Increase of new HIV diagnoses among men who have sex with men in Poland, 2000 to 2011

the increasing rate of new HIV diagnoses suggests ongoing transmission


Introduction
The global HIV epidemic is considered to have peaked during the second half of the 1990s [1]. However, several regions are experiencing increases of new infections, notably eastern Europe and central Asia, but also western Europe and North America. In particular, re-emergence of the HIV epidemic among men who have sex with men (MSM) has been observed in western Europe, North America and Australia since the mid-1990s [2]. In the European Union, the number of new HIV diagnoses among MSM increased by 33% between 2004 and 2011 [3]. In 2011, 39% of newly detected HIV infections were attributed to sex between men, 23% to heterosexual contacts and 5.4% to injecting drug use [3].
In contrast, the HIV epidemic in Poland remained stable until approximately 2005, at relatively low levels (<20 newly diagnosed HIV infections per million population), with injecting drug use as the major mode of transmission [4,5]. People who inject drugs accounted for over 75% of HIV cases with known transmission category [4,5]. New HIV diagnoses among people infected through sexual intercourse were uncommon. However, the testing rate among the general population was among the lowest in Europe (at 8.3 per 1,000 population in 2011) [3] and the high proportion of late-presenting cases among sexually infected individuals (20-32%) indicated possible underestimation in the surveillance data of the sexual spread of the virus [4][5][6]. Until 2005, a slight increase in detection rate was attributed to epidemic maturation [7]. However, a significant increase in new HIV diagnoses across age cohorts occurred after 2005, suggesting emergence of the epidemic [7]. Changes in distribution of the main transmission categories among the people with new diagnoses were also reported, with increasing importance of MSM and heterosexual women and men who do not inject drugs [5,6,8]. However, conclusions were hampered by the high proportion of diagnoses with missing transmission category (over 60%) in routine surveillance data [3,5,6].
HIV prevention services for MSM in Poland remain limited. Monitoring of prevention efforts revealed that in 2007-10, only a few local outreach or community programmes, including education and condom distribution, were targeted at MSM [9]. Additional education was provided through a network of HIV voluntary counselling and testing centres located throughout the country, reaching an estimated 1,500-4,500 MSM clients annually [9]. Internet campaigns were also used. Novel methods such as pre-exposure prophylaxis, which has proved effective and safe [10], are unlikely to be implemented in Poland in near future.
The aim of our study was to analyse the trends in new HIV diagnoses among MSM in Poland, adjusting for missing transmission category in routine surveillance data.

Data sources
We used data from routine HIV/acquired immunodeficiency syndrome (AIDS) surveillance in Poland, which is a comprehensive case-based national system using European Union case definitions [3]. New HIV diagnoses are reported by laboratories and clinicians to regional public health departments. Regional departments collate the clinical and laboratory reports, collect additional information and forward the forms to the National Institute of Public Health -National Institute of Hygiene. The notifications may be namebased, but the patient has the right to refuse to reveal personal information on the report form. The system relies on paper forms and an electronic registry using a name-based identifier is maintained only at the central level. Clinical and epidemiological information (transmission category) is usually not available in the reporting laboratories. As the laboratories report more cases than clinicians do, there is a high proportion of patients with missing data (clinical characteristics and transmission category). The first AIDS-indicative disease is also reportable, by the diagnosing clinician, and reports of HIV infection and AIDS related to the same person are linked at the regional and/or central level. Therefore, reported HIV cases who develop AIDS are more likely to have complete information regarding transmission category.
In order to minimise the effects of reporting delay, we used data on cases diagnosed up to the end of 2011, but reported by July 2012. The reporting delay was estimated once using all reported cases, and a second time excluding those cases that were notified only as a result of reporting completeness control activities.
We also used the results of an annual survey sent by mail to laboratories that offer HIV screening tests (as opposed to the laboratories performing confirmatory HIV tests, who report new diagnoses), for which data were available for 2003-11. In 2003-11, the number of participating laboratories ranged from 140 to 266, reflecting changes in the market of diagnostic services for HIV. The survey is voluntary and the response rate varied from 59% to 85% in 2003-11. Aggregated data on the number of diagnostic HIV tests performed in each laboratory were collected, excluding tests of blood donors. The reporting form allows disaggregation by transmission category and sex, but only several laboratories are able to provide this information. Only laboratories reporting 10 or more tests among MSM in a given year were included in the analysis in order to exclude data from laboratories not routinely collecting information on transmission category, acknowledging that we will also exclude those who test only a few MSM per year.

Adjustment for missing information on transmission category
Transmission category may be missing in the surveillance data because it is not known to clinicians or, even if known, may not be reported to the public health system by clinicians. In data reported by clinicians (e.g. AIDS case reports), transmission category is not specified for approximately 12% of cases, which suggests that lack of these data may be mostly due to inadequate completeness of reporting. The fact of reporting by a clinician is unlikely differential to transmission category. However, clinicians may be more likely to report cases diagnosed with AIDS, thus bias could result from a more common occurrence of AIDS in one of the categories. As the proportion of missing values for transmission category increased over time and as we expect regional differences related to local collaboration of public health services with clinicians, we assumed that the information is missing at random in any given the year of diagnosis, region and time of diagnosis (late presenter or other). We used a multiple imputation procedure imputing concurrently age (years), sex and transmission category (MSM vs other, conditional on sex) with five imputed datasets [11]. Iterative chained equations were used to account for non-normal distribution of the imputed variables [12]. In this method, the variables with missing values are iteratively imputed on the basis of regression equations estimated on an imputed dataset from the previous step, including as explanatory variables other variables that are being imputed as well as other available covariates.
We used logistic regression for sex and transmission category (MSM vs other) and linear regression for age. Additional explanatory variables included year of diagnosis, region and late presentation (yes vs no). Due to lack of data on CD4 count at diagnosis, we could not use the late presenter consensus definition as outlined in [13]. In our study, a late presenter was defined as person who developed AIDS within a year of HIV diagnosis. Estimates were then produced for each imputed dataset and combined using the Rubin's combination rule [11,12]. The analysis was conducted in STATA 10.1 using 'ice' and 'mim' commands [12].
The evolution of characteristics of new HIV diagnoses among men in the study period is shown in Table 1. We note a remarkable shift in the transmission category distribution. Among cases with known transmission       (Table 2).

Discussion
Our study suggests a 14-fold increase in the number of new diagnoses among MSM in Poland from 2000 to 2011, when imputing data on missing transmission category. This increase occurred concomitantly in all age groups, but was more pronounced among younger MSM (aged 25-44 years).
There were no systematic changes in the surveillance system or testing policy during the study period. However, testing for HIV, especially self-initiated testing was repeatedly encouraged. Additionally, the number of voluntary testing and counselling sites offering free-of-charge, anonymous testing increased [14]: such Previous studies (in 2004 and 2005) estimated that in MSM, a large proportion of HIV infections -larger than that among people who inject drugs -might have been undiagnosed [16,17]. Improving access to testing over the study period could have led to a rise in the testing rates among MSM and, in consequence, contributed to the observed increase in notification rate. However, we also noted an increase in HIV prevalence among MSM coming for testing (from 5.4% in 2003-05 to 16.1% in 2006-08 and 7.9% in 2009-11). Although this increase alone would not fully explain the notification trend, it indicates that an upsurge of new HIV infections may also play a role. To further support the hypothesis of increasing HIV incidence among MSM, other studies found a high proportion (>30%) of recent infections confirmed by recent infection testing algorithm (RITA) testing among new diagnoses among MSM [18,19].
On the other hand, the proportions of respondents in behavioural surveys who reported having ever been tested were comparable in 2004 and 2010, 57% and 62% respectively, although it must be borne in mind that the study designs were different [15,16]. Either increased frequency of testing among those who were tested or a possible increase in disclosure of MSM status when testing for HIV may be additional factors that affect the surveillance trend. The observed trend is likely to be influenced by a number of factors, but the available information supports the hypothesis of emergence of the HIV epidemic among MSM in Poland.
Re-emergence of the HIV epidemic in MSM was noted in several western countries and attributed to increasing frequency of high-risk behaviours among MSM [2,[20][21][22]. Inconsistent condom use and a high number of partners was also confirmed in Poland in a community-based survey among clients of HIV voluntary counselling and testing sites in 2004-07 [14,16] as well as (more recently) by an Internet survey among MSM (EMIS) [15]. The proportion of MSM who reported 10 or more partners in the past 12 months in these studies was estimated at 20%. In the communitybased survey, past-12 months consistent condom use with non-stable or casual partners was approximately a third; in the Internet survey, the level rose to a half of the respondents. This level of risky behaviour is similar to that seen in western European countries [23], some of which noted an increasing HIV incidence in the recent past [2]. In contrast, the proportion of non-testers is higher in Poland than it is in western European countries. An important proportion of undiagnosed infections reaching an advanced phase could have contributed to increased spread of the virus. It has been suggested that in longer-lasting epidemics, individuals in the advanced stages of disease who are not diagnosed (and not treated) could contribute greatly to re-emergence of the epidemic [24]. Once the new infections start to appear, the epidemic could pick up with transmission at the acute infection stage. A recent molecular study showed clustering of strains from recently infected MSM, suggesting transmission events often occur from a recently infected MSM as a source [25]. Transmission from long-standing undiagnosed infections followed by intensified transmission from recently infected individuals was proposed to explain the increasing spread of HIV among MSM in Germany in the early 2000s [26].
Our results show that the highest diagnosis rate and the fastest, over 40-fold, increase during the past decade occurred in one region, Mazowieckie, which includes the capital, Warsaw. The increase could be related to the higher proportion of MSM in the capital and more extensive sexual networks. Although we have no data on this for Mazowieckie, a higher concentration of MSM was noted in London, for example, as compared with other part of the United Kingdom [27].
Our study has several limitations. Firstly, we based on surveillance data, which are prone to various biases related to case ascertainment and collection of sensitive information. We cannot exclude misclassification of transmission category by reporting clinicians. Secondly, we relied on the small fraction of cases for whom we obtained reliable clinical reports: the rest was assumed to be missing at random. However, taking into consideration stigmatisation of same-sex relationships in Poland, MSM are rather less likely to reveal their exposures than heterosexual persons or people who inject drugs, so we are more likely to underestimate than to overestimate the HIV rates among MSM. Thirdly, we did not fully account for reporting delay. However, we noted that less than 10% of cases were notified with a delay of over six months, so only a small correction is expected after this time.
Our findings call for enhanced, comprehensive prevention among MSM, including revision of testing strategies, treatment of other sexually transmitted infections and possibly implementation of recently proven techniques such as pre-exposure prophylaxis, along with expanding individual-and community-level behavioural interventions for the prevention of HIV and other sexually transmitted infections and continued access to antiretroviral therapy for those already infected [28][29][30][31]. These efforts should be a priority especially in regions with the highest transmission; however, more data are needed on possible regional differences and specific behaviours driving the epidemic in Poland as well as more generally in countries with marked regional differences.
It is of note that the increase in number of new HIV diagnoses among MSM was evident several years earlier when using the imputed dataset compared with the raw surveillance data. This underscores the need to assure data quality and improve the analysis of routine data by correcting for missing data on important descriptors such as transmission category.