Which serological assay to use

A systematic review (see Web annex 5.3) compared the diagnostic performance (sensitivity, specificity, positive and negative predictive values) of commercially available serological assays (RDTs and EIAs¹) for the detection of HBsAg, when compared to a laboratory-based immunoassay reference standard (with or without a neutralization step). The review identified 30 studies (197–226) from 23 countries with varying prevalence of hepatitis B and evaluated 33 different RDTs. There were five studies of eight different EIAs against an immunoassay reference standard (214, 223, 227–229). A mixture of serum, plasma, capillary and venous whole blood specimens were used for RDTs, but only serum or plasma was used for EIAs. Seven studies assessed performance using capillary or venous whole blood (202, 206, 210, 215, 216, 218, 226). Sample size varied from 25 to 3928, and populations studied included healthy volunteers and blood donors, at-risk populations, pregnant women, incarcerated adults, and patients with confirmed hepatitis B.

RDTs. In 30 studies (197–226) of 33 different RDTs, the pooled clinical sensitivity of RDTs against different EIA reference standards was 90.0% (95% CI: 89.1–90.8) and pooled specificity was 99.5% (95% CI: 99.4–99.5) (Table 7.1).

TABLE 7.1

Summary test accuracy of RDTs and EIAs for HBsAg (different assay formats and comparators, populations and specimen types).

Brands: there was significant variation in performance between RDT brands and within the same brand of RDT, with sensitivity ranging from 50% to 100% and specificity from 69% to 100%.

Specimen type: results for capillary whole blood specimens were comparable to serum but less heterogeneous.

EIAs. In five studies (214, 223, 227–229) of eight EIAs there was wide variation in EIA performance, with sensitivity ranging from 74% to 100% and specificity from 88% to 100%. The pooled sensitivity was 88.9% (95% CI: 87–90.6) and pooled specificity was 98.4% (95% CI: 97.8–98.8).

RDTs and EIAs in HIV-positive persons. Five studies (212, 214, 215, 218, 222) evaluated three different RDTs against different EIA reference standards. The pooled clinical sensitivity of RDTs was 72.3% (95% CI: 67.9–76.4), but specificity was 99.8% (95% CI: 99.5–99.9), compared to a pooled clinical sensitivity and specificity of 92.6% (95% CI: 89.8, 94.8) and 99.6% (95% CI: 99, 99.9), respectively, among HIV-negative persons. Possible explanations for this reduced sensitivity include an increased incidence of occult hepatitis B in HIV-positive persons (i.e. presence of HBV DNA with undetectable HBsAg levels, such that HBsAg might not be detected using the RDTs evaluated), and the use of tenofovir- or lamivudine-based antiretroviral regimens, which are active against HBV and may suppress HBV DNA and HBsAg levels. In the one study (214) that evaluated three EIAs against an EIA reference with neutralization, the overall pooled sensitivity in HIV-positive individuals was 97.9% (95% CI: 96.0–99.0) and specificity was 99.4% (95% CI: 99.0–99.7), suggesting that EIAs perform better in HIV-positive persons.

Analytical sensitivity/limit of detection. The analytical sensitivity or limit of detection (LoD) is another important performance criteria, but there were insufficient data in the included studies to undertake a systematic comparison. However, no RDTs met the levels of analytical sensitivity (i.e. LoD of 0.130 IU/mL) required by the European Union through its Common Technical Specifications. Data from WHO prequalification assessment studies indicate that the LoD of EIAs for HBsAg was 50–100-fold better compared to RDTs (230). However, despite this difference in analytical sensitivity, clinical sensitivity is unlikely to be greatly reduced because the vast majority of chronic HBV infection is associated with blood HBsAg concentrations well over 10 IU/mL. This is important, as it has been suggested that false-negative RDTs for HBsAg are due to low HBsAg viral load levels, the presence of HBsAg mutants or specific genotypes, and the use of lamivudine- or tenofovir-based ART regimens (208, 214, 216, 230).

The overall quality of the evidence for the recommendation of which serological assay to use was rated as low to moderate, with downgrading mainly due to serious risk of bias based on cross-sectional study design, and heterogeneity in results.

Which testing strategy to use

No studies were identified that directly compared the diagnostic accuracy of a one- versus two-assay serological testing strategy in high- and low-prevalence settings (see Web annex 5.5). A predictive modelling analysis was therefore undertaken, which examined diagnostic accuracy of a one- or two-assay strategy based on a hypothetical population of 1000 individuals across both a range of HBsAg seroprevalence levels (10%, 2%, 0.4% representing typical high-, medium- and low-seroprevalence settings or populations, respectively) and a range of assay performance characteristics (sensitivity of 98% and 90%, and specificity of 99% and 98% derived from the systematic review pooled sensitivity and specificity for HBsAg RDTs.

Prevalence had a strong impact on the PPV and the ratio of true-positive to false-positive results (see Web annex 6.1). The introduction of a second assay of similar sensitivity to be applied to all specimens reactive in the initial serological assay provides substantial potential gains in the PPV across all prevalence levels (>97%), but particularly at a low prevalence (0.4%) and with an assay that has a lower specificity.

The overall quality of the evidence for the recommendation on use of a one- or two-assay serological testing strategy was rated as low, as this was based on predictive modelling simulation and hypothetical scenarios.

Balance of benefits and harms

Use of EIAs. In settings where existing laboratory testing infrastructure is available and there is good access to laboratory services, EIAs were recommended as the preferred testing method for several reasons:

1. Although RDTs and EIAs for HBsAg had similar clinical sensitivity and specificity when compared to an EIA reference standard, the sensitivity of different RDTs was highly variable, and some RDTs had suboptimal sensitivity.
2. In HIV-infected individuals, clinical sensitivity of RDTs was poor (72.3%) and appears to be better for EIAs.
3. The analytical sensitivity is much higher for EIAs (50- to 100-fold higher). The benefit of more analytically sensitive assays with better limits of detection is that it improves detection in persons with primary infection, and in individuals in whom HBsAg levels are extremely low.
4. A confirmatory test using a neutralization step can be incorporated into laboratory-based EIAs.
5. Testing using laboratory-based EIAs can be automated and may be more appropriate and cost–effective in settings where there are many tests being performed per day (>40 per day per operator).

Balance of benefits and harms

When to use a one-assay strategy. A one-assay testing strategy is applicable to most testing settings in resource-limited countries based on simplicity and in populations where prevalence is ≥0.4%, and so PPVs are high.

The Guidelines Development Group made an overall conditional recommendation for a one-assay serological testing strategy to diagnose chronic HBV infection based on low-quality evidence for the following reasons:

1. This approach will efficiently identify (rule in) most individuals likely to be infected and in need of further evaluation, and will rule out those who are uninfected.
2. Although a one-assay serological testing strategy has a lower PPV than a two-assay serological testing strategy, particularly at lower levels of prevalence (0.4% and 2%), and will therefore generate more false-positive results, the Guidelines Development Group considered that the consequences of this would not be clinically significant. This is because all HBsAg-positive patients will have further evaluation with staging of liver disease and HBV DNA measurement to assess eligibility for treatment (i.e. presence of cirrhosis or evidence of raised HBV DNA levels). Therefore, no patient would be initiated on lifelong antiviral therapy on the basis of a single serological test.
3. The Guidelines Development Group noted that it is also common practice in many settings to perform a second test after 6 months to confirm a diagnosis of CHB and so distinguish it from acute hepatitis B. This provides an additional approach to confirm a diagnosis of chronic hepatitis infection.
4. If one uses a first test with high specificity then very few would require a second test.
5. It would considerably simplify the process of testing and reduce costs, especially if delivered at the point of care.
6. More rapid reporting of test results (ideally same day) will help improve access and linkage to care.

When to use a two-assay strategy. The Guidelines Development Group made a conditional recommendation to consider a second serological assay in very low-prevalence settings (<0.4%) to improve the PPV. In low prevalence settings, there will be more false-positive than true-positive results with a single serological assay, even with a test of 99% specificity. Employing two assays with a specificity of around 99% increases the ratio of true-positive to false-positive diagnoses from 0.2 to 32–40.

The recommendation is to confirm with a neutralization step if using laboratory-based immunoassays for detection of HBsAg, as per the assay manufacturer's instructions. Where an RDT for HBsAg is used and no neutralization reagents are available or for EIAs with no neutralization reagents, a second different RDT assay may be used (231). However, there has been limited evaluation of the added value of a second RDT, and there are several challenges: (i) “different” RDT assays may fundamentally be the same, and therefore prone to similar inaccuracies and false-positive reactions; (ii) if the analytical or clinical sensitivity of the assay used is poor (high LoD), then a larger proportion of individuals who are truly HBsAg positive will not be identified, regardless of whether a one- or two-test strategy is used.

Acceptability, values and preferences

In a values and preferences survey among 104 respondents from 43 (20 high-income, 23 low- and middle-income) countries, overall, there was strong support from patient groups for simplified testing strategies that would improve access to testing including for high-risk groups. Seventy-seven per cent expressed a strong preference for a one-serological assay testing strategy with same-day results using RDTs to reduce loss to follow up.

Feasibility

In a survey of programmatic experience with hepatitis testing across 19 LMICs, implementing partners reported widespread use of RDTs in all settings, and use of a single HBsAg RDT assay by 68% of respondents.

Resource considerations

The reagent costs for HBsAg assays are similar for RDTs (between US$ 0.95 and US$ 3.00) and EIAs (between US$ 0.40 and US$ 2.80). High-throughput EIAs require additional laboratory infrastructure and equipment, and precision and expertise in operation. In contrast, RDTs do not require capital investment in laboratory infrastructure, and so there is a concurrent reduction in maintenance costs for equipment.