Reported Chlamydia trachomatis (CT) cases have increased substantially in the past 10 years and have become by far the most common sexually transmitted infection (STI) in Sweden (Figure 1) . The number of cases reported to the national surveillance system increased from 13,905 (157 per 100,000) in 1997 to 32,281 cases (359 per 100,000) in 2004, representing a rise of over 120%. In 2005, the annual reported incidence increased only by 2%, and even decreased by 2% in 2006. One reason for this decrease may have been the emergence of a new genetic variant of Chlamydia trachomatis (nvCT) in October 2006 that can not be detected by some of the diagnostic tests commonly used in Sweden [2,3]. As a result, chlamydia diagnoses were missed and the national rates of chlamydia cases were underestimated in 2006 .
The nvCT was found to be widely spread in Sweden and its proportion varied between counties from 10% to 65%, leading to false negative results [3,5]. Laboratories in 13 of the 21 counties in Sweden had used diagnostic kits in 2006 that did not detect nvCT (Roche Diagnostics and Abbott Laboratories), while laboratories in eight counties had used diagnostic kits by Becton Dickinson that could detect both wild-type CT and nvCT. In order to improve diagnosis of the nvCT, the use of other PCR testing kits (Becton Dickinson or Artus) and/or culture was recommended . An overview by the Swedish Institute for Infectious Disease Control showed that, by March 2007, all laboratories (except one) had switched to one of the suggested diagnostic kits. This change made it possible to diagnose chlamydia infections caused by nvCT and to perform non-interrupted contact tracing, resulting in a renewed increase in reported cases.
In the beginning of 2007, the Swedish Institute for Infectious Disease Control noticed a sharp increase in reported chlamydia cases through the electronic reporting system. The incidence of chlamydia in the two-month period of January and February 2007 was 38% higher than the incidence during the same period in 2006. This raised the question: Can this increase be explained only by better diagnosis of nvCT infections? The objective of this study was to explore alternative explanations for the increase and propose further action if needed. Based on available surveillance data several alternative hypotheses were developed. One of the alternative hypotheses is an increase in the testing activity in the beginning of 2007. Another alternative hypothesis is a continued increase in the prevalence of chlamydia infection. Before the hypotheses were tested, we considered the data quality with regards to double reporting or delayed reporting to the system.
Genital chlamydia infection is a mandatorily notifiable disease in Sweden under the Communicable Disease Act from 1988 . Partner notification and contact tracing are also routinely performed . The report of a chlamydia case to the national surveillance system contains an individual laboratory notification from the diagnostic laboratory and an individual clinical notification from the health care professional. Notifications do not contain the name of the patient but are coded, based on the social security number (personnummer). In addition, all laboratories that perform testing for CT report on a voluntary basis the number of people tested and the number found positive for CT every six months. These data are available in electronical format since 2000.
Quality check for reported cases
Double reporting or delayed reporting of chlamydia cases was checked for every month in 2006 and for January and February in 2007. The time between clinical diagnosis and reporting was compared. Reporting of a clinical case more than one week after diagnosis was defined as a delay. In Sweden, all positive laboratory findings for CT with the same code within a three-month period are considered as new infections.
Grouping of counties according to diagnostic methods
We divided all 21 Swedish counties into two groups based on the diagnostic kits used by their laboratories in 2006: Group A/R used Cobas Amplicor (Roche Diagnostics), Cobas TagMan48 (Roche Diagnostics) or Abbott m2000 (Abbott Laboratories) and were unable to detect nvCT. Group BD used the ProbecTec ET kit by Becton Dickinson that is able to detect nvCT. According to this division, 13 counties were included in Group A/R and eight counties in Group BD (including county Västra Götaland, where three out of four laboratories had used Becton Dickinson diagnostic kit and one the Roche diagnostic kit).
Chlamydia cases and incidence
Reported cases in the period of January and February were described in terms of the total number, proportion of males and females, median age, and the reporting county. The incidence of chlamydia was calculated as all reported chlamydia cases per 100,000 population during January and February in the years 2000 to 2007. The national incidence and the incidence per group (A/R and BD) were calculated as geometric means of the incidence of the respective counties.
Testing for C. trachomatis
In order to quantify to what degree the different counties invested in finding new chlamydia cases, we calculated the proportion of the population between 15 and 49 years of age that was tested for chlamydia. This particular age group is tested most frequently and with the highest incidence (ca. 90% of all reported cases). Since it was not possible to obtain the specific data on tests performed in January and February, the annual number of tests was used instead.
A negative binomial regression model was used to study the time trend of chlamydia cases in January and February in 2000 to 2005. The year 2006 was excluded due to underreporting of nvCT. To model the incidence of chlamydia in January and February, the following variables were included in the model:
a) county group A/R or BD (according to diagnostic kits used),
b) proportion of the population in age group 15 to 49 years tested in each county,
We also added an interaction effect of method and time, as differences between the two diagnostic kits could have been exacerbated by the spread of the nvCT over time. The initial model with i=1, ...21 (county) and j=1,...6 (year) was:
log(casesij / popij) = β0 + β1yearj + β2proportion testedij + β3groupi + β4groupi * yearj .
All calculations were based on data from individual counties. Based on the model, a prediction of cases was done for January and February 2006 and 2007. Since the proportion of tested individuals is not yet available for 2007, two scenarios were used. The proportion of persons tested in 2007 was assumed to be:
1) 5% more than in 2006 in each county, which represents the average annual increase.
2) 20% more than in 2006 in each county (extreme scenario).
The differences between the observed and predicted incidence were summarized as mean values.
The quality check for reported chlamydia cases revealed that every month, 1-2% of cases were reported with a delay. This was consistent throughout the year 2006 and also in January and February 2007. No double reporting of chlamydia cases was discovered.
Description of cases
During January and February 2007, a total of 6,903 chlamydia cases were reported to the national surveillance system. Compared to the same period in 2006, this was an increase of 38%. The distribution of the cases by sex and median age was similar to that observed in the previous years (Table 1). The median age was 21.4 years for females and 24.1 years for males.
Between 2000 and 2005, the trend of reported chlamydia incidence in the period of January and February was increasing in all counties (Figure 2). In 2006, however, the reported chlamydia incidence decreased both in the counties of group A/R and in those of group BD, and then increased again in 2007.
Chlamydia cases were reported in all 21 counties (Table 2). Some variation in reported incidence was observed in each county year by year (Table 2). The 2006 decrease in incidence was apparent in 13 counties and in the national incidence, while the increase in reported incidence observed in 2007, affected all counties.
Testing for C. trachomatis
Figure 3 shows the proportion of the population aged between 15 and 49 years that were tested in both groups of counties. From 2000 to 2006 there was, on average, 2% more testing in Group A/R than in Group BD. In both groups of counties there was an upward trend in the proportion of the population tested for chlamydia.
We found that neither the effect for 'group' nor that for the interaction 'group*year' were significant in the model, meaning there were no differences in the trend between groups of counties. However, the general trend ('year', p-value < 0.001) and the proportion of the population tested (p-value < 0.001) were highly significant. Therefore the final model included only the significant factors, with i=1, ...21 (county) and j=1,...6 (year):
log(casesij / popij) = β0 + β1yearj + β2proportion testedij .
The model estimated an increase of 8.4% (95% confidence interval 5.8%-11.0%) in incidence per year, given a constant proportion of tested individuals. An assumed increase of 5% in testing in the same year would result in an increase in incidence of 24%. Figure 4 shows the estimated versus the reported incidence for 2000-2005 in all counties in Sweden according to this model.
We predicted the national incidence in 2006 to be 63 per 100,000 population, using a proportion of tested individuals reported in that year. The model overestimated the reported incidence in almost all counties, as well as at national level (observed incidence 55 per 100,000 population). The mean error, however, was smaller among BD counties with –3.6 compared to –10.1 in A/R counties.
The incidence for 2007 was estimated using two scenarios. When it was assumed that 5% more people were tested in each county in 2007 than in 2006, the model estimated a national incidence of 70 per 100,000 population in January and February. When it was assumed that 20% more people were tested in 2007, the predicted incidence was 75 per 100,000 population. The latter gives a prediction close to what was actually observed this year (mean error per county: 2.5).
Discussion and conclusions
The emergence of the new genetic variant of C. trachomatis (nvCT) in 2006 led to a temporary decrease in the number of diagnosed cases. In early 2007, a renewed increase in chlamydia incidence was observed. This was expected after the change to diagnostic kits that were able to detect nvCT. The cases did not differ from previous years in terms of age and sex distribution or geographical distribution. We also excluded the possibility of delayed and double reporting as a reason for the increase.
However, our comparison of counties using different diagnostic kits showed that the sharp increase in 2007 could not be solely explained by switching the diagnostic method, since rising numbers of CT were also noted in those counties that had already in 2006 used kits that can detect nvCT. This suggests that other factors could have played a role, such as a higher number of persons being tested and/or a higher CT prevalence in the population.
In almost all counties, our statistical model predicted a higher incidence for 2006 than that actually observed. This supports an effect of underreporting due to undetected cases of nvCT already in January and February 2006. When we assumed that 20% more people were tested in 2007 than in 2006, the predicted incidence for January-February 2007 was the same as the observed incidence. The situation with the newly emerged CT variant was widely covered by mass media in Sweden by the end of 2006, contributing to better knowledge on chlamydia diagnostic problems and possible false negative results. This could have led to increased testing for CT in the beginning of 2007, induced both by health professionals and patients themselves.
An additional explanation for the higher incidence could be a continuous increase in the prevalence of chlamydia in the population, as has been described earlier in Sweden . This explanation was also supported by our model.
Several limitations could influence our results. Firstly, our model did not take into consideration size of population, age distribution, testing policy, or the degree of partner tracing in the different counties, which could influence our results. In addition, we assumed that the number of tests performed during the entire year was proportional to the number of tests performed in the period of January and February. Neither did we investigate other possible explanations such as a change in sexual behaviour that could contribute to increased spreading of CT.
The sharp increase in January and February 2007 is misleading if compared to the same period in 2006 without taking into consideration the underestimated rates in 2006. Due to the fact that the diagnostic methods failed to detect nvCT in 2006, cases remained undiagnosed and as a result the contacts of these cases were not traced. This led to an accumulation of chlamydia cases and further spread. We can expect to see this effect in those 13 counties in Sweden that had used diagnostic kits unable to detect nvCT. However, more active testing due to the reasons described above or an increase in the prevalence of CT are likely to have contributed to the increased incidence in January and February 2007.
Published reports from other European countries have so far shown limited evidence of spread of the nvCT outside of Sweden [9,10]. Sporadic cases were reported from neighbouring countries such as Denmark and Norway [11,12]. However, sexual contacts during international travels could lead to spread of this genetic variant to other countries as well. Detection of the nvCT through the surveillance system can take time, as was the case in Sweden where the decrease of chlamydia notifications in some counties was masked by the overall national rates. Therefore epidemiological and laboratory vigilance are important not only at national but also at local level. Continuous evaluation of diagnostic tests is necessary. Sexual health promotion needs to be intensified in order to effectively control the spread of sexually transmitted diseases in general. Sweden has intensified prevention campaigns with information in mass media, Internet and cinemas, condom distribution to teenagers, etc. in the summer of 2007 .