National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)
National Kidney Disease Education Program (NKDEP)
Laboratory Working Group (LWG) Meeting
AACC Annual Meeting – Atlanta, GA
A joint meeting with IFCC Working Groups for Standardization of Albumin in Urine (WG-SAU) and for Glomerular Filtration Rate Assessment (WG-GFRA)
July 28, 2011
8:00 a.m. to 11:00 a.m.
Participants: Greg Miller, Lori Bachmann, Diana Blanco, David Bruns, Glenn Carlisle, Joris Delanghe, Anne Dawnay, John Eckfeldt, Mary Lou Gantzer, Neil Greenberg, Yoshihisa Itoh, Hans-Joachim Kytzia, Jack Levine, John Lieske, Dave McDiarmic, Anjana Nair, Andrew Narva, Eileen Newman, Karen Phinney, Mary Robinson, Murray Rosenthal, Scott Isbell, Miriam Stella, Dave Torrens, Jack Zakowski
Summary of Action Items:
Welcome and Introductions
- Andrew Narva will write a draft educational update on estimating equations for GFR, circulate the draft for comment, and schedule a conference call, if needed, for any follow up discussion. Subsequently, the update will be posted to the LWG web page.
- Neil Greenberg will appoint a task force to develop draft recommendations regarding creatinine method specificity. Schedule a conference call to review the draft recommendations. Be prepared to approve the recommendations at the July 2012 meeting.
- Lori Bachmann will continue data analysis for the harmonization assessment of urine albumin procedures and included assessment of reference materials from JSCC and IRMM.
- John Lieske, David Bunk and Karen Phinney will continue to develop IDMS candidate reference measurement procedures for urine albumin.
- Karen Phinney and David Bunk will continue development of reference materials for creatinine and albumin in urine.
Greg Miller (Chair, LWG), Virginia Commonwealth University
The NKDEP Laboratory Working Group meeting is held in collaboration with two IFCC working groups: 1) Glomerular Filtration Rate Assessment chaired by Neil Greenberg; and 2) Standardization of Albumin in Urine co-chaired by Greg Miller as a joint committee between the organizations.
A. Narva announced that Nancy Accetta, who has supported the Lab Working Group since 2005, is leaving to pursue her passion in a health-related area. He thanked her for the great work she has done for the working group. Taryn Dorsey will assume Nancy's responsibilities.
Status on eGFR Reporting:
CAP asks several questions of their participants each year. Results for 2011 show 84% are reporting eGFR suggesting that educational efforts on reporting eGFR have been very successful. The status of reporting practices: 82% report eGFR with all creatinines, 11% report when requested, 1% report only for outpatients, and 6% report eGFR in other situations. There were 39% still using the original MDRD equation. Therefore, more education is needed since all major manufacturers now provide assays with calibration traceable to IDMS. Only 4% used the CKD-EPI equation although 30% were reporting numeric eGFR values above 60 mL/min/1,73 m2 suggesting that many are extending the MDRD equation beyond its appropriate limits. Thus, more education is also needed regarding the limitations of different equations. The remaining 70% of the respondents reported ">60 mL/min/1,73 m2" when the actual numerical value was above 60 mL/min/1,73 m2.
CKD-EPI Equation and Other Newer Equations in Development
Andrew Narva, NKDEP Director
It is a major step forward that virtually all clinicians understand that we need a better measure than creatinine to assess kidney function. There are other equations that use markers other than creatinine and other ways to assess risk. In NKDEP, we have put strong emphasis on both eGFR and urine albumin in the assessment of chronic kidney disease. We want to address broader challenges as we move forward and develop criteria by which newly developed equations can be considered acceptable. Rather than endorse a single equation, NKDEP will list those equations that can be used and under what circumstances a specific equation is useful. As Greg's earlier comments about the inappropriate use of the MDRD study equation indicated, we need to develop language to help non-clinical chemists understand the limits of estimating equations. It is also important to coordinate an approach to kidney disease across the key federal agencies (NIH, CDC, CMS) with a single way of assessing the burden of chronic kidney disease. We need to find better ways to bring the community up to speed with regard to appropriate uses of the equations, and this is not a simple task.
Josef Coresh, Johns Hopkins University
This group has made major strides in stabilizing the bias in creatinine from +/- 20% to less than 5%. Also, equal emphasis on eGFR and albuminurea is very good and consistent with the latest data. The new CKD-EPI equation was published in 2009 and new data continues to be published on its use. In order to measure improvement, two approaches should be emphasized. First, it is important to determine the accuracy of estimating equations to the measured GFR. The CKD-EPI equation cuts the bias by half and improves the imprecision some. Since the equation was published, eight of 10 studies published report finding the CKD-EPI equation to be better than MDRD, one study reported equal results between the two equations and one study in a transplant population with steroid use found the MDRD equation to give better results. Secondly, risk prediction by eGFR categories is a second way to measure improvement using a new equation. Five studies (ARIC, AusDiab, KEEP, NHANES and AKDN) all find CKD-EPI better in terms of classifying incidence of CKD. This is because the CKD-EPI equation raises the GFR for younger people and women, thus decreasing the risk prediction for CKD for people in these two groups.
One requirement for deciding if one equation is superior to another is to require reporting eGFR standardized for body surface area (eGFR mL/min/1.73m2) and that the equation rely on a standardized marker such as creatinine. Cystatin C is close to being standardized. A new equation is good if it improves the accuracy in estimating the measured GFR using a statistical tool such as the P30, which is the percentage of estimates that are within 30% of the measured GFR. The P30 is simple and integrates bias and imprecision, but there are many statistical tools that can be used. It is difficult to define how much improvement in imprecision and accuracy is enough. Eventually, it has to make a clinical difference, but that is hard to quantify. Using P30, the improvement for MDRD versus CKD-EPI is from P30=80% to P30=84% and this has been replicated in eight of 10 studies. Another criterion to use for determining improvement is that the new equation reclassifies eGFR categories consistent with clinical conditions for the patients reclassified. Clinicians often act on categories and the recommendations have been made for categories. So an improved equation should move a substantial number of people across categories. Looking at eGFR of less than 60 and greater than 60 categories for the NHANES population, the CKD-EPI equation moves three million people (1.5% of the US adult population) from just below 60 to above 60. "Positive net reclassification improvement" is a statistic borrowed from the CRP and cardiovascular literature. Three published and two presented studies have shown that CKD-EPI is a better equation than MDRD based on reclassification of people into lower risk categories. The CKD-EPI equation was developed using a more diverse group and broader GFR range than was used for MDRD. The ideal criteria for determining superiority of a new equation should be a clinical trial and a cost/benefit analysis.
Quantifying the difference between MDRD and CKD-EPI shows that for younger people the eGFR is increased by approximately 5-7 mL; for over 80 years of age, the eGFR using the CKD-EPI equation is slightly lower. Based on creatinine concentration, the CKD-EPI equation flattens for men with creatinine less than 0.9 mg/dL and for women with creatinine less than 0.7 mg/dL while the MDRD equation has poor performance at concentrations below 0.9 mg/dL. When evaluating the effect of imprecision in the creatinine assay on the imprecision of the eGFR using both equations, the CKD-EPI spline reduces the influence of creatinine imprecision at lower concentrations of creatinine while the MDRD equation is more substantially affected. Reclassification using unpublished NHANES data shows that 25% of those classified as 45-59 by MDRD would be higher using CKD-EPI and only 1.6% of those in the 60-89 group would be lower. In terms of risk relationships using the ARIC population, and eGFR by MDRD of 30-59, CKD-EPI categorizes 151 people (mostly younger and female) at a higher eGFR with lower risk of ESRD (10-fold), mortality (three-fold), CHD (three-fold), and stroke (two-fold). Another study in Canada of 920,000 subjects showed that the CKD-EPI reclassified 36,000 people (3.9%) from 45-59 to 60-89 with a 6-fold lower risk of death and three-fold lower risk of AMI and ESRD. There were some classified to the lower category of 30-44 by CKD-EPI, but they were on average 85 years old so their risk of death and AMI is higher due to their age.
Desmond Williams, Centers for Disease Control and Prevention
The ideal measure for kidney function should have these characteristics: be simple and inexpensive; be valid, reliable, and reproducible across the spectrum of values; be able to accurately delineate disease status or risk, possess prognostic significance in terms of progression to kidney failure or development of complications and premature death, and be applicable to broad and diverse populations in different settings. Measured GFR is the gold standard, but it is expensive, cumbersome, and time consuming. Serum creatinine has reliability and reproducibility issues, with coefficients of variation in the range of 0.9-4.3%, and high inter-individual variation. Creatinine based estimating equations are good, but they are still estimations and although the assays have been standardized through the efforts of this working group, there is still error in the measurement. Therefore all eGFR values have an accompanying error and range; they are dependent on demographic variables and coefficients, which when unknown introduce the potential for additional errors. Some equations adjust for muscle mass and some do not. Cystatin C based equations are under development but assays are still being standardized. When looking at the models used in the development of the original MDRD equation, all have relatively large confidence intervals. While the models used in developing the CKD-EPI equation have lower confidence intervals, they are still present. These ranges for the estimating equations probably obliterate any differences that have been seen so far.
Significant resources have gone into the MDRD equation and many other equations are under development. It is not possible to change equations without a good reason. So we need to have guidelines for determining when a new standard is appropriate for us to adopt. Qualifications needed include: 1) a statistically significant improvement in the delineation of disease status and risk compared to the gold standard; 2) significant improvement in the prognostic value for progression to ESRD and to development of complications and premature death using outcomes related to kidney disease (and not all-cause mortality as was done with MDRD and CKD-EPI); 3) reliability and reproducibility of the new standard; and 4) other factors such as simplification (CKD-EPI is more complicated because of the spline), less expensive, and addresses some of the disadvantages of the current standard.
A case study using a paper (Madder R D et al. Circ Cardiovasc Interv 2011;4:219-225) comparing MDRD and CKD-EPI was presented. Correlation with single 125I-iothalamate GFR measurement versus the estimating equations showed a similar correlation coefficient of 0.83 for MDRD and 0.85 for CKD-EPI. When comparing the percent change in serial mGFR to eGFR for each equation, the correlation coefficient was the same for both equations and more importantly, was very weak. The conclusion from this is that irrespective of which equation is used, the result will be the same as long as the same equation is used from one time point to the next. Estimates of GFR greater than 60 are problematic. So this is the best area to evaluate the performance of the equations compared to mGFR. Looking at the performance of each equation at 20% decrease and 20% increase from 60 as compared to mGFR shows that both equations perform equally (AUCs of 65% for both equations when a 20% decrease is used and 63% for both equations when a 20% increase is used) and both are lacking. Interestingly, when the CKD-EPI equation was published (Levey AS et al. Ann Intern Med. 2009;150:604-612), the AUC was calculated for both MDRD and CKD-EPI as compared to mGFR, and the AUC was 0.96 for both equations. So, in comparing the MDRD and CKD-EPI equations based on the list of qualifications, there is no convincing evidence that we need to adopt a new equation: 1) both equations are similar with respect to improvement in delineation of disease as compared to the gold standard; 2) because all-cause mortality rather than a kidney-related outcome was used by studies to evaluate the equations for progression to ESRD and progression to complications and premature death, this qualification has not yet been adequately evaluated for either equation; 3) both equations have similar issues regarding reliability and reproducibility because both equations use the same variables; 4) MDRD is a simpler equation than CKD-EPI; 5) expense of converting to CKD-EPI is unknown; and 6) CKD-EPI does not address the disadvantages of MDRD and is not yet tested in all populations as is MDRD.
Joe Coresh: Agrees with 90% of the data. Error is important and measured GFR is the gold standard because it has no bias. Limitations of mGFR have been published and in practice, mGFR has a CV of about 13% which means that a change would have a CV of the square root of two times 13% which is 18%. Any study that looks at change in less than one year is basically looking at noise. AUCs are about discrimination and not about absolute bias. The biggest improvement of CKD-EPI versus MDRD is that it improves the bias. The precision is improved by the lab, the equation improves the bias and then it is a matter of caring about people moving between categories.
A. Narva: eGFR by itself is not a good a way of predicting risk for progression or identifying clinical triggers for intervention. NKDEP in all of its literature does not use CKD stages as defined by eGFR. It is reasonable for the NKDEP web site to list both equations as usable. The major challenge is to do a better job of helping clinicians understand the real limitations of creatinine measurement and these estimating equations. Everything about urine albumin is going to be more difficult than creatinine and eGFR because clinicians have a poorer understanding and the technical issues are harder and that challenge is still ahead.
John Eckfeldt: There is not a need to explain the equations to clinicians; it is only necessary to state that it is a prediction equation. Also, it is a cost/benefit decision and the cost to convert to a new equation is in the laboratory information system (LIS) cost to program the changes. Based on the discussions, the CKD-EPI equation seems to be a slight improvement and it makes sense to make the change if it is easy to do in the LIS. In contrast to Desmond's summary slide, CKD-EPI does seem to statistically improve the prediction of ESRD and death.
Desmond Williams: There is not a statistically significant difference in MDRD versus CKD-EPI. The studies use all-cause mortality for an outcome which makes them flawed. The cost is not just on the laboratory, but also it is on the public health system because once changed, all previous eGFR results using the old equation will have to be re-calculated to compare to the results using the new equation. Also, given that some will have higher results and some will have lower results with CKD-EPI, there is a need to interpret based on the old equation. The benefits seen so far do not justify the cost issue.
Greg Miller: From the laboratory practice side, many labs are requested to report numbers above 60 for a variety of reasons. If numbers above 60 are reported, probably both presenters will agree that the CKD-EPI equation is better for that purpose. Another issue that needs to be explained to laboratorians is the creatinine imprecision influence on CKD-EPI uncertainty at higher eGFR values. A laboratory using a method with a higher creatinine CV will have a larger and possibly unacceptable imprecision in the eGFR at higher values. Andrew Narva will develop an educational statement on using estimating equations which will be circulated by email for comments. If necessary, we will have a conference call to discuss the updated educational statement to go on the web site.
Serum Creatinine Method Specificity
Neil Greenberg, Chair, IFCC WG-GFRA
This is a report on the outcome of the study to evaluate specificity of several commonly used commercial creatinine methods in routine service laboratories. Contributors to this study include Greg Miller, Lori Bachmann, Bill Roberts, Libby Wright, Jack Zakowski, and Neil Dalton. While much progress has been made on standardization and calibration of creatinine, that has no impact on the specificity of the methods. The focus of this study was to look at individual patient samples (N=365) across an array of 19 disease categories. Also, there were some volatile or labile substances that we could not collect in individual stored sera, so sera at 1.0 and 3.8 mg/dL creatinine were supplemented with these substances: acetoacetate, acetone, ascorbate, and pyruvate. Creatinine in the samples was measured by four enzymatic and three Jaffe commercial procedures, plus a liquid chromatography isotope dilution tandem mass spectrometry measurement procedure.
For the control samples, 98% of results were within ±0.10 mg/dL for both enzymatic and Jaffe methods. Thus, a clinically meaningful bias in terms of influence of an interfering substance was defined as exceeding ±0.10 mg/dL (or 10% for values greater than 1.0 mg/dL). The number of instances when three or more results out of approximately 20 individual samples in a disease category had biases greater than the criteria was 49% for Jaffe and 18% for enzymatic procedures. For the aggregate group of 59 diabetes samples (elevated ß hydroxybutyrate, glucose or HbA1c), enzymatic procedures had one biased instance compared to seven instances for the Jaffe procedures. For supplemented sera (acetoacetate, acetone, ascorbate or pyruvate), interferences occurred in 46% of the samples for Jaffe vs. 25% of the samples for enzymatic procedures. Interferences were highly procedure dependent and were different at low or high creatinine concentrations. The results lead to these conclusions: 1) There were differences in both magnitude and direction of bias among measurement procedures, whether enzymatic or Jaffe; 2) The influence of interfering substances was less frequent with the enzymatic procedures, but no procedure was unaffected; and 3) the details of the specific implementation of a method principle (Jaffe and/or enzymatic) influenced its susceptibility to interfering substances.
Limitations of the Study:
1) We were unable to include all manufacturers because of the limited volumes available as residual samples from clinical laboratories, given standard blood collection practices; 2) The clinical samples were selected to have a high probability of containing various interfering substances. However, the identity and concentrations of the substances responsible for a given interference were either unknown or only partially known based on the selection parameters; 3) The number of samples included in each clinical category was relatively small (n ~ 20), and in some cases, different samples from the same individual were included more than once in a clinical group; and 4) Clinical samples were not handled uniformly before aliquotting and freezing with variable time spent at room, refrigerated or frozen temperatures with possible metabolic changes or loss of labile components (e.g., dopamine and dobutamine).
1) Interfering substances should be identified in all creatinine commercial methods' product labeling whenever they cause a bias that exceeds +/- 0.1 mg/dL (+/- 8.84 µmol/L) or +/- 10%; 2) Manufacturers should ensure that specificity/interference studies (and labeled claims) evaluate and report on impact of potential interfering substances at both "low" and "high" creatinine concentrations; 3) Follow up studies should be performed to collect more complete data on at-risk subjects (e.g., early and pre-diabetic children) with "pediatric" concentrations of serum creatinine (<0.6 mg/dL), especially with regard to protein/albumin concentrations below "normal" ranges; and 4) To confirm findings from this study for the "diabetic" patient sample group, we propose undertaking studies of archived epidemiologic study samples (e.g., frozen CKD-EPI sample aliquots; N > 1000) using selected sub-populations (e.g., known diabetics), comparing creatinine results among several commercial Jaffe and enzymatic methods (and an IDMS DCMP), to confirm the findings for the "diabetic" patient sample group in the present study.
A manuscript describing these results is in review at Clinical Chemistry with the hope of publication by the end of the year.
J. Coresh: Inquired about the cost for a laboratory to change from a Jaffe to an enzymatic method. John Eckfedlt estimated the cost for enzymatic is about $1 more. David Bruns said it would cost his laboratory about $100,000/yr and Christa Colbbaert reported that there was no change in cost in her laboratory based on negotiation with the manufacturer. Graham Jones estimated $0.20/test but cautioned that one should look at clinical decision making to know if the increase is cost-effective. Neil Greenberg added that, for example, elevated beta-hydroxybutarate can cause an elevation of creatinine for Jaffe methods, thus lowering the eGFR. If patients are mis-classified they may be more likely to receive follow up investigation and treatment that increases the total cost of care.
Joris Delagne: Commented that creatinine will be lowered in muscle wasting diseases resulting in false increases in eGFR. He suggested that labs could measure Cystatin C for use in special cases like high bilirubin and ketones; then, the lab would not have to change to a more expensive enzymatic creatinine for all patients. Anne Dawnay commented that acute kidney injury is probably over-diagnosed in patients in ICUs and post cardio-pulmonary bypass because they have all sorts of potential interfering substances with the Jaffe assays.
G. Jones: Reported between laboratory bias is important and evidence from the Netherlands EQA showed there is less bias between labs with enzymatic methods than Jaffe methods; A. Dawnay agreed that this has been found to be true in the UK also. Because things are moving to models of integrated care, it is important to have a creatinine result that is transferable across the health care settings.
Jim Fleming: Stated that there is no profit in clinical chemistry testing. Lab Corp runs 25 million creatinine assays per year and it would cost them about $8 million to convert to an enzymatic method. They do offer enzymatic assays under certain circumstances.
A. Dawnay: At her institution, they did a side-by-side comparison of Jaffe and enzymatic methods in 5,000 nephrology patient samples and found that 25% exceeded total allowable error limits.
David Seccombe: Stated that in resource-limited countries, they use Jaffe assays. Also, there is still work to be done on method performance at the pediatric low concentrations of creatinine. They see a 30% failure rate in PT results at the lower concentrations as compared to a 6% failure rate at higher concentrations.
J. Eckfeldt: Felt that the idea of a diabetic study is interesting but cautioned that it needs to be done at a baseline state; eGFR should not be done in a state of ketoacidosis.
G. Miller: Now that the data for creatinine specificity influences has been gathered, there is a need to have conference calls to discuss recommendations so that we can provide education and advice to laboratory users on choice of methods. It is not easy to make recommendations. It is possible that a small task force will be needed to develop education about selecting methods and the impact on clinical decisions.
Urine Albumin Method Harmonization Assessment
Lori Bachmann Virginia Commonwealth University
The objectives of this project were to: 1) Evaluate harmonization among methods for urine albumin using native human urine samples; 2) assess agreement of routine methods with the IDMS cRMP; 3) Evaluate commutability characteristics of JSCC cRM and diluted ERM-DA470k/IFCC reference materials for urine albumin; 4) Evaluate basic performance characteristics of the urine albumin methods; and 5) Compare harmonization of urine total protein and creatinine methods. Seventeen methods (16 quantitative) from five manufacturers were evaluated. Fresh, non-frozen human urine samples (343) that had been submitted for routine urine albumin measurement were used, and non-frozen aliquots were measured by 17 routine methods. Frozen aliquots of each sample were measured by IDMS at the Mayo Clinic and 41 fresh and frozen paired samples were also assayed to assess freeze-thaw effects. JSCC cRM and diluted ERM DA470k/IFCC RM were prepared by each manufacturer and included in each analytical run along with patient samples. Centrally prepared IRMM reference materials and pooled urine QC materials were also included. To evaluate the effect of different matrices on the methods, urine samples were spiked with potential interfering substances or adjustments were made to the pH and ionic strength. A dipstick urinalysis and urine electrolytes was performed to characterize the matrix of each sample.
Comparison of method medians over the range of 12-30 mg/L, the LCMS median was 21mg/L as was the average routine method median. The range of the differences of medians was 45% or 9 mg/L. For the clinically relevant range of 20 - 300 mg/L, again the median result for LCMS (75 mg/L) and the average median for the routine methods (79 mg/L) were similar. The range of differences of medians was 38% or 30 mg/L. Approximately the same results were seen for the high concentration group of samples. Overall, across concentrations, there was an average difference of 40% between the medians for the methods. While this is not terrible, there is room for improvement with standardization. Observations made when plotting each method versus the LCMS method include: some methods compared very well and others had more scatter and/or bias; there were some LCMS method-specific outliers and some routine method-specific outliers and the reasons for these is still under investigation; centrally prepared diluted ERM-DA470k/IFCC materials tracked closely with patient samples for 11 of 16 commercial methods.
The overall agreement among all methods appears promising with method medians differing by approximately 40%; multiple methods exhibited agreement with the IDMS procedure and a few methods appear to have biases vs. IDMS; method specificity issues may be present for select samples, but specificity does not appear to be a major issue; diluted ERM-DA470k/IFCC reference materials appear to be commutable with patient samples for most methods. The next steps are: complete the formal statistical analysis including error assessments and imprecision; evaluate the freeze-thaw effects and potential interfering substances; assess commutability characteristics of JSCC candidate RM and diluted IRMM ERM-DA470k/IFCC; finish the data analysis for creatinine and total protein.
Ingrid Zegers: Asked if the data for the reference material was from centrally prepared samples and L. Bachmann confirmed that it was. It is likely that data for materials prepared by the manufacturer will have more variation because they were given the option of using their own dilution buffer or saline buffer for making the dilution and this would result in some matrix effects.
G. Jones: Commented that a quality standard for bias is needed. The approach taken in Australia is to have a quality standard around the decision point of +/- 10% and that is taken for the percentage of samples that would be misclassified on the expected distribution of samples. Therefore, quickly eyeballing this data for the number of methods within +/- 10%, it looks like about 80% of the methods meet that criterion. L. Bachmann agreed and stated that the worst method was 15-20% different than the IDMS procedure. G. Jones also commented that the scatter around the line for at least one of the methods was greater than would be expected from just imprecision. So there will be some non-specificity even in those results that are clustered relatively close to the line. L. Bachmann stated that this is being investigated.
A. Dawnay: Commented that a study they published (ACB 1989;26:560-2) had an outlier similar to that seen with this data. They found that it related to the amount of sediment. The problem went away if the sample analyzed was well mixed and included the sediment. The outlier was seen when the sediment settled out and was not included in the sample analyzed. L. Bachmann explained that all the samples were centrifuged before they were aliquoted. However, it is possible that the samples continued to precipitate during shipping and manufacturers were asked to centrifuge samples that looked cloudy upon receipt.
Mary Robinson: Asked if the pH was measured in the samples and L. Bachmann replied that a strip urinalysis was done on all of the samples, so they will have information for those that had pH extremes. Another question was asked about total protein and L. Bachmann replied that there is semi-quantitative results from the strips, but also quantitative protein was assayed on all of the platforms that had that method available. Protein data has not yet been analyzed.
Serum Albumin BCG vs. BCP Standardization Issues
David Bruns, University of Virginia
Low serum albumin is a strong predictor of mortality in renal dialysis patients. National guidelines (KDOQI) call for monitoring of serum albumin as a quality indicator for dialysis centers. Reimbursements could be affected by how many patients meet/do not meet the target values. Target values (BCG assay): 3.5 g/dL and (better) 4.0 g/dL. Unfortunately BCG assays are not all equal. A change in assay procedure in 2005 resulted in many more patients meeting the target levels. Prior to the method change, 22% met the goal of 4.0 g/dL; after the change, 74% met that goal. Another method change in 2009 caused a change from 60% of patients meeting the 4.0 target to 17%. This effect is not seen in only dialysis patients because all patient weekly means dropped after the change in 2009. The results of a sample exchange comparing the Architect (UVA) versus the Vitros assay used at WashU and a Siemens assay used at VCU suggests that there is a problem with BCG assays.
G. Miller, Stan Lo and Basil Doumas are preparing a report of an analysis in 2003 of CAP fresh frozen pooled serum comparing peer group mean values versus the designated comparison method and the results varied from +0.3 to -0.3. In addition, a survey compiled by B. Doumas of reference values for serum/plasma albumin by method shows that the ranges are extremely varied.
Five-step Proposal for How to Proceed:
1) Assess agreement among methods; 2) Depending on results in step one, may need a commutable reference preparation for dye-binding assays; 3) Assess commutability of new material; 4) Make material available; and 5) Reassess agreement among methods. Considerations for step one: check sera and heparin plasmas (or pools), use samples from patients including those (a) with low, middle, and high albumin concentrations, (b) on hemodialysis, (c) on peritoneal dialysis, (d) with high creatinine but not on dialysis, (e) with liver disease (or increased ALT for inclusion), (f) receiving drugs that bind to albumin (inclusion criteria: drug was measured in lab). Also include a pure human albumin solution (Ortho Diagnostics may have a suitable preparation), ERM-DA470k/IFCC (may be used to calibrate all or some assays), and include admixtures of a high concentration pool of serum with DA470k/IFCC.
A. Narva: Suggested that the majority of physicians are not aware of the serum albumin assay differences. Perhaps only nephrologists are aware that there are two types of assays available. There is no ESRD measure that uses serum albumin. Clinically, it is not a good nutritional marker, it is a good marker for inflammation. There is a need for the laboratory community to respond to this issue and albumin is measured for a wide range of clinical conditions. However, it is not a high priority for NKDEP LWG to take this on, but we could provide moral support.
Yoshi Itoh: In Japan, a new improved BCP method has been developed. Before measuring samples, albumin is reduced by ethanol.
A. Dawnay: In the UK, the Renal Registry monitors dialysis patients and has abandoned the idea of comparing albumin due to method differences. However, there are targets for achieving adjusted calcium in dialysis patients and the labs are not using the right adjusting equation for the albumin method used. A pilot project is starting to work with the labs.
David Bruns: This is not just an issue of BCP vs. BCG, but within BCG assays, the equation can be different.
N. Greenberg: Defining the ERM DA470k/IFCC with respect to commutability is interesting because there is a problem with albumin reference materials due to contamination with dimerized albumin. Calibration error is related to this reference material issue.
Ingrid Zegers: Suggested that the ERM-DA470k/IFCC shows commutability with many of the methods and the problem may be the way the material is being used to assign values to in-house calibrators. D. Bruns commented that they have measured the material on the same assay that appears to give lower patient results, and they get the target value for the reference material.
G. Jones: Asked the manufacturers if there is an equivalent body in the hepatic world where manufacturers, clinicians, and laboratorians are meeting to discuss this type of problem. There is definitely a clinical need.
G. Miller: Does not know of anyone who can take on this project, but it is a project worth doing. However, it cannot be done within the context of the LWG due to other commitments.
Urine Albumin Reference Measurement Procedure
Karen Phinney, National Institute of Standards and Technology and John Lieske, Mayo Clinic
NIST has prepared SRM 3667 (Creatinine in Human Urine), which is a single concentration of pooled urine, collected from healthy males/females (minimum of 10 donors). There are 1,000 bottles with 10 mL urine in each. They will perform isotope dilution LC-MS reference measurement procedure to quantify the creatinine. Certification measurements are expected to be finished by fall 2011. In addition, SRM 2925 (Human Serum Albumin Solution), which is a primary certified reference material for use with reference measurement procedures for albumin in urine, has been prepared. It is 1 g/L aqueous solution of recombinant human serum albumin with 1 mL per vial. Quantification by amino acid analysis using isotope dilution LC-MS/MS should be completed by fall 2011. The NIST developed LC-MS/MS reference measurement procedure for urine albumin measures eight tryptic peptides of human serum albumin with ~1% CV for replicates. They will validate this procedure by comparison with Mayo's method.
The method uses five tryptic peptides that are consistently seen by LCMS and averages the results for a final value. Inter-assay precision over 20 days using several samples representing a range of albumin concentrations is about 1-2%, and intra-assay precision for a low, medium, and high concentration control is also 1-2%. In an intra-lab comparison study in which 340 patient samples, 38 spiked samples, 31 IRMM reference material samples, 24 Japanese reference material samples, and 20 serum QC samples were assayed and compared to a single immunoassay result, there was only a -3% bias. Looking at comparison of samples in the clinically relevant range less than 250 mg/L, the correlation is very good with a slightly lower bias for patient samples. A study looking at influence on precision found with increasing the number of peptides used showed that there is improved precision when using two peptides, but there was not an increase in precision seen when three, four or five peptides were used to determine a final value. Also, the bias between the different peptides was studied. It appears that there is a subtle but definite bias seen with the C-terminus and N-terminus peptides yielding slightly lower results than the peptide in the middle of the molecule. Approximately 1% of the samples in the urine albumin methods harmonization assessment study had loss of C or N terminal peptides.
1) Value assign a primary reference material based in a urine matrix; 2) Perform a comparison study between the Mayo and NIST methods (n=five patient samples at various concentrations and using the same calibrators); 3) Publish the Mayo and NIST methods; and 4) Obtain Joint Committee on Traceability in Laboratory Medicine (JCTLM) listing.
L. Bachmann: Asked if the bias over the five peptides was done in clinical samples and if so, does the data shown mean that only 1% of patient samples are missing the N and/or C terminus and the others are the full molecule. John Lieske replied that the 340 patient samples were used and that he agrees with L. Bachmann's interpretation.
J. Eckfeldt: Inquired if the outliers have any specific issue like degradation and J. Lieske replied that if one of the peptides was off, the others were also off for a given sample.
J. Coresh: Commented that it is important to help clinicians understand progression for both eGFR and albuminurea. At this time, a couple of groups think that ~50% increase in albuminurea represents a real clinical increase that is beyond the lab error. How does this relate to the Mayo measurements? J. Lieske feels that there is not a lot of difference between his method and others; per L. Bachmann's report, the difference is about 40% across methods. J. Coresh acknowledged that there is a lot of physiological variability which should be dominant. L. Bachmann commented that 40% was the extreme and that the LCMS method ran at the mean of all methods. J. Coresh then asked about the precision of the methods and L. Bachmann stated that the data has not been evaluated yet.
J. Eckfeldt: Commented that looking at clinical changes in the presence of a methodological change can appear as a real change when in fact it is a bias issue.
A. Narva: Stated that part of the effort with regard to urine albumin includes recommendations on reporting and cut-offs. That will have to span more than kidney disease because urine albumin is also looked at in cardiovascular disease. J. Coresh commented that it is important to coordinate across groups. The gold standard is 24 hour urine albumin, but the simple approach is using the surrogate spot urine and cut-offs of 30 and 300 mg/g Cr. A. Narva stated that there are other ideas and it will probably be based more on a value that is a continuous risk factor. The decision regarding this will be made in consultation with the rest of the medical community.
G. Miller: Asked how much time before the method will be ready to submit to JCTLM. J. Lieske answered that the major hurdle is the NIST comparison study. K. Phinney commented that nominations to JCTLM happen only once per year around April, and a procedure cannot be nominated until it has been published.
JSCC Urine Albumin Reference Material
Yoshi Itoh, Asahikawa Medical University
The candidate urine albumin reference material is a buffer based purified monomeric human serum albumin in lyophilized form with an inter-vial difference of 1.31%. It is stable at 4-10 ºC for up to 10 hours and -80 ºC for 40 months. The assigned value (mean ± U) made by value transfer from ERM-DA470 is 225.1± 9.11 mg/L when reconstituted in 3 mL of purified water. This material is defined as a "working reference material" in the traceability chain for urine albumin. A commutability study using 129 urine samples was performed with seven different methods. Using the manufacturers' calibrators, the between method CV was 8.8%; using the new reference material to calibrate the assays, the between method CV was 6.6%. NEDO (New Energy and Industrial Technology Development Organization) in collaboration with RECS (Reference Material Institute for Clinical Laboratory Standards) has recently assigned the value for albumin in ERM-DA470 by ID/MS. It has proved to be very close to the existing assigned value.
N. Greenberg: Asked if the reference material is available to manufacturers and Y. Itoh replied that it is not yet available, perhaps after the publication is ready. It has been shipped to the US and China for evaluation of the effectiveness for standardization; JCCM or some other society will decide when it is available.
G. Miller: Reminded the group that this material was included in L. Bachmann's harmonization study and the data has not yet been analyzed.
Other items, next steps
We will plan to organize two conference calls: 1) to address the type of educational materials that are appropriate with regard to the CKD-EPI and other eGFR equations; and 2) to discuss creatinine specificity issues and the type of educational recommendations that can be made based on the available data and what additional investigations are needed regarding the relationship between the enzymatic and Jaffe procedures.
Statements summarizing the non-specificity data and the issues that need resolution need to be prepared and moved forward in the next 6 to 8 months in order to develop recommendations and have discussions by next year's meeting.
A. Narva thanked everyone for their participation and contributions to the LWG and to improving laboratory medicine for patients with kidney disease. This type of collaboration between the clinical lab community and clinical education is a unique and important model. It will be increasingly important to future assays to help physicians understand issues specific to new methods and how to use them.
G. Miller also thanked everyone for volunteering so many hours to make this all happen.
Meeting adjourned at 10:45 am.