✅ Working Principle

For every test run, the Delvotest® Accelerator Smart (DAS) is able to automatically figure out the right Control Time per plate or even plate-strips (i.e., every 96-well Delvotest® plate can be broken into 6 'strips' of 16 wells each), accurately determining whether each sample is NEGATIVE or POSITIVE. So, how's it able to do this so effortlessly? It's in great part thanks to how its algorithm works, the DAS's 'Working Principle'. In this article, we will explore that logic in detail, highlighting some of the key implications you will need to keep in mind when testing — in particular, the importance that reference controls play in ensuring your DAS performs as expected. 

Z-values, 'Control Time' -10, 'Cut-off Value' -4 and 𝑓(N)

Every minute during an incubation, the DAS takes a picture of your Delvotest® plate or plate-strips. It then converts this image into a 'digital map', translating the colour of every well into a colour value. This colour value is based on a proprietary three-dimensional colour scale (i.e., the so-called Z-scale), from which Z-values are calculated. In this system, low (-, minus) values represent 'yellow' colours, and high (+, plus) numbers represent 'purple' colours. As the DAS is calibrated with an external, validated colourcard, these colour values are absolute measurements from which all plate-wells can be robustly compared.

However, the crucial relational dimension which the DAS has to navigate is time. Indeed, as microbial tests like Delvotest® are not 'mechanical' in the sense that every batch (and even every plate) will have a slightly different 'Control Time', the challenge is not so much about just defining what is the right 'colour threshold' between a POSITIVE versus a NEGATIVE result, but rather about figuring out when this threshold has or has not been passed. This is why NEGATIVE samples play such an important role in how the DAS operates — without them, it would not be able to figure out the right Control Time from which all wells can be defined as either POSITIVE or NEGATIVE.

So, this is how it works:

• For every test run, the DAS presumes that at least a representative proportion of plate-wells will produce NEGATIVE results. For an entire plate (i.e., 96 wells), this presumption is set at 10% of all total plate-wells rounded up to the next integer (i.e., 10 wells); and, for when only one or two plate-strips are incubated, this 10% principle is overridden by a minimum expectation that at least 4 plate-wells will produce NEGATIVE results. Mathematically, where 'N' represents the total number of plate-wells being incubated, this function can be expressed as follows:

𝑓(N) = max(⌈0.10×N⌉,4)

• This representative proportion of plate-wells becomes a sort of leading cohort. Their average Z-values are closely followed every minute.
  
  
• When this leading cohort's average Z-value meets or crosses a threshold of Z = -10, the DAS 'knows' what the correct Control Time should be. It carries on incubating for a little longer just to check more generally if everything is working as expected, and then generalises this 'timestamp' as Control Time.
  
  
• So, if we map this Control Time threshold of Z = -10 to the standard Delvotest® T visual reading colourcard, this is where it would fall:
  
  
• Of course, plate-wells falling under the '-' square on the image above would still constitute a NEGATIVE result, as the NEGATIVE colour range spectrum (i.e., '---' to '-') is broader than just a Z = -10 result. For this reason, the DAS has also established a cut-off value of Z = -4. Thus, all results below this level at Control Time are considered NEGATIVE, and all others above are considered POSITIVE.
  
  

Altogether, this is the DAS's underlying Working Principle. However, because of the way it operates, it is very important to take some key implications into account when testing.

Key Implications

The first obvious implication is that the DAS will not be able to define Control Time without the help of sufficient NEGATIVE samples to represent a leading cohort. To illustrate, consider this scenario:

Let's say you want to incubate a whole Delvotest® plate (i.e., 96 wells) with as many known/expected POSITIVE samples as possible.

10% of 96 wells, rounded up to the next integer equals 10 — which is greater than the overriding 'minimum of 4 samples' principle explained earlier i.e.,

𝑓(N) = max(⌈0.10×N⌉,4)
N = 96
∴
𝑓(96) = max(⌈0.10×96⌉,4) = max(10,4) = 10

Therefore at least 10 of the plate-wells being incubated must be NEGATIVE in order for the DAS to be able to successfully establish the correct Control Time for all the other 86 wells.

In this scenario, in order to achieve the goal of incubating as many known/expected POSITIVE samples as possible, you'd need to make sure at least 10 NEGATIVE reference controls (i.e., known beforehand to produce a NEGATIVE result) would be included as samples.

In reality however POSITIVE results are rare. This is why it's reasonable to generally expect at least 10 samples in any given full plate to produce negative results. However, when testing one or two plate-strips, because of 𝑓(N) and the fact that the total pool of samples is smaller, we advise making sure at least 4 samples are validated NEGATIVE reference controls. Importantly, in such cases the use of NEGATIVE reference controls are about operational performance and verifiability. This distinction is helpful because it unveils a second, key implication.

Even though users can 'tag' whether an individual plate-well has either a NEGATIVE or POSITIVE reference control added to it, the DAS does not use this data at all to figure out Control Time. Instead, it will also measure these plate-wells, determining if they produce 'TRUE NEGATIVE' or 'TRUE POSITIVE' results — in other words, it doesn't take that data from the user at face value. That means the DAS is also a great system for continuously monitoring and verifying your reference control samples are performing as expected — and for this reason, we advise you to add at least one NEGATIVE and one POSITIVE reference control sample to every plate or plate-strip(s) you test, regardless of 𝑓(N). The point of adding these reference control samples is not to form a 'base' from which results are determined (i.e., you can in principle do without them); rather, they're simply there to help you make sure things are running as they're expected. So, if there's ever a problem with your reference controls, this level of continuous verifiability helps you conduct a Root Cause Analysis (RCA).

So, summarising these two implications, here's a layout guide: 

Incidentally, this is why we advise to never leave any plate-wells empty — not only the DAS will believe them to be POSITIVE, but more generally it's just a missed opportunity. If instead a NEGATIVE reference control would be added, you'd be helping the DAS improve its accuracy because of 𝑓(N).

The third implication is that although Z-values can be an enticing, 'quantified indication' of 'how negative' or 'how positive' a result is, it's not that simple. Broad-spectrum microbial inhibition tests can only at best be a semi-qualitative detection method, and therefore never act as a semi-quantitative or quantitative technology. 'How negative' or 'how positive' Z-value results are as indications, they can't ever be correlated to the type (e.g., is it betalactam or tetracycline?), level (e.g., is it 3 ppb or 5 ppb?) or even presence of antibiotic residue at all. To put simply with an example, consider a sample with no antibiotic residue at all but with something in its composition which completely inhibits any bacterial growth — meaning that this sample would never turn a Delvotest® plate-well 'yellow' / produce a NEGATIVE result. The resulting Z-value would be quite 'high' as a POSTIVE, and yet there were no antibiotic residue present. Therefore, we strongly advise against taking any action based on a misleading premise of 'how positive' or 'how negative' a sample may be indicated by its resulting Z-value.

Last but not least, if the matrix being tested deviates from the standard, raw co-mingled cow milk as the DAS has been certified for use with Delvotest® T by AFNOR🔗, as with, for example, raw milk with preservatives, the DAS's standard analysis 'method' may need to be adapted with our help. Typically, this requires a more extensive local validation framework🔗, from which reproducible data forms the basis for targeted, auditable parametric changes. This is why we advise under no circumstances altering 'methods' without our guidance, as doing so will generate results and outcomes we will not be able to support.