Invitations for the first round of the Chlamydia screening in Amsterdam, Rotterdam, and South Limburg were completed by the end of February 2009. Nearly one of six 16 to 29-year-olds in the three participating regions have been screened in this first round, and one in 25 participants tested positive for Chlamydia. The final impact evaluation of the programme is expected when two screening rounds have been completed in 2010. However, the findings from the 1st year allow us to review (the expected power of) our evaluation design.
The overall participation (16%) was lower than the 30% in the Pilot Ct (both register based) which was expected as our programme is selective in its nature: only for sexually actives and for Limburg a further selection by the prediction rule was incorporated in the programme. Moreover, the programme used test kit requests via internet instead of directly sending test kits. We want to caution about (inter)national comparisons of participation rate as programme designs may differ and denominators and test kit offering may not be comparable . For instance, in opportunistic programmes, acceptance of test-offer may be presented as coverage; sometimes denominators include only sexually actives or are not population-based but reflect attendees of health facilities. The overall 16% participation rate is based on the total number of invitees which includes people who are not sexually active yet.
The overall positivity rate of 4.2% was higher than that of the pilot CT study (2.0% overall and 3.2% in highly urbanized regions) , the Amsterdam programme (3.4%)  and the UK programme (3.1%) , suggesting an adequate (self) selection which will benefit cost-effectiveness of the programme.
Using simulations, we have confirmed that, with participation and positivity rates as observed in the 1st year of screening, we will be able to show a significant 'overall' effect of one screening round on the Chlamydia prevalence if it decreases by about 15% in the cities and 25% in South Limburg. At the level of the blocks in the stepped wedge design, an effect might not be detectable with statistical significance after two screening rounds, unless the prevalence diminishes by one-third or more.
The effect of screening might better be evaluated separately in the three regions of Amsterdam, Rotterdam, and South Limburg because participation and positivity rates in the regions were significantly different. Blocks A and B of the cluster-randomized stepped wedge design also showed different participation and positivity rates in this first round. The randomization of clusters was stratified by risk level, but this apparently did not result in completely similar groups within one region. These differences should be taken into account, and multilevel analysis should be used to compare pre- and post-screening prevalence (extrapolation of findings in blocks A, B, and C).
There was a large variation in participation and positivity rates at the lowest level among the clusters. Beforehand, we assumed a positivity rate per cluster between 0% and 10% (for simulations). Most clusters remained within this range (only one cluster in Amsterdam had a positivity rate above 10%). The assignment of the risk levels (high, medium and low community risk, based on the criteria of age, ethnicity, income profile, and urbanization) created different groups of clusters, at least in the two cities. Participation was lower and positivity was higher in high-risk clusters, as can be expected, and as was also found in the UK . Studying the effect of screening within these different risk levels will provide food for thought for further population screening targets.
The stepped wedge design facilitates the analysis of the effect of multiple screening rounds on test positivity in the light of modelling Chlamydia transmission and estimating cost-effectiveness, but will also have limitations. First, equal sizes of the three blocks in Amsterdam and Rotterdam would have increased the power of the comparisons between blocks, but would deviate too much from the agreed two screening rounds for everyone. Secondly, potential bias is caused by the fact that the participation and the receiving of a Chlamydia positive or -negative test result in the first screening round may affect the participation in the subsequent rounds. This can be reviewed by comparing consistent non responders, repeat participants, and one-time participants.
The characteristics of participants and non responders (gender, age, ethnic background, sexual activity, et cetera) will be used to predict the Chlamydia population prevalence from the Chlamydia positivity rate by modelling. The success of these predictions will depend on the outcomes of the non response study: non responders are the largest group in our population, and estimating the proportion of those not responding because they are not at risk (because they are not, or not yet, sexually active or are in a long-term faithful relationship) is very important. One in 15 non responders received a questionnaire, and 15% replied. For those, information about the sexual history is available. However, basic characteristics (gender, age, ethnic group, and cluster) from the population registers are available for all non responders. The participant response rates to the online general questionnaire, the acceptability study via e-mail (60% or more) and the online treatment questionnaire were higher than response rates to the treatment questionnaire among those who tested positive (43%). The questionnaires will provide detailed insight into 'who participated when and why' (and who did not), as well as information about behavioural characteristics such as sexual behaviour, number of sexual partners, and condom use.
The stepped wedge design has the advantage that the first screened block(s) can be used for a baseline prevalence measurement, so that extra surveys (before and after screening) that could seriously interfere with participation and positivity in the screening population, could be avoided.
A reduction in Chlamydia transmission attributable to screening would provide good primary evidence of effectiveness . Further interpretation and modelling are needed to predict the effects of multiple (more than two) screening rounds. Existing predictive models [15, 16] will be updated with the findings from the screening and recent literature. Much information will be obtained from the CSI database, which will include around 300,000 invitees and, as expected on the basis of the first results, 40,000 to 50,000 participants per year. The large-scale CSI-programme, as implemented in the Netherlands, provides a unique opportunity to assess the effectiveness of a systematic, population-based screening offered to young people for reducing the Chlamydia prevalence - and self-reported PID prevalence - in the screened population and the population at large.
The outcomes can be compared to other ongoing initiatives such as opportunistic screening as practised in the UK, Sweden, and the USA , population-based studies such as those in Sweden , Denmark , and the Netherlands [7, 24], and various RCTs [4, 5]. This may provide the necessary evidence for the effectiveness of register-based programmes . In conclusion, we will use a combination of methodologies to investigate innovative aspects of the design of CSI. We expect the project to yield new insights into the impact of Chlamydia screening, epidemiological trends, and screening that makes use of the Internet. The comprehensive evaluation will enable national health authorities to decide whether to implement nationwide Chlamydia screening in the Netherlands.