipoproteins using defined deconvolution models that characterize deconvolution components for protein and lipoproteins, including HDL, LDL, VLDL/chylos. This type of analysis can provide for a rapid scan acquisition time of under 2 minutes, typically between about 20s- 90s, and corresponding rapid programmatic calculations to generate measurements of the model components, then programmatic calculation of one or more DRI risk scores using one or more defined risk models.
Additionally, in some embodiments, it is contemplated that CPMG (water suppression) pulse sequences can be used to suppress proteins and reveal branched chain amino acids (BCAA)s such as Valine and/or small metabolites that can be quantified when used as components in a DRI risk score model. As is known to those of skill in the art, Valine is an α-amino acid with the chemical formula HO 2
. When measured by NMR, the value can be unitless. The Valine measurement may be multiplied by a defined conversion factor to convert the value into concentration units.
 Figure 2B
illustrates examples of lipoprotein subclass groupings, including those with concentrations that can be summed to determine HDL-P and LDL-P numbers according to some particular embodiments of the present invention. It is noted that the "small, large and medium" size ranges noted can vary or be redefined to widen or narrow the upper or lower end values thereof or even to exclude certain ranges within the noted ranges. The particle sizes noted above typically refer to average measurements, but other demarcations may be used.
Embodiments of the invention classify lipoprotein particles into subclasses grouped by size ranges based on functional/metabolic relatedness as assessed by their correlations with lipid and metabolic variables. Thus, as noted above, the evaluations can measure over 20 discrete subpopulations (sizes) of lipoprotein particles, typically between about 30-80 different size subpopulations (or even more). Figure 2B
also shows these discrete sub-populations can be grouped into defined subclasses for VLDL and HDL and LDL (IDL can be combined with VLDL or LDL or as a separate category, e.g., with one of the three identified as IDL in the size range between large LDL and small VLDL).
The term "LDL-P" refers to a low density lipoprotein particle number (LDL-P) measurement (e.g., LDL-P number) that sums the concentration of defined LDL subclasses. Total LDL-P can be generated using a total low density lipoprotein particle (LDL-P) measurement that sums the concentration (µmol/L) of all the LDL subclasses (large and small) including sizes between 18-23 nm. In some embodiments, the LDL-P measurement may employ selected combinations of the LDL subclasses (rather than the total of all LDL subclass subpopulations).
As used herein, the term "small LDL particles" typically includes particles whose sizes range from between about 18 to less than 20.5 nm, typically between 19-20 nm. The term "large LDL particles" includes particles ranging in diameter from between about 20.5-23 nm. It is noted that the LDL subclasses of particles can be divided in other size ranges. For example, the "small" size may be between about 19-20.5 nm, intermediate may be between about 20.5-21.2 nm, and large may be between about 21.2-23 nm. In addition, intermediate-density lipoprotein particles ("IDL" or "IDL-P"), which range in diameter from between about 23-29 nm, can be included among the particles defined as "large" LDL (or even small VLDL). Thus, for example, the LDL subclasses can be between 19-28 nm.
The term "HDL-P" refers to a high density lipoprotein particle number (HDL-P) measurement (e.g., HDL-P number) that sums the concentration of defined HDL subclasses. Total HDL-P can be generated using a total high density lipoprotein particle (HDL-P) measurement that sums the concentration (µmol/L) of all the HDL subclasses (which may be grouped based on size into different size categories such as large, medium and small) in the size range between about 7 nm (on average) to about 14 nm (on average), typically between 7.4-13.5 nm.
HDL subclass particles typically range (on average) from between about 7 nm to about 15 nm, more typically about 7.3 nm to about 14 nm (e.g., 7. 4 nm-13.5 nm). The HDL-P concentration is the sum of the particle concentrations of the respective subpopulations of its HDL-subclasses. The different subpopulations of HDL-P can be identified by a number from 1-26, with "1" representing the lowest size subpopulation in the HDL subclass and "26" being the largest size subpopulation in the HDL subclass. Figure 2C
illustrates estimated HDL size for each HDL lipoprotein component.
For type 2 diabetes risk indexes, the HDL subpopulations can be grouped as HDM
groupings as shown in Figure 2C.
The subpopulations can be grouped into small, medium and large, for example. HSDM
can include H3-H8, HMDM
can include H9-H17, and HLDM
can include H18-H26, respective subpopulations, for example. These size categories were selected to optimize risk stratification for individuals having intermediate risk in a population norm (FPG 90-110 mg/dL). That is, the subclass groups can be selected based on a statistical analysis of study populations such as MESA and/or Framingham to determine how the various subpopulations should be grouped based on risk association with T2DM (rather than LP-IR or insulin resistance alone as described, for example, in U.S. Patent No. 8,386,187
Thus, in some embodiments, HDL can be identified as a number of discrete size components, e.g., 26 subpopulations (H1-H26) of different sizes of HDL-P ranging from a smallest HDL-P size associated with H1 to a largest HDL-P size associated with H26. Concentrations of a defined subset of the sub-populations can be calculated to generate a HMDM
value. Typically HMDM
is calculated using some or all of H9-H17 (Figure 5).
Optionally, concentrations of H3-H8 (and optionally H1-H2) can be calculated to generate HSDM
and also optionally, concentrations of some or all of H18-H26 can be calculated to generate HLDM
The term "large VLDL particles" refers to particles at or above 60 nm such as between 60-260 nm. The term "medium VLDL particles" refers to particles with sizes between 35-60 nm. The term "small VLDL particles" refers to particles between 29-35 nm. The term "VLDL-P" refers to a very low density lipoprotein particle number (VLDL-P) measurement (e.g., VLDL-P number) that sums the concentration of defined VLDL subclasses. Total VLDL-P can be generated by summing the concentrations (nmol/L) of all the VLDL subclasses (large, medium and small).
As noted above, embodiments of the present invention provide at least one Diabetes Risk Index (DRI) using one or more defined mathematical models of risk of different defined biomarkers or parameters of an in vitro
biosample of a patient to identify at-risk patients before onset of T2DM who may benefit from pharmaceutical, medical, diet, exercise or other intervention.
The DRI evaluation can be decoupled from a glucose measurement and can relatively easily be generated as a screening tool and may be able to identify at-risk individuals earlier in time than with conventional tests.
 Figure 3A
is a graph showing diabetes conversion rates (6 year) based only on subgroup assignment by fasting glucose level, from the MESA study with 411 subjects converting to diabetes. This graph reflects current categorizations of risk, e.g., low risk below 100 and high risk above 100 FPG values. Figure 3B
illustrates division into three categories of risk, again, however, using only glucose levels for this categorization. Embodiments of the invention provide DRI scores that can further stratify risk of developing type 2 diabetes within any glucose subgroup or at any particular glucose level. The new DRI evaluation may be particularly useful for those individuals in the intermediate risk range.
For example, where scores are used, e.g., a DRI score of "1" or in the first quartile or quintile (Q1) can indicate an individual that is considered low risk as shown in Figures 4A
For a DRI score at the high end of the range, e.g., at 5 (Q5 or 5th
Quartile) as shown in Figure 4A
or at "10" (which is further refined as an upper 10% of the population norm) as shown in 4B,
the individual is considered high risk. Figures IB, 4A
were generated using the model components shown in Figure 6. Figure 1A
was generated using a different DRI model using components described in EXAMPLE 1 below. The 1-10 range reflects a finer risk stratification over Figure 4A, e.g.,
the DRI values 1-10 are associated with 10% increments of the population norm instead of 20% increments.
 Figure 5
is a graph that illustrates HDL subpopulation relations with diabetes risk, showing those subpopulation groupings with positive (above the "0" line) and negative (below the "0" line) risk associations according to embodiments of the present invention. These groupings are based on associations with incident diabetes in MESA. Thus, HSDM
has a positive risk association while HMDM
have negative risk associations. It is contemplated that some of the subpopulations can be omitted from the noted groupings, e.g., those components having values close to the "0" line in Figure 5,
for example. Exemplary HSDM
size ranges were discussed above with respect to Figure 2C.
These size groupings are different from size groupings used for LP-IR, for example.
 Figure 6
is a table that illustrates exemplary DRI prediction model components according to some embodiments of the present invention. The table shows statistical evaluations of the parameters included in a logistic regression model, adjusted for age, gender and race, applied to the MESA subgroup with glucose levels between 90-110 mg/dL. There were 2038 subjects in this subgroup, and 210 converted to diabetes during the 6-year follow-up period. While not wishing to be bound to any particular theory, the table also indicates, for each parameter, a proposed pathophysiologic link to diabetes development. The parameters include glucose, LP-IR™, HMDM
multiplied by HZ, GlycA, GlycA multiplied by HMDM
and Valine. The term "HZ" refers to average HDL-size.
 Figure 7
is a chart that shows a statistical evaluation of the prediction of incident diabetes in MESA by 4 different prediction models, beyond that given by glucose (plus age, gender, and race). As indicated by the incremental LR χ2 statistic, the multi-parameter DRI prediction model on the far right provides almost double the prediction given by a model including insulin, BMI, and hs-CRP or a model with LP-IR alone.
To be clear, Figure 6
illustrates one particularly suitable set of DRI model parameters. However, the DRI model can include less than all those parameters shown or even other parameters including GlycA and/or GlycB, lipoprotein parameters and other metabolites and/or interaction parameters as will be discussed further below.
 Table 3
below summarizes some examples of lipoprotein components that may be included in the DRI model.
TABLE 3 MESA, New
|Logistic regression, models adjusted on gender, age, ethnicity||All, N=411 (4983)||Glucose 90-110 mg/dL N=210 (2038)|
||0.0003| Figure 8
is a flow chart of exemplary operations that can carry out embodiments of the present invention. As shown, at least one defined mathematical model of risk of progression to type 2 diabetes in a future time frame (e.g., 5-7 years) can be provided (block 400).
Measurements of components of the at least one mathematical model of a respective patient biosample can be obtained (block 403).
A diabetes risk index score can be calculated using the model and the measurements (block 405).
A patient risk report (paper and/or electronic) with the DRI score can be provided to desired recipients (e.g., a patient and/or clinician) (block 412).
The DRI score can stratify risk for different patients having the same glucose measurement (block 404).
The measurements can include NMR measurements of lipoprotein subclasses and either or both GlycA and Valine (block 405).
The measurements can include one or more of HMDM
, VLDL size, ratio of medium HDL-P to total HDL-P and/or sum of VLDL-P (block 408).
A graph of risk of conversion versus glucose level and comparative Q1 and/or Q4/Q5 or upper decile references can be provided as a relative risk summary (block 413).
A graph of risk of conversion versus glucose level and comparative patient risk with different diabetes risk score measurement and/or different glucose measurement can be provided (block 414).
The risk summary can provide a relative comparison to defined population norms.
In some embodiments, values of some or all of the different parameters of the DRI risk model(s) can be derived from a single nuclear magnetic resonance (NMR) spectrum of a biosample, typically a fasting blood plasma or serum sample.
The DRI models can include lipoprotein subclass parameters and GlycA. In some preferred embodiments, the DRI mathematical model also includes the branched-chain amino acid Valine. Optionally, glucose may also be considered as a risk parameter in the DRI model. Where used, a patient's glucose measurement may also be obtained from the NMR spectrum of the biosample or may be obtained in other conventional manners.
The DRI model(s) can incorporate lipoprotein subclass parameters that are known to be more or less drug-sensitive. The DRI mathematical model(s) can include gender as a variable or may be configured as different models for different genders. The DRI models for a respective patient can be electronically adjusted or selected depending on one or more factors associated with the patient, e.g., gender of the patient, age of the patient, whether the patient is on a certain type of medication and the like.
In some embodiments, gender may be a factor in the diabetes risk index model. In some embodiments, age may be a factor in the diabetes risk index model. In other embodiments, the diabetes risk index can exclude either gender or age considerations so as to avoid generating false negatives or false positives based on data corruption of such ancillary data not directly tied to a biosample, for example.
The DRI index can be generated in one or more than one way for a particular patient and two or more DRI indexes can be generated for comparison, for elective use of one or both by a clinician and/or presented as a ratio of the two. For example, the NMR data can be evaluated using a plurality of different DRI index calculation models where one model can employ lipoprotein components that are sensitive to a particular drug therapy or class of therapies (e.g., statins) and one that includes lipoprotein components that are insensitive (or immune) to such particular therapy or class of therapies.
Lipoprotein Components/Parameters Statin Sensitivity/Resistance
| ||Affected by statins - YES||Affected by statins - NO|
By way of another example the DRI risk score can be calculated in two or more ways, one using components that are more robust where non-fasting (blood plasma or serum) biosamples are analyzed and one that may have a better risk prediction value but is for fasting biosamples. For example, VLDL components may be reduced or excluded from DRI models for non-fasting samples. Table 5
lists lipoprotein components that can be affected by nonfasting.
Lipoprotein components sensitive to nonfasting.
|Lipoprotein Components||Sensitivity to Nonfasting|
Similarly, the DRI risk evaluations can be calculated using both the statin-sensitive and statin-insensitive DRI models allowing the clinician rather than the testing protocol analysis circuit to assess which applies to a particular patient. The different values can be presented in a report or on a display with a comment on the different results.
In some embodiments, only one DRI score is provided to the clinician based on data electronically correlated to the sample or based on clinician or intake lab input, e.g., fasting "F" or non-fasting "NF," and statin "S" or non-statin "NS" characterizations of the patient which data can be provided on labels associated with the biosample to be electronically associated with the sample at the NMR analyzer 22 (Figure 22).
Alternatively, the patient characterization data can be held in a computer database (remote or via server or other defined pathway) and can include a patient identifier, sample type, test type, and the like entered into an electronic correlation file by a clinician or intake laboratory that can be accessed by or hosted by the intake laboratory that communicates with the NMR analyzer 22. The patient characterization data can allow the appropriate DRI model to be used for a particular patient.
HDL-P can be in µmol/L units. VLDL-P and LDL-P can be in units of nmol/L. Valine and/or GlycA, where used, can be provided as a unitless measure of intensity or one or each can be multiplied by a respective defined conversion factor to provide the number in units of µmol/L (see, e.g., Figure 17
It is contemplated that a diabetes risk index can be used to monitor subjects in clinical trials and/or on drug therapies, to identify drug contradictions, and/or to monitor for changes in risk status (positive or negative) that may be associated with a particular drug, a patient's lifestyle and the like, which may be patient-specific.
Further discussion of diabetes risk prediction models is provided below after discussion of examples of NMR signal deconvolution methods that can be used to obtain the GlycA/GycB measurement and, where used, a Valine measurement.
 Figure 9
illustrates the resonant peak regions for GA associated with GlycA and GB associated with GlycB in the plasma NMR spectrum. One or both of these peak regions can include signal that can be defined as inflammation markers in the plasma NMR spectrum.
 Figures 10A-10C
illustrate metabolites that may be quantified. The metabolites can be quantified in NMR spectra, typically using defined deconvolution models. The DRI models may include one or more branched chain amino acids (BCAAs) including one or more of isoleucine, leucine, and Valine (as discussed above) and/or one or more NMR detectable metabolites such as lactate alanine, acetone, acetoacetic acid, and beta-hydroxybutiric acid with closely aligned calculated and measured lines (for contemplated associated deconvolution models) for each illustrated. The associated centers of peak regions for the respective metabolites isoleucine, leucine, Valine, lactate, and alanine are shown in Figure 10C
(0.90 ppm, 0.87 ppm, 1.00 ppm, 1.24 ppm, and 1.54 ppm, respectively).
It is contemplated that other metabolites, which can optionally be measured by NMR using a CPMG or other suitable pulse sequence, can include choline, phosphocholine, glycine, glycerol, alpha- and beta-hydroxybutyrate and carnitine.
Recent studies have shown the levels of branched chain amino acids (BCAAs) in serum are associated with the risk of incident diabetes. Wang et al.
showed in the Framingham longitudinal cohort that a metabolic signature composed of BCAAs as well as aromatic amino acids had a significant correlation with incident diabetes over a span of up to 12 years. See, Wang et al., Nature Med., 17, 448, 2011
. In a six month behavioral/dietary intervention study carried out on 500 obese subjects the amount of weight loss was very poorly correlated with improvements in HOMA score and lipid related factors also showed no significant correlation, but the BCAA signature was a strong predictor of improved insulin sensitivity. See, Svetkey, et al., JAMA, 299, 1139, 2008
and Shah, et al, Diabetologia, 55,321, 2012
One or more of the set of three BCAAs (valine, leucine and isoleucine) can be quantified by NMR with the use of a CPMG sequence. The term "CPMG" refers to a Carr-Purcel-Meiboom-Gill pulse sequence. This is a series of phase defined radiofrequency pulses that provide means to attenuate signals from large, rapidly relaxing molecules such as proteins and lipoprotein particles. This sequence involves a series of radiofrequency pulses in which the signals from the proteins and lipoproteins are attenuated, thereby allowing the detection of a number of additional metabolite which would be otherwise buried under the large protein/lipoprotein signals. A robust signature of all three branched chain amino acids can be obtained from a biosample with a suitable diluents in a defined percentage relative to the sample, which may be a 50:50 mixture of serum to diluents. The BCAA signal can be acquired with about 90 seconds of acquisition time using a 16 scan CMPG sequence and can be appended to a standard Lipoprofile® lipoprotein analysis for high-throughput analysis. This same 16 scan experiment also contains signals of a number of additional high clinical value metabolites including amino and organic acids.
The BCAA quantification model can include computationally derived basis functions for one or each of the three BCAAs as well as experimental protein and lipoprotein functions. In modeling the BCAA region of the NMR spectrum, an experimental and/or synthetic model may be used for a BCAA. The baseline may be modeled by CPMG processed experimental spectra of the individual protein and lipoprotein components that are known to contribute to this region. The model could also use synthetic functions such as linear, quadratic & polynomial functions to fit the baseline component. The fitting approach may utilize the Lawson-Hanson non-negative linear least fitting method to achieve the best agreement between the experimental and modeled spectra.
The NMR method can measure all three amino acids in unprocessed serum without the need for a calibration standard as would be needed in a mass-spectrometry assay. It has been found that the accurate quantitation of these analytes in unprocessed serum biosample can be affected by the shimming of the spectrometer. Clinical quantitation therefore may depend upon a more sophisticated BCAA analysis model which can simultaneously evaluate a linewidth of a defined reference peak in the spectrum, such as an EDTA, citrate peak/line or other defined reference peak. Once a linewidth is determined, a predefined correction factor or algorithm associated with a respective linewidth can be applied to generate the quantified BCAA measurement. That is, a set of defined correction factors or an algorithm that calculates the corrections/adjustments for clinical quantification can be identified for different linewidths associated with a particular reference peak and the BCAA value calculated using curve fit functions can be adjusted by the defined correction factor.
An "unprocessed biosample" as used herein refers to a biosample that, unlike sample preparation for mass spectrometry analysis, is not subjected to processing that causes the biosample to be physically or chemically altered after it is obtained (but buffers and diluents can be used). Thus, once the biosample is obtained, components from the biosample are not altered or removed. For example, once a blood serum biosample is obtained, the serum is not subjected to processing that removes components from the serum. In some embodiments, an unprocessed biosample is not subjected to filtering and/or ultrafiltration processes.
A respective NMR analyzer 22 (Figure 22)
may be configured to obtain at least 10 scans per biosample, typically between 10-256 scans, such as 16 scans, and possibly ≥ 96, such as 96 scans or 128 scans with at least about 16K data points collected over a 4400 Hz sweep width, per sample, to obtain the NMR data used to measure one or more BCAA. The BCAA scan can be carried out before or after a lipoprotein scan sequence, which typically employs a different pulse sequence to allow for quantification of lipoproteins.
 Figures 12A/12B
illustrates the chemical structure of the carbohydrate portion of N-acetylglycosylated proteins showing the CH3
group that gives rise to the GlycA NMR signal. Figures 13A/13B
illustrates the chemical structure of the carbohydrate portion of N-acetylneuraminic acid (also called sialic acid) modified glycoproteins showing the CH3
group that gives rise to the GlycB NMR signal.
 Figure 10A
is an example of a deconvolved signal shown in Figure 10B
with a quartet of Valine signals identified to generate a calculated Vc and measured Vm spectrum of the Valine peaks of Valine signal that can be used to measure Valine (in the embodiment shown, the biosample is blood plasma or serum. One or more of the quartet of peaks of Valine are located within a region between about 0.72 ppm to about 1.07 ppm. Typically, one or more of the quartet of peaks of Valine are located between 0.9 ppm and 1.03 ppm, e.g., at a center of the multiplet region at about 1.00 ppm of the downfield methyl doublets. One or more of the quartet of peaks of Valine can be used to measure Valine in the biosample. The concentration of Valine in urine is significantly higher than that in serum/plasma, and peak position and amplitude will be different.
 Figure 10B
is an enlarged region of plasma NMR spectrum (1.07-0.62 ppm) containing methyl signals from lipoproteins and branched-chain amino acids (leucine, Valine and isoleucine).
As shown in Figure 10C
,the DRI model may include one or more of isoleucine, leucine, Valine (as discussed above), lactate, and alanine with closely aligned calculated and measured lines (for contemplated associated deconvolution models) for each illustrated. The associated centers of peak regions of the respective metabolites are shown in Figure 10C
(0.90 ppm, 0.87 ppm, 1.00 ppm, 1.24 ppm, and 1.54 ppm, respectively).
 Figure 11A
is an NMR spectrum of a serum sample with a glucose spectrum shown below the upper spectrum. There is a glucose multiplet at 5.2 ppm, and glucose mulitplets centered at 4.6, 3.9, 3.8, 3.7, 3.5, 3.4 and 3.2ppm which could be used for measuring glucose using NMR derived measurements.
 Figure 11B
shows the proton NMR spectrum of blood plasma, with the two regions (region 1 and region 2) containing the signals produced by glucose indicated. In some particular embodiments, the peaks in region 1 in the range of 3.81-4.04 ppm can be used for glucose analysis according to some embodiments of the present invention. Alternatively, the peaks in region 2 in the range of 3.50-3.63 ppm can be used for the glucose analysis of the present invention. Additionally, the combination of the peaks in region 1 and region 2, may be used for the quantitative determination of glucose according to the present invention. The data points in the reference or standard spectrum and patient glucose sample spectra are aligned using a line-shape fitting process as described herein to find the "best fit," and the intensity of the standard spectrum is scaled to match the sample spectrum. The glucose concentration of the standard is multiplied by the scaling factor used to match the sample lineshape to give the glucose concentration of the blood sample. See,
e.g., U.S. Patent No. 6,518,069
for further discussion of the glucose and lipoprotein measurements for assessing risk of developing Type diabetes.
Thus, in some embodiments, a patient glucose measurement can be obtained via NMR analysis of the biosample NMR spectrum, along with lipoprotein particle measurements, GlycA and Valine measurements. However, glucose measurements, where used, can alternatively be obtained in conventional chemical ways.
It is contemplated that NMR measurements of GlycA, Valine, and lipoproteins of a single (blood/plasma) in vitro
biosample can provide important clinical information and/or further improve a prediction or evaluation of a patient or subject's risk of developing type 2 diabetes and/or having prediabetes.
 Figure 14A
illustrates an enlarged chemical shift portion of the NMR spectrum between 2.080 and 1.845 ppm as shown in Figure 9. Figure 14A
also illustrates both the calculated C signal and the measured (composite) signal envelope Cm from the allylic protons of the lipids in VLDL, LDL and HDL, with underlying deconvolved GlycA and GlycB and other resonant peaks. GlycA can include contributions from 2-NAcGlc and 2-NAcGal methyl groups. GlycB includes signal from the N-acetyl methyl groups on the sialic acid moieties of glycoproteins.
A defined lineshape GlycA mathematical deconvolution model can be used to measure the GlycA. The "composite" or measured signal envelope Cm can be deconvolved to quantify the signal contributions of GlycA and other contributing components such as lipoprotein subclass components. The deconvolution calculates signal amplitudes of the components contributing to the measured signal shapes and calculates the sum of the components. A close match between the calculated signal C and the measured signal Cm indicates the deconvolution successfully modeled the components that make up the NMR signal.
The peak region of the GlycA region GA and the peak of the GlycB region GB are shown by the peaks centered at 2.00 ppm and 2.04 ppm (at about 47 deg C sample temperature), respectively, underlying the composite (upper) envelope signal line Cm. In some embodiments, the peak regions for GlycA and GlycB can include adjacent smaller nearby signals in the deconvolution model to account for GlycA and GlycB signals of slightly different frequency.
The protein signal Ps includes "humps" or peaks PGA
that align with GA and GB, respectively. GlycA can be calculated using the difference between total plasma GlycA signal or "GA" as given by the total peak area of the plasma GlycA signal and "PGa
", that portion of GlycA that may derive from the non-inflammatory glycoproteins in the protein (d>1.21 g/L) component of plasma. The deconvolution can be carried out to subtract out the (patient/subject) variable "clinically non-informative" part of the total NMR signal at the GA region to leave the more informative disease association measure of GlycA.
Stated differently, while not being bound to any particular theory, in some embodiments, the measured GlycA signal at 2.00 ppm can be referred to as GA, the deconvolution can separate it into three parts: 1) the part contributed to by the protein (d>1.21 g/L) chosen to be largely devoid of inflammatory proteins, 2) the part contributed to by the "non-inflammatory" lipoproteins (d<1.21 g/L), and 3) the inflammatory glycoproteins (both lipoprotein and protein), the latter modeled by the overlapping Lorentzians (LGA) or other curve fit functions. The clinically informative GlycA from the deconvolution can be defined as GA minus PGA
and minus the non-inflammatory lipoprotein components = LGA. GlycB can be determined in a similar manner using the GB minus PGB
signal contribution minus the non-inflammatory lipoprotein components.
The lineshape deconvolution can be achieved with a non-negative least squares fitting program (Lawson, CL, Hanson RJ, Solving Least Squares Problems, Englewood Cliffs, NJ, Prentice-Hall, 1974
). This avoids the use of negative concentrations which can lead to error especially in low signal to noise spectra. Mathematically, a suitable lineshape analysis is described in detail for lipoproteins in the paper by Otvos, JD, Jeyarajah, EJ and Bennett, DW, Clin Chem, 37, 377, 1991
. A synthetic baseline correction function may also be used to account for baseline offsets from residual protein components. This can take the form of a quadratic or other polynomial function. Weighting factors are determined and the fit can be optimized by minimizing the root mean squared deviation between the experimental and calculated spectrum. See, e.g.,
U.S. Patent Nos. U.S. Patent Nos. 4,933,844
for a description of deconvolving composite NMR spectra to measure subclasses of lipoproteins. See also, U.S. Patent No. 7,243,030
for a description of a protocol to deconvolve chemical constituents with overlapping signal contribution.
 Figures 14B
illustrate the composite (measured) signal "Cm" of the NMR spectra of Figure 14A
with a fitting region FR
corresponding to the NMR spectrum between 2.080 and 1.845 ppm. The fitting region FR
typically comprises 315 data points but more or less may be used, such as between about 200-400 data points, for example. The GlycA quantification/deconvolution model includes VLDL/chylos components, LDL components, and HDL components.
 Table 6
shows various TRLs that may be used in an exemplary DRI model.
Characteristics of Triglyceride Rich Lipoprotein Subclasses Measured by NMR LipoProfile
|Subclass||TRL Subclass Components||NMR Chemical Shift (ppm)||Estimated Diameter (nm)|
The term "TRL V6" refers to TRL (triglyceride rich lipoprotein) particles or sub-fractions having a diameter between about 90 nm up to as much as about 170 nm, more typically having diameters between about 100-140 nm. The term "TRL V6" can also be defined with respect to the lipid methyl group NMR signal chemical shifts (ppm) corresponding to the estimated diameters as provided in Table I below.
The term "TRL V5" refers to large TRL particles having a diameter of between about 60 nm and about 80 nm (see Table 6 above for the associated NMR chemical shifts).
The terms "chylomicron" and "chylos" refer to very large TRL particles having diameters that are larger than TRL V6. As such chylomicrons reters to TRL particles or sub-fractions having a diameter between from about 170 nm up to about 260 nm (see Table 5 above for their associated NMR chemical shifts). There is not a clear demarcation between TRL V5 and TRL V6 nor between TRL V6 and chylomicrons, such that there is a distribution of particle sizes for each subgroup that overlaps in the range between about 80-90 nm for TRL V5-6 and between about 140-170 nm for TRL V6 & chylomicrons.
When the TRLs are quantified, the concentrations in particle concentration units (nmol/L) or triglyceride concentration units (mg/dL) can be expressed. Thus, for each of the different definitions of "large VLDL", either the particle concentrations or triglyceride concentrations could be used in the DRI model. Without wishing to be bound to any particular theory, based on linear regression analysis, the triglyceride concentration units may yield marginally better diabetes risk prediction.
 Figure 14D
is a table of different components in a GlycA/B deconvolution model according to embodiments of the present invention. Metabolite A is one component that can be measured in a GlycA/B deconvolution model and may be used clinically. As illustrated in Figures 14E
metabolite A can be present in a spectrum as a singlet peak and is typically present in a sample at low concentrations (Figure 14E),
but a high concentration of metabolite A may be present in a sample (Figures 7F).
A plurality of curve fitting functions for the metabolite A peak region can be used to quantitatively evaluate a level of metabolite A and/or to deconvolve the NMR spectrum for quantification of GlycA and/or GlycB, for example.
The deconvolving model components shown in Figure 14D
list a plurality of curve fit functions Glyc1-Glyc46 that can be applied to a fitting region that includes the GlycA peak region and extends to a GlycB peak region (typically with between about 40-50 curve fit functions, shown as with 46, but less or more such curve fit functions may be used, e.g., between 30-70). As will be discussed further below, the GlycA measurement can be carried out by summing values of a defined first subset of the curve fit functions, values associated with all or some of the Glycl-Glyc26 components, for example. The GlycB measurement can be carried out by summing values of a second (typically smaller) defined subset of the curve fit functions, such as some or all components between Glyc27 and Glyc 46, for example.
 Figures 15A-15D
illustrate spectral overlaps from triglyceride rich lipoproteins as the TG (triglyceride) values increase which can be challenging to reliably deconvolve in a manner that provides precise and reliable GlycA and GlycB measurements.
The GlycA/B model provides sufficient HDL, LDL and VLDL/chylos components to be able to provide a good fit of the experimental signal as indicated by a close match between calculated signal C and experimental or measured composite signal Cm. Typically, the Glyc model will have more of the closely spaced VLDL/chylos components than either LDL or HDL components as these TRL contribute more signal to the left side of the spectrum. The model can include 20-50 VLDL/chylos components, typically about 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39 or 40. In a preferred embodiment, the model includes 30 VLDL/chylos components.
The Glyc model can include a plurality "N" of (typically overlapping) curve fit components N that populate a sub-region Fs of the fitting region FR
that extends from a few data points (e.g., about 10 or less) to the right of the GlycA measurement region R1
(e.g., starting at about 1.9 ppm or higher) to at least a few data points to the left of the GlycB region R2
(and can extend to the end of the fitting region FR
to 2.080 ppm), at least where GlycB is measured or evaluated. Each component N, in this embodiment, can be a Lorentzian-shaped signal with a line width about 1.4 Hz. Also, in particular embodiments, each data point can be about 0.275 Hz apart as determined by the digital resolution of the spectrum. The tail portion of the region Fs on the left side may include more (Lorentzian) components than the tail portion on the right side. The number of components N in the region Fs n can be about 46 (e.g., about 46 Lorentzians) but more or less components "N" can be used. For example, the region Fs can include, but is not limited to, between 30-70 Lorentzians, or n=30, 35, 36. 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, or 60. The curves N are typically Lorentzian functions with line widths at half-height of between 2-10 data points (0.55-2.75 Hz at 400 MHz), more typically between 4-6 data points, and are offset from each other by a defined amount, such as, for example, 2 data points (0.55 Hz).
The GlycA and GlycB Lorentzians (or other curve fitting components N) can have the same or different numbers of data points. The GlycB Lorentzians N can have the same, less or more data points than the GlycA Lorentzians N. The Lorentzian fit components "N" can have peak line widths (LW) of about 1.4 Hz (at half height). However, other LWs can be used including, but not limited to, 1.1, 1.2, 1.3, 1.5 and the like. To be clear, data for GlycB can be omitted from the evaluation for the DRI risk index.
GlycA can be calculated using a defined subset of the number of curve fit components N that fill the entire region R1
. GlycB can be calculated using a suitable number of curve fit (e.g., Lorentzian fit) components N that fill the entire region R2
. The region R1
can be between 5-6 Hz. The GlycB region R2
can be 7-8 Hz. Optionally, the GlycA components N can be offset by 2 data points. The GlycB components N can be offset by 4 data points.
GlycA can be calculated using a sum of adjacent Lorentzian components N, typically between 9-15, such as 9, 10, 11, 12, 13, 14 and 15 components. GlycB can be the sum of adjacent Lorentzian fit components N, with the same, more, or less, typically less, than that used for GlycA measurements, such as between about 5-10 components N, typically about 7, about 8 or about 9 components. The Lorentzians between R1
are not included in the quantified measurement of either GlycA or GlycB. Figure 14B
illustrates the sum of 7 adjacent Lorentzians used to calculate the GlycB measurement and the sum of 10 (more narrow) Lorentzians can be used to calculate the GlycA measurements. Figure 14C
illustrates the sum of 9 adjacent Lorentzians used to calculate the GlycB measurement and the sum of 12 (more closely spaced) Lorentzians can be used to calculate the GlycA measurements.
The number of HDL, LDL and VLDL components may vary for the GlycA calculation. As shown, the HDL components can be 20 HDL components (spanning the range of HDL subclass diameters), but more or less can be used, e.g., between about 10-26. As shown, the number of LDL components is 9 components (representing different LDL diameters), but more or less can be used, e.g., between about 5-20. As shown, the number of VLDLs/Chylos components is 30, but more or less can be used, e.g., 25-60 of different size ranges.
To be clear, while a preferred embodiment describes the curve fit components as Lorentzian fit components, other fitting components may be used including, but not limited to, experimental N-acetyl methyl group signals or Gaussian lineshape functions. Thus, any suitable curve fit function can be used.
 Figure 16A
is a Table of different protein components (Protein 1, Protein 2 and Protein 3) that, when used in the Glyc deconvolution model, yields different GlycA concentrations and different GlycA associations with CHD events and All-Cause Death in MESA. Figures 16B, 16C
illustrate the respective protein signal Ps in the deconvolved spectrum and the differences they exhibit in the amplitudes of the signals in the GlycA and GlycB peak regions. To optimize the calculated GlycA and/or GlycB measurement, in some embodiments, the deconvolution model includes a defined protein signal component as discussed above. This protein signal component Ps is for protein other than lipoproteins, e.g., other than HDL, LDL, VLDL/chylos, e.g., and may be associated with the >1.21 g/L density fraction of plasma obtained by ultracentrifugation (which includes albumin and other non-lipoprotein proteins in plasma).
This signal component "Ps" is shown in Figures 14A-14C.
Surprisingly, although this protein signal Ps does include a peak (PGA
, respectively) aligned with the peak at the chemical shift for both GlycA and GlycB, eliminating this portion of the protein NMR signal from the deconvolution model (by, for example, digital manipulation or signal processing) was found to make the calculated GlycA and GlycB measurements relatively less clinically informative (weaker disease associations). At the other extreme, including in the deconvolution model a protein component with a relatively large signal at the GlycA and GlycB positions results in lower GlycA and GlycB concentrations that are also less clinically informative, as shown for Protein #2 and Protein #3 in Figures 16C
Thus, by selecting an appropriate protein component with an intermediate signal amplitude at the GlycA and GlycB positions, such as Protein #1 in Figure 16B,
the deconvolution model may be "tuned" to produce GlycA and GlycB concentrations that are improved and/or optimized with respect to their clinical associations with inflammation and related disease states.
Thus, in some embodiments, it is contemplated that the GlycA measurement will provide a better clinical indicator if it does not include the lipoprotein signal (accounted for in the deconvolution model with the VLDL/chylo, LDL and HDL components) and if it includes only a portion of the remaining NMR signal, e.g., it does not include all other NMR protein signal at the GlycA peak region. This subset of the NMR signal at the GlycA peak region may be more reflective of inflammatory protein activity, e.g., N-acetyl methyl signals from glycosylated acute phase proteins.
 Figure 17
is a screen shot of the deconvolution of a 10 mmol/L reference standard sample of N-acetylglucosamine, from which a conversion factor of 17.8 was determined to transform signal area concentrations of GlycA and GlycB to µmol/L glycoprotein N-acetyl methyl group concentrations. In some embodiments, according to MESA subjects, first to fourth quartile (mean) levels of GlycA can be: Q1: 21.6*17.8, Q2: 25.8*17.8, Q3: 29.3*17.8 and Q4: 35.3*17.8.
GlycA measurement precision using the model shown in Figure 14B
was shown to be good. A within-run (5 pools from 2009) analysis of lowest GlycA=40.5 (CV =2.47%) and highest GlycA =58.4 (CV=1.6%). Within-lab results from 13 pools from 2010 and 2011 had a lowest GlycA =25.6 (CV=4.08%) and highest GlycA=69.1 (CV=1.87%). These concentrations are expressed as "arbitrary units" of NMR signal areas and can be multiplied by 17.8 to convert them to µmol/L N-acetyl methyl group concentrations.
It is believed that the measured amplitude of the GlycA signal in any one sample may have the advantage of providing a more stable and "time-integrated" measure of the patient's inflammation state than is provided by measurements of hs-CRP or other individual inflammatory proteins.
As noted above, Figure 17
illustrates a conversion factor that can be used to calculate measurements of GlycA. The GlycA measurement can also be a unit less parameter as assessed by NMR by calculating an area under a peak region at a defined peak in NMR spectra. In any event, measures of GlycA with respect to a known population (such as MESA) can be used to define the level or risk for certain subgroups, e.g., those having values within the upper half of a defined range, including values in the third and fourth quartiles, or the upper 3-5 quintiles and the like.
 Figures 18A-18C
are exemplary flow diagrams of operations that can be used to obtain NMR signal associated with Valine according to embodiments of the present invention.
 Figure 18A
illustrates that a pre-analytical evaluation (block 710)
can occur before a Valine region of the NMR signal is determined (block 725),
then deconvolved (block 750). Figure 18B
illustrates an exemplary pre-analytical evaluation 810
which includes delivery verification of the sample into the flow cell as either complete failure (block 812)
or partial injection failure (block 813),
shimming verification (block 815),
temperature verification (block 817)
and a citrate tube detection (failure) (block 819),
all using defined characteristics of signal associated with a defined diluent added to the sample.
Referring again to Figure 18A,
once the defined parameters are confirmed within limits, the pre-analytical quality control analysis can end (block 720)
and the determination of the Valine region can be identified (block 725)
and the spectrum deconvolved and Valine level calculated (block 750).
Optionally, a post-analytical quality control can be electronically performed (block 755)
and the results output (block 760).
The results can be included in a test report with comments, visual indicia of high or low and the like (block 765).
Referring to Figure 18C,
NMR signal can be electronically obtained of an in vitro biosample with a defined added diluent (block 902).
The QC evaluation can be carried out (block 910).
The Valine region is determined (block 925).
The Valine region is deconvolved (block 950d)
and an NMR derived value of Valine is calculated (950c).
The diluent can comprise calcium ethylenediamine tetraacetic acid (Ca EDta) (block 903)
or other suitable diluent that creates a reliable peak and behaves in a predictable manner. Well established chemical shift or quantitation references include, for example, formate, trimethylsilylpropionate (and isotopically labeled isomers), and EDTA.
The pre-analytical quality control evaluation 810
can be based on inspection of characteristics of the CaEDTA reference peak and the system or processor can be configured not to perform the Valine test unless the NMR spectra have been acquired under specified conditions such as those shown in Figure 18B.
The sample temperature can be 47 ± 0.5 °C in the flow cell for NMR scans/signal acquisition. The sample can comprise diluents in a 1:1 ratio (block 905)
or other defined ratio (e.g., more sample, less diluents or more diluent; less sample, e.g., 2:1 or more sample, less diluents, e.g. 1:2).
The test sample can be rejected with a defined error code if CaEDTA height > 140 for any acquired spectrum (block 919).
This high value is indicative of detection of the citrate peak in conjunction with the CaEDTA peak. The citrate peak is introduced by collection of the specimen in an improper citrate tube. By disrupting the ability to locate the exact position of the CaEDTA peak, the citrate peak can disrupt the process for determining the Valine region.
The Valine region is located upfield relative to the position of the CaEDTA peak. The broad peaks beneath Valine are various methyl (-CH3
-) protons of lipoproteins. The CaEDTA location can be determined at approximately 22258 ± 398 data points (block 921).
The Valine region can be determined independently for each acquired spectrum. The Valine signal can be modeled with suitable data points using, for example, 25 data points (center ±12 data points) for each peak of the quartet or 300 data points for the Valine region of both doublets, but other numbers of data points may be used. The measurement of Valine can be carried out using one, two, three or all four peaks of the Valine peak quartet.
All basis set spectra can be linearly interpolated before utilized by the non-negative least squares algorithm. The spectra to be analyzed and the basis set spectra can have a zero baseline offset modification before utilized by the non-negative least squares algorithm.
The start of the Valine region can be at about 2196-4355 data points, typically the latter when including both doublets, (the "Valine region offset") upfield from the location of the CaEDTA peak (block 922).
In some embodiments, the start of the Valine region is at 4355 data points upfield from the location of the CaEDTA peak.
In some embodiments, the Valine quantification is carried out by characterizing the Valine resonances at between 0.0-1.01 ppm as two doublets. Three or more Valine experimental spectra stepped by two data points can be used as basis sets to model Valine signal. The center Valine peaks can be located by sliding three Valine components +/- 15 data points and determined through a least squares sum minimization. The Valine signal can be modeled with a total of about 300 data points.
Each basis set, including those used for the baseline but excluding the DC offset, can be offset such that the lowest value is subtracted from the function (making the lowest point equal to 0). This prevents inclusion of a DC offset in the shapes they represent.
The Valine region from each acquired spectrum is deconvolved with a series of analyte and baseline functions which have been treated to the same type of pre-processing as the acquired spectra. Shim status of the NMR spectrometer during the signal AT for Valine or other BCAA can be monitored by concurrently evaluating linewidth of a reference peak. A defined adjustment factor based on the linewidth of the reference peak can be applied to the calculated Valine concentration.
The deconvolution coefficient for each component can be multiplied by an associated conversion factor. The current Valine embodiment has a conversion factor of 2271 to report Valine in µM units; however, this value can vary by ± 10% without unduly affecting the reported value significantly. Other conversion factors may be used in the future.
|Component Name||Filename||Conversion Factor||Basis Function Starting position relative to CaEDTA|
The resulting values are summed. Result values produced independently for each acquired spectrum can be averaged to generate final values to use in the measurement.
Data can be acquired using presaturation water suppression from a 1:1 diluted sample and can include between 5-20 scans, typically about 10 scans stored as 5 blocks of 2 (5 FIDs consisting of 2 scans each) (block 926).
The pulse sequence used in conjunction with presaturation water suppression can optionally include a presaturation (water suppression) pulse and a suitable excitation pulse. FIDs can be acquired with 9024 data points with a sweep width of 4496.4 Hz. Each FID can be multiplied with a shifted Gaussian function:
or in computer terms, exp(-((t-gfs)/gf)^2), where gfs = 0.2 seconds and gf = 0.2 seconds.
This can be performed prior to Fourier transformation with zero-filling which yields the frequency-domain GM spectrum for each FID consisting of 16,384 data points (block 927).
The spectra can be phased using the calibration-specified phase value. The spectra can be scaled (multiplied) by a calibration-specified scaling factor. All basis set spectra can be linearly interpolated before utilized by the non-negative least squares algorithm. The spectra to be analyzed and the basis set spectra can have a zero baseline offset modification before utilized by the non-negative least squares algorithm (e.g., all components used for the model and the spectrum that will be analyzed can be linearly interpolated) (block 928).
To determine the center of the Valine fitting region, the Valine resonances between 0.9 and 1.0 as two doublets can be characterized and the center peaks can be identified by sliding three Valine components ± 15 data points (block 929).
 Figure 19
is a chart of prospective associations of hs-CRP and NMR measured GlycA and Valine levels with various exemplary disease outcomes based on MESA data (n ≈5680). The chart was generated from logistic regression analyses adjusted for age, gender, race, smoking, systolic blood pressure, hypertension medications, body mass index, diabetes, LDL-P and HDL-P. The likelihood ratio statistic χ2 gives a quantitative measure of the extent to which the indicated variable improves disease prediction when added to the 10 covariates in the regression model. The analyses used GlycA measurement values from the deconvolution model shown in Figure 7B.
The right side column shows that GlycA and Valine are additive in their associations with disease when they both have significant associations examined separately.
 Figure 20
is a Table of Characteristics of MESA subjects by NMR measured GlycA quartile. The mean GlycA level of those in the 3rd quartile is 29.3. This table shows that people with higher GlycA levels have characteristics associated with higher inflammation (more smoking, hypertension, hs-CRP, etc). NMR signal area units can be called "arbitrary" units. The GlycA levels in this table are in these "arbitrary units" that may be converted to methyl group concentration units (umol/L) by multiplying by 17.8.
Referring now to Figure 21,
it is contemplated that most, if not all, the measurements can be carried out on or using a system 10
in communication with or at least partially onboard an NMR clinical analyzer 22
as described, for example, with respect to Figure 22
below and/or in U.S. Patent No. 8,013,602
The system 10
can include a Diabetes Risk Index Module 370
to collect data suitable for determining the DRI (e.g., HDL subpopulations, GlycA, Valine). The system 10
can include an analysis circuit 20
that includes at least one processor 20p
that can be onboard the analyzer 22
or at least partially remote from the analyzer 22.
If the latter, the Module 370
and/or circuit 20
can reside totally or partially on a server 150.
The server 150
can be provided using cloud computing which includes the provision of computational resources on demand via a computer network. The resources can be embodied as various infrastructure services (e.g. computer, storage, etc.) as well as applications, databases, file services, email, etc. In the traditional model of computing, both data and software are typically fully contained on the user's computer; in cloud computing, the user's computer may contain little software or data (perhaps an operating system and/or web browser), and may serve as little more than a display terminal for processes occurring on a network of external computers. A cloud computing service (or an aggregation of multiple cloud resources) may be generally referred to as the "Cloud". Cloud storage may include a model of networked computer data storage where data is stored on multiple virtual servers, rather than being hosted on one or more dedicated servers. Data transfer can be encrypted and can be done via the Internet using any appropriate firewalls to comply with industry or regulatory standards such as HIPAA. The term "HIPAA" refers to the United States laws defined by the Health Insurance Portability and Accountability Act. The patient data can include an accession number or identifier, gender, age and test data.
The results of the analysis can be transmitted via a computer network, such as the Internet, via email or the like to a patient, clinician site 50,
to a health insurance agency 52
or a pharmacy 51.
The results can be sent directly from the analysis site or may be sent indirectly. The results may be printed out and sent via conventional mail. This information can also be transmitted to pharmacies and/or medical insurance companies, or even patients that monitor for prescriptions or drug use that may result in an increase risk of an adverse event or to place a medical alert to prevent prescription of a contradicted pharmaceutical agent. The results can be sent to a patient via email to a "home" computer or to a pervasive computing device such as a smart phone or notepad and the like. The results can be as an email attachment of the overall report or as a text message alert, for example.
Still referring to Figure 21,
one or more electronic devices 50D, 51D, 60D
associated with the different users, e.g., a clinician site 50,
and/or a test or lab site 60
can be configured to access an electronic analysis circuit 155
in communication with a display of a respective electronic device. The analysis circuit 155
can be hosted on a server and can provide an internet portal or downloadable APP or other computer program for various devices. The circuit 155
can configured to allow a user, e.g., a clinician to enter one or more of: (i) a glucose value of a patient, (ii) a glucose value of a patient and a diabetes risk index score, or (iii) a diabetes risk index score. The circuit can automatically populate different data fields based on a patient identifier or other password at sign-in or allow a user to enter both the DRI score and the glucose measurement for a respective patient. The analysis circuit can be configured to track changes in the DRI score over time and generate electronic reports that can be sent to clinicians, patients or other users. The analysis circuit can also send notices for recommendations on retests, follow-up tests and the like, e.g., if a DRI risk score is elevated or above a low risk value, e.g., in an intermediate risk category, the circuit can notify the clinician that a glucose test may be appropriate or send a notice to the patient to confer with the doctor to see if a glucose test is appropriate or whether increased monitoring intervals for follow-on DRI tests may be desirable. The analysis circuit can generate a risk progression pathway or analysis to provide graphic information that stratifies risk of developing type 2 diabetes in the future for patients having the same glucose value when the glucose value is in an intermediate risk range, when fasting plasma glucose levels are between 90-110 mg/dL, A1C % levels are between 5.7-6.4 or oral glucose tolerance levels are between 140-199 mg/dL. The electronic analysis circuit can be onboard the server 150
in the Cloud or otherwise accessible via the Internet 227
or can be associated with a different client architecture as will be appreciated by one of skill in the art. Thus, a clinician, patient or other user can generate a customized report on risk progression or otherwise obtain risk stratification information.
Referring now to Figure 22,
a system 207
for acquiring at least one NMR spectrum for a respective biosample is illustrated. The system 207
includes an NMR spectrometer 22s
and/or analyzer 22
for obtaining NMR data for NMR measurements of a sample. In one embodiment, the spectrometer 22s
is configured so that the NMR signal acquisition is conducted at about 400 MHz for proton signals; in other embodiments the measurements may be carried out at between about 200MHz to about 900 MHz or other suitable frequency. Other frequencies corresponding to a desired operational magnetic field strength may also be employed. Typically, a proton flow probe is installed, as is a temperature controller to maintain the sample temperature at 47 +/- 0.5 degrees C. The spectrometer 22
is controlled by a digital computer 214
or other signal processing unit. The computer 211
should be capable of performing rapid Fourier transformations. It may also include a data link 212
to another processor or computer 213,
and a direct-memory-access channel 214
which can connects to a hard memory storage unit 215
and/or remote server 150 (Figure 15).
The digital computer 211
may also include a set of analog-to-digital converters, digital-to-analog converters and slow device I/O ports which connect through a pulse control and interface circuit 216
to the operating elements of the spectrometer. These elements include an RF transmitter 217
which produces an RF excitation pulse of the duration, frequency and magnitude directed by the digital computer 211,
and an RF power amplifier 218
which amplifies the pulse and couples it to the RF transmit coil 219
that surrounds sample cell 220
and/or flow probe 220p.
The NMR signal produced by the excited sample in the presence of a 9.4 Tesla polarizing magnetic field produced by superconducting magnet 221
is received by a coil 222
and applied to an RF receiver 223.
The amplified and filtered NMR signal is demodulated at 224
and the resulting quadrature signals are applied to the interface circuit 216
where they are digitized and input through the digital computer 211.
The DRI risk evaluation Module 370
or analysis circuit 20 (Figures 21, 22)
or module 350 (Figure 23)
or can be located in one or more processors associated with the digital computer 211
and/or in a secondary computer 213
or other computers that may be on-site or remote, accessible via a worldwide network such as the Internet 227.
After the NMR data are acquired from the sample in the measurement cell 220,
processing by the computer 211
produces another file that can, as desired, be stored in the storage 215.
This second file is a digital representation of the chemical shift spectrum and it is subsequently read out to the computer 213
for storage in its storage 225
or a database associated with one or more servers. Under the direction of a program stored in its memory, the computer 213,
which may be a personal, laptop, desktop, workstation, notepad, tablet or other computer, processes the chemical shift spectrum in accordance with the teachings of the present invention to generate a report which may be output to a printer 226
or electronically stored and relayed to a desired email address or URL. Those skilled in this art will recognize that other output devices, such as a computer display screen, notepad, smart phone and the like, may also be employed for the display of results.
It should be apparent to those skilled in the art that the functions performed by the computer 213
and its separate storage 225
may also be incorporated into the functions performed by the spectrometer's digital computer 211.
In such case, the printer 226
may be connected directly to the digital computer 211.
Other interfaces and output devices may also be employed, as are well-known to those skilled in this art.
Embodiments of the present invention may take the form of an entirely software embodiment or an embodiment combining software and hardware aspects, all generally referred to herein as a "circuit" or "module."
The computer-implemented steps of the present invention may make use of a computer program product on a computer-usable storage medium having computer-usable program code means embodied in the medium. Any suitable computer readable medium may be utilized including hard disks, CD-ROMs, optical storage devices, or magnetic storage devices.
The computer-usable or computer-readable medium may be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CD-ROM). Note that the computer-usable or computer-readable medium could even be paper or another suitable medium, upon which the program is printed, as the program can be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
Computer program code for carrying out operations of the present invention may be written in an object oriented programming language such as Java7, Smalltalk, Python, Labview, C++, or VisualBasic. However, the computer program code for carrying out operations of the present invention may also be written in conventional procedural programming languages, such as the "C" programming language or even assembly language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer. In the latter scenario, the remote computer may be connected to the user's computer through a local area network (LAN) or a wide area network (WAN) or secured area network (SAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).
The flowcharts and block diagrams of certain of the figures herein illustrate the architecture, functionality, and operation of possible implementations of analysis models and evaluation systems and/or programs according to the present invention. In this regard, each block in the flow charts or block diagrams represents a module, segment, operation, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks might occur out of the order noted in the figures. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.
 Figure 23
is a block diagram of exemplary embodiments of data processing systems 305
that illustrates systems, methods, and computer program products in accordance with embodiments of the present invention. The processor 310
communicates with the memory 314
via an address/data bus 348.
The processor 310
can be any commercially available or custom microprocessor. The memory 314
is representative of the overall hierarchy of memory devices containing the software and data used to implement the functionality of the data processing system 305.
The memory 314
can include, but is not limited to, the following types of devices: cache, ROM, PROM, EPROM, EEPROM, flash memory, SRAM, and DRAM.
As shown in Figure 23,
the memory 314
may include several categories of software and data used in the data processing system 305:
the operating system 352;
the application programs 354;
the input/output (I/O) device drivers 358;
a DRI module 370
and the data 356.
The Diabetes Risk Index Evaluation Module 370
can consider the level of the measured GlycA, lipoprotein components and optionally also Valine and also optionally, glucose, in a multi-parameter mathematical model of risk of developing type 2 diabetes in the future, e.g., the next 5-7 years, and/or a likelihood of having prediabetes.
The data 356
may include signal (constituent and/or composite spectrum lineshape) data 362
which may be obtained from a data or signal acquisition system 320.
As will be appreciated by those of skill in the art, the operating system 352
may be any operating system suitable for use with a data processing system, such as OS/2, AIX or OS/390 from International Business Machines Corporation, Armonk, NY, WindowsCE, WindowsNT, Windows95, Windows98, Windows2000 or WindowsXP from Microsoft Corporation, Redmond, WA, PalmOS from Palm, Inc., MacOS from Apple Computer, UNIX, FreeBSD, or Linux, proprietary operating systems or dedicated operating systems, for example, for embedded data processing systems.
The I/O device drivers 358
typically include software routines accessed through the operating system 352
by the application programs 354
to communicate with devices such as I/O data port(s), data storage 356
and certain memory 314
components and/or the NMR spectrometer or analyzer 22.
The application programs 354
are illustrative of the programs that implement the various features of the data processing system 305
and can include at least one application, which supports operations according to embodiments of the present invention. Finally, the data 356
represents the static and dynamic data used by the application programs 354,
the operating system 352,
the I/O device drivers 358,
and other software programs that may reside in the memory 314.
While the present invention is illustrated, for example, with reference to the Modules 350, 370
being an application program in Figure 23,
as will be appreciated by those of skill in the art, other configurations may also be utilized while still benefiting from the teachings of the present invention. For example, the GlycA Module 350
and/or the DRI Module 370
may also be incorporated into the operating system 352,
the I/O device drivers 358
or other such logical division of the data processing system 305.
Thus, the present invention should not be construed as limited to the configuration of Figure 23,
which is intended to encompass any configuration capable of carrying out the operations described herein.
 Figure 24
is a flow chart of exemplary operations that can carry out embodiments of the present invention. A (measured) composite envelope NMR spectrum of NMR spectra of a fitting region of a biosample (e.g., blood plasma or serum) can be obtained (block 500).
The NMR composite signal envelope is electronically deconvolved using a defined model having HDL, LDL and VLDL/Chylos components and a plurality of curve fit (e.g., Lorentzian) functions associated with at least a GlycA peak region centered at a defined chemical shift location (e.g., 2.00 ppm) associated with GlycA (block 502).
A defined number of curve fitting functions for the peak region associated with GlycA can be summed (block 515).
A conversion factor can be applied to the summed functions to generate a calculated measurement of GlycA (block 520).
The method can include providing at least one defined mathematical model of risk of progression to type 2 diabetes (e.g., within 5-7 years) that is used to calculate a diabetes risk score (block 523).
The method can include identifying whether the patient is at risk of developing type 2 diabetes and/or has prediabetes based on the defined mathematical risk model that includes a plurality of lipoprotein components and GlycA or Valine to generate the DRI score (block 524).
Optionally, the DRI and/or GlycA calculations can be provided in a patient and/or clinical trial report (block 522).
The defined GlycA deconvolution model can include a protein signal component at a density greater than about 1.21g/L that can be deconvolved/ separated from the signal composite envelope (block 503).
 Figure 25A
is a schematic illustration of an exemplary patient test report 100
that can include various lipoprotein parameters such as two or more of DRI, GlycA, Valine, HDL-P, LDL-P 101.
The DRI and/or GlycA number 101
can be presented with risk assessment summary 101s
correlated to population norms, graphs, typical ranges, and/or degree of risk (e.g., high, increased or low risk), shown as a sliding scale graph in Figure 25A.
However, other risk summary configurations may be used including ranges, high to low or low to high, or just noting whether the associated risk is low, medium or increased and/or high.
 Figure 25B
is another example of a patient report with a visual (typically color-coded) graphic summary of a continuum of risk from low to high.
 Figure 26
illustrates that a graph 130
of DRI values over time can be provided to illustrate a change in patient health and/or inflammatory status over time due to age, medical intervention or a therapy according to some embodiments. Tracking this parameter may provide a clinical indicator of efficacy of a therapy and/or a better risk predictor for type 2 diabetes for patients. As shown in Figure 26,
the graph analysis can be used to monitor a patient over time to correlate known start or use of a drug or other therapy. Future drugs or uses of known drugs can be identified, screened or tested in patients identified using DRI and/or GlycA evaluations of drugs or therapies of any suitable disease state including, for example, anti-diabetes and anti-obesity therapies.
The tracking can be provided via a tracking module that can be provided as an APPLICATION ("APP") for a smartphone or other electronic format for ease in tracking and/or for facilitating patient compliance with a therapy. Similarly, an APP for clinicians and/or patients can be provided that allows input of glucose measurement when known along with a DRI score to generate a risk trajectory for evaluating risk and providing an easy to understand risk stratification comparison for a clinician or patient. Passwords or other security measures that comply with privacy and/or HIPPA guidelines can be used with such an APP.
 Figures 27A
are graphical patient/clinical reports of % risk of diabetes versus FPG level and DRI score and risk pathway. Figure 27A
shows patient #1's score while Figure 27B
shows patient #1's score in comparison with a lesser risk patient (patient number 2) having the same FPG. While each patient has the same FPG, they have different metabolic issues identified by the DRI scores stratifying risk according to embodiments of the present invention.
 Figures 28A- 28C
are graphical patient/clinical reports of diabetes conversion rate (%) versus FPG level and DRI score (high DRI, Q4 and low DRI, Q1) according to embodiments of the present invention. Figure 28A
is for a 4-year risk of conversion to diabetes. Figure 28B
is a 5-year risk of conversion and Figure 28C
is a 6 year risk of conversion.
 Figure 29
is a graphical patient/clinical report of a Q4/Q1 relative risk of diabetes conversion (1-8) versus FPG level and DRI score for both 6-year (upper line) and 2-year conversions periods.
 Figure 30
is a graphical patient/clinical report of log scale 5 year conversion with a diabetes conversion rate (%) versus FPG level and DRI score (high DRI, Q4 and low DRI, Q1) color coded from green, yellow, pink/orange to red and with a legend textually correlating the risk as very high, high, moderate and low.
 Figure 31
is a graphical patient/clinical report of 5 year conversion to diabetes with a diabetes conversion rate (%) versus FPG level and DRI score (high DRI, Q4 and low DRI, Q1) color coded from green, yellow, pink/orange to red and with a legend textually correlating the risk as very high, high, moderate and low.
Further embodiments of the invention will now be described by way of the following non-limiting Examples.
A Diabetes Risk index (DRI) was developed using MESA that uses only information derived from a single nuclear magnetic resonance (NMR) spectrum of a fasting plasma sample. The DRI can identify the highest-risk patients who are likely to benefit the most from intervention. This information includes glucose and lipoprotein subclass/size parameters previously linked to insulin resistance, as well as Valine, and GlycA. Figure 1A
was generated using NMR spectra collected at baseline from the Multi-Ethnic Study of Atherosclerosis (MESA). The MESA dataset consisted of 3185 participants, 280 of whom developed diabetes during 5 years of follow-up. Figure 1A
presents the diabetes conversion rates of subjects within 6 glucose subgroups (dotted line), and those for subjects in the upper and lower quartile of DRI within each glucose stratum. As shown, the risk of developing diabetes at any given glucose level is substantially greater for DRI in Q4 vs Q1. The NMR-based diabetes risk score can effectively stratify risk without the need for additional clinical information.
In this analysis, predictive modeling techniques were used to identify a "best" logistic regression model of five year diabetes conversion. The modeling used clinical data from the MESA study as well as NMR derived lipid and metabolite data. Although the final model selection restricted the data to subjects with baseline glucose less than or equal to 115, initial modeling considered the possibility of different "best" models for subjects with baseline glucose less than or equal to 100, subjects with baseline glucose greater than 100 but less than or equal to 115, and subjects with baseline glucose greater than 115 but less than or equal to 125.
One selected model for predicted five year conversion to diabetes included baseline glucose (glucos1c), VLDL size (vsz3), the ratio of medium HDL-P to total HDL-P (HMP HDLP), medium HDL-P (hmp3), the sum of large VLDL-P and medium VLDL-P (vlmp3), GlycA, and Valine. This model included two interactions: HMP_HDLP by GlycA and vsz3 by vlmp3. It was built using data from subjects with baseline glucose less than or equal to 115.
During development of these parameters, large VLDL-P (vlp3) was replaced by vlmp3 (sum of vmp3 and vlp3) in the NMR model as it was found to give better predictive results. Further exploration showed that the NMR model could include an interaction between VLDL size and vlmp3.
The final series of analyses showed that VLDL size and vlp3 (prior to calculation of vlmp3) could be appropriately truncated at low and high values without degrading the predictive accuracy. The benefit of such truncation may be model robustness. Also, the analysis showed that accounting for possible non-linear effects of VLDL size and vlmp3 on five year conversion was unnecessary.
The DRI index model can employ lipoproteins, Valine and GlycA as seven different parameters (including 5 lipoprotein parameters): VLDL size, large+medium VLDL particle number, total HDL and medium HDL subclass particle number, Valine, and GlycA.
Another study of the MESA dataset consisted of 4985 non-diabetic participants, 411 of whom developed diabetes during 6 years of follow-up. The MESA data restricted to the 1832 individuals with intermediate glucose 90-110 mg/dL, 198 of whom converted to diabetes. Conversion rates by quintile of a baseline DRI index and the four component parts of the DRI model: lipoproteins, GlycA, Valine, and glucose were assessed. Relative rates for those in the extreme quintiles were 2.2 for lipoproteins, 1.9 for GlycA, 1.7 for Valine, 6.3 for glucose, and 10.7 for DRI (2.2% in Q1; 23.0% in Q5).
The results indicate that the DRI score, without any additional clinical information, can identify among patients with intermediate glucose levels those with diabetes risk differing >10-fold. Ratios between the first quintile and the fifth quintile establish that there is a 10.7 ratio for DRI which indicates that patients can have a 10 fold difference in diabetes risk when the FPG is in the range of 90 -110 mg/dL.
It is believed that the new DRI scores can allow at-risk patients to be targeted for intervention before the onset of substantial beta-cell dysfunction.
MESA and IRAS Comparison
 Figure 32
is a table of data that shows the performance of DRI (with glucose) in the IRAS dataset, the MESA dataset, and IRAS dataset (using the glucose subgroups from MESA). When IRAS samples were collected, the definitions of pre-diabetes and diabetes were different than that of MESA, which occurred years later.
 Figure 33
shows the performance of DRI (without glucose) in the same dataset criteria as Figure 32.
The highlighted values show the difference between the 5th quintile and the first quintile.
Observational study of middle-aged Hispanic, non-Hispanic white, and African-American men and women. Blood samples obtained 1992-1994. NMR analyses of thawed-refrozen heparin plasma performed in 2001 on Varian instruments (preceding Profilers or current generation NMR analyzers used by LipoScience, Inc., Raleigh, NC) using WET water suppression. NMR dataset population was 46% normoglycemic (old definition, glucose <110 mg/dL), 22% impaired glucose tolerance, 32% diabetic. These analyses are for the n=982 nondiabetic subjects, 134 of whom developed diabetes during a mean 5.2 years of follow-up.
Spearman correlation in MESA and IRAS
Logistic regression results: Conversion to New Diabetes in IRAS and MESA
| ||IRAS, 134 New Diab (N=961)||MESA, 411 Newdiab (N=4985)|
|parameters||adjusted on glucose age gender race|
| ||model X2||param X2||Param P||model X2||param X2||Param P|
|DRInmr (no glucose)
|DRI (LPIR) no glucose
|lsp lsz hsz hlp vlp vsz
The model parameters below were derived from regression models that included glucose/size/metabolite/ratio/SizePlus family. The initial form of these models included baseline glucose, VLDL size (vsz3), GlycA, medium HDL-P ratio (HMP_HDLP), medium HDL-P (hmp3), large VLDL-P (vlp3), and the interaction between GlycA and medium HDL-P ratio. Additional variable investigation determined that gender and Valine could be added to this model, but the addition of alanine did not improve predictive accuracy. Further study and discussion determined that large VLDL-P (vlp3) should be replaced by the sum of large and medium VLDL-P (vlmp3). Also, exploratory analyses determined that an interaction between VLDL size (vsz3) and vlmp3 was statistically significant in the model. The modeling used subjects in the training dataset with baseline glucose less than or equal to 115 for model training, and subjects in the test dataset with baseline glucose less than or equal to 115 for model testing.
DRI models can be based on a likelihood and/or predicted five year conversion to diabetes. The risk evaluation models can include VLDL size (vsz3), a ratio of medium HDL-P to total HDL-P (HMP_HDLP), medium HDL-P (hmp3), a sum of large VLDL-P and medium VLDL-P (vlmp3), GlycA, and Valine.
This model includes two interactions: HMP_HDLP by GlycA and vsz3 by vlmp3. The model can optionally also include a baseline glucose (glucos1c).
Additional modeling determined that VLDL size (vsz3) and large VLDL-P (vlp3, prior to calculation of vlmp3) could be truncated without degradation to predictive accuracy. For vsz3, any value less than 39.2 was truncated to 39.2, and any value greater than 65.1 was truncated to 65.1. For vlp3, any value less than 0.7 was truncated to 0.7, and any value greater than 7.9 was truncated to 7.9.
DRIp Non LPIR Model C-statistic 0.840
|Parameter||DF||Estimate||Standard Error||Wald Chi-Square||Pr > ChiSq|
DRI-LPIR Model C-statistic 0.839
|Parameter||DF||Estimate||Standard Error||Wald Chi-Square||Pr > ChiSq|
The Diabetes Risk Index (DRI) test can be a lab-based multivariate assay that employs a defined mathematical model to yield a single composite score of one's risk of developing type II diabetes in 5 years. Predictive biomarkers included in the assay include: fasting glucose, lipoprotein sub-fractions, branched chain amino acid(s) and one or more inflammatory biomarker(s). Clinical performance is believed to be superior to fasting glucose alone in individuals with FPG<125, and is continually more predictive of diabetes risk vs. FPG with lowering levels of FPG. This is because the DRI risk score captures one's underlying metabolic defects by including these other predictive analytes, which glucose alone does not capture.
Data used to develop and validate DRI scores can be based on retrospective analysis of data from a multi-center and multi-ethnic prospective observational study with nearly 5,000 non-diabetic subjects at baseline. The patient report for this assay can show a fasting glucose value and a single 5-year diabetes risk prediction rate (which includes the risk predictive value of fasting glucose along with the other biomarkers).
The Diabetes Risk Index (DRI) model can be calculated using a plurality of components including at least one lipoprotein component, Valine and/or another branched chain amino acid, and GlycA and/or another inflammatory marker. The model can be adjusted to use different components based on whether the patient is on a statin or other drug therapy known to impact DRI risk scores and/or based on whether the biosample is a fasting or non-fasting biosample.
The DRI risk score can be calculated in a plurality of different manners and filtered before sent to a clinician (or sent to the clinician for correct reporting) and patient based on defined patient criteria so that the patient is provided the appropriate score for the test and patient-specific parameters for that biosample.
The DRI model is configured to stratify risk in subjects having the same A1C, oral glucose tolerance or FPG measurement and a different diabetes risk score. The diabetes risk index score can be a numerical score within a defined score range, with scores associated with a fourth quartile (4Q) or fifth quintile (5Q) of a population norm reflecting an increased or high risk of developing type 2 diabetes within 5-7 years. Respective subjects that are at increased risk of developing type 2 diabetes prior to onset of type-2 diabetes when glucose is in a range that is below prediabetes to the high end of prediabetes, e.g., a fasting blood glucose (FPG) levels are between 90-125 mg/dL can be identified. The DRI score provides information to stratify risk in subjects having the same glucose measurement and a different diabetes risk score based on underlying metabolic issues in different patients.
 Table 7
lists potential alternative VLDL parameters that may be used in a DRI model based on a logistic regression analysis on VLDL. One or more of the alternative VLDL parameters may optionally be used with one or more of the components described above with respect to any of the other DRI models and/or components thereof described in the Examples and/or Detailed Description part of the specification. Large VLDL may be referred to as "VLP (V5p+V6p+CHYp)", which are TRL particles ranging from 60-260 nm diameter. Different definitions of "large VLDL" could be used in a DRI model. For example, the chylomicrons could be excluded and may be called TRL60-140 based on Table 5.
Alternatively or in addition, TRL60-120 (without V6-140) may be used in a DRI model.
Potential VLDL parameters that may be used in a DRI model with other components.
|Logistic regression on New Diabetes||MESA N=210/2038, model unadjusted, glucose 90-110 mg/dL||MESA N=210/2038, model adjusted on glucose (glucose 90-110 mg/dL)|
|NMR parameters||X2 parameter (model adjusted on glucose)||X2 parameter (model not-adjusted)||X2 parameter (model adjusted on glucose)||X2 parameter (model not-adjusted)|
The foregoing is illustrative of the present invention and is not to be construed as limiting thereof. Although a few exemplary embodiments of this invention have been described, those skilled in the art will readily appreciate that many modifications are possible in the exemplary embodiments without departing from the scope of the invention as defined by the appended claims. Accordingly, all such modifications are intended to be included within the scope of this invention as defined in the claims. In the claims, means-plus-function clauses, where used, are intended to cover the structures described herein as performing the recited function and not only structural equivalents but also equivalent structures. Therefore, it is to be understood that the foregoing is illustrative of the present invention and is not to be construed as limited to the specific embodiments disclosed, and that modifications to the disclosed embodiments, as well as other embodiments, are intended to be included within the scope of the appended claims. The invention is defined by the following claims.