IE920897A1 - Normalizing analyzer systems to a standard analyzer - Google Patents

Abstract

NORMALIZING ANALYZER SYSTEMS TO A STANDARD ANALYZER There is described a method of calibrating an analyzer in the field, using parameters designed for a "standard" analyzer that may be different from the field analyzer. In this way, it is not necessary to use only certain slide elements of a given chemistry on one analyzer, and certain other slide elements of that same chemistry on another analyzer. Instead, the slide elements are made interchangeable by calibration math which allow the field analyzer to convert its raw response RF into a response Rstandard of the selected "standard" analyzer, using the equation (1) Rstandard = B0 + B1 ? RF + B2 ? (RF)k . The coefficients of this equation are established by comparing, at at least 1 different level of response, RF against Rstandard, optionally using an intermediate spline function gF(RF) in place of RF.

Description

NORMALIZING ANALYZER SYSTEMS TO A STANDARD ANALYZER This invention relates to a method of calibrating a clinical analyzer so that it can be used with slide elements calibrated for use on a drastically different type of analyzer than the one actually used.

Certain dried, slide test elements, such as the elements available under the trademark Ektachem from Eastman Kodak Company, are capable of use on a variety of different analyzers. These latter include the Ektachem 700 analyzer and the Ektachem DT-60 analyzer, both available from Eastman Kodak Company. However, when a test element is used on one type of analyzer, for example, one that uses one kind of reflectometer, it tends to produce a different correlation between detected response and analyte concentration, than it produces when read on a second type of analyzer, for example, one that uses a second type of reflectometer. As a result, the correlation between detected response and concentration is different, for any given test element, when read on for example, the Ektachem DT-60 analyzer, compared to its reading on the Ektachem 700 analyzer.

Such different correlations require that a different calibration curve establishing the correlation, be used for each different kind of analyzer. Different calibration mathematics has to be established for each different type of analyzer. This requires the proper type analyzer be used at the factory to determine the calibration, rather than any possible analyzer. Such different calibration curves and the related calibration mathematics have to be carried with the test element in question such as by bar coding and/or magnetic disk. Importantly, the correction for such variation is different depending on what kind of analyzer is to be used for detection. As -2a result, the test elements have to be somehow segregated, using such a system, based on which type of analyzer the elements have been tested on and are destined for. In sum, the test elements have to be paired with a particular type of analyzer, especially if all or some of the calibration mathematics is being passed along with the test elements.

Such pairings or segregation has not been a problem when there are only two basic types of analyzers to choose from, for example, the ’'Ektachem 700 type and the Ektachem DT-60 type. The reason is that the DT-60 type elements have already been packaged differently (individually) from the packaging of the 700 types (by cartridge), and the calibration mathematics is passed differently, so that segregation occurs naturally. Where a problem arises is when yet a third type of analyzer is introduced that also packages test elements in a manner that is similar to either of the first two noted above. As a result, the test elements can end up being used on an analyzer for which the test elements bear the wrong calibration mathematics. Keeping track of which test element is to go to what type of analyzer, and thus is to have what calibration mathematics carried with it, becomes a horrendous logistics problem.

Therefore, there has been a problem prior to this invention of devising correction factors for the calibration of test elements which will not be different due to which type of analyzer the element ends up being tested on.

It is known to have a correction method which relates an aged test element to that element when fresh, or a lot-varied element of a given assay to a standard element of that assay, when always used on the same analyzer. Such a method is shown for optical -3density in US-A-4 884 213. However, that method makes no correction for analyzer-to-analyzer variations, and furthermore incorrectly asserts that the variations which are discussed can always be corrected by a linear relationship.

The invention is based on the discovery that a relationship exists between how a given test element will perform on one type of analyzer, compared to how it will perform on a standard analyzer, such as on the Kodak Ektachem 700™ , hereinafter ’Έ700, analyzer, so that this relationship can be programmed into that one type of analyzer to make its response LOOK like it is a E700 analyzer response. Once that occurs, correction factors can be sent with all test elements of an assay as though they were all to be tested on the one single type of analyzer, namely (in this case) the E700 analyzer.

More specifically, in accordance with the present invention, there is provided a method of quantitatively determining an analyte in a liquid sample applied to a slide element by measuring in a read station of a field analyzer a response developed in the slide element and correlating said response to an analyte concentration using a calibration curve produced from calibrators having known concentrations, such slide elements being selected from a set of slide elements that all produce the same response in said field analyzer except for deviations due to lot-to-lot variations or aging, said same response being a function of the kind of analyzer used; characterized in that said response on said analyzer is corrected substantially to the response that would have been detected on a standard analyzer of a type different from said field analyzer and having IE 920897 -4one of said calibration curves for correlating the response to an analyte concentration, by the steps of: a) applying at least one level of said calibrators on at least one of said set of slide elements and reading the response in said standard analyzer, b) repeating step (a) but by reading their response in said field analyzer; c) correlating the relationship of the different responses in said two analyzers by using the equation (I): Rstandard = Bo + Bi.gF(R2) + B2.gF(RF)k (I) where Rstandard is the response detected in said standard analyzer, Rf is the response detected in said field analyzer; gF(RF) is a spline function used to transform the responses Rf; Bo, Bi and B2 are standardizing coefficients; and k is an exponent corresponding to the nonlinearity of the equation, d) applying an unknown patient liquid sample to a slide element selected from said set of slide elements and reading a response in said read station of said field analyzer; e) ascertaining from said correlation of step (c) what said read response in said field analyzer in step (d) would produce as the corresponding response in said standard analyzer, and f) correlating said ascertained corresponding response to the predicted analyte concentration using said calibration curve of said standard analyzer.

As used herein, field analyzer refers to the analyzer used by the customer, which may be the same or a different type as was selected for the standard analyzer. -5Thus, it is an advantageous feature of the invention that the calibration of the field analyzer being used is expressed in terms of a single standard analyzer, so that differences introduced by variations in the analyzer type are automatically corrected and the same slide elements can be used in both analyzers even though the calibration mathematics is derived for only a single type of analyzer.

It is a related advantageous feature of the 10 invention that the slide test element bearing a set of calibration parameters is useful in a variety of analyzers having different constructions, rather than just one, and such test elements with their calibration mathematics need not be supplied to only one type of analyzer.

For a better understanding of the present invention, reference will now be made, by way of example only, to the accompanying drawings in which:Figure 1 is a plot of the optical density 20 detected from certain reference elements of known density, using on one hand the standardization analyzer (here, the ’Έ700 analyzer), and, on the other hand, various other analyzers which can be used as field analyzers; Figure 2 is a schematic plot which is illustrative of the steps of the invention; Figure 3 is a plot of Quadrants I and II of Figure 2 for a given example of chemistry and a given field analyzer where the gr function is unity and contributes no correction; Figure 4 is a plot of Quadrant III of Figure 2, for the chemistry of Figure 3; Figure 5 is a plot similar to Figure 3, but for a different chemistry; -6Figure 6 is a plot corresponding to the chemistry of Figure 5 using, however, a gF correction provided in the manner indicated in Quadrant I; Figure 7 is a plot similar to that of Figure 5 5, after the correction of Figure 6 has been applied; Figure 8 is a plot of Quadrant III of Figure 2, for the chemistry and results obtained from Figure 7; Figure 9 is a plot similar to that of Figure 10 3, but for yet another field analyzer; Figure 10 is a plot similar to that of Figure 4, for the analyzer of Figure 9; Figure 11 is a plot similar to that of Figure 3, illustrating the invention when using a rate assay; Figure 12 is a plot of Quadrant III of Figure 2, but for the rate assay of Figure 11; Figure 13 is a plot similar to that of Figure 2, but illustrating an alternative embodiment, namely factory calibration; and Figures 14 and 15 are respective plots of distributions of predictions of concentration using either the conventional K-Model method or the calibration method of the invention.

The invention is hereinafter described in connection with certain preferred embodiments, featuring preferred slide test elements tested in preferred clinical analyzers, wherein calibration mathematics is transmitted by bar codes and/or magnetic disk. In addition, the invention is useful regardless of the form of the test elements, regardless of which types of analyzers are used, and regardless of the mode of transmission of the calibration mathematics, so long as one kind of analyzer is selected as a standard, producing a response against which the response of any -7other analyzer is compared and corrected as part of the calibration.

The preferred test elements are the slide test elements available from Eastman Kodak Company under the trademark Ektachem slides. Such slide elements provide a raw response in the analyzer which is either a reflectance R, an optical density OD where OD = log(l/R), the rate of change of either R, OD, or a transformed OD, or an electrical potential created by a differential measurement of ion concentration in two ion-selective electrodes. Any one of these responses can be calibrated by this invention. (A transformed OD is an optical density value obtained through a spline function from raw optical density values to correct for interferences or to convert to other densities, for example, by using a transform such as the Clapper-Williams transform.) The preferred analyzers are any of the analyzers available from Eastman Kodak Company under the trademark Ektachem analyzer as well as the analyzer available under the tradename Vettest 8008 from Vettest Corporation.

The reflectometer or detecting station of any analyzer can be used to provide the response Rstandard/ that is any analyzer can be the standard analyzer.

For convenience, due to their common occurrence in industry, the E700 analyzers available from Eastman Kodak Company have been selected as the standard analyzer for this invention.

Any other analyzer used as the field analyzer, is then programmed pursuant to this invention to convert the detected response into the corresponding value that would have been produced on the E700.

Thus, calibration mathematics which is passed along with the slide elements of any given chemistry, are -8presented as though those slide elements were to be read on an actual E700 as the field analyzer, even though they may not be. As used herein, calibration mathematics includes not only the calibration parameters but also any relevant spline information and concentrations associated with calibrators.

The problem created by different types of analyzers being used to read the same slide elements using only E700 calibration mathematics, is illustrated in Figure 1. A series of magenta slides of permanent, differing optical density were read both on the E700 reflectometer or read station, and then on one of the following reflectometers (or read stations): the reflectometer for the DT-60™ analyzer of Eastman Kodak Company, or the reflectometer of the E400™ analyzer of Eastman Kodak Company. The X=Y line indicates the substantial identity which is achieved if the other analyzer is in fact the E700 analyzer. Because the results obtained on the E400 analyzer fall substantially on the X=Y line, it can be assumed, and indeed has been shown, that the E400 is in fact just another E700 as far as its calibration mathematics is concerned.

However, the other reflectometer, that of the DT-60, produces results which deviate significantly from the X=Y line. It can be further shown that the Vettest 8008, if plotted on Figure 1, would produce substantial deviation from the X=Y line. A still further example would be a reflectometer identical to that of the E700 analyzer, except for items such as the incubator for that reflectometer being maintained at a temperature substantially different from that of the E700 analyzer, or the dispensing station being different, and so forth. -9These deviations are not surprising due to the differences noted. The Vettest 8008 reflectometer uses LEDs for its visible light and mercury vapor bulbs for UV, which illuminate the slide elements at wavelengths which differ from those used for any given chemistry on the E700. The DT-60 reflectometer differs even more in that it uses a contact read station, instead of the non-contact read station of the E700, created by the fiber optics construction shown, for example, in US-A-4 302 420.

The function then of the invention is to correct the raw reading of, say, the DT-60 analyzer or the Vettest 8008 analyzer to the corresponding reading which would have been obtained on the E700 analyzer. Although this is not as important for the DT-60 analyzer because its slide elements are packaged differently and its calibration mathematics is passed on different media, as noted above, such packaging or media could change at a future date and be so similar to the packaging and media of the slide elements for the E700 as to warrant the use of this invention.

The procedure of the invention is schematically illustrated in Figure 2. The plot of this Figure correlates the response obtained on the field analyzer, Rf, to the response Rstandard obtained or expected on the standard analyzer, here, the E700. Because the Rstandard values are conventionally correlated to the concentration, as defined by Quadrant III of this plot, the correlation of Rf to Rstandard will act to convert the field analyzer responses to the concentration predictable as though the analyzer were a standard analyzer.

Quadrant II of the plot takes a function gF(RF)/ and correlates that function to Rstandard by -10the equation (I) noted above, where it is assumed that k is a value that is either 0 or 2. (If the graph is curved as shown, k is 2.) If, in fact, the raw response Rf is a quadratic function, or a linear function (k=0 and B2=0), of Rstandard/ then gF(RF) is in fact simply Rf and the curve gF of Quadrant I becomes the identity line shown in phantom in quadrant I. The only time gF/ or Quadrant I, is not unity and must be used, is if k truly is not equal to zero or 2 in equation (I) of Quadrant II. For such occasions, such as the blood urea nitrogen (BUN) example set forth below, a spline function is generated for the reading Rf to create a gF(RF) value that most closely fits the data, as is conventional for spline functions and is described in, for example, Industrial Applications of Cubic Spline Functions, by N. J. Barosi, October 26, 1973, pp. 3-6, (A Presentation to the 17th Annual Technical Conference of The American Society for Quality Control and The American Statistical Association), and Splines and Statistics, by Edward J. Wegman and Ian W. Wright, Journal of the American Statistical Association, June 1983, Volume 78, Number 382, Theory and Methods Section, pp. 351-352. Most preferably, the spline is adjusted so as to render the Rstandard a quadratic function of gF(RF)· it is then this gF(RF) value that can be correlated, quadrant II, to Rstandard by the quadratic equation (I) where k = 2. The coefficients Bo, Bi and B2 are the resulting solution of the equation that best fits the data of the plot of gF(RF) versus Rstandard· Ideally, the preceding is handled using fresh slide elements from a standard lot of a given chemistry, the same lot of which is to be used on the field analyzer, also fresh. However, as has been shown in, for example, Us-A-4 884 213, slight variances can — 11 — be introduced in the raw response Rf and thus, gF(RF), when either a new lot of slide elements for the same chemistry is prepared, or when the slide elements actually tested on the field analyzer are aged.

Unfortunately, in most cases the slide elements actually used in the field analyzer will either be aged or from a different lot, or both, and further correction is required. Such correction is obtained, by plotting Rvariance against Rstandard'» where Rvariance is the response of the aged or lot-varied slide element, and Rstandard’ is the response of a fresh slide element from the lot selected to be the standard lot, and determining the best fit in accordance with equation (II): Rstandard = Ao + Ai.Rvariance + A2.(Rvariance)dD It can be shown that, for Rvariance produced by aged slide elements of the same lot as the standard lot, k of equation (II) will be 0 (and A2 = 0). For Rvariance produced by fresh slide elements from a lot different than the standard lot, k of equation (II) is 2. The important factor is, equation (II) is of the same form as equation (I), so that Ao, Ai and A2 can be incorporated into Bo', Bi' and B2', thus correlating for analyzer variation, slide element aging, and lot25 to-lot variances, all at once. For example, the procedure described in the aforesaid US-A-4 884 213 is useful, except that k = 0 and A2 = 0 only if the variance is due to aging alone, and otherwise k is not equal to 0, but most preferably is equal to 2. Once the Rvariance corrections have been included in the coefficients of equation (I) of Quadrant II, to provide coefficients Bo', Bi' and B2', respectively, then the curve may be expected to have shifted to a position such as is shown in phantom. -12Once the curve shape of Quadrant II is ascertained, then it is possible to pass on the calibration mathematics obtained from the standard analyzer for any given chemistry to the field analyzer, using the field analyzer gp and by computing Bo, Bi and B2 at the customer site. Thus, each new generation of slide elements need not be tested on all the different analyzer types, nor would separate calibration mathematics for each type of analyzer, and slide elements with such mathematics, be carefully sheparded only to that type of analyzer. Instead, the new generation and a single set of calibration mathematics becomes applicable to all analyzer types.

The steps of the process of the invention involving reading a response of calibrator liquid of known concentration on a slide element, first on the standard analyzer and then on the field analyzer, is preferably done using three different levels of calibrator liquid. However, one or two levels of calibrators are sufficient if the only difference to be adjusted for is a difference within a single type of field analyzer so as to be just a slope or intercept change or both.

Examples: The following examples further illustrate the scope of the invention.

In all of these examples, the standard analyzer was an E700 analyzer available from Eastman Kodak Company.

Example 1: Glucose Slide Elements in a Vettest 8008 Analyser In this example, the field analyzer was a Vettest 8008 analyzer, and the slide elements were -13'’Ektachem glucose slide elements from Eastman Kodak Company. A set of 20 fluids with glucose concentrations ranging from 30 to 600 mg/dL were collected. These fluids were spotted onto slides from one lot of glucose test slides and the amount of color formed was read by the two analyzers. The standard responses were plotted versus the responses of the field analyzer and a quadratic polynomial was fit to the data resulting in Bo, Βχ, B2 = -0.01263, 0.46095, 0.62659 respectively. A graph of the data and the quadratic function can be found in Figure 3. Standardized responses for the field analyzer were found by evaluating: Standard Response = Bo + Βχ .Rf + B2-Rf^ The resulting responses are tabulated in Table 1. -14TABLE 1: ^standard (E700 Reflectance) Vettest Rp Reflectance Corrected Rp Reflectance Difference from Rstandard 0.1489 0.2603 0.1505 +0.0016 0.0982 0.1949 0.1014 +0.0032 0.2103 0.3299 0.2087 -0.0016 0.2546 0.3805 0.2549 +0.0003 0.3356 0.4641 0.3384 +0.0028 0.3025 0.4281 0.3014 -0.0011 0.5392 0.6357 0.5376 -0.0016 0.5526 0.6484 0.5538 +0.0012 0.0537 0.1256· 0.0553 +0.0016 0.0322 0.0825 0.0297 -0.0025 0.2195 0.3420 0.2195 NONE 0.0555 0.1270 0.0562 +0.0007 0.1481 0.2597 0.1500 +0.0019 0.1976 0.3180 0.1983 +0.0017 0.1742 0.2860 0.1712 -0.0030 0.1073 0.2067 0.1098 +0.0025 0.1270 0.2321 0.1287 +0.0017 0.1690 0.2768 0.1637 -0.0053 0.1380 0.2416 0.1359 -0.0021 0.0394 0.0976 0.0384 -0.0010 Predicted concentrations were found using the known E700 glucose response to concentration relationship for the lot of glucose slides. A plot of the standard glucose response to concentration relationship for the lot of glucose slides is located in Figure 4. The predicted concentrations are printed in Table 2. For comparison, the concentration predicted from the Rf value, uncorrected, is also shown. -15TABLE 2: E700 Concentration Concentration from Rp Uncorrected Response Concentration from Rf Corrected % Error in Corrected Response Compared to E700 180 106 178 1.1 252 142 246 2.4 132 80 133 0.8 108 66 108 0 78 46 77 1.2 89 55 89 0 37 27 38 2.6 36 26 38 5.6 85 207 378 1.8 564 287 600 6.4 126 76 126 0 377 206 374 0.8 181 106 179 1.1 140 84 139 0.7 157 95 160 1.9 235 134 231 1.7 205 119 203 1.0 161 99 166 3.1 192 115 195 1.6 481 253 490 1.9 The % error that remains in Table 2 is not clinically significant to most clinical chemists.

Thus, the calibration of this invention successfully converted the response (here reflectance) of the field analyzer (here, a Vettest 8008”) to the values that would have occurred if such slide elements (here, for glucose) were read on the standard analyzer. -16Because the equation of quadrant II of Figure 2 was essentially quadratic in nature, there was no need to use quadrant I in this example. Equivalently, gF(RF)=RF in this example.

Example 2: BUN Slide Elements In a Vettest 8008 Analyzer The same analyzers were used in this test as in Example 1. A set of 20 fluids with BUN concentrations ranging from 10 to 90 mg/dL were collected. These fluids were spotted onto slides from one lot of BUN test slides and the amount of color formed was read by the two analyzers. The standard responses were plotted versus the responses of the field analyzer and a quadratic polynomial was fitted to the data resulting in Bo, Bi, B2= 0.02845, 2.2517, -3.0861 respectively. A graph of the data and the quadratic function can be found in Figure 5. Standardized responses for the field analyzer were found by evaluating: Standard Response = Bo + Bi.Rf + B2-Rf^ for each Rf (that is, response on the field analyzer). The resulting responses are tabulated in Table 3. -17TABLE 3: ^standard E700 Reflectance Vettest Reflectance Corrected Rf Reflectance 0.2013 0.0845 0.1967 0.3654 0.2153 0.3701 0.3899 0.2344 0.3867 0.2881 0.1455 0.2908 0.3127 0.1668 0.3182 0.2197 0.0958 0.2158 0.1298 0.0429 0.1193 0.3917 0.2392 0.3905 0.2525 0.1207 0.2553 0.3082 0.1622 0.3125 0.3749 0.2199 0.3744 0.4162 0.2668 0.4095 0.1722 0.0652 0.1620 0.1137 0.0324 0.0982 0.0319 0.0065 0.0430 0.0342 0.0068 0.0435 0.0312 0.0066 0.0431 0.1373 0.0462 0.1258 0.2871 0.1479 0.2940 0.2864 0.1488 0.2952 From this, predicted concentrations were found using the known E700 BUN response to concentration relationship for the lot of BUN slides. The predicted concentrations are printed in Table 4, and for comparison, a predicted concentration from Rf uncorrected, is also shown. -18TABLE 4: E700 Concentration Concentration from Rf* Uncorrected Concentration from Rp Corrected 30 58 31 14 28 13 12 26 12 20 40 20 18 36 17 28 54 28 44 80 47 12 25 12 24 46 23 18 37 18 13 28 13 11 22 11 35 66 37 48 89 53 89 135 80 87 134 79 90 135 80 42 77 45 20 40 19 20 40 19 Although the concentration from Rf, corrected, is close to the true result shown in the left hand column, there is substantial deviation for concentrations above 45. This means that the equation (I) of quadrant II that best fits the raw reflectance data is not truly quadratic.

Hence, a gF function was found to remove the non-quadratic response differences that existed between -19the E700 and the field analyzer. The gF function was represented as the following cubic spline: X-Value Y-Value 2nd Derivative 0.0 0.0 0.0 0.02 0.067684 -31.76737 0.12 0.275344 -2.2153 1.0 1.0 0.0 to be used to provide a curve such as forms quadrant I of Figure 2. More specifically, the relation of Rf of Table 3, and this gF(Rp) is shown in Figure 6.

The standard responses, Rstandard/ were plotted versus the responses of the field analyzer.

The field analyzer responses were evaluated through gF and then a quadratic polynomial, which changed transformed field analyzer responses to standard responses, was found. This resulted in Bo, Bj, B2 = 0.013317, 0.881242, -0.029543 respectively. A graph of the data and the correction curve can be found in Figure 7. Standardized responses for the field analyzer were found by evaluating: Standard Response = Bo + Bi.gF(RF) + B2.gF(RF)^ for each Rf (that is, response on the field analyzer).

The resulting responses are tabulated in Table 5. -20TABLE 5: Rstandard E700 Reflectance Vettest Rp Reflectance Of(Rf) Corrected Reflectance from gp (Rf) 0.2013 0.0845 0.2195 0.2053 0.3654 0.2153 0.4060 0.3662 0.3899 0.2344 0.4300 0.3868 0.2881 0.1455 0.3122 0.2856 0.3127 0.1668 0.3419 0.3112 0.2197 0.0958 0.2383 0.2217 0.1298 0.0429 0.1325 0.1296 0.3917 0.2392 0.4359 0.3919 0.2525 0.1207 0.2764 0.2546 0.3082 0.1622 0.3356 0.3057 0.3749 0.2199 0.4119 0.3713 0.4162 0.2668 0.4960 0.4201 0.1722 0.0652 0.1831 0.1737 0.1137 0.0324 0.1048 0.1053 0.0319 0.0065 0.0227 0.0333 0.0342 0.0068 0.0235 0.0340 0.0312 0.0066 0.0228 0.0334 0.1373 0.0462 0.1407 0.1367 0.2871 0.1479 0.3156 0.2885 0.2864 0.1488 0.3169 0.2896 Predicted concentrations were found using the known E700'’ BUN response to concentration relationship for the lot of BUN slides. A plot of the standard BUN response to concentration relationship for the lot of BUN slides is located in Figure 8. The predicted concentrations are printed in Table 6. For comparison, a predicted concentration from Rf, uncorrected, is also shown. -21TABLE 6: E700 Concentration Concentration from Rp Uncorrected Concentration from gF(Rp) % Error Compared to Cstandard 30 58 30 0 14 28 14 0 12 26 12 0 20 40 20 0 18 36 18 0 28 54 27 3.7 44 80 44 0 12 25 12 0 24 46 23 4.2 18 37 18 0 13 28 13 0 11 22 10 9.1 35 66 35 0 48 89 51 6.3 89 135 88 1.1 87 134 87 0 90 135 88 2.2 42 77 42 0 20 40 20 0 20 40 20 0 It is clear from Table 6 that the gF function together with the quadratic correction effectively standardizes the responses of the field analyzer to the E700 analyzer. -22Example 3: Glucose Slide Elements in a DT-60 Analyzer The procedure of Example 1 was repeated, except that the field analyzer was a ’'DT-60'' analyzer from Eastman Kodak Company. A set of 19 fluids with glucose concentrations ranging from 48 to 710 mg/dL were collected. These fluids were spotted onto slides from one lot of glucose test slides and the amount of color formed was read by the two analyzers. The standard responses were plotted versus the responses of the field analyzer and a quadratic polynomial was fit to the data resulting in (Bo, Βχ, B2 = -0.005724, 0.602386, 0.412017). A graph of the data and the quadratic function can be found in Figure 9.

Standardized responses for the field analyzer were found by evaluating: Standard Response = Ro + Bi.Rf + B2-Rf^ for each Rf (that is, response on the field analyzer). The resulting responses are tabulated in Table 7. -23TABLE 7: E700 Reflectance DT60 Reflectance Corrected Rp (DT60) Reflectance 0.4482 0.5487 0.4488 0.0828 0.1348 0.0830 0.0298 0.0554 0.0289 0.3233 0.4238 0.3236 0.3108 0.4100 0.3105 0.3368 0.4359 0.3351 0.3190 0.4207 0.3206 0.2231 0.3086 0.2194 0.2984 0.3978 0.2991 0.1593 0.2370 0.1602 0.2141 0.3019 0.2137 0.0655 0.1132 0.0677 0.0668 0.1134 0.0679 0.0877 0.1420 0.0881 0.0372 0.0705 0.0388 0.0375 0.0700 0.0385 0.0350 0.0658 0.0357 0.0225 0.0411 0.0198 0.0255 0.0473 0.0237 Predicted concentrations were found using the known E700 glucose response to concentration relationship for the lot of glucose slides. A plot of the standard glucose response to concentration relationship for this lot of glucose slides is located in Figure 10. The predicted concentrations are printed in Table 8. For comparison, the concentration predicted from the Rf value, uncorrected, is also shown. -24TABLE 8: E700 Concentration Concentration from Rf Uncorrected Concentration from Rp (Corrected) 48 28 48 286 197 286 580 375 594 81 54 81 85 57 85 76 51 77 82 55 82 123 86 125 89 60 89 170 116 169 129 88 129 335 227 328 331 227 327 275 188 274 493 319 478 490 321 481 516 334 508 711 459 774 652 417 686 Thus, the calibration of this invention successfully converted the response (here reflectance) of the field analyzer (here, an Ektachem DT60) to the values that would have occurred if such slide elements (here, for glucose) were read on the standard analyzer. Because the equation of quadrant II of Figure 2 was essentially quadratic in nature, there was no need to use quadrant I in this example. Equivalently, gF(RF> = RF in this example. -25Example 4: Rate Assay Slide Elements in a Vettest 8008 Analyzer The procedure of Example 1 was repeated, except that the field analyzer was a Vettest 8008 and the slide elements were Ektachem LDH slides from Eastman Kodak Company. A set of 20 fluids with LDH activities ranging from 182 to 1420 U/L were collected. These fluids were spotted onto slides from one lot of LDH test slides and the rate of change of optical density (OD) was read by the two analyzers. The standard responses were plotted versus the responses of the field analyzer and a quadratic polynomial was fit to the data resulting in (Bo, Bi, B2 = -0.00811, 1.20522, -1.51568). A graph of the data and the quadratic function can be found in Figure 11.

Standardized responses for the field analyzer were found by evaluating: Standard Response = Bo + Bi.Rf + B2.Rf2 for each Rf (that is, response on the field analyzer).

The resulting responses are tabulated in Table 9.

!E 920897 -26TABLE 9: E700 Rate Vettest Rate Corrected Vettest Rate -0.0330 -0.0202 -0.0331 -0.0525 -0.0355 -0.0527 -0.0217 -0.0121 -0.0229 -0.0434 -0.0275 -0.0424 -0.0326 -0.0189 -0.0315 -0.0677 -0.0470 -0.0682 -0.0733 -0.0499 -0.0720 -0.0389 -0.0244 -0.0384 -0.0593 -0.0404 -0.0592 -0.0321 -0.0153 -0.0269 -0.0376 -0.0241 -0.0381 -0.0668 -0.0466 -0.0675 -0.0480 -0.0322 -0.0485 -0.0294 -0.0174 -0.0295 -0.0294 -0.0176 -0.0298 -0.0802 -0.0562 -0.0807 -0.0570 -0.0396 -0.0582 -0.0882 -0.0613 -0.0876 -0.0237 -0.0138 -0.0250 -0.0216 -0.0130 -0.0241 Predicted activities were found using the known E700 LDH responses to activity relationship for the lot of LDH slides. A plot of the standard LDH response to activity relationship for this lot of LDH slides is located in Figure 12. The predicted activities are printed in Table 10. For comparison, the activity predicted from the Rf value, uncorrected, is also shown. -27TABLE 10: E700 Activity Uncorrected Vettest Activity Corrected Vettest Activity 418 154 419 791 466 796 184 -23 210 620 307 602 408 127 387 1065 689 1074 1163 743 1141 533 241 523 916 562 915 398 48 294 509 236 517 1050 681 1063 708 402 717 345 93 346 344 99 353 1283 860 1291 875 547 895 1419 951 1409 227 15 254 183 -1 235 Thus, the calibration of this invention successfully converted the response (here rate) of the field analyzer (here, a Vettest 8008) to the values that would have occurred if such slide elements (here, for LDH) were read on the standard analyzer. Because the equation of quadrant II of Figure 2 was essentially quadratic in nature, there was no need to use quadrant I in this example. Equivalently, gF(RF)=RF in this example. -28FACTORY CALIBRATION It is not necessary that the second and third steps of this invention (of reading the field analyzer response and establishing the coefficients of equation no. (I)) be done at the field site, called wet calibration, although this is one use of the invention. In addition, it can all be done at the manufacturing site of the field analyzer. In many cases we have found that factory calibration results are comparable to results obtained using conventional wet calibration. (During such conventional wet calibration, the analyzer and slide elements are characterized at the customer site by reading the slide response of a number of calibrator fluids on the site's analyzer. This information is then used to define the relationship between analyte concentration and slide response for all patient samples on the site's analyzer.) Figure 13 is a graphical representation of the use of the invention in factory calibration. The left hand quadrant depicts the analyzer characterization and the right hand quadrant depicts the test element characterization. In factory calibration, equation (I) is a correction of the field analyzer response of whatever type (Dr, rate or mv) to the standard response. Conveniently, the quadratic form of equation (I) above is used, for example, Responsestandard = Bq + Bi.Rf + B2-Rf2 where Rf is the response of the chemistry on the field analyzer and Bo, Βχ, and B2 are the coefficients of the quadratic correction found in the factory or manufacturing site for the field analyzer and chemistry. The same coefficients could be used for multiple chemistries. -29The curve FLot is found by calibrating the test element on the analyzer chosen as the standard analyzer for factory calibration using the test element's wet calibration model. The mathematical relationship of this curve is then passed on to the customer in suitable coded form, for example, a bar code associated with the test element.

The following examples demonstrate the effectiveness of this factory calibration technique.

Example 5: Factory Calibration of Glucose Colorimetric Slide Element This test used the following: Nine E700™ analyzers manufactured by Eastman Kodak Company, which were intentionally varied, for example, by altering the volume to be metered or by changing the position of the spotted liquid on the test element from the optimum, to cause analyzer response differences; Ektachem™ elements from Eastman Kodak Company, from five different glucose coatings; three calibrator fluids; three control fluids; and two patient pools.

Eight repetitions of each fluid were run on every analyzer/coating combination for a total of 360 (8X5X9) repetitions of each fluid.

CONTROL A calibration curve was found for each analyzer/coating combination using a conventional wet calibration protocol of 2 repetitions of each calibrator fluid. All repetitions of fluids, run on a given analyzer/coating combination, were predicted through that analyzer/coating combination's K-Model wet calibration curve. (Such K-Model calibration is well known and is the method currently used in the E400™ -30and E700™ analyzers. Figure 14 shows the distribution of these predictions for the control fluid, 24102, which is comprised of a human serum base pool with normal glucose concentration.

Analyzer 6 of the nine analyzers was used as the standard analyzer for the factory calibration model since this analyzer was a typical E700 analyzer. Therefore, the K-Model wet calibration curve found for each coating on analyzer 6 served as the web characterization curve FLot» Figure 13, for each coating. (All remaining analyzers are treated as field analyzers.) Table 11 contains calibrator response information on coating 1 used in computing the analyzer characterizations .

TABLE 11: Analyzer Cal 1 Dr Cal 2 Dr Cal 3 Dr. 1 0.32250 1.17297 1.58588 2 0.33128 1.18523 1.60577 3 0.33210 1.19432 1.59229 4 033587 1.19345 1.59562 5 0.33847 1.20931 1.62825 6 0.32895 1.18602 1.60482 7 0.33482 1.21462 1.63463 8 0.32409 1.17470 1.59108 9 0.31762 0.17360 1.58352 Since analyzer 6 was chosen as the standard, 20 a quadratic function was found for each analyzer which converted the given analyzer's Dr to analyzer 6 Dr.

The coefficients of the quadratic correction for each analyzer are listed in Table 12. -31TABLE 12: Analyzer BO Bl B2 1 0.00590 1.000057 0.005153 2 -0.00595 1.012941 -0.006120 3 0.01719 0.923388 0.046278 4 0.00662 0.948520 0.033277 5 0.00075 0.965598 0.012007 6 0.0 1.0 0.0 7 0.00997 0.946787 0.017671 8 0.00187 1.009693 -0.001402 9 0.01693 0.977254 0.016108 To provide a factory calibration curve for coating 4 on field analyzer 8, the following two steps are used: 1. The analyzer 8 characterization is used to convert the Dr of the sample on analyzer 8 to the Dr of the sample on the standard analyzer, that is, analyzer 6, using: Standard Dr = 0.00187 + 1.009693Dr + -0.001402Dr2 2. The K-Model calibration curve for coating 4 on the standard analyzer, that is, analyzer 6, is used to convert standard Dr to concentration.

The factory calibration curves for all analyzer/coating combinations are constructed in a similar manner. If a markedly different generation of slide elements, for example, of the Ektachem glucose elements, produces a drastically different correlation on the standard analyzer, new spline transformations will be then supplied to the field analyzers with that new generation of slide elements, for re-adjustment of the memory of the field analyzer. This is currently IE 92089/ -32done by magnetic floppies for the E700 and the E400 analyzers manufactured by Eastman Kodak Company.

All repetitions of fluids run on a given analyzer/coating combination are predicted through that analyzer/coating combination's factor calibration curve. Figure 15 shows the distribution of predictions for the control fluid 24102 using the factory calibration technique, that is, after converting the Dr readings on analyzer 8 as the Field Response, left quadrant of Figure 13, to predicted Concentration in the right quadrant.

VARIANCES The variance of predicted concentration 15 across all analyzer/coating combinations for the three controls and two patient pools are printed in table 13. Kodatrol 1 Kodatrol 2 are lyophilized human serumbased control fluids available under these trade names from the Eastman Kodak Company, and are included to illustrate the ability of the factory calibration to work with a variety of liquids.

TABLE 13: Fluid Concentration K-Model Variance Factory Calibration Variance Kodatrol I 88.31 1.84 1.53 Kodatrol II 303.33 7.32 7.04 Control 24102 126.48 3.474 2.486 Human Pool 1 60.04 0.697 0.566 Human Pool 2 412.11 26.31 29.89 The variance of predicted concentration using the factory calibration of this invention is similar to -33the variance obtained when one uses the K-Model conventional three-level wet calibration method.

Example 6: Factory Calibration of Amylase (Two-Point Rate) This example used the following: Nine ’Έ700 analyzers from Eastman Kodak Company, which were intentionally varied to cause analyzer differences; five lots of Ektachem amylase slide elements from three different coatings; three calibrator fluids; three control fluids; and two patient pools.

Eight repetitions of each fluid were run on every analyzer/slide lot combination for a total of 360 (8x5x9) repetitions of each fluid.

The same control was used as for Example 5, that is, a calibration curve was found for each analyzer/slide lot combination using a conventional wet calibration protocol of 2 repetitions of each calibrator fluid. All repetitions of fluids, run on a given analyzer/slide lot combination, were predicted through that analyzer/slide lot combination's K-Model calibration curve.

In this example, analyzer 9 was used as the standard analyzer for the factory calibration since this analyzer was a typical E700 analyzer.

Therefore, the K-Model wet calibration curve found for each slide lot on analyzer 9 served as the web characterization curve FLot for each lot. Since analyzer 9 was chosen as the standard, a quadratic function, Equation (I), was found for each analyzer which converted the given analyzer's rate to analyzer 9 rate. -34To provide a factory calibration curve for slide lot 4 on field analyzer 8 the following two steps were used: 1. The analyzer 8 characterization was used 5 to convert the rate of the sample on analyzer 8 to the rate of the sample on the standard analyzer, that is, analyzer 9, using: Standard Rate = Bo + Bi.Rate + B2.Rate2 2. The K-Model calibration curve for slide 10 lot 4 on the standard analyzer, that is, analyzer 9, was used to convert standard rate to concentration.

The factory calibration curves for all analyzer/slide lot combinations are constructed in a similar manner.

All repetitions of fluids run on a given analyzer/slide lot combination were predicted through that analyzer/slide lot combination's factory calibration curve.

VARIANCES The variance of predicted concentration across all analyzer/slide lot combinations for the three controls and two patient pools are printed in table 14.

TABLE 14: Fluid Concentration K-Model Variance Factory Calibration Variance Kodatrol I 83.27 16.31 24.96 Kodatrol II 370.30 1110.33 1273.48 Control 24102 53.64 8.39 7.99 Human Pool 1 706.31 974.10 993.70 Human Pool 2 379.32 140.57 212.34 -35The variances of predicted concentration using the factory calibration model are only slightly greater than the variance obtained when one uses the conventional K-Model three-level customer wet calibration model.

CONCLUSIONS This data shows that the factory calibration method of the invention yields results similar to the conventional K-model three-level customer calibration model for glucose and amylase. Similar test data can be shown for most other Ektachem slide element chemistries.

Claims (8)

-36CLAIMS:

1. A method of quantitatively determining an analyte in a liquid sample applied to a slide element by measuring in a read station of a field

2. A method as defined in claim 1, wherein said step (d) includes the further step of correcting 15 for any deviation produced in said field analyzer due to lot-to-lot variations or aging variations so that said read response from said field analyzer corresponds to a standard fresh slide element of the type actually used in said step (a). 20

3. A method as defined in claim 1 or 2, wherein said response is selected from the group consisting of reflectance R, optical density OD where OD = log (1/R); rate of change in R, OD or a transformed OD; or electrical potential created by a 25 differential measurement of ion concentration in two ion-selective electrodes.

4. A method as defined in claim 1, wherein said response is reflectance or optical density and said k of equation (I) is a value of 2. 30 5. A method as defined in claim 1, wherein said response is the rate of change of either R or OD or a transformed OD.

5. Wherein said steps a), b) and c) are done at a manufacturing facility for said field analyzer and wherein said equation (I) is stored in memory in said field analyzer, so that said field analyzer is factorycalibrated and does not need field calibration. 5 e) ascertaining from said correlation of step (c) what said read response in said field analyzer in step (d) would produce as the corresponding response in said standard analyzer, and f) correlating said ascertained 10 corresponding response to the predicted analyte concentration using said calibration curve of said standard analyzer. 5 analyzer a response developed in the slide element and correlating said response to an analyte concentration using a calibration curve produced from calibrators having known concentrations, such slide elements being selected from a set of slide elements that all produce 10 the same response in said field analyzer except for deviations due to lot-to-lot variations or aging, said same response being a function of the kind of analyzer used; characterized in that said response on said 15 analyzer is corrected substantially to the response that would have been detected on a standard analyzer of a type different from said field analyzer and having one of said calibration curves for correlating the response to an analyte concentration, by the steps of: 20 a) applying at least one level of said calibrators on at least one of said set of slide elements and reading the response in said standard analyzer, b) repeating step (a) but by reading their 25 response in said field analyzer; c) correlating the relationship of the different responses in said two analyzers by using the equation (I): Rstandard = Bo + Bi.gF(R2) + B2.gF(RF) k (I) 30 where Rstandard is the response detected in said standard analyzer, Rf is the response detected in said field analyzer; gF(RF) is a spline function used to transform the responses Rf; Bo, Bi and B2 are standardizing coefficients; and k is an exponent 35 corresponding to the nonlinearity of the equation, -37IE 92089' d) applying an unknown patient liquid sample to a slide element selected from said set of slide elements and reading a response in said read station of said field analyzer;

6. A method as defined in claim 1 or 2, wherein calibration mathematics is provided as a set of 35 machine-readable digits with each set of slide elements -38sent to any one of said field analyzers, that are suitable to establish for said one field analyzer the value of said coefficients Bo, Bi and B2.

7. A method as defined in claim 1 or 2,

8. A method of quantitatively determining an analyte in a liquid sample applied to a slide element substantially as hereinbefore described by way of Example and/or with reference to the accompanying drawings.