H. Reiber
Quality Assessment in Laboratory Medicine
Accuracy and precision in laboratory medicine are maintained at a high level, controlled mainly by the following activities:
• Internal quality control.
• External quality assurance (EQA).
• Knowledge-based interpretation concepts. In CSF analysis, this is supplied by the integrated CSF report (Chap. 19).
• Laboratory protocols (laboratory accreditation).
Internal Quality Control
Internal quality control is part of the daily analysis protocol, using certified commercial control samples for assurance of accuracy in an individual analytical series and of precision calculated from a series of daily control values. The mean of these data measured between two calibrations of an analyzer may be used to control the calibration-dependent variation in accuracy.
External Quality Assurance
EQA is a major attempt to improve the quality standards in laboratory medicine and to keep interlaboratory variation low. Many national and international institutions have defined rules and regulations, and also organize meetings (see links at http://www.instand-ev.de/en/links/). There may be one or two official assessors per country which offer interlaboratory surveys. In Germany, INSTAND e. V. is one of two institutions for quality assurance working from mandatory guidelines defined by the German Federal Medical Association (BÄK, 2008) (Table 20.1); it has been a collaborating center for quality assurance and standardization in laboratory medicine of the World Health Organization since 1994. The following description of EQA for CSF analysis is based on the development of INSTAND with the German Society for CSF Analysis and Clinical Neurochemistry (Deutsche Gesellschaft für Liquordiagnostik und Klinische Neurochemie, www.dgln.de) and represents an implicit consensus of 400 participating laboratories from Germany and 12 other European countries (Reiber, 1995; Reiber and Uhr, 2003).
Interpretation Concepts and the Quality of Information
A generally accepted definition of EQA is given in ISO 15189 (“Medical laboratories: particular requirements for quality and competence”): “External quality assessment programmes should, as far as possible, provide clinically relevant challenges that mimic patient samples and have the effect of checking the entire examination process, including pre- and post-analytical procedures.”
EQA relates not just to finding appropriate samples for an interlaboratory survey, which can in itself be demanding: it also considers postanalytical procedures, i. e., data handling and interpretation.
General guidelines (BÄK, 2008) regulate the quality of results (mostly numbers for absolute concentrations in a sample), but have problems with very particular rules for quality of information or do not consider them at all. However, this is the issue that is important for the patient.
Laboratory Accreditation
Accreditation of a laboratory by an accreditation body is a big business driven by competing laboratories. This costly and time-consuming activity leads to a tremendous expenditure of time on the documentation of everything. This concept pretends that human action is completely to be formalized. Adaptations and improvements of commercial assays—albeit essential to the specialty laboratory, like the CSF laboratory—are practically disencouraged.
Special Features for Quality of CSF Protein Analysis
1. The CSF/serum quotient of serum proteins in CSF—a mathematically normalized CSF concentration—is based on a natural biological relation.
Most proteins in CSF are blood-derived. As shown in Chaps. 1–3, the CSF concentration of a serum protein is modulated by its serum concentration (and the CSF flow rate). Thus, the CSF/serum quotient represents a normalized CSF concentration, i. e., a value independent of the varying concentration in blood. This is why CSF/serum concentration quotients of serum proteins, which represent a biologically based entity, are different from any other calculated ratio in clinical chemistry, e. g., the percentage of albumin in total protein calculated from serum electrophoresis.
The CSF/serum quotient has a smaller biological variability then the absolute CSF concentration, and therefore offers a more sensitive and specific reference value for the discrimination of a pathological brain-derived fraction from a normal blood-derived fraction in CSF.
2. The CSF/serum quotient is a method-independent value if derived after paired protein analysis of CSF and (appropriately diluted) serum samples.
If CSF and serum samples are analyzed in the same analytical run, compared to analysis with reference to two different calibration curves (and hence different accuracies), the coefficient of variation (CV) of CSF/serum quotients is also smaller (Andersson et al., 1994). However, a basic precondition must be fulfilled for these statements: the calibration curve must give concentration-independent accuracy. This is controlled by measuring a serially diluted serum sample with dilutions covering the complete analytical range of the corresponding calibration curve. With these preconditions, the quotient is independent of absolute accuracy of the CSF and serum concentration values, i. e., it is a method- and calibrator-independent value.
3. QAlb-related interpretations of immunoglobulin quotients are most specific for detection of an intrathecal immunoglobulin fraction in CSF if referred to a hyperbolic discrimination function in quotient diagrams.
In contrast to the linear IgG index or linear functions, which are subject to many false-positive interpretations (Reiber et al., 2001; Reiber et al., 2009; Stauch et al., 2010), the hyperbolic discrimination line is an empirically and theoretically founded function (Chap. 2).
4. The age-related evaluation of the QAlb to detect a barrier dysfunction must take into account the existence of different reference values for ventricular, cisternal, and lumbar CSF.
5. The integrated CSF report (Chap. 19) is part of a quality assessment because it adds reliability for the data analysis due to physiological plausibility of data combinations and clinical plausibility from disease-related patterns.
The physiological connection between different analyzed CSF parameters allows plausibility control of the value of any individual analyte if all the data from one patient are reported in a combined “integrated” data report (Chap. 19). This is an improvement for quality of results. With the disease-related data patterns of an integrated report we also gain reliability from complementary clinical information (differential diagnostic question of the clinician): any contradiction here indicates problems with the analytical results or with the differential diagnostic question.
Internal Quality Control for CSF proteins (Total Protein, Albumin, IgG, IgA, IgM)
CSF assays must take into account the low protein concentrations in normal CSF and subsequently the large variation of concentrations when intrathecal synthesis takes place, which is 10- to 100-fold larger than in the case of pathological changes in serum. This is the basic requirement made of industrial development: to allow sufficiently high sensitivity and a reasonable protocol, so that default dilution can be changed to avoid extreme positions on the calibration curve. These are the main problems in immunochemical nephelometry, in particular for IgM, that lead to concentration-dependent differences between the various analysis methods (Reiber et al., 2010). The individual laboratory has to control for this by internal quality control.
The use of an integrated CSF data report (Chap. 19; “Special Features for Quality of CSF Protein Analysis,” above) to report laboratory data to the clinician is a basic part of the quality assessment in the CSF laboratory: it improves quality control by adding plausibility control in the framework of disease-related data patterns. Discrepancies in the analytical data (Chap. 21) or discrepancies between the suggested diagnosis and the analytical results (Chap. 19) may prompt a repetition or extension of the analysis.
Precision. For interassay precision, CSF and serum control samples are measured with each analytical run and documented together with the CSF/serum quotients. Variation and trends between two calibrations can be read directly from the charts or by statistical calculation.
Accuracy. Control for accuracy of absolute values is ensured by two certified reference serum samples (normal range and pathological range) and two CSF samples in the upper and lower range of the calibration curve (see below). A calibration-dependent variability is recognized from mean values of the precision controls.
With each new assay or new batch of reagents, the laboratory must ensure method-independent accuracy of the paired CSF and serum analysis: a serum control sample is serially diluted down to CSF concentrations with dilutions covering the complete analytical range of the corresponding calibration curve in the CSF assay.
Control Material for CSF Analysis
Certified control samples for proteins in serum are used as available from commercial sources. However, to date (2010) no certified commercial protein control is available that is suited to normal CSF protein concentration (in particular for IgA and IgM), as the concentrations of analytes in commercial control samples are too high. This leads to serious control problems (Reiber et al., 2010) and makes the following particular proposals necessary.
Diluted control serum may be used (1:200 to 1:2000, depending on the analyte) and stored frozen in aliquots. In combination with a certified reference serum sample, an inhouse CSF pool may also be used with sample aliquots stored at 4°C, provided an antimicrobial agent has been added for stabilization (e. g., NaAzid).
A suitable control sample for normal CSF should yield (Andersson et al., 1994):
• IgM values between 0.5 and 1.5 mg/L.
• IgA values between 1.0 and 3.0 mg/L.
• IgG values between 10 and 30 mg/L.
• Albumin values between 100 and 300 mg/L.
The INSTAND Interlaboratory CSF Survey
Twenty years ago the interlaboratory CSF survey instituted by INSTAND in Germany established clinical interpretation of data patterns in addition to single result control. This EQA concept takes the postanalytical element—i. e., data interpretation—seriously as an aspect of general quality assessment. In this kind of training program for the interpretation of CSF data, participants also improve their alertness for methodological discrepancies. For a full account of the original survey, see Reiber (1995).
EQA for CSF proteins (Total Protein, Albumin, IgG, IgA, IgM)
The survey for these combined analytes has three levels.
CSF Survey: 1. Evaluation of Data Interpretation
With the five elements listed above (“Special Features for Quality Assessment of CSF Protein Analysis”), and based on the interpretation of the CSF/serum quotients in the quotient diagrams (Chap. 19), the participants in the CSF survey have to judge:
• Barrier function: normal or pathological.
• Intrathecal IgG and/or IgA and/or IgM synthesis: detectable or absent.
• Inflammatory process in CNS: detectable or absent.
The certificate of participation (see, e. g., Table 3 in Reiber, 1995) documents as primary result the correctness or otherwise of participants’ patient-oriented interpretation of the data.
CSF Survey: 2. Evaluation of CSF/Serum Quotients
In respect of quantitative results, we refer with priority to the CSF/serum quotients as the specific feature for quality of CSF analysis (see above) and judge:
• The accuracy of CSF/serum quotients and their deviation from target values (Table 20.1).
Any serious deviation in the accuracy of a quotient is commented on by indicating the source of the deviation in the absolute value of the analyte (i. e., CSF, or serum, or both).
CSF Survey: 3. Evaluation of Single Values in CSF and Serum
• Absolute values in CSF and serum are individually certified to meet the specifications of the German Federal Medical Association (BÄK, 2008) with the new permissible limits shown in Table 20.1.
The target value, consensus value of the group after subtraction of the outliers, CV of the interlaboratory variation, and the individual deviation of the participating laboratory's value from the target value are reported, including a graphical demonstration of the laboratory's position in the group performance in quotient diagrams. This established protocol was reported earlier (Reiber, 1995). Numerical statistics and method-related performance are reported to the participating laboratories in a summarizing letter, recent examples of which can be obtained from the INSTAND website (www.instandev.de/en/eqas/legends/, category: Cerebrospinal fluid analysis).
CSF Survey: Total Protein
Total protein in CSF has lost its diagnostic relevance, but does offer a plausibility control for the correctness of the albumin analysis (< 85% of total protein) or helps to find suitable predilutions in the automated analysis of single proteins. Evaluation of the absolute value is sufficient. Samples of one normal and one pathological total protein concentration (CSF pool) are tested. The method-related evaluation shows large differences of the medians between methods which led to an improvement of the calibrator from one manufacturer (reports 2006–2010, at www.instandev.de/en/eqas/legends/, category: Cerebrospinal fluid analysis).
CSF Survey: Specific Antibodies in CSF and Serum
As with total immunoglobulins, the detection of specific antibodies in CSF and serum by evaluation of the CSF/serum quotients is better than evaluation of the absolute values. Titers are not suitable for CSF analysis.
A basic challenge for the standard microbiological laboratory is the low concentration of the antibodies in blood—frequently below the cut-off for a systemic infection in blood. Accurate analysis of serum antibody concentrations, also below the cut-off value, are the challenge for the usual methods used in serology: in the case of a CSF-related serum analysis, the usual report in serology for such a case with a comment like “blood titer is negative” is obviously not helpful at all.
In the interlaboratory CSF survey, where the calculated antibody indices (Chap. 5, “Antibody Index”) are interpreted, only the QSpec has an impact on the result, because the IgG and albumin values are passed on to the participating laboratories in advance to ensure that uniform IgG values serve as a basis for calculating the antibody index (AI) (quality control of IgG analysis is carried out separately). The mean CV for QSpec is 5–10%, whereas the mean CV for total AI values (including the imprecision for IgG analysis) is 16% (Reiber and Lange, 1991; Quentin et al., 2004). For a correct result participating laboratories also have to calculate and choose the relevant reference quotient, QIgG or QLim (IgG). The AI values are evaluated for deviation from target values and certified as right or wrong. In addition, the participants’ interpretations are compared with the correct interpretation, which contains the following spectrum of options:
• Normal AI values.
• Intrathecal antibody synthesis with specification of the antibody species that is pathologically increased.
• Chronic inflammation (when certain combinations of antibodies are increased, e. g., MR, MZ, or RZ in the MRZ antibody response, Chap. 19, “Evaluation”).
• Implausible data combination (a single value with AI < 0.5 among other different, normal values).
CSF Survey: Oligoclonal IgG
The collection of samples suitable for the control of oligoclonal IgG in the interlaboratory CSF survey is a challenge, as these samples cannot be obtained from pooled CSF, but need to be from a single patient. The only way to collect the volume necessary for a group of 200 participating laboratories in the INSTAND survey is to take it from patients with a ventricular catheter in a neurosurgical intensive care unit. Multiple lumbar punctures are rare, but with predilution of the samples it is occasionally possible to collect samples in this way.
The bands in paired CSF and serum samples are evaluated according to the international criteria (Andersson et al., 1994) (see also Fig. 4.13):
• Type 1: No oligoclonal IgG detectable by isoelectric focusing.
• Type 2: Oligoclonal IgG in CSF (intrathecal synthesis).
• Type 3: Oligoclonal IgG in CSF, additional identical bands in both CSF and serum (intrathecal synthesis).
• Type 4: Identical bands in both CSF and serum, no isolated oligoclonal IgG in CSF (no intrathecal synthesis).
• Type 5: Monoclonal IgG in both CSF and serum (paraprotein, no intrathecal synthesis).
Manufacturer-Related Versus Patient-Related Evaluation in Interlaboratory Surveys
The evaluation of data from laboratories participating in a survey refers to target values (defined for analytes with a reference method, such as glucose or total protein in serum) or assigned values (defined for analytes without a reference method). The assigned values for nephelometric analysis in the CSF survey are detected by two laboratories with Beckman and two with Behring nephelometers, each performing tenfold interassay analysis. The mean in each group is corrected if necessary by the consensus value of the group (for detailed procedures see www.instandev.de).
Occasionally the method-related evaluation shows concentration-dependant discrepancies which do not allow a common evaluation of all participants. In reality these discrepancies are a problem for the patient's diagnosis and may well have consequences for therapy. In particular, the points listed above under “Special Features” clearly show that these problems in protein analysis could be avoided by the manufacturer (Reiber et al., 2010) and in any case must be controlled by each individual laboratory (see above, “Internal Quality Control for CSF Proteins”).
It has been proposed by some EQA authorities that the assigned values for data control could be found as the median within a group using the same method. Taking this approach, however, the survey is not testing the real accuracy of reported results, with serious consequences for patients: in particular, in the decision range between normal and pathological values this concept promotes falsepositive or false-negative interpretations, which in the end must lead to incorrect information for the clinician (Reiber et al., 2010). This debate is actually also current in relation to the antibody index of microorganism-specific antibodies. It is easier to find a method-independent common target value for the antibody index for viral antibody assays than for bacterial antibodies, which have a much larger variation of coating antigens in the various ELISAs (Fig. 4.12). When the VlsE antigen was introduced into most Borreliaantibody assays, the large variation in group-related target values for the Borrelia AI diminished, as shown in the reports of CSF surveys for neuroborreliosis 2007–2010 (see www.instand ev.de/en/eqas/legends/, category: Cerebrospinal fluid analysis).
It must become the target for the manufacturers to evaluate and adapt new assays sufficiently to allow patient-related accuracy instead of a relative, method-related result. To allow the highest accuracy of interpretation is the responsibility of the manufacturers; but the users in the laboratories are responsible for the decision of choosing an assay found to be reliable by their own testing.
CSF Survey: Lactate, Glucose and Surrogate Markers
Glucose and lactate concentrations in CSF are of similar magnitude and there are no different matrix effects in the assays. This allows the interlaboratory survey samples for serum to be applied as CSF samples as well.
Surrogate marker proteins for differential diagnosis of dementia. The increasing spectrum of surrogate markers with a large variation of laboratory-dependent reference values and tremendous preanalytical problems urgently require a quality control scheme. A first approach is underway for a new interlaboratory survey (Lewczuk et al., 2006) to be launched by INSTAND in the near future.
CSF Cytology: Quality Control and Interpretation Training
The internal and external quality assessment for CSF cytology is very demanding. The excellent training program founded by E. Linke (Linke et al., 2005) in Stadtroda, Germany, is unique in the world. This event is organized under the name “Ringversuch vor Ort” (on-site interlaboratory survey); it regularly offers education and, at the same time, examination of each participant on four cytological CSF preparations. The written examination is then certified by INSTAND. This survey is performed by a reviewer of the German Society for CSF Analysis and Clinical Neurochemistry confirmed by the board of INSTAND.
References
Andersson M, Alvarez-Cermeño J, Bernardi G, et al. Cerebrospinal fluid in the diagnosis of multiple sclerosis: a consensus report. J Neurol Neurosurg Psychiatry 1994;57:897–902
BÄK. Richtlinie der Bundesärztekammer zur Qualitätssicherung quantitativer laboratoriums-medizinischer Untersuchungen. Dtsch Ärztebl 2008;105:A341–A355
Lewczuk P, Beck G, Ganslandt O, et al. International quality control survey of neurochemical dementia diagnostics. Neurosci Lett 2006;409:1–4
Linke E, Wieczorek V, Zimmermann K. Qualitätskontrolle in der Liquorzytodiagnostik. In: Zettl UK, Lehmitz R, Mix E, eds. Klinische Liquordiagnostik. Berlin: de Gruyter; 2005:380–391
Quentin CD, Reiber H. Fuchs’ Heterochromic Cyclitis – rubella virus antibodies and genome in aqueous humor. AJO 2004;138:46–54
Reiber H. External quality assessment in clinical neurochemistry: survey of analysis for cerebrospinal fluid (CSF) proteins based on CSF/serum quotients. Clin Chem 1995;41:256–263
Reiber H, Lange P. Quantification of virus-specific antibodies in cerebrospinal fluid and serum: sensitive and specific detection of antibody synthesis in brain. Clin Chem 1991;37:1153–1160
Reiber H, Peter JB. Cerebrospinal fluid analysis—disease-related data patterns and evaluation programs. J Neurol Sci 2001;184:101–122
Reiber H, Uhr M. Liquordiagnostik-Ausbildung und Fachqualifikation:
Richtlinien der Deutschen Gesellschaft für Liquordiagnostik und Klinische Neurochemie (DGLN). J Lab Med 2003;27:322–328
Reiber H, Thompson EJ, Grimsley G, et al. Quality assurance for cerebrospinal fluid protein analysis: international consensus by an internetbased group discussion. Clin Chem Lab Med 2003;41:331–337
Reiber H, Lange P, Albrecht K, Möller E, Wormek A. IgM analysis in CSF. A review of clinical relevance and method-related problems in immunochemical nephelometry. Clin Chem 2010, submitted
Stauch C, Reiber H, Rauchenzauner M, Pohl D, Hanefeld F, Gärtner J, Rostásy KM. Intrathecal IgM synthesis in children with multiple sclerosis is associated with a slower progression. Arch Neurol 2010, submitted