Methods
Data were obtained from the RKI nationwide blood donation infection surveillance and included more than 99% of all donations in 2000 and 100% of all donations in 2001 and 2002. Blood and plasma donation centres reported aggregated data on number and type of donations from new and repeat donors and the number of confirmed HIV, HCV and HBV infections. Detailed serological results from all positive donors were available. An infection was considered confirmed positive if a reactive screening result was verified by an appropriate supplementary test in a different test system and/or NAT. During the study period all blood donations were screened for anti-HIV 1/2, anti-HCV, HCV genome and hepatitis B surface antigen (HbsAg). A large number of donations was also screened with HIV-1 NAT, HBV NAT and to a lesser extent, tested for anti-HBc on a voluntary basis. A minimum sensitivity of 5000 IU/ml with respect to the individual donation was required for HCV-NAT testing. Sensitivity of NAT had to be validated with limiting dilutions followed by probit analysis as recommended by the German licensing agency, the Paul-Ehrlich-Institute [1]. The majority of donations were screened with an in-house Taqman PCR in minipools of up to 96 samples [2,3]. To a lesser extent, donations were tested using commercially available NAT tests or in-house NAT with small pool sizes or with individual donation-NAT. All NAT-only positive results had to be confirmed either by later seroconversion or by positive NAT from a second independently drawn blood sample.
Prevalence was calculated as number of infections in all individuals who presented at the blood donation centre for the first time (new donors). Seroconversions refer to all confirmed infections found in donations from repeat donors. Infection rates were compared to data from previous nationwide studies on infectious disease markers in blood donors [4,5,6]. Trends were calculated using a Chi Square test for linear trends, 95% confidence intervals (CI) were determined using a binomial distribution. Additional data on NAT-only positives from the NAT-study of the German Red Cross (GRC) blood donor service were included (Roth, written communication). Residual risk calculations were performed using a modified incidence rate/window period model [7,8]: Briefly, the residual risk attributable to window period donations was calculated as

(window period) x (adjusted incidence/person years at risk).

Window periods for testing procedures were derived from the literature [7,9]. Incidence was calculated as number of seroconversions for HIV, HCV and HBV reported to the RKI in the study period, respectively (“crude incidence”). Donations which would not have entered the blood supply due to an additional positive test result (ALT, syphilis) or a confidential self exclusion were subtracted from the number of seroconversions to calculate the ”adjusted incidence” used in the model. For HBsAg, risk was calculated both with and without the correction factor to compensate for the transient nature of HBsAg [10]. The correction factor was determined to be 2.73 calculated from the individual interdonation intervals of the HBV-positive donations from German blood donors. Person years at risk were derived from the number of repeat whole blood donations from donors who had given at least 2 donations within the 2 year study periods (“regular donors”) between 2000/2001 and 2001/2002 divided by the mean interdonation interval length (0.52 years).The window period for HBsAg was reduced by 9 days to account for the higher sensitivity of HBsAg tests used in Germany compared with FDA licensed tests commonly used for the determination of the window period [11]. Residual risks were calculated for the 2 overlapping periods 2000/2001 and 2001/2002.

Results
German blood donation services tested 17 925 610 donations during the 3 year study period from 2000 to 2002. Of these, 91.2% were donations from repeat donors. The proportion of whole blood donations was 77.9%. Test results from new donors and repeat donors respectively are given in Table 1 including data from previous studies [5,6].

Comparing the results of blood donor screening in Germany from 1997 to 2002 the prevalence of HBV infections remains relatively stable whereas HIV prevalence increased in 2002. Seroconversion rates for both infections did not change significantly over time. HCV infections, however, demonstrate a significant decrease since 1997, both for prevalence (from 148.8 to 97.4 infections/105 new donors, p<0.000) and for the rate of seroconversions (from 2.6 to 1.5 infections/105 donations from repeat donors, p<0.000).
From 2000 to 2002, more than 17 million donations were reported to the RKI representing > 99 % of all collected donations including those of the GRC. All donations were tested with HCV NAT. With HIV-1 and HBV NAT not being mandatory in the study period, the proportion of donations screened for HIV-1 and HBV genome could not be determined exactly but certainly exceeded 60%.
The GRC blood donor service collects about 75 % of all whole blood donations in Germany and implemented NAT testing as early as 1996 in some centres for all three viruses [12]. The NAT study of the GRC included more than 21 million donations from January 1997 to October 2003. The number of NAT-only positive donations for both studies is given in Table 2.

The residual risk of an infectious window period donation entering the blood supply unrecognised was calculated using the epidemiological data reported to the RKI. Data are shown for two overlapping two-year periods 2000/2001 and 2001/2002. With the same test systems in place the estimated window periods remained the same in both observation periods. The decrease of the adjusted incidence of HCV and to a lesser extend also of HIV lead to a reduction of the estimated residual risk of window period donations. In 2001/2002 it was calculated to be 1 in 2 770 000 for HIV, 1 in 670 000 for HCV and 1 in 230 000 for HBV (corrected) without NAT and 1 in 5 540 000 for HIV, 1 in 4 400 000 for HCV and 1 in 620 000 for HBV with minipool NAT. The risk of an undetected window period donation could be further reduced to 1 in 820,000 for HBV with ID NAT. The results are shown in Table 3:

Discussion
Infection rates among blood donors in Germany are low and since 1997, a significant decrease with regard to HCV infections among new and repeat donors has been observed. Similar trends were also found in other countries [13,14]. The recent rise in HIV prevalence has to be investigated carefully to reveal possible changes in donor characteristics. People seeking free-of-charge HIV tests results by donating blood might contribute to the observed rise in prevalence. Case control studies are necessary to verify this hypothesis.
The implementation of HCV NAT has lead to the identification of 11 otherwise unrecognised HCV-positive donations as reported to the RKI between 2000 and 2002. The benefit of the introduction of HCV NAT was also reflected in the national haemovigilance report [15]. No HCV transmissions have been reported to the Paul-Ehrlich-Institute since HCV NAT testing became mandatory. The additional gain in safety achieved by introduction of HIV-1 NAT is not quite as marked due to the smaller reduction in the diagnostic window period compared with HCV NAT. Still, HIV-1 NAT did identify some otherwise undetected infectious donations which might have led to transmissions – an important result with respect to the severity of the disease. HBV NAT proved helpful in reducing HBV transmissions but this depends largely on the sensitivity of the NAT performed. With the highly sensitive PCR minipool testing following virus enrichment as performed by the GRC [2], 47 HBV NAT-only positive donations could be identified including preseroconversion donors as well as chronic HBsAg-negative HBV carriers. Still, some infectious are missed by minipool NAT after enrichment or even by individual donation NAT [1]. Compared to sensitive HBsAg tests standard minipool NAT can only add little to reduce the window period for HBV infections [16]. Due to the slow replication rate of HBV in the early phase of infection, only a very sensitive individual donation HBV NAT (e.g. with a detection limit of 50 copies/ml or less) would help to avoid a greater number of undetected infectious donations [17]. Another approach to reduce HBV-transmissions is to introduce additional anti-HBc testing to identify chronic HBV carriers with a very low viral load. There is evidence that blood components containing anti-HBc and anti-HBs do not transmit HBV [18]. Therefore re-entry of donors with anti-HBc and anti-HBs (>100 IU/l) who are negative in individual donation HBV NAT should be taken into consideration to minimise the prospective loss of donors if anti-HBc screening were introduced in Germany. Finally both measures, individual donation-NAT and anti-HBc testing should be carefully evaluated in terms of cost-benefit [19,20]. The observed difference between the RKI’s reported numbers and GRC data with respect to HBV NAT-only donations can be explained by the fact that the reporting of an (initially non confirmed) NAT-only positive result is not yet mandatory in Germany. Obviously, these infections are mainly reported after follow up testing revealed seroconversion or presence of HBsAg.

The residual risk of infectious window period donations entering the blood supply in Germany is low. The implementation of HCV NAT and the significant decrease in HCV incidence among repeat donors has lead to a measurable fall in the estimated residual risk. Also the implementation of HIV-1 and HBV NAT has an impact on the risk of undetected infectious donations because of the shortening of the window period. Comparing risk estimates between countries remains difficult as the mathematical models used are commonly adapted to the specific national data characteristics leading to significant differences in risk estimates [21].
Residual risk estimates always have limitations. The determining factor in the equation is the length of the window period which may vary considerably depending on the specificity and sensitivity of the test used. This might also hold true for the German data with different NAT tests and different pool sizes or individual donation-NAT in place. The used window period derived from the literature reflect average sensitivity of minipool NAT which is higher in some blood donation services especially when individual donation NAT is performed and consequently leads to a smaller residual risk. Furthermore, in our model we considered all window period donations to be infectious although during the early ramp-up phase of viral replication, this might not be the case [22]. It must also be kept in mind that given risk estimates are derived from repeat whole blood donors only and might therefore underestimate the true number of undetected infectious donations, as it has been shown that new donors might pose a greater risk of infectious donations than repeat donors [23]. Also, other influencing factors such as test or process errors or mutant viruses that are not detected by blood donor screening are not considered in the model. Still, keeping those limitations in mind, the residual risk model was able to demonstrate the benefit of NAT techniques in reducing window period donations especially for HCV and HIV.

Acknowledgements
We thank Prof. Dr. Roth, German Red Cross, Frankfurt, for providing current data on NAT-only positive donations from the NAT study of the German Red Cross blood donor service.