Despite a considerable reduction of the risk of HBV-infected blood donation entering blood supply (residual risk) due to improved screening by HBV NAT in the developed countries, the bulk of the people with HBV living in the developing countries still needs to be screened by serologic tests such as HBsAg and anti-HBc. Many of these countries lack resources for implementing NAT and are likely to remain so in the next decade or longer, thus depending on the HBV residual risk monitoring based on serologic testing and corresponding estimation methods. This paper reviews main HBV residual risk findings worldwide and the methods based on serology used for their calculation with repeat donors, as well as their extension to the first-time donors. Two artificial datasets with high (4.36%) and low (0.48%) HBV prevalence were generated to test the performance of five methods: the original incidence/window-period model based solely on HBsAg, its modification by Soldan in 2003, the Müller-Breitkreutz model, the HBsAg yield model, and its extension to include anti-HBc seroconversions within a year. The last model was closest to the true values of residual risk and had smallest variation of the estimates in both high and low prevalence data. It may be used for residual risk evaluation in relatively small samples, such as regional blood banks data.
The consequences of infection by Hepatitis B virus (HBV) remain among the most devastating ones for an immunopreventable disease, particularly in the developing countries. The World Health Organization (WHO) estimates of the disease burden worldwide are widely cited in the literature. A decade ago, three quarters of the world population lived in the areas of high HBV prevalence [
The above situation has led to a paradoxal situation where the disease burden of HBV infection which can be prevented by affordable vaccine remains grossly underestimated, even when HBV is one of the principal causes of mortality among infectious diseases. Apart from the relatively low incidence (<1%) of fulminant hepatitis B with high mortality rates, most of the HBV-related deaths are attributed to the end-point chronic diseases, mainly liver cancer and cirrhosis. In the beginning of this century, the HBV residual risk in Europe was estimated to be the cause of almost 1% of posttransfusion deaths due to a liver disease, mostly because of a fulminant form of the hepatitis B [
It is against this general background of the public health policies to combat HBV that the measures to reduce its transmission by blood transfusion and organ transplantation need to be evaluated. Although serologic screening for HBV had been introduced before HIV pandemic took place, it was the latter that increased the general public perception of the pressing need to improve blood safety. During the decade of 1990, all developed and the majority of the developing countries have improved HIV screening with p24 component added to the enzyme immunoassays. The second half of the decade was marked by the development of nucleic acid testing (NAT) technology in blood bank setting, mainly for HIV and hepatitis C virus (HCV), but also for HBV in the developed countries [
NAT era has opened the possibility to directly verify the number of NAT-positive and serology-negative blood donors, denominated as “NAT yield”, and to calculate its cost-utility with greater precision [
The text that follows reviews some major issues in the socalled residual risk estimation for HBV, which refers to the risk of HBV-infected donation entering the blood supply, trying to show geographical diversity and common grounds of worldwide experiences in dealing with this issue. It also provides an example of how to calculate the risk estimates with a variety of methods and shows their performance using two artificial data sets: one with high and the other with low HBV prevalence. Particular attention is given to the methods suitable for relatively small samples (of the order of 100.000) in low versus high HBV prevalence settings, in order to encourage their use for systematic risk monitoring in regional blood banks.
This section reviews the reports on the HBV residual risk from all over the world. As the majority of the reports referred to the developed countries, a separate review was made regarding the level of economic development. By and large, this division also corresponds to the levels of HBV prevalence; important exceptions to the rule are also reviewed. An effort was made to organize the major results by continents, countries, and sometimes even by regions within a country when such data were available. The testing algorithms for HBV prevalence were specified in most cases.
The USA have been leading the residual risk research since the REDS study provided its initial impulse [
In Canada, HBV incidence and residual risk were estimated at 12.4 and 1.4 per 100.000 donors, respectively [
In Europe, a mathematical model of residual risk and NAT yield for HIV, HCV, and HBV projected the yield of 1.2 per million by ID (individual donation) NAT for the year 1997, considered a rather small gain [
Other European countries produced their HBV residual risk estimates too. In Spain, NAT screening for HIV and HCV was introduced in 1999 and halved residual risk per million from 18.67 to 9.78 during the 1997–2002 period [
In Italy, first estimates for the second half of the 1990 decade reported HBV incidence of 10 per 100.000 and associated residual risk of 1 : 62.500 [
The Swiss blood bank estimates 1 : 115.000 for the HBV residual risk in the first years of the 2000 decade were relatively high [
In Germany, minipool NAT with 96 samples was estimated to yield 1 : 600.000 over HBsAg in 1997, with perceived utility of anti-HBc testing in reducing the HBV residual risk [
In France, NAT for HIV and HCV was introduced in 2001 and yielded one infected case per 200.000 donations in the first decade of its implementation [
In England, the HBV residual risk was hugely reduced from 1 : 260.000 in 1993 to 1 : 8,000.000 in 2001 [
In Japan, OBI has become a great concern for safety of the blood supply in the last decade. By analyzing repeat-donor repository samples for the 1997–2004 period, HBV ID NAT found 1.08% reactive samples, 60% of which had low anti-HBs titer and were considered potentially infectious [
Taiwan is a rare developed country with relatively high HBV prevalence. In 2000, a six-month followup of a cohort of blood recipients screened solely by HBsAg showed posttransfusion transmission rate of 20 per 100.000 confirmed by HBV-DNA-positive results, providing an argument for HBV NAT in highly endemic areas [
In the Republic of Korea, the HBV residual risk remained stable during the 2000 decade, with 1 : 45.896 in the beginning and 1 : 43.666 by the end of the decade [
The UAE introduced HBV NAT in 2008. Before its introduction, the HBV residual risk per million was estimated at 1.41 as compared to 0.92 during the HBV NAT era [
In Australia, the HBV residual risk models produced an estimate of 1 : 483.000 in the beginning of the 2000 decade, with predicted HBV NAT yield of one in million [
Among developing countries, the African continent has shown the most dramatic situation regarding residual risk not only for HIV but also for HBV. In the Sub-Saharan Africa, about half of the blood donors were deferred because of HBsAg-reactive rapid test result [
South Africa was the first African country to introduce HBV ID NAT screening, thus being able to observe in the first year of its implementation a yield of 1 : 36.612 for HBV and 1 : 5.200 for OBI among HBsAg-positive donors [
In the Middle East, an analysis from Egypt showed that, in highly endemic region with anti-HBc prevalence of 7.8%, adding this marker to routine HBsAg screening of blood donors yielded 0.5% HBV-DNA-positive test results [
In China, residual risk of 1 : 17.501 was calculated in Shenzhen, with predicted yield per million donors of 6.9 with more sensitive HBsAg tests, compared to 9.5 for MP NAT and 28.3 for ID NAT [
In Latin America, Brazil has been the leading country in residual risk research, including the HBV-related one. The first didactic introduction of the incidence/window-period model and the residual risk calculation appeared in 1998 [
The Retroviral Epidemiology Donor Study (REDS) has extensively used the incidence/window-period model [
A synthesis of the methods for calculating the HBV residual risk based on routinely used serological markers HBsAg and/or anti-HBc (IgG + IgM) is presented (Table
Principal methods for the calculation of HBV residual risk based on serological markers HBsAg and/or anti-HBc.
Method (reference) | Probability of HBV infection | Incidence type | Probability of window-period donation | WP (days) |
---|---|---|---|---|
Standalone HBsAg [ |
|
Rate | 0.7 ( |
59 |
Müller-Breitkreutz model [ |
|
Cumulative | WP/Mdn( |
59 |
HBsAg yield [ |
|
Cumulative | WP/Mdn( |
44 |
HBsAg and anti-HBc yield [ |
|
Rate | 1 if anti-HBc positive, |
44 |
Modified standalone HBsAg [ |
|
Rate | WP/Mean ( |
59 |
WP: duration of immunologic window period.
IDI: interdonation interval (between last seronegative and first seropositive donation).
Pyrs: person-years at risk for HBV infection (
Mdn: median.
The following section reviews main models proposed to estimate the HBV RR.
This method is based solely on HBsAg screening and is restricted to repeat blood donors. It includes an adjustment proposed by Korelitz and colleagues [
The adjustment factor for the incidence density (rate) is then
The probability of an infected repeat donor may be estimated by cumulative incidence, that is, the proportion of incident cases among all repeat donors (
Another method for estimating HBV incidence and residual risk was named “HBsAg yield method” [
Recently, an extension of the yield method was proposed to include anti-HBc (IgM and IgG) seroconverting donors within last year in addition to the HBsAg only seroconverters [
There are several additional methodological issues to be considered. First, the misclassification of seroconverting cases depends on their confirmation algorithm. For a HBsAg reactive test, it may require the same result on subsequent independent blood samples or anti-HBc or HBV NAT positive results [
For the first-time donors, a variety of methods to calculate the HBV seroconversion risk ratio of these to repeat donors were reviewed by Soldan et al. [
Most of the above estimation methods have been developed for the HIV whose pandemic circulation could be estimated with reasonable precision for the purpose of these calculations. However, no such analogy exists for the HBV, thus assuming that the time of risk for HBV starting at birth is heavily dependent on the HBV vertical transmission rate which varies hugely between the countries and regions. For example, in the countries with low HBV prevalence and universal child vaccination, the risk of HBV infection in childhood is extremely low as compared to hyperendemic areas with high vertical transmission rate. Given all these uncertainties regarding the first-time donors time at risk for HBV infection, a simulation of various parameters seems a sensible approach [
This section describes a simulation study to test the performance of the five incidence/window models based on routine serologic screening (Table
Two artificial datasets were created in order to test the accuracy of the selected incidence/window-period models based on serological markers HBsAg and anti-HBc (Table
Simulated data: serologic profile and followup for high versus low HBV prevalence datasets.
First-time | Repeat | ||||
---|---|---|---|---|---|
|
Prevalence (%) |
|
Prevalence (%) | Total years at risk | |
High HBV prevalence: serologic profile | |||||
HBsAg−, anti-HBc− | 47820 | NA | 47820 | NA | 23586.63 |
HBsAg−, anti-HBc+ | 4000 | 7.70 | 200 | 0.42 | 50.43 |
HBsAg+, anti-HBc− | 100 | 0.19 | 20 | 0.04 | 3.29 |
HBsAg+, anti-HBc+ | 30 | 0.06 | 10 | 0.02 | 1.38 |
| |||||
All | 51950 | 7.95 | 48050 | 0.48 | 23641.73 |
| |||||
Low HBV prevalence: serologic profile | |||||
HBsAg−, anti-HBc− | 49587 | NA | 49977 | NA | 24634.49 |
HBsAg−, anti-HBc+ | 400 | 0.800 | 20 | 0.040 | 4.38 |
HBsAg+, anti-HBc− | 10 | 0.020 | 2 | 0.004 | 0.47 |
HBsAg+, anti-HBc+ | 3 | 0.006 | 1 | 0.001 | 0.14 |
| |||||
All | 50000 | 0.826 | 50000 | 0.046 | 24639.48 |
Since residual risk in the first-time donors has been found at a considerably higher level compared to the repeat donors in the vast majority of the studies, the former were set to approximately 16 times higher HBV (either marker) and about 4 times higher HBsAg prevalence than the latter (Table
Total and person-years of interdonation interval (IDI) were 23693 for high and 24646 for low HBV prevalence, respectively. The largest parts of these totals were contributed by seronegative repeat donors (23587 and 24634 person-years), whereas rare HBsAg-positive and anti-HBc-negative donors contributed with 5.6 and 0.6 years for high versus low prevalence scenario. With anti-HBc-positive seroconversions in the last year counted in as incident HBV cases in the extended yield model, these figures increased to 94 and 11 person-years, in the same order. The median IDI were 100 and 110 days for the HBsAg seroconverting donors in low and high HBV prevalence groups, respectively.
As all HBV cases and their time at risk were true by definition, so was the HBV incidence calculated from these data, otherwise uncertain with real-life data. The discrepancy between the true and model-based incidence produced bias estimates for each incidence/window-period model (Table
HBV incidence adjusted for the probability of detecting transient HBsAg marker (PWP) and residual risk (RR) for repeat donors.
HBV prevalence | Method (reference) | Probability of HBV infected1 | PWP | RR (95% CI) |
---|---|---|---|---|
High | True values | 0.0097286 | 1.0000 | 1 : 636 |
Standalone HBsAg [ |
0.0012664 |
0.2600 | 1 : 1270 | |
Müller-Breitkreutz [ |
0.0047867 |
0.5906 | 1 : 354 | |
Modified standalone HBsAg [ |
0.0012664 |
0.5506 | 1 : 1434 | |
HBsAg yield2 [ |
0.000416 | 0.2991, 0.3789 | 1 : 7092 | |
HBsAg and anti-HBc yield [ |
0.0092818 |
0.99 | 1 : 667 | |
| ||||
Low | True values | 0.0009438 | 1.0000 | 1 : 6555 |
Standalone HBsAg [ |
0.0001217 |
0.19856 | 1 : 1632 | |
Müller-Breitkreutz [ |
0.000600 |
0.5444 | 1 : 30615 | |
Modified standalone HBsAg [ |
0.0001217 |
0.5444 | 1 : 15093 | |
HBsAg yield2 [ |
0.000400 |
0.2731, 0.3460 | 1 : 8052 | |
HBsAg and anti-HBc yield [ |
0.000893 |
0.99 | 1 : 6929 |
2Instead of 95% CI, a range between the extremes of the window period (30–38 days) is given.
For the first-time donors, no assumption was made about their HBV infection date in this paper, so no true time with HBV was known and therefore no time at risk or incidence rate could be calculated. It is possible to speculate that the first-time donors may have the same distribution of time at risk as the repeat donors and extrapolate the average (mean or median) of the latter to the former. However, this assumption may be unrealistic for many blood bank settings. Therefore, only the ratio of HBsAg-seroconverting fractions and the socalled yield rate ratio [
For the first-time-to-repeat-donor risk ratio, the variation of the HBV residual risk estimate was calculated by substituting the central estimate by the extremes of the plausible HBV seroconversion window period (30–38 days) as in the source paper [
Data simulation and all calculations were performed by Stata software [
Main results of simulated residual risk for HBV according to the estimation methods and prevalence level are presented (Table
For the high HBV prevalence (4.36%) data, both standalone [
For the low prevalence (0.48%) scenario, the standalone HBsAg model [
For the first-time donors, the ratio of HBsAg-seroconverting fractions and the socalled yield rate ratio [
HBV residual risk (RR) estimates for the first-time donors.
Method | HBV prevalence | First-time donors | Repeat donors | IR (variation)c | RR for the first-time donors (variation)d | ||||
---|---|---|---|---|---|---|---|---|---|
|
|
(%)b |
|
|
(%)b | ||||
HBsAg yield | High | 100 | 51950 | 0.1925 | 20 | 48050 | 0.0416 | 4.62 | 1 : 1533 |
Low | 10 | 50000 | 0.0200 | 2 | 50000 | 0.0040 | 5.00 | 1 : 1610 | |
| |||||||||
HBsAg and anti-HBc yield | High | 130 | 51950 | 0.2500 | 30 | 48050 | 0.0624 | 4.01 |
1 : 166 |
Low | 13 | 50000 | 0.0260 | 3 | 50000 | 0.0060 | 4.33 |
1 : 1599 |
aNumber of HBV-seroconverting donors.
bPercentage of HBV-seroconverting donors.
cIC: Incidence ratio for the first-time to repeat donor; no variation was calculated for the yield method whereas 95% CI of the risk ratio was used for the HBsAg and anti-HBc yield method.
dMultiplies of the plausible range for the yield method versus 95% CI of the risk ratio for the HBsAg and anti-HBc yield method.
The incidence ratio estimates for the first-time to repeat donors were similar for both methods, but corresponding residual risk estimates remained much further apart (Table
Comparison of HBV residual risk (RR) estimates for the first-time and repeat donors.
Residual risk estimation method | HBsAg yield | HBsAg and anti-HBc yield | ||
---|---|---|---|---|
HBV prevalence | High | Low | High | Low |
Repeat donor RR | 1 : 7092 | 1 : 8052 | 1 : 667 | 1 : 6929 |
First-time donor RR | 1 : 1533 | 1 : 1610 | 1 : 1599 | 1 : 1610 |
Estimated RRRa | 12.82 | 5.00 | 4.08 | 4.33 |
True RRRb | 16.56 | 17.96 | 16.56 | 17.96 |
aResidual risk ratio of first-time to repeat donors.
bBest estimate based on HBV prevalence ratio of first-time to repeat donor.
Both methods grossly underestimated the “true” estimate, thus illustrating the difficulties in extrapolating the residual risk from repeat to first-time donors solely on the basis of their HBV prevalence ratio. Better estimates remain a challenge for the researchers in this area.
The utility of anti-HBc marker in HBV screening of blood donors has been pointed out in various publications [
Although by the first half of the 2000 decade NAT screening for HIV and HCV showed poor yield in many countries which had implemented it for routine blood screening [
A range of methodological improvements for estimating the HBV residual risk deserves to be mentioned. The shift to probabilistic modeling of residual risk parameters [
However, for the vast majority of HBV infections located in the developing countries, routine serologic screening with HBsAg and anti-HBc remains the best affordable strategy for the time being and probably for at least a decade ahead. As the countries that have initiated universal child vaccination against HBV did so a decade or two ago, the influx of the vaccinated blood donors has only started to reduce the HBV residual risk very slowly due to the vast majority of unvaccinated donors. Targeting blood donors with HBV vaccine has been recommended to increase the blood safety in such situation [
Whatever the national strategy to increase the blood safety is, it is essential to base it on sound epidemiological data and systematic monitoring of the residual risk. To that end, extended yield model including both HBsAg and recent anti-HBc seroconversions in repeat donors provides a simple, accurate, and affordable method. It is suitable for smaller blood bank settings of the order of hundred thousand donors as opposed to huge samples of the order of several million donors employed by the other methods; all of which were developed in the developed countries with low HBV prevalence. The extended yield method was shown to both reduce the bias and improve the HBV residual risk precision, so it may be sensitive enough for monitoring the residual risk time trend in a particular setting.
Although the criterion for the definition of HBV prevalence as high, intermediate, or low in the general population (WHO) are often extended to the blood donor population, it should be avoided for two reasons. First, the latter population should have a considerably lower risk for a variety of blood-borne diseases, with HBV at the top of this list, as it is predominantly sexually transmitted and therefore a marker of risky sexual behavior. Second, HBV infection is preventable by immunization affordable to virtually every country, so the criterion of what is high or low prevalence should be set against the real opportunity to reduce its prevalence to the order of few cases per 100.000 in the adult population and perhaps even per million in the blood donor population. Some developed countries like Great Britain have achieved this level of HBV risk reduction [
There are several limitations of this work which should be borne in mind. First, it is beyond its scope to address the question of HBV infectivity regarding either the phase of HBV infection or the host factors. Residual risk is limited to the probability of an infected blood unit being erroneously considered noninfectious and thus transfusable, so it fell short of taking into account other factors related to posttransfusion HBV infection, let alone its severity and clinical evolution.
Second, many incidence/window-period model parameters are pretty variable both between regions and between blood donors but their variation is typically underestimated in the calculation. For example, the repeat donors vary considerably regarding their IDI and consequently the time at risk of HBV infection. However, only average IDI is normally used for calculations, so this work followed the same tradition. IDI distribution is often heavy tailed to the right because of a small number of large values, thus adding the choice of the distribution to the uncertainties of the model. Again, this issue was ignored here to make the results comparable to those published in the literature.
Third, as the host response to HBV infection is highly variable, so is the time for HBsAg to reach the threshold of detectability for a given sensibility of a screening test, as well as the duration of the of socalled eclipse phase [
Fourth, the extrapolation of the HBV residual risk from the repeat to the first-time donors is based on a number of simplifying assumptions mentioned earlier, whose impact on the estimate variation has not been quantified except in a very simple way by applying the 95% CI of the relative risk of HBV seroconversion in the extended yield model. The inference about incidence rate of a viral disease based on its prevalence is prone to considerable amount of error [
All of these limitations result in a considerable underestimation of the residual risk variation by traditional methods of calculation. This may seem paradoxal as many published results already present wide confidence intervals even in the samples of the order of hundred thousands due to the very low frequency of seroconversion among repeat donors. However, virtually all of them account only for the random variation in the number of seroconversions and leave out other sources of variation discussed above. Simulation of a joint impact of various sources of variation was used a decade ago [
The underestimation of the HBV residual risk variation may have different impact for low versus high prevalence area. Most of the research has been done in the developed countries with low HBV prevalence, where the number of HBV seroconversions among repeat donors is very rare. On the other hand, with mean being equal to the variance in Poisson distribution used to model the variation in the number of seroconversions, high HBV prevalence makes larger means of seroconverting donors per unit time more likely, as well as their variance.
Although artificial data are the only way to guarantee certainty of a residual risk model parameters, eventually the utility of any model has to be tested with real data. It is important to monitor the HBV residual risk on a regular basis and look for its regional and sociodemographic variations in order to better understand the blood donor behavior and to direct preventive actions accordingly. Routine gathering of relevant epidemiologic data is necessary for both the blood donor and the general population to provide a solid basis for the incidence/window-period models. The knowledge of hepatitis B viral dynamics gathered with evermore sensitive NAT can be used to further improve the incidence/window-period model calculations and validate its robustness and precision.
Despite its limitations compared to the HBV DNA testing, the incidence/window-period model continues to play an important role in blood screening. It is the only affordable method for many developing countries which contain vast majority of the people with HBV in the world. Even for the countries with resources for the HBV DNA blood screening, serologic testing with HBsAg and anti-HBc still prevents some posttransfusion hepatitis B caused by OBI and chronic HBV carriers, although deferral of the donors with transfusable blood may be high. More sensitive serologic tests and targeting HBV vaccination of blood donors could be cost-effective alternatives in some cases [
In conclusion, the incidence/window-period model for HBV has been around for 15 years and remains relevant for evaluating the HBV residual risk in vast majority of the developing countries which concentrate the bulk of the people with HBV worldwide. The model has been improved over the years, benefiting from the insights on viral dynamics based on the research with highly sensitive HBV NAT. An adaptation of the incidence/window-period model termed “extended yield model”, which considers both HBsAg and recently (within last year) seroconverting anti-HBc repeat donors, seems particularly suitable for evaluating the HBV residual risk in smaller blood bank settings due to its robustness against bias and increased precision. Systematic monitoring of this risk and corresponding incidence are indispensable for improving the blood safety regarding HBV, still being the most common posttransfusional infection.
Hepatitis B core antibody
Hepatitis B e antibody
Antibody to hepatitis B surface antigen
Hepatitis B surface antigen
Hepatitis B virus
Hepatitis C virus
Individual donation nucleic acid testing
Minipool nucleic acid testing
Nucleic acid testing
Occult hepatitis B infection.
This work was partially supported by the Brazilian Ministry of Science (CNPQ) Grant no. PQ 300578/2007-5.