BACKGROUND: Cytomegalovirus (CMV) antibody donor screening assays have predominantly included both immunoglobulin G (IgG) and immunoglobulin M (IgM) detection. However, since in the majority of cases both CMV IgG and IgM are detected concomitantly during early seroconversion, CMV assays based only on IgG are now widely applied for donor screening.
STUDY DESIGN AND METHODS: The performance of an automated microparticle CMV IgG assay (Abbott AxSYM CMV IgG microparticle enzyme immunoassay [MEIA]) was compared with an established total antibody blood screening assay (Abbott CMV Total AB EIA). Sensitivity and specificity were assessed using 5050 random blood donors and 13 seroconversion panels. A risk analysis was undertaken to estimate the residual risk of transfusion-transmitted CMV (TT-CMV) from presumptive seronegative blood components.
RESULTS: The EIA achieved marginally (but not significantly) better resolved sensitivity (100%) than the AxSYM IgG assay (99.93%). The AxSYM IgG resolved specificity (99.34%) was superior to the EIA (96.4%). This superiority was maintained (98.61%) when a modified cutoff was applied to the AxSYM IgG assay to achieve 100 percent resolved sensitivity. The seroconversion sensitivities of the EIA and the AxSYM IgG were equivalent, detecting the same bleed as positive in the majority of the seroconversion panels tested. The median TT-CMV residual risk estimate for the two assays was approximately 1 in 66,000 (range, 42,000-165,000).
CONCLUSION: The AxSYM IgG MEIA is suitable for blood donor screening and was optimized by applying a modified cutoff of 9 AU per mL. The modeling predicts that implementing the AxSYM IgG assay would not negatively impact the already very low risk of TT-CMV associated with seronegative blood components in Australia.
Human cytomegalovirus (CMV) is a ubiquitous betaherpesvirus that generally causes asymptomatic infection in the 40 to 90 percent of individuals it infects.1 However, transfusion-transmitted CMV (TT-CMV) can cause serious morbidity and mortality in susceptible patients particularly when immunocompromised. At-risk patient populations include preterm CMV-seronegative neonates, seronegative recipients of autologous or allogeneic marrow or peripheral blood stem cell transplantation, solid organ transplant recipients, and CMV-seronegative AIDS patients.2-4 The primary mechanism for TT-CMV is thought to be the infusion of latently infected mononuclear cells; however, reinfection or reactivation in seropositive recipients has also been reported.5,6 The detection of CMV DNA in the plasma of some newly infected blood donors suggests the possibility that plasma viremia might also contribute to the incidence of TT-CMV.7
The incidence of TT-CMV in susceptible patients, which was reported in the range of 10 to 40 percent in the 1970s and 1980s, has declined substantially to 1 to 4 percent as a result of CMV antibody screening and more recently leukodepletion of donated blood.8,9 There has been substantial debate about the value of maintaining CMV antibody screening where blood components are also leukodepleted.10 The panelists of a Canadian conference held in 2001 considered that both offered the ability to significantly reduce the risk of TT-CMV but could not reach a consensus on whether CMV antibody screening was more effective than universal leukodepletion.11 However, in a subsequent meta-analysis, Vamvakas8 concluded that the available data from controlled studies indicated that in bone marrow transplant (BMT) recipients at least, CMV-seronegative components are more efficacious than white blood cell (WBC)-reduced blood components in preventing TT-CMV infection. Despite the recent development of CMV nucleic acid test (NAT) assays suitable for donor screening, they do not appear at present to offer a significant sensitivity advantage for detecting infectious blood components over serologic assays.7,12
TT-CMV infection subsequent to transfusion with seronegative blood components occurs predominantly as a result of “false-negative” donor serology. Such testing failures may occur for several reasons including test insensitivity or procedural errors or because of the existence of a 6- to 8-week “window period” (WP) in which CMV antibody is absent in the presence of CMV DNA in plasma and WBCs.13,14 Judicious test selection to minimize the rate of false-negative results is therefore an important TT-CMV risk reduction strategy. Since modern CMV antibody tests have significantly improved sensitivity and process control, presumably false-negative results are now most often associated with seroconverting donors in the WP.12 Minimizing the length of the WP is therefore important in CMV antibody assay optimization in respect of detecting potentially infectious (viremic) donations. Although CMV immunoglobulin G (IgG), immunoglobulin M (IgM), and immunoglobulin A (IgA) are all produced after primary CMV infection, it is generally accepted that in the majority of infections, IgG and IgM antibodies are produced almost simultaneously as the first detectable antibody markers.15 In immunocompromised patients the balance of evidence suggests that IgM antibody detection is not superior to IgG for the diagnosis of primary CMV infection. However, in the rare cases in which IgM is detected before IgG, the delay is small (1-2 days).15,16 In consideration of the requirement for optimal sensitivity, blood donor screening assays have traditionally been designed to detect at least CMV IgG and IgM with some also including IgA detection.17 The relatively high rate of false-positive results associated with IgM detection is however a disincentive to its inclusion in blood screening assays.18,19 There is now evidence that CMV antibody screening of blood donors can be performed reliably with highly sensitive assays based on detection of CMV-specific IgG only.20
The primary aim of this study was to assess the relative sensitivity and specificity of an automated IgG-only CMV screening assay against the existing total antibody (IgG, IgM, and IgA) assay to determine its suitability as an alternative for screening blood donors. To investigate if there was any identifiable sensitivity deficit associated with the lack of IgM detection on the candidate assay, the study protocol included parallel assessment of all samples with an IgM CMV antibody assay. This allowed an assessment of the relative sensitivity and specificity of the individual assays as well as a combined strategy of IgG and IgM antibody testing. The predicted impact of implementing the candidate CMV IgG–only assay in terms of the residual risk of TT-CMV was also estimated using a novel approach based on a published probabilistic model of WP infection.
MATERIALS AND METHODS
Donor population and CMV issue policy
The Australian Red Cross Blood Service (ARCBS) collects approximately 1.2 million homologous blood donations annually of which a subset (approx. 10%) are selected for anti-CMV testing to maintain an inventory of CMV-seronegative components. For indications where CMV-negative blood components are required, ARCBS currently recommends the following:21
- 1Select CMV-seronegative components whenever possible.
- 2If not available, leukoreduced components are considered to offer a high level of safety in preventing CMV transmission, but are not universally believed to be equivalent to CMV-seronegative components.
- 3Careful monitoring for CMV infection and disease in high-risk patients.
At the time of the study only a proportion of red blood cell (RBC) and platelet (PLT) components were leukoreduced at source. However, patients with a clinical requirement for CMV-seronegative blood components often also require leukoreduced components for other reasons. Thus, the majority of high-risk patients would have received RBC and PLT components that were both CMV-seronegative and leukoreduced (to below 1 × 106 WBCs per unit). By end of 2008, all RBC and PLT components in Australia will be provided leukoreduced at source.
Donors with an anti-CMV–negative history are retested at each donation and if the donor's CMV status remains negative associated RBC and PLT components may be issued as seronegative. If, however, the donor's profile changes to anti-CMV–reactive (presumed seroconversion), his or her computer record is annotated and future donations are not tested for CMV. CMV-reactive donors are not subjected to “confirmatory” testing and are not routinely notified of their anti-CMV status.
ARCBS random volunteer blood donors (n = 5050) attending to donate in Adelaide, Australia, were prospectively recruited and ethylenediaminetetraacetate plasma samples drawn for testing by an extended anti-CMV testing protocol that included:
- 1The Abbott CMV Total AB enzyme immunoassay (EIA; Abbott, Diagnostics Division, Abbott Park, IL);
- 2The Abbott AxSYM IgG microparticle enzyme immunoassay (MEIA; Abbott, Diagnostics Division); and
- 3The Abbott AxSYM IgM MEIA (Abbott, Diagnostics Division).
Before commencement of the study, the protocol was submitted to and approved by the ARCBS Human Research Ethics Committee. In accordance with the approval conditions, each recruited blood donor was provided with an information document and given the option of participation without the requirement of written consent. If a donor did not wish to be included in the study, he or she was requested to advise collection staff and samples were not taken.
The Abbott CMV Total AB EIA is licensed by the Food and Drug Administration for blood donor screening and was performed according to manufacturer's instructions using the Abbott Parallel Processing Center (Abbott, Diagnostics Division). The Abbott AxSYM CMV IgG and IgM assays are automated MEIAs with dedicated instrumentation. Both were performed using the Abbott AxSYM analyzer (Abbott, Diagnostics Division) according to the manufacturer's instructions.
Confirmatory testing was performed on specimens with AxSYM CMV IgG (15 AU/mL) and CMV Total AB EIA discordant results. These specimens were divided into seven categories, based on the combination of reactivity in each of the three assays, and were further tested and resolved as shown in Table 1. Discordant specimens were tested by the CMV IgG immunoblot, the CMV IgM immunoblot,22 or by the Enzygnost CMV IgA EIA (Behring Diagnostics, Marburg, Germany) as indicated. The sensitivity of the CMV IgG immunoblot was assessed against the Enzygnost CMV IgG EIA (Behring Diagnostics, Marburg, Germany) by parallel testing of 831 seropositive samples from pregnant women. The relative sensitivity of the CMV IgG immunoblot was 99.88 percent (830/831; 95% confidence interval [CI] = 99.33-100.00%; manuscript in preparation).
Table 1. Confirmatory testing strategy
|1||POS||NEG/POS/EQV||NON||CMV IgG immunoblot POS||REA|
| || || ||CMV IgG immunoblot NEG||NON|
|2||NEG||NEG||REA||CMV IgA EIA POS||REA|
| || || ||CMV IgA EIA NEG||NON|
|3||NEG (5-15)||NEG||REA||CMV IgG immunoblot POS and/or CMV IgA EIA POS||REA|
| || || ||CMV IgG immunoblot NEG and CMV IgA EIA NEG||NON|
|4||NEG||POS/EQV||NON||Donor recall 2-4 weeks later and retest:|| |
| || || ||AxSYM CMV IgG POS and/or CMV Total AB EIA REA||REA|
| || || ||AxSYM CMV IgG NEG and CMV Total AB EIA NEG||NON|
|5||NEG (5-15)||POS/EQV||NON||CMV IgG immunoblot POS and/or CMV IgM immunoblot POS||REA|
| || || ||CMV IgG immunoblot NEG and CMV IgM immunoblot NEG||NON|
|6||NEG (5-15)||POS/EQV||REA||CMV IgG immunoblot POS and/or CMV IgM immunoblot POS||REA|
| || || ||CMV IgG immunoblot NEG and CMV IgM immunoblot NEG||NON|
|7||NEG||POS/EQV||REA||CMV IgG immunoblot POS and/or CMV IgM immunoblot POS||REA|
| || || ||CMV IgG immunoblot NEG and CMV IgM immunoblot NEG||NON|
Specimens not available for confirmatory testing were not included in the calculation of resolved sensitivity, specificity, or agreement. Additional specimens that were discordant when the AxSYM CMV IgG cutoff was lowered to 9 AU per mL were not tested by the algorithm in Table 1. Recall donor confirmatory testing was performed on patient specimens drawn 2 to 4 weeks later from patients with initial negative AxSYM CMV IgG serologic results that were positive due to the addition of AxSYM CMV IgM results to the specimen interpretation, that is, samples that were initially AxSYM CMV IgG–negative but were positive by AxSYM CMV IgG plus IgM combined serology. Patient specimens drawn 2 to 4 weeks later were tested by the AxSYM CMV IgG and CMV Total AB EIA assays and were classified as resolved reactive if seroconversion occurred as measured by either assay, that is, the specimen was AxSYM CMV IgG–positive and/or CMV Total AB EIA-reactive. Recall specimens with negative and nonreactive results by the AxSYM CMV IgG and CMV Total AB EIA assays, respectively, were classified as resolved not reactive. Specimens not available for recall donor confirmatory testing were not included in the calculation of resolved sensitivity, specificity, or agreement.
Calculation of relative and resolved sensitivity and specificity
Relative sensitivity and specificity were calculated as described by Griner and coworkers.23 Resolved sensitivity and specificity were calculated after resolution of discordant specimens as described above.
Receiver operator characteristic analysis
To determine the impact of varying the manufacturer's stated cutoff (15 AU/mL) on AxSYM IgG assay sensitivity and specificity, a receiver operator characteristic analysis was undertaken for the range 5 to 15 AU per mL.24
The ability of each assay to detect CMV seroconversion was evaluated using 13 seroconversion panels (n = 52 samples) sourced from pregnant women who were routinely screened by CMV serology during gestation. To parallel routine donor screening where a sample is tested in singlicate, each panel sample was tested only once by each assay and the final status assigned based on this result. Notably the median bleed interval for all specimens in the panels was 52 (range, 4-1052) days, substantially longer than the 7 to 10 days usually targeted in commercial plasma seroconversion panels. However, the mean number of days from the last seronegative bleed to the first seropositive bleed ranged from 40 to 61 days depending on the assay used.
In addition to assessing the clinical sensitivity of each assay, the predicted impact of changing the CMV assay in terms of the residual risk of TT-CMV was also assessed (refer to Appendix for full details, available as supporting information in the online version of this article).
Fisher's exact test (two-sided) was used to evaluate the statistical significance of the resolved sensitivity and specificity data. The WP data were fitted to a normal distribution model. Shappiro-Wilk goodness-of-fit test for normality was performed and accepted at α = 0.05. The mean WP and 95 percent CIs were estimated.
Random volunteer blood donor testing
ARCBS random volunteer blood donors (n = 5050) were tested by the AxSYM CMV IgG and IgM and CMV Total AB EIA assays per approved protocol with the results shown in Tables 2A through 2D. A comparison of the AxSYM CMV IgG assay to the CMV Total AB EIA (Table 2A) generated a total of 103 discordant specimens (2.0%) with a CMV seroprevalence of approximately 60 percent. Following the confirmatory testing strategy in Table 1, 92 of the 103 discordant specimens were tested and resolved as follows. A total of 78 of the 85 specimens with AxSYM CMV IgG-negative/CMV Total AB EIA-reactive discordant results were available for confirmatory testing by a combination of CMV IgG and CMV IgM immunoblot and CMV IgA EIA. These 78 samples were placed into Category 2 (n = 48) and Category 3 (n = 30) based on their initial test results as shown in Table 1. The only difference between Category 2 and Category 3 samples with discordant results was that Category 3 discordant samples have an elevated but negative AxSYM CMV IgG result (5-15 AU/mL) when using the manufacturer's recommended cutoff, whereas there was no CMV IgG antibody detected by AxSYM CMV IgG in Category 2. In addition, there was also no CMV IgM antibody detected in these samples in both Category 2 and Category 3 by the AxSYM CMV IgM assay. Hence, Category 2 samples were only tested by the IgA EIA. However, Category 3 samples were tested by both the CMV IgG immunoblot and the IgA EIA since the AxSYM CMV IgG result was elevated, suggesting that these samples were discordant due to the presence of low levels of CMV IgG and/or CMV IgA. All Category 2 samples tested IgA-negative. Of the 30 Category 3 samples tested, 28 were negative by both CMV IgG immunoblot and CMV IgA EIA, 1 sample was positive for CMV IgG by immunoblot, and 1 sample was positive for CMV IgA by EIA.
Table 2A. Comparison of the AxSYM CMV IgG assay (15 AU/mL) to the CMV Total AB EIA and resolved interpretation
|AxSYM CMV IgG|| || || || || |
| Positive||2928||18||2946||97.18 (96.52-97.74)|| |
| Negative||85||2019||2104|| ||99.12 (98.61-99.48)|
| Total ||3013||2037||5050|| || |
|AxSYM CMV IgG|| || || || || |
| Positive||2928||14||2942||99.93 (99.75-99.99)|| |
| Negative||2||2095||2097|| ||99.34 (98.89-99.64)|
| Total||2930||2109||5039|| || |
Table 2B. Comparison of the AxSYM CMV IgG plus IgM assay to the CMV Total AB EIA and resolved interpretation
|AxSYM CMV IgG + IgM|| || || || || |
| Positive||2929||61||2990||97.21 (96.56-97.77)|| |
| Negative||84||1976||2060|| ||97.01 (96.17-97.70)|
| Total ||3013||2037||5050|| || |
|AxSYM CMV IgG + IgM|| || || || || |
| Positive||2928||54||2982||99.93 (99.75-99.99)|| |
| Negative||2||2052||2054|| ||97.44 (96.67-98.07)|
| Total||2930||2106||5036|| || |
Table 2C. Comparison of the AxSYM CMV IgG (9 AU/mL) assay to the CMV Total AB EIA and resolved interpretation
|AxSYM CMV IgG (9 AU/mL)|| || || || || |
| Positive||2943||46||2989||97.68 (97.07-98.18)|| |
| Negative||70||1991||2061|| ||97.74 (97.00-98.34)|
| Total ||3013||2037||5050|| || |
|AxSYM CMV IgG (9 AU/mL)|| || || || || |
| Positive||2930||29||2959||100.00 (99.87-100.00)|| |
| Negative||0||2054||2054|| ||98.61 (98.01-99.07)|
| Total||2930||2083||5013|| || |
Table 2D. Comparison of the CMV Total AB EIA assay to the resolved interpretation
|CMV total AB EIA|| || || || || |
| REA||2930||76||3006||100.00 (99.87-100.00)|| |
| NON||0||2033||2033|| ||96.40 (95.51-97.15)|
| Total||2930||2109||5039|| || |
Although the AxSYM CMV IgG assay failed to detect these two specimens during routine testing, the CMV IgG titers of these specimens were just below the assay cutoff of 15 AU per mL (14.0 and 10.7 AU/mL). Fourteen of the 18 specimens with AxSYM CMV IgG–positive/CMV Total AB EIA–nonreactive discordant results were tested by the CMV IgG immunoblot and were all resolved to be nonreactive, that is, all tested CMV IgG immunoblot–negative. To evaluate the potential benefit of testing random volunteer blood donors with both a CMV IgG and a CMV IgM test for disposition of blood units as CMV antibody–negative, a comparison was made between the AxSYM CMV IgG plus IgM combined assay interpretation and the CMV Total AB EIA. A specimen was considered positive by this combined interpretation if either assay was positive (equivocal AxSYM CMV IgM results were counted as positive) and negative if both test results were negative. The results are shown in Table 2B. The combined assay interpretation resulted in an additional 44 specimens scored as AxSYM CMV IgG plus IgM positive over the AxSYM CMV IgG–only interpretation. Owing to the concern of the potential high number of false-positive CMV IgM test results in a healthy blood donor population, specimens that were AxSYM CMV IgG plus IgM–positive were resolved by recall donor confirmatory testing. A total of 36 of the 44 donors were recalled 2 to 4 weeks later; their blood was drawn and retested by all three original CMV assays. After testing on the second specimen, the serologic test results indicated that no seroconversion to CMV IgG occurred in all of the 36 donors tested. Since seroconversion to CMV IgG–positive would have occurred in this time frame if these donors were in early primary CMV infection at the first donation, the original first specimens were resolved nonreactive by recall donor confirmatory testing. A comparison of the resolved interpretation results in Tables 2A and 2B shows that two specimens were not detected in this blood donor population by the AxSYM CMV IgG assay or by the combined interpretation of AxSYM CMV IgG plus IgM. These two specimens with false-negative results by the AxSYM CMV assays were the same specimens described earlier with elevated but negative AxSYM CMV IgG results (14.0 and 10.7 AU/mL). Receiver operator characteristic curves performed for the AxSYM CMV IgG assay cross at a cutoff between 8 and 9 AU per mL (plot not shown). Therefore, the effect of lowering the AxSYM CMV IgG assay cutoff from the manufacturer's recommendation of 15 to 9 AU per mL was analyzed in Table 2C. Lowering the assay cutoff to 9 AU per mL results in an additional 43 specimens scored as AxSYM CMV IgG–positive. Taking into account the confirmatory testing of specimens with discordant results previously performed with the assay cutoff at 15 AU per mL, there were no specimens missed by the AxSYM CMV IgG assay at a cutoff of 9 AU per mL relative to the CMV Total AB EIA.
The final assay comparison made was between the CMV Total AB EIA and the resolved interpretation as shown in Table 2D taking into account the previous confirmatory testing for resolution of discordant specimens. The resolved sensitivities and specificities for the CMV Total AB EIA, AxSYM CMV IgG assay, AxSYM CMV IgG assay (reduced cutoff of 9 AU/mL), and AxSYM CMV IgG plus IgM combined interpretation are compiled in Table 3. There were no significant differences between the resolved sensitivity for all four assays. In contrast, the resolved specificity for the AxSYM CMV IgG (99.34%) and reduced cutoff AxSYM CMV IgG (98.61%) assays was superior to CMV Total AB EIA (96.40%; p < 0.0001).
Table 3. Resolved sensitivity and specificity compared for AxSYM CMV assays versus CMV Total AB EIA
|CMV Total AB EIA||100.00 (99.87-100.00)|| ||96.40 (95.51-97.15)|| |
|AxSYM CMV IgG||99.93 (99.75-99.99)||0.50 (NS)||99.34 (98.89-99.64)||<0.0001|
|AxSYM CMV IgG (9 AU/mL)||100.00 (99.87-100.00)||NA||98.61 (98.01-99.07)||<0.0001|
|AxSYM CMV IgG plus IgM||99.93 (99.75-99.99)||0.50 (NS)||97.44 (96.67-98.07)||0.06 (NS)|
Seroconversion panel testing
To evaluate the relative seroconversion sensitivity of the CMV assays, 13 seroconversion panels (n = 52 specimens) were sourced from a Swiss pregnant women population and tested by all three assays. The data obtained with the AxSYM CMV IgG assay was evaluated with the reduced assay cutoff of 9 AU per mL. Of the 13 panels tested, in 7 of 13 panels the same serial bleed was detected as positive or reactive by all three assays. In 3 of 13 panels the first serial bleed was detected as positive by the AxSYM CMV IgM assay only, and the subsequent serial bleed in these panels was detected as positive and reactive by the AxSYM CMV IgG assay (9 AU/mL) and CMV Total AB EIA, respectively. There was 1 of 13 panels where the first serial bleed was detected as positive only by the AxSYM CMV IgG assay (9 AU/mL) with a sample-to-cutoff ratio (S/Co) of 1.3 while the corresponding S/Co result for the Total AB EIA was 0.82. The subsequent serial bleed in this panel was detected as positive and reactive by the AxSYM CMV IgM assay and CMV Total AB EIA, respectively. Finally, there were 2 of 13 panels where the first serial bleed was detected as positive by the AxSYM CMV IgM assay and CMV Total AB EIA. Notably, the Total AB EIA results for these two bleeds were very close to the cutoff with S/Co's of 1.0 and 1.24, while the respective AxSYM IgG (9 AU/mL) S/Co's were 0.21 and 0.49. The AxSYM CMV IgG assay (9 AU/mL) was positive on the subsequent serial bleeds. Analysis of these data in terms of residual risk of TT-CMV is described in the Appendix and the results are summarized in Table 4.
Table 4. Estimated residual risk of TT-CMV (from seronegative blood) for each assay
|CMV Total AB EIA||53.9 (22.7-85.2)||886.7||0.0115||1 in 66,410 (42,013-157,687)|
|AxSYM CMV IgG (9 AU/mL cutoff)||53.7 (21.7-85.7)||886.7||0.0115||1 in 66,657 (41,768-164,954)|
|AxSYM CMV IgG (15 AU/mL cutoff)‖||60.6 (28.9-92.2)||886.7||0.0115||1 in 59,068 (38,823-123,858)|
|AxSYM CMV IgG plus IgM||40.0 (15.7-64.3)||886.7||0.0115||1 in 89,487 (55,669-227,993)|
Blood donor CMV antibody screening remains an important strategy for the provision of CMV-“safe” blood components. Judicious test selection underpins the efficacy of this strategy and in the transfusion setting must strike a balance between optimal sensitivity and specificity. In this context, our study assesses the sensitivity and specificity of the AxSYM CMV IgG MEIA as a candidate blood donor screening assay in comparison with an approved CMV Total AB EIA used extensively for donor screening. Furthermore, it establishes a novel risk estimate for the residual risk of TT-CMV in Australia associated with transfusion of seronegative blood components.
In terms of clinical sensitivity based on performance in the resolved seropositive blood donor samples, the AxSYM CMV IgG demonstrated equivalence to the CMV Total AB EIA. Although the CMV Total AB EIA achieved a slightly better sensitivity (100%) than the AxSYM CMV IgG assay (99.93%), this difference was not significant. Notably the sensitivity of the AxSYM CMV IgG and IgM combined strategy (99.3%) was the same as that for the AxSYM CMV IgG assay alone. Furthermore, the sensitivity of the AxSYM CMV IgG assay could be increased to 100 percent to match the CMV Total AB EIA by modifying the assay cutoff from the recommended 15 AU per mL to 9 AU per mL. Importantly, this optimized sensitivity could be achieved whilst maintaining a superior specificity than the EIA. Our findings are consistent with those of Weber and coworkers20 who similarly investigated the comparative sensitivity of an IgG only and IgG plus IgM combination EIA. In their study, the IgG-only EIA was 100 percent sensitive in seropositive blood donors (n = 100) compared with 98.7 percent for the IgG plus IgM EIA but this difference was not significant.
In respect of assay specificity the AxSYM CMV IgG (99.34%) was superior to both the CMV Total AB EIA (96.4%) and the dual strategy of AxSYM CMV IgG plus IgM (97.44%). This superiority was maintained (98.61%) even when the modified cutoff was applied to achieve 100 percent sensitivity. While not providing any measurable improvement in sensitivity over the AxSYM CMV IgG assay alone, parallel screening using the AxSYM CMV IgG and IgM assays led to a marked decline in assay specificity generating an additional 44 presumed “false-positive” results. Importantly, there was no evidence of seroconversion among these samples where donor follow-up testing was possible. Consistent with other studies demonstrating that CMV IgM antibody tends to cross-react with other herpes viruses, we found 12 of 16 of these false-positive samples reactive for Epstein-Barr virus antibodies (data not shown). There are also a number of other conditions unrelated to CMV infection, which reportedly can result in false-positive IgM reactions including polyclonal antibody production during infection with other herpes viruses, autoimmune diseases, hepatitis B virus (HBV), and pregnancy.18
Although there is certainly a potential limitation of CMV IgG–only screening assays where detectable CMV IgM is produced first during seroconversion, in practice this appears to be a rare event. Supporting this we did not identify any individuals with isolated confirmed IgM antibody among our cohort of 5050 randomly selected donors. Several studies indicate that in the majority of cases, IgG and IgM appear to be produced concomitantly and where they are not the delay can occur either way (i.e., CMV IgG or IgM detectable first) and is small enough to be clinically insignificant. Weber and coworkers18 found CMV IgG and IgM concomitantly detectable in serial samples from five patients suffering from primary infection. In renal transplant recipients, CMV IgG seroconversion was observed to occur on average 1 day earlier than IgM.16 In a separate study by Lazzarotto and coworkers,25 the CMV IgG response was diagnostically superior to IgM for detecting CMV infection in liver transplant recipients.
In contrast, the superior performance of the AxSYM CMV IgM assay in our seroconversion panels did indicate a potential advantage of CMV IgM detection in early primary infection. The AxSYM CMV IgM assay was positive before the CMV Total AB EIA and AxSYM CMV IgG (9 AU/mL) assay in 3 and 5 of the 13 panels, respectively. Importantly though the AxSYM CMV IgG (9 AU/mL cutoff) and the CMV Total AB EIA demonstrated equivalent performance both detecting the first positive bleed in 10 per 13 seroconversion panels. In the remaining 3 panels, the first seropositive bleed was detected in 1 of 3 panels first by AxSYM CMV IgG and in 2 of 3 panels first by the CMV Total AB EIA. Assays that show this relative “flip-flop” pattern of first detection of antibody across different seroconversion panels are considered to have equivalent sensitivity. Notably, our panels were sourced from pregnant women, a population that may be considered to be partially immunosuppressed; therefore, the dynamics of seroconversion may not be equivalent to healthy blood donors. Accordingly, caution is required when interpreting the observed advantage of AxSYM CMV IgM from our data in the context of blood donor screening. Overall our data support the conclusion drawn by Weber and coworkers in their study,20 namely that blood donor screening for CMV antibody can be reliably achieved by detection of specific IgG only conditional on the use of a highly sensitive assay system.
Having established the clinical sensitivity and specificity of the assays we were interested in translating this performance into a residual risk estimate for TT-CMV. Our novel approach derives a risk estimate for each assay which is proportional to the duration of the antibody WP and therefore unique to each assay. Importantly, our risk estimate only considers the risk of transmission to a seronegative recipient from an infectious blood component resulting from primary (WP) CMV infection in a donor. The analysis does not consider other potential sources of TT-CMV which include CMV reinfection or reactivation in a seropositive recipient. The risk analysis indicates that a dual strategy of AxSYM CMV IgG and IgM carries the lowest residual risk of WP infectivity per transfused component (1 in 89,487), while the CMV Total AB EIA and AxSYM CMV IgG (9 AU/mL) are intermediate (1 in 66,410 and 1 in 66,657 respectively) and AxSYM CMV IgG (15 AU/mL) carries the greatest risk (1 in 59,068). The lack of comparable published data makes interpreting these estimates problematic. Considering that ARCBS issued approximately 138,000 CMV antibody-negative components in the 2006 to 2007 financial year (ARCBS unpublished) the risk estimate of approximately 1 in 66,000 for either the CMV Total AB EIA or the AxSYM CMV IgG at 9 AU per mL would predict the issue of on average one to three viremic components per annum.
To provide context around these estimates it is useful to compare them with the other transfusion-transmitted viral infections for which ARCBS screens. Although the AxSYM CMV IgG/CMV Total AB EIA point estimate for TT-CMV of approximately 1 in 66,000 is higher than the comparable ARCBS figures for human immunodeficiency virus (HIV; 1 in 9.2 million), hepatitis C virus (HCV; 1 in 6.4 million), HBV (1 in 633,000), and human T-lymphotropic virus (HTLV; 1 in 6.8 million),26 perhaps the most appropriate comparison would be the estimated frequency of issuing a potentially infectious unit. The predicted rate for CMV of 1 to 3 per annum is similar to that predicted for HBV where the residual risk of 1 in 662,000 in the context of approximately 1.2 million donations translates to approximately two to four potentially infectious components issued per annum (assuming an average of two components per donation). Whether or not CMV infection would establish itself in the recipient of a potentially infectious component is dependent on several factors including the viral load of the component and whether or not it is leukoreduced, as well as the immune status of the recipient.27
The rate of TT-CMV infection in Australian blood recipients is unknown but appears to be very low. Based on his meta-analysis of international published studies, Vamvakas8 concluded that the risk of transfusion-acquired CMV infection could be approximately 1.5 percent in recipients of CMV-seronegative components. Since our estimate is per component, not per recipient, it is not directly comparable unless the 1.5 percent derived in the meta-analysis is divided by the mean number of transfused components per recipient. Deriving the latter is problematic given that it varies markedly based on the recipient population. For this reason it may be more informative to focus on the risk in BMT recipients who constituted the majority of patients in the Vamvakas meta-analysis–derived figure of 1.5 percent. The mean number of transfusions per patient in the randomized control trial of Bowden and coworkers,28 which constituted 674 of 829 recipients was 24. Dividing the 1.5 percent then by 24, the estimated risk per component calculates as 0.0625 percent. The ARCBS residual risk estimate predicts a rate more than 40 times lower (0.0014%) but without incidence data from Australian recipient populations it is difficult to validate the accuracy of this estimate. Assuming that the risk estimate is accurate, then there are several possible explanations for the lower predicted rate of TT-CMV associated with seronegative components in Australia. First, the Vamvakas estimate is based on the rate of established infection in recipients which includes all sources of TT-CMV infection, whereas ours is restricted to the risk of issuing a potentially infectious unit from a seroconverting donor. Even if one considers only the latter, as noted previously there are a number of factors including component viral load, leukoreduction, and recipient immune status, which may impact whether or not a recipient develops overt CMV infection. Second, our estimate is based on the most up-to-date antibody screening techniques, whereas the Vamvakas estimate incorporates studies where less sensitive systems were used. Furthermore, the overall fidelity of the test system is improved by the process control inherent in fully automated instruments like AxSYM due to the virtual elimination of “human error.” This reduces the residual risk but the magnitude of this reduction is difficult to estimate.
There are a number of assumptions inherent in our modeling which impact the accuracy of the estimates: first, that it is appropriate to apply the incidence/WP model to primary CMV infection. Supporting this assumption, the central premise of the incidence rate/WP (Inc/WP) model, namely, that WP infection constitutes the majority of the risk,29 seems to hold true in the context of its application to primary CMV infection. This assertion is based on the following rationale. The current study uses fully (AxSYM) or semiautomated (EIA), highly sensitive testing systems that will minimize false-negative serology results due to testing limitations other than the absence of antibody in the WP. This contrasts with other dated studies using less-sensitive and more-error-prone antibody testing systems where latently infected “falsely negative” donors likely occurred at a higher frequency. Indirect evidence for the infrequency of false-negative serology results associated with latently infected donors can be drawn from studies assessing the rate of DNAemia in seronegative donors tested by more sensitive antibody assays. Combining the data from four such studies using validated CMV polymerase chain reaction (PCR) assays to detect DNA in whole blood or RBCs, the overall rate of detectable DNA among seronegative donors not in the WP was 0 of 954 (0%).12,30-32 In contrast, the rate of detectable plasma DNA in seroconverting (WP) donors from two studies was 3 of 260 (1.15%), indicating that they constitute a substantially greater risk. Our second assumption is that detection of CMV DNA in plasma correlates with competent virus and therefore potential infectivity. Published evidence supports this possibility given the documented detection of CMV DNA in both plasma and cellular components from persons with acute infection weeks to months before seroconversion.13,14,33 Our third assumption is that unlike traditional applications of the Inc/WP model, which assume that all donations made during the WP may be viremic and therefore infectious, we have chosen to correct for the rate of observed viremia in seroconverting blood donors. Inherent in this assumption is that this rate, which we derived from two published studies, is accurate and can be extrapolated to ARCBS blood donors. Should the derived rate underestimate the true rate of viremia during seroconversion, then our risk estimates will accordingly underestimate the residual risk. Our fourth assumption is that all cellular blood components (PLTs, RBCs, or whole blood) from a WP donation each carry the same potential to infect and do so 100 percent of the time. Although this assumption is conservative and almost certainly leads to an overestimate of the true residual risk, it is appropriate given the complex epidemiology of CMV and paucity of data on the relative infectivity of different components and the unknown minimum infectious dose in humans and is consistent with need for caution where blood safety policy is concerned. We further assume that our WP duration estimates derived from seroconverting pregnant women can be extrapolated to seroconverting blood donors. Supporting this assumption, our mean estimate of 54 days for the AxSYM CMV IgG (9 AU/mL) and CMV Total AB EIA falls within the range 42 to 56 days for seroconverting adolescents reported by Zanghellini and coworkers.13 Finally, we assume that the observed rate of seroconversion in the cohort of donors we followed from South Australia extends to the entire ARCBS donor population. In support of this, our observed seroconversion rate of 0.54 percent agrees well with two recent blood donor studies reporting rates of 0.55 and 0.8 percent.34,35
A recent study by Ziemann and coworkers35 provides provocative data suggesting that a radical new approach to CMV antibody screening could further improve CMV safety. In their study using a highly sensitive DNA PCR assay, they demonstrated that CMV DNAemia occurred in 2 of 68 (3%) of seroconverting donors but 0 of 598 donors who had been seropositive for 12 months or more. They interpret their data to imply that transfusion of leukoreduced blood from seronegative donors could in fact carry a greater risk of TT-CMV than transfusion with leukoreduced blood from donors seropositive for at least 12 months. In consideration of their observation that the risk of free CMV appears to be almost exclusively related to newly seroconverted donors, they suggest that “. . . it is necessary to discuss whether those donors should be transiently excluded from blood donations.” In an accompanying editorial, Drew and Roback33 rightly highlight the significance of these findings, specifically the potential for a new strategy for prevention of TT-CMV utilizing CMV antibody testing “in reverse” to identify established (>1 year) seropositive, rather than seronegative donors. They suggest that leukoreduction of units from these donors may be a safer and more cost-effective approach to preventing TT-CMV than current practice of filtering seronegative or CMV-untested units. Although recognizing that the study of Ziemann and coworkers35 provides important new data on the dynamics of DNAemia, they themselves suggest an urgent need for further studies to confirm their findings. In the light of our risk analysis, which indicates a very low WP TT-CMV residual risk in Australia, we contend that before considering any policy change to the current strategy to provide CMV-safe blood, there is a clear need to more accurately define the risk level associated with the proposed new strategy. In particular, the number of established seropositive donors screened for CMV DNA using assays of comparable sensitivity to that used in the Ziemann study needs to be increased to improve the level of confidence in respect of the rate of DNAemia in these donors. Notwithstanding this, as Drew and Roback point out it is likely the Ziemann study will ignite “an episode of energetic debate” surrounding the prevention of TT-CMV. Whatever the outcome, it seems CMV antibody screening in one form or another will be an ongoing requirement in the immediate future.
In summary, based on its performance in randomly selected blood donors and seroconversion panels from pregnant women, we conclude that the AxSYM CMV IgG assay is a suitable alternative to the CMV Total AB EIA for blood donor screening. Furthermore, the modeling we conducted predicts that implementing the AxSYM CMV IgG assay would not negatively impact the already very low risk of TT-CMV associated with seronegative blood components in Australia.