Using field data to refine/finish your sensor. biosensor and medical diagnostic

ZP is a world leading contract developer of sensors, biosensors and IVDs. In this note we discuss a strategy ZP can use on your programs to accelerate the time it takes to get to to market.. The strategy discussed  works in a number of businesses and technical  scenarios, including where the strategy is to perform a 510 K submission.


The traditional way of developing a sensor, biosensor or IVD is to spend several years developing the sensors and then going into the field and/or clinical setting and starting testing on real samples. This is a sequential effort and means  that too much time is spent tuning the assay on samples that are not representative of the the real world samples. In addition the methods used to extract the signal from the raw data are often highly manual with many teams recycling techniques used in the past. For example glucose signals are often measured by using chronoamperometry and chronocoulometry, but this ignores the plethora of other techniques available to the development team.  Further with the slow traditional strategy it ignore the simple fact that investors give a company a higher valuation if it has data on real samples; therefore at ZP we think it is important to get to real sample data ASAP, and rather than sequentially building the sensor technology and then gathering real world data, the two efforts should be in parallel and/or overlapping.


At ZP we rapidly move you to real world samples, using a three phase approach.  The advantage that the signal extraction algorithm is tuned on real sampes, whilst also gather real world sample data, the phases are::


1) PHASE ONE - ZP develops your proof-of-principle assay on our pre-existing electrodes and our AnaPot electronics; upon completion of a proof-of-principle study we can translate the results onto a some hundreds/thousands of functionalized electrodes and start testing with real samples.


2) PHASE TWO - ZP and you perform some initial test with real samples to ensure that the new diagnostic/sensor is approximately working.


3) PHASE THREE - This is a preclinical study on some hundreds of real samples. This PHASE THREE involves two parts:


For this strategy to work we do need three pieces in place, these are: 


ONE - There is an existing 'gold standard'/predicative device or assay available. 


TWO - Ideally ZP should have developed or supplied your initial sensors and  electronics; the reasons why this is important is that we will programme the electronics to analyse the sample with several electrochemical waveforms. The strategy is based on the fact that there are multiple ways of extracting an electrochemical signal for an analyte,  and so to maximise the effectiveness of the Training Phase we will over analyse the sample with multiple techniques and gather all the data. The more data we gather than the higher the probability of success when we mine the data for the analyte signal later.



THREE - We will require authentic real world samples , be it urine, blood, red wine etc. The samples should  cover a range of analyte concentrations, with the analyte of interest being present in the samples from a  low concentration through to a high concentrations.



As discussed above ZP collects data on hundreds of real world samples using both the new technology and the predicative technology, subsequently we split the data randomly into a training data set and a test data set.


With the training set of data we use the ZP Signal Extraction Loop Training Phase Strategy for developing the calibration/extraction algorithm for your device.The output from the training phase is an algorithm to extract the analyte signal from the raw data.


As discussed above the technology provided by ZP will over analyse the samples so that we can extensively mine the raw data. The strategy is to develop a signal extraction algorithm which leads to a  minimum error between the new diagnostic and the existing/predictive diagnostic.


The ZP signal extraction algorithm will automatically loop/iterate until the error betwen the existing/predicative device and the new device is minimised, see adjacent figure.



 As discussed above, he data for the testing phase was in fact collected during the training phase, but now the signal extraction algorithm developed in the training phase is used to extract the signal form the test data set, and a Deming regression analysis performed to ensure equivalence between the predicative assay and the new assay.


If you have any questions regarding how ZP can help in the development and manufacturing of your sensor, biosensor IVD, please don't hesitate to contact us.