WO2015021441A1 - Assay and method for determining insulin-on board - Google Patents

Abstract

The present invention provides for a home test assay or device that can calculate insulin on board levels, and uses thereof for improving insulin bolus calculation and prevention of hypoglycemia in e.g. Type-1 diabetes patients.

Description

ASSAY AND METHOD FOR DETERMINING INSULIN-ON-BOARD

[001] This application claims the benefit of U.S. Provisional Application No. 61/864,234 filed August 9, 2013, which is hereby incorporated in its entirety including all tables, figures, and claims.

FIELD OF THE INVENTION

[002] The present invention is situated in the field of medical diagnostics, more in particular in the field of calculation of insulin on board (IOB), based on patient- specific information and on the real-time measurement of insulin. Said IOB is important for calculating more accurate insulin bolus and can optionally be combined with real-time glucose measurements in a whole blood sample of the subject to get a better understanding of the patients glucose metabolism and actual insulin need.

BACKGROUND OF THE INVENTION

[003] In diabetes patients, and especially in type-1 diabetes mellitus (T1DM) patients, knowing one's insulin need when and where needed is today still largely based on inaccurate assumptions. This is because insulin tests are cumbersome and expensive, and have to be carried out in a lab. The insulin quantity for the bolus injection just before the meal is nowadays calculated using the actual glucose level measured using a home glucose test, in combination with the insulin sensitivity, the amount of carbohydrates in the meal, the glucose correction factor and the insulin to carbohydrates ratio. Insulin sensitivity is translated into the glucose correction factor (the reduction of the glucose level due to the working of 1 unit of insulin) and the Insulin to carbohydrate ratio (the amount of carbohydrates from the meal that will be metabolised with 1 unit of insulin) Furthermore, due to the regular or continuous administrations of insulin in a T1DM subject, there is a need of knowing the amount of insulin that is yet to become active and/or enter the blood stream (e.g. still residing in the sub-cutis) from the previous injection(s). This so called insulin-on-board (IOB) should be taken into account at the time of the new bolus calculation to avoid too much insulin action. Today, estimated IOB values are based on population- wide averages such as the linear 1 -fifth rule of insulin absorption per hour since the last injection and are generally inaccurate. The resulting output of the calculated insulin bolus quantity (or dosage) is hence also often inaccurate and may lead to non-healthy or dangerous glucose and/or insulin levels, with dangerous or even life-threatening results. Today's insulin pumps or point of care devices generally can only monitor or measure the patient's blood glucose levels, rather than actual insulin levels, which makes it impossible to have a good estimation of the IOB.

[004] The present invention intends to overcome these inaccuracies by providing a new way of calculating the IOB or adjusted IOB in a subject.

SUMMARY OF THE INVENTION

[005] The present invention provides products and methods for measuring insulin levels in a blood sample of a subject and calculating or estimating therefrom the insulin on board (IOB). The product and method is to be seen as a home self test or assay, or as a point of care device or test for the patient or medical practitioner.

[006] The present invention allows for more accurate calculation and estimation of Insulin On Board. Insulin On Board is the amount of insulin still residing in the sub-cutis at the place of the last injection(s). It is the amount of insulin that still has to appear into the blood stream over the next few hours. Rather than relying solely on the time that has elapsed since the last insulin injection to calculate the IOB, the method and device according to the present invention will determine the IOB by measuring the concentration of insulin in blood. That is, based on the period of time and the amount of insulin introduced in the last injection and the currently measured insulin concentration, the device and method according to the invention will calculate the Insulin On Board.

[007] Furthermore, a part of the injected insulin will never appear in the bloodstream, due to fixation in the sub-cutis, or due to enzymatic degradation of the insulin or due to complexation with other molecules present in the patient's body. This is typically expressed by the absolute bioavailability (F), such that (1-F) is the fraction of total injected insulin that is lost in transport. Properly accounting for the lost insulin results in an adjusted amount of residing insulin, called the "adjusted IOB" or "alOB" hereinafter. The alOB may be compared directly to a patient's total injected insulin. The alOB is the amount of injected insulin that would be required to yield a given physical IOB. The physical IOB is itself based on the plasma insulin concentration which is already accounting for the loss-in-transport. The alOB is adjusted by dividing the physical IOB by E[F], where E[F] is the expected value of F. If Dose is the total amount of insulin injected, then E[F] Dose is the amount that is expected to arrive in the plasma and F Dose is the amount that actually arrives in the plasma.

[008] To properly determine the amount of insulin needed for the next bolus injection, the determined IOB or alOB should be subtracted from the calculated bolus injection to avoid over-insulinisation. Too much insulin is to be avoided since it may result in hypoglycemia later. This is particularly important and dangerous during the time when an individual is sleeping when no carbohydrate intake counteracts the excess in insulin activity. Alternatively, a too low dose of insulin may result in hyperglycemia, which over time can lead to multiple co-morbities, including kidney damage, neurological damage, cardiovascular damage, damage to the retina or damage to feet and legs of the patient.

[009] The present invention further provides for better dosing the insulin administration in insulin pump users and self-injecting patients, comprising the step of measuring or estimating IOB or alOB as disclosed herein, particularly using a device according to any one of the embodiments described herein, or using any other device that can measure insulin concentrations and optionally glucose concentrations in a small sample of blood (e.g. a drop of blood).

[0010] The present invention further provides for a better dosing of insulin in insulin pump users and self-injecting patients by avoiding over- and under-insulinisation. By measuring the circulating concentration of insulin in blood the device and method according to the invention enables a more accurate calculation of the Insulin On Board or adjusted Insulin On Board. The device or method may thus alert a user of potential hypoglycemia in cases where excessive amounts of insulin are likely to appear in the bloodstream even before a drop in glucose levels occurs. For example at bedtime, the user would be in a position to suspend the insulin pump delivery temporarily when too much Insulin On Board is detected. Alternatively the user may decide to consume additional carbohydrate before going to sleep in order to compensate for any excess residual insulin action within the body, which may otherwise lead to hypoglycemia. The Insulin On Board calculated according to the method of the current invention is based on actual measurement(s) of insulin in a patient blood sample (e.g. a blood drop, e.g. through a finger prick). The device or method according to the invention may further prevent potential hyperglycemia, by avoiding overestimation of insulin on board. Overestimation of IOB can result in the administration of a too low bolus injection of insulin, i.e. insufficient to metabolise the carbohydrates present in the meal to be taken in.

[0011] The device and method according to the invention enables the calculation or estimation of IOB or alOB based on two insulin measurements, i.e. a first measurement of the amount of insulin near the insulin peak concentration time after the previous insulin administration, and a second insulin measurement at the time of calculating the IOB. Said "peak concentration time" is determined a priori based on population pharmacokinetic data depending on the kind of insulin used and typically for "fast-acting" insulins is between 0.5 and 1.5 hours after the injection, with a mean of about 1 hour after the injection. For other insulins engineered to act even more rapidly, this peak concentration time can be significantly earlier i.e. a few minutes in case of aerosol administration. These peak times are usually well documented in the product insert of the used insulin.

[0012] In addition to the expected (population) peak concentration time, several other pharmacokinetic parameters relevant to subcutaneous insulin administration must be set a priori based on relevant population pharmacokinetic data. The estimation method of the present invention demonstrates how to set these parameters, without which the IOB or adjusted Insulin On Board calculation would be impossible because the required pharmacokinetic parameters cannot be identified from measurements of the plasma insulin concentration following subcutaneous injection alone. Data collected via intravenous administration would be required to obtain said other parameters. Furthermore, all methods described in the literature for estimating the pharmacokinetic parameters to calculate IOB require a dense set of serial samples following subcutaneous injection, e.g., measurements of plasma insulin concentration every 30 minutes for up to 6 hours. The estimation method of the present invention demonstrates that a clinically relevant IOB (or adjusted IOB) can be calculated based on two insulin measurements in blood, following insulin injection. While the physical IOB represents the number of units of insulin that reside in the subcutaneous from prior administration of insulin, this value may be adjusted (adjusted IOB, or alOB) to account for subcutaneous losses (i.e. part of the injected insulin that will never arrive in plasma and hence never be active), or to account for the effective bioactivity of the plasma insulin in a given physiological context (active vs. resting, eating vs. fasting, etc.). The adjusted IOB (alOB) may be calculated in standard units of insulin dosage (e.g. IU) and may be used by the patient to modulate future glucose metabolism control (e.g. insulin administration and/or carbohydrate consumption). Methods for calculating both IOB and adjusted IOB are described herein.

[0013] As summarized above, the Insulin On Board, calculated according to the methods of the current invention, is based on actual measurement(s) of insulin in a patient blood sample. The estimation model generally uses two insulin measurements, one at the peak concentration time of the insulin administered, and one at the time of calculating IOB. Of course it will be clear that more insulin concentration measurements may be performed, which can further increase the accuracy of the estimation model.

[0014] In contrast, in order to further reduce the required number of serial samples from two samples per injection to a single sample per injection, a method is presented herein that requires the prior determination of an empirical model relating the insulin concentration measured at a single time-point to the IOB at said time-point. The method of determining this empirical model is based on collecting a multitude of patient specific data of the insulin concentration based on two (or more) serial samples per injection according to the methods described herein. In essence, the method as disclosed herein using the two samples per injection (e.g. at the peak concentration time and at the time of calculating the next bolus or IOB) is used for establishing an empirical model, e.g. at the beginning of the use of the method or device according to the invention. Said empirical model can then be used to replace the measurement at the peak concentration time, and hence further reduce the number of sampling steps needed for estimating IOB or adjusted IOB. The determination of this empirical model will typically require the patient to record at least 2 insulin measurements per injected bolus, 3 times per day (breakfast, lunch and dinner) for a period of at least 15, preferably 20, more preferably 30 days or any period in between. This "training" process will establish the empirical model, which, from that point onward will allow the patient to estimate his IOB based on a single sample per injection.

[0015] The inventors have started from the published Three Compartment Model of Insulin Pharmacokinetics, as disclosed in Wong et al., 2008 (J. Diabetes Sci Technol, Vol.2(3):436-449). This model (Wong et al.) has 6 pharmacokinetic parameters which may be set based on published literature and additional clinical studies. However, the variation of these pharmacokinetic parameters across the population and within an individual from injection to injection is so large (coefficients of variation form 20% to 40%) that use of this model (Wong et al.) with a single fixed set of coefficients cannot provide an accurate estimate of IOB. The present invention shows how to simplify the three compartment model (Wong et al., 2008) so that the parameters of the simplified pharmacokinetic model can be identified by serial measurements of the plasma insulin concentration (e.g. two or more measurements per injection). Furthermore, the present invention shows how the parameters of the simplified pharmacokinetic model (thus identified) may be used to estimate the IOB.

[0016] In essence, the simplified pharmacokinetic model states the plasma insulin concentration I(t) as a time dependent profile with only two free parameters, an amplitude parameter and a rate parameter, that may vary from patient to patient and injection to injection. For a given injection, the two parameters are identified by making two serial measurements of the plasma insulin concentration and using standard mathematical techniques to estimate the parameters (e.g., nonlinear least squares regression). The IOB is then calculated from I(t) in a series of steps.

[0017] The first step is to integrate I(t) from t=0 to t=infinity thereby calculating the AUC (area under the curve).

[0018] The second step is to integrate I(t) from t=T to t=infinity thereby calculating the partial AUC which can be denoted as pAUC(T), where τ is the time-point of the desired IOB estimate (which will typically be equal to the second of the two serial measurements of the plasma insulin concentration).

[0019] The third step is multiplication of pAUC(T) by a constant that converts the result into units of insulin dose (IU), e.g., IOB = CL x pAUC(T), where CL is the plasma clearance of insulin. If CL were determined a priori for the given patient (e.g., determined by intravenous administration of insulin followed by suitable serial measurements of the plasma insulin concentration), then the IOB thus calculated would represent the physical amount of insulin still in transport (but not yet in the plasma) that will arrive in the plasma for t > τ. In patients using subcutaneous injection however, the CL cannot be exactly measured.

[0020] In the limit τ=0, this physical IOB is equal to CL x AUC which is equivalent to Dose x F, where F is the absolute bioavailability via the subcutaneous route of administration and Dose is the total quantity of insulin injected. (Note, F is a dimensionless proportion from 0 to 1). The missing insulin, equal to (1-F) x Dose, is lost in transport (presumably near the site of injection) prior to reaching the plasma and will never become biologically active in the patient. F may vary considerably from injection to injection, even within an individual. Let E[F] denote the expected value of F over the patient's distribution of F. Next, the so-called "adjusted IOB" or "alOB" is defined as: alOB = CL x PAUC(T) / E[F]. The adjusted IOB represents the number of injected units of insulin (IU) that are required to achieve a given level of physical IOB. In this way, the adjusted IOB accounts for the expected losses of insulin in the subcutaneous, i.e., units of insulin that will never make it into the plasma. Noting that AUC/Dose = F/CL (a general pharmacokinetic relationship), the equation for adjusted IOB may be re- written as follows: alOB = pAUC(T) / E[ AUC/Dose], where E[AUC/Dose] is the expected value of AUC/Dose.

[0021] This formulation of the adjusted IOB is particularly useful, because while the values of CL and F are not generally known for an individual and cannot be identified by direct measurement of plasma insulin concentrations following subcutaneous injection, the ratio of F/CL may be identified and therefore the adjusted IOB is practical to measure.

[0022] The adjusted IOB is also easy to interpret. For example, in the case that τ = 0, the adjusted IOB is equal to the dose multiplied by the ratio of the actual AUC to the expected AUC. If the actual AUC is larger than expected, the adjusted IOB is larger than the applied dose, indicating that the patient has more plasma insulin than expected, as if he had applied a larger dose. Similarly, if the actual AUC is smaller than expected, the adjusted IOB is smaller than the dose applied, and the patient has less plasma insulin than expected, as if he had applied a smaller dose. More generally, the calculation of adjusted IOB will be made for a typical value of τ in the range of approximately 3 to 5 hours following the previous injection at which point the patient would like to have a meal and is estimating the amount of insulin required for the mealtime bolus. In this case, pAUC(T) is smaller than AUC because of the time elapsed from t=0 to t=T.

[0023] The adjusted IOB can readily be calculated by the methods described in this invention. Following the steps above, two measurements of plasma insulin are made following each injection, allowing estimation of the amplitude and rate parameters of the simplified pharmacokinetic model. As a generalization, three or more serial measurements of the plasma insulin concentration may be made, allowing identification of the parameters of the pharmacokinetic model. As a further generalization, an alternative pharmacokinetic model with greater than two free parameters may be estimated from three or more serial measurements of the plasma insulin concentration. Following this, the values of AUC and pAUC(T) are readily calculated for each injection. After the first injection, E[AUC/Dose] may be estimated as AUC/Dose. After two injections, E[AUC/Dose] may be estimated as the mean over two injections. After three injections, E[AUC/Dose] may be estimated as the mean over three injections, etc. For example, after one week of applying the method 3 times per day, E[AUC/Dose] may be calculated as the mean over 21 measurements. Over an even longer time period, E[AUC/Dose] may be calculated as a moving average, or weighted moving average, or Kalman filter. Based on this method, the adjusted IOB may be estimated following each dose. The moving average allows the expected value to slowly vary over time, as the patient may undergo physiological changes.

[0024] As a further specification for calculating the IOB or alOB, a known constant (called the background concentration) is subtracted from the measured plasma insulin concentrations prior to performing the analysis described above. This step is termed background subtraction. The background concentration may be determined prior to all IOB calculations by measuring the patient's morning plasma insulin concentration at fasting levels, prior to any mealtime bolus. Preferably, the insulin assay used to measure the plasma insulin concentration is specific to the rapid acting insulin isoform, with minimal cross-reactivity to the long acting isoform, in which case the background concentration will be relatively small compared to the measured insulin levels and will therefore have only a small effect on the IOB calculations. In this case, a suitable constant can be selected based on population studies. Furthermore, the constant may be biased high (relative to the population mean) because higher values of the background concentration will result in higher IOB estimates, which may be safer (as compared to lower IOB estimates) in view of overdosing insulin. More preferably, the device used to measure the plasma insulin concentration can feature two assays or sensors, each detecting the levels of each of the relevant insulin isoforms, rapid acting and long acting, allowing accurate background subtraction.

[0025] As will be shown in the examples and figures below, the IOB or alOB estimation model according to the invention estimates IOB or alOB far more accurate than standard models without sampling (cf. Figures 9, 11, and 13). This is a major improvement for T1DM patients, since one of the most frequent issues with bolus administration is an overdosing of insulin due to erroneous estimation of the amount of insulin still present in the body, or still to become active. The underestimation of the IOB can lead to over-insulination and hypoglycemia. Alternatively, an overestimation of IOB can lead to the administration of an insufficient amount of insulin to e.g. metabolize the next meal, leading to hyperglycemia. Both situations are dangerous and the more accurate IOB estimation method of the present invention overcomes these difficulties.

[0026] The present invention therefore provides for the following aspects:

[0027] Aspect 1. A method for determining the Insulin On Board (IOB) in a diabetes mellitus patient comprising the steps of: a) in a first blood sample taken or obtained from the patient near the time of peak insulin concentration resulting from the last insulin administration, determining the amount of insulin in said sample, b) in a second blood sample taken or obtained from the patient at the time of calculating the IOB after said last insulin administration, determining the amount of insulin in said sample, c) calculating the IOB at a given moment based on the previously administered amount of insulin, the time of previous insulin administration(s) and the two determined amounts of insulin from steps a) and b).

[0028] Aspect 2. The method according to aspect 1, wherein the IOB is calculated based on a two-parameter continuous distribution, fitted to the two insulin concentration measurements of step c). Examples of such a distribution may be constructed from a Gamma density function, any probability density function, or compartmental pharmacokinetic (multi-exponential) density function. In each example, the two- parameter distribution is constructed as an amplitude (the first unknown parameter) multiplied by a density function, wherein the time integral (from 0 to infinity) of the density function is unity and where the second unknown parameter determines both the location and scale of the density function (with all other parameters of the density function held fixed).

[0029] Aspect 3. The method according to aspect 2, wherein the IOB at time τ equals pAUC(T) CL, where pAUC(T) is the partial area under the curve from time τ to infinity calculated from the integral of said distribution, where AUC is the total area under the curve of said distribution, and CL is the plasma clearance of insulin. [0030] Aspect 4. The method according to aspect 3, wherein the adjusted IOB (alOB) at time τ is calculated as follows: alOB = pAUC(T) CL/E[F], wherein E[F] is the expected bioavailability of insulin via the subcutaneous route of administration. If Dose is the total amount of insulin injected, then E[F] Dose is the amount that is expected to arrive in the plasma and F Dose is the amount that actually arrives in the plasma.

[0031] Aspect 5. The method according to aspect 4, wherein E[F]/CL is identified as the expected value of AUC/Dose and calculated based on the AUC/Dose values (the single most recent AUC/Dose value, or a set of prior AUC/Dose values) obtained by the insulin measurements made at the two time points in steps a) and b).

[0032] Said method of calculating IOB or adjusted IOB as outlined in aspects 1 through 5 is referred to as the "two-sample" method of calculating IOB or alOB. It is however clear that additional sampling and insulin-measurement steps could be done in order to further improve the accuracy of the estimation method of calculating IOB or alOB, as is reflected in Aspect 6.

[0033] Aspect 6. The method according to anyone of the previous aspects, wherein additional insulin measurements may be performed on additional blood samples obtained at different time points after an insulin bolus injection and where the two-parameter continuous distribution may be replaced by a distribution with two or more free parameters.

[0034] Aspect 7. The method according to aspect 1, wherein said first measurement near the peak concentration time in step a) is replaced by a patient specific empirical model relating the insulin concentration measured at a single time-point to the adjusted IOB at the same time-point, thereby providing a one-sample method of calculating IOB or alOB.

[0035] Aspect 8. The method according to aspect 7, wherein said empirical model is determined based on aggregated data obtained from a plurality of patient specific insulin measurements obtained in a manner described in the two-sample method of aspects 2 to 5.

[0036] Aspect 9. The method according to any one of the previous aspects, wherein said patient is an insulin-pump patient, or wherein said patient administers insulin through injections. [0037] Aspect 10. The method according to any one of the previous aspects, for use in detecting changes in pharmacokinetics of insulin plasma appearance in insulin pump using patients using a catheter for insulin administration that stays in situ for at least one day, at least two days, at least three days, at least four days, at least five days, or for multiple days.

[0038] Aspect 11. The method according to any one of the previous aspects, for use in regulating insulin release by an insulin pump.

[0039] Aspect 12. The method according to aspect 11, wherein said insulin release by said insulin pump activity is reduced or halted when the IOB level becomes too high, or increased when the insulin concentration is predicted to become too low.

[0040] Aspect 13. A method for determining the amount of insulin needed in a diabetes mellitus patient comprising the steps of: a) determining the amount of Insulin On Board (IOB) using the method according to any one of aspects 1 to 12, b) calculating the amount of insulin needed in said patient, based on the pre-meal glucose concentration in the patient and the quantity of carbohydrates in the next meal, the glucose concentration correction factor, the insulin to carbohydrate ratio and subtracting the Insulin On Board determined in step a).

[0041] Aspect 14. A method for better serving the actual basal insulin need of a subject, comprising the step of measuring the actual IOB in said patient, using the method according to any one of aspects 1 to 12.

[0042] Aspect 15. A method for better dosing the insulin administration in insulin pump users, comprising the step of measuring the actual IOB in said patient, using the method according to any one of aspects 1 to 12.

[0043] Aspect 16. A method of alerting the patient of increased future risk for hypoglycemia by balancing the need for insulin in the near future with the real insulin action in the near future by: a) measuring the actual IOB in said patient, using the method according to any one of aspects 1 to 12, or b) measuring the long-acting insulin part and adding it to a) above, c) determining the blood glucose concentration in said patient, d) evaluating whether the insulin action will be balanced with the glucose appearance in blood, and e) alerting the patient of possible hypoglycemia when the IOB exceeds the available glucose in said patient; or possible hyperglycemia when the IOB is insufficient for the glucose appearing in the plasma in said patient.

[0044] Aspect 17. A method of calculating the cumulative IOB of several previous injections, using the method according to anyone of aspects 1 to 12, and comparing it with the need of insulin in the near future, thus allowing to detect future risk hypoglycemia and hyperglycemia.

[0045] Aspect 18. A method of measuring insulin in blood for fast feedback on how much insulin reached the blood after aerosol administration (nasal or inhaled insulin), comprising the step of: in a sample taken from the patient measuring the amount of insulin, and comparing said amount with the administered amount of insulin.

[0046] Aspect 19. A method for determining the adjusted Insulin On Board (alOB) in a diabetes mellitus patient comprising the steps of:

1) establishing a patient specific empirical model as follows: a) in a first blood sample obtained or taken from the patient near the peak concentration time of the last insulin administration, measuring the amount of insulin in said sample, b) in a second sample obtained or taken from the patient at the time of calculating the Insulin On Board, measuring the amount of insulin in said sample, c) calculating the amount of IOB (or alOB) at a given moment based on the previous administered amount of insulin, the time of previous insulin administrations and the two measured insulin amounts in steps a) and b), wherein said steps a) to c) are repeated at least 5 times, preferably at least 10 times, more preferably at least 15, 20, 25, or 30 times; where the IOB (or alOB) at time τ following each injection is calculated using the algorithm of the present invention, i.e., IOB = pAUC(T) CL, or alOB = pAUC(T) CL/E[F], where E[F]/CL = E[AUC/Dose], Dose is the total injected insulin, PAUC(T) is the partial area under the curve from time τ to infinity calculated from the integral of the distribution obtained through the empirical model, AUC is the total area under the curve of said distribution, CL is the plasma clearance of insulin, and E[F] is the expected bioavailability of insulin via the subcutaneous route of administration; and d) fitting a regression model to the data of IOB (or alOB) versus insulin concentration collected in step c) using either a linear model, or preferably a quadratic model of log IOB versus log concentration, or possibly a higher order (or piecewise) polynomial to represent the relationship between the IOB (the response) and the insulin concentration (the predictor); and

2) calculating the IOB by: e) in a sample taken from the patient at the time of calculating the Insulin On Board, measuring the amount of insulin in said sample, f) calculating the IOB, based on the insulin measurement of step e), and the empirical model obtained in step 1). This method of calculating IOB or adjusted IOB is referred to as the "one-sample" method of calculating IOB or alOB.

[0047] The present invention further provides a test device and method using the realtime Insulin-On-Board calculation in order to more accurately determine the insulin bolus in a TIDM patient. In such as case, next to the insulin measurement(s), also the glucose level can be determined in said patient (e.g. in the same sample, simultaneously, sequentially, or separately). Measuring blood glucose and insulin can be done simultaneously in the same sample, or can be done subsequently with an interval of e.g. 1 second or more, 5 seconds, 10 seconds, 15 seconds, 20 seconds, 25 seconds, 30 seconds or more, 1 minute, 2, 3, 4, or 5 minutes or more, 10 minutes or slightly more than 10 minutes.

[0048] The calculation of the bolus injection or administration needed is then typically done using:

1. the patient's insulin to carb ratio, to calculate how much insulin is needed to absorb the carbohydrates from the next meal,

2. the patient's glucose correction factor to calculate how much insulin is needed to correct the pre-meal or fasting glucose level,

3. and the subtraction of Insulin On Board, i.e. the amount of insulin left over in the subcutis from the previous injection, calculated using the IOB or alOB estimation method according to the present invention. [0049] The insulin to carb ratio is the amount of insulin needed to absorb a standard quantity of carbohydrates from his next meal in said subject, and the glucose correction factor is the factor of insulin needed to lower the pre-meal blood glucose level in said subject to a target range.

[0050] With the term "algorithm" that can calculate the IOB or alOB using the estimation method according to the present invention is meant the algorithm: alOB = pAUC(T)/E[AUC/Dose], or any extrapolation thereof. The Dose is the total injected insulin (a known quantity). The AUC and the pAUC(T) can be calculated based on the insulin concentrations measured in two patient samples, i.e. one sample taken at the peak concentration time of the administered insulin, and one at the time of IOB calculation, typically around 3 to 4 hours, after the last injection. The value of E[AUC/Dose] can be calculated as the average, or moving average, or Kalman filter, or other estimator of the expected value of the AUC/Dose. The term "software program" that can calculate the IOB or alOB using the estimation method according to the present invention is meant to encompass said algorithm, and is capable of receiving input from the user and can provide output to said user. The input needed in the algorithm is the concentration(s) of insulin measured in the sample(s) of the subject, at time-point(s) τ, and the time(s) of measurement, or the time(s) lapsed between τ and the last insulin injection, as well as the total amount of insulin injected. The software next can fit a distribution curve to said insulin concentration(s), wherefrom the AUC value and the pAUC(T) can be calculated and where the AUC/Dose can be extracted and used in the algorithm (together with the previous collected AUC/Dose values). In the case of IOB, the estimation method according to the present invention is based on the equation IOB = CL pAUC(T), where CL is the patient specific plasma clearance of insulin determined in advance (e.g., by intravenous administration following standard pharmacokinetic methodology).

[0051] The present invention further provides a computer readable medium comprising an algorithm or software program that can calculate the IOB or alOB using the estimation method according to the present invention.

[0052] The present invention further provides a point of care test device comprising an algorithm or a software program that can calculate the IOB or alOB using the estimation method according to the present invention. [0053] Alternatively, the algorithm or software program that can calculate the IOB or alOB using the estimation method according to the present invention can be present on a remote controller, e.g. on a server, a laptop, a PC, a smart-phone, or the like, wherein said remote controller receives input from an insulin-measuring device, and calculates the IOB based on said input of insulin concentration(s) at specific time point(s). Said remote controller can either communicate directly with the measurement device through e.g. Short messaging service (SMS), Wifi, Bluetooth, near field communication, wireless, cable, or any other type of connection. Alternatively, the insulin concentrations received from the measuring device can be entered into the remote controller application by the user or a care professional.

[0054] In further embodiments, said algorithm or software program additionally can calculate the bolus amount to be administered to the subject, based on patient blood glucose levels, meal carbohydrate content, insulin resistance values, insulin/carb ratio, glucose correction factors and the like. Said blood glucose concentration(s) can be entered into the algorithm or software program similarly as the insulin concentration(s) as indicated above.

[0055] In a preferred embodiment, said algorithm or software program is embedded in, or connected to a measurement device that can measure both the glucose and insulin concentration in a small blood sample of the patient (e.g. a blood drop obtained through a finger prick).

BRIEF DESCRIPTION OF THE FIGURES

[0056] The present invention is illustrated by the following figures which are to be considered for illustrative purposes only and in no way limit the invention to the embodiments disclosed therein:

[0057] Figures 1 to 7: Time-dependent insulin concentration curves of 7 subjects from the clinical study (cf. Examples 2-4), with theoretical profiles fit to the measured insulin values at different time points (circles indicating the actual measured insulin levels), using two different methods: (1) the full fit (continuous line) of I(t) to all serial measures adjusting all four parameters (A, Θ, k, C) and (2) the simplified fit (dotted line) of I(t) to only two serial measures (at 1 hr and 3 hr) adjusting only two parameters (A, Θ) and holding the other two parameters fixed (k=2, C=0). [0058] Figure 8: Normalized Plasma Insulin Concentration Profiles from 3 Compartment Model by Wong et al., 2008 (Journal of Diabetes Science and Technology, Vol.2(3):436-449). The Three Compartment Model of Wong* was adapted as follows: k₂ = k₃ = 1.07 hr^"1 and k_d = k_di = 0 (i.e., no loss on route to plasma). The value of elimination rate (k_e) was varied form 0.5 x 9.6 hr^"1 to 4.0 x 9.6 hr^"1. The value of k₂ and k were adjusted to make t_max ~ 60 min. The normalized profiles (AUC = 1) vary little with the plasma elimination rate. However, the AUC is proportional to l/k_e.

[0059] Figure 9: Plasma Insulin Concentration from 3 Compartment Model (Wong et al., 2008) with random variation of model parameters, including a background concentration. Wong et al., 2008 with k₂ = k = 1.07 hr^"1, k_e = 9.6 hr^"1, k_d = k_di = 0, and a constant background plasma insulin concentration of C = 3.0 μΐυ/mL. The volume of distribution V_d and the dose are held fixed at 142.1 ml/kg and 0.10 IU/kg, respectively. The other parameters (k₂, k , k_e,C) are log normally distributed with a CV of 40% about their respective medians.

[0060] Figure 10: Parameter Variation Combined with Simulated Measurement Error. Simulation based on Wong et al., 2008 with the parameter distributions described in Figure 9. Profiles are sampled at 15 minute increments with log normally distributed errors of 10%.

[0061] Figure 11: Simulation based on Wong et al., 2008 modified as in Figure 10, and Estimation of PK Parameters Given Two Samples per Bolus, examples of estimated profiles that closely approximate the simulated-true profile.

[0062] Figure 12: Simulation based on Wong et al., 2008, modified as in Figure 10 and Estimation of PK Parameters Given Two Samples per Bolus, examples of estimated profiles that roughly approximate the simulated-true profile.

[0063] Figure 13: Estimation of adjusted IOB (alOB) for 1000 simulated profiles with 2-sample estimation method as disclosed herein (with samples at 1 and 3 hours). Only 1 profile could not be estimated based on this two sample method. The scatter plot shows the estimated alOB/Dose (y-axis) versus the simulated-true values of the alOB/Dose (x- axis).

[0064] Figure 14: Comparison of the adjusted IOB accuracy using no sampling (where the estimate is equal to the population median of alOB) or using the 2-sample estimation method as disclosed herein. The figure shows the cumulative distribution of the difference between the simulated-true alOB/Dose and the estimated alOB/Dose (same data points as in Figure 13). The ideal situation is a steep S-curve, wherein the simulated alOB and estimated IOB are the same (approximating 0). Clearly, the 2-sample method (samples taken at 1 and 3 hrs after bolus) is closer to the actual alOB values, especially in the dangerous underestimation zone (positive values).

[0065] Figure 15: Same comparison as in Figure 13, but now with the second sample taken at 4 hrs after the bolus. Only 3 profiles could not be estimated based on this two sample method.

[0066] Figure 16: Same comparison as in Figure 14, but now with the second sample taken at 4 hrs after the bolus.

[0067] Figure 17: Estimation of 1000 simulated profiles with a 1-sample estimation model as disclosed herein, wherein the single sample is taken at 3 hrs after bolus and wherein the model is trained by recording the insulin twice after each bolus (1 hr and 3 hrs after bolus), for 3 times per day, for 25 days, resulting in 75 profiles. The black curve shows an empirical model using the concentration at a single sample (3 hr) to predict the estimated adjusted IOB based on two samples (1 hr, 3 hr). Whereas the black curve is based on all the data (1000 profiles), the thick grey curve is an estimate based on a limited amount of training data (75 profiles, corresponding to 25 days x 3 bolus per day). The median alOB/Dose of the training data corresponds to a patient specific model that is static and requires no additional sampling.

[0068] Figure 18: Comparison of IOB accuracy using no sampling, adjusted IOB using the 2-sample estimation model as disclosed herein, or adjusted IOB using the trained 1 sample estimation model (empirical model) as disclosed herein. The ideal situation is a steep S-curve, wherein the simulated IOB and estimated adjusted IOB are the same (approximating 0). The adjusted IOB calculated using the 1-sample method (trained, and sample taken at 3 hrs after bolus), or the 2-sample method (samples taken at 1 and 3 hrs after bolus) is closer to the actual alOB values than the fixed method using no sampling, especially in the dangerous underestimation zone (positive values).

[0069] Figure 19: Flow-chart of how the test works for type-I diabetes mellitus patients. A blood sample is deposited on the test strip, which is placed in the test device. The test device measures the blood insulin level and optionally the blood glucose level in said sample. The user can interact with the device to enter the amount of carbohydrates in the meal to be digested and the target glucose level to be achieved by the user. The device then calculates the insulin need by using the glucose correction factor and the insulin to carbohydrate ratio. The device can also calculate the adjusted Insulin On Board based on the amount and time of a previous insulin injection(s) and the current concentration of insulin measured in the sample of blood obtained from the patient, using the IOB or alOB estimation model or method as disclosed herein. The device determines the amount of insulin required to accommodate the carbohydrate load in the next meal to be consumed and thereby seeks to bring the fasting glucose level to the target level. The calculated Insulin On Board is subtracted from the required dose of insulin and the next bolus amount is displayed. The user can also interact with the device to e.g. enter the date and time of the measurement.

[0070] Figure 20: Schematic representation of an exemplary disposable test strip for detecting glucose and insulin in a single drop of blood to be used in the methods according to the invention. This schematic represents a disposable test strip (5), comprising a sample receiving means (501), which is capable of distributing the sample into multiple microfluidic channels (502 to 503), for simultaneous detection of blood glucose level and insulin level. Each channel is accompanied with a pair of electrodes, a working electrode (508) and a counter/reference electrode (509). The test strip has four zones: a sample receiving zone (510), a sample distribution zone (511), a reaction zone (512) and an analyte detection zone (513). Each working electrode has a certain output signal (502a to 503a) and each counter/reference electrode has a certain output signal (502b to 503b), which can be read by a controlling device, designed to be in contact with said different electrodes and that can control the operation of the device and analyse the data obtained from the biosensor system. The number of channels is not to be seen as limited to the 2 channels represented herein, but may include more channels according to the function of the device. These channels can be used to measure glucose and one or more different kinds of insulin: for example the rapid acting forms of insulin, comprising insulin Lispro (from Eli Lily and company), insulin Aspart (from Novo Nordisk), insulin Gluisine (from Sanofi-Aventis); or long acting forms of insulin, comprising insulin Glargine (from Sanofi-Aventis), insulin Detemir (from Novo Nordisk).

[0071] Figure 21: Schematic representation of an exemplary glucose detecting sensor on one microfluidic channel of the test strip to be used in the methods according to the invention, a) The sample comprising glucose (603), is directed towards the sample reaction zone (512) through capillary force, b) In the reaction zone (512), glucose is oxidized (603') by a suitable oxidoreductase enzyme, for example glucose oxidase or glucose dehydrogenase (601), which is present in reaction zone (512). Said oxidation process releases electrons, which are transferred to the working electrode, e.g. by means of a suitable electron mediator (602).

[0072] The number of electrons liberated during the oxidation of glucose by the oxidoreductase enzyme system is proportional to the amount of glucose present in the sample and is measured as an output signal (502a*).

[0073] Figure 22: Schematic representation of an exemplary insulin detecting sensor on another microfluidic channel of the test strip to be used in the methods according to the invention, a) The sample comprising insulin (703), is directed towards the sample reaction zone (512) through capillary force, b) In the reaction zone (512), the insulin is bound by two antibodies: a first antibody, complexed with an enzyme label (701), and a second antibody, complexed with a magnetic particle (702), both present in the reaction zone. Upon metabolizing its substrate (704), present at the detection zone, the enzyme label (701) will generate an electrochemical signal, i.e. releasing one or more electrons, which are detected by the working electrode (508), placed in the detection zone, c) Outside the detection zone (513), e.g. in the reaction zone (512), a magnet (514) can be placed, which upon activation (514*), will draw away all magnetic bead-second antibody complexes from the detection zone. When insulin is present, it will be bound to the second antibody-magnetic bead and will hence be attracted to the magnet as well, together with the first antibody-enzyme complex. This reduces the amount of electrons produced at the site of the working electrode (508) and detection zone (513). Both signals 503a and 503a* can be detected by a reader. The difference in number of electrons formed at the working electrode before and after activation of the magnet is proportional to the amount of insulin in the sample, d) Alternatively, the magnet (514') can be situated at the working electrode (508) in the detection zone (513). e) When activated (514'*) said magnet can now attract the second antibody-magnetic bead complexes to generate electrons at the working electrode (508) where the substrate (704) is present in an amount proportional to the amount of insulin present in the sample. Such an assay may include a step to eliminate the non-bound enzyme label in order to increase the sensitivity and accuracy. [0074] Figure 23: Measurement of insulin (C-peptide) in 5 microliter whole blood samples from healthy subjects (n = 6). The blood samples were spiked with a known concentration of C-peptide, indicating the measurements are accurate in a range of 0 to ΙΟ.ΟΟΟρΜ, using the test device of Figure 20 and 21 (cf. Examples section).

[0075] Figure 24: Measurement of glucose in 5 microliter whole blood samples (n = 6, same subjects as in Figure 18), using the test device of Figure 20 and 22(cf. Examples section).

DETAILED DESCRIPTION OF THE INVENTION

[0076] As used herein, the singular forms "a", "an", and "the" include both singular and plural referents unless the context clearly dictates otherwise.

[0077] The terms "comprising", "comprises" and "comprised of" as used herein are synonymous with "including", "includes" or "containing", "contains", and are inclusive or open-ended and do not exclude additional, non-recited members, elements or method steps. The term also encompasses "consisting of and "consisting essentially of.

[0078] The recitation of numerical ranges by endpoints includes all numbers and fractions subsumed within the respective ranges, as well as the recited endpoints.

[0079] The term "about" as used herein when referring to a measurable value such as a level, an amount, a parameter, a temporal duration, and the like, is meant to encompass variations of and from the specified value, in particular variations of +/-10% or less, preferably +1-5% or less, more preferably +/-1% or less, and still more preferably +/-0.1% or less of and from the specified value, insofar such variations are appropriate to perform in the disclosed invention. It is to be understood that the value to which the modifier "about" refers is itself also specifically, and preferably, disclosed.

[0080] Whereas the term "one or more", such as one or more members of a group of members, is clear per se, by means of further exemplification, the term encompasses inter alia a reference to any one of said members, or to any two or more of said members, such as, e.g., any >3, >4, >5, >6 or >7 etc. of said members, and up to all said members.

[0081] All documents cited in the present specification are hereby incorporated by reference in their entirety. [0082] Unless otherwise specified, all terms used in disclosing the invention, including technical and scientific terms, have the meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. By means of further guidance, term definitions may be included to better appreciate the teaching of the present invention.

[0083] By means of further explanation and without limitation, "predicting" or "prediction" generally refer to an advance declaration, indication or foretelling of a disease or condition in a subject not (yet) having said disease or condition.

[0084] The terms "quantity", "amount", "concentration" and "level" are used as synonyms herein and are generally well-understood in the art. The terms as used herein may particularly refer to an absolute quantification of a molecule or an analyte in a sample, or to a relative quantification of a molecule or analyte in a sample, i.e., relative to another value such as relative to a reference value as taught herein, or to a range of values indicating a base-line expression of the biomarker. These values or ranges can be obtained from a single patient or from a group of patients.

[0085] An absolute quantity of a molecule or analyte in a sample may be advantageously expressed as weight or as molar amount, or more commonly as a concentration, e.g., weight per volume or mol per volume.

[0086] The term "sample" as used herein refers to a sample wherein the insulin and optionally glucose concentration can be measured such as to represent a physiological concentration or amount of insulin or glucose in said subject or patient from which the sample is obtained. Typical examples are blood or plasma samples, or interstitial fluid samples.

[0087] A relative quantity of a molecule or analyte in a sample may be advantageously expressed as an increase or decrease or as a fold-increase or fold-decrease relative to said another value, such as relative to a reference value as taught herein. Performing a relative comparison between first and second parameters (e.g., first and second quantities) may but need not require to first determine the absolute values of said first and second parameters. For example, a measurement method can produce quantifiable readouts (such as, e.g., signal intensities) for said first and second parameters, wherein said readouts are a function of the value of said parameters, and wherein said readouts can be directly compared to produce a relative value for the first parameter vs. the second parameter, without the actual need to first convert the readouts to absolute values of the respective parameters.

[0088] The term "insulin-on-board" or "IOB", or "physical IOB" is the amount of insulin present in the body, which has not been metabolized or has not entered the blood stream (i.e. is still present sub-cutis). Alternative names are Bolus On Board (BOB), Active Insulin, and Insulin Remaining. It refers to the part of insulin previously injected or delivered that is still (to become) active (working) in the body. The "Insulin On Board", is typically calculated using the "one-fifth rule", which assumes a consumption of l/5th of the administered amount of insulin per hour since the injection. This linear PK model is a crude approach, since the insulin absorption into the blood stream and hence its activity is far from linear. Various factors will influence insulin uptake from the subcutaneous injection site into the bloodstream (injection in the sub-cutis of the abdomen versus arm, cicatrized tissue versus vital tissue, vasodilatation due to ambient temperature versus vasoconstriction, stressful moment versus a relaxed moment, as well as losses of insulin through degradation, fixation or complexation in the body). Therefore the residual IOB determined solely based on the amount of time elapsed since last injection is often inaccurate. PK models with static coefficients (population based, or patient specific) will not capture the intra-individual variations in the both the rate and magnitude of insulin appearance in blood from injection to injection.

[0089] The term "adjusted IOB" or "alOB" refers to the estimated IOB, calculated based on the estimation method according to the invention. In essence, this is the "physical IOB" adjusted for the estimated loss of insulin e.g. at the injection site, which will not reach the plasma or will not become active. The adjusted IOB may be compared directly with the total injected insulin and is hence a corrected IOB for patients receiving insulin bolus through subcutaneous administration, either though self-injection, inhalation or through a catheter-linked insulin pump.

[0090] The term "insulin" as used herein in general encompasses all detectable forms and fragments of insulin and can be produced by the subject (endogenous) or can have been administered exogenously. In combination with the IOB or alOB estimation model according to the present invention, the term "insulin" refers to the exogenous insulin delivered to the subject as a bolus through injection, inhalation or through the insulin pump. Typically, this is a fast-acting type of insulin, needed to digest the meal. [0091] The term "close to the peak concentration time" refers to the time at which the bolus insulin concentration in the blood reaches its maximum. This time will differ from patient to patient, and sometimes even from injection to injection. Said time is also dependent on the type of insulin. For example using Lispro, the peak concentration time implies approximately 1 hour after the last bolus injection, but can differ from patient to patient. In general, the peak concentration time will be somewhere between 0.5 and 2 hours, preferably between 0.5 and 1.5 hours, most preferably at about 1 hour, after the last bolus injection when e.g. using Lispro, but will be depending on the type on insulin (very fast acting, fast-acting, regular,..) and the route of administration (subcutaneous injection, in situ catheter, aerosol, nasal spray, etc.). In some cases the peak concentration time can hence be around 30 minutes, i.e. 15 to 45 minutes after bolus administration, or faster. Alternatively, the peak concentration time can patient specific, as shown in Example 2, Table 2, wherein said peak concentration time is equivalent to ((k-l)9) based on the full fit parameters, wherein theta is a scale parameter and k is the shape parameter of the Gamma distribution model (based on the full fit) of a patient's insulin in blood concentration measurement vs. time curve. The population average peak concentration time will be indicated on the product sheet or patient information sheet of the insulin used for the bolus administration. Some exemplary averaged peak concentration times ("peak of action") are given in the table below.

[0092] The exogenous insulin can be administered in basically four formats:

1. Human insulin that appears in the blood undistinguishable from the endogenous insulin

2. Insulins that have been recombinantly modified and are hence also distinguishable using specific antibodies directed to the modified amino acids. One such a recombinant is a extra-short and fast working insulin such as: Humalog (Lispro), NovoLog (Aspart), Apidra (Glulisine)

3. Other recombinant forms are the extra-long acting insulins, such as: Lantus (Glargine).

4. Levemir (Detemir) is an insulin where a fatty acid chain is bound to prolong its half-life in the sub-cutis. The slower release makes it a long acting insulin. Once it enters the blood stream, it becomes indistinguishable from human insulin. [0093] Any combination of the insulins above can be used in one patient. One can therefore decide to measure all types using a single general antibody-pair, or one can decide to detect the amount of long- or short- acting insulin separately, depending on the condition or disease state of the subject, using specific antibody pairs that uniquely detect the particular insulin form. Some examples can be:

[0094] - "rapid-onset insulin" or "fast-acting insulin", which has a peak time of about one hour and lasting for three to five hours. This type of insulin is typically used directly before eating: the bolus insulin.

[0095] - "short acting insulin" begins to lower blood glucose levels within 30 minutes. It has peak effect of four hours and works for about six hours.

[0096] - "Intermediate acting insulin" has either protamine or zinc added to delay their action. This human insulin starts to show its effect about 90 minutes after injection, has a peak at 4 to 12 hours, and lasts for 16 to 24hours.

[0097] - "Mixed insulin" is a combination of either a rapid onset-fast acting or a short acting insulin and intermediate acting insulin. Advantage of it is that, two types of insulin can be given in one injection. When it shows 30/70 then it means 30% of short acting is mixed with 70% of intermediate acting insulin.

[0098] - "Long acting insulin" There are two kinds of long acting insulin available in the market: Lantus (Glargine) - It has no peak period as it works constantly when released into the bloodstream at a relatively constant rate, (full 24 hours) and Levemir (Detemir) - It has a relatively flat action, and can last up to 24 hours and may be given once or twice during the day.

[0099] The table below provides some exemplary but non-limiting insulins that are suitable for treating patients with diabetes and that could be measured using the device and method according to the present invention: Types of Peak of Duration

Examples Onset of Action

Insulin Action of Action

Rapid- acting Humalog (lispro) 15 minutes 30-90 3-5 hours

Eli Lilly minutes

NovoLog (aspart) 15 minutes 40-50 3-5 hours Novo Nordisk minutes

Short- acting Humulin R 30-60 minutes 50-120 5-8 hours (Regular) Eli Lilly minutes

Novolin R

Novo Nordisk

Intermediate- Humulin N 1-3 hours 8 hours 20 hours acting (NPH) Eli Lilly

Novolin N

Novo Nordisk

Humulin L 1-2.5 hours 7-15 hours 18-24 Eli Lilly hours Novolin L

Novo Nordisk

Mixed acting Humulin 50/50 The onset, peak, and

Humulin 70/30 duration of action of

Humalog Mix 75/25 these mixtures would

Humalog Mix 50/50 reflect a composite of

Eli Lilly the intermediate and

Novolin 70/30 short- or rapid- acting

Novolog Mix 70/30 components, with one

Novo Nordisk peak of action.

Long-acting Ultralente 4-8 hours 8-12 hours 36 hours

Eli Lilly

Lantus (glargine) 1 hour None 24 hours Aventis [00100] It is important in certain situations to know the origin of low or high glucose levels in a patient. By measuring the different types of insulin being used by such patient, one may identify the specific problem; that is which of the insulin products, if any, being used is causing the extremes of glucose. Without measuring the different types of insulin, it is difficult to accurately alter the specific dose of a given insulin form to achieve desired glucose profile.

[00101] Type-l-diabetes mellitus

[00102] Type-l-diabetes mellitus (T1DM), is typically characterized by recurrent or persistent hyperglycemia, and is diagnosed by demonstrating any one of the following:

[00103] - Fasting plasma glucose level at or above 7.0 mmol/L (126 mg/dL),

[00104] - Plasma glucose at or above 11.1 mmol/L (200 mg/dL) two hours after a 75 g oral glucose load as in a glucose tolerance test,

[00105] - Symptoms of hyperglycemia and casual plasma glucose at or above 11.1 mmol/L (200 mg/dL).

[00106] (cf. World Health Organisation: Department of Noncommunicable Disease Surveillance (1999). "Definition, Diagnosis and Classification of Diabetes Mellitus and its Complications")

[00107] T1DM is usually treated with insulin replacement therapy. This can be done using subcutaneous injection of insulin, with insulin inhalation or with an insulin pump, along with attention to dietary management (especially carbohydrates), and monitoring of blood glucose levels using glucose meters which can be simply operated by the patient himself. Insulin administration/injection is generally managed by the patient. Untreated T1DM commonly leads to coma, often from diabetic ketoacidosis, which can be fatal.

[00108] Once T1DM is fully established, the remaining beta-cell activity of the patients is often non-existent or too low to sufficiently regulate the blood-sugar homeostasis and administration of exogenous insulin is needed. The device and method according to the present invention can be used to determine the actual need for insulin at the time of blood glucose monitoring and calculation of the insulin bolus dose, e.g. before every meal. Typically, T1DM patients will have to administer a certain amount of long-acting insulin to have a base line level of insulin in their system and a bolus amount of short-acting insulin just before each meal. The base level is given by long acting Insulin which is administered once per day. The bolus short acting insulin needs to be given before each meal, usually 3 times a day. This bolus needs to be injected before every meal, in order to be able to properly take up the sugars released from the meal. Nowadays, complicated schemes exist that allow T1DM subjects to calculate the bolus dose of insulin needed before a meal, based on their current glucose level, the amount of carbohydrates in their anticipated meal and two factors:

1. the insulin/carb ratio,

2. the glucose correction factor, and

3. subtracting the IOB.

[00109] Based on this scheme, the subject calculates the amount of (short acting) insulin needed for the bolus injection (cf. e.g. more information on https://dpg- storage.s3.amazonaws.com/dce/resources/Insulin to Carb Slick.pdf; Zisser et al., 2008, Diabetes Technology & Therapeutics. Vol.10(6) :441-444; McKeown et al., 2004, Diabetes Care, Vol.27(2):538-546); Bevier et al., 2007, Diabetes/Metabolism Research and Reviews, Vol. 23(6):472-478; 2005, Keskin et al., 2005, Pediatrics Vol.115(4):e500- e503).

[00110] The "insulin to carb ratio" is used to determine the amount of insulin needed to absorb the carbohydrates in the next meal. For example, the amount of insulin needed to absorb each 15 grams of carbohydrate in the next meal can be used as a determining factor. If the insulin to carb ratio is 1.5; then the patient needs 1.5 units of insulin for each 15 grams of carbohydrates in his next meal. Should the patient consume 60 grams of carbohydrate, they would require 60grams/15grams X 1.5units = 6 Units of insulin. This insulin/carb ratio is given to the patient by the doctor at the time of diagnosis of his diabetes and is changed when the doctor sees the need for it, at a future consultation. In reality however, this value changes from individual to individual and from day to day within an individual, depending on several factors, including the individual's insulin resistance. Thus the insulin to carbohydrate ratio changes as a function of time and between individuals.

[00111] The "glucose correction factor" is the amount of insulin needed to bring down the measured glucose level to the target range. For example when a patient has a glucose level of 250mg/dL, and his target upper- limit is 150mg/dL, he needs to lower his blood glucose level by lOOmg/dL. When the glucose correction factor is 30, then the patient needs to inject 100/30 = 3.3 Units insulin to lower his glucose to the target range.

[00112] To know the total bolus amount, these 3.3 Units of insulin need to be added to the amount of insulin needed to digest the meal, as explained above for the calculation of the insulin to carbohydrate ratio.

[00113] The glucose correction factor is nowadays set by the healthcare consultant but is in fact a measure for insulin resistance. It is crudely calculated based on the patient's Total Daily Dose of insulin (long acting + all short acting boluses) and a number from 1800 to 2200 (depending on the kind of insulin the patients uses). The glucose correction dose in a patient who uses 20 units would be 1800/20 = 90 to 2200/20 = 110. The glucose correction dose is the number of mg/L that the blood glucose will drop for every unit of insulin injected. For a patient on 20 units of insulin/day: 1800 / 20U = a 90 mg/dL drop per unit of insulin (Humalog). Whether the doctor would use 1800, 2200 or any number there between to determine the glucose correction factor depends on the patient's insulin sensitivity and the kind of insulin that is used.

[00114] The present invention provides means and methods for more accurate estimation of IOB and alOB, based on one or more actual measurement(s) of blood insulin. Based on the time and the amount of insulin injected at the previous injection, coupled with the measured level of insulin concentration determined immediately prior to the next injection, a more accurate and real time determination of IOB and alOB leads to more accurate insulin dosing. This will result in reduced number of hypoglycaemic events, particularly during sleep and in reduced numbers of hyperglycaemic events in exogenous insulin dependent patients.

[00115] Exemplary testing device used

[00116] The present invention provides in the use of test devices for the estimation of IOB or alOB as taught herein comprising means for detecting the level of insulin and optionally glucose in a blood, serum, or interstitial fluid sample of the patient. In a more preferred embodiment, such device can be used in clinical settings or at home.

[00117] The device can be used for calculating the correct insulin bolus needed for a subject in order to safely digest a coming meal, or can be used to monitor or predict the glucose metabolism that will take place in the next hours. The device can be in the form of a home test device or a point of care test device (POC). The device can assist a medical practitioner, or nurse to decide whether the patient under observation is being correctly treated, or whether treatment schemes and/or insulin bolus regimens should be adjusted.

[00118] The device can be used to assist a subject having diabetes to control or fine-tune the amount of insulin needed during the day or before a meal or allows him to monitor his insulin on board throughout the day, e.g. according to the physical state or condition of the subject.

[00119] Typical devices comprise a means for measuring the amount or level of insulin in a blood sample, visualizing the amount of insulin in said sample, and means to calculate the IOB in said patient, based on the amount of insulin measured. Optionally, said device additionally can measure and visualise the concentration of glucose in a blood sample and use this information to adjust the insulin bolus amount needed throughout the day or before taking in a carbohydrate-containing meal.

[00120] In a preferred embodiment, the device is a lateral flow device. Such lateral flow device comprises a test strip allowing migration of a sample by capillary flow from one end of the strip where the sample is applied to the other end of such strip where presence of an analyte in said sample is measured. In another embodiment, the invention provides a device comprising a reagent strip, encompassing a reaction zone which will yield a quantitative signal upon interaction with the analyte. This signal can be generated by electrochemical or optical/photometric systems.

[00121] A "binding molecule" as intended herein is any substance that binds specifically to its target. In the light of the present invention, said binding molecule is intended to be specifically binding a certain type of insulin, i.e. corresponding to the type(s) of insulin that is (are) exogenously administered to the patient. Examples of a binding molecule useful according to the present invention, include, but are not limited to an antibody, an antibody fragment, a polypeptide, a peptide, a lipid, a carbohydrate, a nucleic acid (aptamer, spiegelmer), peptide-nucleic acid, peptide-aptamer, small molecule, small organic molecule, or other any other binding agent.

[00122] According to an aspect of the invention, a "binding molecule" preferably binds specifically to said one or more markers with an affinity of at least, or better than 10^~6 M. A suitable binding molecule can be determined from its binding with a standard sample of said one or more markers. Methods for determining the binding between binding molecule and said any one or more markers are known in the art. As used herein, the term antibody includes, but is not limited to, polyclonal antibodies, monoclonal antibodies, humanised or chimeric antibodies, engineered antibodies, and biologically functional antibody fragments (e.g. scFv, nanobodies, Fv, etc) sufficient for binding of the antibody fragment to the protein. Such antibody may be commercially available antibody against said one or more markers, such as, for example, a mouse, rat, human or humanised polyclonal or monoclonal antibody.

[00123] Electrochemical analyte detection

[00124] In currently available home tests or POC tests, the blood glucose level is typically measured using electrochemical detection methods. Many glucose meters employ the oxidation of glucose to gluconolactone catalyzed by glucose oxidase or glucose dehydrogenase.

[00125] Test strips typically contain a capillary channel that adsorbs a reproducible amount of the blood sample. The glucose in the blood reacts with an enzyme electrode containing glucose oxidase or dehydrogenase and the enzyme is oxidized with an excess of an electron-mediator. The mediator in turn is oxidized by reaction at the electrode, which generates an electrical current. The total charge passing through the electrode is proportional to the amount of glucose in the blood that has reacted with the enzyme. There are two ways of analyzing the charge yielded: a coulometric method (total amount of charge generated by the glucose oxidation reaction over a period of time), or an amperometric method (measures the electrical current generated at a specific point in time by the glucose reaction). The coulometric method can have variable test times, whereas the test time on a meter using the amperometric method is fixed. Both methods give an estimation of the concentration of glucose in the blood sample.

[00126] In essence, the amount of glucose is detected by measuring the charge yielded between two tiny electrodes, which can e.g. be printed on a disposable test strip to which a drop of blood of the subject is added. One of these electrodes encompasses an amount of the glucose oxidase or dehydrogenase enzyme and a certain amount of electron transfer mediator. The glucose present in the blood drop is oxidized by the oxidase or dehydrogenase, which releases (an) electron(s) proportionate to the amount of glucose that is present in the sample. These electrons are then transferred to the second electrode and the current is measured by a simple charge (Volt-Ampero)-meter, and the amount of measured electrons is then extrapolated to the blood glucose level of the subject doing the test.

[00127] Insulin blood level home tests or POC tests are currently being developed. One possible test device for use according to the present invention detects insulin based on an electrochemical immunoassay detection system. [00128] In essence, any electrochemical system can be used. One example is to label the analyte- specific antibody with any charged molecule or particle.

[00129] Preferred examples could be metal particles such as Al³⁺, Ag⁺, Au³⁺, Cu²⁺, and the like. Non-magnetic particles may be preferred for reasons set out below. The antibody-analyte complexes can then be detected by using a second antibody specific for the analyte, which can e.g. be fixed to an analyte detection zone on the test strip, or which is attracted to said zone by other means such as e.g. magnetism (see below). The analyte detection zone comprises a set of 2, 3 or more electrodes, with at least two oppositely polarized electrodes (a working or detection electrode and counter electrode) forming an electrode couple and optionally monitored by a reference electrode. The now fixed antibody-analyte-antibody-charged-label complex is then directed to an opposite charged electrode by inducing a charge or electric current between both electrodes. The antibody- analyte complexes are now attracted to the opposite charged electrode (e.g. positive charged particles will be attracted to the negative pole of the electrode couple). The charge or current is then reversed, thereby releasing the complexes and moving them to the opposite electrode and the current resulting from this change is measured. The measured total current received at the second electrode or at the reference electrode is proportional to the amount of complex that was displaced from the first electrode. In between the two working electrodes, a reference electrode may be placed, in order to simplify the distinction between the induced current and the current caused by the displacement of the labeled antibody-analyte complexes.

[00130] In said embodiments, the charged particle-antibody-analyte complex can be attracted to the reaction zone by using a second antibody which carries a magnetic particle. Inducing magnetism at the reaction zone will attract all second-antibody-antigen- antibody-charged-label complexes and the non-bound reagents will no longer interact with the test.

[00131] In a preferred embodiment of the testing device used in the present invention, detecting both glucose and insulin levels in a blood sample of a subject comprises a disposable test strip which can receive a drop of blood. Said strip preferably comprises a) a sample receiving part; and b) an analyte reaction zone comprising: bl) a first electrochemical or optical sensor for detecting the blood glucose level in said sample, and b2) a second electrochemical or optical sensor for detecting the blood insulin level in said sample. The sample is directed to the different zones through multiple microfluidic channels on the strip. The testing device further comprises c) a controlling device that can control the operation of the device and analyze the data obtained from the biosensor systems; and d) a user interface, displaying the data to the user. The schematic in Figure 20 represents an exemplary disposable test strip (5), comprising a sample receiving means (501), which is capable of distributing the sample into two or more multiple microfluidic channels (502 to 503), for simultaneous detection of blood glucose level (e.g. 502) and insulin level (503). Each channel is equipped with a pair of electrodes, a working electrode (508) and a counter/reference electrode (509). The test strip comprises four zones: a sample receiving zone (510), a sample distribution zone (511), a reaction zone (512) and an analyte detection zone (513). Each working electrode has a certain output signal (502a and 503a) and each counter/reference electrode has a certain output signal (502b and 503b), which can be read by a controlling device, designed to be in contact with said different electrodes and that can control the operation of the device and analyze the data obtained from the biosensor systems. The number of channels is not to be seen as limited to the 2 channels represented by the exemplary embodiment described herein with respect to Figure 20. In principle, two channels will suffice, since two analytes, namely glucose and insulin need to be detected. Other channels can be supplied for detecting other interesting blood analytes, or can be used as control channels, or to permit multiple measurements e.g. in different concentration ranges of the same analyte. Multiple measurements of glucose and insulin can be made in multiple channels, in order to reduce the error margin and increase the accuracy of the measurements. In addition, multiple channels can be used to measure different types of insulin; e.g. fast acting or long acting insulin.

[00132] In a preferred embodiment said first sensor bl) (e.g. 502 in Figures 20 and 21) for detecting glucose typically comprises a screen printed working and counter/reference electrode on the disposable test strip. To the working electrode, an amount of oxidoreductase, such as glucose oxidase or glucose dehydrogenase is attached, in combination with an amount of electron-transfer mediator. The glucose in the blood sample brought onto the test strip is oxidized by the oxidoreductase present on the working electrode, thereby releasing a proportional amount of electrons, transferred by the mediator to the counter/reference electrode. The current measured between both electrodes is proportional to the amount of glucose in the blood sample. Figure 21 exemplifies this process: a) The sample comprising glucose (603), is directed towards the sample reaction zone (512) through capillary force, b) In the reaction zone, it is oxidized (603 ') by glucose oxidase (601) present in the reaction zone. Said oxidation process releases electrons, which are transferred to the working electrode, e.g. by means of an electron mediator (602). The electron production of the glucose oxidase system is proportional to the amount of glucose present in the sample and is measured as an output signal (502a*).

[00133] In a preferred embodiment said second sensor b2) (e.g. 503 in Figures 20 and 22) for detecting insulin is an electrochemical sensor, measuring a change in charge or current due to enzymatic reaction with a substrate upon binding of insulin, more particularly an enzyme-linked immunomagnetic electrochemical assay. Said assay comprises: an electron-releasing enzyme system coupled to an insulin-specific antibody and secondary insulin-specific antibodies, linked to magnetic particles.

[00134] Upon contact with its substrate, an electron is formed by said enzyme and the current obtained through said enzymatic activity is measured. In the presence of an electron transfer mediator the electron-transfer mediated by the enzyme complex is monitored using for example a screen printed working (and counter/reference) electrode on the disposable test strip.

[00135] In order to avoid any washing steps, magnetic particles, linked to the second anti-insulin antibodies, are used to withdraw any insulin-bound enzyme complexes (complexed through a first anti-insulin antibody). The subsequent reduction in current signal generated at the working electrode versus the initial current signal prior to withdrawal of magnetic particle/insulin complexes is proportional to the amount of insulin present in the sample. Figure 22 exemplifies this process: a) The sample comprising insulin (703), is directed towards the sample reaction zone (512) through capillary force, b) In the reaction zone, the insulin is bound by two antibodies: a first antibody, complexed with an enzyme label (701), and a second antibody, complexed with a magnetic particle (702), both present in the reaction zone. The enzyme will produce electrons upon metabolizing its substrate (704), present in the detection zone, which in the presence of an electron mediator, will be detected by the working electrode (508), placed in the detection zone, c) Outside the detection zone (513), e.g. in the reaction zone (512), a magnet (514) is placed, which upon activation (514*), will draw away all magnetic bead-second antibody complexes from the detection zone. When insulin is present, antibody-magnetic particle-insulin will form. Such complexes are susceptible to a localized magnetic field, and as such will be attracted to the magnet (514) along with any of the first antibody-enzyme complex that has formed "sandwich" complexes with the target, insulin. Removal of first antibody-enzyme complexes from the reaction zone (512) leads to a reduction in reaction between enzyme label and substrate at the working electrode (508). This reduces the amount of electrons produced at the site of the working electrode (508) and detection zone (513). Both signals 503a and 503a* can be detected by a reader. The difference in number of electrons formed at the working electrode before and after activation of the magnet is proportional to the amount of insulin in the sample. The greater the amount or concentration of insulin present in the sample, the larger the reduction in signal measured at working electrode (508) following removal of immuno- complexes by magnet (514). Conversely, when little or no insulin is present in the sample, little or no reduction in signal occurs at working electrode (508) upon activation of magnet (514). d) Alternatively, the magnet (514') can be situated at the working electrode in the detection zone (513). e) When activated, said magnet (514'*) can now attract the second antibody-magnetic bead complexes to generate electrons at the working electrode, where the substrate (704) is present, in an amount proportional to the amount of insulin present in the sample. Such an assay may include a step to eliminate the non- bound enzyme label in order to increase the sensitivity and accuracy. This can be done through capillary forces for generating flow of the blood sample through the reaction zone and/or for eliminating non-bound complexes, or can be done by additionally adding an absorption pad or a capillary flow inducing means (e.g. the test strip itself) at the end of the detection zone (513) or capillary tract (503). Optionally, a reservoir with fluid, connected to said reaction zone (512) can be present to allow a better washing step.

[00136] In addition, any known insulin-measuring device or assay can be used to estimate the IOB or alOB according to the method or model of the present invention. Even laboratory assays for measuring blood insulin levels could be used. Further non- limiting examples of devices or assays are e.g. those disclosed in US patent US 6,893,552 Bl by Wang et al., and in PCT applications WO2012032294 and WO2012107717 from Multisense, or the devices disclosed in Xu et al., 2013 (Biosensors and Bioelectronics, vol.39:21-25), or in Kennedy et al., 2006 (Diabetes Care Vol. 29(1): 1- 8).

[00137] The invention further provides uses of a device for detecting both the glucose and insulin level in a whole blood sample of a subject comprising: a) a sample receiving part; b) an analyte reaction zone comprising bl) a first sensor for detecting the blood glucose level in said sample, b2) a second sensor for detecting the blood insulin level in said sample, c) a controlling device that can control the operation of the device and analyse the data obtained from the biosensor systems, and d) a user interface, displaying the data to the user, wherein said controlling device comprises the means to estimate IOB or alOB using the algorithm or estimation method according to the present invention.

[00138] Alternatively, said device, or its controller or controlling device, communicates with a remote controller comprising the means to estimate IOB or alOB using the algorithm or estimation method according to the present invention. This communication may be through any type of internet or wireless connection. Said remote controller can be any type of external server, or can be a laptop, personal computer, smart-phone, or any other device that can communicate with the measuring device and can calculate the IOB or alOB.

[00139] In a preferred embodiment, said second sensor b2) comprises two separate sensors, one for detecting endogenous insulin or its cleaved C-peptide fragment, and one for detecting exogenous insulin. The exogenous insulin can be fast-acting or slow acting. Preferably fast-acting and slow acting (or long acting or basal insulin) can be measured separately by the device.

[00140] In another preferred embodiment, said second sensor b2) comprises two separate sensors, one for detecting fast-acting insulin and one for detecting slow acting insulin (long acting or basal insulin).

[00141] In a preferred embodiment, the analyte reaction zone b) comprises at least two tracts, one for detecting blood glucose, and one for detecting blood insulin, wherein the latter can also comprise different tracts, for detecting different types of insulin (endogenous, short-acting, and/or long-acting).

[00142] In a preferred embodiment of the device according to the invention, the glucose and insulin are measured using a single sensor system, or using two separate sensor systems to detect each analyte separately. [00143] In a preferred embodiment, the device according to the invention is a home test device or a point of care device.

[00144] In a preferred embodiment of the device according to the invention, said insulin sensor is specifically detecting the type of insulin that has been administered to the subject (e.g. long-acting insulin, short-acting insulin, or both). Additionally, said device can specifically detect C-peptide cleaved from endogenously produced insulin, if the subject still has some endogenous insulin production. This is helpful to calculate the bolus amount needed, since the endogenous insulin will also still have some activity in metabolising the carbohydrates present in the meal.

[00145] Preferably, said first sensor is an electrochemical or optical sensor, and/or said second sensor is an electrochemical or optical sensor. Preferably, both sensors are electrochemical sensors. Alternatively, both sensors are optical sensors. Combined optical/electrochemical sensors are also envisaged by the invention.

[00146] In a preferred embodiment of the device according to the invention, the detection of both the glucose and insulin level is done in a sample volume of less than 1ml, preferably less than 0.5ml, more preferably in less than ΙΟΟμΙ, most preferably in less than ΙΟμΙ, or in about 5μ1 of whole blood.

[00147] In a preferred embodiment, the device according to the invention has a sensitivity of lOOpmol/1, preferably of 50pmol/l, more preferably of 20pmol/l or less for insulin.

[00148] In a preferred embodiment, the device according to the invention has a sensitivity of 20mmol/L or less for glucose.

[00149] In said device, the controller device can calculate the IOB, and alOB based on the signal(s) obtained from sensor b2, which are integrated in the estimation model according to the invention. In combination with the information received from sensor bl, the controller can then also calculate the bolus needed.

[00150] In a preferred embodiment, the device according to the invention additionally comprises an input means for introducing user-specific data such as time of measurement, time of last meal, time after exercise etc. into said controller, preferably comprising a keypad or a touch-screen, or any other means for feeding data to said device such as e.g. a wireless connection or a cable port. Said data could be fed from a PC, a portable computer, a smart phone or the like, communicating with said device or the controller thereof.

[00151] In a preferred embodiment, the device according to the invention, additionally comprises a connection with a computer, portable or mobile processing device, or a smart phone, to enable the user or medical practitioner to follow up his status, insulin need and/or beta-cell function. Said connection can be through a cable or wireless.

[00152] The above aspects and embodiments are further supported by the following non- limiting examples.

[00153] EXAMPLES

[00154] Example 1: Examples of electrochemical blood glucose and insulin detection test strips for use in calculation of IOB and/or insulin bolus.

[00155] a) Blood-glucose detection strip:

[00156] Screen printed working and reference electrodes are prepared on a disposable test strip which can receive a drop of blood. To the working electrode, an amount of glucose-oxidase is attached, in combination with an amount of electron-transfer mediator. The glucose in the blood sample brought onto the test strip is oxidized by the glucose- oxidase present on the working electrode, thereby releasing a proportional amount of electrons, transferred by the mediator to the reference electrode. The current measured between both electrodes is proportional to the amount of glucose in the blood sample.

[00157] b) Blood-insulin detection strip:

[00158] In this example, insulin detection based on an electrochemical immunoassay detection system is described, wherein an insulin- specific antibody is labeled with a charged molecule or particle. Said antibody is present in the reaction zone of the test device and is brought into contact with the blood sample through capillary forces. Upon binding of the insulin with the labeled-antibody, said complexes are trapped by a second insulin- specific antibody, linked to a magnetic particle, which is attracted to the reaction zone by magnetism.

[00159] The analyte detection zone comprises a set of electrodes, capable of inducing and receiving an electric charge and/or current between them. Two opposite charged electrodes form an electrode couple and optionally a reference electrode in the middle of said couple is present for ease of detection of the current produced. [00160] The fixed antibody-insulin-antibody-charged-label complex is then drawn to an opposite charged electrode by inducing an electric charge between both electrodes.

[00161] The antibody-analyte complexes are now attracted to the opposite charged electrode (e.g. positive charged particles will be attracted to the negative pole of the electrode couple).

[00162] The polarity of the electrodes is then reversed, thereby releasing the complexes and moving them to the opposite electrode. At the moment of the release, the current is measured between both electrodes. The measured total current received at the second electrode or at the reference electrode is proportional to the amount of complex that was displaced from the first electrode, since it will be the sum of the current induced and that caused by the complexes attracted thereto.

[00163] c) Combined insulin-glucose detection device

[00164] In this example, a device for detecting both the glucose and insulin level in a whole blood sample of a subject is described comprising a disposable test strip which can receive a drop of blood. Said strip comprises a) a sample receiving part; and b) an analyte reaction zone comprising: bl) a first electrochemical or optical sensor for detecting the blood glucose level in said sample, and b2) a second electrochemical or optical sensor for detecting the blood insulin level in said sample. The sample is directed to the different zones through multiple microfluidic channels on the strip. The device further comprises c) a controlling device that can control the operation of the device and analyze the data obtained from the biosensor systems; and d) a user interface, displaying the data to the user.

[00165] Said first sensor bl) for detecting glucose comprises a screen printed working and counter/reference electrode on the disposable test strip. To the working electrode, an amount of oxidoreductase enzyme, for example glucose oxidase or glucose dehydrogenase is attached, in combination with an amount of electron-transfer mediator. The glucose in the blood sample brought onto the test strip is oxidized by the oxidoreductase present on the working electrode, thereby releasing a proportional amount of electrons, which are transferred by the mediator to the counter/reference electrode. The current measured between the working and counter/reference electrodes is indicative to the amount of glucose in the blood sample. Figure 21 exemplifies this process.

[00166] Said second sensor b2) for detecting insulin is an electrochemical sensor, measuring a change in charge or current due to enzymatic reaction with a substrate upon binding of insulin, more particularly an enzyme-linked immunomagnetic electrochemical assay. Said assay comprises: an electron-releasing enzyme system coupled to an insulin- specific antibody and secondary insulin-specific antibodies, linked to magnetic particles.

[00167] Upon contact with its substrate, an electron is formed by said enzyme and the current obtained through said enzymatic activity is measured. The electron-transfer mediated by this enzyme system is then registered on a screen printed working (and counter/reference) electrode on the disposable test strip. Figure 22 exemplifies this process.

[00168] In order to avoid any washing steps, magnetic particles, linked to the second anti-insulin antibodies, are used to withdraw any insulin-bound enzyme complexes (complexed through a first anti-insulin antibody). The subsequent reduction in current signal generated at the working electrode versus the initial current signal prior to withdrawal of magnetic particle/insulin complexes is proportional to the amount of insulin present in the sample. Figure 22 exemplifies this process: a) The sample comprising insulin (703), is directed towards the sample reaction zone (512). b) In the reaction zone, the insulin is bound by two antibodies: a first antibody, complexed with the enzyme label (701), and a second antibody, complexed with a magnetic particle (702), both present in the reaction zone. The enzyme label (701) will metabolize its substrate (704) present in the detection zone in the presence of an electron mediator, thereby releasing electrons, which are detected by the working electrode (508), placed in the detection zone, c) Outside the detection zone (513), e.g. in the reaction zone (512), a magnet (514) is placed, which upon activation (514*), will draw away all magnetic bead- second antibody complexes from the detection zone. When insulin is present, antibody- magnetic particle-insulin will form. Such complexes are susceptible to a localized magnetic field, and as such will be attracted to the activated magnet (514*) along with any of the first antibody-enzyme complex that has formed "sandwich" complexes with the target, insulin. Removal of first antibody-enzyme complexes from the reaction zone (512) leads to a reduction in reaction between enzyme label and substrate at the working electrode (508). This reduces the amount of electrons produced at the site of the working electrode (508) and detection zone (513). Both signals 503a and 503a* can be detected by a reader. The difference in number of electrons formed at the working electrode before and after activation of the magnet is proportional to the amount of insulin in the sample. The greater the amount or concentration of insulin present in the sample, the larger the reduction in signal measured at working electrode (508) following removal of immuno- complexes by magnet (514). Conversely, when little or no insulin is present in the sample, little or no reduction in signal occurs at working electrode (508) upon activation of magnet (514).

[00169] d) Actual test measurement of insulin and glucose level in a small blood sample:

[00170] For this initial test, a small volume of whole blood (5 microliter) was spiked with a known concentration of C-peptide (part of insulin) and said samples (6 in total) were introduced at the sample receiving part (501) of the device as outlined in point c) above. Subsequently, the reagents were left to incubate for about 2 to 3 minutes and the concentration of insulin (Figure 23) and glucose (Figure 24) was measured in the reader using the steps as outlined in point c) above (sensor b2 and bl respectively). The blood samples were taken from healthy subjects. As can be seen from Figure 23, insulin concentrations from 0 to 10.000 pM could be measured in a 5 microliter blood sample. The amount of insulin (C-peptide) was calculated based on the difference of electrochemical current measured after the magnetic field is activated and withdraws bound magnetic -bead-antibody-insulin-antibody-label complexes from the reaction zone and the total electrochemical current measured before said magnetic field is activated and all label is still present. The blood glucose concentration was calculated based on the electrochemical signal obtained using the methodology outlined in step c) above (sensor bl).

[00171] This example provides the proof of concept that quantitative electrochemical measurement of insulin (or C-peptide) and glucose can be done in a small volume of whole blood (5 microliter).

[00172] Example 2: Estimating Insulin PK Profiles

[00173] A small clinical study of subcutaneous insulin pharmacokinetics was undertaken in January-February of 2013. A total of 9 T1DM subjects on combined Lispro (rapid- acting) and Lantus (long-acting) insulin therapy were enrolled. Fasting subjects provided an initial blood sample at t=0 and then injected their usual therapeutic amount of subcutaneous Lispro bolus (within 15 minutes), followed by breakfast. Serial blood draws were taken at 30 minute intervals, i.e., t=0, 30, 60, 90, etc., through 6 hours. The blood samples were processed to plasma and insulin concentrations were measured using the Lispro RIA by Millipore with a lower limit of detection of 2.0 μΐυ/mL. This assay is specific to the monomeric insulin Lispro. Although there were 9 subjects in the study, for reasons unknown, 2 (out of 9) were not evaluable due to extremely low (undetectable) insulin concentrations. In the remaining 7 subjects, evaluable plasma insulin concentration time- series were obtained allowing pharmacokinetic analysis.

[00174] Each set of serial measurements is fit to a theoretical profile I(t) representing the plasma insulin concentration as a function of time. The profile is written as

I(t) = A p(t; k, 9) + C,

[00175] with four parameters: an amplitude parameter (represented by A), a shape parameter (represented by k), a scale parameter (represented by Θ), and an additive background concentration that is not associated with the subcutaneous injection (represented by C). The function p(t; k, Θ) is a density function whose time integral (from 0 to infinity) is equal to unity. Therefore, the amplitude A is equal to the Area Under the Curve (AUC) of the subcutaneous injection. The function p(t; k, Θ) is taken as the Gamma probability density function with shape k and scale Θ (following standard mathematical notation). The scale Θ characterizes the rate of transport (or absorption) from the subcutaneous tissue into the plasma. The shape k is a dimensionless constant (in this case k>l) which tends to be approximately 2.0, but may be smaller than 2.0 for relatively fast profiles (mean/SD < 1) and larger than 2.0 for relatively delayed profiles (mean/SD > 1). The Gamma density function is peaked at (k-l)9, has a mean of k9, and a standard deviation (SD) of k^1/2 Θ, and therefore the mean/SD = k^1/2. Although k is a continuous parameter, it's theoretical interpretation is the integer number of compartments in the transport model (typically from 1 to 4, depending on the model). In the case of k=2, the density function is p(t) = (t/θ²) exp(-t/9). The density function p(t) need not be limited to the Gamma density function, as other probability density functions, or compartmental pharmacokinetic (multi-exponential) density functions may be used.

[00176] Figures 1-7 show the seven evaluable subjects from the clinical study with theoretical profiles fit to the measured values using two different methods: (1) the full fit of I(t) to all serial measures adjusting all four parameters (A, Θ, k, C) and (2) the simplified fit of I(t) to only two serial measures (at 1 hr and 3 hr) adjusting only two parameters (A, Θ) and holding the other two parameters fixed (k=2, C=0). Table 1 shows the relevant demographic and baseline characteristics of the seven evaluable subjects. Table 2 shows the model parameters estimated by the full fit (to all serial measures) and the model parameters estimated by the simple fit (to two serial measures).

[00177] Table 1. Demographics and baseline characteristics of evaluable subjects with T1DM from a clinical study of subcutaneous insulin pharmacokinetics.

[00178] Table 2. Parameter estimates of the Full Fit model and the Simple Fit model. Insulin concentration is in units of μΐυ/mL, A is equal to the AUC and has units of concentration x hours, theta has units of hours, k is dimensionless, and C has units of concentration. The Full Fit model estimates 4 free parameters using 13 serial measures. The Simple Fit model estimates 2 free parameters using only 2 serial measures (at 1 hour and at 3 hours).

Full Fit Parameters Simple Fit Parameters

(k=2, C=0)

0153-0003 67.3 2.38 0.67 2.9 119.7 0.80

0153-0006 30.6 3.36 1.04 3.1 106.9 4.09

0153-0007 74.2 2.45 0.88 2.4 92.2 1.26

0153-0008 69.5 3.09 0.69 1.4 90.3 1.19

0153-0009 217.0 1.98 0.96 6.5 253.8 1.02

[00179] Note, the background concentration (estimated by C in the Full Fit model in Table 2) is typically small compared to the peak concentration (Figures 1-7). This makes sense because the measurements were made with the Lispro specific assay (Lispro RIA). Therefore the cross-reactivity between the Lispro RIA and a patient's basal insulin is expected to be small and it appears that, to a large extent, there is no other source of plasma insulin (e.g., via sequestering by antibodies) that cross-reacts with the Lispro RIA.

[00180] Generally the Simple Fit is an excellent approximation of the Full Fit (as shown in Figures 1-7). However, this is not the case for Patient 0153-0006 (Figure 4) where the background constant is estimated as 3.1 μΐυ/mL and the peak concentration is only about 10 to 14 μΐυ/mL. Furthermore, in this same patient, the profile shape is relatively delayed (k = 3.36). Therefore, this particular Simple Fit based on only two measurements of plasma insulin and relatively poor assumptions for k and C would not be appropriate for the management of this particular patient. This would be corrected in practice by measuring the subject's fasting insulin level multiple times (over several days) to establish an optimal assumption for C and then also measuring several injections with three or more measures per injection to determine an optimal assumption for k and also an optimal time of the first insulin measure (which need not be 1 hour following injection, but may be delayed).

[00181] Example 3: Estimating IOB and a!OB [00182] The Insulin-on-Board (IOB) and the adjusted IOB can be estimated based on actual insulin measurements following the methods of this invention. The Simple Fit in the previous example (Table 2 and Figures 1-7) shows that the plasma insulin concentration I(t) may be modeled as a time dependent profile with only two free parameters, an amplitude parameter and a rate parameter, that may vary from patient to patient and injection to injection. While this is practical (because the patient only needs to make two measurements of plasma insulin per injection), it is not meant as a limitation. The method applies equally well to sampling with more than two measures per injection and equally well to fitting models with more than two free parameters (as shown by the Full Fit in the previous example, Table 2 and Figures 1-7).

[00183] Define the background subtracted plasma insulin concentration profile Γ (t) = I(t) - C, where C is the background insulin concentration. The background concentration may be determined in advance of all IOB calculations by measuring the patient's morning plasma insulin concentration at fasting levels, prior to any mealtime bolus. Preferably, the insulin assay used to measure the plasma insulin concentration is specific to the rapid acting insulin isoform, with minimal cross-reactivity to the long acting isoform, in which case the background concentration will be relatively small compared to the measured insulin levels and will therefore have only a small effect on the IOB calculations. This is shown, for example, in Table 2 where in one case (Full Fit) the values of C were estimated by fitting the model and in the other case (Simple Fit) the value of C=0 was assumed (based on the population).

[00184] The IOB is calculated from the background subtracted plasma insulin concentration profile I'(t) in a series of steps. The first step is to integrate I'(t) from t=0 to t=infinity thereby calculating the AUC (area under the curve). The second step is to integrate I'(t) from t=T to t=infinity thereby calculating the partial AUC which can be denoted as pAUC(T), where τ is the time-point of the desired IOB estimate (which will typically be equal to the second of the two serial measurements of the plasma insulin concentration). The third step is multiplication of pAUC(T) by a constant that converts the result into units of insulin dose (IU), e.g., IOB = CL x pAUC(T), where CL is the plasma clearance of insulin. If CL were determined a priori for the given patient (e.g., determined by intravenous administration of insulin followed by suitable serial measurements of the plasma insulin concentration), then the IOB thus calculated represents the physical amount of insulin still in transport (but not yet in the plasma) that will arrive in the plasma for time t > τ.

[00185] In the limit τ=0, this physical IOB is equal to CL x AUC which is equivalent to Dose x F, where F is the absolute bioavailability via the subcutaneous route of administration and Dose is the total quantity of insulin injected. (Note, F is a dimensionless proportion from 0 to 1). The missing insulin, equal to (1-F) x Dose, is lost in transport (presumably near the site of injection) prior to reaching the plasma and will never become biologically active in the patient. F may vary considerably from injection to injection, even within an individual. Let E[F] denote the expected value of F over the patient's distribution of F. Now define alOB = CL x pAUC(T) / E[F]. The Adjusted IOB (alOB) represents the number of injected units of insulin that are required to achieve a given level of physical IOB. In this way, the adjusted IOB accounts for the expected losses of insulin in the subcutaneous, i.e., units of insulin that will never make it into the plasma. Noting that AUC/Dose = F/CL (a general pharmacokinetic relationship), the equation for adjusted IOB may be re- written as follows: alOB = pAUC(T) / E[ AUC Dose], where E[AUC/Dose] is the expected value of AUC/Dose.

[00186] This formulation of the Adjusted IOB is particularly useful, because while the values of CL and F are not generally known for an individual and cannot be identified by direct measurement of plasma insulin concentrations following subcutaneous injection, the ratio of F/CL may be identified and therefore the Adjusted IOB is practical to measure. The Adjusted IOB is also easy to interpret. For example, in the case that τ = 0, the Adjusted IOB is equal to the dose multiplied by the ratio of the actual AUC to the expected AUC. If the actual AUC is larger than expected, the Adjusted IOB is larger than the applied dose, indicating that patient has more plasma insulin than expected, as if they had applied a larger dose. Similarly, if the actual AUC is smaller than expected, the Adjusted IOB is smaller than the dose applied, and the patient has less plasma insulin than expected, as if they had applied a smaller dose. More generally, the calculation of Adjusted IOB will be made for a typical value of τ in the range of approximately 3 to 5 hours following the previous injection at which point the patient would like to have a meal and is estimating the amount of insulin required for the mealtime bolus. In this case, PAUC(T) is smaller than AUC because of the time elapsed from t=0 to t=T.

[00187] Table 3 shows the adjusted IOB calculated at 3 hours for each patient in the clinical study based on the PK profiles described in Table 2. In this example, each patient was followed for only a single injection and therefore, without historical data, a practical assumption is that the expected AUC is equal to the observed AUC. In this case, the adjusted IOB is given by IOB/Dose = pAUC/AUC as shown in Table 3. For the Simple Fit, only two measurements of plasma insulin were made (at 1 hour and at 3 hours). In the case of the Full Fit, a total of 13 measurements of plasma insulin were made at 30 minute increments from t=0 to 6 hours following injection enabling a more accurate assessment of pAUC and AUC. Based on this data and method, additional examples of calculating the adjusted IOB can be generated using 2 (or more) measurements taken from t=0 through t=T, where τ is the desired time-point for the IOB calculation (typically 3 - 5 hours).

[00188] Table 3. The adjusted IOB calculated at 3 hours for each patient in the clinical study based on the PK profiles described in Table 2 (the Simple Fit and the Full Fit) where the adjusted IOB is given by IOB/Dose = pAUC/AUC and expressed in units of percent relative to each subject's injected dose.

[00189] Example 4: Estimating IOB with Historical Data

[00190] Following examples 2 and 3 above, the values of AUC and pAUC(T) are readily calculated for each injection. After the first injection, E[AUC/Dose] may be estimated as AUC/Dose. After two injections, E[AUC/Dose] may be estimated as the mean over two injections. After three injections, E[AUC/Dose] may be estimated as the mean over three injections, etc. For example, after one week of applying the method 3 times per day, E[AUC/Dose] may be calculated as the mean over 21 measurements. Over an even longer time period, E[AUC/Dose] may be calculated as a moving average, or weighted moving average, or Kalman filter. Based on this method, the Adjusted IOB may be estimated following each dose. The moving average allows the expected value to slowly vary over time, as the patient may undergo physiological changes.

[00191] Table 4 shows the adjusted IOB calculated at 3 hours for 7 injections of a single model subject based on the clinical study population of PK profiles described in Table 2. In this example, each injection represents a single model subject of 190 lb taking D units of insulin. The value of D was calculated as Dose x (190 lb)/Weight based on the clinical study (Table 1). The expected value of AUC/D is denoted as E[AUC/D] and is calculated as the arithmetic mean over the 7 injections. In this case, the adjusted IOB is given by IOB/D = (pAUC/D)/E[AUC/D] as shown in Table 4. For the Simple Fit, only two measurements of plasma insulin were made (at 1 hour and at 3 hours). In the case of the Full Fit, a total of 13 measurements of plasma insulin were made at 30 minute increments from t=0 to 6 hours following injection enabling a more accurate assessment of pAUC and AUC. Based on this data and method, additional examples of calculating the adjusted IOB can be generated using 2 (or more) measurements taken from t=0 through t=T, where T is the desired time-point for the IOB calculation (typically 3 - 5 hours).

[00192] Table 4. The adjusted IOB calculated at 3 hours for 7 injections of a model subject based on the PK profiles described in Table 2 (the Simple Fit and the Full Fit) where the adjusted IOB is given by IOB/D = (pAUC/D)/E[AUC/D] and expressed in units of percent relative to each injected dose. The model subject represents a subject of 190 lb injecting D units of insulin, where the value of D was calculated as Dose*(190 lbyWeight based on the clinical study (Table 1). Each injection is labeled according to the patient from which the PK model was derived.

Simple Fit, E[AUC/D] = 14.8 Full Fit, E[AUC/D] = 11.2

0153-0002 4.3 11.57 1.67 11.3% 14.11 1.69 15.1%

0153-0003 6.2 19.33 2.16 14.6% 10.86 1.09 9.7%

0153-0006 7.2 14.86 12.37 83.6% 4.26 2.29 20.4%

0153-0007 7.7 11.94 3.72 25.2% 9.61 2.15 19.2%

0153-0008 8.5 10.64 3.03 20.5% 8.19 1.71 15.2%

0153-0009 11.4 22.29 4.65 31.4% 19.06 3.38 30.2%

[00193] Example 5: Numerical Simulations of Insulin PK Profiles with IOB Estimation

[00194] In this experiment, numerical modeling (based on literature) is used to simulate a large number of monomeric (rapid acting) insulin PK profiles via subcutaneous administration. The selected model parameters (and the variability of these parameters) are aligned with the literature and with the observations from clinical studies, including our own observations in the clinical study described above (Examples 2-4).

[00195] The simulated plasma insulin profiles (representing the true insulin concentrations) are sampled with random (simulated) measurement error at two time- points and subsequently re-constructed using a simplified model (the method described in Example 2). For each simulated true profile and each re-constructed sampled profile, the adjusted IOB is estimated using the method described in Examples 3 and 4, i.e., alOB = pAUC(T) / E[ AUC/Dose], where E[AUC/Dose] is the expected value of AUC/Dose. This allows a direct comparison of the distribution of alOB True (based on the simulated profiles) to the distribution of alOB Estimate (based on the sampled profiles), as well as to the distribution of a fixed estimate (population based, without sampling) like the rule- of-fifths.

[00196] The modeling is starting from the Three Compartment Model of Insulin Pharmacokinetics, as disclosed in Wong et al., 2008 (J. Diabetes Sci Technol, Vol.2(3):436-449). Said Three Compartment Model of Insulin PK is represented as follows:

Simulated Plasma Insulin Concentration = AUC p(t) + C, wherein AUC = Dose F/CL;

CL is the plasma clearance = k_e x V_d, where k_e is the plasma elimination rate, and V_d is the volume of distribution (plasma plus peripheral);

F is the absolute bioavailability (includes all losses prior to plasma); p(t) is a normalized multi-exponential density profile whose time integral (0 to∞) = 1. It depends on three rate parameters describing transport.

[00197] The additive constant C represents a background concentration including basal insulin, or endogenous insulin, or slow release of antibody bound insulin, or any other mechanisms. First, the Three Compartment Model of Wong* was adapted as follows: k₂ = 1.07 hr^"1, k₃ = 1.07 hr^"1, k_e = 9.6 hr^"1, and k_d = k_di = 0 (i.e., no loss on route to plasma). Although, the mean transit time through the two subcutaneous compartments is (l/k₂ + l/k₃) = 112 min (per Wong), the values of k₂ and k₃ were adjusted first to make them equal (on average) and to make tmax ~ 60 min, which is aligned with our own clinical study (Example 2) and with the Lispro package insert.

[00198] Figure 8 shows 4 simulated plasma insulin profiles as a function of time from 0 to 6 hours following injection. In this figure, the profiles are normalized to AUC=1 (so the units of the y-axis are hr^"1) and the value of k_e was varied form 0.5 x 9.6 hr^"1 to 4.0 x 9.6 hr^"1. The normalized profiles vary little with the plasma elimination rate because k_e » k₂ (or k ). However, the AUC is proportional to l/k_e, so the amplitude of each curve will vary dramatically with k_e (not shown in Figure 8).

[00199] Next, the Plasma Insulin Concentration was simulated using the 3 Compartment Model* with random variation of the model parameters as log normally distributed with a CV of 40% around their respective median values k₂ = 1.07 hr^"1, k = 1.07 hr^"1, k_e = 9.6 hr^"1, and k_d = k_di = 0. In addition, V_d and Dose were held fixed at 142.1 mL/kg and 0.10 IU/kg, respectively (where the value of V_d is taken from Wong et al.). Noting that AUC/Dose = F/CL, the variation of F/CL is modeled by the assumed variation of k_e (so even if CL is nearly constant, this models the situation where F may vary substantially from injection to injection, despite the assumed parameters of k_d = k_di = 0). This approach is equivalent to (but simpler) than explicitly modeling the subcutaneous losses described by k_d and k_<ji in the model of Wong et al. In a further step, a background concentration (additive constant) is also included in each simulated profile, also with a CV of 40% about a median of 3.0 μΙΙ/mL. This distribution of the background constant is taken from our own clinical study (as shown in Example 2, Table 2, Full Fit). Figure 9 shows 10 plasma insulin profiles (in physical units of μυ/mL) selected at random from the simulations thus generated. With the underlying parameters varying by CVs of 40%, there is considerable variation between profiles. We believe this is consistent with the data collected in clinical studies (0sterberg et al., 2003, Journal of Pharmacokinetics and Pharmacodynamics, Vol. 30(3):221-235, especially Table 1; Heineman et al.,1998 Diabetes Care, Vol21(l l): 1910-1914, especially Table 1) demonstrating a high level of intra-individual variability from injection-to-injection.

[00200] In addition, the parameter variation was combined with simulated measurement error where each profile is sampled with log normally distributed errors with a CV of 10% about the simulated-true value at each time-point. Whereas the "true" simulated profiles are shown in Figure 9, the corresponding sampled profiles are shown in Figure 10 with sampling every 15 minutes.

[00201] The re-construction of the PK profiles is based on two serial samples (with measurement error) from each profile, e.g., with sampling at 1 and 3 hours. The first sample is taken near the expected peak of the profile (1 hour) and the 2nd sample is taken at the time desired for calculating the IOB (pre-prandial, prior to next insulin injection). The re-construction method follows the PK parameter estimation method (Example 2) using the Gamma probability density function with fixed k=2 and with fixed C=3.0 μΐυ/mL. If an alternative estimate of k was available and known in advance, then k≠ 2 may also be assumed (and may be appropriate for monitoring certain subjects). Otherwise k=2 is a reasonable assumption, based on empirical data (as shown in Example 1, Table 2, Full Fit) as well as theory (a simplification of the three compartment model, because the plasma elimination rate is large compared to the rate of subcutaneous transport). The specific choice of C=3.0 μΐυ/mL corresponds to the situation where a population estimate of C is available (it corresponds to the average of C from our clinical study data, Example 1, Table 2, Full Fit). Figure 11 shows three examples where the above method has been followed and where each re-constructed profile compares closely with the corresponding simulated-true profile. Figures 12(a) and 12(b) show examples of re-constructed profiles that only roughly approximate their corresponding simulated-true profiles due to relatively large error in one, or both of the serial samples (at 1 and 3 hours). These figures show that the re-constructed profiles are reasonable estimates of the simulated-true profiles despite measurement error, sparse sampling, and the simplified functional form of the re-construction.

[00202] Next, the re-construction method based on two samples (at 1 hr and 3 hr after bolus) was used to estimate the adjusted IOB at 3 hr following injection. The definition of adjusted IOB requires that E[AUC/Dose] be estimated. Here we take E[AUC/Dose] to be the known population median of the simulation. This models the situation where the patient was followed by the method for sufficient time to establish this expected value (as explained and demonstrated in Example 4). The results are shown in Figure 13 as a scatter plot of the simulated-true alOB/Dose based on the simulated-true profiles (x-axis) versus the estimated alOB/Dose based on the re-constructed profiles. Out of 1000 simulated profiles, only 1 profile could not be re-constructed by the method. Figure 13 shows that the estimated alOB correlates well with the simulated-true alOB (Pearson correlation coefficient of R=0.924), with relatively small systematic bias (shown by the linear regression, slope = 1.14 and intercept ~ 0). To interpret the quality of this alignment, it is important to recall that a population model of alOB without any measurements of the insulin concentration provides only a single value for the estimated alOB. For example, using the median of the simulated- true alOB data gives a fixed estimate of alOB/Dose = 0.20. Alternatively, the rule-of-fifths gives a fixed estimate of IOB/Dose = 0.40 at 3 hours. Note, the rule-of-fifths also does not distinguish between physical IOB and alOB and hence does not consider bioavailability in the estimation. The scatter plot in Figure 13 demonstrates significant improvement over such population estimates.

[00203] Similarly, the re-construction method based on two samples (at 1 hr and 4 hr after bolus) was used to estimate the adjusted IOB at 4 hr following injection and the results are shown in Figure 15. Out of 1000 simulated profiles, only 3 could not be estimated by two samples. Again there is high correlation (Pearson R = 0.910) and low bias (slope = 1.06 and intercept ~ 0) between the simulated-true alOB and the estimated alOB based on the re-construction method. At 4 hours, the median of the simulated-true alOB data gives a fixed estimate of alOB/Dose = 0.098 and the rule-of-fifths gives a fixed estimate of IOB/Dose = 0.20. [00204] The following analysis demonstrates the improvement of the estimated alOB based on sampling of the insulin concentration (two serial samples) over a fixed alOB estimate based on the population median (without sampling). The distribution of the difference between the simulated-true alOB/Dose and the estimated alOB/Dose is shown in the cumulative distribution plots for the alOB estimated at 3 hours (Figure 14) and for the alOB estimated at 4 hours (Figure 16). From Figures 14 and 16 it becomes clear that estimation based on sampling is more accurate than estimation without sampling (i.e. approaches the steep S-curve of almost no difference between "true" and "estimated" IOB). This is an important improvement, since the no- sampling method significantly underestimates the alOB in a large portion of the population (i.e. there is a higher amount of insulin present in the body than expected), which may lead to dangerous situations when the next bolus is administered, taking into account a too low value of alOB. The excess of insulin can lead to dangerous situations such as hypoglycemia and potentially even coma is this error is continuously used. Other populations estimates (e.g., the rule- of-fifths) may be biased high relative to the population median with the goal of reducing the underestimation of the IOB (or alOB), but with the inevitable consequence of significantly overestimating the IOB (or alOB) for the bulk of the population. Therefore, a population method without sampling of the plasma insulin concentration cannot hope to track the large variation from injection-to-injection.

[00205] From this experiment, it can hence be concluded that:

• Based on the theoretical 3 compartment PK model for subcutaneous injection of monomeric insulin, it was determined that the plasma elimination rate is large compared to the absorption rate and therefore the normalized profile (AUC=1) is closely approximated by a simplified model that is effectively described by only three free parameters, two rate constants (describing the transport from the subcutaneous tissue into the plasma) and an amplitude (that depends on the plasma elimination rate, the volume of distribution, and the bioavailability).

• The amplitude is given by AUC = Dose*F/CL, where F is the absolute bioavailability and CL is the plasma clearance.

• The Estimation assumes a single rate constant for the two subcutaneous compartments and uses the following functional form I'(t) = A p(t; k, Θ), where Γ (t) is the background subtracted plasma insulin concentration, A is the amplitude (equivalent to the AUC), and p(t; k, Θ) is the Gamma probability density function with shape k and scale Θ (equivalent to the inverse of subcutaneous absorption rate constant).

• The Estimation assumes that the background concentration is known and that the shape k is also known.

• For each bolus, the two parameters A and Θ are estimated from 2 plasma insulin samples and the Adjusted IOB is estimated from the parameters A and Θ.

• Simulated-true PK profiles were generated via the 3 compartment model with 40% CV in the three free parameters, plus 40% CV in the background concentration,

• Re-constructed profiles were based on two insulin samples of each profile where each sample has 10% CV (measurement error) relative to the simulated-true insulin concentration.

• The 1st sample should be approximately at the peak of the profile, i.e., at about 1 hour when using Lispro (may vary amongst types of insulins).

• The 2nd sample is taken at the time used for estimating the IOB.

• The metric S = (alOB True - alOB Est)/Dose was calculated and compared to fixed estimation (population median) without insulin sampling.

• The insulin sampling leads to improved IOB estimates, i.e., tighter values of S, particularly when the IOB is larger than expected.

• For 3 hr post injection, the 75th% of S is 0.152 without sampling compared to 0.031 with sampling.

• For 4 hr post injection, the 75th% of S is 0.075 without sampling compared to 0.022 with sampling.

[00206] It was further evaluated whether an Estimation method of alOB could be established using a single sample per bolus. For this, an empirical model was developed using the concentration at a single sample (3 hr) to predict the estimated alOB based on two samples (1 hr, 3 hr). Figure 17 shows the results as a scatter plot of 1000 simulated profiles, where each point relates the log of the insulin concentration at 3 hours (x-axis) to the log of the alOB/Dose estimated after taking the two samples (y-axis). In Figure 17, the black curve is a polynomial (quadratic) regression model fit to all the data (1000 profiles) predicting log alOB/Dose given log insulin concentration. The thick grey curve is the same regression model fit to a subset of the data representing a limited amount of training data (75 profiles, corresponding to 25 days x 3 bolus per day). The median of alOB/Dose in the training data corresponds to a patient specific model (estimated during training) that requires no additional sampling. These empirical models (based on one sample per profile and based on no samples per profile) are therefore completely determined based on the limited data collected during the training period.

[00207] This one sample method was then compared to the two-sample and the non- sampling method (Figure 18). As for Figures 14 and 16, the difference between "IOB True" (the simulated value of alOB) and "IOB Estimate" (the estimated value of alOB) is compared for the three methods of estimation. The parameters of the IOB Estimate of the one sampling method are calculated during the training period (75 profiles sampled twice per bolus). From this, it is clear that also the one sampling method is more accurate in estimating the alOB versus the no-sampling method.

Claims

1. A method for determining the Insulin On Board (IOB) in a diabetes mellitus patient comprising the steps of: a) in a first blood sample obtained from the patient near the time of peak insulin concentration resulting from the last insulin administration, determining the amount of insulin in said sample, b) in a second blood sample obtained from the patient at the time of calculating the IOB after said last insulin administration, determining the amount of insulin in said sample, c) calculating the IOB at a given moment based on the previously administered amount of insulin, the time of previous insulin administration(s) and the two determined amounts of insulin from steps a) and b).

2. The method according to claim 1, wherein the IOB is calculated based on a two- parameter continuous distribution, fitted to the two insulin concentration measurements of step c).

3. The method according to claim 2, wherein the IOB at time τ equals pAUC(T) CL, where pAUC(T) is the partial area under the curve from time τ to infinity calculated from the integral of said distribution, where AUC is the total area under the curve of said distribution, and CL is the plasma clearance of insulin.

4. The method according to claim 3, wherein the adjusted IOB (alOB) at time τ is calculated as follows: alOB = pAUC(T) CL/E[F], wherein E[F] is the expected bioavailability of insulin via the subcutaneous route of administration.

5. The method according to claim 4, wherein E[F]/CL is identified as the expected value of AUC/Dose and calculated based on the AUC/Dose values (the single most recent AUC/Dose value, or a set of prior AUC/Dose values) obtained by the insulin measurements made at the two time points in steps a) and b).

6. The method according to anyone of the previous claims, wherein additional insulin measurements may be performed on additional blood samples obtained at different time points after an insulin bolus injection and where the two-parameter continuous distribution may be replaced by a distribution with two or more free parameters.

7. The method according to claim 1, wherein said first measurement near the peak concentration time is replaced by a patient specific empirical model relating the insulin concentration measured at a single time-point to the adjusted IOB at the same time-point.

8. The method according to claim 2, wherein said empirical model is determined based on aggregated data obtained from a plurality of patient specific insulin measurements obtained in a manner described in claims 2 to 6.

9. The method according to any one of the previous claims, wherein said patient is an insulin-pump patient, or wherein said patient administers insulin through injections.

10. The method according to any one of the previous claims, for use in detecting changes in pharmacokinetics of insulin appearance in plasma in insulin pump using patients using a catheter for insulin administration that stays in situ for at least one day, at least two days, at least three days, at least four days, at least five days, or for multiple days.

11. The method according to any one of the previous claims, for use in regulating insulin release by an insulin pump.

12. The method according to claim 11, wherein said insulin release by said insulin pump activity is reduced or halted when the IOB level becomes too high, or increased when the insulin concentration is predicted to become too low.

13. A method for determining the amount of insulin needed in a diabetes mellitus patient comprising the steps of: a) determining the amount of Insulin On Board (IOB) using the method according to any one of claims 1 to 9, b) calculating the amount of insulin needed in said patient, based on the pre-meal glucose concentration in the patient and the quantity of carbohydrates in the next meal, the glucose concentration correction factor, the insulin to carbohydrate ratio and subtracting the Insulin On Board determined in step a).

14. A method for better serving the actual basal insulin need of a subject, comprising the step of measuring the actual IOB in said patient, using the method according to any one of claims 1 to 9.

15. A method for improving the control of insulin delivery from an insulin pump to a user of said pump, comprising the step of measuring the actual IOB in said patient, using the method according to any one of claims 1 to 9, and adjusting the amount of inslulin delivered according to the IOB.

16. A method of alerting the patient of increased future risk for hypoglycemia by balancing the need for insulin in the near future with the real insulin action in the near future by: a) measuring the actual IOB in said patient, using the method according to any one of claims 1 to 9, or b) measuring the long-acting insulin part and adding it to a) above, c) determining the blood glucose concentration in said patient, d) evaluating whether the insulin action will be balanced with the glucose appearance in blood, and e) alerting the patient of possible hypoglycemia when the IOB exceeds the available glucose in said patient; or possible hyperglycemia when the IOB is insufficient for the glucose appearing in the plasma in said patient.

17. A method of calculating the cumulative IOB of several previous injections, using the method according to anyone of claims 1 to 9, and comparing it with the need of insulin in the near future, thus allowing to detect future risk hypoglycemia and hyperglycemia.

18. A method of measuring insulin in blood for fast feedback on how much insulin reached the blood after aerosol administration (nasal or inhaled insulin), comprising the step of: in a sample taken from the patient measuring the amount of insulin, and comparing said amount with the administered amount of insulin.

19. A method for determining the adjusted Insulin On Board (alOB) in a diabetes mellitus patient comprising the steps of:

1) establishing a patient specific empirical model as follows: a) taking a first patient blood sample near the peak concentration time of the last insulin administration and measuring the amount of insulin in said sample, b) taking a second patient sample at the time of calculating the Insulin On Board and measuring the amount of insulin in said sample, c) calculating the amount of IOB at a given moment based on the previous administered amount of insulin, the time of previous insulin administrations and the two measured insulin amounts in steps a) and b), wherein said steps a) to c) are repeated at least 5 times, preferably at least 10 times, more preferably at least 15, 20, 25, or 30 times; and d) fitting a regression model to the data of IOB (or alOB) versus insulin concentration collected in step c) using either a linear model, or preferably a quadratic model of log IOB versus log concentration, or possibly a higher order (or piecewise) polynomial to represent the relationship between the IOB (the response) and the insulin concentration (the predictor); and

2) calculating the IOB by: e) taking a patient sample at the time of calculating the Insulin On Board, and measuring the amount of insulin in said sample, f) calculating the IOB or alOB, based on the insulin measurement of step e), and the empirical model obtained in step 1).

20. The method according to claim 19, wherein said calculation of IOB or IOB is done as follows:

IOB at time τ = pAUC(T) CL, where pAUC(T) is the partial area under the curve from time T to infinity calculated from the integral of the distribution obtained in the emprirical model of step 1), wherein AUC is the total area under the curve of said distribution, and wherein CL is the plasma clearance of insulin and/or alOB at time τ = pAUC(T) CL/E[F], wherein E[F] is the expected bioavailability of insulin via the subcutaneous route of administration, and E[F]/CL = E[AUC/Dose], where Dose is the total amount of injected insulin.