Quality and best practice in medical laboratories: specific requests for autoimmunity testing

Today, most European medical laboratories must undergo EN/ISO 15189:2012 [2] accreditation and in most countries, this has been implemented or is ongoing. The European cooperation for Accreditation (EA) is the roof organization of the national accreditation bodies. One of the central targets of EA is to harmonize the requirements laboratories must meet when applying for and maintaining EN/ISO 15189:2012 [2] accreditation. Meeting these requirements is less of a challenge in Clinical Chemistry but is a topic of interpretation for specialized laboratories. For the accreditation of specialized laboratories, such as in the field of immunology, accreditation bodies need scientific advice how to interpret the EN/ISO 15189:2012 [2]. Furthermore, because some countries do not have enough specialists in autoimmune testing or even in immune diagnostics, assessments are done by auditors without specific experience in this field.

There are some international guidelines including, at least in part, autoimmune testing. They have been developed by expert panels, represent state of the art diagnostics, and influence national practice. Most guidelines deal with the testing algorithm and give rather clinical advice. Some of the most relevant ones with respect to ANA have been published by IFCC [11], CLSI [12], ACR [13], EULAR [14], EASI/IUIS [15], and ICAP [16‐18]. Though authoritative expert panels issued these guidelines, there are distinctions amongst the provided recommendations, leaving room for interpretation. As a result of this, national documents were derived in most countries (see also [15]). They have been published as guidelines, recommendations, or checklists. These various national rules have a lot in common. We collected information from Austria, Belgium, Croatia, Estonia, Finland, France, Germany, Greece, Hungary, Israel, Italy, Norway, Poland, Portugal, Russia, South Africa, Spain, Sweden, Switzerland, The Netherlands, and Ukraine. In most countries there are no specific rules for autoimmune testing in accredited laboratories and use of international recommendations is common, but not specified. This is true for Finland, France, Israel, Norway, Poland, Portugal, Russia, Sweden, and Ukraine. Belgium [19], Germany [20‐24], Italy [25, 26], The Netherlands (https://​medischeimmunolo​gie.​nl/​cml/​richtlijnen) and Switzerland (https://​ssai.​ch/​organisation/​commissions/​cld/​) have one or more national guidelines or recommendations for autoantibody testing. In Austria [27], France [28‐30], Italy (http://​www.​gruppofirma.​com/​eng/​), and Spain [31] there are publications by the national EASI groups (https://​www.​gfid-ev.​de/​easi/​), but most of these documents have no formal regulatory role.

Autoantibody testing must be performed by authorized and trained staff. EN/ISO 15189 accreditation not only requires determining the level of competence but also to assess and reassess the competence of each person. This can be done by pairwise examining EQA samples (“consensus training”) or common and rare samples under supervision and should be done and documented half-yearly. There are some further challenges:

There is increasing automation of autoantibody testing, including binding assays, testing by IIFA, and line/dot blots. Space requirements for automated instruments are not different from Clinical Chemistry or infectious serology. Nevertheless, manual processing of microscope slides, blots, ELISAs and even RIAs is still common practice in many laboratories and requires specific conditions and minimal space. These manual assays need a dedicated space for sample preparation and often require an adequate microscopy cabinet that can be dimmed.

If possible, high-throughput tests should be performed on fully automated systems that are linked to the laboratory information system. This avoids errors and gives more time for performing specialty tests and confirmation steps. The proper maintenance of these automated systems is crucial too. Fluorescence microscopes should be equipped with a LED light source to reduce maintenance efforts. If reading is not automated, data should be entered directly into laboratory systems; a middleware is the most common solution applied by laboratories today.

A correct test request that is based on a pre-test clinical suspicion is critical to ensure optimal value of a test result. Request forms must be suitable to provide proper clinical information. Direct interaction with clinicians is difficult for commercial laboratories, but the requesting clinician should be encouraged to formulate clinical questions of relevance to differential diagnosis and management and to give tentative clinical diagnosis. In some settings, a specific form is implemented for antibody requests, providing clinical information to help interpreting test results.

Pre-analytical requirements for autoantibody detection are not time-critical, but quality should be as good as specified by manufacturer (icterus, lipemia, hemolysis, or other contamination; no repeated freeze-thawing). The sample quality also depends on patients’ status (prolongation of coagulation time in certain clinical situations; pregnancy etc.).

Most autoantibodies are detected in serum and are validated by the companies only for this matrix. Important exception is a series of neurological antibodies that may also be detected in cerebrospinal fluid. From laboratory processes perspective, plasma is favored since the lower centrifugation time and many manufacturers pointed out in their protocols that plasma (lithium heparin? citrate? EDTA?) can be used. If other body fluids are used, this will in general be outside the scope of the assay (unless the insert states otherwise).

The correct screening strategy for autoantibodies depends, at least in part, on the pre-test probability of the patient population. This varies by the requesting clinical discipline with different prevalence of positive test results [32, 33]. Furthermore, the number of positive findings also depends on the test characteristics (more positives for highly sensitive but less specific assays). In case of ANA, the prevalence of positive test results is < 30% [34]. For patients with a relatively high pre-test probability for SARD (requested by rheumatologists or clinical immunologists) or patients for which a SARD is to be excluded, the sensitive ANA IIFA may be the most adequate screening test. On the other hand, patients with a relatively low pre-test probability for SARD (requested by general practitioners or non-rheumatology/non-clinical immunology specialists) or for patients suspected of anti-SSA/Ro related diseases or IIM, alternative assays may be more appropriate [35, 36]. Finally, a patient may present with clinical manifestations completely in line with a specified diagnosis; testing in such patient is rather redundant unless the patient is to be included in a study or if—in rare situations—the test must be performed at presentation and during follow up to access response to treatment. Analytical sensitivity and specificity of IIFA ANA tests and binding assays are well known and result in different performances (Fig. 1). A special situation is the detection of antibodies that give information on prognosis. For example, some MSAs may not be useful for diagnosis, but they may give insight into prognosis (e.g., TIF-1gamma) [37‐39]. This illustrates that clinical information is of utmost importance in order to perform tailor-made diagnostics. The alternative is to perform a combination of tests. There is evidence that combining IIFA with solid phase assays gives the most clinically useful information requested by general practitioners or non-rheumatology/non-clinical immunology specialists [36, 40‐42].

Although there may be good arguments favoring immunoassays with defined antigens for screening instead of IIFA, there still are strong arguments for initial IIFA: reagents are generally cheaper, well-established systems for automation are available, the number of target antigens presented is higher and not all of them are available in commercially available solid phase assays (some may not even have been characterized), and several recommendations and disease criteria request IIFA.

Clinical validation must be distinguished from analytical validation. According to the new IVD regulations, clinical validation (i.e. diagnostic sensitivity and specificity) is the responsibility of the company. This information must be shared with the customers.

Analytical validation can occur at a multicenter level and can be supported by existing literature, but local verification must be performed as well. In this respect, laboratories sometimes have not enough positive samples. For a certain period, this can be overcome by proficiency test samples and control material, but the verification process must be finalized with real samples within around half a year. It may otherwise be questionable if these laboratories should perform the assay.

The choice of the analytics should primarily be based on the test characteristics and independent from methodology. Before implementing IVD tests, laboratories must verify them. The following test characteristics should be considered in such a verification:

For analytical validation, qualitative and quantitative assays require different approaches. Until today, there are no true quantitative tests for autoantibody detection. Because such tests detect a polyclonal mixture of antibodies, even well-standardized parameters such as RF are not quantitative; there is no reference method. This also means that results are test kit-specific. It is possible to perform the analytical verification semi-quantitatively and define discrete classes (negative, intermediate, positive).

The challenges here include (but are not limited to):

The Dutch initiative for national validation/verification of autoantibody assays gives clear advice [44]. A common and not yet solved challenge is the verification and quality control in multiplex assays. The typical procedure assumes that the testing of exemplary antigens is enough.

Slide reading should ideally be done with direct importing of results into a laboratory information system through middle ware or must be double-checked. Pattern recognition software is not (yet) fully reliable and can only replace one of two mandatory readers if at all.

Dot-blots and line-blots are not really blotted but are rather conventional immunoassays. They are intended to analyze autoantibodies to multiple antigens in parallel. For quantification, control lanes and internal standards can be part of the test kit. Some manufacturers only provide a control lane showing IgG content in the strip; thus, there are no graded internal controls for the individual autoantibodies analyzed in the test system. In case of disease specific antibody panels, i.e. screening for myositis/scleroderma-typical autoantibodies, these assays are increasingly used, but verification of the rare specificities is often lacking.

Immunological binding assays for ANA, ENA, ANCA, RF or ACPA are available as microtiter plates or as test kits for common analyzers in Clinical Chemistry. While microtiter plates are processed as a batch with calibrators and controls per plate, random access analyzers follow the common quality rules in Clinical Chemistry (lot-specific master curves).

Diagnostic laboratories are bound to kit-intern quality controls of CE-labeled IVD. Replacing a kit-intern control means modifying the test and requires validation efforts. Additional third-party controls are highly recommended. One important reason to have in-house internal controls in the lab is the longevity of stored properly. In-house internal controls might last a decade and will still be the same when the companies change their kit controls in an unpredictable way. In-house controls based on positive findings can be applied but must be profoundly characterized and well documented. Internal control level should be in the measuring range, preferentially near the cut-off.

For solid phase assays and IIFA, internal controls must cover as many antigens as possible. For IIFA, control sera can also be applied as titer-control to proof stability of reagents, slides, microscope, and any other component influencing microscopic imaging.

Lot approval should be done with the use of confirmed samples with known laboratory results. If possible, clinical information about these samples should be known to the laboratory specialist.

Results of internal controls should be transferred into the quality control charts of the laboratory and managed accordingly.

In some countries, there is a difference between external quality assessment (EQA), which is performed on a voluntary basis, and proficiency testing, which is obligatory and involves restrictive measures. Participation in proficiency testing is mandatory for medical laboratories. They should cover the full scope of parameters offered by the laboratory within a defined time period. For some kinds of specific test kits this is a challenge: myositis blots, neuronal blots etc. This often means that proficiency testing providers from other countries must also be involved. A good database for EQA providers can be found here: https://​www.​eptis.​bam.​de/​eptis/​WebSearch/​main. For rare antibodies there is often no EQA. Sample exchange with other laboratories can assist efforts to overcome this issue. Clinically validated specimens should be used but must be defined thoroughly. The rules for such exchanges must be defined in advance and should consider various test systems if possible.

Turn-around time depends on the setting. In a hospital, a preliminary result of ANA/ENA analysis should be available within 48 to 72 h, but this can result in need for correction by unforeseen results of additional follow-up tests. ANCA or GBM are more urgent than other autoantibodies and must be reported within 24 h if requested as such; therefore, rapid test performance must be guaranteed.

For out-patients, there is in general more time to finalize everything until the follow-up consultation. To remain within an acceptable timeframe, rare tests can be outsourced.