Revvity, Inc. (RVTY) Earnings Call Transcript & Summary

Hello, everyone, and thank you for joining our webinar on Mastering Calibrations for Liquid Dairy Analysis From Basics to Best Practices. My name is Jackie Trudell, and I'm the senior Global Market Manager for Dairy and Beverages at PerkinElmer, and I'm honored to be your moderator for today's session. Calibration is a truly essential component of any dairy testing protocol as it ensures that your instruments are producing both accurate and reliable results. As such, adopting best practices into your operations for efficient and effective calibration is incredibly important to keep your instruments and your production processes running smoothly. The webinar today will cover a wide grade of topics, offering an overview of calibrations for dairy instruments and how to perform these calibrations, the challenges posed by more complex matrices during calibration and how to properly maintain and verify your calibration. I'm joined today by a great panel of experts in the industry that will be guiding us through these topics. First, we'll hear from Jennie Inman, the Technical Support Director at QCL Scientific, a distributor of scientific equipment, supplies and support based in the U.K. Next, we'll hear from Ryan Taylor, PerkinElmer's Senior Field Application Scientist based in Springfield, Illinois, focusing on food applications using IR and NIR technologies. We'll then finish after presented content with Sofia Karlberg, our Stockholm-based Global Product Manager for our liquid dairy analyzers at PerkinElmer. We'll close out the webinar with a question-and-answer session, where our panel will answer questions from the audience. And before we get started, I'd like to cover a few of these common housekeeping items for webinars. At the bottom of your screen are multiple application widgets that you can use. All of the widgets are resizable and movable, so feel free to move them around to meet your preferences. If you have any questions for our panel during the webinar, we encourage you to submit them through the Q&A widget on the left side of your screen. We'll try to answer these during the webinar. But if a more detailed answer is needed or if we just run out of time, we'll reach out by e-mail to ensure that we get you an answer to your question. We'd also love to hear your feedback on the webinar today. So please share your feedback about the webinar as well as topics you'd be interested in learning about in future webinars using the survey link in the menu dock. An on-demand version of this webcast will be available in a few hours using the same registration link that we use to log on today, so be on the lookout for that. And as we've got all that covered, I'm happy to hand over the presentation to Jennie Inman, and I'll be back with you for the question-and-answer session following the presentation. Now Jennie, over to you.

Thank you, Jackie, and welcome, everyone, an introduction to calibration. There are many forms of calibration, but today, we are focusing on linear regression for liquid and analyzers. Different technologies are available for liquid analysis. Most commonly today, FT and IR is used. Fully transformed mid-infrared systems have PLS models sat in the background, which will comprise thousands of spectra across a wide range of the parameters required such as fat, protein, et cetera. Filter instruments have a filter wheel which contain individual filters that select IR imaging at specific wavelength for each parameter. Inter-correction factors are then applied to determine the concentration of the parameters. These base settings are applied by the manufacturer. However, both technologies require linear regression calibrations to be applied to align the instrument to reference values with variations such as seasonality, breed of animal and analyzer status being considered. Liquid dairy analyzers all have flow systems. Components within the flow system such as pumps, cuvettes, cells or homogenizers all wear over time. This wear generally causes the calibration to drift upwards so the increment overreads. It is important to correct this drift with a recalibration so that the instrument remains aligned to reference values. This ensures that processing is both optimized and compliant to legislation. For example, semi-skimmed milk should not read 1.45% fat as this legally too low. But equally, shouldn't have a fat value of 1.6% as this means the factory has given away valuable fat. All IR instruments are classed as secondary testing methods and can only be considered to be 1.5x as precise as the uncertainty of the reference method on a single analysis. It is important, therefore, that the precision of the reference standards used is known. With some reference materials, the given accuracy of the reference test is assigned, i.e., plus or minus 0.25% for fat by Rose-Gotlieb on CRMs. Generally, the more accurately uncertainty of measurement is, then the more robust the calibration data becomes. Caution needs to be exercised if calibrations are to be built using results generated by an internal laboratory. For example, using the Gerber method for analyzing cream, which has an accuracy of 0.5%, does not give a robust uncertainty. Where possible, reference standards should be used, which are as close as possible to the product channels being calibrated. If we use raw milk in the U.K., as an example, the calibration should range from 2% to 5% fat. The standards should be made using more milk as this is an unhomogenized product and we ensure the homogenizer within the instruments is fully utilized, and the calibration can capture the efficiency of the homogenizer at the point of measurement. As we mentioned before, the homogenizer is a wearing part. If pasteurized milk, which is generally homogenized, is used to calibrate a more milk channel, then the instrument homogenizer doesn't have much work to do. And this is not representative of raw milk. Some reference materials contain bronopol or an alternative preservative and whilst there is little effect on measurement, it should be noted. It is good to ensure that your analyzer is in optimum condition before calibrating. There are a few things you can do to ensure you are good to calibrate. The analyzers should have a deep clean before commencing calibration to ensure any deposits are released from the flow system. Carryout a repeatability test to ensure there is no carryover. Carryover can be seen by looking at the first to second replicate. If the second replicate is significantly higher than the first, that could be an issue with the previous sample being partially left in the cell or cuvette. Check the zero history of the instrument to make sure that the zeros have been stable for the last week or so. Then finally, perform zero measurement before calibrating each product channel. The zero measurement should be 0.01% or less depending on the technology used.

Thank you, Jennie. My goal is to help today with the practical walk-through of the instrument and the software for calibration adjustments. We will first talk about the instrument capabilities and how that applies to calibration management. Secondly, we will show a short video and discuss the instrument functionality. Finally, I will take you through a practical example of a calibration set, what you may expect to see in the Results Plus platform. The LactoScope FT-A dairy product analyzer can measure and monitor a wide variety of products. It is important in calibration maintenance to be sure that your calibration sets and further pilot samples reflect the products that you will measure. For example, while you could calibrate milk-protein concentrate product with a raw milk data set, it may not lead to the most accurate results. If not commercially available, pilot samples would need to be made in-house and validated with reference chemistry methods to verify performance for your analyzer. It is also important that calibration sets reflect your intended range, running low solid level samples on a product that requires higher solids, again may not necessarily give the best performance. Sample temperature may also affect product flow. So running your calibration standards in a manner consistent with production is crucial. Now let's switch to a short video where I describe the functionality of the FT-A and how it helps the calibration process. [Presentation]

Now that our samples are analyzed, we have a set of 10 where lab chemistry needs to be added. On the left, you can highlight your sample and then press the details button. Pressing the [indiscernible] request tab you can enter your known reference chemistry values. These are either provided by your lab or with a commercially manufactured set. Once the reference values have been input into the system, you can revert to the home screen and choose our setup menu. A password, which is available through our support team is necessary to enter the setup menu. Once there, you can choose the validation option. And in the screen on the right, you can then choose analysis profile, parameter, date, time and then search for your data. After verifying all samples that are shown, you can press the validate button to proceed. The scatter plot that you will see will show the current performance and then you can view various plots and calculations. When you are ready to deploy these values, should the statistics warrant them, you can choose on the right which method you want to use. The options are current, slope and intercept, slope only and intercept only. Make your choice and press deploy to auto-bias correct that parameter. Then you can go back and repeat the process for your other parameters. The data presented here is an ideal situation, but we know that most situations are not ideal and in fact, can be rather tricky. For that, I will hand you off to Jennie to describe more on statistics and what we might see in non-ideal situations.

Thank you, Ryan. Ryan showed how a good calibration will look. There can be many different reasons as to why bad calibrations can occur. In this slide, we can see a clear outlier. The R squared value which shows how well the instrument values correlate to the reference values should be close to 1. In this example, it is down to 0.0116 and the slope is also too low at 0.5066. The root mean squared error predictions should be less than 0.05 for milk. And here we can see it is at 0.376. An outlier can occur and many commonly due to sample repeatability or incorrect reference data inputs. In this example, it is clear that we need to remove the outlier from the data set. Doing so gives us a far better correlation with an R squared of [ 0.999 ] and root mean squared error of prediction of 0.0506. In this example, we can see that the majority of the results fit the linear regression line. However, there is one result that is skewed in the linear regression, and this gives us an R squared of 0. 919. Again, removing this outlier improves the R squared to 0.999, although the overall accuracy of both sets of data is similar. Not all analyzers will have in-built calibration software. So here, we have a simple linear regression in Excel. Using a program such as Excel to build a regression is perfectly good. It's an easy way to see outliers and, in this way, to work with data. The new slope and intercept values would need to be added to the instrument and the Excel data kept for evidence to back up decision-making process for calibration. Before we go to the next slide, let's take a quick poll. Please answer the following questions. How frequently do you perform calibration for your analyzer? Please select one of the following or the closest to your frequency: monthly, every quarter, every 6 months, once a year. [Voting]

We will present the results at the end of the webinar. Calibration frequency. It's the golden question that I'm asked all the time: how often should I calibrate my analyzer? In reality, instruments vary for many reasons, but the heavily used analyzer will require more frequent calibrations. Generally, with a new installation, we recommend the calibration interval of 6 weeks. This will then allow you to see how much the instrument has moved in that time. If, for example, the instrument has only moved 0.02% on milk fat, you could push the frequency to 8 weeks and reassess. Some sites have specific products that are more important than others, and the site may choose to calibrate this product more frequently. For example, cream is a common product for extra calibration as the losses have a greater consequence. If an analyzer is left to drift and not corrected, the losses are significant as well as causing issues to production. Therefore, the real art of calibration is to correct the drift before it becomes noticeable. Bespoke calibrations. Reference standards are not always available for all products. Examples include yogurt, whey and milkshakes. In this instance, a minimum of 6 samples should be taken across the working range, including outer-specification samples, if possible, so samples that are too high or too low. The sample should then be analyzed on the instrument and sent for reference analysis to an accredited laboratory for the parameters required, for example, Rose Gottlieb, Kjeldahl, et cetera. It is important to ensure that as much of the reference analysis is conducted [indiscernible] to ensure that reference results are repeatable. The reference results then need to be added to the instrument software and the calibration built where we can carry out in linear regression in Excel, as I mentioned earlier. There is lots of terminology when it comes to calibration, but here are some of the most important terms. SD, standard deviation, a measurement of how dispersed the data set is in relation to the mean. Generally, the lower the SD the better the value for calibration. CV, coefficient variation. This is used to describe the overall accuracy of an analyzer. So when we say the LactoScope FT-A operates to less than or equal to 1% CV, this means that milk with a fat of 3.5% will be plus or minus 0.035%. SEP, standard error of prediction. This is the standard deviation of the current calibration. SEC, standard error of calibration, is standard deviation of the new calibration or accuracy. Note that the SEC should be lower than the SEP or the current calibration in place is actually better. R squared, this is the correlation coefficient of the variances between the analyzer results and the reference values. Thank you. I will now hand you over to Sofia.

Thanks, Jennie. Jennie and Ryan walked you through the calibration best practices. And I will just make a short summary before we move to the Q&A session. As you now know, the whole purpose with the calibration work is to get your analyzer to read as close to the truth as possible. And the truth in this case should be the results you get from reference chemistry. If properly done, calibration will give a slope and intercept recommendation to give you accurate and precise results from the analyzer. Just as you can see in this slide here where there are yellow dots close to the foci in the middle. But with experience, you will know what to do in different situations. In the graphic with inaccurate but precise measurements and adjustment will move the results back to the foci. A trickier situation is when results are found to be both inaccurate and imprecise. So here, an adjustment will not help, but further troubleshooting is required to understand the underlying problem. So it can be different things. Is it the reference material, perhaps something with the analyzer or maybe how the samples were analyzed? So by doing this many times, you will gain that experience that is needed to make the right decisions. We would like you to remember at least these 3 important things from this webinar. One, treat your analyzer as a hard-working member of your team and make sure it's kept in good condition. Number two, take some time to understand the reference material and the measurement's uncertainty. So this is the way you know what you can expect from your calibration. And the last thing, depending on how many samples are run per day and what the main product is, the need for calibration will differ and the best way to determine when it's time to recalibrate is to regularly run the pilot sample and see how the results change over time. Results can, for example, be viewed in the trend charts. And when the results fall outside the limits, it's a warning that now it's time to do a new calibration. Okay. So those are the 3 important things to remember. And that was my quick summary. So we will now open up the Q&A session. And also, if you have other topics you would like us to cover in future webinars, please let us know. Thank you, and I will now hand over to Jackie, who will moderate the Q&A session.

Thank you, Sofia, and thank you to the presenters on our panel today for this great content. I hope that you found it worthwhile and that everybody learned quite a bit from our panel. So now let's get started on the question-and-answer session with our panel. If you have any questions that you haven't submitted yet, please do so using the question-and-answer panel on your screen. So let's start with the first question. Ryan, I think this is a good one for you. For products where commercial standards are not available, how can standards be developed?

Thanks, Jackie. Yes. Thank you. It depends on the type of sample you're going to look at and, I guess, your production type. But if the commercial samples aren't available, you're going to have to build them from your production. Most likely you would not want to be in production where it's [ tight ] you want to have a range because your standards, you want to make sure they're going to cover your production range. Otherwise, you're just going to standardize to an optimal position and possibly have a poor correlation. So no magic answer on that, but you're going to -- you'll have to find some variation within your production and then make sure you have accurate reference chemistry to validate those samples. And hopefully, your samples will be large enough that you can store them and create pilot samples out of them for future use.

All right. Thanks. So a couple of things that you said there and maybe that came up just for clarification here. What's a good reference sample? And what's a pilot sample, Ryan? And just a couple of definitions here to make sure that we're all on the same page.

Of course. We'll go with pilot sample first. That's the one that was top of my mind. That is going to be a sample you might run on a daily basis. Something to look at, a quick spot check of how your production -- or how your sample or your analyzer is running. Usually, a pilot sample is a storage sample. So it might be refrigerated or even frozen and thawed out, warmed to temperature, and so you can run it. And so you can compare daily your pilot sample hasn't changed. And if it changes past a certain deviation, do you need a recalibration set? Good reference chemistry would be repeatable reference chemistry. So if we're looking at the Gerber method or if you're doing some sort of nitrogen combustion analyzer for protein or a moisture method for your solids value, you want to make sure that those are repeatable and reproducible. And so you're looking to make sure that your reference chemistry is done in duplicate, sometimes even triplicate, just to get a good feel for how tight of a measurement you can make.

Great. Thanks, Ryan. So hopefully, the audience here remembers we asked that poll question during Jennie's presentation, where we were asking how frequently do you perform your calibration? Jennie, I think, we've got those results on the screen now. Do you want to share those with the audience and maybe some comments?

Yes. Thanks, Jackie. So by far, in votes, 52.4% of our audience calibrates their analyzer monthly. I would say this is a pretty common frequency. And depending, again, as we've said on use we generally ensure that the instrument doesn't drift too far beyond what you'd like. Then we have 21.4% of the audience calibrating quarterly. Again, that's perfectly fine. If you have an instrument that is not working too hard, 7.1% of our audience, every 6 months and then 19%, once a year. And I'm guessing these guys are probably not using their analyzers quite so hard. So, therefore, the calibrations aren't drifting so quickly.

Yes. That's great. Thanks. Jennie. So Sofia, I got a question for you now. Does PerkinElmer provide reference samples to customers?

Thanks, Jackie. No, we don't really supply these reference samples. However, we do have good collaborations with other companies that do provide these samples. So we can always help out if you have questions or you don't know where to get them. We can always give recommendations, but we don't provide them at PerkinElmer samples.

All right. And what about calibration services? Is this something that PerkinElmer offers?

I mean...

That's for you, Sofia.

Yes. Of course, we want to give our expertise if we install an instrument, for example, that we can help you get going and we can help out with the first calibrations and maybe set up the calibration scheme and get input to that. But I mean, for the routine calibration work, that should be done within the site and nothing really that we will come out and do for you.

All right. Jennie, a question for you. Do raw milk, pasteurized milk and cream need to be warmed prior to running the samples? And does that warming make a big difference in the butter fat outcome?

It's a good question, Jackie. So raw milk and pasteurized milk don't generally need to be warmed. Most instruments will preheat the sample to 40 degrees. Cream, depending on the fat content, it may be easier if that sample is warmed up, particularly if it's a cold one taken from a fridge as opposed to warm of production. What you generally see, and it's common with my customers that they will microwave their sample, to be fair, but only to about 36 degrees. You don't ever want to take it higher than the instrument preheater. It just makes it a little bit more easier for the instrument to pump the sample. If you overheat the sample though, you will obviously see a variation in butterfat, and the sample won't repeat either. So that's the biggest thing to see from that. If the sample doesn't repeat, there's probably something wrong with the sample in the first place.

Okay. And then whipped cream that you mentioned, why is cream oftentimes split into 2 calibrations, a single and a double? Why is that usually done?

So we do that because there is actually a nonlinearity in the calibration between sorts of 30% to 33%. So if you try to take that into account, you then move the top and bottom end of the calibration, it becomes less stable, whereas if you split them into say, 11% to 30% and then 33% to 50%, you get far more robust calibration.

Okay. That makes sense. Ryan, I have a question for you. How do blockages affect FT-IR results? Ryan, I should say, sorry -- yes, with the calibration, how would blockages affect the calibration?

Sure. If there is a partial blockage in the system, if it's in the cell or if in the milk filter, we would see essentially a distribution in fats or in other particulates in the milk sample would probably be affected. So if there is a partial blockage, we probably would see a skew in the data, unknown exactly how to quantify that without knowing what type of blockage. But usually, if we're doing a deep clean before calibration process and a clean and 0, we'll know -- we should be in a maybe even a manual inspection, we should know that the instrument is free of blockages at that point. But if there is a blockage or a partial blockage in the cell or in the milk filter before the calibration, it probably will skew the data.