| Author | Message | ||
     Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 27 Registered: 9-2001 |
Jonathan - depending on the instrument, and the accompanying software package's capabilities, some of the other tests might be automated, and can be performed in a few minutes with practically no effort just by selecting the function. In these cases, the recommendation for a testing interval is daily, and an SOP will often call for those tests to be performed every morning, before starting the day's analyses. These days, most instruments have a "repeat scan" function of some sort, which would allow data for a noise test to be collected equally easily; it would then be a matter of selecting the noise calculation function. So in this case a noise test could also be performed on a daily basis with not much extra intrusion into regular operation. A full linearity test might take more time and effort, but to measure just a high and low reflectance sample and verifying that the readings are within some reasonable limits is only a matter of measuring two samples. By including this you could have the basis for a daily test procedure that would take minimal time and effort, but would detect most actual and even incipient instrument failures, before much damage was done. And you would catch the problems before more than one day's measurements were affected and invalidated - pretty good return for a small investment, I think. Howard \o/ /_\ | ||
| Jonathan Hoggatt (jhoggatt) Junior Member Username: jhoggatt Post Number: 10 Registered: 6-2006 |
Howard, That makes sense... one of the vendors I was looking at recommended the tests every 6 months, so I just cut that in half to get my three month figure. I completely agree with you and Gary... this is essentially a risk assessment. In all practicality, it doesn't take all that much time to do the photometric linearity... so I think maybe I will just do it every Monday morning and then I should be good. If I think of it compared to other traditional methods which have system suitability before every run (read: HPLC) then perhaps maybe even every day would be good (though possibly slightly excessive for NIR). How far are we from having a monograph or set qualification standards for NIR in the pharmaceutical industry? Are we looking at many months, or many years? | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 26 Registered: 9-2001 |
Jonathan - the exact interval at which to measure noise and linearity tests is up in the air since no official or standards-setting organization specifies it. However, three months seems a very long time frame for that - a lot can happen. My "gut" tells me that one month is a much more reasonable interval, but there is no objective justification for that. It is possible to develop an objective justification - not easy, perhaps, but possible, by optimizing two conflicting considerations. The first consderation is easy to describe: the more often you run the tests, the more time and effort you will spend doing that. The conflicting one is a bit harder to describe, but here it is: during the interval between tests, the instrument is in routine use, performing whatever analyses the instruments was obtained to do. The results are valid only as long as the instrument has been verified to be valid. Now, suppose a problem with the instrument occurs, at some unknown time between two tests. The presence of the problem is not revealed until the next time the tests are run. Therefore the instrument is found to not be valid, and consequently no valid results could have been obtained since the problem occurred. However, since it is not known when the problem occurred, I believe that in this situation, regulatory authorities will require that all analyses run since the last time the instrument was shown to be valid, be considered invalid. The consequences of this are obvious: all samples must at least be re-analyzed, after the problem is corrected and the instrumnent revalidated. The regulatory authorities may require product recalls, or other very expensive actions, be taken as a consequence of this, for all materials analyzed in the interval between the two tests. The view that the validation AFTER a measurement is the one that certifies the validity of the measurement indicates that the instrument test should be run frequently, to minimize the number of samples that would be affected by a finding that the instrument was out-of-spec, the next time it is checked. Thus, it becomes necessary to trade off the time and expense of doing frequent checking against the potential expenses incurred by testing infreqently, when an out-of-spec condition is found, and find the interval that will minimize the total expense. Howard \o/ /_\ | ||
| Jonathan Hoggatt (jhoggatt) Junior Member Username: jhoggatt Post Number: 9 Registered: 6-2006 |
Gary, Thank you for the reply... I certainly intended on using the USP guidelines, I was mainly just unclear as to how other people have interpreted them since there is currently no monograph available. My current thought is to do a wavelength uncertainty test every day I use the instrument with NIST SRM 2035, and do the Photometric Linearity and noise tests every three months. This is, of course, in addition to any vendor suggested qualification procedures. Since the USP does not have any Relfectance standards, and there seems to be some debate about how to actually perform these tests, and from Howard: "If you want to measure those parameters in order to check whether a commercial instrument is in spec, your best approach is to ask the manufacturer how they do it, and then use their same procedures." I guess my question to Gary or anyone else familiar with FDA submission (as I am new to this) would be: If I follow my qualification procedure: Wavelength Uncertainity everyday with NIST SRM 2035, and a Photometric Linearity and noise test every three months using vendor supplied reflectance standards and a vendor recommended method, and do any vendor recommended qualification protocols... will this be satisfactory to comply with USP 1119 and the FDA? | ||
| Gary E. Ritchie (gary) New member Username: gary Post Number: 1 Registered: 1-2006 |
Jonathan, My name is Gary Ritchie and I am the liaison to the General Chapters Committee and Scientific Fellow for PAT at the USP. I am closely involved with the USP standard you are discussing here and would like to clarify some points. You said: "For the other three tests the USP chapter recommends carbon doped polymer reflectance standards ranging from 10% - 90%. After just doing an extensive vendor evaluation for my company I can tell you that most vendors sell some sort of standards for this measurement. After reading the responses on this topic, it sounds like these USP guidelines are moot... that it isn't going to help with anything for reflectance measurements. Does anyone here, working under GMP practices in a pharmaceutical setting, follow these USP guidelines for instrument qualification? If so, how often do you do these tests and how are they reported?" The USP standard, which is official in <1119> and will be made available following the committee�s assessment of the collaborative study data. USP standards are in fact traceable and are enforceable. Vendor standards may or may not be traceable, depending on what NIST or any other organization who makes claims to providing standards, agrees to for certification or recertification purposes, but a USP standard is a certified, traceble and legal standard. This is not a moot point. In fact, you may visit the USP website and note that the FDA is celebrating its 100th birthday today, and will see the USP message acknowledging that and the legal relationship between the FDA and the USP. With respect to chapter <1119>, while the chapter does not currently have any citation to any monographs which would require it to be enforced, the requirement for checking wavelength uncertainty of your equipment is something that the FDA would certainly pay close attention to, and that the USP also provides a standard for doing so is also something that the agency will take into consideration. I hope this clarifies your understanding, or lack of, of the USP. There is no recommendation for frequency of running the test, accept that it should be done at intervals as required by the needs of the user However, good science and risk assessment as to the firms requirement for the quality of the data and its impact on the patient is what you should consider as you establish a routine for NIR measurements at your firm. The reporting of the data (for wavelelngth Uncertainty) will be based on the final wavelengths selected for evaluation and the decision of the committee. The reporting of the data from the noise and photometric accuracy/precision are self evident. If you have any further questions, please feel free to contact me at 301-816-8353. By the way, I am the General Chairman of the International Diffuse Reflectance Conference, being held from Aug 4-11, 2006 in Chambersburg, PA, at Wilson College. Please visit the website (www.idrc-chambersburg.org). Consider attending the conference where I will be available for further discussion. Thank you for your consideration. GER | ||
| Jonathan Hoggatt (jhoggatt) Junior Member Username: jhoggatt Post Number: 8 Registered: 6-2006 |
I too had similar questions regarding calibration standards... but more from an instrument qualification standpoint in a pharmaceutical setting. USP General Chapter 1119 describes four tests for instrument qualification: Wavelength Uncertainty Photometric Linearity Spectrophotometric Noise High Flux Spectrophotometric Noise Low Flux For Wavelength Uncertainty in reflectance mode, the USP chapter recommends using their standard (USP Near-Infrared Calibrator RS) or NIST Standard SRM 2036. For the USP standard the chapter even gives the expected wavelengths and the tolerances for qualification. For the other three tests the USP chapter recommends carbon doped polymer reflectance standards ranging from 10% - 90%. After just doing an extensive vendor evaluation for my company I can tell you that most vendors sell some sort of standards for this measurement. After reading the responses on this topic, it sounds like these USP guidelines are moot... that it isn't going to help with anything for reflectance measurements. Does anyone here, working under GMP practices in a pharmaceutical setting, follow these USP guidelines for instrument qualification? If so, how often do you do these tests and how are they reported? By many of my posts in the last couple of weeks I am obviously new to NIR and I appreciate all of the help I have been getting from many of you. I am interning this summer before going on to start my Ph.D. (just finished my Masters) and for some crazy reason I chose to implement NIR methods even though I knew nothing about it, for a company who has never owned an NIR. But, I have learned much in the last two months and am fascinated by all of the varied uses and potential. | ||
| prathap (Prathap) |
hai, can any body help me to sort a list of NIR calibration standards as per NIST for the below mention parameters to evaluate the prformance of my nir in transmision and reflectance modes 1. wavelength accuracy 2. photometric accuracy 3. stray light thanks for the reply regards prathap | ||
| hlmark |
Prathap - as last week's discussion showed, a proper determination of those parameters involves more than simply measuring a couple of samples, even if suitable ones are available. Among other things, it will depend on the type of measurement you want to make (transmission or reflection, for example) as well as the purpose you want to use them for (check whether a commercial instrument is within spec, or to design an instrument of your own, for example). I noticed that you did not respond to Tony's question as to why you want to make these measurements. Incidentally, here's a comment on Tony's message about using mixtures with carbon for stray light testing: you can (sometimes) depend on Beer's law when measuring in transmission, but not for reflection. The correct physical epxression for reflection measurments is still unknown, therefore simply making measurements as he describes and using them to tell you about stray light will almost certainly mislead you, since you will likely attribute non-linearity of response to stray light when it is actually due to the innate behavior of the sample. That said, NIST does not make very many standards for instrument checking. They do provide certified wavelength standards; those can be used to check wavelength accuracy, but currently the ones they provide are useful mainly for transmission measurements. They used to provide a reflection standard (CRM-1920a) but that has been discontinued for several years now. Standards similar to CRM-1920a are available commercially, however, some with enhanced wavelength ranges (you can send an e-mail message to Gabor Kemeny ([email protected]) to ask about the one he has patented). NIST does not provide reflectance standards. However, they will measure and certify the reflectance of a standard that you provide. Labsphere pretty much has a lock on the market for reflectance standards these days. You can check their web site to find out about the accuracy and whether they will provide NIST-traceable certification for them. As last week's discussion showed, there's no "standard" for stray light. Measuring that is mainly a matter of trying to make sure that your sample does not transmit (or reflect) anything, and then see if you can measure any energy. There are some electronic effects in instruments that can masquerade as stray light, so you have to know how to tell the difference. The bottom line is that instrument design and testing is a specialized field a specialized field, and you cannot expect to enter it casually and expect to get meaningful results. Good luck Howard \o/ /_\ | ||
| hlmark |
Prathap - two more comments: 1) The measured reflectance of a sample, even from a NIST-certified sample, depends on how you make the measurement, as well as on the sample. Therefore you have to make the measurement the same way NIST does in order for it to be comparable. 2) If you want to measure those parameters in order to check whether a commercial instrument is in spec, your best approach is to ask the manufacturer how they do it, and then use their same procedures. Howard \o/ /_\ | ||
| Edward Stark |
Prathap and Howard I suggest contacting Art Springsteen at Avian Technogies ([email protected])for NRC and NIST traceable reflectance and transmission standards as well as rare earth wavelength standards. It is incorrect to state that Labsphere pretty much has a lock on the market now although they are a major supplier. I agree that it requires great care to measure instrument performance and the techniques vary among instrument types and models. Please remember that in reflectance other forms of stray light, eg the specular component of reflection, usually outweigh the instrument spectral stray light. Regards, Ed | ||
| hlmark |
Good point, Ed; I should have at least remembered about Art and certainly have included mention of him. Since Prathap likely doesn't know Art, he's the guy who created probably most of Labsphere's standards, before he formed his own company, and probably knows more about standards for NIR than anybody else. EARTH TO ART - why don't you chime in here? Howard \o/ /_\ | ||
| djdahm |
I feel compelled to comment on Howard�s comment about Tony�s message in this chain. Specifically: �The correct physical expression for reflection measurements is still unknown, therefore simply making measurements as he describes and using them to tell you about stray light will almost certainly mislead you, since you will likely attribute non-linearity of response to stray light when it is actually due to the innate behavior of the sample.� The correct physical expression for reflection (I prefer remission) measurements is in fact known. Howard is reluctant to acknowledge that, because the he wants the expression for remission to use the same absorption coefficient as transmission measurements, and it does not. But the form is known, and it is unambiguously non-linear. Therefore his cautionary comment that �you will likely attribute non-linearity of response to stray light when it is actually due to the innate behavior of the sample� is right on. I would go on and be a bit stronger than he was. Trying to determine the stray light in the presence of front surface reflection is a fool�s errand. The non-linear function will have a maximum absorbance at the value of log (1/Rfront), where Rfront is the front surface reflection. That term refers to the light that strikes the first layer of particles and is remitted back without ever having entered the particle, and therefore does not carry the absorption signature of the sample. There has been good work done (by Harrick and others) on removing the front surface reflection using polarized light. I don�t know the experimental details, but I assume that there will still be a problem at telling the difference between stray light from outside the sample and light that has bounced off the sample in some fashion. For the purpose of evaluating your instrument, I would be tempted to say that the stray light in the reflectance mode is probably irrelevant unless you have a major light leak. | ||
| hlmark |
Don - you misunderstand my point whan you say "... he wants the expression for remission to use the same absorption coefficient as transmission measurements, and it does not." While that would be the ideal, it is not what I insist on. I do not necessarily want to see THE absorption coefficient, but I do want to see AN absorption coefficient. That is, a quantity that can be put into an equation that is a constant, and that represents an intrinsic absorption property of the material, and it's magnitude is not affected by the measurement conditions or any other extrinsic or extraneous conditions. The absorption coefficient for clear transmission (i.e., "Beer's Law" coefficient) is not just a number that makes analysis come out right. It represents a fundamental property of the material that can be calculated (in principle, at least) through quantum mechanics, and therefore represents a fundamental intrinsic property of the material. Unless I misunderstand you, we're not quite there yet, for reflection measurements. Howard \o/ /_\ | ||
| hlmark |
Don - The principle behind the work of Harrick and the others who use polarization to get rid of front-surface reflection is not hard to figure out. From the Rayleigh equations for reflection from a dielectric, we find that at a particular angle of incidence (called Brewster's angle, where the angle of incidence obeys the rule that tan{theta} = n (where n is the refractive index of the material)) the reflected light is polarized perpendicularly to the plane of incidence. The other polarization is comletely refracted into the material. Therefore if you illuminate the sample at Brewster's angle, and measure only a narrow beam at the "mirror angle", the only front-surface light reflected in the direction of the detector will be light polarized perpendicular to the plane of incidence. Therefore, if you put a polarizer in the beam you can reject that polarization as well, and no front surface reflection will make it to the detector. Front-surface rays reflected at other angles will be rejected geometrically, since you're measuring only a narrow beam and those rays will not be in the field of view. \o/ /_\ | ||
| prathap (Prathap) |
hai every body, thanks for your valuable suggestions Basically we are the manfacturers of spectrophotometers (UV, VIS, NIR)(www.elico-india.com) being in the design department i want to know in detail the performance of my instrument. prathap |