ARCHIVED - Elisa Systems Casein: Performance Evaluation

Disclaimer: Inclusion of this method in the compendium does not imply endorsement or approval by Health Canada.


The purpose of this study was to generate performance data on the Elisa Systems Casein test kit for the detection of casein, one of the major allergens in cow's milk, as part of ongoing efforts to evaluate food allergen detection methodologies and further their introduction into the Compendium of Methodologies. A representative non-fat milk powder was chosen to be the designated reference material, according to the definition established by the AMC.

This evaluation involved the analysis of selected food matrices which had been artificially fortified (spiked samples) with the non-fat milk powder.

Method Evaluated

The Elisa Systems Casein kit.

Evaluation Level

Full evaluation under the guidelines developed for the Compendium of Food Allergen Methodologies.

Designated Reference Material

Non-fat milk powder obtained from the National Institute of Standards & Technology (NIST) was used as a reference material. The milk powder is joint material of Agriculture and Agri-Food Canada and NIST with a reference material # 1549.

Participating Laboratories

Health Canada Food Allergen Research Laboratory
Banting Building,
Ottawa, ON

Health Canada Western Regional Laboratory
Burnaby, BC

CFIA Québec Regional Laboratory
Longueuil, QC

CFIA Western Regional Laboratory
Burnaby, BC

Spiking Levels

Cereal samples were spiked with 0, 4 ppm and 10 ppm of the NIST 1549 standard. The other matrices were spiked at 0, 2 ppm and 5 ppm of the NIST 1549 standard. The reason for this will be discussed in more detail in the Results/Discussions section but these levels were found to conform with the evaluation guidelines of the Compendium which call for spiking levels that give a response from the kit of approximately 0, 2 and 5 times the limit of quantitation (LOQ) when applied to the designated reference material.

Spiking Conditions

Spiking was done using a solution of the NIST 1549 non-fat milk powder in phosphate buffered saline with 0.1% Tween 20 (PBS-T) solution. Samples were left blank, or spiked at one of two levels and given a blind identification code.

Matrices of Interest

Three different matrices were included in the evaluation: Cereal, Cookies and Dark Chocolate. These commodities were included as representatives of some of the matrices which could be most likely to contain undeclared milk proteins. The potential for cross contamination with milk products at low levels exists when commodities that do not contain milk or components of milk as an ingredient are manufactured on the same equipment as commodities that do contain milk or milk components.

Materials and Resources

Each participating lab was provided with:

  • ElisaSystems Casein kits with all kit components including the kit insert
  • blind-coded samples (30 of each matrix - 90 samples in total)


Spiking Procedure

The spiking levels were chosen in order to obtain a response of approximately 2 and 5 times the LOQ from the ElisaSystems Casein kit. The ElisaSystems Casein kit uses a different standard than the NIST 1549 Milk Powder for its calibration curve. However, the two standards behaved very similarly on the kit and therefore it was found that no correlation factor was needed. In other words when 5 ppm of the NIST 1549 non-fat milk powder was spiked in a sample, the sample gave a response of 5 ppm with the ElisaSystems Casein kit.

Samples were spiked using a 50 µg/ml working solution of NIST 1549 non-fat milk powder. The 50 µg/ml working stock solution of NIST 1549 was prepared by a series of dilutions. A 1g aliquot of the non-fat milk powder was dissolved in 10 ml of PBS-tween 20 0.1%v/v (100 mg/ml), then diluted 1ml of 100mg/ml + 9ml additional PBS-Tween (10 mg/ml) and, then, 5 ml of 10 mg/ml was diluted in a total volume of 50 ml for a 1mg/ml solution. Next 500ml of spiking solution at 50ug/ml was prepared by diluting 25ml of the 1.0mg/ml in a total volume of 500ml.

The spiking of the samples was performed as follows:

10ppm was done with 1.0ml of Spiking solution 50ug/ml
5ppm with 0.5ml of Spiking solution 50ug/ml + 0.5ml PBS-Tween
4ppm with 0.4ml of Spiking solution 50ug/ml + 0.6ml PBS-T
2ppm with 0.2ml of Spiking solution 50ug/ml + 0.8ml PBS-T

All blank samples were spiked with 1.0ml of PBS-T

Preparation of Samples

The 360 samples required (10 replicates at each of 3 levels in 3 commodities for 4 laboratories) were prepared at the Health Canada's Food Allergen Research Laboratory in Ottawa. 5 g samples were weighed out into 250 ml screw cap bottles (120 samples for each commodity). The samples were separated into groups of 40, then each group spiked at one of the three spiking levels. Each sample was given a code number, then the samples were grouped together for each of the four participating laboratories and shipped by courier.

Sample Extraction and Analysis

Each laboratory extracted the samples following the extraction procedure outlined in the ElisaSystems BLG Kit instructions. Dark chocolate samples were extracted using the enhanced extraction buffer provided for use with this matrix. The cereal samples were extracted using twice the normal volume of extraction buffer (100ml instead of 50 ml). The extraction was performed directly in the sample bottles provided. The sample extracts were then analyzed using the ElisaSystems BLG kit, following the instructions provided in the kit insert.

Results / Discussion

In testing that was performed prior to this evaluation study some issues were encountered with low recoveries for the cereal samples. These issues were resolved by doubling the amount of extraction buffer used for these samples; a change in methodology which was suggested by the kit manufacturer. One effect of doubling the amount of extraction buffer is that this impacts the overall sensitivity of the assay. In effect, this modification changes the calibrators that come with the kit rendering them the equivalent of twice their stated concentrations. For example, when comparing a sample extracted with 100ml of extraction buffer (instead of the usual 50 ml), the 1ppm calibrator that is supplied with the kit becomes a 2 ppm calibrator, the 2.5 ppm calibrator becomes a 5 ppm calibrator and so on. Due to this change, the cereal samples had to be spiked at twice the concentration of the other samples (4ppm and 10ppm instead of 2ppm and 5ppm) in order to achieve a result of 2 and 5 times the LOQ.

The results for the 360 samples in the study showed good inter and intra laboratory consistency. A small number of false positives (3 out of 120 blanks; 2.5%) were reported. Two false negatives were reported out of the 120 (1.66%) lower level spiked samples. Both of these were reported in the cereal samples. It should be noted that, although below the 2 ppm cut-off, the two samples gave results close to 2 ppm and had O.D.'s considerably higher than the blank sample. All of the higher level spiked samples gave positive results.
For each group of 10 samples at one of the three spiking levels in a particular matrix analyzed at the same lab (for example the 2 ppm dark chocolate samples from lab#2) a mean and standard deviation were calculated. Any results more than two standard deviations from the mean were considered an outlier and were excluded. Out of 360 samples only 11 samples were excluded for this reason.
The Z score for an item indicates how far, and in what direction, that item deviates from its distribution's mean, expressed in units of its distribution's standard deviation. A Z-score of 2.0 indicates a result that was two standard deviations above the mean, while a score of -2.0 would indicate a result two standard deviations below the mean. Z-scores were calculated for each lab at each commodity and spiking level. None of the results were above 2.0 and the majority were below 1.0, which shows agreement between all the participating labs.

A summary of the results from each of the four labs is presented here.

As mentioned earlier, the NIST non-fat milk powder behaved very similar to the milk powder used as a calibrator for this kit. For this reason no correlation factor is required and the spiking level and expected result values in the tables are the same.
The choice of the NIST non-fat milk powder was made in order that a well characterized and readily available material was used as the reference material for this evaluation. Although the NIST non-fat milk powder was not originally designed for allergen testing, it did perform well for this application.


The ElisaSystems Casein kit has delivered satisfactory results for the matrices and at the levels tested in this evaluation. Dark chocolate and cookies had similar recoveries in the 75-85% range on average. For the cereal samples a modification to the extraction procedure was required, such that the amount of extraction buffer was doubled. After this modification, the cereal samples over-quantitated slightly at 112-118% on average.

The data from this evaluation, based on spiking samples with the non-fat milk powder from NIST, would be complemented by other data based on "naturally incurred" samples which were manufactured using a known amount of milk or casein. Data from any future studies involving these kinds of samples will be added to this report as it becomes available, in order to further document the performance of this commercial test kit.

Page details

Date modified: