Question and Answer Session

December 7, 2004 Presentation:
Validating GFR Information via Data Mining and Practical Aspects of Embedding GFR Data in the LIS/HIS


Welcome to AACC’s Expert Access Live Online program

Our topic this month is Validating GFR Information via Data Mining and Practical Aspects of Embedding GFR Data in the LIS/HIS

This month's expert is Jay Jones, PhD, DABCC, Director of Chemistry and Regional Laboratories, Geisinger Health System. View the presentation and direct your questions to our experts.

AACC would like to thank Bayer HealthCare Diagnostics for making this program possible.

Is it possible to measure creatinine with different kits (manufacturers) and still get comparable GFR estimations ?
Aarau, Switzerand

Jay Jones, PhD, DABCC: Yes. If the creatinine methods correlate well, the estGFR should be comparable. Data mining archived patient populations of creatinine/estGFR results generated with separate methods may be an alternate means of judging the closeness of the comparison (as opposed to the classical means of method correlation with a few set aside specimens).

We report a calc GFR with all our Met panels and renal panels, but are now being asked to report a GFR on a single BUN/Creatinine order. Is it becoming common place to report a GFR anytime a creatinine is performed, whether part of a panel or ordered individually? This request is coming from the imaging center where they want to assess kidney function prior to the scan. Thank You
San Diego, CA

Jay Jones, PhD, DABCC: The NKF's original purpose was to help assess risk for Chronic Kidney Disease in an apparently healthy population with estGFR. To use estGFR to judge safety of an imaging procedure is not something I as a laboratorian don't feel comfortable with. We do not suppress estGFR results going to imaging locations also do not state this as an application we support.

I'm medical director of a 13 hospital consolidated system in southeast Florida (IRL-FL)with very diverse ethnic populations. We do not know the ethnicity (African vs non-African) nor the weight of the patient, which I understand are important for this GFR estimation. I am also worried that some of our 3000 physicians will use the estimated GFR for drug dosage decisions, which I understand should not be done. For these reasons, I have not instituted the estimated GFR. Would you please address these very important implementation issues? Thank you, James Robb, MD
Ft. Lauderdale, Florida

Jay Jones, PhD, DABCC: We prefer providers add 21% to the estGFR of their African American patients during interpretation of the results. Phlebotomists are in a less advantageous position to glean ethnicity as it is often missing from registration information. The key is to educate providers how to use estGFR (improving population assessment of CKD) and also how not to misuse estGFR (calculating drug doses). Pharmacy must help enforce using the correct GFR formula, including those used in PDAs, web sites, etc.

I do not agree that the way to control the reporting of alarming GFR's is to set the limit of abnormality at 60ml/min. This prejudices against young patients with renal disease whose creatinines are raised, and so reported with an 'abnormal' highlight on the creatinine result, but have a GFR of 60-90 and so not flagged as abnormal. I think this would be confusing to non-specialist clinicians. I would be interested in a comparison of GFR and serum creatinine values to identify how many patients fall into this category. We have been reporting estimated GFR's since January 2004 and only report the GFR when the creatinine is raised. This is more logical and can be justified on the basis of the source of the MDRD equation which is patients with renal disease and raised creatinines. Best wishes Hugh Rayner Consultant Nephrologist
Birmingham UK

Jay Jones, PhD, DABCC: Hopefully estGFR calculations will be refined in the future as evidence from population studies accumulates. Clearly estGFR decreases with age and "one size fits all" is faulty. I believe the AKF guidelines will increase the detection of CKD especially in the more elderly adult population and refinements for detection in young adults will be made.

I think the estimated GFR would be very useful in the light of biological variation of creatinine values and the limited value of reference limits. But in Brazil we have not "whites"and "non whintes", only "in betweens". What could we do?
Belo Horizonte, MG, Brazil

Jay Jones, PhD, DABCC: You present a very interesting dilemma. The "in between" patients are the norm so you need not add in the 21% correction factor. Since the estGFR is for population risk assessment, perhaps data mining could help you establish slightly different cutoffs for your specific patient population. Modeling an estGFR cutoff at, say, 40 mL/min and correlating those patients less than 40 with outcome of CKD may help you validate Brazilian norms. There is nothing magic about the 30 and 60 mL/min cutoffs and they may be refined for specific patient populations over time.

do you think the calibration procedure for enzymatic creatinine is important to obtain a correct estimated GFR? if so, how can we verify that the lab method is properly calibrated?

Jay Jones, PhD, DABCC: Standardization of creatinine measurements is set by the NKF as a goal as the estGFR guidelines are rolled out. The first major discrepancy to address is the recognized differences between the enzymatic and Jaffe methods. Calibration of individual methods to a recognized and readily available reference material would be a large step forward.

Has anyone using the Cerner Millennium computer system succeeded in setting up an automatic GFR_EST to chart on patients >/=18 when GFR_EST<60 using Age, Sex & Race? Would they share?
Portland, Oregon

Jay Jones, PhD, DABCC: Cerner uses table-driven test codes and has calculation capabilities similar to most Lab Information Systems (LISes). I'm sure by now someone must have implemented estGFR calculations in this broadly used LIS. Perhaps your Cerner technical support department can connect you with one of their customers.

Would increasing the precision of Creatinine testing impact patient care. I.E there are new enzyamtic creatinine methods that can provide Cr result to two decimal places. Would these assays offer any advantage over the Jaffe and currently aviable enz methods. Do you believe there will be a trend towards increasing enzymatic Cr testing in general ?

Jay Jones, PhD, DABCC: I don't believe precision is as important a factor as accuracy in impacting patient care. Two decimal place numeric precision although significant does not affect estGFR calculation greatly. Although the sensitivity and specificity of various creatinine methods have not been demonstrated with broad usage of estGFR, our data demonstrate a data mining means to judge clinical yield of our enzymatic method. Peer comparison of population histograms (and eventual receiver operator curves) should help validate the various creatinine methods in the long run.

We've recently added the estimated GFR to our renal panel and also offer it as an orderable test. Since we don't have the patient's race available, we report two values (one if African American and one if caucasian). Are there significant differences in the estimated GFR of other races? If so, are there accepted modifications of the calculation to take this into account? We use the MDRD equation for patients 16 years and older. Also, which equation do you use for pediatric patients?
Fond du Lac, WI

Jay Jones, PhD, DABCC: You have implemented estGFR to a greater degree than we have. Since we do not have a race field in our LIS and we decided not to add comments to EMR reports, we leave it up to the provider to add in factors for ethnicity (race). We also report estGFR for all patients greater than 18 years and do not apply additional pediatric calculations. There we several questions like yours asking advice for "taking the next steps" in applying estGFR. We are taking an incremental approach, validating with data-mined population studies as we go. It is likely we will stratify calculations and/or clinical cutoffs in the future as they become better studied.

Is there data that supports the relation between reporting GFR estimations health cost savings as a result of diagnosing and treating CKD in earlier stages?
Indianapolis, IN

Jay Jones, PhD, DABCC: I have not seen published outcome studies showing financial benefits of estGFR identifying CKD earlier and ameliorating its complications. The NKF guidelines point in this direction but in my opinion, it will take long-term outcome studies (probably over a few decades) to quantitate the impact.

I have been reporting eGFR on all serum creatinines for some time using the abreviated MDRD study equation. I statistically compared the eGFR with Creatinine Clearances and found it acceptable. Now the pharmacy has started calculating an eGFR for patients that they can calculate a body surface area. The pharmacy results sometimes are quite different than the labs. Sometimes one is normal and the other is quite abnormal. Is this typical between calculations and what do you suggest to resolve?
Walnut Creek CA

Jay Jones, PhD, DABCC: We ran into the same situation with pharmacy complaining that their calculated GFR was being shortcut by the lab's estGFR. We jointly told providers that lab's GFR was "estimated" and pharmacy's was "calculated" and that they should use pharmacy's for calculating drug dosage. It is important to limit estGFR to its intended purpose - inexpensively and conveniently identifying patients at risk for CKD.

We also use enzymatic creatinine (Vitros). I understand the MDRD equation is based on Jaffe creatinine, which is about 0.3 mg/dL higher than Vitros enzymatic. Should we add this back in before estimating GFR?
Salt Lake City, Utah

Jay Jones, PhD, DABCC: We use an enzymatic creatinine (Roche) as our primary method without adding in an offset or slope factor. We have taken the tact of validating our primary method to calculate est GFR with our own patient population.

"Validation of the est-GFR with the MDRD formula should include validation of its reference range". The original MDRD studies that derived the equation and clinical decision values, were performed on Beckman LX20 analyzer with Jaffe method. Most labs cannot afford the enzymatic determination. Considering all this, how does the 60 ml/min cutoff point relate to the different methods of creatinine determination (Roche-compensated and non-compensated, enzymatic, other manufacturers, etc.)?
Petach-Tikva, ISRAEL

Jay Jones, PhD, DABCC: This is a key question. Different creatinine methods will predictably create different estGFR population histograms. The primary message of this web conference is that each lab should validate its own reference range from its own archived patient results using readily available data mining techniques (i.e. "doing the right thing"). The 60 mL/min cutoff point for different methods will produce different clinical yields of at risk patients. We need to collectively compare our population histograms as a "reality check" of the recommended cutoff points. A nice feature of data mining is that manageable patient lists may be extracted from thousands of histogrammed data points (e.g. 0-30 mL/min) and used for clinical correlation. I am not advocating that all labs should convert to enzymatic creatinine, only that they validate their estGFR histogram prior to turning on calculated parameters in their LIS.

We have been using the Estimated Creatinine Clearance (ECC) calculation for a number of years. Except for taking into account the adjustment for African Americans, it would seem to be superior because it adjusts for height and weight. Please discuss which would be preferred and why. Is there an accepted adjustment for African Americans in the ECC?
Bradenton, FL

Jay Jones, PhD, DABCC: Adjusting for height and weight is preferable but typically cannot be used directly from the LIS to produce an estGFR. A 6'4" robust professional wrestler and a 5'4" invalid of the same age and gender would produce the same correction factor in the MDRD formula. The serum creatinine would likely be different in these two patients but the *estimated* GFR would be equivocal. EstGFR should not be taken as an absolute lab test result but rather as a useful estimate to perform CKD risk assessment.

What role do you see for Cystatin C?
Scarborough Maine

Jay Jones, PhD, DABCC: Yes, I foresee that more direct markers like Cystatin C will play a role as they become proven in population studies and commercially available. Also other lab tests such as urine microalbumin are already recommended by AKF to assess CKD. Informational calculations and computer algorithms are here to stay but more as an extension of direct lab testing, in my opinion.

What is your opinion about de problem of significant influence of method's differences on the GFR estimation by an equation. Can we use this type of mathematical formula in regard of the problem of standardisation of the creatinine measured by different methods ?
Montréal, Qc, Canada

Jay Jones, PhD, DABCC: Theoretically if we can choose our study populations to be nearly identical in regards to relevant demographics we are studying, we should be able to compare methods. In the case at hand, we should be able to select healthy outpatient populations and compare estGFR histograms with different creatinine methods. Although confounding variables between populations (most relatively minor) will likely appear, nevertheless this would be a good reality check with real field data. After all, the reason this screening is being done is to identify as many patients with occult CKD as practicable with as little burden as possible.

How do you apply estGFR (or data mining) to sub-populations where some of the more common formulae for calculation of GFR are not valid (e.g. renal transplant recipients)? How many data points do you need to ensure a reasonably accurate estimate? Especially if you start data mining sub-populations?
Toronto, Ontario

Jay Jones, PhD, DABCC: Except for age we have not stratified to subpopulations. Since such a large pool of original data is typically available, data mining sub-subpopulations becomes feasible. The nice thing about the data mining approach I describe is that one can even back track a single data point in a field of 100,000. Discordance analysis ("finding a needle in a haystack") becomes practicable.

1.Many hospital pharmacies have been using the Cockcroft-Gault equation for estimating GFR to help with initial drug dosing. Do you have any concerns regarding having different calculations present in different services and possibly available in the patient's chart? 2. - I am aware the pediatricians have varying preferences for different calculations. How does the MDRD equation fit into this and do you use different equations for the adult and pediatric population?
Bangor ME

Jay Jones, PhD, DABCC: We educate providers not to use the *estimated* GFR to calculate drug doasages. This is another reason not to report estGFR >60 (but rather just as "estGFR >60 mL/min").

Where can I find information regarding clinical use of the chronic kidney disease guidelines?
Washington, DC

AACC Moderator: For information regarding clinical use of the K/DOQI Clinical Practice Guidelines for Chronic Kidney Disease: Evaluation, Classification, and Stratification, you can view the guideline on the National Kidney Foundation web site (

Some variation exists in serum creatinine measurement depending on the laboratory method. How significant do you think this variation will be in the estimation of creatinine clearance and did you make an effort to baseline your lab's creatinine against a "gold standard" method?
Bangor , ME

Jay Jones, PhD, DABCC: We have always used enzymatic creatinine in our clinical enterprise and clinical protocol (e.g. kidney biopsies) have adapted to these somewhat lower reference ranges over the years. We did not baseline our method other than to compare our method to peers on CAP proficiency surveys.

Instead of exporting my data to an excel, can I keep my information in an access table and get similar results?
Washington, DC

Jay Jones, PhD, DABCC: I have not tried to create histograms and statistics directly from MS Access, but if it could be done it would simplify the data mining process.

- Considering the experience obtained, would you recommend reporting all GFR above 60 or above 90 as such and not as actual result? - Is there any effect of this change on subsequent laboratory testing, for instance on protein c.q. albumin in urin?
Erasmus MC, Rotterdam, The Netherlands

Jay Jones, PhD, DABCC: We report all estGFR above 60 mL/min as ">60 mL/min" with the same listed as reference range. We are hoping that subsequent testing for urine microalbumin/creatinine ratio increases for those estGFR <60 mL/min. At some point in the future we plan to data mine aggregated data from these patients identified at risk for CKD.

How do you handle the problem of the difference in serum creatinine values obtained by enzymatic, compensated and older Jaffe methods? Afula, Israel

Jay Jones, PhD, DABCC: We have standardized our clinical enterprise to a single enzymatic creatinine method run on multiple "harmonized" instuments in a single LIS on a wide area network.

How many institutions are using GFR and how was it received by providers?
Holland, MI

Jay Jones, PhD, DABCC: From what I can tell, a minority of labs (probably less than 10%) have implemented the estGFR. Of those who have, they tend to be larger institutions (e.g. Quest Labs, some Canadian provinces, our regional system). So far provider acceptance, senior medical management buy in, and Nephrology support have been very accepatable in our 8 months of experience.

In Canada we report creatinine is SI units (umols/L). In a recent study involving 450 clinical laboratories in Canada, we found a significant degree of variation in reported creatinine results between analyzers (124 to 94 umoles/L on a human serum sample with a true creatinine result of 100 umols/L (as set by ID/GC.MS). The extent of this variation is of sufficient magnitude to adversely impact the estimation of GFR resul;ting in the miss-classification of patients. Have you standardized the measurement of creatinine across platforms within your network in conjunction with your routine reporting of GFR?
Vancouver Canada

Jay Jones, PhD, DABCC: Yes, we use the same enzymatic creatinine from the same vendor on three instrument platforms (also from the same vendor) across our networked regional enterprise. We also share the same lot number of daily QC material to monitor ongoing concordance. It appears you have done sound foundation building with your 450 lab study. Kudos to you.

1. Are only 2 calculations needed for GFREST, 1 for male and 1 for female? Also due to not entering race as part of the calculation is the final result X 1.21? 2.How is order and calculation handled when Creatinine is part of a panel, ie BMP, Comp Metabolic, Renal etc. 3. Are any values for the calculation ie age, race, gender, or creatinine entered manually in the LIS or is entire calculation done automatically by LIS using age, gender, and race obtained from HIS? Thank you.
Williamsport, Pa.

Jay Jones, PhD, DABCC: We make GFREST an individual battery member of all Basic Metabolic Panels (BMP) and other panels. All values exist in the LIS as part of the lab accession that are necessary to calculate GFREST. No individual values are manually entered so this was an entirely seamless implementation to the lab. All lab results pass from the LIS to the EMR so providers just stated seeing the new estGFR result nested with the rest of the BMP on the date we promised.

Do you recommend reporting out the eGFR equation so that the eGFR result ages with the patient or should it "freeze" the result for that date of the draw?
Buffalo, NY, USA

Jay Jones, PhD, DABCC: The age of the patient is determined as of the collection date. It is "frozen" once the result is verified in the LIS and passed to the EMR. If a patient has 2 creatinines drawn a day before and a day after their birthday and the creatinine is the same, they will have 2 slightly different results. So far this has not created provider concern.

Page Access: