By using the data from the German HIV-1 Seroconverter Cohort, we estimated the risk of AIDS in calendar periods with different HAART regimens as well as during mono and dual antiretroviral therapy (pre-1997). We found an overall reduction of 80% in the risk of AIDS over the calendar periods, which was not linear but included two main drops, one after 1997 and one after 2007. This is the first study to our knowledge showing a reduction in AIDS risk after 2007. However, these results might be influenced by the fact that the study population includes a large proportion of MSM who are more likely to be closely monitored by HIV specialists and may therefore not be representative of the entire German population infected by HIV. A similar cohort study in Spain reported significant reduction in the risk of AIDS over calendar periods in comparison with the pre-HAART era, but there were no differences between the calendar periods in the post-HAART era . However, the last observed calendar period was 20002003 and there were more IDU in the composition of this cohort than in our study. In another study combining 22 cohorts of people living with HIV-1 from Europe, Australia and Canada, the authors showed an even bigger reduction in the risk of AIDS, with a relative risk of 0.46 and 0.13 after 1997 and after 2001, respectively .
Our results showed an overall increase in HAART uptake, which likely parallels the decrease in AIDS risk. These trends have also been seen in the UK, where the proportion of patients being treated by antiretroviral therapy increased from less than 2% in pre-1996 to 58% in 20042006 . The decrease in HAART uptake in 20052006 in our study, associated with an increase in AIDS risk and AIDS incidence in the same period, supports the hypothesis of this association between HAART uptake and AIDS risk. Increasing HAART uptake benefits could have been balanced by an increasing prevalence of Transmitted Drug Resistance (TDR) over the calendar periods. However, one previous study in the same cohort of seroconverters indicated a stable prevalence of TDR over the time . Other factors related to the improvements in drug quality, including safety and galenic formulation easing intake, together with increases in HIV-knowledge among health professionals have probably also contributed to the decrease in AIDS risk.
We observed a trend towards initiating treatment later following seroconversion in later calendar periods. This could be explained by the different recommendations in Germany since 1997. Following the Vancouver Conference  and the increasing uptake of HAART, the main attitude in Germany was to hit hard and early . In 2002, following the large proportion of people with antiretroviral therapy side effects/toxicity, the German AIDS Society recommended to start therapy below a threshold of 200 CD4+ cells per L . This recommendation might partially explain the slight decrease in HAART uptake and consequently, the slight increase in AIDS risk for the period 20052006 in our study. In 2008, recommendations were revised to initiate the antiretroviral therapy below a threshold of 350 CD4+ cells per L . Surprisingly, this last recommendation was not associated with an increase, neither in the median time to treatment initiation nor in the proportion of people being treated less than 2years and between 2 and 5years after seroconversion. However, this last period was associated with a decrease in AIDS risk, indicating that factors other than time to treatment initiation may play a more important role in disease progression.
The main determinant that affects disease progression in our cohort was age at seroconversion. This factor, together with duration of infection, was already known to be a crucial determinant of HIV disease progression in developed countries before the introduction of HAART . However, age disparities seem to diminish over the calendar periods in our cohort, as already shown in another study including 22 cohorts .
Our results did not show an effect of sex on AIDS risk. This result contrasts with recent studies that reported a slower disease progression among women than men [8, 18]. However, the number of women in our study population was too small to detect any potential differences.
Comparing disease progression rates between transmission risk groups in our study was difficult, considering the large proportion of MSM compared to the other transmission groups. However, the trend over the calendar periods showed that IDU had a smaller reduction in AIDS risk as compared with the others risk groups. This result is supported by Porter et al., reporting a smaller reduction in AIDS risk in the post- HAART era among IDU . However, other causes of mortality, such as hepatitis C, could confound this result. People from high HIV prevalence countries had at higher, although non-significant risk for progression to AIDS. Furthermore, the risk seems to decrease with the calendar periods. This result should reflect a heterogeneous population and be interpreted with caution because of the small number of people in this group. Nevertheless, access to health services, health insurance status, poor adherence to treatment  and other co-morbidities could have played a role and may require special attention for this group.
The presence of seroconversion illness, typically characterized by flu-like symptoms, has been reported to be associated with a more rapid disease progression [20, 21]. In order to avoid recall bias, short HIV test interval has been reported to be a good proxy for the presence of seroconversion illness . In our study, short HIV test interval factor showed a slight effect, which appears to increase over time. However, this result depends on the quality of the tests used and on the time interval chosen to define a short test interval.
This studys principal strength is its good quality data on date of seroconversion, permitting reliable approximation of AIDS incidence and risk. This study also has the advantage of providing data for more than 13years of prospective follow-up from different settings, representing roughly 10% of all new HIV diagnoses reported to the national HIV surveillance since 2004. However, some limitations have to be raised. First, the data for the pre-1997 period are retrospective and may not be as reliable as the data from the prospective period. Patients in this period were included regardless of their AIDS outcome; a survivorship bias might have been introduced as patients have to sign an informed consent at study entry. Regarding the pre-AIDS mortality, this bias might have been limited as the pre-AIDS mortality rates were relatively stable and at a very low level during the prospective periods. Regarding the lost to follow up, rates were relatively stable over the time periods but were at a high level. If lost to follow up subjects have a more rapid progression, the pre-1997 period in our study might underestimate the risk of AIDS in comparison to the other periods but our results remain conservative to this regard. Alternatively, we might have overestimate the first decrease in the AIDS risk if lost to follow up subjects have a slower disease progression. Considering only the periods after 1997, a bias might have been introduced if lost of follow up subjects differ between the periods. Further analyses (e.g. competing risk analysis) should be performed to better answer that issue. Another issue is that women are largely underrepresented in this study (6.2%) in comparison with the national reporting system, in which women accounted for as many as 20% of all new diagnoses between 1997 and 2010. This proportion does not allow us to analyze the effects of sex and HIV transmission categories with enough power. Finally, we assumed continuous HAART intake after treatment initiation. This limitation might be important for the interpretation of the proportion of people being treated by calendar period. However, this does not change the interpretation of the effectiveness over time. The treatment was considered to be HAART (i.e., three or more drugs that are from two or more classes, or that contain abacavir) in every calendar period from 1997 onwards, even if this might not be the case for one or two percent.