A Review of Laboratory Practices Using the HLA-B27 Survey by the College of American Pathologists: How Important Is Allele-Level Typing?

The CAP provides proficiency surveys for HLA-B27 testing twice per year to clinical laboratory subscribers. Each survey includes 5 challenges (ie, unknown samples). Participants submit their testing results for each challenge as HLA-B27 either “present” or “absent.” Participants are graded on the basis of the referral laboratory typing result (the expected) and/or the consensus response of all participants (consensus must be ≥90%). In addition to reporting the presence or absence of HLA-B27 in each challenge, beginning in the second survey (“B” survey) of 2012 and onward, each challenge also included a follow-up question of whether the participant believed their particular HLA-B27 test result is associated with AS. Laboratories are able to respond with “yes,” “no,” or “cannot determine.”

Along with the challenge result, participants are also asked to indicate what testing method(s) is used for the HLA-B27 testing. Participants are able to report the following as methods for HLA-B27 testing: flow cytometry; microcytotoxicity; and/or molecular methods, namely sequence-specific oligonucleotide hybridization (SSO), sequence-specific primer (SSP) polymerase chain reaction (PCR) amplification, and real-time PCR amplification. Of note, from 2010 to 2011 all molecular methods could only be reported in the single category of “molecular.” Starting in 2012 and beyond, the molecular methods category was expanded to differentiate the specific types of molecular assays (ie, SSO, SSP, real-time PCR/sequencing). Earlier surveys also allowed less frequently used methods to be reported (eg, enzyme-linked immunoassays) but were removed from later surveys owing to lack of participants who used these methods. Laboratories could also report “other” when their testing method did not fit into any of the listed methods. For these less frequently used methods as well as those reported as “other,” we combined them into the “other” category in this study. Survey data from an 11-year period (2010–2020) were included.

The number of participants for each survey and the number challenges answered by participants within each survey fluctuated (ie, there is not a 100% response rate). Therefore, to examine the trends of which testing methods were used by laboratories over time, we calculated the percentage that each method is used for each survey and plotted them over the entire study period. We also determined the analytical performance of each method by determining false-positive and false-negative results. The false-positive and false-negative rates for each method were calculated by adding all false results by each method while the denominator used was the total number of challenges tested by that specific method. The χ² test was used to determine if errors were different by method, with P < .05 considered significant. All percentages were rounded to the whole number.

To better understand if laboratories recognize and use the evolving scientific data around HLA-B27 risk alleles, for the 2018 surveys, we included additional queries on practices related to HLA-B27 testing. For survey “A,” 3 dry laboratory challenges (case scenarios) were presented with brief clinical data and HLA test results by both flow cytometry and molecular methods. Laboratories were then asked whether these findings were consistent with AS. For survey “B” we delved deeper into specific reporting practices and inquired which HLA-B27 alleles laboratories believed to be associated with AS and whether laboratories performed allele-level typing.

A total of 110 HLA-B27 typing challenges (unknown samples) were included in 22 proficiency testing surveys for HLA-B27 typing reviewed (5 challenges per survey, 2 surveys per year) from 2010 to 2020. The average number of participating laboratories during this period was 173.31 (range, 156–186) and the average number of laboratories that returned responses to surveys was 151.36 (range, 131–178). On average, there were 149.2 responses per challenge/unknown sample (range, 128–170).

Overall, the trends in testing methods for HLA-B27 have shown a slow shift during the past 11 years, with a gradual increase in the use of molecular testing methods (see Figure). Of note, in 2010 and part of 2011, all molecular methods now in use (SSP, SSO, and real-time PCR) were originally combined into the broad category of “molecular.” Only in the second survey (survey “B”) of 2011 were the distinct molecular methods parsed out. Serologic/antibody methods (ie, microcytotoxicity and flow cytometry) represented almost 70% (908 of 1307 responses) of HLA-B27 testing methods in 2010 but decreased to 54% (782 of 1450 responses) by the 2020 survey. Flow cytometry continually comprised the most common method used for HLA-B27 testing throughout the 11-year period (range of 47% to 62% or 361 of 769 and 409 of 664 responses, respectively). Notably, the specific methods that account for most of this change include microcytotoxicity, SSO, and real-time PCR. Microcytotoxicity has decreased over time from a peak of about 10% (135 of 1307 responses) down to 2% (28 of 1450 responses) currently. For the same period, SSO has increased (from 8% to a peak of 17%), while real-time PCR has shown the greatest increase from 2% to 15% (216 of 1450 responses).

Table 1 shows a comparison of error rates for the most common methods (flow cytometry, molecular, and microcytotoxicity). Of note, the specific molecular methods (ie, SSO, SSP, or real-time PCR) were not distinguished in the proficiency surveys until the second survey in 2011. It is notable that the single method with the highest error rate was HLA-B27 typing by flow cytometry and the most accurate was by SSO (P < .001). Because SSO had no errors, we used this as a comparator against other molecular methods reported in the survey. While SSO had no errors, real-time PCR (with 2 errors) was not significantly different (P = .11) but significantly better than SSP (P = .02).

Beginning in 2012, each survey included an additional question for each of the 5 challenges to determine if laboratories were able to further assess whether their results are associated with the development of AS or not. During the 9 years (18 surveys) that included these questions, an average of 22% (528 of 2398 responses) of respondents indicated that they are unable to determine if their result is associated with the development of AS or not, based on their reported results.

To further understand why approximately one-fifth of participants were unable to determine if their result was associated with AS or not, for the 2 mailings in 2018, we included dry challenges/supplemental questions regarding HLA-B27 alleles. For these specific 2018 surveys, 358 laboratories subscribed to the CAP HLA-B27 survey (HLA-B27 A and HLA-B27 B survey, n = 179 each), but only 316 laboratories returned answers to the survey. A little more than half of respondents reported using flow cytometry to test for HLA-B27 status for the two 2018 surveys (Table 2), whereas the balance of respondents used molecular methods.

Based on 3 case scenarios presented in the 2018 survey “A,” there appears to be broad understanding that allele-level HLA-B27 typing results inform clinical interpretation (Table 3). This is best seen for the third case presented: despite the flow cytometry result being “positive/present” for HLA-B27, when also provided information that the allele typing is HLA-B*27:06, most laboratories recognize that this allele is not associated with AS. This contrasts with the other cases where both flow and molecular results were concordant (positive).

For survey “B” we listed some of the common, well-documented alleles of HLA-B27 and queried participants to indicate which alleles were associated with AS and which were not. We also inquired whether laboratories performed allele-level typing and if they included clinical interpretations with their test results. In this survey, 8% (11 of 134) of laboratories reported allele-level typing for HLA-B27 testing and most laboratories (78% or 115 of 148) do not or are unable to provide additional clinical interpretation beyond “present/positive” or “absent/negative” for HLA-B27. Interestingly, laboratories appear to agree that certain HLA-B27 alleles are associated with AS, with most respondents indicating that the following alleles are associated with AS: *27:02, *27:04, and *27:05, whereas certain alleles are not associated with AS: *27:06 and *27:09. Notably, there are other B27 alleles for which laboratories are less clear on risk for development of AS (Table 4).

During the past 11 years, an average of 173 clinical laboratories have subscribed to the CAP HLA-B27 survey; this likely represents broad practices by clinical laboratories performing HLA-B27 testing. However, we note that CAP data may not be fully reflective of the state of HLA-B27 typing nationwide, as other clinical laboratories may subscribe to other proficiency testing providers. Each survey is shipped with 5 challenges (unknowns). We would expect that participants perform testing on all challenges, but some laboratories did not return results for all 5 challenges. It is not clear why laboratories do not submit responses for all 5 challenges. Based on individual experiences in our laboratories, it may be laboratory error (eg, sample was spilled/contaminated/exhausted) or that some assays require more technical expertise. SSP generally requires additional sample DNA compared to other molecular methods. Serologic assays require intact (and in the case of microcytotoxicity, live) cells and sample handling/processing errors or delays may impact testing. Interestingly, laboratories are more likely to not report results for all 5 challenges when their sole testing method is serology and requires intact cells (ie, flow cytometry and/or microcytotoxicity). In some circumstances, laboratories whose sole method is SSP also do not submit responses for challenges. It is also notable that flow cytometry had the highest rate of errors for HLA-B27 testing, while among the molecular methods, SSP trailed behind real-time PCR and SSO. Perhaps when laboratories have challenges with indeterminate/questionable results, they opt not to report those for fear of having a poor performance on proficiency testing, as you are only graded on submitted results.

The most common method for HLA-B27 testing reported by our survey participants continues to be flow cytometry. However, there is a noted trend toward molecular methods with the combined molecular methods representing less than 25% of testing methods in 2010 and, in the past 2 years’ data (2019–2020), now making up 44% of methods used to test for HLA-B27. We view this as a positive trend, as HLA-B27 testing by molecular methods probably provides more reliable and additional information for clinical use.

Flow cytometry for HLA-B27 testing is used only to determine the presence or absence of HLA-B27 on patient cells. This (and all serology/antibody-based methods) is broadly recognized by the clinical histocompatibility and immunogenetics community as achieving only antigen-level (low-resolution) and not allelic-level (high-resolution) HLA typing. The most widely reported commercial kits used by our participants include HLA-B27 detection reagents/kits from BD Biosciences, One Lambda/Thermo Fisher, and Beckman Coulter.^4,5  All 3 kits/reagents use monoclonal antibodies to detect HLA-B27; the most commonly used is the ABC-m3 clone, which is recognized to have false-positive reactivity, especially against HLA-B7 and other related antigens.^6,7  In fact, in a study comparing the specificities of HLA-B27 monoclonal antibodies, the ABC-m3 clone showed the worst false-positive rate when compared to 2 other HLA-B27 antibody clones, FD704 and GS145.2.⁷  The CAP HLA-B27 proficiency survey data support and expand on the limitation of HLA-B27 testing by serology. Our experience in reviewing proficiency testing results shows that antibody-based testing (ie, flow cytometry and microcytotoxicity) has its pitfalls, including a higher error rate in both false-positive and false-negative results compared with molecular methods. The CAP HIT Committee previously reported false-positive results associated with flow cytometry, especially in the presence of the common HLA-B7 antigen.⁸  In our 11-year cumulative data, the false-negative rate by flow cytometry actually exceeds the false-positive rate (3.33% versus 2.03%; Table 1). Importantly, another pitfall to flow cytometry and not reflected in the specific method error rates is that many laboratories when using flow cytometry report “indeterminate/other.” The causes of “indeterminate/other” results may be due to several factors. For serologic methods, the presence of cross-reactive HLA-B antigens may lead to indeterminate results. There may also be technical problems that participants encountered. Indeterminate results generally need to be confirmed by an alternative method. The CAP HIT Committee recognizes that laboratories may have policies/protocols whereby for clinical samples with indeterminate results, these samples may be reflexed or sent for referral testing. Clinical Laboratory Improvement Amendments (CLIA) rules require proficiency testing samples be treated like clinical specimens but explicitly prohibit referral testing of proficiency samples. Therefore, in this situation where laboratories report “indeterminate,” laboratories cannot be penalized and therefore such results are not considered incorrect but instead considered “acceptable.” A snapshot review of the last 5 years of survey results shows that 101 responses (of 7690 responses to 50 challenges) were reported as “indeterminate/other.” Ninety of these 101 “indeterminate” responses were by flow cytometry, 2 were by molecular (1 each SSO and SSP), 1 by immunoassay, and the remaining 8 were by cytotoxicity. Notably, HLA-B27 testing by flow cytometry does have a distinct advantage over molecular methods of being performed on whole blood without requiring DNA extraction and amplification. In clinical laboratories without the proper facilities for molecular methods–based testing, this may be an advantage. Another advantage of flow cytometry is the ability to automate sample handling and thus help in laboratories with high throughput.

This is in contrast to some molecular methods (eg, high-resolution SSP, intermediate-resolution SSO, or real-time PCR) that are less likely to mistype HLA-B27 and can provide higher level typing (eg, high-resolution SSP, SSO, or real-time PCR). Depending on the molecular method used and the genotype of the patient, it is possible to infer HLA-B27 alleles by ruling-in (or out) certain HLA-B27 alleles based on typing results. This may be sufficient to interpret whether a patient has a risk allele or not. These methods all together have a lower false-positive and false-negative rate than flow cytometry, with SSO showing the best sensitivity and specificity among the 3 (no false-positive or false-negative results). Molecular-based testing does require more stringent clinical laboratory practices owing in part to the need to mitigate contamination of samples by prior nucleic acid amplification products. Molecular testing, relative to flow cytometry, also takes longer to perform and may cost more per test. There are molecular kits available outside the United States for performing nucleic acid amplification directly from whole blood without DNA extraction⁹  and may prove to be an important consideration as laboratories are challenged to improve turnaround times and efficiency. Each specific molecular method has advantages and disadvantages that may impact laboratory-specific workflow. Although technical comparison of methods is beyond the scope of this article, SSO is amenable to higher-volume testing, whereas SSP may be helpful for smaller test volumes but requires the use of gel electrophoresis. Depending on the specific assay, SSO and SSP may yield low- to intermediate-level resolution typing. Real-time PCR provides a very rapid result but also requires real-time PCR instruments that may not be as widely available as generic thermocyclers. We also note that sequencing-based assays such as Sanger and next-generation sequencing methods could also be used for typing but may be impractical and cost-prohibitive for single gene/locus typing, such as HLA-B27. HLA typing by sequencing, particularly next-generation sequencing, is considered the gold standard for allele-level or high-resolution HLA typing, particularly for hematopoietic cell transplant.¹⁰ 

It is not surprising then that when queried, most laboratories that perform HLA-B27 testing do not, and cannot, include additional interpretative comments other than HLA-B27 “positive/present” or “negative/absent,” as antibody-based testing is unlikely to provide this additional information. It is apparent from our participant responses, especially with the 2018 survey cases and supplemental questions, that there is broad clinical laboratory agreement and recognition that certain HLA-B27 alleles have stronger association with the development of AS. The consensus among participants is that well-described alleles in the literature, especially HLA-B*27:05, are highly associated with AS. Indeed, in White populations, HLA-B*27:05 and *27:02 are the most common alleles associated with AS, whereas in Asian populations, B*27:04 and *27:07 are more common. HLA-B*27:06 and *27:09 are not associated with development of AS.¹¹  However, the survey responses also appear to show a spectrum of uncertainty about whether or not other HLA-B27 alleles (HLA-B*27:03, *27:07, and *27:08) play a role in the development of AS.

It is also not clear whether a patient’s race is taken into consideration when interpreting data. Studies have shown that in White Europeans, HLA-B*27:05 has the strongest association with development of AS,¹²  yet this association is not true for all races/populations as the same HLA-B*27:05 allele has instead a negative association with AS development in the south of China.¹³  Specific HLA-B27 allele risk also seems to be dependent on population as this influences the disease presentation in SpA. For instance, presentation with uveitis is associated with B*27:04 in the Indian population,¹⁴  whereas in the Chinese population, it is B*27:05 that shows association with orbital inflammation.¹⁵  Some data suggest that HLA-B27 zygosity can influence disease phenotype. In vitro studies have shown that leukocytes from patients with AS express more HLA-B27 antigen than controls¹⁶  and that homozygosity for HLA-B27 increases risk of disease.¹⁷  Currently, testing for HLA-B27 is recommended to aid in the diagnosis of AS, and the US Food and Drug Administration Blood Product Advisory Committee indicates that incorrect HLA-B27 test results can delay or adversely impact diagnoses of immune disorders.¹⁸  The disease risks attributed to HLA-B27 are clearly recognized as being dependent on both the HLA-B27 allele and the racial/ethnic group under consideration.¹⁹  Nonetheless, HLA accounts for just less than half of AS cases.¹¹  Other immune-related gene polymorphisms, such as in IL-1A and IL-23R,²⁰  may be an important part of a future panel to fully understand genetic predisposition for AS.

In summary, the trends in clinical testing for HLA-B27 show a persistent use of flow cytometry with a gradual trend for adoption of molecular methods. Our data suggest that flow cytometry results in more errors and does not provide additional diagnostic information, compared with molecular methods (ie, regarding HLA-B27 allele-specific disease development risk). As autoimmune disease rates have reportedly been increasing,²¹  the convergence of broader availability and adoption of molecular methods for higher-resolution HLA typing may be a ripe opportunity for laboratories to modify or extend their testing modalities and better enhance personalized and precision medicine. The ability to perform HLA typing at the allelic level for disease association may provide histocompatibility and immunogenetics laboratories an opportunity to contribute in increasing our understanding of the importance of allele-level typing in the biology of disease and also enable laboratories to provide interpretative results for better patient care.

We would like to thank the College of American Pathologists staff who support the Histocompatibility and Identity Committee, especially Jacques Mobille, BS, MLS(ASCP) and Ann Nwosu, MS, for data review.