NGSP News Archive: Summer 2009 Meetings

Clinical Advisory Committee | NGSP Steering Committee | Manufacturer Forum

2009 NGSP Steering Committee Meeting

American Association for Clinical Chemistry, July 2009

Below is a brief summary of the issues discussed:

1) NGSP Progress Report—Randie Little , NGSP Network Coordinator

  • NGSP Network Monitoring
    • The PRLs and SRLs continue to demonstrate excellent comparability.
    • Monthly between-lab CVs for the NGSP network were all well under 2.5% over the past year.
  • Certification
    • The number of certified methods and laboratories has continued to increase.
    • In addition to the U.S. and Europe there are now a number of NGSP-certified laboratories in Asia, including China, India and Japan, as well as several in South Africa, Australia and South America.
  • Reducing variability (CAP data)
    • The CAP data show much improvement in the comparability of HbA1c results in the field between 1993 and 2009.
    • 2009A CAP survey
      • Bias
        1. Many methods (19/27) showed less than 0.3% mean bias at all three levels.
        2. For the low and mid-range samples (5-6% HbA1c) many laboratories reported results within ± 6% of the target (CAP criteria beginning in 2011).
        3. Overall the bias vs. the NGSP targets among methods at the low and middle HbA1c levels has shown some improvement since 2006.
      • Imprecision
        1. On the 2009A survey method-specific, between-laboratory CVs ranged from 1.2 to 8.0%.
        2. Approximately 95% of labs were using methods that had CVs ≤ 5% at all three levels.
      • Pass Rates
        1. Overall pass rates on the 2009A survey at the current limits of ±10% were 95.2%, 96.6% and 97.1% at HbA1c levels of 5.1, 6.0 and 8.4, respectively.
        2. If the 2011 limits of ±6% are used the pass rates would be 92.1%, 90.5% and 92.9% respectively at these three levels.
    • Critical Difference (CD), or Reference Change Value: CD is the change in a patient’s serial test results that can be considered significantly different.
      • Many clinicians now consider a difference of 0.5%HbA1c to be clinically significant.
      • Assuming a 1% biological variability in HbA1c, the assay CV must be in the range of 2-3% in order for a change of 0.5% HbA1c to be significant (95% probability).
      • If a patient’s specimens may be sent to more than one lab then the all-method CV must be considered.  The current all-method CV based on the CAP survey is about 4.5%; this needs to be improved.
    • Changes in NGSP Manufacturer Certification Criteria:  The NGSP assessment of agreement criteria for the 95% CI of the differences between the method and the NGSP will be further reduced from ±0.85%  to ±0.75% HbA1c and the HbA1c range will be reduced to 4-10% starting January 2010.
  • CAP HbA1c Survey Grading: Changes in Acceptable Limits for Laboratories
    • The CAP began accuracy grading for the GH2 survey in 2007 with initial acceptable limits of ±15% of the NGSP target.  These limits were lowered to ±12% in 2008 and ±10% in 2009, and will be further reduced to ±8% in 2010 and ±6% in 2011.
  • Interferences from Hb Variants
    • 6.5% of labs are using methods with clinically significant HbAS and/or HbAC interference.  This will be reduced to 5% when the original Integra method is completely phased out.
    • At least 20% of labs are using methods with clinically significant HbAE interference, and at least 6% of labs are using methods with clinically significant HbAD interference.
    • Interference from Hb variants is a concern but at the present time most labs are using methods that do not show interference from the most common variants,
    • The NGSP will continue to investigate new assay methods for interference from common variants.

Discussion: The CAP began dividing methods differently on the survey in 2009 to better distinguish the different assay methods; this resulted in some method groups being too small for the CAP to calculate summary statistics.  This was requested by manufacturers and is useful to them.  The CAP will attempt to make the high HbA1c level be approximately 10% on future surveys.  The tightening of the CAP criteria to ±6% by 2011 was an important factor in the decision of the Expert Committee to recommend HbA1c for diagnosing diabetes.  Given the recommendation to use a HbA1c level of 6.5% as the cutoff for diabetes diagnosis, this level is very critical.  Since HbA1c is not a CLIA-regulated analyte, the penalty for failing a CAP survey is less severe than for analytes that are CLIA-regulated.

2) Other NGSP Issues—Randie Little, Curt Rohlfing—NGSP

  • Manufacturer Certification—Randie Little
    • What should be done in cases where a manufacturer submits certification data, fails, then submits new data and passes?
    • It is not always clear what factors caused the initial failure, and we do not always have a good sense of how these issues translate to method performance in the field.

Discussion: Since this situation is infrequent the Committee decided not to address this issue at this time but it will be addressed by the full Committee if the need arises.

  • Monitoring of Network and Level 1 Laboratories—Curt Rohlfing
    • Additional criteria will be used to monitor the SRLs beginning July 2009
      • Based on Konnert A, et al.  J Testing Evaluation 2006;34:1-7.
      • Based on 7 years of monthly SRL data.
      • Derived limits for within laboratory and between laboratory individual replicates as well as overall laboratory slope/intercept.
      • Will be used in addition to the current criteria.
      • Determination of  criteria
        1. Fixed limits (%HbA1c) were set for replicates.
          • >99% of within-SRL replicates differed by ≤0.4% HbA1c
          • >99% of the individual SRL results differed from the medians of all SRLs by ≤0.5% HbA1c.
        2. An acceptance “ellipse” based on the slope and intercepts (representing proportional and systematic bias) of the differences between each SRL and the medians of all SRLs for each specimen was determined.
      • The criteria
        1. The criteria will follow the same basic scheme as the IFCC network criteria.
        2. SRLs that fall outside the acceptance ellipse in a given month will have all results from that month excluded from the “approved mean” calculations.
        3. Any SRL that falls outside the ellipse in two consecutive months will not be allowed to certify laboratories until the issue is resolved.
        4. Any within-SRL replicates that differ by >0.4% HbA1c and any results that differ from the median of the SRLs by >0.5% will be excluded from the final mean calculations.
    • Level I Laboratories
      • Criteria for Level 1 laboratories will remain the same but they will be assessed against the “approved” means of the SRLs rather than CPRL results.
      • The possibility of utilizing criteria based on the acceptance ellipse to evaluate the performance of Level 1 laboratories will be further investigated.

 The committee agreed to the new monitoring criteria.

3)   IFCC Meeting Report—David Sacks: The IFCC Working Group met in April, issues discussed:

  • The master equations (ME) between the IFCC network and the NGSP and Sweden DCMs have remained stable from 2001 thru 2008 but the ME between the IFCC and Japan DCM has shown a shift at the high end.
  • There has been a problem with the IFCC calibrator for the first time in 2008.  The values were ~15% lower than expected.  CW has been working with Roche to determine the cause of this, there is a three-year supply of backup calibrator lots that can be used in the meantime.
  • Global harmonization:
    • Germany will use only IFCC numbers to report HbA1c results by the end of the year.
    • The UK has decided to report NGSP and IFCC numbers until 2011 when only IFCC numbers will be reported.
    • Several other countries including the Netherlands, Italy and New Zealand have decided to go the same way as the UK.
    • At this point it appears that Japan will report JDS and IFCC.
    • Only the U.S. is planning to adopt reporting of eAG (along with NGSP numbers).

Discussion: The shift in the IFCC/Japan DCM relationship is not an NGSP issue; the relationship between the IFCC and NGSP networks has remained stable.  There is concern about different countries deciding to report HbA1c differently but there does not seem to be a way to resolve this.

4)   NGSP Clinical Advisory Committee Report—David Sacks

  • The CAC met in New Orleans in June at the ADA Annual Meeting.
  • The CAC is composed of members of the major clinical diabetes organizations and advises the NGSP on clinical aspects of HbA1c testing.
  • There was general agreement that variability needs to be reduced as much as possible.
  • Matt Petersen of the ADA gave an update on ADA efforts to implement the reporting of eAG.  The ADA has put a lot of time and money into educating providers and patients; they have a web site and have put out flyers and handed out wheels and calculators that allow conversion of HbA1c to eAG.
  • On the last CAP survey labs were asked whether they report eAG and if so which equation they used.
    • 500 (17%) of them report eAG.
    • Of these labs reporting eAG, 200 stated that they report eAG using the equation established by the ADAG study; many of the other laboratories are using the old DCCT equation.
    • The CAP put a notice in the reports sent back to laboratories telling them that the ADAG equation should be used.
    • Since some labs are much larger than others we do not have a good sense of what percentage of results get reported as eAG along with HbA1c, the CAP will try to find out in the next survey later this year.
  • David Nathan presented the new recommendations from the Expert Committee that HbA1c be used for diagnosis.
  • A Perspective will be published in Clinical Chemistry concerning the implications of the new diagnostic criteria (using HbA1c).

5)  HbA1c for Diagnosis—Michael Steffes, NGSP Steering Committee

  • The decision to recommend HbA1c for diagnosis
  • The tightening of the CAP criteria was very important in convincing the committee to recommend the use of HbA1c to diagnose diabetes.
  • Arguments against glucose
    • In some ways glucose may be a better analyte in terms of variability between manufacturers, but not by much.
    • There are significant pre-analytical issues with glucose.  For example the guidelines call for a high-carb diet for three days before performing an OGTT, no one actually does this.
    • Biological variability is much greater with glucose.
    • One argument against HbA1c was that some patients will have abnormal red cell lifespan; I argued that such patients are sick and thus will have glucose values that cannot be properly interpreted.
  • Determination of diagnostic cutoff for HbA1c
  • HbA1c is a continuum in terms of indicating risks for complications.
    • A cutoff of 6.5% HbA1c was chosen because data from the NHANES, Egypt and other studies looking at retinopathy showed an increase in retinopathy above this level.
    • In terms of pre-diabetes HbA1c levels between 6.0 and 6.5% represent a “gray zone”.
    • Our own data looking at over 14,000 HbA1c done over a number of years show that risks for macrovascular complications begin at ~5.5% HbA1c.
  • WHO considerations
    • The idea that HbA1c is not practical in third world countries predominates in the WHO, they are sensitive to cost issues.
    • WHO will likely state that HbA1c and glucose are both fine and that each country should decide for themselves.

Discussion: There have been few studies that compare the OGTT, FPG and HbA1c for diagnosing diabetes and those that have tended to use the OGTT as the gold standard.  Each test diagnoses about the same number of people but they are not all the same people.  Long-term retrospective studies comparing outcomes for patients diagnosed by the different tests will be coming out in the next few years.  Many physicians are already using HbA1c to diagnose diabetes even without an official recommendation.  The relationship between HbA1c and outcome risks is a continuum.  Recent data show that cardiovascular risks increase even at HbA1c levels over 5.5%, and HbA1c will likely be added to the list of risk factors for cardiovascular disease in the near future.