Estimating the regional distribution of men who have sex with men (MSM) based on Internet surveys
© Marcus et al; licensee BioMed Central Ltd. 2009
Received: 14 August 2008
Accepted: 11 June 2009
Published: 11 June 2009
Measurement of prevalence and incidence of infections in a hard to reach population like men who have sex with men (MSM) is hampered by its unknown size and regional distribution. Population-based surveys have recently been used to estimate the total number of MSM, but these surveys are usually not large enough to measure regional differences in the proportion of MSM in the population. We explored the use of the proportional regional distribution of participants of large internet-based surveys among MSM from Germany to estimate the regional distribution of MSM in Germany.
We compared participants from two separate MSM behavioural surveys with each other and with the distribution of user profiles of the largest contact and dating website for gay and other MSM in Germany in terms of the representativeness of the regional distribution. In addition, we compared the regional distribution of reportedly HIV positive survey participants with the regional distribution of HIV notifications within the national surveillance system that can be attributed to transmission through homosexual contacts.
Regional distribution of survey participants was almost identical in both surveys, despite little overlap between survey participants. Slight discrepancies between surveys and user profiles could be observed. Proportional regional distribution of survey participants with HIV diagnosis resembled national surveillance data.
Considering the difficulties to obtain representative data by other sampling methods for "hidden" populations like MSM, internet-based surveys may provide an easy and low cost tool to estimate the regional population distribution – at least in Western post-industrialized countries. Some uncertainties remain about the exact place of residence of MSM in larger cities or catchment areas of these cities. Slightly different results from different datasets may be due to unequal popularity of MSM websites in different regions. The total population size of the MSM population can be estimated based on e.g. data from representative national population surveys. Both estimates can then be combined to calculate the absolute size of regional MSM populations.
National infectious disease surveillance data for sexually transmitted infections, if they collect information on the gender of sexual partners, usually refer diagnosis or infection incidence in MSM to the general population or the general male population. However, this approach neglects specific migration processes that lead to a concentration of sexual minorities such as men who have sex with men in larger cities, especially if they self-identify as gay.
The measurement of prevalence and incidence of health conditions or infections in subpopulations such as men who have sex with men (MSM) is hampered by the unknown size and regional distribution of this "hidden" population. Problems arise from the definition of MSM, the identification of MSM in population based surveys, and the unequal distribution of MSM in the general population, which would require very large sample sizes for general population surveys to establish a valid pattern of the regional distribution of MSM .
Previous efforts to estimate subpopulation-specific incidence and prevalence of HIV therefore often remained restricted to metropolitan areas with higher proportions of MSM in the population . Some approaches rely on a multitude of local and regional studies with differing methodologies , and some use samples of gay men recruited in health care institutions or community venues, which may contain numerous biases and are not representative for MSM.
Other methods that have been explored to estimate regional distribution of MSM are the use of markers like clustering of registered civil partnerships of same sex couples in national census data and participation in telephone surveys for MSM [4–7].
Population-based surveys have been used to explore the percentage of MSM in the general population [8, 9], but the validity of such estimates relies on the acceptance of talking about same sex experience. Fortunately, it seems that in Western post-industrialized countries, asking people about their sexual preferences is becoming more and more acceptable [8, 10]. It has also been reported that in computer-assisted interviews respondents are more willing to answer sensitive questions, such as on sexual preferences, than in person-to-person interviews . MSM in population based surveys are usually identified by self-reported sexual contact with other men, either during lifetime or during more recent time periods. Men who are ambiguous about their desire to have sex with other men, who are sexually abstinent or who have predominantly heterosexual contacts will only partly self identify as MSM in such surveys, and the willingness to report homosexual contacts or sexual preferences depends on the wording of questions, survey methods and settings . Some level of underreporting of stigmatized behaviour will therefore inevitably persist, and the proportion of the male population that can be identified as MSM in surveys is strongly depending on the social acceptance of same sex sexual relations and may thus change over time and differ between countries and regions. However, the data derived from such surveys can still be quite useful to estimate the "identifiable" size of a population at risk for sexually transmitted infections (STIs), because men who are reluctant to report sexual experiences with other men in the relatively anonymous context of a survey, may even more rarely report homosexual contacts to a physician if diagnosed positive for HIV or other STIs.
We hypothesize that due to the rapid adaptation of online chatrooms and other internet contact sites among MSM, the regional distribution of participants of larger MSM internet surveys can be used to estimate the regional distribution of the whole "identifiable" MSM population, at least in countries where communication infrastructure allows wide spread and equal internet access in all regions.
To verify this hypothesis we explored the consistency of the regional distribution of participants of two separate MSM behavioural surveys and compared these distributions with one of the largest available dataset on MSM: the user profiles of Germany's leading MSM dating and contact site, called GayRomeo ™. Nearly 300,000 user profiles (as of early 2008) can be sorted by federal state and 'city' – which is usually the next bigger city with more than 100,000 inhabitants.
In order to validate our findings with a data source that is independent of internet access, internet use, or study participation (and therefore lacks participation bias), we compared regional distribution of HIV notifications (attributed to homosexual contacts) in the national infectious disease surveillance system with the regional distribution of survey participants who reported to be HIV positive. The underlying argument is that if the regional distribution of a subgroup – MSM with a diagnosis of HIV – is representative, then the regional distribution of the whole convenience sample is likely to be representative as well.
Data on MSM from behavioural surveys
The KABaSTI-study (Knowledge, Attitudes and Behaviour as to Sexually Transmitted Infections) was a behavioural survey among MSM in Germany that focussed on STIs and was conducted in 2006. For the biggest of three different study arms, participants were recruited online: German language MSM contact-, dating-, and chat-websites were asked to provide a link to the questionnaire website for their users. Of 19 websites approached, seven websites participated, including the largest German language websites. The mode of advertisement for participating in the KABaSTI study ranged from individual information to all website users by a newsletter, providing a prominent link on the start page, to providing a more hidden link on an information resources page. The referring website was recorded for evaluation. Postal code and country of residence data were collected in early questions, while HIV testing history and HIV status data were collected in the last part of the questionnaire. The drop-out between the first and the last questions was 35–40%. There were minor differences in regional distribution of participants who dropped out and those who completed the questionnaire. However, these minor differences did not result in any appreciable regional bias concerning distribution by federal state or by rural – metropolitan areas.
Recruitment of the offline sample
Offline participants were recruited in 76 private medical practices, who since 2001 had reported syphilis infections in MSM to the German infectious disease surveillance system. A considerable proportion of these practices have specialized in HIV care. Most participating practices were located in larger cities. Since offline participants reflect the geographical distribution of HIV specialised practices rather than geographical distribution of MSM in Germany the offline sample of the study was excluded from this analysis.
The KABaSTI study was approved by the ethical committee of the Charité University Clinic in Berlin.
The Gay Men and AIDS (GMA) survey of 2007 was the latest in a series of periodic behavioural surveys among MSM in Germany, started in 1987 . In GMA-2007, participants were recruited online (basically like in the KABaSTI-study) and offline through nine magazines for gay men, that included the print questionnaire in their May edition. All magazines but one are available free of charge in gay venues in the larger cities. The questionnaire was self administered, and no incentives were offered for participation.
GMA participants were also asked whether they had participated in the KABaSTI study or other similar online surveys in 2006, and offline participants were asked whether they already had filled out an online questionnaire. Both surveys – KABaSTI and GMA – collected demographic data, including the first two digits of the five digit postal code of the place of residence, data on population size of the place of residence, educational status, sexual identity, HIV testing behaviour and HIV status [13, 14].
Questionnaire data were analysed for consistency and validity (e.g. online questionnaires filled out in less than ten minutes were excluded). For statistical analysis we used SPSS, versions 14.0 and 15.0.
GayRomeo User data 2008
The regional distribution of participants of the above described surveys was compared with publicly available data on GayRomeo user regional distribution, which is provided on the website and is based on self-reported data from the user profiles. Data were accessed in January 2008 and refer to almost 290,000 currently active profiles of users living in Germany. We consider this the largest publicly available dataset on regional distribution of MSM in Germany.
Data on the size and regional distribution of the general and MSM Population
For the size and distribution of the general population in postal code areas we used commercially available data provided by Infas Geodaten for the adult population aged 20–59 years dated from 2001. These data are not available stratified by sex. For this analysis it was assumed that men and women in this age group are equally distributed in the postal code areas. The difference of the total male adult population in the age group 20–59 years between 2001 and 2007 was 31,000 (0.14%) and was neglected. We assumed no significant change in the distribution by postal code areas between 2001 and 2006/07
Estimates of the proportion of MSM in the adult male population in Germany: the Federal Agency for Health Promotion (Bundeszentrale für gesundheitliche Aufklärung) collected data on sexual preference and experience in representative telephone surveys with 3,100 adult male participants in late 2007 for regular evaluation of HIV related health promotion activities. 2.5% (95%CI 1.5 – 3.4) of the male participants in the age group 20–50 years reported sexual contacts with men in the previous 12 months [, BZgA, personal communication]. Since sex between persons of the same gender is still a stigmatised behaviour, we believe that the upper range of the confidence interval provides a more realistic estimate than the lower range. Based on an estimated proportion of 2.5 to 3.4% and population statistics provided by the Federal statistics agency (Statistisches Bundesamt), we estimated the number of MSM in the adult population between 20 and 59 years of age in Germany at 575,000 to 785,000 persons. For calculation of regional MSM populations we used an estimate of 600,000.
Comparison of study samples with another internet convenience sample and with national surveillance data
To evaluate the extent to which survey participants are representative for German MSM in terms of regional distribution, we compared the two study samples with each other and with the Gay Romeo user profiles. In addition, the proportional regional distribution of survey participants who reported an HIV diagnosis later than 1995 was compared with HIV surveillance data from 1996 through 2007 for regional distribution. For the analysis, HIV notifications within the surveillance system had to fulfil two conditions: (1) attributed to MSM by the notifying physician (2) the postal code of the residence place (PCA) was available for the patient. 11,657 notifications were included. Case reports with postal code available only from the health care provider (n = 3,956) or the laboratory (n = 2,132) were excluded. For the comparison of the proportional regional distribution between surveillance and survey data both survey samples were combined.
Details of the German infectious disease surveillance system for HIV have been described elsewhere .
For the KABaSTI-study we received 6,958 online questionnaires of which 1,600 (23%) were excluded. The most frequent reason for exclusion was residence outside of Germany (35%), followed by early termination of the online questionnaire after the first or second question (12%), fill-out times of less than ten minutes, highly implausible or contradictory responses, and reported age below 16 or above 85 years. Young persons and persons with low educational status were overrepresented among the excluded questionnaires.
For the GMA-2007-survey, 9.724 questionnaires were completed. Of those, 886 from participants residing in Austria were evaluated separately. 357 were excluded because no country of residence was reported. 311 questionnaires were excluded for other reasons (completion in too short time, no age specified, highly implausible answers, other countries of residence than Germany). Offline participants who reported participation in the internet survey were also excluded (n = 88). 8,170 remaining questionnaires could be evaluated, 1,975 of them were submitted offline and 6,195 online.
Basic characteristics of KABaSTI and GMA-2007 survey participants
KABaSTI (online) (2006)
(n = 5,358)
(n = 8,170)
No qualification for university access
Qualification for university access
> 1 million
HIV testing history
Self-reported HIV status among tested
no test result specified
Funkyboys (young MSM)
DBNA (young MSM)
Bareback City (bareback)
Gay print media
The composition of the online samples differed according to the websites from which the participants were linked to the questionnaire.
The most obvious differences existed for participants recruited on a website for young MSM (younger, more often living outside of larger cities, lower educational status, higher proportion of lower paid jobs), and for participants recruited on a bareback website (older, highest proportion of participants with HIV or previous STIs, a high proportion living in metropolitan areas, highest proportion of unemployed participants).
Comparing all online participants of the KABaSTI-study with all GMA-2007 participants, no relevant differences were found with respect to age, educational or professional status, population size of the place of residence, sexual identity, and HIV testing history (see Table 1). A significantly higher proportion of KABaSTI participants reported to be HIV positive. This can be explained by one bareback website in the KABaSTI study, which contributed almost 50% of all HIV positive KABaSTI participants. Excluding these participants, the proportion of HIV positive participants were virtually the same in both surveys.
In comparison with the general population, the study survey populations were younger (men aged 50 years or more are strongly underrepresented), better educated (the lowest educational degree achieved by 51% in the general and 14% (KABaSTI) resp. 13% (GMA) in the study population, highest degree achieved by 22% in the general and 51% (KABaSTI) resp. 58% (GMA) in the study population and had a higher professional status. The unemployment rate was comparable with the general population.
Comparison of the regional distribution of survey participants and GayRomeo user profiles
Estimates* for the regional population size of MSM in Federal States and largest cities in Germany
Source for estimate
GayRomeo User Profiles 2008
German Federal States, in alphabetical order
(postal code regions)
(68 – 79, 88, 89)
Stuttgart (70, 71)
(80 – 87, 90–97)
Munich (80 – 81)
(10 – 14)
(14 – 17)
(20 – 22)
(34–36, 60 – 65)
Frankfurt (60 – 61)
Mecklenburg-Vorpommern (17 – 19)
North Rhine- Westphalia
(32, 33, 40–48, 50 – 53)
(54 – 58, 67)
We calculated MSM population sizes of federal states and the largest cities of Germany based on the proportional regional distribution of survey participants and user profiles, assuming a total population size of 600,000 for MSM in the age range between 20 and 59 years (Table 2), which represents 2.6% of the total male population in these age groups in Germany. With few exceptions the ranges of the estimates for the 16 federal states calculated from the three different data sources were small. Noteworthy are the larger discrepancies between survey-based estimates und user profile based estimates in Bavaria and North-Rhine Westfalia. (see Table 2).
Comparison of the relative regional distribution of prevalent HIV infection in internet survey participants and national HIV surveillance data
Estimating regional MSM concentration factors
Potential Bias in the offline sample of the GMA survey
Because the GMA-offline sample was recruited by gay magazines which are primarily distributed in gay venues of the largest cities, this sample may be slightly biased towards men older than 25–30 years living in metropolitan cities. However, while the KABaSTI-offline sample represents patients (with a high probability of being HIV-positive or having a history of previous STIs) and the geographical location of practices, the GMA-offline sample represents middle class middle age gay men in larger cities and the respective catchment areas. Because of these differences, we feel it was justified to exclude the offline sample of the KABaSTI-study while we included the offline sample of the GAM survey for the estimation of the regional distribution of MSM.
Representativeness of Internet surveys for regional MSM distribution
High precision in estimating MSM population sizes is an elusive goal due to the lack of a precise definition of MSM and measurement problems. We demonstrate that in a post-industrialized country like Germany, quite representative samples of sexually active middle-class MSM can be reached with web-based surveys. Participation biases for MSM subpopulations are likely to be much smaller for internet convenience samples of MSM than for MSM convenience samples approached in any other setting; however biases probably still remain for MSM who do not self-define as gay men, older MSM (> 50 years of age), MSM from migrant communities, and MSM with lower educational or professional status. Web-recruited MSM samples are also likely to be more sexually active than MSM in a probability sample of the total population, which is true also for other gay venue based samples. We can detect a significant impact of educational status on the population size of the place of residence, and it has been described that men with a lower educational status less often self-identify as gay  and seem to participate in Internet surveys to a lower degree than MSM with a higher educational status. Thus it may be that the regional distribution of lower class MSM who are reached less well by Internet surveys or are less willing to participate in such surveys differs from our sample.
For the higher population size estimates based on user profiles compared with survey participants for most of the larger cities there may be several interpretations. The most likely explanation is that the regional allocation in the user profiles is not as much based on the real place of residence, but rather on the "MSM catchment area" of the respective cities. This would result in inflated estimates for large cities in densely populated areas (e.g. Munich, Frankfurt, Hanover and Stuttgart), while estimates for cities with less densely populated surroundings (e.g. Berlin, Leipzig) would be less variable (see Table 2). For Hamburg, Bremen and Cologne we reduced this "catchment effect" already by including postal code areas which only partly cover city areas.
Other possible reasons might be that MSM from metropolitan areas are less willing to participate in behavioural surveys than MSM outside these areas, or that a higher proportion of metropolitan MSM is visiting MSM websites, i.e. that the penetration of MSM websites in the urban MSM population is higher, while the proportion of MSM from different areas willing to participate in behavioural surveys is similar or equal. Thus, it remains an unresolved question, whether MSM population estimates for some of the largest cities derived from survey participant distribution reflect reality more or less precise than estimates derived from the user profile distribution.
Another potential bias can be regional differences in the preferred websites for MSM, which we tried to minimize by including a mix of websites for the recruitment of survey participants. However, we had no influence on the method of advertising the surveys. The website with the largest user number had decided to promote the surveys only by a relatively hidden link on an information resource page, while other websites with a more skewed regional popularity had promoted the surveys more prominently. As we can show by comparing the survey participant distribution between participants recruited on Gay Romeo and the other websites, this may have contributed to some bias in regional participation rates which may explain the discrepancies in the estimates for Bavaria and North Rhine-Westphalia derived from the surveys and the user profiles: From the KABaSTI participants who were recruited on GayRomeo 14.6% reported residence in Bavaria, 19.3% residence in North Rhine-Westphalia, compared with 10.6% and 25% of all participants. Similarly, by inclusion of a regionally popular website in the KABaSTI study, the KABaSTI-based estimate for Saxony and Saxony-Anhalt may be slightly inflated (see Table 2).
While we found little published literature on geographical representativeness of internet convenience samples of MSM, Evans et al., who compared an internet convenience sample of MSM from Great Britain with a national probability sample , also found a similar geographical distribution of MSM in both samples.
The comparison between the regional distribution of diagnosed HIV infections in KABaSTI participants and the regional distribution of diagnosed infections reported in the national surveillance system (Fig. 2) reveals some minor and some more pronounced differences. To minimise biases due to the low absolute numbers of diagnosed infections in the internet survey participants we combined both surveys for this analysis. Differences therefore exist in regions with low numbers of internet survey participants, or low numbers of reported infections. But also for several larger cities discrepancies exist: in Munich, Cologne, Duesseldorf, Stuttgart and Hanover (PCA 80, 50, 40, 70, 30) the proportion of HIV reports in the surveillance data is higher than the proportion of survey participants, while in Berlin and Frankfurt (PCA 10–14, PCA 60–61) it is the other way round. We believe that for Munich, Cologne, and Duesseldorf this may still be partly due to medical tourism from the surrounding areas, although we tried to reduce this confounder by restricting the analysis to cases reported with the PCA of the patient's place of residence. However, there may still be some level of misreporting the PCA of the health care provider as the PCA of the patient. This interpretation is supported by the observation that when we use the distribution of surveillance reports for comparison, where the PCA refers to both the place of residence of the patient and of the care provider, the differences become larger. For Stuttgart and Hanover however there is no difference in this regard. For these cities internal migration to other cities seems the most plausible explanation to us. Berlin would be expected to be the most attractive destination city for such internal migration movements because of its large gay community and well developed medical and subculture infrastructure. Also Frankfurt may be a destination city for the same reasons.
Another factor that could contribute some bias would be regional differences in internet access. However, after 2003, there are no major regional gaps in internet access, and availability of high speed or broad band internet connections in Germany is highly common.
To summarize, the spatial distribution of participants of internet convenience samples may be used as a tool to estimate the regional distribution of a "hidden" population like MSM. Potential biases should be considered, which may arise from subtle differences in regional participation rates and recruitment on websites with skewed user characteristics. Compared to infectious disease surveillance data, self reported infection status data from surveys have the potential to detect post-HIV diagnosis migration patterns.
The KABaSTI study has been funded by the German Federal Ministry of Health, the GMA survey by the Federal Agency for Health Promotion (BZgA).
- Michaels S, Lhomond B: Conceptualization and measurement of homosexuality in sex surveys: a critical review. Cad Saúde Pública. 2006, 22: 1365-1374. 10.1590/S0102-311X2006000700002.View ArticlePubMedGoogle Scholar
- Nguyen TQ, Gwynn RC, Kellerman SE, Begier E, Garg RK, Pfeiffer MR, Konty KJ, Torian L, Frieden TR, Thorpe LE: Population prevalence of reported and unreported HIV and related behaviors among the household adult population in New York City, 2004. AIDS. 2008, 22: 281-287. 10.1097/QAD.0b013e3282f2ef58.View ArticlePubMedGoogle Scholar
- Holmberg SD: The estimated prevalence and incidence of HIV in 96 large US metropolitan areas. Am J Public Health. 1996, 86: 642-654. 10.2105/AJPH.86.5.642.View ArticlePubMedPubMed CentralGoogle Scholar
- Manning SE, Thorpe LE, Ramaswamy C, Hajat A, Marx MA, Karpati AM, Mostashari F, Pfeiffer MR, Nash D: Estimation of HIV Prevalence, Risk Factors, and Testing Frequency among Sexually Active Men Who Have Sex with Men, Aged 18–64 Years-New York City, 2002. J Urban Health. 2007, 84: 212-225. 10.1007/s11524-006-9135-5.View ArticlePubMedPubMed CentralGoogle Scholar
- Prestage G, Ferris J, Grierson J, Thorpe R, Zablotska I, Imrie J, Smith A, Grulich AE: Homosexual men in Australia: population, distribution and HIV prevalence. Sexual Health. 2008, 5: 97-102. 10.1071/SH07080.View ArticlePubMedGoogle Scholar
- Hughes A, Saxton P: Geographic micro-clustering of homosexual men: implications for research and social policy. Social Policy J New Zealand. 2006, 28: 158-178.Google Scholar
- Kippax S, Crawford J, Rodden P, Benton K: Report on Project Male-Call. Australian Government Publishing Services. Canberra. 2004Google Scholar
- Mercer CH, Fenton KA, Copas AJ, Wellings K, Erens B, McManus S, Nanchahal K, Macdowall W, Johnson AM: Increasing prevalence of male homosexual partnerships and practices in Britain 1990–2000: evidence from national probability surveys. AIDS. 2004, 18: 1453-1458. 10.1097/01.aids.0000131331.36386.de.View ArticlePubMedGoogle Scholar
- Sandfort T: Homosexual and bisexual behaviour in European countries. Sexual behaviour and HIV/AIDS in Europe: Comparison of National Surveys. Edited by: Hubert M, Bajos N, Sandfort T. 1998, UCL Press, LondonGoogle Scholar
- Smith AMA, Rissel CE, Richters J, Grulich AE, de Visser RO: Sexual identity, sexual attraction and sexual experience among a representative sample of adults. Aust N Z J Public Health. 2003, 27: 138-145. 10.1111/j.1467-842X.2003.tb00801.x.View ArticlePubMedGoogle Scholar
- Metzger DS, Koblin B, Turner C, Navaline H, Valenti F, Holte S, Gross M, Sheon A, Miller H, Cooley P, Seage GR: Randomized controlled trial of audio computer-assisted self-interviewing: utility and acceptability in longitudinal studies. HIVNET Vaccine Preparedness Study Protocol Team. Am J Epidemiol. 2000, 152: 99-106. 10.1093/aje/152.2.99.View ArticlePubMedGoogle Scholar
- Bochow M: AIDS – Wie leben schwule Männer heute?. AIDS-Forum DAH, Deutsche AIDS-Hilfe e.V., Berlin. 1989Google Scholar
- Schmidt AJ, Marcus U, Hamouda O: KABaSTI-Studie – Wissen, Einstellungen und Verhalten bezüglich sexuell übertragbarer Infektionen bei Männern mit gleichgeschlechtlichem Sex. 2007, [http://www.rki.de/nn_271502/DE/Content/InfAZ/S/STD/Studien/KABaSTI/KABaSTI.html]Google Scholar
- Bochow M, Schmidt AJ, Grote S: Schwule Männer und AIDS. 8. Wiederholungsbefragung im Auftrag der Bundeszentrale für gesundheitliche Aufklärung. 2007, mimeo/Social Science Research Center, Berlin, [http://www.wzb.eu/bal/ph/pdf/SMA-2007_gesamtbericht.pdf]Google Scholar
- BZgA: AIDS im öffentlichen Bewusstsein der Bundesrepublik Deutschland 2007, Kurzfassung Mai 2008. [http://www.bzga.de/studien]
- Hamouda O: HIV/AIDS Surveillance in Germany. J Acquir Immune Defic Syndr. 2003, 32 (Suppl 1): S49-54.View ArticlePubMedGoogle Scholar
- Pollak Michael: Les Homosexuels et le Sida. 1988, Métailié, ParisGoogle Scholar
- Evans AR, Wiggins RD, Mercer CH, Bolding GJ, Elford J: Men who have sex with men in Great Britain: comparison of a self-selected internet sample with a national probability sample. Sex Transm Infect. 2007, 83: 200-205. 10.1136/sti.2006.023283.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2458/9/180/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.