This study suggests there was a dramatic increase in the prevalence of circumcision among young men in rural areas of Mwanza Region, Northern Tanzania, between the late 1990s and 2008 despite no active health promotion of circumcision occurring in the Region during that time period. The proportion of young men who were circumcised was 41% in 2007/08, and had more than doubled since the 2001/02 survey at all ages studied (16–23 years). The median reported age at circumcision was 16 years, with some indication of an earlier age at circumcision among the youngest birth cohorts.
The increase in circumcision prevalence is notable, particularly considering that most participants belong to the Sukuma ethnic group, who are traditionally non-circumcising, and that this occurred before active promotion of circumcision in the Region. Four independent studies carried out in rural Mwanza Region during the 1990s suggested there was a low prevalence of circumcision among adult men of 10–15% 
. Data from the 2001/02 survey within the MEMA kwa Vijana trial in rural Mwanza Region also suggested the prevalence was around 17% among men aged 16–20 
. Another survey in selected villages in rural Mwanza Region conducted in 2004 reported a prevalence of 30% among 15–44 year olds 
. National surveys have reported a relatively high prevalence of circumcision in Mwanza Region of around 55% in 2003 and again in 2007; however these estimates included urban areas and were based on very small sample sizes of around 300 men 
. The strength of this study in determining the prevalence of circumcision over time is that it uses data from the long-term follow-up of the same study populations, rather than comparing estimates from studies in different populations at different time periods.
This rise in prevalence in male circumcision in rural Mwanza Region is also supported by qualitative evidence from the 1990s 
indicating that circumcision was becoming more acceptable and widely practised within Mwanza Region. It was suggested that interaction with circumcising ethnic groups and changing local perceptions that associate male circumcision with modernity and sexual hygiene were driving the change in attitudes 
. Together, the qualitative and quantitative evidence suggests that the increase in circumcision prevalence observed here among youth in rural Mwanza Region is a continuation of a trend that began in the 1990s, and has then accelerated substantially.
Since the results of the circumcision trials have been published, the Tanzanian government has developed a national circumcision strategic plan which aims to provide free circumcision to 2.8 million men and boys aged 10–34 years from 2010–2015 
. Prior to 2008 however there were no large-scale formal circumcision promotion campaigns in Tanzania, so public health initiatives cannot have been responsible for the observed trends. The first trial evidence confirming that circumcision conferred protection against HIV was not published until 2005. It seems unlikely that this knowledge would have reached sufficient numbers of young men within this rural area by 2007/08 to have driven this increase in the absence of a publicity campaign and active promotion of circumcision. It may be that the younger generation are more health-aware and this may have led to an increase in circumcision. Qualitative work in Mwanza Region and elsewhere in East Africa shows that circumcision is perceived as a hygienic practice 
. This is supported by our finding that circumcision was associated with safer sexual behaviours in this population. The rapid expansion of secondary education in most of rural Tanzania which occurred after 2001/2 may also have contributed to the observed increase in circumcision prevalence, given that higher than primary education has been associated with greater circumcision 
However, it is important to explore whether the observed increase in circumcision prevalence could be due to an artefact of the data, rather than a true effect. Firstly, if participants at the various surveys were not comparable, this might explain the apparent increase. However, analysis of the socio-demographic characteristics of participants in the 2001/02 and 2007/08 surveys showed no such differences [data not shown]. There were more Christians in the 2007/08 survey compared to the 2001/02 survey (80.7% versus 70.4%); self-identification as a Christian may increase with age. However, as self-identifying as a Christian did not determine circumcision status, and as there was no difference in the proportion of Muslims, this is not likely to have driven the increase.
Secondly, if study clinicians were better at recognising circumcised men at the 2007/08 survey, this could have led to an apparent increase in circumcision prevalence. Problems of clinician-assessed circumcision status have been raised, particularly regarding the difficulty in recognising partial circumcisions 
. However, the training given to the clinicians was similar for each of the MEMA kwa Vijana surveys. A comparison of self-reported and clinician-assessed circumcision showed a reasonably high level of agreement in the 1998 survey (97.1% concurrence) 
. Moreover, similar trends were obtained when analysing self-reported circumcision status. It therefore seems unlikely that the observed increase in male circumcision is attributable to differential reporting of circumcision by clinicians.
Within our study cohort, circumcised men reported less risky sexual behaviour, particularly regarding condom use. In contrast, a cross-sectional study in Mbale, Uganda, found circumcised men engaged in riskier sexual behaviours; circumcised men had more extra-marital partners, and more sex in exchange for gifts or money 
. However a cohort study in Western Kenya found no evidence for any difference in risky behaviours between recently circumcised and non-circumcised men 
, and a review of studies from Mwanza Region found condom use tended to be higher in circumcised men 
. Further studies are needed to understand the sexual behaviour of circumcised compared to non-circumcised men in different settings, as behaviours associated with circumcision may be locally specific and not generalisable to other settings.
The associations between circumcision and STIs in this study are in line with results from other studies in Africa. Those circumcised before sexual debut had a 50% lower odds of having HIV, compared to non-circumcised men, comparable with the risk reduction of 50–60% found in circumcision trials 
. As might be expected, there was less association with HIV among those circumcised at or after sexual debut, as it is possible that these men became infected before circumcision. Being circumcised was associated with reduced odds of having HSV-2, supporting evidence from two of the RCTs 
and previous observational studies 
, that circumcision protects against the acquisition of HSV-2. Syphilis and circumcision were not associated in this study after adjustment for confounders, which although contradicting other observational data, 
was also seen in trial data from Uganda 
. Among those circumcised before sexual debut, there was some evidence of an association with GUS after adjusting for confounders, which is consistent with other evidence suggesting circumcision protects against genital ulcer disease 
. As with trial data 
, there was no evidence that circumcision protects against the non-ulcerative STIs, chlamydia and gonorrhoea.
Our study had some limitations. First, the estimated age at circumcision was likely to be an approximation for many participants, limiting interpretation of age at circumcision. Missing circumcision status (with circumcision divided into three categories), for the 505 circumcised individuals in whom age at circumcision and/or age at sexual debut was unknown, could also potentially have biased the results, and a sensitivity analyses did demonstrate a less strong association between circumcision and HIV that was no longer statistically significant. There is also the possibility of reverse causality; the cross-sectional design could not establish the sequence of circumcision and STIs, and some men may have been circumcised as a result of having an STI, but this would tend to underestimate any protective effect.
In conclusion, the dramatic increase in circumcision prevalence over a relatively short period of time in this population, in the absence of any circumcision promotion campaigns, demonstrates that traditionally non-circumcising groups are amenable to change regarding their attitude toward circumcision. In this study, circumcised men reported safer, rather than riskier sexual behaviours, which is encouraging. However, our data were collected prior to widespread knowledge from the RCTs that circumcision can reduce risk of HIV infection and behavioural counselling prior to adult circumcision remains an integral and essential component of circumcision scale-up.