This is the largest community based study which has measured the prevalence and incidence of HTLV-1 and its associations with HIV. These data show a high HTLV-1 prevalence of approximately 5% and a stable incidence of approximately 1.7 per 1000 pyo in the adult population of Caió in rural Guinea-Bissau between 1990 and 2007. The prevalence increased with age and was higher in women than in men. A significant association between prevalent HIV and HTLV-1 infections was observed among women, which persisted after adjustment for potentially confounding risk factors. In the longitudinal analysis, HIV positive individuals tended to have a higher risk of acquiring HTLV-1 infection than HIV negative people, while HTLV-1 infection did not increase the risk of becoming infected with HIV.
Although the HTLV-1 prevalence in the study area declined from 5.9% in 1997 to 4.6% in 2007, there was little difference between the prevalence in 1990 and 2007. Also, it was still twice as high as the prevalence in the capital Bissau, situated at a distance of 100 km (2.3% in 2006) . Quite big differences in HTLV-1 prevalence have been described between areas that are relatively close geographically, which could be related to cultural and/or ethnic differences . The prevalence is also high compared to other community-based studies in Africa . Studies from Bissau have shown HTLV-1 to be associated with an increased mortality and tuberculosis among HIV infected individuals [11, 30] and in the study area with tropical spastic paresis . Hence, it is likely that HTLV-1 has an important impact on this community.
While HTLV-1 and HIV-2 prevalences have decreased, HIV-1 prevalence has increased in both the study area and the capital [16, 23, 26]. Therefore, it seems unlikely that safer sex practices have played an important role in this decline. An increase in risk behavior and blood transfusions during the War of Independence (1963-74) is thought to have enabled the spread of HIV-2 [32, 33]. A concomitant iatrogenic spread through vaccination campaigns and large-scale parenteral treatment programs might have also contributed to the initial spread [34, 35]. A similar scenario has been proposed for the spread of hepatitis C virus . These events may also have contributed to the higher prevalence of HTLV-1 observed in earlier studies and the decrease in prevalence observed in the current study.
In this study, sexual risk factors (HIV and TPHA positivity) were mainly identified for women and prevalence was higher among women, which suggests greater susceptibility to HTLV-1 infection by women [37, 38]. A positive TPHA test was used as an indication of exposure to syphilis at some point, but does not indicate acute infection.
Mother-to-child-transmission (MTCT) of HTLV-1 has been clearly documented in Japan and Jamaica, but has only been described in one cohort from Africa [39, 40]. In the current study, a strong association was observed between HTLV-1 status of mothers and their offspring (OR 4.6, CI 2.6-8.1), indicating that MTCT contributes to maintaining the HTLV-1 epidemic in this community.
Screening of blood transfusions for HIV since 1989 may have lead indirectly to a decrease of HTLV-1 in Bissau [16, 41]. This mechanism seems unlikely to have played a role in the Caió area, where having received a blood transfusion was not associated with HTLV-1 infection in either 1990  nor in 1997 (this risk factor was not assessed in 2007).
Striking was the fact that among men, 6 out of the 7 incident cases occurred in 15-16 year olds in 1997-2007. In this area, the incidence of HIV-1 and HIV-2 was low in 15-24 year old men , so it remains to be elucidated how these young men acquired HTLV-1.
HTLV-1 and HIV dual-infection
In this study, HIV and HTLV-1 infections showed a cross-sectional association, as has been shown before in this study area in 1990  and in the general adult population [16, 41], elderly people , an occupational cohort  and pregnant women  in Bissau. Why this association is stronger for women than for men remains unclear. Some studies have demonstrated a more efficient male-to-female transmission of HTLV-1 [37, 44, 45]. An increased susceptibility among older women due to biological changes has been suggested, such as post-menopausal changes in the vaginal mucosa [10, 28, 38, 46]. Higher mortality in men with HIV/HTLV-1 dual infections could also contribute to the observed sex difference; however, mortality rates were similar for men and women in the >35 years cohort from Bissau .
It is unknown whether HTLV-1 is a risk factor for HIV infection or vice versa. Therefore, it was interesting to find that pre-existing HTLV-1 infection was not associated with incident HIV infection, but prevalent HIV infection appeared to increase the risk of acquiring HTLV-1. An increased susceptibility could be due to the higher level of immune activation induced by HIV, thereby enhancing the susceptibility of the host to other retroviral infections which are dependent on active immune cells as targets [47–49]. If a substantial number of individuals became HTLV-1 infected perinatally, their HTLV-1 status would not represent a sexual risk factor and this could explain the similar HIV incidence rates observed among HTLV-1 positive and HTLV-1 negative people.
With the current roll-out of anti-retroviral treatment in Guinea-Bissau, it is important to realize that HTLV-1 co-infection may increase the CD4 counts, which is the main indicator for start of treatment  (reviewed in ).
This study has several limitations. First, in the 1990 and 1997 surveys a number of samples could not be tested for HTLV and these samples were more often of HIV infected people; therefore, the prevalence may have been underestimated. The adjusted prevalences that were reported were based on the assumption that the prevalence of HTLV-1 was distributed the same as among subjects with a known HTLV result. These missing HTLV-1 results may also have led to an underestimation of the association between HTLV-1 and HIV in the reported ORs. Second, the association between HTLV-1 and HIV may have been partly caused by residual confounding, since the factors used in this study may not have controlled completely for (sexual) risk behavior. Third, the HTLV-1 infected adult children of HTLV-1 infected mothers may not have been infected by their mother but might have acquired the virus later in life. However, HIV status was not a confounder in this analysis, suggesting that sexual transmission played a much less important role in this group. Fourth, HIV infected subjects will have had a higher chance of dying before a follow-up blood sample was obtained, especially since the periods between survey rounds were long (median 7.3 years for the first and 9.4 years for the second period). Therefore, the HTLV-1 incidence among HIV positive people is likely to be an underestimate.
Finally, the IRR from the analysis of HIV and HTLV-1 incidence by retroviral status should be interpreted with caution since the number of incident cases among retroviral infected people was small.