WO2012122606A1

WO2012122606A1 - Disease and disease predisposition risk indication

Info

Publication number: WO2012122606A1
Application number: PCT/AU2012/000278
Authority: WO
Inventors: Jing Xie; Ecosse Luc Edouard Denis LAMOUREUX; Tien Yin Wong
Original assignee: Centre For Eye Research Australia
Priority date: 2011-03-16
Filing date: 2012-03-16
Publication date: 2012-09-20

Abstract

The invention provides a method of obtaining an indication of risk of a disease or predisposition thereto the method comprising processing one or more non-invasive risk prediction variables and data based on one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto. The processing may comprise assessing the received one or more non-invasive risk prediction variables and the data based on one or more retinal features to calculate the indication of risk of a disease or predisposition thereto. The calculation may comprise summing the received variables and data. Also provided is an apparatus and computer instruction code for obtaining an indication of risk of a disease or predisposition thereto along with a method of screening.

Description

TITLE

DISEASE AND DISEASE PREDISPOSITION RISK INDICATION

FIELD

The present invention relates generally to disease and disease predisposition risk indication. More particularly, the invention is directed to a method for indicating disease risk or disease predisposition including processing one or more non-invasive risk prediction variables and data based on one or more retinal features, although the scope of the invention is not necessarily limited thereto.

BACKGROUND

Curative and palliative care of diseases are maj or components of national economies. For example, cardiovascular disease (CVD) is the leading cause of death and constitutes a major public health problem worldwide.

It is highly desirable that people at risk of developing a disease be identified before disease onset or early in disease progression. Disease and disease predisposition prediction play an important role in clinical medicine which can help in identifying individuals at high risk, educating patients and guiding treatment decisions (1). The use of a "risk score" (e.g., Framingham) allows clinicians to identify persons at risk of diseases such as, CVD. However, in the case of CVD, current risk prediction models lack precision and up to 50% of CVD cases cannot be captured by the Framingham risk score alone. Furthermore, there is no clear evidence that such risk assessment actually improves primary prevention or health outcomes.

Traditionally, "risk scores" have been calculated from multivariable logistic/Cox regression models (2-6). However, this approach may be problematic due to the following limitations: (a) it does not incorporate an individual's susceptibility to these risk factors and thus treats all individuals at the same level of risk given their having the same numbers and same levels of the risk factors. Due to that individual's tolerance and response to risk factors on some individual-specific characteristics (age, genetic susceptibility, lifestyle-related factors, etc), it would be logical and more precise if such characteristics are taken into consideration when estimating an individual's disease risk; (b) the regression model becomes very difficult to interpret for the purpose of classification when there exists many interaction terms; (c) some risk groups become empty because no patient is included in some covariate profiles; and (d) traditional logistic/Cox regression analysis cannot handle missing values directly and it is also not robust in dealing with outliers. Improved methods are required.

OBJECT OF THE INVENTION

It is an object of the present invention to provide a method for obtaining an _Λ indication of risk of disease or predisposition thereto. A preferred object is to provide a method for obtaining an indication of risk of disease or predisposition thereto including processing one or more non-invasive risk prediction variables and data based on one or more retinal features.

Another preferred object is to provide a method for obtaining an indication of risk of disease or predisposition thereto without requiring further testing to be performed on the subject.

Still another preferred object is to overcome and/or alleviate one or more of the above disadvantages of the prior art and/or provide a useful commercial choice.

SUMMARY

The present invention is broadly directed to a method for indicating risk of disease or predisposition thereto. A preferred advantage of the method is that risk of disease is indicated without a further test having to be performed on the subject.

In a first aspect, the invention provides a method for obtaining an indication of risk of a disease or predisposition thereto the method comprising:

processing one or more non-invasive risk prediction variables and data based upon one or more retinal features to thereby obtain the indication of risk of disease or predisposition thereto.

Suitably, the method may be performed by a computer.

Suitably, the method may further comprise receiving the one or more non- invasive risk prediction variables and the data based upon one or more retinal features.

The one or more non-invasive risk prediction variables and the data based upon one or more retinal features may be received via a computer network or may be received by direct input into the computer.

Suitably, the processing may comprise assessing the received one or more non-invasive risk prediction variables and the data based upon one or more retinal features to calculate the indication of risk of a disease or predisposition thereto.

Suitably, the calculation may comprise summing the received one or more noh-invasive risk prediction variables and the data based upon one or more retinal features.

Suitably, me indication may be a relative indication of the risk of developing a disease or predisposition thereto.

The relative indication may be associated with a percentage value or percentage range.

The percentage value or percentage range may be selected based on a disease or predisposition thereto for which the indication is being obtained.

Preferably, the relative indication may be selected from low, intermediate and high.

The low indication may be less than 10%.

The intermediate indication may be between 10 and 20%.

The high indication may be greater than 20%.

Suitably, the disease or predisposition may be a circulatory disease or predisposition or a cognitive disease or predisposition.

Suitably, the circulatory disease may be a cardiovascular disease, a coronary heart disease, stroke and/or a predisposition thereto.

Suitably, the cognitive disease or predisposition may be Alzheimer' s disease, Parkinson's disease or a predisposition thereto.

Suitably, the one or more non-invasive risk prediction variables may be selected from the group consisting of age, gender, smoking, diabetes and medication.

The age may be a subject's age.

The age may be classified into an age range.

The age range may be 40-60; 61 -70; or 71 -80 years.

Gender may be male or female.

Smoking may be whether the subject is a smoker or not. A smoker may comprise a current smoker, a former smoker or any history of smoking.

Diabetes may be presence or absence of diabetes. The presence or absence of diabetes may be obtained by questioning the subject. If a further test such as, a glucose tolerance test, is required to determine the presence or absence of diabetes this test may also be comprised in the present invention.

Medication may mean on medication or not.

On medication may comprise currently on medication or history of taking prescription medication.

On medication may comprise taking insulin or other medication for diabetes and/or taking medication for hypertension.

Suitably, the one or more non-invasive risk prediction variables may comprise age, gender, diabetes and smoking.

Suitably, the data based on one or more retinal features may be selected from the group consisting of central retinal artery equivalent (CRAE); central retinal vein equivalent (CRVE) and/or retinopathy.

The data based on one or more retinal feature may be associated with a risk of disease or predisposition therefor.

Preferably, the one or more non-invasive risk prediction variables is not a feature of an eye or retina.

Suitably, the method may further comprise processing one or more invasive risk prediction variable.

Suitably, the assessment may comprise a predictive analytics assessment.

The predictive analytics assessment may comprise a classification and regression tree (CART) analysis.

The CART analysis may comprise one or more rules.

The one or more rules of the CART assessment may be based on selected one or more non-invasive risk prediction variables and/or the data based on one or more retinal features.

The one or more rules may be dichotomous or may comprise three or more parts.

Wherein one of the one or more rules comprises three or more parts the rule may comprise three or more sub-ranges within a range.

The one or more rules may comprise a true/false question and/or election of a sub-range within a range.

The CRAE rule may comprise selection of less than or equal to 135.5; 135.5- 150.63 ; or greater than 150.63 based on the subject' s CRAE.

The age rule may comprise classifying the subject according to their age into an age-related group. The age related group may be a group for ages 40-60; a group for ages 61-70; or a group for ages 71 -80 years.

In a second aspect the invention provides an apparatus for obtaining an indication of risk of a disease or predisposition thereto the apparatus comprising: a processor for processing one or more non-invasive risk prediction variables and data based on one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto.

Suitably, the apparatus may further comprise an input for receiving the one or more non-invasive risk prediction variables and the data based on one or more retinal features.

The input may be via a computer network or may be via direct input into the apparatus.

Suitably, the processor may calculate from the received one or more noninvasive risk prediction variables and the data based on one or more retinal features the indication of risk of a disease or predisposition thereto.

Suitably, the calculation may comprise summing the received one or more non-invasive risk prediction variables and the data based on the one or more retinal features.

Suitably, indication may be a relative indication of the risk of developing a disease or predisposition thereto.

The low indication may be less than 10%.

The intermediate indication may be between 10 and 20%.

The high indication may be greater than 20%.

Suitably, the circulatory disease may be a cardiovascular disease, a coronary heart disease, stroke and or a predisposition thereto.

Suitably, the cognitive disease or predisposition may be Alzheimer's disease, Parkinson's disease or a predisposition thereto.

The age may be subject's age.

The age may be classified into an age range.

The age range may be 40-60; 61-70; or 71-80 years.

Gender may be male or female.

Diabetes may be presence or absence of diabetes. The presence or absence of diabetes may be obtained by questioning the subject. If a further test such as, a glucose tolerance test, is required to determine the presence or absence of diabetes this test may also be comprised in the second aspect of the invention.

Medication may mean on medication or not.

On medication may comprise taking insulin or other medication for diabetes and/or taking medication for hypertension. ^'

The data based on retinopathy may be presence or absence of retinopathy.

Suitably, the assessment may comprise a predictive analytics assessment. The predictive analytics assessment may comprise a classification and regression tree (CART) analysis.

The CART analysis may comprise one or more rules.

The one or more rules of the CART assessment may be based on selected one or more non-invasive risk prediction variables and/or the data based on the one or more retinal features.

The one or more rules may be dichotomous or may have three or more parts.

Wherein the rules have three or more parts the rule may comprise three or more sub-ranges within a range.

The CRAE rule may comprise selection of less than or equal to 135.5; 135.5- 150.63; or greater than 150.63 based on the subject's CRAE.

The age rule may comprise classifying the subject according to their age into an age-related group. The age related group may be a group for ages 40-60; a group for ages 61-70; or a group for ages 71-80 years.

In a third aspect the invention provides computer instruction code for obtaining an indication of risk of a disease or predisposition thereto, comprising: computer instruction code operable to process one or more non-invasive risk prediction variables and data based on one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto.

The computer instruction code may be carried on a suitable carrier medium, including in particular a tangible carrier medium such as, a disk. The computer instruction code may also be carried by a non-tangible carrier medium such as a communication signal. The invention therefore provides a computer readable medium carrying computer instruction code as provided by the third aspect of the invention.

Suitably, the computer instruction code further comprise computer instruction code operable to receive the one or more non-invasive risk prediction variables and the data based on one or more retinal features.

The one or more non-invasive risk prediction variables and the data based on one or more retinal features may be received via a computer network or may be received by direct input into a computer operating the computer instruction code.

Suitably, the computer instruction code may further comprise computer instruction code to assess the received one or more non-invasive risk prediction variables and the data based on one or more retinal features to calculate the indication of risk of a disease or predisposition thereto. Suitably, the calculation may comprise summing the received one or more non-invasive risk prediction variables and the data based on one or more retinal features.

The low indication may be less than 10%.

The intermediate indication may be between 10 and 20%.

The high indication may be greater than 20%.

The age may be subject's age.

The age may be classified into an age range. ·

The age range may be 40-60; 61-70; or 71-80 years.

Gender may be male or female.

Diabetes may be presence or absence of diabetes . The presence or absence of diabetes may be obtained by questioning the subject. If a further test such as, a glucose tolerance test, is required to determine the presence or absence of diabetes this invasive test may also be comprised in the third aspect of the invention.

Medication may mean on medication or not. On medication may comprise currently on medication or history of taking prescription medication.

The data based on retinopathy may be presence or absence of retinopathy.

Suitably, the assessment may comprise a predictive analytics assessment.

The CART analysis may comprise one or more rules.

The one or more rules of the CART assessment may be based on the one or more non-invasive risk prediction variables and/or the data based on one or more retinal features.

The one or more rules may be dichotomous or may have three or more parts. Wherein the rules have three or more parts the rule may comprise three or more sub-ranges within a range.

The one or more rules may comprise a true/falsequestion and/or election of a sub-range within a range.

The CRAE rule may comprise selection ofless than or equal to 135.5; 135.5- 150.63 ; or greater than 150.63 based on the subject' s CRAE.

The age rule may comprise classifying the subject according to their age into an age-related group. The age related group may be a group for ages 40-60; a group for ages 61 -70; or a group for ages 71 -80 years. In a fourth aspect the invention provides a method for screening for risk of a disease or predisposition thereto the method comprising:

processing one or more non-invasive risk prediction variables and data based on one or more retinal features to thereby obtain the indication of risk of disease or predisposition thereto.

Suitably, the screening method may be performed by a computer.

Suitably, the screening method may further comprise receiving the one or more non-invasive risk prediction variables and the data based on one or more retinal features.

The one or more non-invasive risk prediction variables and the data based on one or more retinal features may be received via a computer network or may be received by direct input into the computer.

Suitably, the processing may comprise assessing the received one or more non-invasive risk prediction variables and the data based on one or more retinal features to calculate the indication of risk of a disease or predisposition thereto.

Suitably, the calculation may comprise summing the received one or more non-invasive risk prediction variables and the data based on one or more retinal features.

The low indication may be less than 10%.

The intermediate indication may be between 10 and 20%.

The high indication may be greater than 20%.

Suitably, the circulatory disease may be a cardiovascular disease, a coronary heart disease, stroke and/or a predisposition thereto. Suitably, the cognitive disease or predisposition may be Alzheimer's disease, Parkinson's disease or a predisposition thereto.

The age may be subject's age.

The age may be classified into an age range.

The age range may be 40-60; 61-70; or 71-80 years.

Gender may be male or female.

Diabetes may be presence or absence of diabetes. The presence or absence of diabetes may be obtained by questioning the subject. If a further test such as, a glucose tolerance test, is required to determine the presence or absence of diabetes this invasive test may also be comprised in the fourth aspect of the invention.

Medication may mean on medication or not.

Suitably, the data based on one or more retinal features may be selected from the group consisting of be central retinal artery equivalent (CRAE); central retinal vein equivalent (CRVE) and/or retinopathy.

The data based on retinopathy may be presence or absence of retinopathy.

Suitably, the assessment may comprise a predictive analytics assessment.

The CART analysis may comprise one or more rules.

The one or more rules of the CART assessment may be based on the one or more non-invasive risk prediction variables and or the data based on one or more retinal features.

The one or more rules may be dichotomous or may have three or more parts.

Any discussion of the prior art throughout the specification should in no way be considered as an admission that such prior art is widely known or forms part of the common general knowledge in the field.

US patents do not constitute common general knowledge in Australia or other countries.

As used herein, except where the context requires otherwise, the term "comprise" and variations of the term, such as "comprising", "comprises" and "comprised", are not intended to exclude further additives, components, integers or steps.

BRIEF DESCRIPTION OF THE DRAWINGS AND TABLES

In order that the present invention may be readily understood and put into practical effect, reference will now be made to the accompanying illustrations and tables where:

FIG. 1 shows a schematic diagram illustrating an apparatus according to one embodiment of the invention.

FIG. 2 is a chart illustrating one embodiment of the invention.

TABLE 1 is a table showing characteristics of study participants included in the

CART analysis for CVD risk predication. Note: this analysis is limited in the White population and aged 40 and above.

Table 2 is a table showing the area under Receiver Operating Characteristic (ROC) curves for prediction of CVD, CHD and stroke.

Table 3 is a table showing risk reclassification by CART for prediction CVD, CHD and stroke. Ref: The national Cholesterol Education Program Report (low: <10%, intermediate: 10-20%, and high: >20%).

DETAILED DESCRIPTION

The inventors have provided a novel and inventive method of obtaining an indication of risk of a disease or predisposition thereto which includes processing one or more non-invasive risk prediction variable and data based on one or more retinal feature. The processing may include assessing the one or more non-invasive risk prediction variable and data based on one or more retinal feature.

Surprisingly, as exemplified herein, the inventors have provided a method that has equivalent predictive ability to methods based on the traditional risk factors without requiring a further test to be performed on the subject.

The further test may be any test that requires a physical interaction with the subject. The further test may be an invasive test. An invasive test is one which requires a break in the skin to be created or contact with the mucosa, or internal body cavity beyond a natural or artificial body orifice. An example of an invasive test is taking a blood sample. It is to be understood that questioning the subject is not a physical interaction.

In this way the present invention allows an accurate determination of risk of disease or predisposition from data based on a retinal feature and other readily available information such as, age, gender, smoking, diabetes and medication.

The indication may be a relative indication compared to normal. The relative indication may stratify relative risk for developing disease or predisposition thereto. The relative indication may be associated with a percentage value or percentage range. The percentage value or percentage range may be selected based on a diseaseor predisposition thereto for which the indication is being obtained. The relative indication may be selected from low, intermediate and high. For example, when the disease or predisposition is CVD, the low indication may be less than 10%; the intermediate indication may be between 10 and 20%; and the high indication may be greater than 20% (Ref: The national Cholesterol Education Program Report (low: <10%, intermediate: 10-20%, and high: >20%)).

The present invention may be applied to any disease or predisposition thereto.

As used herein disease means any disease, condition or symptom thereof. The invention may find particular application to diseases such as circulatory diseases and conditions, cognitive diseases and conditions and predispositions therefor.

The circulatory disease or predisposition may be a cardiovascular disease, a coronary heart disease, stroke and/or a predisposition thereto.

The cognitive disease or predisposition may be Alzheimer's disease, Parkinson's disease or a predisposition.

By non-invasive risk prediction variable is meant a factor not requiring any direct contact or intervention with the subject to be obtained. The non-invasive risk prediction variable may be obtained merely by questioning the subject or from a subject's records. Preferably, the one or more non-invasive risk prediction variable is a non-optical risk prediction variable.

The one or more non-invasive risk prediction variable may be age, gender or . sex, smoking, diabetes, medication and/or a non-invasive Framingham risk factor. While age, gender and diabetes are self-explanatory and the Framingham risk factors are well known the other factors require further explanation.

Although some of these risk prediction variables are self-explanatory, for the avoidance of confusion they are further elucidated below.

Age may be the subject's age. The age may be classified into an age range. The age range may be 40-60; 61 -70; or 71 -80 years.

Gender may be male or female.

Diabetes may be presence or absence of diabetes. The presence or absence of diabetes may be obtained by questioning the subject. If a further test such as, a glucose tolerance test, is required to determine the presence or absence of diabetes this invasive test may also be comprised in the method of the invention.

Smoking means whether the subject is a smoker or not. A smoker may comprise a current smoker, a former smoker or any history of smoking.

Medication means on medication or not. On medication may comprise currently on medication or history of taking prescription medication. The factor medication includes more specific factors such as: currently taking any medication for a circulatory disease, a cognitive disease and/or other disease. In particular embodiments the medication may be for hypertension and/or diabetes. The medication for diabetes may be insulin.

The Framingham risk factors are well known in the art and include LDL cholesterol, HDL cholesterol, total cholesterol, blood pressure, whether the patient is treated or not for hypertension, and/or dyslipidaemia.

The data based on one or more retinal feature may be derived from any feature of the retina including a retinal parameter and/or retinopathy. Examples of retinal parameters are CRAE and CRVE. The data based on one or more retinal feature may also comprise a combination of one or more of retinopathy; CRAE and CRVE.

CRAE and CRVE are measures of average retinal arteriolar and venular calibre, respectively. The measurement may be computer-assisted. In one embodiment the retinal microvascular calibre is measured using a computer-based program (TV AN, University of Wisconsin, Madison) based on a detailed protocol. According to this protocol optic-disc centred eye photographs are selected for measurement. All arterioles and venules coursing through an area 0.5 to 1 disc diameter from the optic disc margin are measured and the biggest 6 arteriolar and venular calibres are summarized as the central retinal artery equivalent (CRAE) and central retinal venular equivalent (CRVE) based on the Knudtson formula [ref: Knudtson MD, Lee KE, Hubbard LD, Wong TY, Klein R, Klein BE. Revised formulas for summarizing retinal vessel diameters. Curr Eye Res. Sep 2003;27(3):143-149].

In one embodiment retinopathy may be considered to be present if any characteristic features or lesions as defined by the Early Treatment Diabetic Retinopathy Study (ETDRS) severity scale is present. The features or lesions may comprise one or more of a microaneurysm, a haemorrhage, a cotton wool spot, an intraretinal microvascular abnormality, a hard exudate, venous beading, and/or a new vessel. This list is indicative only and is not exhaustive.

These features or lesions may be defined as present if graded as either definite or probable [ref: Grading diabetic retinopathy from stereoscopic color fundus photographs—an extension of the modified Airlie House classification. ETDRS report number 10. Early Treatment Diabetic Retinopathy Study Research Group. Ophthalmology. May 1991 ;98(5 Suppl):786-806]. These features or lesions may be identified either by manual grading or by automated image analysis algorithms. The automated image analysis may be as described in US patent 12/631,515 published as US 2010/0142767.

Although obtaining the one or more retinal parameter requires an interaction with the subject, as discussed above, once obtained no further interaction with the subject is required.

The one or more non-invasive risk prediction variable may be selected by a logistic regression model (LM) and/or a tree-based model (TBM). The selection may be from the non-invasive risk prediction variables discussed herein or other suitable non-invasive risk prediction variables. From the information provided herein a skilled person is readily able to select other suitable non-invasive risk prediction variables.

Such a logistic regression model (LM) and/or a tree-based model (TBM) may also be used in the generation of an algorithm for use in the invention. The algorithm preferably includes suitable non-invasive risk prediction variables and suitable data based on one or more retinal feature. The algorithm may also include selection of one or more invasive risk prediction variables.

As noted above, the method may also include an invasive risk prediction variable. The invasive risk prediction variable may have been obtained previously in which case no further invasive interaction with the subject is required.

By invasive risk prediction variable is meant any variable that requires direct contact and/or intervention with the subject. Examples of invasive risk prediction variable include total cholesterol, low-density lipoprotein (LDL) cholesterol, high density lipoprotein (HDL) cholesterol, blood pressure (BP), systolic blood pressure (SBP) and or an invasive Framingham risk factor.

The cholesterol variable may be an indication of hypercholesterolemia.

The blood pressure factor may be an indication of hypertension.

The invention may also make use of one or more new risk prediction variable and or one or more biomarkers that have only recently been identified and proposed. Such one or more new risk prediction variable may be based on retinal vascular calibre. The and one or more biomarkers may include C-reactive protein and/or fibrinogen. The one or more new risk factors and/or biomarkers may be a non- invasive risk prediction variable or an invasive risk prediction variable. The assessment used in the method of the invention may be a statistical analysis such as a predictive analytics assessment. The predictive analytics assessment may be a classification and regression tree (CART) analysis. The CART analysis may include one or more rules. The one or more rules of the CART assessment may be the one or more non-invasive risk prediction variables and the one or more retinal feature.

The one or more rules may be in the form of true/false questions and/or election of a sub-range within a range.

The CRAE rule may comprise choosing either of a) less than or equal to 135.5; B) 135.5-150.63 or C) greater than 150.63 based on the subject's CRAE.

The age rule may comprise classifying the subject according to their age into an age-related group. The age related group may be a group for ages 40-60; a group for ages 61 -70 ; or a group for ages 71-80 years.

The one or more rule may be used stratify the population into different risk levels for a disease.

The one or more rule may also be used to build a model of risk prediction by grouping specific variables and features. The following exemplary models, Model 1 , Model 2 and/or Model 3, may be used.

Model 1 may include the one or more non-invasive risk prediction variables of age, gender, diabetes and smoking and data based on the one or more retinal features of CRAE.

Model 2 may include the one or more non-invasive risk prediction variables of age, gender, diabetes and smoking; and data based on the one or more retinal features of CRVE.

Model 3 may include the one or more non-invasive risk prediction variables of age, gender, diabetes and smoking; and data based on the one or more retinal features of retinopathy.

Other models may include suitable one or more non-invasive risk prediction variables and data based on retinopathy and data based on CRAE and/or CRVE.

The invention may also provide an apparatus for obtaining an indication of risk of a disease or predisposition thereto. The apparatus may take any suitable form such as, a computer or paper-based.

In one embodiment the apparatus may display on a screen associated with the computer or on the paper check boxes for the answers to the one or more rules. The sum or weighted sum of the answers to the one or more rules may map to a risk of having the disease.

In another embodiment the apparatus may display on a screen associated with the computer or on the paper a decision tree. The decision tree may be navigated from root to a terminal node through a series of branching internal nodes, each internal node comprising a rule of the one or more rules. The path followed may depend on the answers to each rule at each internal node encountered along the path. The final terminal node arrived at determines the risk of disease or predisposition.

An internal node may be dichotomous or may be divided into three or more parts. ,

When the internal node is dichotomous it may be true or false. Whether the subject has diabetes (true or false); the subject's gender or sex (male or female); whether the subject is a smoker (true or false); and whether the subject is currently taking any medication (true or false) are examples of a dichotomous node.

When the internal node is divided into three more or parts the node may be divided into three or more sub-ranges within a range. CRAE, CRVE and age are examples of questions at a node that may have three or more paths. An example of sub-ranges within a range is the CRAE sub-ranges of: less than 135.5; between 135.5 and 150.63; and greater than 150.63; within the range of all possible CRAE values. Another example is the sub-ranges of an age of 40-60; 61-70; and 71-80; within the range of 40-80 years.

The invention may also provide a computer program product comprising a computer usable medium and computer readable program code embodied on said computer usable medium for obtaining an indication of risk of a disease or predisposition thereto. The computer readable code may comprise computer readable program code devices (i) configured to cause the computer to assess one or more non-invasive risk prediction variables and one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto.

The one or more non-invasive risk prediction variable and the one or more retinal feature may be received by the computer, for example by direct input from by the subject or computer operator or from a database through a network such as, the internet or local area network. With reference to FIG. 1 , one embodiment of an apparatus 10 according to the invention is shown comprising a processor 12 operatively coupled to a storage medium in the form of a memory 14. One or more input device 16, such as a keyboard, mouse and/or pointer, is operatively coupled to the processor 12 and one or more output device 18, such as a computer screen, is operatively coupled to the processor 12.

Memory 14 comprises a computer or machine readable medium 22, such as a read only memory (e.g., programmable read only memory (PROM), or electrically erasable programmable read only memory (EEPROM)), a random access memory (e.g. static random access memory (SRAM), or synchronous dynamic random access memory (SDRAM)), or hybrid memory (e.g., FLASH), or other types of memory as is well known in the art. The computer readable medium 22 comprises computer readable program code components 24 for performing the methods in accordance with the teachings of the present invention, at least some of which are selectively executed by the processor 12 and are configured to cause the execution of the embodiments of the present invention described herein. Hence, the machine readable medium 22 may have recorded thereon a program of instructions for causing the machine 10 to perform methods in accordance with embodiments of the present invention described herein.

The CART assessment may be conducted in a stepwise fashion in which the most significant predictor at each step is used to split the sample into highly homogenous subgroups. This assessment may be continued until differences are no longer statistically significant based on an adjusted chi-square statistic.

The results may be presented as a tree that requires no calculations for use, but identifies groups that are expected to experience a given outcome.

Nodes in the CART analysis may be constrained to have a minimum size. For example, in one embodiment a node may be constrained to have a minimum size of 100 records in parent nodes and 20 records in final child nodes. In another embodiment a node may be constrained to have a minimum size of 200 records in parent nodes and 60 records in final child nodes. In another embodiment a node may be constrained to have a minimum size of 400 records in parent nodes and 100 records in final child nodes. Based on the teaching here a person of skill in the art is readily able to vary the constraints for a suitable model. The proportion of subjects having the disease or predisposition may be reported in each node of the tree. A risk stratification system may be developed based on the final child nodes in CART analysis.

The present inventors have provided a method for providing an indication of disease risk. The invention is of significant advantage because the indication may be provided without an invasive technique. The invention is of further advantage because it may be applied as a simple screening tool to indicate disease risk in a population, a group within a population and/or an individual.

Although not restricted thereto the present invention may be applied to the broad categories of circulatory and cognitive disease and symptoms.

Within circulatory disease the present invention has particular application to the risk of CVD, CHD and stroke based on health information readily known to the average person..

Of further significant advantage the desired predictor variables and the prediction algorithms may be incorporated into a screening program.

The prediction algorithms and the predictor variables may also be incorporated into a computer software program or medical record system enabling the health care practitioner to evaluate a patient's situation and advise the best course of action.

The following non-limiting examples illustrate the invention. These examples should not be construed as limiting: the examples are included for the purposes of illustration only. The Examples will be understood to represent an exemplification of the invention. For example, the invention is illustrated with reference to providing an indication of risk of a CVD, CHD and stroke but is not limited thereto.

Examples

EXAMPLE 1: Risk prediction for Cardiovascular Disease (CVD), Coronary Heart Disease (CHD) and Stroke-A simple tool based on regression and classification tree:

Objective: Retinal vessel calibre has been shown to be associated with stroke, a major public health problem. Using modern statistical methods (i.e. decision tree), our aims were: (1) to identify subgroups that are at risk of stroke; and (2) using our new method, to compare the predictive performance of a combination of retinal features (e.g. CRAE, CRVE and retinopathy) with non-invasive risk prediction variables compared with traditional risk factors alone. Traditional risk factors are those such as age, gender, total cholesterol, HDL cholesterol, smoking status, diabetes and SBP.

Methods : The data, of 21075 white adults aged >40 years, were obtained from seven large population-based studies. Three major predictive models were built by modern statistical methods to identify individuals at high risk for CHD, CVD and stroke, respectively, see Table 2. Model performances were assessed by using area of ROC, sensitivity and specificity analyses.

Results: The categories generated by CART analysis stratified the population into 18 different risk levels for stroke. Older people (>70 years old) with retinopathy had the highest risk for stroke (19.1%). Comparatively, the significant predictors based on traditional risk factors were age, SBP, diabetes, smoking, medication for hypertension, and total cholesterol. Model 1 and Model 2 had comparable predictive ability (Area under the ROC [AUC] for Models 1: 0.754 (95% CI: 0.740, 0.769); AUC for Model 2: 0.747 (95% CI: 0.733, 0.761) (^=1.38, 0.24).

Statistical analysis: Classification and regression tree (CART):

Classification and regression tree (CART) analysis may be used to identify a subject's risk of a disease or predisposition thereto. A subject at high risk for stroke may be examined. The target variable has a value of 0 or 1 depending on stoke status (0 if absent, or 1 if present). The explanatory variables selected to predict stroke consist of age, gender, smoking status, CRVE, CRAE, and any retinopathy. CART is conducted in a stepwise fashion in which the most significant predictor at each step is used to split the sample into highly homogenous subgroups. This process continues until differences are no longer statistically significant based on an adjusted chi-square statistic.

Validation method of CART analysis (split-sample validation):

With split-sample validation, the model is generated using a training sample and tested on a hold-out sample. The total sample was randomly divided into training and test samples. The training sample was used to generate the predictive algorithm, which later on being tested in the testing sample.

To test the performance of the models, prediction algorithms were applied to the testing dataset, and the values of the predicted probability of stroke were generated and compared to actual values of stroke outcome. The following measures were used for the validation of the prediction models.

Probability of stroke predicted on a testing set was categorized into deciles and for each category the rate of stroke was calculated and compared with the : predicted value.

Predictive accuracy of the model was assessed using ROC curves and by comparing observed and expected survival curves. The ROC curve shows the ability of a risk score to discriminate between high- and low-risk subjects. Any specific value of predicted risk can be used to divide people into high- and low-risk groups. For any given value, the true positive rate (sensitivity) was the proportion of those with strokes who were classified as high risk, and the false positive rate (1- specificity) was the proportion without strokes that were classified as high risk. These numbers were plotted against each other for each possible cut point of predicted risk. A larger area under the curve indicates better discrimination.

Reclassification improvement is defined as an increase in risk category for individuals who develop events and as a decrease for those who do not. Net reclassification improvement accounts for movement between categories in the wrong direction and applies different weights to events and non-events. We used 0% to 10%, 10% to 20%, and 20% as risk categories.

The procedure ROCCOM in the software package STATA (Stata Corporation, College Station, Tex.) was used to calculate and compare the area under the ROC curves.

EXAMPLE 2: Exemplary CART Analysis:

CART analysis was used as shown in FIG. 2. Decision tree 100 was used to analyse a subject for risk indication for a cardiovascular event 102.

In decision tree 100 all listed no and yes % values are percentage indications of the subject having a cardiovascular event.

The first decision in tree 100 is CRAE 104. If the CRAE 104 is less than or equal to (≤) 135.5 106, Node 1 112 is reached from where it can be seen the subject has a 13 % yes value. If the CRAE 104 is between 135.5 and 150.63 108, Node 2 114 is reached with a 9.1 % yes value. If the CRAE 104 is greater than 150.63 110, Node 3 116 is reached with an 11.7 % yes value.

From Node 1 112, the second decision is gender 118 being either female 120 or male 122. Taking female 120 arrives at Node 4 124 with a 9.4% yes value. Taking male 122 arrives at Node 5 140 with a 17.2% yes value.

From Node 4 124 the third decision is age 126. If age 126 is 71-80 128, Node 11 134 is reached with a 21.2% yes value. If the age 126 is 40-60 130, Node 12 136 is reached with a 5.4% yes value. If the age 126 is 61-70 132, Node 13 138 is reached with a 9.3% yes value.

Returning to Node 5 140, the third decision is diabetes 142, being either no 144 or yes 146. Taking no 144 arrives at Node 14 148 with a 15.7% yes value. Taking yes 146 arrives at Node 15 150 with a 27.2% yes value.

Going back to Node 2 114 the second decision is diabetes 152 being no 154 oryes 156. Ifno 154, Node 6 158 is reached with a 7.8% yes value. If yes 156, Node 7 160 is reached with a 20.8% yes value.

From Node 6 158 the third decision is age 162. If the age 162 is 61-80 164, Node 16 170 is reached with a 10.9% yes value. If the age 162 is 40-50 166, Node 17 172 is reached with a 0% yes. If the age 162 is 51-60 168, Node 18 174 is reached with a 5.5% yes value.

From Node 16 170 the fourth decision is gender 176, being female 178 or male 180. If female 178, Node 23 182 is reached with an 8.7% yes value. If male 180, Node 24 184 is reached with a 14% yes value.

From Node 18 174 the fourth decision is currently smoking 186, being either no 188 or yes 190. If no 188, Node 25 192 is reached with a 4.1 % yes value. If yes 190, Node 26 194 is reached with a 12.6% yes value.

Returning to Node 3 116, the second decision is age 196. If the age 196 is 71 - 80 198, Node 8 204 is reached with a 23.4% yes value. If the age 196 is 40-60 200, Node 9 206 is reached with a 6.4% yes value. If the age 196 is 61-70 202, Node 10 208 is reached with a 14.3% yes value.

From Node 9 206 the third decision is smoking 210, being either no 212 or yes 214. If no 212, Node 19 216 is reached with a 4.9% yes value. If yes 214, Node 20 218 is reached with an 11.2% yes value.

From Node 19216 the fourth decision is diabetes 220, being either no 222 or yes 224. If no 222, Node 27 226 is reached with a 3.9% yes value. If yes 224, Node 28 228 is reached with a 15.4% yes value.

Returning to Node 10 208 the third decision is sex 230, being either female 232 or male 234. If female 232, Node 21 236 is reached with a 10.3% yes value. If male 234, Node 22 238 is reached with a 20.5% yes value.

EXAMPLE 3: Results, Discussion and Conclusions:

Results: The use of classification trees using age, gender, diabetes, smoking status and the retinal parameters has revealed a number of high-risk subgroups for CVD, CHD and stroke, see Table 3. Importantly the method of the invention has the equivalent predictive ability to the traditional risk factors (see Table 2) and the advantage thereover of not requiring an invasive interaction with the subject. Discussion: Unlike traditional methods, the tree-building techniques are ideally suited for the development of a reliable clinical decision rule that can be used to classify new patients into categories according to predicted outcome, where traditional statistical methods are sometimes cumbersome to use or of limited utility. Tree-based modeling works well when the regression variables are a mixture of categorical and continuous variables. The algorithm is non-parametric, so no assumptions are made regarding the underlying distribution of values of the predictor variables. Tree-based modeling requires relatively little input from the analyst; the outcome is presented in a form of binary trees and is easy to interpret by a non- statistician. As further information such as, further prediction variables, becomes available the tree may be refined and analysis repeated. By repeating the analysis with further information a more accurate risk prediction may be provided.

CART has advantages over logistic regression for the creation of a simple screening tool for the outcome prediction. The method of the invention provides a solution that is simple to understand and interpret and requires no mathematical calculations like those for logistic regression, e.g., Equation. Data do not need preprocessing such as normalization, creation of dummy variables, or removal of records with missing fields, and both nominal and categorical variables are handled.

Conclusion: Model I, which includes age, gender and retinal parameters, has comparable predictive ability for stroke, compared to using the traditional invasive and non-invasive risk factors. This suggests that there is great potential for applying the method of the invention as a non-invasive predictive tool as a replacement for conventional risk scores to screen people at high risk of a disease in health care limited settings. ,

The findings herein show that retinal factors can provide additional information of use in disease prediction, such as stroke prediction, among asymptomatic people.

The present invention is of significant advantage because for intervention to be most effective in clinical settings, it should be easy to implement and tailored to meet the needs and risks of particular patients.

The present invention can be used to identify subgroups and individuals who are at risk of a disease.

The present invention is able to stratify disease risk, detect interactions among variables, evaluate misclassification of diagnosis, and handle missing values.

The present invention is also flexible and can utilise new risk prediction variables and/or biomarkers which may be assessed and incorporated into current risk prediction algorithms. In this manner the present invention provides novel personalised risk prediction models incorporating new biomarkers, which has a great potential to improve risk prediction of diseases events and predisposition.

Throughout the specification the aim has been to describe the preferred embodiments of the invention without limiting the invention to any one embodiment or specific collection of features. It will therefore be appreciated by those of skill in the art that, in light of the instant disclosure, various modifications and changes can be made in the particular embodiments exemplified without departing from the scope of the present invention.

All computer programs, algorithms, patent and scientific literature referred to herein is incorporated herein by reference.

REFERENCES

1. Altman DG, Royston P: What do we mean by validating a prognostic model? Stat Med 19:453-473, 2000.

2. Brindle P, Beswick A, Fahey T, Ebrahim S. Accuracy and impact of risk assessment in the primary prevention of cardiovascular disease: a systematic review. Heart 2006; 92:1752-9.

3. Pettila V. Predicting risk of death from cardiovascular disease. Outcome prediction is hampered by methodological problems. BMJ 2001 ; 323 : 1000.

4. Brindle P, Emberson J, Lampe F et al. Predictive accuracy of the Framingham coronary risk score in British men: prospective cohort study. BMJ 2003; 327:1267. Brindle P, May M, Gill P et al. Primary prevention of cardiovascular disease: a web-based risk score for seven British black and minority ethnic groups. Heart 2006; 92:1595-602. Panagiotakos DB, Pitsavos C, Stefanadis C. Inclusion of Dietary Evaluation in Cardiovascular Disease Risk Prediction Models Increases Accuracy and Reduces Bias of the Estimations. Risk Anal. 2008.

Table 1 Characteristics of study participants included in the CART analysis for CVD risk predication

Note: this analysis is limited in the White population and aged 40 and above.

Table 2 Area under Roc curves for prediction of CVD, CHD and stroke

Table 3 Risk reclassification by CART for prediction CVD, CHD and stroke

Ref: The national Cholesterol Education Program Report (low: <10%, intermediate: 10-20%, and high: >20%)

Claims

1. A method of obtaining an indication of risk of a disease or predisposition thereto the method comprising:

processing one or more non-invasive risk prediction variables and data based on one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto.

2. The method of claim 1 wherein the method is performed by a computer.

3. The method of any preceding claim further comprising receiving the one or more non-invasive risk prediction variables and the data based on the one or more retinal features.

4. The method of any preceding claim wherein the one or more non-invasive risk prediction variables and the data based on the one or more retinal features are received via a computer network.

5. The method of any preceding claim wherein the one or more non-invasive risk prediction variables and the data based on the one or more retinal features are received by direct input into the computer.

6. The method of any preceding claim wherein processing comprises assessing the received one or more non-invasive risk prediction variables and the data based on one or more retinal features to calculate the indication of risk of a disease or predisposition thereto.

7. The method of any preceding claim wherein the calculation comprises summing the received one or more non-invasive risk prediction variables and the data based on the one or more retinal features.

8. An apparatus for obtaining an indication of risk of a disease or predisposition thereto the apparatus comprising:

a processor for processing one or more non-invasive risk prediction variables and data based on the one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto.

9. The apparatus of claim 8 further comprising an input for receiving the one or more non-invasive risk prediction variables and the data based on the one or more retinal features.

10. The apparatus of claim 9 wherein the input is via a computer network.

11. The apparatus of claim 9 wherein the input is by direct input into the apparatus.

12. The apparatus of any one of claims 8 to 11 wherein the processor calculates from the received one or more non-invasive risk prediction variables and the data based on the one or more retinal features the indication of risk of a disease or predisposition thereto.

13. The apparatus of claim 12 wherein the calculation comprises summing the received one or more non-invasive risk prediction variables and the data based on the one or more retinal features.

14. Computer instruction code for obtaining an indication of risk of a disease or predisposition thereto, comprising:

computer instruction code operable to process one or more non-invasive risk prediction variables and data based on one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto.

15. The computer instruction code of claim 14 further comprising computer instruction code operable to receive the one or more non-invasive risk prediction variables and the data based on one or more retinal features.

16. The computer instruction code of claim 14 or claim 15 wherein the one or more non-invasive risk prediction variables and the data based one or more retinal features are received via a computer network

17. The computer instruction code of any one of claims 14 to 16 wherein the one or more non-invasive risk prediction variables and the data based on one or more retinal features are received by direct input into a computer operating the computer instruction code.

18. The computer instruction code of any one of claims 14 to 17 further comprising computer instruction code to process the received one or more noninvasive risk prediction variables and the data based on one or more retinal features to calculate the indication of risk of a disease or predisposition thereto.

19. The computer instruction code of claim 18 wherein the calculation comprises summing the received one or more non-invasive risk prediction variables and the data based on the one or more retinal features.

20. A method of screening for risk of a disease or predisposition thereto the method comprising:

processing one or more non-invasive risk prediction variables and data based o one or more retinal features to thereby obtain the indication of risk a disease or predisposition thereto.

21. The method of claim 20 wherein the method is performed by a computer.

22. The method of claim 20 or claim 21 further comprising receiving the one or more non-invasive risk prediction variables and the data based on the one or more retinal features.

23. The method of claim 22 wherein the one or more non-invasive risk prediction variables and the data based on one or more retinal features are received via a computer network.

24. The method of claim 22 wherein the one or more non-invasive risk prediction variables and the data based on one or more retinal features are received by direct input into the computer.

25. The method of any one of claims 20 to 24 wherein the processing comprises assessing the received one or more non-invasive risk prediction variables and the data based on one or more retinal features to calculate the indication of risk of a disease or predisposition thereto.

26. The method of claim 25, wherein the calculation comprises summing the received one or more non-invasive risk prediction variables and the data based on one or more retinal features.